what drives deregulation? economics and ......charters, and they often owned or purchased shares in...
TRANSCRIPT
WHAT DRIVES DEREGULATIONECONOMICS AND POLITICS OF THE RELAXATION
OF BANK BRANCHING RESTRICTIONS
RANDALL S KROSZNER AND PHILIP E STRAHAN
This paper investigates private-interest public-interest and political-institutional theories of regulatory change to analyze state-level deregulation ofbank branching restrictions Using a hazard model we nd that interest groupfactors related to the relative strength of potential winners (large banks and smallbank-dependent rms) and losers (small banks and the rival insurance rms) canexplain the timing of branching deregulation across states during the last quartercentury The same factors also explain congressional voting on interstate branch-ing deregulation While we nd some support for each theory the private interestapproach provides the most compelling overall explanation of our results
I INTRODUCTION
The pervasive economic deregulation of the last quartercentury poses a key challenge for positive theories of regulatorychange The private-interest theory of regulation also called theeconomic theory characterizes the regulatory process as one inwhich well-organized groups use the coercive power of the state tocapture rents at the expense of more dispersed groups (egStigler [1971] Peltzman [1976 1989] and Becker [1983]) Thisapproach contrasts with the public-interest theory of regulation inwhich government intervention corrects market failures andmaximizes social welfare (see Joskow and Noll [1981]) Otherapproaches emphasize the importance of beliefs and ideology (egPoole and Rosenthal [1997]) and the institutional arrangementsof the decision-making process (eg North [1990] Dixit [1996]and Irwin and Kroszner [1999])
While the private-interest theory has had much success inexplaining a wide variety of regulatory interventions that aredifficult to rationalize on public-interest grounds (see Stigler
We thank Rebecca Demsetz Edward Glaeser Jeffrey Gunther JosephHarrington Anil Kashyap George Kaufman Lawrence Kenny David MarshallJohn Matsusaka Wally Mullin Sam Peltzman Raghuram Rajan Andrei ShleiferToby Stuart Joel Waldfogel an anonymous referee and seminar participants atthe Federal Reserve Bank of New York University of Chicago University ofRochester University of British ColumbiaAmerican Law and Economics Associa-tion and NBER Summer Institute for helpful comments Kroszner is grateful forresearch support from grants to the George J Stigler Center for the Study of theEconomy and the State by the Lynde and Harry Bradley Foundation The viewsexpressed here are strictly those of the authors and do not necessarily reect theposition of the Federal Reserve Bank of New York or the Federal Reserve System
r 1999 by the President and Fellows of Harvard College and the Massachusetts Institute ofTechnologyThe Quarterly Journal of Economics November 1999
1437
[1988]) it has been less effective in explaining the removal of suchregulations [Peltzman 1989 Noll 1989] In contrast the public-interest theory can account for welfare-enhancing deregulationbut not for the origin of regulations that reduced competition andthat had few if any welfare benets
This paper examines the elimination of restrictions on bankbranching since the 1970s to assess the relative explanatorypower of these theories of regulatory change Unlike most otherrecent episodes of deregulation that occurred at a nationallevelmdashsuch as in railroads trucking airlines long-distance tele-communications securities brokerage petroleum and naturalgasmdashbank branching regulation operated on a state-by-statebasis and deregulation has taken place gradually across thestates Branching deregulation thus provides a much greatersource of cross-sectional and time-series variation than othertypes of deregulation
To investigate what drives deregulation we use a hazardmodel to explain the timing of intrastate branching deregulationWe incorporate proxies for interest-group public-interest andpolitical-institutional factors to understand how changes in thesevariables over time and across states affect the likelihood ofderegulation An additional benet of focusing on branchingderegulation is that the state-by-state reform culminates in thepassage of the 1994 Riegle-Neal Interstate Banking and Branch-ing Efficiency Act that effectively eliminated branching restric-tions nationwide This allows us to determine whether the samefactors driving state-level deregulation can also explain congres-sional voting on federal repeal of interstate branching restrictions
We nd that deregulation occurs earlier in states with fewersmall banks in states where small banks are nancially weakerand in states with more small presumably bank-dependentrms Also a larger insurance industry delays deregulation whenbanks may compete in the sale of insurance products Interestgroup factors related to the relative strength of potential winners(large banks and small rms) and losers (small banks and therival insurance rms) thus can explain the timing of branchingderegulation across states The same interest group variables alsocan explain the voting pattern of legislators in the U S House ofRepresentatives on interstate banking deregulation We thenexamine broad technological legal and nancial innovations thataltered the costs and benets of the regulations and can explainwhy deregulation began in the 1970s rather than earlier [Kane
QUARTERLY JOURNAL OF ECONOMICS1438
1996] While some of our ndings can be rationalized on public-interest grounds the private-interest theory provides a simpleway to understand these results as a whole1
II THE ORIGINS AND DEMISE OF GEOGRAPHICAL
RESTRICTIONS ON BANKING
A Origins in Public Financing Strategies
After the United States Constitution prevented the statesfrom issuing at money and from taxing interstate commercestates used their powers over banks to generate substantialrevenues [Sylla Legler and Wallis 1987]2 To enter the businessone had to obtain a bank charter States received fees for grantingcharters and they often owned or purchased shares in banks orlevied taxes on banks To enhance these revenues each state hadan interest in restricting competition among banks and many ofthe restrictions on the geographical expansion of banks originatein this period3 Since states received no charter fees from banksincorporated in other states the states prohibited out-of-statebanks from operating in their territoriesmdashhence the origin of theprohibition on interstate banking4
Legislatures also often restricted intrastate expansion Stateswould grant a charter for a specic location or limit bank branchesto that city or country By adopting branching restrictions statescreated a series of local monopolies from which they could extractpart of the rents Some state legislatures even passed lsquolsquounitbankingrsquorsquo laws that prevented a bank from having any branchesSuch regulations naturally produce beneciaries who are loatheto give up their protections and privileges Benets tend to be
1 Jarrell [1984] tests the private interest theory of regulatory exit using theend of xed commissions at the New York Stock Exchange Also see Peltzman[1989] and Noll [1989] Kanersquos [1996] lsquolsquoregulatory dialecticrsquorsquo emphasizes technologi-cal change interacting with private interests
2 During the rst third of the nineteenth century for example the bank-related share of total state revenues exceeded 10 percent in a dozen states InMassachusetts and Delaware a majority of total state revenue was bank-related
3 Noll [1989] has characterized conceiving of governments as distinctinterest groups concerned about nancing their expenditures as the LeviathanApproach see Niskanen [1971] and Brennan and Buchanan [1977]
4 With the passage of the National Banking Act in 1864 the federalgovernment also began to charter banks (motivated by a desire to use suchinstitutions to help to fund the Civil War see Kroszner [1997]) While there hadbeen some initial ambiguity concerning state authority over these institutions the1927 McFadden Act claried that until the 1994 Riegle-NealAct states effectivelyhad the right to prevent interstate branching and to force national banks toconform to state branching regulations See White [1983]
WHAT DRIVES DEREGULATION 1439
concentrated while costs to consumers of a less efficient andcompetitive nancial sector tend to be diffuse (eg Stigler [1971]and Peltzman [1976])
B A Brief History of Recent Branching Deregulation
Prior to the 1970s most states had laws restricting within-state branching and all states forbade interstate branching(Table I and Figure I) Although there had been some changes instate branching laws during the late nineteenth and early twenti-eth centuries these laws have remained stable since the GreatDepression Some of the statutes were essentially unchanged formore than a century Since the early 1970s however all but one ofthese states have relaxed restrictions on intrastate branching
Deregulation of these restrictions typically involves threetypes of reforms The rst concerns the formation of multibankholding companies (MBHCs) that were permitted to own multiplebanks but had to operate them separately The offices of the banksin an MBHC could not be integrated into a single network so adepositor at one bank would not have access to her deposits atanother The banks in an MBHC also could not consolidate theirback-office operations and each had to meet all regulatoryobligations eg capital requirements as if it were a stand-aloneinstitution
The second and most important step toward deregulationoccurs when states permit branching by merger and acquisitionthereby allowing MBHCs to convert offices of subsidiary banks(existing or acquired) into branches of a single bank An MBHCcould then integrate its banking offices into a single branchnetwork A third reform occurs when states permit full statewidebranching whereby banks could open new branches anywherewithin state borders5
The Douglas amendment to the Bank Holding Company Actof 1956 prevented holding companies from buying out-of-statebanks unless that state explicitly permitted such acquisitions bystatute (see Macey and Miller [1992]) Since no state allowed suchacquisitions holding companies were effectively prohibited fromcrossing state lines Deregulation began in 1975 when Mainepassed legislation permitting out-of-state bank holding compa-
5 Permitting branching only through merger and acquisition before fullstatewide branching could be interpreted as states allowing incumbent banks tomaintain the ability to extract at least a portion of the rents associated withbarriers to entry from purchasing banks
QUARTERLY JOURNAL OF ECONOMICS1440
TABLE IYEAR OF DEREGULATION OF RESTRICTIONS ON GEOGRAPHICAL EXPANSION BY STATE
State
Intrastatebranching
through MampA
Full intrastatebranchingpermitted
Interstatebanking
permitted
Multibank holdingcompaniespermitted
AL 1981 1990 1987 1970AK 1970 1970 1982 1970AZ 1970 1970 1986 1970AR 1994 1989 1985CA 1970 1970 1987 1970CO 1991 1988 1970CT 1980 1988 1983 1970DE 1970 1970 1988 1970DC 1970 1970 1985 1970FL 1988 1988 1985 1970GA 1983 1985 1976HI 1986 1986 1970ID 1970 1970 1985 1970IL 1988 1993 1986 1982IN 1989 1991 1986 1985IA 1991 1984KS 1987 1990 1992 1985KY 1990 1984 1984LA 1988 1988 1987 1985ME 1975 1975 1978 1970MD 1970 1970 1985 1970MA 1984 1984 1983 1970MI 1987 1988 1986 1971MN 1993 1986 1970MS 1986 1989 1988 1990MO 1990 1990 1986 1970MT 1990 1993 1970NE 1985 1990 1983NV 1970 1970 1985 1970NH 1987 1987 1987 1970NJ 1977 1986 1970NM 1991 1991 1989 1970NY 1976 1976 1982 1976NC 1970 1970 1985 1970ND 1987 1991 1970OH 1979 1989 1985 1970OK 1988 1987 1983OR 1985 1985 1986 1970PA 1982 1990 1986 1982RI 1970 1970 1984 1970SC 1970 1970 1986 1970SD 1970 1970 1988 1970TN 1985 1990 1985 1970TX 1988 1988 1987 1970UT 1981 1981 1984 1970VT 1970 1970 1988 1970VA 1978 1987 1985 1970WA 1985 1985 1987 1981WV 1987 1987 1988 1982WI 1990 1990 1987 1970WY 1988 1987 1970
States not yet deregulated Source Amel [1993] and updates by authors
WHAT DRIVES DEREGULATION 1441
nies to acquire Maine banks As part of the Garn-St GermainAct federal legislators in 1982 amended the Bank HoldingCompany Act to allow failed banks and thrifts to be acquired byany bank holding company regardless of state laws (see egKroszner and Strahan [1996]) Many states then entered regionalor national reciprocal arrangements whereby their banks could bebought by any other state in the arrangement Between 1984 and1988 38 states joined one of these arrangements (see Amel[1993])
Table I and Figure I illustrate the history of state deregula-tion of geographical restrictions since 1970 The state-by-statederegulation culminates in the phaseout of interstate bank-ing restrictions with the passage of the 1994 Riegle-Neal Inter-state Banking and Branching Efficiency Act which effectivelycodied at the national level what had been occurring at the statelevel6
6 The Act permitted states to pass legislation to opt out of the interstatebanking provisions if the legislature did so before the provisions were to go intoeffect in mid-1997 Since only Texas and Montana have passed opt-out legislationthe United States will now have nearly complete interstate banking and branching
FIGURE IDeregulation of Restrictions on Intrastate Branching
Permitted Intrastate Branching before 1970
QUARTERLY JOURNAL OF ECONOMICS1442
III HYPOTHESES AND VARIABLE DEFINITIONS
Our empirical tests are based on the timing of intrastatebranching deregulation through merger and acquisition We focuson this form of branching deregulation because it has a muchgreater economic impact than the other forms of branchingderegulation (see the Appendix) Deregulation through mergerand acquisition is the only type of branching deregulation thatconsistently has a statistically signicant effect on bankingstructure bank efficiency and overall economic growth and theestimated magnitudes of the effects are greatest for this type ofderegulation We also analyze the interstate deregulation thattakes place at the national level by modeling voting behavior inthe House of Representatives on federal interstate branchinglegislation7
In the remainder of this section we describe our hypothesesabout the private-interest public-interest and political-institu-tional theories along with the data sources used to construct thevariables Some of the variables help us to distinguish between apublic- and private-interest approach while others will be consis-tent with both approaches The political-institutional approachhas little overlap in its implications with the other theories Ourmethod is to assess the relative importance of factors representingeach approach in speeding or slowing deregulation
A Hypotheses
Intraindustry Rivalry Small banks have fought to maintainand extend branching restrictions both historically and in therecent debates8 Smaller banks appear to have been the mainwinners from antibranching laws since these restrictions protectthem from competition from larger and more efficient bankingorganizations (see Flannery [1984] Jayaratne and Strahan [1998]
7 We do not employ a hazard model to explain the timing of interstatederegulation for two reasons the time clustering of interstate deregulation andcomplications of the strategic interdependence of the statesrsquo behavior As thenumber of states in an interstate arrangement increases for example the effectson potential acquirers is ambiguous they might prefer a larger pool of potentialtarget banks but might have to pay more as the number of competing acquirersexpands The benet to potential targets tends to increase with the number statesin an arrangement since the expected selling price increases with the number ofpotential bidders [Brickley and James 1987]
8 Economides Hubbard and Palia [1996] provide evidence that voting inCongress for the 1927 McFadden Act responded to small state banksrsquo interest inlimiting competition from large national banks See also White [1983] and Abramsand Settle [1993] for historical opposition On the small bank opposition to therecent branching deregulation see Kane [1996] and The Economist [8694 p 59]
WHAT DRIVES DEREGULATION 1443
and Winston [1993]) Branching restrictions thus tend to reducethe efficiency and consumer convenience of the banking system9
The private-interest theory therefore predicts that reformshould occur later in those states where the strength of the smallbanks relative to that of the large banks is greater According tothe public-interest theory deregulation should take place earlierwhere small banks are relatively important because the socialcosts of the regulation are directly related to the size of theprotected sector The social costs include deadweight costs andlosses associated with inefficiencies in the production of bankingservices relative to production without geographical restrictions
Interindustry Rivalry A number of states permit state-chartered commercial banks to sell insurance The insurancelobby would thus oppose the relaxation of branching restrictionswhen banks can sell insurance because such deregulation mightpermit banks to provide a more efficient insurance distributionnetwork10 According to the private-interest theory reform shouldoccur later in states where banks can sell insurance and theinsurance industry is important relative to the banking industryThe efficiency costs of the branching restrictions however risewith the size of the insurance sector since the restrictions preventthe exploitation of scope economies Under the public-interesttheory reform should therefore occur earlier where banks can sellinsurance and the insurance industry is relatively large11
Consumers of Banking Services Banks are a major source ofcredit for small rms [Cole and Wolken 1994] Branching deregu-lation tends to reduce banksrsquo local market power [Jayaratne andStrahan 1998] In addition Strahan and Weston [1998] nd thatlending to small businesses increases on average when smallbanks are purchased by other banking organizations and Bergeret al [1998] nd that credit availability to small businessesincreases in the years following a takeover of a small bank by alarger banking organization Since bank borrowers tend to benet
9 Flannery [1984] shows that small banks in states with branching restric-tions have higher costs than small banks in states without such restrictions andthe Appendix shows how the production of banking services becomes more efficientfollowing deregulation
10 The Economist [8694 p 59] describes the insurance industryrsquos opposi-tion to branching deregulation
11 Another rival interest we considered was the Savings and Loan industryThe share of assets in Savings and Loans relative to banking plus Savings andLoan assets in the state however had only small and statistically insignicanteffects in the models we estimate below
QUARTERLY JOURNAL OF ECONOMICS1444
from branching deregulation in particular and bank consolidationin general the private-interest theory would predict that stateswith numerous small bank-dependent rms would deregulateearlier12 This prediction however is also consistent with thepublic-interest theory since the social costs of the restrictions arehigher in states with more small bank-dependent rms
Another effect of branching restrictions on bank customerscan be related to the prices paid for bank services High loanprices before deregulation may reect for example a largemarket share for inefficient high-cost banks or high rents beingearned by the banks protected from competition In either casethe public-interest theory would suggest that states with rela-tively high initial loan interest rates should deregulate earlierThe private-interest theory however does not have a clearimplication high rents could lead the beneciaries to ght harderto maintain them (eg Stigler [1971]) or the deadweight costsassociated with these high rents could make it more difficult forthe beneciaries to maintain the restrictions (eg Becker [1983])13
Bank Stability Geographic diversication through branch-ing could mitigate instability problems that were importantduring much of our sample period thereby improving welfareInstability also may reduce the incentives of banks to lobby tomaintain protections because unstable banks are less likely tosurvive to reap the benets of the restrictions politicians also mayanticipate that such banks are less likely to be able to providefuture contributions and support [Gunther 1994 1996]14 Both theprivate-interest and public-interest theories thus suggest thatderegulation is most likely in states where banking instability isgreatest
12 On the other hand local banking monopolies created by branchingrestrictions could strengthen relationshipsbetween banks and small and medium-sized rms and increase the availability of credit to these rms [Petersen andRajan 1994] Also some have argued that small business lending declines whenlarge banks take over small banks (eg Berger Kashyap Scalise [1995])
13 High loan interest rates also could reect compensation for higher risksand fewer opportunities for banks to diversify If state-specic risks explaindifferences in average loans rates then neither theory would have any predictionsfor the sign of this effect Following branching deregulation however loan interestrates tend to fall but bank risk proles do not appear to change [Jayaratne andStrahan 1998 Kroszner and Strahan 1998] This suggests that high loan rates inantibranching states are not simply reecting state-specic risks
14 Also deregulation can arise as a response to banking instability under atheory of ignorant or misinformed voters in which a banking crisis acts as aneducational device to make the previously ignorant public aware of the costs of theantibranching policy (see Kane [1996])
WHAT DRIVES DEREGULATION 1445
Political-Institutional Factors Republicans are typically per-ceived as more likely to favor deregulation than Democrats so thepolitical-institutional theories suggest that states that are con-trolled by Democrats deregulate later than those controlled byRepublicans15 In addition we investigate whether reform be-comes more likely when one party controls the legislature and thegovernorship Note that these political effects must be interpretedwith caution since the views of the politicians may simply reectthe economic interests of the constituents in the state (seePeltzman [1984])
B Variable Denitions and Data Sources
Our main proxy for the strength of the small banks is thefraction of banking assets in the state in lsquolsquosmallrsquorsquo banks We denesmall banks as banks with assets below the median size in eachstate By allowing the denition of small to vary across states wetake into account cross-state heterogeneity in bank sizes16 An-nual data on bank size are from the fourth quarter Reports ofIncome and Conditions (lsquolsquoCall Reportsrsquorsquo) from the Federal ReserveBoard We also include the capital-to-asset ratio of small banksrelative to large as a measure of their relative nancial healthSpecically this variable is the asset-weighted average capital-asset ratio for small banks minus the asset-weighted averagecapital-asset ratio for large banks Annual data on bank capitalare from the Call Reports
To measure the effects of the rival insurance industry we rstconstruct an indicator variable that is one if the state permitsbanks to sell insurance For each state we then measure the sizeof the insurance sector (total value added in the state) relative tothe sum of the banking plus insurance sectors We will examine
15 For more detail on the importance of legislative structures party politicsand ideology see Poole and Rosenthal [1997] Kahn and Matsusaka [1997] Irwinand Kroszner [1999] and Kroszner and Stratmann [1999]
16 We also considered a variety of other denitions of relative size but therelative size results reported below are not sensitive to which denition we useFirst we used a xed measure of small bank that did not vary across statesdening small banks as those with assets below $100 million in 1994 dollarsSecond we applied the state-varying and xed denitions at the level of lsquolsquobankingorganizationsrsquorsquo (which include multibank holding companies) instead of at the levelof banks Third we calculated gini coefficients of bank size inequality and bankconcentration indices as alternative proxies for small versus large bank power inthe state Finally we also considered banks in lsquolsquoruralrsquorsquo areas (that is not located in aBureau of the Census lsquolsquostandard metropolitan statistical arearsquorsquo) as small since thevalue of the restrictions might be greatest in protecting banks outside of cities fromentry by the city banks All of the alternative denitions are highly correlated andyield the same results on the importance of intraindustry differences
QUARTERLY JOURNAL OF ECONOMICS1446
the effect separately for states that permit banks to sell insuranceand those that do not Data on value-added by industry are fromU S Commerce Department Bureau of Economic Analysis Sur-vey of Current Business [August 1994]
To measure the relative importance of small bank-dependentborrowers we include the proportion of all establishments operat-ing in the state with fewer than twenty employees These data arecompiled by the Bureau of the Census17 As a measure of priceswe use the difference between the average interest rate on loans inthe state and the prevailing federal funds rate where the loanrate equals the ratio of total interest income on all domestic loansdivided by total domestic loans held by banks operating in thestate18 These data are from the end-of-year Call Reports butbecome available beginning only in 1976 Our proxy for bankinstability is the failure rate of banks measured as the percent oftotal state banking assets in failed institutions from the FederalReserve Boardrsquos National Information Center database19
We include two political variables to adjust for any indepen-dent inuence of party politics First we measure the degree ofparty control of the state government by the fraction of the threebodies of the state government (the assembly senate and gover-norship) controlled by Democrats This variable for example isone-third if the Democrats have a majority in the assembly andthe Republicans have a majority in the senate and hold thegovernorship Second we include an indicator variable equal toone if the same party controls the governorrsquos office and hasmajorities in both chambers of the state legislature20
17 We collected the establishment data by the state for three cross sections(1976 1982 and 1987) and interpolated the data in the intermediate years SeeState and Metropolitan Data Book 1982 1986 and 1991
18 A measure of how individual customers would be affected is interest ratesthey receive on deposits We calculated the average deposit interest rate in thestate from the end-of-year Call Reports and did not nd any relationship betweenthis variable and the timing of deregulation in our analysis below Since Jayaratneand Strahan [1998] nd that branching deregulation has no effect on deposit ratesour nding of no relationship between timing and rates result could be due toeither individual bank customers anticipating little pricing effect and not lobbyingor an inability of individual depositors to form an effective lobby
19 As alternative measures of distress we used (1) the bank failure ratemeasured as the number of failed institutions in the state relative to the totalnumber of banks (2) the overall capital-to-asset ratio of all banks in the state and(3) the growth in state personal income Our results are not sensitive to the choiceof distress proxy
20 We also estimated the models with three indicator variables to reect theparty control of the assembly senate and governorship separately The coefficientestimates on these variables are qualitatively similar to those we present below(that is Democratic control tends to slow deregulation) and the results on theinterest group variables are unaffected by the specication of the political
WHAT DRIVES DEREGULATION 1447
IV METHODS AND RESULTS
In this section we rst estimate how the variables describedabove inuence the timing of state-level branching deregulationusing a hazard model Next we discuss the robustness of theresults including a test of whether the ex post impact of deregula-tion on the different interest groups is consistent with ourinterpretation of the coefficients We then investigate federalinterstate branching deregulation with a probit model to explorewhether the factors that affect the timing of intrastate deregula-tion also inuence voting in the House of Representatives oninterstate deregulation
A Hazard Model
We consider the period from the beginning of our sample(1970) until deregulation as the lsquolsquoduration of regulationrsquorsquo or thelsquolsquotime until deregulationrsquorsquo The hazard rate h(t) is the likelihoodthat a state deregulates at time t given that the state has not yetderegulated [Kalbeisch and Prentice 1980 Kiefer 1988 Greene1997] To model the duration of regulation we must decide whatstructure if any to impose on the hazard function The Kaplan-Meier product-limit estimator provides a simple nonparametricway to estimate the shape of the hazard function over time[Greene 1997]21 Figure II graphs this estimate for our data andshows that the hazard function is relatively at in the early yearsand then grows steeper in the later years
A duration model that can approximate this shape is theWeibull proportional hazards model The hazard rate functiontakes the form
h[t x(t) b] 5 h0(t) exp[x(t)8b]
where x(t) is a vector of time-varying covariates b is a vector ofunknown parameters to be estimated and the baseline hazard
variables Note also that Nebraska has a nonpartisan unicameral legislature Weassume that the party of the Governor controls the state government for Nebraskaand set the unied government indicator to one As another alternative we triedincluding a variable that is one if party control of the legislature or governorshipchanges but it did not have an economically or statistically signicant effect anddid not affect any of the other results
21 Denote each of the K years in our sample period Tk and order them suchthat T1 T2 and so on Let nk be the number of states that have not yet deregulatedby Tk and dk be the number of states that deregulate in year Tk The Kaplan-Meierestimate of the hazard rate in each year is dknk In other words the hazard rate isthe ratio of the number of states that actually deregulate in each year to thenumber of states that have not yet deregulated in that year
QUARTERLY JOURNAL OF ECONOMICS1448
rate h0(t) is ptp 2 1 with shape parameter p that will be estimatedfrom the data When p 1 this model displays a monotonicallyincreasing hazard rate To estimate the parameters we maximizethe following log-likelihood function [Kiefer 1988]
L(b) 5 oi 5 1
N
di ln h[ti x(ti) b] 2 oi 5 1
N
e 0
tih[u x(u) b] du
where N indexes the number of states at risk at the beginning ofthe sample period di is an indicator equal to one for the statesthat deregulated by the end of the sample period and zero for thestates that do not deregulate (ie the censored observations) andti is the time of deregulation for the ith state This formulationassumes that the time until deregulation is independent acrossstates but the observations of the same state over time are notassumed to be independent In addition we use a robust estima-tion procedure to calculate the standard errors [Lin and Wei1989]
We exclude states that deregulated before 1970 from theanalysis so N equals 39 states During the sample period 36 ofthe 39 states deregulate We have three censored observations forwhich di is zero (Arkansas Iowa and Minnesota) but our results
FIGURE IIKaplan-Meier Estimate of Hazard Rate for Branching Deregulation
WHAT DRIVES DEREGULATION 1449
are not sensitive to the inclusion of these censored observationsSince we observe each state in each year up to and including theyear of deregulation we have a total of 637 observations Table IIreports summary statistics and correlation coefficients for theexplanatory variables we use in our analysis
We use the Weibull model rather than a model that does notimpose any structure on the baseline hazard rate so that we cancalculate the change in the expected time to deregulation for agiven change in the levels of the covariates22 In the Weibullmodel we can invert the hazard function and map it into the timedomain Rewriting the Weibull model in this way the log of thetime to deregulation T is a linear function of the political economyfactors and an error term ln (T ) 5 x8b 1 e23 Because we assumethat the baseline hazard rate is Weibull with a shape characterizedby p the new coefficients will be scaled by p that is b 5 2 bp
The b coefficients represent the percentage change in thetime to deregulation for a one-unit change in the covariates Apositive coefficient implies that an increase in the variableincreases the expected time until deregulation To gauge theeconomic importance of the effects we will multiply the b by thestandard deviation of the explanatory variables and then evaluatehow much the changes in the variables raise or lower the expectedtime to deregulation24 In Table III we report the b coefficients
22 The Cox proportional hazards model does not impose any structure on thebaseline hazard rate h0(t) and takes the following form
h[t x(t) b] 5 h0(t) exp [x(t)8b]
Common factors that affect the probability of deregulation in the same way for allstates therefore do not inuence the estimated coefficients The cost of the Coxmodel is that because there is no structure on the baseline we can calculatechanges in only relative hazard rates associated with changes in the covariatesThe Weibull model however provides sufficient structure so that we can translateour estimates into a log expected time metric that allows us to calculate the changein the expected time to deregulation for a given change in the covariates SeeKiefer [1988] Greene [1997] and Stata [1997] When we compare the coefficientsestimated by the Cox and Weibull models on our data they are very close (seeKroszner and Strahan [1998]) so imposing a Weibull structure does not appear todistort our estimates
23 The error term e is independent of x and has an extreme value distributionscaled by 1p (see Kiefer [1988] Greene [1997] and Stata [1997] for more details onthis log expected time parameterization of the Weibull model)
24 The transformation to the log expected time domain is only strictly correctwhen the covariates are constant over time Our interpretation of the coefficients ispossible in the time-varying model if we assume that a one-standard-deviationdifference would hold over the entire sample period that is the deregulationwould occur so many years earlier or later if the variable in question were higherby one standard deviation from 1970 to 1992 Note that the mean number of yearsto deregulation in the sample is 158
QUARTERLY JOURNAL OF ECONOMICS1450
TABLE IISUMMARY STATISTICS FOR VARIABLES IN HAZARD MODEL 1970ndash1992
Panel A Univariate statistics MeanStandarddeviation Minimum Maximum
(1) (2) (3) (4)Small bank asset share of all
banking assets in state 011 005 001 023Relative size of insurance to
banking plus insurance inthe state 047 009 024 080
Indicator is 1 if banks can sellinsurance in the state 020 040 0 1
Small rm share of the numberof rms in the state 088 002 078 095
Capital ratio of small banksrelative to large in the state 002 001 2 001 009
Share of state government con-trolled by Democrats 065 036 0 1
Indicator is 1 if state controlledby one party 054 050 0 1
Average yield on bank loans inthe state minus Fed fundsrate 002 002 2 004 008
Panel B CorrelationsSmallbank
Relativesize of
insuranceInsuranceindicator
Smallrm
Smallbank
capitalDem
controlSingleparty
(1) (2) (3) (4) (5) (6) (7)Small bank asset share
of all banking assetsin state 1
Relative size of insur-ance to banking plusinsurance in the state 2 034 1
Indicator is 1 if bankscan sell insurance inthe state 020 2 012 1
Small rm share of thenumber of rms inthe state 030 2 011 005 1
Capital ratio of smallbanks relative tolarge in the state 2 066 020 2 010 2 006 1
Share of state govern-ment controlled byDemocrats 2 006 001 2 018 2 018 005 1
Indicator is 1 if statecontrolled by oneparty 007 012 001 2 006 2 006 030 1
Average yield on bankloans in the stateminus Fed funds rate 2 001 2 015 003 022 2 010 2 003 2 006
The number of observations is 637 except for the average interest rate on bank loans which becomesavailable only in 1976 so the number of observations is 408
TABLE IIIHAZARD MODEL OF POLITICAL ECONOMY FACTORS AFFECTING THE TIMING OF STATE
BRANCHING DEREGULATION 1970ndash1992
(1) (2) (3) (4) (5) (6)Small bank asset share of
all banking assets instate
648(122)
519(137)
632(135)
516(118)
488(133)
445(088)
Capital ratio of smallbanks relative to largein the state
1325(348)
921(319)
1301(362)
1067(346)
1144(375)
906(381)
Relative size of insurancein states where banksmay sell insurance 0otherwise
324(161)
259(149)
315(165)
285(154)
272(159)
055(134)
Indicator is 1 if banksmay sell insurance inthe state
045(010)
037(011)
045(010)
038(008)
036(009)
001(012)
Relative size of insurancein states where banksmay not sell insurance0 otherwise
2 093(043)
2 090(033)
2 102(051)
2 093(048)
2 096(053)
2 004(039)
Small rm share of thenumber of rms in thestate
2 972(211)
2 614(215)
2 952(211)
2 943(230)
2 952(270)
2 1509(276)
Share of state govern-ment controlled byDemocrats
031(012)
023(011)
030(012)
027(011)
026(011)
011(013)
Indicator is 1 if state con-trolled by one party
2 004(010)
2 004(007)
2 004(007)
2 001(009)
2 002(010)
017(009)
Average yield on bankloans in the stateminus Fed funds rate mdash
023(319) mdash mdash mdash mdash
Bank failure rate mdash mdash092
(111) mdash mdash mdash
Indicator is 1 if state hasunit banking law mdash mdash mdash
018(010)
019(010)
030(009)
Indicator is 1 if statechanges bank insur-ance powers mdash mdash mdash mdash
2 008(012)
2 026(013)
Includes regionalindicators No No No No No Yes
N 637 408 637 637 637 637Log likelihood 2 374 900 2 360 2 212 2 191 1192p-value of x 2 for
regression 001 001 001 001 001 001
The hazard model is Weibull where the dependent variable is the log expected time to branchingderegulation ln (T) 5 bx 1 eAll variables are measured for each state in each year Small bank asset share isthe percent of banking assets in the state held by banks below the median size of banks in each state in eachyear Relative capital ratio is the capital to assets ratio of small banks minus that of large banks Size ofinsurance relative to banking plus insurance in the state is measured as gross state product from insurancedivided by gross state product from insurance plus banking Insurance indicator is 1 if state law permits banksto enter the insurance business Small rm share is the percent of all establishments in the state that havefewer than twenty employees Party control variable is the share of the three bodies of state governmentcontrolled by Democrats One party control indicator is 1 if the same party controlrsquos the governorship and hasmajorities in both chambers of the state legislature Average yield on bank loans equals total interest incomeon all domestic loans made by banks in the state divided by total loans Bank failure rate is the ratio of assetsin failed banks during a given state and year divided by total bank assets in that state and year Unit bankingindicator equals 1 for states with unit banking restrictions The change in insurance powers indicator equalsone if the state changed the law to permit banks to sell insurance during the sample period We divide thecountry into four regions and include regional indicators in column (5) Robust standard errors are inparentheses and denote statistically signicant at the 1 5 and 10 percent levels
QUARTERLY JOURNAL OF ECONOMICS1452
The positive and statistically signicant coefficient on smallbank share implies that a greater small bank share delaysderegulation and this effect is economically important A one-standard-deviation increase in the small bank share results in a30 percent increase in the time until deregulation or about 47years This result supports the implications of private-interesttheory concerning intraindustry rivalry and is contrary to thepublic-interest theory
The positive and statistically signicant coefficient on therelative performance of small banks in the state implies thatstates deregulate later when small banks are relatively strongnancially25 A one-standard-deviation increase in the relativecapital-to-asset ratio results in a 15 percent rise in the time untilderegulation or about 24 years This result is consistent with theprivate-interest theory but may also be consistent with a public-interest interpretation because the risk of failures (and a costlytaxpayer-funded bailout) increases as small bank nancial healthdeclines
Interindustry competition also helps explain the timing ofderegulation In states where banks can sell insurance a rela-tively large insurance sector is associated with an increase in theexpected time to deregulation26 A one-standard-deviation in-crease in the relative size of the insurance sector in those statesthat permit banks to sell insurance leads to a 22 percent increasein the time until deregulation or about 35 years27 This result isconsistent with the private-interest theory but not the public-interest theory
Deregulation occurs earlier in states where small bank-dependent rms are relatively numerous A one-standard-
25 We also allowed small bank and large bank capital-to-asset ratios to enterseparately Their coefficient estimates were nearly equal in absolute value and ofopposite signs and so the data do not reject the relative small to large bank capitalratio specication we have chosen
26 When we use the relative size of the insurance sector in the state as awhole rather than insurance relative to banking plus insurance in the state weobtain similar results When we simultaneously include both measures of therelative size of insurance the interactions of the lsquolsquoinsurance relative to bankingplus insurancersquorsquo variable continue to be statisticallysignicant but the interactionsof lsquolsquoinsurance relative to the state as a wholersquorsquo variable are not
27 We have also estimated the model with the indicator variable equal to onefor states where banks may sell insurance but without the variables measuringtheimportance of insurance relative to banking In this model the coefficient on theinsurance indicator variablemdashan estimate of the difference in the expected time toderegulation in states where banks may sell insurance relative to states wherebanks may not sell insurancemdashequals 027 and is statistically signicant at the 5percent level This suggests that branching deregulation takes about 45 yearslonger in states where banks may sell insurance
WHAT DRIVES DEREGULATION 1453
deviation increase in the share of small rms reduces the timeuntil deregulation by 18 percent or about three years This resultconcerning the interests of consumers of banking services isconsistent with both the private- and public-interest theories
Column (2) of Table III introduces the average interest rateon loans to the model as a rough proxy for the cost to bankborrowers of these regulations These data become available onlyin 1976 so our sample size shrinks to 408 observations Thecoefficient on this variable is small and not statistically signi-cant The public-interest theory would imply that the coefficientshould be negative while the implication of the economic theorydoes not have a clear sign prediction
When we add the failure rate of banks to the specication incolumn (3) its coefficient is small and statistically insignicantWe do not nd a linkage between the timing of state deregulationand statewide banking distress Given that deposit insurance isfunded federally however bank failures may raise the publicrsquosconcerns about the costs of the regulations at the national ratherthan the state level This interpretation can explain why the rateof bank failure does not affect the timing of state-level deregula-tion even though bank instability does appear to rise generallyduring the period of increasing deregulation (see Section V below)
Finally the partisan structure of the state government alsoinuences when states deregulate As expected a higher propor-tion of Democrats in the government tends to delay deregulationA one-standard-deviation rise in the share of the governmentcontrolled by Democrats slows the deregulation by about twoyears Whether the state is dominated by one party however doesnot appear to affect the timing of the deregulation
To summarize our ndings from the hazard model privateinterests appear to play an important role in the deregulatoryprocess While private interests and public interests do sometimescoincide the results on the relative share of small banks and largebanks and on the relative size of insurance where banks competeare consistent with a private-interest approach but are difficult toexplain on public-interest grounds
B Robustness Tests
Omitted Variables States begin the sample period withdifferent degrees of restrictions on bank expansion These initialconditions may proxy for omitted factors determining the political-economy equilibrium in the state that could then affect the
QUARTERLY JOURNAL OF ECONOMICS1454
subsequent timing of deregulation For instance some statesbegan the period with unit banking while other states allowedlimited branching and the formation of multibank holdingcompanies
To test whether the stringency of initial restrictions onbranching matters we rst divide the sample into states with andwithout the most extreme form of branching regulation unitbanking28 Column (4) of Table III includes an indicator variableequal to one if the state began the period with unit bankingrestrictions The coefficient on this variable is positive andstatistically signicant but the other results remain virtuallyunchanged Next we include an indicator variable that is one ifthe state permitted multibank holding companies (MBHCs) toexist at the beginning of our sample period Thirty-four statesbegan the period allowing MBHCs The coefficient estimate forthis indicator variable (not reported) is small and not statisticallysignicant and its inclusion does not change the effects of theother variables
Eleven states expand bank powers to include the sale ofinsurance products during our sample period This could proxy foromitted factors about the relative strength of the insuranceindustry in the state In column (5) we include an indicatorvariable that is one if the state changes the law during our sampleperiod to permit banks to sell insurance business and zerootherwise The coefficient estimate is small and statisticallyinsignicant and does not affect the other results29
In column (6) we introduce four regional indicator variablesNorth South Midwest and West30 The coefficients on small bankshare the relative capital-to-assets ratio and small rm sharechange little Although the coefficient estimate still suggests that
28 We classify states that prohibited branching but permitted banks toestablish facilities as unit banking states The unit banking indicator is one for thefollowing states CO AR FL IL IA KS MN MO MT NE ND OK TX WI WVand WY
29 We conducted a likelihood ratio test of whether the unit banking statesthe MBHC states and the states that changed their regulations of bank insurancepowers could be pooled with the other states In all three cases we could not rejectpooling of the data
30 Since intrastate deregulation generally preceded interstate deregulationand the latter typically took the form of regional interstate compacts thepossibility of participating in a regional interstate banking compact could haveinuenced the decision to deregulate intrastate branching Our denitions of theregions are region 1 (South) contains AL AR DC FL GA KY LA MS NC OKSC TN TX and VA region 2 (Northeast) contains CT MA MD ME NH NJ NYPA RI VT and WV region 3 (Midwest) contains IA IL IN KS MI MN MO NEND OH SD and WI region 4 (West) contains the other states
WHAT DRIVES DEREGULATION 1455
in states where banks can sell insurance a larger relative size ofthe insurance sector delays deregulation the effect is no longerstatistically signicant The coefficient on the indicator variablefor whether a state expands bank powers to permit insurancesales during the sample period however now becomes statisti-cally signicantly negative This implies that deregulation occursroughly 41 years earlier in states where the insurance lobby maybe relatively weak While the inclusion of the regional indicatorsdoes not affect the other results the results on insurance doappear to be sensitive to the specication of the hazard model31
Estimation Technique To check the robustness of using ahazard model and to investigate whether the time-series or thecross-sectional variation in each of the factors accounts for thehazard results we estimate two simple linear probability models(neither are reported) First we run a pooled time-series cross-section OLS regression using the 637 observations from thehazard model The dependent variable equals one in the year ofderegulation and zero otherwise We include the same set ofregressors that were used in the hazard model and add state-levelxed effects This specication thus removes the cross-statevariation from the independent variables and focuses on theirmovements over time In this model the signs of the coefficientson the interest group factors are consistent with the results fromthe hazard model and the coefficients for the small bank relativecapital ratio and for the relative size of insurance where states cansell insurance are statistically signicant
In the second model we eliminate the time-series variationby averaging each of the explanatory variables over time We thushave one observation per state that had branching restrictions asof 1970 (N 5 39) The dependent variable equals the inverse of thenumber of years until deregulation which represents an estimateof the statersquos probability of deregulating in a year under theassumption that this probability is constant over time and eachyear provides an independent observation32 Although we haveonly 39 observations the coefficients on the small bank marketshare small bank relative capital ratio and small rm share
31 As further robustness check we also tried a population density variablewhich Abrams and Settle [1993] found relevant for regulatory change during the1920s and 1930s but its coefficient was not statistically signicant and did notaffect the other results
32 For instance the dependent variable would take the value of 01 for astate that deregulated in 1980 For the three states that did not deregulate thedependent variable equals 0
QUARTERLY JOURNAL OF ECONOMICS1456
variables are statistically signicant The signs of the effects ofthe insurance variables are the same as those reported in thehazard model but not statistically signicant
The hazard model and linear probability models thus producevery similar results The estimates from the latter suggest thatthe effects for the small bank market share and small rm sharevariables are driven primarily by cross-state variation the effectof the insurance variable is driven primarily by time-seriesvariation and the small bank capital effect is driven by bothcross-sectional and time-series variation
Ex Post Consequences of Deregulation Finally to check theplausibility of our interpretation of the hazard results we con-sider whether the ex post consequences of deregulation areconsistent with the ex ante lobbying positions we attribute to eachinterest group (see Kroszner and Strahan [1998] for details) Wend that small banks lose market share following deregulationand in states where banks can enter the insurance business theinsurance sector shrinks relative to the banking sector followingderegulation Borrowers also benet from deregulation throughlower average interest rates on loans These ndings support theprivate-interest interpretation of the coefficients in the hazardmodel groups that will benet lobby to speed deregulation andthose that will be harmed lobby to slow it
C Voting on Interstate Branching Deregulation in the U SHouse of Representatives
We now examine whether the forces driving intrastate branch-ing deregulation also drive interstate deregulation Financialservices interests are active contributors and lobbyists in Washing-ton Their political action committees constitute the largest groupof contributors to legislators providing nearly 20 percent of totalcongressional campaign contributions [Makinson 1992] and muchof their lobbying effort involves competition among rival interestswithin nancial services (see Kroszner and Stratmann [1998])
As noted above after virtually all states adopted intra- andinterstate branching deregulation the 1994 Riegle-Neal Act re-pealed the 1927 McFadden Act to eliminate all barriers tointerstate banking and branching by 1997 The key votes concern-ing the Riegle-Neal Act were either voice votes or extremelylopsided so we could not estimate a voting model from them Anumber of bills and amendments related to interstate branching
WHAT DRIVES DEREGULATION 1457
had been debated in Congress during the years prior to thepassage of the Riegle-Neal Act but a search of the weekly BNABanking Reporter and the Congressional Record produced onlyone roll-call vote related to interstate branching that was notlopsided This vote occurred in the House on November 14 1991and concerned an amendment sponsored by Wylie (R-OH) andNeal (D-NC) to introduce interstate banking and branchingderegulation to a nancial services reform package33 Althoughthe amendment passed by 210 to 208 the bill to which it wasattached subsequently was defeated
To check for the inuence of the factors we considered in thestate-level reforms we examine both the sponsorship of interstatebanking legislation and voting on the amendment The sponsorsof the Wylie-Neal amendment are from states with low small bankshares 004 in Ohio (Wylie) and 002 in North Carolina (Neal) Incontrast the sample mean in 1991 is 008 (median 5 007)Michigan home state of the Senatersquos sponsor of the 1994 Riegle-Neal Act also had relatively low small bank strength (small bankshare of 005)
Table IV reports estimates of a probit model where thedependent variable equals one if the legislator voted in favor ofthe amendment and zero otherwise The explanatory variablesare the same as those in the hazard model except that we use aparty affiliation indicator variable equal to one for Democrats andzero for Republicans as our proxy for political factors Thecoefficients in Table IV are the marginal effects (lsquolsquoslopesrsquorsquo) of aone-unit change of each variable on the probability that alegislator will vote for the amendment34 Note that a forcefavoring deregulation has a negative coefficient in the hazardmodel (Table III) but a positive coefficient in the probit model
Consistent with the state-level deregulation process legisla-tors are more likely to support the amendment if their states havea relatively low share of small banks As in the hazard model thefraction of small banks is the most important interest groupinuence on a legislatorrsquos voting decision A one-standard-deviation increase in small banksrsquo market share (from the mean)is associated with a decline in the probability of voting in favor of
33 The Wylie-Neal amendment also included provisions limiting certaininsurance and real estate powers of national banks [Congressional RecordNovember 14 1991 pp 10239ndash10242]
34 Since we have multiple legislators from each state we adjust the standarderrors to correct for the lack of independence among observations clustered in thesame state
QUARTERLY JOURNAL OF ECONOMICS1458
TAB
LE
IVM
AR
GIN
AL
EF
FE
CT
SF
RO
MA
PR
OB
ITM
OD
EL
OF
TH
EIN
FL
UE
NC
EO
FP
OL
ITIC
AL
EC
ON
OM
YF
AC
TO
RS
ON
TH
EH
OU
SE
VO
TE
ON
TH
EW
YL
IE-N
EA
L
AM
EN
DM
EN
TT
OP
ER
MIT
INT
ER
ST
AT
EB
AN
KB
RA
NC
HIN
G
NO
VE
MB
ER
14
1991
(1)
(2)
(3)
(4)
(5)
(6)
Smal
lban
kas
set
shar
eof
allb
anki
ng
asse
tsin
stat
e2
417
(12
4)2
454
(12
2)2
407
(12
0)2
443
(14
7)2
450
(14
9)2
490
(14
7)C
apit
alra
tio
ofsm
allb
anks
rela
tive
tola
rge
inth
est
ate
180
(37
5)2
68(3
42)
256
(37
5)1
69(3
78)
153
(38
0)2
45(4
25)
Rel
ativ
esi
zeof
insu
ranc
ein
stat
esw
here
bank
sm
ayse
llin
sura
nce
0ot
herw
ise
22
34
(10
2)2
249
(1
14)
22
33
(10
1)2
249
(1
04)
22
51
(10
2)2
154
(11
1)In
dica
tor
is1
ifba
nks
may
sell
insu
ranc
ein
the
stat
e0
79
(0
06)
074
(00
6)0
69
(0
06)
070
(00
6)0
70
(0
06)
067
(00
7)R
elat
ive
size
ofin
sura
nce
inst
ates
whe
reba
nks
may
not
sell
insu
ranc
e0
oth
erw
ise
180
(06
6)2
42
(0
72)
154
(0
67)
173
(06
6)1
72
(0
67)
203
(0
91)
Smal
lrm
shar
eof
the
num
ber
ofr
ms
inth
est
ate
21
23(0
92)
20
83(0
89)
21
62(1
07)
21
32(0
95)
21
30(1
00)
20
94(1
16)
Indi
cato
ris
1if
Rep
rese
ntat
ive
isa
Dem
ocra
t2
035
(00
7)2
035
(00
7)2
036
(00
7)2
035
(00
7)2
035
(00
7)2
036
(00
7)A
vera
geyi
eld
onba
nklo
ans
inth
est
ate
min
usF
edfu
nds
rate
mdash10
01
(51
5)mdash
mdashmdash
mdash
Ave
rage
fail
ure
rate
198
5to
1991
mdashmdash
228
(14
1)mdash
mdashmdash
Indi
cato
ris
1if
stat
eh
asun
itba
nkin
gla
wmdash
mdashmdash
004
(00
7)0
04(0
07)
20
01(0
08)
Indi
cato
ris
1if
stat
ech
ange
sba
nk
insu
ran
cepo
wer
smdash
mdashmdash
mdash2
001
(00
7)2
003
(00
7)In
clud
esre
gion
alin
dica
tors
N
oN
oN
oN
oN
oYe
sP
seud
o-R
20
170
180
170
170
170
18L
ogli
keli
hood
224
024
223
826
223
951
224
013
224
010
223
808
p-va
lue
ofx2
for
regr
essi
on
001
0
01
001
0
01
001
0
01
Th
ede
pen
den
tvar
iabl
eis
one
ifth
ele
gisl
ator
vote
sfo
rth
eam
endm
enta
nd
zero
ifag
ains
tT
he
repo
rted
coef
fici
ents
are
the
effe
cts
ofa
uni
tch
ange
ofth
ein
depe
nde
ntva
riab
le(f
rom
the
mea
n)on
the
prob
abili
tyof
voti
ng
infa
vor
ofth
eam
endm
ent
All
vari
able
sar
em
easu
red
byst
ate
N5
418
the
num
ber
ofR
epre
sent
ativ
esvo
tin
gon
the
amen
dmen
tS
mal
lban
kas
sets
har
eis
the
perc
ent
ofba
nki
ng
asse
tsin
the
stat
eh
eld
byba
nks
belo
wth
em
edia
nsi
zeof
ban
kin
each
stat
ein
each
year
Rel
ativ
eca
pita
lrat
iois
the
capi
talt
oas
sets
rati
oof
smal
lban
ksm
inu
sth
atof
larg
eba
nks
Siz
eof
insu
ran
cere
lati
veto
bank
ing
plu
sin
sura
nce
inth
est
ate
ism
easu
red
asgr
oss
stat
epr
oduc
tfro
min
sura
nce
divi
ded
bygr
oss
stat
epr
odu
ctfr
omin
sura
nce
plu
sba
nkin
gIn
dica
tor
ison
eif
stat
ela
wpe
rmit
sba
nks
toen
ter
the
insu
ran
cebu
sin
ess
Sm
all
rmsh
are
isth
epe
rcen
tofa
lles
tabl
ishm
ents
inth
est
ate
that
hav
efe
wer
than
twen
tyem
ploy
ees
Par
tyin
dica
tor
ison
eif
the
Rep
rese
nta
tive
isa
Dem
ocra
tA
vera
geyi
eld
onba
nklo
ans
equa
lsto
tali
nter
est
inco
me
onal
ldom
esti
clo
ans
mad
eby
ban
ksin
the
stat
edi
vide
dby
tota
lloa
ns
The
aver
age
bank
failu
rera
teis
the
rati
oof
asse
tsin
faile
dba
nks
duri
ng
agi
ven
stat
edi
vide
dby
tota
lban
kas
sets
inth
atst
ate
aver
aged
over
the
year
s19
85to
1991
Un
itba
nki
ng
indi
cato
req
ual
son
efo
rst
ates
wit
hun
itba
nkin
gre
stri
ctio
nsT
hech
ange
inin
sura
nce
pow
ers
indi
cato
req
ual
son
eif
the
stat
ech
ange
dth
ela
wto
perm
itba
nks
tose
llin
sura
nce
duri
ngth
esa
mpl
epe
riod
We
divi
deth
eco
unt
ryin
tofo
urre
gion
san
din
clu
dere
gion
alin
dica
tors
inco
lum
n(5
)R
obu
stst
anda
rder
rors
are
inpa
ren
thes
es
an
d
den
ote
stat
isti
cally
sign
ica
nta
tth
e1
5an
d10
perc
entl
evel
s
WHAT DRIVES DEREGULATION 1459
branching of approximately 17 percent The relative capital-to-assets ratio of small and large banks however does not have astatistically signicant effect in the probit model
The impact of rival interests outside of banking is alsoconsistent with intrastate deregulation results Where banks cansell insurance legislators from states with larger insurancesectors relative to banking are less likely to vote for the amend-ment35 A one-standard-deviation increase in the relative size ofinsurance in those states which permit banks to sell insurancedecreases the probability that a legislator will favor the amend-ment by about 13 percent Turning to consumer interests thecoefficient on small rm share is not statistically signicant butthe coefficient on interest rates in the state is A one-standard-deviation increase in the average interest rate on loans raises theprobability that a legislator will support the amendment byroughly 8 percent The coefficient estimate for the bankingdistress variable measured as the average bank failure rate from1985 to 1991 is small and statistically signicant Overall theprobit analysis of the vote on national branching deregulationsupports the private-interest theory of deregulation and providesa consistency check that the importance of interests operating onthe state legislatures are very similar to those operating at thefederal level
V WHY DID BRANCHING DEREGULATION BEGIN DURING THE
1970S SHOCKS TO THE EQUILIBRIUM SUPPORTING BANK
BRANCHING RESTRICTIONS
A complete explanation of regulatory exit should account forwhy deregulation begins in the 1970s as well as the timing of thestate-by-state reforms [Kane 1996] In this section we describebroad technological legal and economic shocks that altered thepolitical-economy equilibrium which had kept antibranching regu-lations little changed for at least 30 years
Beginning in the 1970s three innovations reduced the valueto the protected banks of local geographic monopolies (see TableV) First automatic teller machines (ATMs) helped to erode thegeographic ties between customers and banks Second checkablemoney market mutual funds and the Merrill Lynch Cash Manage-ment Account demonstrated that banking by mail and telephone
35 The positive and statistically signicant estimates on the other insurancevariables may reect insurance industry support for the amendmentrsquos provisionslimiting banksrsquo insurance powers
QUARTERLY JOURNAL OF ECONOMICS1460
provided a convenient alternative to local banks36 Third techno-logical innovation and deregulation reduced transportation andcommunication costs particularly since the 1970s thereby lower-ing the costs for customers to use distant banks By increasing theelasticity of deposits supplied to banks these innovations reducedthe value of geographical restrictions to their traditional benecia-ries and thereby reduced their incentive to ght to maintain them[Peltzman 1976]
On the lending side increasing sophistication of credit-scoring techniques following innovations in information process-ing nancial theory and the development of large credit databases began to diminish the value of knowledge that localbankers had about the risks of borrowers in the community As a
36 Regulation Q which limited the interest rates that banks could pay ondeposits may have helped to drive depositors away from banks when the gapbetween market rates and deposit ceilings grew during the 1970s The eliminationof interest rate ceilings on large denomination certicates of deposit during the1970s appears to have hurt smaller and retail-oriented banks relative to largerwholesale banks [James 1983]
TABLE VBROAD TRENDS IN COMMERCIAL BANKING 1950ndash1995
YearNumberof ATMs
Domesticbank
deposits(billions)
Moneymarketmutual
fund(billions)
Percent ofdeposits 1
moneyfunds heldby banks
Smallbanksrsquo
percent ofbankingassets
Averagenumber of
bankfailures
(1) (2) (3) (4) (5) (6)1950 0 $ 154 $ 0 100 NA 41955 0 191 0 100 NA 31960 0 228 0 100 24 21965 0 330 0 100 20 41970 0 479 0 100 18 61975 9750 775 4 99 18 61980 18500 1182 76 94 17 101985 61117 1787 242 88 14 601990 80156 2339 493 83 11 1791995 122706 2552 745 77 8 61
Column (1) ATM gures are from Bank Network News The EFT Network Data Book (New YorkFaulkner and Gray Inc) The 1975 gure was unavailable 9750 is the number of ATMs in 1978 the rst yearfor which complete data are available Columns (2)ndash(4) Banksrsquo domestic deposits are from the Reports ofIncome and Condition money market mutual funds are from the Flow of Funds Data on all bank depositsforeign plus domestic are only available beginning in 1970 The trend in banksrsquo share (column (4)) is the sameusing total deposits instead of domestic deposits Column (5) Percent of banking assets held by small bankswhere a small bank is dened as a commercial bank with less than $100 million in assets in 1994 dollarsThese data are based on the Reports of Income and Condition Data on small banks are not available before1960 Column (6) Five-year average number of bank failures where the nal year is indicated in the rstcolumn These data are from FDIC Annual Report and the Quarterly Banking Prole
WHAT DRIVES DEREGULATION 1461
result of these innovations a national market developed forresidential mortgages credit card receivables have been securi-tized and bank lending to small business now relies less on thejudgment of loan officers and more on standardized scoringmodels
These changes have increased the potential protability forlarge banks to enter what had been the core of small bankactivities Large banksrsquo incentive to increase their lobbying pres-sure to be able to expand into these markets has thus beenincreasing over time In fact small banksrsquo market share began todecline even prior to the branching deregulation (Table V) As thevalue of a local banking relationships declined small rms thatwere the main borrowers from the small banks also probablybecame more likely to favor the entry of large banks into localmarkets With the deadweight costs of preventing large bankentry rising the private-interest theory predicts that small localbanks would become less likely able to maintain the branchingrestrictions [Becker 1983] Deregulation that reduces deadweightcosts of regulation is also consistent with the public-interesttheory
Kane [1996] argues that another major shock to the oldequilibrium was the increasing public awareness of the costlinessof having government-insured but (geographically) undiversiednancial institutions In the late 1970s the failure rate of banksbegan to rise (Table V) In the 1980s the Savings and Loan crisisand taxpayer bailout further heightened public awareness aboutthe costs of restrictions that make depository institutions morelikely to require infusions of taxpayer funds The failures thusmay have heightened public support for branching deregulation(see also Abrams and Settle [1993]) As noted above we did notnd any evidence that banking failures or distress in a stateaffected the speed with which the state deregulated but distressstill may be a common factor affecting all states
These technological economic and legal shocks generatedconditions that changed the long-standing balance favoring theantibranching forces The marginal value of lobbying to repealbranching restrictions increased just as the relative value to thesmall banks of maintaining branching restrictions was decliningThese nationwide shocks are common factors across states andthey are consistent with the positive duration dependence associ-ated with state-level deregulation that we demonstrated in FigureII In addition a broad change in lsquolsquoideologyrsquorsquo against government
QUARTERLY JOURNAL OF ECONOMICS1462
regulation could explain the positive duration dependence duringthe sample period
VI CONCLUSIONS
The private-interest theory of regulation can account for thepattern of bank branching deregulation of the 1970s and 1980sBeneciaries of branching regulation had supported a coalitionfavoring geographical restrictions despite their costs to consum-ers of nancial services Innovations that began in the 1970saltered the value of the restrictions to the affected parties and theresulting competition among interest groups can explain thesubsequent deregulation While some of our results are alsoconsistent with the public interest theorymdashfor example deregula-tion occurs earlier when small banks are in a relatively weaknancial positionmdashother results particularly evidence on theimportance of rivalries between small and large banks andbetween banking and insurance are difficult to explain with thepublic-interest approach We also nd that political-institutionalfactors affect deregulation although these variables may act asproxies for unmeasured economic interests The greater success ofthe private-interest theory to explain deregulation here than inprevious work may be due to the richer cross-sectional andtime-series variation in branching deregulation that can beexploited with a hazard model Future empirical studies ofendogenous deregulation thus may be most fruitful where changehas occurred across states over time such as in franchisinginsurance and public utilities (eg Joskow [1996])
Technological and nancial innovations will continue to erodethe benets to any interest group of maintaining regulatorybarriers in nancial services These forces are likely to bringabout reforms both domestically for example through legislationthat would increase bank powers (see Kroszner [1996] Krosznerand Rajan [1994 1997] and Kroszner and Stratmann [1998]) andinternationally for example through the extension of nancialservices provisions of NAFTAto reduce geographic barriers acrosscountries [Kroszner 1997]
APPENDIX EFFECTS OF DIFFERENT TYPES OF BRANCHING
DEREGULATION ON ECONOMIC GROWTH BANKING STRUCTUREAND BANK EFFICIENCY
We measure the importance of alternative forms of deregula-tion by their consequences for the economic growth and the
WHAT DRIVES DEREGULATION 1463
structure and efficiency of the banking sector across the states(see Jayaratne and Strahan [1996 1998]) To do so we run anOLS panel regression relating indicators for each type of deregula-tion to potential consequences Our dependent variables includeproxies for economic growth (growth in personal income) bankstructure (the log of the number of banking companies and thenumber of branches per banking company) and efficiency (the logof total noninterest expenses and the ratio of loan loss provisionsto total loans) The four deregulation indicators (MBHC branch-ing through merger and acquisition unrestricted statewide branch-ing and interstate banking) are equal to one in the years afterderegulation and zero before If a state deregulated before thesample period begins in 1970 the indicator is one throughout ifthe state does not deregulate then the indicator is zero through-out All regressions include state xed effects and time effects
Table VI shows the results of including the set of four
TABLE VIMULTIVARIATE REGRESSIONS OF THE EFFECTS OF DIFFERENT TYPES OF BRANCHING
DEREGULATION ON ECONOMIC GROWTH BANKING STRUCTUREAND BANK EFFICIENCY
Dependent variable
Growthin
personalincome
Log of of
bankingcompanies
Branchesper
bankingcompany
Log ofnon-
interestexpenses
Loan lossprovisions
(loans 1leases)
Merger amp acquisitionbranchingindicator
0013(0004)
2 0162(0037)
1146(0453)
2 0107(0035)
2 0007(0001)
Unrestricted state-wide branchingindicator
2 0008(0005)
0086(0059)
2 0043(0574)
0070(0037)
0005(0002)
Interstate bankingindicator
0006(0005)
2 0052(0057)
1064(0511)
0005(0030)
2 0001(0001)
Multibank holdingcompany indicator
2 0005(0006)
2 0096(0052)
2 0364(0529)
2 0095(0036)
2 0001(0001)
Dependent variablemean [median]
0086[0082]
4881[5136]
6649[5055]
13812[13803]
0009[0006]
This table contains regressions of state-level economic growth and state-level measures of the structureand efficiency of the banking industry Each dependent variable is regressed on a set of state indicatorvariables a set of year indicator variables and the set of four deregulation indicators these deregulationindicators are equal to one after deregulation and zero before If a state was always regulated the indicator iszero throughout if it was always deregulated the indicator is one throughout Sample period is 1975 to 1991and Delaware and South Dakota are dropped so N 5 833 Noninterest expenses become available only in1984 so N 5 392 The coefficient on the deregulation indicators is reported with robust standard errors inparentheses and denote statistically signicant at the 1 5 and 10 percent levels
QUARTERLY JOURNAL OF ECONOMICS1464
deregulation indicators simultaneously The branching by mergerand acquisition indicator is statistically signicant and of theexpected sign in all ve of the equations and larger in absolutevalue than any of the other coefficients on the indicators for theother types of deregulation Since there may be multicollinearityamong our indicators we also ran univariate specications whichincluded each indicator separately and the results are verysimilar These results imply that branching by merger andacquisition is the most important type of deregulation measuredby its consequences for economic growth banking structure andbank efficiency See Kroszner and Strahan [1998] for more details
UNIVERSITY OF CHICAGO GRADUATE SCHOOL OF BUSINESS AND NATIONAL BUREAU OF
ECONOMIC RESEARCH
FEDERAL RESERVE BANK OF NEW YORK RESEARCH amp MARKET ANALYSIS GROUP
REFERENCES
Abrams Burton A and Russell F Settle lsquolsquoPressure-Group Inuence and Institu-tional Change Branch-Banking Legislation during the Great DepressionrsquorsquoPublic Choice LXXVII (1993) 687ndash705
Amel Dean lsquolsquoState Laws Affecting the Geographic Expansion of CommercialBanksrsquorsquo unpublished manuscript Board of Governors of the Federal ReserveSystem 1993
Becker Gary S lsquolsquoA Theory of Competition among Pressure Groups for PoliticalInuencersquorsquo Quarterly Journal of Economics XCVIII (1983) 371ndash400
Berger Allen Anil Kashyap Joseph Scalise lsquolsquoThe Transformation of the U SBanking Industry What a Long Strange Trip Itrsquos Beenrsquorsquo Brookings Papers onEconomic Activity 2 (1995) 55ndash218
Berger Allen Anthony Saunders Joseph Scalise and Gregory Udell lsquolsquoThe Effectsof Bank Mergers and Acquisitions on Small Business Lendingrsquorsquo Journal ofFinancial Economics L (1998) 187ndash229
Brennan Geoffrey and James Buchanan lsquolsquoTowards a Tax Constitution forLeviathanrsquorsquo Journal of Public Economics VIII (1977) 255ndash273
Brickley James A and Christopher M James lsquolsquoThe Takeover Market CorporateBoard Composition and Ownership Structure The Case of Bankingrsquorsquo Journalof Law and Economics XXX (1987) 161ndash180
Cole Rebel and John Wolken lsquolsquoFinancial Services Used by Small BusinessesEvidence from the 1993 Survey of Small Business Financesrsquorsquo Federal ReserveBulletin LXXXI (1994) 629ndash676
Dixit Avinash The Making of Economic Policy A Transactions-Cost PoliticsPerspective (Cambridge MA MIT Press 1996)
Economides Nicolas R Glenn Hubbard and Darius Palia lsquolsquoThe Political Economyof Branching Restrictions and Deposit Insurancersquorsquo Journal of Law andEconomics XXIX (1996) 667ndash704
Flannery Mark lsquolsquoThe Social Costs of Unit Banking Restrictionsrsquorsquo Journal ofMonetary Economics XIII (1984) 237ndash249
Greene William H Econometric Analysis third edition (Upper Saddle River NJPrentice-Hall 1997)
Gunther Jeffery lsquolsquoRegional Capital Imbalances and the Removal of InterstateBanking Restrictionsrsquorsquo Economics Letters XLIV (1994) 439ndash442 lsquolsquoThe Erosion of Intrastate Branch Banking Restrictionsrsquorsquo unpublishedmanuscript Federal Reserve Bank of Dallas 1996
Irwin Douglas and Randall Kroszner lsquolsquoInterests Institutions and Ideology inSecuring Policy Change The Republican Conversion to Trade Liberalizationafter Smoot-Hawleyrsquorsquo Journal of Law and Economics XLII (1999) 643ndash673
WHAT DRIVES DEREGULATION 1465
James Christopher lsquolsquoAn Analysis of Intra-Industry Differences in the Effect ofRegulation The Case of Deposit Rate Ceilingsrsquorsquo Journal of Monetary Econom-ics XII (1983) 417ndash442
Jarrell Gregg lsquolsquoChange at the Exchange The Causes and Effects of DeregulationrsquorsquoJournal of Law and Economics XXVII (1984) 273ndash312
Jayaratne Jith and Philip E Strahan lsquolsquoThe Finance-Growth Nexus Evidencefrom Bank Branch Deregulationrsquorsquo Quarterly Journal of Economics CI (1996)639ndash670
Jayaratne Jith and Philip E Strahan lsquolsquoEntry Restrictions Industry Evolutionand Dynamic Efficiency Evidence from Commercial Bankingrsquorsquo Journal of Lawand Economics XLIX (1998) 239ndash274
Joskow Paul lsquolsquoComment on lsquoPower Struggles Explaining Deregulatory Reformsin Electricity Marketsrsquo rsquorsquo Brookings Papers on Economic Activity Microeconom-ics (1996) 251ndash264
Joskow Paul and Roger Noll lsquolsquoRegulation in Theory and Practice An Overviewrsquorsquoin Studies in Public Regulation Gary Fromm ed (Cambridge MA MITPress 1981) pp 1ndash65
Kahn Matthew E and John G Matsusaka lsquolsquoDemand for Environmental GoodsEvidence from Voting Patterns on California Initiativesrsquorsquo Journal of Law andEconomics XL (1997) 137ndash173
Kalbeisch J and R Prentice The Statistical Analysis of Failure Time Data (NewYork John Wiley amp Sons 1980)
Kane Edward J lsquolsquoDe Jure Interstate Banking Why Only NowrsquorsquoJournal of MoneyCredit and Banking XXVIII (1996) 141ndash161
Kiefer Nicholas lsquolsquoEconomic Duration Data and Hazard Functionsrsquorsquo Journal ofEconomic Literature XXVI (1988) 646ndash679
Kroszner Randall S lsquolsquoThe Evolution of Universal Banking and Its Regulation inTwentieth Century Americarsquorsquo in Universal Banking Financial System DesignReconsidered Anthony Saunders and Ingo Walter eds (New York IrwinProfessional Publishers 1996) pp 70ndash99 lsquolsquoThe Political Economy of Banking and Financial Regulation in the U Srsquorsquo inThe Banking and Financial Structure in the NAFTA Countries and ChileGeorge M von Furstenberg ed (Boston Kluwer Academic Publishers 1997)pp 200ndash213
Kroszner Randall S and Raghuram G Rajan lsquolsquoIs the Glass-Steagall ActJustied A Study of the U S Experience with Universal Banking before1933rsquorsquo American Economic Review LXXXIV (1994) 810ndash832
Kroszner Randall S and Raghuram G Rajan lsquolsquoOrganization Structure andCredibility Evidence from Bank Securities Activities before the Glass-Steagall Actrsquorsquo Journal of Monetary Economics XXXIX (1997) 475ndash516
Kroszner Randall S and Philip E Strahan lsquolsquoRegulatory Incentives and the ThriftCrisis Dividends Mutual-to-Stock Conversions and Financial DistressrsquorsquoJournal of Finance LI (1996) 1285ndash1320
Kroszner Randall S and Philip E Strahan lsquolsquoWhat Drives Deregulation Econom-ics and Politics of the Relaxation of Bank Branching Restrictionsrsquorsquo NBERWorking Paper No 6637 1998
Kroszner Randall S and Thomas Stratmann lsquolsquoInterest Group Competition andthe Organization of Congress Theory and Evidence from Financial ServicesPolitical Action Committeesrsquorsquo American Economic Review LXXXVIII (1998)pp 1163ndash1187
Kroszner Randall S and Thomas Stratmann lsquolsquoDoes Political Ambiguity PayCorporate Campaign Contributions and Rewards to Legislator Reputationrsquorsquounpublishedmanuscript University of Chicago Graduate School of Business1999
Lin D Y and L J Wei lsquolsquoThe Robust Inference for the Cox Proportional HazardsModel Journal of the American Statistical Association LXXXIV (1989)1074ndash1078
Macey Jonathan R and Geoffrey P Miller Banking Law and Regulation (BostonLittle Brown and Company 1992)
Makinson Larry Open Secrets The Encyclopedia of Congressional Money andPolitics (Washington DC Congressional Quarterly 1992)
Niskanen WilliamBureaucracy and Representative Government (Chicago Aldine-Atherton 1971)
QUARTERLY JOURNAL OF ECONOMICS1466
Noll Roger lsquolsquoComment on Peltzmanrsquorsquo Brookings Papers Microeconomics SpecialIssue (1989) 48ndash58
North Douglass lsquolsquoA Transactions Cost Theory of Politicsrsquorsquo Journal of TheoreticalPolitics II (1990) 355ndash367
Peltzman Sam lsquolsquoToward a More General Theory of Regulationrsquorsquo Journal of Lawand Economics XIX (1976) 109ndash148 lsquolsquoConstituent Interest and Congressional Votingrsquorsquo Journal of Law andEconomics XXVII (1984) 181ndash210 lsquolsquoThe Economic Theory of Regulation after a Decade of DeregulationrsquorsquoBrookings Papers Microeconomics Special Issue (1989) 1ndash41
Peterson Mitchell and Raghuram Rajan lsquolsquoThe Benets of Lending RelationshipsEvidence from Small Business Datarsquorsquo Journal of Finance XLIX (1994) 3ndash38
Poole Keith and Howard Rosenthal Congress A Political-Economic History ofRoll Call Voting (Oxford Oxford University Press 1997)
Stata Stata Reference Manual Release 5 (College Station TX Stata Press 1997)Stigler George J lsquolsquoThe Theory of Economic Regulationrsquorsquo Bell Journal of Economics
and Management Science II (1971) 3ndash21Stigler George J Chicago Studies in the Political Economy (Chicago University
of Chicago Press 1988)Strahan Philip E and James Weston lsquolsquoSmall Business Lending and the Changing
Structure of the Banking Industryrsquorsquo Journal of Banking and Finance XXII(1998) 821ndash845
Sylla Richard John Legler and John Wallis lsquolsquoBanks and State Public Finance inthe New Republic The United States 1790ndash1860rsquorsquo Journal of EconomicHistory XLVII (1987) 391ndash403
White Eugene The Regulation and Reform of the American Banking System1900ndash1929 (Princeton Princeton University Press 1983)
Winston Clifford lsquolsquoEconomic Deregulation Days of Reckoning for Microecono-mistsrsquorsquo Journal of Economic Literature XXXI (1993) 1263ndash1289
WHAT DRIVES DEREGULATION 1467
[1988]) it has been less effective in explaining the removal of suchregulations [Peltzman 1989 Noll 1989] In contrast the public-interest theory can account for welfare-enhancing deregulationbut not for the origin of regulations that reduced competition andthat had few if any welfare benets
This paper examines the elimination of restrictions on bankbranching since the 1970s to assess the relative explanatorypower of these theories of regulatory change Unlike most otherrecent episodes of deregulation that occurred at a nationallevelmdashsuch as in railroads trucking airlines long-distance tele-communications securities brokerage petroleum and naturalgasmdashbank branching regulation operated on a state-by-statebasis and deregulation has taken place gradually across thestates Branching deregulation thus provides a much greatersource of cross-sectional and time-series variation than othertypes of deregulation
To investigate what drives deregulation we use a hazardmodel to explain the timing of intrastate branching deregulationWe incorporate proxies for interest-group public-interest andpolitical-institutional factors to understand how changes in thesevariables over time and across states affect the likelihood ofderegulation An additional benet of focusing on branchingderegulation is that the state-by-state reform culminates in thepassage of the 1994 Riegle-Neal Interstate Banking and Branch-ing Efficiency Act that effectively eliminated branching restric-tions nationwide This allows us to determine whether the samefactors driving state-level deregulation can also explain congres-sional voting on federal repeal of interstate branching restrictions
We nd that deregulation occurs earlier in states with fewersmall banks in states where small banks are nancially weakerand in states with more small presumably bank-dependentrms Also a larger insurance industry delays deregulation whenbanks may compete in the sale of insurance products Interestgroup factors related to the relative strength of potential winners(large banks and small rms) and losers (small banks and therival insurance rms) thus can explain the timing of branchingderegulation across states The same interest group variables alsocan explain the voting pattern of legislators in the U S House ofRepresentatives on interstate banking deregulation We thenexamine broad technological legal and nancial innovations thataltered the costs and benets of the regulations and can explainwhy deregulation began in the 1970s rather than earlier [Kane
QUARTERLY JOURNAL OF ECONOMICS1438
1996] While some of our ndings can be rationalized on public-interest grounds the private-interest theory provides a simpleway to understand these results as a whole1
II THE ORIGINS AND DEMISE OF GEOGRAPHICAL
RESTRICTIONS ON BANKING
A Origins in Public Financing Strategies
After the United States Constitution prevented the statesfrom issuing at money and from taxing interstate commercestates used their powers over banks to generate substantialrevenues [Sylla Legler and Wallis 1987]2 To enter the businessone had to obtain a bank charter States received fees for grantingcharters and they often owned or purchased shares in banks orlevied taxes on banks To enhance these revenues each state hadan interest in restricting competition among banks and many ofthe restrictions on the geographical expansion of banks originatein this period3 Since states received no charter fees from banksincorporated in other states the states prohibited out-of-statebanks from operating in their territoriesmdashhence the origin of theprohibition on interstate banking4
Legislatures also often restricted intrastate expansion Stateswould grant a charter for a specic location or limit bank branchesto that city or country By adopting branching restrictions statescreated a series of local monopolies from which they could extractpart of the rents Some state legislatures even passed lsquolsquounitbankingrsquorsquo laws that prevented a bank from having any branchesSuch regulations naturally produce beneciaries who are loatheto give up their protections and privileges Benets tend to be
1 Jarrell [1984] tests the private interest theory of regulatory exit using theend of xed commissions at the New York Stock Exchange Also see Peltzman[1989] and Noll [1989] Kanersquos [1996] lsquolsquoregulatory dialecticrsquorsquo emphasizes technologi-cal change interacting with private interests
2 During the rst third of the nineteenth century for example the bank-related share of total state revenues exceeded 10 percent in a dozen states InMassachusetts and Delaware a majority of total state revenue was bank-related
3 Noll [1989] has characterized conceiving of governments as distinctinterest groups concerned about nancing their expenditures as the LeviathanApproach see Niskanen [1971] and Brennan and Buchanan [1977]
4 With the passage of the National Banking Act in 1864 the federalgovernment also began to charter banks (motivated by a desire to use suchinstitutions to help to fund the Civil War see Kroszner [1997]) While there hadbeen some initial ambiguity concerning state authority over these institutions the1927 McFadden Act claried that until the 1994 Riegle-NealAct states effectivelyhad the right to prevent interstate branching and to force national banks toconform to state branching regulations See White [1983]
WHAT DRIVES DEREGULATION 1439
concentrated while costs to consumers of a less efficient andcompetitive nancial sector tend to be diffuse (eg Stigler [1971]and Peltzman [1976])
B A Brief History of Recent Branching Deregulation
Prior to the 1970s most states had laws restricting within-state branching and all states forbade interstate branching(Table I and Figure I) Although there had been some changes instate branching laws during the late nineteenth and early twenti-eth centuries these laws have remained stable since the GreatDepression Some of the statutes were essentially unchanged formore than a century Since the early 1970s however all but one ofthese states have relaxed restrictions on intrastate branching
Deregulation of these restrictions typically involves threetypes of reforms The rst concerns the formation of multibankholding companies (MBHCs) that were permitted to own multiplebanks but had to operate them separately The offices of the banksin an MBHC could not be integrated into a single network so adepositor at one bank would not have access to her deposits atanother The banks in an MBHC also could not consolidate theirback-office operations and each had to meet all regulatoryobligations eg capital requirements as if it were a stand-aloneinstitution
The second and most important step toward deregulationoccurs when states permit branching by merger and acquisitionthereby allowing MBHCs to convert offices of subsidiary banks(existing or acquired) into branches of a single bank An MBHCcould then integrate its banking offices into a single branchnetwork A third reform occurs when states permit full statewidebranching whereby banks could open new branches anywherewithin state borders5
The Douglas amendment to the Bank Holding Company Actof 1956 prevented holding companies from buying out-of-statebanks unless that state explicitly permitted such acquisitions bystatute (see Macey and Miller [1992]) Since no state allowed suchacquisitions holding companies were effectively prohibited fromcrossing state lines Deregulation began in 1975 when Mainepassed legislation permitting out-of-state bank holding compa-
5 Permitting branching only through merger and acquisition before fullstatewide branching could be interpreted as states allowing incumbent banks tomaintain the ability to extract at least a portion of the rents associated withbarriers to entry from purchasing banks
QUARTERLY JOURNAL OF ECONOMICS1440
TABLE IYEAR OF DEREGULATION OF RESTRICTIONS ON GEOGRAPHICAL EXPANSION BY STATE
State
Intrastatebranching
through MampA
Full intrastatebranchingpermitted
Interstatebanking
permitted
Multibank holdingcompaniespermitted
AL 1981 1990 1987 1970AK 1970 1970 1982 1970AZ 1970 1970 1986 1970AR 1994 1989 1985CA 1970 1970 1987 1970CO 1991 1988 1970CT 1980 1988 1983 1970DE 1970 1970 1988 1970DC 1970 1970 1985 1970FL 1988 1988 1985 1970GA 1983 1985 1976HI 1986 1986 1970ID 1970 1970 1985 1970IL 1988 1993 1986 1982IN 1989 1991 1986 1985IA 1991 1984KS 1987 1990 1992 1985KY 1990 1984 1984LA 1988 1988 1987 1985ME 1975 1975 1978 1970MD 1970 1970 1985 1970MA 1984 1984 1983 1970MI 1987 1988 1986 1971MN 1993 1986 1970MS 1986 1989 1988 1990MO 1990 1990 1986 1970MT 1990 1993 1970NE 1985 1990 1983NV 1970 1970 1985 1970NH 1987 1987 1987 1970NJ 1977 1986 1970NM 1991 1991 1989 1970NY 1976 1976 1982 1976NC 1970 1970 1985 1970ND 1987 1991 1970OH 1979 1989 1985 1970OK 1988 1987 1983OR 1985 1985 1986 1970PA 1982 1990 1986 1982RI 1970 1970 1984 1970SC 1970 1970 1986 1970SD 1970 1970 1988 1970TN 1985 1990 1985 1970TX 1988 1988 1987 1970UT 1981 1981 1984 1970VT 1970 1970 1988 1970VA 1978 1987 1985 1970WA 1985 1985 1987 1981WV 1987 1987 1988 1982WI 1990 1990 1987 1970WY 1988 1987 1970
States not yet deregulated Source Amel [1993] and updates by authors
WHAT DRIVES DEREGULATION 1441
nies to acquire Maine banks As part of the Garn-St GermainAct federal legislators in 1982 amended the Bank HoldingCompany Act to allow failed banks and thrifts to be acquired byany bank holding company regardless of state laws (see egKroszner and Strahan [1996]) Many states then entered regionalor national reciprocal arrangements whereby their banks could bebought by any other state in the arrangement Between 1984 and1988 38 states joined one of these arrangements (see Amel[1993])
Table I and Figure I illustrate the history of state deregula-tion of geographical restrictions since 1970 The state-by-statederegulation culminates in the phaseout of interstate bank-ing restrictions with the passage of the 1994 Riegle-Neal Inter-state Banking and Branching Efficiency Act which effectivelycodied at the national level what had been occurring at the statelevel6
6 The Act permitted states to pass legislation to opt out of the interstatebanking provisions if the legislature did so before the provisions were to go intoeffect in mid-1997 Since only Texas and Montana have passed opt-out legislationthe United States will now have nearly complete interstate banking and branching
FIGURE IDeregulation of Restrictions on Intrastate Branching
Permitted Intrastate Branching before 1970
QUARTERLY JOURNAL OF ECONOMICS1442
III HYPOTHESES AND VARIABLE DEFINITIONS
Our empirical tests are based on the timing of intrastatebranching deregulation through merger and acquisition We focuson this form of branching deregulation because it has a muchgreater economic impact than the other forms of branchingderegulation (see the Appendix) Deregulation through mergerand acquisition is the only type of branching deregulation thatconsistently has a statistically signicant effect on bankingstructure bank efficiency and overall economic growth and theestimated magnitudes of the effects are greatest for this type ofderegulation We also analyze the interstate deregulation thattakes place at the national level by modeling voting behavior inthe House of Representatives on federal interstate branchinglegislation7
In the remainder of this section we describe our hypothesesabout the private-interest public-interest and political-institu-tional theories along with the data sources used to construct thevariables Some of the variables help us to distinguish between apublic- and private-interest approach while others will be consis-tent with both approaches The political-institutional approachhas little overlap in its implications with the other theories Ourmethod is to assess the relative importance of factors representingeach approach in speeding or slowing deregulation
A Hypotheses
Intraindustry Rivalry Small banks have fought to maintainand extend branching restrictions both historically and in therecent debates8 Smaller banks appear to have been the mainwinners from antibranching laws since these restrictions protectthem from competition from larger and more efficient bankingorganizations (see Flannery [1984] Jayaratne and Strahan [1998]
7 We do not employ a hazard model to explain the timing of interstatederegulation for two reasons the time clustering of interstate deregulation andcomplications of the strategic interdependence of the statesrsquo behavior As thenumber of states in an interstate arrangement increases for example the effectson potential acquirers is ambiguous they might prefer a larger pool of potentialtarget banks but might have to pay more as the number of competing acquirersexpands The benet to potential targets tends to increase with the number statesin an arrangement since the expected selling price increases with the number ofpotential bidders [Brickley and James 1987]
8 Economides Hubbard and Palia [1996] provide evidence that voting inCongress for the 1927 McFadden Act responded to small state banksrsquo interest inlimiting competition from large national banks See also White [1983] and Abramsand Settle [1993] for historical opposition On the small bank opposition to therecent branching deregulation see Kane [1996] and The Economist [8694 p 59]
WHAT DRIVES DEREGULATION 1443
and Winston [1993]) Branching restrictions thus tend to reducethe efficiency and consumer convenience of the banking system9
The private-interest theory therefore predicts that reformshould occur later in those states where the strength of the smallbanks relative to that of the large banks is greater According tothe public-interest theory deregulation should take place earlierwhere small banks are relatively important because the socialcosts of the regulation are directly related to the size of theprotected sector The social costs include deadweight costs andlosses associated with inefficiencies in the production of bankingservices relative to production without geographical restrictions
Interindustry Rivalry A number of states permit state-chartered commercial banks to sell insurance The insurancelobby would thus oppose the relaxation of branching restrictionswhen banks can sell insurance because such deregulation mightpermit banks to provide a more efficient insurance distributionnetwork10 According to the private-interest theory reform shouldoccur later in states where banks can sell insurance and theinsurance industry is important relative to the banking industryThe efficiency costs of the branching restrictions however risewith the size of the insurance sector since the restrictions preventthe exploitation of scope economies Under the public-interesttheory reform should therefore occur earlier where banks can sellinsurance and the insurance industry is relatively large11
Consumers of Banking Services Banks are a major source ofcredit for small rms [Cole and Wolken 1994] Branching deregu-lation tends to reduce banksrsquo local market power [Jayaratne andStrahan 1998] In addition Strahan and Weston [1998] nd thatlending to small businesses increases on average when smallbanks are purchased by other banking organizations and Bergeret al [1998] nd that credit availability to small businessesincreases in the years following a takeover of a small bank by alarger banking organization Since bank borrowers tend to benet
9 Flannery [1984] shows that small banks in states with branching restric-tions have higher costs than small banks in states without such restrictions andthe Appendix shows how the production of banking services becomes more efficientfollowing deregulation
10 The Economist [8694 p 59] describes the insurance industryrsquos opposi-tion to branching deregulation
11 Another rival interest we considered was the Savings and Loan industryThe share of assets in Savings and Loans relative to banking plus Savings andLoan assets in the state however had only small and statistically insignicanteffects in the models we estimate below
QUARTERLY JOURNAL OF ECONOMICS1444
from branching deregulation in particular and bank consolidationin general the private-interest theory would predict that stateswith numerous small bank-dependent rms would deregulateearlier12 This prediction however is also consistent with thepublic-interest theory since the social costs of the restrictions arehigher in states with more small bank-dependent rms
Another effect of branching restrictions on bank customerscan be related to the prices paid for bank services High loanprices before deregulation may reect for example a largemarket share for inefficient high-cost banks or high rents beingearned by the banks protected from competition In either casethe public-interest theory would suggest that states with rela-tively high initial loan interest rates should deregulate earlierThe private-interest theory however does not have a clearimplication high rents could lead the beneciaries to ght harderto maintain them (eg Stigler [1971]) or the deadweight costsassociated with these high rents could make it more difficult forthe beneciaries to maintain the restrictions (eg Becker [1983])13
Bank Stability Geographic diversication through branch-ing could mitigate instability problems that were importantduring much of our sample period thereby improving welfareInstability also may reduce the incentives of banks to lobby tomaintain protections because unstable banks are less likely tosurvive to reap the benets of the restrictions politicians also mayanticipate that such banks are less likely to be able to providefuture contributions and support [Gunther 1994 1996]14 Both theprivate-interest and public-interest theories thus suggest thatderegulation is most likely in states where banking instability isgreatest
12 On the other hand local banking monopolies created by branchingrestrictions could strengthen relationshipsbetween banks and small and medium-sized rms and increase the availability of credit to these rms [Petersen andRajan 1994] Also some have argued that small business lending declines whenlarge banks take over small banks (eg Berger Kashyap Scalise [1995])
13 High loan interest rates also could reect compensation for higher risksand fewer opportunities for banks to diversify If state-specic risks explaindifferences in average loans rates then neither theory would have any predictionsfor the sign of this effect Following branching deregulation however loan interestrates tend to fall but bank risk proles do not appear to change [Jayaratne andStrahan 1998 Kroszner and Strahan 1998] This suggests that high loan rates inantibranching states are not simply reecting state-specic risks
14 Also deregulation can arise as a response to banking instability under atheory of ignorant or misinformed voters in which a banking crisis acts as aneducational device to make the previously ignorant public aware of the costs of theantibranching policy (see Kane [1996])
WHAT DRIVES DEREGULATION 1445
Political-Institutional Factors Republicans are typically per-ceived as more likely to favor deregulation than Democrats so thepolitical-institutional theories suggest that states that are con-trolled by Democrats deregulate later than those controlled byRepublicans15 In addition we investigate whether reform be-comes more likely when one party controls the legislature and thegovernorship Note that these political effects must be interpretedwith caution since the views of the politicians may simply reectthe economic interests of the constituents in the state (seePeltzman [1984])
B Variable Denitions and Data Sources
Our main proxy for the strength of the small banks is thefraction of banking assets in the state in lsquolsquosmallrsquorsquo banks We denesmall banks as banks with assets below the median size in eachstate By allowing the denition of small to vary across states wetake into account cross-state heterogeneity in bank sizes16 An-nual data on bank size are from the fourth quarter Reports ofIncome and Conditions (lsquolsquoCall Reportsrsquorsquo) from the Federal ReserveBoard We also include the capital-to-asset ratio of small banksrelative to large as a measure of their relative nancial healthSpecically this variable is the asset-weighted average capital-asset ratio for small banks minus the asset-weighted averagecapital-asset ratio for large banks Annual data on bank capitalare from the Call Reports
To measure the effects of the rival insurance industry we rstconstruct an indicator variable that is one if the state permitsbanks to sell insurance For each state we then measure the sizeof the insurance sector (total value added in the state) relative tothe sum of the banking plus insurance sectors We will examine
15 For more detail on the importance of legislative structures party politicsand ideology see Poole and Rosenthal [1997] Kahn and Matsusaka [1997] Irwinand Kroszner [1999] and Kroszner and Stratmann [1999]
16 We also considered a variety of other denitions of relative size but therelative size results reported below are not sensitive to which denition we useFirst we used a xed measure of small bank that did not vary across statesdening small banks as those with assets below $100 million in 1994 dollarsSecond we applied the state-varying and xed denitions at the level of lsquolsquobankingorganizationsrsquorsquo (which include multibank holding companies) instead of at the levelof banks Third we calculated gini coefficients of bank size inequality and bankconcentration indices as alternative proxies for small versus large bank power inthe state Finally we also considered banks in lsquolsquoruralrsquorsquo areas (that is not located in aBureau of the Census lsquolsquostandard metropolitan statistical arearsquorsquo) as small since thevalue of the restrictions might be greatest in protecting banks outside of cities fromentry by the city banks All of the alternative denitions are highly correlated andyield the same results on the importance of intraindustry differences
QUARTERLY JOURNAL OF ECONOMICS1446
the effect separately for states that permit banks to sell insuranceand those that do not Data on value-added by industry are fromU S Commerce Department Bureau of Economic Analysis Sur-vey of Current Business [August 1994]
To measure the relative importance of small bank-dependentborrowers we include the proportion of all establishments operat-ing in the state with fewer than twenty employees These data arecompiled by the Bureau of the Census17 As a measure of priceswe use the difference between the average interest rate on loans inthe state and the prevailing federal funds rate where the loanrate equals the ratio of total interest income on all domestic loansdivided by total domestic loans held by banks operating in thestate18 These data are from the end-of-year Call Reports butbecome available beginning only in 1976 Our proxy for bankinstability is the failure rate of banks measured as the percent oftotal state banking assets in failed institutions from the FederalReserve Boardrsquos National Information Center database19
We include two political variables to adjust for any indepen-dent inuence of party politics First we measure the degree ofparty control of the state government by the fraction of the threebodies of the state government (the assembly senate and gover-norship) controlled by Democrats This variable for example isone-third if the Democrats have a majority in the assembly andthe Republicans have a majority in the senate and hold thegovernorship Second we include an indicator variable equal toone if the same party controls the governorrsquos office and hasmajorities in both chambers of the state legislature20
17 We collected the establishment data by the state for three cross sections(1976 1982 and 1987) and interpolated the data in the intermediate years SeeState and Metropolitan Data Book 1982 1986 and 1991
18 A measure of how individual customers would be affected is interest ratesthey receive on deposits We calculated the average deposit interest rate in thestate from the end-of-year Call Reports and did not nd any relationship betweenthis variable and the timing of deregulation in our analysis below Since Jayaratneand Strahan [1998] nd that branching deregulation has no effect on deposit ratesour nding of no relationship between timing and rates result could be due toeither individual bank customers anticipating little pricing effect and not lobbyingor an inability of individual depositors to form an effective lobby
19 As alternative measures of distress we used (1) the bank failure ratemeasured as the number of failed institutions in the state relative to the totalnumber of banks (2) the overall capital-to-asset ratio of all banks in the state and(3) the growth in state personal income Our results are not sensitive to the choiceof distress proxy
20 We also estimated the models with three indicator variables to reect theparty control of the assembly senate and governorship separately The coefficientestimates on these variables are qualitatively similar to those we present below(that is Democratic control tends to slow deregulation) and the results on theinterest group variables are unaffected by the specication of the political
WHAT DRIVES DEREGULATION 1447
IV METHODS AND RESULTS
In this section we rst estimate how the variables describedabove inuence the timing of state-level branching deregulationusing a hazard model Next we discuss the robustness of theresults including a test of whether the ex post impact of deregula-tion on the different interest groups is consistent with ourinterpretation of the coefficients We then investigate federalinterstate branching deregulation with a probit model to explorewhether the factors that affect the timing of intrastate deregula-tion also inuence voting in the House of Representatives oninterstate deregulation
A Hazard Model
We consider the period from the beginning of our sample(1970) until deregulation as the lsquolsquoduration of regulationrsquorsquo or thelsquolsquotime until deregulationrsquorsquo The hazard rate h(t) is the likelihoodthat a state deregulates at time t given that the state has not yetderegulated [Kalbeisch and Prentice 1980 Kiefer 1988 Greene1997] To model the duration of regulation we must decide whatstructure if any to impose on the hazard function The Kaplan-Meier product-limit estimator provides a simple nonparametricway to estimate the shape of the hazard function over time[Greene 1997]21 Figure II graphs this estimate for our data andshows that the hazard function is relatively at in the early yearsand then grows steeper in the later years
A duration model that can approximate this shape is theWeibull proportional hazards model The hazard rate functiontakes the form
h[t x(t) b] 5 h0(t) exp[x(t)8b]
where x(t) is a vector of time-varying covariates b is a vector ofunknown parameters to be estimated and the baseline hazard
variables Note also that Nebraska has a nonpartisan unicameral legislature Weassume that the party of the Governor controls the state government for Nebraskaand set the unied government indicator to one As another alternative we triedincluding a variable that is one if party control of the legislature or governorshipchanges but it did not have an economically or statistically signicant effect anddid not affect any of the other results
21 Denote each of the K years in our sample period Tk and order them suchthat T1 T2 and so on Let nk be the number of states that have not yet deregulatedby Tk and dk be the number of states that deregulate in year Tk The Kaplan-Meierestimate of the hazard rate in each year is dknk In other words the hazard rate isthe ratio of the number of states that actually deregulate in each year to thenumber of states that have not yet deregulated in that year
QUARTERLY JOURNAL OF ECONOMICS1448
rate h0(t) is ptp 2 1 with shape parameter p that will be estimatedfrom the data When p 1 this model displays a monotonicallyincreasing hazard rate To estimate the parameters we maximizethe following log-likelihood function [Kiefer 1988]
L(b) 5 oi 5 1
N
di ln h[ti x(ti) b] 2 oi 5 1
N
e 0
tih[u x(u) b] du
where N indexes the number of states at risk at the beginning ofthe sample period di is an indicator equal to one for the statesthat deregulated by the end of the sample period and zero for thestates that do not deregulate (ie the censored observations) andti is the time of deregulation for the ith state This formulationassumes that the time until deregulation is independent acrossstates but the observations of the same state over time are notassumed to be independent In addition we use a robust estima-tion procedure to calculate the standard errors [Lin and Wei1989]
We exclude states that deregulated before 1970 from theanalysis so N equals 39 states During the sample period 36 ofthe 39 states deregulate We have three censored observations forwhich di is zero (Arkansas Iowa and Minnesota) but our results
FIGURE IIKaplan-Meier Estimate of Hazard Rate for Branching Deregulation
WHAT DRIVES DEREGULATION 1449
are not sensitive to the inclusion of these censored observationsSince we observe each state in each year up to and including theyear of deregulation we have a total of 637 observations Table IIreports summary statistics and correlation coefficients for theexplanatory variables we use in our analysis
We use the Weibull model rather than a model that does notimpose any structure on the baseline hazard rate so that we cancalculate the change in the expected time to deregulation for agiven change in the levels of the covariates22 In the Weibullmodel we can invert the hazard function and map it into the timedomain Rewriting the Weibull model in this way the log of thetime to deregulation T is a linear function of the political economyfactors and an error term ln (T ) 5 x8b 1 e23 Because we assumethat the baseline hazard rate is Weibull with a shape characterizedby p the new coefficients will be scaled by p that is b 5 2 bp
The b coefficients represent the percentage change in thetime to deregulation for a one-unit change in the covariates Apositive coefficient implies that an increase in the variableincreases the expected time until deregulation To gauge theeconomic importance of the effects we will multiply the b by thestandard deviation of the explanatory variables and then evaluatehow much the changes in the variables raise or lower the expectedtime to deregulation24 In Table III we report the b coefficients
22 The Cox proportional hazards model does not impose any structure on thebaseline hazard rate h0(t) and takes the following form
h[t x(t) b] 5 h0(t) exp [x(t)8b]
Common factors that affect the probability of deregulation in the same way for allstates therefore do not inuence the estimated coefficients The cost of the Coxmodel is that because there is no structure on the baseline we can calculatechanges in only relative hazard rates associated with changes in the covariatesThe Weibull model however provides sufficient structure so that we can translateour estimates into a log expected time metric that allows us to calculate the changein the expected time to deregulation for a given change in the covariates SeeKiefer [1988] Greene [1997] and Stata [1997] When we compare the coefficientsestimated by the Cox and Weibull models on our data they are very close (seeKroszner and Strahan [1998]) so imposing a Weibull structure does not appear todistort our estimates
23 The error term e is independent of x and has an extreme value distributionscaled by 1p (see Kiefer [1988] Greene [1997] and Stata [1997] for more details onthis log expected time parameterization of the Weibull model)
24 The transformation to the log expected time domain is only strictly correctwhen the covariates are constant over time Our interpretation of the coefficients ispossible in the time-varying model if we assume that a one-standard-deviationdifference would hold over the entire sample period that is the deregulationwould occur so many years earlier or later if the variable in question were higherby one standard deviation from 1970 to 1992 Note that the mean number of yearsto deregulation in the sample is 158
QUARTERLY JOURNAL OF ECONOMICS1450
TABLE IISUMMARY STATISTICS FOR VARIABLES IN HAZARD MODEL 1970ndash1992
Panel A Univariate statistics MeanStandarddeviation Minimum Maximum
(1) (2) (3) (4)Small bank asset share of all
banking assets in state 011 005 001 023Relative size of insurance to
banking plus insurance inthe state 047 009 024 080
Indicator is 1 if banks can sellinsurance in the state 020 040 0 1
Small rm share of the numberof rms in the state 088 002 078 095
Capital ratio of small banksrelative to large in the state 002 001 2 001 009
Share of state government con-trolled by Democrats 065 036 0 1
Indicator is 1 if state controlledby one party 054 050 0 1
Average yield on bank loans inthe state minus Fed fundsrate 002 002 2 004 008
Panel B CorrelationsSmallbank
Relativesize of
insuranceInsuranceindicator
Smallrm
Smallbank
capitalDem
controlSingleparty
(1) (2) (3) (4) (5) (6) (7)Small bank asset share
of all banking assetsin state 1
Relative size of insur-ance to banking plusinsurance in the state 2 034 1
Indicator is 1 if bankscan sell insurance inthe state 020 2 012 1
Small rm share of thenumber of rms inthe state 030 2 011 005 1
Capital ratio of smallbanks relative tolarge in the state 2 066 020 2 010 2 006 1
Share of state govern-ment controlled byDemocrats 2 006 001 2 018 2 018 005 1
Indicator is 1 if statecontrolled by oneparty 007 012 001 2 006 2 006 030 1
Average yield on bankloans in the stateminus Fed funds rate 2 001 2 015 003 022 2 010 2 003 2 006
The number of observations is 637 except for the average interest rate on bank loans which becomesavailable only in 1976 so the number of observations is 408
TABLE IIIHAZARD MODEL OF POLITICAL ECONOMY FACTORS AFFECTING THE TIMING OF STATE
BRANCHING DEREGULATION 1970ndash1992
(1) (2) (3) (4) (5) (6)Small bank asset share of
all banking assets instate
648(122)
519(137)
632(135)
516(118)
488(133)
445(088)
Capital ratio of smallbanks relative to largein the state
1325(348)
921(319)
1301(362)
1067(346)
1144(375)
906(381)
Relative size of insurancein states where banksmay sell insurance 0otherwise
324(161)
259(149)
315(165)
285(154)
272(159)
055(134)
Indicator is 1 if banksmay sell insurance inthe state
045(010)
037(011)
045(010)
038(008)
036(009)
001(012)
Relative size of insurancein states where banksmay not sell insurance0 otherwise
2 093(043)
2 090(033)
2 102(051)
2 093(048)
2 096(053)
2 004(039)
Small rm share of thenumber of rms in thestate
2 972(211)
2 614(215)
2 952(211)
2 943(230)
2 952(270)
2 1509(276)
Share of state govern-ment controlled byDemocrats
031(012)
023(011)
030(012)
027(011)
026(011)
011(013)
Indicator is 1 if state con-trolled by one party
2 004(010)
2 004(007)
2 004(007)
2 001(009)
2 002(010)
017(009)
Average yield on bankloans in the stateminus Fed funds rate mdash
023(319) mdash mdash mdash mdash
Bank failure rate mdash mdash092
(111) mdash mdash mdash
Indicator is 1 if state hasunit banking law mdash mdash mdash
018(010)
019(010)
030(009)
Indicator is 1 if statechanges bank insur-ance powers mdash mdash mdash mdash
2 008(012)
2 026(013)
Includes regionalindicators No No No No No Yes
N 637 408 637 637 637 637Log likelihood 2 374 900 2 360 2 212 2 191 1192p-value of x 2 for
regression 001 001 001 001 001 001
The hazard model is Weibull where the dependent variable is the log expected time to branchingderegulation ln (T) 5 bx 1 eAll variables are measured for each state in each year Small bank asset share isthe percent of banking assets in the state held by banks below the median size of banks in each state in eachyear Relative capital ratio is the capital to assets ratio of small banks minus that of large banks Size ofinsurance relative to banking plus insurance in the state is measured as gross state product from insurancedivided by gross state product from insurance plus banking Insurance indicator is 1 if state law permits banksto enter the insurance business Small rm share is the percent of all establishments in the state that havefewer than twenty employees Party control variable is the share of the three bodies of state governmentcontrolled by Democrats One party control indicator is 1 if the same party controlrsquos the governorship and hasmajorities in both chambers of the state legislature Average yield on bank loans equals total interest incomeon all domestic loans made by banks in the state divided by total loans Bank failure rate is the ratio of assetsin failed banks during a given state and year divided by total bank assets in that state and year Unit bankingindicator equals 1 for states with unit banking restrictions The change in insurance powers indicator equalsone if the state changed the law to permit banks to sell insurance during the sample period We divide thecountry into four regions and include regional indicators in column (5) Robust standard errors are inparentheses and denote statistically signicant at the 1 5 and 10 percent levels
QUARTERLY JOURNAL OF ECONOMICS1452
The positive and statistically signicant coefficient on smallbank share implies that a greater small bank share delaysderegulation and this effect is economically important A one-standard-deviation increase in the small bank share results in a30 percent increase in the time until deregulation or about 47years This result supports the implications of private-interesttheory concerning intraindustry rivalry and is contrary to thepublic-interest theory
The positive and statistically signicant coefficient on therelative performance of small banks in the state implies thatstates deregulate later when small banks are relatively strongnancially25 A one-standard-deviation increase in the relativecapital-to-asset ratio results in a 15 percent rise in the time untilderegulation or about 24 years This result is consistent with theprivate-interest theory but may also be consistent with a public-interest interpretation because the risk of failures (and a costlytaxpayer-funded bailout) increases as small bank nancial healthdeclines
Interindustry competition also helps explain the timing ofderegulation In states where banks can sell insurance a rela-tively large insurance sector is associated with an increase in theexpected time to deregulation26 A one-standard-deviation in-crease in the relative size of the insurance sector in those statesthat permit banks to sell insurance leads to a 22 percent increasein the time until deregulation or about 35 years27 This result isconsistent with the private-interest theory but not the public-interest theory
Deregulation occurs earlier in states where small bank-dependent rms are relatively numerous A one-standard-
25 We also allowed small bank and large bank capital-to-asset ratios to enterseparately Their coefficient estimates were nearly equal in absolute value and ofopposite signs and so the data do not reject the relative small to large bank capitalratio specication we have chosen
26 When we use the relative size of the insurance sector in the state as awhole rather than insurance relative to banking plus insurance in the state weobtain similar results When we simultaneously include both measures of therelative size of insurance the interactions of the lsquolsquoinsurance relative to bankingplus insurancersquorsquo variable continue to be statisticallysignicant but the interactionsof lsquolsquoinsurance relative to the state as a wholersquorsquo variable are not
27 We have also estimated the model with the indicator variable equal to onefor states where banks may sell insurance but without the variables measuringtheimportance of insurance relative to banking In this model the coefficient on theinsurance indicator variablemdashan estimate of the difference in the expected time toderegulation in states where banks may sell insurance relative to states wherebanks may not sell insurancemdashequals 027 and is statistically signicant at the 5percent level This suggests that branching deregulation takes about 45 yearslonger in states where banks may sell insurance
WHAT DRIVES DEREGULATION 1453
deviation increase in the share of small rms reduces the timeuntil deregulation by 18 percent or about three years This resultconcerning the interests of consumers of banking services isconsistent with both the private- and public-interest theories
Column (2) of Table III introduces the average interest rateon loans to the model as a rough proxy for the cost to bankborrowers of these regulations These data become available onlyin 1976 so our sample size shrinks to 408 observations Thecoefficient on this variable is small and not statistically signi-cant The public-interest theory would imply that the coefficientshould be negative while the implication of the economic theorydoes not have a clear sign prediction
When we add the failure rate of banks to the specication incolumn (3) its coefficient is small and statistically insignicantWe do not nd a linkage between the timing of state deregulationand statewide banking distress Given that deposit insurance isfunded federally however bank failures may raise the publicrsquosconcerns about the costs of the regulations at the national ratherthan the state level This interpretation can explain why the rateof bank failure does not affect the timing of state-level deregula-tion even though bank instability does appear to rise generallyduring the period of increasing deregulation (see Section V below)
Finally the partisan structure of the state government alsoinuences when states deregulate As expected a higher propor-tion of Democrats in the government tends to delay deregulationA one-standard-deviation rise in the share of the governmentcontrolled by Democrats slows the deregulation by about twoyears Whether the state is dominated by one party however doesnot appear to affect the timing of the deregulation
To summarize our ndings from the hazard model privateinterests appear to play an important role in the deregulatoryprocess While private interests and public interests do sometimescoincide the results on the relative share of small banks and largebanks and on the relative size of insurance where banks competeare consistent with a private-interest approach but are difficult toexplain on public-interest grounds
B Robustness Tests
Omitted Variables States begin the sample period withdifferent degrees of restrictions on bank expansion These initialconditions may proxy for omitted factors determining the political-economy equilibrium in the state that could then affect the
QUARTERLY JOURNAL OF ECONOMICS1454
subsequent timing of deregulation For instance some statesbegan the period with unit banking while other states allowedlimited branching and the formation of multibank holdingcompanies
To test whether the stringency of initial restrictions onbranching matters we rst divide the sample into states with andwithout the most extreme form of branching regulation unitbanking28 Column (4) of Table III includes an indicator variableequal to one if the state began the period with unit bankingrestrictions The coefficient on this variable is positive andstatistically signicant but the other results remain virtuallyunchanged Next we include an indicator variable that is one ifthe state permitted multibank holding companies (MBHCs) toexist at the beginning of our sample period Thirty-four statesbegan the period allowing MBHCs The coefficient estimate forthis indicator variable (not reported) is small and not statisticallysignicant and its inclusion does not change the effects of theother variables
Eleven states expand bank powers to include the sale ofinsurance products during our sample period This could proxy foromitted factors about the relative strength of the insuranceindustry in the state In column (5) we include an indicatorvariable that is one if the state changes the law during our sampleperiod to permit banks to sell insurance business and zerootherwise The coefficient estimate is small and statisticallyinsignicant and does not affect the other results29
In column (6) we introduce four regional indicator variablesNorth South Midwest and West30 The coefficients on small bankshare the relative capital-to-assets ratio and small rm sharechange little Although the coefficient estimate still suggests that
28 We classify states that prohibited branching but permitted banks toestablish facilities as unit banking states The unit banking indicator is one for thefollowing states CO AR FL IL IA KS MN MO MT NE ND OK TX WI WVand WY
29 We conducted a likelihood ratio test of whether the unit banking statesthe MBHC states and the states that changed their regulations of bank insurancepowers could be pooled with the other states In all three cases we could not rejectpooling of the data
30 Since intrastate deregulation generally preceded interstate deregulationand the latter typically took the form of regional interstate compacts thepossibility of participating in a regional interstate banking compact could haveinuenced the decision to deregulate intrastate branching Our denitions of theregions are region 1 (South) contains AL AR DC FL GA KY LA MS NC OKSC TN TX and VA region 2 (Northeast) contains CT MA MD ME NH NJ NYPA RI VT and WV region 3 (Midwest) contains IA IL IN KS MI MN MO NEND OH SD and WI region 4 (West) contains the other states
WHAT DRIVES DEREGULATION 1455
in states where banks can sell insurance a larger relative size ofthe insurance sector delays deregulation the effect is no longerstatistically signicant The coefficient on the indicator variablefor whether a state expands bank powers to permit insurancesales during the sample period however now becomes statisti-cally signicantly negative This implies that deregulation occursroughly 41 years earlier in states where the insurance lobby maybe relatively weak While the inclusion of the regional indicatorsdoes not affect the other results the results on insurance doappear to be sensitive to the specication of the hazard model31
Estimation Technique To check the robustness of using ahazard model and to investigate whether the time-series or thecross-sectional variation in each of the factors accounts for thehazard results we estimate two simple linear probability models(neither are reported) First we run a pooled time-series cross-section OLS regression using the 637 observations from thehazard model The dependent variable equals one in the year ofderegulation and zero otherwise We include the same set ofregressors that were used in the hazard model and add state-levelxed effects This specication thus removes the cross-statevariation from the independent variables and focuses on theirmovements over time In this model the signs of the coefficientson the interest group factors are consistent with the results fromthe hazard model and the coefficients for the small bank relativecapital ratio and for the relative size of insurance where states cansell insurance are statistically signicant
In the second model we eliminate the time-series variationby averaging each of the explanatory variables over time We thushave one observation per state that had branching restrictions asof 1970 (N 5 39) The dependent variable equals the inverse of thenumber of years until deregulation which represents an estimateof the statersquos probability of deregulating in a year under theassumption that this probability is constant over time and eachyear provides an independent observation32 Although we haveonly 39 observations the coefficients on the small bank marketshare small bank relative capital ratio and small rm share
31 As further robustness check we also tried a population density variablewhich Abrams and Settle [1993] found relevant for regulatory change during the1920s and 1930s but its coefficient was not statistically signicant and did notaffect the other results
32 For instance the dependent variable would take the value of 01 for astate that deregulated in 1980 For the three states that did not deregulate thedependent variable equals 0
QUARTERLY JOURNAL OF ECONOMICS1456
variables are statistically signicant The signs of the effects ofthe insurance variables are the same as those reported in thehazard model but not statistically signicant
The hazard model and linear probability models thus producevery similar results The estimates from the latter suggest thatthe effects for the small bank market share and small rm sharevariables are driven primarily by cross-state variation the effectof the insurance variable is driven primarily by time-seriesvariation and the small bank capital effect is driven by bothcross-sectional and time-series variation
Ex Post Consequences of Deregulation Finally to check theplausibility of our interpretation of the hazard results we con-sider whether the ex post consequences of deregulation areconsistent with the ex ante lobbying positions we attribute to eachinterest group (see Kroszner and Strahan [1998] for details) Wend that small banks lose market share following deregulationand in states where banks can enter the insurance business theinsurance sector shrinks relative to the banking sector followingderegulation Borrowers also benet from deregulation throughlower average interest rates on loans These ndings support theprivate-interest interpretation of the coefficients in the hazardmodel groups that will benet lobby to speed deregulation andthose that will be harmed lobby to slow it
C Voting on Interstate Branching Deregulation in the U SHouse of Representatives
We now examine whether the forces driving intrastate branch-ing deregulation also drive interstate deregulation Financialservices interests are active contributors and lobbyists in Washing-ton Their political action committees constitute the largest groupof contributors to legislators providing nearly 20 percent of totalcongressional campaign contributions [Makinson 1992] and muchof their lobbying effort involves competition among rival interestswithin nancial services (see Kroszner and Stratmann [1998])
As noted above after virtually all states adopted intra- andinterstate branching deregulation the 1994 Riegle-Neal Act re-pealed the 1927 McFadden Act to eliminate all barriers tointerstate banking and branching by 1997 The key votes concern-ing the Riegle-Neal Act were either voice votes or extremelylopsided so we could not estimate a voting model from them Anumber of bills and amendments related to interstate branching
WHAT DRIVES DEREGULATION 1457
had been debated in Congress during the years prior to thepassage of the Riegle-Neal Act but a search of the weekly BNABanking Reporter and the Congressional Record produced onlyone roll-call vote related to interstate branching that was notlopsided This vote occurred in the House on November 14 1991and concerned an amendment sponsored by Wylie (R-OH) andNeal (D-NC) to introduce interstate banking and branchingderegulation to a nancial services reform package33 Althoughthe amendment passed by 210 to 208 the bill to which it wasattached subsequently was defeated
To check for the inuence of the factors we considered in thestate-level reforms we examine both the sponsorship of interstatebanking legislation and voting on the amendment The sponsorsof the Wylie-Neal amendment are from states with low small bankshares 004 in Ohio (Wylie) and 002 in North Carolina (Neal) Incontrast the sample mean in 1991 is 008 (median 5 007)Michigan home state of the Senatersquos sponsor of the 1994 Riegle-Neal Act also had relatively low small bank strength (small bankshare of 005)
Table IV reports estimates of a probit model where thedependent variable equals one if the legislator voted in favor ofthe amendment and zero otherwise The explanatory variablesare the same as those in the hazard model except that we use aparty affiliation indicator variable equal to one for Democrats andzero for Republicans as our proxy for political factors Thecoefficients in Table IV are the marginal effects (lsquolsquoslopesrsquorsquo) of aone-unit change of each variable on the probability that alegislator will vote for the amendment34 Note that a forcefavoring deregulation has a negative coefficient in the hazardmodel (Table III) but a positive coefficient in the probit model
Consistent with the state-level deregulation process legisla-tors are more likely to support the amendment if their states havea relatively low share of small banks As in the hazard model thefraction of small banks is the most important interest groupinuence on a legislatorrsquos voting decision A one-standard-deviation increase in small banksrsquo market share (from the mean)is associated with a decline in the probability of voting in favor of
33 The Wylie-Neal amendment also included provisions limiting certaininsurance and real estate powers of national banks [Congressional RecordNovember 14 1991 pp 10239ndash10242]
34 Since we have multiple legislators from each state we adjust the standarderrors to correct for the lack of independence among observations clustered in thesame state
QUARTERLY JOURNAL OF ECONOMICS1458
TAB
LE
IVM
AR
GIN
AL
EF
FE
CT
SF
RO
MA
PR
OB
ITM
OD
EL
OF
TH
EIN
FL
UE
NC
EO
FP
OL
ITIC
AL
EC
ON
OM
YF
AC
TO
RS
ON
TH
EH
OU
SE
VO
TE
ON
TH
EW
YL
IE-N
EA
L
AM
EN
DM
EN
TT
OP
ER
MIT
INT
ER
ST
AT
EB
AN
KB
RA
NC
HIN
G
NO
VE
MB
ER
14
1991
(1)
(2)
(3)
(4)
(5)
(6)
Smal
lban
kas
set
shar
eof
allb
anki
ng
asse
tsin
stat
e2
417
(12
4)2
454
(12
2)2
407
(12
0)2
443
(14
7)2
450
(14
9)2
490
(14
7)C
apit
alra
tio
ofsm
allb
anks
rela
tive
tola
rge
inth
est
ate
180
(37
5)2
68(3
42)
256
(37
5)1
69(3
78)
153
(38
0)2
45(4
25)
Rel
ativ
esi
zeof
insu
ranc
ein
stat
esw
here
bank
sm
ayse
llin
sura
nce
0ot
herw
ise
22
34
(10
2)2
249
(1
14)
22
33
(10
1)2
249
(1
04)
22
51
(10
2)2
154
(11
1)In
dica
tor
is1
ifba
nks
may
sell
insu
ranc
ein
the
stat
e0
79
(0
06)
074
(00
6)0
69
(0
06)
070
(00
6)0
70
(0
06)
067
(00
7)R
elat
ive
size
ofin
sura
nce
inst
ates
whe
reba
nks
may
not
sell
insu
ranc
e0
oth
erw
ise
180
(06
6)2
42
(0
72)
154
(0
67)
173
(06
6)1
72
(0
67)
203
(0
91)
Smal
lrm
shar
eof
the
num
ber
ofr
ms
inth
est
ate
21
23(0
92)
20
83(0
89)
21
62(1
07)
21
32(0
95)
21
30(1
00)
20
94(1
16)
Indi
cato
ris
1if
Rep
rese
ntat
ive
isa
Dem
ocra
t2
035
(00
7)2
035
(00
7)2
036
(00
7)2
035
(00
7)2
035
(00
7)2
036
(00
7)A
vera
geyi
eld
onba
nklo
ans
inth
est
ate
min
usF
edfu
nds
rate
mdash10
01
(51
5)mdash
mdashmdash
mdash
Ave
rage
fail
ure
rate
198
5to
1991
mdashmdash
228
(14
1)mdash
mdashmdash
Indi
cato
ris
1if
stat
eh
asun
itba
nkin
gla
wmdash
mdashmdash
004
(00
7)0
04(0
07)
20
01(0
08)
Indi
cato
ris
1if
stat
ech
ange
sba
nk
insu
ran
cepo
wer
smdash
mdashmdash
mdash2
001
(00
7)2
003
(00
7)In
clud
esre
gion
alin
dica
tors
N
oN
oN
oN
oN
oYe
sP
seud
o-R
20
170
180
170
170
170
18L
ogli
keli
hood
224
024
223
826
223
951
224
013
224
010
223
808
p-va
lue
ofx2
for
regr
essi
on
001
0
01
001
0
01
001
0
01
Th
ede
pen
den
tvar
iabl
eis
one
ifth
ele
gisl
ator
vote
sfo
rth
eam
endm
enta
nd
zero
ifag
ains
tT
he
repo
rted
coef
fici
ents
are
the
effe
cts
ofa
uni
tch
ange
ofth
ein
depe
nde
ntva
riab
le(f
rom
the
mea
n)on
the
prob
abili
tyof
voti
ng
infa
vor
ofth
eam
endm
ent
All
vari
able
sar
em
easu
red
byst
ate
N5
418
the
num
ber
ofR
epre
sent
ativ
esvo
tin
gon
the
amen
dmen
tS
mal
lban
kas
sets
har
eis
the
perc
ent
ofba
nki
ng
asse
tsin
the
stat
eh
eld
byba
nks
belo
wth
em
edia
nsi
zeof
ban
kin
each
stat
ein
each
year
Rel
ativ
eca
pita
lrat
iois
the
capi
talt
oas
sets
rati
oof
smal
lban
ksm
inu
sth
atof
larg
eba
nks
Siz
eof
insu
ran
cere
lati
veto
bank
ing
plu
sin
sura
nce
inth
est
ate
ism
easu
red
asgr
oss
stat
epr
oduc
tfro
min
sura
nce
divi
ded
bygr
oss
stat
epr
odu
ctfr
omin
sura
nce
plu
sba
nkin
gIn
dica
tor
ison
eif
stat
ela
wpe
rmit
sba
nks
toen
ter
the
insu
ran
cebu
sin
ess
Sm
all
rmsh
are
isth
epe
rcen
tofa
lles
tabl
ishm
ents
inth
est
ate
that
hav
efe
wer
than
twen
tyem
ploy
ees
Par
tyin
dica
tor
ison
eif
the
Rep
rese
nta
tive
isa
Dem
ocra
tA
vera
geyi
eld
onba
nklo
ans
equa
lsto
tali
nter
est
inco
me
onal
ldom
esti
clo
ans
mad
eby
ban
ksin
the
stat
edi
vide
dby
tota
lloa
ns
The
aver
age
bank
failu
rera
teis
the
rati
oof
asse
tsin
faile
dba
nks
duri
ng
agi
ven
stat
edi
vide
dby
tota
lban
kas
sets
inth
atst
ate
aver
aged
over
the
year
s19
85to
1991
Un
itba
nki
ng
indi
cato
req
ual
son
efo
rst
ates
wit
hun
itba
nkin
gre
stri
ctio
nsT
hech
ange
inin
sura
nce
pow
ers
indi
cato
req
ual
son
eif
the
stat
ech
ange
dth
ela
wto
perm
itba
nks
tose
llin
sura
nce
duri
ngth
esa
mpl
epe
riod
We
divi
deth
eco
unt
ryin
tofo
urre
gion
san
din
clu
dere
gion
alin
dica
tors
inco
lum
n(5
)R
obu
stst
anda
rder
rors
are
inpa
ren
thes
es
an
d
den
ote
stat
isti
cally
sign
ica
nta
tth
e1
5an
d10
perc
entl
evel
s
WHAT DRIVES DEREGULATION 1459
branching of approximately 17 percent The relative capital-to-assets ratio of small and large banks however does not have astatistically signicant effect in the probit model
The impact of rival interests outside of banking is alsoconsistent with intrastate deregulation results Where banks cansell insurance legislators from states with larger insurancesectors relative to banking are less likely to vote for the amend-ment35 A one-standard-deviation increase in the relative size ofinsurance in those states which permit banks to sell insurancedecreases the probability that a legislator will favor the amend-ment by about 13 percent Turning to consumer interests thecoefficient on small rm share is not statistically signicant butthe coefficient on interest rates in the state is A one-standard-deviation increase in the average interest rate on loans raises theprobability that a legislator will support the amendment byroughly 8 percent The coefficient estimate for the bankingdistress variable measured as the average bank failure rate from1985 to 1991 is small and statistically signicant Overall theprobit analysis of the vote on national branching deregulationsupports the private-interest theory of deregulation and providesa consistency check that the importance of interests operating onthe state legislatures are very similar to those operating at thefederal level
V WHY DID BRANCHING DEREGULATION BEGIN DURING THE
1970S SHOCKS TO THE EQUILIBRIUM SUPPORTING BANK
BRANCHING RESTRICTIONS
A complete explanation of regulatory exit should account forwhy deregulation begins in the 1970s as well as the timing of thestate-by-state reforms [Kane 1996] In this section we describebroad technological legal and economic shocks that altered thepolitical-economy equilibrium which had kept antibranching regu-lations little changed for at least 30 years
Beginning in the 1970s three innovations reduced the valueto the protected banks of local geographic monopolies (see TableV) First automatic teller machines (ATMs) helped to erode thegeographic ties between customers and banks Second checkablemoney market mutual funds and the Merrill Lynch Cash Manage-ment Account demonstrated that banking by mail and telephone
35 The positive and statistically signicant estimates on the other insurancevariables may reect insurance industry support for the amendmentrsquos provisionslimiting banksrsquo insurance powers
QUARTERLY JOURNAL OF ECONOMICS1460
provided a convenient alternative to local banks36 Third techno-logical innovation and deregulation reduced transportation andcommunication costs particularly since the 1970s thereby lower-ing the costs for customers to use distant banks By increasing theelasticity of deposits supplied to banks these innovations reducedthe value of geographical restrictions to their traditional benecia-ries and thereby reduced their incentive to ght to maintain them[Peltzman 1976]
On the lending side increasing sophistication of credit-scoring techniques following innovations in information process-ing nancial theory and the development of large credit databases began to diminish the value of knowledge that localbankers had about the risks of borrowers in the community As a
36 Regulation Q which limited the interest rates that banks could pay ondeposits may have helped to drive depositors away from banks when the gapbetween market rates and deposit ceilings grew during the 1970s The eliminationof interest rate ceilings on large denomination certicates of deposit during the1970s appears to have hurt smaller and retail-oriented banks relative to largerwholesale banks [James 1983]
TABLE VBROAD TRENDS IN COMMERCIAL BANKING 1950ndash1995
YearNumberof ATMs
Domesticbank
deposits(billions)
Moneymarketmutual
fund(billions)
Percent ofdeposits 1
moneyfunds heldby banks
Smallbanksrsquo
percent ofbankingassets
Averagenumber of
bankfailures
(1) (2) (3) (4) (5) (6)1950 0 $ 154 $ 0 100 NA 41955 0 191 0 100 NA 31960 0 228 0 100 24 21965 0 330 0 100 20 41970 0 479 0 100 18 61975 9750 775 4 99 18 61980 18500 1182 76 94 17 101985 61117 1787 242 88 14 601990 80156 2339 493 83 11 1791995 122706 2552 745 77 8 61
Column (1) ATM gures are from Bank Network News The EFT Network Data Book (New YorkFaulkner and Gray Inc) The 1975 gure was unavailable 9750 is the number of ATMs in 1978 the rst yearfor which complete data are available Columns (2)ndash(4) Banksrsquo domestic deposits are from the Reports ofIncome and Condition money market mutual funds are from the Flow of Funds Data on all bank depositsforeign plus domestic are only available beginning in 1970 The trend in banksrsquo share (column (4)) is the sameusing total deposits instead of domestic deposits Column (5) Percent of banking assets held by small bankswhere a small bank is dened as a commercial bank with less than $100 million in assets in 1994 dollarsThese data are based on the Reports of Income and Condition Data on small banks are not available before1960 Column (6) Five-year average number of bank failures where the nal year is indicated in the rstcolumn These data are from FDIC Annual Report and the Quarterly Banking Prole
WHAT DRIVES DEREGULATION 1461
result of these innovations a national market developed forresidential mortgages credit card receivables have been securi-tized and bank lending to small business now relies less on thejudgment of loan officers and more on standardized scoringmodels
These changes have increased the potential protability forlarge banks to enter what had been the core of small bankactivities Large banksrsquo incentive to increase their lobbying pres-sure to be able to expand into these markets has thus beenincreasing over time In fact small banksrsquo market share began todecline even prior to the branching deregulation (Table V) As thevalue of a local banking relationships declined small rms thatwere the main borrowers from the small banks also probablybecame more likely to favor the entry of large banks into localmarkets With the deadweight costs of preventing large bankentry rising the private-interest theory predicts that small localbanks would become less likely able to maintain the branchingrestrictions [Becker 1983] Deregulation that reduces deadweightcosts of regulation is also consistent with the public-interesttheory
Kane [1996] argues that another major shock to the oldequilibrium was the increasing public awareness of the costlinessof having government-insured but (geographically) undiversiednancial institutions In the late 1970s the failure rate of banksbegan to rise (Table V) In the 1980s the Savings and Loan crisisand taxpayer bailout further heightened public awareness aboutthe costs of restrictions that make depository institutions morelikely to require infusions of taxpayer funds The failures thusmay have heightened public support for branching deregulation(see also Abrams and Settle [1993]) As noted above we did notnd any evidence that banking failures or distress in a stateaffected the speed with which the state deregulated but distressstill may be a common factor affecting all states
These technological economic and legal shocks generatedconditions that changed the long-standing balance favoring theantibranching forces The marginal value of lobbying to repealbranching restrictions increased just as the relative value to thesmall banks of maintaining branching restrictions was decliningThese nationwide shocks are common factors across states andthey are consistent with the positive duration dependence associ-ated with state-level deregulation that we demonstrated in FigureII In addition a broad change in lsquolsquoideologyrsquorsquo against government
QUARTERLY JOURNAL OF ECONOMICS1462
regulation could explain the positive duration dependence duringthe sample period
VI CONCLUSIONS
The private-interest theory of regulation can account for thepattern of bank branching deregulation of the 1970s and 1980sBeneciaries of branching regulation had supported a coalitionfavoring geographical restrictions despite their costs to consum-ers of nancial services Innovations that began in the 1970saltered the value of the restrictions to the affected parties and theresulting competition among interest groups can explain thesubsequent deregulation While some of our results are alsoconsistent with the public interest theorymdashfor example deregula-tion occurs earlier when small banks are in a relatively weaknancial positionmdashother results particularly evidence on theimportance of rivalries between small and large banks andbetween banking and insurance are difficult to explain with thepublic-interest approach We also nd that political-institutionalfactors affect deregulation although these variables may act asproxies for unmeasured economic interests The greater success ofthe private-interest theory to explain deregulation here than inprevious work may be due to the richer cross-sectional andtime-series variation in branching deregulation that can beexploited with a hazard model Future empirical studies ofendogenous deregulation thus may be most fruitful where changehas occurred across states over time such as in franchisinginsurance and public utilities (eg Joskow [1996])
Technological and nancial innovations will continue to erodethe benets to any interest group of maintaining regulatorybarriers in nancial services These forces are likely to bringabout reforms both domestically for example through legislationthat would increase bank powers (see Kroszner [1996] Krosznerand Rajan [1994 1997] and Kroszner and Stratmann [1998]) andinternationally for example through the extension of nancialservices provisions of NAFTAto reduce geographic barriers acrosscountries [Kroszner 1997]
APPENDIX EFFECTS OF DIFFERENT TYPES OF BRANCHING
DEREGULATION ON ECONOMIC GROWTH BANKING STRUCTUREAND BANK EFFICIENCY
We measure the importance of alternative forms of deregula-tion by their consequences for the economic growth and the
WHAT DRIVES DEREGULATION 1463
structure and efficiency of the banking sector across the states(see Jayaratne and Strahan [1996 1998]) To do so we run anOLS panel regression relating indicators for each type of deregula-tion to potential consequences Our dependent variables includeproxies for economic growth (growth in personal income) bankstructure (the log of the number of banking companies and thenumber of branches per banking company) and efficiency (the logof total noninterest expenses and the ratio of loan loss provisionsto total loans) The four deregulation indicators (MBHC branch-ing through merger and acquisition unrestricted statewide branch-ing and interstate banking) are equal to one in the years afterderegulation and zero before If a state deregulated before thesample period begins in 1970 the indicator is one throughout ifthe state does not deregulate then the indicator is zero through-out All regressions include state xed effects and time effects
Table VI shows the results of including the set of four
TABLE VIMULTIVARIATE REGRESSIONS OF THE EFFECTS OF DIFFERENT TYPES OF BRANCHING
DEREGULATION ON ECONOMIC GROWTH BANKING STRUCTUREAND BANK EFFICIENCY
Dependent variable
Growthin
personalincome
Log of of
bankingcompanies
Branchesper
bankingcompany
Log ofnon-
interestexpenses
Loan lossprovisions
(loans 1leases)
Merger amp acquisitionbranchingindicator
0013(0004)
2 0162(0037)
1146(0453)
2 0107(0035)
2 0007(0001)
Unrestricted state-wide branchingindicator
2 0008(0005)
0086(0059)
2 0043(0574)
0070(0037)
0005(0002)
Interstate bankingindicator
0006(0005)
2 0052(0057)
1064(0511)
0005(0030)
2 0001(0001)
Multibank holdingcompany indicator
2 0005(0006)
2 0096(0052)
2 0364(0529)
2 0095(0036)
2 0001(0001)
Dependent variablemean [median]
0086[0082]
4881[5136]
6649[5055]
13812[13803]
0009[0006]
This table contains regressions of state-level economic growth and state-level measures of the structureand efficiency of the banking industry Each dependent variable is regressed on a set of state indicatorvariables a set of year indicator variables and the set of four deregulation indicators these deregulationindicators are equal to one after deregulation and zero before If a state was always regulated the indicator iszero throughout if it was always deregulated the indicator is one throughout Sample period is 1975 to 1991and Delaware and South Dakota are dropped so N 5 833 Noninterest expenses become available only in1984 so N 5 392 The coefficient on the deregulation indicators is reported with robust standard errors inparentheses and denote statistically signicant at the 1 5 and 10 percent levels
QUARTERLY JOURNAL OF ECONOMICS1464
deregulation indicators simultaneously The branching by mergerand acquisition indicator is statistically signicant and of theexpected sign in all ve of the equations and larger in absolutevalue than any of the other coefficients on the indicators for theother types of deregulation Since there may be multicollinearityamong our indicators we also ran univariate specications whichincluded each indicator separately and the results are verysimilar These results imply that branching by merger andacquisition is the most important type of deregulation measuredby its consequences for economic growth banking structure andbank efficiency See Kroszner and Strahan [1998] for more details
UNIVERSITY OF CHICAGO GRADUATE SCHOOL OF BUSINESS AND NATIONAL BUREAU OF
ECONOMIC RESEARCH
FEDERAL RESERVE BANK OF NEW YORK RESEARCH amp MARKET ANALYSIS GROUP
REFERENCES
Abrams Burton A and Russell F Settle lsquolsquoPressure-Group Inuence and Institu-tional Change Branch-Banking Legislation during the Great DepressionrsquorsquoPublic Choice LXXVII (1993) 687ndash705
Amel Dean lsquolsquoState Laws Affecting the Geographic Expansion of CommercialBanksrsquorsquo unpublished manuscript Board of Governors of the Federal ReserveSystem 1993
Becker Gary S lsquolsquoA Theory of Competition among Pressure Groups for PoliticalInuencersquorsquo Quarterly Journal of Economics XCVIII (1983) 371ndash400
Berger Allen Anil Kashyap Joseph Scalise lsquolsquoThe Transformation of the U SBanking Industry What a Long Strange Trip Itrsquos Beenrsquorsquo Brookings Papers onEconomic Activity 2 (1995) 55ndash218
Berger Allen Anthony Saunders Joseph Scalise and Gregory Udell lsquolsquoThe Effectsof Bank Mergers and Acquisitions on Small Business Lendingrsquorsquo Journal ofFinancial Economics L (1998) 187ndash229
Brennan Geoffrey and James Buchanan lsquolsquoTowards a Tax Constitution forLeviathanrsquorsquo Journal of Public Economics VIII (1977) 255ndash273
Brickley James A and Christopher M James lsquolsquoThe Takeover Market CorporateBoard Composition and Ownership Structure The Case of Bankingrsquorsquo Journalof Law and Economics XXX (1987) 161ndash180
Cole Rebel and John Wolken lsquolsquoFinancial Services Used by Small BusinessesEvidence from the 1993 Survey of Small Business Financesrsquorsquo Federal ReserveBulletin LXXXI (1994) 629ndash676
Dixit Avinash The Making of Economic Policy A Transactions-Cost PoliticsPerspective (Cambridge MA MIT Press 1996)
Economides Nicolas R Glenn Hubbard and Darius Palia lsquolsquoThe Political Economyof Branching Restrictions and Deposit Insurancersquorsquo Journal of Law andEconomics XXIX (1996) 667ndash704
Flannery Mark lsquolsquoThe Social Costs of Unit Banking Restrictionsrsquorsquo Journal ofMonetary Economics XIII (1984) 237ndash249
Greene William H Econometric Analysis third edition (Upper Saddle River NJPrentice-Hall 1997)
Gunther Jeffery lsquolsquoRegional Capital Imbalances and the Removal of InterstateBanking Restrictionsrsquorsquo Economics Letters XLIV (1994) 439ndash442 lsquolsquoThe Erosion of Intrastate Branch Banking Restrictionsrsquorsquo unpublishedmanuscript Federal Reserve Bank of Dallas 1996
Irwin Douglas and Randall Kroszner lsquolsquoInterests Institutions and Ideology inSecuring Policy Change The Republican Conversion to Trade Liberalizationafter Smoot-Hawleyrsquorsquo Journal of Law and Economics XLII (1999) 643ndash673
WHAT DRIVES DEREGULATION 1465
James Christopher lsquolsquoAn Analysis of Intra-Industry Differences in the Effect ofRegulation The Case of Deposit Rate Ceilingsrsquorsquo Journal of Monetary Econom-ics XII (1983) 417ndash442
Jarrell Gregg lsquolsquoChange at the Exchange The Causes and Effects of DeregulationrsquorsquoJournal of Law and Economics XXVII (1984) 273ndash312
Jayaratne Jith and Philip E Strahan lsquolsquoThe Finance-Growth Nexus Evidencefrom Bank Branch Deregulationrsquorsquo Quarterly Journal of Economics CI (1996)639ndash670
Jayaratne Jith and Philip E Strahan lsquolsquoEntry Restrictions Industry Evolutionand Dynamic Efficiency Evidence from Commercial Bankingrsquorsquo Journal of Lawand Economics XLIX (1998) 239ndash274
Joskow Paul lsquolsquoComment on lsquoPower Struggles Explaining Deregulatory Reformsin Electricity Marketsrsquo rsquorsquo Brookings Papers on Economic Activity Microeconom-ics (1996) 251ndash264
Joskow Paul and Roger Noll lsquolsquoRegulation in Theory and Practice An Overviewrsquorsquoin Studies in Public Regulation Gary Fromm ed (Cambridge MA MITPress 1981) pp 1ndash65
Kahn Matthew E and John G Matsusaka lsquolsquoDemand for Environmental GoodsEvidence from Voting Patterns on California Initiativesrsquorsquo Journal of Law andEconomics XL (1997) 137ndash173
Kalbeisch J and R Prentice The Statistical Analysis of Failure Time Data (NewYork John Wiley amp Sons 1980)
Kane Edward J lsquolsquoDe Jure Interstate Banking Why Only NowrsquorsquoJournal of MoneyCredit and Banking XXVIII (1996) 141ndash161
Kiefer Nicholas lsquolsquoEconomic Duration Data and Hazard Functionsrsquorsquo Journal ofEconomic Literature XXVI (1988) 646ndash679
Kroszner Randall S lsquolsquoThe Evolution of Universal Banking and Its Regulation inTwentieth Century Americarsquorsquo in Universal Banking Financial System DesignReconsidered Anthony Saunders and Ingo Walter eds (New York IrwinProfessional Publishers 1996) pp 70ndash99 lsquolsquoThe Political Economy of Banking and Financial Regulation in the U Srsquorsquo inThe Banking and Financial Structure in the NAFTA Countries and ChileGeorge M von Furstenberg ed (Boston Kluwer Academic Publishers 1997)pp 200ndash213
Kroszner Randall S and Raghuram G Rajan lsquolsquoIs the Glass-Steagall ActJustied A Study of the U S Experience with Universal Banking before1933rsquorsquo American Economic Review LXXXIV (1994) 810ndash832
Kroszner Randall S and Raghuram G Rajan lsquolsquoOrganization Structure andCredibility Evidence from Bank Securities Activities before the Glass-Steagall Actrsquorsquo Journal of Monetary Economics XXXIX (1997) 475ndash516
Kroszner Randall S and Philip E Strahan lsquolsquoRegulatory Incentives and the ThriftCrisis Dividends Mutual-to-Stock Conversions and Financial DistressrsquorsquoJournal of Finance LI (1996) 1285ndash1320
Kroszner Randall S and Philip E Strahan lsquolsquoWhat Drives Deregulation Econom-ics and Politics of the Relaxation of Bank Branching Restrictionsrsquorsquo NBERWorking Paper No 6637 1998
Kroszner Randall S and Thomas Stratmann lsquolsquoInterest Group Competition andthe Organization of Congress Theory and Evidence from Financial ServicesPolitical Action Committeesrsquorsquo American Economic Review LXXXVIII (1998)pp 1163ndash1187
Kroszner Randall S and Thomas Stratmann lsquolsquoDoes Political Ambiguity PayCorporate Campaign Contributions and Rewards to Legislator Reputationrsquorsquounpublishedmanuscript University of Chicago Graduate School of Business1999
Lin D Y and L J Wei lsquolsquoThe Robust Inference for the Cox Proportional HazardsModel Journal of the American Statistical Association LXXXIV (1989)1074ndash1078
Macey Jonathan R and Geoffrey P Miller Banking Law and Regulation (BostonLittle Brown and Company 1992)
Makinson Larry Open Secrets The Encyclopedia of Congressional Money andPolitics (Washington DC Congressional Quarterly 1992)
Niskanen WilliamBureaucracy and Representative Government (Chicago Aldine-Atherton 1971)
QUARTERLY JOURNAL OF ECONOMICS1466
Noll Roger lsquolsquoComment on Peltzmanrsquorsquo Brookings Papers Microeconomics SpecialIssue (1989) 48ndash58
North Douglass lsquolsquoA Transactions Cost Theory of Politicsrsquorsquo Journal of TheoreticalPolitics II (1990) 355ndash367
Peltzman Sam lsquolsquoToward a More General Theory of Regulationrsquorsquo Journal of Lawand Economics XIX (1976) 109ndash148 lsquolsquoConstituent Interest and Congressional Votingrsquorsquo Journal of Law andEconomics XXVII (1984) 181ndash210 lsquolsquoThe Economic Theory of Regulation after a Decade of DeregulationrsquorsquoBrookings Papers Microeconomics Special Issue (1989) 1ndash41
Peterson Mitchell and Raghuram Rajan lsquolsquoThe Benets of Lending RelationshipsEvidence from Small Business Datarsquorsquo Journal of Finance XLIX (1994) 3ndash38
Poole Keith and Howard Rosenthal Congress A Political-Economic History ofRoll Call Voting (Oxford Oxford University Press 1997)
Stata Stata Reference Manual Release 5 (College Station TX Stata Press 1997)Stigler George J lsquolsquoThe Theory of Economic Regulationrsquorsquo Bell Journal of Economics
and Management Science II (1971) 3ndash21Stigler George J Chicago Studies in the Political Economy (Chicago University
of Chicago Press 1988)Strahan Philip E and James Weston lsquolsquoSmall Business Lending and the Changing
Structure of the Banking Industryrsquorsquo Journal of Banking and Finance XXII(1998) 821ndash845
Sylla Richard John Legler and John Wallis lsquolsquoBanks and State Public Finance inthe New Republic The United States 1790ndash1860rsquorsquo Journal of EconomicHistory XLVII (1987) 391ndash403
White Eugene The Regulation and Reform of the American Banking System1900ndash1929 (Princeton Princeton University Press 1983)
Winston Clifford lsquolsquoEconomic Deregulation Days of Reckoning for Microecono-mistsrsquorsquo Journal of Economic Literature XXXI (1993) 1263ndash1289
WHAT DRIVES DEREGULATION 1467
1996] While some of our ndings can be rationalized on public-interest grounds the private-interest theory provides a simpleway to understand these results as a whole1
II THE ORIGINS AND DEMISE OF GEOGRAPHICAL
RESTRICTIONS ON BANKING
A Origins in Public Financing Strategies
After the United States Constitution prevented the statesfrom issuing at money and from taxing interstate commercestates used their powers over banks to generate substantialrevenues [Sylla Legler and Wallis 1987]2 To enter the businessone had to obtain a bank charter States received fees for grantingcharters and they often owned or purchased shares in banks orlevied taxes on banks To enhance these revenues each state hadan interest in restricting competition among banks and many ofthe restrictions on the geographical expansion of banks originatein this period3 Since states received no charter fees from banksincorporated in other states the states prohibited out-of-statebanks from operating in their territoriesmdashhence the origin of theprohibition on interstate banking4
Legislatures also often restricted intrastate expansion Stateswould grant a charter for a specic location or limit bank branchesto that city or country By adopting branching restrictions statescreated a series of local monopolies from which they could extractpart of the rents Some state legislatures even passed lsquolsquounitbankingrsquorsquo laws that prevented a bank from having any branchesSuch regulations naturally produce beneciaries who are loatheto give up their protections and privileges Benets tend to be
1 Jarrell [1984] tests the private interest theory of regulatory exit using theend of xed commissions at the New York Stock Exchange Also see Peltzman[1989] and Noll [1989] Kanersquos [1996] lsquolsquoregulatory dialecticrsquorsquo emphasizes technologi-cal change interacting with private interests
2 During the rst third of the nineteenth century for example the bank-related share of total state revenues exceeded 10 percent in a dozen states InMassachusetts and Delaware a majority of total state revenue was bank-related
3 Noll [1989] has characterized conceiving of governments as distinctinterest groups concerned about nancing their expenditures as the LeviathanApproach see Niskanen [1971] and Brennan and Buchanan [1977]
4 With the passage of the National Banking Act in 1864 the federalgovernment also began to charter banks (motivated by a desire to use suchinstitutions to help to fund the Civil War see Kroszner [1997]) While there hadbeen some initial ambiguity concerning state authority over these institutions the1927 McFadden Act claried that until the 1994 Riegle-NealAct states effectivelyhad the right to prevent interstate branching and to force national banks toconform to state branching regulations See White [1983]
WHAT DRIVES DEREGULATION 1439
concentrated while costs to consumers of a less efficient andcompetitive nancial sector tend to be diffuse (eg Stigler [1971]and Peltzman [1976])
B A Brief History of Recent Branching Deregulation
Prior to the 1970s most states had laws restricting within-state branching and all states forbade interstate branching(Table I and Figure I) Although there had been some changes instate branching laws during the late nineteenth and early twenti-eth centuries these laws have remained stable since the GreatDepression Some of the statutes were essentially unchanged formore than a century Since the early 1970s however all but one ofthese states have relaxed restrictions on intrastate branching
Deregulation of these restrictions typically involves threetypes of reforms The rst concerns the formation of multibankholding companies (MBHCs) that were permitted to own multiplebanks but had to operate them separately The offices of the banksin an MBHC could not be integrated into a single network so adepositor at one bank would not have access to her deposits atanother The banks in an MBHC also could not consolidate theirback-office operations and each had to meet all regulatoryobligations eg capital requirements as if it were a stand-aloneinstitution
The second and most important step toward deregulationoccurs when states permit branching by merger and acquisitionthereby allowing MBHCs to convert offices of subsidiary banks(existing or acquired) into branches of a single bank An MBHCcould then integrate its banking offices into a single branchnetwork A third reform occurs when states permit full statewidebranching whereby banks could open new branches anywherewithin state borders5
The Douglas amendment to the Bank Holding Company Actof 1956 prevented holding companies from buying out-of-statebanks unless that state explicitly permitted such acquisitions bystatute (see Macey and Miller [1992]) Since no state allowed suchacquisitions holding companies were effectively prohibited fromcrossing state lines Deregulation began in 1975 when Mainepassed legislation permitting out-of-state bank holding compa-
5 Permitting branching only through merger and acquisition before fullstatewide branching could be interpreted as states allowing incumbent banks tomaintain the ability to extract at least a portion of the rents associated withbarriers to entry from purchasing banks
QUARTERLY JOURNAL OF ECONOMICS1440
TABLE IYEAR OF DEREGULATION OF RESTRICTIONS ON GEOGRAPHICAL EXPANSION BY STATE
State
Intrastatebranching
through MampA
Full intrastatebranchingpermitted
Interstatebanking
permitted
Multibank holdingcompaniespermitted
AL 1981 1990 1987 1970AK 1970 1970 1982 1970AZ 1970 1970 1986 1970AR 1994 1989 1985CA 1970 1970 1987 1970CO 1991 1988 1970CT 1980 1988 1983 1970DE 1970 1970 1988 1970DC 1970 1970 1985 1970FL 1988 1988 1985 1970GA 1983 1985 1976HI 1986 1986 1970ID 1970 1970 1985 1970IL 1988 1993 1986 1982IN 1989 1991 1986 1985IA 1991 1984KS 1987 1990 1992 1985KY 1990 1984 1984LA 1988 1988 1987 1985ME 1975 1975 1978 1970MD 1970 1970 1985 1970MA 1984 1984 1983 1970MI 1987 1988 1986 1971MN 1993 1986 1970MS 1986 1989 1988 1990MO 1990 1990 1986 1970MT 1990 1993 1970NE 1985 1990 1983NV 1970 1970 1985 1970NH 1987 1987 1987 1970NJ 1977 1986 1970NM 1991 1991 1989 1970NY 1976 1976 1982 1976NC 1970 1970 1985 1970ND 1987 1991 1970OH 1979 1989 1985 1970OK 1988 1987 1983OR 1985 1985 1986 1970PA 1982 1990 1986 1982RI 1970 1970 1984 1970SC 1970 1970 1986 1970SD 1970 1970 1988 1970TN 1985 1990 1985 1970TX 1988 1988 1987 1970UT 1981 1981 1984 1970VT 1970 1970 1988 1970VA 1978 1987 1985 1970WA 1985 1985 1987 1981WV 1987 1987 1988 1982WI 1990 1990 1987 1970WY 1988 1987 1970
States not yet deregulated Source Amel [1993] and updates by authors
WHAT DRIVES DEREGULATION 1441
nies to acquire Maine banks As part of the Garn-St GermainAct federal legislators in 1982 amended the Bank HoldingCompany Act to allow failed banks and thrifts to be acquired byany bank holding company regardless of state laws (see egKroszner and Strahan [1996]) Many states then entered regionalor national reciprocal arrangements whereby their banks could bebought by any other state in the arrangement Between 1984 and1988 38 states joined one of these arrangements (see Amel[1993])
Table I and Figure I illustrate the history of state deregula-tion of geographical restrictions since 1970 The state-by-statederegulation culminates in the phaseout of interstate bank-ing restrictions with the passage of the 1994 Riegle-Neal Inter-state Banking and Branching Efficiency Act which effectivelycodied at the national level what had been occurring at the statelevel6
6 The Act permitted states to pass legislation to opt out of the interstatebanking provisions if the legislature did so before the provisions were to go intoeffect in mid-1997 Since only Texas and Montana have passed opt-out legislationthe United States will now have nearly complete interstate banking and branching
FIGURE IDeregulation of Restrictions on Intrastate Branching
Permitted Intrastate Branching before 1970
QUARTERLY JOURNAL OF ECONOMICS1442
III HYPOTHESES AND VARIABLE DEFINITIONS
Our empirical tests are based on the timing of intrastatebranching deregulation through merger and acquisition We focuson this form of branching deregulation because it has a muchgreater economic impact than the other forms of branchingderegulation (see the Appendix) Deregulation through mergerand acquisition is the only type of branching deregulation thatconsistently has a statistically signicant effect on bankingstructure bank efficiency and overall economic growth and theestimated magnitudes of the effects are greatest for this type ofderegulation We also analyze the interstate deregulation thattakes place at the national level by modeling voting behavior inthe House of Representatives on federal interstate branchinglegislation7
In the remainder of this section we describe our hypothesesabout the private-interest public-interest and political-institu-tional theories along with the data sources used to construct thevariables Some of the variables help us to distinguish between apublic- and private-interest approach while others will be consis-tent with both approaches The political-institutional approachhas little overlap in its implications with the other theories Ourmethod is to assess the relative importance of factors representingeach approach in speeding or slowing deregulation
A Hypotheses
Intraindustry Rivalry Small banks have fought to maintainand extend branching restrictions both historically and in therecent debates8 Smaller banks appear to have been the mainwinners from antibranching laws since these restrictions protectthem from competition from larger and more efficient bankingorganizations (see Flannery [1984] Jayaratne and Strahan [1998]
7 We do not employ a hazard model to explain the timing of interstatederegulation for two reasons the time clustering of interstate deregulation andcomplications of the strategic interdependence of the statesrsquo behavior As thenumber of states in an interstate arrangement increases for example the effectson potential acquirers is ambiguous they might prefer a larger pool of potentialtarget banks but might have to pay more as the number of competing acquirersexpands The benet to potential targets tends to increase with the number statesin an arrangement since the expected selling price increases with the number ofpotential bidders [Brickley and James 1987]
8 Economides Hubbard and Palia [1996] provide evidence that voting inCongress for the 1927 McFadden Act responded to small state banksrsquo interest inlimiting competition from large national banks See also White [1983] and Abramsand Settle [1993] for historical opposition On the small bank opposition to therecent branching deregulation see Kane [1996] and The Economist [8694 p 59]
WHAT DRIVES DEREGULATION 1443
and Winston [1993]) Branching restrictions thus tend to reducethe efficiency and consumer convenience of the banking system9
The private-interest theory therefore predicts that reformshould occur later in those states where the strength of the smallbanks relative to that of the large banks is greater According tothe public-interest theory deregulation should take place earlierwhere small banks are relatively important because the socialcosts of the regulation are directly related to the size of theprotected sector The social costs include deadweight costs andlosses associated with inefficiencies in the production of bankingservices relative to production without geographical restrictions
Interindustry Rivalry A number of states permit state-chartered commercial banks to sell insurance The insurancelobby would thus oppose the relaxation of branching restrictionswhen banks can sell insurance because such deregulation mightpermit banks to provide a more efficient insurance distributionnetwork10 According to the private-interest theory reform shouldoccur later in states where banks can sell insurance and theinsurance industry is important relative to the banking industryThe efficiency costs of the branching restrictions however risewith the size of the insurance sector since the restrictions preventthe exploitation of scope economies Under the public-interesttheory reform should therefore occur earlier where banks can sellinsurance and the insurance industry is relatively large11
Consumers of Banking Services Banks are a major source ofcredit for small rms [Cole and Wolken 1994] Branching deregu-lation tends to reduce banksrsquo local market power [Jayaratne andStrahan 1998] In addition Strahan and Weston [1998] nd thatlending to small businesses increases on average when smallbanks are purchased by other banking organizations and Bergeret al [1998] nd that credit availability to small businessesincreases in the years following a takeover of a small bank by alarger banking organization Since bank borrowers tend to benet
9 Flannery [1984] shows that small banks in states with branching restric-tions have higher costs than small banks in states without such restrictions andthe Appendix shows how the production of banking services becomes more efficientfollowing deregulation
10 The Economist [8694 p 59] describes the insurance industryrsquos opposi-tion to branching deregulation
11 Another rival interest we considered was the Savings and Loan industryThe share of assets in Savings and Loans relative to banking plus Savings andLoan assets in the state however had only small and statistically insignicanteffects in the models we estimate below
QUARTERLY JOURNAL OF ECONOMICS1444
from branching deregulation in particular and bank consolidationin general the private-interest theory would predict that stateswith numerous small bank-dependent rms would deregulateearlier12 This prediction however is also consistent with thepublic-interest theory since the social costs of the restrictions arehigher in states with more small bank-dependent rms
Another effect of branching restrictions on bank customerscan be related to the prices paid for bank services High loanprices before deregulation may reect for example a largemarket share for inefficient high-cost banks or high rents beingearned by the banks protected from competition In either casethe public-interest theory would suggest that states with rela-tively high initial loan interest rates should deregulate earlierThe private-interest theory however does not have a clearimplication high rents could lead the beneciaries to ght harderto maintain them (eg Stigler [1971]) or the deadweight costsassociated with these high rents could make it more difficult forthe beneciaries to maintain the restrictions (eg Becker [1983])13
Bank Stability Geographic diversication through branch-ing could mitigate instability problems that were importantduring much of our sample period thereby improving welfareInstability also may reduce the incentives of banks to lobby tomaintain protections because unstable banks are less likely tosurvive to reap the benets of the restrictions politicians also mayanticipate that such banks are less likely to be able to providefuture contributions and support [Gunther 1994 1996]14 Both theprivate-interest and public-interest theories thus suggest thatderegulation is most likely in states where banking instability isgreatest
12 On the other hand local banking monopolies created by branchingrestrictions could strengthen relationshipsbetween banks and small and medium-sized rms and increase the availability of credit to these rms [Petersen andRajan 1994] Also some have argued that small business lending declines whenlarge banks take over small banks (eg Berger Kashyap Scalise [1995])
13 High loan interest rates also could reect compensation for higher risksand fewer opportunities for banks to diversify If state-specic risks explaindifferences in average loans rates then neither theory would have any predictionsfor the sign of this effect Following branching deregulation however loan interestrates tend to fall but bank risk proles do not appear to change [Jayaratne andStrahan 1998 Kroszner and Strahan 1998] This suggests that high loan rates inantibranching states are not simply reecting state-specic risks
14 Also deregulation can arise as a response to banking instability under atheory of ignorant or misinformed voters in which a banking crisis acts as aneducational device to make the previously ignorant public aware of the costs of theantibranching policy (see Kane [1996])
WHAT DRIVES DEREGULATION 1445
Political-Institutional Factors Republicans are typically per-ceived as more likely to favor deregulation than Democrats so thepolitical-institutional theories suggest that states that are con-trolled by Democrats deregulate later than those controlled byRepublicans15 In addition we investigate whether reform be-comes more likely when one party controls the legislature and thegovernorship Note that these political effects must be interpretedwith caution since the views of the politicians may simply reectthe economic interests of the constituents in the state (seePeltzman [1984])
B Variable Denitions and Data Sources
Our main proxy for the strength of the small banks is thefraction of banking assets in the state in lsquolsquosmallrsquorsquo banks We denesmall banks as banks with assets below the median size in eachstate By allowing the denition of small to vary across states wetake into account cross-state heterogeneity in bank sizes16 An-nual data on bank size are from the fourth quarter Reports ofIncome and Conditions (lsquolsquoCall Reportsrsquorsquo) from the Federal ReserveBoard We also include the capital-to-asset ratio of small banksrelative to large as a measure of their relative nancial healthSpecically this variable is the asset-weighted average capital-asset ratio for small banks minus the asset-weighted averagecapital-asset ratio for large banks Annual data on bank capitalare from the Call Reports
To measure the effects of the rival insurance industry we rstconstruct an indicator variable that is one if the state permitsbanks to sell insurance For each state we then measure the sizeof the insurance sector (total value added in the state) relative tothe sum of the banking plus insurance sectors We will examine
15 For more detail on the importance of legislative structures party politicsand ideology see Poole and Rosenthal [1997] Kahn and Matsusaka [1997] Irwinand Kroszner [1999] and Kroszner and Stratmann [1999]
16 We also considered a variety of other denitions of relative size but therelative size results reported below are not sensitive to which denition we useFirst we used a xed measure of small bank that did not vary across statesdening small banks as those with assets below $100 million in 1994 dollarsSecond we applied the state-varying and xed denitions at the level of lsquolsquobankingorganizationsrsquorsquo (which include multibank holding companies) instead of at the levelof banks Third we calculated gini coefficients of bank size inequality and bankconcentration indices as alternative proxies for small versus large bank power inthe state Finally we also considered banks in lsquolsquoruralrsquorsquo areas (that is not located in aBureau of the Census lsquolsquostandard metropolitan statistical arearsquorsquo) as small since thevalue of the restrictions might be greatest in protecting banks outside of cities fromentry by the city banks All of the alternative denitions are highly correlated andyield the same results on the importance of intraindustry differences
QUARTERLY JOURNAL OF ECONOMICS1446
the effect separately for states that permit banks to sell insuranceand those that do not Data on value-added by industry are fromU S Commerce Department Bureau of Economic Analysis Sur-vey of Current Business [August 1994]
To measure the relative importance of small bank-dependentborrowers we include the proportion of all establishments operat-ing in the state with fewer than twenty employees These data arecompiled by the Bureau of the Census17 As a measure of priceswe use the difference between the average interest rate on loans inthe state and the prevailing federal funds rate where the loanrate equals the ratio of total interest income on all domestic loansdivided by total domestic loans held by banks operating in thestate18 These data are from the end-of-year Call Reports butbecome available beginning only in 1976 Our proxy for bankinstability is the failure rate of banks measured as the percent oftotal state banking assets in failed institutions from the FederalReserve Boardrsquos National Information Center database19
We include two political variables to adjust for any indepen-dent inuence of party politics First we measure the degree ofparty control of the state government by the fraction of the threebodies of the state government (the assembly senate and gover-norship) controlled by Democrats This variable for example isone-third if the Democrats have a majority in the assembly andthe Republicans have a majority in the senate and hold thegovernorship Second we include an indicator variable equal toone if the same party controls the governorrsquos office and hasmajorities in both chambers of the state legislature20
17 We collected the establishment data by the state for three cross sections(1976 1982 and 1987) and interpolated the data in the intermediate years SeeState and Metropolitan Data Book 1982 1986 and 1991
18 A measure of how individual customers would be affected is interest ratesthey receive on deposits We calculated the average deposit interest rate in thestate from the end-of-year Call Reports and did not nd any relationship betweenthis variable and the timing of deregulation in our analysis below Since Jayaratneand Strahan [1998] nd that branching deregulation has no effect on deposit ratesour nding of no relationship between timing and rates result could be due toeither individual bank customers anticipating little pricing effect and not lobbyingor an inability of individual depositors to form an effective lobby
19 As alternative measures of distress we used (1) the bank failure ratemeasured as the number of failed institutions in the state relative to the totalnumber of banks (2) the overall capital-to-asset ratio of all banks in the state and(3) the growth in state personal income Our results are not sensitive to the choiceof distress proxy
20 We also estimated the models with three indicator variables to reect theparty control of the assembly senate and governorship separately The coefficientestimates on these variables are qualitatively similar to those we present below(that is Democratic control tends to slow deregulation) and the results on theinterest group variables are unaffected by the specication of the political
WHAT DRIVES DEREGULATION 1447
IV METHODS AND RESULTS
In this section we rst estimate how the variables describedabove inuence the timing of state-level branching deregulationusing a hazard model Next we discuss the robustness of theresults including a test of whether the ex post impact of deregula-tion on the different interest groups is consistent with ourinterpretation of the coefficients We then investigate federalinterstate branching deregulation with a probit model to explorewhether the factors that affect the timing of intrastate deregula-tion also inuence voting in the House of Representatives oninterstate deregulation
A Hazard Model
We consider the period from the beginning of our sample(1970) until deregulation as the lsquolsquoduration of regulationrsquorsquo or thelsquolsquotime until deregulationrsquorsquo The hazard rate h(t) is the likelihoodthat a state deregulates at time t given that the state has not yetderegulated [Kalbeisch and Prentice 1980 Kiefer 1988 Greene1997] To model the duration of regulation we must decide whatstructure if any to impose on the hazard function The Kaplan-Meier product-limit estimator provides a simple nonparametricway to estimate the shape of the hazard function over time[Greene 1997]21 Figure II graphs this estimate for our data andshows that the hazard function is relatively at in the early yearsand then grows steeper in the later years
A duration model that can approximate this shape is theWeibull proportional hazards model The hazard rate functiontakes the form
h[t x(t) b] 5 h0(t) exp[x(t)8b]
where x(t) is a vector of time-varying covariates b is a vector ofunknown parameters to be estimated and the baseline hazard
variables Note also that Nebraska has a nonpartisan unicameral legislature Weassume that the party of the Governor controls the state government for Nebraskaand set the unied government indicator to one As another alternative we triedincluding a variable that is one if party control of the legislature or governorshipchanges but it did not have an economically or statistically signicant effect anddid not affect any of the other results
21 Denote each of the K years in our sample period Tk and order them suchthat T1 T2 and so on Let nk be the number of states that have not yet deregulatedby Tk and dk be the number of states that deregulate in year Tk The Kaplan-Meierestimate of the hazard rate in each year is dknk In other words the hazard rate isthe ratio of the number of states that actually deregulate in each year to thenumber of states that have not yet deregulated in that year
QUARTERLY JOURNAL OF ECONOMICS1448
rate h0(t) is ptp 2 1 with shape parameter p that will be estimatedfrom the data When p 1 this model displays a monotonicallyincreasing hazard rate To estimate the parameters we maximizethe following log-likelihood function [Kiefer 1988]
L(b) 5 oi 5 1
N
di ln h[ti x(ti) b] 2 oi 5 1
N
e 0
tih[u x(u) b] du
where N indexes the number of states at risk at the beginning ofthe sample period di is an indicator equal to one for the statesthat deregulated by the end of the sample period and zero for thestates that do not deregulate (ie the censored observations) andti is the time of deregulation for the ith state This formulationassumes that the time until deregulation is independent acrossstates but the observations of the same state over time are notassumed to be independent In addition we use a robust estima-tion procedure to calculate the standard errors [Lin and Wei1989]
We exclude states that deregulated before 1970 from theanalysis so N equals 39 states During the sample period 36 ofthe 39 states deregulate We have three censored observations forwhich di is zero (Arkansas Iowa and Minnesota) but our results
FIGURE IIKaplan-Meier Estimate of Hazard Rate for Branching Deregulation
WHAT DRIVES DEREGULATION 1449
are not sensitive to the inclusion of these censored observationsSince we observe each state in each year up to and including theyear of deregulation we have a total of 637 observations Table IIreports summary statistics and correlation coefficients for theexplanatory variables we use in our analysis
We use the Weibull model rather than a model that does notimpose any structure on the baseline hazard rate so that we cancalculate the change in the expected time to deregulation for agiven change in the levels of the covariates22 In the Weibullmodel we can invert the hazard function and map it into the timedomain Rewriting the Weibull model in this way the log of thetime to deregulation T is a linear function of the political economyfactors and an error term ln (T ) 5 x8b 1 e23 Because we assumethat the baseline hazard rate is Weibull with a shape characterizedby p the new coefficients will be scaled by p that is b 5 2 bp
The b coefficients represent the percentage change in thetime to deregulation for a one-unit change in the covariates Apositive coefficient implies that an increase in the variableincreases the expected time until deregulation To gauge theeconomic importance of the effects we will multiply the b by thestandard deviation of the explanatory variables and then evaluatehow much the changes in the variables raise or lower the expectedtime to deregulation24 In Table III we report the b coefficients
22 The Cox proportional hazards model does not impose any structure on thebaseline hazard rate h0(t) and takes the following form
h[t x(t) b] 5 h0(t) exp [x(t)8b]
Common factors that affect the probability of deregulation in the same way for allstates therefore do not inuence the estimated coefficients The cost of the Coxmodel is that because there is no structure on the baseline we can calculatechanges in only relative hazard rates associated with changes in the covariatesThe Weibull model however provides sufficient structure so that we can translateour estimates into a log expected time metric that allows us to calculate the changein the expected time to deregulation for a given change in the covariates SeeKiefer [1988] Greene [1997] and Stata [1997] When we compare the coefficientsestimated by the Cox and Weibull models on our data they are very close (seeKroszner and Strahan [1998]) so imposing a Weibull structure does not appear todistort our estimates
23 The error term e is independent of x and has an extreme value distributionscaled by 1p (see Kiefer [1988] Greene [1997] and Stata [1997] for more details onthis log expected time parameterization of the Weibull model)
24 The transformation to the log expected time domain is only strictly correctwhen the covariates are constant over time Our interpretation of the coefficients ispossible in the time-varying model if we assume that a one-standard-deviationdifference would hold over the entire sample period that is the deregulationwould occur so many years earlier or later if the variable in question were higherby one standard deviation from 1970 to 1992 Note that the mean number of yearsto deregulation in the sample is 158
QUARTERLY JOURNAL OF ECONOMICS1450
TABLE IISUMMARY STATISTICS FOR VARIABLES IN HAZARD MODEL 1970ndash1992
Panel A Univariate statistics MeanStandarddeviation Minimum Maximum
(1) (2) (3) (4)Small bank asset share of all
banking assets in state 011 005 001 023Relative size of insurance to
banking plus insurance inthe state 047 009 024 080
Indicator is 1 if banks can sellinsurance in the state 020 040 0 1
Small rm share of the numberof rms in the state 088 002 078 095
Capital ratio of small banksrelative to large in the state 002 001 2 001 009
Share of state government con-trolled by Democrats 065 036 0 1
Indicator is 1 if state controlledby one party 054 050 0 1
Average yield on bank loans inthe state minus Fed fundsrate 002 002 2 004 008
Panel B CorrelationsSmallbank
Relativesize of
insuranceInsuranceindicator
Smallrm
Smallbank
capitalDem
controlSingleparty
(1) (2) (3) (4) (5) (6) (7)Small bank asset share
of all banking assetsin state 1
Relative size of insur-ance to banking plusinsurance in the state 2 034 1
Indicator is 1 if bankscan sell insurance inthe state 020 2 012 1
Small rm share of thenumber of rms inthe state 030 2 011 005 1
Capital ratio of smallbanks relative tolarge in the state 2 066 020 2 010 2 006 1
Share of state govern-ment controlled byDemocrats 2 006 001 2 018 2 018 005 1
Indicator is 1 if statecontrolled by oneparty 007 012 001 2 006 2 006 030 1
Average yield on bankloans in the stateminus Fed funds rate 2 001 2 015 003 022 2 010 2 003 2 006
The number of observations is 637 except for the average interest rate on bank loans which becomesavailable only in 1976 so the number of observations is 408
TABLE IIIHAZARD MODEL OF POLITICAL ECONOMY FACTORS AFFECTING THE TIMING OF STATE
BRANCHING DEREGULATION 1970ndash1992
(1) (2) (3) (4) (5) (6)Small bank asset share of
all banking assets instate
648(122)
519(137)
632(135)
516(118)
488(133)
445(088)
Capital ratio of smallbanks relative to largein the state
1325(348)
921(319)
1301(362)
1067(346)
1144(375)
906(381)
Relative size of insurancein states where banksmay sell insurance 0otherwise
324(161)
259(149)
315(165)
285(154)
272(159)
055(134)
Indicator is 1 if banksmay sell insurance inthe state
045(010)
037(011)
045(010)
038(008)
036(009)
001(012)
Relative size of insurancein states where banksmay not sell insurance0 otherwise
2 093(043)
2 090(033)
2 102(051)
2 093(048)
2 096(053)
2 004(039)
Small rm share of thenumber of rms in thestate
2 972(211)
2 614(215)
2 952(211)
2 943(230)
2 952(270)
2 1509(276)
Share of state govern-ment controlled byDemocrats
031(012)
023(011)
030(012)
027(011)
026(011)
011(013)
Indicator is 1 if state con-trolled by one party
2 004(010)
2 004(007)
2 004(007)
2 001(009)
2 002(010)
017(009)
Average yield on bankloans in the stateminus Fed funds rate mdash
023(319) mdash mdash mdash mdash
Bank failure rate mdash mdash092
(111) mdash mdash mdash
Indicator is 1 if state hasunit banking law mdash mdash mdash
018(010)
019(010)
030(009)
Indicator is 1 if statechanges bank insur-ance powers mdash mdash mdash mdash
2 008(012)
2 026(013)
Includes regionalindicators No No No No No Yes
N 637 408 637 637 637 637Log likelihood 2 374 900 2 360 2 212 2 191 1192p-value of x 2 for
regression 001 001 001 001 001 001
The hazard model is Weibull where the dependent variable is the log expected time to branchingderegulation ln (T) 5 bx 1 eAll variables are measured for each state in each year Small bank asset share isthe percent of banking assets in the state held by banks below the median size of banks in each state in eachyear Relative capital ratio is the capital to assets ratio of small banks minus that of large banks Size ofinsurance relative to banking plus insurance in the state is measured as gross state product from insurancedivided by gross state product from insurance plus banking Insurance indicator is 1 if state law permits banksto enter the insurance business Small rm share is the percent of all establishments in the state that havefewer than twenty employees Party control variable is the share of the three bodies of state governmentcontrolled by Democrats One party control indicator is 1 if the same party controlrsquos the governorship and hasmajorities in both chambers of the state legislature Average yield on bank loans equals total interest incomeon all domestic loans made by banks in the state divided by total loans Bank failure rate is the ratio of assetsin failed banks during a given state and year divided by total bank assets in that state and year Unit bankingindicator equals 1 for states with unit banking restrictions The change in insurance powers indicator equalsone if the state changed the law to permit banks to sell insurance during the sample period We divide thecountry into four regions and include regional indicators in column (5) Robust standard errors are inparentheses and denote statistically signicant at the 1 5 and 10 percent levels
QUARTERLY JOURNAL OF ECONOMICS1452
The positive and statistically signicant coefficient on smallbank share implies that a greater small bank share delaysderegulation and this effect is economically important A one-standard-deviation increase in the small bank share results in a30 percent increase in the time until deregulation or about 47years This result supports the implications of private-interesttheory concerning intraindustry rivalry and is contrary to thepublic-interest theory
The positive and statistically signicant coefficient on therelative performance of small banks in the state implies thatstates deregulate later when small banks are relatively strongnancially25 A one-standard-deviation increase in the relativecapital-to-asset ratio results in a 15 percent rise in the time untilderegulation or about 24 years This result is consistent with theprivate-interest theory but may also be consistent with a public-interest interpretation because the risk of failures (and a costlytaxpayer-funded bailout) increases as small bank nancial healthdeclines
Interindustry competition also helps explain the timing ofderegulation In states where banks can sell insurance a rela-tively large insurance sector is associated with an increase in theexpected time to deregulation26 A one-standard-deviation in-crease in the relative size of the insurance sector in those statesthat permit banks to sell insurance leads to a 22 percent increasein the time until deregulation or about 35 years27 This result isconsistent with the private-interest theory but not the public-interest theory
Deregulation occurs earlier in states where small bank-dependent rms are relatively numerous A one-standard-
25 We also allowed small bank and large bank capital-to-asset ratios to enterseparately Their coefficient estimates were nearly equal in absolute value and ofopposite signs and so the data do not reject the relative small to large bank capitalratio specication we have chosen
26 When we use the relative size of the insurance sector in the state as awhole rather than insurance relative to banking plus insurance in the state weobtain similar results When we simultaneously include both measures of therelative size of insurance the interactions of the lsquolsquoinsurance relative to bankingplus insurancersquorsquo variable continue to be statisticallysignicant but the interactionsof lsquolsquoinsurance relative to the state as a wholersquorsquo variable are not
27 We have also estimated the model with the indicator variable equal to onefor states where banks may sell insurance but without the variables measuringtheimportance of insurance relative to banking In this model the coefficient on theinsurance indicator variablemdashan estimate of the difference in the expected time toderegulation in states where banks may sell insurance relative to states wherebanks may not sell insurancemdashequals 027 and is statistically signicant at the 5percent level This suggests that branching deregulation takes about 45 yearslonger in states where banks may sell insurance
WHAT DRIVES DEREGULATION 1453
deviation increase in the share of small rms reduces the timeuntil deregulation by 18 percent or about three years This resultconcerning the interests of consumers of banking services isconsistent with both the private- and public-interest theories
Column (2) of Table III introduces the average interest rateon loans to the model as a rough proxy for the cost to bankborrowers of these regulations These data become available onlyin 1976 so our sample size shrinks to 408 observations Thecoefficient on this variable is small and not statistically signi-cant The public-interest theory would imply that the coefficientshould be negative while the implication of the economic theorydoes not have a clear sign prediction
When we add the failure rate of banks to the specication incolumn (3) its coefficient is small and statistically insignicantWe do not nd a linkage between the timing of state deregulationand statewide banking distress Given that deposit insurance isfunded federally however bank failures may raise the publicrsquosconcerns about the costs of the regulations at the national ratherthan the state level This interpretation can explain why the rateof bank failure does not affect the timing of state-level deregula-tion even though bank instability does appear to rise generallyduring the period of increasing deregulation (see Section V below)
Finally the partisan structure of the state government alsoinuences when states deregulate As expected a higher propor-tion of Democrats in the government tends to delay deregulationA one-standard-deviation rise in the share of the governmentcontrolled by Democrats slows the deregulation by about twoyears Whether the state is dominated by one party however doesnot appear to affect the timing of the deregulation
To summarize our ndings from the hazard model privateinterests appear to play an important role in the deregulatoryprocess While private interests and public interests do sometimescoincide the results on the relative share of small banks and largebanks and on the relative size of insurance where banks competeare consistent with a private-interest approach but are difficult toexplain on public-interest grounds
B Robustness Tests
Omitted Variables States begin the sample period withdifferent degrees of restrictions on bank expansion These initialconditions may proxy for omitted factors determining the political-economy equilibrium in the state that could then affect the
QUARTERLY JOURNAL OF ECONOMICS1454
subsequent timing of deregulation For instance some statesbegan the period with unit banking while other states allowedlimited branching and the formation of multibank holdingcompanies
To test whether the stringency of initial restrictions onbranching matters we rst divide the sample into states with andwithout the most extreme form of branching regulation unitbanking28 Column (4) of Table III includes an indicator variableequal to one if the state began the period with unit bankingrestrictions The coefficient on this variable is positive andstatistically signicant but the other results remain virtuallyunchanged Next we include an indicator variable that is one ifthe state permitted multibank holding companies (MBHCs) toexist at the beginning of our sample period Thirty-four statesbegan the period allowing MBHCs The coefficient estimate forthis indicator variable (not reported) is small and not statisticallysignicant and its inclusion does not change the effects of theother variables
Eleven states expand bank powers to include the sale ofinsurance products during our sample period This could proxy foromitted factors about the relative strength of the insuranceindustry in the state In column (5) we include an indicatorvariable that is one if the state changes the law during our sampleperiod to permit banks to sell insurance business and zerootherwise The coefficient estimate is small and statisticallyinsignicant and does not affect the other results29
In column (6) we introduce four regional indicator variablesNorth South Midwest and West30 The coefficients on small bankshare the relative capital-to-assets ratio and small rm sharechange little Although the coefficient estimate still suggests that
28 We classify states that prohibited branching but permitted banks toestablish facilities as unit banking states The unit banking indicator is one for thefollowing states CO AR FL IL IA KS MN MO MT NE ND OK TX WI WVand WY
29 We conducted a likelihood ratio test of whether the unit banking statesthe MBHC states and the states that changed their regulations of bank insurancepowers could be pooled with the other states In all three cases we could not rejectpooling of the data
30 Since intrastate deregulation generally preceded interstate deregulationand the latter typically took the form of regional interstate compacts thepossibility of participating in a regional interstate banking compact could haveinuenced the decision to deregulate intrastate branching Our denitions of theregions are region 1 (South) contains AL AR DC FL GA KY LA MS NC OKSC TN TX and VA region 2 (Northeast) contains CT MA MD ME NH NJ NYPA RI VT and WV region 3 (Midwest) contains IA IL IN KS MI MN MO NEND OH SD and WI region 4 (West) contains the other states
WHAT DRIVES DEREGULATION 1455
in states where banks can sell insurance a larger relative size ofthe insurance sector delays deregulation the effect is no longerstatistically signicant The coefficient on the indicator variablefor whether a state expands bank powers to permit insurancesales during the sample period however now becomes statisti-cally signicantly negative This implies that deregulation occursroughly 41 years earlier in states where the insurance lobby maybe relatively weak While the inclusion of the regional indicatorsdoes not affect the other results the results on insurance doappear to be sensitive to the specication of the hazard model31
Estimation Technique To check the robustness of using ahazard model and to investigate whether the time-series or thecross-sectional variation in each of the factors accounts for thehazard results we estimate two simple linear probability models(neither are reported) First we run a pooled time-series cross-section OLS regression using the 637 observations from thehazard model The dependent variable equals one in the year ofderegulation and zero otherwise We include the same set ofregressors that were used in the hazard model and add state-levelxed effects This specication thus removes the cross-statevariation from the independent variables and focuses on theirmovements over time In this model the signs of the coefficientson the interest group factors are consistent with the results fromthe hazard model and the coefficients for the small bank relativecapital ratio and for the relative size of insurance where states cansell insurance are statistically signicant
In the second model we eliminate the time-series variationby averaging each of the explanatory variables over time We thushave one observation per state that had branching restrictions asof 1970 (N 5 39) The dependent variable equals the inverse of thenumber of years until deregulation which represents an estimateof the statersquos probability of deregulating in a year under theassumption that this probability is constant over time and eachyear provides an independent observation32 Although we haveonly 39 observations the coefficients on the small bank marketshare small bank relative capital ratio and small rm share
31 As further robustness check we also tried a population density variablewhich Abrams and Settle [1993] found relevant for regulatory change during the1920s and 1930s but its coefficient was not statistically signicant and did notaffect the other results
32 For instance the dependent variable would take the value of 01 for astate that deregulated in 1980 For the three states that did not deregulate thedependent variable equals 0
QUARTERLY JOURNAL OF ECONOMICS1456
variables are statistically signicant The signs of the effects ofthe insurance variables are the same as those reported in thehazard model but not statistically signicant
The hazard model and linear probability models thus producevery similar results The estimates from the latter suggest thatthe effects for the small bank market share and small rm sharevariables are driven primarily by cross-state variation the effectof the insurance variable is driven primarily by time-seriesvariation and the small bank capital effect is driven by bothcross-sectional and time-series variation
Ex Post Consequences of Deregulation Finally to check theplausibility of our interpretation of the hazard results we con-sider whether the ex post consequences of deregulation areconsistent with the ex ante lobbying positions we attribute to eachinterest group (see Kroszner and Strahan [1998] for details) Wend that small banks lose market share following deregulationand in states where banks can enter the insurance business theinsurance sector shrinks relative to the banking sector followingderegulation Borrowers also benet from deregulation throughlower average interest rates on loans These ndings support theprivate-interest interpretation of the coefficients in the hazardmodel groups that will benet lobby to speed deregulation andthose that will be harmed lobby to slow it
C Voting on Interstate Branching Deregulation in the U SHouse of Representatives
We now examine whether the forces driving intrastate branch-ing deregulation also drive interstate deregulation Financialservices interests are active contributors and lobbyists in Washing-ton Their political action committees constitute the largest groupof contributors to legislators providing nearly 20 percent of totalcongressional campaign contributions [Makinson 1992] and muchof their lobbying effort involves competition among rival interestswithin nancial services (see Kroszner and Stratmann [1998])
As noted above after virtually all states adopted intra- andinterstate branching deregulation the 1994 Riegle-Neal Act re-pealed the 1927 McFadden Act to eliminate all barriers tointerstate banking and branching by 1997 The key votes concern-ing the Riegle-Neal Act were either voice votes or extremelylopsided so we could not estimate a voting model from them Anumber of bills and amendments related to interstate branching
WHAT DRIVES DEREGULATION 1457
had been debated in Congress during the years prior to thepassage of the Riegle-Neal Act but a search of the weekly BNABanking Reporter and the Congressional Record produced onlyone roll-call vote related to interstate branching that was notlopsided This vote occurred in the House on November 14 1991and concerned an amendment sponsored by Wylie (R-OH) andNeal (D-NC) to introduce interstate banking and branchingderegulation to a nancial services reform package33 Althoughthe amendment passed by 210 to 208 the bill to which it wasattached subsequently was defeated
To check for the inuence of the factors we considered in thestate-level reforms we examine both the sponsorship of interstatebanking legislation and voting on the amendment The sponsorsof the Wylie-Neal amendment are from states with low small bankshares 004 in Ohio (Wylie) and 002 in North Carolina (Neal) Incontrast the sample mean in 1991 is 008 (median 5 007)Michigan home state of the Senatersquos sponsor of the 1994 Riegle-Neal Act also had relatively low small bank strength (small bankshare of 005)
Table IV reports estimates of a probit model where thedependent variable equals one if the legislator voted in favor ofthe amendment and zero otherwise The explanatory variablesare the same as those in the hazard model except that we use aparty affiliation indicator variable equal to one for Democrats andzero for Republicans as our proxy for political factors Thecoefficients in Table IV are the marginal effects (lsquolsquoslopesrsquorsquo) of aone-unit change of each variable on the probability that alegislator will vote for the amendment34 Note that a forcefavoring deregulation has a negative coefficient in the hazardmodel (Table III) but a positive coefficient in the probit model
Consistent with the state-level deregulation process legisla-tors are more likely to support the amendment if their states havea relatively low share of small banks As in the hazard model thefraction of small banks is the most important interest groupinuence on a legislatorrsquos voting decision A one-standard-deviation increase in small banksrsquo market share (from the mean)is associated with a decline in the probability of voting in favor of
33 The Wylie-Neal amendment also included provisions limiting certaininsurance and real estate powers of national banks [Congressional RecordNovember 14 1991 pp 10239ndash10242]
34 Since we have multiple legislators from each state we adjust the standarderrors to correct for the lack of independence among observations clustered in thesame state
QUARTERLY JOURNAL OF ECONOMICS1458
TAB
LE
IVM
AR
GIN
AL
EF
FE
CT
SF
RO
MA
PR
OB
ITM
OD
EL
OF
TH
EIN
FL
UE
NC
EO
FP
OL
ITIC
AL
EC
ON
OM
YF
AC
TO
RS
ON
TH
EH
OU
SE
VO
TE
ON
TH
EW
YL
IE-N
EA
L
AM
EN
DM
EN
TT
OP
ER
MIT
INT
ER
ST
AT
EB
AN
KB
RA
NC
HIN
G
NO
VE
MB
ER
14
1991
(1)
(2)
(3)
(4)
(5)
(6)
Smal
lban
kas
set
shar
eof
allb
anki
ng
asse
tsin
stat
e2
417
(12
4)2
454
(12
2)2
407
(12
0)2
443
(14
7)2
450
(14
9)2
490
(14
7)C
apit
alra
tio
ofsm
allb
anks
rela
tive
tola
rge
inth
est
ate
180
(37
5)2
68(3
42)
256
(37
5)1
69(3
78)
153
(38
0)2
45(4
25)
Rel
ativ
esi
zeof
insu
ranc
ein
stat
esw
here
bank
sm
ayse
llin
sura
nce
0ot
herw
ise
22
34
(10
2)2
249
(1
14)
22
33
(10
1)2
249
(1
04)
22
51
(10
2)2
154
(11
1)In
dica
tor
is1
ifba
nks
may
sell
insu
ranc
ein
the
stat
e0
79
(0
06)
074
(00
6)0
69
(0
06)
070
(00
6)0
70
(0
06)
067
(00
7)R
elat
ive
size
ofin
sura
nce
inst
ates
whe
reba
nks
may
not
sell
insu
ranc
e0
oth
erw
ise
180
(06
6)2
42
(0
72)
154
(0
67)
173
(06
6)1
72
(0
67)
203
(0
91)
Smal
lrm
shar
eof
the
num
ber
ofr
ms
inth
est
ate
21
23(0
92)
20
83(0
89)
21
62(1
07)
21
32(0
95)
21
30(1
00)
20
94(1
16)
Indi
cato
ris
1if
Rep
rese
ntat
ive
isa
Dem
ocra
t2
035
(00
7)2
035
(00
7)2
036
(00
7)2
035
(00
7)2
035
(00
7)2
036
(00
7)A
vera
geyi
eld
onba
nklo
ans
inth
est
ate
min
usF
edfu
nds
rate
mdash10
01
(51
5)mdash
mdashmdash
mdash
Ave
rage
fail
ure
rate
198
5to
1991
mdashmdash
228
(14
1)mdash
mdashmdash
Indi
cato
ris
1if
stat
eh
asun
itba
nkin
gla
wmdash
mdashmdash
004
(00
7)0
04(0
07)
20
01(0
08)
Indi
cato
ris
1if
stat
ech
ange
sba
nk
insu
ran
cepo
wer
smdash
mdashmdash
mdash2
001
(00
7)2
003
(00
7)In
clud
esre
gion
alin
dica
tors
N
oN
oN
oN
oN
oYe
sP
seud
o-R
20
170
180
170
170
170
18L
ogli
keli
hood
224
024
223
826
223
951
224
013
224
010
223
808
p-va
lue
ofx2
for
regr
essi
on
001
0
01
001
0
01
001
0
01
Th
ede
pen
den
tvar
iabl
eis
one
ifth
ele
gisl
ator
vote
sfo
rth
eam
endm
enta
nd
zero
ifag
ains
tT
he
repo
rted
coef
fici
ents
are
the
effe
cts
ofa
uni
tch
ange
ofth
ein
depe
nde
ntva
riab
le(f
rom
the
mea
n)on
the
prob
abili
tyof
voti
ng
infa
vor
ofth
eam
endm
ent
All
vari
able
sar
em
easu
red
byst
ate
N5
418
the
num
ber
ofR
epre
sent
ativ
esvo
tin
gon
the
amen
dmen
tS
mal
lban
kas
sets
har
eis
the
perc
ent
ofba
nki
ng
asse
tsin
the
stat
eh
eld
byba
nks
belo
wth
em
edia
nsi
zeof
ban
kin
each
stat
ein
each
year
Rel
ativ
eca
pita
lrat
iois
the
capi
talt
oas
sets
rati
oof
smal
lban
ksm
inu
sth
atof
larg
eba
nks
Siz
eof
insu
ran
cere
lati
veto
bank
ing
plu
sin
sura
nce
inth
est
ate
ism
easu
red
asgr
oss
stat
epr
oduc
tfro
min
sura
nce
divi
ded
bygr
oss
stat
epr
odu
ctfr
omin
sura
nce
plu
sba
nkin
gIn
dica
tor
ison
eif
stat
ela
wpe
rmit
sba
nks
toen
ter
the
insu
ran
cebu
sin
ess
Sm
all
rmsh
are
isth
epe
rcen
tofa
lles
tabl
ishm
ents
inth
est
ate
that
hav
efe
wer
than
twen
tyem
ploy
ees
Par
tyin
dica
tor
ison
eif
the
Rep
rese
nta
tive
isa
Dem
ocra
tA
vera
geyi
eld
onba
nklo
ans
equa
lsto
tali
nter
est
inco
me
onal
ldom
esti
clo
ans
mad
eby
ban
ksin
the
stat
edi
vide
dby
tota
lloa
ns
The
aver
age
bank
failu
rera
teis
the
rati
oof
asse
tsin
faile
dba
nks
duri
ng
agi
ven
stat
edi
vide
dby
tota
lban
kas
sets
inth
atst
ate
aver
aged
over
the
year
s19
85to
1991
Un
itba
nki
ng
indi
cato
req
ual
son
efo
rst
ates
wit
hun
itba
nkin
gre
stri
ctio
nsT
hech
ange
inin
sura
nce
pow
ers
indi
cato
req
ual
son
eif
the
stat
ech
ange
dth
ela
wto
perm
itba
nks
tose
llin
sura
nce
duri
ngth
esa
mpl
epe
riod
We
divi
deth
eco
unt
ryin
tofo
urre
gion
san
din
clu
dere
gion
alin
dica
tors
inco
lum
n(5
)R
obu
stst
anda
rder
rors
are
inpa
ren
thes
es
an
d
den
ote
stat
isti
cally
sign
ica
nta
tth
e1
5an
d10
perc
entl
evel
s
WHAT DRIVES DEREGULATION 1459
branching of approximately 17 percent The relative capital-to-assets ratio of small and large banks however does not have astatistically signicant effect in the probit model
The impact of rival interests outside of banking is alsoconsistent with intrastate deregulation results Where banks cansell insurance legislators from states with larger insurancesectors relative to banking are less likely to vote for the amend-ment35 A one-standard-deviation increase in the relative size ofinsurance in those states which permit banks to sell insurancedecreases the probability that a legislator will favor the amend-ment by about 13 percent Turning to consumer interests thecoefficient on small rm share is not statistically signicant butthe coefficient on interest rates in the state is A one-standard-deviation increase in the average interest rate on loans raises theprobability that a legislator will support the amendment byroughly 8 percent The coefficient estimate for the bankingdistress variable measured as the average bank failure rate from1985 to 1991 is small and statistically signicant Overall theprobit analysis of the vote on national branching deregulationsupports the private-interest theory of deregulation and providesa consistency check that the importance of interests operating onthe state legislatures are very similar to those operating at thefederal level
V WHY DID BRANCHING DEREGULATION BEGIN DURING THE
1970S SHOCKS TO THE EQUILIBRIUM SUPPORTING BANK
BRANCHING RESTRICTIONS
A complete explanation of regulatory exit should account forwhy deregulation begins in the 1970s as well as the timing of thestate-by-state reforms [Kane 1996] In this section we describebroad technological legal and economic shocks that altered thepolitical-economy equilibrium which had kept antibranching regu-lations little changed for at least 30 years
Beginning in the 1970s three innovations reduced the valueto the protected banks of local geographic monopolies (see TableV) First automatic teller machines (ATMs) helped to erode thegeographic ties between customers and banks Second checkablemoney market mutual funds and the Merrill Lynch Cash Manage-ment Account demonstrated that banking by mail and telephone
35 The positive and statistically signicant estimates on the other insurancevariables may reect insurance industry support for the amendmentrsquos provisionslimiting banksrsquo insurance powers
QUARTERLY JOURNAL OF ECONOMICS1460
provided a convenient alternative to local banks36 Third techno-logical innovation and deregulation reduced transportation andcommunication costs particularly since the 1970s thereby lower-ing the costs for customers to use distant banks By increasing theelasticity of deposits supplied to banks these innovations reducedthe value of geographical restrictions to their traditional benecia-ries and thereby reduced their incentive to ght to maintain them[Peltzman 1976]
On the lending side increasing sophistication of credit-scoring techniques following innovations in information process-ing nancial theory and the development of large credit databases began to diminish the value of knowledge that localbankers had about the risks of borrowers in the community As a
36 Regulation Q which limited the interest rates that banks could pay ondeposits may have helped to drive depositors away from banks when the gapbetween market rates and deposit ceilings grew during the 1970s The eliminationof interest rate ceilings on large denomination certicates of deposit during the1970s appears to have hurt smaller and retail-oriented banks relative to largerwholesale banks [James 1983]
TABLE VBROAD TRENDS IN COMMERCIAL BANKING 1950ndash1995
YearNumberof ATMs
Domesticbank
deposits(billions)
Moneymarketmutual
fund(billions)
Percent ofdeposits 1
moneyfunds heldby banks
Smallbanksrsquo
percent ofbankingassets
Averagenumber of
bankfailures
(1) (2) (3) (4) (5) (6)1950 0 $ 154 $ 0 100 NA 41955 0 191 0 100 NA 31960 0 228 0 100 24 21965 0 330 0 100 20 41970 0 479 0 100 18 61975 9750 775 4 99 18 61980 18500 1182 76 94 17 101985 61117 1787 242 88 14 601990 80156 2339 493 83 11 1791995 122706 2552 745 77 8 61
Column (1) ATM gures are from Bank Network News The EFT Network Data Book (New YorkFaulkner and Gray Inc) The 1975 gure was unavailable 9750 is the number of ATMs in 1978 the rst yearfor which complete data are available Columns (2)ndash(4) Banksrsquo domestic deposits are from the Reports ofIncome and Condition money market mutual funds are from the Flow of Funds Data on all bank depositsforeign plus domestic are only available beginning in 1970 The trend in banksrsquo share (column (4)) is the sameusing total deposits instead of domestic deposits Column (5) Percent of banking assets held by small bankswhere a small bank is dened as a commercial bank with less than $100 million in assets in 1994 dollarsThese data are based on the Reports of Income and Condition Data on small banks are not available before1960 Column (6) Five-year average number of bank failures where the nal year is indicated in the rstcolumn These data are from FDIC Annual Report and the Quarterly Banking Prole
WHAT DRIVES DEREGULATION 1461
result of these innovations a national market developed forresidential mortgages credit card receivables have been securi-tized and bank lending to small business now relies less on thejudgment of loan officers and more on standardized scoringmodels
These changes have increased the potential protability forlarge banks to enter what had been the core of small bankactivities Large banksrsquo incentive to increase their lobbying pres-sure to be able to expand into these markets has thus beenincreasing over time In fact small banksrsquo market share began todecline even prior to the branching deregulation (Table V) As thevalue of a local banking relationships declined small rms thatwere the main borrowers from the small banks also probablybecame more likely to favor the entry of large banks into localmarkets With the deadweight costs of preventing large bankentry rising the private-interest theory predicts that small localbanks would become less likely able to maintain the branchingrestrictions [Becker 1983] Deregulation that reduces deadweightcosts of regulation is also consistent with the public-interesttheory
Kane [1996] argues that another major shock to the oldequilibrium was the increasing public awareness of the costlinessof having government-insured but (geographically) undiversiednancial institutions In the late 1970s the failure rate of banksbegan to rise (Table V) In the 1980s the Savings and Loan crisisand taxpayer bailout further heightened public awareness aboutthe costs of restrictions that make depository institutions morelikely to require infusions of taxpayer funds The failures thusmay have heightened public support for branching deregulation(see also Abrams and Settle [1993]) As noted above we did notnd any evidence that banking failures or distress in a stateaffected the speed with which the state deregulated but distressstill may be a common factor affecting all states
These technological economic and legal shocks generatedconditions that changed the long-standing balance favoring theantibranching forces The marginal value of lobbying to repealbranching restrictions increased just as the relative value to thesmall banks of maintaining branching restrictions was decliningThese nationwide shocks are common factors across states andthey are consistent with the positive duration dependence associ-ated with state-level deregulation that we demonstrated in FigureII In addition a broad change in lsquolsquoideologyrsquorsquo against government
QUARTERLY JOURNAL OF ECONOMICS1462
regulation could explain the positive duration dependence duringthe sample period
VI CONCLUSIONS
The private-interest theory of regulation can account for thepattern of bank branching deregulation of the 1970s and 1980sBeneciaries of branching regulation had supported a coalitionfavoring geographical restrictions despite their costs to consum-ers of nancial services Innovations that began in the 1970saltered the value of the restrictions to the affected parties and theresulting competition among interest groups can explain thesubsequent deregulation While some of our results are alsoconsistent with the public interest theorymdashfor example deregula-tion occurs earlier when small banks are in a relatively weaknancial positionmdashother results particularly evidence on theimportance of rivalries between small and large banks andbetween banking and insurance are difficult to explain with thepublic-interest approach We also nd that political-institutionalfactors affect deregulation although these variables may act asproxies for unmeasured economic interests The greater success ofthe private-interest theory to explain deregulation here than inprevious work may be due to the richer cross-sectional andtime-series variation in branching deregulation that can beexploited with a hazard model Future empirical studies ofendogenous deregulation thus may be most fruitful where changehas occurred across states over time such as in franchisinginsurance and public utilities (eg Joskow [1996])
Technological and nancial innovations will continue to erodethe benets to any interest group of maintaining regulatorybarriers in nancial services These forces are likely to bringabout reforms both domestically for example through legislationthat would increase bank powers (see Kroszner [1996] Krosznerand Rajan [1994 1997] and Kroszner and Stratmann [1998]) andinternationally for example through the extension of nancialservices provisions of NAFTAto reduce geographic barriers acrosscountries [Kroszner 1997]
APPENDIX EFFECTS OF DIFFERENT TYPES OF BRANCHING
DEREGULATION ON ECONOMIC GROWTH BANKING STRUCTUREAND BANK EFFICIENCY
We measure the importance of alternative forms of deregula-tion by their consequences for the economic growth and the
WHAT DRIVES DEREGULATION 1463
structure and efficiency of the banking sector across the states(see Jayaratne and Strahan [1996 1998]) To do so we run anOLS panel regression relating indicators for each type of deregula-tion to potential consequences Our dependent variables includeproxies for economic growth (growth in personal income) bankstructure (the log of the number of banking companies and thenumber of branches per banking company) and efficiency (the logof total noninterest expenses and the ratio of loan loss provisionsto total loans) The four deregulation indicators (MBHC branch-ing through merger and acquisition unrestricted statewide branch-ing and interstate banking) are equal to one in the years afterderegulation and zero before If a state deregulated before thesample period begins in 1970 the indicator is one throughout ifthe state does not deregulate then the indicator is zero through-out All regressions include state xed effects and time effects
Table VI shows the results of including the set of four
TABLE VIMULTIVARIATE REGRESSIONS OF THE EFFECTS OF DIFFERENT TYPES OF BRANCHING
DEREGULATION ON ECONOMIC GROWTH BANKING STRUCTUREAND BANK EFFICIENCY
Dependent variable
Growthin
personalincome
Log of of
bankingcompanies
Branchesper
bankingcompany
Log ofnon-
interestexpenses
Loan lossprovisions
(loans 1leases)
Merger amp acquisitionbranchingindicator
0013(0004)
2 0162(0037)
1146(0453)
2 0107(0035)
2 0007(0001)
Unrestricted state-wide branchingindicator
2 0008(0005)
0086(0059)
2 0043(0574)
0070(0037)
0005(0002)
Interstate bankingindicator
0006(0005)
2 0052(0057)
1064(0511)
0005(0030)
2 0001(0001)
Multibank holdingcompany indicator
2 0005(0006)
2 0096(0052)
2 0364(0529)
2 0095(0036)
2 0001(0001)
Dependent variablemean [median]
0086[0082]
4881[5136]
6649[5055]
13812[13803]
0009[0006]
This table contains regressions of state-level economic growth and state-level measures of the structureand efficiency of the banking industry Each dependent variable is regressed on a set of state indicatorvariables a set of year indicator variables and the set of four deregulation indicators these deregulationindicators are equal to one after deregulation and zero before If a state was always regulated the indicator iszero throughout if it was always deregulated the indicator is one throughout Sample period is 1975 to 1991and Delaware and South Dakota are dropped so N 5 833 Noninterest expenses become available only in1984 so N 5 392 The coefficient on the deregulation indicators is reported with robust standard errors inparentheses and denote statistically signicant at the 1 5 and 10 percent levels
QUARTERLY JOURNAL OF ECONOMICS1464
deregulation indicators simultaneously The branching by mergerand acquisition indicator is statistically signicant and of theexpected sign in all ve of the equations and larger in absolutevalue than any of the other coefficients on the indicators for theother types of deregulation Since there may be multicollinearityamong our indicators we also ran univariate specications whichincluded each indicator separately and the results are verysimilar These results imply that branching by merger andacquisition is the most important type of deregulation measuredby its consequences for economic growth banking structure andbank efficiency See Kroszner and Strahan [1998] for more details
UNIVERSITY OF CHICAGO GRADUATE SCHOOL OF BUSINESS AND NATIONAL BUREAU OF
ECONOMIC RESEARCH
FEDERAL RESERVE BANK OF NEW YORK RESEARCH amp MARKET ANALYSIS GROUP
REFERENCES
Abrams Burton A and Russell F Settle lsquolsquoPressure-Group Inuence and Institu-tional Change Branch-Banking Legislation during the Great DepressionrsquorsquoPublic Choice LXXVII (1993) 687ndash705
Amel Dean lsquolsquoState Laws Affecting the Geographic Expansion of CommercialBanksrsquorsquo unpublished manuscript Board of Governors of the Federal ReserveSystem 1993
Becker Gary S lsquolsquoA Theory of Competition among Pressure Groups for PoliticalInuencersquorsquo Quarterly Journal of Economics XCVIII (1983) 371ndash400
Berger Allen Anil Kashyap Joseph Scalise lsquolsquoThe Transformation of the U SBanking Industry What a Long Strange Trip Itrsquos Beenrsquorsquo Brookings Papers onEconomic Activity 2 (1995) 55ndash218
Berger Allen Anthony Saunders Joseph Scalise and Gregory Udell lsquolsquoThe Effectsof Bank Mergers and Acquisitions on Small Business Lendingrsquorsquo Journal ofFinancial Economics L (1998) 187ndash229
Brennan Geoffrey and James Buchanan lsquolsquoTowards a Tax Constitution forLeviathanrsquorsquo Journal of Public Economics VIII (1977) 255ndash273
Brickley James A and Christopher M James lsquolsquoThe Takeover Market CorporateBoard Composition and Ownership Structure The Case of Bankingrsquorsquo Journalof Law and Economics XXX (1987) 161ndash180
Cole Rebel and John Wolken lsquolsquoFinancial Services Used by Small BusinessesEvidence from the 1993 Survey of Small Business Financesrsquorsquo Federal ReserveBulletin LXXXI (1994) 629ndash676
Dixit Avinash The Making of Economic Policy A Transactions-Cost PoliticsPerspective (Cambridge MA MIT Press 1996)
Economides Nicolas R Glenn Hubbard and Darius Palia lsquolsquoThe Political Economyof Branching Restrictions and Deposit Insurancersquorsquo Journal of Law andEconomics XXIX (1996) 667ndash704
Flannery Mark lsquolsquoThe Social Costs of Unit Banking Restrictionsrsquorsquo Journal ofMonetary Economics XIII (1984) 237ndash249
Greene William H Econometric Analysis third edition (Upper Saddle River NJPrentice-Hall 1997)
Gunther Jeffery lsquolsquoRegional Capital Imbalances and the Removal of InterstateBanking Restrictionsrsquorsquo Economics Letters XLIV (1994) 439ndash442 lsquolsquoThe Erosion of Intrastate Branch Banking Restrictionsrsquorsquo unpublishedmanuscript Federal Reserve Bank of Dallas 1996
Irwin Douglas and Randall Kroszner lsquolsquoInterests Institutions and Ideology inSecuring Policy Change The Republican Conversion to Trade Liberalizationafter Smoot-Hawleyrsquorsquo Journal of Law and Economics XLII (1999) 643ndash673
WHAT DRIVES DEREGULATION 1465
James Christopher lsquolsquoAn Analysis of Intra-Industry Differences in the Effect ofRegulation The Case of Deposit Rate Ceilingsrsquorsquo Journal of Monetary Econom-ics XII (1983) 417ndash442
Jarrell Gregg lsquolsquoChange at the Exchange The Causes and Effects of DeregulationrsquorsquoJournal of Law and Economics XXVII (1984) 273ndash312
Jayaratne Jith and Philip E Strahan lsquolsquoThe Finance-Growth Nexus Evidencefrom Bank Branch Deregulationrsquorsquo Quarterly Journal of Economics CI (1996)639ndash670
Jayaratne Jith and Philip E Strahan lsquolsquoEntry Restrictions Industry Evolutionand Dynamic Efficiency Evidence from Commercial Bankingrsquorsquo Journal of Lawand Economics XLIX (1998) 239ndash274
Joskow Paul lsquolsquoComment on lsquoPower Struggles Explaining Deregulatory Reformsin Electricity Marketsrsquo rsquorsquo Brookings Papers on Economic Activity Microeconom-ics (1996) 251ndash264
Joskow Paul and Roger Noll lsquolsquoRegulation in Theory and Practice An Overviewrsquorsquoin Studies in Public Regulation Gary Fromm ed (Cambridge MA MITPress 1981) pp 1ndash65
Kahn Matthew E and John G Matsusaka lsquolsquoDemand for Environmental GoodsEvidence from Voting Patterns on California Initiativesrsquorsquo Journal of Law andEconomics XL (1997) 137ndash173
Kalbeisch J and R Prentice The Statistical Analysis of Failure Time Data (NewYork John Wiley amp Sons 1980)
Kane Edward J lsquolsquoDe Jure Interstate Banking Why Only NowrsquorsquoJournal of MoneyCredit and Banking XXVIII (1996) 141ndash161
Kiefer Nicholas lsquolsquoEconomic Duration Data and Hazard Functionsrsquorsquo Journal ofEconomic Literature XXVI (1988) 646ndash679
Kroszner Randall S lsquolsquoThe Evolution of Universal Banking and Its Regulation inTwentieth Century Americarsquorsquo in Universal Banking Financial System DesignReconsidered Anthony Saunders and Ingo Walter eds (New York IrwinProfessional Publishers 1996) pp 70ndash99 lsquolsquoThe Political Economy of Banking and Financial Regulation in the U Srsquorsquo inThe Banking and Financial Structure in the NAFTA Countries and ChileGeorge M von Furstenberg ed (Boston Kluwer Academic Publishers 1997)pp 200ndash213
Kroszner Randall S and Raghuram G Rajan lsquolsquoIs the Glass-Steagall ActJustied A Study of the U S Experience with Universal Banking before1933rsquorsquo American Economic Review LXXXIV (1994) 810ndash832
Kroszner Randall S and Raghuram G Rajan lsquolsquoOrganization Structure andCredibility Evidence from Bank Securities Activities before the Glass-Steagall Actrsquorsquo Journal of Monetary Economics XXXIX (1997) 475ndash516
Kroszner Randall S and Philip E Strahan lsquolsquoRegulatory Incentives and the ThriftCrisis Dividends Mutual-to-Stock Conversions and Financial DistressrsquorsquoJournal of Finance LI (1996) 1285ndash1320
Kroszner Randall S and Philip E Strahan lsquolsquoWhat Drives Deregulation Econom-ics and Politics of the Relaxation of Bank Branching Restrictionsrsquorsquo NBERWorking Paper No 6637 1998
Kroszner Randall S and Thomas Stratmann lsquolsquoInterest Group Competition andthe Organization of Congress Theory and Evidence from Financial ServicesPolitical Action Committeesrsquorsquo American Economic Review LXXXVIII (1998)pp 1163ndash1187
Kroszner Randall S and Thomas Stratmann lsquolsquoDoes Political Ambiguity PayCorporate Campaign Contributions and Rewards to Legislator Reputationrsquorsquounpublishedmanuscript University of Chicago Graduate School of Business1999
Lin D Y and L J Wei lsquolsquoThe Robust Inference for the Cox Proportional HazardsModel Journal of the American Statistical Association LXXXIV (1989)1074ndash1078
Macey Jonathan R and Geoffrey P Miller Banking Law and Regulation (BostonLittle Brown and Company 1992)
Makinson Larry Open Secrets The Encyclopedia of Congressional Money andPolitics (Washington DC Congressional Quarterly 1992)
Niskanen WilliamBureaucracy and Representative Government (Chicago Aldine-Atherton 1971)
QUARTERLY JOURNAL OF ECONOMICS1466
Noll Roger lsquolsquoComment on Peltzmanrsquorsquo Brookings Papers Microeconomics SpecialIssue (1989) 48ndash58
North Douglass lsquolsquoA Transactions Cost Theory of Politicsrsquorsquo Journal of TheoreticalPolitics II (1990) 355ndash367
Peltzman Sam lsquolsquoToward a More General Theory of Regulationrsquorsquo Journal of Lawand Economics XIX (1976) 109ndash148 lsquolsquoConstituent Interest and Congressional Votingrsquorsquo Journal of Law andEconomics XXVII (1984) 181ndash210 lsquolsquoThe Economic Theory of Regulation after a Decade of DeregulationrsquorsquoBrookings Papers Microeconomics Special Issue (1989) 1ndash41
Peterson Mitchell and Raghuram Rajan lsquolsquoThe Benets of Lending RelationshipsEvidence from Small Business Datarsquorsquo Journal of Finance XLIX (1994) 3ndash38
Poole Keith and Howard Rosenthal Congress A Political-Economic History ofRoll Call Voting (Oxford Oxford University Press 1997)
Stata Stata Reference Manual Release 5 (College Station TX Stata Press 1997)Stigler George J lsquolsquoThe Theory of Economic Regulationrsquorsquo Bell Journal of Economics
and Management Science II (1971) 3ndash21Stigler George J Chicago Studies in the Political Economy (Chicago University
of Chicago Press 1988)Strahan Philip E and James Weston lsquolsquoSmall Business Lending and the Changing
Structure of the Banking Industryrsquorsquo Journal of Banking and Finance XXII(1998) 821ndash845
Sylla Richard John Legler and John Wallis lsquolsquoBanks and State Public Finance inthe New Republic The United States 1790ndash1860rsquorsquo Journal of EconomicHistory XLVII (1987) 391ndash403
White Eugene The Regulation and Reform of the American Banking System1900ndash1929 (Princeton Princeton University Press 1983)
Winston Clifford lsquolsquoEconomic Deregulation Days of Reckoning for Microecono-mistsrsquorsquo Journal of Economic Literature XXXI (1993) 1263ndash1289
WHAT DRIVES DEREGULATION 1467
concentrated while costs to consumers of a less efficient andcompetitive nancial sector tend to be diffuse (eg Stigler [1971]and Peltzman [1976])
B A Brief History of Recent Branching Deregulation
Prior to the 1970s most states had laws restricting within-state branching and all states forbade interstate branching(Table I and Figure I) Although there had been some changes instate branching laws during the late nineteenth and early twenti-eth centuries these laws have remained stable since the GreatDepression Some of the statutes were essentially unchanged formore than a century Since the early 1970s however all but one ofthese states have relaxed restrictions on intrastate branching
Deregulation of these restrictions typically involves threetypes of reforms The rst concerns the formation of multibankholding companies (MBHCs) that were permitted to own multiplebanks but had to operate them separately The offices of the banksin an MBHC could not be integrated into a single network so adepositor at one bank would not have access to her deposits atanother The banks in an MBHC also could not consolidate theirback-office operations and each had to meet all regulatoryobligations eg capital requirements as if it were a stand-aloneinstitution
The second and most important step toward deregulationoccurs when states permit branching by merger and acquisitionthereby allowing MBHCs to convert offices of subsidiary banks(existing or acquired) into branches of a single bank An MBHCcould then integrate its banking offices into a single branchnetwork A third reform occurs when states permit full statewidebranching whereby banks could open new branches anywherewithin state borders5
The Douglas amendment to the Bank Holding Company Actof 1956 prevented holding companies from buying out-of-statebanks unless that state explicitly permitted such acquisitions bystatute (see Macey and Miller [1992]) Since no state allowed suchacquisitions holding companies were effectively prohibited fromcrossing state lines Deregulation began in 1975 when Mainepassed legislation permitting out-of-state bank holding compa-
5 Permitting branching only through merger and acquisition before fullstatewide branching could be interpreted as states allowing incumbent banks tomaintain the ability to extract at least a portion of the rents associated withbarriers to entry from purchasing banks
QUARTERLY JOURNAL OF ECONOMICS1440
TABLE IYEAR OF DEREGULATION OF RESTRICTIONS ON GEOGRAPHICAL EXPANSION BY STATE
State
Intrastatebranching
through MampA
Full intrastatebranchingpermitted
Interstatebanking
permitted
Multibank holdingcompaniespermitted
AL 1981 1990 1987 1970AK 1970 1970 1982 1970AZ 1970 1970 1986 1970AR 1994 1989 1985CA 1970 1970 1987 1970CO 1991 1988 1970CT 1980 1988 1983 1970DE 1970 1970 1988 1970DC 1970 1970 1985 1970FL 1988 1988 1985 1970GA 1983 1985 1976HI 1986 1986 1970ID 1970 1970 1985 1970IL 1988 1993 1986 1982IN 1989 1991 1986 1985IA 1991 1984KS 1987 1990 1992 1985KY 1990 1984 1984LA 1988 1988 1987 1985ME 1975 1975 1978 1970MD 1970 1970 1985 1970MA 1984 1984 1983 1970MI 1987 1988 1986 1971MN 1993 1986 1970MS 1986 1989 1988 1990MO 1990 1990 1986 1970MT 1990 1993 1970NE 1985 1990 1983NV 1970 1970 1985 1970NH 1987 1987 1987 1970NJ 1977 1986 1970NM 1991 1991 1989 1970NY 1976 1976 1982 1976NC 1970 1970 1985 1970ND 1987 1991 1970OH 1979 1989 1985 1970OK 1988 1987 1983OR 1985 1985 1986 1970PA 1982 1990 1986 1982RI 1970 1970 1984 1970SC 1970 1970 1986 1970SD 1970 1970 1988 1970TN 1985 1990 1985 1970TX 1988 1988 1987 1970UT 1981 1981 1984 1970VT 1970 1970 1988 1970VA 1978 1987 1985 1970WA 1985 1985 1987 1981WV 1987 1987 1988 1982WI 1990 1990 1987 1970WY 1988 1987 1970
States not yet deregulated Source Amel [1993] and updates by authors
WHAT DRIVES DEREGULATION 1441
nies to acquire Maine banks As part of the Garn-St GermainAct federal legislators in 1982 amended the Bank HoldingCompany Act to allow failed banks and thrifts to be acquired byany bank holding company regardless of state laws (see egKroszner and Strahan [1996]) Many states then entered regionalor national reciprocal arrangements whereby their banks could bebought by any other state in the arrangement Between 1984 and1988 38 states joined one of these arrangements (see Amel[1993])
Table I and Figure I illustrate the history of state deregula-tion of geographical restrictions since 1970 The state-by-statederegulation culminates in the phaseout of interstate bank-ing restrictions with the passage of the 1994 Riegle-Neal Inter-state Banking and Branching Efficiency Act which effectivelycodied at the national level what had been occurring at the statelevel6
6 The Act permitted states to pass legislation to opt out of the interstatebanking provisions if the legislature did so before the provisions were to go intoeffect in mid-1997 Since only Texas and Montana have passed opt-out legislationthe United States will now have nearly complete interstate banking and branching
FIGURE IDeregulation of Restrictions on Intrastate Branching
Permitted Intrastate Branching before 1970
QUARTERLY JOURNAL OF ECONOMICS1442
III HYPOTHESES AND VARIABLE DEFINITIONS
Our empirical tests are based on the timing of intrastatebranching deregulation through merger and acquisition We focuson this form of branching deregulation because it has a muchgreater economic impact than the other forms of branchingderegulation (see the Appendix) Deregulation through mergerand acquisition is the only type of branching deregulation thatconsistently has a statistically signicant effect on bankingstructure bank efficiency and overall economic growth and theestimated magnitudes of the effects are greatest for this type ofderegulation We also analyze the interstate deregulation thattakes place at the national level by modeling voting behavior inthe House of Representatives on federal interstate branchinglegislation7
In the remainder of this section we describe our hypothesesabout the private-interest public-interest and political-institu-tional theories along with the data sources used to construct thevariables Some of the variables help us to distinguish between apublic- and private-interest approach while others will be consis-tent with both approaches The political-institutional approachhas little overlap in its implications with the other theories Ourmethod is to assess the relative importance of factors representingeach approach in speeding or slowing deregulation
A Hypotheses
Intraindustry Rivalry Small banks have fought to maintainand extend branching restrictions both historically and in therecent debates8 Smaller banks appear to have been the mainwinners from antibranching laws since these restrictions protectthem from competition from larger and more efficient bankingorganizations (see Flannery [1984] Jayaratne and Strahan [1998]
7 We do not employ a hazard model to explain the timing of interstatederegulation for two reasons the time clustering of interstate deregulation andcomplications of the strategic interdependence of the statesrsquo behavior As thenumber of states in an interstate arrangement increases for example the effectson potential acquirers is ambiguous they might prefer a larger pool of potentialtarget banks but might have to pay more as the number of competing acquirersexpands The benet to potential targets tends to increase with the number statesin an arrangement since the expected selling price increases with the number ofpotential bidders [Brickley and James 1987]
8 Economides Hubbard and Palia [1996] provide evidence that voting inCongress for the 1927 McFadden Act responded to small state banksrsquo interest inlimiting competition from large national banks See also White [1983] and Abramsand Settle [1993] for historical opposition On the small bank opposition to therecent branching deregulation see Kane [1996] and The Economist [8694 p 59]
WHAT DRIVES DEREGULATION 1443
and Winston [1993]) Branching restrictions thus tend to reducethe efficiency and consumer convenience of the banking system9
The private-interest theory therefore predicts that reformshould occur later in those states where the strength of the smallbanks relative to that of the large banks is greater According tothe public-interest theory deregulation should take place earlierwhere small banks are relatively important because the socialcosts of the regulation are directly related to the size of theprotected sector The social costs include deadweight costs andlosses associated with inefficiencies in the production of bankingservices relative to production without geographical restrictions
Interindustry Rivalry A number of states permit state-chartered commercial banks to sell insurance The insurancelobby would thus oppose the relaxation of branching restrictionswhen banks can sell insurance because such deregulation mightpermit banks to provide a more efficient insurance distributionnetwork10 According to the private-interest theory reform shouldoccur later in states where banks can sell insurance and theinsurance industry is important relative to the banking industryThe efficiency costs of the branching restrictions however risewith the size of the insurance sector since the restrictions preventthe exploitation of scope economies Under the public-interesttheory reform should therefore occur earlier where banks can sellinsurance and the insurance industry is relatively large11
Consumers of Banking Services Banks are a major source ofcredit for small rms [Cole and Wolken 1994] Branching deregu-lation tends to reduce banksrsquo local market power [Jayaratne andStrahan 1998] In addition Strahan and Weston [1998] nd thatlending to small businesses increases on average when smallbanks are purchased by other banking organizations and Bergeret al [1998] nd that credit availability to small businessesincreases in the years following a takeover of a small bank by alarger banking organization Since bank borrowers tend to benet
9 Flannery [1984] shows that small banks in states with branching restric-tions have higher costs than small banks in states without such restrictions andthe Appendix shows how the production of banking services becomes more efficientfollowing deregulation
10 The Economist [8694 p 59] describes the insurance industryrsquos opposi-tion to branching deregulation
11 Another rival interest we considered was the Savings and Loan industryThe share of assets in Savings and Loans relative to banking plus Savings andLoan assets in the state however had only small and statistically insignicanteffects in the models we estimate below
QUARTERLY JOURNAL OF ECONOMICS1444
from branching deregulation in particular and bank consolidationin general the private-interest theory would predict that stateswith numerous small bank-dependent rms would deregulateearlier12 This prediction however is also consistent with thepublic-interest theory since the social costs of the restrictions arehigher in states with more small bank-dependent rms
Another effect of branching restrictions on bank customerscan be related to the prices paid for bank services High loanprices before deregulation may reect for example a largemarket share for inefficient high-cost banks or high rents beingearned by the banks protected from competition In either casethe public-interest theory would suggest that states with rela-tively high initial loan interest rates should deregulate earlierThe private-interest theory however does not have a clearimplication high rents could lead the beneciaries to ght harderto maintain them (eg Stigler [1971]) or the deadweight costsassociated with these high rents could make it more difficult forthe beneciaries to maintain the restrictions (eg Becker [1983])13
Bank Stability Geographic diversication through branch-ing could mitigate instability problems that were importantduring much of our sample period thereby improving welfareInstability also may reduce the incentives of banks to lobby tomaintain protections because unstable banks are less likely tosurvive to reap the benets of the restrictions politicians also mayanticipate that such banks are less likely to be able to providefuture contributions and support [Gunther 1994 1996]14 Both theprivate-interest and public-interest theories thus suggest thatderegulation is most likely in states where banking instability isgreatest
12 On the other hand local banking monopolies created by branchingrestrictions could strengthen relationshipsbetween banks and small and medium-sized rms and increase the availability of credit to these rms [Petersen andRajan 1994] Also some have argued that small business lending declines whenlarge banks take over small banks (eg Berger Kashyap Scalise [1995])
13 High loan interest rates also could reect compensation for higher risksand fewer opportunities for banks to diversify If state-specic risks explaindifferences in average loans rates then neither theory would have any predictionsfor the sign of this effect Following branching deregulation however loan interestrates tend to fall but bank risk proles do not appear to change [Jayaratne andStrahan 1998 Kroszner and Strahan 1998] This suggests that high loan rates inantibranching states are not simply reecting state-specic risks
14 Also deregulation can arise as a response to banking instability under atheory of ignorant or misinformed voters in which a banking crisis acts as aneducational device to make the previously ignorant public aware of the costs of theantibranching policy (see Kane [1996])
WHAT DRIVES DEREGULATION 1445
Political-Institutional Factors Republicans are typically per-ceived as more likely to favor deregulation than Democrats so thepolitical-institutional theories suggest that states that are con-trolled by Democrats deregulate later than those controlled byRepublicans15 In addition we investigate whether reform be-comes more likely when one party controls the legislature and thegovernorship Note that these political effects must be interpretedwith caution since the views of the politicians may simply reectthe economic interests of the constituents in the state (seePeltzman [1984])
B Variable Denitions and Data Sources
Our main proxy for the strength of the small banks is thefraction of banking assets in the state in lsquolsquosmallrsquorsquo banks We denesmall banks as banks with assets below the median size in eachstate By allowing the denition of small to vary across states wetake into account cross-state heterogeneity in bank sizes16 An-nual data on bank size are from the fourth quarter Reports ofIncome and Conditions (lsquolsquoCall Reportsrsquorsquo) from the Federal ReserveBoard We also include the capital-to-asset ratio of small banksrelative to large as a measure of their relative nancial healthSpecically this variable is the asset-weighted average capital-asset ratio for small banks minus the asset-weighted averagecapital-asset ratio for large banks Annual data on bank capitalare from the Call Reports
To measure the effects of the rival insurance industry we rstconstruct an indicator variable that is one if the state permitsbanks to sell insurance For each state we then measure the sizeof the insurance sector (total value added in the state) relative tothe sum of the banking plus insurance sectors We will examine
15 For more detail on the importance of legislative structures party politicsand ideology see Poole and Rosenthal [1997] Kahn and Matsusaka [1997] Irwinand Kroszner [1999] and Kroszner and Stratmann [1999]
16 We also considered a variety of other denitions of relative size but therelative size results reported below are not sensitive to which denition we useFirst we used a xed measure of small bank that did not vary across statesdening small banks as those with assets below $100 million in 1994 dollarsSecond we applied the state-varying and xed denitions at the level of lsquolsquobankingorganizationsrsquorsquo (which include multibank holding companies) instead of at the levelof banks Third we calculated gini coefficients of bank size inequality and bankconcentration indices as alternative proxies for small versus large bank power inthe state Finally we also considered banks in lsquolsquoruralrsquorsquo areas (that is not located in aBureau of the Census lsquolsquostandard metropolitan statistical arearsquorsquo) as small since thevalue of the restrictions might be greatest in protecting banks outside of cities fromentry by the city banks All of the alternative denitions are highly correlated andyield the same results on the importance of intraindustry differences
QUARTERLY JOURNAL OF ECONOMICS1446
the effect separately for states that permit banks to sell insuranceand those that do not Data on value-added by industry are fromU S Commerce Department Bureau of Economic Analysis Sur-vey of Current Business [August 1994]
To measure the relative importance of small bank-dependentborrowers we include the proportion of all establishments operat-ing in the state with fewer than twenty employees These data arecompiled by the Bureau of the Census17 As a measure of priceswe use the difference between the average interest rate on loans inthe state and the prevailing federal funds rate where the loanrate equals the ratio of total interest income on all domestic loansdivided by total domestic loans held by banks operating in thestate18 These data are from the end-of-year Call Reports butbecome available beginning only in 1976 Our proxy for bankinstability is the failure rate of banks measured as the percent oftotal state banking assets in failed institutions from the FederalReserve Boardrsquos National Information Center database19
We include two political variables to adjust for any indepen-dent inuence of party politics First we measure the degree ofparty control of the state government by the fraction of the threebodies of the state government (the assembly senate and gover-norship) controlled by Democrats This variable for example isone-third if the Democrats have a majority in the assembly andthe Republicans have a majority in the senate and hold thegovernorship Second we include an indicator variable equal toone if the same party controls the governorrsquos office and hasmajorities in both chambers of the state legislature20
17 We collected the establishment data by the state for three cross sections(1976 1982 and 1987) and interpolated the data in the intermediate years SeeState and Metropolitan Data Book 1982 1986 and 1991
18 A measure of how individual customers would be affected is interest ratesthey receive on deposits We calculated the average deposit interest rate in thestate from the end-of-year Call Reports and did not nd any relationship betweenthis variable and the timing of deregulation in our analysis below Since Jayaratneand Strahan [1998] nd that branching deregulation has no effect on deposit ratesour nding of no relationship between timing and rates result could be due toeither individual bank customers anticipating little pricing effect and not lobbyingor an inability of individual depositors to form an effective lobby
19 As alternative measures of distress we used (1) the bank failure ratemeasured as the number of failed institutions in the state relative to the totalnumber of banks (2) the overall capital-to-asset ratio of all banks in the state and(3) the growth in state personal income Our results are not sensitive to the choiceof distress proxy
20 We also estimated the models with three indicator variables to reect theparty control of the assembly senate and governorship separately The coefficientestimates on these variables are qualitatively similar to those we present below(that is Democratic control tends to slow deregulation) and the results on theinterest group variables are unaffected by the specication of the political
WHAT DRIVES DEREGULATION 1447
IV METHODS AND RESULTS
In this section we rst estimate how the variables describedabove inuence the timing of state-level branching deregulationusing a hazard model Next we discuss the robustness of theresults including a test of whether the ex post impact of deregula-tion on the different interest groups is consistent with ourinterpretation of the coefficients We then investigate federalinterstate branching deregulation with a probit model to explorewhether the factors that affect the timing of intrastate deregula-tion also inuence voting in the House of Representatives oninterstate deregulation
A Hazard Model
We consider the period from the beginning of our sample(1970) until deregulation as the lsquolsquoduration of regulationrsquorsquo or thelsquolsquotime until deregulationrsquorsquo The hazard rate h(t) is the likelihoodthat a state deregulates at time t given that the state has not yetderegulated [Kalbeisch and Prentice 1980 Kiefer 1988 Greene1997] To model the duration of regulation we must decide whatstructure if any to impose on the hazard function The Kaplan-Meier product-limit estimator provides a simple nonparametricway to estimate the shape of the hazard function over time[Greene 1997]21 Figure II graphs this estimate for our data andshows that the hazard function is relatively at in the early yearsand then grows steeper in the later years
A duration model that can approximate this shape is theWeibull proportional hazards model The hazard rate functiontakes the form
h[t x(t) b] 5 h0(t) exp[x(t)8b]
where x(t) is a vector of time-varying covariates b is a vector ofunknown parameters to be estimated and the baseline hazard
variables Note also that Nebraska has a nonpartisan unicameral legislature Weassume that the party of the Governor controls the state government for Nebraskaand set the unied government indicator to one As another alternative we triedincluding a variable that is one if party control of the legislature or governorshipchanges but it did not have an economically or statistically signicant effect anddid not affect any of the other results
21 Denote each of the K years in our sample period Tk and order them suchthat T1 T2 and so on Let nk be the number of states that have not yet deregulatedby Tk and dk be the number of states that deregulate in year Tk The Kaplan-Meierestimate of the hazard rate in each year is dknk In other words the hazard rate isthe ratio of the number of states that actually deregulate in each year to thenumber of states that have not yet deregulated in that year
QUARTERLY JOURNAL OF ECONOMICS1448
rate h0(t) is ptp 2 1 with shape parameter p that will be estimatedfrom the data When p 1 this model displays a monotonicallyincreasing hazard rate To estimate the parameters we maximizethe following log-likelihood function [Kiefer 1988]
L(b) 5 oi 5 1
N
di ln h[ti x(ti) b] 2 oi 5 1
N
e 0
tih[u x(u) b] du
where N indexes the number of states at risk at the beginning ofthe sample period di is an indicator equal to one for the statesthat deregulated by the end of the sample period and zero for thestates that do not deregulate (ie the censored observations) andti is the time of deregulation for the ith state This formulationassumes that the time until deregulation is independent acrossstates but the observations of the same state over time are notassumed to be independent In addition we use a robust estima-tion procedure to calculate the standard errors [Lin and Wei1989]
We exclude states that deregulated before 1970 from theanalysis so N equals 39 states During the sample period 36 ofthe 39 states deregulate We have three censored observations forwhich di is zero (Arkansas Iowa and Minnesota) but our results
FIGURE IIKaplan-Meier Estimate of Hazard Rate for Branching Deregulation
WHAT DRIVES DEREGULATION 1449
are not sensitive to the inclusion of these censored observationsSince we observe each state in each year up to and including theyear of deregulation we have a total of 637 observations Table IIreports summary statistics and correlation coefficients for theexplanatory variables we use in our analysis
We use the Weibull model rather than a model that does notimpose any structure on the baseline hazard rate so that we cancalculate the change in the expected time to deregulation for agiven change in the levels of the covariates22 In the Weibullmodel we can invert the hazard function and map it into the timedomain Rewriting the Weibull model in this way the log of thetime to deregulation T is a linear function of the political economyfactors and an error term ln (T ) 5 x8b 1 e23 Because we assumethat the baseline hazard rate is Weibull with a shape characterizedby p the new coefficients will be scaled by p that is b 5 2 bp
The b coefficients represent the percentage change in thetime to deregulation for a one-unit change in the covariates Apositive coefficient implies that an increase in the variableincreases the expected time until deregulation To gauge theeconomic importance of the effects we will multiply the b by thestandard deviation of the explanatory variables and then evaluatehow much the changes in the variables raise or lower the expectedtime to deregulation24 In Table III we report the b coefficients
22 The Cox proportional hazards model does not impose any structure on thebaseline hazard rate h0(t) and takes the following form
h[t x(t) b] 5 h0(t) exp [x(t)8b]
Common factors that affect the probability of deregulation in the same way for allstates therefore do not inuence the estimated coefficients The cost of the Coxmodel is that because there is no structure on the baseline we can calculatechanges in only relative hazard rates associated with changes in the covariatesThe Weibull model however provides sufficient structure so that we can translateour estimates into a log expected time metric that allows us to calculate the changein the expected time to deregulation for a given change in the covariates SeeKiefer [1988] Greene [1997] and Stata [1997] When we compare the coefficientsestimated by the Cox and Weibull models on our data they are very close (seeKroszner and Strahan [1998]) so imposing a Weibull structure does not appear todistort our estimates
23 The error term e is independent of x and has an extreme value distributionscaled by 1p (see Kiefer [1988] Greene [1997] and Stata [1997] for more details onthis log expected time parameterization of the Weibull model)
24 The transformation to the log expected time domain is only strictly correctwhen the covariates are constant over time Our interpretation of the coefficients ispossible in the time-varying model if we assume that a one-standard-deviationdifference would hold over the entire sample period that is the deregulationwould occur so many years earlier or later if the variable in question were higherby one standard deviation from 1970 to 1992 Note that the mean number of yearsto deregulation in the sample is 158
QUARTERLY JOURNAL OF ECONOMICS1450
TABLE IISUMMARY STATISTICS FOR VARIABLES IN HAZARD MODEL 1970ndash1992
Panel A Univariate statistics MeanStandarddeviation Minimum Maximum
(1) (2) (3) (4)Small bank asset share of all
banking assets in state 011 005 001 023Relative size of insurance to
banking plus insurance inthe state 047 009 024 080
Indicator is 1 if banks can sellinsurance in the state 020 040 0 1
Small rm share of the numberof rms in the state 088 002 078 095
Capital ratio of small banksrelative to large in the state 002 001 2 001 009
Share of state government con-trolled by Democrats 065 036 0 1
Indicator is 1 if state controlledby one party 054 050 0 1
Average yield on bank loans inthe state minus Fed fundsrate 002 002 2 004 008
Panel B CorrelationsSmallbank
Relativesize of
insuranceInsuranceindicator
Smallrm
Smallbank
capitalDem
controlSingleparty
(1) (2) (3) (4) (5) (6) (7)Small bank asset share
of all banking assetsin state 1
Relative size of insur-ance to banking plusinsurance in the state 2 034 1
Indicator is 1 if bankscan sell insurance inthe state 020 2 012 1
Small rm share of thenumber of rms inthe state 030 2 011 005 1
Capital ratio of smallbanks relative tolarge in the state 2 066 020 2 010 2 006 1
Share of state govern-ment controlled byDemocrats 2 006 001 2 018 2 018 005 1
Indicator is 1 if statecontrolled by oneparty 007 012 001 2 006 2 006 030 1
Average yield on bankloans in the stateminus Fed funds rate 2 001 2 015 003 022 2 010 2 003 2 006
The number of observations is 637 except for the average interest rate on bank loans which becomesavailable only in 1976 so the number of observations is 408
TABLE IIIHAZARD MODEL OF POLITICAL ECONOMY FACTORS AFFECTING THE TIMING OF STATE
BRANCHING DEREGULATION 1970ndash1992
(1) (2) (3) (4) (5) (6)Small bank asset share of
all banking assets instate
648(122)
519(137)
632(135)
516(118)
488(133)
445(088)
Capital ratio of smallbanks relative to largein the state
1325(348)
921(319)
1301(362)
1067(346)
1144(375)
906(381)
Relative size of insurancein states where banksmay sell insurance 0otherwise
324(161)
259(149)
315(165)
285(154)
272(159)
055(134)
Indicator is 1 if banksmay sell insurance inthe state
045(010)
037(011)
045(010)
038(008)
036(009)
001(012)
Relative size of insurancein states where banksmay not sell insurance0 otherwise
2 093(043)
2 090(033)
2 102(051)
2 093(048)
2 096(053)
2 004(039)
Small rm share of thenumber of rms in thestate
2 972(211)
2 614(215)
2 952(211)
2 943(230)
2 952(270)
2 1509(276)
Share of state govern-ment controlled byDemocrats
031(012)
023(011)
030(012)
027(011)
026(011)
011(013)
Indicator is 1 if state con-trolled by one party
2 004(010)
2 004(007)
2 004(007)
2 001(009)
2 002(010)
017(009)
Average yield on bankloans in the stateminus Fed funds rate mdash
023(319) mdash mdash mdash mdash
Bank failure rate mdash mdash092
(111) mdash mdash mdash
Indicator is 1 if state hasunit banking law mdash mdash mdash
018(010)
019(010)
030(009)
Indicator is 1 if statechanges bank insur-ance powers mdash mdash mdash mdash
2 008(012)
2 026(013)
Includes regionalindicators No No No No No Yes
N 637 408 637 637 637 637Log likelihood 2 374 900 2 360 2 212 2 191 1192p-value of x 2 for
regression 001 001 001 001 001 001
The hazard model is Weibull where the dependent variable is the log expected time to branchingderegulation ln (T) 5 bx 1 eAll variables are measured for each state in each year Small bank asset share isthe percent of banking assets in the state held by banks below the median size of banks in each state in eachyear Relative capital ratio is the capital to assets ratio of small banks minus that of large banks Size ofinsurance relative to banking plus insurance in the state is measured as gross state product from insurancedivided by gross state product from insurance plus banking Insurance indicator is 1 if state law permits banksto enter the insurance business Small rm share is the percent of all establishments in the state that havefewer than twenty employees Party control variable is the share of the three bodies of state governmentcontrolled by Democrats One party control indicator is 1 if the same party controlrsquos the governorship and hasmajorities in both chambers of the state legislature Average yield on bank loans equals total interest incomeon all domestic loans made by banks in the state divided by total loans Bank failure rate is the ratio of assetsin failed banks during a given state and year divided by total bank assets in that state and year Unit bankingindicator equals 1 for states with unit banking restrictions The change in insurance powers indicator equalsone if the state changed the law to permit banks to sell insurance during the sample period We divide thecountry into four regions and include regional indicators in column (5) Robust standard errors are inparentheses and denote statistically signicant at the 1 5 and 10 percent levels
QUARTERLY JOURNAL OF ECONOMICS1452
The positive and statistically signicant coefficient on smallbank share implies that a greater small bank share delaysderegulation and this effect is economically important A one-standard-deviation increase in the small bank share results in a30 percent increase in the time until deregulation or about 47years This result supports the implications of private-interesttheory concerning intraindustry rivalry and is contrary to thepublic-interest theory
The positive and statistically signicant coefficient on therelative performance of small banks in the state implies thatstates deregulate later when small banks are relatively strongnancially25 A one-standard-deviation increase in the relativecapital-to-asset ratio results in a 15 percent rise in the time untilderegulation or about 24 years This result is consistent with theprivate-interest theory but may also be consistent with a public-interest interpretation because the risk of failures (and a costlytaxpayer-funded bailout) increases as small bank nancial healthdeclines
Interindustry competition also helps explain the timing ofderegulation In states where banks can sell insurance a rela-tively large insurance sector is associated with an increase in theexpected time to deregulation26 A one-standard-deviation in-crease in the relative size of the insurance sector in those statesthat permit banks to sell insurance leads to a 22 percent increasein the time until deregulation or about 35 years27 This result isconsistent with the private-interest theory but not the public-interest theory
Deregulation occurs earlier in states where small bank-dependent rms are relatively numerous A one-standard-
25 We also allowed small bank and large bank capital-to-asset ratios to enterseparately Their coefficient estimates were nearly equal in absolute value and ofopposite signs and so the data do not reject the relative small to large bank capitalratio specication we have chosen
26 When we use the relative size of the insurance sector in the state as awhole rather than insurance relative to banking plus insurance in the state weobtain similar results When we simultaneously include both measures of therelative size of insurance the interactions of the lsquolsquoinsurance relative to bankingplus insurancersquorsquo variable continue to be statisticallysignicant but the interactionsof lsquolsquoinsurance relative to the state as a wholersquorsquo variable are not
27 We have also estimated the model with the indicator variable equal to onefor states where banks may sell insurance but without the variables measuringtheimportance of insurance relative to banking In this model the coefficient on theinsurance indicator variablemdashan estimate of the difference in the expected time toderegulation in states where banks may sell insurance relative to states wherebanks may not sell insurancemdashequals 027 and is statistically signicant at the 5percent level This suggests that branching deregulation takes about 45 yearslonger in states where banks may sell insurance
WHAT DRIVES DEREGULATION 1453
deviation increase in the share of small rms reduces the timeuntil deregulation by 18 percent or about three years This resultconcerning the interests of consumers of banking services isconsistent with both the private- and public-interest theories
Column (2) of Table III introduces the average interest rateon loans to the model as a rough proxy for the cost to bankborrowers of these regulations These data become available onlyin 1976 so our sample size shrinks to 408 observations Thecoefficient on this variable is small and not statistically signi-cant The public-interest theory would imply that the coefficientshould be negative while the implication of the economic theorydoes not have a clear sign prediction
When we add the failure rate of banks to the specication incolumn (3) its coefficient is small and statistically insignicantWe do not nd a linkage between the timing of state deregulationand statewide banking distress Given that deposit insurance isfunded federally however bank failures may raise the publicrsquosconcerns about the costs of the regulations at the national ratherthan the state level This interpretation can explain why the rateof bank failure does not affect the timing of state-level deregula-tion even though bank instability does appear to rise generallyduring the period of increasing deregulation (see Section V below)
Finally the partisan structure of the state government alsoinuences when states deregulate As expected a higher propor-tion of Democrats in the government tends to delay deregulationA one-standard-deviation rise in the share of the governmentcontrolled by Democrats slows the deregulation by about twoyears Whether the state is dominated by one party however doesnot appear to affect the timing of the deregulation
To summarize our ndings from the hazard model privateinterests appear to play an important role in the deregulatoryprocess While private interests and public interests do sometimescoincide the results on the relative share of small banks and largebanks and on the relative size of insurance where banks competeare consistent with a private-interest approach but are difficult toexplain on public-interest grounds
B Robustness Tests
Omitted Variables States begin the sample period withdifferent degrees of restrictions on bank expansion These initialconditions may proxy for omitted factors determining the political-economy equilibrium in the state that could then affect the
QUARTERLY JOURNAL OF ECONOMICS1454
subsequent timing of deregulation For instance some statesbegan the period with unit banking while other states allowedlimited branching and the formation of multibank holdingcompanies
To test whether the stringency of initial restrictions onbranching matters we rst divide the sample into states with andwithout the most extreme form of branching regulation unitbanking28 Column (4) of Table III includes an indicator variableequal to one if the state began the period with unit bankingrestrictions The coefficient on this variable is positive andstatistically signicant but the other results remain virtuallyunchanged Next we include an indicator variable that is one ifthe state permitted multibank holding companies (MBHCs) toexist at the beginning of our sample period Thirty-four statesbegan the period allowing MBHCs The coefficient estimate forthis indicator variable (not reported) is small and not statisticallysignicant and its inclusion does not change the effects of theother variables
Eleven states expand bank powers to include the sale ofinsurance products during our sample period This could proxy foromitted factors about the relative strength of the insuranceindustry in the state In column (5) we include an indicatorvariable that is one if the state changes the law during our sampleperiod to permit banks to sell insurance business and zerootherwise The coefficient estimate is small and statisticallyinsignicant and does not affect the other results29
In column (6) we introduce four regional indicator variablesNorth South Midwest and West30 The coefficients on small bankshare the relative capital-to-assets ratio and small rm sharechange little Although the coefficient estimate still suggests that
28 We classify states that prohibited branching but permitted banks toestablish facilities as unit banking states The unit banking indicator is one for thefollowing states CO AR FL IL IA KS MN MO MT NE ND OK TX WI WVand WY
29 We conducted a likelihood ratio test of whether the unit banking statesthe MBHC states and the states that changed their regulations of bank insurancepowers could be pooled with the other states In all three cases we could not rejectpooling of the data
30 Since intrastate deregulation generally preceded interstate deregulationand the latter typically took the form of regional interstate compacts thepossibility of participating in a regional interstate banking compact could haveinuenced the decision to deregulate intrastate branching Our denitions of theregions are region 1 (South) contains AL AR DC FL GA KY LA MS NC OKSC TN TX and VA region 2 (Northeast) contains CT MA MD ME NH NJ NYPA RI VT and WV region 3 (Midwest) contains IA IL IN KS MI MN MO NEND OH SD and WI region 4 (West) contains the other states
WHAT DRIVES DEREGULATION 1455
in states where banks can sell insurance a larger relative size ofthe insurance sector delays deregulation the effect is no longerstatistically signicant The coefficient on the indicator variablefor whether a state expands bank powers to permit insurancesales during the sample period however now becomes statisti-cally signicantly negative This implies that deregulation occursroughly 41 years earlier in states where the insurance lobby maybe relatively weak While the inclusion of the regional indicatorsdoes not affect the other results the results on insurance doappear to be sensitive to the specication of the hazard model31
Estimation Technique To check the robustness of using ahazard model and to investigate whether the time-series or thecross-sectional variation in each of the factors accounts for thehazard results we estimate two simple linear probability models(neither are reported) First we run a pooled time-series cross-section OLS regression using the 637 observations from thehazard model The dependent variable equals one in the year ofderegulation and zero otherwise We include the same set ofregressors that were used in the hazard model and add state-levelxed effects This specication thus removes the cross-statevariation from the independent variables and focuses on theirmovements over time In this model the signs of the coefficientson the interest group factors are consistent with the results fromthe hazard model and the coefficients for the small bank relativecapital ratio and for the relative size of insurance where states cansell insurance are statistically signicant
In the second model we eliminate the time-series variationby averaging each of the explanatory variables over time We thushave one observation per state that had branching restrictions asof 1970 (N 5 39) The dependent variable equals the inverse of thenumber of years until deregulation which represents an estimateof the statersquos probability of deregulating in a year under theassumption that this probability is constant over time and eachyear provides an independent observation32 Although we haveonly 39 observations the coefficients on the small bank marketshare small bank relative capital ratio and small rm share
31 As further robustness check we also tried a population density variablewhich Abrams and Settle [1993] found relevant for regulatory change during the1920s and 1930s but its coefficient was not statistically signicant and did notaffect the other results
32 For instance the dependent variable would take the value of 01 for astate that deregulated in 1980 For the three states that did not deregulate thedependent variable equals 0
QUARTERLY JOURNAL OF ECONOMICS1456
variables are statistically signicant The signs of the effects ofthe insurance variables are the same as those reported in thehazard model but not statistically signicant
The hazard model and linear probability models thus producevery similar results The estimates from the latter suggest thatthe effects for the small bank market share and small rm sharevariables are driven primarily by cross-state variation the effectof the insurance variable is driven primarily by time-seriesvariation and the small bank capital effect is driven by bothcross-sectional and time-series variation
Ex Post Consequences of Deregulation Finally to check theplausibility of our interpretation of the hazard results we con-sider whether the ex post consequences of deregulation areconsistent with the ex ante lobbying positions we attribute to eachinterest group (see Kroszner and Strahan [1998] for details) Wend that small banks lose market share following deregulationand in states where banks can enter the insurance business theinsurance sector shrinks relative to the banking sector followingderegulation Borrowers also benet from deregulation throughlower average interest rates on loans These ndings support theprivate-interest interpretation of the coefficients in the hazardmodel groups that will benet lobby to speed deregulation andthose that will be harmed lobby to slow it
C Voting on Interstate Branching Deregulation in the U SHouse of Representatives
We now examine whether the forces driving intrastate branch-ing deregulation also drive interstate deregulation Financialservices interests are active contributors and lobbyists in Washing-ton Their political action committees constitute the largest groupof contributors to legislators providing nearly 20 percent of totalcongressional campaign contributions [Makinson 1992] and muchof their lobbying effort involves competition among rival interestswithin nancial services (see Kroszner and Stratmann [1998])
As noted above after virtually all states adopted intra- andinterstate branching deregulation the 1994 Riegle-Neal Act re-pealed the 1927 McFadden Act to eliminate all barriers tointerstate banking and branching by 1997 The key votes concern-ing the Riegle-Neal Act were either voice votes or extremelylopsided so we could not estimate a voting model from them Anumber of bills and amendments related to interstate branching
WHAT DRIVES DEREGULATION 1457
had been debated in Congress during the years prior to thepassage of the Riegle-Neal Act but a search of the weekly BNABanking Reporter and the Congressional Record produced onlyone roll-call vote related to interstate branching that was notlopsided This vote occurred in the House on November 14 1991and concerned an amendment sponsored by Wylie (R-OH) andNeal (D-NC) to introduce interstate banking and branchingderegulation to a nancial services reform package33 Althoughthe amendment passed by 210 to 208 the bill to which it wasattached subsequently was defeated
To check for the inuence of the factors we considered in thestate-level reforms we examine both the sponsorship of interstatebanking legislation and voting on the amendment The sponsorsof the Wylie-Neal amendment are from states with low small bankshares 004 in Ohio (Wylie) and 002 in North Carolina (Neal) Incontrast the sample mean in 1991 is 008 (median 5 007)Michigan home state of the Senatersquos sponsor of the 1994 Riegle-Neal Act also had relatively low small bank strength (small bankshare of 005)
Table IV reports estimates of a probit model where thedependent variable equals one if the legislator voted in favor ofthe amendment and zero otherwise The explanatory variablesare the same as those in the hazard model except that we use aparty affiliation indicator variable equal to one for Democrats andzero for Republicans as our proxy for political factors Thecoefficients in Table IV are the marginal effects (lsquolsquoslopesrsquorsquo) of aone-unit change of each variable on the probability that alegislator will vote for the amendment34 Note that a forcefavoring deregulation has a negative coefficient in the hazardmodel (Table III) but a positive coefficient in the probit model
Consistent with the state-level deregulation process legisla-tors are more likely to support the amendment if their states havea relatively low share of small banks As in the hazard model thefraction of small banks is the most important interest groupinuence on a legislatorrsquos voting decision A one-standard-deviation increase in small banksrsquo market share (from the mean)is associated with a decline in the probability of voting in favor of
33 The Wylie-Neal amendment also included provisions limiting certaininsurance and real estate powers of national banks [Congressional RecordNovember 14 1991 pp 10239ndash10242]
34 Since we have multiple legislators from each state we adjust the standarderrors to correct for the lack of independence among observations clustered in thesame state
QUARTERLY JOURNAL OF ECONOMICS1458
TAB
LE
IVM
AR
GIN
AL
EF
FE
CT
SF
RO
MA
PR
OB
ITM
OD
EL
OF
TH
EIN
FL
UE
NC
EO
FP
OL
ITIC
AL
EC
ON
OM
YF
AC
TO
RS
ON
TH
EH
OU
SE
VO
TE
ON
TH
EW
YL
IE-N
EA
L
AM
EN
DM
EN
TT
OP
ER
MIT
INT
ER
ST
AT
EB
AN
KB
RA
NC
HIN
G
NO
VE
MB
ER
14
1991
(1)
(2)
(3)
(4)
(5)
(6)
Smal
lban
kas
set
shar
eof
allb
anki
ng
asse
tsin
stat
e2
417
(12
4)2
454
(12
2)2
407
(12
0)2
443
(14
7)2
450
(14
9)2
490
(14
7)C
apit
alra
tio
ofsm
allb
anks
rela
tive
tola
rge
inth
est
ate
180
(37
5)2
68(3
42)
256
(37
5)1
69(3
78)
153
(38
0)2
45(4
25)
Rel
ativ
esi
zeof
insu
ranc
ein
stat
esw
here
bank
sm
ayse
llin
sura
nce
0ot
herw
ise
22
34
(10
2)2
249
(1
14)
22
33
(10
1)2
249
(1
04)
22
51
(10
2)2
154
(11
1)In
dica
tor
is1
ifba
nks
may
sell
insu
ranc
ein
the
stat
e0
79
(0
06)
074
(00
6)0
69
(0
06)
070
(00
6)0
70
(0
06)
067
(00
7)R
elat
ive
size
ofin
sura
nce
inst
ates
whe
reba
nks
may
not
sell
insu
ranc
e0
oth
erw
ise
180
(06
6)2
42
(0
72)
154
(0
67)
173
(06
6)1
72
(0
67)
203
(0
91)
Smal
lrm
shar
eof
the
num
ber
ofr
ms
inth
est
ate
21
23(0
92)
20
83(0
89)
21
62(1
07)
21
32(0
95)
21
30(1
00)
20
94(1
16)
Indi
cato
ris
1if
Rep
rese
ntat
ive
isa
Dem
ocra
t2
035
(00
7)2
035
(00
7)2
036
(00
7)2
035
(00
7)2
035
(00
7)2
036
(00
7)A
vera
geyi
eld
onba
nklo
ans
inth
est
ate
min
usF
edfu
nds
rate
mdash10
01
(51
5)mdash
mdashmdash
mdash
Ave
rage
fail
ure
rate
198
5to
1991
mdashmdash
228
(14
1)mdash
mdashmdash
Indi
cato
ris
1if
stat
eh
asun
itba
nkin
gla
wmdash
mdashmdash
004
(00
7)0
04(0
07)
20
01(0
08)
Indi
cato
ris
1if
stat
ech
ange
sba
nk
insu
ran
cepo
wer
smdash
mdashmdash
mdash2
001
(00
7)2
003
(00
7)In
clud
esre
gion
alin
dica
tors
N
oN
oN
oN
oN
oYe
sP
seud
o-R
20
170
180
170
170
170
18L
ogli
keli
hood
224
024
223
826
223
951
224
013
224
010
223
808
p-va
lue
ofx2
for
regr
essi
on
001
0
01
001
0
01
001
0
01
Th
ede
pen
den
tvar
iabl
eis
one
ifth
ele
gisl
ator
vote
sfo
rth
eam
endm
enta
nd
zero
ifag
ains
tT
he
repo
rted
coef
fici
ents
are
the
effe
cts
ofa
uni
tch
ange
ofth
ein
depe
nde
ntva
riab
le(f
rom
the
mea
n)on
the
prob
abili
tyof
voti
ng
infa
vor
ofth
eam
endm
ent
All
vari
able
sar
em
easu
red
byst
ate
N5
418
the
num
ber
ofR
epre
sent
ativ
esvo
tin
gon
the
amen
dmen
tS
mal
lban
kas
sets
har
eis
the
perc
ent
ofba
nki
ng
asse
tsin
the
stat
eh
eld
byba
nks
belo
wth
em
edia
nsi
zeof
ban
kin
each
stat
ein
each
year
Rel
ativ
eca
pita
lrat
iois
the
capi
talt
oas
sets
rati
oof
smal
lban
ksm
inu
sth
atof
larg
eba
nks
Siz
eof
insu
ran
cere
lati
veto
bank
ing
plu
sin
sura
nce
inth
est
ate
ism
easu
red
asgr
oss
stat
epr
oduc
tfro
min
sura
nce
divi
ded
bygr
oss
stat
epr
odu
ctfr
omin
sura
nce
plu
sba
nkin
gIn
dica
tor
ison
eif
stat
ela
wpe
rmit
sba
nks
toen
ter
the
insu
ran
cebu
sin
ess
Sm
all
rmsh
are
isth
epe
rcen
tofa
lles
tabl
ishm
ents
inth
est
ate
that
hav
efe
wer
than
twen
tyem
ploy
ees
Par
tyin
dica
tor
ison
eif
the
Rep
rese
nta
tive
isa
Dem
ocra
tA
vera
geyi
eld
onba
nklo
ans
equa
lsto
tali
nter
est
inco
me
onal
ldom
esti
clo
ans
mad
eby
ban
ksin
the
stat
edi
vide
dby
tota
lloa
ns
The
aver
age
bank
failu
rera
teis
the
rati
oof
asse
tsin
faile
dba
nks
duri
ng
agi
ven
stat
edi
vide
dby
tota
lban
kas
sets
inth
atst
ate
aver
aged
over
the
year
s19
85to
1991
Un
itba
nki
ng
indi
cato
req
ual
son
efo
rst
ates
wit
hun
itba
nkin
gre
stri
ctio
nsT
hech
ange
inin
sura
nce
pow
ers
indi
cato
req
ual
son
eif
the
stat
ech
ange
dth
ela
wto
perm
itba
nks
tose
llin
sura
nce
duri
ngth
esa
mpl
epe
riod
We
divi
deth
eco
unt
ryin
tofo
urre
gion
san
din
clu
dere
gion
alin
dica
tors
inco
lum
n(5
)R
obu
stst
anda
rder
rors
are
inpa
ren
thes
es
an
d
den
ote
stat
isti
cally
sign
ica
nta
tth
e1
5an
d10
perc
entl
evel
s
WHAT DRIVES DEREGULATION 1459
branching of approximately 17 percent The relative capital-to-assets ratio of small and large banks however does not have astatistically signicant effect in the probit model
The impact of rival interests outside of banking is alsoconsistent with intrastate deregulation results Where banks cansell insurance legislators from states with larger insurancesectors relative to banking are less likely to vote for the amend-ment35 A one-standard-deviation increase in the relative size ofinsurance in those states which permit banks to sell insurancedecreases the probability that a legislator will favor the amend-ment by about 13 percent Turning to consumer interests thecoefficient on small rm share is not statistically signicant butthe coefficient on interest rates in the state is A one-standard-deviation increase in the average interest rate on loans raises theprobability that a legislator will support the amendment byroughly 8 percent The coefficient estimate for the bankingdistress variable measured as the average bank failure rate from1985 to 1991 is small and statistically signicant Overall theprobit analysis of the vote on national branching deregulationsupports the private-interest theory of deregulation and providesa consistency check that the importance of interests operating onthe state legislatures are very similar to those operating at thefederal level
V WHY DID BRANCHING DEREGULATION BEGIN DURING THE
1970S SHOCKS TO THE EQUILIBRIUM SUPPORTING BANK
BRANCHING RESTRICTIONS
A complete explanation of regulatory exit should account forwhy deregulation begins in the 1970s as well as the timing of thestate-by-state reforms [Kane 1996] In this section we describebroad technological legal and economic shocks that altered thepolitical-economy equilibrium which had kept antibranching regu-lations little changed for at least 30 years
Beginning in the 1970s three innovations reduced the valueto the protected banks of local geographic monopolies (see TableV) First automatic teller machines (ATMs) helped to erode thegeographic ties between customers and banks Second checkablemoney market mutual funds and the Merrill Lynch Cash Manage-ment Account demonstrated that banking by mail and telephone
35 The positive and statistically signicant estimates on the other insurancevariables may reect insurance industry support for the amendmentrsquos provisionslimiting banksrsquo insurance powers
QUARTERLY JOURNAL OF ECONOMICS1460
provided a convenient alternative to local banks36 Third techno-logical innovation and deregulation reduced transportation andcommunication costs particularly since the 1970s thereby lower-ing the costs for customers to use distant banks By increasing theelasticity of deposits supplied to banks these innovations reducedthe value of geographical restrictions to their traditional benecia-ries and thereby reduced their incentive to ght to maintain them[Peltzman 1976]
On the lending side increasing sophistication of credit-scoring techniques following innovations in information process-ing nancial theory and the development of large credit databases began to diminish the value of knowledge that localbankers had about the risks of borrowers in the community As a
36 Regulation Q which limited the interest rates that banks could pay ondeposits may have helped to drive depositors away from banks when the gapbetween market rates and deposit ceilings grew during the 1970s The eliminationof interest rate ceilings on large denomination certicates of deposit during the1970s appears to have hurt smaller and retail-oriented banks relative to largerwholesale banks [James 1983]
TABLE VBROAD TRENDS IN COMMERCIAL BANKING 1950ndash1995
YearNumberof ATMs
Domesticbank
deposits(billions)
Moneymarketmutual
fund(billions)
Percent ofdeposits 1
moneyfunds heldby banks
Smallbanksrsquo
percent ofbankingassets
Averagenumber of
bankfailures
(1) (2) (3) (4) (5) (6)1950 0 $ 154 $ 0 100 NA 41955 0 191 0 100 NA 31960 0 228 0 100 24 21965 0 330 0 100 20 41970 0 479 0 100 18 61975 9750 775 4 99 18 61980 18500 1182 76 94 17 101985 61117 1787 242 88 14 601990 80156 2339 493 83 11 1791995 122706 2552 745 77 8 61
Column (1) ATM gures are from Bank Network News The EFT Network Data Book (New YorkFaulkner and Gray Inc) The 1975 gure was unavailable 9750 is the number of ATMs in 1978 the rst yearfor which complete data are available Columns (2)ndash(4) Banksrsquo domestic deposits are from the Reports ofIncome and Condition money market mutual funds are from the Flow of Funds Data on all bank depositsforeign plus domestic are only available beginning in 1970 The trend in banksrsquo share (column (4)) is the sameusing total deposits instead of domestic deposits Column (5) Percent of banking assets held by small bankswhere a small bank is dened as a commercial bank with less than $100 million in assets in 1994 dollarsThese data are based on the Reports of Income and Condition Data on small banks are not available before1960 Column (6) Five-year average number of bank failures where the nal year is indicated in the rstcolumn These data are from FDIC Annual Report and the Quarterly Banking Prole
WHAT DRIVES DEREGULATION 1461
result of these innovations a national market developed forresidential mortgages credit card receivables have been securi-tized and bank lending to small business now relies less on thejudgment of loan officers and more on standardized scoringmodels
These changes have increased the potential protability forlarge banks to enter what had been the core of small bankactivities Large banksrsquo incentive to increase their lobbying pres-sure to be able to expand into these markets has thus beenincreasing over time In fact small banksrsquo market share began todecline even prior to the branching deregulation (Table V) As thevalue of a local banking relationships declined small rms thatwere the main borrowers from the small banks also probablybecame more likely to favor the entry of large banks into localmarkets With the deadweight costs of preventing large bankentry rising the private-interest theory predicts that small localbanks would become less likely able to maintain the branchingrestrictions [Becker 1983] Deregulation that reduces deadweightcosts of regulation is also consistent with the public-interesttheory
Kane [1996] argues that another major shock to the oldequilibrium was the increasing public awareness of the costlinessof having government-insured but (geographically) undiversiednancial institutions In the late 1970s the failure rate of banksbegan to rise (Table V) In the 1980s the Savings and Loan crisisand taxpayer bailout further heightened public awareness aboutthe costs of restrictions that make depository institutions morelikely to require infusions of taxpayer funds The failures thusmay have heightened public support for branching deregulation(see also Abrams and Settle [1993]) As noted above we did notnd any evidence that banking failures or distress in a stateaffected the speed with which the state deregulated but distressstill may be a common factor affecting all states
These technological economic and legal shocks generatedconditions that changed the long-standing balance favoring theantibranching forces The marginal value of lobbying to repealbranching restrictions increased just as the relative value to thesmall banks of maintaining branching restrictions was decliningThese nationwide shocks are common factors across states andthey are consistent with the positive duration dependence associ-ated with state-level deregulation that we demonstrated in FigureII In addition a broad change in lsquolsquoideologyrsquorsquo against government
QUARTERLY JOURNAL OF ECONOMICS1462
regulation could explain the positive duration dependence duringthe sample period
VI CONCLUSIONS
The private-interest theory of regulation can account for thepattern of bank branching deregulation of the 1970s and 1980sBeneciaries of branching regulation had supported a coalitionfavoring geographical restrictions despite their costs to consum-ers of nancial services Innovations that began in the 1970saltered the value of the restrictions to the affected parties and theresulting competition among interest groups can explain thesubsequent deregulation While some of our results are alsoconsistent with the public interest theorymdashfor example deregula-tion occurs earlier when small banks are in a relatively weaknancial positionmdashother results particularly evidence on theimportance of rivalries between small and large banks andbetween banking and insurance are difficult to explain with thepublic-interest approach We also nd that political-institutionalfactors affect deregulation although these variables may act asproxies for unmeasured economic interests The greater success ofthe private-interest theory to explain deregulation here than inprevious work may be due to the richer cross-sectional andtime-series variation in branching deregulation that can beexploited with a hazard model Future empirical studies ofendogenous deregulation thus may be most fruitful where changehas occurred across states over time such as in franchisinginsurance and public utilities (eg Joskow [1996])
Technological and nancial innovations will continue to erodethe benets to any interest group of maintaining regulatorybarriers in nancial services These forces are likely to bringabout reforms both domestically for example through legislationthat would increase bank powers (see Kroszner [1996] Krosznerand Rajan [1994 1997] and Kroszner and Stratmann [1998]) andinternationally for example through the extension of nancialservices provisions of NAFTAto reduce geographic barriers acrosscountries [Kroszner 1997]
APPENDIX EFFECTS OF DIFFERENT TYPES OF BRANCHING
DEREGULATION ON ECONOMIC GROWTH BANKING STRUCTUREAND BANK EFFICIENCY
We measure the importance of alternative forms of deregula-tion by their consequences for the economic growth and the
WHAT DRIVES DEREGULATION 1463
structure and efficiency of the banking sector across the states(see Jayaratne and Strahan [1996 1998]) To do so we run anOLS panel regression relating indicators for each type of deregula-tion to potential consequences Our dependent variables includeproxies for economic growth (growth in personal income) bankstructure (the log of the number of banking companies and thenumber of branches per banking company) and efficiency (the logof total noninterest expenses and the ratio of loan loss provisionsto total loans) The four deregulation indicators (MBHC branch-ing through merger and acquisition unrestricted statewide branch-ing and interstate banking) are equal to one in the years afterderegulation and zero before If a state deregulated before thesample period begins in 1970 the indicator is one throughout ifthe state does not deregulate then the indicator is zero through-out All regressions include state xed effects and time effects
Table VI shows the results of including the set of four
TABLE VIMULTIVARIATE REGRESSIONS OF THE EFFECTS OF DIFFERENT TYPES OF BRANCHING
DEREGULATION ON ECONOMIC GROWTH BANKING STRUCTUREAND BANK EFFICIENCY
Dependent variable
Growthin
personalincome
Log of of
bankingcompanies
Branchesper
bankingcompany
Log ofnon-
interestexpenses
Loan lossprovisions
(loans 1leases)
Merger amp acquisitionbranchingindicator
0013(0004)
2 0162(0037)
1146(0453)
2 0107(0035)
2 0007(0001)
Unrestricted state-wide branchingindicator
2 0008(0005)
0086(0059)
2 0043(0574)
0070(0037)
0005(0002)
Interstate bankingindicator
0006(0005)
2 0052(0057)
1064(0511)
0005(0030)
2 0001(0001)
Multibank holdingcompany indicator
2 0005(0006)
2 0096(0052)
2 0364(0529)
2 0095(0036)
2 0001(0001)
Dependent variablemean [median]
0086[0082]
4881[5136]
6649[5055]
13812[13803]
0009[0006]
This table contains regressions of state-level economic growth and state-level measures of the structureand efficiency of the banking industry Each dependent variable is regressed on a set of state indicatorvariables a set of year indicator variables and the set of four deregulation indicators these deregulationindicators are equal to one after deregulation and zero before If a state was always regulated the indicator iszero throughout if it was always deregulated the indicator is one throughout Sample period is 1975 to 1991and Delaware and South Dakota are dropped so N 5 833 Noninterest expenses become available only in1984 so N 5 392 The coefficient on the deregulation indicators is reported with robust standard errors inparentheses and denote statistically signicant at the 1 5 and 10 percent levels
QUARTERLY JOURNAL OF ECONOMICS1464
deregulation indicators simultaneously The branching by mergerand acquisition indicator is statistically signicant and of theexpected sign in all ve of the equations and larger in absolutevalue than any of the other coefficients on the indicators for theother types of deregulation Since there may be multicollinearityamong our indicators we also ran univariate specications whichincluded each indicator separately and the results are verysimilar These results imply that branching by merger andacquisition is the most important type of deregulation measuredby its consequences for economic growth banking structure andbank efficiency See Kroszner and Strahan [1998] for more details
UNIVERSITY OF CHICAGO GRADUATE SCHOOL OF BUSINESS AND NATIONAL BUREAU OF
ECONOMIC RESEARCH
FEDERAL RESERVE BANK OF NEW YORK RESEARCH amp MARKET ANALYSIS GROUP
REFERENCES
Abrams Burton A and Russell F Settle lsquolsquoPressure-Group Inuence and Institu-tional Change Branch-Banking Legislation during the Great DepressionrsquorsquoPublic Choice LXXVII (1993) 687ndash705
Amel Dean lsquolsquoState Laws Affecting the Geographic Expansion of CommercialBanksrsquorsquo unpublished manuscript Board of Governors of the Federal ReserveSystem 1993
Becker Gary S lsquolsquoA Theory of Competition among Pressure Groups for PoliticalInuencersquorsquo Quarterly Journal of Economics XCVIII (1983) 371ndash400
Berger Allen Anil Kashyap Joseph Scalise lsquolsquoThe Transformation of the U SBanking Industry What a Long Strange Trip Itrsquos Beenrsquorsquo Brookings Papers onEconomic Activity 2 (1995) 55ndash218
Berger Allen Anthony Saunders Joseph Scalise and Gregory Udell lsquolsquoThe Effectsof Bank Mergers and Acquisitions on Small Business Lendingrsquorsquo Journal ofFinancial Economics L (1998) 187ndash229
Brennan Geoffrey and James Buchanan lsquolsquoTowards a Tax Constitution forLeviathanrsquorsquo Journal of Public Economics VIII (1977) 255ndash273
Brickley James A and Christopher M James lsquolsquoThe Takeover Market CorporateBoard Composition and Ownership Structure The Case of Bankingrsquorsquo Journalof Law and Economics XXX (1987) 161ndash180
Cole Rebel and John Wolken lsquolsquoFinancial Services Used by Small BusinessesEvidence from the 1993 Survey of Small Business Financesrsquorsquo Federal ReserveBulletin LXXXI (1994) 629ndash676
Dixit Avinash The Making of Economic Policy A Transactions-Cost PoliticsPerspective (Cambridge MA MIT Press 1996)
Economides Nicolas R Glenn Hubbard and Darius Palia lsquolsquoThe Political Economyof Branching Restrictions and Deposit Insurancersquorsquo Journal of Law andEconomics XXIX (1996) 667ndash704
Flannery Mark lsquolsquoThe Social Costs of Unit Banking Restrictionsrsquorsquo Journal ofMonetary Economics XIII (1984) 237ndash249
Greene William H Econometric Analysis third edition (Upper Saddle River NJPrentice-Hall 1997)
Gunther Jeffery lsquolsquoRegional Capital Imbalances and the Removal of InterstateBanking Restrictionsrsquorsquo Economics Letters XLIV (1994) 439ndash442 lsquolsquoThe Erosion of Intrastate Branch Banking Restrictionsrsquorsquo unpublishedmanuscript Federal Reserve Bank of Dallas 1996
Irwin Douglas and Randall Kroszner lsquolsquoInterests Institutions and Ideology inSecuring Policy Change The Republican Conversion to Trade Liberalizationafter Smoot-Hawleyrsquorsquo Journal of Law and Economics XLII (1999) 643ndash673
WHAT DRIVES DEREGULATION 1465
James Christopher lsquolsquoAn Analysis of Intra-Industry Differences in the Effect ofRegulation The Case of Deposit Rate Ceilingsrsquorsquo Journal of Monetary Econom-ics XII (1983) 417ndash442
Jarrell Gregg lsquolsquoChange at the Exchange The Causes and Effects of DeregulationrsquorsquoJournal of Law and Economics XXVII (1984) 273ndash312
Jayaratne Jith and Philip E Strahan lsquolsquoThe Finance-Growth Nexus Evidencefrom Bank Branch Deregulationrsquorsquo Quarterly Journal of Economics CI (1996)639ndash670
Jayaratne Jith and Philip E Strahan lsquolsquoEntry Restrictions Industry Evolutionand Dynamic Efficiency Evidence from Commercial Bankingrsquorsquo Journal of Lawand Economics XLIX (1998) 239ndash274
Joskow Paul lsquolsquoComment on lsquoPower Struggles Explaining Deregulatory Reformsin Electricity Marketsrsquo rsquorsquo Brookings Papers on Economic Activity Microeconom-ics (1996) 251ndash264
Joskow Paul and Roger Noll lsquolsquoRegulation in Theory and Practice An Overviewrsquorsquoin Studies in Public Regulation Gary Fromm ed (Cambridge MA MITPress 1981) pp 1ndash65
Kahn Matthew E and John G Matsusaka lsquolsquoDemand for Environmental GoodsEvidence from Voting Patterns on California Initiativesrsquorsquo Journal of Law andEconomics XL (1997) 137ndash173
Kalbeisch J and R Prentice The Statistical Analysis of Failure Time Data (NewYork John Wiley amp Sons 1980)
Kane Edward J lsquolsquoDe Jure Interstate Banking Why Only NowrsquorsquoJournal of MoneyCredit and Banking XXVIII (1996) 141ndash161
Kiefer Nicholas lsquolsquoEconomic Duration Data and Hazard Functionsrsquorsquo Journal ofEconomic Literature XXVI (1988) 646ndash679
Kroszner Randall S lsquolsquoThe Evolution of Universal Banking and Its Regulation inTwentieth Century Americarsquorsquo in Universal Banking Financial System DesignReconsidered Anthony Saunders and Ingo Walter eds (New York IrwinProfessional Publishers 1996) pp 70ndash99 lsquolsquoThe Political Economy of Banking and Financial Regulation in the U Srsquorsquo inThe Banking and Financial Structure in the NAFTA Countries and ChileGeorge M von Furstenberg ed (Boston Kluwer Academic Publishers 1997)pp 200ndash213
Kroszner Randall S and Raghuram G Rajan lsquolsquoIs the Glass-Steagall ActJustied A Study of the U S Experience with Universal Banking before1933rsquorsquo American Economic Review LXXXIV (1994) 810ndash832
Kroszner Randall S and Raghuram G Rajan lsquolsquoOrganization Structure andCredibility Evidence from Bank Securities Activities before the Glass-Steagall Actrsquorsquo Journal of Monetary Economics XXXIX (1997) 475ndash516
Kroszner Randall S and Philip E Strahan lsquolsquoRegulatory Incentives and the ThriftCrisis Dividends Mutual-to-Stock Conversions and Financial DistressrsquorsquoJournal of Finance LI (1996) 1285ndash1320
Kroszner Randall S and Philip E Strahan lsquolsquoWhat Drives Deregulation Econom-ics and Politics of the Relaxation of Bank Branching Restrictionsrsquorsquo NBERWorking Paper No 6637 1998
Kroszner Randall S and Thomas Stratmann lsquolsquoInterest Group Competition andthe Organization of Congress Theory and Evidence from Financial ServicesPolitical Action Committeesrsquorsquo American Economic Review LXXXVIII (1998)pp 1163ndash1187
Kroszner Randall S and Thomas Stratmann lsquolsquoDoes Political Ambiguity PayCorporate Campaign Contributions and Rewards to Legislator Reputationrsquorsquounpublishedmanuscript University of Chicago Graduate School of Business1999
Lin D Y and L J Wei lsquolsquoThe Robust Inference for the Cox Proportional HazardsModel Journal of the American Statistical Association LXXXIV (1989)1074ndash1078
Macey Jonathan R and Geoffrey P Miller Banking Law and Regulation (BostonLittle Brown and Company 1992)
Makinson Larry Open Secrets The Encyclopedia of Congressional Money andPolitics (Washington DC Congressional Quarterly 1992)
Niskanen WilliamBureaucracy and Representative Government (Chicago Aldine-Atherton 1971)
QUARTERLY JOURNAL OF ECONOMICS1466
Noll Roger lsquolsquoComment on Peltzmanrsquorsquo Brookings Papers Microeconomics SpecialIssue (1989) 48ndash58
North Douglass lsquolsquoA Transactions Cost Theory of Politicsrsquorsquo Journal of TheoreticalPolitics II (1990) 355ndash367
Peltzman Sam lsquolsquoToward a More General Theory of Regulationrsquorsquo Journal of Lawand Economics XIX (1976) 109ndash148 lsquolsquoConstituent Interest and Congressional Votingrsquorsquo Journal of Law andEconomics XXVII (1984) 181ndash210 lsquolsquoThe Economic Theory of Regulation after a Decade of DeregulationrsquorsquoBrookings Papers Microeconomics Special Issue (1989) 1ndash41
Peterson Mitchell and Raghuram Rajan lsquolsquoThe Benets of Lending RelationshipsEvidence from Small Business Datarsquorsquo Journal of Finance XLIX (1994) 3ndash38
Poole Keith and Howard Rosenthal Congress A Political-Economic History ofRoll Call Voting (Oxford Oxford University Press 1997)
Stata Stata Reference Manual Release 5 (College Station TX Stata Press 1997)Stigler George J lsquolsquoThe Theory of Economic Regulationrsquorsquo Bell Journal of Economics
and Management Science II (1971) 3ndash21Stigler George J Chicago Studies in the Political Economy (Chicago University
of Chicago Press 1988)Strahan Philip E and James Weston lsquolsquoSmall Business Lending and the Changing
Structure of the Banking Industryrsquorsquo Journal of Banking and Finance XXII(1998) 821ndash845
Sylla Richard John Legler and John Wallis lsquolsquoBanks and State Public Finance inthe New Republic The United States 1790ndash1860rsquorsquo Journal of EconomicHistory XLVII (1987) 391ndash403
White Eugene The Regulation and Reform of the American Banking System1900ndash1929 (Princeton Princeton University Press 1983)
Winston Clifford lsquolsquoEconomic Deregulation Days of Reckoning for Microecono-mistsrsquorsquo Journal of Economic Literature XXXI (1993) 1263ndash1289
WHAT DRIVES DEREGULATION 1467
TABLE IYEAR OF DEREGULATION OF RESTRICTIONS ON GEOGRAPHICAL EXPANSION BY STATE
State
Intrastatebranching
through MampA
Full intrastatebranchingpermitted
Interstatebanking
permitted
Multibank holdingcompaniespermitted
AL 1981 1990 1987 1970AK 1970 1970 1982 1970AZ 1970 1970 1986 1970AR 1994 1989 1985CA 1970 1970 1987 1970CO 1991 1988 1970CT 1980 1988 1983 1970DE 1970 1970 1988 1970DC 1970 1970 1985 1970FL 1988 1988 1985 1970GA 1983 1985 1976HI 1986 1986 1970ID 1970 1970 1985 1970IL 1988 1993 1986 1982IN 1989 1991 1986 1985IA 1991 1984KS 1987 1990 1992 1985KY 1990 1984 1984LA 1988 1988 1987 1985ME 1975 1975 1978 1970MD 1970 1970 1985 1970MA 1984 1984 1983 1970MI 1987 1988 1986 1971MN 1993 1986 1970MS 1986 1989 1988 1990MO 1990 1990 1986 1970MT 1990 1993 1970NE 1985 1990 1983NV 1970 1970 1985 1970NH 1987 1987 1987 1970NJ 1977 1986 1970NM 1991 1991 1989 1970NY 1976 1976 1982 1976NC 1970 1970 1985 1970ND 1987 1991 1970OH 1979 1989 1985 1970OK 1988 1987 1983OR 1985 1985 1986 1970PA 1982 1990 1986 1982RI 1970 1970 1984 1970SC 1970 1970 1986 1970SD 1970 1970 1988 1970TN 1985 1990 1985 1970TX 1988 1988 1987 1970UT 1981 1981 1984 1970VT 1970 1970 1988 1970VA 1978 1987 1985 1970WA 1985 1985 1987 1981WV 1987 1987 1988 1982WI 1990 1990 1987 1970WY 1988 1987 1970
States not yet deregulated Source Amel [1993] and updates by authors
WHAT DRIVES DEREGULATION 1441
nies to acquire Maine banks As part of the Garn-St GermainAct federal legislators in 1982 amended the Bank HoldingCompany Act to allow failed banks and thrifts to be acquired byany bank holding company regardless of state laws (see egKroszner and Strahan [1996]) Many states then entered regionalor national reciprocal arrangements whereby their banks could bebought by any other state in the arrangement Between 1984 and1988 38 states joined one of these arrangements (see Amel[1993])
Table I and Figure I illustrate the history of state deregula-tion of geographical restrictions since 1970 The state-by-statederegulation culminates in the phaseout of interstate bank-ing restrictions with the passage of the 1994 Riegle-Neal Inter-state Banking and Branching Efficiency Act which effectivelycodied at the national level what had been occurring at the statelevel6
6 The Act permitted states to pass legislation to opt out of the interstatebanking provisions if the legislature did so before the provisions were to go intoeffect in mid-1997 Since only Texas and Montana have passed opt-out legislationthe United States will now have nearly complete interstate banking and branching
FIGURE IDeregulation of Restrictions on Intrastate Branching
Permitted Intrastate Branching before 1970
QUARTERLY JOURNAL OF ECONOMICS1442
III HYPOTHESES AND VARIABLE DEFINITIONS
Our empirical tests are based on the timing of intrastatebranching deregulation through merger and acquisition We focuson this form of branching deregulation because it has a muchgreater economic impact than the other forms of branchingderegulation (see the Appendix) Deregulation through mergerand acquisition is the only type of branching deregulation thatconsistently has a statistically signicant effect on bankingstructure bank efficiency and overall economic growth and theestimated magnitudes of the effects are greatest for this type ofderegulation We also analyze the interstate deregulation thattakes place at the national level by modeling voting behavior inthe House of Representatives on federal interstate branchinglegislation7
In the remainder of this section we describe our hypothesesabout the private-interest public-interest and political-institu-tional theories along with the data sources used to construct thevariables Some of the variables help us to distinguish between apublic- and private-interest approach while others will be consis-tent with both approaches The political-institutional approachhas little overlap in its implications with the other theories Ourmethod is to assess the relative importance of factors representingeach approach in speeding or slowing deregulation
A Hypotheses
Intraindustry Rivalry Small banks have fought to maintainand extend branching restrictions both historically and in therecent debates8 Smaller banks appear to have been the mainwinners from antibranching laws since these restrictions protectthem from competition from larger and more efficient bankingorganizations (see Flannery [1984] Jayaratne and Strahan [1998]
7 We do not employ a hazard model to explain the timing of interstatederegulation for two reasons the time clustering of interstate deregulation andcomplications of the strategic interdependence of the statesrsquo behavior As thenumber of states in an interstate arrangement increases for example the effectson potential acquirers is ambiguous they might prefer a larger pool of potentialtarget banks but might have to pay more as the number of competing acquirersexpands The benet to potential targets tends to increase with the number statesin an arrangement since the expected selling price increases with the number ofpotential bidders [Brickley and James 1987]
8 Economides Hubbard and Palia [1996] provide evidence that voting inCongress for the 1927 McFadden Act responded to small state banksrsquo interest inlimiting competition from large national banks See also White [1983] and Abramsand Settle [1993] for historical opposition On the small bank opposition to therecent branching deregulation see Kane [1996] and The Economist [8694 p 59]
WHAT DRIVES DEREGULATION 1443
and Winston [1993]) Branching restrictions thus tend to reducethe efficiency and consumer convenience of the banking system9
The private-interest theory therefore predicts that reformshould occur later in those states where the strength of the smallbanks relative to that of the large banks is greater According tothe public-interest theory deregulation should take place earlierwhere small banks are relatively important because the socialcosts of the regulation are directly related to the size of theprotected sector The social costs include deadweight costs andlosses associated with inefficiencies in the production of bankingservices relative to production without geographical restrictions
Interindustry Rivalry A number of states permit state-chartered commercial banks to sell insurance The insurancelobby would thus oppose the relaxation of branching restrictionswhen banks can sell insurance because such deregulation mightpermit banks to provide a more efficient insurance distributionnetwork10 According to the private-interest theory reform shouldoccur later in states where banks can sell insurance and theinsurance industry is important relative to the banking industryThe efficiency costs of the branching restrictions however risewith the size of the insurance sector since the restrictions preventthe exploitation of scope economies Under the public-interesttheory reform should therefore occur earlier where banks can sellinsurance and the insurance industry is relatively large11
Consumers of Banking Services Banks are a major source ofcredit for small rms [Cole and Wolken 1994] Branching deregu-lation tends to reduce banksrsquo local market power [Jayaratne andStrahan 1998] In addition Strahan and Weston [1998] nd thatlending to small businesses increases on average when smallbanks are purchased by other banking organizations and Bergeret al [1998] nd that credit availability to small businessesincreases in the years following a takeover of a small bank by alarger banking organization Since bank borrowers tend to benet
9 Flannery [1984] shows that small banks in states with branching restric-tions have higher costs than small banks in states without such restrictions andthe Appendix shows how the production of banking services becomes more efficientfollowing deregulation
10 The Economist [8694 p 59] describes the insurance industryrsquos opposi-tion to branching deregulation
11 Another rival interest we considered was the Savings and Loan industryThe share of assets in Savings and Loans relative to banking plus Savings andLoan assets in the state however had only small and statistically insignicanteffects in the models we estimate below
QUARTERLY JOURNAL OF ECONOMICS1444
from branching deregulation in particular and bank consolidationin general the private-interest theory would predict that stateswith numerous small bank-dependent rms would deregulateearlier12 This prediction however is also consistent with thepublic-interest theory since the social costs of the restrictions arehigher in states with more small bank-dependent rms
Another effect of branching restrictions on bank customerscan be related to the prices paid for bank services High loanprices before deregulation may reect for example a largemarket share for inefficient high-cost banks or high rents beingearned by the banks protected from competition In either casethe public-interest theory would suggest that states with rela-tively high initial loan interest rates should deregulate earlierThe private-interest theory however does not have a clearimplication high rents could lead the beneciaries to ght harderto maintain them (eg Stigler [1971]) or the deadweight costsassociated with these high rents could make it more difficult forthe beneciaries to maintain the restrictions (eg Becker [1983])13
Bank Stability Geographic diversication through branch-ing could mitigate instability problems that were importantduring much of our sample period thereby improving welfareInstability also may reduce the incentives of banks to lobby tomaintain protections because unstable banks are less likely tosurvive to reap the benets of the restrictions politicians also mayanticipate that such banks are less likely to be able to providefuture contributions and support [Gunther 1994 1996]14 Both theprivate-interest and public-interest theories thus suggest thatderegulation is most likely in states where banking instability isgreatest
12 On the other hand local banking monopolies created by branchingrestrictions could strengthen relationshipsbetween banks and small and medium-sized rms and increase the availability of credit to these rms [Petersen andRajan 1994] Also some have argued that small business lending declines whenlarge banks take over small banks (eg Berger Kashyap Scalise [1995])
13 High loan interest rates also could reect compensation for higher risksand fewer opportunities for banks to diversify If state-specic risks explaindifferences in average loans rates then neither theory would have any predictionsfor the sign of this effect Following branching deregulation however loan interestrates tend to fall but bank risk proles do not appear to change [Jayaratne andStrahan 1998 Kroszner and Strahan 1998] This suggests that high loan rates inantibranching states are not simply reecting state-specic risks
14 Also deregulation can arise as a response to banking instability under atheory of ignorant or misinformed voters in which a banking crisis acts as aneducational device to make the previously ignorant public aware of the costs of theantibranching policy (see Kane [1996])
WHAT DRIVES DEREGULATION 1445
Political-Institutional Factors Republicans are typically per-ceived as more likely to favor deregulation than Democrats so thepolitical-institutional theories suggest that states that are con-trolled by Democrats deregulate later than those controlled byRepublicans15 In addition we investigate whether reform be-comes more likely when one party controls the legislature and thegovernorship Note that these political effects must be interpretedwith caution since the views of the politicians may simply reectthe economic interests of the constituents in the state (seePeltzman [1984])
B Variable Denitions and Data Sources
Our main proxy for the strength of the small banks is thefraction of banking assets in the state in lsquolsquosmallrsquorsquo banks We denesmall banks as banks with assets below the median size in eachstate By allowing the denition of small to vary across states wetake into account cross-state heterogeneity in bank sizes16 An-nual data on bank size are from the fourth quarter Reports ofIncome and Conditions (lsquolsquoCall Reportsrsquorsquo) from the Federal ReserveBoard We also include the capital-to-asset ratio of small banksrelative to large as a measure of their relative nancial healthSpecically this variable is the asset-weighted average capital-asset ratio for small banks minus the asset-weighted averagecapital-asset ratio for large banks Annual data on bank capitalare from the Call Reports
To measure the effects of the rival insurance industry we rstconstruct an indicator variable that is one if the state permitsbanks to sell insurance For each state we then measure the sizeof the insurance sector (total value added in the state) relative tothe sum of the banking plus insurance sectors We will examine
15 For more detail on the importance of legislative structures party politicsand ideology see Poole and Rosenthal [1997] Kahn and Matsusaka [1997] Irwinand Kroszner [1999] and Kroszner and Stratmann [1999]
16 We also considered a variety of other denitions of relative size but therelative size results reported below are not sensitive to which denition we useFirst we used a xed measure of small bank that did not vary across statesdening small banks as those with assets below $100 million in 1994 dollarsSecond we applied the state-varying and xed denitions at the level of lsquolsquobankingorganizationsrsquorsquo (which include multibank holding companies) instead of at the levelof banks Third we calculated gini coefficients of bank size inequality and bankconcentration indices as alternative proxies for small versus large bank power inthe state Finally we also considered banks in lsquolsquoruralrsquorsquo areas (that is not located in aBureau of the Census lsquolsquostandard metropolitan statistical arearsquorsquo) as small since thevalue of the restrictions might be greatest in protecting banks outside of cities fromentry by the city banks All of the alternative denitions are highly correlated andyield the same results on the importance of intraindustry differences
QUARTERLY JOURNAL OF ECONOMICS1446
the effect separately for states that permit banks to sell insuranceand those that do not Data on value-added by industry are fromU S Commerce Department Bureau of Economic Analysis Sur-vey of Current Business [August 1994]
To measure the relative importance of small bank-dependentborrowers we include the proportion of all establishments operat-ing in the state with fewer than twenty employees These data arecompiled by the Bureau of the Census17 As a measure of priceswe use the difference between the average interest rate on loans inthe state and the prevailing federal funds rate where the loanrate equals the ratio of total interest income on all domestic loansdivided by total domestic loans held by banks operating in thestate18 These data are from the end-of-year Call Reports butbecome available beginning only in 1976 Our proxy for bankinstability is the failure rate of banks measured as the percent oftotal state banking assets in failed institutions from the FederalReserve Boardrsquos National Information Center database19
We include two political variables to adjust for any indepen-dent inuence of party politics First we measure the degree ofparty control of the state government by the fraction of the threebodies of the state government (the assembly senate and gover-norship) controlled by Democrats This variable for example isone-third if the Democrats have a majority in the assembly andthe Republicans have a majority in the senate and hold thegovernorship Second we include an indicator variable equal toone if the same party controls the governorrsquos office and hasmajorities in both chambers of the state legislature20
17 We collected the establishment data by the state for three cross sections(1976 1982 and 1987) and interpolated the data in the intermediate years SeeState and Metropolitan Data Book 1982 1986 and 1991
18 A measure of how individual customers would be affected is interest ratesthey receive on deposits We calculated the average deposit interest rate in thestate from the end-of-year Call Reports and did not nd any relationship betweenthis variable and the timing of deregulation in our analysis below Since Jayaratneand Strahan [1998] nd that branching deregulation has no effect on deposit ratesour nding of no relationship between timing and rates result could be due toeither individual bank customers anticipating little pricing effect and not lobbyingor an inability of individual depositors to form an effective lobby
19 As alternative measures of distress we used (1) the bank failure ratemeasured as the number of failed institutions in the state relative to the totalnumber of banks (2) the overall capital-to-asset ratio of all banks in the state and(3) the growth in state personal income Our results are not sensitive to the choiceof distress proxy
20 We also estimated the models with three indicator variables to reect theparty control of the assembly senate and governorship separately The coefficientestimates on these variables are qualitatively similar to those we present below(that is Democratic control tends to slow deregulation) and the results on theinterest group variables are unaffected by the specication of the political
WHAT DRIVES DEREGULATION 1447
IV METHODS AND RESULTS
In this section we rst estimate how the variables describedabove inuence the timing of state-level branching deregulationusing a hazard model Next we discuss the robustness of theresults including a test of whether the ex post impact of deregula-tion on the different interest groups is consistent with ourinterpretation of the coefficients We then investigate federalinterstate branching deregulation with a probit model to explorewhether the factors that affect the timing of intrastate deregula-tion also inuence voting in the House of Representatives oninterstate deregulation
A Hazard Model
We consider the period from the beginning of our sample(1970) until deregulation as the lsquolsquoduration of regulationrsquorsquo or thelsquolsquotime until deregulationrsquorsquo The hazard rate h(t) is the likelihoodthat a state deregulates at time t given that the state has not yetderegulated [Kalbeisch and Prentice 1980 Kiefer 1988 Greene1997] To model the duration of regulation we must decide whatstructure if any to impose on the hazard function The Kaplan-Meier product-limit estimator provides a simple nonparametricway to estimate the shape of the hazard function over time[Greene 1997]21 Figure II graphs this estimate for our data andshows that the hazard function is relatively at in the early yearsand then grows steeper in the later years
A duration model that can approximate this shape is theWeibull proportional hazards model The hazard rate functiontakes the form
h[t x(t) b] 5 h0(t) exp[x(t)8b]
where x(t) is a vector of time-varying covariates b is a vector ofunknown parameters to be estimated and the baseline hazard
variables Note also that Nebraska has a nonpartisan unicameral legislature Weassume that the party of the Governor controls the state government for Nebraskaand set the unied government indicator to one As another alternative we triedincluding a variable that is one if party control of the legislature or governorshipchanges but it did not have an economically or statistically signicant effect anddid not affect any of the other results
21 Denote each of the K years in our sample period Tk and order them suchthat T1 T2 and so on Let nk be the number of states that have not yet deregulatedby Tk and dk be the number of states that deregulate in year Tk The Kaplan-Meierestimate of the hazard rate in each year is dknk In other words the hazard rate isthe ratio of the number of states that actually deregulate in each year to thenumber of states that have not yet deregulated in that year
QUARTERLY JOURNAL OF ECONOMICS1448
rate h0(t) is ptp 2 1 with shape parameter p that will be estimatedfrom the data When p 1 this model displays a monotonicallyincreasing hazard rate To estimate the parameters we maximizethe following log-likelihood function [Kiefer 1988]
L(b) 5 oi 5 1
N
di ln h[ti x(ti) b] 2 oi 5 1
N
e 0
tih[u x(u) b] du
where N indexes the number of states at risk at the beginning ofthe sample period di is an indicator equal to one for the statesthat deregulated by the end of the sample period and zero for thestates that do not deregulate (ie the censored observations) andti is the time of deregulation for the ith state This formulationassumes that the time until deregulation is independent acrossstates but the observations of the same state over time are notassumed to be independent In addition we use a robust estima-tion procedure to calculate the standard errors [Lin and Wei1989]
We exclude states that deregulated before 1970 from theanalysis so N equals 39 states During the sample period 36 ofthe 39 states deregulate We have three censored observations forwhich di is zero (Arkansas Iowa and Minnesota) but our results
FIGURE IIKaplan-Meier Estimate of Hazard Rate for Branching Deregulation
WHAT DRIVES DEREGULATION 1449
are not sensitive to the inclusion of these censored observationsSince we observe each state in each year up to and including theyear of deregulation we have a total of 637 observations Table IIreports summary statistics and correlation coefficients for theexplanatory variables we use in our analysis
We use the Weibull model rather than a model that does notimpose any structure on the baseline hazard rate so that we cancalculate the change in the expected time to deregulation for agiven change in the levels of the covariates22 In the Weibullmodel we can invert the hazard function and map it into the timedomain Rewriting the Weibull model in this way the log of thetime to deregulation T is a linear function of the political economyfactors and an error term ln (T ) 5 x8b 1 e23 Because we assumethat the baseline hazard rate is Weibull with a shape characterizedby p the new coefficients will be scaled by p that is b 5 2 bp
The b coefficients represent the percentage change in thetime to deregulation for a one-unit change in the covariates Apositive coefficient implies that an increase in the variableincreases the expected time until deregulation To gauge theeconomic importance of the effects we will multiply the b by thestandard deviation of the explanatory variables and then evaluatehow much the changes in the variables raise or lower the expectedtime to deregulation24 In Table III we report the b coefficients
22 The Cox proportional hazards model does not impose any structure on thebaseline hazard rate h0(t) and takes the following form
h[t x(t) b] 5 h0(t) exp [x(t)8b]
Common factors that affect the probability of deregulation in the same way for allstates therefore do not inuence the estimated coefficients The cost of the Coxmodel is that because there is no structure on the baseline we can calculatechanges in only relative hazard rates associated with changes in the covariatesThe Weibull model however provides sufficient structure so that we can translateour estimates into a log expected time metric that allows us to calculate the changein the expected time to deregulation for a given change in the covariates SeeKiefer [1988] Greene [1997] and Stata [1997] When we compare the coefficientsestimated by the Cox and Weibull models on our data they are very close (seeKroszner and Strahan [1998]) so imposing a Weibull structure does not appear todistort our estimates
23 The error term e is independent of x and has an extreme value distributionscaled by 1p (see Kiefer [1988] Greene [1997] and Stata [1997] for more details onthis log expected time parameterization of the Weibull model)
24 The transformation to the log expected time domain is only strictly correctwhen the covariates are constant over time Our interpretation of the coefficients ispossible in the time-varying model if we assume that a one-standard-deviationdifference would hold over the entire sample period that is the deregulationwould occur so many years earlier or later if the variable in question were higherby one standard deviation from 1970 to 1992 Note that the mean number of yearsto deregulation in the sample is 158
QUARTERLY JOURNAL OF ECONOMICS1450
TABLE IISUMMARY STATISTICS FOR VARIABLES IN HAZARD MODEL 1970ndash1992
Panel A Univariate statistics MeanStandarddeviation Minimum Maximum
(1) (2) (3) (4)Small bank asset share of all
banking assets in state 011 005 001 023Relative size of insurance to
banking plus insurance inthe state 047 009 024 080
Indicator is 1 if banks can sellinsurance in the state 020 040 0 1
Small rm share of the numberof rms in the state 088 002 078 095
Capital ratio of small banksrelative to large in the state 002 001 2 001 009
Share of state government con-trolled by Democrats 065 036 0 1
Indicator is 1 if state controlledby one party 054 050 0 1
Average yield on bank loans inthe state minus Fed fundsrate 002 002 2 004 008
Panel B CorrelationsSmallbank
Relativesize of
insuranceInsuranceindicator
Smallrm
Smallbank
capitalDem
controlSingleparty
(1) (2) (3) (4) (5) (6) (7)Small bank asset share
of all banking assetsin state 1
Relative size of insur-ance to banking plusinsurance in the state 2 034 1
Indicator is 1 if bankscan sell insurance inthe state 020 2 012 1
Small rm share of thenumber of rms inthe state 030 2 011 005 1
Capital ratio of smallbanks relative tolarge in the state 2 066 020 2 010 2 006 1
Share of state govern-ment controlled byDemocrats 2 006 001 2 018 2 018 005 1
Indicator is 1 if statecontrolled by oneparty 007 012 001 2 006 2 006 030 1
Average yield on bankloans in the stateminus Fed funds rate 2 001 2 015 003 022 2 010 2 003 2 006
The number of observations is 637 except for the average interest rate on bank loans which becomesavailable only in 1976 so the number of observations is 408
TABLE IIIHAZARD MODEL OF POLITICAL ECONOMY FACTORS AFFECTING THE TIMING OF STATE
BRANCHING DEREGULATION 1970ndash1992
(1) (2) (3) (4) (5) (6)Small bank asset share of
all banking assets instate
648(122)
519(137)
632(135)
516(118)
488(133)
445(088)
Capital ratio of smallbanks relative to largein the state
1325(348)
921(319)
1301(362)
1067(346)
1144(375)
906(381)
Relative size of insurancein states where banksmay sell insurance 0otherwise
324(161)
259(149)
315(165)
285(154)
272(159)
055(134)
Indicator is 1 if banksmay sell insurance inthe state
045(010)
037(011)
045(010)
038(008)
036(009)
001(012)
Relative size of insurancein states where banksmay not sell insurance0 otherwise
2 093(043)
2 090(033)
2 102(051)
2 093(048)
2 096(053)
2 004(039)
Small rm share of thenumber of rms in thestate
2 972(211)
2 614(215)
2 952(211)
2 943(230)
2 952(270)
2 1509(276)
Share of state govern-ment controlled byDemocrats
031(012)
023(011)
030(012)
027(011)
026(011)
011(013)
Indicator is 1 if state con-trolled by one party
2 004(010)
2 004(007)
2 004(007)
2 001(009)
2 002(010)
017(009)
Average yield on bankloans in the stateminus Fed funds rate mdash
023(319) mdash mdash mdash mdash
Bank failure rate mdash mdash092
(111) mdash mdash mdash
Indicator is 1 if state hasunit banking law mdash mdash mdash
018(010)
019(010)
030(009)
Indicator is 1 if statechanges bank insur-ance powers mdash mdash mdash mdash
2 008(012)
2 026(013)
Includes regionalindicators No No No No No Yes
N 637 408 637 637 637 637Log likelihood 2 374 900 2 360 2 212 2 191 1192p-value of x 2 for
regression 001 001 001 001 001 001
The hazard model is Weibull where the dependent variable is the log expected time to branchingderegulation ln (T) 5 bx 1 eAll variables are measured for each state in each year Small bank asset share isthe percent of banking assets in the state held by banks below the median size of banks in each state in eachyear Relative capital ratio is the capital to assets ratio of small banks minus that of large banks Size ofinsurance relative to banking plus insurance in the state is measured as gross state product from insurancedivided by gross state product from insurance plus banking Insurance indicator is 1 if state law permits banksto enter the insurance business Small rm share is the percent of all establishments in the state that havefewer than twenty employees Party control variable is the share of the three bodies of state governmentcontrolled by Democrats One party control indicator is 1 if the same party controlrsquos the governorship and hasmajorities in both chambers of the state legislature Average yield on bank loans equals total interest incomeon all domestic loans made by banks in the state divided by total loans Bank failure rate is the ratio of assetsin failed banks during a given state and year divided by total bank assets in that state and year Unit bankingindicator equals 1 for states with unit banking restrictions The change in insurance powers indicator equalsone if the state changed the law to permit banks to sell insurance during the sample period We divide thecountry into four regions and include regional indicators in column (5) Robust standard errors are inparentheses and denote statistically signicant at the 1 5 and 10 percent levels
QUARTERLY JOURNAL OF ECONOMICS1452
The positive and statistically signicant coefficient on smallbank share implies that a greater small bank share delaysderegulation and this effect is economically important A one-standard-deviation increase in the small bank share results in a30 percent increase in the time until deregulation or about 47years This result supports the implications of private-interesttheory concerning intraindustry rivalry and is contrary to thepublic-interest theory
The positive and statistically signicant coefficient on therelative performance of small banks in the state implies thatstates deregulate later when small banks are relatively strongnancially25 A one-standard-deviation increase in the relativecapital-to-asset ratio results in a 15 percent rise in the time untilderegulation or about 24 years This result is consistent with theprivate-interest theory but may also be consistent with a public-interest interpretation because the risk of failures (and a costlytaxpayer-funded bailout) increases as small bank nancial healthdeclines
Interindustry competition also helps explain the timing ofderegulation In states where banks can sell insurance a rela-tively large insurance sector is associated with an increase in theexpected time to deregulation26 A one-standard-deviation in-crease in the relative size of the insurance sector in those statesthat permit banks to sell insurance leads to a 22 percent increasein the time until deregulation or about 35 years27 This result isconsistent with the private-interest theory but not the public-interest theory
Deregulation occurs earlier in states where small bank-dependent rms are relatively numerous A one-standard-
25 We also allowed small bank and large bank capital-to-asset ratios to enterseparately Their coefficient estimates were nearly equal in absolute value and ofopposite signs and so the data do not reject the relative small to large bank capitalratio specication we have chosen
26 When we use the relative size of the insurance sector in the state as awhole rather than insurance relative to banking plus insurance in the state weobtain similar results When we simultaneously include both measures of therelative size of insurance the interactions of the lsquolsquoinsurance relative to bankingplus insurancersquorsquo variable continue to be statisticallysignicant but the interactionsof lsquolsquoinsurance relative to the state as a wholersquorsquo variable are not
27 We have also estimated the model with the indicator variable equal to onefor states where banks may sell insurance but without the variables measuringtheimportance of insurance relative to banking In this model the coefficient on theinsurance indicator variablemdashan estimate of the difference in the expected time toderegulation in states where banks may sell insurance relative to states wherebanks may not sell insurancemdashequals 027 and is statistically signicant at the 5percent level This suggests that branching deregulation takes about 45 yearslonger in states where banks may sell insurance
WHAT DRIVES DEREGULATION 1453
deviation increase in the share of small rms reduces the timeuntil deregulation by 18 percent or about three years This resultconcerning the interests of consumers of banking services isconsistent with both the private- and public-interest theories
Column (2) of Table III introduces the average interest rateon loans to the model as a rough proxy for the cost to bankborrowers of these regulations These data become available onlyin 1976 so our sample size shrinks to 408 observations Thecoefficient on this variable is small and not statistically signi-cant The public-interest theory would imply that the coefficientshould be negative while the implication of the economic theorydoes not have a clear sign prediction
When we add the failure rate of banks to the specication incolumn (3) its coefficient is small and statistically insignicantWe do not nd a linkage between the timing of state deregulationand statewide banking distress Given that deposit insurance isfunded federally however bank failures may raise the publicrsquosconcerns about the costs of the regulations at the national ratherthan the state level This interpretation can explain why the rateof bank failure does not affect the timing of state-level deregula-tion even though bank instability does appear to rise generallyduring the period of increasing deregulation (see Section V below)
Finally the partisan structure of the state government alsoinuences when states deregulate As expected a higher propor-tion of Democrats in the government tends to delay deregulationA one-standard-deviation rise in the share of the governmentcontrolled by Democrats slows the deregulation by about twoyears Whether the state is dominated by one party however doesnot appear to affect the timing of the deregulation
To summarize our ndings from the hazard model privateinterests appear to play an important role in the deregulatoryprocess While private interests and public interests do sometimescoincide the results on the relative share of small banks and largebanks and on the relative size of insurance where banks competeare consistent with a private-interest approach but are difficult toexplain on public-interest grounds
B Robustness Tests
Omitted Variables States begin the sample period withdifferent degrees of restrictions on bank expansion These initialconditions may proxy for omitted factors determining the political-economy equilibrium in the state that could then affect the
QUARTERLY JOURNAL OF ECONOMICS1454
subsequent timing of deregulation For instance some statesbegan the period with unit banking while other states allowedlimited branching and the formation of multibank holdingcompanies
To test whether the stringency of initial restrictions onbranching matters we rst divide the sample into states with andwithout the most extreme form of branching regulation unitbanking28 Column (4) of Table III includes an indicator variableequal to one if the state began the period with unit bankingrestrictions The coefficient on this variable is positive andstatistically signicant but the other results remain virtuallyunchanged Next we include an indicator variable that is one ifthe state permitted multibank holding companies (MBHCs) toexist at the beginning of our sample period Thirty-four statesbegan the period allowing MBHCs The coefficient estimate forthis indicator variable (not reported) is small and not statisticallysignicant and its inclusion does not change the effects of theother variables
Eleven states expand bank powers to include the sale ofinsurance products during our sample period This could proxy foromitted factors about the relative strength of the insuranceindustry in the state In column (5) we include an indicatorvariable that is one if the state changes the law during our sampleperiod to permit banks to sell insurance business and zerootherwise The coefficient estimate is small and statisticallyinsignicant and does not affect the other results29
In column (6) we introduce four regional indicator variablesNorth South Midwest and West30 The coefficients on small bankshare the relative capital-to-assets ratio and small rm sharechange little Although the coefficient estimate still suggests that
28 We classify states that prohibited branching but permitted banks toestablish facilities as unit banking states The unit banking indicator is one for thefollowing states CO AR FL IL IA KS MN MO MT NE ND OK TX WI WVand WY
29 We conducted a likelihood ratio test of whether the unit banking statesthe MBHC states and the states that changed their regulations of bank insurancepowers could be pooled with the other states In all three cases we could not rejectpooling of the data
30 Since intrastate deregulation generally preceded interstate deregulationand the latter typically took the form of regional interstate compacts thepossibility of participating in a regional interstate banking compact could haveinuenced the decision to deregulate intrastate branching Our denitions of theregions are region 1 (South) contains AL AR DC FL GA KY LA MS NC OKSC TN TX and VA region 2 (Northeast) contains CT MA MD ME NH NJ NYPA RI VT and WV region 3 (Midwest) contains IA IL IN KS MI MN MO NEND OH SD and WI region 4 (West) contains the other states
WHAT DRIVES DEREGULATION 1455
in states where banks can sell insurance a larger relative size ofthe insurance sector delays deregulation the effect is no longerstatistically signicant The coefficient on the indicator variablefor whether a state expands bank powers to permit insurancesales during the sample period however now becomes statisti-cally signicantly negative This implies that deregulation occursroughly 41 years earlier in states where the insurance lobby maybe relatively weak While the inclusion of the regional indicatorsdoes not affect the other results the results on insurance doappear to be sensitive to the specication of the hazard model31
Estimation Technique To check the robustness of using ahazard model and to investigate whether the time-series or thecross-sectional variation in each of the factors accounts for thehazard results we estimate two simple linear probability models(neither are reported) First we run a pooled time-series cross-section OLS regression using the 637 observations from thehazard model The dependent variable equals one in the year ofderegulation and zero otherwise We include the same set ofregressors that were used in the hazard model and add state-levelxed effects This specication thus removes the cross-statevariation from the independent variables and focuses on theirmovements over time In this model the signs of the coefficientson the interest group factors are consistent with the results fromthe hazard model and the coefficients for the small bank relativecapital ratio and for the relative size of insurance where states cansell insurance are statistically signicant
In the second model we eliminate the time-series variationby averaging each of the explanatory variables over time We thushave one observation per state that had branching restrictions asof 1970 (N 5 39) The dependent variable equals the inverse of thenumber of years until deregulation which represents an estimateof the statersquos probability of deregulating in a year under theassumption that this probability is constant over time and eachyear provides an independent observation32 Although we haveonly 39 observations the coefficients on the small bank marketshare small bank relative capital ratio and small rm share
31 As further robustness check we also tried a population density variablewhich Abrams and Settle [1993] found relevant for regulatory change during the1920s and 1930s but its coefficient was not statistically signicant and did notaffect the other results
32 For instance the dependent variable would take the value of 01 for astate that deregulated in 1980 For the three states that did not deregulate thedependent variable equals 0
QUARTERLY JOURNAL OF ECONOMICS1456
variables are statistically signicant The signs of the effects ofthe insurance variables are the same as those reported in thehazard model but not statistically signicant
The hazard model and linear probability models thus producevery similar results The estimates from the latter suggest thatthe effects for the small bank market share and small rm sharevariables are driven primarily by cross-state variation the effectof the insurance variable is driven primarily by time-seriesvariation and the small bank capital effect is driven by bothcross-sectional and time-series variation
Ex Post Consequences of Deregulation Finally to check theplausibility of our interpretation of the hazard results we con-sider whether the ex post consequences of deregulation areconsistent with the ex ante lobbying positions we attribute to eachinterest group (see Kroszner and Strahan [1998] for details) Wend that small banks lose market share following deregulationand in states where banks can enter the insurance business theinsurance sector shrinks relative to the banking sector followingderegulation Borrowers also benet from deregulation throughlower average interest rates on loans These ndings support theprivate-interest interpretation of the coefficients in the hazardmodel groups that will benet lobby to speed deregulation andthose that will be harmed lobby to slow it
C Voting on Interstate Branching Deregulation in the U SHouse of Representatives
We now examine whether the forces driving intrastate branch-ing deregulation also drive interstate deregulation Financialservices interests are active contributors and lobbyists in Washing-ton Their political action committees constitute the largest groupof contributors to legislators providing nearly 20 percent of totalcongressional campaign contributions [Makinson 1992] and muchof their lobbying effort involves competition among rival interestswithin nancial services (see Kroszner and Stratmann [1998])
As noted above after virtually all states adopted intra- andinterstate branching deregulation the 1994 Riegle-Neal Act re-pealed the 1927 McFadden Act to eliminate all barriers tointerstate banking and branching by 1997 The key votes concern-ing the Riegle-Neal Act were either voice votes or extremelylopsided so we could not estimate a voting model from them Anumber of bills and amendments related to interstate branching
WHAT DRIVES DEREGULATION 1457
had been debated in Congress during the years prior to thepassage of the Riegle-Neal Act but a search of the weekly BNABanking Reporter and the Congressional Record produced onlyone roll-call vote related to interstate branching that was notlopsided This vote occurred in the House on November 14 1991and concerned an amendment sponsored by Wylie (R-OH) andNeal (D-NC) to introduce interstate banking and branchingderegulation to a nancial services reform package33 Althoughthe amendment passed by 210 to 208 the bill to which it wasattached subsequently was defeated
To check for the inuence of the factors we considered in thestate-level reforms we examine both the sponsorship of interstatebanking legislation and voting on the amendment The sponsorsof the Wylie-Neal amendment are from states with low small bankshares 004 in Ohio (Wylie) and 002 in North Carolina (Neal) Incontrast the sample mean in 1991 is 008 (median 5 007)Michigan home state of the Senatersquos sponsor of the 1994 Riegle-Neal Act also had relatively low small bank strength (small bankshare of 005)
Table IV reports estimates of a probit model where thedependent variable equals one if the legislator voted in favor ofthe amendment and zero otherwise The explanatory variablesare the same as those in the hazard model except that we use aparty affiliation indicator variable equal to one for Democrats andzero for Republicans as our proxy for political factors Thecoefficients in Table IV are the marginal effects (lsquolsquoslopesrsquorsquo) of aone-unit change of each variable on the probability that alegislator will vote for the amendment34 Note that a forcefavoring deregulation has a negative coefficient in the hazardmodel (Table III) but a positive coefficient in the probit model
Consistent with the state-level deregulation process legisla-tors are more likely to support the amendment if their states havea relatively low share of small banks As in the hazard model thefraction of small banks is the most important interest groupinuence on a legislatorrsquos voting decision A one-standard-deviation increase in small banksrsquo market share (from the mean)is associated with a decline in the probability of voting in favor of
33 The Wylie-Neal amendment also included provisions limiting certaininsurance and real estate powers of national banks [Congressional RecordNovember 14 1991 pp 10239ndash10242]
34 Since we have multiple legislators from each state we adjust the standarderrors to correct for the lack of independence among observations clustered in thesame state
QUARTERLY JOURNAL OF ECONOMICS1458
TAB
LE
IVM
AR
GIN
AL
EF
FE
CT
SF
RO
MA
PR
OB
ITM
OD
EL
OF
TH
EIN
FL
UE
NC
EO
FP
OL
ITIC
AL
EC
ON
OM
YF
AC
TO
RS
ON
TH
EH
OU
SE
VO
TE
ON
TH
EW
YL
IE-N
EA
L
AM
EN
DM
EN
TT
OP
ER
MIT
INT
ER
ST
AT
EB
AN
KB
RA
NC
HIN
G
NO
VE
MB
ER
14
1991
(1)
(2)
(3)
(4)
(5)
(6)
Smal
lban
kas
set
shar
eof
allb
anki
ng
asse
tsin
stat
e2
417
(12
4)2
454
(12
2)2
407
(12
0)2
443
(14
7)2
450
(14
9)2
490
(14
7)C
apit
alra
tio
ofsm
allb
anks
rela
tive
tola
rge
inth
est
ate
180
(37
5)2
68(3
42)
256
(37
5)1
69(3
78)
153
(38
0)2
45(4
25)
Rel
ativ
esi
zeof
insu
ranc
ein
stat
esw
here
bank
sm
ayse
llin
sura
nce
0ot
herw
ise
22
34
(10
2)2
249
(1
14)
22
33
(10
1)2
249
(1
04)
22
51
(10
2)2
154
(11
1)In
dica
tor
is1
ifba
nks
may
sell
insu
ranc
ein
the
stat
e0
79
(0
06)
074
(00
6)0
69
(0
06)
070
(00
6)0
70
(0
06)
067
(00
7)R
elat
ive
size
ofin
sura
nce
inst
ates
whe
reba
nks
may
not
sell
insu
ranc
e0
oth
erw
ise
180
(06
6)2
42
(0
72)
154
(0
67)
173
(06
6)1
72
(0
67)
203
(0
91)
Smal
lrm
shar
eof
the
num
ber
ofr
ms
inth
est
ate
21
23(0
92)
20
83(0
89)
21
62(1
07)
21
32(0
95)
21
30(1
00)
20
94(1
16)
Indi
cato
ris
1if
Rep
rese
ntat
ive
isa
Dem
ocra
t2
035
(00
7)2
035
(00
7)2
036
(00
7)2
035
(00
7)2
035
(00
7)2
036
(00
7)A
vera
geyi
eld
onba
nklo
ans
inth
est
ate
min
usF
edfu
nds
rate
mdash10
01
(51
5)mdash
mdashmdash
mdash
Ave
rage
fail
ure
rate
198
5to
1991
mdashmdash
228
(14
1)mdash
mdashmdash
Indi
cato
ris
1if
stat
eh
asun
itba
nkin
gla
wmdash
mdashmdash
004
(00
7)0
04(0
07)
20
01(0
08)
Indi
cato
ris
1if
stat
ech
ange
sba
nk
insu
ran
cepo
wer
smdash
mdashmdash
mdash2
001
(00
7)2
003
(00
7)In
clud
esre
gion
alin
dica
tors
N
oN
oN
oN
oN
oYe
sP
seud
o-R
20
170
180
170
170
170
18L
ogli
keli
hood
224
024
223
826
223
951
224
013
224
010
223
808
p-va
lue
ofx2
for
regr
essi
on
001
0
01
001
0
01
001
0
01
Th
ede
pen
den
tvar
iabl
eis
one
ifth
ele
gisl
ator
vote
sfo
rth
eam
endm
enta
nd
zero
ifag
ains
tT
he
repo
rted
coef
fici
ents
are
the
effe
cts
ofa
uni
tch
ange
ofth
ein
depe
nde
ntva
riab
le(f
rom
the
mea
n)on
the
prob
abili
tyof
voti
ng
infa
vor
ofth
eam
endm
ent
All
vari
able
sar
em
easu
red
byst
ate
N5
418
the
num
ber
ofR
epre
sent
ativ
esvo
tin
gon
the
amen
dmen
tS
mal
lban
kas
sets
har
eis
the
perc
ent
ofba
nki
ng
asse
tsin
the
stat
eh
eld
byba
nks
belo
wth
em
edia
nsi
zeof
ban
kin
each
stat
ein
each
year
Rel
ativ
eca
pita
lrat
iois
the
capi
talt
oas
sets
rati
oof
smal
lban
ksm
inu
sth
atof
larg
eba
nks
Siz
eof
insu
ran
cere
lati
veto
bank
ing
plu
sin
sura
nce
inth
est
ate
ism
easu
red
asgr
oss
stat
epr
oduc
tfro
min
sura
nce
divi
ded
bygr
oss
stat
epr
odu
ctfr
omin
sura
nce
plu
sba
nkin
gIn
dica
tor
ison
eif
stat
ela
wpe
rmit
sba
nks
toen
ter
the
insu
ran
cebu
sin
ess
Sm
all
rmsh
are
isth
epe
rcen
tofa
lles
tabl
ishm
ents
inth
est
ate
that
hav
efe
wer
than
twen
tyem
ploy
ees
Par
tyin
dica
tor
ison
eif
the
Rep
rese
nta
tive
isa
Dem
ocra
tA
vera
geyi
eld
onba
nklo
ans
equa
lsto
tali
nter
est
inco
me
onal
ldom
esti
clo
ans
mad
eby
ban
ksin
the
stat
edi
vide
dby
tota
lloa
ns
The
aver
age
bank
failu
rera
teis
the
rati
oof
asse
tsin
faile
dba
nks
duri
ng
agi
ven
stat
edi
vide
dby
tota
lban
kas
sets
inth
atst
ate
aver
aged
over
the
year
s19
85to
1991
Un
itba
nki
ng
indi
cato
req
ual
son
efo
rst
ates
wit
hun
itba
nkin
gre
stri
ctio
nsT
hech
ange
inin
sura
nce
pow
ers
indi
cato
req
ual
son
eif
the
stat
ech
ange
dth
ela
wto
perm
itba
nks
tose
llin
sura
nce
duri
ngth
esa
mpl
epe
riod
We
divi
deth
eco
unt
ryin
tofo
urre
gion
san
din
clu
dere
gion
alin
dica
tors
inco
lum
n(5
)R
obu
stst
anda
rder
rors
are
inpa
ren
thes
es
an
d
den
ote
stat
isti
cally
sign
ica
nta
tth
e1
5an
d10
perc
entl
evel
s
WHAT DRIVES DEREGULATION 1459
branching of approximately 17 percent The relative capital-to-assets ratio of small and large banks however does not have astatistically signicant effect in the probit model
The impact of rival interests outside of banking is alsoconsistent with intrastate deregulation results Where banks cansell insurance legislators from states with larger insurancesectors relative to banking are less likely to vote for the amend-ment35 A one-standard-deviation increase in the relative size ofinsurance in those states which permit banks to sell insurancedecreases the probability that a legislator will favor the amend-ment by about 13 percent Turning to consumer interests thecoefficient on small rm share is not statistically signicant butthe coefficient on interest rates in the state is A one-standard-deviation increase in the average interest rate on loans raises theprobability that a legislator will support the amendment byroughly 8 percent The coefficient estimate for the bankingdistress variable measured as the average bank failure rate from1985 to 1991 is small and statistically signicant Overall theprobit analysis of the vote on national branching deregulationsupports the private-interest theory of deregulation and providesa consistency check that the importance of interests operating onthe state legislatures are very similar to those operating at thefederal level
V WHY DID BRANCHING DEREGULATION BEGIN DURING THE
1970S SHOCKS TO THE EQUILIBRIUM SUPPORTING BANK
BRANCHING RESTRICTIONS
A complete explanation of regulatory exit should account forwhy deregulation begins in the 1970s as well as the timing of thestate-by-state reforms [Kane 1996] In this section we describebroad technological legal and economic shocks that altered thepolitical-economy equilibrium which had kept antibranching regu-lations little changed for at least 30 years
Beginning in the 1970s three innovations reduced the valueto the protected banks of local geographic monopolies (see TableV) First automatic teller machines (ATMs) helped to erode thegeographic ties between customers and banks Second checkablemoney market mutual funds and the Merrill Lynch Cash Manage-ment Account demonstrated that banking by mail and telephone
35 The positive and statistically signicant estimates on the other insurancevariables may reect insurance industry support for the amendmentrsquos provisionslimiting banksrsquo insurance powers
QUARTERLY JOURNAL OF ECONOMICS1460
provided a convenient alternative to local banks36 Third techno-logical innovation and deregulation reduced transportation andcommunication costs particularly since the 1970s thereby lower-ing the costs for customers to use distant banks By increasing theelasticity of deposits supplied to banks these innovations reducedthe value of geographical restrictions to their traditional benecia-ries and thereby reduced their incentive to ght to maintain them[Peltzman 1976]
On the lending side increasing sophistication of credit-scoring techniques following innovations in information process-ing nancial theory and the development of large credit databases began to diminish the value of knowledge that localbankers had about the risks of borrowers in the community As a
36 Regulation Q which limited the interest rates that banks could pay ondeposits may have helped to drive depositors away from banks when the gapbetween market rates and deposit ceilings grew during the 1970s The eliminationof interest rate ceilings on large denomination certicates of deposit during the1970s appears to have hurt smaller and retail-oriented banks relative to largerwholesale banks [James 1983]
TABLE VBROAD TRENDS IN COMMERCIAL BANKING 1950ndash1995
YearNumberof ATMs
Domesticbank
deposits(billions)
Moneymarketmutual
fund(billions)
Percent ofdeposits 1
moneyfunds heldby banks
Smallbanksrsquo
percent ofbankingassets
Averagenumber of
bankfailures
(1) (2) (3) (4) (5) (6)1950 0 $ 154 $ 0 100 NA 41955 0 191 0 100 NA 31960 0 228 0 100 24 21965 0 330 0 100 20 41970 0 479 0 100 18 61975 9750 775 4 99 18 61980 18500 1182 76 94 17 101985 61117 1787 242 88 14 601990 80156 2339 493 83 11 1791995 122706 2552 745 77 8 61
Column (1) ATM gures are from Bank Network News The EFT Network Data Book (New YorkFaulkner and Gray Inc) The 1975 gure was unavailable 9750 is the number of ATMs in 1978 the rst yearfor which complete data are available Columns (2)ndash(4) Banksrsquo domestic deposits are from the Reports ofIncome and Condition money market mutual funds are from the Flow of Funds Data on all bank depositsforeign plus domestic are only available beginning in 1970 The trend in banksrsquo share (column (4)) is the sameusing total deposits instead of domestic deposits Column (5) Percent of banking assets held by small bankswhere a small bank is dened as a commercial bank with less than $100 million in assets in 1994 dollarsThese data are based on the Reports of Income and Condition Data on small banks are not available before1960 Column (6) Five-year average number of bank failures where the nal year is indicated in the rstcolumn These data are from FDIC Annual Report and the Quarterly Banking Prole
WHAT DRIVES DEREGULATION 1461
result of these innovations a national market developed forresidential mortgages credit card receivables have been securi-tized and bank lending to small business now relies less on thejudgment of loan officers and more on standardized scoringmodels
These changes have increased the potential protability forlarge banks to enter what had been the core of small bankactivities Large banksrsquo incentive to increase their lobbying pres-sure to be able to expand into these markets has thus beenincreasing over time In fact small banksrsquo market share began todecline even prior to the branching deregulation (Table V) As thevalue of a local banking relationships declined small rms thatwere the main borrowers from the small banks also probablybecame more likely to favor the entry of large banks into localmarkets With the deadweight costs of preventing large bankentry rising the private-interest theory predicts that small localbanks would become less likely able to maintain the branchingrestrictions [Becker 1983] Deregulation that reduces deadweightcosts of regulation is also consistent with the public-interesttheory
Kane [1996] argues that another major shock to the oldequilibrium was the increasing public awareness of the costlinessof having government-insured but (geographically) undiversiednancial institutions In the late 1970s the failure rate of banksbegan to rise (Table V) In the 1980s the Savings and Loan crisisand taxpayer bailout further heightened public awareness aboutthe costs of restrictions that make depository institutions morelikely to require infusions of taxpayer funds The failures thusmay have heightened public support for branching deregulation(see also Abrams and Settle [1993]) As noted above we did notnd any evidence that banking failures or distress in a stateaffected the speed with which the state deregulated but distressstill may be a common factor affecting all states
These technological economic and legal shocks generatedconditions that changed the long-standing balance favoring theantibranching forces The marginal value of lobbying to repealbranching restrictions increased just as the relative value to thesmall banks of maintaining branching restrictions was decliningThese nationwide shocks are common factors across states andthey are consistent with the positive duration dependence associ-ated with state-level deregulation that we demonstrated in FigureII In addition a broad change in lsquolsquoideologyrsquorsquo against government
QUARTERLY JOURNAL OF ECONOMICS1462
regulation could explain the positive duration dependence duringthe sample period
VI CONCLUSIONS
The private-interest theory of regulation can account for thepattern of bank branching deregulation of the 1970s and 1980sBeneciaries of branching regulation had supported a coalitionfavoring geographical restrictions despite their costs to consum-ers of nancial services Innovations that began in the 1970saltered the value of the restrictions to the affected parties and theresulting competition among interest groups can explain thesubsequent deregulation While some of our results are alsoconsistent with the public interest theorymdashfor example deregula-tion occurs earlier when small banks are in a relatively weaknancial positionmdashother results particularly evidence on theimportance of rivalries between small and large banks andbetween banking and insurance are difficult to explain with thepublic-interest approach We also nd that political-institutionalfactors affect deregulation although these variables may act asproxies for unmeasured economic interests The greater success ofthe private-interest theory to explain deregulation here than inprevious work may be due to the richer cross-sectional andtime-series variation in branching deregulation that can beexploited with a hazard model Future empirical studies ofendogenous deregulation thus may be most fruitful where changehas occurred across states over time such as in franchisinginsurance and public utilities (eg Joskow [1996])
Technological and nancial innovations will continue to erodethe benets to any interest group of maintaining regulatorybarriers in nancial services These forces are likely to bringabout reforms both domestically for example through legislationthat would increase bank powers (see Kroszner [1996] Krosznerand Rajan [1994 1997] and Kroszner and Stratmann [1998]) andinternationally for example through the extension of nancialservices provisions of NAFTAto reduce geographic barriers acrosscountries [Kroszner 1997]
APPENDIX EFFECTS OF DIFFERENT TYPES OF BRANCHING
DEREGULATION ON ECONOMIC GROWTH BANKING STRUCTUREAND BANK EFFICIENCY
We measure the importance of alternative forms of deregula-tion by their consequences for the economic growth and the
WHAT DRIVES DEREGULATION 1463
structure and efficiency of the banking sector across the states(see Jayaratne and Strahan [1996 1998]) To do so we run anOLS panel regression relating indicators for each type of deregula-tion to potential consequences Our dependent variables includeproxies for economic growth (growth in personal income) bankstructure (the log of the number of banking companies and thenumber of branches per banking company) and efficiency (the logof total noninterest expenses and the ratio of loan loss provisionsto total loans) The four deregulation indicators (MBHC branch-ing through merger and acquisition unrestricted statewide branch-ing and interstate banking) are equal to one in the years afterderegulation and zero before If a state deregulated before thesample period begins in 1970 the indicator is one throughout ifthe state does not deregulate then the indicator is zero through-out All regressions include state xed effects and time effects
Table VI shows the results of including the set of four
TABLE VIMULTIVARIATE REGRESSIONS OF THE EFFECTS OF DIFFERENT TYPES OF BRANCHING
DEREGULATION ON ECONOMIC GROWTH BANKING STRUCTUREAND BANK EFFICIENCY
Dependent variable
Growthin
personalincome
Log of of
bankingcompanies
Branchesper
bankingcompany
Log ofnon-
interestexpenses
Loan lossprovisions
(loans 1leases)
Merger amp acquisitionbranchingindicator
0013(0004)
2 0162(0037)
1146(0453)
2 0107(0035)
2 0007(0001)
Unrestricted state-wide branchingindicator
2 0008(0005)
0086(0059)
2 0043(0574)
0070(0037)
0005(0002)
Interstate bankingindicator
0006(0005)
2 0052(0057)
1064(0511)
0005(0030)
2 0001(0001)
Multibank holdingcompany indicator
2 0005(0006)
2 0096(0052)
2 0364(0529)
2 0095(0036)
2 0001(0001)
Dependent variablemean [median]
0086[0082]
4881[5136]
6649[5055]
13812[13803]
0009[0006]
This table contains regressions of state-level economic growth and state-level measures of the structureand efficiency of the banking industry Each dependent variable is regressed on a set of state indicatorvariables a set of year indicator variables and the set of four deregulation indicators these deregulationindicators are equal to one after deregulation and zero before If a state was always regulated the indicator iszero throughout if it was always deregulated the indicator is one throughout Sample period is 1975 to 1991and Delaware and South Dakota are dropped so N 5 833 Noninterest expenses become available only in1984 so N 5 392 The coefficient on the deregulation indicators is reported with robust standard errors inparentheses and denote statistically signicant at the 1 5 and 10 percent levels
QUARTERLY JOURNAL OF ECONOMICS1464
deregulation indicators simultaneously The branching by mergerand acquisition indicator is statistically signicant and of theexpected sign in all ve of the equations and larger in absolutevalue than any of the other coefficients on the indicators for theother types of deregulation Since there may be multicollinearityamong our indicators we also ran univariate specications whichincluded each indicator separately and the results are verysimilar These results imply that branching by merger andacquisition is the most important type of deregulation measuredby its consequences for economic growth banking structure andbank efficiency See Kroszner and Strahan [1998] for more details
UNIVERSITY OF CHICAGO GRADUATE SCHOOL OF BUSINESS AND NATIONAL BUREAU OF
ECONOMIC RESEARCH
FEDERAL RESERVE BANK OF NEW YORK RESEARCH amp MARKET ANALYSIS GROUP
REFERENCES
Abrams Burton A and Russell F Settle lsquolsquoPressure-Group Inuence and Institu-tional Change Branch-Banking Legislation during the Great DepressionrsquorsquoPublic Choice LXXVII (1993) 687ndash705
Amel Dean lsquolsquoState Laws Affecting the Geographic Expansion of CommercialBanksrsquorsquo unpublished manuscript Board of Governors of the Federal ReserveSystem 1993
Becker Gary S lsquolsquoA Theory of Competition among Pressure Groups for PoliticalInuencersquorsquo Quarterly Journal of Economics XCVIII (1983) 371ndash400
Berger Allen Anil Kashyap Joseph Scalise lsquolsquoThe Transformation of the U SBanking Industry What a Long Strange Trip Itrsquos Beenrsquorsquo Brookings Papers onEconomic Activity 2 (1995) 55ndash218
Berger Allen Anthony Saunders Joseph Scalise and Gregory Udell lsquolsquoThe Effectsof Bank Mergers and Acquisitions on Small Business Lendingrsquorsquo Journal ofFinancial Economics L (1998) 187ndash229
Brennan Geoffrey and James Buchanan lsquolsquoTowards a Tax Constitution forLeviathanrsquorsquo Journal of Public Economics VIII (1977) 255ndash273
Brickley James A and Christopher M James lsquolsquoThe Takeover Market CorporateBoard Composition and Ownership Structure The Case of Bankingrsquorsquo Journalof Law and Economics XXX (1987) 161ndash180
Cole Rebel and John Wolken lsquolsquoFinancial Services Used by Small BusinessesEvidence from the 1993 Survey of Small Business Financesrsquorsquo Federal ReserveBulletin LXXXI (1994) 629ndash676
Dixit Avinash The Making of Economic Policy A Transactions-Cost PoliticsPerspective (Cambridge MA MIT Press 1996)
Economides Nicolas R Glenn Hubbard and Darius Palia lsquolsquoThe Political Economyof Branching Restrictions and Deposit Insurancersquorsquo Journal of Law andEconomics XXIX (1996) 667ndash704
Flannery Mark lsquolsquoThe Social Costs of Unit Banking Restrictionsrsquorsquo Journal ofMonetary Economics XIII (1984) 237ndash249
Greene William H Econometric Analysis third edition (Upper Saddle River NJPrentice-Hall 1997)
Gunther Jeffery lsquolsquoRegional Capital Imbalances and the Removal of InterstateBanking Restrictionsrsquorsquo Economics Letters XLIV (1994) 439ndash442 lsquolsquoThe Erosion of Intrastate Branch Banking Restrictionsrsquorsquo unpublishedmanuscript Federal Reserve Bank of Dallas 1996
Irwin Douglas and Randall Kroszner lsquolsquoInterests Institutions and Ideology inSecuring Policy Change The Republican Conversion to Trade Liberalizationafter Smoot-Hawleyrsquorsquo Journal of Law and Economics XLII (1999) 643ndash673
WHAT DRIVES DEREGULATION 1465
James Christopher lsquolsquoAn Analysis of Intra-Industry Differences in the Effect ofRegulation The Case of Deposit Rate Ceilingsrsquorsquo Journal of Monetary Econom-ics XII (1983) 417ndash442
Jarrell Gregg lsquolsquoChange at the Exchange The Causes and Effects of DeregulationrsquorsquoJournal of Law and Economics XXVII (1984) 273ndash312
Jayaratne Jith and Philip E Strahan lsquolsquoThe Finance-Growth Nexus Evidencefrom Bank Branch Deregulationrsquorsquo Quarterly Journal of Economics CI (1996)639ndash670
Jayaratne Jith and Philip E Strahan lsquolsquoEntry Restrictions Industry Evolutionand Dynamic Efficiency Evidence from Commercial Bankingrsquorsquo Journal of Lawand Economics XLIX (1998) 239ndash274
Joskow Paul lsquolsquoComment on lsquoPower Struggles Explaining Deregulatory Reformsin Electricity Marketsrsquo rsquorsquo Brookings Papers on Economic Activity Microeconom-ics (1996) 251ndash264
Joskow Paul and Roger Noll lsquolsquoRegulation in Theory and Practice An Overviewrsquorsquoin Studies in Public Regulation Gary Fromm ed (Cambridge MA MITPress 1981) pp 1ndash65
Kahn Matthew E and John G Matsusaka lsquolsquoDemand for Environmental GoodsEvidence from Voting Patterns on California Initiativesrsquorsquo Journal of Law andEconomics XL (1997) 137ndash173
Kalbeisch J and R Prentice The Statistical Analysis of Failure Time Data (NewYork John Wiley amp Sons 1980)
Kane Edward J lsquolsquoDe Jure Interstate Banking Why Only NowrsquorsquoJournal of MoneyCredit and Banking XXVIII (1996) 141ndash161
Kiefer Nicholas lsquolsquoEconomic Duration Data and Hazard Functionsrsquorsquo Journal ofEconomic Literature XXVI (1988) 646ndash679
Kroszner Randall S lsquolsquoThe Evolution of Universal Banking and Its Regulation inTwentieth Century Americarsquorsquo in Universal Banking Financial System DesignReconsidered Anthony Saunders and Ingo Walter eds (New York IrwinProfessional Publishers 1996) pp 70ndash99 lsquolsquoThe Political Economy of Banking and Financial Regulation in the U Srsquorsquo inThe Banking and Financial Structure in the NAFTA Countries and ChileGeorge M von Furstenberg ed (Boston Kluwer Academic Publishers 1997)pp 200ndash213
Kroszner Randall S and Raghuram G Rajan lsquolsquoIs the Glass-Steagall ActJustied A Study of the U S Experience with Universal Banking before1933rsquorsquo American Economic Review LXXXIV (1994) 810ndash832
Kroszner Randall S and Raghuram G Rajan lsquolsquoOrganization Structure andCredibility Evidence from Bank Securities Activities before the Glass-Steagall Actrsquorsquo Journal of Monetary Economics XXXIX (1997) 475ndash516
Kroszner Randall S and Philip E Strahan lsquolsquoRegulatory Incentives and the ThriftCrisis Dividends Mutual-to-Stock Conversions and Financial DistressrsquorsquoJournal of Finance LI (1996) 1285ndash1320
Kroszner Randall S and Philip E Strahan lsquolsquoWhat Drives Deregulation Econom-ics and Politics of the Relaxation of Bank Branching Restrictionsrsquorsquo NBERWorking Paper No 6637 1998
Kroszner Randall S and Thomas Stratmann lsquolsquoInterest Group Competition andthe Organization of Congress Theory and Evidence from Financial ServicesPolitical Action Committeesrsquorsquo American Economic Review LXXXVIII (1998)pp 1163ndash1187
Kroszner Randall S and Thomas Stratmann lsquolsquoDoes Political Ambiguity PayCorporate Campaign Contributions and Rewards to Legislator Reputationrsquorsquounpublishedmanuscript University of Chicago Graduate School of Business1999
Lin D Y and L J Wei lsquolsquoThe Robust Inference for the Cox Proportional HazardsModel Journal of the American Statistical Association LXXXIV (1989)1074ndash1078
Macey Jonathan R and Geoffrey P Miller Banking Law and Regulation (BostonLittle Brown and Company 1992)
Makinson Larry Open Secrets The Encyclopedia of Congressional Money andPolitics (Washington DC Congressional Quarterly 1992)
Niskanen WilliamBureaucracy and Representative Government (Chicago Aldine-Atherton 1971)
QUARTERLY JOURNAL OF ECONOMICS1466
Noll Roger lsquolsquoComment on Peltzmanrsquorsquo Brookings Papers Microeconomics SpecialIssue (1989) 48ndash58
North Douglass lsquolsquoA Transactions Cost Theory of Politicsrsquorsquo Journal of TheoreticalPolitics II (1990) 355ndash367
Peltzman Sam lsquolsquoToward a More General Theory of Regulationrsquorsquo Journal of Lawand Economics XIX (1976) 109ndash148 lsquolsquoConstituent Interest and Congressional Votingrsquorsquo Journal of Law andEconomics XXVII (1984) 181ndash210 lsquolsquoThe Economic Theory of Regulation after a Decade of DeregulationrsquorsquoBrookings Papers Microeconomics Special Issue (1989) 1ndash41
Peterson Mitchell and Raghuram Rajan lsquolsquoThe Benets of Lending RelationshipsEvidence from Small Business Datarsquorsquo Journal of Finance XLIX (1994) 3ndash38
Poole Keith and Howard Rosenthal Congress A Political-Economic History ofRoll Call Voting (Oxford Oxford University Press 1997)
Stata Stata Reference Manual Release 5 (College Station TX Stata Press 1997)Stigler George J lsquolsquoThe Theory of Economic Regulationrsquorsquo Bell Journal of Economics
and Management Science II (1971) 3ndash21Stigler George J Chicago Studies in the Political Economy (Chicago University
of Chicago Press 1988)Strahan Philip E and James Weston lsquolsquoSmall Business Lending and the Changing
Structure of the Banking Industryrsquorsquo Journal of Banking and Finance XXII(1998) 821ndash845
Sylla Richard John Legler and John Wallis lsquolsquoBanks and State Public Finance inthe New Republic The United States 1790ndash1860rsquorsquo Journal of EconomicHistory XLVII (1987) 391ndash403
White Eugene The Regulation and Reform of the American Banking System1900ndash1929 (Princeton Princeton University Press 1983)
Winston Clifford lsquolsquoEconomic Deregulation Days of Reckoning for Microecono-mistsrsquorsquo Journal of Economic Literature XXXI (1993) 1263ndash1289
WHAT DRIVES DEREGULATION 1467
nies to acquire Maine banks As part of the Garn-St GermainAct federal legislators in 1982 amended the Bank HoldingCompany Act to allow failed banks and thrifts to be acquired byany bank holding company regardless of state laws (see egKroszner and Strahan [1996]) Many states then entered regionalor national reciprocal arrangements whereby their banks could bebought by any other state in the arrangement Between 1984 and1988 38 states joined one of these arrangements (see Amel[1993])
Table I and Figure I illustrate the history of state deregula-tion of geographical restrictions since 1970 The state-by-statederegulation culminates in the phaseout of interstate bank-ing restrictions with the passage of the 1994 Riegle-Neal Inter-state Banking and Branching Efficiency Act which effectivelycodied at the national level what had been occurring at the statelevel6
6 The Act permitted states to pass legislation to opt out of the interstatebanking provisions if the legislature did so before the provisions were to go intoeffect in mid-1997 Since only Texas and Montana have passed opt-out legislationthe United States will now have nearly complete interstate banking and branching
FIGURE IDeregulation of Restrictions on Intrastate Branching
Permitted Intrastate Branching before 1970
QUARTERLY JOURNAL OF ECONOMICS1442
III HYPOTHESES AND VARIABLE DEFINITIONS
Our empirical tests are based on the timing of intrastatebranching deregulation through merger and acquisition We focuson this form of branching deregulation because it has a muchgreater economic impact than the other forms of branchingderegulation (see the Appendix) Deregulation through mergerand acquisition is the only type of branching deregulation thatconsistently has a statistically signicant effect on bankingstructure bank efficiency and overall economic growth and theestimated magnitudes of the effects are greatest for this type ofderegulation We also analyze the interstate deregulation thattakes place at the national level by modeling voting behavior inthe House of Representatives on federal interstate branchinglegislation7
In the remainder of this section we describe our hypothesesabout the private-interest public-interest and political-institu-tional theories along with the data sources used to construct thevariables Some of the variables help us to distinguish between apublic- and private-interest approach while others will be consis-tent with both approaches The political-institutional approachhas little overlap in its implications with the other theories Ourmethod is to assess the relative importance of factors representingeach approach in speeding or slowing deregulation
A Hypotheses
Intraindustry Rivalry Small banks have fought to maintainand extend branching restrictions both historically and in therecent debates8 Smaller banks appear to have been the mainwinners from antibranching laws since these restrictions protectthem from competition from larger and more efficient bankingorganizations (see Flannery [1984] Jayaratne and Strahan [1998]
7 We do not employ a hazard model to explain the timing of interstatederegulation for two reasons the time clustering of interstate deregulation andcomplications of the strategic interdependence of the statesrsquo behavior As thenumber of states in an interstate arrangement increases for example the effectson potential acquirers is ambiguous they might prefer a larger pool of potentialtarget banks but might have to pay more as the number of competing acquirersexpands The benet to potential targets tends to increase with the number statesin an arrangement since the expected selling price increases with the number ofpotential bidders [Brickley and James 1987]
8 Economides Hubbard and Palia [1996] provide evidence that voting inCongress for the 1927 McFadden Act responded to small state banksrsquo interest inlimiting competition from large national banks See also White [1983] and Abramsand Settle [1993] for historical opposition On the small bank opposition to therecent branching deregulation see Kane [1996] and The Economist [8694 p 59]
WHAT DRIVES DEREGULATION 1443
and Winston [1993]) Branching restrictions thus tend to reducethe efficiency and consumer convenience of the banking system9
The private-interest theory therefore predicts that reformshould occur later in those states where the strength of the smallbanks relative to that of the large banks is greater According tothe public-interest theory deregulation should take place earlierwhere small banks are relatively important because the socialcosts of the regulation are directly related to the size of theprotected sector The social costs include deadweight costs andlosses associated with inefficiencies in the production of bankingservices relative to production without geographical restrictions
Interindustry Rivalry A number of states permit state-chartered commercial banks to sell insurance The insurancelobby would thus oppose the relaxation of branching restrictionswhen banks can sell insurance because such deregulation mightpermit banks to provide a more efficient insurance distributionnetwork10 According to the private-interest theory reform shouldoccur later in states where banks can sell insurance and theinsurance industry is important relative to the banking industryThe efficiency costs of the branching restrictions however risewith the size of the insurance sector since the restrictions preventthe exploitation of scope economies Under the public-interesttheory reform should therefore occur earlier where banks can sellinsurance and the insurance industry is relatively large11
Consumers of Banking Services Banks are a major source ofcredit for small rms [Cole and Wolken 1994] Branching deregu-lation tends to reduce banksrsquo local market power [Jayaratne andStrahan 1998] In addition Strahan and Weston [1998] nd thatlending to small businesses increases on average when smallbanks are purchased by other banking organizations and Bergeret al [1998] nd that credit availability to small businessesincreases in the years following a takeover of a small bank by alarger banking organization Since bank borrowers tend to benet
9 Flannery [1984] shows that small banks in states with branching restric-tions have higher costs than small banks in states without such restrictions andthe Appendix shows how the production of banking services becomes more efficientfollowing deregulation
10 The Economist [8694 p 59] describes the insurance industryrsquos opposi-tion to branching deregulation
11 Another rival interest we considered was the Savings and Loan industryThe share of assets in Savings and Loans relative to banking plus Savings andLoan assets in the state however had only small and statistically insignicanteffects in the models we estimate below
QUARTERLY JOURNAL OF ECONOMICS1444
from branching deregulation in particular and bank consolidationin general the private-interest theory would predict that stateswith numerous small bank-dependent rms would deregulateearlier12 This prediction however is also consistent with thepublic-interest theory since the social costs of the restrictions arehigher in states with more small bank-dependent rms
Another effect of branching restrictions on bank customerscan be related to the prices paid for bank services High loanprices before deregulation may reect for example a largemarket share for inefficient high-cost banks or high rents beingearned by the banks protected from competition In either casethe public-interest theory would suggest that states with rela-tively high initial loan interest rates should deregulate earlierThe private-interest theory however does not have a clearimplication high rents could lead the beneciaries to ght harderto maintain them (eg Stigler [1971]) or the deadweight costsassociated with these high rents could make it more difficult forthe beneciaries to maintain the restrictions (eg Becker [1983])13
Bank Stability Geographic diversication through branch-ing could mitigate instability problems that were importantduring much of our sample period thereby improving welfareInstability also may reduce the incentives of banks to lobby tomaintain protections because unstable banks are less likely tosurvive to reap the benets of the restrictions politicians also mayanticipate that such banks are less likely to be able to providefuture contributions and support [Gunther 1994 1996]14 Both theprivate-interest and public-interest theories thus suggest thatderegulation is most likely in states where banking instability isgreatest
12 On the other hand local banking monopolies created by branchingrestrictions could strengthen relationshipsbetween banks and small and medium-sized rms and increase the availability of credit to these rms [Petersen andRajan 1994] Also some have argued that small business lending declines whenlarge banks take over small banks (eg Berger Kashyap Scalise [1995])
13 High loan interest rates also could reect compensation for higher risksand fewer opportunities for banks to diversify If state-specic risks explaindifferences in average loans rates then neither theory would have any predictionsfor the sign of this effect Following branching deregulation however loan interestrates tend to fall but bank risk proles do not appear to change [Jayaratne andStrahan 1998 Kroszner and Strahan 1998] This suggests that high loan rates inantibranching states are not simply reecting state-specic risks
14 Also deregulation can arise as a response to banking instability under atheory of ignorant or misinformed voters in which a banking crisis acts as aneducational device to make the previously ignorant public aware of the costs of theantibranching policy (see Kane [1996])
WHAT DRIVES DEREGULATION 1445
Political-Institutional Factors Republicans are typically per-ceived as more likely to favor deregulation than Democrats so thepolitical-institutional theories suggest that states that are con-trolled by Democrats deregulate later than those controlled byRepublicans15 In addition we investigate whether reform be-comes more likely when one party controls the legislature and thegovernorship Note that these political effects must be interpretedwith caution since the views of the politicians may simply reectthe economic interests of the constituents in the state (seePeltzman [1984])
B Variable Denitions and Data Sources
Our main proxy for the strength of the small banks is thefraction of banking assets in the state in lsquolsquosmallrsquorsquo banks We denesmall banks as banks with assets below the median size in eachstate By allowing the denition of small to vary across states wetake into account cross-state heterogeneity in bank sizes16 An-nual data on bank size are from the fourth quarter Reports ofIncome and Conditions (lsquolsquoCall Reportsrsquorsquo) from the Federal ReserveBoard We also include the capital-to-asset ratio of small banksrelative to large as a measure of their relative nancial healthSpecically this variable is the asset-weighted average capital-asset ratio for small banks minus the asset-weighted averagecapital-asset ratio for large banks Annual data on bank capitalare from the Call Reports
To measure the effects of the rival insurance industry we rstconstruct an indicator variable that is one if the state permitsbanks to sell insurance For each state we then measure the sizeof the insurance sector (total value added in the state) relative tothe sum of the banking plus insurance sectors We will examine
15 For more detail on the importance of legislative structures party politicsand ideology see Poole and Rosenthal [1997] Kahn and Matsusaka [1997] Irwinand Kroszner [1999] and Kroszner and Stratmann [1999]
16 We also considered a variety of other denitions of relative size but therelative size results reported below are not sensitive to which denition we useFirst we used a xed measure of small bank that did not vary across statesdening small banks as those with assets below $100 million in 1994 dollarsSecond we applied the state-varying and xed denitions at the level of lsquolsquobankingorganizationsrsquorsquo (which include multibank holding companies) instead of at the levelof banks Third we calculated gini coefficients of bank size inequality and bankconcentration indices as alternative proxies for small versus large bank power inthe state Finally we also considered banks in lsquolsquoruralrsquorsquo areas (that is not located in aBureau of the Census lsquolsquostandard metropolitan statistical arearsquorsquo) as small since thevalue of the restrictions might be greatest in protecting banks outside of cities fromentry by the city banks All of the alternative denitions are highly correlated andyield the same results on the importance of intraindustry differences
QUARTERLY JOURNAL OF ECONOMICS1446
the effect separately for states that permit banks to sell insuranceand those that do not Data on value-added by industry are fromU S Commerce Department Bureau of Economic Analysis Sur-vey of Current Business [August 1994]
To measure the relative importance of small bank-dependentborrowers we include the proportion of all establishments operat-ing in the state with fewer than twenty employees These data arecompiled by the Bureau of the Census17 As a measure of priceswe use the difference between the average interest rate on loans inthe state and the prevailing federal funds rate where the loanrate equals the ratio of total interest income on all domestic loansdivided by total domestic loans held by banks operating in thestate18 These data are from the end-of-year Call Reports butbecome available beginning only in 1976 Our proxy for bankinstability is the failure rate of banks measured as the percent oftotal state banking assets in failed institutions from the FederalReserve Boardrsquos National Information Center database19
We include two political variables to adjust for any indepen-dent inuence of party politics First we measure the degree ofparty control of the state government by the fraction of the threebodies of the state government (the assembly senate and gover-norship) controlled by Democrats This variable for example isone-third if the Democrats have a majority in the assembly andthe Republicans have a majority in the senate and hold thegovernorship Second we include an indicator variable equal toone if the same party controls the governorrsquos office and hasmajorities in both chambers of the state legislature20
17 We collected the establishment data by the state for three cross sections(1976 1982 and 1987) and interpolated the data in the intermediate years SeeState and Metropolitan Data Book 1982 1986 and 1991
18 A measure of how individual customers would be affected is interest ratesthey receive on deposits We calculated the average deposit interest rate in thestate from the end-of-year Call Reports and did not nd any relationship betweenthis variable and the timing of deregulation in our analysis below Since Jayaratneand Strahan [1998] nd that branching deregulation has no effect on deposit ratesour nding of no relationship between timing and rates result could be due toeither individual bank customers anticipating little pricing effect and not lobbyingor an inability of individual depositors to form an effective lobby
19 As alternative measures of distress we used (1) the bank failure ratemeasured as the number of failed institutions in the state relative to the totalnumber of banks (2) the overall capital-to-asset ratio of all banks in the state and(3) the growth in state personal income Our results are not sensitive to the choiceof distress proxy
20 We also estimated the models with three indicator variables to reect theparty control of the assembly senate and governorship separately The coefficientestimates on these variables are qualitatively similar to those we present below(that is Democratic control tends to slow deregulation) and the results on theinterest group variables are unaffected by the specication of the political
WHAT DRIVES DEREGULATION 1447
IV METHODS AND RESULTS
In this section we rst estimate how the variables describedabove inuence the timing of state-level branching deregulationusing a hazard model Next we discuss the robustness of theresults including a test of whether the ex post impact of deregula-tion on the different interest groups is consistent with ourinterpretation of the coefficients We then investigate federalinterstate branching deregulation with a probit model to explorewhether the factors that affect the timing of intrastate deregula-tion also inuence voting in the House of Representatives oninterstate deregulation
A Hazard Model
We consider the period from the beginning of our sample(1970) until deregulation as the lsquolsquoduration of regulationrsquorsquo or thelsquolsquotime until deregulationrsquorsquo The hazard rate h(t) is the likelihoodthat a state deregulates at time t given that the state has not yetderegulated [Kalbeisch and Prentice 1980 Kiefer 1988 Greene1997] To model the duration of regulation we must decide whatstructure if any to impose on the hazard function The Kaplan-Meier product-limit estimator provides a simple nonparametricway to estimate the shape of the hazard function over time[Greene 1997]21 Figure II graphs this estimate for our data andshows that the hazard function is relatively at in the early yearsand then grows steeper in the later years
A duration model that can approximate this shape is theWeibull proportional hazards model The hazard rate functiontakes the form
h[t x(t) b] 5 h0(t) exp[x(t)8b]
where x(t) is a vector of time-varying covariates b is a vector ofunknown parameters to be estimated and the baseline hazard
variables Note also that Nebraska has a nonpartisan unicameral legislature Weassume that the party of the Governor controls the state government for Nebraskaand set the unied government indicator to one As another alternative we triedincluding a variable that is one if party control of the legislature or governorshipchanges but it did not have an economically or statistically signicant effect anddid not affect any of the other results
21 Denote each of the K years in our sample period Tk and order them suchthat T1 T2 and so on Let nk be the number of states that have not yet deregulatedby Tk and dk be the number of states that deregulate in year Tk The Kaplan-Meierestimate of the hazard rate in each year is dknk In other words the hazard rate isthe ratio of the number of states that actually deregulate in each year to thenumber of states that have not yet deregulated in that year
QUARTERLY JOURNAL OF ECONOMICS1448
rate h0(t) is ptp 2 1 with shape parameter p that will be estimatedfrom the data When p 1 this model displays a monotonicallyincreasing hazard rate To estimate the parameters we maximizethe following log-likelihood function [Kiefer 1988]
L(b) 5 oi 5 1
N
di ln h[ti x(ti) b] 2 oi 5 1
N
e 0
tih[u x(u) b] du
where N indexes the number of states at risk at the beginning ofthe sample period di is an indicator equal to one for the statesthat deregulated by the end of the sample period and zero for thestates that do not deregulate (ie the censored observations) andti is the time of deregulation for the ith state This formulationassumes that the time until deregulation is independent acrossstates but the observations of the same state over time are notassumed to be independent In addition we use a robust estima-tion procedure to calculate the standard errors [Lin and Wei1989]
We exclude states that deregulated before 1970 from theanalysis so N equals 39 states During the sample period 36 ofthe 39 states deregulate We have three censored observations forwhich di is zero (Arkansas Iowa and Minnesota) but our results
FIGURE IIKaplan-Meier Estimate of Hazard Rate for Branching Deregulation
WHAT DRIVES DEREGULATION 1449
are not sensitive to the inclusion of these censored observationsSince we observe each state in each year up to and including theyear of deregulation we have a total of 637 observations Table IIreports summary statistics and correlation coefficients for theexplanatory variables we use in our analysis
We use the Weibull model rather than a model that does notimpose any structure on the baseline hazard rate so that we cancalculate the change in the expected time to deregulation for agiven change in the levels of the covariates22 In the Weibullmodel we can invert the hazard function and map it into the timedomain Rewriting the Weibull model in this way the log of thetime to deregulation T is a linear function of the political economyfactors and an error term ln (T ) 5 x8b 1 e23 Because we assumethat the baseline hazard rate is Weibull with a shape characterizedby p the new coefficients will be scaled by p that is b 5 2 bp
The b coefficients represent the percentage change in thetime to deregulation for a one-unit change in the covariates Apositive coefficient implies that an increase in the variableincreases the expected time until deregulation To gauge theeconomic importance of the effects we will multiply the b by thestandard deviation of the explanatory variables and then evaluatehow much the changes in the variables raise or lower the expectedtime to deregulation24 In Table III we report the b coefficients
22 The Cox proportional hazards model does not impose any structure on thebaseline hazard rate h0(t) and takes the following form
h[t x(t) b] 5 h0(t) exp [x(t)8b]
Common factors that affect the probability of deregulation in the same way for allstates therefore do not inuence the estimated coefficients The cost of the Coxmodel is that because there is no structure on the baseline we can calculatechanges in only relative hazard rates associated with changes in the covariatesThe Weibull model however provides sufficient structure so that we can translateour estimates into a log expected time metric that allows us to calculate the changein the expected time to deregulation for a given change in the covariates SeeKiefer [1988] Greene [1997] and Stata [1997] When we compare the coefficientsestimated by the Cox and Weibull models on our data they are very close (seeKroszner and Strahan [1998]) so imposing a Weibull structure does not appear todistort our estimates
23 The error term e is independent of x and has an extreme value distributionscaled by 1p (see Kiefer [1988] Greene [1997] and Stata [1997] for more details onthis log expected time parameterization of the Weibull model)
24 The transformation to the log expected time domain is only strictly correctwhen the covariates are constant over time Our interpretation of the coefficients ispossible in the time-varying model if we assume that a one-standard-deviationdifference would hold over the entire sample period that is the deregulationwould occur so many years earlier or later if the variable in question were higherby one standard deviation from 1970 to 1992 Note that the mean number of yearsto deregulation in the sample is 158
QUARTERLY JOURNAL OF ECONOMICS1450
TABLE IISUMMARY STATISTICS FOR VARIABLES IN HAZARD MODEL 1970ndash1992
Panel A Univariate statistics MeanStandarddeviation Minimum Maximum
(1) (2) (3) (4)Small bank asset share of all
banking assets in state 011 005 001 023Relative size of insurance to
banking plus insurance inthe state 047 009 024 080
Indicator is 1 if banks can sellinsurance in the state 020 040 0 1
Small rm share of the numberof rms in the state 088 002 078 095
Capital ratio of small banksrelative to large in the state 002 001 2 001 009
Share of state government con-trolled by Democrats 065 036 0 1
Indicator is 1 if state controlledby one party 054 050 0 1
Average yield on bank loans inthe state minus Fed fundsrate 002 002 2 004 008
Panel B CorrelationsSmallbank
Relativesize of
insuranceInsuranceindicator
Smallrm
Smallbank
capitalDem
controlSingleparty
(1) (2) (3) (4) (5) (6) (7)Small bank asset share
of all banking assetsin state 1
Relative size of insur-ance to banking plusinsurance in the state 2 034 1
Indicator is 1 if bankscan sell insurance inthe state 020 2 012 1
Small rm share of thenumber of rms inthe state 030 2 011 005 1
Capital ratio of smallbanks relative tolarge in the state 2 066 020 2 010 2 006 1
Share of state govern-ment controlled byDemocrats 2 006 001 2 018 2 018 005 1
Indicator is 1 if statecontrolled by oneparty 007 012 001 2 006 2 006 030 1
Average yield on bankloans in the stateminus Fed funds rate 2 001 2 015 003 022 2 010 2 003 2 006
The number of observations is 637 except for the average interest rate on bank loans which becomesavailable only in 1976 so the number of observations is 408
TABLE IIIHAZARD MODEL OF POLITICAL ECONOMY FACTORS AFFECTING THE TIMING OF STATE
BRANCHING DEREGULATION 1970ndash1992
(1) (2) (3) (4) (5) (6)Small bank asset share of
all banking assets instate
648(122)
519(137)
632(135)
516(118)
488(133)
445(088)
Capital ratio of smallbanks relative to largein the state
1325(348)
921(319)
1301(362)
1067(346)
1144(375)
906(381)
Relative size of insurancein states where banksmay sell insurance 0otherwise
324(161)
259(149)
315(165)
285(154)
272(159)
055(134)
Indicator is 1 if banksmay sell insurance inthe state
045(010)
037(011)
045(010)
038(008)
036(009)
001(012)
Relative size of insurancein states where banksmay not sell insurance0 otherwise
2 093(043)
2 090(033)
2 102(051)
2 093(048)
2 096(053)
2 004(039)
Small rm share of thenumber of rms in thestate
2 972(211)
2 614(215)
2 952(211)
2 943(230)
2 952(270)
2 1509(276)
Share of state govern-ment controlled byDemocrats
031(012)
023(011)
030(012)
027(011)
026(011)
011(013)
Indicator is 1 if state con-trolled by one party
2 004(010)
2 004(007)
2 004(007)
2 001(009)
2 002(010)
017(009)
Average yield on bankloans in the stateminus Fed funds rate mdash
023(319) mdash mdash mdash mdash
Bank failure rate mdash mdash092
(111) mdash mdash mdash
Indicator is 1 if state hasunit banking law mdash mdash mdash
018(010)
019(010)
030(009)
Indicator is 1 if statechanges bank insur-ance powers mdash mdash mdash mdash
2 008(012)
2 026(013)
Includes regionalindicators No No No No No Yes
N 637 408 637 637 637 637Log likelihood 2 374 900 2 360 2 212 2 191 1192p-value of x 2 for
regression 001 001 001 001 001 001
The hazard model is Weibull where the dependent variable is the log expected time to branchingderegulation ln (T) 5 bx 1 eAll variables are measured for each state in each year Small bank asset share isthe percent of banking assets in the state held by banks below the median size of banks in each state in eachyear Relative capital ratio is the capital to assets ratio of small banks minus that of large banks Size ofinsurance relative to banking plus insurance in the state is measured as gross state product from insurancedivided by gross state product from insurance plus banking Insurance indicator is 1 if state law permits banksto enter the insurance business Small rm share is the percent of all establishments in the state that havefewer than twenty employees Party control variable is the share of the three bodies of state governmentcontrolled by Democrats One party control indicator is 1 if the same party controlrsquos the governorship and hasmajorities in both chambers of the state legislature Average yield on bank loans equals total interest incomeon all domestic loans made by banks in the state divided by total loans Bank failure rate is the ratio of assetsin failed banks during a given state and year divided by total bank assets in that state and year Unit bankingindicator equals 1 for states with unit banking restrictions The change in insurance powers indicator equalsone if the state changed the law to permit banks to sell insurance during the sample period We divide thecountry into four regions and include regional indicators in column (5) Robust standard errors are inparentheses and denote statistically signicant at the 1 5 and 10 percent levels
QUARTERLY JOURNAL OF ECONOMICS1452
The positive and statistically signicant coefficient on smallbank share implies that a greater small bank share delaysderegulation and this effect is economically important A one-standard-deviation increase in the small bank share results in a30 percent increase in the time until deregulation or about 47years This result supports the implications of private-interesttheory concerning intraindustry rivalry and is contrary to thepublic-interest theory
The positive and statistically signicant coefficient on therelative performance of small banks in the state implies thatstates deregulate later when small banks are relatively strongnancially25 A one-standard-deviation increase in the relativecapital-to-asset ratio results in a 15 percent rise in the time untilderegulation or about 24 years This result is consistent with theprivate-interest theory but may also be consistent with a public-interest interpretation because the risk of failures (and a costlytaxpayer-funded bailout) increases as small bank nancial healthdeclines
Interindustry competition also helps explain the timing ofderegulation In states where banks can sell insurance a rela-tively large insurance sector is associated with an increase in theexpected time to deregulation26 A one-standard-deviation in-crease in the relative size of the insurance sector in those statesthat permit banks to sell insurance leads to a 22 percent increasein the time until deregulation or about 35 years27 This result isconsistent with the private-interest theory but not the public-interest theory
Deregulation occurs earlier in states where small bank-dependent rms are relatively numerous A one-standard-
25 We also allowed small bank and large bank capital-to-asset ratios to enterseparately Their coefficient estimates were nearly equal in absolute value and ofopposite signs and so the data do not reject the relative small to large bank capitalratio specication we have chosen
26 When we use the relative size of the insurance sector in the state as awhole rather than insurance relative to banking plus insurance in the state weobtain similar results When we simultaneously include both measures of therelative size of insurance the interactions of the lsquolsquoinsurance relative to bankingplus insurancersquorsquo variable continue to be statisticallysignicant but the interactionsof lsquolsquoinsurance relative to the state as a wholersquorsquo variable are not
27 We have also estimated the model with the indicator variable equal to onefor states where banks may sell insurance but without the variables measuringtheimportance of insurance relative to banking In this model the coefficient on theinsurance indicator variablemdashan estimate of the difference in the expected time toderegulation in states where banks may sell insurance relative to states wherebanks may not sell insurancemdashequals 027 and is statistically signicant at the 5percent level This suggests that branching deregulation takes about 45 yearslonger in states where banks may sell insurance
WHAT DRIVES DEREGULATION 1453
deviation increase in the share of small rms reduces the timeuntil deregulation by 18 percent or about three years This resultconcerning the interests of consumers of banking services isconsistent with both the private- and public-interest theories
Column (2) of Table III introduces the average interest rateon loans to the model as a rough proxy for the cost to bankborrowers of these regulations These data become available onlyin 1976 so our sample size shrinks to 408 observations Thecoefficient on this variable is small and not statistically signi-cant The public-interest theory would imply that the coefficientshould be negative while the implication of the economic theorydoes not have a clear sign prediction
When we add the failure rate of banks to the specication incolumn (3) its coefficient is small and statistically insignicantWe do not nd a linkage between the timing of state deregulationand statewide banking distress Given that deposit insurance isfunded federally however bank failures may raise the publicrsquosconcerns about the costs of the regulations at the national ratherthan the state level This interpretation can explain why the rateof bank failure does not affect the timing of state-level deregula-tion even though bank instability does appear to rise generallyduring the period of increasing deregulation (see Section V below)
Finally the partisan structure of the state government alsoinuences when states deregulate As expected a higher propor-tion of Democrats in the government tends to delay deregulationA one-standard-deviation rise in the share of the governmentcontrolled by Democrats slows the deregulation by about twoyears Whether the state is dominated by one party however doesnot appear to affect the timing of the deregulation
To summarize our ndings from the hazard model privateinterests appear to play an important role in the deregulatoryprocess While private interests and public interests do sometimescoincide the results on the relative share of small banks and largebanks and on the relative size of insurance where banks competeare consistent with a private-interest approach but are difficult toexplain on public-interest grounds
B Robustness Tests
Omitted Variables States begin the sample period withdifferent degrees of restrictions on bank expansion These initialconditions may proxy for omitted factors determining the political-economy equilibrium in the state that could then affect the
QUARTERLY JOURNAL OF ECONOMICS1454
subsequent timing of deregulation For instance some statesbegan the period with unit banking while other states allowedlimited branching and the formation of multibank holdingcompanies
To test whether the stringency of initial restrictions onbranching matters we rst divide the sample into states with andwithout the most extreme form of branching regulation unitbanking28 Column (4) of Table III includes an indicator variableequal to one if the state began the period with unit bankingrestrictions The coefficient on this variable is positive andstatistically signicant but the other results remain virtuallyunchanged Next we include an indicator variable that is one ifthe state permitted multibank holding companies (MBHCs) toexist at the beginning of our sample period Thirty-four statesbegan the period allowing MBHCs The coefficient estimate forthis indicator variable (not reported) is small and not statisticallysignicant and its inclusion does not change the effects of theother variables
Eleven states expand bank powers to include the sale ofinsurance products during our sample period This could proxy foromitted factors about the relative strength of the insuranceindustry in the state In column (5) we include an indicatorvariable that is one if the state changes the law during our sampleperiod to permit banks to sell insurance business and zerootherwise The coefficient estimate is small and statisticallyinsignicant and does not affect the other results29
In column (6) we introduce four regional indicator variablesNorth South Midwest and West30 The coefficients on small bankshare the relative capital-to-assets ratio and small rm sharechange little Although the coefficient estimate still suggests that
28 We classify states that prohibited branching but permitted banks toestablish facilities as unit banking states The unit banking indicator is one for thefollowing states CO AR FL IL IA KS MN MO MT NE ND OK TX WI WVand WY
29 We conducted a likelihood ratio test of whether the unit banking statesthe MBHC states and the states that changed their regulations of bank insurancepowers could be pooled with the other states In all three cases we could not rejectpooling of the data
30 Since intrastate deregulation generally preceded interstate deregulationand the latter typically took the form of regional interstate compacts thepossibility of participating in a regional interstate banking compact could haveinuenced the decision to deregulate intrastate branching Our denitions of theregions are region 1 (South) contains AL AR DC FL GA KY LA MS NC OKSC TN TX and VA region 2 (Northeast) contains CT MA MD ME NH NJ NYPA RI VT and WV region 3 (Midwest) contains IA IL IN KS MI MN MO NEND OH SD and WI region 4 (West) contains the other states
WHAT DRIVES DEREGULATION 1455
in states where banks can sell insurance a larger relative size ofthe insurance sector delays deregulation the effect is no longerstatistically signicant The coefficient on the indicator variablefor whether a state expands bank powers to permit insurancesales during the sample period however now becomes statisti-cally signicantly negative This implies that deregulation occursroughly 41 years earlier in states where the insurance lobby maybe relatively weak While the inclusion of the regional indicatorsdoes not affect the other results the results on insurance doappear to be sensitive to the specication of the hazard model31
Estimation Technique To check the robustness of using ahazard model and to investigate whether the time-series or thecross-sectional variation in each of the factors accounts for thehazard results we estimate two simple linear probability models(neither are reported) First we run a pooled time-series cross-section OLS regression using the 637 observations from thehazard model The dependent variable equals one in the year ofderegulation and zero otherwise We include the same set ofregressors that were used in the hazard model and add state-levelxed effects This specication thus removes the cross-statevariation from the independent variables and focuses on theirmovements over time In this model the signs of the coefficientson the interest group factors are consistent with the results fromthe hazard model and the coefficients for the small bank relativecapital ratio and for the relative size of insurance where states cansell insurance are statistically signicant
In the second model we eliminate the time-series variationby averaging each of the explanatory variables over time We thushave one observation per state that had branching restrictions asof 1970 (N 5 39) The dependent variable equals the inverse of thenumber of years until deregulation which represents an estimateof the statersquos probability of deregulating in a year under theassumption that this probability is constant over time and eachyear provides an independent observation32 Although we haveonly 39 observations the coefficients on the small bank marketshare small bank relative capital ratio and small rm share
31 As further robustness check we also tried a population density variablewhich Abrams and Settle [1993] found relevant for regulatory change during the1920s and 1930s but its coefficient was not statistically signicant and did notaffect the other results
32 For instance the dependent variable would take the value of 01 for astate that deregulated in 1980 For the three states that did not deregulate thedependent variable equals 0
QUARTERLY JOURNAL OF ECONOMICS1456
variables are statistically signicant The signs of the effects ofthe insurance variables are the same as those reported in thehazard model but not statistically signicant
The hazard model and linear probability models thus producevery similar results The estimates from the latter suggest thatthe effects for the small bank market share and small rm sharevariables are driven primarily by cross-state variation the effectof the insurance variable is driven primarily by time-seriesvariation and the small bank capital effect is driven by bothcross-sectional and time-series variation
Ex Post Consequences of Deregulation Finally to check theplausibility of our interpretation of the hazard results we con-sider whether the ex post consequences of deregulation areconsistent with the ex ante lobbying positions we attribute to eachinterest group (see Kroszner and Strahan [1998] for details) Wend that small banks lose market share following deregulationand in states where banks can enter the insurance business theinsurance sector shrinks relative to the banking sector followingderegulation Borrowers also benet from deregulation throughlower average interest rates on loans These ndings support theprivate-interest interpretation of the coefficients in the hazardmodel groups that will benet lobby to speed deregulation andthose that will be harmed lobby to slow it
C Voting on Interstate Branching Deregulation in the U SHouse of Representatives
We now examine whether the forces driving intrastate branch-ing deregulation also drive interstate deregulation Financialservices interests are active contributors and lobbyists in Washing-ton Their political action committees constitute the largest groupof contributors to legislators providing nearly 20 percent of totalcongressional campaign contributions [Makinson 1992] and muchof their lobbying effort involves competition among rival interestswithin nancial services (see Kroszner and Stratmann [1998])
As noted above after virtually all states adopted intra- andinterstate branching deregulation the 1994 Riegle-Neal Act re-pealed the 1927 McFadden Act to eliminate all barriers tointerstate banking and branching by 1997 The key votes concern-ing the Riegle-Neal Act were either voice votes or extremelylopsided so we could not estimate a voting model from them Anumber of bills and amendments related to interstate branching
WHAT DRIVES DEREGULATION 1457
had been debated in Congress during the years prior to thepassage of the Riegle-Neal Act but a search of the weekly BNABanking Reporter and the Congressional Record produced onlyone roll-call vote related to interstate branching that was notlopsided This vote occurred in the House on November 14 1991and concerned an amendment sponsored by Wylie (R-OH) andNeal (D-NC) to introduce interstate banking and branchingderegulation to a nancial services reform package33 Althoughthe amendment passed by 210 to 208 the bill to which it wasattached subsequently was defeated
To check for the inuence of the factors we considered in thestate-level reforms we examine both the sponsorship of interstatebanking legislation and voting on the amendment The sponsorsof the Wylie-Neal amendment are from states with low small bankshares 004 in Ohio (Wylie) and 002 in North Carolina (Neal) Incontrast the sample mean in 1991 is 008 (median 5 007)Michigan home state of the Senatersquos sponsor of the 1994 Riegle-Neal Act also had relatively low small bank strength (small bankshare of 005)
Table IV reports estimates of a probit model where thedependent variable equals one if the legislator voted in favor ofthe amendment and zero otherwise The explanatory variablesare the same as those in the hazard model except that we use aparty affiliation indicator variable equal to one for Democrats andzero for Republicans as our proxy for political factors Thecoefficients in Table IV are the marginal effects (lsquolsquoslopesrsquorsquo) of aone-unit change of each variable on the probability that alegislator will vote for the amendment34 Note that a forcefavoring deregulation has a negative coefficient in the hazardmodel (Table III) but a positive coefficient in the probit model
Consistent with the state-level deregulation process legisla-tors are more likely to support the amendment if their states havea relatively low share of small banks As in the hazard model thefraction of small banks is the most important interest groupinuence on a legislatorrsquos voting decision A one-standard-deviation increase in small banksrsquo market share (from the mean)is associated with a decline in the probability of voting in favor of
33 The Wylie-Neal amendment also included provisions limiting certaininsurance and real estate powers of national banks [Congressional RecordNovember 14 1991 pp 10239ndash10242]
34 Since we have multiple legislators from each state we adjust the standarderrors to correct for the lack of independence among observations clustered in thesame state
QUARTERLY JOURNAL OF ECONOMICS1458
TAB
LE
IVM
AR
GIN
AL
EF
FE
CT
SF
RO
MA
PR
OB
ITM
OD
EL
OF
TH
EIN
FL
UE
NC
EO
FP
OL
ITIC
AL
EC
ON
OM
YF
AC
TO
RS
ON
TH
EH
OU
SE
VO
TE
ON
TH
EW
YL
IE-N
EA
L
AM
EN
DM
EN
TT
OP
ER
MIT
INT
ER
ST
AT
EB
AN
KB
RA
NC
HIN
G
NO
VE
MB
ER
14
1991
(1)
(2)
(3)
(4)
(5)
(6)
Smal
lban
kas
set
shar
eof
allb
anki
ng
asse
tsin
stat
e2
417
(12
4)2
454
(12
2)2
407
(12
0)2
443
(14
7)2
450
(14
9)2
490
(14
7)C
apit
alra
tio
ofsm
allb
anks
rela
tive
tola
rge
inth
est
ate
180
(37
5)2
68(3
42)
256
(37
5)1
69(3
78)
153
(38
0)2
45(4
25)
Rel
ativ
esi
zeof
insu
ranc
ein
stat
esw
here
bank
sm
ayse
llin
sura
nce
0ot
herw
ise
22
34
(10
2)2
249
(1
14)
22
33
(10
1)2
249
(1
04)
22
51
(10
2)2
154
(11
1)In
dica
tor
is1
ifba
nks
may
sell
insu
ranc
ein
the
stat
e0
79
(0
06)
074
(00
6)0
69
(0
06)
070
(00
6)0
70
(0
06)
067
(00
7)R
elat
ive
size
ofin
sura
nce
inst
ates
whe
reba
nks
may
not
sell
insu
ranc
e0
oth
erw
ise
180
(06
6)2
42
(0
72)
154
(0
67)
173
(06
6)1
72
(0
67)
203
(0
91)
Smal
lrm
shar
eof
the
num
ber
ofr
ms
inth
est
ate
21
23(0
92)
20
83(0
89)
21
62(1
07)
21
32(0
95)
21
30(1
00)
20
94(1
16)
Indi
cato
ris
1if
Rep
rese
ntat
ive
isa
Dem
ocra
t2
035
(00
7)2
035
(00
7)2
036
(00
7)2
035
(00
7)2
035
(00
7)2
036
(00
7)A
vera
geyi
eld
onba
nklo
ans
inth
est
ate
min
usF
edfu
nds
rate
mdash10
01
(51
5)mdash
mdashmdash
mdash
Ave
rage
fail
ure
rate
198
5to
1991
mdashmdash
228
(14
1)mdash
mdashmdash
Indi
cato
ris
1if
stat
eh
asun
itba
nkin
gla
wmdash
mdashmdash
004
(00
7)0
04(0
07)
20
01(0
08)
Indi
cato
ris
1if
stat
ech
ange
sba
nk
insu
ran
cepo
wer
smdash
mdashmdash
mdash2
001
(00
7)2
003
(00
7)In
clud
esre
gion
alin
dica
tors
N
oN
oN
oN
oN
oYe
sP
seud
o-R
20
170
180
170
170
170
18L
ogli
keli
hood
224
024
223
826
223
951
224
013
224
010
223
808
p-va
lue
ofx2
for
regr
essi
on
001
0
01
001
0
01
001
0
01
Th
ede
pen
den
tvar
iabl
eis
one
ifth
ele
gisl
ator
vote
sfo
rth
eam
endm
enta
nd
zero
ifag
ains
tT
he
repo
rted
coef
fici
ents
are
the
effe
cts
ofa
uni
tch
ange
ofth
ein
depe
nde
ntva
riab
le(f
rom
the
mea
n)on
the
prob
abili
tyof
voti
ng
infa
vor
ofth
eam
endm
ent
All
vari
able
sar
em
easu
red
byst
ate
N5
418
the
num
ber
ofR
epre
sent
ativ
esvo
tin
gon
the
amen
dmen
tS
mal
lban
kas
sets
har
eis
the
perc
ent
ofba
nki
ng
asse
tsin
the
stat
eh
eld
byba
nks
belo
wth
em
edia
nsi
zeof
ban
kin
each
stat
ein
each
year
Rel
ativ
eca
pita
lrat
iois
the
capi
talt
oas
sets
rati
oof
smal
lban
ksm
inu
sth
atof
larg
eba
nks
Siz
eof
insu
ran
cere
lati
veto
bank
ing
plu
sin
sura
nce
inth
est
ate
ism
easu
red
asgr
oss
stat
epr
oduc
tfro
min
sura
nce
divi
ded
bygr
oss
stat
epr
odu
ctfr
omin
sura
nce
plu
sba
nkin
gIn
dica
tor
ison
eif
stat
ela
wpe
rmit
sba
nks
toen
ter
the
insu
ran
cebu
sin
ess
Sm
all
rmsh
are
isth
epe
rcen
tofa
lles
tabl
ishm
ents
inth
est
ate
that
hav
efe
wer
than
twen
tyem
ploy
ees
Par
tyin
dica
tor
ison
eif
the
Rep
rese
nta
tive
isa
Dem
ocra
tA
vera
geyi
eld
onba
nklo
ans
equa
lsto
tali
nter
est
inco
me
onal
ldom
esti
clo
ans
mad
eby
ban
ksin
the
stat
edi
vide
dby
tota
lloa
ns
The
aver
age
bank
failu
rera
teis
the
rati
oof
asse
tsin
faile
dba
nks
duri
ng
agi
ven
stat
edi
vide
dby
tota
lban
kas
sets
inth
atst
ate
aver
aged
over
the
year
s19
85to
1991
Un
itba
nki
ng
indi
cato
req
ual
son
efo
rst
ates
wit
hun
itba
nkin
gre
stri
ctio
nsT
hech
ange
inin
sura
nce
pow
ers
indi
cato
req
ual
son
eif
the
stat
ech
ange
dth
ela
wto
perm
itba
nks
tose
llin
sura
nce
duri
ngth
esa
mpl
epe
riod
We
divi
deth
eco
unt
ryin
tofo
urre
gion
san
din
clu
dere
gion
alin
dica
tors
inco
lum
n(5
)R
obu
stst
anda
rder
rors
are
inpa
ren
thes
es
an
d
den
ote
stat
isti
cally
sign
ica
nta
tth
e1
5an
d10
perc
entl
evel
s
WHAT DRIVES DEREGULATION 1459
branching of approximately 17 percent The relative capital-to-assets ratio of small and large banks however does not have astatistically signicant effect in the probit model
The impact of rival interests outside of banking is alsoconsistent with intrastate deregulation results Where banks cansell insurance legislators from states with larger insurancesectors relative to banking are less likely to vote for the amend-ment35 A one-standard-deviation increase in the relative size ofinsurance in those states which permit banks to sell insurancedecreases the probability that a legislator will favor the amend-ment by about 13 percent Turning to consumer interests thecoefficient on small rm share is not statistically signicant butthe coefficient on interest rates in the state is A one-standard-deviation increase in the average interest rate on loans raises theprobability that a legislator will support the amendment byroughly 8 percent The coefficient estimate for the bankingdistress variable measured as the average bank failure rate from1985 to 1991 is small and statistically signicant Overall theprobit analysis of the vote on national branching deregulationsupports the private-interest theory of deregulation and providesa consistency check that the importance of interests operating onthe state legislatures are very similar to those operating at thefederal level
V WHY DID BRANCHING DEREGULATION BEGIN DURING THE
1970S SHOCKS TO THE EQUILIBRIUM SUPPORTING BANK
BRANCHING RESTRICTIONS
A complete explanation of regulatory exit should account forwhy deregulation begins in the 1970s as well as the timing of thestate-by-state reforms [Kane 1996] In this section we describebroad technological legal and economic shocks that altered thepolitical-economy equilibrium which had kept antibranching regu-lations little changed for at least 30 years
Beginning in the 1970s three innovations reduced the valueto the protected banks of local geographic monopolies (see TableV) First automatic teller machines (ATMs) helped to erode thegeographic ties between customers and banks Second checkablemoney market mutual funds and the Merrill Lynch Cash Manage-ment Account demonstrated that banking by mail and telephone
35 The positive and statistically signicant estimates on the other insurancevariables may reect insurance industry support for the amendmentrsquos provisionslimiting banksrsquo insurance powers
QUARTERLY JOURNAL OF ECONOMICS1460
provided a convenient alternative to local banks36 Third techno-logical innovation and deregulation reduced transportation andcommunication costs particularly since the 1970s thereby lower-ing the costs for customers to use distant banks By increasing theelasticity of deposits supplied to banks these innovations reducedthe value of geographical restrictions to their traditional benecia-ries and thereby reduced their incentive to ght to maintain them[Peltzman 1976]
On the lending side increasing sophistication of credit-scoring techniques following innovations in information process-ing nancial theory and the development of large credit databases began to diminish the value of knowledge that localbankers had about the risks of borrowers in the community As a
36 Regulation Q which limited the interest rates that banks could pay ondeposits may have helped to drive depositors away from banks when the gapbetween market rates and deposit ceilings grew during the 1970s The eliminationof interest rate ceilings on large denomination certicates of deposit during the1970s appears to have hurt smaller and retail-oriented banks relative to largerwholesale banks [James 1983]
TABLE VBROAD TRENDS IN COMMERCIAL BANKING 1950ndash1995
YearNumberof ATMs
Domesticbank
deposits(billions)
Moneymarketmutual
fund(billions)
Percent ofdeposits 1
moneyfunds heldby banks
Smallbanksrsquo
percent ofbankingassets
Averagenumber of
bankfailures
(1) (2) (3) (4) (5) (6)1950 0 $ 154 $ 0 100 NA 41955 0 191 0 100 NA 31960 0 228 0 100 24 21965 0 330 0 100 20 41970 0 479 0 100 18 61975 9750 775 4 99 18 61980 18500 1182 76 94 17 101985 61117 1787 242 88 14 601990 80156 2339 493 83 11 1791995 122706 2552 745 77 8 61
Column (1) ATM gures are from Bank Network News The EFT Network Data Book (New YorkFaulkner and Gray Inc) The 1975 gure was unavailable 9750 is the number of ATMs in 1978 the rst yearfor which complete data are available Columns (2)ndash(4) Banksrsquo domestic deposits are from the Reports ofIncome and Condition money market mutual funds are from the Flow of Funds Data on all bank depositsforeign plus domestic are only available beginning in 1970 The trend in banksrsquo share (column (4)) is the sameusing total deposits instead of domestic deposits Column (5) Percent of banking assets held by small bankswhere a small bank is dened as a commercial bank with less than $100 million in assets in 1994 dollarsThese data are based on the Reports of Income and Condition Data on small banks are not available before1960 Column (6) Five-year average number of bank failures where the nal year is indicated in the rstcolumn These data are from FDIC Annual Report and the Quarterly Banking Prole
WHAT DRIVES DEREGULATION 1461
result of these innovations a national market developed forresidential mortgages credit card receivables have been securi-tized and bank lending to small business now relies less on thejudgment of loan officers and more on standardized scoringmodels
These changes have increased the potential protability forlarge banks to enter what had been the core of small bankactivities Large banksrsquo incentive to increase their lobbying pres-sure to be able to expand into these markets has thus beenincreasing over time In fact small banksrsquo market share began todecline even prior to the branching deregulation (Table V) As thevalue of a local banking relationships declined small rms thatwere the main borrowers from the small banks also probablybecame more likely to favor the entry of large banks into localmarkets With the deadweight costs of preventing large bankentry rising the private-interest theory predicts that small localbanks would become less likely able to maintain the branchingrestrictions [Becker 1983] Deregulation that reduces deadweightcosts of regulation is also consistent with the public-interesttheory
Kane [1996] argues that another major shock to the oldequilibrium was the increasing public awareness of the costlinessof having government-insured but (geographically) undiversiednancial institutions In the late 1970s the failure rate of banksbegan to rise (Table V) In the 1980s the Savings and Loan crisisand taxpayer bailout further heightened public awareness aboutthe costs of restrictions that make depository institutions morelikely to require infusions of taxpayer funds The failures thusmay have heightened public support for branching deregulation(see also Abrams and Settle [1993]) As noted above we did notnd any evidence that banking failures or distress in a stateaffected the speed with which the state deregulated but distressstill may be a common factor affecting all states
These technological economic and legal shocks generatedconditions that changed the long-standing balance favoring theantibranching forces The marginal value of lobbying to repealbranching restrictions increased just as the relative value to thesmall banks of maintaining branching restrictions was decliningThese nationwide shocks are common factors across states andthey are consistent with the positive duration dependence associ-ated with state-level deregulation that we demonstrated in FigureII In addition a broad change in lsquolsquoideologyrsquorsquo against government
QUARTERLY JOURNAL OF ECONOMICS1462
regulation could explain the positive duration dependence duringthe sample period
VI CONCLUSIONS
The private-interest theory of regulation can account for thepattern of bank branching deregulation of the 1970s and 1980sBeneciaries of branching regulation had supported a coalitionfavoring geographical restrictions despite their costs to consum-ers of nancial services Innovations that began in the 1970saltered the value of the restrictions to the affected parties and theresulting competition among interest groups can explain thesubsequent deregulation While some of our results are alsoconsistent with the public interest theorymdashfor example deregula-tion occurs earlier when small banks are in a relatively weaknancial positionmdashother results particularly evidence on theimportance of rivalries between small and large banks andbetween banking and insurance are difficult to explain with thepublic-interest approach We also nd that political-institutionalfactors affect deregulation although these variables may act asproxies for unmeasured economic interests The greater success ofthe private-interest theory to explain deregulation here than inprevious work may be due to the richer cross-sectional andtime-series variation in branching deregulation that can beexploited with a hazard model Future empirical studies ofendogenous deregulation thus may be most fruitful where changehas occurred across states over time such as in franchisinginsurance and public utilities (eg Joskow [1996])
Technological and nancial innovations will continue to erodethe benets to any interest group of maintaining regulatorybarriers in nancial services These forces are likely to bringabout reforms both domestically for example through legislationthat would increase bank powers (see Kroszner [1996] Krosznerand Rajan [1994 1997] and Kroszner and Stratmann [1998]) andinternationally for example through the extension of nancialservices provisions of NAFTAto reduce geographic barriers acrosscountries [Kroszner 1997]
APPENDIX EFFECTS OF DIFFERENT TYPES OF BRANCHING
DEREGULATION ON ECONOMIC GROWTH BANKING STRUCTUREAND BANK EFFICIENCY
We measure the importance of alternative forms of deregula-tion by their consequences for the economic growth and the
WHAT DRIVES DEREGULATION 1463
structure and efficiency of the banking sector across the states(see Jayaratne and Strahan [1996 1998]) To do so we run anOLS panel regression relating indicators for each type of deregula-tion to potential consequences Our dependent variables includeproxies for economic growth (growth in personal income) bankstructure (the log of the number of banking companies and thenumber of branches per banking company) and efficiency (the logof total noninterest expenses and the ratio of loan loss provisionsto total loans) The four deregulation indicators (MBHC branch-ing through merger and acquisition unrestricted statewide branch-ing and interstate banking) are equal to one in the years afterderegulation and zero before If a state deregulated before thesample period begins in 1970 the indicator is one throughout ifthe state does not deregulate then the indicator is zero through-out All regressions include state xed effects and time effects
Table VI shows the results of including the set of four
TABLE VIMULTIVARIATE REGRESSIONS OF THE EFFECTS OF DIFFERENT TYPES OF BRANCHING
DEREGULATION ON ECONOMIC GROWTH BANKING STRUCTUREAND BANK EFFICIENCY
Dependent variable
Growthin
personalincome
Log of of
bankingcompanies
Branchesper
bankingcompany
Log ofnon-
interestexpenses
Loan lossprovisions
(loans 1leases)
Merger amp acquisitionbranchingindicator
0013(0004)
2 0162(0037)
1146(0453)
2 0107(0035)
2 0007(0001)
Unrestricted state-wide branchingindicator
2 0008(0005)
0086(0059)
2 0043(0574)
0070(0037)
0005(0002)
Interstate bankingindicator
0006(0005)
2 0052(0057)
1064(0511)
0005(0030)
2 0001(0001)
Multibank holdingcompany indicator
2 0005(0006)
2 0096(0052)
2 0364(0529)
2 0095(0036)
2 0001(0001)
Dependent variablemean [median]
0086[0082]
4881[5136]
6649[5055]
13812[13803]
0009[0006]
This table contains regressions of state-level economic growth and state-level measures of the structureand efficiency of the banking industry Each dependent variable is regressed on a set of state indicatorvariables a set of year indicator variables and the set of four deregulation indicators these deregulationindicators are equal to one after deregulation and zero before If a state was always regulated the indicator iszero throughout if it was always deregulated the indicator is one throughout Sample period is 1975 to 1991and Delaware and South Dakota are dropped so N 5 833 Noninterest expenses become available only in1984 so N 5 392 The coefficient on the deregulation indicators is reported with robust standard errors inparentheses and denote statistically signicant at the 1 5 and 10 percent levels
QUARTERLY JOURNAL OF ECONOMICS1464
deregulation indicators simultaneously The branching by mergerand acquisition indicator is statistically signicant and of theexpected sign in all ve of the equations and larger in absolutevalue than any of the other coefficients on the indicators for theother types of deregulation Since there may be multicollinearityamong our indicators we also ran univariate specications whichincluded each indicator separately and the results are verysimilar These results imply that branching by merger andacquisition is the most important type of deregulation measuredby its consequences for economic growth banking structure andbank efficiency See Kroszner and Strahan [1998] for more details
UNIVERSITY OF CHICAGO GRADUATE SCHOOL OF BUSINESS AND NATIONAL BUREAU OF
ECONOMIC RESEARCH
FEDERAL RESERVE BANK OF NEW YORK RESEARCH amp MARKET ANALYSIS GROUP
REFERENCES
Abrams Burton A and Russell F Settle lsquolsquoPressure-Group Inuence and Institu-tional Change Branch-Banking Legislation during the Great DepressionrsquorsquoPublic Choice LXXVII (1993) 687ndash705
Amel Dean lsquolsquoState Laws Affecting the Geographic Expansion of CommercialBanksrsquorsquo unpublished manuscript Board of Governors of the Federal ReserveSystem 1993
Becker Gary S lsquolsquoA Theory of Competition among Pressure Groups for PoliticalInuencersquorsquo Quarterly Journal of Economics XCVIII (1983) 371ndash400
Berger Allen Anil Kashyap Joseph Scalise lsquolsquoThe Transformation of the U SBanking Industry What a Long Strange Trip Itrsquos Beenrsquorsquo Brookings Papers onEconomic Activity 2 (1995) 55ndash218
Berger Allen Anthony Saunders Joseph Scalise and Gregory Udell lsquolsquoThe Effectsof Bank Mergers and Acquisitions on Small Business Lendingrsquorsquo Journal ofFinancial Economics L (1998) 187ndash229
Brennan Geoffrey and James Buchanan lsquolsquoTowards a Tax Constitution forLeviathanrsquorsquo Journal of Public Economics VIII (1977) 255ndash273
Brickley James A and Christopher M James lsquolsquoThe Takeover Market CorporateBoard Composition and Ownership Structure The Case of Bankingrsquorsquo Journalof Law and Economics XXX (1987) 161ndash180
Cole Rebel and John Wolken lsquolsquoFinancial Services Used by Small BusinessesEvidence from the 1993 Survey of Small Business Financesrsquorsquo Federal ReserveBulletin LXXXI (1994) 629ndash676
Dixit Avinash The Making of Economic Policy A Transactions-Cost PoliticsPerspective (Cambridge MA MIT Press 1996)
Economides Nicolas R Glenn Hubbard and Darius Palia lsquolsquoThe Political Economyof Branching Restrictions and Deposit Insurancersquorsquo Journal of Law andEconomics XXIX (1996) 667ndash704
Flannery Mark lsquolsquoThe Social Costs of Unit Banking Restrictionsrsquorsquo Journal ofMonetary Economics XIII (1984) 237ndash249
Greene William H Econometric Analysis third edition (Upper Saddle River NJPrentice-Hall 1997)
Gunther Jeffery lsquolsquoRegional Capital Imbalances and the Removal of InterstateBanking Restrictionsrsquorsquo Economics Letters XLIV (1994) 439ndash442 lsquolsquoThe Erosion of Intrastate Branch Banking Restrictionsrsquorsquo unpublishedmanuscript Federal Reserve Bank of Dallas 1996
Irwin Douglas and Randall Kroszner lsquolsquoInterests Institutions and Ideology inSecuring Policy Change The Republican Conversion to Trade Liberalizationafter Smoot-Hawleyrsquorsquo Journal of Law and Economics XLII (1999) 643ndash673
WHAT DRIVES DEREGULATION 1465
James Christopher lsquolsquoAn Analysis of Intra-Industry Differences in the Effect ofRegulation The Case of Deposit Rate Ceilingsrsquorsquo Journal of Monetary Econom-ics XII (1983) 417ndash442
Jarrell Gregg lsquolsquoChange at the Exchange The Causes and Effects of DeregulationrsquorsquoJournal of Law and Economics XXVII (1984) 273ndash312
Jayaratne Jith and Philip E Strahan lsquolsquoThe Finance-Growth Nexus Evidencefrom Bank Branch Deregulationrsquorsquo Quarterly Journal of Economics CI (1996)639ndash670
Jayaratne Jith and Philip E Strahan lsquolsquoEntry Restrictions Industry Evolutionand Dynamic Efficiency Evidence from Commercial Bankingrsquorsquo Journal of Lawand Economics XLIX (1998) 239ndash274
Joskow Paul lsquolsquoComment on lsquoPower Struggles Explaining Deregulatory Reformsin Electricity Marketsrsquo rsquorsquo Brookings Papers on Economic Activity Microeconom-ics (1996) 251ndash264
Joskow Paul and Roger Noll lsquolsquoRegulation in Theory and Practice An Overviewrsquorsquoin Studies in Public Regulation Gary Fromm ed (Cambridge MA MITPress 1981) pp 1ndash65
Kahn Matthew E and John G Matsusaka lsquolsquoDemand for Environmental GoodsEvidence from Voting Patterns on California Initiativesrsquorsquo Journal of Law andEconomics XL (1997) 137ndash173
Kalbeisch J and R Prentice The Statistical Analysis of Failure Time Data (NewYork John Wiley amp Sons 1980)
Kane Edward J lsquolsquoDe Jure Interstate Banking Why Only NowrsquorsquoJournal of MoneyCredit and Banking XXVIII (1996) 141ndash161
Kiefer Nicholas lsquolsquoEconomic Duration Data and Hazard Functionsrsquorsquo Journal ofEconomic Literature XXVI (1988) 646ndash679
Kroszner Randall S lsquolsquoThe Evolution of Universal Banking and Its Regulation inTwentieth Century Americarsquorsquo in Universal Banking Financial System DesignReconsidered Anthony Saunders and Ingo Walter eds (New York IrwinProfessional Publishers 1996) pp 70ndash99 lsquolsquoThe Political Economy of Banking and Financial Regulation in the U Srsquorsquo inThe Banking and Financial Structure in the NAFTA Countries and ChileGeorge M von Furstenberg ed (Boston Kluwer Academic Publishers 1997)pp 200ndash213
Kroszner Randall S and Raghuram G Rajan lsquolsquoIs the Glass-Steagall ActJustied A Study of the U S Experience with Universal Banking before1933rsquorsquo American Economic Review LXXXIV (1994) 810ndash832
Kroszner Randall S and Raghuram G Rajan lsquolsquoOrganization Structure andCredibility Evidence from Bank Securities Activities before the Glass-Steagall Actrsquorsquo Journal of Monetary Economics XXXIX (1997) 475ndash516
Kroszner Randall S and Philip E Strahan lsquolsquoRegulatory Incentives and the ThriftCrisis Dividends Mutual-to-Stock Conversions and Financial DistressrsquorsquoJournal of Finance LI (1996) 1285ndash1320
Kroszner Randall S and Philip E Strahan lsquolsquoWhat Drives Deregulation Econom-ics and Politics of the Relaxation of Bank Branching Restrictionsrsquorsquo NBERWorking Paper No 6637 1998
Kroszner Randall S and Thomas Stratmann lsquolsquoInterest Group Competition andthe Organization of Congress Theory and Evidence from Financial ServicesPolitical Action Committeesrsquorsquo American Economic Review LXXXVIII (1998)pp 1163ndash1187
Kroszner Randall S and Thomas Stratmann lsquolsquoDoes Political Ambiguity PayCorporate Campaign Contributions and Rewards to Legislator Reputationrsquorsquounpublishedmanuscript University of Chicago Graduate School of Business1999
Lin D Y and L J Wei lsquolsquoThe Robust Inference for the Cox Proportional HazardsModel Journal of the American Statistical Association LXXXIV (1989)1074ndash1078
Macey Jonathan R and Geoffrey P Miller Banking Law and Regulation (BostonLittle Brown and Company 1992)
Makinson Larry Open Secrets The Encyclopedia of Congressional Money andPolitics (Washington DC Congressional Quarterly 1992)
Niskanen WilliamBureaucracy and Representative Government (Chicago Aldine-Atherton 1971)
QUARTERLY JOURNAL OF ECONOMICS1466
Noll Roger lsquolsquoComment on Peltzmanrsquorsquo Brookings Papers Microeconomics SpecialIssue (1989) 48ndash58
North Douglass lsquolsquoA Transactions Cost Theory of Politicsrsquorsquo Journal of TheoreticalPolitics II (1990) 355ndash367
Peltzman Sam lsquolsquoToward a More General Theory of Regulationrsquorsquo Journal of Lawand Economics XIX (1976) 109ndash148 lsquolsquoConstituent Interest and Congressional Votingrsquorsquo Journal of Law andEconomics XXVII (1984) 181ndash210 lsquolsquoThe Economic Theory of Regulation after a Decade of DeregulationrsquorsquoBrookings Papers Microeconomics Special Issue (1989) 1ndash41
Peterson Mitchell and Raghuram Rajan lsquolsquoThe Benets of Lending RelationshipsEvidence from Small Business Datarsquorsquo Journal of Finance XLIX (1994) 3ndash38
Poole Keith and Howard Rosenthal Congress A Political-Economic History ofRoll Call Voting (Oxford Oxford University Press 1997)
Stata Stata Reference Manual Release 5 (College Station TX Stata Press 1997)Stigler George J lsquolsquoThe Theory of Economic Regulationrsquorsquo Bell Journal of Economics
and Management Science II (1971) 3ndash21Stigler George J Chicago Studies in the Political Economy (Chicago University
of Chicago Press 1988)Strahan Philip E and James Weston lsquolsquoSmall Business Lending and the Changing
Structure of the Banking Industryrsquorsquo Journal of Banking and Finance XXII(1998) 821ndash845
Sylla Richard John Legler and John Wallis lsquolsquoBanks and State Public Finance inthe New Republic The United States 1790ndash1860rsquorsquo Journal of EconomicHistory XLVII (1987) 391ndash403
White Eugene The Regulation and Reform of the American Banking System1900ndash1929 (Princeton Princeton University Press 1983)
Winston Clifford lsquolsquoEconomic Deregulation Days of Reckoning for Microecono-mistsrsquorsquo Journal of Economic Literature XXXI (1993) 1263ndash1289
WHAT DRIVES DEREGULATION 1467
III HYPOTHESES AND VARIABLE DEFINITIONS
Our empirical tests are based on the timing of intrastatebranching deregulation through merger and acquisition We focuson this form of branching deregulation because it has a muchgreater economic impact than the other forms of branchingderegulation (see the Appendix) Deregulation through mergerand acquisition is the only type of branching deregulation thatconsistently has a statistically signicant effect on bankingstructure bank efficiency and overall economic growth and theestimated magnitudes of the effects are greatest for this type ofderegulation We also analyze the interstate deregulation thattakes place at the national level by modeling voting behavior inthe House of Representatives on federal interstate branchinglegislation7
In the remainder of this section we describe our hypothesesabout the private-interest public-interest and political-institu-tional theories along with the data sources used to construct thevariables Some of the variables help us to distinguish between apublic- and private-interest approach while others will be consis-tent with both approaches The political-institutional approachhas little overlap in its implications with the other theories Ourmethod is to assess the relative importance of factors representingeach approach in speeding or slowing deregulation
A Hypotheses
Intraindustry Rivalry Small banks have fought to maintainand extend branching restrictions both historically and in therecent debates8 Smaller banks appear to have been the mainwinners from antibranching laws since these restrictions protectthem from competition from larger and more efficient bankingorganizations (see Flannery [1984] Jayaratne and Strahan [1998]
7 We do not employ a hazard model to explain the timing of interstatederegulation for two reasons the time clustering of interstate deregulation andcomplications of the strategic interdependence of the statesrsquo behavior As thenumber of states in an interstate arrangement increases for example the effectson potential acquirers is ambiguous they might prefer a larger pool of potentialtarget banks but might have to pay more as the number of competing acquirersexpands The benet to potential targets tends to increase with the number statesin an arrangement since the expected selling price increases with the number ofpotential bidders [Brickley and James 1987]
8 Economides Hubbard and Palia [1996] provide evidence that voting inCongress for the 1927 McFadden Act responded to small state banksrsquo interest inlimiting competition from large national banks See also White [1983] and Abramsand Settle [1993] for historical opposition On the small bank opposition to therecent branching deregulation see Kane [1996] and The Economist [8694 p 59]
WHAT DRIVES DEREGULATION 1443
and Winston [1993]) Branching restrictions thus tend to reducethe efficiency and consumer convenience of the banking system9
The private-interest theory therefore predicts that reformshould occur later in those states where the strength of the smallbanks relative to that of the large banks is greater According tothe public-interest theory deregulation should take place earlierwhere small banks are relatively important because the socialcosts of the regulation are directly related to the size of theprotected sector The social costs include deadweight costs andlosses associated with inefficiencies in the production of bankingservices relative to production without geographical restrictions
Interindustry Rivalry A number of states permit state-chartered commercial banks to sell insurance The insurancelobby would thus oppose the relaxation of branching restrictionswhen banks can sell insurance because such deregulation mightpermit banks to provide a more efficient insurance distributionnetwork10 According to the private-interest theory reform shouldoccur later in states where banks can sell insurance and theinsurance industry is important relative to the banking industryThe efficiency costs of the branching restrictions however risewith the size of the insurance sector since the restrictions preventthe exploitation of scope economies Under the public-interesttheory reform should therefore occur earlier where banks can sellinsurance and the insurance industry is relatively large11
Consumers of Banking Services Banks are a major source ofcredit for small rms [Cole and Wolken 1994] Branching deregu-lation tends to reduce banksrsquo local market power [Jayaratne andStrahan 1998] In addition Strahan and Weston [1998] nd thatlending to small businesses increases on average when smallbanks are purchased by other banking organizations and Bergeret al [1998] nd that credit availability to small businessesincreases in the years following a takeover of a small bank by alarger banking organization Since bank borrowers tend to benet
9 Flannery [1984] shows that small banks in states with branching restric-tions have higher costs than small banks in states without such restrictions andthe Appendix shows how the production of banking services becomes more efficientfollowing deregulation
10 The Economist [8694 p 59] describes the insurance industryrsquos opposi-tion to branching deregulation
11 Another rival interest we considered was the Savings and Loan industryThe share of assets in Savings and Loans relative to banking plus Savings andLoan assets in the state however had only small and statistically insignicanteffects in the models we estimate below
QUARTERLY JOURNAL OF ECONOMICS1444
from branching deregulation in particular and bank consolidationin general the private-interest theory would predict that stateswith numerous small bank-dependent rms would deregulateearlier12 This prediction however is also consistent with thepublic-interest theory since the social costs of the restrictions arehigher in states with more small bank-dependent rms
Another effect of branching restrictions on bank customerscan be related to the prices paid for bank services High loanprices before deregulation may reect for example a largemarket share for inefficient high-cost banks or high rents beingearned by the banks protected from competition In either casethe public-interest theory would suggest that states with rela-tively high initial loan interest rates should deregulate earlierThe private-interest theory however does not have a clearimplication high rents could lead the beneciaries to ght harderto maintain them (eg Stigler [1971]) or the deadweight costsassociated with these high rents could make it more difficult forthe beneciaries to maintain the restrictions (eg Becker [1983])13
Bank Stability Geographic diversication through branch-ing could mitigate instability problems that were importantduring much of our sample period thereby improving welfareInstability also may reduce the incentives of banks to lobby tomaintain protections because unstable banks are less likely tosurvive to reap the benets of the restrictions politicians also mayanticipate that such banks are less likely to be able to providefuture contributions and support [Gunther 1994 1996]14 Both theprivate-interest and public-interest theories thus suggest thatderegulation is most likely in states where banking instability isgreatest
12 On the other hand local banking monopolies created by branchingrestrictions could strengthen relationshipsbetween banks and small and medium-sized rms and increase the availability of credit to these rms [Petersen andRajan 1994] Also some have argued that small business lending declines whenlarge banks take over small banks (eg Berger Kashyap Scalise [1995])
13 High loan interest rates also could reect compensation for higher risksand fewer opportunities for banks to diversify If state-specic risks explaindifferences in average loans rates then neither theory would have any predictionsfor the sign of this effect Following branching deregulation however loan interestrates tend to fall but bank risk proles do not appear to change [Jayaratne andStrahan 1998 Kroszner and Strahan 1998] This suggests that high loan rates inantibranching states are not simply reecting state-specic risks
14 Also deregulation can arise as a response to banking instability under atheory of ignorant or misinformed voters in which a banking crisis acts as aneducational device to make the previously ignorant public aware of the costs of theantibranching policy (see Kane [1996])
WHAT DRIVES DEREGULATION 1445
Political-Institutional Factors Republicans are typically per-ceived as more likely to favor deregulation than Democrats so thepolitical-institutional theories suggest that states that are con-trolled by Democrats deregulate later than those controlled byRepublicans15 In addition we investigate whether reform be-comes more likely when one party controls the legislature and thegovernorship Note that these political effects must be interpretedwith caution since the views of the politicians may simply reectthe economic interests of the constituents in the state (seePeltzman [1984])
B Variable Denitions and Data Sources
Our main proxy for the strength of the small banks is thefraction of banking assets in the state in lsquolsquosmallrsquorsquo banks We denesmall banks as banks with assets below the median size in eachstate By allowing the denition of small to vary across states wetake into account cross-state heterogeneity in bank sizes16 An-nual data on bank size are from the fourth quarter Reports ofIncome and Conditions (lsquolsquoCall Reportsrsquorsquo) from the Federal ReserveBoard We also include the capital-to-asset ratio of small banksrelative to large as a measure of their relative nancial healthSpecically this variable is the asset-weighted average capital-asset ratio for small banks minus the asset-weighted averagecapital-asset ratio for large banks Annual data on bank capitalare from the Call Reports
To measure the effects of the rival insurance industry we rstconstruct an indicator variable that is one if the state permitsbanks to sell insurance For each state we then measure the sizeof the insurance sector (total value added in the state) relative tothe sum of the banking plus insurance sectors We will examine
15 For more detail on the importance of legislative structures party politicsand ideology see Poole and Rosenthal [1997] Kahn and Matsusaka [1997] Irwinand Kroszner [1999] and Kroszner and Stratmann [1999]
16 We also considered a variety of other denitions of relative size but therelative size results reported below are not sensitive to which denition we useFirst we used a xed measure of small bank that did not vary across statesdening small banks as those with assets below $100 million in 1994 dollarsSecond we applied the state-varying and xed denitions at the level of lsquolsquobankingorganizationsrsquorsquo (which include multibank holding companies) instead of at the levelof banks Third we calculated gini coefficients of bank size inequality and bankconcentration indices as alternative proxies for small versus large bank power inthe state Finally we also considered banks in lsquolsquoruralrsquorsquo areas (that is not located in aBureau of the Census lsquolsquostandard metropolitan statistical arearsquorsquo) as small since thevalue of the restrictions might be greatest in protecting banks outside of cities fromentry by the city banks All of the alternative denitions are highly correlated andyield the same results on the importance of intraindustry differences
QUARTERLY JOURNAL OF ECONOMICS1446
the effect separately for states that permit banks to sell insuranceand those that do not Data on value-added by industry are fromU S Commerce Department Bureau of Economic Analysis Sur-vey of Current Business [August 1994]
To measure the relative importance of small bank-dependentborrowers we include the proportion of all establishments operat-ing in the state with fewer than twenty employees These data arecompiled by the Bureau of the Census17 As a measure of priceswe use the difference between the average interest rate on loans inthe state and the prevailing federal funds rate where the loanrate equals the ratio of total interest income on all domestic loansdivided by total domestic loans held by banks operating in thestate18 These data are from the end-of-year Call Reports butbecome available beginning only in 1976 Our proxy for bankinstability is the failure rate of banks measured as the percent oftotal state banking assets in failed institutions from the FederalReserve Boardrsquos National Information Center database19
We include two political variables to adjust for any indepen-dent inuence of party politics First we measure the degree ofparty control of the state government by the fraction of the threebodies of the state government (the assembly senate and gover-norship) controlled by Democrats This variable for example isone-third if the Democrats have a majority in the assembly andthe Republicans have a majority in the senate and hold thegovernorship Second we include an indicator variable equal toone if the same party controls the governorrsquos office and hasmajorities in both chambers of the state legislature20
17 We collected the establishment data by the state for three cross sections(1976 1982 and 1987) and interpolated the data in the intermediate years SeeState and Metropolitan Data Book 1982 1986 and 1991
18 A measure of how individual customers would be affected is interest ratesthey receive on deposits We calculated the average deposit interest rate in thestate from the end-of-year Call Reports and did not nd any relationship betweenthis variable and the timing of deregulation in our analysis below Since Jayaratneand Strahan [1998] nd that branching deregulation has no effect on deposit ratesour nding of no relationship between timing and rates result could be due toeither individual bank customers anticipating little pricing effect and not lobbyingor an inability of individual depositors to form an effective lobby
19 As alternative measures of distress we used (1) the bank failure ratemeasured as the number of failed institutions in the state relative to the totalnumber of banks (2) the overall capital-to-asset ratio of all banks in the state and(3) the growth in state personal income Our results are not sensitive to the choiceof distress proxy
20 We also estimated the models with three indicator variables to reect theparty control of the assembly senate and governorship separately The coefficientestimates on these variables are qualitatively similar to those we present below(that is Democratic control tends to slow deregulation) and the results on theinterest group variables are unaffected by the specication of the political
WHAT DRIVES DEREGULATION 1447
IV METHODS AND RESULTS
In this section we rst estimate how the variables describedabove inuence the timing of state-level branching deregulationusing a hazard model Next we discuss the robustness of theresults including a test of whether the ex post impact of deregula-tion on the different interest groups is consistent with ourinterpretation of the coefficients We then investigate federalinterstate branching deregulation with a probit model to explorewhether the factors that affect the timing of intrastate deregula-tion also inuence voting in the House of Representatives oninterstate deregulation
A Hazard Model
We consider the period from the beginning of our sample(1970) until deregulation as the lsquolsquoduration of regulationrsquorsquo or thelsquolsquotime until deregulationrsquorsquo The hazard rate h(t) is the likelihoodthat a state deregulates at time t given that the state has not yetderegulated [Kalbeisch and Prentice 1980 Kiefer 1988 Greene1997] To model the duration of regulation we must decide whatstructure if any to impose on the hazard function The Kaplan-Meier product-limit estimator provides a simple nonparametricway to estimate the shape of the hazard function over time[Greene 1997]21 Figure II graphs this estimate for our data andshows that the hazard function is relatively at in the early yearsand then grows steeper in the later years
A duration model that can approximate this shape is theWeibull proportional hazards model The hazard rate functiontakes the form
h[t x(t) b] 5 h0(t) exp[x(t)8b]
where x(t) is a vector of time-varying covariates b is a vector ofunknown parameters to be estimated and the baseline hazard
variables Note also that Nebraska has a nonpartisan unicameral legislature Weassume that the party of the Governor controls the state government for Nebraskaand set the unied government indicator to one As another alternative we triedincluding a variable that is one if party control of the legislature or governorshipchanges but it did not have an economically or statistically signicant effect anddid not affect any of the other results
21 Denote each of the K years in our sample period Tk and order them suchthat T1 T2 and so on Let nk be the number of states that have not yet deregulatedby Tk and dk be the number of states that deregulate in year Tk The Kaplan-Meierestimate of the hazard rate in each year is dknk In other words the hazard rate isthe ratio of the number of states that actually deregulate in each year to thenumber of states that have not yet deregulated in that year
QUARTERLY JOURNAL OF ECONOMICS1448
rate h0(t) is ptp 2 1 with shape parameter p that will be estimatedfrom the data When p 1 this model displays a monotonicallyincreasing hazard rate To estimate the parameters we maximizethe following log-likelihood function [Kiefer 1988]
L(b) 5 oi 5 1
N
di ln h[ti x(ti) b] 2 oi 5 1
N
e 0
tih[u x(u) b] du
where N indexes the number of states at risk at the beginning ofthe sample period di is an indicator equal to one for the statesthat deregulated by the end of the sample period and zero for thestates that do not deregulate (ie the censored observations) andti is the time of deregulation for the ith state This formulationassumes that the time until deregulation is independent acrossstates but the observations of the same state over time are notassumed to be independent In addition we use a robust estima-tion procedure to calculate the standard errors [Lin and Wei1989]
We exclude states that deregulated before 1970 from theanalysis so N equals 39 states During the sample period 36 ofthe 39 states deregulate We have three censored observations forwhich di is zero (Arkansas Iowa and Minnesota) but our results
FIGURE IIKaplan-Meier Estimate of Hazard Rate for Branching Deregulation
WHAT DRIVES DEREGULATION 1449
are not sensitive to the inclusion of these censored observationsSince we observe each state in each year up to and including theyear of deregulation we have a total of 637 observations Table IIreports summary statistics and correlation coefficients for theexplanatory variables we use in our analysis
We use the Weibull model rather than a model that does notimpose any structure on the baseline hazard rate so that we cancalculate the change in the expected time to deregulation for agiven change in the levels of the covariates22 In the Weibullmodel we can invert the hazard function and map it into the timedomain Rewriting the Weibull model in this way the log of thetime to deregulation T is a linear function of the political economyfactors and an error term ln (T ) 5 x8b 1 e23 Because we assumethat the baseline hazard rate is Weibull with a shape characterizedby p the new coefficients will be scaled by p that is b 5 2 bp
The b coefficients represent the percentage change in thetime to deregulation for a one-unit change in the covariates Apositive coefficient implies that an increase in the variableincreases the expected time until deregulation To gauge theeconomic importance of the effects we will multiply the b by thestandard deviation of the explanatory variables and then evaluatehow much the changes in the variables raise or lower the expectedtime to deregulation24 In Table III we report the b coefficients
22 The Cox proportional hazards model does not impose any structure on thebaseline hazard rate h0(t) and takes the following form
h[t x(t) b] 5 h0(t) exp [x(t)8b]
Common factors that affect the probability of deregulation in the same way for allstates therefore do not inuence the estimated coefficients The cost of the Coxmodel is that because there is no structure on the baseline we can calculatechanges in only relative hazard rates associated with changes in the covariatesThe Weibull model however provides sufficient structure so that we can translateour estimates into a log expected time metric that allows us to calculate the changein the expected time to deregulation for a given change in the covariates SeeKiefer [1988] Greene [1997] and Stata [1997] When we compare the coefficientsestimated by the Cox and Weibull models on our data they are very close (seeKroszner and Strahan [1998]) so imposing a Weibull structure does not appear todistort our estimates
23 The error term e is independent of x and has an extreme value distributionscaled by 1p (see Kiefer [1988] Greene [1997] and Stata [1997] for more details onthis log expected time parameterization of the Weibull model)
24 The transformation to the log expected time domain is only strictly correctwhen the covariates are constant over time Our interpretation of the coefficients ispossible in the time-varying model if we assume that a one-standard-deviationdifference would hold over the entire sample period that is the deregulationwould occur so many years earlier or later if the variable in question were higherby one standard deviation from 1970 to 1992 Note that the mean number of yearsto deregulation in the sample is 158
QUARTERLY JOURNAL OF ECONOMICS1450
TABLE IISUMMARY STATISTICS FOR VARIABLES IN HAZARD MODEL 1970ndash1992
Panel A Univariate statistics MeanStandarddeviation Minimum Maximum
(1) (2) (3) (4)Small bank asset share of all
banking assets in state 011 005 001 023Relative size of insurance to
banking plus insurance inthe state 047 009 024 080
Indicator is 1 if banks can sellinsurance in the state 020 040 0 1
Small rm share of the numberof rms in the state 088 002 078 095
Capital ratio of small banksrelative to large in the state 002 001 2 001 009
Share of state government con-trolled by Democrats 065 036 0 1
Indicator is 1 if state controlledby one party 054 050 0 1
Average yield on bank loans inthe state minus Fed fundsrate 002 002 2 004 008
Panel B CorrelationsSmallbank
Relativesize of
insuranceInsuranceindicator
Smallrm
Smallbank
capitalDem
controlSingleparty
(1) (2) (3) (4) (5) (6) (7)Small bank asset share
of all banking assetsin state 1
Relative size of insur-ance to banking plusinsurance in the state 2 034 1
Indicator is 1 if bankscan sell insurance inthe state 020 2 012 1
Small rm share of thenumber of rms inthe state 030 2 011 005 1
Capital ratio of smallbanks relative tolarge in the state 2 066 020 2 010 2 006 1
Share of state govern-ment controlled byDemocrats 2 006 001 2 018 2 018 005 1
Indicator is 1 if statecontrolled by oneparty 007 012 001 2 006 2 006 030 1
Average yield on bankloans in the stateminus Fed funds rate 2 001 2 015 003 022 2 010 2 003 2 006
The number of observations is 637 except for the average interest rate on bank loans which becomesavailable only in 1976 so the number of observations is 408
TABLE IIIHAZARD MODEL OF POLITICAL ECONOMY FACTORS AFFECTING THE TIMING OF STATE
BRANCHING DEREGULATION 1970ndash1992
(1) (2) (3) (4) (5) (6)Small bank asset share of
all banking assets instate
648(122)
519(137)
632(135)
516(118)
488(133)
445(088)
Capital ratio of smallbanks relative to largein the state
1325(348)
921(319)
1301(362)
1067(346)
1144(375)
906(381)
Relative size of insurancein states where banksmay sell insurance 0otherwise
324(161)
259(149)
315(165)
285(154)
272(159)
055(134)
Indicator is 1 if banksmay sell insurance inthe state
045(010)
037(011)
045(010)
038(008)
036(009)
001(012)
Relative size of insurancein states where banksmay not sell insurance0 otherwise
2 093(043)
2 090(033)
2 102(051)
2 093(048)
2 096(053)
2 004(039)
Small rm share of thenumber of rms in thestate
2 972(211)
2 614(215)
2 952(211)
2 943(230)
2 952(270)
2 1509(276)
Share of state govern-ment controlled byDemocrats
031(012)
023(011)
030(012)
027(011)
026(011)
011(013)
Indicator is 1 if state con-trolled by one party
2 004(010)
2 004(007)
2 004(007)
2 001(009)
2 002(010)
017(009)
Average yield on bankloans in the stateminus Fed funds rate mdash
023(319) mdash mdash mdash mdash
Bank failure rate mdash mdash092
(111) mdash mdash mdash
Indicator is 1 if state hasunit banking law mdash mdash mdash
018(010)
019(010)
030(009)
Indicator is 1 if statechanges bank insur-ance powers mdash mdash mdash mdash
2 008(012)
2 026(013)
Includes regionalindicators No No No No No Yes
N 637 408 637 637 637 637Log likelihood 2 374 900 2 360 2 212 2 191 1192p-value of x 2 for
regression 001 001 001 001 001 001
The hazard model is Weibull where the dependent variable is the log expected time to branchingderegulation ln (T) 5 bx 1 eAll variables are measured for each state in each year Small bank asset share isthe percent of banking assets in the state held by banks below the median size of banks in each state in eachyear Relative capital ratio is the capital to assets ratio of small banks minus that of large banks Size ofinsurance relative to banking plus insurance in the state is measured as gross state product from insurancedivided by gross state product from insurance plus banking Insurance indicator is 1 if state law permits banksto enter the insurance business Small rm share is the percent of all establishments in the state that havefewer than twenty employees Party control variable is the share of the three bodies of state governmentcontrolled by Democrats One party control indicator is 1 if the same party controlrsquos the governorship and hasmajorities in both chambers of the state legislature Average yield on bank loans equals total interest incomeon all domestic loans made by banks in the state divided by total loans Bank failure rate is the ratio of assetsin failed banks during a given state and year divided by total bank assets in that state and year Unit bankingindicator equals 1 for states with unit banking restrictions The change in insurance powers indicator equalsone if the state changed the law to permit banks to sell insurance during the sample period We divide thecountry into four regions and include regional indicators in column (5) Robust standard errors are inparentheses and denote statistically signicant at the 1 5 and 10 percent levels
QUARTERLY JOURNAL OF ECONOMICS1452
The positive and statistically signicant coefficient on smallbank share implies that a greater small bank share delaysderegulation and this effect is economically important A one-standard-deviation increase in the small bank share results in a30 percent increase in the time until deregulation or about 47years This result supports the implications of private-interesttheory concerning intraindustry rivalry and is contrary to thepublic-interest theory
The positive and statistically signicant coefficient on therelative performance of small banks in the state implies thatstates deregulate later when small banks are relatively strongnancially25 A one-standard-deviation increase in the relativecapital-to-asset ratio results in a 15 percent rise in the time untilderegulation or about 24 years This result is consistent with theprivate-interest theory but may also be consistent with a public-interest interpretation because the risk of failures (and a costlytaxpayer-funded bailout) increases as small bank nancial healthdeclines
Interindustry competition also helps explain the timing ofderegulation In states where banks can sell insurance a rela-tively large insurance sector is associated with an increase in theexpected time to deregulation26 A one-standard-deviation in-crease in the relative size of the insurance sector in those statesthat permit banks to sell insurance leads to a 22 percent increasein the time until deregulation or about 35 years27 This result isconsistent with the private-interest theory but not the public-interest theory
Deregulation occurs earlier in states where small bank-dependent rms are relatively numerous A one-standard-
25 We also allowed small bank and large bank capital-to-asset ratios to enterseparately Their coefficient estimates were nearly equal in absolute value and ofopposite signs and so the data do not reject the relative small to large bank capitalratio specication we have chosen
26 When we use the relative size of the insurance sector in the state as awhole rather than insurance relative to banking plus insurance in the state weobtain similar results When we simultaneously include both measures of therelative size of insurance the interactions of the lsquolsquoinsurance relative to bankingplus insurancersquorsquo variable continue to be statisticallysignicant but the interactionsof lsquolsquoinsurance relative to the state as a wholersquorsquo variable are not
27 We have also estimated the model with the indicator variable equal to onefor states where banks may sell insurance but without the variables measuringtheimportance of insurance relative to banking In this model the coefficient on theinsurance indicator variablemdashan estimate of the difference in the expected time toderegulation in states where banks may sell insurance relative to states wherebanks may not sell insurancemdashequals 027 and is statistically signicant at the 5percent level This suggests that branching deregulation takes about 45 yearslonger in states where banks may sell insurance
WHAT DRIVES DEREGULATION 1453
deviation increase in the share of small rms reduces the timeuntil deregulation by 18 percent or about three years This resultconcerning the interests of consumers of banking services isconsistent with both the private- and public-interest theories
Column (2) of Table III introduces the average interest rateon loans to the model as a rough proxy for the cost to bankborrowers of these regulations These data become available onlyin 1976 so our sample size shrinks to 408 observations Thecoefficient on this variable is small and not statistically signi-cant The public-interest theory would imply that the coefficientshould be negative while the implication of the economic theorydoes not have a clear sign prediction
When we add the failure rate of banks to the specication incolumn (3) its coefficient is small and statistically insignicantWe do not nd a linkage between the timing of state deregulationand statewide banking distress Given that deposit insurance isfunded federally however bank failures may raise the publicrsquosconcerns about the costs of the regulations at the national ratherthan the state level This interpretation can explain why the rateof bank failure does not affect the timing of state-level deregula-tion even though bank instability does appear to rise generallyduring the period of increasing deregulation (see Section V below)
Finally the partisan structure of the state government alsoinuences when states deregulate As expected a higher propor-tion of Democrats in the government tends to delay deregulationA one-standard-deviation rise in the share of the governmentcontrolled by Democrats slows the deregulation by about twoyears Whether the state is dominated by one party however doesnot appear to affect the timing of the deregulation
To summarize our ndings from the hazard model privateinterests appear to play an important role in the deregulatoryprocess While private interests and public interests do sometimescoincide the results on the relative share of small banks and largebanks and on the relative size of insurance where banks competeare consistent with a private-interest approach but are difficult toexplain on public-interest grounds
B Robustness Tests
Omitted Variables States begin the sample period withdifferent degrees of restrictions on bank expansion These initialconditions may proxy for omitted factors determining the political-economy equilibrium in the state that could then affect the
QUARTERLY JOURNAL OF ECONOMICS1454
subsequent timing of deregulation For instance some statesbegan the period with unit banking while other states allowedlimited branching and the formation of multibank holdingcompanies
To test whether the stringency of initial restrictions onbranching matters we rst divide the sample into states with andwithout the most extreme form of branching regulation unitbanking28 Column (4) of Table III includes an indicator variableequal to one if the state began the period with unit bankingrestrictions The coefficient on this variable is positive andstatistically signicant but the other results remain virtuallyunchanged Next we include an indicator variable that is one ifthe state permitted multibank holding companies (MBHCs) toexist at the beginning of our sample period Thirty-four statesbegan the period allowing MBHCs The coefficient estimate forthis indicator variable (not reported) is small and not statisticallysignicant and its inclusion does not change the effects of theother variables
Eleven states expand bank powers to include the sale ofinsurance products during our sample period This could proxy foromitted factors about the relative strength of the insuranceindustry in the state In column (5) we include an indicatorvariable that is one if the state changes the law during our sampleperiod to permit banks to sell insurance business and zerootherwise The coefficient estimate is small and statisticallyinsignicant and does not affect the other results29
In column (6) we introduce four regional indicator variablesNorth South Midwest and West30 The coefficients on small bankshare the relative capital-to-assets ratio and small rm sharechange little Although the coefficient estimate still suggests that
28 We classify states that prohibited branching but permitted banks toestablish facilities as unit banking states The unit banking indicator is one for thefollowing states CO AR FL IL IA KS MN MO MT NE ND OK TX WI WVand WY
29 We conducted a likelihood ratio test of whether the unit banking statesthe MBHC states and the states that changed their regulations of bank insurancepowers could be pooled with the other states In all three cases we could not rejectpooling of the data
30 Since intrastate deregulation generally preceded interstate deregulationand the latter typically took the form of regional interstate compacts thepossibility of participating in a regional interstate banking compact could haveinuenced the decision to deregulate intrastate branching Our denitions of theregions are region 1 (South) contains AL AR DC FL GA KY LA MS NC OKSC TN TX and VA region 2 (Northeast) contains CT MA MD ME NH NJ NYPA RI VT and WV region 3 (Midwest) contains IA IL IN KS MI MN MO NEND OH SD and WI region 4 (West) contains the other states
WHAT DRIVES DEREGULATION 1455
in states where banks can sell insurance a larger relative size ofthe insurance sector delays deregulation the effect is no longerstatistically signicant The coefficient on the indicator variablefor whether a state expands bank powers to permit insurancesales during the sample period however now becomes statisti-cally signicantly negative This implies that deregulation occursroughly 41 years earlier in states where the insurance lobby maybe relatively weak While the inclusion of the regional indicatorsdoes not affect the other results the results on insurance doappear to be sensitive to the specication of the hazard model31
Estimation Technique To check the robustness of using ahazard model and to investigate whether the time-series or thecross-sectional variation in each of the factors accounts for thehazard results we estimate two simple linear probability models(neither are reported) First we run a pooled time-series cross-section OLS regression using the 637 observations from thehazard model The dependent variable equals one in the year ofderegulation and zero otherwise We include the same set ofregressors that were used in the hazard model and add state-levelxed effects This specication thus removes the cross-statevariation from the independent variables and focuses on theirmovements over time In this model the signs of the coefficientson the interest group factors are consistent with the results fromthe hazard model and the coefficients for the small bank relativecapital ratio and for the relative size of insurance where states cansell insurance are statistically signicant
In the second model we eliminate the time-series variationby averaging each of the explanatory variables over time We thushave one observation per state that had branching restrictions asof 1970 (N 5 39) The dependent variable equals the inverse of thenumber of years until deregulation which represents an estimateof the statersquos probability of deregulating in a year under theassumption that this probability is constant over time and eachyear provides an independent observation32 Although we haveonly 39 observations the coefficients on the small bank marketshare small bank relative capital ratio and small rm share
31 As further robustness check we also tried a population density variablewhich Abrams and Settle [1993] found relevant for regulatory change during the1920s and 1930s but its coefficient was not statistically signicant and did notaffect the other results
32 For instance the dependent variable would take the value of 01 for astate that deregulated in 1980 For the three states that did not deregulate thedependent variable equals 0
QUARTERLY JOURNAL OF ECONOMICS1456
variables are statistically signicant The signs of the effects ofthe insurance variables are the same as those reported in thehazard model but not statistically signicant
The hazard model and linear probability models thus producevery similar results The estimates from the latter suggest thatthe effects for the small bank market share and small rm sharevariables are driven primarily by cross-state variation the effectof the insurance variable is driven primarily by time-seriesvariation and the small bank capital effect is driven by bothcross-sectional and time-series variation
Ex Post Consequences of Deregulation Finally to check theplausibility of our interpretation of the hazard results we con-sider whether the ex post consequences of deregulation areconsistent with the ex ante lobbying positions we attribute to eachinterest group (see Kroszner and Strahan [1998] for details) Wend that small banks lose market share following deregulationand in states where banks can enter the insurance business theinsurance sector shrinks relative to the banking sector followingderegulation Borrowers also benet from deregulation throughlower average interest rates on loans These ndings support theprivate-interest interpretation of the coefficients in the hazardmodel groups that will benet lobby to speed deregulation andthose that will be harmed lobby to slow it
C Voting on Interstate Branching Deregulation in the U SHouse of Representatives
We now examine whether the forces driving intrastate branch-ing deregulation also drive interstate deregulation Financialservices interests are active contributors and lobbyists in Washing-ton Their political action committees constitute the largest groupof contributors to legislators providing nearly 20 percent of totalcongressional campaign contributions [Makinson 1992] and muchof their lobbying effort involves competition among rival interestswithin nancial services (see Kroszner and Stratmann [1998])
As noted above after virtually all states adopted intra- andinterstate branching deregulation the 1994 Riegle-Neal Act re-pealed the 1927 McFadden Act to eliminate all barriers tointerstate banking and branching by 1997 The key votes concern-ing the Riegle-Neal Act were either voice votes or extremelylopsided so we could not estimate a voting model from them Anumber of bills and amendments related to interstate branching
WHAT DRIVES DEREGULATION 1457
had been debated in Congress during the years prior to thepassage of the Riegle-Neal Act but a search of the weekly BNABanking Reporter and the Congressional Record produced onlyone roll-call vote related to interstate branching that was notlopsided This vote occurred in the House on November 14 1991and concerned an amendment sponsored by Wylie (R-OH) andNeal (D-NC) to introduce interstate banking and branchingderegulation to a nancial services reform package33 Althoughthe amendment passed by 210 to 208 the bill to which it wasattached subsequently was defeated
To check for the inuence of the factors we considered in thestate-level reforms we examine both the sponsorship of interstatebanking legislation and voting on the amendment The sponsorsof the Wylie-Neal amendment are from states with low small bankshares 004 in Ohio (Wylie) and 002 in North Carolina (Neal) Incontrast the sample mean in 1991 is 008 (median 5 007)Michigan home state of the Senatersquos sponsor of the 1994 Riegle-Neal Act also had relatively low small bank strength (small bankshare of 005)
Table IV reports estimates of a probit model where thedependent variable equals one if the legislator voted in favor ofthe amendment and zero otherwise The explanatory variablesare the same as those in the hazard model except that we use aparty affiliation indicator variable equal to one for Democrats andzero for Republicans as our proxy for political factors Thecoefficients in Table IV are the marginal effects (lsquolsquoslopesrsquorsquo) of aone-unit change of each variable on the probability that alegislator will vote for the amendment34 Note that a forcefavoring deregulation has a negative coefficient in the hazardmodel (Table III) but a positive coefficient in the probit model
Consistent with the state-level deregulation process legisla-tors are more likely to support the amendment if their states havea relatively low share of small banks As in the hazard model thefraction of small banks is the most important interest groupinuence on a legislatorrsquos voting decision A one-standard-deviation increase in small banksrsquo market share (from the mean)is associated with a decline in the probability of voting in favor of
33 The Wylie-Neal amendment also included provisions limiting certaininsurance and real estate powers of national banks [Congressional RecordNovember 14 1991 pp 10239ndash10242]
34 Since we have multiple legislators from each state we adjust the standarderrors to correct for the lack of independence among observations clustered in thesame state
QUARTERLY JOURNAL OF ECONOMICS1458
TAB
LE
IVM
AR
GIN
AL
EF
FE
CT
SF
RO
MA
PR
OB
ITM
OD
EL
OF
TH
EIN
FL
UE
NC
EO
FP
OL
ITIC
AL
EC
ON
OM
YF
AC
TO
RS
ON
TH
EH
OU
SE
VO
TE
ON
TH
EW
YL
IE-N
EA
L
AM
EN
DM
EN
TT
OP
ER
MIT
INT
ER
ST
AT
EB
AN
KB
RA
NC
HIN
G
NO
VE
MB
ER
14
1991
(1)
(2)
(3)
(4)
(5)
(6)
Smal
lban
kas
set
shar
eof
allb
anki
ng
asse
tsin
stat
e2
417
(12
4)2
454
(12
2)2
407
(12
0)2
443
(14
7)2
450
(14
9)2
490
(14
7)C
apit
alra
tio
ofsm
allb
anks
rela
tive
tola
rge
inth
est
ate
180
(37
5)2
68(3
42)
256
(37
5)1
69(3
78)
153
(38
0)2
45(4
25)
Rel
ativ
esi
zeof
insu
ranc
ein
stat
esw
here
bank
sm
ayse
llin
sura
nce
0ot
herw
ise
22
34
(10
2)2
249
(1
14)
22
33
(10
1)2
249
(1
04)
22
51
(10
2)2
154
(11
1)In
dica
tor
is1
ifba
nks
may
sell
insu
ranc
ein
the
stat
e0
79
(0
06)
074
(00
6)0
69
(0
06)
070
(00
6)0
70
(0
06)
067
(00
7)R
elat
ive
size
ofin
sura
nce
inst
ates
whe
reba
nks
may
not
sell
insu
ranc
e0
oth
erw
ise
180
(06
6)2
42
(0
72)
154
(0
67)
173
(06
6)1
72
(0
67)
203
(0
91)
Smal
lrm
shar
eof
the
num
ber
ofr
ms
inth
est
ate
21
23(0
92)
20
83(0
89)
21
62(1
07)
21
32(0
95)
21
30(1
00)
20
94(1
16)
Indi
cato
ris
1if
Rep
rese
ntat
ive
isa
Dem
ocra
t2
035
(00
7)2
035
(00
7)2
036
(00
7)2
035
(00
7)2
035
(00
7)2
036
(00
7)A
vera
geyi
eld
onba
nklo
ans
inth
est
ate
min
usF
edfu
nds
rate
mdash10
01
(51
5)mdash
mdashmdash
mdash
Ave
rage
fail
ure
rate
198
5to
1991
mdashmdash
228
(14
1)mdash
mdashmdash
Indi
cato
ris
1if
stat
eh
asun
itba
nkin
gla
wmdash
mdashmdash
004
(00
7)0
04(0
07)
20
01(0
08)
Indi
cato
ris
1if
stat
ech
ange
sba
nk
insu
ran
cepo
wer
smdash
mdashmdash
mdash2
001
(00
7)2
003
(00
7)In
clud
esre
gion
alin
dica
tors
N
oN
oN
oN
oN
oYe
sP
seud
o-R
20
170
180
170
170
170
18L
ogli
keli
hood
224
024
223
826
223
951
224
013
224
010
223
808
p-va
lue
ofx2
for
regr
essi
on
001
0
01
001
0
01
001
0
01
Th
ede
pen
den
tvar
iabl
eis
one
ifth
ele
gisl
ator
vote
sfo
rth
eam
endm
enta
nd
zero
ifag
ains
tT
he
repo
rted
coef
fici
ents
are
the
effe
cts
ofa
uni
tch
ange
ofth
ein
depe
nde
ntva
riab
le(f
rom
the
mea
n)on
the
prob
abili
tyof
voti
ng
infa
vor
ofth
eam
endm
ent
All
vari
able
sar
em
easu
red
byst
ate
N5
418
the
num
ber
ofR
epre
sent
ativ
esvo
tin
gon
the
amen
dmen
tS
mal
lban
kas
sets
har
eis
the
perc
ent
ofba
nki
ng
asse
tsin
the
stat
eh
eld
byba
nks
belo
wth
em
edia
nsi
zeof
ban
kin
each
stat
ein
each
year
Rel
ativ
eca
pita
lrat
iois
the
capi
talt
oas
sets
rati
oof
smal
lban
ksm
inu
sth
atof
larg
eba
nks
Siz
eof
insu
ran
cere
lati
veto
bank
ing
plu
sin
sura
nce
inth
est
ate
ism
easu
red
asgr
oss
stat
epr
oduc
tfro
min
sura
nce
divi
ded
bygr
oss
stat
epr
odu
ctfr
omin
sura
nce
plu
sba
nkin
gIn
dica
tor
ison
eif
stat
ela
wpe
rmit
sba
nks
toen
ter
the
insu
ran
cebu
sin
ess
Sm
all
rmsh
are
isth
epe
rcen
tofa
lles
tabl
ishm
ents
inth
est
ate
that
hav
efe
wer
than
twen
tyem
ploy
ees
Par
tyin
dica
tor
ison
eif
the
Rep
rese
nta
tive
isa
Dem
ocra
tA
vera
geyi
eld
onba
nklo
ans
equa
lsto
tali
nter
est
inco
me
onal
ldom
esti
clo
ans
mad
eby
ban
ksin
the
stat
edi
vide
dby
tota
lloa
ns
The
aver
age
bank
failu
rera
teis
the
rati
oof
asse
tsin
faile
dba
nks
duri
ng
agi
ven
stat
edi
vide
dby
tota
lban
kas
sets
inth
atst
ate
aver
aged
over
the
year
s19
85to
1991
Un
itba
nki
ng
indi
cato
req
ual
son
efo
rst
ates
wit
hun
itba
nkin
gre
stri
ctio
nsT
hech
ange
inin
sura
nce
pow
ers
indi
cato
req
ual
son
eif
the
stat
ech
ange
dth
ela
wto
perm
itba
nks
tose
llin
sura
nce
duri
ngth
esa
mpl
epe
riod
We
divi
deth
eco
unt
ryin
tofo
urre
gion
san
din
clu
dere
gion
alin
dica
tors
inco
lum
n(5
)R
obu
stst
anda
rder
rors
are
inpa
ren
thes
es
an
d
den
ote
stat
isti
cally
sign
ica
nta
tth
e1
5an
d10
perc
entl
evel
s
WHAT DRIVES DEREGULATION 1459
branching of approximately 17 percent The relative capital-to-assets ratio of small and large banks however does not have astatistically signicant effect in the probit model
The impact of rival interests outside of banking is alsoconsistent with intrastate deregulation results Where banks cansell insurance legislators from states with larger insurancesectors relative to banking are less likely to vote for the amend-ment35 A one-standard-deviation increase in the relative size ofinsurance in those states which permit banks to sell insurancedecreases the probability that a legislator will favor the amend-ment by about 13 percent Turning to consumer interests thecoefficient on small rm share is not statistically signicant butthe coefficient on interest rates in the state is A one-standard-deviation increase in the average interest rate on loans raises theprobability that a legislator will support the amendment byroughly 8 percent The coefficient estimate for the bankingdistress variable measured as the average bank failure rate from1985 to 1991 is small and statistically signicant Overall theprobit analysis of the vote on national branching deregulationsupports the private-interest theory of deregulation and providesa consistency check that the importance of interests operating onthe state legislatures are very similar to those operating at thefederal level
V WHY DID BRANCHING DEREGULATION BEGIN DURING THE
1970S SHOCKS TO THE EQUILIBRIUM SUPPORTING BANK
BRANCHING RESTRICTIONS
A complete explanation of regulatory exit should account forwhy deregulation begins in the 1970s as well as the timing of thestate-by-state reforms [Kane 1996] In this section we describebroad technological legal and economic shocks that altered thepolitical-economy equilibrium which had kept antibranching regu-lations little changed for at least 30 years
Beginning in the 1970s three innovations reduced the valueto the protected banks of local geographic monopolies (see TableV) First automatic teller machines (ATMs) helped to erode thegeographic ties between customers and banks Second checkablemoney market mutual funds and the Merrill Lynch Cash Manage-ment Account demonstrated that banking by mail and telephone
35 The positive and statistically signicant estimates on the other insurancevariables may reect insurance industry support for the amendmentrsquos provisionslimiting banksrsquo insurance powers
QUARTERLY JOURNAL OF ECONOMICS1460
provided a convenient alternative to local banks36 Third techno-logical innovation and deregulation reduced transportation andcommunication costs particularly since the 1970s thereby lower-ing the costs for customers to use distant banks By increasing theelasticity of deposits supplied to banks these innovations reducedthe value of geographical restrictions to their traditional benecia-ries and thereby reduced their incentive to ght to maintain them[Peltzman 1976]
On the lending side increasing sophistication of credit-scoring techniques following innovations in information process-ing nancial theory and the development of large credit databases began to diminish the value of knowledge that localbankers had about the risks of borrowers in the community As a
36 Regulation Q which limited the interest rates that banks could pay ondeposits may have helped to drive depositors away from banks when the gapbetween market rates and deposit ceilings grew during the 1970s The eliminationof interest rate ceilings on large denomination certicates of deposit during the1970s appears to have hurt smaller and retail-oriented banks relative to largerwholesale banks [James 1983]
TABLE VBROAD TRENDS IN COMMERCIAL BANKING 1950ndash1995
YearNumberof ATMs
Domesticbank
deposits(billions)
Moneymarketmutual
fund(billions)
Percent ofdeposits 1
moneyfunds heldby banks
Smallbanksrsquo
percent ofbankingassets
Averagenumber of
bankfailures
(1) (2) (3) (4) (5) (6)1950 0 $ 154 $ 0 100 NA 41955 0 191 0 100 NA 31960 0 228 0 100 24 21965 0 330 0 100 20 41970 0 479 0 100 18 61975 9750 775 4 99 18 61980 18500 1182 76 94 17 101985 61117 1787 242 88 14 601990 80156 2339 493 83 11 1791995 122706 2552 745 77 8 61
Column (1) ATM gures are from Bank Network News The EFT Network Data Book (New YorkFaulkner and Gray Inc) The 1975 gure was unavailable 9750 is the number of ATMs in 1978 the rst yearfor which complete data are available Columns (2)ndash(4) Banksrsquo domestic deposits are from the Reports ofIncome and Condition money market mutual funds are from the Flow of Funds Data on all bank depositsforeign plus domestic are only available beginning in 1970 The trend in banksrsquo share (column (4)) is the sameusing total deposits instead of domestic deposits Column (5) Percent of banking assets held by small bankswhere a small bank is dened as a commercial bank with less than $100 million in assets in 1994 dollarsThese data are based on the Reports of Income and Condition Data on small banks are not available before1960 Column (6) Five-year average number of bank failures where the nal year is indicated in the rstcolumn These data are from FDIC Annual Report and the Quarterly Banking Prole
WHAT DRIVES DEREGULATION 1461
result of these innovations a national market developed forresidential mortgages credit card receivables have been securi-tized and bank lending to small business now relies less on thejudgment of loan officers and more on standardized scoringmodels
These changes have increased the potential protability forlarge banks to enter what had been the core of small bankactivities Large banksrsquo incentive to increase their lobbying pres-sure to be able to expand into these markets has thus beenincreasing over time In fact small banksrsquo market share began todecline even prior to the branching deregulation (Table V) As thevalue of a local banking relationships declined small rms thatwere the main borrowers from the small banks also probablybecame more likely to favor the entry of large banks into localmarkets With the deadweight costs of preventing large bankentry rising the private-interest theory predicts that small localbanks would become less likely able to maintain the branchingrestrictions [Becker 1983] Deregulation that reduces deadweightcosts of regulation is also consistent with the public-interesttheory
Kane [1996] argues that another major shock to the oldequilibrium was the increasing public awareness of the costlinessof having government-insured but (geographically) undiversiednancial institutions In the late 1970s the failure rate of banksbegan to rise (Table V) In the 1980s the Savings and Loan crisisand taxpayer bailout further heightened public awareness aboutthe costs of restrictions that make depository institutions morelikely to require infusions of taxpayer funds The failures thusmay have heightened public support for branching deregulation(see also Abrams and Settle [1993]) As noted above we did notnd any evidence that banking failures or distress in a stateaffected the speed with which the state deregulated but distressstill may be a common factor affecting all states
These technological economic and legal shocks generatedconditions that changed the long-standing balance favoring theantibranching forces The marginal value of lobbying to repealbranching restrictions increased just as the relative value to thesmall banks of maintaining branching restrictions was decliningThese nationwide shocks are common factors across states andthey are consistent with the positive duration dependence associ-ated with state-level deregulation that we demonstrated in FigureII In addition a broad change in lsquolsquoideologyrsquorsquo against government
QUARTERLY JOURNAL OF ECONOMICS1462
regulation could explain the positive duration dependence duringthe sample period
VI CONCLUSIONS
The private-interest theory of regulation can account for thepattern of bank branching deregulation of the 1970s and 1980sBeneciaries of branching regulation had supported a coalitionfavoring geographical restrictions despite their costs to consum-ers of nancial services Innovations that began in the 1970saltered the value of the restrictions to the affected parties and theresulting competition among interest groups can explain thesubsequent deregulation While some of our results are alsoconsistent with the public interest theorymdashfor example deregula-tion occurs earlier when small banks are in a relatively weaknancial positionmdashother results particularly evidence on theimportance of rivalries between small and large banks andbetween banking and insurance are difficult to explain with thepublic-interest approach We also nd that political-institutionalfactors affect deregulation although these variables may act asproxies for unmeasured economic interests The greater success ofthe private-interest theory to explain deregulation here than inprevious work may be due to the richer cross-sectional andtime-series variation in branching deregulation that can beexploited with a hazard model Future empirical studies ofendogenous deregulation thus may be most fruitful where changehas occurred across states over time such as in franchisinginsurance and public utilities (eg Joskow [1996])
Technological and nancial innovations will continue to erodethe benets to any interest group of maintaining regulatorybarriers in nancial services These forces are likely to bringabout reforms both domestically for example through legislationthat would increase bank powers (see Kroszner [1996] Krosznerand Rajan [1994 1997] and Kroszner and Stratmann [1998]) andinternationally for example through the extension of nancialservices provisions of NAFTAto reduce geographic barriers acrosscountries [Kroszner 1997]
APPENDIX EFFECTS OF DIFFERENT TYPES OF BRANCHING
DEREGULATION ON ECONOMIC GROWTH BANKING STRUCTUREAND BANK EFFICIENCY
We measure the importance of alternative forms of deregula-tion by their consequences for the economic growth and the
WHAT DRIVES DEREGULATION 1463
structure and efficiency of the banking sector across the states(see Jayaratne and Strahan [1996 1998]) To do so we run anOLS panel regression relating indicators for each type of deregula-tion to potential consequences Our dependent variables includeproxies for economic growth (growth in personal income) bankstructure (the log of the number of banking companies and thenumber of branches per banking company) and efficiency (the logof total noninterest expenses and the ratio of loan loss provisionsto total loans) The four deregulation indicators (MBHC branch-ing through merger and acquisition unrestricted statewide branch-ing and interstate banking) are equal to one in the years afterderegulation and zero before If a state deregulated before thesample period begins in 1970 the indicator is one throughout ifthe state does not deregulate then the indicator is zero through-out All regressions include state xed effects and time effects
Table VI shows the results of including the set of four
TABLE VIMULTIVARIATE REGRESSIONS OF THE EFFECTS OF DIFFERENT TYPES OF BRANCHING
DEREGULATION ON ECONOMIC GROWTH BANKING STRUCTUREAND BANK EFFICIENCY
Dependent variable
Growthin
personalincome
Log of of
bankingcompanies
Branchesper
bankingcompany
Log ofnon-
interestexpenses
Loan lossprovisions
(loans 1leases)
Merger amp acquisitionbranchingindicator
0013(0004)
2 0162(0037)
1146(0453)
2 0107(0035)
2 0007(0001)
Unrestricted state-wide branchingindicator
2 0008(0005)
0086(0059)
2 0043(0574)
0070(0037)
0005(0002)
Interstate bankingindicator
0006(0005)
2 0052(0057)
1064(0511)
0005(0030)
2 0001(0001)
Multibank holdingcompany indicator
2 0005(0006)
2 0096(0052)
2 0364(0529)
2 0095(0036)
2 0001(0001)
Dependent variablemean [median]
0086[0082]
4881[5136]
6649[5055]
13812[13803]
0009[0006]
This table contains regressions of state-level economic growth and state-level measures of the structureand efficiency of the banking industry Each dependent variable is regressed on a set of state indicatorvariables a set of year indicator variables and the set of four deregulation indicators these deregulationindicators are equal to one after deregulation and zero before If a state was always regulated the indicator iszero throughout if it was always deregulated the indicator is one throughout Sample period is 1975 to 1991and Delaware and South Dakota are dropped so N 5 833 Noninterest expenses become available only in1984 so N 5 392 The coefficient on the deregulation indicators is reported with robust standard errors inparentheses and denote statistically signicant at the 1 5 and 10 percent levels
QUARTERLY JOURNAL OF ECONOMICS1464
deregulation indicators simultaneously The branching by mergerand acquisition indicator is statistically signicant and of theexpected sign in all ve of the equations and larger in absolutevalue than any of the other coefficients on the indicators for theother types of deregulation Since there may be multicollinearityamong our indicators we also ran univariate specications whichincluded each indicator separately and the results are verysimilar These results imply that branching by merger andacquisition is the most important type of deregulation measuredby its consequences for economic growth banking structure andbank efficiency See Kroszner and Strahan [1998] for more details
UNIVERSITY OF CHICAGO GRADUATE SCHOOL OF BUSINESS AND NATIONAL BUREAU OF
ECONOMIC RESEARCH
FEDERAL RESERVE BANK OF NEW YORK RESEARCH amp MARKET ANALYSIS GROUP
REFERENCES
Abrams Burton A and Russell F Settle lsquolsquoPressure-Group Inuence and Institu-tional Change Branch-Banking Legislation during the Great DepressionrsquorsquoPublic Choice LXXVII (1993) 687ndash705
Amel Dean lsquolsquoState Laws Affecting the Geographic Expansion of CommercialBanksrsquorsquo unpublished manuscript Board of Governors of the Federal ReserveSystem 1993
Becker Gary S lsquolsquoA Theory of Competition among Pressure Groups for PoliticalInuencersquorsquo Quarterly Journal of Economics XCVIII (1983) 371ndash400
Berger Allen Anil Kashyap Joseph Scalise lsquolsquoThe Transformation of the U SBanking Industry What a Long Strange Trip Itrsquos Beenrsquorsquo Brookings Papers onEconomic Activity 2 (1995) 55ndash218
Berger Allen Anthony Saunders Joseph Scalise and Gregory Udell lsquolsquoThe Effectsof Bank Mergers and Acquisitions on Small Business Lendingrsquorsquo Journal ofFinancial Economics L (1998) 187ndash229
Brennan Geoffrey and James Buchanan lsquolsquoTowards a Tax Constitution forLeviathanrsquorsquo Journal of Public Economics VIII (1977) 255ndash273
Brickley James A and Christopher M James lsquolsquoThe Takeover Market CorporateBoard Composition and Ownership Structure The Case of Bankingrsquorsquo Journalof Law and Economics XXX (1987) 161ndash180
Cole Rebel and John Wolken lsquolsquoFinancial Services Used by Small BusinessesEvidence from the 1993 Survey of Small Business Financesrsquorsquo Federal ReserveBulletin LXXXI (1994) 629ndash676
Dixit Avinash The Making of Economic Policy A Transactions-Cost PoliticsPerspective (Cambridge MA MIT Press 1996)
Economides Nicolas R Glenn Hubbard and Darius Palia lsquolsquoThe Political Economyof Branching Restrictions and Deposit Insurancersquorsquo Journal of Law andEconomics XXIX (1996) 667ndash704
Flannery Mark lsquolsquoThe Social Costs of Unit Banking Restrictionsrsquorsquo Journal ofMonetary Economics XIII (1984) 237ndash249
Greene William H Econometric Analysis third edition (Upper Saddle River NJPrentice-Hall 1997)
Gunther Jeffery lsquolsquoRegional Capital Imbalances and the Removal of InterstateBanking Restrictionsrsquorsquo Economics Letters XLIV (1994) 439ndash442 lsquolsquoThe Erosion of Intrastate Branch Banking Restrictionsrsquorsquo unpublishedmanuscript Federal Reserve Bank of Dallas 1996
Irwin Douglas and Randall Kroszner lsquolsquoInterests Institutions and Ideology inSecuring Policy Change The Republican Conversion to Trade Liberalizationafter Smoot-Hawleyrsquorsquo Journal of Law and Economics XLII (1999) 643ndash673
WHAT DRIVES DEREGULATION 1465
James Christopher lsquolsquoAn Analysis of Intra-Industry Differences in the Effect ofRegulation The Case of Deposit Rate Ceilingsrsquorsquo Journal of Monetary Econom-ics XII (1983) 417ndash442
Jarrell Gregg lsquolsquoChange at the Exchange The Causes and Effects of DeregulationrsquorsquoJournal of Law and Economics XXVII (1984) 273ndash312
Jayaratne Jith and Philip E Strahan lsquolsquoThe Finance-Growth Nexus Evidencefrom Bank Branch Deregulationrsquorsquo Quarterly Journal of Economics CI (1996)639ndash670
Jayaratne Jith and Philip E Strahan lsquolsquoEntry Restrictions Industry Evolutionand Dynamic Efficiency Evidence from Commercial Bankingrsquorsquo Journal of Lawand Economics XLIX (1998) 239ndash274
Joskow Paul lsquolsquoComment on lsquoPower Struggles Explaining Deregulatory Reformsin Electricity Marketsrsquo rsquorsquo Brookings Papers on Economic Activity Microeconom-ics (1996) 251ndash264
Joskow Paul and Roger Noll lsquolsquoRegulation in Theory and Practice An Overviewrsquorsquoin Studies in Public Regulation Gary Fromm ed (Cambridge MA MITPress 1981) pp 1ndash65
Kahn Matthew E and John G Matsusaka lsquolsquoDemand for Environmental GoodsEvidence from Voting Patterns on California Initiativesrsquorsquo Journal of Law andEconomics XL (1997) 137ndash173
Kalbeisch J and R Prentice The Statistical Analysis of Failure Time Data (NewYork John Wiley amp Sons 1980)
Kane Edward J lsquolsquoDe Jure Interstate Banking Why Only NowrsquorsquoJournal of MoneyCredit and Banking XXVIII (1996) 141ndash161
Kiefer Nicholas lsquolsquoEconomic Duration Data and Hazard Functionsrsquorsquo Journal ofEconomic Literature XXVI (1988) 646ndash679
Kroszner Randall S lsquolsquoThe Evolution of Universal Banking and Its Regulation inTwentieth Century Americarsquorsquo in Universal Banking Financial System DesignReconsidered Anthony Saunders and Ingo Walter eds (New York IrwinProfessional Publishers 1996) pp 70ndash99 lsquolsquoThe Political Economy of Banking and Financial Regulation in the U Srsquorsquo inThe Banking and Financial Structure in the NAFTA Countries and ChileGeorge M von Furstenberg ed (Boston Kluwer Academic Publishers 1997)pp 200ndash213
Kroszner Randall S and Raghuram G Rajan lsquolsquoIs the Glass-Steagall ActJustied A Study of the U S Experience with Universal Banking before1933rsquorsquo American Economic Review LXXXIV (1994) 810ndash832
Kroszner Randall S and Raghuram G Rajan lsquolsquoOrganization Structure andCredibility Evidence from Bank Securities Activities before the Glass-Steagall Actrsquorsquo Journal of Monetary Economics XXXIX (1997) 475ndash516
Kroszner Randall S and Philip E Strahan lsquolsquoRegulatory Incentives and the ThriftCrisis Dividends Mutual-to-Stock Conversions and Financial DistressrsquorsquoJournal of Finance LI (1996) 1285ndash1320
Kroszner Randall S and Philip E Strahan lsquolsquoWhat Drives Deregulation Econom-ics and Politics of the Relaxation of Bank Branching Restrictionsrsquorsquo NBERWorking Paper No 6637 1998
Kroszner Randall S and Thomas Stratmann lsquolsquoInterest Group Competition andthe Organization of Congress Theory and Evidence from Financial ServicesPolitical Action Committeesrsquorsquo American Economic Review LXXXVIII (1998)pp 1163ndash1187
Kroszner Randall S and Thomas Stratmann lsquolsquoDoes Political Ambiguity PayCorporate Campaign Contributions and Rewards to Legislator Reputationrsquorsquounpublishedmanuscript University of Chicago Graduate School of Business1999
Lin D Y and L J Wei lsquolsquoThe Robust Inference for the Cox Proportional HazardsModel Journal of the American Statistical Association LXXXIV (1989)1074ndash1078
Macey Jonathan R and Geoffrey P Miller Banking Law and Regulation (BostonLittle Brown and Company 1992)
Makinson Larry Open Secrets The Encyclopedia of Congressional Money andPolitics (Washington DC Congressional Quarterly 1992)
Niskanen WilliamBureaucracy and Representative Government (Chicago Aldine-Atherton 1971)
QUARTERLY JOURNAL OF ECONOMICS1466
Noll Roger lsquolsquoComment on Peltzmanrsquorsquo Brookings Papers Microeconomics SpecialIssue (1989) 48ndash58
North Douglass lsquolsquoA Transactions Cost Theory of Politicsrsquorsquo Journal of TheoreticalPolitics II (1990) 355ndash367
Peltzman Sam lsquolsquoToward a More General Theory of Regulationrsquorsquo Journal of Lawand Economics XIX (1976) 109ndash148 lsquolsquoConstituent Interest and Congressional Votingrsquorsquo Journal of Law andEconomics XXVII (1984) 181ndash210 lsquolsquoThe Economic Theory of Regulation after a Decade of DeregulationrsquorsquoBrookings Papers Microeconomics Special Issue (1989) 1ndash41
Peterson Mitchell and Raghuram Rajan lsquolsquoThe Benets of Lending RelationshipsEvidence from Small Business Datarsquorsquo Journal of Finance XLIX (1994) 3ndash38
Poole Keith and Howard Rosenthal Congress A Political-Economic History ofRoll Call Voting (Oxford Oxford University Press 1997)
Stata Stata Reference Manual Release 5 (College Station TX Stata Press 1997)Stigler George J lsquolsquoThe Theory of Economic Regulationrsquorsquo Bell Journal of Economics
and Management Science II (1971) 3ndash21Stigler George J Chicago Studies in the Political Economy (Chicago University
of Chicago Press 1988)Strahan Philip E and James Weston lsquolsquoSmall Business Lending and the Changing
Structure of the Banking Industryrsquorsquo Journal of Banking and Finance XXII(1998) 821ndash845
Sylla Richard John Legler and John Wallis lsquolsquoBanks and State Public Finance inthe New Republic The United States 1790ndash1860rsquorsquo Journal of EconomicHistory XLVII (1987) 391ndash403
White Eugene The Regulation and Reform of the American Banking System1900ndash1929 (Princeton Princeton University Press 1983)
Winston Clifford lsquolsquoEconomic Deregulation Days of Reckoning for Microecono-mistsrsquorsquo Journal of Economic Literature XXXI (1993) 1263ndash1289
WHAT DRIVES DEREGULATION 1467
and Winston [1993]) Branching restrictions thus tend to reducethe efficiency and consumer convenience of the banking system9
The private-interest theory therefore predicts that reformshould occur later in those states where the strength of the smallbanks relative to that of the large banks is greater According tothe public-interest theory deregulation should take place earlierwhere small banks are relatively important because the socialcosts of the regulation are directly related to the size of theprotected sector The social costs include deadweight costs andlosses associated with inefficiencies in the production of bankingservices relative to production without geographical restrictions
Interindustry Rivalry A number of states permit state-chartered commercial banks to sell insurance The insurancelobby would thus oppose the relaxation of branching restrictionswhen banks can sell insurance because such deregulation mightpermit banks to provide a more efficient insurance distributionnetwork10 According to the private-interest theory reform shouldoccur later in states where banks can sell insurance and theinsurance industry is important relative to the banking industryThe efficiency costs of the branching restrictions however risewith the size of the insurance sector since the restrictions preventthe exploitation of scope economies Under the public-interesttheory reform should therefore occur earlier where banks can sellinsurance and the insurance industry is relatively large11
Consumers of Banking Services Banks are a major source ofcredit for small rms [Cole and Wolken 1994] Branching deregu-lation tends to reduce banksrsquo local market power [Jayaratne andStrahan 1998] In addition Strahan and Weston [1998] nd thatlending to small businesses increases on average when smallbanks are purchased by other banking organizations and Bergeret al [1998] nd that credit availability to small businessesincreases in the years following a takeover of a small bank by alarger banking organization Since bank borrowers tend to benet
9 Flannery [1984] shows that small banks in states with branching restric-tions have higher costs than small banks in states without such restrictions andthe Appendix shows how the production of banking services becomes more efficientfollowing deregulation
10 The Economist [8694 p 59] describes the insurance industryrsquos opposi-tion to branching deregulation
11 Another rival interest we considered was the Savings and Loan industryThe share of assets in Savings and Loans relative to banking plus Savings andLoan assets in the state however had only small and statistically insignicanteffects in the models we estimate below
QUARTERLY JOURNAL OF ECONOMICS1444
from branching deregulation in particular and bank consolidationin general the private-interest theory would predict that stateswith numerous small bank-dependent rms would deregulateearlier12 This prediction however is also consistent with thepublic-interest theory since the social costs of the restrictions arehigher in states with more small bank-dependent rms
Another effect of branching restrictions on bank customerscan be related to the prices paid for bank services High loanprices before deregulation may reect for example a largemarket share for inefficient high-cost banks or high rents beingearned by the banks protected from competition In either casethe public-interest theory would suggest that states with rela-tively high initial loan interest rates should deregulate earlierThe private-interest theory however does not have a clearimplication high rents could lead the beneciaries to ght harderto maintain them (eg Stigler [1971]) or the deadweight costsassociated with these high rents could make it more difficult forthe beneciaries to maintain the restrictions (eg Becker [1983])13
Bank Stability Geographic diversication through branch-ing could mitigate instability problems that were importantduring much of our sample period thereby improving welfareInstability also may reduce the incentives of banks to lobby tomaintain protections because unstable banks are less likely tosurvive to reap the benets of the restrictions politicians also mayanticipate that such banks are less likely to be able to providefuture contributions and support [Gunther 1994 1996]14 Both theprivate-interest and public-interest theories thus suggest thatderegulation is most likely in states where banking instability isgreatest
12 On the other hand local banking monopolies created by branchingrestrictions could strengthen relationshipsbetween banks and small and medium-sized rms and increase the availability of credit to these rms [Petersen andRajan 1994] Also some have argued that small business lending declines whenlarge banks take over small banks (eg Berger Kashyap Scalise [1995])
13 High loan interest rates also could reect compensation for higher risksand fewer opportunities for banks to diversify If state-specic risks explaindifferences in average loans rates then neither theory would have any predictionsfor the sign of this effect Following branching deregulation however loan interestrates tend to fall but bank risk proles do not appear to change [Jayaratne andStrahan 1998 Kroszner and Strahan 1998] This suggests that high loan rates inantibranching states are not simply reecting state-specic risks
14 Also deregulation can arise as a response to banking instability under atheory of ignorant or misinformed voters in which a banking crisis acts as aneducational device to make the previously ignorant public aware of the costs of theantibranching policy (see Kane [1996])
WHAT DRIVES DEREGULATION 1445
Political-Institutional Factors Republicans are typically per-ceived as more likely to favor deregulation than Democrats so thepolitical-institutional theories suggest that states that are con-trolled by Democrats deregulate later than those controlled byRepublicans15 In addition we investigate whether reform be-comes more likely when one party controls the legislature and thegovernorship Note that these political effects must be interpretedwith caution since the views of the politicians may simply reectthe economic interests of the constituents in the state (seePeltzman [1984])
B Variable Denitions and Data Sources
Our main proxy for the strength of the small banks is thefraction of banking assets in the state in lsquolsquosmallrsquorsquo banks We denesmall banks as banks with assets below the median size in eachstate By allowing the denition of small to vary across states wetake into account cross-state heterogeneity in bank sizes16 An-nual data on bank size are from the fourth quarter Reports ofIncome and Conditions (lsquolsquoCall Reportsrsquorsquo) from the Federal ReserveBoard We also include the capital-to-asset ratio of small banksrelative to large as a measure of their relative nancial healthSpecically this variable is the asset-weighted average capital-asset ratio for small banks minus the asset-weighted averagecapital-asset ratio for large banks Annual data on bank capitalare from the Call Reports
To measure the effects of the rival insurance industry we rstconstruct an indicator variable that is one if the state permitsbanks to sell insurance For each state we then measure the sizeof the insurance sector (total value added in the state) relative tothe sum of the banking plus insurance sectors We will examine
15 For more detail on the importance of legislative structures party politicsand ideology see Poole and Rosenthal [1997] Kahn and Matsusaka [1997] Irwinand Kroszner [1999] and Kroszner and Stratmann [1999]
16 We also considered a variety of other denitions of relative size but therelative size results reported below are not sensitive to which denition we useFirst we used a xed measure of small bank that did not vary across statesdening small banks as those with assets below $100 million in 1994 dollarsSecond we applied the state-varying and xed denitions at the level of lsquolsquobankingorganizationsrsquorsquo (which include multibank holding companies) instead of at the levelof banks Third we calculated gini coefficients of bank size inequality and bankconcentration indices as alternative proxies for small versus large bank power inthe state Finally we also considered banks in lsquolsquoruralrsquorsquo areas (that is not located in aBureau of the Census lsquolsquostandard metropolitan statistical arearsquorsquo) as small since thevalue of the restrictions might be greatest in protecting banks outside of cities fromentry by the city banks All of the alternative denitions are highly correlated andyield the same results on the importance of intraindustry differences
QUARTERLY JOURNAL OF ECONOMICS1446
the effect separately for states that permit banks to sell insuranceand those that do not Data on value-added by industry are fromU S Commerce Department Bureau of Economic Analysis Sur-vey of Current Business [August 1994]
To measure the relative importance of small bank-dependentborrowers we include the proportion of all establishments operat-ing in the state with fewer than twenty employees These data arecompiled by the Bureau of the Census17 As a measure of priceswe use the difference between the average interest rate on loans inthe state and the prevailing federal funds rate where the loanrate equals the ratio of total interest income on all domestic loansdivided by total domestic loans held by banks operating in thestate18 These data are from the end-of-year Call Reports butbecome available beginning only in 1976 Our proxy for bankinstability is the failure rate of banks measured as the percent oftotal state banking assets in failed institutions from the FederalReserve Boardrsquos National Information Center database19
We include two political variables to adjust for any indepen-dent inuence of party politics First we measure the degree ofparty control of the state government by the fraction of the threebodies of the state government (the assembly senate and gover-norship) controlled by Democrats This variable for example isone-third if the Democrats have a majority in the assembly andthe Republicans have a majority in the senate and hold thegovernorship Second we include an indicator variable equal toone if the same party controls the governorrsquos office and hasmajorities in both chambers of the state legislature20
17 We collected the establishment data by the state for three cross sections(1976 1982 and 1987) and interpolated the data in the intermediate years SeeState and Metropolitan Data Book 1982 1986 and 1991
18 A measure of how individual customers would be affected is interest ratesthey receive on deposits We calculated the average deposit interest rate in thestate from the end-of-year Call Reports and did not nd any relationship betweenthis variable and the timing of deregulation in our analysis below Since Jayaratneand Strahan [1998] nd that branching deregulation has no effect on deposit ratesour nding of no relationship between timing and rates result could be due toeither individual bank customers anticipating little pricing effect and not lobbyingor an inability of individual depositors to form an effective lobby
19 As alternative measures of distress we used (1) the bank failure ratemeasured as the number of failed institutions in the state relative to the totalnumber of banks (2) the overall capital-to-asset ratio of all banks in the state and(3) the growth in state personal income Our results are not sensitive to the choiceof distress proxy
20 We also estimated the models with three indicator variables to reect theparty control of the assembly senate and governorship separately The coefficientestimates on these variables are qualitatively similar to those we present below(that is Democratic control tends to slow deregulation) and the results on theinterest group variables are unaffected by the specication of the political
WHAT DRIVES DEREGULATION 1447
IV METHODS AND RESULTS
In this section we rst estimate how the variables describedabove inuence the timing of state-level branching deregulationusing a hazard model Next we discuss the robustness of theresults including a test of whether the ex post impact of deregula-tion on the different interest groups is consistent with ourinterpretation of the coefficients We then investigate federalinterstate branching deregulation with a probit model to explorewhether the factors that affect the timing of intrastate deregula-tion also inuence voting in the House of Representatives oninterstate deregulation
A Hazard Model
We consider the period from the beginning of our sample(1970) until deregulation as the lsquolsquoduration of regulationrsquorsquo or thelsquolsquotime until deregulationrsquorsquo The hazard rate h(t) is the likelihoodthat a state deregulates at time t given that the state has not yetderegulated [Kalbeisch and Prentice 1980 Kiefer 1988 Greene1997] To model the duration of regulation we must decide whatstructure if any to impose on the hazard function The Kaplan-Meier product-limit estimator provides a simple nonparametricway to estimate the shape of the hazard function over time[Greene 1997]21 Figure II graphs this estimate for our data andshows that the hazard function is relatively at in the early yearsand then grows steeper in the later years
A duration model that can approximate this shape is theWeibull proportional hazards model The hazard rate functiontakes the form
h[t x(t) b] 5 h0(t) exp[x(t)8b]
where x(t) is a vector of time-varying covariates b is a vector ofunknown parameters to be estimated and the baseline hazard
variables Note also that Nebraska has a nonpartisan unicameral legislature Weassume that the party of the Governor controls the state government for Nebraskaand set the unied government indicator to one As another alternative we triedincluding a variable that is one if party control of the legislature or governorshipchanges but it did not have an economically or statistically signicant effect anddid not affect any of the other results
21 Denote each of the K years in our sample period Tk and order them suchthat T1 T2 and so on Let nk be the number of states that have not yet deregulatedby Tk and dk be the number of states that deregulate in year Tk The Kaplan-Meierestimate of the hazard rate in each year is dknk In other words the hazard rate isthe ratio of the number of states that actually deregulate in each year to thenumber of states that have not yet deregulated in that year
QUARTERLY JOURNAL OF ECONOMICS1448
rate h0(t) is ptp 2 1 with shape parameter p that will be estimatedfrom the data When p 1 this model displays a monotonicallyincreasing hazard rate To estimate the parameters we maximizethe following log-likelihood function [Kiefer 1988]
L(b) 5 oi 5 1
N
di ln h[ti x(ti) b] 2 oi 5 1
N
e 0
tih[u x(u) b] du
where N indexes the number of states at risk at the beginning ofthe sample period di is an indicator equal to one for the statesthat deregulated by the end of the sample period and zero for thestates that do not deregulate (ie the censored observations) andti is the time of deregulation for the ith state This formulationassumes that the time until deregulation is independent acrossstates but the observations of the same state over time are notassumed to be independent In addition we use a robust estima-tion procedure to calculate the standard errors [Lin and Wei1989]
We exclude states that deregulated before 1970 from theanalysis so N equals 39 states During the sample period 36 ofthe 39 states deregulate We have three censored observations forwhich di is zero (Arkansas Iowa and Minnesota) but our results
FIGURE IIKaplan-Meier Estimate of Hazard Rate for Branching Deregulation
WHAT DRIVES DEREGULATION 1449
are not sensitive to the inclusion of these censored observationsSince we observe each state in each year up to and including theyear of deregulation we have a total of 637 observations Table IIreports summary statistics and correlation coefficients for theexplanatory variables we use in our analysis
We use the Weibull model rather than a model that does notimpose any structure on the baseline hazard rate so that we cancalculate the change in the expected time to deregulation for agiven change in the levels of the covariates22 In the Weibullmodel we can invert the hazard function and map it into the timedomain Rewriting the Weibull model in this way the log of thetime to deregulation T is a linear function of the political economyfactors and an error term ln (T ) 5 x8b 1 e23 Because we assumethat the baseline hazard rate is Weibull with a shape characterizedby p the new coefficients will be scaled by p that is b 5 2 bp
The b coefficients represent the percentage change in thetime to deregulation for a one-unit change in the covariates Apositive coefficient implies that an increase in the variableincreases the expected time until deregulation To gauge theeconomic importance of the effects we will multiply the b by thestandard deviation of the explanatory variables and then evaluatehow much the changes in the variables raise or lower the expectedtime to deregulation24 In Table III we report the b coefficients
22 The Cox proportional hazards model does not impose any structure on thebaseline hazard rate h0(t) and takes the following form
h[t x(t) b] 5 h0(t) exp [x(t)8b]
Common factors that affect the probability of deregulation in the same way for allstates therefore do not inuence the estimated coefficients The cost of the Coxmodel is that because there is no structure on the baseline we can calculatechanges in only relative hazard rates associated with changes in the covariatesThe Weibull model however provides sufficient structure so that we can translateour estimates into a log expected time metric that allows us to calculate the changein the expected time to deregulation for a given change in the covariates SeeKiefer [1988] Greene [1997] and Stata [1997] When we compare the coefficientsestimated by the Cox and Weibull models on our data they are very close (seeKroszner and Strahan [1998]) so imposing a Weibull structure does not appear todistort our estimates
23 The error term e is independent of x and has an extreme value distributionscaled by 1p (see Kiefer [1988] Greene [1997] and Stata [1997] for more details onthis log expected time parameterization of the Weibull model)
24 The transformation to the log expected time domain is only strictly correctwhen the covariates are constant over time Our interpretation of the coefficients ispossible in the time-varying model if we assume that a one-standard-deviationdifference would hold over the entire sample period that is the deregulationwould occur so many years earlier or later if the variable in question were higherby one standard deviation from 1970 to 1992 Note that the mean number of yearsto deregulation in the sample is 158
QUARTERLY JOURNAL OF ECONOMICS1450
TABLE IISUMMARY STATISTICS FOR VARIABLES IN HAZARD MODEL 1970ndash1992
Panel A Univariate statistics MeanStandarddeviation Minimum Maximum
(1) (2) (3) (4)Small bank asset share of all
banking assets in state 011 005 001 023Relative size of insurance to
banking plus insurance inthe state 047 009 024 080
Indicator is 1 if banks can sellinsurance in the state 020 040 0 1
Small rm share of the numberof rms in the state 088 002 078 095
Capital ratio of small banksrelative to large in the state 002 001 2 001 009
Share of state government con-trolled by Democrats 065 036 0 1
Indicator is 1 if state controlledby one party 054 050 0 1
Average yield on bank loans inthe state minus Fed fundsrate 002 002 2 004 008
Panel B CorrelationsSmallbank
Relativesize of
insuranceInsuranceindicator
Smallrm
Smallbank
capitalDem
controlSingleparty
(1) (2) (3) (4) (5) (6) (7)Small bank asset share
of all banking assetsin state 1
Relative size of insur-ance to banking plusinsurance in the state 2 034 1
Indicator is 1 if bankscan sell insurance inthe state 020 2 012 1
Small rm share of thenumber of rms inthe state 030 2 011 005 1
Capital ratio of smallbanks relative tolarge in the state 2 066 020 2 010 2 006 1
Share of state govern-ment controlled byDemocrats 2 006 001 2 018 2 018 005 1
Indicator is 1 if statecontrolled by oneparty 007 012 001 2 006 2 006 030 1
Average yield on bankloans in the stateminus Fed funds rate 2 001 2 015 003 022 2 010 2 003 2 006
The number of observations is 637 except for the average interest rate on bank loans which becomesavailable only in 1976 so the number of observations is 408
TABLE IIIHAZARD MODEL OF POLITICAL ECONOMY FACTORS AFFECTING THE TIMING OF STATE
BRANCHING DEREGULATION 1970ndash1992
(1) (2) (3) (4) (5) (6)Small bank asset share of
all banking assets instate
648(122)
519(137)
632(135)
516(118)
488(133)
445(088)
Capital ratio of smallbanks relative to largein the state
1325(348)
921(319)
1301(362)
1067(346)
1144(375)
906(381)
Relative size of insurancein states where banksmay sell insurance 0otherwise
324(161)
259(149)
315(165)
285(154)
272(159)
055(134)
Indicator is 1 if banksmay sell insurance inthe state
045(010)
037(011)
045(010)
038(008)
036(009)
001(012)
Relative size of insurancein states where banksmay not sell insurance0 otherwise
2 093(043)
2 090(033)
2 102(051)
2 093(048)
2 096(053)
2 004(039)
Small rm share of thenumber of rms in thestate
2 972(211)
2 614(215)
2 952(211)
2 943(230)
2 952(270)
2 1509(276)
Share of state govern-ment controlled byDemocrats
031(012)
023(011)
030(012)
027(011)
026(011)
011(013)
Indicator is 1 if state con-trolled by one party
2 004(010)
2 004(007)
2 004(007)
2 001(009)
2 002(010)
017(009)
Average yield on bankloans in the stateminus Fed funds rate mdash
023(319) mdash mdash mdash mdash
Bank failure rate mdash mdash092
(111) mdash mdash mdash
Indicator is 1 if state hasunit banking law mdash mdash mdash
018(010)
019(010)
030(009)
Indicator is 1 if statechanges bank insur-ance powers mdash mdash mdash mdash
2 008(012)
2 026(013)
Includes regionalindicators No No No No No Yes
N 637 408 637 637 637 637Log likelihood 2 374 900 2 360 2 212 2 191 1192p-value of x 2 for
regression 001 001 001 001 001 001
The hazard model is Weibull where the dependent variable is the log expected time to branchingderegulation ln (T) 5 bx 1 eAll variables are measured for each state in each year Small bank asset share isthe percent of banking assets in the state held by banks below the median size of banks in each state in eachyear Relative capital ratio is the capital to assets ratio of small banks minus that of large banks Size ofinsurance relative to banking plus insurance in the state is measured as gross state product from insurancedivided by gross state product from insurance plus banking Insurance indicator is 1 if state law permits banksto enter the insurance business Small rm share is the percent of all establishments in the state that havefewer than twenty employees Party control variable is the share of the three bodies of state governmentcontrolled by Democrats One party control indicator is 1 if the same party controlrsquos the governorship and hasmajorities in both chambers of the state legislature Average yield on bank loans equals total interest incomeon all domestic loans made by banks in the state divided by total loans Bank failure rate is the ratio of assetsin failed banks during a given state and year divided by total bank assets in that state and year Unit bankingindicator equals 1 for states with unit banking restrictions The change in insurance powers indicator equalsone if the state changed the law to permit banks to sell insurance during the sample period We divide thecountry into four regions and include regional indicators in column (5) Robust standard errors are inparentheses and denote statistically signicant at the 1 5 and 10 percent levels
QUARTERLY JOURNAL OF ECONOMICS1452
The positive and statistically signicant coefficient on smallbank share implies that a greater small bank share delaysderegulation and this effect is economically important A one-standard-deviation increase in the small bank share results in a30 percent increase in the time until deregulation or about 47years This result supports the implications of private-interesttheory concerning intraindustry rivalry and is contrary to thepublic-interest theory
The positive and statistically signicant coefficient on therelative performance of small banks in the state implies thatstates deregulate later when small banks are relatively strongnancially25 A one-standard-deviation increase in the relativecapital-to-asset ratio results in a 15 percent rise in the time untilderegulation or about 24 years This result is consistent with theprivate-interest theory but may also be consistent with a public-interest interpretation because the risk of failures (and a costlytaxpayer-funded bailout) increases as small bank nancial healthdeclines
Interindustry competition also helps explain the timing ofderegulation In states where banks can sell insurance a rela-tively large insurance sector is associated with an increase in theexpected time to deregulation26 A one-standard-deviation in-crease in the relative size of the insurance sector in those statesthat permit banks to sell insurance leads to a 22 percent increasein the time until deregulation or about 35 years27 This result isconsistent with the private-interest theory but not the public-interest theory
Deregulation occurs earlier in states where small bank-dependent rms are relatively numerous A one-standard-
25 We also allowed small bank and large bank capital-to-asset ratios to enterseparately Their coefficient estimates were nearly equal in absolute value and ofopposite signs and so the data do not reject the relative small to large bank capitalratio specication we have chosen
26 When we use the relative size of the insurance sector in the state as awhole rather than insurance relative to banking plus insurance in the state weobtain similar results When we simultaneously include both measures of therelative size of insurance the interactions of the lsquolsquoinsurance relative to bankingplus insurancersquorsquo variable continue to be statisticallysignicant but the interactionsof lsquolsquoinsurance relative to the state as a wholersquorsquo variable are not
27 We have also estimated the model with the indicator variable equal to onefor states where banks may sell insurance but without the variables measuringtheimportance of insurance relative to banking In this model the coefficient on theinsurance indicator variablemdashan estimate of the difference in the expected time toderegulation in states where banks may sell insurance relative to states wherebanks may not sell insurancemdashequals 027 and is statistically signicant at the 5percent level This suggests that branching deregulation takes about 45 yearslonger in states where banks may sell insurance
WHAT DRIVES DEREGULATION 1453
deviation increase in the share of small rms reduces the timeuntil deregulation by 18 percent or about three years This resultconcerning the interests of consumers of banking services isconsistent with both the private- and public-interest theories
Column (2) of Table III introduces the average interest rateon loans to the model as a rough proxy for the cost to bankborrowers of these regulations These data become available onlyin 1976 so our sample size shrinks to 408 observations Thecoefficient on this variable is small and not statistically signi-cant The public-interest theory would imply that the coefficientshould be negative while the implication of the economic theorydoes not have a clear sign prediction
When we add the failure rate of banks to the specication incolumn (3) its coefficient is small and statistically insignicantWe do not nd a linkage between the timing of state deregulationand statewide banking distress Given that deposit insurance isfunded federally however bank failures may raise the publicrsquosconcerns about the costs of the regulations at the national ratherthan the state level This interpretation can explain why the rateof bank failure does not affect the timing of state-level deregula-tion even though bank instability does appear to rise generallyduring the period of increasing deregulation (see Section V below)
Finally the partisan structure of the state government alsoinuences when states deregulate As expected a higher propor-tion of Democrats in the government tends to delay deregulationA one-standard-deviation rise in the share of the governmentcontrolled by Democrats slows the deregulation by about twoyears Whether the state is dominated by one party however doesnot appear to affect the timing of the deregulation
To summarize our ndings from the hazard model privateinterests appear to play an important role in the deregulatoryprocess While private interests and public interests do sometimescoincide the results on the relative share of small banks and largebanks and on the relative size of insurance where banks competeare consistent with a private-interest approach but are difficult toexplain on public-interest grounds
B Robustness Tests
Omitted Variables States begin the sample period withdifferent degrees of restrictions on bank expansion These initialconditions may proxy for omitted factors determining the political-economy equilibrium in the state that could then affect the
QUARTERLY JOURNAL OF ECONOMICS1454
subsequent timing of deregulation For instance some statesbegan the period with unit banking while other states allowedlimited branching and the formation of multibank holdingcompanies
To test whether the stringency of initial restrictions onbranching matters we rst divide the sample into states with andwithout the most extreme form of branching regulation unitbanking28 Column (4) of Table III includes an indicator variableequal to one if the state began the period with unit bankingrestrictions The coefficient on this variable is positive andstatistically signicant but the other results remain virtuallyunchanged Next we include an indicator variable that is one ifthe state permitted multibank holding companies (MBHCs) toexist at the beginning of our sample period Thirty-four statesbegan the period allowing MBHCs The coefficient estimate forthis indicator variable (not reported) is small and not statisticallysignicant and its inclusion does not change the effects of theother variables
Eleven states expand bank powers to include the sale ofinsurance products during our sample period This could proxy foromitted factors about the relative strength of the insuranceindustry in the state In column (5) we include an indicatorvariable that is one if the state changes the law during our sampleperiod to permit banks to sell insurance business and zerootherwise The coefficient estimate is small and statisticallyinsignicant and does not affect the other results29
In column (6) we introduce four regional indicator variablesNorth South Midwest and West30 The coefficients on small bankshare the relative capital-to-assets ratio and small rm sharechange little Although the coefficient estimate still suggests that
28 We classify states that prohibited branching but permitted banks toestablish facilities as unit banking states The unit banking indicator is one for thefollowing states CO AR FL IL IA KS MN MO MT NE ND OK TX WI WVand WY
29 We conducted a likelihood ratio test of whether the unit banking statesthe MBHC states and the states that changed their regulations of bank insurancepowers could be pooled with the other states In all three cases we could not rejectpooling of the data
30 Since intrastate deregulation generally preceded interstate deregulationand the latter typically took the form of regional interstate compacts thepossibility of participating in a regional interstate banking compact could haveinuenced the decision to deregulate intrastate branching Our denitions of theregions are region 1 (South) contains AL AR DC FL GA KY LA MS NC OKSC TN TX and VA region 2 (Northeast) contains CT MA MD ME NH NJ NYPA RI VT and WV region 3 (Midwest) contains IA IL IN KS MI MN MO NEND OH SD and WI region 4 (West) contains the other states
WHAT DRIVES DEREGULATION 1455
in states where banks can sell insurance a larger relative size ofthe insurance sector delays deregulation the effect is no longerstatistically signicant The coefficient on the indicator variablefor whether a state expands bank powers to permit insurancesales during the sample period however now becomes statisti-cally signicantly negative This implies that deregulation occursroughly 41 years earlier in states where the insurance lobby maybe relatively weak While the inclusion of the regional indicatorsdoes not affect the other results the results on insurance doappear to be sensitive to the specication of the hazard model31
Estimation Technique To check the robustness of using ahazard model and to investigate whether the time-series or thecross-sectional variation in each of the factors accounts for thehazard results we estimate two simple linear probability models(neither are reported) First we run a pooled time-series cross-section OLS regression using the 637 observations from thehazard model The dependent variable equals one in the year ofderegulation and zero otherwise We include the same set ofregressors that were used in the hazard model and add state-levelxed effects This specication thus removes the cross-statevariation from the independent variables and focuses on theirmovements over time In this model the signs of the coefficientson the interest group factors are consistent with the results fromthe hazard model and the coefficients for the small bank relativecapital ratio and for the relative size of insurance where states cansell insurance are statistically signicant
In the second model we eliminate the time-series variationby averaging each of the explanatory variables over time We thushave one observation per state that had branching restrictions asof 1970 (N 5 39) The dependent variable equals the inverse of thenumber of years until deregulation which represents an estimateof the statersquos probability of deregulating in a year under theassumption that this probability is constant over time and eachyear provides an independent observation32 Although we haveonly 39 observations the coefficients on the small bank marketshare small bank relative capital ratio and small rm share
31 As further robustness check we also tried a population density variablewhich Abrams and Settle [1993] found relevant for regulatory change during the1920s and 1930s but its coefficient was not statistically signicant and did notaffect the other results
32 For instance the dependent variable would take the value of 01 for astate that deregulated in 1980 For the three states that did not deregulate thedependent variable equals 0
QUARTERLY JOURNAL OF ECONOMICS1456
variables are statistically signicant The signs of the effects ofthe insurance variables are the same as those reported in thehazard model but not statistically signicant
The hazard model and linear probability models thus producevery similar results The estimates from the latter suggest thatthe effects for the small bank market share and small rm sharevariables are driven primarily by cross-state variation the effectof the insurance variable is driven primarily by time-seriesvariation and the small bank capital effect is driven by bothcross-sectional and time-series variation
Ex Post Consequences of Deregulation Finally to check theplausibility of our interpretation of the hazard results we con-sider whether the ex post consequences of deregulation areconsistent with the ex ante lobbying positions we attribute to eachinterest group (see Kroszner and Strahan [1998] for details) Wend that small banks lose market share following deregulationand in states where banks can enter the insurance business theinsurance sector shrinks relative to the banking sector followingderegulation Borrowers also benet from deregulation throughlower average interest rates on loans These ndings support theprivate-interest interpretation of the coefficients in the hazardmodel groups that will benet lobby to speed deregulation andthose that will be harmed lobby to slow it
C Voting on Interstate Branching Deregulation in the U SHouse of Representatives
We now examine whether the forces driving intrastate branch-ing deregulation also drive interstate deregulation Financialservices interests are active contributors and lobbyists in Washing-ton Their political action committees constitute the largest groupof contributors to legislators providing nearly 20 percent of totalcongressional campaign contributions [Makinson 1992] and muchof their lobbying effort involves competition among rival interestswithin nancial services (see Kroszner and Stratmann [1998])
As noted above after virtually all states adopted intra- andinterstate branching deregulation the 1994 Riegle-Neal Act re-pealed the 1927 McFadden Act to eliminate all barriers tointerstate banking and branching by 1997 The key votes concern-ing the Riegle-Neal Act were either voice votes or extremelylopsided so we could not estimate a voting model from them Anumber of bills and amendments related to interstate branching
WHAT DRIVES DEREGULATION 1457
had been debated in Congress during the years prior to thepassage of the Riegle-Neal Act but a search of the weekly BNABanking Reporter and the Congressional Record produced onlyone roll-call vote related to interstate branching that was notlopsided This vote occurred in the House on November 14 1991and concerned an amendment sponsored by Wylie (R-OH) andNeal (D-NC) to introduce interstate banking and branchingderegulation to a nancial services reform package33 Althoughthe amendment passed by 210 to 208 the bill to which it wasattached subsequently was defeated
To check for the inuence of the factors we considered in thestate-level reforms we examine both the sponsorship of interstatebanking legislation and voting on the amendment The sponsorsof the Wylie-Neal amendment are from states with low small bankshares 004 in Ohio (Wylie) and 002 in North Carolina (Neal) Incontrast the sample mean in 1991 is 008 (median 5 007)Michigan home state of the Senatersquos sponsor of the 1994 Riegle-Neal Act also had relatively low small bank strength (small bankshare of 005)
Table IV reports estimates of a probit model where thedependent variable equals one if the legislator voted in favor ofthe amendment and zero otherwise The explanatory variablesare the same as those in the hazard model except that we use aparty affiliation indicator variable equal to one for Democrats andzero for Republicans as our proxy for political factors Thecoefficients in Table IV are the marginal effects (lsquolsquoslopesrsquorsquo) of aone-unit change of each variable on the probability that alegislator will vote for the amendment34 Note that a forcefavoring deregulation has a negative coefficient in the hazardmodel (Table III) but a positive coefficient in the probit model
Consistent with the state-level deregulation process legisla-tors are more likely to support the amendment if their states havea relatively low share of small banks As in the hazard model thefraction of small banks is the most important interest groupinuence on a legislatorrsquos voting decision A one-standard-deviation increase in small banksrsquo market share (from the mean)is associated with a decline in the probability of voting in favor of
33 The Wylie-Neal amendment also included provisions limiting certaininsurance and real estate powers of national banks [Congressional RecordNovember 14 1991 pp 10239ndash10242]
34 Since we have multiple legislators from each state we adjust the standarderrors to correct for the lack of independence among observations clustered in thesame state
QUARTERLY JOURNAL OF ECONOMICS1458
TAB
LE
IVM
AR
GIN
AL
EF
FE
CT
SF
RO
MA
PR
OB
ITM
OD
EL
OF
TH
EIN
FL
UE
NC
EO
FP
OL
ITIC
AL
EC
ON
OM
YF
AC
TO
RS
ON
TH
EH
OU
SE
VO
TE
ON
TH
EW
YL
IE-N
EA
L
AM
EN
DM
EN
TT
OP
ER
MIT
INT
ER
ST
AT
EB
AN
KB
RA
NC
HIN
G
NO
VE
MB
ER
14
1991
(1)
(2)
(3)
(4)
(5)
(6)
Smal
lban
kas
set
shar
eof
allb
anki
ng
asse
tsin
stat
e2
417
(12
4)2
454
(12
2)2
407
(12
0)2
443
(14
7)2
450
(14
9)2
490
(14
7)C
apit
alra
tio
ofsm
allb
anks
rela
tive
tola
rge
inth
est
ate
180
(37
5)2
68(3
42)
256
(37
5)1
69(3
78)
153
(38
0)2
45(4
25)
Rel
ativ
esi
zeof
insu
ranc
ein
stat
esw
here
bank
sm
ayse
llin
sura
nce
0ot
herw
ise
22
34
(10
2)2
249
(1
14)
22
33
(10
1)2
249
(1
04)
22
51
(10
2)2
154
(11
1)In
dica
tor
is1
ifba
nks
may
sell
insu
ranc
ein
the
stat
e0
79
(0
06)
074
(00
6)0
69
(0
06)
070
(00
6)0
70
(0
06)
067
(00
7)R
elat
ive
size
ofin
sura
nce
inst
ates
whe
reba
nks
may
not
sell
insu
ranc
e0
oth
erw
ise
180
(06
6)2
42
(0
72)
154
(0
67)
173
(06
6)1
72
(0
67)
203
(0
91)
Smal
lrm
shar
eof
the
num
ber
ofr
ms
inth
est
ate
21
23(0
92)
20
83(0
89)
21
62(1
07)
21
32(0
95)
21
30(1
00)
20
94(1
16)
Indi
cato
ris
1if
Rep
rese
ntat
ive
isa
Dem
ocra
t2
035
(00
7)2
035
(00
7)2
036
(00
7)2
035
(00
7)2
035
(00
7)2
036
(00
7)A
vera
geyi
eld
onba
nklo
ans
inth
est
ate
min
usF
edfu
nds
rate
mdash10
01
(51
5)mdash
mdashmdash
mdash
Ave
rage
fail
ure
rate
198
5to
1991
mdashmdash
228
(14
1)mdash
mdashmdash
Indi
cato
ris
1if
stat
eh
asun
itba
nkin
gla
wmdash
mdashmdash
004
(00
7)0
04(0
07)
20
01(0
08)
Indi
cato
ris
1if
stat
ech
ange
sba
nk
insu
ran
cepo
wer
smdash
mdashmdash
mdash2
001
(00
7)2
003
(00
7)In
clud
esre
gion
alin
dica
tors
N
oN
oN
oN
oN
oYe
sP
seud
o-R
20
170
180
170
170
170
18L
ogli
keli
hood
224
024
223
826
223
951
224
013
224
010
223
808
p-va
lue
ofx2
for
regr
essi
on
001
0
01
001
0
01
001
0
01
Th
ede
pen
den
tvar
iabl
eis
one
ifth
ele
gisl
ator
vote
sfo
rth
eam
endm
enta
nd
zero
ifag
ains
tT
he
repo
rted
coef
fici
ents
are
the
effe
cts
ofa
uni
tch
ange
ofth
ein
depe
nde
ntva
riab
le(f
rom
the
mea
n)on
the
prob
abili
tyof
voti
ng
infa
vor
ofth
eam
endm
ent
All
vari
able
sar
em
easu
red
byst
ate
N5
418
the
num
ber
ofR
epre
sent
ativ
esvo
tin
gon
the
amen
dmen
tS
mal
lban
kas
sets
har
eis
the
perc
ent
ofba
nki
ng
asse
tsin
the
stat
eh
eld
byba
nks
belo
wth
em
edia
nsi
zeof
ban
kin
each
stat
ein
each
year
Rel
ativ
eca
pita
lrat
iois
the
capi
talt
oas
sets
rati
oof
smal
lban
ksm
inu
sth
atof
larg
eba
nks
Siz
eof
insu
ran
cere
lati
veto
bank
ing
plu
sin
sura
nce
inth
est
ate
ism
easu
red
asgr
oss
stat
epr
oduc
tfro
min
sura
nce
divi
ded
bygr
oss
stat
epr
odu
ctfr
omin
sura
nce
plu
sba
nkin
gIn
dica
tor
ison
eif
stat
ela
wpe
rmit
sba
nks
toen
ter
the
insu
ran
cebu
sin
ess
Sm
all
rmsh
are
isth
epe
rcen
tofa
lles
tabl
ishm
ents
inth
est
ate
that
hav
efe
wer
than
twen
tyem
ploy
ees
Par
tyin
dica
tor
ison
eif
the
Rep
rese
nta
tive
isa
Dem
ocra
tA
vera
geyi
eld
onba
nklo
ans
equa
lsto
tali
nter
est
inco
me
onal
ldom
esti
clo
ans
mad
eby
ban
ksin
the
stat
edi
vide
dby
tota
lloa
ns
The
aver
age
bank
failu
rera
teis
the
rati
oof
asse
tsin
faile
dba
nks
duri
ng
agi
ven
stat
edi
vide
dby
tota
lban
kas
sets
inth
atst
ate
aver
aged
over
the
year
s19
85to
1991
Un
itba
nki
ng
indi
cato
req
ual
son
efo
rst
ates
wit
hun
itba
nkin
gre
stri
ctio
nsT
hech
ange
inin
sura
nce
pow
ers
indi
cato
req
ual
son
eif
the
stat
ech
ange
dth
ela
wto
perm
itba
nks
tose
llin
sura
nce
duri
ngth
esa
mpl
epe
riod
We
divi
deth
eco
unt
ryin
tofo
urre
gion
san
din
clu
dere
gion
alin
dica
tors
inco
lum
n(5
)R
obu
stst
anda
rder
rors
are
inpa
ren
thes
es
an
d
den
ote
stat
isti
cally
sign
ica
nta
tth
e1
5an
d10
perc
entl
evel
s
WHAT DRIVES DEREGULATION 1459
branching of approximately 17 percent The relative capital-to-assets ratio of small and large banks however does not have astatistically signicant effect in the probit model
The impact of rival interests outside of banking is alsoconsistent with intrastate deregulation results Where banks cansell insurance legislators from states with larger insurancesectors relative to banking are less likely to vote for the amend-ment35 A one-standard-deviation increase in the relative size ofinsurance in those states which permit banks to sell insurancedecreases the probability that a legislator will favor the amend-ment by about 13 percent Turning to consumer interests thecoefficient on small rm share is not statistically signicant butthe coefficient on interest rates in the state is A one-standard-deviation increase in the average interest rate on loans raises theprobability that a legislator will support the amendment byroughly 8 percent The coefficient estimate for the bankingdistress variable measured as the average bank failure rate from1985 to 1991 is small and statistically signicant Overall theprobit analysis of the vote on national branching deregulationsupports the private-interest theory of deregulation and providesa consistency check that the importance of interests operating onthe state legislatures are very similar to those operating at thefederal level
V WHY DID BRANCHING DEREGULATION BEGIN DURING THE
1970S SHOCKS TO THE EQUILIBRIUM SUPPORTING BANK
BRANCHING RESTRICTIONS
A complete explanation of regulatory exit should account forwhy deregulation begins in the 1970s as well as the timing of thestate-by-state reforms [Kane 1996] In this section we describebroad technological legal and economic shocks that altered thepolitical-economy equilibrium which had kept antibranching regu-lations little changed for at least 30 years
Beginning in the 1970s three innovations reduced the valueto the protected banks of local geographic monopolies (see TableV) First automatic teller machines (ATMs) helped to erode thegeographic ties between customers and banks Second checkablemoney market mutual funds and the Merrill Lynch Cash Manage-ment Account demonstrated that banking by mail and telephone
35 The positive and statistically signicant estimates on the other insurancevariables may reect insurance industry support for the amendmentrsquos provisionslimiting banksrsquo insurance powers
QUARTERLY JOURNAL OF ECONOMICS1460
provided a convenient alternative to local banks36 Third techno-logical innovation and deregulation reduced transportation andcommunication costs particularly since the 1970s thereby lower-ing the costs for customers to use distant banks By increasing theelasticity of deposits supplied to banks these innovations reducedthe value of geographical restrictions to their traditional benecia-ries and thereby reduced their incentive to ght to maintain them[Peltzman 1976]
On the lending side increasing sophistication of credit-scoring techniques following innovations in information process-ing nancial theory and the development of large credit databases began to diminish the value of knowledge that localbankers had about the risks of borrowers in the community As a
36 Regulation Q which limited the interest rates that banks could pay ondeposits may have helped to drive depositors away from banks when the gapbetween market rates and deposit ceilings grew during the 1970s The eliminationof interest rate ceilings on large denomination certicates of deposit during the1970s appears to have hurt smaller and retail-oriented banks relative to largerwholesale banks [James 1983]
TABLE VBROAD TRENDS IN COMMERCIAL BANKING 1950ndash1995
YearNumberof ATMs
Domesticbank
deposits(billions)
Moneymarketmutual
fund(billions)
Percent ofdeposits 1
moneyfunds heldby banks
Smallbanksrsquo
percent ofbankingassets
Averagenumber of
bankfailures
(1) (2) (3) (4) (5) (6)1950 0 $ 154 $ 0 100 NA 41955 0 191 0 100 NA 31960 0 228 0 100 24 21965 0 330 0 100 20 41970 0 479 0 100 18 61975 9750 775 4 99 18 61980 18500 1182 76 94 17 101985 61117 1787 242 88 14 601990 80156 2339 493 83 11 1791995 122706 2552 745 77 8 61
Column (1) ATM gures are from Bank Network News The EFT Network Data Book (New YorkFaulkner and Gray Inc) The 1975 gure was unavailable 9750 is the number of ATMs in 1978 the rst yearfor which complete data are available Columns (2)ndash(4) Banksrsquo domestic deposits are from the Reports ofIncome and Condition money market mutual funds are from the Flow of Funds Data on all bank depositsforeign plus domestic are only available beginning in 1970 The trend in banksrsquo share (column (4)) is the sameusing total deposits instead of domestic deposits Column (5) Percent of banking assets held by small bankswhere a small bank is dened as a commercial bank with less than $100 million in assets in 1994 dollarsThese data are based on the Reports of Income and Condition Data on small banks are not available before1960 Column (6) Five-year average number of bank failures where the nal year is indicated in the rstcolumn These data are from FDIC Annual Report and the Quarterly Banking Prole
WHAT DRIVES DEREGULATION 1461
result of these innovations a national market developed forresidential mortgages credit card receivables have been securi-tized and bank lending to small business now relies less on thejudgment of loan officers and more on standardized scoringmodels
These changes have increased the potential protability forlarge banks to enter what had been the core of small bankactivities Large banksrsquo incentive to increase their lobbying pres-sure to be able to expand into these markets has thus beenincreasing over time In fact small banksrsquo market share began todecline even prior to the branching deregulation (Table V) As thevalue of a local banking relationships declined small rms thatwere the main borrowers from the small banks also probablybecame more likely to favor the entry of large banks into localmarkets With the deadweight costs of preventing large bankentry rising the private-interest theory predicts that small localbanks would become less likely able to maintain the branchingrestrictions [Becker 1983] Deregulation that reduces deadweightcosts of regulation is also consistent with the public-interesttheory
Kane [1996] argues that another major shock to the oldequilibrium was the increasing public awareness of the costlinessof having government-insured but (geographically) undiversiednancial institutions In the late 1970s the failure rate of banksbegan to rise (Table V) In the 1980s the Savings and Loan crisisand taxpayer bailout further heightened public awareness aboutthe costs of restrictions that make depository institutions morelikely to require infusions of taxpayer funds The failures thusmay have heightened public support for branching deregulation(see also Abrams and Settle [1993]) As noted above we did notnd any evidence that banking failures or distress in a stateaffected the speed with which the state deregulated but distressstill may be a common factor affecting all states
These technological economic and legal shocks generatedconditions that changed the long-standing balance favoring theantibranching forces The marginal value of lobbying to repealbranching restrictions increased just as the relative value to thesmall banks of maintaining branching restrictions was decliningThese nationwide shocks are common factors across states andthey are consistent with the positive duration dependence associ-ated with state-level deregulation that we demonstrated in FigureII In addition a broad change in lsquolsquoideologyrsquorsquo against government
QUARTERLY JOURNAL OF ECONOMICS1462
regulation could explain the positive duration dependence duringthe sample period
VI CONCLUSIONS
The private-interest theory of regulation can account for thepattern of bank branching deregulation of the 1970s and 1980sBeneciaries of branching regulation had supported a coalitionfavoring geographical restrictions despite their costs to consum-ers of nancial services Innovations that began in the 1970saltered the value of the restrictions to the affected parties and theresulting competition among interest groups can explain thesubsequent deregulation While some of our results are alsoconsistent with the public interest theorymdashfor example deregula-tion occurs earlier when small banks are in a relatively weaknancial positionmdashother results particularly evidence on theimportance of rivalries between small and large banks andbetween banking and insurance are difficult to explain with thepublic-interest approach We also nd that political-institutionalfactors affect deregulation although these variables may act asproxies for unmeasured economic interests The greater success ofthe private-interest theory to explain deregulation here than inprevious work may be due to the richer cross-sectional andtime-series variation in branching deregulation that can beexploited with a hazard model Future empirical studies ofendogenous deregulation thus may be most fruitful where changehas occurred across states over time such as in franchisinginsurance and public utilities (eg Joskow [1996])
Technological and nancial innovations will continue to erodethe benets to any interest group of maintaining regulatorybarriers in nancial services These forces are likely to bringabout reforms both domestically for example through legislationthat would increase bank powers (see Kroszner [1996] Krosznerand Rajan [1994 1997] and Kroszner and Stratmann [1998]) andinternationally for example through the extension of nancialservices provisions of NAFTAto reduce geographic barriers acrosscountries [Kroszner 1997]
APPENDIX EFFECTS OF DIFFERENT TYPES OF BRANCHING
DEREGULATION ON ECONOMIC GROWTH BANKING STRUCTUREAND BANK EFFICIENCY
We measure the importance of alternative forms of deregula-tion by their consequences for the economic growth and the
WHAT DRIVES DEREGULATION 1463
structure and efficiency of the banking sector across the states(see Jayaratne and Strahan [1996 1998]) To do so we run anOLS panel regression relating indicators for each type of deregula-tion to potential consequences Our dependent variables includeproxies for economic growth (growth in personal income) bankstructure (the log of the number of banking companies and thenumber of branches per banking company) and efficiency (the logof total noninterest expenses and the ratio of loan loss provisionsto total loans) The four deregulation indicators (MBHC branch-ing through merger and acquisition unrestricted statewide branch-ing and interstate banking) are equal to one in the years afterderegulation and zero before If a state deregulated before thesample period begins in 1970 the indicator is one throughout ifthe state does not deregulate then the indicator is zero through-out All regressions include state xed effects and time effects
Table VI shows the results of including the set of four
TABLE VIMULTIVARIATE REGRESSIONS OF THE EFFECTS OF DIFFERENT TYPES OF BRANCHING
DEREGULATION ON ECONOMIC GROWTH BANKING STRUCTUREAND BANK EFFICIENCY
Dependent variable
Growthin
personalincome
Log of of
bankingcompanies
Branchesper
bankingcompany
Log ofnon-
interestexpenses
Loan lossprovisions
(loans 1leases)
Merger amp acquisitionbranchingindicator
0013(0004)
2 0162(0037)
1146(0453)
2 0107(0035)
2 0007(0001)
Unrestricted state-wide branchingindicator
2 0008(0005)
0086(0059)
2 0043(0574)
0070(0037)
0005(0002)
Interstate bankingindicator
0006(0005)
2 0052(0057)
1064(0511)
0005(0030)
2 0001(0001)
Multibank holdingcompany indicator
2 0005(0006)
2 0096(0052)
2 0364(0529)
2 0095(0036)
2 0001(0001)
Dependent variablemean [median]
0086[0082]
4881[5136]
6649[5055]
13812[13803]
0009[0006]
This table contains regressions of state-level economic growth and state-level measures of the structureand efficiency of the banking industry Each dependent variable is regressed on a set of state indicatorvariables a set of year indicator variables and the set of four deregulation indicators these deregulationindicators are equal to one after deregulation and zero before If a state was always regulated the indicator iszero throughout if it was always deregulated the indicator is one throughout Sample period is 1975 to 1991and Delaware and South Dakota are dropped so N 5 833 Noninterest expenses become available only in1984 so N 5 392 The coefficient on the deregulation indicators is reported with robust standard errors inparentheses and denote statistically signicant at the 1 5 and 10 percent levels
QUARTERLY JOURNAL OF ECONOMICS1464
deregulation indicators simultaneously The branching by mergerand acquisition indicator is statistically signicant and of theexpected sign in all ve of the equations and larger in absolutevalue than any of the other coefficients on the indicators for theother types of deregulation Since there may be multicollinearityamong our indicators we also ran univariate specications whichincluded each indicator separately and the results are verysimilar These results imply that branching by merger andacquisition is the most important type of deregulation measuredby its consequences for economic growth banking structure andbank efficiency See Kroszner and Strahan [1998] for more details
UNIVERSITY OF CHICAGO GRADUATE SCHOOL OF BUSINESS AND NATIONAL BUREAU OF
ECONOMIC RESEARCH
FEDERAL RESERVE BANK OF NEW YORK RESEARCH amp MARKET ANALYSIS GROUP
REFERENCES
Abrams Burton A and Russell F Settle lsquolsquoPressure-Group Inuence and Institu-tional Change Branch-Banking Legislation during the Great DepressionrsquorsquoPublic Choice LXXVII (1993) 687ndash705
Amel Dean lsquolsquoState Laws Affecting the Geographic Expansion of CommercialBanksrsquorsquo unpublished manuscript Board of Governors of the Federal ReserveSystem 1993
Becker Gary S lsquolsquoA Theory of Competition among Pressure Groups for PoliticalInuencersquorsquo Quarterly Journal of Economics XCVIII (1983) 371ndash400
Berger Allen Anil Kashyap Joseph Scalise lsquolsquoThe Transformation of the U SBanking Industry What a Long Strange Trip Itrsquos Beenrsquorsquo Brookings Papers onEconomic Activity 2 (1995) 55ndash218
Berger Allen Anthony Saunders Joseph Scalise and Gregory Udell lsquolsquoThe Effectsof Bank Mergers and Acquisitions on Small Business Lendingrsquorsquo Journal ofFinancial Economics L (1998) 187ndash229
Brennan Geoffrey and James Buchanan lsquolsquoTowards a Tax Constitution forLeviathanrsquorsquo Journal of Public Economics VIII (1977) 255ndash273
Brickley James A and Christopher M James lsquolsquoThe Takeover Market CorporateBoard Composition and Ownership Structure The Case of Bankingrsquorsquo Journalof Law and Economics XXX (1987) 161ndash180
Cole Rebel and John Wolken lsquolsquoFinancial Services Used by Small BusinessesEvidence from the 1993 Survey of Small Business Financesrsquorsquo Federal ReserveBulletin LXXXI (1994) 629ndash676
Dixit Avinash The Making of Economic Policy A Transactions-Cost PoliticsPerspective (Cambridge MA MIT Press 1996)
Economides Nicolas R Glenn Hubbard and Darius Palia lsquolsquoThe Political Economyof Branching Restrictions and Deposit Insurancersquorsquo Journal of Law andEconomics XXIX (1996) 667ndash704
Flannery Mark lsquolsquoThe Social Costs of Unit Banking Restrictionsrsquorsquo Journal ofMonetary Economics XIII (1984) 237ndash249
Greene William H Econometric Analysis third edition (Upper Saddle River NJPrentice-Hall 1997)
Gunther Jeffery lsquolsquoRegional Capital Imbalances and the Removal of InterstateBanking Restrictionsrsquorsquo Economics Letters XLIV (1994) 439ndash442 lsquolsquoThe Erosion of Intrastate Branch Banking Restrictionsrsquorsquo unpublishedmanuscript Federal Reserve Bank of Dallas 1996
Irwin Douglas and Randall Kroszner lsquolsquoInterests Institutions and Ideology inSecuring Policy Change The Republican Conversion to Trade Liberalizationafter Smoot-Hawleyrsquorsquo Journal of Law and Economics XLII (1999) 643ndash673
WHAT DRIVES DEREGULATION 1465
James Christopher lsquolsquoAn Analysis of Intra-Industry Differences in the Effect ofRegulation The Case of Deposit Rate Ceilingsrsquorsquo Journal of Monetary Econom-ics XII (1983) 417ndash442
Jarrell Gregg lsquolsquoChange at the Exchange The Causes and Effects of DeregulationrsquorsquoJournal of Law and Economics XXVII (1984) 273ndash312
Jayaratne Jith and Philip E Strahan lsquolsquoThe Finance-Growth Nexus Evidencefrom Bank Branch Deregulationrsquorsquo Quarterly Journal of Economics CI (1996)639ndash670
Jayaratne Jith and Philip E Strahan lsquolsquoEntry Restrictions Industry Evolutionand Dynamic Efficiency Evidence from Commercial Bankingrsquorsquo Journal of Lawand Economics XLIX (1998) 239ndash274
Joskow Paul lsquolsquoComment on lsquoPower Struggles Explaining Deregulatory Reformsin Electricity Marketsrsquo rsquorsquo Brookings Papers on Economic Activity Microeconom-ics (1996) 251ndash264
Joskow Paul and Roger Noll lsquolsquoRegulation in Theory and Practice An Overviewrsquorsquoin Studies in Public Regulation Gary Fromm ed (Cambridge MA MITPress 1981) pp 1ndash65
Kahn Matthew E and John G Matsusaka lsquolsquoDemand for Environmental GoodsEvidence from Voting Patterns on California Initiativesrsquorsquo Journal of Law andEconomics XL (1997) 137ndash173
Kalbeisch J and R Prentice The Statistical Analysis of Failure Time Data (NewYork John Wiley amp Sons 1980)
Kane Edward J lsquolsquoDe Jure Interstate Banking Why Only NowrsquorsquoJournal of MoneyCredit and Banking XXVIII (1996) 141ndash161
Kiefer Nicholas lsquolsquoEconomic Duration Data and Hazard Functionsrsquorsquo Journal ofEconomic Literature XXVI (1988) 646ndash679
Kroszner Randall S lsquolsquoThe Evolution of Universal Banking and Its Regulation inTwentieth Century Americarsquorsquo in Universal Banking Financial System DesignReconsidered Anthony Saunders and Ingo Walter eds (New York IrwinProfessional Publishers 1996) pp 70ndash99 lsquolsquoThe Political Economy of Banking and Financial Regulation in the U Srsquorsquo inThe Banking and Financial Structure in the NAFTA Countries and ChileGeorge M von Furstenberg ed (Boston Kluwer Academic Publishers 1997)pp 200ndash213
Kroszner Randall S and Raghuram G Rajan lsquolsquoIs the Glass-Steagall ActJustied A Study of the U S Experience with Universal Banking before1933rsquorsquo American Economic Review LXXXIV (1994) 810ndash832
Kroszner Randall S and Raghuram G Rajan lsquolsquoOrganization Structure andCredibility Evidence from Bank Securities Activities before the Glass-Steagall Actrsquorsquo Journal of Monetary Economics XXXIX (1997) 475ndash516
Kroszner Randall S and Philip E Strahan lsquolsquoRegulatory Incentives and the ThriftCrisis Dividends Mutual-to-Stock Conversions and Financial DistressrsquorsquoJournal of Finance LI (1996) 1285ndash1320
Kroszner Randall S and Philip E Strahan lsquolsquoWhat Drives Deregulation Econom-ics and Politics of the Relaxation of Bank Branching Restrictionsrsquorsquo NBERWorking Paper No 6637 1998
Kroszner Randall S and Thomas Stratmann lsquolsquoInterest Group Competition andthe Organization of Congress Theory and Evidence from Financial ServicesPolitical Action Committeesrsquorsquo American Economic Review LXXXVIII (1998)pp 1163ndash1187
Kroszner Randall S and Thomas Stratmann lsquolsquoDoes Political Ambiguity PayCorporate Campaign Contributions and Rewards to Legislator Reputationrsquorsquounpublishedmanuscript University of Chicago Graduate School of Business1999
Lin D Y and L J Wei lsquolsquoThe Robust Inference for the Cox Proportional HazardsModel Journal of the American Statistical Association LXXXIV (1989)1074ndash1078
Macey Jonathan R and Geoffrey P Miller Banking Law and Regulation (BostonLittle Brown and Company 1992)
Makinson Larry Open Secrets The Encyclopedia of Congressional Money andPolitics (Washington DC Congressional Quarterly 1992)
Niskanen WilliamBureaucracy and Representative Government (Chicago Aldine-Atherton 1971)
QUARTERLY JOURNAL OF ECONOMICS1466
Noll Roger lsquolsquoComment on Peltzmanrsquorsquo Brookings Papers Microeconomics SpecialIssue (1989) 48ndash58
North Douglass lsquolsquoA Transactions Cost Theory of Politicsrsquorsquo Journal of TheoreticalPolitics II (1990) 355ndash367
Peltzman Sam lsquolsquoToward a More General Theory of Regulationrsquorsquo Journal of Lawand Economics XIX (1976) 109ndash148 lsquolsquoConstituent Interest and Congressional Votingrsquorsquo Journal of Law andEconomics XXVII (1984) 181ndash210 lsquolsquoThe Economic Theory of Regulation after a Decade of DeregulationrsquorsquoBrookings Papers Microeconomics Special Issue (1989) 1ndash41
Peterson Mitchell and Raghuram Rajan lsquolsquoThe Benets of Lending RelationshipsEvidence from Small Business Datarsquorsquo Journal of Finance XLIX (1994) 3ndash38
Poole Keith and Howard Rosenthal Congress A Political-Economic History ofRoll Call Voting (Oxford Oxford University Press 1997)
Stata Stata Reference Manual Release 5 (College Station TX Stata Press 1997)Stigler George J lsquolsquoThe Theory of Economic Regulationrsquorsquo Bell Journal of Economics
and Management Science II (1971) 3ndash21Stigler George J Chicago Studies in the Political Economy (Chicago University
of Chicago Press 1988)Strahan Philip E and James Weston lsquolsquoSmall Business Lending and the Changing
Structure of the Banking Industryrsquorsquo Journal of Banking and Finance XXII(1998) 821ndash845
Sylla Richard John Legler and John Wallis lsquolsquoBanks and State Public Finance inthe New Republic The United States 1790ndash1860rsquorsquo Journal of EconomicHistory XLVII (1987) 391ndash403
White Eugene The Regulation and Reform of the American Banking System1900ndash1929 (Princeton Princeton University Press 1983)
Winston Clifford lsquolsquoEconomic Deregulation Days of Reckoning for Microecono-mistsrsquorsquo Journal of Economic Literature XXXI (1993) 1263ndash1289
WHAT DRIVES DEREGULATION 1467
from branching deregulation in particular and bank consolidationin general the private-interest theory would predict that stateswith numerous small bank-dependent rms would deregulateearlier12 This prediction however is also consistent with thepublic-interest theory since the social costs of the restrictions arehigher in states with more small bank-dependent rms
Another effect of branching restrictions on bank customerscan be related to the prices paid for bank services High loanprices before deregulation may reect for example a largemarket share for inefficient high-cost banks or high rents beingearned by the banks protected from competition In either casethe public-interest theory would suggest that states with rela-tively high initial loan interest rates should deregulate earlierThe private-interest theory however does not have a clearimplication high rents could lead the beneciaries to ght harderto maintain them (eg Stigler [1971]) or the deadweight costsassociated with these high rents could make it more difficult forthe beneciaries to maintain the restrictions (eg Becker [1983])13
Bank Stability Geographic diversication through branch-ing could mitigate instability problems that were importantduring much of our sample period thereby improving welfareInstability also may reduce the incentives of banks to lobby tomaintain protections because unstable banks are less likely tosurvive to reap the benets of the restrictions politicians also mayanticipate that such banks are less likely to be able to providefuture contributions and support [Gunther 1994 1996]14 Both theprivate-interest and public-interest theories thus suggest thatderegulation is most likely in states where banking instability isgreatest
12 On the other hand local banking monopolies created by branchingrestrictions could strengthen relationshipsbetween banks and small and medium-sized rms and increase the availability of credit to these rms [Petersen andRajan 1994] Also some have argued that small business lending declines whenlarge banks take over small banks (eg Berger Kashyap Scalise [1995])
13 High loan interest rates also could reect compensation for higher risksand fewer opportunities for banks to diversify If state-specic risks explaindifferences in average loans rates then neither theory would have any predictionsfor the sign of this effect Following branching deregulation however loan interestrates tend to fall but bank risk proles do not appear to change [Jayaratne andStrahan 1998 Kroszner and Strahan 1998] This suggests that high loan rates inantibranching states are not simply reecting state-specic risks
14 Also deregulation can arise as a response to banking instability under atheory of ignorant or misinformed voters in which a banking crisis acts as aneducational device to make the previously ignorant public aware of the costs of theantibranching policy (see Kane [1996])
WHAT DRIVES DEREGULATION 1445
Political-Institutional Factors Republicans are typically per-ceived as more likely to favor deregulation than Democrats so thepolitical-institutional theories suggest that states that are con-trolled by Democrats deregulate later than those controlled byRepublicans15 In addition we investigate whether reform be-comes more likely when one party controls the legislature and thegovernorship Note that these political effects must be interpretedwith caution since the views of the politicians may simply reectthe economic interests of the constituents in the state (seePeltzman [1984])
B Variable Denitions and Data Sources
Our main proxy for the strength of the small banks is thefraction of banking assets in the state in lsquolsquosmallrsquorsquo banks We denesmall banks as banks with assets below the median size in eachstate By allowing the denition of small to vary across states wetake into account cross-state heterogeneity in bank sizes16 An-nual data on bank size are from the fourth quarter Reports ofIncome and Conditions (lsquolsquoCall Reportsrsquorsquo) from the Federal ReserveBoard We also include the capital-to-asset ratio of small banksrelative to large as a measure of their relative nancial healthSpecically this variable is the asset-weighted average capital-asset ratio for small banks minus the asset-weighted averagecapital-asset ratio for large banks Annual data on bank capitalare from the Call Reports
To measure the effects of the rival insurance industry we rstconstruct an indicator variable that is one if the state permitsbanks to sell insurance For each state we then measure the sizeof the insurance sector (total value added in the state) relative tothe sum of the banking plus insurance sectors We will examine
15 For more detail on the importance of legislative structures party politicsand ideology see Poole and Rosenthal [1997] Kahn and Matsusaka [1997] Irwinand Kroszner [1999] and Kroszner and Stratmann [1999]
16 We also considered a variety of other denitions of relative size but therelative size results reported below are not sensitive to which denition we useFirst we used a xed measure of small bank that did not vary across statesdening small banks as those with assets below $100 million in 1994 dollarsSecond we applied the state-varying and xed denitions at the level of lsquolsquobankingorganizationsrsquorsquo (which include multibank holding companies) instead of at the levelof banks Third we calculated gini coefficients of bank size inequality and bankconcentration indices as alternative proxies for small versus large bank power inthe state Finally we also considered banks in lsquolsquoruralrsquorsquo areas (that is not located in aBureau of the Census lsquolsquostandard metropolitan statistical arearsquorsquo) as small since thevalue of the restrictions might be greatest in protecting banks outside of cities fromentry by the city banks All of the alternative denitions are highly correlated andyield the same results on the importance of intraindustry differences
QUARTERLY JOURNAL OF ECONOMICS1446
the effect separately for states that permit banks to sell insuranceand those that do not Data on value-added by industry are fromU S Commerce Department Bureau of Economic Analysis Sur-vey of Current Business [August 1994]
To measure the relative importance of small bank-dependentborrowers we include the proportion of all establishments operat-ing in the state with fewer than twenty employees These data arecompiled by the Bureau of the Census17 As a measure of priceswe use the difference between the average interest rate on loans inthe state and the prevailing federal funds rate where the loanrate equals the ratio of total interest income on all domestic loansdivided by total domestic loans held by banks operating in thestate18 These data are from the end-of-year Call Reports butbecome available beginning only in 1976 Our proxy for bankinstability is the failure rate of banks measured as the percent oftotal state banking assets in failed institutions from the FederalReserve Boardrsquos National Information Center database19
We include two political variables to adjust for any indepen-dent inuence of party politics First we measure the degree ofparty control of the state government by the fraction of the threebodies of the state government (the assembly senate and gover-norship) controlled by Democrats This variable for example isone-third if the Democrats have a majority in the assembly andthe Republicans have a majority in the senate and hold thegovernorship Second we include an indicator variable equal toone if the same party controls the governorrsquos office and hasmajorities in both chambers of the state legislature20
17 We collected the establishment data by the state for three cross sections(1976 1982 and 1987) and interpolated the data in the intermediate years SeeState and Metropolitan Data Book 1982 1986 and 1991
18 A measure of how individual customers would be affected is interest ratesthey receive on deposits We calculated the average deposit interest rate in thestate from the end-of-year Call Reports and did not nd any relationship betweenthis variable and the timing of deregulation in our analysis below Since Jayaratneand Strahan [1998] nd that branching deregulation has no effect on deposit ratesour nding of no relationship between timing and rates result could be due toeither individual bank customers anticipating little pricing effect and not lobbyingor an inability of individual depositors to form an effective lobby
19 As alternative measures of distress we used (1) the bank failure ratemeasured as the number of failed institutions in the state relative to the totalnumber of banks (2) the overall capital-to-asset ratio of all banks in the state and(3) the growth in state personal income Our results are not sensitive to the choiceof distress proxy
20 We also estimated the models with three indicator variables to reect theparty control of the assembly senate and governorship separately The coefficientestimates on these variables are qualitatively similar to those we present below(that is Democratic control tends to slow deregulation) and the results on theinterest group variables are unaffected by the specication of the political
WHAT DRIVES DEREGULATION 1447
IV METHODS AND RESULTS
In this section we rst estimate how the variables describedabove inuence the timing of state-level branching deregulationusing a hazard model Next we discuss the robustness of theresults including a test of whether the ex post impact of deregula-tion on the different interest groups is consistent with ourinterpretation of the coefficients We then investigate federalinterstate branching deregulation with a probit model to explorewhether the factors that affect the timing of intrastate deregula-tion also inuence voting in the House of Representatives oninterstate deregulation
A Hazard Model
We consider the period from the beginning of our sample(1970) until deregulation as the lsquolsquoduration of regulationrsquorsquo or thelsquolsquotime until deregulationrsquorsquo The hazard rate h(t) is the likelihoodthat a state deregulates at time t given that the state has not yetderegulated [Kalbeisch and Prentice 1980 Kiefer 1988 Greene1997] To model the duration of regulation we must decide whatstructure if any to impose on the hazard function The Kaplan-Meier product-limit estimator provides a simple nonparametricway to estimate the shape of the hazard function over time[Greene 1997]21 Figure II graphs this estimate for our data andshows that the hazard function is relatively at in the early yearsand then grows steeper in the later years
A duration model that can approximate this shape is theWeibull proportional hazards model The hazard rate functiontakes the form
h[t x(t) b] 5 h0(t) exp[x(t)8b]
where x(t) is a vector of time-varying covariates b is a vector ofunknown parameters to be estimated and the baseline hazard
variables Note also that Nebraska has a nonpartisan unicameral legislature Weassume that the party of the Governor controls the state government for Nebraskaand set the unied government indicator to one As another alternative we triedincluding a variable that is one if party control of the legislature or governorshipchanges but it did not have an economically or statistically signicant effect anddid not affect any of the other results
21 Denote each of the K years in our sample period Tk and order them suchthat T1 T2 and so on Let nk be the number of states that have not yet deregulatedby Tk and dk be the number of states that deregulate in year Tk The Kaplan-Meierestimate of the hazard rate in each year is dknk In other words the hazard rate isthe ratio of the number of states that actually deregulate in each year to thenumber of states that have not yet deregulated in that year
QUARTERLY JOURNAL OF ECONOMICS1448
rate h0(t) is ptp 2 1 with shape parameter p that will be estimatedfrom the data When p 1 this model displays a monotonicallyincreasing hazard rate To estimate the parameters we maximizethe following log-likelihood function [Kiefer 1988]
L(b) 5 oi 5 1
N
di ln h[ti x(ti) b] 2 oi 5 1
N
e 0
tih[u x(u) b] du
where N indexes the number of states at risk at the beginning ofthe sample period di is an indicator equal to one for the statesthat deregulated by the end of the sample period and zero for thestates that do not deregulate (ie the censored observations) andti is the time of deregulation for the ith state This formulationassumes that the time until deregulation is independent acrossstates but the observations of the same state over time are notassumed to be independent In addition we use a robust estima-tion procedure to calculate the standard errors [Lin and Wei1989]
We exclude states that deregulated before 1970 from theanalysis so N equals 39 states During the sample period 36 ofthe 39 states deregulate We have three censored observations forwhich di is zero (Arkansas Iowa and Minnesota) but our results
FIGURE IIKaplan-Meier Estimate of Hazard Rate for Branching Deregulation
WHAT DRIVES DEREGULATION 1449
are not sensitive to the inclusion of these censored observationsSince we observe each state in each year up to and including theyear of deregulation we have a total of 637 observations Table IIreports summary statistics and correlation coefficients for theexplanatory variables we use in our analysis
We use the Weibull model rather than a model that does notimpose any structure on the baseline hazard rate so that we cancalculate the change in the expected time to deregulation for agiven change in the levels of the covariates22 In the Weibullmodel we can invert the hazard function and map it into the timedomain Rewriting the Weibull model in this way the log of thetime to deregulation T is a linear function of the political economyfactors and an error term ln (T ) 5 x8b 1 e23 Because we assumethat the baseline hazard rate is Weibull with a shape characterizedby p the new coefficients will be scaled by p that is b 5 2 bp
The b coefficients represent the percentage change in thetime to deregulation for a one-unit change in the covariates Apositive coefficient implies that an increase in the variableincreases the expected time until deregulation To gauge theeconomic importance of the effects we will multiply the b by thestandard deviation of the explanatory variables and then evaluatehow much the changes in the variables raise or lower the expectedtime to deregulation24 In Table III we report the b coefficients
22 The Cox proportional hazards model does not impose any structure on thebaseline hazard rate h0(t) and takes the following form
h[t x(t) b] 5 h0(t) exp [x(t)8b]
Common factors that affect the probability of deregulation in the same way for allstates therefore do not inuence the estimated coefficients The cost of the Coxmodel is that because there is no structure on the baseline we can calculatechanges in only relative hazard rates associated with changes in the covariatesThe Weibull model however provides sufficient structure so that we can translateour estimates into a log expected time metric that allows us to calculate the changein the expected time to deregulation for a given change in the covariates SeeKiefer [1988] Greene [1997] and Stata [1997] When we compare the coefficientsestimated by the Cox and Weibull models on our data they are very close (seeKroszner and Strahan [1998]) so imposing a Weibull structure does not appear todistort our estimates
23 The error term e is independent of x and has an extreme value distributionscaled by 1p (see Kiefer [1988] Greene [1997] and Stata [1997] for more details onthis log expected time parameterization of the Weibull model)
24 The transformation to the log expected time domain is only strictly correctwhen the covariates are constant over time Our interpretation of the coefficients ispossible in the time-varying model if we assume that a one-standard-deviationdifference would hold over the entire sample period that is the deregulationwould occur so many years earlier or later if the variable in question were higherby one standard deviation from 1970 to 1992 Note that the mean number of yearsto deregulation in the sample is 158
QUARTERLY JOURNAL OF ECONOMICS1450
TABLE IISUMMARY STATISTICS FOR VARIABLES IN HAZARD MODEL 1970ndash1992
Panel A Univariate statistics MeanStandarddeviation Minimum Maximum
(1) (2) (3) (4)Small bank asset share of all
banking assets in state 011 005 001 023Relative size of insurance to
banking plus insurance inthe state 047 009 024 080
Indicator is 1 if banks can sellinsurance in the state 020 040 0 1
Small rm share of the numberof rms in the state 088 002 078 095
Capital ratio of small banksrelative to large in the state 002 001 2 001 009
Share of state government con-trolled by Democrats 065 036 0 1
Indicator is 1 if state controlledby one party 054 050 0 1
Average yield on bank loans inthe state minus Fed fundsrate 002 002 2 004 008
Panel B CorrelationsSmallbank
Relativesize of
insuranceInsuranceindicator
Smallrm
Smallbank
capitalDem
controlSingleparty
(1) (2) (3) (4) (5) (6) (7)Small bank asset share
of all banking assetsin state 1
Relative size of insur-ance to banking plusinsurance in the state 2 034 1
Indicator is 1 if bankscan sell insurance inthe state 020 2 012 1
Small rm share of thenumber of rms inthe state 030 2 011 005 1
Capital ratio of smallbanks relative tolarge in the state 2 066 020 2 010 2 006 1
Share of state govern-ment controlled byDemocrats 2 006 001 2 018 2 018 005 1
Indicator is 1 if statecontrolled by oneparty 007 012 001 2 006 2 006 030 1
Average yield on bankloans in the stateminus Fed funds rate 2 001 2 015 003 022 2 010 2 003 2 006
The number of observations is 637 except for the average interest rate on bank loans which becomesavailable only in 1976 so the number of observations is 408
TABLE IIIHAZARD MODEL OF POLITICAL ECONOMY FACTORS AFFECTING THE TIMING OF STATE
BRANCHING DEREGULATION 1970ndash1992
(1) (2) (3) (4) (5) (6)Small bank asset share of
all banking assets instate
648(122)
519(137)
632(135)
516(118)
488(133)
445(088)
Capital ratio of smallbanks relative to largein the state
1325(348)
921(319)
1301(362)
1067(346)
1144(375)
906(381)
Relative size of insurancein states where banksmay sell insurance 0otherwise
324(161)
259(149)
315(165)
285(154)
272(159)
055(134)
Indicator is 1 if banksmay sell insurance inthe state
045(010)
037(011)
045(010)
038(008)
036(009)
001(012)
Relative size of insurancein states where banksmay not sell insurance0 otherwise
2 093(043)
2 090(033)
2 102(051)
2 093(048)
2 096(053)
2 004(039)
Small rm share of thenumber of rms in thestate
2 972(211)
2 614(215)
2 952(211)
2 943(230)
2 952(270)
2 1509(276)
Share of state govern-ment controlled byDemocrats
031(012)
023(011)
030(012)
027(011)
026(011)
011(013)
Indicator is 1 if state con-trolled by one party
2 004(010)
2 004(007)
2 004(007)
2 001(009)
2 002(010)
017(009)
Average yield on bankloans in the stateminus Fed funds rate mdash
023(319) mdash mdash mdash mdash
Bank failure rate mdash mdash092
(111) mdash mdash mdash
Indicator is 1 if state hasunit banking law mdash mdash mdash
018(010)
019(010)
030(009)
Indicator is 1 if statechanges bank insur-ance powers mdash mdash mdash mdash
2 008(012)
2 026(013)
Includes regionalindicators No No No No No Yes
N 637 408 637 637 637 637Log likelihood 2 374 900 2 360 2 212 2 191 1192p-value of x 2 for
regression 001 001 001 001 001 001
The hazard model is Weibull where the dependent variable is the log expected time to branchingderegulation ln (T) 5 bx 1 eAll variables are measured for each state in each year Small bank asset share isthe percent of banking assets in the state held by banks below the median size of banks in each state in eachyear Relative capital ratio is the capital to assets ratio of small banks minus that of large banks Size ofinsurance relative to banking plus insurance in the state is measured as gross state product from insurancedivided by gross state product from insurance plus banking Insurance indicator is 1 if state law permits banksto enter the insurance business Small rm share is the percent of all establishments in the state that havefewer than twenty employees Party control variable is the share of the three bodies of state governmentcontrolled by Democrats One party control indicator is 1 if the same party controlrsquos the governorship and hasmajorities in both chambers of the state legislature Average yield on bank loans equals total interest incomeon all domestic loans made by banks in the state divided by total loans Bank failure rate is the ratio of assetsin failed banks during a given state and year divided by total bank assets in that state and year Unit bankingindicator equals 1 for states with unit banking restrictions The change in insurance powers indicator equalsone if the state changed the law to permit banks to sell insurance during the sample period We divide thecountry into four regions and include regional indicators in column (5) Robust standard errors are inparentheses and denote statistically signicant at the 1 5 and 10 percent levels
QUARTERLY JOURNAL OF ECONOMICS1452
The positive and statistically signicant coefficient on smallbank share implies that a greater small bank share delaysderegulation and this effect is economically important A one-standard-deviation increase in the small bank share results in a30 percent increase in the time until deregulation or about 47years This result supports the implications of private-interesttheory concerning intraindustry rivalry and is contrary to thepublic-interest theory
The positive and statistically signicant coefficient on therelative performance of small banks in the state implies thatstates deregulate later when small banks are relatively strongnancially25 A one-standard-deviation increase in the relativecapital-to-asset ratio results in a 15 percent rise in the time untilderegulation or about 24 years This result is consistent with theprivate-interest theory but may also be consistent with a public-interest interpretation because the risk of failures (and a costlytaxpayer-funded bailout) increases as small bank nancial healthdeclines
Interindustry competition also helps explain the timing ofderegulation In states where banks can sell insurance a rela-tively large insurance sector is associated with an increase in theexpected time to deregulation26 A one-standard-deviation in-crease in the relative size of the insurance sector in those statesthat permit banks to sell insurance leads to a 22 percent increasein the time until deregulation or about 35 years27 This result isconsistent with the private-interest theory but not the public-interest theory
Deregulation occurs earlier in states where small bank-dependent rms are relatively numerous A one-standard-
25 We also allowed small bank and large bank capital-to-asset ratios to enterseparately Their coefficient estimates were nearly equal in absolute value and ofopposite signs and so the data do not reject the relative small to large bank capitalratio specication we have chosen
26 When we use the relative size of the insurance sector in the state as awhole rather than insurance relative to banking plus insurance in the state weobtain similar results When we simultaneously include both measures of therelative size of insurance the interactions of the lsquolsquoinsurance relative to bankingplus insurancersquorsquo variable continue to be statisticallysignicant but the interactionsof lsquolsquoinsurance relative to the state as a wholersquorsquo variable are not
27 We have also estimated the model with the indicator variable equal to onefor states where banks may sell insurance but without the variables measuringtheimportance of insurance relative to banking In this model the coefficient on theinsurance indicator variablemdashan estimate of the difference in the expected time toderegulation in states where banks may sell insurance relative to states wherebanks may not sell insurancemdashequals 027 and is statistically signicant at the 5percent level This suggests that branching deregulation takes about 45 yearslonger in states where banks may sell insurance
WHAT DRIVES DEREGULATION 1453
deviation increase in the share of small rms reduces the timeuntil deregulation by 18 percent or about three years This resultconcerning the interests of consumers of banking services isconsistent with both the private- and public-interest theories
Column (2) of Table III introduces the average interest rateon loans to the model as a rough proxy for the cost to bankborrowers of these regulations These data become available onlyin 1976 so our sample size shrinks to 408 observations Thecoefficient on this variable is small and not statistically signi-cant The public-interest theory would imply that the coefficientshould be negative while the implication of the economic theorydoes not have a clear sign prediction
When we add the failure rate of banks to the specication incolumn (3) its coefficient is small and statistically insignicantWe do not nd a linkage between the timing of state deregulationand statewide banking distress Given that deposit insurance isfunded federally however bank failures may raise the publicrsquosconcerns about the costs of the regulations at the national ratherthan the state level This interpretation can explain why the rateof bank failure does not affect the timing of state-level deregula-tion even though bank instability does appear to rise generallyduring the period of increasing deregulation (see Section V below)
Finally the partisan structure of the state government alsoinuences when states deregulate As expected a higher propor-tion of Democrats in the government tends to delay deregulationA one-standard-deviation rise in the share of the governmentcontrolled by Democrats slows the deregulation by about twoyears Whether the state is dominated by one party however doesnot appear to affect the timing of the deregulation
To summarize our ndings from the hazard model privateinterests appear to play an important role in the deregulatoryprocess While private interests and public interests do sometimescoincide the results on the relative share of small banks and largebanks and on the relative size of insurance where banks competeare consistent with a private-interest approach but are difficult toexplain on public-interest grounds
B Robustness Tests
Omitted Variables States begin the sample period withdifferent degrees of restrictions on bank expansion These initialconditions may proxy for omitted factors determining the political-economy equilibrium in the state that could then affect the
QUARTERLY JOURNAL OF ECONOMICS1454
subsequent timing of deregulation For instance some statesbegan the period with unit banking while other states allowedlimited branching and the formation of multibank holdingcompanies
To test whether the stringency of initial restrictions onbranching matters we rst divide the sample into states with andwithout the most extreme form of branching regulation unitbanking28 Column (4) of Table III includes an indicator variableequal to one if the state began the period with unit bankingrestrictions The coefficient on this variable is positive andstatistically signicant but the other results remain virtuallyunchanged Next we include an indicator variable that is one ifthe state permitted multibank holding companies (MBHCs) toexist at the beginning of our sample period Thirty-four statesbegan the period allowing MBHCs The coefficient estimate forthis indicator variable (not reported) is small and not statisticallysignicant and its inclusion does not change the effects of theother variables
Eleven states expand bank powers to include the sale ofinsurance products during our sample period This could proxy foromitted factors about the relative strength of the insuranceindustry in the state In column (5) we include an indicatorvariable that is one if the state changes the law during our sampleperiod to permit banks to sell insurance business and zerootherwise The coefficient estimate is small and statisticallyinsignicant and does not affect the other results29
In column (6) we introduce four regional indicator variablesNorth South Midwest and West30 The coefficients on small bankshare the relative capital-to-assets ratio and small rm sharechange little Although the coefficient estimate still suggests that
28 We classify states that prohibited branching but permitted banks toestablish facilities as unit banking states The unit banking indicator is one for thefollowing states CO AR FL IL IA KS MN MO MT NE ND OK TX WI WVand WY
29 We conducted a likelihood ratio test of whether the unit banking statesthe MBHC states and the states that changed their regulations of bank insurancepowers could be pooled with the other states In all three cases we could not rejectpooling of the data
30 Since intrastate deregulation generally preceded interstate deregulationand the latter typically took the form of regional interstate compacts thepossibility of participating in a regional interstate banking compact could haveinuenced the decision to deregulate intrastate branching Our denitions of theregions are region 1 (South) contains AL AR DC FL GA KY LA MS NC OKSC TN TX and VA region 2 (Northeast) contains CT MA MD ME NH NJ NYPA RI VT and WV region 3 (Midwest) contains IA IL IN KS MI MN MO NEND OH SD and WI region 4 (West) contains the other states
WHAT DRIVES DEREGULATION 1455
in states where banks can sell insurance a larger relative size ofthe insurance sector delays deregulation the effect is no longerstatistically signicant The coefficient on the indicator variablefor whether a state expands bank powers to permit insurancesales during the sample period however now becomes statisti-cally signicantly negative This implies that deregulation occursroughly 41 years earlier in states where the insurance lobby maybe relatively weak While the inclusion of the regional indicatorsdoes not affect the other results the results on insurance doappear to be sensitive to the specication of the hazard model31
Estimation Technique To check the robustness of using ahazard model and to investigate whether the time-series or thecross-sectional variation in each of the factors accounts for thehazard results we estimate two simple linear probability models(neither are reported) First we run a pooled time-series cross-section OLS regression using the 637 observations from thehazard model The dependent variable equals one in the year ofderegulation and zero otherwise We include the same set ofregressors that were used in the hazard model and add state-levelxed effects This specication thus removes the cross-statevariation from the independent variables and focuses on theirmovements over time In this model the signs of the coefficientson the interest group factors are consistent with the results fromthe hazard model and the coefficients for the small bank relativecapital ratio and for the relative size of insurance where states cansell insurance are statistically signicant
In the second model we eliminate the time-series variationby averaging each of the explanatory variables over time We thushave one observation per state that had branching restrictions asof 1970 (N 5 39) The dependent variable equals the inverse of thenumber of years until deregulation which represents an estimateof the statersquos probability of deregulating in a year under theassumption that this probability is constant over time and eachyear provides an independent observation32 Although we haveonly 39 observations the coefficients on the small bank marketshare small bank relative capital ratio and small rm share
31 As further robustness check we also tried a population density variablewhich Abrams and Settle [1993] found relevant for regulatory change during the1920s and 1930s but its coefficient was not statistically signicant and did notaffect the other results
32 For instance the dependent variable would take the value of 01 for astate that deregulated in 1980 For the three states that did not deregulate thedependent variable equals 0
QUARTERLY JOURNAL OF ECONOMICS1456
variables are statistically signicant The signs of the effects ofthe insurance variables are the same as those reported in thehazard model but not statistically signicant
The hazard model and linear probability models thus producevery similar results The estimates from the latter suggest thatthe effects for the small bank market share and small rm sharevariables are driven primarily by cross-state variation the effectof the insurance variable is driven primarily by time-seriesvariation and the small bank capital effect is driven by bothcross-sectional and time-series variation
Ex Post Consequences of Deregulation Finally to check theplausibility of our interpretation of the hazard results we con-sider whether the ex post consequences of deregulation areconsistent with the ex ante lobbying positions we attribute to eachinterest group (see Kroszner and Strahan [1998] for details) Wend that small banks lose market share following deregulationand in states where banks can enter the insurance business theinsurance sector shrinks relative to the banking sector followingderegulation Borrowers also benet from deregulation throughlower average interest rates on loans These ndings support theprivate-interest interpretation of the coefficients in the hazardmodel groups that will benet lobby to speed deregulation andthose that will be harmed lobby to slow it
C Voting on Interstate Branching Deregulation in the U SHouse of Representatives
We now examine whether the forces driving intrastate branch-ing deregulation also drive interstate deregulation Financialservices interests are active contributors and lobbyists in Washing-ton Their political action committees constitute the largest groupof contributors to legislators providing nearly 20 percent of totalcongressional campaign contributions [Makinson 1992] and muchof their lobbying effort involves competition among rival interestswithin nancial services (see Kroszner and Stratmann [1998])
As noted above after virtually all states adopted intra- andinterstate branching deregulation the 1994 Riegle-Neal Act re-pealed the 1927 McFadden Act to eliminate all barriers tointerstate banking and branching by 1997 The key votes concern-ing the Riegle-Neal Act were either voice votes or extremelylopsided so we could not estimate a voting model from them Anumber of bills and amendments related to interstate branching
WHAT DRIVES DEREGULATION 1457
had been debated in Congress during the years prior to thepassage of the Riegle-Neal Act but a search of the weekly BNABanking Reporter and the Congressional Record produced onlyone roll-call vote related to interstate branching that was notlopsided This vote occurred in the House on November 14 1991and concerned an amendment sponsored by Wylie (R-OH) andNeal (D-NC) to introduce interstate banking and branchingderegulation to a nancial services reform package33 Althoughthe amendment passed by 210 to 208 the bill to which it wasattached subsequently was defeated
To check for the inuence of the factors we considered in thestate-level reforms we examine both the sponsorship of interstatebanking legislation and voting on the amendment The sponsorsof the Wylie-Neal amendment are from states with low small bankshares 004 in Ohio (Wylie) and 002 in North Carolina (Neal) Incontrast the sample mean in 1991 is 008 (median 5 007)Michigan home state of the Senatersquos sponsor of the 1994 Riegle-Neal Act also had relatively low small bank strength (small bankshare of 005)
Table IV reports estimates of a probit model where thedependent variable equals one if the legislator voted in favor ofthe amendment and zero otherwise The explanatory variablesare the same as those in the hazard model except that we use aparty affiliation indicator variable equal to one for Democrats andzero for Republicans as our proxy for political factors Thecoefficients in Table IV are the marginal effects (lsquolsquoslopesrsquorsquo) of aone-unit change of each variable on the probability that alegislator will vote for the amendment34 Note that a forcefavoring deregulation has a negative coefficient in the hazardmodel (Table III) but a positive coefficient in the probit model
Consistent with the state-level deregulation process legisla-tors are more likely to support the amendment if their states havea relatively low share of small banks As in the hazard model thefraction of small banks is the most important interest groupinuence on a legislatorrsquos voting decision A one-standard-deviation increase in small banksrsquo market share (from the mean)is associated with a decline in the probability of voting in favor of
33 The Wylie-Neal amendment also included provisions limiting certaininsurance and real estate powers of national banks [Congressional RecordNovember 14 1991 pp 10239ndash10242]
34 Since we have multiple legislators from each state we adjust the standarderrors to correct for the lack of independence among observations clustered in thesame state
QUARTERLY JOURNAL OF ECONOMICS1458
TAB
LE
IVM
AR
GIN
AL
EF
FE
CT
SF
RO
MA
PR
OB
ITM
OD
EL
OF
TH
EIN
FL
UE
NC
EO
FP
OL
ITIC
AL
EC
ON
OM
YF
AC
TO
RS
ON
TH
EH
OU
SE
VO
TE
ON
TH
EW
YL
IE-N
EA
L
AM
EN
DM
EN
TT
OP
ER
MIT
INT
ER
ST
AT
EB
AN
KB
RA
NC
HIN
G
NO
VE
MB
ER
14
1991
(1)
(2)
(3)
(4)
(5)
(6)
Smal
lban
kas
set
shar
eof
allb
anki
ng
asse
tsin
stat
e2
417
(12
4)2
454
(12
2)2
407
(12
0)2
443
(14
7)2
450
(14
9)2
490
(14
7)C
apit
alra
tio
ofsm
allb
anks
rela
tive
tola
rge
inth
est
ate
180
(37
5)2
68(3
42)
256
(37
5)1
69(3
78)
153
(38
0)2
45(4
25)
Rel
ativ
esi
zeof
insu
ranc
ein
stat
esw
here
bank
sm
ayse
llin
sura
nce
0ot
herw
ise
22
34
(10
2)2
249
(1
14)
22
33
(10
1)2
249
(1
04)
22
51
(10
2)2
154
(11
1)In
dica
tor
is1
ifba
nks
may
sell
insu
ranc
ein
the
stat
e0
79
(0
06)
074
(00
6)0
69
(0
06)
070
(00
6)0
70
(0
06)
067
(00
7)R
elat
ive
size
ofin
sura
nce
inst
ates
whe
reba
nks
may
not
sell
insu
ranc
e0
oth
erw
ise
180
(06
6)2
42
(0
72)
154
(0
67)
173
(06
6)1
72
(0
67)
203
(0
91)
Smal
lrm
shar
eof
the
num
ber
ofr
ms
inth
est
ate
21
23(0
92)
20
83(0
89)
21
62(1
07)
21
32(0
95)
21
30(1
00)
20
94(1
16)
Indi
cato
ris
1if
Rep
rese
ntat
ive
isa
Dem
ocra
t2
035
(00
7)2
035
(00
7)2
036
(00
7)2
035
(00
7)2
035
(00
7)2
036
(00
7)A
vera
geyi
eld
onba
nklo
ans
inth
est
ate
min
usF
edfu
nds
rate
mdash10
01
(51
5)mdash
mdashmdash
mdash
Ave
rage
fail
ure
rate
198
5to
1991
mdashmdash
228
(14
1)mdash
mdashmdash
Indi
cato
ris
1if
stat
eh
asun
itba
nkin
gla
wmdash
mdashmdash
004
(00
7)0
04(0
07)
20
01(0
08)
Indi
cato
ris
1if
stat
ech
ange
sba
nk
insu
ran
cepo
wer
smdash
mdashmdash
mdash2
001
(00
7)2
003
(00
7)In
clud
esre
gion
alin
dica
tors
N
oN
oN
oN
oN
oYe
sP
seud
o-R
20
170
180
170
170
170
18L
ogli
keli
hood
224
024
223
826
223
951
224
013
224
010
223
808
p-va
lue
ofx2
for
regr
essi
on
001
0
01
001
0
01
001
0
01
Th
ede
pen
den
tvar
iabl
eis
one
ifth
ele
gisl
ator
vote
sfo
rth
eam
endm
enta
nd
zero
ifag
ains
tT
he
repo
rted
coef
fici
ents
are
the
effe
cts
ofa
uni
tch
ange
ofth
ein
depe
nde
ntva
riab
le(f
rom
the
mea
n)on
the
prob
abili
tyof
voti
ng
infa
vor
ofth
eam
endm
ent
All
vari
able
sar
em
easu
red
byst
ate
N5
418
the
num
ber
ofR
epre
sent
ativ
esvo
tin
gon
the
amen
dmen
tS
mal
lban
kas
sets
har
eis
the
perc
ent
ofba
nki
ng
asse
tsin
the
stat
eh
eld
byba
nks
belo
wth
em
edia
nsi
zeof
ban
kin
each
stat
ein
each
year
Rel
ativ
eca
pita
lrat
iois
the
capi
talt
oas
sets
rati
oof
smal
lban
ksm
inu
sth
atof
larg
eba
nks
Siz
eof
insu
ran
cere
lati
veto
bank
ing
plu
sin
sura
nce
inth
est
ate
ism
easu
red
asgr
oss
stat
epr
oduc
tfro
min
sura
nce
divi
ded
bygr
oss
stat
epr
odu
ctfr
omin
sura
nce
plu
sba
nkin
gIn
dica
tor
ison
eif
stat
ela
wpe
rmit
sba
nks
toen
ter
the
insu
ran
cebu
sin
ess
Sm
all
rmsh
are
isth
epe
rcen
tofa
lles
tabl
ishm
ents
inth
est
ate
that
hav
efe
wer
than
twen
tyem
ploy
ees
Par
tyin
dica
tor
ison
eif
the
Rep
rese
nta
tive
isa
Dem
ocra
tA
vera
geyi
eld
onba
nklo
ans
equa
lsto
tali
nter
est
inco
me
onal
ldom
esti
clo
ans
mad
eby
ban
ksin
the
stat
edi
vide
dby
tota
lloa
ns
The
aver
age
bank
failu
rera
teis
the
rati
oof
asse
tsin
faile
dba
nks
duri
ng
agi
ven
stat
edi
vide
dby
tota
lban
kas
sets
inth
atst
ate
aver
aged
over
the
year
s19
85to
1991
Un
itba
nki
ng
indi
cato
req
ual
son
efo
rst
ates
wit
hun
itba
nkin
gre
stri
ctio
nsT
hech
ange
inin
sura
nce
pow
ers
indi
cato
req
ual
son
eif
the
stat
ech
ange
dth
ela
wto
perm
itba
nks
tose
llin
sura
nce
duri
ngth
esa
mpl
epe
riod
We
divi
deth
eco
unt
ryin
tofo
urre
gion
san
din
clu
dere
gion
alin
dica
tors
inco
lum
n(5
)R
obu
stst
anda
rder
rors
are
inpa
ren
thes
es
an
d
den
ote
stat
isti
cally
sign
ica
nta
tth
e1
5an
d10
perc
entl
evel
s
WHAT DRIVES DEREGULATION 1459
branching of approximately 17 percent The relative capital-to-assets ratio of small and large banks however does not have astatistically signicant effect in the probit model
The impact of rival interests outside of banking is alsoconsistent with intrastate deregulation results Where banks cansell insurance legislators from states with larger insurancesectors relative to banking are less likely to vote for the amend-ment35 A one-standard-deviation increase in the relative size ofinsurance in those states which permit banks to sell insurancedecreases the probability that a legislator will favor the amend-ment by about 13 percent Turning to consumer interests thecoefficient on small rm share is not statistically signicant butthe coefficient on interest rates in the state is A one-standard-deviation increase in the average interest rate on loans raises theprobability that a legislator will support the amendment byroughly 8 percent The coefficient estimate for the bankingdistress variable measured as the average bank failure rate from1985 to 1991 is small and statistically signicant Overall theprobit analysis of the vote on national branching deregulationsupports the private-interest theory of deregulation and providesa consistency check that the importance of interests operating onthe state legislatures are very similar to those operating at thefederal level
V WHY DID BRANCHING DEREGULATION BEGIN DURING THE
1970S SHOCKS TO THE EQUILIBRIUM SUPPORTING BANK
BRANCHING RESTRICTIONS
A complete explanation of regulatory exit should account forwhy deregulation begins in the 1970s as well as the timing of thestate-by-state reforms [Kane 1996] In this section we describebroad technological legal and economic shocks that altered thepolitical-economy equilibrium which had kept antibranching regu-lations little changed for at least 30 years
Beginning in the 1970s three innovations reduced the valueto the protected banks of local geographic monopolies (see TableV) First automatic teller machines (ATMs) helped to erode thegeographic ties between customers and banks Second checkablemoney market mutual funds and the Merrill Lynch Cash Manage-ment Account demonstrated that banking by mail and telephone
35 The positive and statistically signicant estimates on the other insurancevariables may reect insurance industry support for the amendmentrsquos provisionslimiting banksrsquo insurance powers
QUARTERLY JOURNAL OF ECONOMICS1460
provided a convenient alternative to local banks36 Third techno-logical innovation and deregulation reduced transportation andcommunication costs particularly since the 1970s thereby lower-ing the costs for customers to use distant banks By increasing theelasticity of deposits supplied to banks these innovations reducedthe value of geographical restrictions to their traditional benecia-ries and thereby reduced their incentive to ght to maintain them[Peltzman 1976]
On the lending side increasing sophistication of credit-scoring techniques following innovations in information process-ing nancial theory and the development of large credit databases began to diminish the value of knowledge that localbankers had about the risks of borrowers in the community As a
36 Regulation Q which limited the interest rates that banks could pay ondeposits may have helped to drive depositors away from banks when the gapbetween market rates and deposit ceilings grew during the 1970s The eliminationof interest rate ceilings on large denomination certicates of deposit during the1970s appears to have hurt smaller and retail-oriented banks relative to largerwholesale banks [James 1983]
TABLE VBROAD TRENDS IN COMMERCIAL BANKING 1950ndash1995
YearNumberof ATMs
Domesticbank
deposits(billions)
Moneymarketmutual
fund(billions)
Percent ofdeposits 1
moneyfunds heldby banks
Smallbanksrsquo
percent ofbankingassets
Averagenumber of
bankfailures
(1) (2) (3) (4) (5) (6)1950 0 $ 154 $ 0 100 NA 41955 0 191 0 100 NA 31960 0 228 0 100 24 21965 0 330 0 100 20 41970 0 479 0 100 18 61975 9750 775 4 99 18 61980 18500 1182 76 94 17 101985 61117 1787 242 88 14 601990 80156 2339 493 83 11 1791995 122706 2552 745 77 8 61
Column (1) ATM gures are from Bank Network News The EFT Network Data Book (New YorkFaulkner and Gray Inc) The 1975 gure was unavailable 9750 is the number of ATMs in 1978 the rst yearfor which complete data are available Columns (2)ndash(4) Banksrsquo domestic deposits are from the Reports ofIncome and Condition money market mutual funds are from the Flow of Funds Data on all bank depositsforeign plus domestic are only available beginning in 1970 The trend in banksrsquo share (column (4)) is the sameusing total deposits instead of domestic deposits Column (5) Percent of banking assets held by small bankswhere a small bank is dened as a commercial bank with less than $100 million in assets in 1994 dollarsThese data are based on the Reports of Income and Condition Data on small banks are not available before1960 Column (6) Five-year average number of bank failures where the nal year is indicated in the rstcolumn These data are from FDIC Annual Report and the Quarterly Banking Prole
WHAT DRIVES DEREGULATION 1461
result of these innovations a national market developed forresidential mortgages credit card receivables have been securi-tized and bank lending to small business now relies less on thejudgment of loan officers and more on standardized scoringmodels
These changes have increased the potential protability forlarge banks to enter what had been the core of small bankactivities Large banksrsquo incentive to increase their lobbying pres-sure to be able to expand into these markets has thus beenincreasing over time In fact small banksrsquo market share began todecline even prior to the branching deregulation (Table V) As thevalue of a local banking relationships declined small rms thatwere the main borrowers from the small banks also probablybecame more likely to favor the entry of large banks into localmarkets With the deadweight costs of preventing large bankentry rising the private-interest theory predicts that small localbanks would become less likely able to maintain the branchingrestrictions [Becker 1983] Deregulation that reduces deadweightcosts of regulation is also consistent with the public-interesttheory
Kane [1996] argues that another major shock to the oldequilibrium was the increasing public awareness of the costlinessof having government-insured but (geographically) undiversiednancial institutions In the late 1970s the failure rate of banksbegan to rise (Table V) In the 1980s the Savings and Loan crisisand taxpayer bailout further heightened public awareness aboutthe costs of restrictions that make depository institutions morelikely to require infusions of taxpayer funds The failures thusmay have heightened public support for branching deregulation(see also Abrams and Settle [1993]) As noted above we did notnd any evidence that banking failures or distress in a stateaffected the speed with which the state deregulated but distressstill may be a common factor affecting all states
These technological economic and legal shocks generatedconditions that changed the long-standing balance favoring theantibranching forces The marginal value of lobbying to repealbranching restrictions increased just as the relative value to thesmall banks of maintaining branching restrictions was decliningThese nationwide shocks are common factors across states andthey are consistent with the positive duration dependence associ-ated with state-level deregulation that we demonstrated in FigureII In addition a broad change in lsquolsquoideologyrsquorsquo against government
QUARTERLY JOURNAL OF ECONOMICS1462
regulation could explain the positive duration dependence duringthe sample period
VI CONCLUSIONS
The private-interest theory of regulation can account for thepattern of bank branching deregulation of the 1970s and 1980sBeneciaries of branching regulation had supported a coalitionfavoring geographical restrictions despite their costs to consum-ers of nancial services Innovations that began in the 1970saltered the value of the restrictions to the affected parties and theresulting competition among interest groups can explain thesubsequent deregulation While some of our results are alsoconsistent with the public interest theorymdashfor example deregula-tion occurs earlier when small banks are in a relatively weaknancial positionmdashother results particularly evidence on theimportance of rivalries between small and large banks andbetween banking and insurance are difficult to explain with thepublic-interest approach We also nd that political-institutionalfactors affect deregulation although these variables may act asproxies for unmeasured economic interests The greater success ofthe private-interest theory to explain deregulation here than inprevious work may be due to the richer cross-sectional andtime-series variation in branching deregulation that can beexploited with a hazard model Future empirical studies ofendogenous deregulation thus may be most fruitful where changehas occurred across states over time such as in franchisinginsurance and public utilities (eg Joskow [1996])
Technological and nancial innovations will continue to erodethe benets to any interest group of maintaining regulatorybarriers in nancial services These forces are likely to bringabout reforms both domestically for example through legislationthat would increase bank powers (see Kroszner [1996] Krosznerand Rajan [1994 1997] and Kroszner and Stratmann [1998]) andinternationally for example through the extension of nancialservices provisions of NAFTAto reduce geographic barriers acrosscountries [Kroszner 1997]
APPENDIX EFFECTS OF DIFFERENT TYPES OF BRANCHING
DEREGULATION ON ECONOMIC GROWTH BANKING STRUCTUREAND BANK EFFICIENCY
We measure the importance of alternative forms of deregula-tion by their consequences for the economic growth and the
WHAT DRIVES DEREGULATION 1463
structure and efficiency of the banking sector across the states(see Jayaratne and Strahan [1996 1998]) To do so we run anOLS panel regression relating indicators for each type of deregula-tion to potential consequences Our dependent variables includeproxies for economic growth (growth in personal income) bankstructure (the log of the number of banking companies and thenumber of branches per banking company) and efficiency (the logof total noninterest expenses and the ratio of loan loss provisionsto total loans) The four deregulation indicators (MBHC branch-ing through merger and acquisition unrestricted statewide branch-ing and interstate banking) are equal to one in the years afterderegulation and zero before If a state deregulated before thesample period begins in 1970 the indicator is one throughout ifthe state does not deregulate then the indicator is zero through-out All regressions include state xed effects and time effects
Table VI shows the results of including the set of four
TABLE VIMULTIVARIATE REGRESSIONS OF THE EFFECTS OF DIFFERENT TYPES OF BRANCHING
DEREGULATION ON ECONOMIC GROWTH BANKING STRUCTUREAND BANK EFFICIENCY
Dependent variable
Growthin
personalincome
Log of of
bankingcompanies
Branchesper
bankingcompany
Log ofnon-
interestexpenses
Loan lossprovisions
(loans 1leases)
Merger amp acquisitionbranchingindicator
0013(0004)
2 0162(0037)
1146(0453)
2 0107(0035)
2 0007(0001)
Unrestricted state-wide branchingindicator
2 0008(0005)
0086(0059)
2 0043(0574)
0070(0037)
0005(0002)
Interstate bankingindicator
0006(0005)
2 0052(0057)
1064(0511)
0005(0030)
2 0001(0001)
Multibank holdingcompany indicator
2 0005(0006)
2 0096(0052)
2 0364(0529)
2 0095(0036)
2 0001(0001)
Dependent variablemean [median]
0086[0082]
4881[5136]
6649[5055]
13812[13803]
0009[0006]
This table contains regressions of state-level economic growth and state-level measures of the structureand efficiency of the banking industry Each dependent variable is regressed on a set of state indicatorvariables a set of year indicator variables and the set of four deregulation indicators these deregulationindicators are equal to one after deregulation and zero before If a state was always regulated the indicator iszero throughout if it was always deregulated the indicator is one throughout Sample period is 1975 to 1991and Delaware and South Dakota are dropped so N 5 833 Noninterest expenses become available only in1984 so N 5 392 The coefficient on the deregulation indicators is reported with robust standard errors inparentheses and denote statistically signicant at the 1 5 and 10 percent levels
QUARTERLY JOURNAL OF ECONOMICS1464
deregulation indicators simultaneously The branching by mergerand acquisition indicator is statistically signicant and of theexpected sign in all ve of the equations and larger in absolutevalue than any of the other coefficients on the indicators for theother types of deregulation Since there may be multicollinearityamong our indicators we also ran univariate specications whichincluded each indicator separately and the results are verysimilar These results imply that branching by merger andacquisition is the most important type of deregulation measuredby its consequences for economic growth banking structure andbank efficiency See Kroszner and Strahan [1998] for more details
UNIVERSITY OF CHICAGO GRADUATE SCHOOL OF BUSINESS AND NATIONAL BUREAU OF
ECONOMIC RESEARCH
FEDERAL RESERVE BANK OF NEW YORK RESEARCH amp MARKET ANALYSIS GROUP
REFERENCES
Abrams Burton A and Russell F Settle lsquolsquoPressure-Group Inuence and Institu-tional Change Branch-Banking Legislation during the Great DepressionrsquorsquoPublic Choice LXXVII (1993) 687ndash705
Amel Dean lsquolsquoState Laws Affecting the Geographic Expansion of CommercialBanksrsquorsquo unpublished manuscript Board of Governors of the Federal ReserveSystem 1993
Becker Gary S lsquolsquoA Theory of Competition among Pressure Groups for PoliticalInuencersquorsquo Quarterly Journal of Economics XCVIII (1983) 371ndash400
Berger Allen Anil Kashyap Joseph Scalise lsquolsquoThe Transformation of the U SBanking Industry What a Long Strange Trip Itrsquos Beenrsquorsquo Brookings Papers onEconomic Activity 2 (1995) 55ndash218
Berger Allen Anthony Saunders Joseph Scalise and Gregory Udell lsquolsquoThe Effectsof Bank Mergers and Acquisitions on Small Business Lendingrsquorsquo Journal ofFinancial Economics L (1998) 187ndash229
Brennan Geoffrey and James Buchanan lsquolsquoTowards a Tax Constitution forLeviathanrsquorsquo Journal of Public Economics VIII (1977) 255ndash273
Brickley James A and Christopher M James lsquolsquoThe Takeover Market CorporateBoard Composition and Ownership Structure The Case of Bankingrsquorsquo Journalof Law and Economics XXX (1987) 161ndash180
Cole Rebel and John Wolken lsquolsquoFinancial Services Used by Small BusinessesEvidence from the 1993 Survey of Small Business Financesrsquorsquo Federal ReserveBulletin LXXXI (1994) 629ndash676
Dixit Avinash The Making of Economic Policy A Transactions-Cost PoliticsPerspective (Cambridge MA MIT Press 1996)
Economides Nicolas R Glenn Hubbard and Darius Palia lsquolsquoThe Political Economyof Branching Restrictions and Deposit Insurancersquorsquo Journal of Law andEconomics XXIX (1996) 667ndash704
Flannery Mark lsquolsquoThe Social Costs of Unit Banking Restrictionsrsquorsquo Journal ofMonetary Economics XIII (1984) 237ndash249
Greene William H Econometric Analysis third edition (Upper Saddle River NJPrentice-Hall 1997)
Gunther Jeffery lsquolsquoRegional Capital Imbalances and the Removal of InterstateBanking Restrictionsrsquorsquo Economics Letters XLIV (1994) 439ndash442 lsquolsquoThe Erosion of Intrastate Branch Banking Restrictionsrsquorsquo unpublishedmanuscript Federal Reserve Bank of Dallas 1996
Irwin Douglas and Randall Kroszner lsquolsquoInterests Institutions and Ideology inSecuring Policy Change The Republican Conversion to Trade Liberalizationafter Smoot-Hawleyrsquorsquo Journal of Law and Economics XLII (1999) 643ndash673
WHAT DRIVES DEREGULATION 1465
James Christopher lsquolsquoAn Analysis of Intra-Industry Differences in the Effect ofRegulation The Case of Deposit Rate Ceilingsrsquorsquo Journal of Monetary Econom-ics XII (1983) 417ndash442
Jarrell Gregg lsquolsquoChange at the Exchange The Causes and Effects of DeregulationrsquorsquoJournal of Law and Economics XXVII (1984) 273ndash312
Jayaratne Jith and Philip E Strahan lsquolsquoThe Finance-Growth Nexus Evidencefrom Bank Branch Deregulationrsquorsquo Quarterly Journal of Economics CI (1996)639ndash670
Jayaratne Jith and Philip E Strahan lsquolsquoEntry Restrictions Industry Evolutionand Dynamic Efficiency Evidence from Commercial Bankingrsquorsquo Journal of Lawand Economics XLIX (1998) 239ndash274
Joskow Paul lsquolsquoComment on lsquoPower Struggles Explaining Deregulatory Reformsin Electricity Marketsrsquo rsquorsquo Brookings Papers on Economic Activity Microeconom-ics (1996) 251ndash264
Joskow Paul and Roger Noll lsquolsquoRegulation in Theory and Practice An Overviewrsquorsquoin Studies in Public Regulation Gary Fromm ed (Cambridge MA MITPress 1981) pp 1ndash65
Kahn Matthew E and John G Matsusaka lsquolsquoDemand for Environmental GoodsEvidence from Voting Patterns on California Initiativesrsquorsquo Journal of Law andEconomics XL (1997) 137ndash173
Kalbeisch J and R Prentice The Statistical Analysis of Failure Time Data (NewYork John Wiley amp Sons 1980)
Kane Edward J lsquolsquoDe Jure Interstate Banking Why Only NowrsquorsquoJournal of MoneyCredit and Banking XXVIII (1996) 141ndash161
Kiefer Nicholas lsquolsquoEconomic Duration Data and Hazard Functionsrsquorsquo Journal ofEconomic Literature XXVI (1988) 646ndash679
Kroszner Randall S lsquolsquoThe Evolution of Universal Banking and Its Regulation inTwentieth Century Americarsquorsquo in Universal Banking Financial System DesignReconsidered Anthony Saunders and Ingo Walter eds (New York IrwinProfessional Publishers 1996) pp 70ndash99 lsquolsquoThe Political Economy of Banking and Financial Regulation in the U Srsquorsquo inThe Banking and Financial Structure in the NAFTA Countries and ChileGeorge M von Furstenberg ed (Boston Kluwer Academic Publishers 1997)pp 200ndash213
Kroszner Randall S and Raghuram G Rajan lsquolsquoIs the Glass-Steagall ActJustied A Study of the U S Experience with Universal Banking before1933rsquorsquo American Economic Review LXXXIV (1994) 810ndash832
Kroszner Randall S and Raghuram G Rajan lsquolsquoOrganization Structure andCredibility Evidence from Bank Securities Activities before the Glass-Steagall Actrsquorsquo Journal of Monetary Economics XXXIX (1997) 475ndash516
Kroszner Randall S and Philip E Strahan lsquolsquoRegulatory Incentives and the ThriftCrisis Dividends Mutual-to-Stock Conversions and Financial DistressrsquorsquoJournal of Finance LI (1996) 1285ndash1320
Kroszner Randall S and Philip E Strahan lsquolsquoWhat Drives Deregulation Econom-ics and Politics of the Relaxation of Bank Branching Restrictionsrsquorsquo NBERWorking Paper No 6637 1998
Kroszner Randall S and Thomas Stratmann lsquolsquoInterest Group Competition andthe Organization of Congress Theory and Evidence from Financial ServicesPolitical Action Committeesrsquorsquo American Economic Review LXXXVIII (1998)pp 1163ndash1187
Kroszner Randall S and Thomas Stratmann lsquolsquoDoes Political Ambiguity PayCorporate Campaign Contributions and Rewards to Legislator Reputationrsquorsquounpublishedmanuscript University of Chicago Graduate School of Business1999
Lin D Y and L J Wei lsquolsquoThe Robust Inference for the Cox Proportional HazardsModel Journal of the American Statistical Association LXXXIV (1989)1074ndash1078
Macey Jonathan R and Geoffrey P Miller Banking Law and Regulation (BostonLittle Brown and Company 1992)
Makinson Larry Open Secrets The Encyclopedia of Congressional Money andPolitics (Washington DC Congressional Quarterly 1992)
Niskanen WilliamBureaucracy and Representative Government (Chicago Aldine-Atherton 1971)
QUARTERLY JOURNAL OF ECONOMICS1466
Noll Roger lsquolsquoComment on Peltzmanrsquorsquo Brookings Papers Microeconomics SpecialIssue (1989) 48ndash58
North Douglass lsquolsquoA Transactions Cost Theory of Politicsrsquorsquo Journal of TheoreticalPolitics II (1990) 355ndash367
Peltzman Sam lsquolsquoToward a More General Theory of Regulationrsquorsquo Journal of Lawand Economics XIX (1976) 109ndash148 lsquolsquoConstituent Interest and Congressional Votingrsquorsquo Journal of Law andEconomics XXVII (1984) 181ndash210 lsquolsquoThe Economic Theory of Regulation after a Decade of DeregulationrsquorsquoBrookings Papers Microeconomics Special Issue (1989) 1ndash41
Peterson Mitchell and Raghuram Rajan lsquolsquoThe Benets of Lending RelationshipsEvidence from Small Business Datarsquorsquo Journal of Finance XLIX (1994) 3ndash38
Poole Keith and Howard Rosenthal Congress A Political-Economic History ofRoll Call Voting (Oxford Oxford University Press 1997)
Stata Stata Reference Manual Release 5 (College Station TX Stata Press 1997)Stigler George J lsquolsquoThe Theory of Economic Regulationrsquorsquo Bell Journal of Economics
and Management Science II (1971) 3ndash21Stigler George J Chicago Studies in the Political Economy (Chicago University
of Chicago Press 1988)Strahan Philip E and James Weston lsquolsquoSmall Business Lending and the Changing
Structure of the Banking Industryrsquorsquo Journal of Banking and Finance XXII(1998) 821ndash845
Sylla Richard John Legler and John Wallis lsquolsquoBanks and State Public Finance inthe New Republic The United States 1790ndash1860rsquorsquo Journal of EconomicHistory XLVII (1987) 391ndash403
White Eugene The Regulation and Reform of the American Banking System1900ndash1929 (Princeton Princeton University Press 1983)
Winston Clifford lsquolsquoEconomic Deregulation Days of Reckoning for Microecono-mistsrsquorsquo Journal of Economic Literature XXXI (1993) 1263ndash1289
WHAT DRIVES DEREGULATION 1467
Political-Institutional Factors Republicans are typically per-ceived as more likely to favor deregulation than Democrats so thepolitical-institutional theories suggest that states that are con-trolled by Democrats deregulate later than those controlled byRepublicans15 In addition we investigate whether reform be-comes more likely when one party controls the legislature and thegovernorship Note that these political effects must be interpretedwith caution since the views of the politicians may simply reectthe economic interests of the constituents in the state (seePeltzman [1984])
B Variable Denitions and Data Sources
Our main proxy for the strength of the small banks is thefraction of banking assets in the state in lsquolsquosmallrsquorsquo banks We denesmall banks as banks with assets below the median size in eachstate By allowing the denition of small to vary across states wetake into account cross-state heterogeneity in bank sizes16 An-nual data on bank size are from the fourth quarter Reports ofIncome and Conditions (lsquolsquoCall Reportsrsquorsquo) from the Federal ReserveBoard We also include the capital-to-asset ratio of small banksrelative to large as a measure of their relative nancial healthSpecically this variable is the asset-weighted average capital-asset ratio for small banks minus the asset-weighted averagecapital-asset ratio for large banks Annual data on bank capitalare from the Call Reports
To measure the effects of the rival insurance industry we rstconstruct an indicator variable that is one if the state permitsbanks to sell insurance For each state we then measure the sizeof the insurance sector (total value added in the state) relative tothe sum of the banking plus insurance sectors We will examine
15 For more detail on the importance of legislative structures party politicsand ideology see Poole and Rosenthal [1997] Kahn and Matsusaka [1997] Irwinand Kroszner [1999] and Kroszner and Stratmann [1999]
16 We also considered a variety of other denitions of relative size but therelative size results reported below are not sensitive to which denition we useFirst we used a xed measure of small bank that did not vary across statesdening small banks as those with assets below $100 million in 1994 dollarsSecond we applied the state-varying and xed denitions at the level of lsquolsquobankingorganizationsrsquorsquo (which include multibank holding companies) instead of at the levelof banks Third we calculated gini coefficients of bank size inequality and bankconcentration indices as alternative proxies for small versus large bank power inthe state Finally we also considered banks in lsquolsquoruralrsquorsquo areas (that is not located in aBureau of the Census lsquolsquostandard metropolitan statistical arearsquorsquo) as small since thevalue of the restrictions might be greatest in protecting banks outside of cities fromentry by the city banks All of the alternative denitions are highly correlated andyield the same results on the importance of intraindustry differences
QUARTERLY JOURNAL OF ECONOMICS1446
the effect separately for states that permit banks to sell insuranceand those that do not Data on value-added by industry are fromU S Commerce Department Bureau of Economic Analysis Sur-vey of Current Business [August 1994]
To measure the relative importance of small bank-dependentborrowers we include the proportion of all establishments operat-ing in the state with fewer than twenty employees These data arecompiled by the Bureau of the Census17 As a measure of priceswe use the difference between the average interest rate on loans inthe state and the prevailing federal funds rate where the loanrate equals the ratio of total interest income on all domestic loansdivided by total domestic loans held by banks operating in thestate18 These data are from the end-of-year Call Reports butbecome available beginning only in 1976 Our proxy for bankinstability is the failure rate of banks measured as the percent oftotal state banking assets in failed institutions from the FederalReserve Boardrsquos National Information Center database19
We include two political variables to adjust for any indepen-dent inuence of party politics First we measure the degree ofparty control of the state government by the fraction of the threebodies of the state government (the assembly senate and gover-norship) controlled by Democrats This variable for example isone-third if the Democrats have a majority in the assembly andthe Republicans have a majority in the senate and hold thegovernorship Second we include an indicator variable equal toone if the same party controls the governorrsquos office and hasmajorities in both chambers of the state legislature20
17 We collected the establishment data by the state for three cross sections(1976 1982 and 1987) and interpolated the data in the intermediate years SeeState and Metropolitan Data Book 1982 1986 and 1991
18 A measure of how individual customers would be affected is interest ratesthey receive on deposits We calculated the average deposit interest rate in thestate from the end-of-year Call Reports and did not nd any relationship betweenthis variable and the timing of deregulation in our analysis below Since Jayaratneand Strahan [1998] nd that branching deregulation has no effect on deposit ratesour nding of no relationship between timing and rates result could be due toeither individual bank customers anticipating little pricing effect and not lobbyingor an inability of individual depositors to form an effective lobby
19 As alternative measures of distress we used (1) the bank failure ratemeasured as the number of failed institutions in the state relative to the totalnumber of banks (2) the overall capital-to-asset ratio of all banks in the state and(3) the growth in state personal income Our results are not sensitive to the choiceof distress proxy
20 We also estimated the models with three indicator variables to reect theparty control of the assembly senate and governorship separately The coefficientestimates on these variables are qualitatively similar to those we present below(that is Democratic control tends to slow deregulation) and the results on theinterest group variables are unaffected by the specication of the political
WHAT DRIVES DEREGULATION 1447
IV METHODS AND RESULTS
In this section we rst estimate how the variables describedabove inuence the timing of state-level branching deregulationusing a hazard model Next we discuss the robustness of theresults including a test of whether the ex post impact of deregula-tion on the different interest groups is consistent with ourinterpretation of the coefficients We then investigate federalinterstate branching deregulation with a probit model to explorewhether the factors that affect the timing of intrastate deregula-tion also inuence voting in the House of Representatives oninterstate deregulation
A Hazard Model
We consider the period from the beginning of our sample(1970) until deregulation as the lsquolsquoduration of regulationrsquorsquo or thelsquolsquotime until deregulationrsquorsquo The hazard rate h(t) is the likelihoodthat a state deregulates at time t given that the state has not yetderegulated [Kalbeisch and Prentice 1980 Kiefer 1988 Greene1997] To model the duration of regulation we must decide whatstructure if any to impose on the hazard function The Kaplan-Meier product-limit estimator provides a simple nonparametricway to estimate the shape of the hazard function over time[Greene 1997]21 Figure II graphs this estimate for our data andshows that the hazard function is relatively at in the early yearsand then grows steeper in the later years
A duration model that can approximate this shape is theWeibull proportional hazards model The hazard rate functiontakes the form
h[t x(t) b] 5 h0(t) exp[x(t)8b]
where x(t) is a vector of time-varying covariates b is a vector ofunknown parameters to be estimated and the baseline hazard
variables Note also that Nebraska has a nonpartisan unicameral legislature Weassume that the party of the Governor controls the state government for Nebraskaand set the unied government indicator to one As another alternative we triedincluding a variable that is one if party control of the legislature or governorshipchanges but it did not have an economically or statistically signicant effect anddid not affect any of the other results
21 Denote each of the K years in our sample period Tk and order them suchthat T1 T2 and so on Let nk be the number of states that have not yet deregulatedby Tk and dk be the number of states that deregulate in year Tk The Kaplan-Meierestimate of the hazard rate in each year is dknk In other words the hazard rate isthe ratio of the number of states that actually deregulate in each year to thenumber of states that have not yet deregulated in that year
QUARTERLY JOURNAL OF ECONOMICS1448
rate h0(t) is ptp 2 1 with shape parameter p that will be estimatedfrom the data When p 1 this model displays a monotonicallyincreasing hazard rate To estimate the parameters we maximizethe following log-likelihood function [Kiefer 1988]
L(b) 5 oi 5 1
N
di ln h[ti x(ti) b] 2 oi 5 1
N
e 0
tih[u x(u) b] du
where N indexes the number of states at risk at the beginning ofthe sample period di is an indicator equal to one for the statesthat deregulated by the end of the sample period and zero for thestates that do not deregulate (ie the censored observations) andti is the time of deregulation for the ith state This formulationassumes that the time until deregulation is independent acrossstates but the observations of the same state over time are notassumed to be independent In addition we use a robust estima-tion procedure to calculate the standard errors [Lin and Wei1989]
We exclude states that deregulated before 1970 from theanalysis so N equals 39 states During the sample period 36 ofthe 39 states deregulate We have three censored observations forwhich di is zero (Arkansas Iowa and Minnesota) but our results
FIGURE IIKaplan-Meier Estimate of Hazard Rate for Branching Deregulation
WHAT DRIVES DEREGULATION 1449
are not sensitive to the inclusion of these censored observationsSince we observe each state in each year up to and including theyear of deregulation we have a total of 637 observations Table IIreports summary statistics and correlation coefficients for theexplanatory variables we use in our analysis
We use the Weibull model rather than a model that does notimpose any structure on the baseline hazard rate so that we cancalculate the change in the expected time to deregulation for agiven change in the levels of the covariates22 In the Weibullmodel we can invert the hazard function and map it into the timedomain Rewriting the Weibull model in this way the log of thetime to deregulation T is a linear function of the political economyfactors and an error term ln (T ) 5 x8b 1 e23 Because we assumethat the baseline hazard rate is Weibull with a shape characterizedby p the new coefficients will be scaled by p that is b 5 2 bp
The b coefficients represent the percentage change in thetime to deregulation for a one-unit change in the covariates Apositive coefficient implies that an increase in the variableincreases the expected time until deregulation To gauge theeconomic importance of the effects we will multiply the b by thestandard deviation of the explanatory variables and then evaluatehow much the changes in the variables raise or lower the expectedtime to deregulation24 In Table III we report the b coefficients
22 The Cox proportional hazards model does not impose any structure on thebaseline hazard rate h0(t) and takes the following form
h[t x(t) b] 5 h0(t) exp [x(t)8b]
Common factors that affect the probability of deregulation in the same way for allstates therefore do not inuence the estimated coefficients The cost of the Coxmodel is that because there is no structure on the baseline we can calculatechanges in only relative hazard rates associated with changes in the covariatesThe Weibull model however provides sufficient structure so that we can translateour estimates into a log expected time metric that allows us to calculate the changein the expected time to deregulation for a given change in the covariates SeeKiefer [1988] Greene [1997] and Stata [1997] When we compare the coefficientsestimated by the Cox and Weibull models on our data they are very close (seeKroszner and Strahan [1998]) so imposing a Weibull structure does not appear todistort our estimates
23 The error term e is independent of x and has an extreme value distributionscaled by 1p (see Kiefer [1988] Greene [1997] and Stata [1997] for more details onthis log expected time parameterization of the Weibull model)
24 The transformation to the log expected time domain is only strictly correctwhen the covariates are constant over time Our interpretation of the coefficients ispossible in the time-varying model if we assume that a one-standard-deviationdifference would hold over the entire sample period that is the deregulationwould occur so many years earlier or later if the variable in question were higherby one standard deviation from 1970 to 1992 Note that the mean number of yearsto deregulation in the sample is 158
QUARTERLY JOURNAL OF ECONOMICS1450
TABLE IISUMMARY STATISTICS FOR VARIABLES IN HAZARD MODEL 1970ndash1992
Panel A Univariate statistics MeanStandarddeviation Minimum Maximum
(1) (2) (3) (4)Small bank asset share of all
banking assets in state 011 005 001 023Relative size of insurance to
banking plus insurance inthe state 047 009 024 080
Indicator is 1 if banks can sellinsurance in the state 020 040 0 1
Small rm share of the numberof rms in the state 088 002 078 095
Capital ratio of small banksrelative to large in the state 002 001 2 001 009
Share of state government con-trolled by Democrats 065 036 0 1
Indicator is 1 if state controlledby one party 054 050 0 1
Average yield on bank loans inthe state minus Fed fundsrate 002 002 2 004 008
Panel B CorrelationsSmallbank
Relativesize of
insuranceInsuranceindicator
Smallrm
Smallbank
capitalDem
controlSingleparty
(1) (2) (3) (4) (5) (6) (7)Small bank asset share
of all banking assetsin state 1
Relative size of insur-ance to banking plusinsurance in the state 2 034 1
Indicator is 1 if bankscan sell insurance inthe state 020 2 012 1
Small rm share of thenumber of rms inthe state 030 2 011 005 1
Capital ratio of smallbanks relative tolarge in the state 2 066 020 2 010 2 006 1
Share of state govern-ment controlled byDemocrats 2 006 001 2 018 2 018 005 1
Indicator is 1 if statecontrolled by oneparty 007 012 001 2 006 2 006 030 1
Average yield on bankloans in the stateminus Fed funds rate 2 001 2 015 003 022 2 010 2 003 2 006
The number of observations is 637 except for the average interest rate on bank loans which becomesavailable only in 1976 so the number of observations is 408
TABLE IIIHAZARD MODEL OF POLITICAL ECONOMY FACTORS AFFECTING THE TIMING OF STATE
BRANCHING DEREGULATION 1970ndash1992
(1) (2) (3) (4) (5) (6)Small bank asset share of
all banking assets instate
648(122)
519(137)
632(135)
516(118)
488(133)
445(088)
Capital ratio of smallbanks relative to largein the state
1325(348)
921(319)
1301(362)
1067(346)
1144(375)
906(381)
Relative size of insurancein states where banksmay sell insurance 0otherwise
324(161)
259(149)
315(165)
285(154)
272(159)
055(134)
Indicator is 1 if banksmay sell insurance inthe state
045(010)
037(011)
045(010)
038(008)
036(009)
001(012)
Relative size of insurancein states where banksmay not sell insurance0 otherwise
2 093(043)
2 090(033)
2 102(051)
2 093(048)
2 096(053)
2 004(039)
Small rm share of thenumber of rms in thestate
2 972(211)
2 614(215)
2 952(211)
2 943(230)
2 952(270)
2 1509(276)
Share of state govern-ment controlled byDemocrats
031(012)
023(011)
030(012)
027(011)
026(011)
011(013)
Indicator is 1 if state con-trolled by one party
2 004(010)
2 004(007)
2 004(007)
2 001(009)
2 002(010)
017(009)
Average yield on bankloans in the stateminus Fed funds rate mdash
023(319) mdash mdash mdash mdash
Bank failure rate mdash mdash092
(111) mdash mdash mdash
Indicator is 1 if state hasunit banking law mdash mdash mdash
018(010)
019(010)
030(009)
Indicator is 1 if statechanges bank insur-ance powers mdash mdash mdash mdash
2 008(012)
2 026(013)
Includes regionalindicators No No No No No Yes
N 637 408 637 637 637 637Log likelihood 2 374 900 2 360 2 212 2 191 1192p-value of x 2 for
regression 001 001 001 001 001 001
The hazard model is Weibull where the dependent variable is the log expected time to branchingderegulation ln (T) 5 bx 1 eAll variables are measured for each state in each year Small bank asset share isthe percent of banking assets in the state held by banks below the median size of banks in each state in eachyear Relative capital ratio is the capital to assets ratio of small banks minus that of large banks Size ofinsurance relative to banking plus insurance in the state is measured as gross state product from insurancedivided by gross state product from insurance plus banking Insurance indicator is 1 if state law permits banksto enter the insurance business Small rm share is the percent of all establishments in the state that havefewer than twenty employees Party control variable is the share of the three bodies of state governmentcontrolled by Democrats One party control indicator is 1 if the same party controlrsquos the governorship and hasmajorities in both chambers of the state legislature Average yield on bank loans equals total interest incomeon all domestic loans made by banks in the state divided by total loans Bank failure rate is the ratio of assetsin failed banks during a given state and year divided by total bank assets in that state and year Unit bankingindicator equals 1 for states with unit banking restrictions The change in insurance powers indicator equalsone if the state changed the law to permit banks to sell insurance during the sample period We divide thecountry into four regions and include regional indicators in column (5) Robust standard errors are inparentheses and denote statistically signicant at the 1 5 and 10 percent levels
QUARTERLY JOURNAL OF ECONOMICS1452
The positive and statistically signicant coefficient on smallbank share implies that a greater small bank share delaysderegulation and this effect is economically important A one-standard-deviation increase in the small bank share results in a30 percent increase in the time until deregulation or about 47years This result supports the implications of private-interesttheory concerning intraindustry rivalry and is contrary to thepublic-interest theory
The positive and statistically signicant coefficient on therelative performance of small banks in the state implies thatstates deregulate later when small banks are relatively strongnancially25 A one-standard-deviation increase in the relativecapital-to-asset ratio results in a 15 percent rise in the time untilderegulation or about 24 years This result is consistent with theprivate-interest theory but may also be consistent with a public-interest interpretation because the risk of failures (and a costlytaxpayer-funded bailout) increases as small bank nancial healthdeclines
Interindustry competition also helps explain the timing ofderegulation In states where banks can sell insurance a rela-tively large insurance sector is associated with an increase in theexpected time to deregulation26 A one-standard-deviation in-crease in the relative size of the insurance sector in those statesthat permit banks to sell insurance leads to a 22 percent increasein the time until deregulation or about 35 years27 This result isconsistent with the private-interest theory but not the public-interest theory
Deregulation occurs earlier in states where small bank-dependent rms are relatively numerous A one-standard-
25 We also allowed small bank and large bank capital-to-asset ratios to enterseparately Their coefficient estimates were nearly equal in absolute value and ofopposite signs and so the data do not reject the relative small to large bank capitalratio specication we have chosen
26 When we use the relative size of the insurance sector in the state as awhole rather than insurance relative to banking plus insurance in the state weobtain similar results When we simultaneously include both measures of therelative size of insurance the interactions of the lsquolsquoinsurance relative to bankingplus insurancersquorsquo variable continue to be statisticallysignicant but the interactionsof lsquolsquoinsurance relative to the state as a wholersquorsquo variable are not
27 We have also estimated the model with the indicator variable equal to onefor states where banks may sell insurance but without the variables measuringtheimportance of insurance relative to banking In this model the coefficient on theinsurance indicator variablemdashan estimate of the difference in the expected time toderegulation in states where banks may sell insurance relative to states wherebanks may not sell insurancemdashequals 027 and is statistically signicant at the 5percent level This suggests that branching deregulation takes about 45 yearslonger in states where banks may sell insurance
WHAT DRIVES DEREGULATION 1453
deviation increase in the share of small rms reduces the timeuntil deregulation by 18 percent or about three years This resultconcerning the interests of consumers of banking services isconsistent with both the private- and public-interest theories
Column (2) of Table III introduces the average interest rateon loans to the model as a rough proxy for the cost to bankborrowers of these regulations These data become available onlyin 1976 so our sample size shrinks to 408 observations Thecoefficient on this variable is small and not statistically signi-cant The public-interest theory would imply that the coefficientshould be negative while the implication of the economic theorydoes not have a clear sign prediction
When we add the failure rate of banks to the specication incolumn (3) its coefficient is small and statistically insignicantWe do not nd a linkage between the timing of state deregulationand statewide banking distress Given that deposit insurance isfunded federally however bank failures may raise the publicrsquosconcerns about the costs of the regulations at the national ratherthan the state level This interpretation can explain why the rateof bank failure does not affect the timing of state-level deregula-tion even though bank instability does appear to rise generallyduring the period of increasing deregulation (see Section V below)
Finally the partisan structure of the state government alsoinuences when states deregulate As expected a higher propor-tion of Democrats in the government tends to delay deregulationA one-standard-deviation rise in the share of the governmentcontrolled by Democrats slows the deregulation by about twoyears Whether the state is dominated by one party however doesnot appear to affect the timing of the deregulation
To summarize our ndings from the hazard model privateinterests appear to play an important role in the deregulatoryprocess While private interests and public interests do sometimescoincide the results on the relative share of small banks and largebanks and on the relative size of insurance where banks competeare consistent with a private-interest approach but are difficult toexplain on public-interest grounds
B Robustness Tests
Omitted Variables States begin the sample period withdifferent degrees of restrictions on bank expansion These initialconditions may proxy for omitted factors determining the political-economy equilibrium in the state that could then affect the
QUARTERLY JOURNAL OF ECONOMICS1454
subsequent timing of deregulation For instance some statesbegan the period with unit banking while other states allowedlimited branching and the formation of multibank holdingcompanies
To test whether the stringency of initial restrictions onbranching matters we rst divide the sample into states with andwithout the most extreme form of branching regulation unitbanking28 Column (4) of Table III includes an indicator variableequal to one if the state began the period with unit bankingrestrictions The coefficient on this variable is positive andstatistically signicant but the other results remain virtuallyunchanged Next we include an indicator variable that is one ifthe state permitted multibank holding companies (MBHCs) toexist at the beginning of our sample period Thirty-four statesbegan the period allowing MBHCs The coefficient estimate forthis indicator variable (not reported) is small and not statisticallysignicant and its inclusion does not change the effects of theother variables
Eleven states expand bank powers to include the sale ofinsurance products during our sample period This could proxy foromitted factors about the relative strength of the insuranceindustry in the state In column (5) we include an indicatorvariable that is one if the state changes the law during our sampleperiod to permit banks to sell insurance business and zerootherwise The coefficient estimate is small and statisticallyinsignicant and does not affect the other results29
In column (6) we introduce four regional indicator variablesNorth South Midwest and West30 The coefficients on small bankshare the relative capital-to-assets ratio and small rm sharechange little Although the coefficient estimate still suggests that
28 We classify states that prohibited branching but permitted banks toestablish facilities as unit banking states The unit banking indicator is one for thefollowing states CO AR FL IL IA KS MN MO MT NE ND OK TX WI WVand WY
29 We conducted a likelihood ratio test of whether the unit banking statesthe MBHC states and the states that changed their regulations of bank insurancepowers could be pooled with the other states In all three cases we could not rejectpooling of the data
30 Since intrastate deregulation generally preceded interstate deregulationand the latter typically took the form of regional interstate compacts thepossibility of participating in a regional interstate banking compact could haveinuenced the decision to deregulate intrastate branching Our denitions of theregions are region 1 (South) contains AL AR DC FL GA KY LA MS NC OKSC TN TX and VA region 2 (Northeast) contains CT MA MD ME NH NJ NYPA RI VT and WV region 3 (Midwest) contains IA IL IN KS MI MN MO NEND OH SD and WI region 4 (West) contains the other states
WHAT DRIVES DEREGULATION 1455
in states where banks can sell insurance a larger relative size ofthe insurance sector delays deregulation the effect is no longerstatistically signicant The coefficient on the indicator variablefor whether a state expands bank powers to permit insurancesales during the sample period however now becomes statisti-cally signicantly negative This implies that deregulation occursroughly 41 years earlier in states where the insurance lobby maybe relatively weak While the inclusion of the regional indicatorsdoes not affect the other results the results on insurance doappear to be sensitive to the specication of the hazard model31
Estimation Technique To check the robustness of using ahazard model and to investigate whether the time-series or thecross-sectional variation in each of the factors accounts for thehazard results we estimate two simple linear probability models(neither are reported) First we run a pooled time-series cross-section OLS regression using the 637 observations from thehazard model The dependent variable equals one in the year ofderegulation and zero otherwise We include the same set ofregressors that were used in the hazard model and add state-levelxed effects This specication thus removes the cross-statevariation from the independent variables and focuses on theirmovements over time In this model the signs of the coefficientson the interest group factors are consistent with the results fromthe hazard model and the coefficients for the small bank relativecapital ratio and for the relative size of insurance where states cansell insurance are statistically signicant
In the second model we eliminate the time-series variationby averaging each of the explanatory variables over time We thushave one observation per state that had branching restrictions asof 1970 (N 5 39) The dependent variable equals the inverse of thenumber of years until deregulation which represents an estimateof the statersquos probability of deregulating in a year under theassumption that this probability is constant over time and eachyear provides an independent observation32 Although we haveonly 39 observations the coefficients on the small bank marketshare small bank relative capital ratio and small rm share
31 As further robustness check we also tried a population density variablewhich Abrams and Settle [1993] found relevant for regulatory change during the1920s and 1930s but its coefficient was not statistically signicant and did notaffect the other results
32 For instance the dependent variable would take the value of 01 for astate that deregulated in 1980 For the three states that did not deregulate thedependent variable equals 0
QUARTERLY JOURNAL OF ECONOMICS1456
variables are statistically signicant The signs of the effects ofthe insurance variables are the same as those reported in thehazard model but not statistically signicant
The hazard model and linear probability models thus producevery similar results The estimates from the latter suggest thatthe effects for the small bank market share and small rm sharevariables are driven primarily by cross-state variation the effectof the insurance variable is driven primarily by time-seriesvariation and the small bank capital effect is driven by bothcross-sectional and time-series variation
Ex Post Consequences of Deregulation Finally to check theplausibility of our interpretation of the hazard results we con-sider whether the ex post consequences of deregulation areconsistent with the ex ante lobbying positions we attribute to eachinterest group (see Kroszner and Strahan [1998] for details) Wend that small banks lose market share following deregulationand in states where banks can enter the insurance business theinsurance sector shrinks relative to the banking sector followingderegulation Borrowers also benet from deregulation throughlower average interest rates on loans These ndings support theprivate-interest interpretation of the coefficients in the hazardmodel groups that will benet lobby to speed deregulation andthose that will be harmed lobby to slow it
C Voting on Interstate Branching Deregulation in the U SHouse of Representatives
We now examine whether the forces driving intrastate branch-ing deregulation also drive interstate deregulation Financialservices interests are active contributors and lobbyists in Washing-ton Their political action committees constitute the largest groupof contributors to legislators providing nearly 20 percent of totalcongressional campaign contributions [Makinson 1992] and muchof their lobbying effort involves competition among rival interestswithin nancial services (see Kroszner and Stratmann [1998])
As noted above after virtually all states adopted intra- andinterstate branching deregulation the 1994 Riegle-Neal Act re-pealed the 1927 McFadden Act to eliminate all barriers tointerstate banking and branching by 1997 The key votes concern-ing the Riegle-Neal Act were either voice votes or extremelylopsided so we could not estimate a voting model from them Anumber of bills and amendments related to interstate branching
WHAT DRIVES DEREGULATION 1457
had been debated in Congress during the years prior to thepassage of the Riegle-Neal Act but a search of the weekly BNABanking Reporter and the Congressional Record produced onlyone roll-call vote related to interstate branching that was notlopsided This vote occurred in the House on November 14 1991and concerned an amendment sponsored by Wylie (R-OH) andNeal (D-NC) to introduce interstate banking and branchingderegulation to a nancial services reform package33 Althoughthe amendment passed by 210 to 208 the bill to which it wasattached subsequently was defeated
To check for the inuence of the factors we considered in thestate-level reforms we examine both the sponsorship of interstatebanking legislation and voting on the amendment The sponsorsof the Wylie-Neal amendment are from states with low small bankshares 004 in Ohio (Wylie) and 002 in North Carolina (Neal) Incontrast the sample mean in 1991 is 008 (median 5 007)Michigan home state of the Senatersquos sponsor of the 1994 Riegle-Neal Act also had relatively low small bank strength (small bankshare of 005)
Table IV reports estimates of a probit model where thedependent variable equals one if the legislator voted in favor ofthe amendment and zero otherwise The explanatory variablesare the same as those in the hazard model except that we use aparty affiliation indicator variable equal to one for Democrats andzero for Republicans as our proxy for political factors Thecoefficients in Table IV are the marginal effects (lsquolsquoslopesrsquorsquo) of aone-unit change of each variable on the probability that alegislator will vote for the amendment34 Note that a forcefavoring deregulation has a negative coefficient in the hazardmodel (Table III) but a positive coefficient in the probit model
Consistent with the state-level deregulation process legisla-tors are more likely to support the amendment if their states havea relatively low share of small banks As in the hazard model thefraction of small banks is the most important interest groupinuence on a legislatorrsquos voting decision A one-standard-deviation increase in small banksrsquo market share (from the mean)is associated with a decline in the probability of voting in favor of
33 The Wylie-Neal amendment also included provisions limiting certaininsurance and real estate powers of national banks [Congressional RecordNovember 14 1991 pp 10239ndash10242]
34 Since we have multiple legislators from each state we adjust the standarderrors to correct for the lack of independence among observations clustered in thesame state
QUARTERLY JOURNAL OF ECONOMICS1458
TAB
LE
IVM
AR
GIN
AL
EF
FE
CT
SF
RO
MA
PR
OB
ITM
OD
EL
OF
TH
EIN
FL
UE
NC
EO
FP
OL
ITIC
AL
EC
ON
OM
YF
AC
TO
RS
ON
TH
EH
OU
SE
VO
TE
ON
TH
EW
YL
IE-N
EA
L
AM
EN
DM
EN
TT
OP
ER
MIT
INT
ER
ST
AT
EB
AN
KB
RA
NC
HIN
G
NO
VE
MB
ER
14
1991
(1)
(2)
(3)
(4)
(5)
(6)
Smal
lban
kas
set
shar
eof
allb
anki
ng
asse
tsin
stat
e2
417
(12
4)2
454
(12
2)2
407
(12
0)2
443
(14
7)2
450
(14
9)2
490
(14
7)C
apit
alra
tio
ofsm
allb
anks
rela
tive
tola
rge
inth
est
ate
180
(37
5)2
68(3
42)
256
(37
5)1
69(3
78)
153
(38
0)2
45(4
25)
Rel
ativ
esi
zeof
insu
ranc
ein
stat
esw
here
bank
sm
ayse
llin
sura
nce
0ot
herw
ise
22
34
(10
2)2
249
(1
14)
22
33
(10
1)2
249
(1
04)
22
51
(10
2)2
154
(11
1)In
dica
tor
is1
ifba
nks
may
sell
insu
ranc
ein
the
stat
e0
79
(0
06)
074
(00
6)0
69
(0
06)
070
(00
6)0
70
(0
06)
067
(00
7)R
elat
ive
size
ofin
sura
nce
inst
ates
whe
reba
nks
may
not
sell
insu
ranc
e0
oth
erw
ise
180
(06
6)2
42
(0
72)
154
(0
67)
173
(06
6)1
72
(0
67)
203
(0
91)
Smal
lrm
shar
eof
the
num
ber
ofr
ms
inth
est
ate
21
23(0
92)
20
83(0
89)
21
62(1
07)
21
32(0
95)
21
30(1
00)
20
94(1
16)
Indi
cato
ris
1if
Rep
rese
ntat
ive
isa
Dem
ocra
t2
035
(00
7)2
035
(00
7)2
036
(00
7)2
035
(00
7)2
035
(00
7)2
036
(00
7)A
vera
geyi
eld
onba
nklo
ans
inth
est
ate
min
usF
edfu
nds
rate
mdash10
01
(51
5)mdash
mdashmdash
mdash
Ave
rage
fail
ure
rate
198
5to
1991
mdashmdash
228
(14
1)mdash
mdashmdash
Indi
cato
ris
1if
stat
eh
asun
itba
nkin
gla
wmdash
mdashmdash
004
(00
7)0
04(0
07)
20
01(0
08)
Indi
cato
ris
1if
stat
ech
ange
sba
nk
insu
ran
cepo
wer
smdash
mdashmdash
mdash2
001
(00
7)2
003
(00
7)In
clud
esre
gion
alin
dica
tors
N
oN
oN
oN
oN
oYe
sP
seud
o-R
20
170
180
170
170
170
18L
ogli
keli
hood
224
024
223
826
223
951
224
013
224
010
223
808
p-va
lue
ofx2
for
regr
essi
on
001
0
01
001
0
01
001
0
01
Th
ede
pen
den
tvar
iabl
eis
one
ifth
ele
gisl
ator
vote
sfo
rth
eam
endm
enta
nd
zero
ifag
ains
tT
he
repo
rted
coef
fici
ents
are
the
effe
cts
ofa
uni
tch
ange
ofth
ein
depe
nde
ntva
riab
le(f
rom
the
mea
n)on
the
prob
abili
tyof
voti
ng
infa
vor
ofth
eam
endm
ent
All
vari
able
sar
em
easu
red
byst
ate
N5
418
the
num
ber
ofR
epre
sent
ativ
esvo
tin
gon
the
amen
dmen
tS
mal
lban
kas
sets
har
eis
the
perc
ent
ofba
nki
ng
asse
tsin
the
stat
eh
eld
byba
nks
belo
wth
em
edia
nsi
zeof
ban
kin
each
stat
ein
each
year
Rel
ativ
eca
pita
lrat
iois
the
capi
talt
oas
sets
rati
oof
smal
lban
ksm
inu
sth
atof
larg
eba
nks
Siz
eof
insu
ran
cere
lati
veto
bank
ing
plu
sin
sura
nce
inth
est
ate
ism
easu
red
asgr
oss
stat
epr
oduc
tfro
min
sura
nce
divi
ded
bygr
oss
stat
epr
odu
ctfr
omin
sura
nce
plu
sba
nkin
gIn
dica
tor
ison
eif
stat
ela
wpe
rmit
sba
nks
toen
ter
the
insu
ran
cebu
sin
ess
Sm
all
rmsh
are
isth
epe
rcen
tofa
lles
tabl
ishm
ents
inth
est
ate
that
hav
efe
wer
than
twen
tyem
ploy
ees
Par
tyin
dica
tor
ison
eif
the
Rep
rese
nta
tive
isa
Dem
ocra
tA
vera
geyi
eld
onba
nklo
ans
equa
lsto
tali
nter
est
inco
me
onal
ldom
esti
clo
ans
mad
eby
ban
ksin
the
stat
edi
vide
dby
tota
lloa
ns
The
aver
age
bank
failu
rera
teis
the
rati
oof
asse
tsin
faile
dba
nks
duri
ng
agi
ven
stat
edi
vide
dby
tota
lban
kas
sets
inth
atst
ate
aver
aged
over
the
year
s19
85to
1991
Un
itba
nki
ng
indi
cato
req
ual
son
efo
rst
ates
wit
hun
itba
nkin
gre
stri
ctio
nsT
hech
ange
inin
sura
nce
pow
ers
indi
cato
req
ual
son
eif
the
stat
ech
ange
dth
ela
wto
perm
itba
nks
tose
llin
sura
nce
duri
ngth
esa
mpl
epe
riod
We
divi
deth
eco
unt
ryin
tofo
urre
gion
san
din
clu
dere
gion
alin
dica
tors
inco
lum
n(5
)R
obu
stst
anda
rder
rors
are
inpa
ren
thes
es
an
d
den
ote
stat
isti
cally
sign
ica
nta
tth
e1
5an
d10
perc
entl
evel
s
WHAT DRIVES DEREGULATION 1459
branching of approximately 17 percent The relative capital-to-assets ratio of small and large banks however does not have astatistically signicant effect in the probit model
The impact of rival interests outside of banking is alsoconsistent with intrastate deregulation results Where banks cansell insurance legislators from states with larger insurancesectors relative to banking are less likely to vote for the amend-ment35 A one-standard-deviation increase in the relative size ofinsurance in those states which permit banks to sell insurancedecreases the probability that a legislator will favor the amend-ment by about 13 percent Turning to consumer interests thecoefficient on small rm share is not statistically signicant butthe coefficient on interest rates in the state is A one-standard-deviation increase in the average interest rate on loans raises theprobability that a legislator will support the amendment byroughly 8 percent The coefficient estimate for the bankingdistress variable measured as the average bank failure rate from1985 to 1991 is small and statistically signicant Overall theprobit analysis of the vote on national branching deregulationsupports the private-interest theory of deregulation and providesa consistency check that the importance of interests operating onthe state legislatures are very similar to those operating at thefederal level
V WHY DID BRANCHING DEREGULATION BEGIN DURING THE
1970S SHOCKS TO THE EQUILIBRIUM SUPPORTING BANK
BRANCHING RESTRICTIONS
A complete explanation of regulatory exit should account forwhy deregulation begins in the 1970s as well as the timing of thestate-by-state reforms [Kane 1996] In this section we describebroad technological legal and economic shocks that altered thepolitical-economy equilibrium which had kept antibranching regu-lations little changed for at least 30 years
Beginning in the 1970s three innovations reduced the valueto the protected banks of local geographic monopolies (see TableV) First automatic teller machines (ATMs) helped to erode thegeographic ties between customers and banks Second checkablemoney market mutual funds and the Merrill Lynch Cash Manage-ment Account demonstrated that banking by mail and telephone
35 The positive and statistically signicant estimates on the other insurancevariables may reect insurance industry support for the amendmentrsquos provisionslimiting banksrsquo insurance powers
QUARTERLY JOURNAL OF ECONOMICS1460
provided a convenient alternative to local banks36 Third techno-logical innovation and deregulation reduced transportation andcommunication costs particularly since the 1970s thereby lower-ing the costs for customers to use distant banks By increasing theelasticity of deposits supplied to banks these innovations reducedthe value of geographical restrictions to their traditional benecia-ries and thereby reduced their incentive to ght to maintain them[Peltzman 1976]
On the lending side increasing sophistication of credit-scoring techniques following innovations in information process-ing nancial theory and the development of large credit databases began to diminish the value of knowledge that localbankers had about the risks of borrowers in the community As a
36 Regulation Q which limited the interest rates that banks could pay ondeposits may have helped to drive depositors away from banks when the gapbetween market rates and deposit ceilings grew during the 1970s The eliminationof interest rate ceilings on large denomination certicates of deposit during the1970s appears to have hurt smaller and retail-oriented banks relative to largerwholesale banks [James 1983]
TABLE VBROAD TRENDS IN COMMERCIAL BANKING 1950ndash1995
YearNumberof ATMs
Domesticbank
deposits(billions)
Moneymarketmutual
fund(billions)
Percent ofdeposits 1
moneyfunds heldby banks
Smallbanksrsquo
percent ofbankingassets
Averagenumber of
bankfailures
(1) (2) (3) (4) (5) (6)1950 0 $ 154 $ 0 100 NA 41955 0 191 0 100 NA 31960 0 228 0 100 24 21965 0 330 0 100 20 41970 0 479 0 100 18 61975 9750 775 4 99 18 61980 18500 1182 76 94 17 101985 61117 1787 242 88 14 601990 80156 2339 493 83 11 1791995 122706 2552 745 77 8 61
Column (1) ATM gures are from Bank Network News The EFT Network Data Book (New YorkFaulkner and Gray Inc) The 1975 gure was unavailable 9750 is the number of ATMs in 1978 the rst yearfor which complete data are available Columns (2)ndash(4) Banksrsquo domestic deposits are from the Reports ofIncome and Condition money market mutual funds are from the Flow of Funds Data on all bank depositsforeign plus domestic are only available beginning in 1970 The trend in banksrsquo share (column (4)) is the sameusing total deposits instead of domestic deposits Column (5) Percent of banking assets held by small bankswhere a small bank is dened as a commercial bank with less than $100 million in assets in 1994 dollarsThese data are based on the Reports of Income and Condition Data on small banks are not available before1960 Column (6) Five-year average number of bank failures where the nal year is indicated in the rstcolumn These data are from FDIC Annual Report and the Quarterly Banking Prole
WHAT DRIVES DEREGULATION 1461
result of these innovations a national market developed forresidential mortgages credit card receivables have been securi-tized and bank lending to small business now relies less on thejudgment of loan officers and more on standardized scoringmodels
These changes have increased the potential protability forlarge banks to enter what had been the core of small bankactivities Large banksrsquo incentive to increase their lobbying pres-sure to be able to expand into these markets has thus beenincreasing over time In fact small banksrsquo market share began todecline even prior to the branching deregulation (Table V) As thevalue of a local banking relationships declined small rms thatwere the main borrowers from the small banks also probablybecame more likely to favor the entry of large banks into localmarkets With the deadweight costs of preventing large bankentry rising the private-interest theory predicts that small localbanks would become less likely able to maintain the branchingrestrictions [Becker 1983] Deregulation that reduces deadweightcosts of regulation is also consistent with the public-interesttheory
Kane [1996] argues that another major shock to the oldequilibrium was the increasing public awareness of the costlinessof having government-insured but (geographically) undiversiednancial institutions In the late 1970s the failure rate of banksbegan to rise (Table V) In the 1980s the Savings and Loan crisisand taxpayer bailout further heightened public awareness aboutthe costs of restrictions that make depository institutions morelikely to require infusions of taxpayer funds The failures thusmay have heightened public support for branching deregulation(see also Abrams and Settle [1993]) As noted above we did notnd any evidence that banking failures or distress in a stateaffected the speed with which the state deregulated but distressstill may be a common factor affecting all states
These technological economic and legal shocks generatedconditions that changed the long-standing balance favoring theantibranching forces The marginal value of lobbying to repealbranching restrictions increased just as the relative value to thesmall banks of maintaining branching restrictions was decliningThese nationwide shocks are common factors across states andthey are consistent with the positive duration dependence associ-ated with state-level deregulation that we demonstrated in FigureII In addition a broad change in lsquolsquoideologyrsquorsquo against government
QUARTERLY JOURNAL OF ECONOMICS1462
regulation could explain the positive duration dependence duringthe sample period
VI CONCLUSIONS
The private-interest theory of regulation can account for thepattern of bank branching deregulation of the 1970s and 1980sBeneciaries of branching regulation had supported a coalitionfavoring geographical restrictions despite their costs to consum-ers of nancial services Innovations that began in the 1970saltered the value of the restrictions to the affected parties and theresulting competition among interest groups can explain thesubsequent deregulation While some of our results are alsoconsistent with the public interest theorymdashfor example deregula-tion occurs earlier when small banks are in a relatively weaknancial positionmdashother results particularly evidence on theimportance of rivalries between small and large banks andbetween banking and insurance are difficult to explain with thepublic-interest approach We also nd that political-institutionalfactors affect deregulation although these variables may act asproxies for unmeasured economic interests The greater success ofthe private-interest theory to explain deregulation here than inprevious work may be due to the richer cross-sectional andtime-series variation in branching deregulation that can beexploited with a hazard model Future empirical studies ofendogenous deregulation thus may be most fruitful where changehas occurred across states over time such as in franchisinginsurance and public utilities (eg Joskow [1996])
Technological and nancial innovations will continue to erodethe benets to any interest group of maintaining regulatorybarriers in nancial services These forces are likely to bringabout reforms both domestically for example through legislationthat would increase bank powers (see Kroszner [1996] Krosznerand Rajan [1994 1997] and Kroszner and Stratmann [1998]) andinternationally for example through the extension of nancialservices provisions of NAFTAto reduce geographic barriers acrosscountries [Kroszner 1997]
APPENDIX EFFECTS OF DIFFERENT TYPES OF BRANCHING
DEREGULATION ON ECONOMIC GROWTH BANKING STRUCTUREAND BANK EFFICIENCY
We measure the importance of alternative forms of deregula-tion by their consequences for the economic growth and the
WHAT DRIVES DEREGULATION 1463
structure and efficiency of the banking sector across the states(see Jayaratne and Strahan [1996 1998]) To do so we run anOLS panel regression relating indicators for each type of deregula-tion to potential consequences Our dependent variables includeproxies for economic growth (growth in personal income) bankstructure (the log of the number of banking companies and thenumber of branches per banking company) and efficiency (the logof total noninterest expenses and the ratio of loan loss provisionsto total loans) The four deregulation indicators (MBHC branch-ing through merger and acquisition unrestricted statewide branch-ing and interstate banking) are equal to one in the years afterderegulation and zero before If a state deregulated before thesample period begins in 1970 the indicator is one throughout ifthe state does not deregulate then the indicator is zero through-out All regressions include state xed effects and time effects
Table VI shows the results of including the set of four
TABLE VIMULTIVARIATE REGRESSIONS OF THE EFFECTS OF DIFFERENT TYPES OF BRANCHING
DEREGULATION ON ECONOMIC GROWTH BANKING STRUCTUREAND BANK EFFICIENCY
Dependent variable
Growthin
personalincome
Log of of
bankingcompanies
Branchesper
bankingcompany
Log ofnon-
interestexpenses
Loan lossprovisions
(loans 1leases)
Merger amp acquisitionbranchingindicator
0013(0004)
2 0162(0037)
1146(0453)
2 0107(0035)
2 0007(0001)
Unrestricted state-wide branchingindicator
2 0008(0005)
0086(0059)
2 0043(0574)
0070(0037)
0005(0002)
Interstate bankingindicator
0006(0005)
2 0052(0057)
1064(0511)
0005(0030)
2 0001(0001)
Multibank holdingcompany indicator
2 0005(0006)
2 0096(0052)
2 0364(0529)
2 0095(0036)
2 0001(0001)
Dependent variablemean [median]
0086[0082]
4881[5136]
6649[5055]
13812[13803]
0009[0006]
This table contains regressions of state-level economic growth and state-level measures of the structureand efficiency of the banking industry Each dependent variable is regressed on a set of state indicatorvariables a set of year indicator variables and the set of four deregulation indicators these deregulationindicators are equal to one after deregulation and zero before If a state was always regulated the indicator iszero throughout if it was always deregulated the indicator is one throughout Sample period is 1975 to 1991and Delaware and South Dakota are dropped so N 5 833 Noninterest expenses become available only in1984 so N 5 392 The coefficient on the deregulation indicators is reported with robust standard errors inparentheses and denote statistically signicant at the 1 5 and 10 percent levels
QUARTERLY JOURNAL OF ECONOMICS1464
deregulation indicators simultaneously The branching by mergerand acquisition indicator is statistically signicant and of theexpected sign in all ve of the equations and larger in absolutevalue than any of the other coefficients on the indicators for theother types of deregulation Since there may be multicollinearityamong our indicators we also ran univariate specications whichincluded each indicator separately and the results are verysimilar These results imply that branching by merger andacquisition is the most important type of deregulation measuredby its consequences for economic growth banking structure andbank efficiency See Kroszner and Strahan [1998] for more details
UNIVERSITY OF CHICAGO GRADUATE SCHOOL OF BUSINESS AND NATIONAL BUREAU OF
ECONOMIC RESEARCH
FEDERAL RESERVE BANK OF NEW YORK RESEARCH amp MARKET ANALYSIS GROUP
REFERENCES
Abrams Burton A and Russell F Settle lsquolsquoPressure-Group Inuence and Institu-tional Change Branch-Banking Legislation during the Great DepressionrsquorsquoPublic Choice LXXVII (1993) 687ndash705
Amel Dean lsquolsquoState Laws Affecting the Geographic Expansion of CommercialBanksrsquorsquo unpublished manuscript Board of Governors of the Federal ReserveSystem 1993
Becker Gary S lsquolsquoA Theory of Competition among Pressure Groups for PoliticalInuencersquorsquo Quarterly Journal of Economics XCVIII (1983) 371ndash400
Berger Allen Anil Kashyap Joseph Scalise lsquolsquoThe Transformation of the U SBanking Industry What a Long Strange Trip Itrsquos Beenrsquorsquo Brookings Papers onEconomic Activity 2 (1995) 55ndash218
Berger Allen Anthony Saunders Joseph Scalise and Gregory Udell lsquolsquoThe Effectsof Bank Mergers and Acquisitions on Small Business Lendingrsquorsquo Journal ofFinancial Economics L (1998) 187ndash229
Brennan Geoffrey and James Buchanan lsquolsquoTowards a Tax Constitution forLeviathanrsquorsquo Journal of Public Economics VIII (1977) 255ndash273
Brickley James A and Christopher M James lsquolsquoThe Takeover Market CorporateBoard Composition and Ownership Structure The Case of Bankingrsquorsquo Journalof Law and Economics XXX (1987) 161ndash180
Cole Rebel and John Wolken lsquolsquoFinancial Services Used by Small BusinessesEvidence from the 1993 Survey of Small Business Financesrsquorsquo Federal ReserveBulletin LXXXI (1994) 629ndash676
Dixit Avinash The Making of Economic Policy A Transactions-Cost PoliticsPerspective (Cambridge MA MIT Press 1996)
Economides Nicolas R Glenn Hubbard and Darius Palia lsquolsquoThe Political Economyof Branching Restrictions and Deposit Insurancersquorsquo Journal of Law andEconomics XXIX (1996) 667ndash704
Flannery Mark lsquolsquoThe Social Costs of Unit Banking Restrictionsrsquorsquo Journal ofMonetary Economics XIII (1984) 237ndash249
Greene William H Econometric Analysis third edition (Upper Saddle River NJPrentice-Hall 1997)
Gunther Jeffery lsquolsquoRegional Capital Imbalances and the Removal of InterstateBanking Restrictionsrsquorsquo Economics Letters XLIV (1994) 439ndash442 lsquolsquoThe Erosion of Intrastate Branch Banking Restrictionsrsquorsquo unpublishedmanuscript Federal Reserve Bank of Dallas 1996
Irwin Douglas and Randall Kroszner lsquolsquoInterests Institutions and Ideology inSecuring Policy Change The Republican Conversion to Trade Liberalizationafter Smoot-Hawleyrsquorsquo Journal of Law and Economics XLII (1999) 643ndash673
WHAT DRIVES DEREGULATION 1465
James Christopher lsquolsquoAn Analysis of Intra-Industry Differences in the Effect ofRegulation The Case of Deposit Rate Ceilingsrsquorsquo Journal of Monetary Econom-ics XII (1983) 417ndash442
Jarrell Gregg lsquolsquoChange at the Exchange The Causes and Effects of DeregulationrsquorsquoJournal of Law and Economics XXVII (1984) 273ndash312
Jayaratne Jith and Philip E Strahan lsquolsquoThe Finance-Growth Nexus Evidencefrom Bank Branch Deregulationrsquorsquo Quarterly Journal of Economics CI (1996)639ndash670
Jayaratne Jith and Philip E Strahan lsquolsquoEntry Restrictions Industry Evolutionand Dynamic Efficiency Evidence from Commercial Bankingrsquorsquo Journal of Lawand Economics XLIX (1998) 239ndash274
Joskow Paul lsquolsquoComment on lsquoPower Struggles Explaining Deregulatory Reformsin Electricity Marketsrsquo rsquorsquo Brookings Papers on Economic Activity Microeconom-ics (1996) 251ndash264
Joskow Paul and Roger Noll lsquolsquoRegulation in Theory and Practice An Overviewrsquorsquoin Studies in Public Regulation Gary Fromm ed (Cambridge MA MITPress 1981) pp 1ndash65
Kahn Matthew E and John G Matsusaka lsquolsquoDemand for Environmental GoodsEvidence from Voting Patterns on California Initiativesrsquorsquo Journal of Law andEconomics XL (1997) 137ndash173
Kalbeisch J and R Prentice The Statistical Analysis of Failure Time Data (NewYork John Wiley amp Sons 1980)
Kane Edward J lsquolsquoDe Jure Interstate Banking Why Only NowrsquorsquoJournal of MoneyCredit and Banking XXVIII (1996) 141ndash161
Kiefer Nicholas lsquolsquoEconomic Duration Data and Hazard Functionsrsquorsquo Journal ofEconomic Literature XXVI (1988) 646ndash679
Kroszner Randall S lsquolsquoThe Evolution of Universal Banking and Its Regulation inTwentieth Century Americarsquorsquo in Universal Banking Financial System DesignReconsidered Anthony Saunders and Ingo Walter eds (New York IrwinProfessional Publishers 1996) pp 70ndash99 lsquolsquoThe Political Economy of Banking and Financial Regulation in the U Srsquorsquo inThe Banking and Financial Structure in the NAFTA Countries and ChileGeorge M von Furstenberg ed (Boston Kluwer Academic Publishers 1997)pp 200ndash213
Kroszner Randall S and Raghuram G Rajan lsquolsquoIs the Glass-Steagall ActJustied A Study of the U S Experience with Universal Banking before1933rsquorsquo American Economic Review LXXXIV (1994) 810ndash832
Kroszner Randall S and Raghuram G Rajan lsquolsquoOrganization Structure andCredibility Evidence from Bank Securities Activities before the Glass-Steagall Actrsquorsquo Journal of Monetary Economics XXXIX (1997) 475ndash516
Kroszner Randall S and Philip E Strahan lsquolsquoRegulatory Incentives and the ThriftCrisis Dividends Mutual-to-Stock Conversions and Financial DistressrsquorsquoJournal of Finance LI (1996) 1285ndash1320
Kroszner Randall S and Philip E Strahan lsquolsquoWhat Drives Deregulation Econom-ics and Politics of the Relaxation of Bank Branching Restrictionsrsquorsquo NBERWorking Paper No 6637 1998
Kroszner Randall S and Thomas Stratmann lsquolsquoInterest Group Competition andthe Organization of Congress Theory and Evidence from Financial ServicesPolitical Action Committeesrsquorsquo American Economic Review LXXXVIII (1998)pp 1163ndash1187
Kroszner Randall S and Thomas Stratmann lsquolsquoDoes Political Ambiguity PayCorporate Campaign Contributions and Rewards to Legislator Reputationrsquorsquounpublishedmanuscript University of Chicago Graduate School of Business1999
Lin D Y and L J Wei lsquolsquoThe Robust Inference for the Cox Proportional HazardsModel Journal of the American Statistical Association LXXXIV (1989)1074ndash1078
Macey Jonathan R and Geoffrey P Miller Banking Law and Regulation (BostonLittle Brown and Company 1992)
Makinson Larry Open Secrets The Encyclopedia of Congressional Money andPolitics (Washington DC Congressional Quarterly 1992)
Niskanen WilliamBureaucracy and Representative Government (Chicago Aldine-Atherton 1971)
QUARTERLY JOURNAL OF ECONOMICS1466
Noll Roger lsquolsquoComment on Peltzmanrsquorsquo Brookings Papers Microeconomics SpecialIssue (1989) 48ndash58
North Douglass lsquolsquoA Transactions Cost Theory of Politicsrsquorsquo Journal of TheoreticalPolitics II (1990) 355ndash367
Peltzman Sam lsquolsquoToward a More General Theory of Regulationrsquorsquo Journal of Lawand Economics XIX (1976) 109ndash148 lsquolsquoConstituent Interest and Congressional Votingrsquorsquo Journal of Law andEconomics XXVII (1984) 181ndash210 lsquolsquoThe Economic Theory of Regulation after a Decade of DeregulationrsquorsquoBrookings Papers Microeconomics Special Issue (1989) 1ndash41
Peterson Mitchell and Raghuram Rajan lsquolsquoThe Benets of Lending RelationshipsEvidence from Small Business Datarsquorsquo Journal of Finance XLIX (1994) 3ndash38
Poole Keith and Howard Rosenthal Congress A Political-Economic History ofRoll Call Voting (Oxford Oxford University Press 1997)
Stata Stata Reference Manual Release 5 (College Station TX Stata Press 1997)Stigler George J lsquolsquoThe Theory of Economic Regulationrsquorsquo Bell Journal of Economics
and Management Science II (1971) 3ndash21Stigler George J Chicago Studies in the Political Economy (Chicago University
of Chicago Press 1988)Strahan Philip E and James Weston lsquolsquoSmall Business Lending and the Changing
Structure of the Banking Industryrsquorsquo Journal of Banking and Finance XXII(1998) 821ndash845
Sylla Richard John Legler and John Wallis lsquolsquoBanks and State Public Finance inthe New Republic The United States 1790ndash1860rsquorsquo Journal of EconomicHistory XLVII (1987) 391ndash403
White Eugene The Regulation and Reform of the American Banking System1900ndash1929 (Princeton Princeton University Press 1983)
Winston Clifford lsquolsquoEconomic Deregulation Days of Reckoning for Microecono-mistsrsquorsquo Journal of Economic Literature XXXI (1993) 1263ndash1289
WHAT DRIVES DEREGULATION 1467
the effect separately for states that permit banks to sell insuranceand those that do not Data on value-added by industry are fromU S Commerce Department Bureau of Economic Analysis Sur-vey of Current Business [August 1994]
To measure the relative importance of small bank-dependentborrowers we include the proportion of all establishments operat-ing in the state with fewer than twenty employees These data arecompiled by the Bureau of the Census17 As a measure of priceswe use the difference between the average interest rate on loans inthe state and the prevailing federal funds rate where the loanrate equals the ratio of total interest income on all domestic loansdivided by total domestic loans held by banks operating in thestate18 These data are from the end-of-year Call Reports butbecome available beginning only in 1976 Our proxy for bankinstability is the failure rate of banks measured as the percent oftotal state banking assets in failed institutions from the FederalReserve Boardrsquos National Information Center database19
We include two political variables to adjust for any indepen-dent inuence of party politics First we measure the degree ofparty control of the state government by the fraction of the threebodies of the state government (the assembly senate and gover-norship) controlled by Democrats This variable for example isone-third if the Democrats have a majority in the assembly andthe Republicans have a majority in the senate and hold thegovernorship Second we include an indicator variable equal toone if the same party controls the governorrsquos office and hasmajorities in both chambers of the state legislature20
17 We collected the establishment data by the state for three cross sections(1976 1982 and 1987) and interpolated the data in the intermediate years SeeState and Metropolitan Data Book 1982 1986 and 1991
18 A measure of how individual customers would be affected is interest ratesthey receive on deposits We calculated the average deposit interest rate in thestate from the end-of-year Call Reports and did not nd any relationship betweenthis variable and the timing of deregulation in our analysis below Since Jayaratneand Strahan [1998] nd that branching deregulation has no effect on deposit ratesour nding of no relationship between timing and rates result could be due toeither individual bank customers anticipating little pricing effect and not lobbyingor an inability of individual depositors to form an effective lobby
19 As alternative measures of distress we used (1) the bank failure ratemeasured as the number of failed institutions in the state relative to the totalnumber of banks (2) the overall capital-to-asset ratio of all banks in the state and(3) the growth in state personal income Our results are not sensitive to the choiceof distress proxy
20 We also estimated the models with three indicator variables to reect theparty control of the assembly senate and governorship separately The coefficientestimates on these variables are qualitatively similar to those we present below(that is Democratic control tends to slow deregulation) and the results on theinterest group variables are unaffected by the specication of the political
WHAT DRIVES DEREGULATION 1447
IV METHODS AND RESULTS
In this section we rst estimate how the variables describedabove inuence the timing of state-level branching deregulationusing a hazard model Next we discuss the robustness of theresults including a test of whether the ex post impact of deregula-tion on the different interest groups is consistent with ourinterpretation of the coefficients We then investigate federalinterstate branching deregulation with a probit model to explorewhether the factors that affect the timing of intrastate deregula-tion also inuence voting in the House of Representatives oninterstate deregulation
A Hazard Model
We consider the period from the beginning of our sample(1970) until deregulation as the lsquolsquoduration of regulationrsquorsquo or thelsquolsquotime until deregulationrsquorsquo The hazard rate h(t) is the likelihoodthat a state deregulates at time t given that the state has not yetderegulated [Kalbeisch and Prentice 1980 Kiefer 1988 Greene1997] To model the duration of regulation we must decide whatstructure if any to impose on the hazard function The Kaplan-Meier product-limit estimator provides a simple nonparametricway to estimate the shape of the hazard function over time[Greene 1997]21 Figure II graphs this estimate for our data andshows that the hazard function is relatively at in the early yearsand then grows steeper in the later years
A duration model that can approximate this shape is theWeibull proportional hazards model The hazard rate functiontakes the form
h[t x(t) b] 5 h0(t) exp[x(t)8b]
where x(t) is a vector of time-varying covariates b is a vector ofunknown parameters to be estimated and the baseline hazard
variables Note also that Nebraska has a nonpartisan unicameral legislature Weassume that the party of the Governor controls the state government for Nebraskaand set the unied government indicator to one As another alternative we triedincluding a variable that is one if party control of the legislature or governorshipchanges but it did not have an economically or statistically signicant effect anddid not affect any of the other results
21 Denote each of the K years in our sample period Tk and order them suchthat T1 T2 and so on Let nk be the number of states that have not yet deregulatedby Tk and dk be the number of states that deregulate in year Tk The Kaplan-Meierestimate of the hazard rate in each year is dknk In other words the hazard rate isthe ratio of the number of states that actually deregulate in each year to thenumber of states that have not yet deregulated in that year
QUARTERLY JOURNAL OF ECONOMICS1448
rate h0(t) is ptp 2 1 with shape parameter p that will be estimatedfrom the data When p 1 this model displays a monotonicallyincreasing hazard rate To estimate the parameters we maximizethe following log-likelihood function [Kiefer 1988]
L(b) 5 oi 5 1
N
di ln h[ti x(ti) b] 2 oi 5 1
N
e 0
tih[u x(u) b] du
where N indexes the number of states at risk at the beginning ofthe sample period di is an indicator equal to one for the statesthat deregulated by the end of the sample period and zero for thestates that do not deregulate (ie the censored observations) andti is the time of deregulation for the ith state This formulationassumes that the time until deregulation is independent acrossstates but the observations of the same state over time are notassumed to be independent In addition we use a robust estima-tion procedure to calculate the standard errors [Lin and Wei1989]
We exclude states that deregulated before 1970 from theanalysis so N equals 39 states During the sample period 36 ofthe 39 states deregulate We have three censored observations forwhich di is zero (Arkansas Iowa and Minnesota) but our results
FIGURE IIKaplan-Meier Estimate of Hazard Rate for Branching Deregulation
WHAT DRIVES DEREGULATION 1449
are not sensitive to the inclusion of these censored observationsSince we observe each state in each year up to and including theyear of deregulation we have a total of 637 observations Table IIreports summary statistics and correlation coefficients for theexplanatory variables we use in our analysis
We use the Weibull model rather than a model that does notimpose any structure on the baseline hazard rate so that we cancalculate the change in the expected time to deregulation for agiven change in the levels of the covariates22 In the Weibullmodel we can invert the hazard function and map it into the timedomain Rewriting the Weibull model in this way the log of thetime to deregulation T is a linear function of the political economyfactors and an error term ln (T ) 5 x8b 1 e23 Because we assumethat the baseline hazard rate is Weibull with a shape characterizedby p the new coefficients will be scaled by p that is b 5 2 bp
The b coefficients represent the percentage change in thetime to deregulation for a one-unit change in the covariates Apositive coefficient implies that an increase in the variableincreases the expected time until deregulation To gauge theeconomic importance of the effects we will multiply the b by thestandard deviation of the explanatory variables and then evaluatehow much the changes in the variables raise or lower the expectedtime to deregulation24 In Table III we report the b coefficients
22 The Cox proportional hazards model does not impose any structure on thebaseline hazard rate h0(t) and takes the following form
h[t x(t) b] 5 h0(t) exp [x(t)8b]
Common factors that affect the probability of deregulation in the same way for allstates therefore do not inuence the estimated coefficients The cost of the Coxmodel is that because there is no structure on the baseline we can calculatechanges in only relative hazard rates associated with changes in the covariatesThe Weibull model however provides sufficient structure so that we can translateour estimates into a log expected time metric that allows us to calculate the changein the expected time to deregulation for a given change in the covariates SeeKiefer [1988] Greene [1997] and Stata [1997] When we compare the coefficientsestimated by the Cox and Weibull models on our data they are very close (seeKroszner and Strahan [1998]) so imposing a Weibull structure does not appear todistort our estimates
23 The error term e is independent of x and has an extreme value distributionscaled by 1p (see Kiefer [1988] Greene [1997] and Stata [1997] for more details onthis log expected time parameterization of the Weibull model)
24 The transformation to the log expected time domain is only strictly correctwhen the covariates are constant over time Our interpretation of the coefficients ispossible in the time-varying model if we assume that a one-standard-deviationdifference would hold over the entire sample period that is the deregulationwould occur so many years earlier or later if the variable in question were higherby one standard deviation from 1970 to 1992 Note that the mean number of yearsto deregulation in the sample is 158
QUARTERLY JOURNAL OF ECONOMICS1450
TABLE IISUMMARY STATISTICS FOR VARIABLES IN HAZARD MODEL 1970ndash1992
Panel A Univariate statistics MeanStandarddeviation Minimum Maximum
(1) (2) (3) (4)Small bank asset share of all
banking assets in state 011 005 001 023Relative size of insurance to
banking plus insurance inthe state 047 009 024 080
Indicator is 1 if banks can sellinsurance in the state 020 040 0 1
Small rm share of the numberof rms in the state 088 002 078 095
Capital ratio of small banksrelative to large in the state 002 001 2 001 009
Share of state government con-trolled by Democrats 065 036 0 1
Indicator is 1 if state controlledby one party 054 050 0 1
Average yield on bank loans inthe state minus Fed fundsrate 002 002 2 004 008
Panel B CorrelationsSmallbank
Relativesize of
insuranceInsuranceindicator
Smallrm
Smallbank
capitalDem
controlSingleparty
(1) (2) (3) (4) (5) (6) (7)Small bank asset share
of all banking assetsin state 1
Relative size of insur-ance to banking plusinsurance in the state 2 034 1
Indicator is 1 if bankscan sell insurance inthe state 020 2 012 1
Small rm share of thenumber of rms inthe state 030 2 011 005 1
Capital ratio of smallbanks relative tolarge in the state 2 066 020 2 010 2 006 1
Share of state govern-ment controlled byDemocrats 2 006 001 2 018 2 018 005 1
Indicator is 1 if statecontrolled by oneparty 007 012 001 2 006 2 006 030 1
Average yield on bankloans in the stateminus Fed funds rate 2 001 2 015 003 022 2 010 2 003 2 006
The number of observations is 637 except for the average interest rate on bank loans which becomesavailable only in 1976 so the number of observations is 408
TABLE IIIHAZARD MODEL OF POLITICAL ECONOMY FACTORS AFFECTING THE TIMING OF STATE
BRANCHING DEREGULATION 1970ndash1992
(1) (2) (3) (4) (5) (6)Small bank asset share of
all banking assets instate
648(122)
519(137)
632(135)
516(118)
488(133)
445(088)
Capital ratio of smallbanks relative to largein the state
1325(348)
921(319)
1301(362)
1067(346)
1144(375)
906(381)
Relative size of insurancein states where banksmay sell insurance 0otherwise
324(161)
259(149)
315(165)
285(154)
272(159)
055(134)
Indicator is 1 if banksmay sell insurance inthe state
045(010)
037(011)
045(010)
038(008)
036(009)
001(012)
Relative size of insurancein states where banksmay not sell insurance0 otherwise
2 093(043)
2 090(033)
2 102(051)
2 093(048)
2 096(053)
2 004(039)
Small rm share of thenumber of rms in thestate
2 972(211)
2 614(215)
2 952(211)
2 943(230)
2 952(270)
2 1509(276)
Share of state govern-ment controlled byDemocrats
031(012)
023(011)
030(012)
027(011)
026(011)
011(013)
Indicator is 1 if state con-trolled by one party
2 004(010)
2 004(007)
2 004(007)
2 001(009)
2 002(010)
017(009)
Average yield on bankloans in the stateminus Fed funds rate mdash
023(319) mdash mdash mdash mdash
Bank failure rate mdash mdash092
(111) mdash mdash mdash
Indicator is 1 if state hasunit banking law mdash mdash mdash
018(010)
019(010)
030(009)
Indicator is 1 if statechanges bank insur-ance powers mdash mdash mdash mdash
2 008(012)
2 026(013)
Includes regionalindicators No No No No No Yes
N 637 408 637 637 637 637Log likelihood 2 374 900 2 360 2 212 2 191 1192p-value of x 2 for
regression 001 001 001 001 001 001
The hazard model is Weibull where the dependent variable is the log expected time to branchingderegulation ln (T) 5 bx 1 eAll variables are measured for each state in each year Small bank asset share isthe percent of banking assets in the state held by banks below the median size of banks in each state in eachyear Relative capital ratio is the capital to assets ratio of small banks minus that of large banks Size ofinsurance relative to banking plus insurance in the state is measured as gross state product from insurancedivided by gross state product from insurance plus banking Insurance indicator is 1 if state law permits banksto enter the insurance business Small rm share is the percent of all establishments in the state that havefewer than twenty employees Party control variable is the share of the three bodies of state governmentcontrolled by Democrats One party control indicator is 1 if the same party controlrsquos the governorship and hasmajorities in both chambers of the state legislature Average yield on bank loans equals total interest incomeon all domestic loans made by banks in the state divided by total loans Bank failure rate is the ratio of assetsin failed banks during a given state and year divided by total bank assets in that state and year Unit bankingindicator equals 1 for states with unit banking restrictions The change in insurance powers indicator equalsone if the state changed the law to permit banks to sell insurance during the sample period We divide thecountry into four regions and include regional indicators in column (5) Robust standard errors are inparentheses and denote statistically signicant at the 1 5 and 10 percent levels
QUARTERLY JOURNAL OF ECONOMICS1452
The positive and statistically signicant coefficient on smallbank share implies that a greater small bank share delaysderegulation and this effect is economically important A one-standard-deviation increase in the small bank share results in a30 percent increase in the time until deregulation or about 47years This result supports the implications of private-interesttheory concerning intraindustry rivalry and is contrary to thepublic-interest theory
The positive and statistically signicant coefficient on therelative performance of small banks in the state implies thatstates deregulate later when small banks are relatively strongnancially25 A one-standard-deviation increase in the relativecapital-to-asset ratio results in a 15 percent rise in the time untilderegulation or about 24 years This result is consistent with theprivate-interest theory but may also be consistent with a public-interest interpretation because the risk of failures (and a costlytaxpayer-funded bailout) increases as small bank nancial healthdeclines
Interindustry competition also helps explain the timing ofderegulation In states where banks can sell insurance a rela-tively large insurance sector is associated with an increase in theexpected time to deregulation26 A one-standard-deviation in-crease in the relative size of the insurance sector in those statesthat permit banks to sell insurance leads to a 22 percent increasein the time until deregulation or about 35 years27 This result isconsistent with the private-interest theory but not the public-interest theory
Deregulation occurs earlier in states where small bank-dependent rms are relatively numerous A one-standard-
25 We also allowed small bank and large bank capital-to-asset ratios to enterseparately Their coefficient estimates were nearly equal in absolute value and ofopposite signs and so the data do not reject the relative small to large bank capitalratio specication we have chosen
26 When we use the relative size of the insurance sector in the state as awhole rather than insurance relative to banking plus insurance in the state weobtain similar results When we simultaneously include both measures of therelative size of insurance the interactions of the lsquolsquoinsurance relative to bankingplus insurancersquorsquo variable continue to be statisticallysignicant but the interactionsof lsquolsquoinsurance relative to the state as a wholersquorsquo variable are not
27 We have also estimated the model with the indicator variable equal to onefor states where banks may sell insurance but without the variables measuringtheimportance of insurance relative to banking In this model the coefficient on theinsurance indicator variablemdashan estimate of the difference in the expected time toderegulation in states where banks may sell insurance relative to states wherebanks may not sell insurancemdashequals 027 and is statistically signicant at the 5percent level This suggests that branching deregulation takes about 45 yearslonger in states where banks may sell insurance
WHAT DRIVES DEREGULATION 1453
deviation increase in the share of small rms reduces the timeuntil deregulation by 18 percent or about three years This resultconcerning the interests of consumers of banking services isconsistent with both the private- and public-interest theories
Column (2) of Table III introduces the average interest rateon loans to the model as a rough proxy for the cost to bankborrowers of these regulations These data become available onlyin 1976 so our sample size shrinks to 408 observations Thecoefficient on this variable is small and not statistically signi-cant The public-interest theory would imply that the coefficientshould be negative while the implication of the economic theorydoes not have a clear sign prediction
When we add the failure rate of banks to the specication incolumn (3) its coefficient is small and statistically insignicantWe do not nd a linkage between the timing of state deregulationand statewide banking distress Given that deposit insurance isfunded federally however bank failures may raise the publicrsquosconcerns about the costs of the regulations at the national ratherthan the state level This interpretation can explain why the rateof bank failure does not affect the timing of state-level deregula-tion even though bank instability does appear to rise generallyduring the period of increasing deregulation (see Section V below)
Finally the partisan structure of the state government alsoinuences when states deregulate As expected a higher propor-tion of Democrats in the government tends to delay deregulationA one-standard-deviation rise in the share of the governmentcontrolled by Democrats slows the deregulation by about twoyears Whether the state is dominated by one party however doesnot appear to affect the timing of the deregulation
To summarize our ndings from the hazard model privateinterests appear to play an important role in the deregulatoryprocess While private interests and public interests do sometimescoincide the results on the relative share of small banks and largebanks and on the relative size of insurance where banks competeare consistent with a private-interest approach but are difficult toexplain on public-interest grounds
B Robustness Tests
Omitted Variables States begin the sample period withdifferent degrees of restrictions on bank expansion These initialconditions may proxy for omitted factors determining the political-economy equilibrium in the state that could then affect the
QUARTERLY JOURNAL OF ECONOMICS1454
subsequent timing of deregulation For instance some statesbegan the period with unit banking while other states allowedlimited branching and the formation of multibank holdingcompanies
To test whether the stringency of initial restrictions onbranching matters we rst divide the sample into states with andwithout the most extreme form of branching regulation unitbanking28 Column (4) of Table III includes an indicator variableequal to one if the state began the period with unit bankingrestrictions The coefficient on this variable is positive andstatistically signicant but the other results remain virtuallyunchanged Next we include an indicator variable that is one ifthe state permitted multibank holding companies (MBHCs) toexist at the beginning of our sample period Thirty-four statesbegan the period allowing MBHCs The coefficient estimate forthis indicator variable (not reported) is small and not statisticallysignicant and its inclusion does not change the effects of theother variables
Eleven states expand bank powers to include the sale ofinsurance products during our sample period This could proxy foromitted factors about the relative strength of the insuranceindustry in the state In column (5) we include an indicatorvariable that is one if the state changes the law during our sampleperiod to permit banks to sell insurance business and zerootherwise The coefficient estimate is small and statisticallyinsignicant and does not affect the other results29
In column (6) we introduce four regional indicator variablesNorth South Midwest and West30 The coefficients on small bankshare the relative capital-to-assets ratio and small rm sharechange little Although the coefficient estimate still suggests that
28 We classify states that prohibited branching but permitted banks toestablish facilities as unit banking states The unit banking indicator is one for thefollowing states CO AR FL IL IA KS MN MO MT NE ND OK TX WI WVand WY
29 We conducted a likelihood ratio test of whether the unit banking statesthe MBHC states and the states that changed their regulations of bank insurancepowers could be pooled with the other states In all three cases we could not rejectpooling of the data
30 Since intrastate deregulation generally preceded interstate deregulationand the latter typically took the form of regional interstate compacts thepossibility of participating in a regional interstate banking compact could haveinuenced the decision to deregulate intrastate branching Our denitions of theregions are region 1 (South) contains AL AR DC FL GA KY LA MS NC OKSC TN TX and VA region 2 (Northeast) contains CT MA MD ME NH NJ NYPA RI VT and WV region 3 (Midwest) contains IA IL IN KS MI MN MO NEND OH SD and WI region 4 (West) contains the other states
WHAT DRIVES DEREGULATION 1455
in states where banks can sell insurance a larger relative size ofthe insurance sector delays deregulation the effect is no longerstatistically signicant The coefficient on the indicator variablefor whether a state expands bank powers to permit insurancesales during the sample period however now becomes statisti-cally signicantly negative This implies that deregulation occursroughly 41 years earlier in states where the insurance lobby maybe relatively weak While the inclusion of the regional indicatorsdoes not affect the other results the results on insurance doappear to be sensitive to the specication of the hazard model31
Estimation Technique To check the robustness of using ahazard model and to investigate whether the time-series or thecross-sectional variation in each of the factors accounts for thehazard results we estimate two simple linear probability models(neither are reported) First we run a pooled time-series cross-section OLS regression using the 637 observations from thehazard model The dependent variable equals one in the year ofderegulation and zero otherwise We include the same set ofregressors that were used in the hazard model and add state-levelxed effects This specication thus removes the cross-statevariation from the independent variables and focuses on theirmovements over time In this model the signs of the coefficientson the interest group factors are consistent with the results fromthe hazard model and the coefficients for the small bank relativecapital ratio and for the relative size of insurance where states cansell insurance are statistically signicant
In the second model we eliminate the time-series variationby averaging each of the explanatory variables over time We thushave one observation per state that had branching restrictions asof 1970 (N 5 39) The dependent variable equals the inverse of thenumber of years until deregulation which represents an estimateof the statersquos probability of deregulating in a year under theassumption that this probability is constant over time and eachyear provides an independent observation32 Although we haveonly 39 observations the coefficients on the small bank marketshare small bank relative capital ratio and small rm share
31 As further robustness check we also tried a population density variablewhich Abrams and Settle [1993] found relevant for regulatory change during the1920s and 1930s but its coefficient was not statistically signicant and did notaffect the other results
32 For instance the dependent variable would take the value of 01 for astate that deregulated in 1980 For the three states that did not deregulate thedependent variable equals 0
QUARTERLY JOURNAL OF ECONOMICS1456
variables are statistically signicant The signs of the effects ofthe insurance variables are the same as those reported in thehazard model but not statistically signicant
The hazard model and linear probability models thus producevery similar results The estimates from the latter suggest thatthe effects for the small bank market share and small rm sharevariables are driven primarily by cross-state variation the effectof the insurance variable is driven primarily by time-seriesvariation and the small bank capital effect is driven by bothcross-sectional and time-series variation
Ex Post Consequences of Deregulation Finally to check theplausibility of our interpretation of the hazard results we con-sider whether the ex post consequences of deregulation areconsistent with the ex ante lobbying positions we attribute to eachinterest group (see Kroszner and Strahan [1998] for details) Wend that small banks lose market share following deregulationand in states where banks can enter the insurance business theinsurance sector shrinks relative to the banking sector followingderegulation Borrowers also benet from deregulation throughlower average interest rates on loans These ndings support theprivate-interest interpretation of the coefficients in the hazardmodel groups that will benet lobby to speed deregulation andthose that will be harmed lobby to slow it
C Voting on Interstate Branching Deregulation in the U SHouse of Representatives
We now examine whether the forces driving intrastate branch-ing deregulation also drive interstate deregulation Financialservices interests are active contributors and lobbyists in Washing-ton Their political action committees constitute the largest groupof contributors to legislators providing nearly 20 percent of totalcongressional campaign contributions [Makinson 1992] and muchof their lobbying effort involves competition among rival interestswithin nancial services (see Kroszner and Stratmann [1998])
As noted above after virtually all states adopted intra- andinterstate branching deregulation the 1994 Riegle-Neal Act re-pealed the 1927 McFadden Act to eliminate all barriers tointerstate banking and branching by 1997 The key votes concern-ing the Riegle-Neal Act were either voice votes or extremelylopsided so we could not estimate a voting model from them Anumber of bills and amendments related to interstate branching
WHAT DRIVES DEREGULATION 1457
had been debated in Congress during the years prior to thepassage of the Riegle-Neal Act but a search of the weekly BNABanking Reporter and the Congressional Record produced onlyone roll-call vote related to interstate branching that was notlopsided This vote occurred in the House on November 14 1991and concerned an amendment sponsored by Wylie (R-OH) andNeal (D-NC) to introduce interstate banking and branchingderegulation to a nancial services reform package33 Althoughthe amendment passed by 210 to 208 the bill to which it wasattached subsequently was defeated
To check for the inuence of the factors we considered in thestate-level reforms we examine both the sponsorship of interstatebanking legislation and voting on the amendment The sponsorsof the Wylie-Neal amendment are from states with low small bankshares 004 in Ohio (Wylie) and 002 in North Carolina (Neal) Incontrast the sample mean in 1991 is 008 (median 5 007)Michigan home state of the Senatersquos sponsor of the 1994 Riegle-Neal Act also had relatively low small bank strength (small bankshare of 005)
Table IV reports estimates of a probit model where thedependent variable equals one if the legislator voted in favor ofthe amendment and zero otherwise The explanatory variablesare the same as those in the hazard model except that we use aparty affiliation indicator variable equal to one for Democrats andzero for Republicans as our proxy for political factors Thecoefficients in Table IV are the marginal effects (lsquolsquoslopesrsquorsquo) of aone-unit change of each variable on the probability that alegislator will vote for the amendment34 Note that a forcefavoring deregulation has a negative coefficient in the hazardmodel (Table III) but a positive coefficient in the probit model
Consistent with the state-level deregulation process legisla-tors are more likely to support the amendment if their states havea relatively low share of small banks As in the hazard model thefraction of small banks is the most important interest groupinuence on a legislatorrsquos voting decision A one-standard-deviation increase in small banksrsquo market share (from the mean)is associated with a decline in the probability of voting in favor of
33 The Wylie-Neal amendment also included provisions limiting certaininsurance and real estate powers of national banks [Congressional RecordNovember 14 1991 pp 10239ndash10242]
34 Since we have multiple legislators from each state we adjust the standarderrors to correct for the lack of independence among observations clustered in thesame state
QUARTERLY JOURNAL OF ECONOMICS1458
TAB
LE
IVM
AR
GIN
AL
EF
FE
CT
SF
RO
MA
PR
OB
ITM
OD
EL
OF
TH
EIN
FL
UE
NC
EO
FP
OL
ITIC
AL
EC
ON
OM
YF
AC
TO
RS
ON
TH
EH
OU
SE
VO
TE
ON
TH
EW
YL
IE-N
EA
L
AM
EN
DM
EN
TT
OP
ER
MIT
INT
ER
ST
AT
EB
AN
KB
RA
NC
HIN
G
NO
VE
MB
ER
14
1991
(1)
(2)
(3)
(4)
(5)
(6)
Smal
lban
kas
set
shar
eof
allb
anki
ng
asse
tsin
stat
e2
417
(12
4)2
454
(12
2)2
407
(12
0)2
443
(14
7)2
450
(14
9)2
490
(14
7)C
apit
alra
tio
ofsm
allb
anks
rela
tive
tola
rge
inth
est
ate
180
(37
5)2
68(3
42)
256
(37
5)1
69(3
78)
153
(38
0)2
45(4
25)
Rel
ativ
esi
zeof
insu
ranc
ein
stat
esw
here
bank
sm
ayse
llin
sura
nce
0ot
herw
ise
22
34
(10
2)2
249
(1
14)
22
33
(10
1)2
249
(1
04)
22
51
(10
2)2
154
(11
1)In
dica
tor
is1
ifba
nks
may
sell
insu
ranc
ein
the
stat
e0
79
(0
06)
074
(00
6)0
69
(0
06)
070
(00
6)0
70
(0
06)
067
(00
7)R
elat
ive
size
ofin
sura
nce
inst
ates
whe
reba
nks
may
not
sell
insu
ranc
e0
oth
erw
ise
180
(06
6)2
42
(0
72)
154
(0
67)
173
(06
6)1
72
(0
67)
203
(0
91)
Smal
lrm
shar
eof
the
num
ber
ofr
ms
inth
est
ate
21
23(0
92)
20
83(0
89)
21
62(1
07)
21
32(0
95)
21
30(1
00)
20
94(1
16)
Indi
cato
ris
1if
Rep
rese
ntat
ive
isa
Dem
ocra
t2
035
(00
7)2
035
(00
7)2
036
(00
7)2
035
(00
7)2
035
(00
7)2
036
(00
7)A
vera
geyi
eld
onba
nklo
ans
inth
est
ate
min
usF
edfu
nds
rate
mdash10
01
(51
5)mdash
mdashmdash
mdash
Ave
rage
fail
ure
rate
198
5to
1991
mdashmdash
228
(14
1)mdash
mdashmdash
Indi
cato
ris
1if
stat
eh
asun
itba
nkin
gla
wmdash
mdashmdash
004
(00
7)0
04(0
07)
20
01(0
08)
Indi
cato
ris
1if
stat
ech
ange
sba
nk
insu
ran
cepo
wer
smdash
mdashmdash
mdash2
001
(00
7)2
003
(00
7)In
clud
esre
gion
alin
dica
tors
N
oN
oN
oN
oN
oYe
sP
seud
o-R
20
170
180
170
170
170
18L
ogli
keli
hood
224
024
223
826
223
951
224
013
224
010
223
808
p-va
lue
ofx2
for
regr
essi
on
001
0
01
001
0
01
001
0
01
Th
ede
pen
den
tvar
iabl
eis
one
ifth
ele
gisl
ator
vote
sfo
rth
eam
endm
enta
nd
zero
ifag
ains
tT
he
repo
rted
coef
fici
ents
are
the
effe
cts
ofa
uni
tch
ange
ofth
ein
depe
nde
ntva
riab
le(f
rom
the
mea
n)on
the
prob
abili
tyof
voti
ng
infa
vor
ofth
eam
endm
ent
All
vari
able
sar
em
easu
red
byst
ate
N5
418
the
num
ber
ofR
epre
sent
ativ
esvo
tin
gon
the
amen
dmen
tS
mal
lban
kas
sets
har
eis
the
perc
ent
ofba
nki
ng
asse
tsin
the
stat
eh
eld
byba
nks
belo
wth
em
edia
nsi
zeof
ban
kin
each
stat
ein
each
year
Rel
ativ
eca
pita
lrat
iois
the
capi
talt
oas
sets
rati
oof
smal
lban
ksm
inu
sth
atof
larg
eba
nks
Siz
eof
insu
ran
cere
lati
veto
bank
ing
plu
sin
sura
nce
inth
est
ate
ism
easu
red
asgr
oss
stat
epr
oduc
tfro
min
sura
nce
divi
ded
bygr
oss
stat
epr
odu
ctfr
omin
sura
nce
plu
sba
nkin
gIn
dica
tor
ison
eif
stat
ela
wpe
rmit
sba
nks
toen
ter
the
insu
ran
cebu
sin
ess
Sm
all
rmsh
are
isth
epe
rcen
tofa
lles
tabl
ishm
ents
inth
est
ate
that
hav
efe
wer
than
twen
tyem
ploy
ees
Par
tyin
dica
tor
ison
eif
the
Rep
rese
nta
tive
isa
Dem
ocra
tA
vera
geyi
eld
onba
nklo
ans
equa
lsto
tali
nter
est
inco
me
onal
ldom
esti
clo
ans
mad
eby
ban
ksin
the
stat
edi
vide
dby
tota
lloa
ns
The
aver
age
bank
failu
rera
teis
the
rati
oof
asse
tsin
faile
dba
nks
duri
ng
agi
ven
stat
edi
vide
dby
tota
lban
kas
sets
inth
atst
ate
aver
aged
over
the
year
s19
85to
1991
Un
itba
nki
ng
indi
cato
req
ual
son
efo
rst
ates
wit
hun
itba
nkin
gre
stri
ctio
nsT
hech
ange
inin
sura
nce
pow
ers
indi
cato
req
ual
son
eif
the
stat
ech
ange
dth
ela
wto
perm
itba
nks
tose
llin
sura
nce
duri
ngth
esa
mpl
epe
riod
We
divi
deth
eco
unt
ryin
tofo
urre
gion
san
din
clu
dere
gion
alin
dica
tors
inco
lum
n(5
)R
obu
stst
anda
rder
rors
are
inpa
ren
thes
es
an
d
den
ote
stat
isti
cally
sign
ica
nta
tth
e1
5an
d10
perc
entl
evel
s
WHAT DRIVES DEREGULATION 1459
branching of approximately 17 percent The relative capital-to-assets ratio of small and large banks however does not have astatistically signicant effect in the probit model
The impact of rival interests outside of banking is alsoconsistent with intrastate deregulation results Where banks cansell insurance legislators from states with larger insurancesectors relative to banking are less likely to vote for the amend-ment35 A one-standard-deviation increase in the relative size ofinsurance in those states which permit banks to sell insurancedecreases the probability that a legislator will favor the amend-ment by about 13 percent Turning to consumer interests thecoefficient on small rm share is not statistically signicant butthe coefficient on interest rates in the state is A one-standard-deviation increase in the average interest rate on loans raises theprobability that a legislator will support the amendment byroughly 8 percent The coefficient estimate for the bankingdistress variable measured as the average bank failure rate from1985 to 1991 is small and statistically signicant Overall theprobit analysis of the vote on national branching deregulationsupports the private-interest theory of deregulation and providesa consistency check that the importance of interests operating onthe state legislatures are very similar to those operating at thefederal level
V WHY DID BRANCHING DEREGULATION BEGIN DURING THE
1970S SHOCKS TO THE EQUILIBRIUM SUPPORTING BANK
BRANCHING RESTRICTIONS
A complete explanation of regulatory exit should account forwhy deregulation begins in the 1970s as well as the timing of thestate-by-state reforms [Kane 1996] In this section we describebroad technological legal and economic shocks that altered thepolitical-economy equilibrium which had kept antibranching regu-lations little changed for at least 30 years
Beginning in the 1970s three innovations reduced the valueto the protected banks of local geographic monopolies (see TableV) First automatic teller machines (ATMs) helped to erode thegeographic ties between customers and banks Second checkablemoney market mutual funds and the Merrill Lynch Cash Manage-ment Account demonstrated that banking by mail and telephone
35 The positive and statistically signicant estimates on the other insurancevariables may reect insurance industry support for the amendmentrsquos provisionslimiting banksrsquo insurance powers
QUARTERLY JOURNAL OF ECONOMICS1460
provided a convenient alternative to local banks36 Third techno-logical innovation and deregulation reduced transportation andcommunication costs particularly since the 1970s thereby lower-ing the costs for customers to use distant banks By increasing theelasticity of deposits supplied to banks these innovations reducedthe value of geographical restrictions to their traditional benecia-ries and thereby reduced their incentive to ght to maintain them[Peltzman 1976]
On the lending side increasing sophistication of credit-scoring techniques following innovations in information process-ing nancial theory and the development of large credit databases began to diminish the value of knowledge that localbankers had about the risks of borrowers in the community As a
36 Regulation Q which limited the interest rates that banks could pay ondeposits may have helped to drive depositors away from banks when the gapbetween market rates and deposit ceilings grew during the 1970s The eliminationof interest rate ceilings on large denomination certicates of deposit during the1970s appears to have hurt smaller and retail-oriented banks relative to largerwholesale banks [James 1983]
TABLE VBROAD TRENDS IN COMMERCIAL BANKING 1950ndash1995
YearNumberof ATMs
Domesticbank
deposits(billions)
Moneymarketmutual
fund(billions)
Percent ofdeposits 1
moneyfunds heldby banks
Smallbanksrsquo
percent ofbankingassets
Averagenumber of
bankfailures
(1) (2) (3) (4) (5) (6)1950 0 $ 154 $ 0 100 NA 41955 0 191 0 100 NA 31960 0 228 0 100 24 21965 0 330 0 100 20 41970 0 479 0 100 18 61975 9750 775 4 99 18 61980 18500 1182 76 94 17 101985 61117 1787 242 88 14 601990 80156 2339 493 83 11 1791995 122706 2552 745 77 8 61
Column (1) ATM gures are from Bank Network News The EFT Network Data Book (New YorkFaulkner and Gray Inc) The 1975 gure was unavailable 9750 is the number of ATMs in 1978 the rst yearfor which complete data are available Columns (2)ndash(4) Banksrsquo domestic deposits are from the Reports ofIncome and Condition money market mutual funds are from the Flow of Funds Data on all bank depositsforeign plus domestic are only available beginning in 1970 The trend in banksrsquo share (column (4)) is the sameusing total deposits instead of domestic deposits Column (5) Percent of banking assets held by small bankswhere a small bank is dened as a commercial bank with less than $100 million in assets in 1994 dollarsThese data are based on the Reports of Income and Condition Data on small banks are not available before1960 Column (6) Five-year average number of bank failures where the nal year is indicated in the rstcolumn These data are from FDIC Annual Report and the Quarterly Banking Prole
WHAT DRIVES DEREGULATION 1461
result of these innovations a national market developed forresidential mortgages credit card receivables have been securi-tized and bank lending to small business now relies less on thejudgment of loan officers and more on standardized scoringmodels
These changes have increased the potential protability forlarge banks to enter what had been the core of small bankactivities Large banksrsquo incentive to increase their lobbying pres-sure to be able to expand into these markets has thus beenincreasing over time In fact small banksrsquo market share began todecline even prior to the branching deregulation (Table V) As thevalue of a local banking relationships declined small rms thatwere the main borrowers from the small banks also probablybecame more likely to favor the entry of large banks into localmarkets With the deadweight costs of preventing large bankentry rising the private-interest theory predicts that small localbanks would become less likely able to maintain the branchingrestrictions [Becker 1983] Deregulation that reduces deadweightcosts of regulation is also consistent with the public-interesttheory
Kane [1996] argues that another major shock to the oldequilibrium was the increasing public awareness of the costlinessof having government-insured but (geographically) undiversiednancial institutions In the late 1970s the failure rate of banksbegan to rise (Table V) In the 1980s the Savings and Loan crisisand taxpayer bailout further heightened public awareness aboutthe costs of restrictions that make depository institutions morelikely to require infusions of taxpayer funds The failures thusmay have heightened public support for branching deregulation(see also Abrams and Settle [1993]) As noted above we did notnd any evidence that banking failures or distress in a stateaffected the speed with which the state deregulated but distressstill may be a common factor affecting all states
These technological economic and legal shocks generatedconditions that changed the long-standing balance favoring theantibranching forces The marginal value of lobbying to repealbranching restrictions increased just as the relative value to thesmall banks of maintaining branching restrictions was decliningThese nationwide shocks are common factors across states andthey are consistent with the positive duration dependence associ-ated with state-level deregulation that we demonstrated in FigureII In addition a broad change in lsquolsquoideologyrsquorsquo against government
QUARTERLY JOURNAL OF ECONOMICS1462
regulation could explain the positive duration dependence duringthe sample period
VI CONCLUSIONS
The private-interest theory of regulation can account for thepattern of bank branching deregulation of the 1970s and 1980sBeneciaries of branching regulation had supported a coalitionfavoring geographical restrictions despite their costs to consum-ers of nancial services Innovations that began in the 1970saltered the value of the restrictions to the affected parties and theresulting competition among interest groups can explain thesubsequent deregulation While some of our results are alsoconsistent with the public interest theorymdashfor example deregula-tion occurs earlier when small banks are in a relatively weaknancial positionmdashother results particularly evidence on theimportance of rivalries between small and large banks andbetween banking and insurance are difficult to explain with thepublic-interest approach We also nd that political-institutionalfactors affect deregulation although these variables may act asproxies for unmeasured economic interests The greater success ofthe private-interest theory to explain deregulation here than inprevious work may be due to the richer cross-sectional andtime-series variation in branching deregulation that can beexploited with a hazard model Future empirical studies ofendogenous deregulation thus may be most fruitful where changehas occurred across states over time such as in franchisinginsurance and public utilities (eg Joskow [1996])
Technological and nancial innovations will continue to erodethe benets to any interest group of maintaining regulatorybarriers in nancial services These forces are likely to bringabout reforms both domestically for example through legislationthat would increase bank powers (see Kroszner [1996] Krosznerand Rajan [1994 1997] and Kroszner and Stratmann [1998]) andinternationally for example through the extension of nancialservices provisions of NAFTAto reduce geographic barriers acrosscountries [Kroszner 1997]
APPENDIX EFFECTS OF DIFFERENT TYPES OF BRANCHING
DEREGULATION ON ECONOMIC GROWTH BANKING STRUCTUREAND BANK EFFICIENCY
We measure the importance of alternative forms of deregula-tion by their consequences for the economic growth and the
WHAT DRIVES DEREGULATION 1463
structure and efficiency of the banking sector across the states(see Jayaratne and Strahan [1996 1998]) To do so we run anOLS panel regression relating indicators for each type of deregula-tion to potential consequences Our dependent variables includeproxies for economic growth (growth in personal income) bankstructure (the log of the number of banking companies and thenumber of branches per banking company) and efficiency (the logof total noninterest expenses and the ratio of loan loss provisionsto total loans) The four deregulation indicators (MBHC branch-ing through merger and acquisition unrestricted statewide branch-ing and interstate banking) are equal to one in the years afterderegulation and zero before If a state deregulated before thesample period begins in 1970 the indicator is one throughout ifthe state does not deregulate then the indicator is zero through-out All regressions include state xed effects and time effects
Table VI shows the results of including the set of four
TABLE VIMULTIVARIATE REGRESSIONS OF THE EFFECTS OF DIFFERENT TYPES OF BRANCHING
DEREGULATION ON ECONOMIC GROWTH BANKING STRUCTUREAND BANK EFFICIENCY
Dependent variable
Growthin
personalincome
Log of of
bankingcompanies
Branchesper
bankingcompany
Log ofnon-
interestexpenses
Loan lossprovisions
(loans 1leases)
Merger amp acquisitionbranchingindicator
0013(0004)
2 0162(0037)
1146(0453)
2 0107(0035)
2 0007(0001)
Unrestricted state-wide branchingindicator
2 0008(0005)
0086(0059)
2 0043(0574)
0070(0037)
0005(0002)
Interstate bankingindicator
0006(0005)
2 0052(0057)
1064(0511)
0005(0030)
2 0001(0001)
Multibank holdingcompany indicator
2 0005(0006)
2 0096(0052)
2 0364(0529)
2 0095(0036)
2 0001(0001)
Dependent variablemean [median]
0086[0082]
4881[5136]
6649[5055]
13812[13803]
0009[0006]
This table contains regressions of state-level economic growth and state-level measures of the structureand efficiency of the banking industry Each dependent variable is regressed on a set of state indicatorvariables a set of year indicator variables and the set of four deregulation indicators these deregulationindicators are equal to one after deregulation and zero before If a state was always regulated the indicator iszero throughout if it was always deregulated the indicator is one throughout Sample period is 1975 to 1991and Delaware and South Dakota are dropped so N 5 833 Noninterest expenses become available only in1984 so N 5 392 The coefficient on the deregulation indicators is reported with robust standard errors inparentheses and denote statistically signicant at the 1 5 and 10 percent levels
QUARTERLY JOURNAL OF ECONOMICS1464
deregulation indicators simultaneously The branching by mergerand acquisition indicator is statistically signicant and of theexpected sign in all ve of the equations and larger in absolutevalue than any of the other coefficients on the indicators for theother types of deregulation Since there may be multicollinearityamong our indicators we also ran univariate specications whichincluded each indicator separately and the results are verysimilar These results imply that branching by merger andacquisition is the most important type of deregulation measuredby its consequences for economic growth banking structure andbank efficiency See Kroszner and Strahan [1998] for more details
UNIVERSITY OF CHICAGO GRADUATE SCHOOL OF BUSINESS AND NATIONAL BUREAU OF
ECONOMIC RESEARCH
FEDERAL RESERVE BANK OF NEW YORK RESEARCH amp MARKET ANALYSIS GROUP
REFERENCES
Abrams Burton A and Russell F Settle lsquolsquoPressure-Group Inuence and Institu-tional Change Branch-Banking Legislation during the Great DepressionrsquorsquoPublic Choice LXXVII (1993) 687ndash705
Amel Dean lsquolsquoState Laws Affecting the Geographic Expansion of CommercialBanksrsquorsquo unpublished manuscript Board of Governors of the Federal ReserveSystem 1993
Becker Gary S lsquolsquoA Theory of Competition among Pressure Groups for PoliticalInuencersquorsquo Quarterly Journal of Economics XCVIII (1983) 371ndash400
Berger Allen Anil Kashyap Joseph Scalise lsquolsquoThe Transformation of the U SBanking Industry What a Long Strange Trip Itrsquos Beenrsquorsquo Brookings Papers onEconomic Activity 2 (1995) 55ndash218
Berger Allen Anthony Saunders Joseph Scalise and Gregory Udell lsquolsquoThe Effectsof Bank Mergers and Acquisitions on Small Business Lendingrsquorsquo Journal ofFinancial Economics L (1998) 187ndash229
Brennan Geoffrey and James Buchanan lsquolsquoTowards a Tax Constitution forLeviathanrsquorsquo Journal of Public Economics VIII (1977) 255ndash273
Brickley James A and Christopher M James lsquolsquoThe Takeover Market CorporateBoard Composition and Ownership Structure The Case of Bankingrsquorsquo Journalof Law and Economics XXX (1987) 161ndash180
Cole Rebel and John Wolken lsquolsquoFinancial Services Used by Small BusinessesEvidence from the 1993 Survey of Small Business Financesrsquorsquo Federal ReserveBulletin LXXXI (1994) 629ndash676
Dixit Avinash The Making of Economic Policy A Transactions-Cost PoliticsPerspective (Cambridge MA MIT Press 1996)
Economides Nicolas R Glenn Hubbard and Darius Palia lsquolsquoThe Political Economyof Branching Restrictions and Deposit Insurancersquorsquo Journal of Law andEconomics XXIX (1996) 667ndash704
Flannery Mark lsquolsquoThe Social Costs of Unit Banking Restrictionsrsquorsquo Journal ofMonetary Economics XIII (1984) 237ndash249
Greene William H Econometric Analysis third edition (Upper Saddle River NJPrentice-Hall 1997)
Gunther Jeffery lsquolsquoRegional Capital Imbalances and the Removal of InterstateBanking Restrictionsrsquorsquo Economics Letters XLIV (1994) 439ndash442 lsquolsquoThe Erosion of Intrastate Branch Banking Restrictionsrsquorsquo unpublishedmanuscript Federal Reserve Bank of Dallas 1996
Irwin Douglas and Randall Kroszner lsquolsquoInterests Institutions and Ideology inSecuring Policy Change The Republican Conversion to Trade Liberalizationafter Smoot-Hawleyrsquorsquo Journal of Law and Economics XLII (1999) 643ndash673
WHAT DRIVES DEREGULATION 1465
James Christopher lsquolsquoAn Analysis of Intra-Industry Differences in the Effect ofRegulation The Case of Deposit Rate Ceilingsrsquorsquo Journal of Monetary Econom-ics XII (1983) 417ndash442
Jarrell Gregg lsquolsquoChange at the Exchange The Causes and Effects of DeregulationrsquorsquoJournal of Law and Economics XXVII (1984) 273ndash312
Jayaratne Jith and Philip E Strahan lsquolsquoThe Finance-Growth Nexus Evidencefrom Bank Branch Deregulationrsquorsquo Quarterly Journal of Economics CI (1996)639ndash670
Jayaratne Jith and Philip E Strahan lsquolsquoEntry Restrictions Industry Evolutionand Dynamic Efficiency Evidence from Commercial Bankingrsquorsquo Journal of Lawand Economics XLIX (1998) 239ndash274
Joskow Paul lsquolsquoComment on lsquoPower Struggles Explaining Deregulatory Reformsin Electricity Marketsrsquo rsquorsquo Brookings Papers on Economic Activity Microeconom-ics (1996) 251ndash264
Joskow Paul and Roger Noll lsquolsquoRegulation in Theory and Practice An Overviewrsquorsquoin Studies in Public Regulation Gary Fromm ed (Cambridge MA MITPress 1981) pp 1ndash65
Kahn Matthew E and John G Matsusaka lsquolsquoDemand for Environmental GoodsEvidence from Voting Patterns on California Initiativesrsquorsquo Journal of Law andEconomics XL (1997) 137ndash173
Kalbeisch J and R Prentice The Statistical Analysis of Failure Time Data (NewYork John Wiley amp Sons 1980)
Kane Edward J lsquolsquoDe Jure Interstate Banking Why Only NowrsquorsquoJournal of MoneyCredit and Banking XXVIII (1996) 141ndash161
Kiefer Nicholas lsquolsquoEconomic Duration Data and Hazard Functionsrsquorsquo Journal ofEconomic Literature XXVI (1988) 646ndash679
Kroszner Randall S lsquolsquoThe Evolution of Universal Banking and Its Regulation inTwentieth Century Americarsquorsquo in Universal Banking Financial System DesignReconsidered Anthony Saunders and Ingo Walter eds (New York IrwinProfessional Publishers 1996) pp 70ndash99 lsquolsquoThe Political Economy of Banking and Financial Regulation in the U Srsquorsquo inThe Banking and Financial Structure in the NAFTA Countries and ChileGeorge M von Furstenberg ed (Boston Kluwer Academic Publishers 1997)pp 200ndash213
Kroszner Randall S and Raghuram G Rajan lsquolsquoIs the Glass-Steagall ActJustied A Study of the U S Experience with Universal Banking before1933rsquorsquo American Economic Review LXXXIV (1994) 810ndash832
Kroszner Randall S and Raghuram G Rajan lsquolsquoOrganization Structure andCredibility Evidence from Bank Securities Activities before the Glass-Steagall Actrsquorsquo Journal of Monetary Economics XXXIX (1997) 475ndash516
Kroszner Randall S and Philip E Strahan lsquolsquoRegulatory Incentives and the ThriftCrisis Dividends Mutual-to-Stock Conversions and Financial DistressrsquorsquoJournal of Finance LI (1996) 1285ndash1320
Kroszner Randall S and Philip E Strahan lsquolsquoWhat Drives Deregulation Econom-ics and Politics of the Relaxation of Bank Branching Restrictionsrsquorsquo NBERWorking Paper No 6637 1998
Kroszner Randall S and Thomas Stratmann lsquolsquoInterest Group Competition andthe Organization of Congress Theory and Evidence from Financial ServicesPolitical Action Committeesrsquorsquo American Economic Review LXXXVIII (1998)pp 1163ndash1187
Kroszner Randall S and Thomas Stratmann lsquolsquoDoes Political Ambiguity PayCorporate Campaign Contributions and Rewards to Legislator Reputationrsquorsquounpublishedmanuscript University of Chicago Graduate School of Business1999
Lin D Y and L J Wei lsquolsquoThe Robust Inference for the Cox Proportional HazardsModel Journal of the American Statistical Association LXXXIV (1989)1074ndash1078
Macey Jonathan R and Geoffrey P Miller Banking Law and Regulation (BostonLittle Brown and Company 1992)
Makinson Larry Open Secrets The Encyclopedia of Congressional Money andPolitics (Washington DC Congressional Quarterly 1992)
Niskanen WilliamBureaucracy and Representative Government (Chicago Aldine-Atherton 1971)
QUARTERLY JOURNAL OF ECONOMICS1466
Noll Roger lsquolsquoComment on Peltzmanrsquorsquo Brookings Papers Microeconomics SpecialIssue (1989) 48ndash58
North Douglass lsquolsquoA Transactions Cost Theory of Politicsrsquorsquo Journal of TheoreticalPolitics II (1990) 355ndash367
Peltzman Sam lsquolsquoToward a More General Theory of Regulationrsquorsquo Journal of Lawand Economics XIX (1976) 109ndash148 lsquolsquoConstituent Interest and Congressional Votingrsquorsquo Journal of Law andEconomics XXVII (1984) 181ndash210 lsquolsquoThe Economic Theory of Regulation after a Decade of DeregulationrsquorsquoBrookings Papers Microeconomics Special Issue (1989) 1ndash41
Peterson Mitchell and Raghuram Rajan lsquolsquoThe Benets of Lending RelationshipsEvidence from Small Business Datarsquorsquo Journal of Finance XLIX (1994) 3ndash38
Poole Keith and Howard Rosenthal Congress A Political-Economic History ofRoll Call Voting (Oxford Oxford University Press 1997)
Stata Stata Reference Manual Release 5 (College Station TX Stata Press 1997)Stigler George J lsquolsquoThe Theory of Economic Regulationrsquorsquo Bell Journal of Economics
and Management Science II (1971) 3ndash21Stigler George J Chicago Studies in the Political Economy (Chicago University
of Chicago Press 1988)Strahan Philip E and James Weston lsquolsquoSmall Business Lending and the Changing
Structure of the Banking Industryrsquorsquo Journal of Banking and Finance XXII(1998) 821ndash845
Sylla Richard John Legler and John Wallis lsquolsquoBanks and State Public Finance inthe New Republic The United States 1790ndash1860rsquorsquo Journal of EconomicHistory XLVII (1987) 391ndash403
White Eugene The Regulation and Reform of the American Banking System1900ndash1929 (Princeton Princeton University Press 1983)
Winston Clifford lsquolsquoEconomic Deregulation Days of Reckoning for Microecono-mistsrsquorsquo Journal of Economic Literature XXXI (1993) 1263ndash1289
WHAT DRIVES DEREGULATION 1467
IV METHODS AND RESULTS
In this section we rst estimate how the variables describedabove inuence the timing of state-level branching deregulationusing a hazard model Next we discuss the robustness of theresults including a test of whether the ex post impact of deregula-tion on the different interest groups is consistent with ourinterpretation of the coefficients We then investigate federalinterstate branching deregulation with a probit model to explorewhether the factors that affect the timing of intrastate deregula-tion also inuence voting in the House of Representatives oninterstate deregulation
A Hazard Model
We consider the period from the beginning of our sample(1970) until deregulation as the lsquolsquoduration of regulationrsquorsquo or thelsquolsquotime until deregulationrsquorsquo The hazard rate h(t) is the likelihoodthat a state deregulates at time t given that the state has not yetderegulated [Kalbeisch and Prentice 1980 Kiefer 1988 Greene1997] To model the duration of regulation we must decide whatstructure if any to impose on the hazard function The Kaplan-Meier product-limit estimator provides a simple nonparametricway to estimate the shape of the hazard function over time[Greene 1997]21 Figure II graphs this estimate for our data andshows that the hazard function is relatively at in the early yearsand then grows steeper in the later years
A duration model that can approximate this shape is theWeibull proportional hazards model The hazard rate functiontakes the form
h[t x(t) b] 5 h0(t) exp[x(t)8b]
where x(t) is a vector of time-varying covariates b is a vector ofunknown parameters to be estimated and the baseline hazard
variables Note also that Nebraska has a nonpartisan unicameral legislature Weassume that the party of the Governor controls the state government for Nebraskaand set the unied government indicator to one As another alternative we triedincluding a variable that is one if party control of the legislature or governorshipchanges but it did not have an economically or statistically signicant effect anddid not affect any of the other results
21 Denote each of the K years in our sample period Tk and order them suchthat T1 T2 and so on Let nk be the number of states that have not yet deregulatedby Tk and dk be the number of states that deregulate in year Tk The Kaplan-Meierestimate of the hazard rate in each year is dknk In other words the hazard rate isthe ratio of the number of states that actually deregulate in each year to thenumber of states that have not yet deregulated in that year
QUARTERLY JOURNAL OF ECONOMICS1448
rate h0(t) is ptp 2 1 with shape parameter p that will be estimatedfrom the data When p 1 this model displays a monotonicallyincreasing hazard rate To estimate the parameters we maximizethe following log-likelihood function [Kiefer 1988]
L(b) 5 oi 5 1
N
di ln h[ti x(ti) b] 2 oi 5 1
N
e 0
tih[u x(u) b] du
where N indexes the number of states at risk at the beginning ofthe sample period di is an indicator equal to one for the statesthat deregulated by the end of the sample period and zero for thestates that do not deregulate (ie the censored observations) andti is the time of deregulation for the ith state This formulationassumes that the time until deregulation is independent acrossstates but the observations of the same state over time are notassumed to be independent In addition we use a robust estima-tion procedure to calculate the standard errors [Lin and Wei1989]
We exclude states that deregulated before 1970 from theanalysis so N equals 39 states During the sample period 36 ofthe 39 states deregulate We have three censored observations forwhich di is zero (Arkansas Iowa and Minnesota) but our results
FIGURE IIKaplan-Meier Estimate of Hazard Rate for Branching Deregulation
WHAT DRIVES DEREGULATION 1449
are not sensitive to the inclusion of these censored observationsSince we observe each state in each year up to and including theyear of deregulation we have a total of 637 observations Table IIreports summary statistics and correlation coefficients for theexplanatory variables we use in our analysis
We use the Weibull model rather than a model that does notimpose any structure on the baseline hazard rate so that we cancalculate the change in the expected time to deregulation for agiven change in the levels of the covariates22 In the Weibullmodel we can invert the hazard function and map it into the timedomain Rewriting the Weibull model in this way the log of thetime to deregulation T is a linear function of the political economyfactors and an error term ln (T ) 5 x8b 1 e23 Because we assumethat the baseline hazard rate is Weibull with a shape characterizedby p the new coefficients will be scaled by p that is b 5 2 bp
The b coefficients represent the percentage change in thetime to deregulation for a one-unit change in the covariates Apositive coefficient implies that an increase in the variableincreases the expected time until deregulation To gauge theeconomic importance of the effects we will multiply the b by thestandard deviation of the explanatory variables and then evaluatehow much the changes in the variables raise or lower the expectedtime to deregulation24 In Table III we report the b coefficients
22 The Cox proportional hazards model does not impose any structure on thebaseline hazard rate h0(t) and takes the following form
h[t x(t) b] 5 h0(t) exp [x(t)8b]
Common factors that affect the probability of deregulation in the same way for allstates therefore do not inuence the estimated coefficients The cost of the Coxmodel is that because there is no structure on the baseline we can calculatechanges in only relative hazard rates associated with changes in the covariatesThe Weibull model however provides sufficient structure so that we can translateour estimates into a log expected time metric that allows us to calculate the changein the expected time to deregulation for a given change in the covariates SeeKiefer [1988] Greene [1997] and Stata [1997] When we compare the coefficientsestimated by the Cox and Weibull models on our data they are very close (seeKroszner and Strahan [1998]) so imposing a Weibull structure does not appear todistort our estimates
23 The error term e is independent of x and has an extreme value distributionscaled by 1p (see Kiefer [1988] Greene [1997] and Stata [1997] for more details onthis log expected time parameterization of the Weibull model)
24 The transformation to the log expected time domain is only strictly correctwhen the covariates are constant over time Our interpretation of the coefficients ispossible in the time-varying model if we assume that a one-standard-deviationdifference would hold over the entire sample period that is the deregulationwould occur so many years earlier or later if the variable in question were higherby one standard deviation from 1970 to 1992 Note that the mean number of yearsto deregulation in the sample is 158
QUARTERLY JOURNAL OF ECONOMICS1450
TABLE IISUMMARY STATISTICS FOR VARIABLES IN HAZARD MODEL 1970ndash1992
Panel A Univariate statistics MeanStandarddeviation Minimum Maximum
(1) (2) (3) (4)Small bank asset share of all
banking assets in state 011 005 001 023Relative size of insurance to
banking plus insurance inthe state 047 009 024 080
Indicator is 1 if banks can sellinsurance in the state 020 040 0 1
Small rm share of the numberof rms in the state 088 002 078 095
Capital ratio of small banksrelative to large in the state 002 001 2 001 009
Share of state government con-trolled by Democrats 065 036 0 1
Indicator is 1 if state controlledby one party 054 050 0 1
Average yield on bank loans inthe state minus Fed fundsrate 002 002 2 004 008
Panel B CorrelationsSmallbank
Relativesize of
insuranceInsuranceindicator
Smallrm
Smallbank
capitalDem
controlSingleparty
(1) (2) (3) (4) (5) (6) (7)Small bank asset share
of all banking assetsin state 1
Relative size of insur-ance to banking plusinsurance in the state 2 034 1
Indicator is 1 if bankscan sell insurance inthe state 020 2 012 1
Small rm share of thenumber of rms inthe state 030 2 011 005 1
Capital ratio of smallbanks relative tolarge in the state 2 066 020 2 010 2 006 1
Share of state govern-ment controlled byDemocrats 2 006 001 2 018 2 018 005 1
Indicator is 1 if statecontrolled by oneparty 007 012 001 2 006 2 006 030 1
Average yield on bankloans in the stateminus Fed funds rate 2 001 2 015 003 022 2 010 2 003 2 006
The number of observations is 637 except for the average interest rate on bank loans which becomesavailable only in 1976 so the number of observations is 408
TABLE IIIHAZARD MODEL OF POLITICAL ECONOMY FACTORS AFFECTING THE TIMING OF STATE
BRANCHING DEREGULATION 1970ndash1992
(1) (2) (3) (4) (5) (6)Small bank asset share of
all banking assets instate
648(122)
519(137)
632(135)
516(118)
488(133)
445(088)
Capital ratio of smallbanks relative to largein the state
1325(348)
921(319)
1301(362)
1067(346)
1144(375)
906(381)
Relative size of insurancein states where banksmay sell insurance 0otherwise
324(161)
259(149)
315(165)
285(154)
272(159)
055(134)
Indicator is 1 if banksmay sell insurance inthe state
045(010)
037(011)
045(010)
038(008)
036(009)
001(012)
Relative size of insurancein states where banksmay not sell insurance0 otherwise
2 093(043)
2 090(033)
2 102(051)
2 093(048)
2 096(053)
2 004(039)
Small rm share of thenumber of rms in thestate
2 972(211)
2 614(215)
2 952(211)
2 943(230)
2 952(270)
2 1509(276)
Share of state govern-ment controlled byDemocrats
031(012)
023(011)
030(012)
027(011)
026(011)
011(013)
Indicator is 1 if state con-trolled by one party
2 004(010)
2 004(007)
2 004(007)
2 001(009)
2 002(010)
017(009)
Average yield on bankloans in the stateminus Fed funds rate mdash
023(319) mdash mdash mdash mdash
Bank failure rate mdash mdash092
(111) mdash mdash mdash
Indicator is 1 if state hasunit banking law mdash mdash mdash
018(010)
019(010)
030(009)
Indicator is 1 if statechanges bank insur-ance powers mdash mdash mdash mdash
2 008(012)
2 026(013)
Includes regionalindicators No No No No No Yes
N 637 408 637 637 637 637Log likelihood 2 374 900 2 360 2 212 2 191 1192p-value of x 2 for
regression 001 001 001 001 001 001
The hazard model is Weibull where the dependent variable is the log expected time to branchingderegulation ln (T) 5 bx 1 eAll variables are measured for each state in each year Small bank asset share isthe percent of banking assets in the state held by banks below the median size of banks in each state in eachyear Relative capital ratio is the capital to assets ratio of small banks minus that of large banks Size ofinsurance relative to banking plus insurance in the state is measured as gross state product from insurancedivided by gross state product from insurance plus banking Insurance indicator is 1 if state law permits banksto enter the insurance business Small rm share is the percent of all establishments in the state that havefewer than twenty employees Party control variable is the share of the three bodies of state governmentcontrolled by Democrats One party control indicator is 1 if the same party controlrsquos the governorship and hasmajorities in both chambers of the state legislature Average yield on bank loans equals total interest incomeon all domestic loans made by banks in the state divided by total loans Bank failure rate is the ratio of assetsin failed banks during a given state and year divided by total bank assets in that state and year Unit bankingindicator equals 1 for states with unit banking restrictions The change in insurance powers indicator equalsone if the state changed the law to permit banks to sell insurance during the sample period We divide thecountry into four regions and include regional indicators in column (5) Robust standard errors are inparentheses and denote statistically signicant at the 1 5 and 10 percent levels
QUARTERLY JOURNAL OF ECONOMICS1452
The positive and statistically signicant coefficient on smallbank share implies that a greater small bank share delaysderegulation and this effect is economically important A one-standard-deviation increase in the small bank share results in a30 percent increase in the time until deregulation or about 47years This result supports the implications of private-interesttheory concerning intraindustry rivalry and is contrary to thepublic-interest theory
The positive and statistically signicant coefficient on therelative performance of small banks in the state implies thatstates deregulate later when small banks are relatively strongnancially25 A one-standard-deviation increase in the relativecapital-to-asset ratio results in a 15 percent rise in the time untilderegulation or about 24 years This result is consistent with theprivate-interest theory but may also be consistent with a public-interest interpretation because the risk of failures (and a costlytaxpayer-funded bailout) increases as small bank nancial healthdeclines
Interindustry competition also helps explain the timing ofderegulation In states where banks can sell insurance a rela-tively large insurance sector is associated with an increase in theexpected time to deregulation26 A one-standard-deviation in-crease in the relative size of the insurance sector in those statesthat permit banks to sell insurance leads to a 22 percent increasein the time until deregulation or about 35 years27 This result isconsistent with the private-interest theory but not the public-interest theory
Deregulation occurs earlier in states where small bank-dependent rms are relatively numerous A one-standard-
25 We also allowed small bank and large bank capital-to-asset ratios to enterseparately Their coefficient estimates were nearly equal in absolute value and ofopposite signs and so the data do not reject the relative small to large bank capitalratio specication we have chosen
26 When we use the relative size of the insurance sector in the state as awhole rather than insurance relative to banking plus insurance in the state weobtain similar results When we simultaneously include both measures of therelative size of insurance the interactions of the lsquolsquoinsurance relative to bankingplus insurancersquorsquo variable continue to be statisticallysignicant but the interactionsof lsquolsquoinsurance relative to the state as a wholersquorsquo variable are not
27 We have also estimated the model with the indicator variable equal to onefor states where banks may sell insurance but without the variables measuringtheimportance of insurance relative to banking In this model the coefficient on theinsurance indicator variablemdashan estimate of the difference in the expected time toderegulation in states where banks may sell insurance relative to states wherebanks may not sell insurancemdashequals 027 and is statistically signicant at the 5percent level This suggests that branching deregulation takes about 45 yearslonger in states where banks may sell insurance
WHAT DRIVES DEREGULATION 1453
deviation increase in the share of small rms reduces the timeuntil deregulation by 18 percent or about three years This resultconcerning the interests of consumers of banking services isconsistent with both the private- and public-interest theories
Column (2) of Table III introduces the average interest rateon loans to the model as a rough proxy for the cost to bankborrowers of these regulations These data become available onlyin 1976 so our sample size shrinks to 408 observations Thecoefficient on this variable is small and not statistically signi-cant The public-interest theory would imply that the coefficientshould be negative while the implication of the economic theorydoes not have a clear sign prediction
When we add the failure rate of banks to the specication incolumn (3) its coefficient is small and statistically insignicantWe do not nd a linkage between the timing of state deregulationand statewide banking distress Given that deposit insurance isfunded federally however bank failures may raise the publicrsquosconcerns about the costs of the regulations at the national ratherthan the state level This interpretation can explain why the rateof bank failure does not affect the timing of state-level deregula-tion even though bank instability does appear to rise generallyduring the period of increasing deregulation (see Section V below)
Finally the partisan structure of the state government alsoinuences when states deregulate As expected a higher propor-tion of Democrats in the government tends to delay deregulationA one-standard-deviation rise in the share of the governmentcontrolled by Democrats slows the deregulation by about twoyears Whether the state is dominated by one party however doesnot appear to affect the timing of the deregulation
To summarize our ndings from the hazard model privateinterests appear to play an important role in the deregulatoryprocess While private interests and public interests do sometimescoincide the results on the relative share of small banks and largebanks and on the relative size of insurance where banks competeare consistent with a private-interest approach but are difficult toexplain on public-interest grounds
B Robustness Tests
Omitted Variables States begin the sample period withdifferent degrees of restrictions on bank expansion These initialconditions may proxy for omitted factors determining the political-economy equilibrium in the state that could then affect the
QUARTERLY JOURNAL OF ECONOMICS1454
subsequent timing of deregulation For instance some statesbegan the period with unit banking while other states allowedlimited branching and the formation of multibank holdingcompanies
To test whether the stringency of initial restrictions onbranching matters we rst divide the sample into states with andwithout the most extreme form of branching regulation unitbanking28 Column (4) of Table III includes an indicator variableequal to one if the state began the period with unit bankingrestrictions The coefficient on this variable is positive andstatistically signicant but the other results remain virtuallyunchanged Next we include an indicator variable that is one ifthe state permitted multibank holding companies (MBHCs) toexist at the beginning of our sample period Thirty-four statesbegan the period allowing MBHCs The coefficient estimate forthis indicator variable (not reported) is small and not statisticallysignicant and its inclusion does not change the effects of theother variables
Eleven states expand bank powers to include the sale ofinsurance products during our sample period This could proxy foromitted factors about the relative strength of the insuranceindustry in the state In column (5) we include an indicatorvariable that is one if the state changes the law during our sampleperiod to permit banks to sell insurance business and zerootherwise The coefficient estimate is small and statisticallyinsignicant and does not affect the other results29
In column (6) we introduce four regional indicator variablesNorth South Midwest and West30 The coefficients on small bankshare the relative capital-to-assets ratio and small rm sharechange little Although the coefficient estimate still suggests that
28 We classify states that prohibited branching but permitted banks toestablish facilities as unit banking states The unit banking indicator is one for thefollowing states CO AR FL IL IA KS MN MO MT NE ND OK TX WI WVand WY
29 We conducted a likelihood ratio test of whether the unit banking statesthe MBHC states and the states that changed their regulations of bank insurancepowers could be pooled with the other states In all three cases we could not rejectpooling of the data
30 Since intrastate deregulation generally preceded interstate deregulationand the latter typically took the form of regional interstate compacts thepossibility of participating in a regional interstate banking compact could haveinuenced the decision to deregulate intrastate branching Our denitions of theregions are region 1 (South) contains AL AR DC FL GA KY LA MS NC OKSC TN TX and VA region 2 (Northeast) contains CT MA MD ME NH NJ NYPA RI VT and WV region 3 (Midwest) contains IA IL IN KS MI MN MO NEND OH SD and WI region 4 (West) contains the other states
WHAT DRIVES DEREGULATION 1455
in states where banks can sell insurance a larger relative size ofthe insurance sector delays deregulation the effect is no longerstatistically signicant The coefficient on the indicator variablefor whether a state expands bank powers to permit insurancesales during the sample period however now becomes statisti-cally signicantly negative This implies that deregulation occursroughly 41 years earlier in states where the insurance lobby maybe relatively weak While the inclusion of the regional indicatorsdoes not affect the other results the results on insurance doappear to be sensitive to the specication of the hazard model31
Estimation Technique To check the robustness of using ahazard model and to investigate whether the time-series or thecross-sectional variation in each of the factors accounts for thehazard results we estimate two simple linear probability models(neither are reported) First we run a pooled time-series cross-section OLS regression using the 637 observations from thehazard model The dependent variable equals one in the year ofderegulation and zero otherwise We include the same set ofregressors that were used in the hazard model and add state-levelxed effects This specication thus removes the cross-statevariation from the independent variables and focuses on theirmovements over time In this model the signs of the coefficientson the interest group factors are consistent with the results fromthe hazard model and the coefficients for the small bank relativecapital ratio and for the relative size of insurance where states cansell insurance are statistically signicant
In the second model we eliminate the time-series variationby averaging each of the explanatory variables over time We thushave one observation per state that had branching restrictions asof 1970 (N 5 39) The dependent variable equals the inverse of thenumber of years until deregulation which represents an estimateof the statersquos probability of deregulating in a year under theassumption that this probability is constant over time and eachyear provides an independent observation32 Although we haveonly 39 observations the coefficients on the small bank marketshare small bank relative capital ratio and small rm share
31 As further robustness check we also tried a population density variablewhich Abrams and Settle [1993] found relevant for regulatory change during the1920s and 1930s but its coefficient was not statistically signicant and did notaffect the other results
32 For instance the dependent variable would take the value of 01 for astate that deregulated in 1980 For the three states that did not deregulate thedependent variable equals 0
QUARTERLY JOURNAL OF ECONOMICS1456
variables are statistically signicant The signs of the effects ofthe insurance variables are the same as those reported in thehazard model but not statistically signicant
The hazard model and linear probability models thus producevery similar results The estimates from the latter suggest thatthe effects for the small bank market share and small rm sharevariables are driven primarily by cross-state variation the effectof the insurance variable is driven primarily by time-seriesvariation and the small bank capital effect is driven by bothcross-sectional and time-series variation
Ex Post Consequences of Deregulation Finally to check theplausibility of our interpretation of the hazard results we con-sider whether the ex post consequences of deregulation areconsistent with the ex ante lobbying positions we attribute to eachinterest group (see Kroszner and Strahan [1998] for details) Wend that small banks lose market share following deregulationand in states where banks can enter the insurance business theinsurance sector shrinks relative to the banking sector followingderegulation Borrowers also benet from deregulation throughlower average interest rates on loans These ndings support theprivate-interest interpretation of the coefficients in the hazardmodel groups that will benet lobby to speed deregulation andthose that will be harmed lobby to slow it
C Voting on Interstate Branching Deregulation in the U SHouse of Representatives
We now examine whether the forces driving intrastate branch-ing deregulation also drive interstate deregulation Financialservices interests are active contributors and lobbyists in Washing-ton Their political action committees constitute the largest groupof contributors to legislators providing nearly 20 percent of totalcongressional campaign contributions [Makinson 1992] and muchof their lobbying effort involves competition among rival interestswithin nancial services (see Kroszner and Stratmann [1998])
As noted above after virtually all states adopted intra- andinterstate branching deregulation the 1994 Riegle-Neal Act re-pealed the 1927 McFadden Act to eliminate all barriers tointerstate banking and branching by 1997 The key votes concern-ing the Riegle-Neal Act were either voice votes or extremelylopsided so we could not estimate a voting model from them Anumber of bills and amendments related to interstate branching
WHAT DRIVES DEREGULATION 1457
had been debated in Congress during the years prior to thepassage of the Riegle-Neal Act but a search of the weekly BNABanking Reporter and the Congressional Record produced onlyone roll-call vote related to interstate branching that was notlopsided This vote occurred in the House on November 14 1991and concerned an amendment sponsored by Wylie (R-OH) andNeal (D-NC) to introduce interstate banking and branchingderegulation to a nancial services reform package33 Althoughthe amendment passed by 210 to 208 the bill to which it wasattached subsequently was defeated
To check for the inuence of the factors we considered in thestate-level reforms we examine both the sponsorship of interstatebanking legislation and voting on the amendment The sponsorsof the Wylie-Neal amendment are from states with low small bankshares 004 in Ohio (Wylie) and 002 in North Carolina (Neal) Incontrast the sample mean in 1991 is 008 (median 5 007)Michigan home state of the Senatersquos sponsor of the 1994 Riegle-Neal Act also had relatively low small bank strength (small bankshare of 005)
Table IV reports estimates of a probit model where thedependent variable equals one if the legislator voted in favor ofthe amendment and zero otherwise The explanatory variablesare the same as those in the hazard model except that we use aparty affiliation indicator variable equal to one for Democrats andzero for Republicans as our proxy for political factors Thecoefficients in Table IV are the marginal effects (lsquolsquoslopesrsquorsquo) of aone-unit change of each variable on the probability that alegislator will vote for the amendment34 Note that a forcefavoring deregulation has a negative coefficient in the hazardmodel (Table III) but a positive coefficient in the probit model
Consistent with the state-level deregulation process legisla-tors are more likely to support the amendment if their states havea relatively low share of small banks As in the hazard model thefraction of small banks is the most important interest groupinuence on a legislatorrsquos voting decision A one-standard-deviation increase in small banksrsquo market share (from the mean)is associated with a decline in the probability of voting in favor of
33 The Wylie-Neal amendment also included provisions limiting certaininsurance and real estate powers of national banks [Congressional RecordNovember 14 1991 pp 10239ndash10242]
34 Since we have multiple legislators from each state we adjust the standarderrors to correct for the lack of independence among observations clustered in thesame state
QUARTERLY JOURNAL OF ECONOMICS1458
TAB
LE
IVM
AR
GIN
AL
EF
FE
CT
SF
RO
MA
PR
OB
ITM
OD
EL
OF
TH
EIN
FL
UE
NC
EO
FP
OL
ITIC
AL
EC
ON
OM
YF
AC
TO
RS
ON
TH
EH
OU
SE
VO
TE
ON
TH
EW
YL
IE-N
EA
L
AM
EN
DM
EN
TT
OP
ER
MIT
INT
ER
ST
AT
EB
AN
KB
RA
NC
HIN
G
NO
VE
MB
ER
14
1991
(1)
(2)
(3)
(4)
(5)
(6)
Smal
lban
kas
set
shar
eof
allb
anki
ng
asse
tsin
stat
e2
417
(12
4)2
454
(12
2)2
407
(12
0)2
443
(14
7)2
450
(14
9)2
490
(14
7)C
apit
alra
tio
ofsm
allb
anks
rela
tive
tola
rge
inth
est
ate
180
(37
5)2
68(3
42)
256
(37
5)1
69(3
78)
153
(38
0)2
45(4
25)
Rel
ativ
esi
zeof
insu
ranc
ein
stat
esw
here
bank
sm
ayse
llin
sura
nce
0ot
herw
ise
22
34
(10
2)2
249
(1
14)
22
33
(10
1)2
249
(1
04)
22
51
(10
2)2
154
(11
1)In
dica
tor
is1
ifba
nks
may
sell
insu
ranc
ein
the
stat
e0
79
(0
06)
074
(00
6)0
69
(0
06)
070
(00
6)0
70
(0
06)
067
(00
7)R
elat
ive
size
ofin
sura
nce
inst
ates
whe
reba
nks
may
not
sell
insu
ranc
e0
oth
erw
ise
180
(06
6)2
42
(0
72)
154
(0
67)
173
(06
6)1
72
(0
67)
203
(0
91)
Smal
lrm
shar
eof
the
num
ber
ofr
ms
inth
est
ate
21
23(0
92)
20
83(0
89)
21
62(1
07)
21
32(0
95)
21
30(1
00)
20
94(1
16)
Indi
cato
ris
1if
Rep
rese
ntat
ive
isa
Dem
ocra
t2
035
(00
7)2
035
(00
7)2
036
(00
7)2
035
(00
7)2
035
(00
7)2
036
(00
7)A
vera
geyi
eld
onba
nklo
ans
inth
est
ate
min
usF
edfu
nds
rate
mdash10
01
(51
5)mdash
mdashmdash
mdash
Ave
rage
fail
ure
rate
198
5to
1991
mdashmdash
228
(14
1)mdash
mdashmdash
Indi
cato
ris
1if
stat
eh
asun
itba
nkin
gla
wmdash
mdashmdash
004
(00
7)0
04(0
07)
20
01(0
08)
Indi
cato
ris
1if
stat
ech
ange
sba
nk
insu
ran
cepo
wer
smdash
mdashmdash
mdash2
001
(00
7)2
003
(00
7)In
clud
esre
gion
alin
dica
tors
N
oN
oN
oN
oN
oYe
sP
seud
o-R
20
170
180
170
170
170
18L
ogli
keli
hood
224
024
223
826
223
951
224
013
224
010
223
808
p-va
lue
ofx2
for
regr
essi
on
001
0
01
001
0
01
001
0
01
Th
ede
pen
den
tvar
iabl
eis
one
ifth
ele
gisl
ator
vote
sfo
rth
eam
endm
enta
nd
zero
ifag
ains
tT
he
repo
rted
coef
fici
ents
are
the
effe
cts
ofa
uni
tch
ange
ofth
ein
depe
nde
ntva
riab
le(f
rom
the
mea
n)on
the
prob
abili
tyof
voti
ng
infa
vor
ofth
eam
endm
ent
All
vari
able
sar
em
easu
red
byst
ate
N5
418
the
num
ber
ofR
epre
sent
ativ
esvo
tin
gon
the
amen
dmen
tS
mal
lban
kas
sets
har
eis
the
perc
ent
ofba
nki
ng
asse
tsin
the
stat
eh
eld
byba
nks
belo
wth
em
edia
nsi
zeof
ban
kin
each
stat
ein
each
year
Rel
ativ
eca
pita
lrat
iois
the
capi
talt
oas
sets
rati
oof
smal
lban
ksm
inu
sth
atof
larg
eba
nks
Siz
eof
insu
ran
cere
lati
veto
bank
ing
plu
sin
sura
nce
inth
est
ate
ism
easu
red
asgr
oss
stat
epr
oduc
tfro
min
sura
nce
divi
ded
bygr
oss
stat
epr
odu
ctfr
omin
sura
nce
plu
sba
nkin
gIn
dica
tor
ison
eif
stat
ela
wpe
rmit
sba
nks
toen
ter
the
insu
ran
cebu
sin
ess
Sm
all
rmsh
are
isth
epe
rcen
tofa
lles
tabl
ishm
ents
inth
est
ate
that
hav
efe
wer
than
twen
tyem
ploy
ees
Par
tyin
dica
tor
ison
eif
the
Rep
rese
nta
tive
isa
Dem
ocra
tA
vera
geyi
eld
onba
nklo
ans
equa
lsto
tali
nter
est
inco
me
onal
ldom
esti
clo
ans
mad
eby
ban
ksin
the
stat
edi
vide
dby
tota
lloa
ns
The
aver
age
bank
failu
rera
teis
the
rati
oof
asse
tsin
faile
dba
nks
duri
ng
agi
ven
stat
edi
vide
dby
tota
lban
kas
sets
inth
atst
ate
aver
aged
over
the
year
s19
85to
1991
Un
itba
nki
ng
indi
cato
req
ual
son
efo
rst
ates
wit
hun
itba
nkin
gre
stri
ctio
nsT
hech
ange
inin
sura
nce
pow
ers
indi
cato
req
ual
son
eif
the
stat
ech
ange
dth
ela
wto
perm
itba
nks
tose
llin
sura
nce
duri
ngth
esa
mpl
epe
riod
We
divi
deth
eco
unt
ryin
tofo
urre
gion
san
din
clu
dere
gion
alin
dica
tors
inco
lum
n(5
)R
obu
stst
anda
rder
rors
are
inpa
ren
thes
es
an
d
den
ote
stat
isti
cally
sign
ica
nta
tth
e1
5an
d10
perc
entl
evel
s
WHAT DRIVES DEREGULATION 1459
branching of approximately 17 percent The relative capital-to-assets ratio of small and large banks however does not have astatistically signicant effect in the probit model
The impact of rival interests outside of banking is alsoconsistent with intrastate deregulation results Where banks cansell insurance legislators from states with larger insurancesectors relative to banking are less likely to vote for the amend-ment35 A one-standard-deviation increase in the relative size ofinsurance in those states which permit banks to sell insurancedecreases the probability that a legislator will favor the amend-ment by about 13 percent Turning to consumer interests thecoefficient on small rm share is not statistically signicant butthe coefficient on interest rates in the state is A one-standard-deviation increase in the average interest rate on loans raises theprobability that a legislator will support the amendment byroughly 8 percent The coefficient estimate for the bankingdistress variable measured as the average bank failure rate from1985 to 1991 is small and statistically signicant Overall theprobit analysis of the vote on national branching deregulationsupports the private-interest theory of deregulation and providesa consistency check that the importance of interests operating onthe state legislatures are very similar to those operating at thefederal level
V WHY DID BRANCHING DEREGULATION BEGIN DURING THE
1970S SHOCKS TO THE EQUILIBRIUM SUPPORTING BANK
BRANCHING RESTRICTIONS
A complete explanation of regulatory exit should account forwhy deregulation begins in the 1970s as well as the timing of thestate-by-state reforms [Kane 1996] In this section we describebroad technological legal and economic shocks that altered thepolitical-economy equilibrium which had kept antibranching regu-lations little changed for at least 30 years
Beginning in the 1970s three innovations reduced the valueto the protected banks of local geographic monopolies (see TableV) First automatic teller machines (ATMs) helped to erode thegeographic ties between customers and banks Second checkablemoney market mutual funds and the Merrill Lynch Cash Manage-ment Account demonstrated that banking by mail and telephone
35 The positive and statistically signicant estimates on the other insurancevariables may reect insurance industry support for the amendmentrsquos provisionslimiting banksrsquo insurance powers
QUARTERLY JOURNAL OF ECONOMICS1460
provided a convenient alternative to local banks36 Third techno-logical innovation and deregulation reduced transportation andcommunication costs particularly since the 1970s thereby lower-ing the costs for customers to use distant banks By increasing theelasticity of deposits supplied to banks these innovations reducedthe value of geographical restrictions to their traditional benecia-ries and thereby reduced their incentive to ght to maintain them[Peltzman 1976]
On the lending side increasing sophistication of credit-scoring techniques following innovations in information process-ing nancial theory and the development of large credit databases began to diminish the value of knowledge that localbankers had about the risks of borrowers in the community As a
36 Regulation Q which limited the interest rates that banks could pay ondeposits may have helped to drive depositors away from banks when the gapbetween market rates and deposit ceilings grew during the 1970s The eliminationof interest rate ceilings on large denomination certicates of deposit during the1970s appears to have hurt smaller and retail-oriented banks relative to largerwholesale banks [James 1983]
TABLE VBROAD TRENDS IN COMMERCIAL BANKING 1950ndash1995
YearNumberof ATMs
Domesticbank
deposits(billions)
Moneymarketmutual
fund(billions)
Percent ofdeposits 1
moneyfunds heldby banks
Smallbanksrsquo
percent ofbankingassets
Averagenumber of
bankfailures
(1) (2) (3) (4) (5) (6)1950 0 $ 154 $ 0 100 NA 41955 0 191 0 100 NA 31960 0 228 0 100 24 21965 0 330 0 100 20 41970 0 479 0 100 18 61975 9750 775 4 99 18 61980 18500 1182 76 94 17 101985 61117 1787 242 88 14 601990 80156 2339 493 83 11 1791995 122706 2552 745 77 8 61
Column (1) ATM gures are from Bank Network News The EFT Network Data Book (New YorkFaulkner and Gray Inc) The 1975 gure was unavailable 9750 is the number of ATMs in 1978 the rst yearfor which complete data are available Columns (2)ndash(4) Banksrsquo domestic deposits are from the Reports ofIncome and Condition money market mutual funds are from the Flow of Funds Data on all bank depositsforeign plus domestic are only available beginning in 1970 The trend in banksrsquo share (column (4)) is the sameusing total deposits instead of domestic deposits Column (5) Percent of banking assets held by small bankswhere a small bank is dened as a commercial bank with less than $100 million in assets in 1994 dollarsThese data are based on the Reports of Income and Condition Data on small banks are not available before1960 Column (6) Five-year average number of bank failures where the nal year is indicated in the rstcolumn These data are from FDIC Annual Report and the Quarterly Banking Prole
WHAT DRIVES DEREGULATION 1461
result of these innovations a national market developed forresidential mortgages credit card receivables have been securi-tized and bank lending to small business now relies less on thejudgment of loan officers and more on standardized scoringmodels
These changes have increased the potential protability forlarge banks to enter what had been the core of small bankactivities Large banksrsquo incentive to increase their lobbying pres-sure to be able to expand into these markets has thus beenincreasing over time In fact small banksrsquo market share began todecline even prior to the branching deregulation (Table V) As thevalue of a local banking relationships declined small rms thatwere the main borrowers from the small banks also probablybecame more likely to favor the entry of large banks into localmarkets With the deadweight costs of preventing large bankentry rising the private-interest theory predicts that small localbanks would become less likely able to maintain the branchingrestrictions [Becker 1983] Deregulation that reduces deadweightcosts of regulation is also consistent with the public-interesttheory
Kane [1996] argues that another major shock to the oldequilibrium was the increasing public awareness of the costlinessof having government-insured but (geographically) undiversiednancial institutions In the late 1970s the failure rate of banksbegan to rise (Table V) In the 1980s the Savings and Loan crisisand taxpayer bailout further heightened public awareness aboutthe costs of restrictions that make depository institutions morelikely to require infusions of taxpayer funds The failures thusmay have heightened public support for branching deregulation(see also Abrams and Settle [1993]) As noted above we did notnd any evidence that banking failures or distress in a stateaffected the speed with which the state deregulated but distressstill may be a common factor affecting all states
These technological economic and legal shocks generatedconditions that changed the long-standing balance favoring theantibranching forces The marginal value of lobbying to repealbranching restrictions increased just as the relative value to thesmall banks of maintaining branching restrictions was decliningThese nationwide shocks are common factors across states andthey are consistent with the positive duration dependence associ-ated with state-level deregulation that we demonstrated in FigureII In addition a broad change in lsquolsquoideologyrsquorsquo against government
QUARTERLY JOURNAL OF ECONOMICS1462
regulation could explain the positive duration dependence duringthe sample period
VI CONCLUSIONS
The private-interest theory of regulation can account for thepattern of bank branching deregulation of the 1970s and 1980sBeneciaries of branching regulation had supported a coalitionfavoring geographical restrictions despite their costs to consum-ers of nancial services Innovations that began in the 1970saltered the value of the restrictions to the affected parties and theresulting competition among interest groups can explain thesubsequent deregulation While some of our results are alsoconsistent with the public interest theorymdashfor example deregula-tion occurs earlier when small banks are in a relatively weaknancial positionmdashother results particularly evidence on theimportance of rivalries between small and large banks andbetween banking and insurance are difficult to explain with thepublic-interest approach We also nd that political-institutionalfactors affect deregulation although these variables may act asproxies for unmeasured economic interests The greater success ofthe private-interest theory to explain deregulation here than inprevious work may be due to the richer cross-sectional andtime-series variation in branching deregulation that can beexploited with a hazard model Future empirical studies ofendogenous deregulation thus may be most fruitful where changehas occurred across states over time such as in franchisinginsurance and public utilities (eg Joskow [1996])
Technological and nancial innovations will continue to erodethe benets to any interest group of maintaining regulatorybarriers in nancial services These forces are likely to bringabout reforms both domestically for example through legislationthat would increase bank powers (see Kroszner [1996] Krosznerand Rajan [1994 1997] and Kroszner and Stratmann [1998]) andinternationally for example through the extension of nancialservices provisions of NAFTAto reduce geographic barriers acrosscountries [Kroszner 1997]
APPENDIX EFFECTS OF DIFFERENT TYPES OF BRANCHING
DEREGULATION ON ECONOMIC GROWTH BANKING STRUCTUREAND BANK EFFICIENCY
We measure the importance of alternative forms of deregula-tion by their consequences for the economic growth and the
WHAT DRIVES DEREGULATION 1463
structure and efficiency of the banking sector across the states(see Jayaratne and Strahan [1996 1998]) To do so we run anOLS panel regression relating indicators for each type of deregula-tion to potential consequences Our dependent variables includeproxies for economic growth (growth in personal income) bankstructure (the log of the number of banking companies and thenumber of branches per banking company) and efficiency (the logof total noninterest expenses and the ratio of loan loss provisionsto total loans) The four deregulation indicators (MBHC branch-ing through merger and acquisition unrestricted statewide branch-ing and interstate banking) are equal to one in the years afterderegulation and zero before If a state deregulated before thesample period begins in 1970 the indicator is one throughout ifthe state does not deregulate then the indicator is zero through-out All regressions include state xed effects and time effects
Table VI shows the results of including the set of four
TABLE VIMULTIVARIATE REGRESSIONS OF THE EFFECTS OF DIFFERENT TYPES OF BRANCHING
DEREGULATION ON ECONOMIC GROWTH BANKING STRUCTUREAND BANK EFFICIENCY
Dependent variable
Growthin
personalincome
Log of of
bankingcompanies
Branchesper
bankingcompany
Log ofnon-
interestexpenses
Loan lossprovisions
(loans 1leases)
Merger amp acquisitionbranchingindicator
0013(0004)
2 0162(0037)
1146(0453)
2 0107(0035)
2 0007(0001)
Unrestricted state-wide branchingindicator
2 0008(0005)
0086(0059)
2 0043(0574)
0070(0037)
0005(0002)
Interstate bankingindicator
0006(0005)
2 0052(0057)
1064(0511)
0005(0030)
2 0001(0001)
Multibank holdingcompany indicator
2 0005(0006)
2 0096(0052)
2 0364(0529)
2 0095(0036)
2 0001(0001)
Dependent variablemean [median]
0086[0082]
4881[5136]
6649[5055]
13812[13803]
0009[0006]
This table contains regressions of state-level economic growth and state-level measures of the structureand efficiency of the banking industry Each dependent variable is regressed on a set of state indicatorvariables a set of year indicator variables and the set of four deregulation indicators these deregulationindicators are equal to one after deregulation and zero before If a state was always regulated the indicator iszero throughout if it was always deregulated the indicator is one throughout Sample period is 1975 to 1991and Delaware and South Dakota are dropped so N 5 833 Noninterest expenses become available only in1984 so N 5 392 The coefficient on the deregulation indicators is reported with robust standard errors inparentheses and denote statistically signicant at the 1 5 and 10 percent levels
QUARTERLY JOURNAL OF ECONOMICS1464
deregulation indicators simultaneously The branching by mergerand acquisition indicator is statistically signicant and of theexpected sign in all ve of the equations and larger in absolutevalue than any of the other coefficients on the indicators for theother types of deregulation Since there may be multicollinearityamong our indicators we also ran univariate specications whichincluded each indicator separately and the results are verysimilar These results imply that branching by merger andacquisition is the most important type of deregulation measuredby its consequences for economic growth banking structure andbank efficiency See Kroszner and Strahan [1998] for more details
UNIVERSITY OF CHICAGO GRADUATE SCHOOL OF BUSINESS AND NATIONAL BUREAU OF
ECONOMIC RESEARCH
FEDERAL RESERVE BANK OF NEW YORK RESEARCH amp MARKET ANALYSIS GROUP
REFERENCES
Abrams Burton A and Russell F Settle lsquolsquoPressure-Group Inuence and Institu-tional Change Branch-Banking Legislation during the Great DepressionrsquorsquoPublic Choice LXXVII (1993) 687ndash705
Amel Dean lsquolsquoState Laws Affecting the Geographic Expansion of CommercialBanksrsquorsquo unpublished manuscript Board of Governors of the Federal ReserveSystem 1993
Becker Gary S lsquolsquoA Theory of Competition among Pressure Groups for PoliticalInuencersquorsquo Quarterly Journal of Economics XCVIII (1983) 371ndash400
Berger Allen Anil Kashyap Joseph Scalise lsquolsquoThe Transformation of the U SBanking Industry What a Long Strange Trip Itrsquos Beenrsquorsquo Brookings Papers onEconomic Activity 2 (1995) 55ndash218
Berger Allen Anthony Saunders Joseph Scalise and Gregory Udell lsquolsquoThe Effectsof Bank Mergers and Acquisitions on Small Business Lendingrsquorsquo Journal ofFinancial Economics L (1998) 187ndash229
Brennan Geoffrey and James Buchanan lsquolsquoTowards a Tax Constitution forLeviathanrsquorsquo Journal of Public Economics VIII (1977) 255ndash273
Brickley James A and Christopher M James lsquolsquoThe Takeover Market CorporateBoard Composition and Ownership Structure The Case of Bankingrsquorsquo Journalof Law and Economics XXX (1987) 161ndash180
Cole Rebel and John Wolken lsquolsquoFinancial Services Used by Small BusinessesEvidence from the 1993 Survey of Small Business Financesrsquorsquo Federal ReserveBulletin LXXXI (1994) 629ndash676
Dixit Avinash The Making of Economic Policy A Transactions-Cost PoliticsPerspective (Cambridge MA MIT Press 1996)
Economides Nicolas R Glenn Hubbard and Darius Palia lsquolsquoThe Political Economyof Branching Restrictions and Deposit Insurancersquorsquo Journal of Law andEconomics XXIX (1996) 667ndash704
Flannery Mark lsquolsquoThe Social Costs of Unit Banking Restrictionsrsquorsquo Journal ofMonetary Economics XIII (1984) 237ndash249
Greene William H Econometric Analysis third edition (Upper Saddle River NJPrentice-Hall 1997)
Gunther Jeffery lsquolsquoRegional Capital Imbalances and the Removal of InterstateBanking Restrictionsrsquorsquo Economics Letters XLIV (1994) 439ndash442 lsquolsquoThe Erosion of Intrastate Branch Banking Restrictionsrsquorsquo unpublishedmanuscript Federal Reserve Bank of Dallas 1996
Irwin Douglas and Randall Kroszner lsquolsquoInterests Institutions and Ideology inSecuring Policy Change The Republican Conversion to Trade Liberalizationafter Smoot-Hawleyrsquorsquo Journal of Law and Economics XLII (1999) 643ndash673
WHAT DRIVES DEREGULATION 1465
James Christopher lsquolsquoAn Analysis of Intra-Industry Differences in the Effect ofRegulation The Case of Deposit Rate Ceilingsrsquorsquo Journal of Monetary Econom-ics XII (1983) 417ndash442
Jarrell Gregg lsquolsquoChange at the Exchange The Causes and Effects of DeregulationrsquorsquoJournal of Law and Economics XXVII (1984) 273ndash312
Jayaratne Jith and Philip E Strahan lsquolsquoThe Finance-Growth Nexus Evidencefrom Bank Branch Deregulationrsquorsquo Quarterly Journal of Economics CI (1996)639ndash670
Jayaratne Jith and Philip E Strahan lsquolsquoEntry Restrictions Industry Evolutionand Dynamic Efficiency Evidence from Commercial Bankingrsquorsquo Journal of Lawand Economics XLIX (1998) 239ndash274
Joskow Paul lsquolsquoComment on lsquoPower Struggles Explaining Deregulatory Reformsin Electricity Marketsrsquo rsquorsquo Brookings Papers on Economic Activity Microeconom-ics (1996) 251ndash264
Joskow Paul and Roger Noll lsquolsquoRegulation in Theory and Practice An Overviewrsquorsquoin Studies in Public Regulation Gary Fromm ed (Cambridge MA MITPress 1981) pp 1ndash65
Kahn Matthew E and John G Matsusaka lsquolsquoDemand for Environmental GoodsEvidence from Voting Patterns on California Initiativesrsquorsquo Journal of Law andEconomics XL (1997) 137ndash173
Kalbeisch J and R Prentice The Statistical Analysis of Failure Time Data (NewYork John Wiley amp Sons 1980)
Kane Edward J lsquolsquoDe Jure Interstate Banking Why Only NowrsquorsquoJournal of MoneyCredit and Banking XXVIII (1996) 141ndash161
Kiefer Nicholas lsquolsquoEconomic Duration Data and Hazard Functionsrsquorsquo Journal ofEconomic Literature XXVI (1988) 646ndash679
Kroszner Randall S lsquolsquoThe Evolution of Universal Banking and Its Regulation inTwentieth Century Americarsquorsquo in Universal Banking Financial System DesignReconsidered Anthony Saunders and Ingo Walter eds (New York IrwinProfessional Publishers 1996) pp 70ndash99 lsquolsquoThe Political Economy of Banking and Financial Regulation in the U Srsquorsquo inThe Banking and Financial Structure in the NAFTA Countries and ChileGeorge M von Furstenberg ed (Boston Kluwer Academic Publishers 1997)pp 200ndash213
Kroszner Randall S and Raghuram G Rajan lsquolsquoIs the Glass-Steagall ActJustied A Study of the U S Experience with Universal Banking before1933rsquorsquo American Economic Review LXXXIV (1994) 810ndash832
Kroszner Randall S and Raghuram G Rajan lsquolsquoOrganization Structure andCredibility Evidence from Bank Securities Activities before the Glass-Steagall Actrsquorsquo Journal of Monetary Economics XXXIX (1997) 475ndash516
Kroszner Randall S and Philip E Strahan lsquolsquoRegulatory Incentives and the ThriftCrisis Dividends Mutual-to-Stock Conversions and Financial DistressrsquorsquoJournal of Finance LI (1996) 1285ndash1320
Kroszner Randall S and Philip E Strahan lsquolsquoWhat Drives Deregulation Econom-ics and Politics of the Relaxation of Bank Branching Restrictionsrsquorsquo NBERWorking Paper No 6637 1998
Kroszner Randall S and Thomas Stratmann lsquolsquoInterest Group Competition andthe Organization of Congress Theory and Evidence from Financial ServicesPolitical Action Committeesrsquorsquo American Economic Review LXXXVIII (1998)pp 1163ndash1187
Kroszner Randall S and Thomas Stratmann lsquolsquoDoes Political Ambiguity PayCorporate Campaign Contributions and Rewards to Legislator Reputationrsquorsquounpublishedmanuscript University of Chicago Graduate School of Business1999
Lin D Y and L J Wei lsquolsquoThe Robust Inference for the Cox Proportional HazardsModel Journal of the American Statistical Association LXXXIV (1989)1074ndash1078
Macey Jonathan R and Geoffrey P Miller Banking Law and Regulation (BostonLittle Brown and Company 1992)
Makinson Larry Open Secrets The Encyclopedia of Congressional Money andPolitics (Washington DC Congressional Quarterly 1992)
Niskanen WilliamBureaucracy and Representative Government (Chicago Aldine-Atherton 1971)
QUARTERLY JOURNAL OF ECONOMICS1466
Noll Roger lsquolsquoComment on Peltzmanrsquorsquo Brookings Papers Microeconomics SpecialIssue (1989) 48ndash58
North Douglass lsquolsquoA Transactions Cost Theory of Politicsrsquorsquo Journal of TheoreticalPolitics II (1990) 355ndash367
Peltzman Sam lsquolsquoToward a More General Theory of Regulationrsquorsquo Journal of Lawand Economics XIX (1976) 109ndash148 lsquolsquoConstituent Interest and Congressional Votingrsquorsquo Journal of Law andEconomics XXVII (1984) 181ndash210 lsquolsquoThe Economic Theory of Regulation after a Decade of DeregulationrsquorsquoBrookings Papers Microeconomics Special Issue (1989) 1ndash41
Peterson Mitchell and Raghuram Rajan lsquolsquoThe Benets of Lending RelationshipsEvidence from Small Business Datarsquorsquo Journal of Finance XLIX (1994) 3ndash38
Poole Keith and Howard Rosenthal Congress A Political-Economic History ofRoll Call Voting (Oxford Oxford University Press 1997)
Stata Stata Reference Manual Release 5 (College Station TX Stata Press 1997)Stigler George J lsquolsquoThe Theory of Economic Regulationrsquorsquo Bell Journal of Economics
and Management Science II (1971) 3ndash21Stigler George J Chicago Studies in the Political Economy (Chicago University
of Chicago Press 1988)Strahan Philip E and James Weston lsquolsquoSmall Business Lending and the Changing
Structure of the Banking Industryrsquorsquo Journal of Banking and Finance XXII(1998) 821ndash845
Sylla Richard John Legler and John Wallis lsquolsquoBanks and State Public Finance inthe New Republic The United States 1790ndash1860rsquorsquo Journal of EconomicHistory XLVII (1987) 391ndash403
White Eugene The Regulation and Reform of the American Banking System1900ndash1929 (Princeton Princeton University Press 1983)
Winston Clifford lsquolsquoEconomic Deregulation Days of Reckoning for Microecono-mistsrsquorsquo Journal of Economic Literature XXXI (1993) 1263ndash1289
WHAT DRIVES DEREGULATION 1467
rate h0(t) is ptp 2 1 with shape parameter p that will be estimatedfrom the data When p 1 this model displays a monotonicallyincreasing hazard rate To estimate the parameters we maximizethe following log-likelihood function [Kiefer 1988]
L(b) 5 oi 5 1
N
di ln h[ti x(ti) b] 2 oi 5 1
N
e 0
tih[u x(u) b] du
where N indexes the number of states at risk at the beginning ofthe sample period di is an indicator equal to one for the statesthat deregulated by the end of the sample period and zero for thestates that do not deregulate (ie the censored observations) andti is the time of deregulation for the ith state This formulationassumes that the time until deregulation is independent acrossstates but the observations of the same state over time are notassumed to be independent In addition we use a robust estima-tion procedure to calculate the standard errors [Lin and Wei1989]
We exclude states that deregulated before 1970 from theanalysis so N equals 39 states During the sample period 36 ofthe 39 states deregulate We have three censored observations forwhich di is zero (Arkansas Iowa and Minnesota) but our results
FIGURE IIKaplan-Meier Estimate of Hazard Rate for Branching Deregulation
WHAT DRIVES DEREGULATION 1449
are not sensitive to the inclusion of these censored observationsSince we observe each state in each year up to and including theyear of deregulation we have a total of 637 observations Table IIreports summary statistics and correlation coefficients for theexplanatory variables we use in our analysis
We use the Weibull model rather than a model that does notimpose any structure on the baseline hazard rate so that we cancalculate the change in the expected time to deregulation for agiven change in the levels of the covariates22 In the Weibullmodel we can invert the hazard function and map it into the timedomain Rewriting the Weibull model in this way the log of thetime to deregulation T is a linear function of the political economyfactors and an error term ln (T ) 5 x8b 1 e23 Because we assumethat the baseline hazard rate is Weibull with a shape characterizedby p the new coefficients will be scaled by p that is b 5 2 bp
The b coefficients represent the percentage change in thetime to deregulation for a one-unit change in the covariates Apositive coefficient implies that an increase in the variableincreases the expected time until deregulation To gauge theeconomic importance of the effects we will multiply the b by thestandard deviation of the explanatory variables and then evaluatehow much the changes in the variables raise or lower the expectedtime to deregulation24 In Table III we report the b coefficients
22 The Cox proportional hazards model does not impose any structure on thebaseline hazard rate h0(t) and takes the following form
h[t x(t) b] 5 h0(t) exp [x(t)8b]
Common factors that affect the probability of deregulation in the same way for allstates therefore do not inuence the estimated coefficients The cost of the Coxmodel is that because there is no structure on the baseline we can calculatechanges in only relative hazard rates associated with changes in the covariatesThe Weibull model however provides sufficient structure so that we can translateour estimates into a log expected time metric that allows us to calculate the changein the expected time to deregulation for a given change in the covariates SeeKiefer [1988] Greene [1997] and Stata [1997] When we compare the coefficientsestimated by the Cox and Weibull models on our data they are very close (seeKroszner and Strahan [1998]) so imposing a Weibull structure does not appear todistort our estimates
23 The error term e is independent of x and has an extreme value distributionscaled by 1p (see Kiefer [1988] Greene [1997] and Stata [1997] for more details onthis log expected time parameterization of the Weibull model)
24 The transformation to the log expected time domain is only strictly correctwhen the covariates are constant over time Our interpretation of the coefficients ispossible in the time-varying model if we assume that a one-standard-deviationdifference would hold over the entire sample period that is the deregulationwould occur so many years earlier or later if the variable in question were higherby one standard deviation from 1970 to 1992 Note that the mean number of yearsto deregulation in the sample is 158
QUARTERLY JOURNAL OF ECONOMICS1450
TABLE IISUMMARY STATISTICS FOR VARIABLES IN HAZARD MODEL 1970ndash1992
Panel A Univariate statistics MeanStandarddeviation Minimum Maximum
(1) (2) (3) (4)Small bank asset share of all
banking assets in state 011 005 001 023Relative size of insurance to
banking plus insurance inthe state 047 009 024 080
Indicator is 1 if banks can sellinsurance in the state 020 040 0 1
Small rm share of the numberof rms in the state 088 002 078 095
Capital ratio of small banksrelative to large in the state 002 001 2 001 009
Share of state government con-trolled by Democrats 065 036 0 1
Indicator is 1 if state controlledby one party 054 050 0 1
Average yield on bank loans inthe state minus Fed fundsrate 002 002 2 004 008
Panel B CorrelationsSmallbank
Relativesize of
insuranceInsuranceindicator
Smallrm
Smallbank
capitalDem
controlSingleparty
(1) (2) (3) (4) (5) (6) (7)Small bank asset share
of all banking assetsin state 1
Relative size of insur-ance to banking plusinsurance in the state 2 034 1
Indicator is 1 if bankscan sell insurance inthe state 020 2 012 1
Small rm share of thenumber of rms inthe state 030 2 011 005 1
Capital ratio of smallbanks relative tolarge in the state 2 066 020 2 010 2 006 1
Share of state govern-ment controlled byDemocrats 2 006 001 2 018 2 018 005 1
Indicator is 1 if statecontrolled by oneparty 007 012 001 2 006 2 006 030 1
Average yield on bankloans in the stateminus Fed funds rate 2 001 2 015 003 022 2 010 2 003 2 006
The number of observations is 637 except for the average interest rate on bank loans which becomesavailable only in 1976 so the number of observations is 408
TABLE IIIHAZARD MODEL OF POLITICAL ECONOMY FACTORS AFFECTING THE TIMING OF STATE
BRANCHING DEREGULATION 1970ndash1992
(1) (2) (3) (4) (5) (6)Small bank asset share of
all banking assets instate
648(122)
519(137)
632(135)
516(118)
488(133)
445(088)
Capital ratio of smallbanks relative to largein the state
1325(348)
921(319)
1301(362)
1067(346)
1144(375)
906(381)
Relative size of insurancein states where banksmay sell insurance 0otherwise
324(161)
259(149)
315(165)
285(154)
272(159)
055(134)
Indicator is 1 if banksmay sell insurance inthe state
045(010)
037(011)
045(010)
038(008)
036(009)
001(012)
Relative size of insurancein states where banksmay not sell insurance0 otherwise
2 093(043)
2 090(033)
2 102(051)
2 093(048)
2 096(053)
2 004(039)
Small rm share of thenumber of rms in thestate
2 972(211)
2 614(215)
2 952(211)
2 943(230)
2 952(270)
2 1509(276)
Share of state govern-ment controlled byDemocrats
031(012)
023(011)
030(012)
027(011)
026(011)
011(013)
Indicator is 1 if state con-trolled by one party
2 004(010)
2 004(007)
2 004(007)
2 001(009)
2 002(010)
017(009)
Average yield on bankloans in the stateminus Fed funds rate mdash
023(319) mdash mdash mdash mdash
Bank failure rate mdash mdash092
(111) mdash mdash mdash
Indicator is 1 if state hasunit banking law mdash mdash mdash
018(010)
019(010)
030(009)
Indicator is 1 if statechanges bank insur-ance powers mdash mdash mdash mdash
2 008(012)
2 026(013)
Includes regionalindicators No No No No No Yes
N 637 408 637 637 637 637Log likelihood 2 374 900 2 360 2 212 2 191 1192p-value of x 2 for
regression 001 001 001 001 001 001
The hazard model is Weibull where the dependent variable is the log expected time to branchingderegulation ln (T) 5 bx 1 eAll variables are measured for each state in each year Small bank asset share isthe percent of banking assets in the state held by banks below the median size of banks in each state in eachyear Relative capital ratio is the capital to assets ratio of small banks minus that of large banks Size ofinsurance relative to banking plus insurance in the state is measured as gross state product from insurancedivided by gross state product from insurance plus banking Insurance indicator is 1 if state law permits banksto enter the insurance business Small rm share is the percent of all establishments in the state that havefewer than twenty employees Party control variable is the share of the three bodies of state governmentcontrolled by Democrats One party control indicator is 1 if the same party controlrsquos the governorship and hasmajorities in both chambers of the state legislature Average yield on bank loans equals total interest incomeon all domestic loans made by banks in the state divided by total loans Bank failure rate is the ratio of assetsin failed banks during a given state and year divided by total bank assets in that state and year Unit bankingindicator equals 1 for states with unit banking restrictions The change in insurance powers indicator equalsone if the state changed the law to permit banks to sell insurance during the sample period We divide thecountry into four regions and include regional indicators in column (5) Robust standard errors are inparentheses and denote statistically signicant at the 1 5 and 10 percent levels
QUARTERLY JOURNAL OF ECONOMICS1452
The positive and statistically signicant coefficient on smallbank share implies that a greater small bank share delaysderegulation and this effect is economically important A one-standard-deviation increase in the small bank share results in a30 percent increase in the time until deregulation or about 47years This result supports the implications of private-interesttheory concerning intraindustry rivalry and is contrary to thepublic-interest theory
The positive and statistically signicant coefficient on therelative performance of small banks in the state implies thatstates deregulate later when small banks are relatively strongnancially25 A one-standard-deviation increase in the relativecapital-to-asset ratio results in a 15 percent rise in the time untilderegulation or about 24 years This result is consistent with theprivate-interest theory but may also be consistent with a public-interest interpretation because the risk of failures (and a costlytaxpayer-funded bailout) increases as small bank nancial healthdeclines
Interindustry competition also helps explain the timing ofderegulation In states where banks can sell insurance a rela-tively large insurance sector is associated with an increase in theexpected time to deregulation26 A one-standard-deviation in-crease in the relative size of the insurance sector in those statesthat permit banks to sell insurance leads to a 22 percent increasein the time until deregulation or about 35 years27 This result isconsistent with the private-interest theory but not the public-interest theory
Deregulation occurs earlier in states where small bank-dependent rms are relatively numerous A one-standard-
25 We also allowed small bank and large bank capital-to-asset ratios to enterseparately Their coefficient estimates were nearly equal in absolute value and ofopposite signs and so the data do not reject the relative small to large bank capitalratio specication we have chosen
26 When we use the relative size of the insurance sector in the state as awhole rather than insurance relative to banking plus insurance in the state weobtain similar results When we simultaneously include both measures of therelative size of insurance the interactions of the lsquolsquoinsurance relative to bankingplus insurancersquorsquo variable continue to be statisticallysignicant but the interactionsof lsquolsquoinsurance relative to the state as a wholersquorsquo variable are not
27 We have also estimated the model with the indicator variable equal to onefor states where banks may sell insurance but without the variables measuringtheimportance of insurance relative to banking In this model the coefficient on theinsurance indicator variablemdashan estimate of the difference in the expected time toderegulation in states where banks may sell insurance relative to states wherebanks may not sell insurancemdashequals 027 and is statistically signicant at the 5percent level This suggests that branching deregulation takes about 45 yearslonger in states where banks may sell insurance
WHAT DRIVES DEREGULATION 1453
deviation increase in the share of small rms reduces the timeuntil deregulation by 18 percent or about three years This resultconcerning the interests of consumers of banking services isconsistent with both the private- and public-interest theories
Column (2) of Table III introduces the average interest rateon loans to the model as a rough proxy for the cost to bankborrowers of these regulations These data become available onlyin 1976 so our sample size shrinks to 408 observations Thecoefficient on this variable is small and not statistically signi-cant The public-interest theory would imply that the coefficientshould be negative while the implication of the economic theorydoes not have a clear sign prediction
When we add the failure rate of banks to the specication incolumn (3) its coefficient is small and statistically insignicantWe do not nd a linkage between the timing of state deregulationand statewide banking distress Given that deposit insurance isfunded federally however bank failures may raise the publicrsquosconcerns about the costs of the regulations at the national ratherthan the state level This interpretation can explain why the rateof bank failure does not affect the timing of state-level deregula-tion even though bank instability does appear to rise generallyduring the period of increasing deregulation (see Section V below)
Finally the partisan structure of the state government alsoinuences when states deregulate As expected a higher propor-tion of Democrats in the government tends to delay deregulationA one-standard-deviation rise in the share of the governmentcontrolled by Democrats slows the deregulation by about twoyears Whether the state is dominated by one party however doesnot appear to affect the timing of the deregulation
To summarize our ndings from the hazard model privateinterests appear to play an important role in the deregulatoryprocess While private interests and public interests do sometimescoincide the results on the relative share of small banks and largebanks and on the relative size of insurance where banks competeare consistent with a private-interest approach but are difficult toexplain on public-interest grounds
B Robustness Tests
Omitted Variables States begin the sample period withdifferent degrees of restrictions on bank expansion These initialconditions may proxy for omitted factors determining the political-economy equilibrium in the state that could then affect the
QUARTERLY JOURNAL OF ECONOMICS1454
subsequent timing of deregulation For instance some statesbegan the period with unit banking while other states allowedlimited branching and the formation of multibank holdingcompanies
To test whether the stringency of initial restrictions onbranching matters we rst divide the sample into states with andwithout the most extreme form of branching regulation unitbanking28 Column (4) of Table III includes an indicator variableequal to one if the state began the period with unit bankingrestrictions The coefficient on this variable is positive andstatistically signicant but the other results remain virtuallyunchanged Next we include an indicator variable that is one ifthe state permitted multibank holding companies (MBHCs) toexist at the beginning of our sample period Thirty-four statesbegan the period allowing MBHCs The coefficient estimate forthis indicator variable (not reported) is small and not statisticallysignicant and its inclusion does not change the effects of theother variables
Eleven states expand bank powers to include the sale ofinsurance products during our sample period This could proxy foromitted factors about the relative strength of the insuranceindustry in the state In column (5) we include an indicatorvariable that is one if the state changes the law during our sampleperiod to permit banks to sell insurance business and zerootherwise The coefficient estimate is small and statisticallyinsignicant and does not affect the other results29
In column (6) we introduce four regional indicator variablesNorth South Midwest and West30 The coefficients on small bankshare the relative capital-to-assets ratio and small rm sharechange little Although the coefficient estimate still suggests that
28 We classify states that prohibited branching but permitted banks toestablish facilities as unit banking states The unit banking indicator is one for thefollowing states CO AR FL IL IA KS MN MO MT NE ND OK TX WI WVand WY
29 We conducted a likelihood ratio test of whether the unit banking statesthe MBHC states and the states that changed their regulations of bank insurancepowers could be pooled with the other states In all three cases we could not rejectpooling of the data
30 Since intrastate deregulation generally preceded interstate deregulationand the latter typically took the form of regional interstate compacts thepossibility of participating in a regional interstate banking compact could haveinuenced the decision to deregulate intrastate branching Our denitions of theregions are region 1 (South) contains AL AR DC FL GA KY LA MS NC OKSC TN TX and VA region 2 (Northeast) contains CT MA MD ME NH NJ NYPA RI VT and WV region 3 (Midwest) contains IA IL IN KS MI MN MO NEND OH SD and WI region 4 (West) contains the other states
WHAT DRIVES DEREGULATION 1455
in states where banks can sell insurance a larger relative size ofthe insurance sector delays deregulation the effect is no longerstatistically signicant The coefficient on the indicator variablefor whether a state expands bank powers to permit insurancesales during the sample period however now becomes statisti-cally signicantly negative This implies that deregulation occursroughly 41 years earlier in states where the insurance lobby maybe relatively weak While the inclusion of the regional indicatorsdoes not affect the other results the results on insurance doappear to be sensitive to the specication of the hazard model31
Estimation Technique To check the robustness of using ahazard model and to investigate whether the time-series or thecross-sectional variation in each of the factors accounts for thehazard results we estimate two simple linear probability models(neither are reported) First we run a pooled time-series cross-section OLS regression using the 637 observations from thehazard model The dependent variable equals one in the year ofderegulation and zero otherwise We include the same set ofregressors that were used in the hazard model and add state-levelxed effects This specication thus removes the cross-statevariation from the independent variables and focuses on theirmovements over time In this model the signs of the coefficientson the interest group factors are consistent with the results fromthe hazard model and the coefficients for the small bank relativecapital ratio and for the relative size of insurance where states cansell insurance are statistically signicant
In the second model we eliminate the time-series variationby averaging each of the explanatory variables over time We thushave one observation per state that had branching restrictions asof 1970 (N 5 39) The dependent variable equals the inverse of thenumber of years until deregulation which represents an estimateof the statersquos probability of deregulating in a year under theassumption that this probability is constant over time and eachyear provides an independent observation32 Although we haveonly 39 observations the coefficients on the small bank marketshare small bank relative capital ratio and small rm share
31 As further robustness check we also tried a population density variablewhich Abrams and Settle [1993] found relevant for regulatory change during the1920s and 1930s but its coefficient was not statistically signicant and did notaffect the other results
32 For instance the dependent variable would take the value of 01 for astate that deregulated in 1980 For the three states that did not deregulate thedependent variable equals 0
QUARTERLY JOURNAL OF ECONOMICS1456
variables are statistically signicant The signs of the effects ofthe insurance variables are the same as those reported in thehazard model but not statistically signicant
The hazard model and linear probability models thus producevery similar results The estimates from the latter suggest thatthe effects for the small bank market share and small rm sharevariables are driven primarily by cross-state variation the effectof the insurance variable is driven primarily by time-seriesvariation and the small bank capital effect is driven by bothcross-sectional and time-series variation
Ex Post Consequences of Deregulation Finally to check theplausibility of our interpretation of the hazard results we con-sider whether the ex post consequences of deregulation areconsistent with the ex ante lobbying positions we attribute to eachinterest group (see Kroszner and Strahan [1998] for details) Wend that small banks lose market share following deregulationand in states where banks can enter the insurance business theinsurance sector shrinks relative to the banking sector followingderegulation Borrowers also benet from deregulation throughlower average interest rates on loans These ndings support theprivate-interest interpretation of the coefficients in the hazardmodel groups that will benet lobby to speed deregulation andthose that will be harmed lobby to slow it
C Voting on Interstate Branching Deregulation in the U SHouse of Representatives
We now examine whether the forces driving intrastate branch-ing deregulation also drive interstate deregulation Financialservices interests are active contributors and lobbyists in Washing-ton Their political action committees constitute the largest groupof contributors to legislators providing nearly 20 percent of totalcongressional campaign contributions [Makinson 1992] and muchof their lobbying effort involves competition among rival interestswithin nancial services (see Kroszner and Stratmann [1998])
As noted above after virtually all states adopted intra- andinterstate branching deregulation the 1994 Riegle-Neal Act re-pealed the 1927 McFadden Act to eliminate all barriers tointerstate banking and branching by 1997 The key votes concern-ing the Riegle-Neal Act were either voice votes or extremelylopsided so we could not estimate a voting model from them Anumber of bills and amendments related to interstate branching
WHAT DRIVES DEREGULATION 1457
had been debated in Congress during the years prior to thepassage of the Riegle-Neal Act but a search of the weekly BNABanking Reporter and the Congressional Record produced onlyone roll-call vote related to interstate branching that was notlopsided This vote occurred in the House on November 14 1991and concerned an amendment sponsored by Wylie (R-OH) andNeal (D-NC) to introduce interstate banking and branchingderegulation to a nancial services reform package33 Althoughthe amendment passed by 210 to 208 the bill to which it wasattached subsequently was defeated
To check for the inuence of the factors we considered in thestate-level reforms we examine both the sponsorship of interstatebanking legislation and voting on the amendment The sponsorsof the Wylie-Neal amendment are from states with low small bankshares 004 in Ohio (Wylie) and 002 in North Carolina (Neal) Incontrast the sample mean in 1991 is 008 (median 5 007)Michigan home state of the Senatersquos sponsor of the 1994 Riegle-Neal Act also had relatively low small bank strength (small bankshare of 005)
Table IV reports estimates of a probit model where thedependent variable equals one if the legislator voted in favor ofthe amendment and zero otherwise The explanatory variablesare the same as those in the hazard model except that we use aparty affiliation indicator variable equal to one for Democrats andzero for Republicans as our proxy for political factors Thecoefficients in Table IV are the marginal effects (lsquolsquoslopesrsquorsquo) of aone-unit change of each variable on the probability that alegislator will vote for the amendment34 Note that a forcefavoring deregulation has a negative coefficient in the hazardmodel (Table III) but a positive coefficient in the probit model
Consistent with the state-level deregulation process legisla-tors are more likely to support the amendment if their states havea relatively low share of small banks As in the hazard model thefraction of small banks is the most important interest groupinuence on a legislatorrsquos voting decision A one-standard-deviation increase in small banksrsquo market share (from the mean)is associated with a decline in the probability of voting in favor of
33 The Wylie-Neal amendment also included provisions limiting certaininsurance and real estate powers of national banks [Congressional RecordNovember 14 1991 pp 10239ndash10242]
34 Since we have multiple legislators from each state we adjust the standarderrors to correct for the lack of independence among observations clustered in thesame state
QUARTERLY JOURNAL OF ECONOMICS1458
TAB
LE
IVM
AR
GIN
AL
EF
FE
CT
SF
RO
MA
PR
OB
ITM
OD
EL
OF
TH
EIN
FL
UE
NC
EO
FP
OL
ITIC
AL
EC
ON
OM
YF
AC
TO
RS
ON
TH
EH
OU
SE
VO
TE
ON
TH
EW
YL
IE-N
EA
L
AM
EN
DM
EN
TT
OP
ER
MIT
INT
ER
ST
AT
EB
AN
KB
RA
NC
HIN
G
NO
VE
MB
ER
14
1991
(1)
(2)
(3)
(4)
(5)
(6)
Smal
lban
kas
set
shar
eof
allb
anki
ng
asse
tsin
stat
e2
417
(12
4)2
454
(12
2)2
407
(12
0)2
443
(14
7)2
450
(14
9)2
490
(14
7)C
apit
alra
tio
ofsm
allb
anks
rela
tive
tola
rge
inth
est
ate
180
(37
5)2
68(3
42)
256
(37
5)1
69(3
78)
153
(38
0)2
45(4
25)
Rel
ativ
esi
zeof
insu
ranc
ein
stat
esw
here
bank
sm
ayse
llin
sura
nce
0ot
herw
ise
22
34
(10
2)2
249
(1
14)
22
33
(10
1)2
249
(1
04)
22
51
(10
2)2
154
(11
1)In
dica
tor
is1
ifba
nks
may
sell
insu
ranc
ein
the
stat
e0
79
(0
06)
074
(00
6)0
69
(0
06)
070
(00
6)0
70
(0
06)
067
(00
7)R
elat
ive
size
ofin
sura
nce
inst
ates
whe
reba
nks
may
not
sell
insu
ranc
e0
oth
erw
ise
180
(06
6)2
42
(0
72)
154
(0
67)
173
(06
6)1
72
(0
67)
203
(0
91)
Smal
lrm
shar
eof
the
num
ber
ofr
ms
inth
est
ate
21
23(0
92)
20
83(0
89)
21
62(1
07)
21
32(0
95)
21
30(1
00)
20
94(1
16)
Indi
cato
ris
1if
Rep
rese
ntat
ive
isa
Dem
ocra
t2
035
(00
7)2
035
(00
7)2
036
(00
7)2
035
(00
7)2
035
(00
7)2
036
(00
7)A
vera
geyi
eld
onba
nklo
ans
inth
est
ate
min
usF
edfu
nds
rate
mdash10
01
(51
5)mdash
mdashmdash
mdash
Ave
rage
fail
ure
rate
198
5to
1991
mdashmdash
228
(14
1)mdash
mdashmdash
Indi
cato
ris
1if
stat
eh
asun
itba
nkin
gla
wmdash
mdashmdash
004
(00
7)0
04(0
07)
20
01(0
08)
Indi
cato
ris
1if
stat
ech
ange
sba
nk
insu
ran
cepo
wer
smdash
mdashmdash
mdash2
001
(00
7)2
003
(00
7)In
clud
esre
gion
alin
dica
tors
N
oN
oN
oN
oN
oYe
sP
seud
o-R
20
170
180
170
170
170
18L
ogli
keli
hood
224
024
223
826
223
951
224
013
224
010
223
808
p-va
lue
ofx2
for
regr
essi
on
001
0
01
001
0
01
001
0
01
Th
ede
pen
den
tvar
iabl
eis
one
ifth
ele
gisl
ator
vote
sfo
rth
eam
endm
enta
nd
zero
ifag
ains
tT
he
repo
rted
coef
fici
ents
are
the
effe
cts
ofa
uni
tch
ange
ofth
ein
depe
nde
ntva
riab
le(f
rom
the
mea
n)on
the
prob
abili
tyof
voti
ng
infa
vor
ofth
eam
endm
ent
All
vari
able
sar
em
easu
red
byst
ate
N5
418
the
num
ber
ofR
epre
sent
ativ
esvo
tin
gon
the
amen
dmen
tS
mal
lban
kas
sets
har
eis
the
perc
ent
ofba
nki
ng
asse
tsin
the
stat
eh
eld
byba
nks
belo
wth
em
edia
nsi
zeof
ban
kin
each
stat
ein
each
year
Rel
ativ
eca
pita
lrat
iois
the
capi
talt
oas
sets
rati
oof
smal
lban
ksm
inu
sth
atof
larg
eba
nks
Siz
eof
insu
ran
cere
lati
veto
bank
ing
plu
sin
sura
nce
inth
est
ate
ism
easu
red
asgr
oss
stat
epr
oduc
tfro
min
sura
nce
divi
ded
bygr
oss
stat
epr
odu
ctfr
omin
sura
nce
plu
sba
nkin
gIn
dica
tor
ison
eif
stat
ela
wpe
rmit
sba
nks
toen
ter
the
insu
ran
cebu
sin
ess
Sm
all
rmsh
are
isth
epe
rcen
tofa
lles
tabl
ishm
ents
inth
est
ate
that
hav
efe
wer
than
twen
tyem
ploy
ees
Par
tyin
dica
tor
ison
eif
the
Rep
rese
nta
tive
isa
Dem
ocra
tA
vera
geyi
eld
onba
nklo
ans
equa
lsto
tali
nter
est
inco
me
onal
ldom
esti
clo
ans
mad
eby
ban
ksin
the
stat
edi
vide
dby
tota
lloa
ns
The
aver
age
bank
failu
rera
teis
the
rati
oof
asse
tsin
faile
dba
nks
duri
ng
agi
ven
stat
edi
vide
dby
tota
lban
kas
sets
inth
atst
ate
aver
aged
over
the
year
s19
85to
1991
Un
itba
nki
ng
indi
cato
req
ual
son
efo
rst
ates
wit
hun
itba
nkin
gre
stri
ctio
nsT
hech
ange
inin
sura
nce
pow
ers
indi
cato
req
ual
son
eif
the
stat
ech
ange
dth
ela
wto
perm
itba
nks
tose
llin
sura
nce
duri
ngth
esa
mpl
epe
riod
We
divi
deth
eco
unt
ryin
tofo
urre
gion
san
din
clu
dere
gion
alin
dica
tors
inco
lum
n(5
)R
obu
stst
anda
rder
rors
are
inpa
ren
thes
es
an
d
den
ote
stat
isti
cally
sign
ica
nta
tth
e1
5an
d10
perc
entl
evel
s
WHAT DRIVES DEREGULATION 1459
branching of approximately 17 percent The relative capital-to-assets ratio of small and large banks however does not have astatistically signicant effect in the probit model
The impact of rival interests outside of banking is alsoconsistent with intrastate deregulation results Where banks cansell insurance legislators from states with larger insurancesectors relative to banking are less likely to vote for the amend-ment35 A one-standard-deviation increase in the relative size ofinsurance in those states which permit banks to sell insurancedecreases the probability that a legislator will favor the amend-ment by about 13 percent Turning to consumer interests thecoefficient on small rm share is not statistically signicant butthe coefficient on interest rates in the state is A one-standard-deviation increase in the average interest rate on loans raises theprobability that a legislator will support the amendment byroughly 8 percent The coefficient estimate for the bankingdistress variable measured as the average bank failure rate from1985 to 1991 is small and statistically signicant Overall theprobit analysis of the vote on national branching deregulationsupports the private-interest theory of deregulation and providesa consistency check that the importance of interests operating onthe state legislatures are very similar to those operating at thefederal level
V WHY DID BRANCHING DEREGULATION BEGIN DURING THE
1970S SHOCKS TO THE EQUILIBRIUM SUPPORTING BANK
BRANCHING RESTRICTIONS
A complete explanation of regulatory exit should account forwhy deregulation begins in the 1970s as well as the timing of thestate-by-state reforms [Kane 1996] In this section we describebroad technological legal and economic shocks that altered thepolitical-economy equilibrium which had kept antibranching regu-lations little changed for at least 30 years
Beginning in the 1970s three innovations reduced the valueto the protected banks of local geographic monopolies (see TableV) First automatic teller machines (ATMs) helped to erode thegeographic ties between customers and banks Second checkablemoney market mutual funds and the Merrill Lynch Cash Manage-ment Account demonstrated that banking by mail and telephone
35 The positive and statistically signicant estimates on the other insurancevariables may reect insurance industry support for the amendmentrsquos provisionslimiting banksrsquo insurance powers
QUARTERLY JOURNAL OF ECONOMICS1460
provided a convenient alternative to local banks36 Third techno-logical innovation and deregulation reduced transportation andcommunication costs particularly since the 1970s thereby lower-ing the costs for customers to use distant banks By increasing theelasticity of deposits supplied to banks these innovations reducedthe value of geographical restrictions to their traditional benecia-ries and thereby reduced their incentive to ght to maintain them[Peltzman 1976]
On the lending side increasing sophistication of credit-scoring techniques following innovations in information process-ing nancial theory and the development of large credit databases began to diminish the value of knowledge that localbankers had about the risks of borrowers in the community As a
36 Regulation Q which limited the interest rates that banks could pay ondeposits may have helped to drive depositors away from banks when the gapbetween market rates and deposit ceilings grew during the 1970s The eliminationof interest rate ceilings on large denomination certicates of deposit during the1970s appears to have hurt smaller and retail-oriented banks relative to largerwholesale banks [James 1983]
TABLE VBROAD TRENDS IN COMMERCIAL BANKING 1950ndash1995
YearNumberof ATMs
Domesticbank
deposits(billions)
Moneymarketmutual
fund(billions)
Percent ofdeposits 1
moneyfunds heldby banks
Smallbanksrsquo
percent ofbankingassets
Averagenumber of
bankfailures
(1) (2) (3) (4) (5) (6)1950 0 $ 154 $ 0 100 NA 41955 0 191 0 100 NA 31960 0 228 0 100 24 21965 0 330 0 100 20 41970 0 479 0 100 18 61975 9750 775 4 99 18 61980 18500 1182 76 94 17 101985 61117 1787 242 88 14 601990 80156 2339 493 83 11 1791995 122706 2552 745 77 8 61
Column (1) ATM gures are from Bank Network News The EFT Network Data Book (New YorkFaulkner and Gray Inc) The 1975 gure was unavailable 9750 is the number of ATMs in 1978 the rst yearfor which complete data are available Columns (2)ndash(4) Banksrsquo domestic deposits are from the Reports ofIncome and Condition money market mutual funds are from the Flow of Funds Data on all bank depositsforeign plus domestic are only available beginning in 1970 The trend in banksrsquo share (column (4)) is the sameusing total deposits instead of domestic deposits Column (5) Percent of banking assets held by small bankswhere a small bank is dened as a commercial bank with less than $100 million in assets in 1994 dollarsThese data are based on the Reports of Income and Condition Data on small banks are not available before1960 Column (6) Five-year average number of bank failures where the nal year is indicated in the rstcolumn These data are from FDIC Annual Report and the Quarterly Banking Prole
WHAT DRIVES DEREGULATION 1461
result of these innovations a national market developed forresidential mortgages credit card receivables have been securi-tized and bank lending to small business now relies less on thejudgment of loan officers and more on standardized scoringmodels
These changes have increased the potential protability forlarge banks to enter what had been the core of small bankactivities Large banksrsquo incentive to increase their lobbying pres-sure to be able to expand into these markets has thus beenincreasing over time In fact small banksrsquo market share began todecline even prior to the branching deregulation (Table V) As thevalue of a local banking relationships declined small rms thatwere the main borrowers from the small banks also probablybecame more likely to favor the entry of large banks into localmarkets With the deadweight costs of preventing large bankentry rising the private-interest theory predicts that small localbanks would become less likely able to maintain the branchingrestrictions [Becker 1983] Deregulation that reduces deadweightcosts of regulation is also consistent with the public-interesttheory
Kane [1996] argues that another major shock to the oldequilibrium was the increasing public awareness of the costlinessof having government-insured but (geographically) undiversiednancial institutions In the late 1970s the failure rate of banksbegan to rise (Table V) In the 1980s the Savings and Loan crisisand taxpayer bailout further heightened public awareness aboutthe costs of restrictions that make depository institutions morelikely to require infusions of taxpayer funds The failures thusmay have heightened public support for branching deregulation(see also Abrams and Settle [1993]) As noted above we did notnd any evidence that banking failures or distress in a stateaffected the speed with which the state deregulated but distressstill may be a common factor affecting all states
These technological economic and legal shocks generatedconditions that changed the long-standing balance favoring theantibranching forces The marginal value of lobbying to repealbranching restrictions increased just as the relative value to thesmall banks of maintaining branching restrictions was decliningThese nationwide shocks are common factors across states andthey are consistent with the positive duration dependence associ-ated with state-level deregulation that we demonstrated in FigureII In addition a broad change in lsquolsquoideologyrsquorsquo against government
QUARTERLY JOURNAL OF ECONOMICS1462
regulation could explain the positive duration dependence duringthe sample period
VI CONCLUSIONS
The private-interest theory of regulation can account for thepattern of bank branching deregulation of the 1970s and 1980sBeneciaries of branching regulation had supported a coalitionfavoring geographical restrictions despite their costs to consum-ers of nancial services Innovations that began in the 1970saltered the value of the restrictions to the affected parties and theresulting competition among interest groups can explain thesubsequent deregulation While some of our results are alsoconsistent with the public interest theorymdashfor example deregula-tion occurs earlier when small banks are in a relatively weaknancial positionmdashother results particularly evidence on theimportance of rivalries between small and large banks andbetween banking and insurance are difficult to explain with thepublic-interest approach We also nd that political-institutionalfactors affect deregulation although these variables may act asproxies for unmeasured economic interests The greater success ofthe private-interest theory to explain deregulation here than inprevious work may be due to the richer cross-sectional andtime-series variation in branching deregulation that can beexploited with a hazard model Future empirical studies ofendogenous deregulation thus may be most fruitful where changehas occurred across states over time such as in franchisinginsurance and public utilities (eg Joskow [1996])
Technological and nancial innovations will continue to erodethe benets to any interest group of maintaining regulatorybarriers in nancial services These forces are likely to bringabout reforms both domestically for example through legislationthat would increase bank powers (see Kroszner [1996] Krosznerand Rajan [1994 1997] and Kroszner and Stratmann [1998]) andinternationally for example through the extension of nancialservices provisions of NAFTAto reduce geographic barriers acrosscountries [Kroszner 1997]
APPENDIX EFFECTS OF DIFFERENT TYPES OF BRANCHING
DEREGULATION ON ECONOMIC GROWTH BANKING STRUCTUREAND BANK EFFICIENCY
We measure the importance of alternative forms of deregula-tion by their consequences for the economic growth and the
WHAT DRIVES DEREGULATION 1463
structure and efficiency of the banking sector across the states(see Jayaratne and Strahan [1996 1998]) To do so we run anOLS panel regression relating indicators for each type of deregula-tion to potential consequences Our dependent variables includeproxies for economic growth (growth in personal income) bankstructure (the log of the number of banking companies and thenumber of branches per banking company) and efficiency (the logof total noninterest expenses and the ratio of loan loss provisionsto total loans) The four deregulation indicators (MBHC branch-ing through merger and acquisition unrestricted statewide branch-ing and interstate banking) are equal to one in the years afterderegulation and zero before If a state deregulated before thesample period begins in 1970 the indicator is one throughout ifthe state does not deregulate then the indicator is zero through-out All regressions include state xed effects and time effects
Table VI shows the results of including the set of four
TABLE VIMULTIVARIATE REGRESSIONS OF THE EFFECTS OF DIFFERENT TYPES OF BRANCHING
DEREGULATION ON ECONOMIC GROWTH BANKING STRUCTUREAND BANK EFFICIENCY
Dependent variable
Growthin
personalincome
Log of of
bankingcompanies
Branchesper
bankingcompany
Log ofnon-
interestexpenses
Loan lossprovisions
(loans 1leases)
Merger amp acquisitionbranchingindicator
0013(0004)
2 0162(0037)
1146(0453)
2 0107(0035)
2 0007(0001)
Unrestricted state-wide branchingindicator
2 0008(0005)
0086(0059)
2 0043(0574)
0070(0037)
0005(0002)
Interstate bankingindicator
0006(0005)
2 0052(0057)
1064(0511)
0005(0030)
2 0001(0001)
Multibank holdingcompany indicator
2 0005(0006)
2 0096(0052)
2 0364(0529)
2 0095(0036)
2 0001(0001)
Dependent variablemean [median]
0086[0082]
4881[5136]
6649[5055]
13812[13803]
0009[0006]
This table contains regressions of state-level economic growth and state-level measures of the structureand efficiency of the banking industry Each dependent variable is regressed on a set of state indicatorvariables a set of year indicator variables and the set of four deregulation indicators these deregulationindicators are equal to one after deregulation and zero before If a state was always regulated the indicator iszero throughout if it was always deregulated the indicator is one throughout Sample period is 1975 to 1991and Delaware and South Dakota are dropped so N 5 833 Noninterest expenses become available only in1984 so N 5 392 The coefficient on the deregulation indicators is reported with robust standard errors inparentheses and denote statistically signicant at the 1 5 and 10 percent levels
QUARTERLY JOURNAL OF ECONOMICS1464
deregulation indicators simultaneously The branching by mergerand acquisition indicator is statistically signicant and of theexpected sign in all ve of the equations and larger in absolutevalue than any of the other coefficients on the indicators for theother types of deregulation Since there may be multicollinearityamong our indicators we also ran univariate specications whichincluded each indicator separately and the results are verysimilar These results imply that branching by merger andacquisition is the most important type of deregulation measuredby its consequences for economic growth banking structure andbank efficiency See Kroszner and Strahan [1998] for more details
UNIVERSITY OF CHICAGO GRADUATE SCHOOL OF BUSINESS AND NATIONAL BUREAU OF
ECONOMIC RESEARCH
FEDERAL RESERVE BANK OF NEW YORK RESEARCH amp MARKET ANALYSIS GROUP
REFERENCES
Abrams Burton A and Russell F Settle lsquolsquoPressure-Group Inuence and Institu-tional Change Branch-Banking Legislation during the Great DepressionrsquorsquoPublic Choice LXXVII (1993) 687ndash705
Amel Dean lsquolsquoState Laws Affecting the Geographic Expansion of CommercialBanksrsquorsquo unpublished manuscript Board of Governors of the Federal ReserveSystem 1993
Becker Gary S lsquolsquoA Theory of Competition among Pressure Groups for PoliticalInuencersquorsquo Quarterly Journal of Economics XCVIII (1983) 371ndash400
Berger Allen Anil Kashyap Joseph Scalise lsquolsquoThe Transformation of the U SBanking Industry What a Long Strange Trip Itrsquos Beenrsquorsquo Brookings Papers onEconomic Activity 2 (1995) 55ndash218
Berger Allen Anthony Saunders Joseph Scalise and Gregory Udell lsquolsquoThe Effectsof Bank Mergers and Acquisitions on Small Business Lendingrsquorsquo Journal ofFinancial Economics L (1998) 187ndash229
Brennan Geoffrey and James Buchanan lsquolsquoTowards a Tax Constitution forLeviathanrsquorsquo Journal of Public Economics VIII (1977) 255ndash273
Brickley James A and Christopher M James lsquolsquoThe Takeover Market CorporateBoard Composition and Ownership Structure The Case of Bankingrsquorsquo Journalof Law and Economics XXX (1987) 161ndash180
Cole Rebel and John Wolken lsquolsquoFinancial Services Used by Small BusinessesEvidence from the 1993 Survey of Small Business Financesrsquorsquo Federal ReserveBulletin LXXXI (1994) 629ndash676
Dixit Avinash The Making of Economic Policy A Transactions-Cost PoliticsPerspective (Cambridge MA MIT Press 1996)
Economides Nicolas R Glenn Hubbard and Darius Palia lsquolsquoThe Political Economyof Branching Restrictions and Deposit Insurancersquorsquo Journal of Law andEconomics XXIX (1996) 667ndash704
Flannery Mark lsquolsquoThe Social Costs of Unit Banking Restrictionsrsquorsquo Journal ofMonetary Economics XIII (1984) 237ndash249
Greene William H Econometric Analysis third edition (Upper Saddle River NJPrentice-Hall 1997)
Gunther Jeffery lsquolsquoRegional Capital Imbalances and the Removal of InterstateBanking Restrictionsrsquorsquo Economics Letters XLIV (1994) 439ndash442 lsquolsquoThe Erosion of Intrastate Branch Banking Restrictionsrsquorsquo unpublishedmanuscript Federal Reserve Bank of Dallas 1996
Irwin Douglas and Randall Kroszner lsquolsquoInterests Institutions and Ideology inSecuring Policy Change The Republican Conversion to Trade Liberalizationafter Smoot-Hawleyrsquorsquo Journal of Law and Economics XLII (1999) 643ndash673
WHAT DRIVES DEREGULATION 1465
James Christopher lsquolsquoAn Analysis of Intra-Industry Differences in the Effect ofRegulation The Case of Deposit Rate Ceilingsrsquorsquo Journal of Monetary Econom-ics XII (1983) 417ndash442
Jarrell Gregg lsquolsquoChange at the Exchange The Causes and Effects of DeregulationrsquorsquoJournal of Law and Economics XXVII (1984) 273ndash312
Jayaratne Jith and Philip E Strahan lsquolsquoThe Finance-Growth Nexus Evidencefrom Bank Branch Deregulationrsquorsquo Quarterly Journal of Economics CI (1996)639ndash670
Jayaratne Jith and Philip E Strahan lsquolsquoEntry Restrictions Industry Evolutionand Dynamic Efficiency Evidence from Commercial Bankingrsquorsquo Journal of Lawand Economics XLIX (1998) 239ndash274
Joskow Paul lsquolsquoComment on lsquoPower Struggles Explaining Deregulatory Reformsin Electricity Marketsrsquo rsquorsquo Brookings Papers on Economic Activity Microeconom-ics (1996) 251ndash264
Joskow Paul and Roger Noll lsquolsquoRegulation in Theory and Practice An Overviewrsquorsquoin Studies in Public Regulation Gary Fromm ed (Cambridge MA MITPress 1981) pp 1ndash65
Kahn Matthew E and John G Matsusaka lsquolsquoDemand for Environmental GoodsEvidence from Voting Patterns on California Initiativesrsquorsquo Journal of Law andEconomics XL (1997) 137ndash173
Kalbeisch J and R Prentice The Statistical Analysis of Failure Time Data (NewYork John Wiley amp Sons 1980)
Kane Edward J lsquolsquoDe Jure Interstate Banking Why Only NowrsquorsquoJournal of MoneyCredit and Banking XXVIII (1996) 141ndash161
Kiefer Nicholas lsquolsquoEconomic Duration Data and Hazard Functionsrsquorsquo Journal ofEconomic Literature XXVI (1988) 646ndash679
Kroszner Randall S lsquolsquoThe Evolution of Universal Banking and Its Regulation inTwentieth Century Americarsquorsquo in Universal Banking Financial System DesignReconsidered Anthony Saunders and Ingo Walter eds (New York IrwinProfessional Publishers 1996) pp 70ndash99 lsquolsquoThe Political Economy of Banking and Financial Regulation in the U Srsquorsquo inThe Banking and Financial Structure in the NAFTA Countries and ChileGeorge M von Furstenberg ed (Boston Kluwer Academic Publishers 1997)pp 200ndash213
Kroszner Randall S and Raghuram G Rajan lsquolsquoIs the Glass-Steagall ActJustied A Study of the U S Experience with Universal Banking before1933rsquorsquo American Economic Review LXXXIV (1994) 810ndash832
Kroszner Randall S and Raghuram G Rajan lsquolsquoOrganization Structure andCredibility Evidence from Bank Securities Activities before the Glass-Steagall Actrsquorsquo Journal of Monetary Economics XXXIX (1997) 475ndash516
Kroszner Randall S and Philip E Strahan lsquolsquoRegulatory Incentives and the ThriftCrisis Dividends Mutual-to-Stock Conversions and Financial DistressrsquorsquoJournal of Finance LI (1996) 1285ndash1320
Kroszner Randall S and Philip E Strahan lsquolsquoWhat Drives Deregulation Econom-ics and Politics of the Relaxation of Bank Branching Restrictionsrsquorsquo NBERWorking Paper No 6637 1998
Kroszner Randall S and Thomas Stratmann lsquolsquoInterest Group Competition andthe Organization of Congress Theory and Evidence from Financial ServicesPolitical Action Committeesrsquorsquo American Economic Review LXXXVIII (1998)pp 1163ndash1187
Kroszner Randall S and Thomas Stratmann lsquolsquoDoes Political Ambiguity PayCorporate Campaign Contributions and Rewards to Legislator Reputationrsquorsquounpublishedmanuscript University of Chicago Graduate School of Business1999
Lin D Y and L J Wei lsquolsquoThe Robust Inference for the Cox Proportional HazardsModel Journal of the American Statistical Association LXXXIV (1989)1074ndash1078
Macey Jonathan R and Geoffrey P Miller Banking Law and Regulation (BostonLittle Brown and Company 1992)
Makinson Larry Open Secrets The Encyclopedia of Congressional Money andPolitics (Washington DC Congressional Quarterly 1992)
Niskanen WilliamBureaucracy and Representative Government (Chicago Aldine-Atherton 1971)
QUARTERLY JOURNAL OF ECONOMICS1466
Noll Roger lsquolsquoComment on Peltzmanrsquorsquo Brookings Papers Microeconomics SpecialIssue (1989) 48ndash58
North Douglass lsquolsquoA Transactions Cost Theory of Politicsrsquorsquo Journal of TheoreticalPolitics II (1990) 355ndash367
Peltzman Sam lsquolsquoToward a More General Theory of Regulationrsquorsquo Journal of Lawand Economics XIX (1976) 109ndash148 lsquolsquoConstituent Interest and Congressional Votingrsquorsquo Journal of Law andEconomics XXVII (1984) 181ndash210 lsquolsquoThe Economic Theory of Regulation after a Decade of DeregulationrsquorsquoBrookings Papers Microeconomics Special Issue (1989) 1ndash41
Peterson Mitchell and Raghuram Rajan lsquolsquoThe Benets of Lending RelationshipsEvidence from Small Business Datarsquorsquo Journal of Finance XLIX (1994) 3ndash38
Poole Keith and Howard Rosenthal Congress A Political-Economic History ofRoll Call Voting (Oxford Oxford University Press 1997)
Stata Stata Reference Manual Release 5 (College Station TX Stata Press 1997)Stigler George J lsquolsquoThe Theory of Economic Regulationrsquorsquo Bell Journal of Economics
and Management Science II (1971) 3ndash21Stigler George J Chicago Studies in the Political Economy (Chicago University
of Chicago Press 1988)Strahan Philip E and James Weston lsquolsquoSmall Business Lending and the Changing
Structure of the Banking Industryrsquorsquo Journal of Banking and Finance XXII(1998) 821ndash845
Sylla Richard John Legler and John Wallis lsquolsquoBanks and State Public Finance inthe New Republic The United States 1790ndash1860rsquorsquo Journal of EconomicHistory XLVII (1987) 391ndash403
White Eugene The Regulation and Reform of the American Banking System1900ndash1929 (Princeton Princeton University Press 1983)
Winston Clifford lsquolsquoEconomic Deregulation Days of Reckoning for Microecono-mistsrsquorsquo Journal of Economic Literature XXXI (1993) 1263ndash1289
WHAT DRIVES DEREGULATION 1467
are not sensitive to the inclusion of these censored observationsSince we observe each state in each year up to and including theyear of deregulation we have a total of 637 observations Table IIreports summary statistics and correlation coefficients for theexplanatory variables we use in our analysis
We use the Weibull model rather than a model that does notimpose any structure on the baseline hazard rate so that we cancalculate the change in the expected time to deregulation for agiven change in the levels of the covariates22 In the Weibullmodel we can invert the hazard function and map it into the timedomain Rewriting the Weibull model in this way the log of thetime to deregulation T is a linear function of the political economyfactors and an error term ln (T ) 5 x8b 1 e23 Because we assumethat the baseline hazard rate is Weibull with a shape characterizedby p the new coefficients will be scaled by p that is b 5 2 bp
The b coefficients represent the percentage change in thetime to deregulation for a one-unit change in the covariates Apositive coefficient implies that an increase in the variableincreases the expected time until deregulation To gauge theeconomic importance of the effects we will multiply the b by thestandard deviation of the explanatory variables and then evaluatehow much the changes in the variables raise or lower the expectedtime to deregulation24 In Table III we report the b coefficients
22 The Cox proportional hazards model does not impose any structure on thebaseline hazard rate h0(t) and takes the following form
h[t x(t) b] 5 h0(t) exp [x(t)8b]
Common factors that affect the probability of deregulation in the same way for allstates therefore do not inuence the estimated coefficients The cost of the Coxmodel is that because there is no structure on the baseline we can calculatechanges in only relative hazard rates associated with changes in the covariatesThe Weibull model however provides sufficient structure so that we can translateour estimates into a log expected time metric that allows us to calculate the changein the expected time to deregulation for a given change in the covariates SeeKiefer [1988] Greene [1997] and Stata [1997] When we compare the coefficientsestimated by the Cox and Weibull models on our data they are very close (seeKroszner and Strahan [1998]) so imposing a Weibull structure does not appear todistort our estimates
23 The error term e is independent of x and has an extreme value distributionscaled by 1p (see Kiefer [1988] Greene [1997] and Stata [1997] for more details onthis log expected time parameterization of the Weibull model)
24 The transformation to the log expected time domain is only strictly correctwhen the covariates are constant over time Our interpretation of the coefficients ispossible in the time-varying model if we assume that a one-standard-deviationdifference would hold over the entire sample period that is the deregulationwould occur so many years earlier or later if the variable in question were higherby one standard deviation from 1970 to 1992 Note that the mean number of yearsto deregulation in the sample is 158
QUARTERLY JOURNAL OF ECONOMICS1450
TABLE IISUMMARY STATISTICS FOR VARIABLES IN HAZARD MODEL 1970ndash1992
Panel A Univariate statistics MeanStandarddeviation Minimum Maximum
(1) (2) (3) (4)Small bank asset share of all
banking assets in state 011 005 001 023Relative size of insurance to
banking plus insurance inthe state 047 009 024 080
Indicator is 1 if banks can sellinsurance in the state 020 040 0 1
Small rm share of the numberof rms in the state 088 002 078 095
Capital ratio of small banksrelative to large in the state 002 001 2 001 009
Share of state government con-trolled by Democrats 065 036 0 1
Indicator is 1 if state controlledby one party 054 050 0 1
Average yield on bank loans inthe state minus Fed fundsrate 002 002 2 004 008
Panel B CorrelationsSmallbank
Relativesize of
insuranceInsuranceindicator
Smallrm
Smallbank
capitalDem
controlSingleparty
(1) (2) (3) (4) (5) (6) (7)Small bank asset share
of all banking assetsin state 1
Relative size of insur-ance to banking plusinsurance in the state 2 034 1
Indicator is 1 if bankscan sell insurance inthe state 020 2 012 1
Small rm share of thenumber of rms inthe state 030 2 011 005 1
Capital ratio of smallbanks relative tolarge in the state 2 066 020 2 010 2 006 1
Share of state govern-ment controlled byDemocrats 2 006 001 2 018 2 018 005 1
Indicator is 1 if statecontrolled by oneparty 007 012 001 2 006 2 006 030 1
Average yield on bankloans in the stateminus Fed funds rate 2 001 2 015 003 022 2 010 2 003 2 006
The number of observations is 637 except for the average interest rate on bank loans which becomesavailable only in 1976 so the number of observations is 408
TABLE IIIHAZARD MODEL OF POLITICAL ECONOMY FACTORS AFFECTING THE TIMING OF STATE
BRANCHING DEREGULATION 1970ndash1992
(1) (2) (3) (4) (5) (6)Small bank asset share of
all banking assets instate
648(122)
519(137)
632(135)
516(118)
488(133)
445(088)
Capital ratio of smallbanks relative to largein the state
1325(348)
921(319)
1301(362)
1067(346)
1144(375)
906(381)
Relative size of insurancein states where banksmay sell insurance 0otherwise
324(161)
259(149)
315(165)
285(154)
272(159)
055(134)
Indicator is 1 if banksmay sell insurance inthe state
045(010)
037(011)
045(010)
038(008)
036(009)
001(012)
Relative size of insurancein states where banksmay not sell insurance0 otherwise
2 093(043)
2 090(033)
2 102(051)
2 093(048)
2 096(053)
2 004(039)
Small rm share of thenumber of rms in thestate
2 972(211)
2 614(215)
2 952(211)
2 943(230)
2 952(270)
2 1509(276)
Share of state govern-ment controlled byDemocrats
031(012)
023(011)
030(012)
027(011)
026(011)
011(013)
Indicator is 1 if state con-trolled by one party
2 004(010)
2 004(007)
2 004(007)
2 001(009)
2 002(010)
017(009)
Average yield on bankloans in the stateminus Fed funds rate mdash
023(319) mdash mdash mdash mdash
Bank failure rate mdash mdash092
(111) mdash mdash mdash
Indicator is 1 if state hasunit banking law mdash mdash mdash
018(010)
019(010)
030(009)
Indicator is 1 if statechanges bank insur-ance powers mdash mdash mdash mdash
2 008(012)
2 026(013)
Includes regionalindicators No No No No No Yes
N 637 408 637 637 637 637Log likelihood 2 374 900 2 360 2 212 2 191 1192p-value of x 2 for
regression 001 001 001 001 001 001
The hazard model is Weibull where the dependent variable is the log expected time to branchingderegulation ln (T) 5 bx 1 eAll variables are measured for each state in each year Small bank asset share isthe percent of banking assets in the state held by banks below the median size of banks in each state in eachyear Relative capital ratio is the capital to assets ratio of small banks minus that of large banks Size ofinsurance relative to banking plus insurance in the state is measured as gross state product from insurancedivided by gross state product from insurance plus banking Insurance indicator is 1 if state law permits banksto enter the insurance business Small rm share is the percent of all establishments in the state that havefewer than twenty employees Party control variable is the share of the three bodies of state governmentcontrolled by Democrats One party control indicator is 1 if the same party controlrsquos the governorship and hasmajorities in both chambers of the state legislature Average yield on bank loans equals total interest incomeon all domestic loans made by banks in the state divided by total loans Bank failure rate is the ratio of assetsin failed banks during a given state and year divided by total bank assets in that state and year Unit bankingindicator equals 1 for states with unit banking restrictions The change in insurance powers indicator equalsone if the state changed the law to permit banks to sell insurance during the sample period We divide thecountry into four regions and include regional indicators in column (5) Robust standard errors are inparentheses and denote statistically signicant at the 1 5 and 10 percent levels
QUARTERLY JOURNAL OF ECONOMICS1452
The positive and statistically signicant coefficient on smallbank share implies that a greater small bank share delaysderegulation and this effect is economically important A one-standard-deviation increase in the small bank share results in a30 percent increase in the time until deregulation or about 47years This result supports the implications of private-interesttheory concerning intraindustry rivalry and is contrary to thepublic-interest theory
The positive and statistically signicant coefficient on therelative performance of small banks in the state implies thatstates deregulate later when small banks are relatively strongnancially25 A one-standard-deviation increase in the relativecapital-to-asset ratio results in a 15 percent rise in the time untilderegulation or about 24 years This result is consistent with theprivate-interest theory but may also be consistent with a public-interest interpretation because the risk of failures (and a costlytaxpayer-funded bailout) increases as small bank nancial healthdeclines
Interindustry competition also helps explain the timing ofderegulation In states where banks can sell insurance a rela-tively large insurance sector is associated with an increase in theexpected time to deregulation26 A one-standard-deviation in-crease in the relative size of the insurance sector in those statesthat permit banks to sell insurance leads to a 22 percent increasein the time until deregulation or about 35 years27 This result isconsistent with the private-interest theory but not the public-interest theory
Deregulation occurs earlier in states where small bank-dependent rms are relatively numerous A one-standard-
25 We also allowed small bank and large bank capital-to-asset ratios to enterseparately Their coefficient estimates were nearly equal in absolute value and ofopposite signs and so the data do not reject the relative small to large bank capitalratio specication we have chosen
26 When we use the relative size of the insurance sector in the state as awhole rather than insurance relative to banking plus insurance in the state weobtain similar results When we simultaneously include both measures of therelative size of insurance the interactions of the lsquolsquoinsurance relative to bankingplus insurancersquorsquo variable continue to be statisticallysignicant but the interactionsof lsquolsquoinsurance relative to the state as a wholersquorsquo variable are not
27 We have also estimated the model with the indicator variable equal to onefor states where banks may sell insurance but without the variables measuringtheimportance of insurance relative to banking In this model the coefficient on theinsurance indicator variablemdashan estimate of the difference in the expected time toderegulation in states where banks may sell insurance relative to states wherebanks may not sell insurancemdashequals 027 and is statistically signicant at the 5percent level This suggests that branching deregulation takes about 45 yearslonger in states where banks may sell insurance
WHAT DRIVES DEREGULATION 1453
deviation increase in the share of small rms reduces the timeuntil deregulation by 18 percent or about three years This resultconcerning the interests of consumers of banking services isconsistent with both the private- and public-interest theories
Column (2) of Table III introduces the average interest rateon loans to the model as a rough proxy for the cost to bankborrowers of these regulations These data become available onlyin 1976 so our sample size shrinks to 408 observations Thecoefficient on this variable is small and not statistically signi-cant The public-interest theory would imply that the coefficientshould be negative while the implication of the economic theorydoes not have a clear sign prediction
When we add the failure rate of banks to the specication incolumn (3) its coefficient is small and statistically insignicantWe do not nd a linkage between the timing of state deregulationand statewide banking distress Given that deposit insurance isfunded federally however bank failures may raise the publicrsquosconcerns about the costs of the regulations at the national ratherthan the state level This interpretation can explain why the rateof bank failure does not affect the timing of state-level deregula-tion even though bank instability does appear to rise generallyduring the period of increasing deregulation (see Section V below)
Finally the partisan structure of the state government alsoinuences when states deregulate As expected a higher propor-tion of Democrats in the government tends to delay deregulationA one-standard-deviation rise in the share of the governmentcontrolled by Democrats slows the deregulation by about twoyears Whether the state is dominated by one party however doesnot appear to affect the timing of the deregulation
To summarize our ndings from the hazard model privateinterests appear to play an important role in the deregulatoryprocess While private interests and public interests do sometimescoincide the results on the relative share of small banks and largebanks and on the relative size of insurance where banks competeare consistent with a private-interest approach but are difficult toexplain on public-interest grounds
B Robustness Tests
Omitted Variables States begin the sample period withdifferent degrees of restrictions on bank expansion These initialconditions may proxy for omitted factors determining the political-economy equilibrium in the state that could then affect the
QUARTERLY JOURNAL OF ECONOMICS1454
subsequent timing of deregulation For instance some statesbegan the period with unit banking while other states allowedlimited branching and the formation of multibank holdingcompanies
To test whether the stringency of initial restrictions onbranching matters we rst divide the sample into states with andwithout the most extreme form of branching regulation unitbanking28 Column (4) of Table III includes an indicator variableequal to one if the state began the period with unit bankingrestrictions The coefficient on this variable is positive andstatistically signicant but the other results remain virtuallyunchanged Next we include an indicator variable that is one ifthe state permitted multibank holding companies (MBHCs) toexist at the beginning of our sample period Thirty-four statesbegan the period allowing MBHCs The coefficient estimate forthis indicator variable (not reported) is small and not statisticallysignicant and its inclusion does not change the effects of theother variables
Eleven states expand bank powers to include the sale ofinsurance products during our sample period This could proxy foromitted factors about the relative strength of the insuranceindustry in the state In column (5) we include an indicatorvariable that is one if the state changes the law during our sampleperiod to permit banks to sell insurance business and zerootherwise The coefficient estimate is small and statisticallyinsignicant and does not affect the other results29
In column (6) we introduce four regional indicator variablesNorth South Midwest and West30 The coefficients on small bankshare the relative capital-to-assets ratio and small rm sharechange little Although the coefficient estimate still suggests that
28 We classify states that prohibited branching but permitted banks toestablish facilities as unit banking states The unit banking indicator is one for thefollowing states CO AR FL IL IA KS MN MO MT NE ND OK TX WI WVand WY
29 We conducted a likelihood ratio test of whether the unit banking statesthe MBHC states and the states that changed their regulations of bank insurancepowers could be pooled with the other states In all three cases we could not rejectpooling of the data
30 Since intrastate deregulation generally preceded interstate deregulationand the latter typically took the form of regional interstate compacts thepossibility of participating in a regional interstate banking compact could haveinuenced the decision to deregulate intrastate branching Our denitions of theregions are region 1 (South) contains AL AR DC FL GA KY LA MS NC OKSC TN TX and VA region 2 (Northeast) contains CT MA MD ME NH NJ NYPA RI VT and WV region 3 (Midwest) contains IA IL IN KS MI MN MO NEND OH SD and WI region 4 (West) contains the other states
WHAT DRIVES DEREGULATION 1455
in states where banks can sell insurance a larger relative size ofthe insurance sector delays deregulation the effect is no longerstatistically signicant The coefficient on the indicator variablefor whether a state expands bank powers to permit insurancesales during the sample period however now becomes statisti-cally signicantly negative This implies that deregulation occursroughly 41 years earlier in states where the insurance lobby maybe relatively weak While the inclusion of the regional indicatorsdoes not affect the other results the results on insurance doappear to be sensitive to the specication of the hazard model31
Estimation Technique To check the robustness of using ahazard model and to investigate whether the time-series or thecross-sectional variation in each of the factors accounts for thehazard results we estimate two simple linear probability models(neither are reported) First we run a pooled time-series cross-section OLS regression using the 637 observations from thehazard model The dependent variable equals one in the year ofderegulation and zero otherwise We include the same set ofregressors that were used in the hazard model and add state-levelxed effects This specication thus removes the cross-statevariation from the independent variables and focuses on theirmovements over time In this model the signs of the coefficientson the interest group factors are consistent with the results fromthe hazard model and the coefficients for the small bank relativecapital ratio and for the relative size of insurance where states cansell insurance are statistically signicant
In the second model we eliminate the time-series variationby averaging each of the explanatory variables over time We thushave one observation per state that had branching restrictions asof 1970 (N 5 39) The dependent variable equals the inverse of thenumber of years until deregulation which represents an estimateof the statersquos probability of deregulating in a year under theassumption that this probability is constant over time and eachyear provides an independent observation32 Although we haveonly 39 observations the coefficients on the small bank marketshare small bank relative capital ratio and small rm share
31 As further robustness check we also tried a population density variablewhich Abrams and Settle [1993] found relevant for regulatory change during the1920s and 1930s but its coefficient was not statistically signicant and did notaffect the other results
32 For instance the dependent variable would take the value of 01 for astate that deregulated in 1980 For the three states that did not deregulate thedependent variable equals 0
QUARTERLY JOURNAL OF ECONOMICS1456
variables are statistically signicant The signs of the effects ofthe insurance variables are the same as those reported in thehazard model but not statistically signicant
The hazard model and linear probability models thus producevery similar results The estimates from the latter suggest thatthe effects for the small bank market share and small rm sharevariables are driven primarily by cross-state variation the effectof the insurance variable is driven primarily by time-seriesvariation and the small bank capital effect is driven by bothcross-sectional and time-series variation
Ex Post Consequences of Deregulation Finally to check theplausibility of our interpretation of the hazard results we con-sider whether the ex post consequences of deregulation areconsistent with the ex ante lobbying positions we attribute to eachinterest group (see Kroszner and Strahan [1998] for details) Wend that small banks lose market share following deregulationand in states where banks can enter the insurance business theinsurance sector shrinks relative to the banking sector followingderegulation Borrowers also benet from deregulation throughlower average interest rates on loans These ndings support theprivate-interest interpretation of the coefficients in the hazardmodel groups that will benet lobby to speed deregulation andthose that will be harmed lobby to slow it
C Voting on Interstate Branching Deregulation in the U SHouse of Representatives
We now examine whether the forces driving intrastate branch-ing deregulation also drive interstate deregulation Financialservices interests are active contributors and lobbyists in Washing-ton Their political action committees constitute the largest groupof contributors to legislators providing nearly 20 percent of totalcongressional campaign contributions [Makinson 1992] and muchof their lobbying effort involves competition among rival interestswithin nancial services (see Kroszner and Stratmann [1998])
As noted above after virtually all states adopted intra- andinterstate branching deregulation the 1994 Riegle-Neal Act re-pealed the 1927 McFadden Act to eliminate all barriers tointerstate banking and branching by 1997 The key votes concern-ing the Riegle-Neal Act were either voice votes or extremelylopsided so we could not estimate a voting model from them Anumber of bills and amendments related to interstate branching
WHAT DRIVES DEREGULATION 1457
had been debated in Congress during the years prior to thepassage of the Riegle-Neal Act but a search of the weekly BNABanking Reporter and the Congressional Record produced onlyone roll-call vote related to interstate branching that was notlopsided This vote occurred in the House on November 14 1991and concerned an amendment sponsored by Wylie (R-OH) andNeal (D-NC) to introduce interstate banking and branchingderegulation to a nancial services reform package33 Althoughthe amendment passed by 210 to 208 the bill to which it wasattached subsequently was defeated
To check for the inuence of the factors we considered in thestate-level reforms we examine both the sponsorship of interstatebanking legislation and voting on the amendment The sponsorsof the Wylie-Neal amendment are from states with low small bankshares 004 in Ohio (Wylie) and 002 in North Carolina (Neal) Incontrast the sample mean in 1991 is 008 (median 5 007)Michigan home state of the Senatersquos sponsor of the 1994 Riegle-Neal Act also had relatively low small bank strength (small bankshare of 005)
Table IV reports estimates of a probit model where thedependent variable equals one if the legislator voted in favor ofthe amendment and zero otherwise The explanatory variablesare the same as those in the hazard model except that we use aparty affiliation indicator variable equal to one for Democrats andzero for Republicans as our proxy for political factors Thecoefficients in Table IV are the marginal effects (lsquolsquoslopesrsquorsquo) of aone-unit change of each variable on the probability that alegislator will vote for the amendment34 Note that a forcefavoring deregulation has a negative coefficient in the hazardmodel (Table III) but a positive coefficient in the probit model
Consistent with the state-level deregulation process legisla-tors are more likely to support the amendment if their states havea relatively low share of small banks As in the hazard model thefraction of small banks is the most important interest groupinuence on a legislatorrsquos voting decision A one-standard-deviation increase in small banksrsquo market share (from the mean)is associated with a decline in the probability of voting in favor of
33 The Wylie-Neal amendment also included provisions limiting certaininsurance and real estate powers of national banks [Congressional RecordNovember 14 1991 pp 10239ndash10242]
34 Since we have multiple legislators from each state we adjust the standarderrors to correct for the lack of independence among observations clustered in thesame state
QUARTERLY JOURNAL OF ECONOMICS1458
TAB
LE
IVM
AR
GIN
AL
EF
FE
CT
SF
RO
MA
PR
OB
ITM
OD
EL
OF
TH
EIN
FL
UE
NC
EO
FP
OL
ITIC
AL
EC
ON
OM
YF
AC
TO
RS
ON
TH
EH
OU
SE
VO
TE
ON
TH
EW
YL
IE-N
EA
L
AM
EN
DM
EN
TT
OP
ER
MIT
INT
ER
ST
AT
EB
AN
KB
RA
NC
HIN
G
NO
VE
MB
ER
14
1991
(1)
(2)
(3)
(4)
(5)
(6)
Smal
lban
kas
set
shar
eof
allb
anki
ng
asse
tsin
stat
e2
417
(12
4)2
454
(12
2)2
407
(12
0)2
443
(14
7)2
450
(14
9)2
490
(14
7)C
apit
alra
tio
ofsm
allb
anks
rela
tive
tola
rge
inth
est
ate
180
(37
5)2
68(3
42)
256
(37
5)1
69(3
78)
153
(38
0)2
45(4
25)
Rel
ativ
esi
zeof
insu
ranc
ein
stat
esw
here
bank
sm
ayse
llin
sura
nce
0ot
herw
ise
22
34
(10
2)2
249
(1
14)
22
33
(10
1)2
249
(1
04)
22
51
(10
2)2
154
(11
1)In
dica
tor
is1
ifba
nks
may
sell
insu
ranc
ein
the
stat
e0
79
(0
06)
074
(00
6)0
69
(0
06)
070
(00
6)0
70
(0
06)
067
(00
7)R
elat
ive
size
ofin
sura
nce
inst
ates
whe
reba
nks
may
not
sell
insu
ranc
e0
oth
erw
ise
180
(06
6)2
42
(0
72)
154
(0
67)
173
(06
6)1
72
(0
67)
203
(0
91)
Smal
lrm
shar
eof
the
num
ber
ofr
ms
inth
est
ate
21
23(0
92)
20
83(0
89)
21
62(1
07)
21
32(0
95)
21
30(1
00)
20
94(1
16)
Indi
cato
ris
1if
Rep
rese
ntat
ive
isa
Dem
ocra
t2
035
(00
7)2
035
(00
7)2
036
(00
7)2
035
(00
7)2
035
(00
7)2
036
(00
7)A
vera
geyi
eld
onba
nklo
ans
inth
est
ate
min
usF
edfu
nds
rate
mdash10
01
(51
5)mdash
mdashmdash
mdash
Ave
rage
fail
ure
rate
198
5to
1991
mdashmdash
228
(14
1)mdash
mdashmdash
Indi
cato
ris
1if
stat
eh
asun
itba
nkin
gla
wmdash
mdashmdash
004
(00
7)0
04(0
07)
20
01(0
08)
Indi
cato
ris
1if
stat
ech
ange
sba
nk
insu
ran
cepo
wer
smdash
mdashmdash
mdash2
001
(00
7)2
003
(00
7)In
clud
esre
gion
alin
dica
tors
N
oN
oN
oN
oN
oYe
sP
seud
o-R
20
170
180
170
170
170
18L
ogli
keli
hood
224
024
223
826
223
951
224
013
224
010
223
808
p-va
lue
ofx2
for
regr
essi
on
001
0
01
001
0
01
001
0
01
Th
ede
pen
den
tvar
iabl
eis
one
ifth
ele
gisl
ator
vote
sfo
rth
eam
endm
enta
nd
zero
ifag
ains
tT
he
repo
rted
coef
fici
ents
are
the
effe
cts
ofa
uni
tch
ange
ofth
ein
depe
nde
ntva
riab
le(f
rom
the
mea
n)on
the
prob
abili
tyof
voti
ng
infa
vor
ofth
eam
endm
ent
All
vari
able
sar
em
easu
red
byst
ate
N5
418
the
num
ber
ofR
epre
sent
ativ
esvo
tin
gon
the
amen
dmen
tS
mal
lban
kas
sets
har
eis
the
perc
ent
ofba
nki
ng
asse
tsin
the
stat
eh
eld
byba
nks
belo
wth
em
edia
nsi
zeof
ban
kin
each
stat
ein
each
year
Rel
ativ
eca
pita
lrat
iois
the
capi
talt
oas
sets
rati
oof
smal
lban
ksm
inu
sth
atof
larg
eba
nks
Siz
eof
insu
ran
cere
lati
veto
bank
ing
plu
sin
sura
nce
inth
est
ate
ism
easu
red
asgr
oss
stat
epr
oduc
tfro
min
sura
nce
divi
ded
bygr
oss
stat
epr
odu
ctfr
omin
sura
nce
plu
sba
nkin
gIn
dica
tor
ison
eif
stat
ela
wpe
rmit
sba
nks
toen
ter
the
insu
ran
cebu
sin
ess
Sm
all
rmsh
are
isth
epe
rcen
tofa
lles
tabl
ishm
ents
inth
est
ate
that
hav
efe
wer
than
twen
tyem
ploy
ees
Par
tyin
dica
tor
ison
eif
the
Rep
rese
nta
tive
isa
Dem
ocra
tA
vera
geyi
eld
onba
nklo
ans
equa
lsto
tali
nter
est
inco
me
onal
ldom
esti
clo
ans
mad
eby
ban
ksin
the
stat
edi
vide
dby
tota
lloa
ns
The
aver
age
bank
failu
rera
teis
the
rati
oof
asse
tsin
faile
dba
nks
duri
ng
agi
ven
stat
edi
vide
dby
tota
lban
kas
sets
inth
atst
ate
aver
aged
over
the
year
s19
85to
1991
Un
itba
nki
ng
indi
cato
req
ual
son
efo
rst
ates
wit
hun
itba
nkin
gre
stri
ctio
nsT
hech
ange
inin
sura
nce
pow
ers
indi
cato
req
ual
son
eif
the
stat
ech
ange
dth
ela
wto
perm
itba
nks
tose
llin
sura
nce
duri
ngth
esa
mpl
epe
riod
We
divi
deth
eco
unt
ryin
tofo
urre
gion
san
din
clu
dere
gion
alin
dica
tors
inco
lum
n(5
)R
obu
stst
anda
rder
rors
are
inpa
ren
thes
es
an
d
den
ote
stat
isti
cally
sign
ica
nta
tth
e1
5an
d10
perc
entl
evel
s
WHAT DRIVES DEREGULATION 1459
branching of approximately 17 percent The relative capital-to-assets ratio of small and large banks however does not have astatistically signicant effect in the probit model
The impact of rival interests outside of banking is alsoconsistent with intrastate deregulation results Where banks cansell insurance legislators from states with larger insurancesectors relative to banking are less likely to vote for the amend-ment35 A one-standard-deviation increase in the relative size ofinsurance in those states which permit banks to sell insurancedecreases the probability that a legislator will favor the amend-ment by about 13 percent Turning to consumer interests thecoefficient on small rm share is not statistically signicant butthe coefficient on interest rates in the state is A one-standard-deviation increase in the average interest rate on loans raises theprobability that a legislator will support the amendment byroughly 8 percent The coefficient estimate for the bankingdistress variable measured as the average bank failure rate from1985 to 1991 is small and statistically signicant Overall theprobit analysis of the vote on national branching deregulationsupports the private-interest theory of deregulation and providesa consistency check that the importance of interests operating onthe state legislatures are very similar to those operating at thefederal level
V WHY DID BRANCHING DEREGULATION BEGIN DURING THE
1970S SHOCKS TO THE EQUILIBRIUM SUPPORTING BANK
BRANCHING RESTRICTIONS
A complete explanation of regulatory exit should account forwhy deregulation begins in the 1970s as well as the timing of thestate-by-state reforms [Kane 1996] In this section we describebroad technological legal and economic shocks that altered thepolitical-economy equilibrium which had kept antibranching regu-lations little changed for at least 30 years
Beginning in the 1970s three innovations reduced the valueto the protected banks of local geographic monopolies (see TableV) First automatic teller machines (ATMs) helped to erode thegeographic ties between customers and banks Second checkablemoney market mutual funds and the Merrill Lynch Cash Manage-ment Account demonstrated that banking by mail and telephone
35 The positive and statistically signicant estimates on the other insurancevariables may reect insurance industry support for the amendmentrsquos provisionslimiting banksrsquo insurance powers
QUARTERLY JOURNAL OF ECONOMICS1460
provided a convenient alternative to local banks36 Third techno-logical innovation and deregulation reduced transportation andcommunication costs particularly since the 1970s thereby lower-ing the costs for customers to use distant banks By increasing theelasticity of deposits supplied to banks these innovations reducedthe value of geographical restrictions to their traditional benecia-ries and thereby reduced their incentive to ght to maintain them[Peltzman 1976]
On the lending side increasing sophistication of credit-scoring techniques following innovations in information process-ing nancial theory and the development of large credit databases began to diminish the value of knowledge that localbankers had about the risks of borrowers in the community As a
36 Regulation Q which limited the interest rates that banks could pay ondeposits may have helped to drive depositors away from banks when the gapbetween market rates and deposit ceilings grew during the 1970s The eliminationof interest rate ceilings on large denomination certicates of deposit during the1970s appears to have hurt smaller and retail-oriented banks relative to largerwholesale banks [James 1983]
TABLE VBROAD TRENDS IN COMMERCIAL BANKING 1950ndash1995
YearNumberof ATMs
Domesticbank
deposits(billions)
Moneymarketmutual
fund(billions)
Percent ofdeposits 1
moneyfunds heldby banks
Smallbanksrsquo
percent ofbankingassets
Averagenumber of
bankfailures
(1) (2) (3) (4) (5) (6)1950 0 $ 154 $ 0 100 NA 41955 0 191 0 100 NA 31960 0 228 0 100 24 21965 0 330 0 100 20 41970 0 479 0 100 18 61975 9750 775 4 99 18 61980 18500 1182 76 94 17 101985 61117 1787 242 88 14 601990 80156 2339 493 83 11 1791995 122706 2552 745 77 8 61
Column (1) ATM gures are from Bank Network News The EFT Network Data Book (New YorkFaulkner and Gray Inc) The 1975 gure was unavailable 9750 is the number of ATMs in 1978 the rst yearfor which complete data are available Columns (2)ndash(4) Banksrsquo domestic deposits are from the Reports ofIncome and Condition money market mutual funds are from the Flow of Funds Data on all bank depositsforeign plus domestic are only available beginning in 1970 The trend in banksrsquo share (column (4)) is the sameusing total deposits instead of domestic deposits Column (5) Percent of banking assets held by small bankswhere a small bank is dened as a commercial bank with less than $100 million in assets in 1994 dollarsThese data are based on the Reports of Income and Condition Data on small banks are not available before1960 Column (6) Five-year average number of bank failures where the nal year is indicated in the rstcolumn These data are from FDIC Annual Report and the Quarterly Banking Prole
WHAT DRIVES DEREGULATION 1461
result of these innovations a national market developed forresidential mortgages credit card receivables have been securi-tized and bank lending to small business now relies less on thejudgment of loan officers and more on standardized scoringmodels
These changes have increased the potential protability forlarge banks to enter what had been the core of small bankactivities Large banksrsquo incentive to increase their lobbying pres-sure to be able to expand into these markets has thus beenincreasing over time In fact small banksrsquo market share began todecline even prior to the branching deregulation (Table V) As thevalue of a local banking relationships declined small rms thatwere the main borrowers from the small banks also probablybecame more likely to favor the entry of large banks into localmarkets With the deadweight costs of preventing large bankentry rising the private-interest theory predicts that small localbanks would become less likely able to maintain the branchingrestrictions [Becker 1983] Deregulation that reduces deadweightcosts of regulation is also consistent with the public-interesttheory
Kane [1996] argues that another major shock to the oldequilibrium was the increasing public awareness of the costlinessof having government-insured but (geographically) undiversiednancial institutions In the late 1970s the failure rate of banksbegan to rise (Table V) In the 1980s the Savings and Loan crisisand taxpayer bailout further heightened public awareness aboutthe costs of restrictions that make depository institutions morelikely to require infusions of taxpayer funds The failures thusmay have heightened public support for branching deregulation(see also Abrams and Settle [1993]) As noted above we did notnd any evidence that banking failures or distress in a stateaffected the speed with which the state deregulated but distressstill may be a common factor affecting all states
These technological economic and legal shocks generatedconditions that changed the long-standing balance favoring theantibranching forces The marginal value of lobbying to repealbranching restrictions increased just as the relative value to thesmall banks of maintaining branching restrictions was decliningThese nationwide shocks are common factors across states andthey are consistent with the positive duration dependence associ-ated with state-level deregulation that we demonstrated in FigureII In addition a broad change in lsquolsquoideologyrsquorsquo against government
QUARTERLY JOURNAL OF ECONOMICS1462
regulation could explain the positive duration dependence duringthe sample period
VI CONCLUSIONS
The private-interest theory of regulation can account for thepattern of bank branching deregulation of the 1970s and 1980sBeneciaries of branching regulation had supported a coalitionfavoring geographical restrictions despite their costs to consum-ers of nancial services Innovations that began in the 1970saltered the value of the restrictions to the affected parties and theresulting competition among interest groups can explain thesubsequent deregulation While some of our results are alsoconsistent with the public interest theorymdashfor example deregula-tion occurs earlier when small banks are in a relatively weaknancial positionmdashother results particularly evidence on theimportance of rivalries between small and large banks andbetween banking and insurance are difficult to explain with thepublic-interest approach We also nd that political-institutionalfactors affect deregulation although these variables may act asproxies for unmeasured economic interests The greater success ofthe private-interest theory to explain deregulation here than inprevious work may be due to the richer cross-sectional andtime-series variation in branching deregulation that can beexploited with a hazard model Future empirical studies ofendogenous deregulation thus may be most fruitful where changehas occurred across states over time such as in franchisinginsurance and public utilities (eg Joskow [1996])
Technological and nancial innovations will continue to erodethe benets to any interest group of maintaining regulatorybarriers in nancial services These forces are likely to bringabout reforms both domestically for example through legislationthat would increase bank powers (see Kroszner [1996] Krosznerand Rajan [1994 1997] and Kroszner and Stratmann [1998]) andinternationally for example through the extension of nancialservices provisions of NAFTAto reduce geographic barriers acrosscountries [Kroszner 1997]
APPENDIX EFFECTS OF DIFFERENT TYPES OF BRANCHING
DEREGULATION ON ECONOMIC GROWTH BANKING STRUCTUREAND BANK EFFICIENCY
We measure the importance of alternative forms of deregula-tion by their consequences for the economic growth and the
WHAT DRIVES DEREGULATION 1463
structure and efficiency of the banking sector across the states(see Jayaratne and Strahan [1996 1998]) To do so we run anOLS panel regression relating indicators for each type of deregula-tion to potential consequences Our dependent variables includeproxies for economic growth (growth in personal income) bankstructure (the log of the number of banking companies and thenumber of branches per banking company) and efficiency (the logof total noninterest expenses and the ratio of loan loss provisionsto total loans) The four deregulation indicators (MBHC branch-ing through merger and acquisition unrestricted statewide branch-ing and interstate banking) are equal to one in the years afterderegulation and zero before If a state deregulated before thesample period begins in 1970 the indicator is one throughout ifthe state does not deregulate then the indicator is zero through-out All regressions include state xed effects and time effects
Table VI shows the results of including the set of four
TABLE VIMULTIVARIATE REGRESSIONS OF THE EFFECTS OF DIFFERENT TYPES OF BRANCHING
DEREGULATION ON ECONOMIC GROWTH BANKING STRUCTUREAND BANK EFFICIENCY
Dependent variable
Growthin
personalincome
Log of of
bankingcompanies
Branchesper
bankingcompany
Log ofnon-
interestexpenses
Loan lossprovisions
(loans 1leases)
Merger amp acquisitionbranchingindicator
0013(0004)
2 0162(0037)
1146(0453)
2 0107(0035)
2 0007(0001)
Unrestricted state-wide branchingindicator
2 0008(0005)
0086(0059)
2 0043(0574)
0070(0037)
0005(0002)
Interstate bankingindicator
0006(0005)
2 0052(0057)
1064(0511)
0005(0030)
2 0001(0001)
Multibank holdingcompany indicator
2 0005(0006)
2 0096(0052)
2 0364(0529)
2 0095(0036)
2 0001(0001)
Dependent variablemean [median]
0086[0082]
4881[5136]
6649[5055]
13812[13803]
0009[0006]
This table contains regressions of state-level economic growth and state-level measures of the structureand efficiency of the banking industry Each dependent variable is regressed on a set of state indicatorvariables a set of year indicator variables and the set of four deregulation indicators these deregulationindicators are equal to one after deregulation and zero before If a state was always regulated the indicator iszero throughout if it was always deregulated the indicator is one throughout Sample period is 1975 to 1991and Delaware and South Dakota are dropped so N 5 833 Noninterest expenses become available only in1984 so N 5 392 The coefficient on the deregulation indicators is reported with robust standard errors inparentheses and denote statistically signicant at the 1 5 and 10 percent levels
QUARTERLY JOURNAL OF ECONOMICS1464
deregulation indicators simultaneously The branching by mergerand acquisition indicator is statistically signicant and of theexpected sign in all ve of the equations and larger in absolutevalue than any of the other coefficients on the indicators for theother types of deregulation Since there may be multicollinearityamong our indicators we also ran univariate specications whichincluded each indicator separately and the results are verysimilar These results imply that branching by merger andacquisition is the most important type of deregulation measuredby its consequences for economic growth banking structure andbank efficiency See Kroszner and Strahan [1998] for more details
UNIVERSITY OF CHICAGO GRADUATE SCHOOL OF BUSINESS AND NATIONAL BUREAU OF
ECONOMIC RESEARCH
FEDERAL RESERVE BANK OF NEW YORK RESEARCH amp MARKET ANALYSIS GROUP
REFERENCES
Abrams Burton A and Russell F Settle lsquolsquoPressure-Group Inuence and Institu-tional Change Branch-Banking Legislation during the Great DepressionrsquorsquoPublic Choice LXXVII (1993) 687ndash705
Amel Dean lsquolsquoState Laws Affecting the Geographic Expansion of CommercialBanksrsquorsquo unpublished manuscript Board of Governors of the Federal ReserveSystem 1993
Becker Gary S lsquolsquoA Theory of Competition among Pressure Groups for PoliticalInuencersquorsquo Quarterly Journal of Economics XCVIII (1983) 371ndash400
Berger Allen Anil Kashyap Joseph Scalise lsquolsquoThe Transformation of the U SBanking Industry What a Long Strange Trip Itrsquos Beenrsquorsquo Brookings Papers onEconomic Activity 2 (1995) 55ndash218
Berger Allen Anthony Saunders Joseph Scalise and Gregory Udell lsquolsquoThe Effectsof Bank Mergers and Acquisitions on Small Business Lendingrsquorsquo Journal ofFinancial Economics L (1998) 187ndash229
Brennan Geoffrey and James Buchanan lsquolsquoTowards a Tax Constitution forLeviathanrsquorsquo Journal of Public Economics VIII (1977) 255ndash273
Brickley James A and Christopher M James lsquolsquoThe Takeover Market CorporateBoard Composition and Ownership Structure The Case of Bankingrsquorsquo Journalof Law and Economics XXX (1987) 161ndash180
Cole Rebel and John Wolken lsquolsquoFinancial Services Used by Small BusinessesEvidence from the 1993 Survey of Small Business Financesrsquorsquo Federal ReserveBulletin LXXXI (1994) 629ndash676
Dixit Avinash The Making of Economic Policy A Transactions-Cost PoliticsPerspective (Cambridge MA MIT Press 1996)
Economides Nicolas R Glenn Hubbard and Darius Palia lsquolsquoThe Political Economyof Branching Restrictions and Deposit Insurancersquorsquo Journal of Law andEconomics XXIX (1996) 667ndash704
Flannery Mark lsquolsquoThe Social Costs of Unit Banking Restrictionsrsquorsquo Journal ofMonetary Economics XIII (1984) 237ndash249
Greene William H Econometric Analysis third edition (Upper Saddle River NJPrentice-Hall 1997)
Gunther Jeffery lsquolsquoRegional Capital Imbalances and the Removal of InterstateBanking Restrictionsrsquorsquo Economics Letters XLIV (1994) 439ndash442 lsquolsquoThe Erosion of Intrastate Branch Banking Restrictionsrsquorsquo unpublishedmanuscript Federal Reserve Bank of Dallas 1996
Irwin Douglas and Randall Kroszner lsquolsquoInterests Institutions and Ideology inSecuring Policy Change The Republican Conversion to Trade Liberalizationafter Smoot-Hawleyrsquorsquo Journal of Law and Economics XLII (1999) 643ndash673
WHAT DRIVES DEREGULATION 1465
James Christopher lsquolsquoAn Analysis of Intra-Industry Differences in the Effect ofRegulation The Case of Deposit Rate Ceilingsrsquorsquo Journal of Monetary Econom-ics XII (1983) 417ndash442
Jarrell Gregg lsquolsquoChange at the Exchange The Causes and Effects of DeregulationrsquorsquoJournal of Law and Economics XXVII (1984) 273ndash312
Jayaratne Jith and Philip E Strahan lsquolsquoThe Finance-Growth Nexus Evidencefrom Bank Branch Deregulationrsquorsquo Quarterly Journal of Economics CI (1996)639ndash670
Jayaratne Jith and Philip E Strahan lsquolsquoEntry Restrictions Industry Evolutionand Dynamic Efficiency Evidence from Commercial Bankingrsquorsquo Journal of Lawand Economics XLIX (1998) 239ndash274
Joskow Paul lsquolsquoComment on lsquoPower Struggles Explaining Deregulatory Reformsin Electricity Marketsrsquo rsquorsquo Brookings Papers on Economic Activity Microeconom-ics (1996) 251ndash264
Joskow Paul and Roger Noll lsquolsquoRegulation in Theory and Practice An Overviewrsquorsquoin Studies in Public Regulation Gary Fromm ed (Cambridge MA MITPress 1981) pp 1ndash65
Kahn Matthew E and John G Matsusaka lsquolsquoDemand for Environmental GoodsEvidence from Voting Patterns on California Initiativesrsquorsquo Journal of Law andEconomics XL (1997) 137ndash173
Kalbeisch J and R Prentice The Statistical Analysis of Failure Time Data (NewYork John Wiley amp Sons 1980)
Kane Edward J lsquolsquoDe Jure Interstate Banking Why Only NowrsquorsquoJournal of MoneyCredit and Banking XXVIII (1996) 141ndash161
Kiefer Nicholas lsquolsquoEconomic Duration Data and Hazard Functionsrsquorsquo Journal ofEconomic Literature XXVI (1988) 646ndash679
Kroszner Randall S lsquolsquoThe Evolution of Universal Banking and Its Regulation inTwentieth Century Americarsquorsquo in Universal Banking Financial System DesignReconsidered Anthony Saunders and Ingo Walter eds (New York IrwinProfessional Publishers 1996) pp 70ndash99 lsquolsquoThe Political Economy of Banking and Financial Regulation in the U Srsquorsquo inThe Banking and Financial Structure in the NAFTA Countries and ChileGeorge M von Furstenberg ed (Boston Kluwer Academic Publishers 1997)pp 200ndash213
Kroszner Randall S and Raghuram G Rajan lsquolsquoIs the Glass-Steagall ActJustied A Study of the U S Experience with Universal Banking before1933rsquorsquo American Economic Review LXXXIV (1994) 810ndash832
Kroszner Randall S and Raghuram G Rajan lsquolsquoOrganization Structure andCredibility Evidence from Bank Securities Activities before the Glass-Steagall Actrsquorsquo Journal of Monetary Economics XXXIX (1997) 475ndash516
Kroszner Randall S and Philip E Strahan lsquolsquoRegulatory Incentives and the ThriftCrisis Dividends Mutual-to-Stock Conversions and Financial DistressrsquorsquoJournal of Finance LI (1996) 1285ndash1320
Kroszner Randall S and Philip E Strahan lsquolsquoWhat Drives Deregulation Econom-ics and Politics of the Relaxation of Bank Branching Restrictionsrsquorsquo NBERWorking Paper No 6637 1998
Kroszner Randall S and Thomas Stratmann lsquolsquoInterest Group Competition andthe Organization of Congress Theory and Evidence from Financial ServicesPolitical Action Committeesrsquorsquo American Economic Review LXXXVIII (1998)pp 1163ndash1187
Kroszner Randall S and Thomas Stratmann lsquolsquoDoes Political Ambiguity PayCorporate Campaign Contributions and Rewards to Legislator Reputationrsquorsquounpublishedmanuscript University of Chicago Graduate School of Business1999
Lin D Y and L J Wei lsquolsquoThe Robust Inference for the Cox Proportional HazardsModel Journal of the American Statistical Association LXXXIV (1989)1074ndash1078
Macey Jonathan R and Geoffrey P Miller Banking Law and Regulation (BostonLittle Brown and Company 1992)
Makinson Larry Open Secrets The Encyclopedia of Congressional Money andPolitics (Washington DC Congressional Quarterly 1992)
Niskanen WilliamBureaucracy and Representative Government (Chicago Aldine-Atherton 1971)
QUARTERLY JOURNAL OF ECONOMICS1466
Noll Roger lsquolsquoComment on Peltzmanrsquorsquo Brookings Papers Microeconomics SpecialIssue (1989) 48ndash58
North Douglass lsquolsquoA Transactions Cost Theory of Politicsrsquorsquo Journal of TheoreticalPolitics II (1990) 355ndash367
Peltzman Sam lsquolsquoToward a More General Theory of Regulationrsquorsquo Journal of Lawand Economics XIX (1976) 109ndash148 lsquolsquoConstituent Interest and Congressional Votingrsquorsquo Journal of Law andEconomics XXVII (1984) 181ndash210 lsquolsquoThe Economic Theory of Regulation after a Decade of DeregulationrsquorsquoBrookings Papers Microeconomics Special Issue (1989) 1ndash41
Peterson Mitchell and Raghuram Rajan lsquolsquoThe Benets of Lending RelationshipsEvidence from Small Business Datarsquorsquo Journal of Finance XLIX (1994) 3ndash38
Poole Keith and Howard Rosenthal Congress A Political-Economic History ofRoll Call Voting (Oxford Oxford University Press 1997)
Stata Stata Reference Manual Release 5 (College Station TX Stata Press 1997)Stigler George J lsquolsquoThe Theory of Economic Regulationrsquorsquo Bell Journal of Economics
and Management Science II (1971) 3ndash21Stigler George J Chicago Studies in the Political Economy (Chicago University
of Chicago Press 1988)Strahan Philip E and James Weston lsquolsquoSmall Business Lending and the Changing
Structure of the Banking Industryrsquorsquo Journal of Banking and Finance XXII(1998) 821ndash845
Sylla Richard John Legler and John Wallis lsquolsquoBanks and State Public Finance inthe New Republic The United States 1790ndash1860rsquorsquo Journal of EconomicHistory XLVII (1987) 391ndash403
White Eugene The Regulation and Reform of the American Banking System1900ndash1929 (Princeton Princeton University Press 1983)
Winston Clifford lsquolsquoEconomic Deregulation Days of Reckoning for Microecono-mistsrsquorsquo Journal of Economic Literature XXXI (1993) 1263ndash1289
WHAT DRIVES DEREGULATION 1467
TABLE IISUMMARY STATISTICS FOR VARIABLES IN HAZARD MODEL 1970ndash1992
Panel A Univariate statistics MeanStandarddeviation Minimum Maximum
(1) (2) (3) (4)Small bank asset share of all
banking assets in state 011 005 001 023Relative size of insurance to
banking plus insurance inthe state 047 009 024 080
Indicator is 1 if banks can sellinsurance in the state 020 040 0 1
Small rm share of the numberof rms in the state 088 002 078 095
Capital ratio of small banksrelative to large in the state 002 001 2 001 009
Share of state government con-trolled by Democrats 065 036 0 1
Indicator is 1 if state controlledby one party 054 050 0 1
Average yield on bank loans inthe state minus Fed fundsrate 002 002 2 004 008
Panel B CorrelationsSmallbank
Relativesize of
insuranceInsuranceindicator
Smallrm
Smallbank
capitalDem
controlSingleparty
(1) (2) (3) (4) (5) (6) (7)Small bank asset share
of all banking assetsin state 1
Relative size of insur-ance to banking plusinsurance in the state 2 034 1
Indicator is 1 if bankscan sell insurance inthe state 020 2 012 1
Small rm share of thenumber of rms inthe state 030 2 011 005 1
Capital ratio of smallbanks relative tolarge in the state 2 066 020 2 010 2 006 1
Share of state govern-ment controlled byDemocrats 2 006 001 2 018 2 018 005 1
Indicator is 1 if statecontrolled by oneparty 007 012 001 2 006 2 006 030 1
Average yield on bankloans in the stateminus Fed funds rate 2 001 2 015 003 022 2 010 2 003 2 006
The number of observations is 637 except for the average interest rate on bank loans which becomesavailable only in 1976 so the number of observations is 408
TABLE IIIHAZARD MODEL OF POLITICAL ECONOMY FACTORS AFFECTING THE TIMING OF STATE
BRANCHING DEREGULATION 1970ndash1992
(1) (2) (3) (4) (5) (6)Small bank asset share of
all banking assets instate
648(122)
519(137)
632(135)
516(118)
488(133)
445(088)
Capital ratio of smallbanks relative to largein the state
1325(348)
921(319)
1301(362)
1067(346)
1144(375)
906(381)
Relative size of insurancein states where banksmay sell insurance 0otherwise
324(161)
259(149)
315(165)
285(154)
272(159)
055(134)
Indicator is 1 if banksmay sell insurance inthe state
045(010)
037(011)
045(010)
038(008)
036(009)
001(012)
Relative size of insurancein states where banksmay not sell insurance0 otherwise
2 093(043)
2 090(033)
2 102(051)
2 093(048)
2 096(053)
2 004(039)
Small rm share of thenumber of rms in thestate
2 972(211)
2 614(215)
2 952(211)
2 943(230)
2 952(270)
2 1509(276)
Share of state govern-ment controlled byDemocrats
031(012)
023(011)
030(012)
027(011)
026(011)
011(013)
Indicator is 1 if state con-trolled by one party
2 004(010)
2 004(007)
2 004(007)
2 001(009)
2 002(010)
017(009)
Average yield on bankloans in the stateminus Fed funds rate mdash
023(319) mdash mdash mdash mdash
Bank failure rate mdash mdash092
(111) mdash mdash mdash
Indicator is 1 if state hasunit banking law mdash mdash mdash
018(010)
019(010)
030(009)
Indicator is 1 if statechanges bank insur-ance powers mdash mdash mdash mdash
2 008(012)
2 026(013)
Includes regionalindicators No No No No No Yes
N 637 408 637 637 637 637Log likelihood 2 374 900 2 360 2 212 2 191 1192p-value of x 2 for
regression 001 001 001 001 001 001
The hazard model is Weibull where the dependent variable is the log expected time to branchingderegulation ln (T) 5 bx 1 eAll variables are measured for each state in each year Small bank asset share isthe percent of banking assets in the state held by banks below the median size of banks in each state in eachyear Relative capital ratio is the capital to assets ratio of small banks minus that of large banks Size ofinsurance relative to banking plus insurance in the state is measured as gross state product from insurancedivided by gross state product from insurance plus banking Insurance indicator is 1 if state law permits banksto enter the insurance business Small rm share is the percent of all establishments in the state that havefewer than twenty employees Party control variable is the share of the three bodies of state governmentcontrolled by Democrats One party control indicator is 1 if the same party controlrsquos the governorship and hasmajorities in both chambers of the state legislature Average yield on bank loans equals total interest incomeon all domestic loans made by banks in the state divided by total loans Bank failure rate is the ratio of assetsin failed banks during a given state and year divided by total bank assets in that state and year Unit bankingindicator equals 1 for states with unit banking restrictions The change in insurance powers indicator equalsone if the state changed the law to permit banks to sell insurance during the sample period We divide thecountry into four regions and include regional indicators in column (5) Robust standard errors are inparentheses and denote statistically signicant at the 1 5 and 10 percent levels
QUARTERLY JOURNAL OF ECONOMICS1452
The positive and statistically signicant coefficient on smallbank share implies that a greater small bank share delaysderegulation and this effect is economically important A one-standard-deviation increase in the small bank share results in a30 percent increase in the time until deregulation or about 47years This result supports the implications of private-interesttheory concerning intraindustry rivalry and is contrary to thepublic-interest theory
The positive and statistically signicant coefficient on therelative performance of small banks in the state implies thatstates deregulate later when small banks are relatively strongnancially25 A one-standard-deviation increase in the relativecapital-to-asset ratio results in a 15 percent rise in the time untilderegulation or about 24 years This result is consistent with theprivate-interest theory but may also be consistent with a public-interest interpretation because the risk of failures (and a costlytaxpayer-funded bailout) increases as small bank nancial healthdeclines
Interindustry competition also helps explain the timing ofderegulation In states where banks can sell insurance a rela-tively large insurance sector is associated with an increase in theexpected time to deregulation26 A one-standard-deviation in-crease in the relative size of the insurance sector in those statesthat permit banks to sell insurance leads to a 22 percent increasein the time until deregulation or about 35 years27 This result isconsistent with the private-interest theory but not the public-interest theory
Deregulation occurs earlier in states where small bank-dependent rms are relatively numerous A one-standard-
25 We also allowed small bank and large bank capital-to-asset ratios to enterseparately Their coefficient estimates were nearly equal in absolute value and ofopposite signs and so the data do not reject the relative small to large bank capitalratio specication we have chosen
26 When we use the relative size of the insurance sector in the state as awhole rather than insurance relative to banking plus insurance in the state weobtain similar results When we simultaneously include both measures of therelative size of insurance the interactions of the lsquolsquoinsurance relative to bankingplus insurancersquorsquo variable continue to be statisticallysignicant but the interactionsof lsquolsquoinsurance relative to the state as a wholersquorsquo variable are not
27 We have also estimated the model with the indicator variable equal to onefor states where banks may sell insurance but without the variables measuringtheimportance of insurance relative to banking In this model the coefficient on theinsurance indicator variablemdashan estimate of the difference in the expected time toderegulation in states where banks may sell insurance relative to states wherebanks may not sell insurancemdashequals 027 and is statistically signicant at the 5percent level This suggests that branching deregulation takes about 45 yearslonger in states where banks may sell insurance
WHAT DRIVES DEREGULATION 1453
deviation increase in the share of small rms reduces the timeuntil deregulation by 18 percent or about three years This resultconcerning the interests of consumers of banking services isconsistent with both the private- and public-interest theories
Column (2) of Table III introduces the average interest rateon loans to the model as a rough proxy for the cost to bankborrowers of these regulations These data become available onlyin 1976 so our sample size shrinks to 408 observations Thecoefficient on this variable is small and not statistically signi-cant The public-interest theory would imply that the coefficientshould be negative while the implication of the economic theorydoes not have a clear sign prediction
When we add the failure rate of banks to the specication incolumn (3) its coefficient is small and statistically insignicantWe do not nd a linkage between the timing of state deregulationand statewide banking distress Given that deposit insurance isfunded federally however bank failures may raise the publicrsquosconcerns about the costs of the regulations at the national ratherthan the state level This interpretation can explain why the rateof bank failure does not affect the timing of state-level deregula-tion even though bank instability does appear to rise generallyduring the period of increasing deregulation (see Section V below)
Finally the partisan structure of the state government alsoinuences when states deregulate As expected a higher propor-tion of Democrats in the government tends to delay deregulationA one-standard-deviation rise in the share of the governmentcontrolled by Democrats slows the deregulation by about twoyears Whether the state is dominated by one party however doesnot appear to affect the timing of the deregulation
To summarize our ndings from the hazard model privateinterests appear to play an important role in the deregulatoryprocess While private interests and public interests do sometimescoincide the results on the relative share of small banks and largebanks and on the relative size of insurance where banks competeare consistent with a private-interest approach but are difficult toexplain on public-interest grounds
B Robustness Tests
Omitted Variables States begin the sample period withdifferent degrees of restrictions on bank expansion These initialconditions may proxy for omitted factors determining the political-economy equilibrium in the state that could then affect the
QUARTERLY JOURNAL OF ECONOMICS1454
subsequent timing of deregulation For instance some statesbegan the period with unit banking while other states allowedlimited branching and the formation of multibank holdingcompanies
To test whether the stringency of initial restrictions onbranching matters we rst divide the sample into states with andwithout the most extreme form of branching regulation unitbanking28 Column (4) of Table III includes an indicator variableequal to one if the state began the period with unit bankingrestrictions The coefficient on this variable is positive andstatistically signicant but the other results remain virtuallyunchanged Next we include an indicator variable that is one ifthe state permitted multibank holding companies (MBHCs) toexist at the beginning of our sample period Thirty-four statesbegan the period allowing MBHCs The coefficient estimate forthis indicator variable (not reported) is small and not statisticallysignicant and its inclusion does not change the effects of theother variables
Eleven states expand bank powers to include the sale ofinsurance products during our sample period This could proxy foromitted factors about the relative strength of the insuranceindustry in the state In column (5) we include an indicatorvariable that is one if the state changes the law during our sampleperiod to permit banks to sell insurance business and zerootherwise The coefficient estimate is small and statisticallyinsignicant and does not affect the other results29
In column (6) we introduce four regional indicator variablesNorth South Midwest and West30 The coefficients on small bankshare the relative capital-to-assets ratio and small rm sharechange little Although the coefficient estimate still suggests that
28 We classify states that prohibited branching but permitted banks toestablish facilities as unit banking states The unit banking indicator is one for thefollowing states CO AR FL IL IA KS MN MO MT NE ND OK TX WI WVand WY
29 We conducted a likelihood ratio test of whether the unit banking statesthe MBHC states and the states that changed their regulations of bank insurancepowers could be pooled with the other states In all three cases we could not rejectpooling of the data
30 Since intrastate deregulation generally preceded interstate deregulationand the latter typically took the form of regional interstate compacts thepossibility of participating in a regional interstate banking compact could haveinuenced the decision to deregulate intrastate branching Our denitions of theregions are region 1 (South) contains AL AR DC FL GA KY LA MS NC OKSC TN TX and VA region 2 (Northeast) contains CT MA MD ME NH NJ NYPA RI VT and WV region 3 (Midwest) contains IA IL IN KS MI MN MO NEND OH SD and WI region 4 (West) contains the other states
WHAT DRIVES DEREGULATION 1455
in states where banks can sell insurance a larger relative size ofthe insurance sector delays deregulation the effect is no longerstatistically signicant The coefficient on the indicator variablefor whether a state expands bank powers to permit insurancesales during the sample period however now becomes statisti-cally signicantly negative This implies that deregulation occursroughly 41 years earlier in states where the insurance lobby maybe relatively weak While the inclusion of the regional indicatorsdoes not affect the other results the results on insurance doappear to be sensitive to the specication of the hazard model31
Estimation Technique To check the robustness of using ahazard model and to investigate whether the time-series or thecross-sectional variation in each of the factors accounts for thehazard results we estimate two simple linear probability models(neither are reported) First we run a pooled time-series cross-section OLS regression using the 637 observations from thehazard model The dependent variable equals one in the year ofderegulation and zero otherwise We include the same set ofregressors that were used in the hazard model and add state-levelxed effects This specication thus removes the cross-statevariation from the independent variables and focuses on theirmovements over time In this model the signs of the coefficientson the interest group factors are consistent with the results fromthe hazard model and the coefficients for the small bank relativecapital ratio and for the relative size of insurance where states cansell insurance are statistically signicant
In the second model we eliminate the time-series variationby averaging each of the explanatory variables over time We thushave one observation per state that had branching restrictions asof 1970 (N 5 39) The dependent variable equals the inverse of thenumber of years until deregulation which represents an estimateof the statersquos probability of deregulating in a year under theassumption that this probability is constant over time and eachyear provides an independent observation32 Although we haveonly 39 observations the coefficients on the small bank marketshare small bank relative capital ratio and small rm share
31 As further robustness check we also tried a population density variablewhich Abrams and Settle [1993] found relevant for regulatory change during the1920s and 1930s but its coefficient was not statistically signicant and did notaffect the other results
32 For instance the dependent variable would take the value of 01 for astate that deregulated in 1980 For the three states that did not deregulate thedependent variable equals 0
QUARTERLY JOURNAL OF ECONOMICS1456
variables are statistically signicant The signs of the effects ofthe insurance variables are the same as those reported in thehazard model but not statistically signicant
The hazard model and linear probability models thus producevery similar results The estimates from the latter suggest thatthe effects for the small bank market share and small rm sharevariables are driven primarily by cross-state variation the effectof the insurance variable is driven primarily by time-seriesvariation and the small bank capital effect is driven by bothcross-sectional and time-series variation
Ex Post Consequences of Deregulation Finally to check theplausibility of our interpretation of the hazard results we con-sider whether the ex post consequences of deregulation areconsistent with the ex ante lobbying positions we attribute to eachinterest group (see Kroszner and Strahan [1998] for details) Wend that small banks lose market share following deregulationand in states where banks can enter the insurance business theinsurance sector shrinks relative to the banking sector followingderegulation Borrowers also benet from deregulation throughlower average interest rates on loans These ndings support theprivate-interest interpretation of the coefficients in the hazardmodel groups that will benet lobby to speed deregulation andthose that will be harmed lobby to slow it
C Voting on Interstate Branching Deregulation in the U SHouse of Representatives
We now examine whether the forces driving intrastate branch-ing deregulation also drive interstate deregulation Financialservices interests are active contributors and lobbyists in Washing-ton Their political action committees constitute the largest groupof contributors to legislators providing nearly 20 percent of totalcongressional campaign contributions [Makinson 1992] and muchof their lobbying effort involves competition among rival interestswithin nancial services (see Kroszner and Stratmann [1998])
As noted above after virtually all states adopted intra- andinterstate branching deregulation the 1994 Riegle-Neal Act re-pealed the 1927 McFadden Act to eliminate all barriers tointerstate banking and branching by 1997 The key votes concern-ing the Riegle-Neal Act were either voice votes or extremelylopsided so we could not estimate a voting model from them Anumber of bills and amendments related to interstate branching
WHAT DRIVES DEREGULATION 1457
had been debated in Congress during the years prior to thepassage of the Riegle-Neal Act but a search of the weekly BNABanking Reporter and the Congressional Record produced onlyone roll-call vote related to interstate branching that was notlopsided This vote occurred in the House on November 14 1991and concerned an amendment sponsored by Wylie (R-OH) andNeal (D-NC) to introduce interstate banking and branchingderegulation to a nancial services reform package33 Althoughthe amendment passed by 210 to 208 the bill to which it wasattached subsequently was defeated
To check for the inuence of the factors we considered in thestate-level reforms we examine both the sponsorship of interstatebanking legislation and voting on the amendment The sponsorsof the Wylie-Neal amendment are from states with low small bankshares 004 in Ohio (Wylie) and 002 in North Carolina (Neal) Incontrast the sample mean in 1991 is 008 (median 5 007)Michigan home state of the Senatersquos sponsor of the 1994 Riegle-Neal Act also had relatively low small bank strength (small bankshare of 005)
Table IV reports estimates of a probit model where thedependent variable equals one if the legislator voted in favor ofthe amendment and zero otherwise The explanatory variablesare the same as those in the hazard model except that we use aparty affiliation indicator variable equal to one for Democrats andzero for Republicans as our proxy for political factors Thecoefficients in Table IV are the marginal effects (lsquolsquoslopesrsquorsquo) of aone-unit change of each variable on the probability that alegislator will vote for the amendment34 Note that a forcefavoring deregulation has a negative coefficient in the hazardmodel (Table III) but a positive coefficient in the probit model
Consistent with the state-level deregulation process legisla-tors are more likely to support the amendment if their states havea relatively low share of small banks As in the hazard model thefraction of small banks is the most important interest groupinuence on a legislatorrsquos voting decision A one-standard-deviation increase in small banksrsquo market share (from the mean)is associated with a decline in the probability of voting in favor of
33 The Wylie-Neal amendment also included provisions limiting certaininsurance and real estate powers of national banks [Congressional RecordNovember 14 1991 pp 10239ndash10242]
34 Since we have multiple legislators from each state we adjust the standarderrors to correct for the lack of independence among observations clustered in thesame state
QUARTERLY JOURNAL OF ECONOMICS1458
TAB
LE
IVM
AR
GIN
AL
EF
FE
CT
SF
RO
MA
PR
OB
ITM
OD
EL
OF
TH
EIN
FL
UE
NC
EO
FP
OL
ITIC
AL
EC
ON
OM
YF
AC
TO
RS
ON
TH
EH
OU
SE
VO
TE
ON
TH
EW
YL
IE-N
EA
L
AM
EN
DM
EN
TT
OP
ER
MIT
INT
ER
ST
AT
EB
AN
KB
RA
NC
HIN
G
NO
VE
MB
ER
14
1991
(1)
(2)
(3)
(4)
(5)
(6)
Smal
lban
kas
set
shar
eof
allb
anki
ng
asse
tsin
stat
e2
417
(12
4)2
454
(12
2)2
407
(12
0)2
443
(14
7)2
450
(14
9)2
490
(14
7)C
apit
alra
tio
ofsm
allb
anks
rela
tive
tola
rge
inth
est
ate
180
(37
5)2
68(3
42)
256
(37
5)1
69(3
78)
153
(38
0)2
45(4
25)
Rel
ativ
esi
zeof
insu
ranc
ein
stat
esw
here
bank
sm
ayse
llin
sura
nce
0ot
herw
ise
22
34
(10
2)2
249
(1
14)
22
33
(10
1)2
249
(1
04)
22
51
(10
2)2
154
(11
1)In
dica
tor
is1
ifba
nks
may
sell
insu
ranc
ein
the
stat
e0
79
(0
06)
074
(00
6)0
69
(0
06)
070
(00
6)0
70
(0
06)
067
(00
7)R
elat
ive
size
ofin
sura
nce
inst
ates
whe
reba
nks
may
not
sell
insu
ranc
e0
oth
erw
ise
180
(06
6)2
42
(0
72)
154
(0
67)
173
(06
6)1
72
(0
67)
203
(0
91)
Smal
lrm
shar
eof
the
num
ber
ofr
ms
inth
est
ate
21
23(0
92)
20
83(0
89)
21
62(1
07)
21
32(0
95)
21
30(1
00)
20
94(1
16)
Indi
cato
ris
1if
Rep
rese
ntat
ive
isa
Dem
ocra
t2
035
(00
7)2
035
(00
7)2
036
(00
7)2
035
(00
7)2
035
(00
7)2
036
(00
7)A
vera
geyi
eld
onba
nklo
ans
inth
est
ate
min
usF
edfu
nds
rate
mdash10
01
(51
5)mdash
mdashmdash
mdash
Ave
rage
fail
ure
rate
198
5to
1991
mdashmdash
228
(14
1)mdash
mdashmdash
Indi
cato
ris
1if
stat
eh
asun
itba
nkin
gla
wmdash
mdashmdash
004
(00
7)0
04(0
07)
20
01(0
08)
Indi
cato
ris
1if
stat
ech
ange
sba
nk
insu
ran
cepo
wer
smdash
mdashmdash
mdash2
001
(00
7)2
003
(00
7)In
clud
esre
gion
alin
dica
tors
N
oN
oN
oN
oN
oYe
sP
seud
o-R
20
170
180
170
170
170
18L
ogli
keli
hood
224
024
223
826
223
951
224
013
224
010
223
808
p-va
lue
ofx2
for
regr
essi
on
001
0
01
001
0
01
001
0
01
Th
ede
pen
den
tvar
iabl
eis
one
ifth
ele
gisl
ator
vote
sfo
rth
eam
endm
enta
nd
zero
ifag
ains
tT
he
repo
rted
coef
fici
ents
are
the
effe
cts
ofa
uni
tch
ange
ofth
ein
depe
nde
ntva
riab
le(f
rom
the
mea
n)on
the
prob
abili
tyof
voti
ng
infa
vor
ofth
eam
endm
ent
All
vari
able
sar
em
easu
red
byst
ate
N5
418
the
num
ber
ofR
epre
sent
ativ
esvo
tin
gon
the
amen
dmen
tS
mal
lban
kas
sets
har
eis
the
perc
ent
ofba
nki
ng
asse
tsin
the
stat
eh
eld
byba
nks
belo
wth
em
edia
nsi
zeof
ban
kin
each
stat
ein
each
year
Rel
ativ
eca
pita
lrat
iois
the
capi
talt
oas
sets
rati
oof
smal
lban
ksm
inu
sth
atof
larg
eba
nks
Siz
eof
insu
ran
cere
lati
veto
bank
ing
plu
sin
sura
nce
inth
est
ate
ism
easu
red
asgr
oss
stat
epr
oduc
tfro
min
sura
nce
divi
ded
bygr
oss
stat
epr
odu
ctfr
omin
sura
nce
plu
sba
nkin
gIn
dica
tor
ison
eif
stat
ela
wpe
rmit
sba
nks
toen
ter
the
insu
ran
cebu
sin
ess
Sm
all
rmsh
are
isth
epe
rcen
tofa
lles
tabl
ishm
ents
inth
est
ate
that
hav
efe
wer
than
twen
tyem
ploy
ees
Par
tyin
dica
tor
ison
eif
the
Rep
rese
nta
tive
isa
Dem
ocra
tA
vera
geyi
eld
onba
nklo
ans
equa
lsto
tali
nter
est
inco
me
onal
ldom
esti
clo
ans
mad
eby
ban
ksin
the
stat
edi
vide
dby
tota
lloa
ns
The
aver
age
bank
failu
rera
teis
the
rati
oof
asse
tsin
faile
dba
nks
duri
ng
agi
ven
stat
edi
vide
dby
tota
lban
kas
sets
inth
atst
ate
aver
aged
over
the
year
s19
85to
1991
Un
itba
nki
ng
indi
cato
req
ual
son
efo
rst
ates
wit
hun
itba
nkin
gre
stri
ctio
nsT
hech
ange
inin
sura
nce
pow
ers
indi
cato
req
ual
son
eif
the
stat
ech
ange
dth
ela
wto
perm
itba
nks
tose
llin
sura
nce
duri
ngth
esa
mpl
epe
riod
We
divi
deth
eco
unt
ryin
tofo
urre
gion
san
din
clu
dere
gion
alin
dica
tors
inco
lum
n(5
)R
obu
stst
anda
rder
rors
are
inpa
ren
thes
es
an
d
den
ote
stat
isti
cally
sign
ica
nta
tth
e1
5an
d10
perc
entl
evel
s
WHAT DRIVES DEREGULATION 1459
branching of approximately 17 percent The relative capital-to-assets ratio of small and large banks however does not have astatistically signicant effect in the probit model
The impact of rival interests outside of banking is alsoconsistent with intrastate deregulation results Where banks cansell insurance legislators from states with larger insurancesectors relative to banking are less likely to vote for the amend-ment35 A one-standard-deviation increase in the relative size ofinsurance in those states which permit banks to sell insurancedecreases the probability that a legislator will favor the amend-ment by about 13 percent Turning to consumer interests thecoefficient on small rm share is not statistically signicant butthe coefficient on interest rates in the state is A one-standard-deviation increase in the average interest rate on loans raises theprobability that a legislator will support the amendment byroughly 8 percent The coefficient estimate for the bankingdistress variable measured as the average bank failure rate from1985 to 1991 is small and statistically signicant Overall theprobit analysis of the vote on national branching deregulationsupports the private-interest theory of deregulation and providesa consistency check that the importance of interests operating onthe state legislatures are very similar to those operating at thefederal level
V WHY DID BRANCHING DEREGULATION BEGIN DURING THE
1970S SHOCKS TO THE EQUILIBRIUM SUPPORTING BANK
BRANCHING RESTRICTIONS
A complete explanation of regulatory exit should account forwhy deregulation begins in the 1970s as well as the timing of thestate-by-state reforms [Kane 1996] In this section we describebroad technological legal and economic shocks that altered thepolitical-economy equilibrium which had kept antibranching regu-lations little changed for at least 30 years
Beginning in the 1970s three innovations reduced the valueto the protected banks of local geographic monopolies (see TableV) First automatic teller machines (ATMs) helped to erode thegeographic ties between customers and banks Second checkablemoney market mutual funds and the Merrill Lynch Cash Manage-ment Account demonstrated that banking by mail and telephone
35 The positive and statistically signicant estimates on the other insurancevariables may reect insurance industry support for the amendmentrsquos provisionslimiting banksrsquo insurance powers
QUARTERLY JOURNAL OF ECONOMICS1460
provided a convenient alternative to local banks36 Third techno-logical innovation and deregulation reduced transportation andcommunication costs particularly since the 1970s thereby lower-ing the costs for customers to use distant banks By increasing theelasticity of deposits supplied to banks these innovations reducedthe value of geographical restrictions to their traditional benecia-ries and thereby reduced their incentive to ght to maintain them[Peltzman 1976]
On the lending side increasing sophistication of credit-scoring techniques following innovations in information process-ing nancial theory and the development of large credit databases began to diminish the value of knowledge that localbankers had about the risks of borrowers in the community As a
36 Regulation Q which limited the interest rates that banks could pay ondeposits may have helped to drive depositors away from banks when the gapbetween market rates and deposit ceilings grew during the 1970s The eliminationof interest rate ceilings on large denomination certicates of deposit during the1970s appears to have hurt smaller and retail-oriented banks relative to largerwholesale banks [James 1983]
TABLE VBROAD TRENDS IN COMMERCIAL BANKING 1950ndash1995
YearNumberof ATMs
Domesticbank
deposits(billions)
Moneymarketmutual
fund(billions)
Percent ofdeposits 1
moneyfunds heldby banks
Smallbanksrsquo
percent ofbankingassets
Averagenumber of
bankfailures
(1) (2) (3) (4) (5) (6)1950 0 $ 154 $ 0 100 NA 41955 0 191 0 100 NA 31960 0 228 0 100 24 21965 0 330 0 100 20 41970 0 479 0 100 18 61975 9750 775 4 99 18 61980 18500 1182 76 94 17 101985 61117 1787 242 88 14 601990 80156 2339 493 83 11 1791995 122706 2552 745 77 8 61
Column (1) ATM gures are from Bank Network News The EFT Network Data Book (New YorkFaulkner and Gray Inc) The 1975 gure was unavailable 9750 is the number of ATMs in 1978 the rst yearfor which complete data are available Columns (2)ndash(4) Banksrsquo domestic deposits are from the Reports ofIncome and Condition money market mutual funds are from the Flow of Funds Data on all bank depositsforeign plus domestic are only available beginning in 1970 The trend in banksrsquo share (column (4)) is the sameusing total deposits instead of domestic deposits Column (5) Percent of banking assets held by small bankswhere a small bank is dened as a commercial bank with less than $100 million in assets in 1994 dollarsThese data are based on the Reports of Income and Condition Data on small banks are not available before1960 Column (6) Five-year average number of bank failures where the nal year is indicated in the rstcolumn These data are from FDIC Annual Report and the Quarterly Banking Prole
WHAT DRIVES DEREGULATION 1461
result of these innovations a national market developed forresidential mortgages credit card receivables have been securi-tized and bank lending to small business now relies less on thejudgment of loan officers and more on standardized scoringmodels
These changes have increased the potential protability forlarge banks to enter what had been the core of small bankactivities Large banksrsquo incentive to increase their lobbying pres-sure to be able to expand into these markets has thus beenincreasing over time In fact small banksrsquo market share began todecline even prior to the branching deregulation (Table V) As thevalue of a local banking relationships declined small rms thatwere the main borrowers from the small banks also probablybecame more likely to favor the entry of large banks into localmarkets With the deadweight costs of preventing large bankentry rising the private-interest theory predicts that small localbanks would become less likely able to maintain the branchingrestrictions [Becker 1983] Deregulation that reduces deadweightcosts of regulation is also consistent with the public-interesttheory
Kane [1996] argues that another major shock to the oldequilibrium was the increasing public awareness of the costlinessof having government-insured but (geographically) undiversiednancial institutions In the late 1970s the failure rate of banksbegan to rise (Table V) In the 1980s the Savings and Loan crisisand taxpayer bailout further heightened public awareness aboutthe costs of restrictions that make depository institutions morelikely to require infusions of taxpayer funds The failures thusmay have heightened public support for branching deregulation(see also Abrams and Settle [1993]) As noted above we did notnd any evidence that banking failures or distress in a stateaffected the speed with which the state deregulated but distressstill may be a common factor affecting all states
These technological economic and legal shocks generatedconditions that changed the long-standing balance favoring theantibranching forces The marginal value of lobbying to repealbranching restrictions increased just as the relative value to thesmall banks of maintaining branching restrictions was decliningThese nationwide shocks are common factors across states andthey are consistent with the positive duration dependence associ-ated with state-level deregulation that we demonstrated in FigureII In addition a broad change in lsquolsquoideologyrsquorsquo against government
QUARTERLY JOURNAL OF ECONOMICS1462
regulation could explain the positive duration dependence duringthe sample period
VI CONCLUSIONS
The private-interest theory of regulation can account for thepattern of bank branching deregulation of the 1970s and 1980sBeneciaries of branching regulation had supported a coalitionfavoring geographical restrictions despite their costs to consum-ers of nancial services Innovations that began in the 1970saltered the value of the restrictions to the affected parties and theresulting competition among interest groups can explain thesubsequent deregulation While some of our results are alsoconsistent with the public interest theorymdashfor example deregula-tion occurs earlier when small banks are in a relatively weaknancial positionmdashother results particularly evidence on theimportance of rivalries between small and large banks andbetween banking and insurance are difficult to explain with thepublic-interest approach We also nd that political-institutionalfactors affect deregulation although these variables may act asproxies for unmeasured economic interests The greater success ofthe private-interest theory to explain deregulation here than inprevious work may be due to the richer cross-sectional andtime-series variation in branching deregulation that can beexploited with a hazard model Future empirical studies ofendogenous deregulation thus may be most fruitful where changehas occurred across states over time such as in franchisinginsurance and public utilities (eg Joskow [1996])
Technological and nancial innovations will continue to erodethe benets to any interest group of maintaining regulatorybarriers in nancial services These forces are likely to bringabout reforms both domestically for example through legislationthat would increase bank powers (see Kroszner [1996] Krosznerand Rajan [1994 1997] and Kroszner and Stratmann [1998]) andinternationally for example through the extension of nancialservices provisions of NAFTAto reduce geographic barriers acrosscountries [Kroszner 1997]
APPENDIX EFFECTS OF DIFFERENT TYPES OF BRANCHING
DEREGULATION ON ECONOMIC GROWTH BANKING STRUCTUREAND BANK EFFICIENCY
We measure the importance of alternative forms of deregula-tion by their consequences for the economic growth and the
WHAT DRIVES DEREGULATION 1463
structure and efficiency of the banking sector across the states(see Jayaratne and Strahan [1996 1998]) To do so we run anOLS panel regression relating indicators for each type of deregula-tion to potential consequences Our dependent variables includeproxies for economic growth (growth in personal income) bankstructure (the log of the number of banking companies and thenumber of branches per banking company) and efficiency (the logof total noninterest expenses and the ratio of loan loss provisionsto total loans) The four deregulation indicators (MBHC branch-ing through merger and acquisition unrestricted statewide branch-ing and interstate banking) are equal to one in the years afterderegulation and zero before If a state deregulated before thesample period begins in 1970 the indicator is one throughout ifthe state does not deregulate then the indicator is zero through-out All regressions include state xed effects and time effects
Table VI shows the results of including the set of four
TABLE VIMULTIVARIATE REGRESSIONS OF THE EFFECTS OF DIFFERENT TYPES OF BRANCHING
DEREGULATION ON ECONOMIC GROWTH BANKING STRUCTUREAND BANK EFFICIENCY
Dependent variable
Growthin
personalincome
Log of of
bankingcompanies
Branchesper
bankingcompany
Log ofnon-
interestexpenses
Loan lossprovisions
(loans 1leases)
Merger amp acquisitionbranchingindicator
0013(0004)
2 0162(0037)
1146(0453)
2 0107(0035)
2 0007(0001)
Unrestricted state-wide branchingindicator
2 0008(0005)
0086(0059)
2 0043(0574)
0070(0037)
0005(0002)
Interstate bankingindicator
0006(0005)
2 0052(0057)
1064(0511)
0005(0030)
2 0001(0001)
Multibank holdingcompany indicator
2 0005(0006)
2 0096(0052)
2 0364(0529)
2 0095(0036)
2 0001(0001)
Dependent variablemean [median]
0086[0082]
4881[5136]
6649[5055]
13812[13803]
0009[0006]
This table contains regressions of state-level economic growth and state-level measures of the structureand efficiency of the banking industry Each dependent variable is regressed on a set of state indicatorvariables a set of year indicator variables and the set of four deregulation indicators these deregulationindicators are equal to one after deregulation and zero before If a state was always regulated the indicator iszero throughout if it was always deregulated the indicator is one throughout Sample period is 1975 to 1991and Delaware and South Dakota are dropped so N 5 833 Noninterest expenses become available only in1984 so N 5 392 The coefficient on the deregulation indicators is reported with robust standard errors inparentheses and denote statistically signicant at the 1 5 and 10 percent levels
QUARTERLY JOURNAL OF ECONOMICS1464
deregulation indicators simultaneously The branching by mergerand acquisition indicator is statistically signicant and of theexpected sign in all ve of the equations and larger in absolutevalue than any of the other coefficients on the indicators for theother types of deregulation Since there may be multicollinearityamong our indicators we also ran univariate specications whichincluded each indicator separately and the results are verysimilar These results imply that branching by merger andacquisition is the most important type of deregulation measuredby its consequences for economic growth banking structure andbank efficiency See Kroszner and Strahan [1998] for more details
UNIVERSITY OF CHICAGO GRADUATE SCHOOL OF BUSINESS AND NATIONAL BUREAU OF
ECONOMIC RESEARCH
FEDERAL RESERVE BANK OF NEW YORK RESEARCH amp MARKET ANALYSIS GROUP
REFERENCES
Abrams Burton A and Russell F Settle lsquolsquoPressure-Group Inuence and Institu-tional Change Branch-Banking Legislation during the Great DepressionrsquorsquoPublic Choice LXXVII (1993) 687ndash705
Amel Dean lsquolsquoState Laws Affecting the Geographic Expansion of CommercialBanksrsquorsquo unpublished manuscript Board of Governors of the Federal ReserveSystem 1993
Becker Gary S lsquolsquoA Theory of Competition among Pressure Groups for PoliticalInuencersquorsquo Quarterly Journal of Economics XCVIII (1983) 371ndash400
Berger Allen Anil Kashyap Joseph Scalise lsquolsquoThe Transformation of the U SBanking Industry What a Long Strange Trip Itrsquos Beenrsquorsquo Brookings Papers onEconomic Activity 2 (1995) 55ndash218
Berger Allen Anthony Saunders Joseph Scalise and Gregory Udell lsquolsquoThe Effectsof Bank Mergers and Acquisitions on Small Business Lendingrsquorsquo Journal ofFinancial Economics L (1998) 187ndash229
Brennan Geoffrey and James Buchanan lsquolsquoTowards a Tax Constitution forLeviathanrsquorsquo Journal of Public Economics VIII (1977) 255ndash273
Brickley James A and Christopher M James lsquolsquoThe Takeover Market CorporateBoard Composition and Ownership Structure The Case of Bankingrsquorsquo Journalof Law and Economics XXX (1987) 161ndash180
Cole Rebel and John Wolken lsquolsquoFinancial Services Used by Small BusinessesEvidence from the 1993 Survey of Small Business Financesrsquorsquo Federal ReserveBulletin LXXXI (1994) 629ndash676
Dixit Avinash The Making of Economic Policy A Transactions-Cost PoliticsPerspective (Cambridge MA MIT Press 1996)
Economides Nicolas R Glenn Hubbard and Darius Palia lsquolsquoThe Political Economyof Branching Restrictions and Deposit Insurancersquorsquo Journal of Law andEconomics XXIX (1996) 667ndash704
Flannery Mark lsquolsquoThe Social Costs of Unit Banking Restrictionsrsquorsquo Journal ofMonetary Economics XIII (1984) 237ndash249
Greene William H Econometric Analysis third edition (Upper Saddle River NJPrentice-Hall 1997)
Gunther Jeffery lsquolsquoRegional Capital Imbalances and the Removal of InterstateBanking Restrictionsrsquorsquo Economics Letters XLIV (1994) 439ndash442 lsquolsquoThe Erosion of Intrastate Branch Banking Restrictionsrsquorsquo unpublishedmanuscript Federal Reserve Bank of Dallas 1996
Irwin Douglas and Randall Kroszner lsquolsquoInterests Institutions and Ideology inSecuring Policy Change The Republican Conversion to Trade Liberalizationafter Smoot-Hawleyrsquorsquo Journal of Law and Economics XLII (1999) 643ndash673
WHAT DRIVES DEREGULATION 1465
James Christopher lsquolsquoAn Analysis of Intra-Industry Differences in the Effect ofRegulation The Case of Deposit Rate Ceilingsrsquorsquo Journal of Monetary Econom-ics XII (1983) 417ndash442
Jarrell Gregg lsquolsquoChange at the Exchange The Causes and Effects of DeregulationrsquorsquoJournal of Law and Economics XXVII (1984) 273ndash312
Jayaratne Jith and Philip E Strahan lsquolsquoThe Finance-Growth Nexus Evidencefrom Bank Branch Deregulationrsquorsquo Quarterly Journal of Economics CI (1996)639ndash670
Jayaratne Jith and Philip E Strahan lsquolsquoEntry Restrictions Industry Evolutionand Dynamic Efficiency Evidence from Commercial Bankingrsquorsquo Journal of Lawand Economics XLIX (1998) 239ndash274
Joskow Paul lsquolsquoComment on lsquoPower Struggles Explaining Deregulatory Reformsin Electricity Marketsrsquo rsquorsquo Brookings Papers on Economic Activity Microeconom-ics (1996) 251ndash264
Joskow Paul and Roger Noll lsquolsquoRegulation in Theory and Practice An Overviewrsquorsquoin Studies in Public Regulation Gary Fromm ed (Cambridge MA MITPress 1981) pp 1ndash65
Kahn Matthew E and John G Matsusaka lsquolsquoDemand for Environmental GoodsEvidence from Voting Patterns on California Initiativesrsquorsquo Journal of Law andEconomics XL (1997) 137ndash173
Kalbeisch J and R Prentice The Statistical Analysis of Failure Time Data (NewYork John Wiley amp Sons 1980)
Kane Edward J lsquolsquoDe Jure Interstate Banking Why Only NowrsquorsquoJournal of MoneyCredit and Banking XXVIII (1996) 141ndash161
Kiefer Nicholas lsquolsquoEconomic Duration Data and Hazard Functionsrsquorsquo Journal ofEconomic Literature XXVI (1988) 646ndash679
Kroszner Randall S lsquolsquoThe Evolution of Universal Banking and Its Regulation inTwentieth Century Americarsquorsquo in Universal Banking Financial System DesignReconsidered Anthony Saunders and Ingo Walter eds (New York IrwinProfessional Publishers 1996) pp 70ndash99 lsquolsquoThe Political Economy of Banking and Financial Regulation in the U Srsquorsquo inThe Banking and Financial Structure in the NAFTA Countries and ChileGeorge M von Furstenberg ed (Boston Kluwer Academic Publishers 1997)pp 200ndash213
Kroszner Randall S and Raghuram G Rajan lsquolsquoIs the Glass-Steagall ActJustied A Study of the U S Experience with Universal Banking before1933rsquorsquo American Economic Review LXXXIV (1994) 810ndash832
Kroszner Randall S and Raghuram G Rajan lsquolsquoOrganization Structure andCredibility Evidence from Bank Securities Activities before the Glass-Steagall Actrsquorsquo Journal of Monetary Economics XXXIX (1997) 475ndash516
Kroszner Randall S and Philip E Strahan lsquolsquoRegulatory Incentives and the ThriftCrisis Dividends Mutual-to-Stock Conversions and Financial DistressrsquorsquoJournal of Finance LI (1996) 1285ndash1320
Kroszner Randall S and Philip E Strahan lsquolsquoWhat Drives Deregulation Econom-ics and Politics of the Relaxation of Bank Branching Restrictionsrsquorsquo NBERWorking Paper No 6637 1998
Kroszner Randall S and Thomas Stratmann lsquolsquoInterest Group Competition andthe Organization of Congress Theory and Evidence from Financial ServicesPolitical Action Committeesrsquorsquo American Economic Review LXXXVIII (1998)pp 1163ndash1187
Kroszner Randall S and Thomas Stratmann lsquolsquoDoes Political Ambiguity PayCorporate Campaign Contributions and Rewards to Legislator Reputationrsquorsquounpublishedmanuscript University of Chicago Graduate School of Business1999
Lin D Y and L J Wei lsquolsquoThe Robust Inference for the Cox Proportional HazardsModel Journal of the American Statistical Association LXXXIV (1989)1074ndash1078
Macey Jonathan R and Geoffrey P Miller Banking Law and Regulation (BostonLittle Brown and Company 1992)
Makinson Larry Open Secrets The Encyclopedia of Congressional Money andPolitics (Washington DC Congressional Quarterly 1992)
Niskanen WilliamBureaucracy and Representative Government (Chicago Aldine-Atherton 1971)
QUARTERLY JOURNAL OF ECONOMICS1466
Noll Roger lsquolsquoComment on Peltzmanrsquorsquo Brookings Papers Microeconomics SpecialIssue (1989) 48ndash58
North Douglass lsquolsquoA Transactions Cost Theory of Politicsrsquorsquo Journal of TheoreticalPolitics II (1990) 355ndash367
Peltzman Sam lsquolsquoToward a More General Theory of Regulationrsquorsquo Journal of Lawand Economics XIX (1976) 109ndash148 lsquolsquoConstituent Interest and Congressional Votingrsquorsquo Journal of Law andEconomics XXVII (1984) 181ndash210 lsquolsquoThe Economic Theory of Regulation after a Decade of DeregulationrsquorsquoBrookings Papers Microeconomics Special Issue (1989) 1ndash41
Peterson Mitchell and Raghuram Rajan lsquolsquoThe Benets of Lending RelationshipsEvidence from Small Business Datarsquorsquo Journal of Finance XLIX (1994) 3ndash38
Poole Keith and Howard Rosenthal Congress A Political-Economic History ofRoll Call Voting (Oxford Oxford University Press 1997)
Stata Stata Reference Manual Release 5 (College Station TX Stata Press 1997)Stigler George J lsquolsquoThe Theory of Economic Regulationrsquorsquo Bell Journal of Economics
and Management Science II (1971) 3ndash21Stigler George J Chicago Studies in the Political Economy (Chicago University
of Chicago Press 1988)Strahan Philip E and James Weston lsquolsquoSmall Business Lending and the Changing
Structure of the Banking Industryrsquorsquo Journal of Banking and Finance XXII(1998) 821ndash845
Sylla Richard John Legler and John Wallis lsquolsquoBanks and State Public Finance inthe New Republic The United States 1790ndash1860rsquorsquo Journal of EconomicHistory XLVII (1987) 391ndash403
White Eugene The Regulation and Reform of the American Banking System1900ndash1929 (Princeton Princeton University Press 1983)
Winston Clifford lsquolsquoEconomic Deregulation Days of Reckoning for Microecono-mistsrsquorsquo Journal of Economic Literature XXXI (1993) 1263ndash1289
WHAT DRIVES DEREGULATION 1467
TABLE IIIHAZARD MODEL OF POLITICAL ECONOMY FACTORS AFFECTING THE TIMING OF STATE
BRANCHING DEREGULATION 1970ndash1992
(1) (2) (3) (4) (5) (6)Small bank asset share of
all banking assets instate
648(122)
519(137)
632(135)
516(118)
488(133)
445(088)
Capital ratio of smallbanks relative to largein the state
1325(348)
921(319)
1301(362)
1067(346)
1144(375)
906(381)
Relative size of insurancein states where banksmay sell insurance 0otherwise
324(161)
259(149)
315(165)
285(154)
272(159)
055(134)
Indicator is 1 if banksmay sell insurance inthe state
045(010)
037(011)
045(010)
038(008)
036(009)
001(012)
Relative size of insurancein states where banksmay not sell insurance0 otherwise
2 093(043)
2 090(033)
2 102(051)
2 093(048)
2 096(053)
2 004(039)
Small rm share of thenumber of rms in thestate
2 972(211)
2 614(215)
2 952(211)
2 943(230)
2 952(270)
2 1509(276)
Share of state govern-ment controlled byDemocrats
031(012)
023(011)
030(012)
027(011)
026(011)
011(013)
Indicator is 1 if state con-trolled by one party
2 004(010)
2 004(007)
2 004(007)
2 001(009)
2 002(010)
017(009)
Average yield on bankloans in the stateminus Fed funds rate mdash
023(319) mdash mdash mdash mdash
Bank failure rate mdash mdash092
(111) mdash mdash mdash
Indicator is 1 if state hasunit banking law mdash mdash mdash
018(010)
019(010)
030(009)
Indicator is 1 if statechanges bank insur-ance powers mdash mdash mdash mdash
2 008(012)
2 026(013)
Includes regionalindicators No No No No No Yes
N 637 408 637 637 637 637Log likelihood 2 374 900 2 360 2 212 2 191 1192p-value of x 2 for
regression 001 001 001 001 001 001
The hazard model is Weibull where the dependent variable is the log expected time to branchingderegulation ln (T) 5 bx 1 eAll variables are measured for each state in each year Small bank asset share isthe percent of banking assets in the state held by banks below the median size of banks in each state in eachyear Relative capital ratio is the capital to assets ratio of small banks minus that of large banks Size ofinsurance relative to banking plus insurance in the state is measured as gross state product from insurancedivided by gross state product from insurance plus banking Insurance indicator is 1 if state law permits banksto enter the insurance business Small rm share is the percent of all establishments in the state that havefewer than twenty employees Party control variable is the share of the three bodies of state governmentcontrolled by Democrats One party control indicator is 1 if the same party controlrsquos the governorship and hasmajorities in both chambers of the state legislature Average yield on bank loans equals total interest incomeon all domestic loans made by banks in the state divided by total loans Bank failure rate is the ratio of assetsin failed banks during a given state and year divided by total bank assets in that state and year Unit bankingindicator equals 1 for states with unit banking restrictions The change in insurance powers indicator equalsone if the state changed the law to permit banks to sell insurance during the sample period We divide thecountry into four regions and include regional indicators in column (5) Robust standard errors are inparentheses and denote statistically signicant at the 1 5 and 10 percent levels
QUARTERLY JOURNAL OF ECONOMICS1452
The positive and statistically signicant coefficient on smallbank share implies that a greater small bank share delaysderegulation and this effect is economically important A one-standard-deviation increase in the small bank share results in a30 percent increase in the time until deregulation or about 47years This result supports the implications of private-interesttheory concerning intraindustry rivalry and is contrary to thepublic-interest theory
The positive and statistically signicant coefficient on therelative performance of small banks in the state implies thatstates deregulate later when small banks are relatively strongnancially25 A one-standard-deviation increase in the relativecapital-to-asset ratio results in a 15 percent rise in the time untilderegulation or about 24 years This result is consistent with theprivate-interest theory but may also be consistent with a public-interest interpretation because the risk of failures (and a costlytaxpayer-funded bailout) increases as small bank nancial healthdeclines
Interindustry competition also helps explain the timing ofderegulation In states where banks can sell insurance a rela-tively large insurance sector is associated with an increase in theexpected time to deregulation26 A one-standard-deviation in-crease in the relative size of the insurance sector in those statesthat permit banks to sell insurance leads to a 22 percent increasein the time until deregulation or about 35 years27 This result isconsistent with the private-interest theory but not the public-interest theory
Deregulation occurs earlier in states where small bank-dependent rms are relatively numerous A one-standard-
25 We also allowed small bank and large bank capital-to-asset ratios to enterseparately Their coefficient estimates were nearly equal in absolute value and ofopposite signs and so the data do not reject the relative small to large bank capitalratio specication we have chosen
26 When we use the relative size of the insurance sector in the state as awhole rather than insurance relative to banking plus insurance in the state weobtain similar results When we simultaneously include both measures of therelative size of insurance the interactions of the lsquolsquoinsurance relative to bankingplus insurancersquorsquo variable continue to be statisticallysignicant but the interactionsof lsquolsquoinsurance relative to the state as a wholersquorsquo variable are not
27 We have also estimated the model with the indicator variable equal to onefor states where banks may sell insurance but without the variables measuringtheimportance of insurance relative to banking In this model the coefficient on theinsurance indicator variablemdashan estimate of the difference in the expected time toderegulation in states where banks may sell insurance relative to states wherebanks may not sell insurancemdashequals 027 and is statistically signicant at the 5percent level This suggests that branching deregulation takes about 45 yearslonger in states where banks may sell insurance
WHAT DRIVES DEREGULATION 1453
deviation increase in the share of small rms reduces the timeuntil deregulation by 18 percent or about three years This resultconcerning the interests of consumers of banking services isconsistent with both the private- and public-interest theories
Column (2) of Table III introduces the average interest rateon loans to the model as a rough proxy for the cost to bankborrowers of these regulations These data become available onlyin 1976 so our sample size shrinks to 408 observations Thecoefficient on this variable is small and not statistically signi-cant The public-interest theory would imply that the coefficientshould be negative while the implication of the economic theorydoes not have a clear sign prediction
When we add the failure rate of banks to the specication incolumn (3) its coefficient is small and statistically insignicantWe do not nd a linkage between the timing of state deregulationand statewide banking distress Given that deposit insurance isfunded federally however bank failures may raise the publicrsquosconcerns about the costs of the regulations at the national ratherthan the state level This interpretation can explain why the rateof bank failure does not affect the timing of state-level deregula-tion even though bank instability does appear to rise generallyduring the period of increasing deregulation (see Section V below)
Finally the partisan structure of the state government alsoinuences when states deregulate As expected a higher propor-tion of Democrats in the government tends to delay deregulationA one-standard-deviation rise in the share of the governmentcontrolled by Democrats slows the deregulation by about twoyears Whether the state is dominated by one party however doesnot appear to affect the timing of the deregulation
To summarize our ndings from the hazard model privateinterests appear to play an important role in the deregulatoryprocess While private interests and public interests do sometimescoincide the results on the relative share of small banks and largebanks and on the relative size of insurance where banks competeare consistent with a private-interest approach but are difficult toexplain on public-interest grounds
B Robustness Tests
Omitted Variables States begin the sample period withdifferent degrees of restrictions on bank expansion These initialconditions may proxy for omitted factors determining the political-economy equilibrium in the state that could then affect the
QUARTERLY JOURNAL OF ECONOMICS1454
subsequent timing of deregulation For instance some statesbegan the period with unit banking while other states allowedlimited branching and the formation of multibank holdingcompanies
To test whether the stringency of initial restrictions onbranching matters we rst divide the sample into states with andwithout the most extreme form of branching regulation unitbanking28 Column (4) of Table III includes an indicator variableequal to one if the state began the period with unit bankingrestrictions The coefficient on this variable is positive andstatistically signicant but the other results remain virtuallyunchanged Next we include an indicator variable that is one ifthe state permitted multibank holding companies (MBHCs) toexist at the beginning of our sample period Thirty-four statesbegan the period allowing MBHCs The coefficient estimate forthis indicator variable (not reported) is small and not statisticallysignicant and its inclusion does not change the effects of theother variables
Eleven states expand bank powers to include the sale ofinsurance products during our sample period This could proxy foromitted factors about the relative strength of the insuranceindustry in the state In column (5) we include an indicatorvariable that is one if the state changes the law during our sampleperiod to permit banks to sell insurance business and zerootherwise The coefficient estimate is small and statisticallyinsignicant and does not affect the other results29
In column (6) we introduce four regional indicator variablesNorth South Midwest and West30 The coefficients on small bankshare the relative capital-to-assets ratio and small rm sharechange little Although the coefficient estimate still suggests that
28 We classify states that prohibited branching but permitted banks toestablish facilities as unit banking states The unit banking indicator is one for thefollowing states CO AR FL IL IA KS MN MO MT NE ND OK TX WI WVand WY
29 We conducted a likelihood ratio test of whether the unit banking statesthe MBHC states and the states that changed their regulations of bank insurancepowers could be pooled with the other states In all three cases we could not rejectpooling of the data
30 Since intrastate deregulation generally preceded interstate deregulationand the latter typically took the form of regional interstate compacts thepossibility of participating in a regional interstate banking compact could haveinuenced the decision to deregulate intrastate branching Our denitions of theregions are region 1 (South) contains AL AR DC FL GA KY LA MS NC OKSC TN TX and VA region 2 (Northeast) contains CT MA MD ME NH NJ NYPA RI VT and WV region 3 (Midwest) contains IA IL IN KS MI MN MO NEND OH SD and WI region 4 (West) contains the other states
WHAT DRIVES DEREGULATION 1455
in states where banks can sell insurance a larger relative size ofthe insurance sector delays deregulation the effect is no longerstatistically signicant The coefficient on the indicator variablefor whether a state expands bank powers to permit insurancesales during the sample period however now becomes statisti-cally signicantly negative This implies that deregulation occursroughly 41 years earlier in states where the insurance lobby maybe relatively weak While the inclusion of the regional indicatorsdoes not affect the other results the results on insurance doappear to be sensitive to the specication of the hazard model31
Estimation Technique To check the robustness of using ahazard model and to investigate whether the time-series or thecross-sectional variation in each of the factors accounts for thehazard results we estimate two simple linear probability models(neither are reported) First we run a pooled time-series cross-section OLS regression using the 637 observations from thehazard model The dependent variable equals one in the year ofderegulation and zero otherwise We include the same set ofregressors that were used in the hazard model and add state-levelxed effects This specication thus removes the cross-statevariation from the independent variables and focuses on theirmovements over time In this model the signs of the coefficientson the interest group factors are consistent with the results fromthe hazard model and the coefficients for the small bank relativecapital ratio and for the relative size of insurance where states cansell insurance are statistically signicant
In the second model we eliminate the time-series variationby averaging each of the explanatory variables over time We thushave one observation per state that had branching restrictions asof 1970 (N 5 39) The dependent variable equals the inverse of thenumber of years until deregulation which represents an estimateof the statersquos probability of deregulating in a year under theassumption that this probability is constant over time and eachyear provides an independent observation32 Although we haveonly 39 observations the coefficients on the small bank marketshare small bank relative capital ratio and small rm share
31 As further robustness check we also tried a population density variablewhich Abrams and Settle [1993] found relevant for regulatory change during the1920s and 1930s but its coefficient was not statistically signicant and did notaffect the other results
32 For instance the dependent variable would take the value of 01 for astate that deregulated in 1980 For the three states that did not deregulate thedependent variable equals 0
QUARTERLY JOURNAL OF ECONOMICS1456
variables are statistically signicant The signs of the effects ofthe insurance variables are the same as those reported in thehazard model but not statistically signicant
The hazard model and linear probability models thus producevery similar results The estimates from the latter suggest thatthe effects for the small bank market share and small rm sharevariables are driven primarily by cross-state variation the effectof the insurance variable is driven primarily by time-seriesvariation and the small bank capital effect is driven by bothcross-sectional and time-series variation
Ex Post Consequences of Deregulation Finally to check theplausibility of our interpretation of the hazard results we con-sider whether the ex post consequences of deregulation areconsistent with the ex ante lobbying positions we attribute to eachinterest group (see Kroszner and Strahan [1998] for details) Wend that small banks lose market share following deregulationand in states where banks can enter the insurance business theinsurance sector shrinks relative to the banking sector followingderegulation Borrowers also benet from deregulation throughlower average interest rates on loans These ndings support theprivate-interest interpretation of the coefficients in the hazardmodel groups that will benet lobby to speed deregulation andthose that will be harmed lobby to slow it
C Voting on Interstate Branching Deregulation in the U SHouse of Representatives
We now examine whether the forces driving intrastate branch-ing deregulation also drive interstate deregulation Financialservices interests are active contributors and lobbyists in Washing-ton Their political action committees constitute the largest groupof contributors to legislators providing nearly 20 percent of totalcongressional campaign contributions [Makinson 1992] and muchof their lobbying effort involves competition among rival interestswithin nancial services (see Kroszner and Stratmann [1998])
As noted above after virtually all states adopted intra- andinterstate branching deregulation the 1994 Riegle-Neal Act re-pealed the 1927 McFadden Act to eliminate all barriers tointerstate banking and branching by 1997 The key votes concern-ing the Riegle-Neal Act were either voice votes or extremelylopsided so we could not estimate a voting model from them Anumber of bills and amendments related to interstate branching
WHAT DRIVES DEREGULATION 1457
had been debated in Congress during the years prior to thepassage of the Riegle-Neal Act but a search of the weekly BNABanking Reporter and the Congressional Record produced onlyone roll-call vote related to interstate branching that was notlopsided This vote occurred in the House on November 14 1991and concerned an amendment sponsored by Wylie (R-OH) andNeal (D-NC) to introduce interstate banking and branchingderegulation to a nancial services reform package33 Althoughthe amendment passed by 210 to 208 the bill to which it wasattached subsequently was defeated
To check for the inuence of the factors we considered in thestate-level reforms we examine both the sponsorship of interstatebanking legislation and voting on the amendment The sponsorsof the Wylie-Neal amendment are from states with low small bankshares 004 in Ohio (Wylie) and 002 in North Carolina (Neal) Incontrast the sample mean in 1991 is 008 (median 5 007)Michigan home state of the Senatersquos sponsor of the 1994 Riegle-Neal Act also had relatively low small bank strength (small bankshare of 005)
Table IV reports estimates of a probit model where thedependent variable equals one if the legislator voted in favor ofthe amendment and zero otherwise The explanatory variablesare the same as those in the hazard model except that we use aparty affiliation indicator variable equal to one for Democrats andzero for Republicans as our proxy for political factors Thecoefficients in Table IV are the marginal effects (lsquolsquoslopesrsquorsquo) of aone-unit change of each variable on the probability that alegislator will vote for the amendment34 Note that a forcefavoring deregulation has a negative coefficient in the hazardmodel (Table III) but a positive coefficient in the probit model
Consistent with the state-level deregulation process legisla-tors are more likely to support the amendment if their states havea relatively low share of small banks As in the hazard model thefraction of small banks is the most important interest groupinuence on a legislatorrsquos voting decision A one-standard-deviation increase in small banksrsquo market share (from the mean)is associated with a decline in the probability of voting in favor of
33 The Wylie-Neal amendment also included provisions limiting certaininsurance and real estate powers of national banks [Congressional RecordNovember 14 1991 pp 10239ndash10242]
34 Since we have multiple legislators from each state we adjust the standarderrors to correct for the lack of independence among observations clustered in thesame state
QUARTERLY JOURNAL OF ECONOMICS1458
TAB
LE
IVM
AR
GIN
AL
EF
FE
CT
SF
RO
MA
PR
OB
ITM
OD
EL
OF
TH
EIN
FL
UE
NC
EO
FP
OL
ITIC
AL
EC
ON
OM
YF
AC
TO
RS
ON
TH
EH
OU
SE
VO
TE
ON
TH
EW
YL
IE-N
EA
L
AM
EN
DM
EN
TT
OP
ER
MIT
INT
ER
ST
AT
EB
AN
KB
RA
NC
HIN
G
NO
VE
MB
ER
14
1991
(1)
(2)
(3)
(4)
(5)
(6)
Smal
lban
kas
set
shar
eof
allb
anki
ng
asse
tsin
stat
e2
417
(12
4)2
454
(12
2)2
407
(12
0)2
443
(14
7)2
450
(14
9)2
490
(14
7)C
apit
alra
tio
ofsm
allb
anks
rela
tive
tola
rge
inth
est
ate
180
(37
5)2
68(3
42)
256
(37
5)1
69(3
78)
153
(38
0)2
45(4
25)
Rel
ativ
esi
zeof
insu
ranc
ein
stat
esw
here
bank
sm
ayse
llin
sura
nce
0ot
herw
ise
22
34
(10
2)2
249
(1
14)
22
33
(10
1)2
249
(1
04)
22
51
(10
2)2
154
(11
1)In
dica
tor
is1
ifba
nks
may
sell
insu
ranc
ein
the
stat
e0
79
(0
06)
074
(00
6)0
69
(0
06)
070
(00
6)0
70
(0
06)
067
(00
7)R
elat
ive
size
ofin
sura
nce
inst
ates
whe
reba
nks
may
not
sell
insu
ranc
e0
oth
erw
ise
180
(06
6)2
42
(0
72)
154
(0
67)
173
(06
6)1
72
(0
67)
203
(0
91)
Smal
lrm
shar
eof
the
num
ber
ofr
ms
inth
est
ate
21
23(0
92)
20
83(0
89)
21
62(1
07)
21
32(0
95)
21
30(1
00)
20
94(1
16)
Indi
cato
ris
1if
Rep
rese
ntat
ive
isa
Dem
ocra
t2
035
(00
7)2
035
(00
7)2
036
(00
7)2
035
(00
7)2
035
(00
7)2
036
(00
7)A
vera
geyi
eld
onba
nklo
ans
inth
est
ate
min
usF
edfu
nds
rate
mdash10
01
(51
5)mdash
mdashmdash
mdash
Ave
rage
fail
ure
rate
198
5to
1991
mdashmdash
228
(14
1)mdash
mdashmdash
Indi
cato
ris
1if
stat
eh
asun
itba
nkin
gla
wmdash
mdashmdash
004
(00
7)0
04(0
07)
20
01(0
08)
Indi
cato
ris
1if
stat
ech
ange
sba
nk
insu
ran
cepo
wer
smdash
mdashmdash
mdash2
001
(00
7)2
003
(00
7)In
clud
esre
gion
alin
dica
tors
N
oN
oN
oN
oN
oYe
sP
seud
o-R
20
170
180
170
170
170
18L
ogli
keli
hood
224
024
223
826
223
951
224
013
224
010
223
808
p-va
lue
ofx2
for
regr
essi
on
001
0
01
001
0
01
001
0
01
Th
ede
pen
den
tvar
iabl
eis
one
ifth
ele
gisl
ator
vote
sfo
rth
eam
endm
enta
nd
zero
ifag
ains
tT
he
repo
rted
coef
fici
ents
are
the
effe
cts
ofa
uni
tch
ange
ofth
ein
depe
nde
ntva
riab
le(f
rom
the
mea
n)on
the
prob
abili
tyof
voti
ng
infa
vor
ofth
eam
endm
ent
All
vari
able
sar
em
easu
red
byst
ate
N5
418
the
num
ber
ofR
epre
sent
ativ
esvo
tin
gon
the
amen
dmen
tS
mal
lban
kas
sets
har
eis
the
perc
ent
ofba
nki
ng
asse
tsin
the
stat
eh
eld
byba
nks
belo
wth
em
edia
nsi
zeof
ban
kin
each
stat
ein
each
year
Rel
ativ
eca
pita
lrat
iois
the
capi
talt
oas
sets
rati
oof
smal
lban
ksm
inu
sth
atof
larg
eba
nks
Siz
eof
insu
ran
cere
lati
veto
bank
ing
plu
sin
sura
nce
inth
est
ate
ism
easu
red
asgr
oss
stat
epr
oduc
tfro
min
sura
nce
divi
ded
bygr
oss
stat
epr
odu
ctfr
omin
sura
nce
plu
sba
nkin
gIn
dica
tor
ison
eif
stat
ela
wpe
rmit
sba
nks
toen
ter
the
insu
ran
cebu
sin
ess
Sm
all
rmsh
are
isth
epe
rcen
tofa
lles
tabl
ishm
ents
inth
est
ate
that
hav
efe
wer
than
twen
tyem
ploy
ees
Par
tyin
dica
tor
ison
eif
the
Rep
rese
nta
tive
isa
Dem
ocra
tA
vera
geyi
eld
onba
nklo
ans
equa
lsto
tali
nter
est
inco
me
onal
ldom
esti
clo
ans
mad
eby
ban
ksin
the
stat
edi
vide
dby
tota
lloa
ns
The
aver
age
bank
failu
rera
teis
the
rati
oof
asse
tsin
faile
dba
nks
duri
ng
agi
ven
stat
edi
vide
dby
tota
lban
kas
sets
inth
atst
ate
aver
aged
over
the
year
s19
85to
1991
Un
itba
nki
ng
indi
cato
req
ual
son
efo
rst
ates
wit
hun
itba
nkin
gre
stri
ctio
nsT
hech
ange
inin
sura
nce
pow
ers
indi
cato
req
ual
son
eif
the
stat
ech
ange
dth
ela
wto
perm
itba
nks
tose
llin
sura
nce
duri
ngth
esa
mpl
epe
riod
We
divi
deth
eco
unt
ryin
tofo
urre
gion
san
din
clu
dere
gion
alin
dica
tors
inco
lum
n(5
)R
obu
stst
anda
rder
rors
are
inpa
ren
thes
es
an
d
den
ote
stat
isti
cally
sign
ica
nta
tth
e1
5an
d10
perc
entl
evel
s
WHAT DRIVES DEREGULATION 1459
branching of approximately 17 percent The relative capital-to-assets ratio of small and large banks however does not have astatistically signicant effect in the probit model
The impact of rival interests outside of banking is alsoconsistent with intrastate deregulation results Where banks cansell insurance legislators from states with larger insurancesectors relative to banking are less likely to vote for the amend-ment35 A one-standard-deviation increase in the relative size ofinsurance in those states which permit banks to sell insurancedecreases the probability that a legislator will favor the amend-ment by about 13 percent Turning to consumer interests thecoefficient on small rm share is not statistically signicant butthe coefficient on interest rates in the state is A one-standard-deviation increase in the average interest rate on loans raises theprobability that a legislator will support the amendment byroughly 8 percent The coefficient estimate for the bankingdistress variable measured as the average bank failure rate from1985 to 1991 is small and statistically signicant Overall theprobit analysis of the vote on national branching deregulationsupports the private-interest theory of deregulation and providesa consistency check that the importance of interests operating onthe state legislatures are very similar to those operating at thefederal level
V WHY DID BRANCHING DEREGULATION BEGIN DURING THE
1970S SHOCKS TO THE EQUILIBRIUM SUPPORTING BANK
BRANCHING RESTRICTIONS
A complete explanation of regulatory exit should account forwhy deregulation begins in the 1970s as well as the timing of thestate-by-state reforms [Kane 1996] In this section we describebroad technological legal and economic shocks that altered thepolitical-economy equilibrium which had kept antibranching regu-lations little changed for at least 30 years
Beginning in the 1970s three innovations reduced the valueto the protected banks of local geographic monopolies (see TableV) First automatic teller machines (ATMs) helped to erode thegeographic ties between customers and banks Second checkablemoney market mutual funds and the Merrill Lynch Cash Manage-ment Account demonstrated that banking by mail and telephone
35 The positive and statistically signicant estimates on the other insurancevariables may reect insurance industry support for the amendmentrsquos provisionslimiting banksrsquo insurance powers
QUARTERLY JOURNAL OF ECONOMICS1460
provided a convenient alternative to local banks36 Third techno-logical innovation and deregulation reduced transportation andcommunication costs particularly since the 1970s thereby lower-ing the costs for customers to use distant banks By increasing theelasticity of deposits supplied to banks these innovations reducedthe value of geographical restrictions to their traditional benecia-ries and thereby reduced their incentive to ght to maintain them[Peltzman 1976]
On the lending side increasing sophistication of credit-scoring techniques following innovations in information process-ing nancial theory and the development of large credit databases began to diminish the value of knowledge that localbankers had about the risks of borrowers in the community As a
36 Regulation Q which limited the interest rates that banks could pay ondeposits may have helped to drive depositors away from banks when the gapbetween market rates and deposit ceilings grew during the 1970s The eliminationof interest rate ceilings on large denomination certicates of deposit during the1970s appears to have hurt smaller and retail-oriented banks relative to largerwholesale banks [James 1983]
TABLE VBROAD TRENDS IN COMMERCIAL BANKING 1950ndash1995
YearNumberof ATMs
Domesticbank
deposits(billions)
Moneymarketmutual
fund(billions)
Percent ofdeposits 1
moneyfunds heldby banks
Smallbanksrsquo
percent ofbankingassets
Averagenumber of
bankfailures
(1) (2) (3) (4) (5) (6)1950 0 $ 154 $ 0 100 NA 41955 0 191 0 100 NA 31960 0 228 0 100 24 21965 0 330 0 100 20 41970 0 479 0 100 18 61975 9750 775 4 99 18 61980 18500 1182 76 94 17 101985 61117 1787 242 88 14 601990 80156 2339 493 83 11 1791995 122706 2552 745 77 8 61
Column (1) ATM gures are from Bank Network News The EFT Network Data Book (New YorkFaulkner and Gray Inc) The 1975 gure was unavailable 9750 is the number of ATMs in 1978 the rst yearfor which complete data are available Columns (2)ndash(4) Banksrsquo domestic deposits are from the Reports ofIncome and Condition money market mutual funds are from the Flow of Funds Data on all bank depositsforeign plus domestic are only available beginning in 1970 The trend in banksrsquo share (column (4)) is the sameusing total deposits instead of domestic deposits Column (5) Percent of banking assets held by small bankswhere a small bank is dened as a commercial bank with less than $100 million in assets in 1994 dollarsThese data are based on the Reports of Income and Condition Data on small banks are not available before1960 Column (6) Five-year average number of bank failures where the nal year is indicated in the rstcolumn These data are from FDIC Annual Report and the Quarterly Banking Prole
WHAT DRIVES DEREGULATION 1461
result of these innovations a national market developed forresidential mortgages credit card receivables have been securi-tized and bank lending to small business now relies less on thejudgment of loan officers and more on standardized scoringmodels
These changes have increased the potential protability forlarge banks to enter what had been the core of small bankactivities Large banksrsquo incentive to increase their lobbying pres-sure to be able to expand into these markets has thus beenincreasing over time In fact small banksrsquo market share began todecline even prior to the branching deregulation (Table V) As thevalue of a local banking relationships declined small rms thatwere the main borrowers from the small banks also probablybecame more likely to favor the entry of large banks into localmarkets With the deadweight costs of preventing large bankentry rising the private-interest theory predicts that small localbanks would become less likely able to maintain the branchingrestrictions [Becker 1983] Deregulation that reduces deadweightcosts of regulation is also consistent with the public-interesttheory
Kane [1996] argues that another major shock to the oldequilibrium was the increasing public awareness of the costlinessof having government-insured but (geographically) undiversiednancial institutions In the late 1970s the failure rate of banksbegan to rise (Table V) In the 1980s the Savings and Loan crisisand taxpayer bailout further heightened public awareness aboutthe costs of restrictions that make depository institutions morelikely to require infusions of taxpayer funds The failures thusmay have heightened public support for branching deregulation(see also Abrams and Settle [1993]) As noted above we did notnd any evidence that banking failures or distress in a stateaffected the speed with which the state deregulated but distressstill may be a common factor affecting all states
These technological economic and legal shocks generatedconditions that changed the long-standing balance favoring theantibranching forces The marginal value of lobbying to repealbranching restrictions increased just as the relative value to thesmall banks of maintaining branching restrictions was decliningThese nationwide shocks are common factors across states andthey are consistent with the positive duration dependence associ-ated with state-level deregulation that we demonstrated in FigureII In addition a broad change in lsquolsquoideologyrsquorsquo against government
QUARTERLY JOURNAL OF ECONOMICS1462
regulation could explain the positive duration dependence duringthe sample period
VI CONCLUSIONS
The private-interest theory of regulation can account for thepattern of bank branching deregulation of the 1970s and 1980sBeneciaries of branching regulation had supported a coalitionfavoring geographical restrictions despite their costs to consum-ers of nancial services Innovations that began in the 1970saltered the value of the restrictions to the affected parties and theresulting competition among interest groups can explain thesubsequent deregulation While some of our results are alsoconsistent with the public interest theorymdashfor example deregula-tion occurs earlier when small banks are in a relatively weaknancial positionmdashother results particularly evidence on theimportance of rivalries between small and large banks andbetween banking and insurance are difficult to explain with thepublic-interest approach We also nd that political-institutionalfactors affect deregulation although these variables may act asproxies for unmeasured economic interests The greater success ofthe private-interest theory to explain deregulation here than inprevious work may be due to the richer cross-sectional andtime-series variation in branching deregulation that can beexploited with a hazard model Future empirical studies ofendogenous deregulation thus may be most fruitful where changehas occurred across states over time such as in franchisinginsurance and public utilities (eg Joskow [1996])
Technological and nancial innovations will continue to erodethe benets to any interest group of maintaining regulatorybarriers in nancial services These forces are likely to bringabout reforms both domestically for example through legislationthat would increase bank powers (see Kroszner [1996] Krosznerand Rajan [1994 1997] and Kroszner and Stratmann [1998]) andinternationally for example through the extension of nancialservices provisions of NAFTAto reduce geographic barriers acrosscountries [Kroszner 1997]
APPENDIX EFFECTS OF DIFFERENT TYPES OF BRANCHING
DEREGULATION ON ECONOMIC GROWTH BANKING STRUCTUREAND BANK EFFICIENCY
We measure the importance of alternative forms of deregula-tion by their consequences for the economic growth and the
WHAT DRIVES DEREGULATION 1463
structure and efficiency of the banking sector across the states(see Jayaratne and Strahan [1996 1998]) To do so we run anOLS panel regression relating indicators for each type of deregula-tion to potential consequences Our dependent variables includeproxies for economic growth (growth in personal income) bankstructure (the log of the number of banking companies and thenumber of branches per banking company) and efficiency (the logof total noninterest expenses and the ratio of loan loss provisionsto total loans) The four deregulation indicators (MBHC branch-ing through merger and acquisition unrestricted statewide branch-ing and interstate banking) are equal to one in the years afterderegulation and zero before If a state deregulated before thesample period begins in 1970 the indicator is one throughout ifthe state does not deregulate then the indicator is zero through-out All regressions include state xed effects and time effects
Table VI shows the results of including the set of four
TABLE VIMULTIVARIATE REGRESSIONS OF THE EFFECTS OF DIFFERENT TYPES OF BRANCHING
DEREGULATION ON ECONOMIC GROWTH BANKING STRUCTUREAND BANK EFFICIENCY
Dependent variable
Growthin
personalincome
Log of of
bankingcompanies
Branchesper
bankingcompany
Log ofnon-
interestexpenses
Loan lossprovisions
(loans 1leases)
Merger amp acquisitionbranchingindicator
0013(0004)
2 0162(0037)
1146(0453)
2 0107(0035)
2 0007(0001)
Unrestricted state-wide branchingindicator
2 0008(0005)
0086(0059)
2 0043(0574)
0070(0037)
0005(0002)
Interstate bankingindicator
0006(0005)
2 0052(0057)
1064(0511)
0005(0030)
2 0001(0001)
Multibank holdingcompany indicator
2 0005(0006)
2 0096(0052)
2 0364(0529)
2 0095(0036)
2 0001(0001)
Dependent variablemean [median]
0086[0082]
4881[5136]
6649[5055]
13812[13803]
0009[0006]
This table contains regressions of state-level economic growth and state-level measures of the structureand efficiency of the banking industry Each dependent variable is regressed on a set of state indicatorvariables a set of year indicator variables and the set of four deregulation indicators these deregulationindicators are equal to one after deregulation and zero before If a state was always regulated the indicator iszero throughout if it was always deregulated the indicator is one throughout Sample period is 1975 to 1991and Delaware and South Dakota are dropped so N 5 833 Noninterest expenses become available only in1984 so N 5 392 The coefficient on the deregulation indicators is reported with robust standard errors inparentheses and denote statistically signicant at the 1 5 and 10 percent levels
QUARTERLY JOURNAL OF ECONOMICS1464
deregulation indicators simultaneously The branching by mergerand acquisition indicator is statistically signicant and of theexpected sign in all ve of the equations and larger in absolutevalue than any of the other coefficients on the indicators for theother types of deregulation Since there may be multicollinearityamong our indicators we also ran univariate specications whichincluded each indicator separately and the results are verysimilar These results imply that branching by merger andacquisition is the most important type of deregulation measuredby its consequences for economic growth banking structure andbank efficiency See Kroszner and Strahan [1998] for more details
UNIVERSITY OF CHICAGO GRADUATE SCHOOL OF BUSINESS AND NATIONAL BUREAU OF
ECONOMIC RESEARCH
FEDERAL RESERVE BANK OF NEW YORK RESEARCH amp MARKET ANALYSIS GROUP
REFERENCES
Abrams Burton A and Russell F Settle lsquolsquoPressure-Group Inuence and Institu-tional Change Branch-Banking Legislation during the Great DepressionrsquorsquoPublic Choice LXXVII (1993) 687ndash705
Amel Dean lsquolsquoState Laws Affecting the Geographic Expansion of CommercialBanksrsquorsquo unpublished manuscript Board of Governors of the Federal ReserveSystem 1993
Becker Gary S lsquolsquoA Theory of Competition among Pressure Groups for PoliticalInuencersquorsquo Quarterly Journal of Economics XCVIII (1983) 371ndash400
Berger Allen Anil Kashyap Joseph Scalise lsquolsquoThe Transformation of the U SBanking Industry What a Long Strange Trip Itrsquos Beenrsquorsquo Brookings Papers onEconomic Activity 2 (1995) 55ndash218
Berger Allen Anthony Saunders Joseph Scalise and Gregory Udell lsquolsquoThe Effectsof Bank Mergers and Acquisitions on Small Business Lendingrsquorsquo Journal ofFinancial Economics L (1998) 187ndash229
Brennan Geoffrey and James Buchanan lsquolsquoTowards a Tax Constitution forLeviathanrsquorsquo Journal of Public Economics VIII (1977) 255ndash273
Brickley James A and Christopher M James lsquolsquoThe Takeover Market CorporateBoard Composition and Ownership Structure The Case of Bankingrsquorsquo Journalof Law and Economics XXX (1987) 161ndash180
Cole Rebel and John Wolken lsquolsquoFinancial Services Used by Small BusinessesEvidence from the 1993 Survey of Small Business Financesrsquorsquo Federal ReserveBulletin LXXXI (1994) 629ndash676
Dixit Avinash The Making of Economic Policy A Transactions-Cost PoliticsPerspective (Cambridge MA MIT Press 1996)
Economides Nicolas R Glenn Hubbard and Darius Palia lsquolsquoThe Political Economyof Branching Restrictions and Deposit Insurancersquorsquo Journal of Law andEconomics XXIX (1996) 667ndash704
Flannery Mark lsquolsquoThe Social Costs of Unit Banking Restrictionsrsquorsquo Journal ofMonetary Economics XIII (1984) 237ndash249
Greene William H Econometric Analysis third edition (Upper Saddle River NJPrentice-Hall 1997)
Gunther Jeffery lsquolsquoRegional Capital Imbalances and the Removal of InterstateBanking Restrictionsrsquorsquo Economics Letters XLIV (1994) 439ndash442 lsquolsquoThe Erosion of Intrastate Branch Banking Restrictionsrsquorsquo unpublishedmanuscript Federal Reserve Bank of Dallas 1996
Irwin Douglas and Randall Kroszner lsquolsquoInterests Institutions and Ideology inSecuring Policy Change The Republican Conversion to Trade Liberalizationafter Smoot-Hawleyrsquorsquo Journal of Law and Economics XLII (1999) 643ndash673
WHAT DRIVES DEREGULATION 1465
James Christopher lsquolsquoAn Analysis of Intra-Industry Differences in the Effect ofRegulation The Case of Deposit Rate Ceilingsrsquorsquo Journal of Monetary Econom-ics XII (1983) 417ndash442
Jarrell Gregg lsquolsquoChange at the Exchange The Causes and Effects of DeregulationrsquorsquoJournal of Law and Economics XXVII (1984) 273ndash312
Jayaratne Jith and Philip E Strahan lsquolsquoThe Finance-Growth Nexus Evidencefrom Bank Branch Deregulationrsquorsquo Quarterly Journal of Economics CI (1996)639ndash670
Jayaratne Jith and Philip E Strahan lsquolsquoEntry Restrictions Industry Evolutionand Dynamic Efficiency Evidence from Commercial Bankingrsquorsquo Journal of Lawand Economics XLIX (1998) 239ndash274
Joskow Paul lsquolsquoComment on lsquoPower Struggles Explaining Deregulatory Reformsin Electricity Marketsrsquo rsquorsquo Brookings Papers on Economic Activity Microeconom-ics (1996) 251ndash264
Joskow Paul and Roger Noll lsquolsquoRegulation in Theory and Practice An Overviewrsquorsquoin Studies in Public Regulation Gary Fromm ed (Cambridge MA MITPress 1981) pp 1ndash65
Kahn Matthew E and John G Matsusaka lsquolsquoDemand for Environmental GoodsEvidence from Voting Patterns on California Initiativesrsquorsquo Journal of Law andEconomics XL (1997) 137ndash173
Kalbeisch J and R Prentice The Statistical Analysis of Failure Time Data (NewYork John Wiley amp Sons 1980)
Kane Edward J lsquolsquoDe Jure Interstate Banking Why Only NowrsquorsquoJournal of MoneyCredit and Banking XXVIII (1996) 141ndash161
Kiefer Nicholas lsquolsquoEconomic Duration Data and Hazard Functionsrsquorsquo Journal ofEconomic Literature XXVI (1988) 646ndash679
Kroszner Randall S lsquolsquoThe Evolution of Universal Banking and Its Regulation inTwentieth Century Americarsquorsquo in Universal Banking Financial System DesignReconsidered Anthony Saunders and Ingo Walter eds (New York IrwinProfessional Publishers 1996) pp 70ndash99 lsquolsquoThe Political Economy of Banking and Financial Regulation in the U Srsquorsquo inThe Banking and Financial Structure in the NAFTA Countries and ChileGeorge M von Furstenberg ed (Boston Kluwer Academic Publishers 1997)pp 200ndash213
Kroszner Randall S and Raghuram G Rajan lsquolsquoIs the Glass-Steagall ActJustied A Study of the U S Experience with Universal Banking before1933rsquorsquo American Economic Review LXXXIV (1994) 810ndash832
Kroszner Randall S and Raghuram G Rajan lsquolsquoOrganization Structure andCredibility Evidence from Bank Securities Activities before the Glass-Steagall Actrsquorsquo Journal of Monetary Economics XXXIX (1997) 475ndash516
Kroszner Randall S and Philip E Strahan lsquolsquoRegulatory Incentives and the ThriftCrisis Dividends Mutual-to-Stock Conversions and Financial DistressrsquorsquoJournal of Finance LI (1996) 1285ndash1320
Kroszner Randall S and Philip E Strahan lsquolsquoWhat Drives Deregulation Econom-ics and Politics of the Relaxation of Bank Branching Restrictionsrsquorsquo NBERWorking Paper No 6637 1998
Kroszner Randall S and Thomas Stratmann lsquolsquoInterest Group Competition andthe Organization of Congress Theory and Evidence from Financial ServicesPolitical Action Committeesrsquorsquo American Economic Review LXXXVIII (1998)pp 1163ndash1187
Kroszner Randall S and Thomas Stratmann lsquolsquoDoes Political Ambiguity PayCorporate Campaign Contributions and Rewards to Legislator Reputationrsquorsquounpublishedmanuscript University of Chicago Graduate School of Business1999
Lin D Y and L J Wei lsquolsquoThe Robust Inference for the Cox Proportional HazardsModel Journal of the American Statistical Association LXXXIV (1989)1074ndash1078
Macey Jonathan R and Geoffrey P Miller Banking Law and Regulation (BostonLittle Brown and Company 1992)
Makinson Larry Open Secrets The Encyclopedia of Congressional Money andPolitics (Washington DC Congressional Quarterly 1992)
Niskanen WilliamBureaucracy and Representative Government (Chicago Aldine-Atherton 1971)
QUARTERLY JOURNAL OF ECONOMICS1466
Noll Roger lsquolsquoComment on Peltzmanrsquorsquo Brookings Papers Microeconomics SpecialIssue (1989) 48ndash58
North Douglass lsquolsquoA Transactions Cost Theory of Politicsrsquorsquo Journal of TheoreticalPolitics II (1990) 355ndash367
Peltzman Sam lsquolsquoToward a More General Theory of Regulationrsquorsquo Journal of Lawand Economics XIX (1976) 109ndash148 lsquolsquoConstituent Interest and Congressional Votingrsquorsquo Journal of Law andEconomics XXVII (1984) 181ndash210 lsquolsquoThe Economic Theory of Regulation after a Decade of DeregulationrsquorsquoBrookings Papers Microeconomics Special Issue (1989) 1ndash41
Peterson Mitchell and Raghuram Rajan lsquolsquoThe Benets of Lending RelationshipsEvidence from Small Business Datarsquorsquo Journal of Finance XLIX (1994) 3ndash38
Poole Keith and Howard Rosenthal Congress A Political-Economic History ofRoll Call Voting (Oxford Oxford University Press 1997)
Stata Stata Reference Manual Release 5 (College Station TX Stata Press 1997)Stigler George J lsquolsquoThe Theory of Economic Regulationrsquorsquo Bell Journal of Economics
and Management Science II (1971) 3ndash21Stigler George J Chicago Studies in the Political Economy (Chicago University
of Chicago Press 1988)Strahan Philip E and James Weston lsquolsquoSmall Business Lending and the Changing
Structure of the Banking Industryrsquorsquo Journal of Banking and Finance XXII(1998) 821ndash845
Sylla Richard John Legler and John Wallis lsquolsquoBanks and State Public Finance inthe New Republic The United States 1790ndash1860rsquorsquo Journal of EconomicHistory XLVII (1987) 391ndash403
White Eugene The Regulation and Reform of the American Banking System1900ndash1929 (Princeton Princeton University Press 1983)
Winston Clifford lsquolsquoEconomic Deregulation Days of Reckoning for Microecono-mistsrsquorsquo Journal of Economic Literature XXXI (1993) 1263ndash1289
WHAT DRIVES DEREGULATION 1467
The positive and statistically signicant coefficient on smallbank share implies that a greater small bank share delaysderegulation and this effect is economically important A one-standard-deviation increase in the small bank share results in a30 percent increase in the time until deregulation or about 47years This result supports the implications of private-interesttheory concerning intraindustry rivalry and is contrary to thepublic-interest theory
The positive and statistically signicant coefficient on therelative performance of small banks in the state implies thatstates deregulate later when small banks are relatively strongnancially25 A one-standard-deviation increase in the relativecapital-to-asset ratio results in a 15 percent rise in the time untilderegulation or about 24 years This result is consistent with theprivate-interest theory but may also be consistent with a public-interest interpretation because the risk of failures (and a costlytaxpayer-funded bailout) increases as small bank nancial healthdeclines
Interindustry competition also helps explain the timing ofderegulation In states where banks can sell insurance a rela-tively large insurance sector is associated with an increase in theexpected time to deregulation26 A one-standard-deviation in-crease in the relative size of the insurance sector in those statesthat permit banks to sell insurance leads to a 22 percent increasein the time until deregulation or about 35 years27 This result isconsistent with the private-interest theory but not the public-interest theory
Deregulation occurs earlier in states where small bank-dependent rms are relatively numerous A one-standard-
25 We also allowed small bank and large bank capital-to-asset ratios to enterseparately Their coefficient estimates were nearly equal in absolute value and ofopposite signs and so the data do not reject the relative small to large bank capitalratio specication we have chosen
26 When we use the relative size of the insurance sector in the state as awhole rather than insurance relative to banking plus insurance in the state weobtain similar results When we simultaneously include both measures of therelative size of insurance the interactions of the lsquolsquoinsurance relative to bankingplus insurancersquorsquo variable continue to be statisticallysignicant but the interactionsof lsquolsquoinsurance relative to the state as a wholersquorsquo variable are not
27 We have also estimated the model with the indicator variable equal to onefor states where banks may sell insurance but without the variables measuringtheimportance of insurance relative to banking In this model the coefficient on theinsurance indicator variablemdashan estimate of the difference in the expected time toderegulation in states where banks may sell insurance relative to states wherebanks may not sell insurancemdashequals 027 and is statistically signicant at the 5percent level This suggests that branching deregulation takes about 45 yearslonger in states where banks may sell insurance
WHAT DRIVES DEREGULATION 1453
deviation increase in the share of small rms reduces the timeuntil deregulation by 18 percent or about three years This resultconcerning the interests of consumers of banking services isconsistent with both the private- and public-interest theories
Column (2) of Table III introduces the average interest rateon loans to the model as a rough proxy for the cost to bankborrowers of these regulations These data become available onlyin 1976 so our sample size shrinks to 408 observations Thecoefficient on this variable is small and not statistically signi-cant The public-interest theory would imply that the coefficientshould be negative while the implication of the economic theorydoes not have a clear sign prediction
When we add the failure rate of banks to the specication incolumn (3) its coefficient is small and statistically insignicantWe do not nd a linkage between the timing of state deregulationand statewide banking distress Given that deposit insurance isfunded federally however bank failures may raise the publicrsquosconcerns about the costs of the regulations at the national ratherthan the state level This interpretation can explain why the rateof bank failure does not affect the timing of state-level deregula-tion even though bank instability does appear to rise generallyduring the period of increasing deregulation (see Section V below)
Finally the partisan structure of the state government alsoinuences when states deregulate As expected a higher propor-tion of Democrats in the government tends to delay deregulationA one-standard-deviation rise in the share of the governmentcontrolled by Democrats slows the deregulation by about twoyears Whether the state is dominated by one party however doesnot appear to affect the timing of the deregulation
To summarize our ndings from the hazard model privateinterests appear to play an important role in the deregulatoryprocess While private interests and public interests do sometimescoincide the results on the relative share of small banks and largebanks and on the relative size of insurance where banks competeare consistent with a private-interest approach but are difficult toexplain on public-interest grounds
B Robustness Tests
Omitted Variables States begin the sample period withdifferent degrees of restrictions on bank expansion These initialconditions may proxy for omitted factors determining the political-economy equilibrium in the state that could then affect the
QUARTERLY JOURNAL OF ECONOMICS1454
subsequent timing of deregulation For instance some statesbegan the period with unit banking while other states allowedlimited branching and the formation of multibank holdingcompanies
To test whether the stringency of initial restrictions onbranching matters we rst divide the sample into states with andwithout the most extreme form of branching regulation unitbanking28 Column (4) of Table III includes an indicator variableequal to one if the state began the period with unit bankingrestrictions The coefficient on this variable is positive andstatistically signicant but the other results remain virtuallyunchanged Next we include an indicator variable that is one ifthe state permitted multibank holding companies (MBHCs) toexist at the beginning of our sample period Thirty-four statesbegan the period allowing MBHCs The coefficient estimate forthis indicator variable (not reported) is small and not statisticallysignicant and its inclusion does not change the effects of theother variables
Eleven states expand bank powers to include the sale ofinsurance products during our sample period This could proxy foromitted factors about the relative strength of the insuranceindustry in the state In column (5) we include an indicatorvariable that is one if the state changes the law during our sampleperiod to permit banks to sell insurance business and zerootherwise The coefficient estimate is small and statisticallyinsignicant and does not affect the other results29
In column (6) we introduce four regional indicator variablesNorth South Midwest and West30 The coefficients on small bankshare the relative capital-to-assets ratio and small rm sharechange little Although the coefficient estimate still suggests that
28 We classify states that prohibited branching but permitted banks toestablish facilities as unit banking states The unit banking indicator is one for thefollowing states CO AR FL IL IA KS MN MO MT NE ND OK TX WI WVand WY
29 We conducted a likelihood ratio test of whether the unit banking statesthe MBHC states and the states that changed their regulations of bank insurancepowers could be pooled with the other states In all three cases we could not rejectpooling of the data
30 Since intrastate deregulation generally preceded interstate deregulationand the latter typically took the form of regional interstate compacts thepossibility of participating in a regional interstate banking compact could haveinuenced the decision to deregulate intrastate branching Our denitions of theregions are region 1 (South) contains AL AR DC FL GA KY LA MS NC OKSC TN TX and VA region 2 (Northeast) contains CT MA MD ME NH NJ NYPA RI VT and WV region 3 (Midwest) contains IA IL IN KS MI MN MO NEND OH SD and WI region 4 (West) contains the other states
WHAT DRIVES DEREGULATION 1455
in states where banks can sell insurance a larger relative size ofthe insurance sector delays deregulation the effect is no longerstatistically signicant The coefficient on the indicator variablefor whether a state expands bank powers to permit insurancesales during the sample period however now becomes statisti-cally signicantly negative This implies that deregulation occursroughly 41 years earlier in states where the insurance lobby maybe relatively weak While the inclusion of the regional indicatorsdoes not affect the other results the results on insurance doappear to be sensitive to the specication of the hazard model31
Estimation Technique To check the robustness of using ahazard model and to investigate whether the time-series or thecross-sectional variation in each of the factors accounts for thehazard results we estimate two simple linear probability models(neither are reported) First we run a pooled time-series cross-section OLS regression using the 637 observations from thehazard model The dependent variable equals one in the year ofderegulation and zero otherwise We include the same set ofregressors that were used in the hazard model and add state-levelxed effects This specication thus removes the cross-statevariation from the independent variables and focuses on theirmovements over time In this model the signs of the coefficientson the interest group factors are consistent with the results fromthe hazard model and the coefficients for the small bank relativecapital ratio and for the relative size of insurance where states cansell insurance are statistically signicant
In the second model we eliminate the time-series variationby averaging each of the explanatory variables over time We thushave one observation per state that had branching restrictions asof 1970 (N 5 39) The dependent variable equals the inverse of thenumber of years until deregulation which represents an estimateof the statersquos probability of deregulating in a year under theassumption that this probability is constant over time and eachyear provides an independent observation32 Although we haveonly 39 observations the coefficients on the small bank marketshare small bank relative capital ratio and small rm share
31 As further robustness check we also tried a population density variablewhich Abrams and Settle [1993] found relevant for regulatory change during the1920s and 1930s but its coefficient was not statistically signicant and did notaffect the other results
32 For instance the dependent variable would take the value of 01 for astate that deregulated in 1980 For the three states that did not deregulate thedependent variable equals 0
QUARTERLY JOURNAL OF ECONOMICS1456
variables are statistically signicant The signs of the effects ofthe insurance variables are the same as those reported in thehazard model but not statistically signicant
The hazard model and linear probability models thus producevery similar results The estimates from the latter suggest thatthe effects for the small bank market share and small rm sharevariables are driven primarily by cross-state variation the effectof the insurance variable is driven primarily by time-seriesvariation and the small bank capital effect is driven by bothcross-sectional and time-series variation
Ex Post Consequences of Deregulation Finally to check theplausibility of our interpretation of the hazard results we con-sider whether the ex post consequences of deregulation areconsistent with the ex ante lobbying positions we attribute to eachinterest group (see Kroszner and Strahan [1998] for details) Wend that small banks lose market share following deregulationand in states where banks can enter the insurance business theinsurance sector shrinks relative to the banking sector followingderegulation Borrowers also benet from deregulation throughlower average interest rates on loans These ndings support theprivate-interest interpretation of the coefficients in the hazardmodel groups that will benet lobby to speed deregulation andthose that will be harmed lobby to slow it
C Voting on Interstate Branching Deregulation in the U SHouse of Representatives
We now examine whether the forces driving intrastate branch-ing deregulation also drive interstate deregulation Financialservices interests are active contributors and lobbyists in Washing-ton Their political action committees constitute the largest groupof contributors to legislators providing nearly 20 percent of totalcongressional campaign contributions [Makinson 1992] and muchof their lobbying effort involves competition among rival interestswithin nancial services (see Kroszner and Stratmann [1998])
As noted above after virtually all states adopted intra- andinterstate branching deregulation the 1994 Riegle-Neal Act re-pealed the 1927 McFadden Act to eliminate all barriers tointerstate banking and branching by 1997 The key votes concern-ing the Riegle-Neal Act were either voice votes or extremelylopsided so we could not estimate a voting model from them Anumber of bills and amendments related to interstate branching
WHAT DRIVES DEREGULATION 1457
had been debated in Congress during the years prior to thepassage of the Riegle-Neal Act but a search of the weekly BNABanking Reporter and the Congressional Record produced onlyone roll-call vote related to interstate branching that was notlopsided This vote occurred in the House on November 14 1991and concerned an amendment sponsored by Wylie (R-OH) andNeal (D-NC) to introduce interstate banking and branchingderegulation to a nancial services reform package33 Althoughthe amendment passed by 210 to 208 the bill to which it wasattached subsequently was defeated
To check for the inuence of the factors we considered in thestate-level reforms we examine both the sponsorship of interstatebanking legislation and voting on the amendment The sponsorsof the Wylie-Neal amendment are from states with low small bankshares 004 in Ohio (Wylie) and 002 in North Carolina (Neal) Incontrast the sample mean in 1991 is 008 (median 5 007)Michigan home state of the Senatersquos sponsor of the 1994 Riegle-Neal Act also had relatively low small bank strength (small bankshare of 005)
Table IV reports estimates of a probit model where thedependent variable equals one if the legislator voted in favor ofthe amendment and zero otherwise The explanatory variablesare the same as those in the hazard model except that we use aparty affiliation indicator variable equal to one for Democrats andzero for Republicans as our proxy for political factors Thecoefficients in Table IV are the marginal effects (lsquolsquoslopesrsquorsquo) of aone-unit change of each variable on the probability that alegislator will vote for the amendment34 Note that a forcefavoring deregulation has a negative coefficient in the hazardmodel (Table III) but a positive coefficient in the probit model
Consistent with the state-level deregulation process legisla-tors are more likely to support the amendment if their states havea relatively low share of small banks As in the hazard model thefraction of small banks is the most important interest groupinuence on a legislatorrsquos voting decision A one-standard-deviation increase in small banksrsquo market share (from the mean)is associated with a decline in the probability of voting in favor of
33 The Wylie-Neal amendment also included provisions limiting certaininsurance and real estate powers of national banks [Congressional RecordNovember 14 1991 pp 10239ndash10242]
34 Since we have multiple legislators from each state we adjust the standarderrors to correct for the lack of independence among observations clustered in thesame state
QUARTERLY JOURNAL OF ECONOMICS1458
TAB
LE
IVM
AR
GIN
AL
EF
FE
CT
SF
RO
MA
PR
OB
ITM
OD
EL
OF
TH
EIN
FL
UE
NC
EO
FP
OL
ITIC
AL
EC
ON
OM
YF
AC
TO
RS
ON
TH
EH
OU
SE
VO
TE
ON
TH
EW
YL
IE-N
EA
L
AM
EN
DM
EN
TT
OP
ER
MIT
INT
ER
ST
AT
EB
AN
KB
RA
NC
HIN
G
NO
VE
MB
ER
14
1991
(1)
(2)
(3)
(4)
(5)
(6)
Smal
lban
kas
set
shar
eof
allb
anki
ng
asse
tsin
stat
e2
417
(12
4)2
454
(12
2)2
407
(12
0)2
443
(14
7)2
450
(14
9)2
490
(14
7)C
apit
alra
tio
ofsm
allb
anks
rela
tive
tola
rge
inth
est
ate
180
(37
5)2
68(3
42)
256
(37
5)1
69(3
78)
153
(38
0)2
45(4
25)
Rel
ativ
esi
zeof
insu
ranc
ein
stat
esw
here
bank
sm
ayse
llin
sura
nce
0ot
herw
ise
22
34
(10
2)2
249
(1
14)
22
33
(10
1)2
249
(1
04)
22
51
(10
2)2
154
(11
1)In
dica
tor
is1
ifba
nks
may
sell
insu
ranc
ein
the
stat
e0
79
(0
06)
074
(00
6)0
69
(0
06)
070
(00
6)0
70
(0
06)
067
(00
7)R
elat
ive
size
ofin
sura
nce
inst
ates
whe
reba
nks
may
not
sell
insu
ranc
e0
oth
erw
ise
180
(06
6)2
42
(0
72)
154
(0
67)
173
(06
6)1
72
(0
67)
203
(0
91)
Smal
lrm
shar
eof
the
num
ber
ofr
ms
inth
est
ate
21
23(0
92)
20
83(0
89)
21
62(1
07)
21
32(0
95)
21
30(1
00)
20
94(1
16)
Indi
cato
ris
1if
Rep
rese
ntat
ive
isa
Dem
ocra
t2
035
(00
7)2
035
(00
7)2
036
(00
7)2
035
(00
7)2
035
(00
7)2
036
(00
7)A
vera
geyi
eld
onba
nklo
ans
inth
est
ate
min
usF
edfu
nds
rate
mdash10
01
(51
5)mdash
mdashmdash
mdash
Ave
rage
fail
ure
rate
198
5to
1991
mdashmdash
228
(14
1)mdash
mdashmdash
Indi
cato
ris
1if
stat
eh
asun
itba
nkin
gla
wmdash
mdashmdash
004
(00
7)0
04(0
07)
20
01(0
08)
Indi
cato
ris
1if
stat
ech
ange
sba
nk
insu
ran
cepo
wer
smdash
mdashmdash
mdash2
001
(00
7)2
003
(00
7)In
clud
esre
gion
alin
dica
tors
N
oN
oN
oN
oN
oYe
sP
seud
o-R
20
170
180
170
170
170
18L
ogli
keli
hood
224
024
223
826
223
951
224
013
224
010
223
808
p-va
lue
ofx2
for
regr
essi
on
001
0
01
001
0
01
001
0
01
Th
ede
pen
den
tvar
iabl
eis
one
ifth
ele
gisl
ator
vote
sfo
rth
eam
endm
enta
nd
zero
ifag
ains
tT
he
repo
rted
coef
fici
ents
are
the
effe
cts
ofa
uni
tch
ange
ofth
ein
depe
nde
ntva
riab
le(f
rom
the
mea
n)on
the
prob
abili
tyof
voti
ng
infa
vor
ofth
eam
endm
ent
All
vari
able
sar
em
easu
red
byst
ate
N5
418
the
num
ber
ofR
epre
sent
ativ
esvo
tin
gon
the
amen
dmen
tS
mal
lban
kas
sets
har
eis
the
perc
ent
ofba
nki
ng
asse
tsin
the
stat
eh
eld
byba
nks
belo
wth
em
edia
nsi
zeof
ban
kin
each
stat
ein
each
year
Rel
ativ
eca
pita
lrat
iois
the
capi
talt
oas
sets
rati
oof
smal
lban
ksm
inu
sth
atof
larg
eba
nks
Siz
eof
insu
ran
cere
lati
veto
bank
ing
plu
sin
sura
nce
inth
est
ate
ism
easu
red
asgr
oss
stat
epr
oduc
tfro
min
sura
nce
divi
ded
bygr
oss
stat
epr
odu
ctfr
omin
sura
nce
plu
sba
nkin
gIn
dica
tor
ison
eif
stat
ela
wpe
rmit
sba
nks
toen
ter
the
insu
ran
cebu
sin
ess
Sm
all
rmsh
are
isth
epe
rcen
tofa
lles
tabl
ishm
ents
inth
est
ate
that
hav
efe
wer
than
twen
tyem
ploy
ees
Par
tyin
dica
tor
ison
eif
the
Rep
rese
nta
tive
isa
Dem
ocra
tA
vera
geyi
eld
onba
nklo
ans
equa
lsto
tali
nter
est
inco
me
onal
ldom
esti
clo
ans
mad
eby
ban
ksin
the
stat
edi
vide
dby
tota
lloa
ns
The
aver
age
bank
failu
rera
teis
the
rati
oof
asse
tsin
faile
dba
nks
duri
ng
agi
ven
stat
edi
vide
dby
tota
lban
kas
sets
inth
atst
ate
aver
aged
over
the
year
s19
85to
1991
Un
itba
nki
ng
indi
cato
req
ual
son
efo
rst
ates
wit
hun
itba
nkin
gre
stri
ctio
nsT
hech
ange
inin
sura
nce
pow
ers
indi
cato
req
ual
son
eif
the
stat
ech
ange
dth
ela
wto
perm
itba
nks
tose
llin
sura
nce
duri
ngth
esa
mpl
epe
riod
We
divi
deth
eco
unt
ryin
tofo
urre
gion
san
din
clu
dere
gion
alin
dica
tors
inco
lum
n(5
)R
obu
stst
anda
rder
rors
are
inpa
ren
thes
es
an
d
den
ote
stat
isti
cally
sign
ica
nta
tth
e1
5an
d10
perc
entl
evel
s
WHAT DRIVES DEREGULATION 1459
branching of approximately 17 percent The relative capital-to-assets ratio of small and large banks however does not have astatistically signicant effect in the probit model
The impact of rival interests outside of banking is alsoconsistent with intrastate deregulation results Where banks cansell insurance legislators from states with larger insurancesectors relative to banking are less likely to vote for the amend-ment35 A one-standard-deviation increase in the relative size ofinsurance in those states which permit banks to sell insurancedecreases the probability that a legislator will favor the amend-ment by about 13 percent Turning to consumer interests thecoefficient on small rm share is not statistically signicant butthe coefficient on interest rates in the state is A one-standard-deviation increase in the average interest rate on loans raises theprobability that a legislator will support the amendment byroughly 8 percent The coefficient estimate for the bankingdistress variable measured as the average bank failure rate from1985 to 1991 is small and statistically signicant Overall theprobit analysis of the vote on national branching deregulationsupports the private-interest theory of deregulation and providesa consistency check that the importance of interests operating onthe state legislatures are very similar to those operating at thefederal level
V WHY DID BRANCHING DEREGULATION BEGIN DURING THE
1970S SHOCKS TO THE EQUILIBRIUM SUPPORTING BANK
BRANCHING RESTRICTIONS
A complete explanation of regulatory exit should account forwhy deregulation begins in the 1970s as well as the timing of thestate-by-state reforms [Kane 1996] In this section we describebroad technological legal and economic shocks that altered thepolitical-economy equilibrium which had kept antibranching regu-lations little changed for at least 30 years
Beginning in the 1970s three innovations reduced the valueto the protected banks of local geographic monopolies (see TableV) First automatic teller machines (ATMs) helped to erode thegeographic ties between customers and banks Second checkablemoney market mutual funds and the Merrill Lynch Cash Manage-ment Account demonstrated that banking by mail and telephone
35 The positive and statistically signicant estimates on the other insurancevariables may reect insurance industry support for the amendmentrsquos provisionslimiting banksrsquo insurance powers
QUARTERLY JOURNAL OF ECONOMICS1460
provided a convenient alternative to local banks36 Third techno-logical innovation and deregulation reduced transportation andcommunication costs particularly since the 1970s thereby lower-ing the costs for customers to use distant banks By increasing theelasticity of deposits supplied to banks these innovations reducedthe value of geographical restrictions to their traditional benecia-ries and thereby reduced their incentive to ght to maintain them[Peltzman 1976]
On the lending side increasing sophistication of credit-scoring techniques following innovations in information process-ing nancial theory and the development of large credit databases began to diminish the value of knowledge that localbankers had about the risks of borrowers in the community As a
36 Regulation Q which limited the interest rates that banks could pay ondeposits may have helped to drive depositors away from banks when the gapbetween market rates and deposit ceilings grew during the 1970s The eliminationof interest rate ceilings on large denomination certicates of deposit during the1970s appears to have hurt smaller and retail-oriented banks relative to largerwholesale banks [James 1983]
TABLE VBROAD TRENDS IN COMMERCIAL BANKING 1950ndash1995
YearNumberof ATMs
Domesticbank
deposits(billions)
Moneymarketmutual
fund(billions)
Percent ofdeposits 1
moneyfunds heldby banks
Smallbanksrsquo
percent ofbankingassets
Averagenumber of
bankfailures
(1) (2) (3) (4) (5) (6)1950 0 $ 154 $ 0 100 NA 41955 0 191 0 100 NA 31960 0 228 0 100 24 21965 0 330 0 100 20 41970 0 479 0 100 18 61975 9750 775 4 99 18 61980 18500 1182 76 94 17 101985 61117 1787 242 88 14 601990 80156 2339 493 83 11 1791995 122706 2552 745 77 8 61
Column (1) ATM gures are from Bank Network News The EFT Network Data Book (New YorkFaulkner and Gray Inc) The 1975 gure was unavailable 9750 is the number of ATMs in 1978 the rst yearfor which complete data are available Columns (2)ndash(4) Banksrsquo domestic deposits are from the Reports ofIncome and Condition money market mutual funds are from the Flow of Funds Data on all bank depositsforeign plus domestic are only available beginning in 1970 The trend in banksrsquo share (column (4)) is the sameusing total deposits instead of domestic deposits Column (5) Percent of banking assets held by small bankswhere a small bank is dened as a commercial bank with less than $100 million in assets in 1994 dollarsThese data are based on the Reports of Income and Condition Data on small banks are not available before1960 Column (6) Five-year average number of bank failures where the nal year is indicated in the rstcolumn These data are from FDIC Annual Report and the Quarterly Banking Prole
WHAT DRIVES DEREGULATION 1461
result of these innovations a national market developed forresidential mortgages credit card receivables have been securi-tized and bank lending to small business now relies less on thejudgment of loan officers and more on standardized scoringmodels
These changes have increased the potential protability forlarge banks to enter what had been the core of small bankactivities Large banksrsquo incentive to increase their lobbying pres-sure to be able to expand into these markets has thus beenincreasing over time In fact small banksrsquo market share began todecline even prior to the branching deregulation (Table V) As thevalue of a local banking relationships declined small rms thatwere the main borrowers from the small banks also probablybecame more likely to favor the entry of large banks into localmarkets With the deadweight costs of preventing large bankentry rising the private-interest theory predicts that small localbanks would become less likely able to maintain the branchingrestrictions [Becker 1983] Deregulation that reduces deadweightcosts of regulation is also consistent with the public-interesttheory
Kane [1996] argues that another major shock to the oldequilibrium was the increasing public awareness of the costlinessof having government-insured but (geographically) undiversiednancial institutions In the late 1970s the failure rate of banksbegan to rise (Table V) In the 1980s the Savings and Loan crisisand taxpayer bailout further heightened public awareness aboutthe costs of restrictions that make depository institutions morelikely to require infusions of taxpayer funds The failures thusmay have heightened public support for branching deregulation(see also Abrams and Settle [1993]) As noted above we did notnd any evidence that banking failures or distress in a stateaffected the speed with which the state deregulated but distressstill may be a common factor affecting all states
These technological economic and legal shocks generatedconditions that changed the long-standing balance favoring theantibranching forces The marginal value of lobbying to repealbranching restrictions increased just as the relative value to thesmall banks of maintaining branching restrictions was decliningThese nationwide shocks are common factors across states andthey are consistent with the positive duration dependence associ-ated with state-level deregulation that we demonstrated in FigureII In addition a broad change in lsquolsquoideologyrsquorsquo against government
QUARTERLY JOURNAL OF ECONOMICS1462
regulation could explain the positive duration dependence duringthe sample period
VI CONCLUSIONS
The private-interest theory of regulation can account for thepattern of bank branching deregulation of the 1970s and 1980sBeneciaries of branching regulation had supported a coalitionfavoring geographical restrictions despite their costs to consum-ers of nancial services Innovations that began in the 1970saltered the value of the restrictions to the affected parties and theresulting competition among interest groups can explain thesubsequent deregulation While some of our results are alsoconsistent with the public interest theorymdashfor example deregula-tion occurs earlier when small banks are in a relatively weaknancial positionmdashother results particularly evidence on theimportance of rivalries between small and large banks andbetween banking and insurance are difficult to explain with thepublic-interest approach We also nd that political-institutionalfactors affect deregulation although these variables may act asproxies for unmeasured economic interests The greater success ofthe private-interest theory to explain deregulation here than inprevious work may be due to the richer cross-sectional andtime-series variation in branching deregulation that can beexploited with a hazard model Future empirical studies ofendogenous deregulation thus may be most fruitful where changehas occurred across states over time such as in franchisinginsurance and public utilities (eg Joskow [1996])
Technological and nancial innovations will continue to erodethe benets to any interest group of maintaining regulatorybarriers in nancial services These forces are likely to bringabout reforms both domestically for example through legislationthat would increase bank powers (see Kroszner [1996] Krosznerand Rajan [1994 1997] and Kroszner and Stratmann [1998]) andinternationally for example through the extension of nancialservices provisions of NAFTAto reduce geographic barriers acrosscountries [Kroszner 1997]
APPENDIX EFFECTS OF DIFFERENT TYPES OF BRANCHING
DEREGULATION ON ECONOMIC GROWTH BANKING STRUCTUREAND BANK EFFICIENCY
We measure the importance of alternative forms of deregula-tion by their consequences for the economic growth and the
WHAT DRIVES DEREGULATION 1463
structure and efficiency of the banking sector across the states(see Jayaratne and Strahan [1996 1998]) To do so we run anOLS panel regression relating indicators for each type of deregula-tion to potential consequences Our dependent variables includeproxies for economic growth (growth in personal income) bankstructure (the log of the number of banking companies and thenumber of branches per banking company) and efficiency (the logof total noninterest expenses and the ratio of loan loss provisionsto total loans) The four deregulation indicators (MBHC branch-ing through merger and acquisition unrestricted statewide branch-ing and interstate banking) are equal to one in the years afterderegulation and zero before If a state deregulated before thesample period begins in 1970 the indicator is one throughout ifthe state does not deregulate then the indicator is zero through-out All regressions include state xed effects and time effects
Table VI shows the results of including the set of four
TABLE VIMULTIVARIATE REGRESSIONS OF THE EFFECTS OF DIFFERENT TYPES OF BRANCHING
DEREGULATION ON ECONOMIC GROWTH BANKING STRUCTUREAND BANK EFFICIENCY
Dependent variable
Growthin
personalincome
Log of of
bankingcompanies
Branchesper
bankingcompany
Log ofnon-
interestexpenses
Loan lossprovisions
(loans 1leases)
Merger amp acquisitionbranchingindicator
0013(0004)
2 0162(0037)
1146(0453)
2 0107(0035)
2 0007(0001)
Unrestricted state-wide branchingindicator
2 0008(0005)
0086(0059)
2 0043(0574)
0070(0037)
0005(0002)
Interstate bankingindicator
0006(0005)
2 0052(0057)
1064(0511)
0005(0030)
2 0001(0001)
Multibank holdingcompany indicator
2 0005(0006)
2 0096(0052)
2 0364(0529)
2 0095(0036)
2 0001(0001)
Dependent variablemean [median]
0086[0082]
4881[5136]
6649[5055]
13812[13803]
0009[0006]
This table contains regressions of state-level economic growth and state-level measures of the structureand efficiency of the banking industry Each dependent variable is regressed on a set of state indicatorvariables a set of year indicator variables and the set of four deregulation indicators these deregulationindicators are equal to one after deregulation and zero before If a state was always regulated the indicator iszero throughout if it was always deregulated the indicator is one throughout Sample period is 1975 to 1991and Delaware and South Dakota are dropped so N 5 833 Noninterest expenses become available only in1984 so N 5 392 The coefficient on the deregulation indicators is reported with robust standard errors inparentheses and denote statistically signicant at the 1 5 and 10 percent levels
QUARTERLY JOURNAL OF ECONOMICS1464
deregulation indicators simultaneously The branching by mergerand acquisition indicator is statistically signicant and of theexpected sign in all ve of the equations and larger in absolutevalue than any of the other coefficients on the indicators for theother types of deregulation Since there may be multicollinearityamong our indicators we also ran univariate specications whichincluded each indicator separately and the results are verysimilar These results imply that branching by merger andacquisition is the most important type of deregulation measuredby its consequences for economic growth banking structure andbank efficiency See Kroszner and Strahan [1998] for more details
UNIVERSITY OF CHICAGO GRADUATE SCHOOL OF BUSINESS AND NATIONAL BUREAU OF
ECONOMIC RESEARCH
FEDERAL RESERVE BANK OF NEW YORK RESEARCH amp MARKET ANALYSIS GROUP
REFERENCES
Abrams Burton A and Russell F Settle lsquolsquoPressure-Group Inuence and Institu-tional Change Branch-Banking Legislation during the Great DepressionrsquorsquoPublic Choice LXXVII (1993) 687ndash705
Amel Dean lsquolsquoState Laws Affecting the Geographic Expansion of CommercialBanksrsquorsquo unpublished manuscript Board of Governors of the Federal ReserveSystem 1993
Becker Gary S lsquolsquoA Theory of Competition among Pressure Groups for PoliticalInuencersquorsquo Quarterly Journal of Economics XCVIII (1983) 371ndash400
Berger Allen Anil Kashyap Joseph Scalise lsquolsquoThe Transformation of the U SBanking Industry What a Long Strange Trip Itrsquos Beenrsquorsquo Brookings Papers onEconomic Activity 2 (1995) 55ndash218
Berger Allen Anthony Saunders Joseph Scalise and Gregory Udell lsquolsquoThe Effectsof Bank Mergers and Acquisitions on Small Business Lendingrsquorsquo Journal ofFinancial Economics L (1998) 187ndash229
Brennan Geoffrey and James Buchanan lsquolsquoTowards a Tax Constitution forLeviathanrsquorsquo Journal of Public Economics VIII (1977) 255ndash273
Brickley James A and Christopher M James lsquolsquoThe Takeover Market CorporateBoard Composition and Ownership Structure The Case of Bankingrsquorsquo Journalof Law and Economics XXX (1987) 161ndash180
Cole Rebel and John Wolken lsquolsquoFinancial Services Used by Small BusinessesEvidence from the 1993 Survey of Small Business Financesrsquorsquo Federal ReserveBulletin LXXXI (1994) 629ndash676
Dixit Avinash The Making of Economic Policy A Transactions-Cost PoliticsPerspective (Cambridge MA MIT Press 1996)
Economides Nicolas R Glenn Hubbard and Darius Palia lsquolsquoThe Political Economyof Branching Restrictions and Deposit Insurancersquorsquo Journal of Law andEconomics XXIX (1996) 667ndash704
Flannery Mark lsquolsquoThe Social Costs of Unit Banking Restrictionsrsquorsquo Journal ofMonetary Economics XIII (1984) 237ndash249
Greene William H Econometric Analysis third edition (Upper Saddle River NJPrentice-Hall 1997)
Gunther Jeffery lsquolsquoRegional Capital Imbalances and the Removal of InterstateBanking Restrictionsrsquorsquo Economics Letters XLIV (1994) 439ndash442 lsquolsquoThe Erosion of Intrastate Branch Banking Restrictionsrsquorsquo unpublishedmanuscript Federal Reserve Bank of Dallas 1996
Irwin Douglas and Randall Kroszner lsquolsquoInterests Institutions and Ideology inSecuring Policy Change The Republican Conversion to Trade Liberalizationafter Smoot-Hawleyrsquorsquo Journal of Law and Economics XLII (1999) 643ndash673
WHAT DRIVES DEREGULATION 1465
James Christopher lsquolsquoAn Analysis of Intra-Industry Differences in the Effect ofRegulation The Case of Deposit Rate Ceilingsrsquorsquo Journal of Monetary Econom-ics XII (1983) 417ndash442
Jarrell Gregg lsquolsquoChange at the Exchange The Causes and Effects of DeregulationrsquorsquoJournal of Law and Economics XXVII (1984) 273ndash312
Jayaratne Jith and Philip E Strahan lsquolsquoThe Finance-Growth Nexus Evidencefrom Bank Branch Deregulationrsquorsquo Quarterly Journal of Economics CI (1996)639ndash670
Jayaratne Jith and Philip E Strahan lsquolsquoEntry Restrictions Industry Evolutionand Dynamic Efficiency Evidence from Commercial Bankingrsquorsquo Journal of Lawand Economics XLIX (1998) 239ndash274
Joskow Paul lsquolsquoComment on lsquoPower Struggles Explaining Deregulatory Reformsin Electricity Marketsrsquo rsquorsquo Brookings Papers on Economic Activity Microeconom-ics (1996) 251ndash264
Joskow Paul and Roger Noll lsquolsquoRegulation in Theory and Practice An Overviewrsquorsquoin Studies in Public Regulation Gary Fromm ed (Cambridge MA MITPress 1981) pp 1ndash65
Kahn Matthew E and John G Matsusaka lsquolsquoDemand for Environmental GoodsEvidence from Voting Patterns on California Initiativesrsquorsquo Journal of Law andEconomics XL (1997) 137ndash173
Kalbeisch J and R Prentice The Statistical Analysis of Failure Time Data (NewYork John Wiley amp Sons 1980)
Kane Edward J lsquolsquoDe Jure Interstate Banking Why Only NowrsquorsquoJournal of MoneyCredit and Banking XXVIII (1996) 141ndash161
Kiefer Nicholas lsquolsquoEconomic Duration Data and Hazard Functionsrsquorsquo Journal ofEconomic Literature XXVI (1988) 646ndash679
Kroszner Randall S lsquolsquoThe Evolution of Universal Banking and Its Regulation inTwentieth Century Americarsquorsquo in Universal Banking Financial System DesignReconsidered Anthony Saunders and Ingo Walter eds (New York IrwinProfessional Publishers 1996) pp 70ndash99 lsquolsquoThe Political Economy of Banking and Financial Regulation in the U Srsquorsquo inThe Banking and Financial Structure in the NAFTA Countries and ChileGeorge M von Furstenberg ed (Boston Kluwer Academic Publishers 1997)pp 200ndash213
Kroszner Randall S and Raghuram G Rajan lsquolsquoIs the Glass-Steagall ActJustied A Study of the U S Experience with Universal Banking before1933rsquorsquo American Economic Review LXXXIV (1994) 810ndash832
Kroszner Randall S and Raghuram G Rajan lsquolsquoOrganization Structure andCredibility Evidence from Bank Securities Activities before the Glass-Steagall Actrsquorsquo Journal of Monetary Economics XXXIX (1997) 475ndash516
Kroszner Randall S and Philip E Strahan lsquolsquoRegulatory Incentives and the ThriftCrisis Dividends Mutual-to-Stock Conversions and Financial DistressrsquorsquoJournal of Finance LI (1996) 1285ndash1320
Kroszner Randall S and Philip E Strahan lsquolsquoWhat Drives Deregulation Econom-ics and Politics of the Relaxation of Bank Branching Restrictionsrsquorsquo NBERWorking Paper No 6637 1998
Kroszner Randall S and Thomas Stratmann lsquolsquoInterest Group Competition andthe Organization of Congress Theory and Evidence from Financial ServicesPolitical Action Committeesrsquorsquo American Economic Review LXXXVIII (1998)pp 1163ndash1187
Kroszner Randall S and Thomas Stratmann lsquolsquoDoes Political Ambiguity PayCorporate Campaign Contributions and Rewards to Legislator Reputationrsquorsquounpublishedmanuscript University of Chicago Graduate School of Business1999
Lin D Y and L J Wei lsquolsquoThe Robust Inference for the Cox Proportional HazardsModel Journal of the American Statistical Association LXXXIV (1989)1074ndash1078
Macey Jonathan R and Geoffrey P Miller Banking Law and Regulation (BostonLittle Brown and Company 1992)
Makinson Larry Open Secrets The Encyclopedia of Congressional Money andPolitics (Washington DC Congressional Quarterly 1992)
Niskanen WilliamBureaucracy and Representative Government (Chicago Aldine-Atherton 1971)
QUARTERLY JOURNAL OF ECONOMICS1466
Noll Roger lsquolsquoComment on Peltzmanrsquorsquo Brookings Papers Microeconomics SpecialIssue (1989) 48ndash58
North Douglass lsquolsquoA Transactions Cost Theory of Politicsrsquorsquo Journal of TheoreticalPolitics II (1990) 355ndash367
Peltzman Sam lsquolsquoToward a More General Theory of Regulationrsquorsquo Journal of Lawand Economics XIX (1976) 109ndash148 lsquolsquoConstituent Interest and Congressional Votingrsquorsquo Journal of Law andEconomics XXVII (1984) 181ndash210 lsquolsquoThe Economic Theory of Regulation after a Decade of DeregulationrsquorsquoBrookings Papers Microeconomics Special Issue (1989) 1ndash41
Peterson Mitchell and Raghuram Rajan lsquolsquoThe Benets of Lending RelationshipsEvidence from Small Business Datarsquorsquo Journal of Finance XLIX (1994) 3ndash38
Poole Keith and Howard Rosenthal Congress A Political-Economic History ofRoll Call Voting (Oxford Oxford University Press 1997)
Stata Stata Reference Manual Release 5 (College Station TX Stata Press 1997)Stigler George J lsquolsquoThe Theory of Economic Regulationrsquorsquo Bell Journal of Economics
and Management Science II (1971) 3ndash21Stigler George J Chicago Studies in the Political Economy (Chicago University
of Chicago Press 1988)Strahan Philip E and James Weston lsquolsquoSmall Business Lending and the Changing
Structure of the Banking Industryrsquorsquo Journal of Banking and Finance XXII(1998) 821ndash845
Sylla Richard John Legler and John Wallis lsquolsquoBanks and State Public Finance inthe New Republic The United States 1790ndash1860rsquorsquo Journal of EconomicHistory XLVII (1987) 391ndash403
White Eugene The Regulation and Reform of the American Banking System1900ndash1929 (Princeton Princeton University Press 1983)
Winston Clifford lsquolsquoEconomic Deregulation Days of Reckoning for Microecono-mistsrsquorsquo Journal of Economic Literature XXXI (1993) 1263ndash1289
WHAT DRIVES DEREGULATION 1467
deviation increase in the share of small rms reduces the timeuntil deregulation by 18 percent or about three years This resultconcerning the interests of consumers of banking services isconsistent with both the private- and public-interest theories
Column (2) of Table III introduces the average interest rateon loans to the model as a rough proxy for the cost to bankborrowers of these regulations These data become available onlyin 1976 so our sample size shrinks to 408 observations Thecoefficient on this variable is small and not statistically signi-cant The public-interest theory would imply that the coefficientshould be negative while the implication of the economic theorydoes not have a clear sign prediction
When we add the failure rate of banks to the specication incolumn (3) its coefficient is small and statistically insignicantWe do not nd a linkage between the timing of state deregulationand statewide banking distress Given that deposit insurance isfunded federally however bank failures may raise the publicrsquosconcerns about the costs of the regulations at the national ratherthan the state level This interpretation can explain why the rateof bank failure does not affect the timing of state-level deregula-tion even though bank instability does appear to rise generallyduring the period of increasing deregulation (see Section V below)
Finally the partisan structure of the state government alsoinuences when states deregulate As expected a higher propor-tion of Democrats in the government tends to delay deregulationA one-standard-deviation rise in the share of the governmentcontrolled by Democrats slows the deregulation by about twoyears Whether the state is dominated by one party however doesnot appear to affect the timing of the deregulation
To summarize our ndings from the hazard model privateinterests appear to play an important role in the deregulatoryprocess While private interests and public interests do sometimescoincide the results on the relative share of small banks and largebanks and on the relative size of insurance where banks competeare consistent with a private-interest approach but are difficult toexplain on public-interest grounds
B Robustness Tests
Omitted Variables States begin the sample period withdifferent degrees of restrictions on bank expansion These initialconditions may proxy for omitted factors determining the political-economy equilibrium in the state that could then affect the
QUARTERLY JOURNAL OF ECONOMICS1454
subsequent timing of deregulation For instance some statesbegan the period with unit banking while other states allowedlimited branching and the formation of multibank holdingcompanies
To test whether the stringency of initial restrictions onbranching matters we rst divide the sample into states with andwithout the most extreme form of branching regulation unitbanking28 Column (4) of Table III includes an indicator variableequal to one if the state began the period with unit bankingrestrictions The coefficient on this variable is positive andstatistically signicant but the other results remain virtuallyunchanged Next we include an indicator variable that is one ifthe state permitted multibank holding companies (MBHCs) toexist at the beginning of our sample period Thirty-four statesbegan the period allowing MBHCs The coefficient estimate forthis indicator variable (not reported) is small and not statisticallysignicant and its inclusion does not change the effects of theother variables
Eleven states expand bank powers to include the sale ofinsurance products during our sample period This could proxy foromitted factors about the relative strength of the insuranceindustry in the state In column (5) we include an indicatorvariable that is one if the state changes the law during our sampleperiod to permit banks to sell insurance business and zerootherwise The coefficient estimate is small and statisticallyinsignicant and does not affect the other results29
In column (6) we introduce four regional indicator variablesNorth South Midwest and West30 The coefficients on small bankshare the relative capital-to-assets ratio and small rm sharechange little Although the coefficient estimate still suggests that
28 We classify states that prohibited branching but permitted banks toestablish facilities as unit banking states The unit banking indicator is one for thefollowing states CO AR FL IL IA KS MN MO MT NE ND OK TX WI WVand WY
29 We conducted a likelihood ratio test of whether the unit banking statesthe MBHC states and the states that changed their regulations of bank insurancepowers could be pooled with the other states In all three cases we could not rejectpooling of the data
30 Since intrastate deregulation generally preceded interstate deregulationand the latter typically took the form of regional interstate compacts thepossibility of participating in a regional interstate banking compact could haveinuenced the decision to deregulate intrastate branching Our denitions of theregions are region 1 (South) contains AL AR DC FL GA KY LA MS NC OKSC TN TX and VA region 2 (Northeast) contains CT MA MD ME NH NJ NYPA RI VT and WV region 3 (Midwest) contains IA IL IN KS MI MN MO NEND OH SD and WI region 4 (West) contains the other states
WHAT DRIVES DEREGULATION 1455
in states where banks can sell insurance a larger relative size ofthe insurance sector delays deregulation the effect is no longerstatistically signicant The coefficient on the indicator variablefor whether a state expands bank powers to permit insurancesales during the sample period however now becomes statisti-cally signicantly negative This implies that deregulation occursroughly 41 years earlier in states where the insurance lobby maybe relatively weak While the inclusion of the regional indicatorsdoes not affect the other results the results on insurance doappear to be sensitive to the specication of the hazard model31
Estimation Technique To check the robustness of using ahazard model and to investigate whether the time-series or thecross-sectional variation in each of the factors accounts for thehazard results we estimate two simple linear probability models(neither are reported) First we run a pooled time-series cross-section OLS regression using the 637 observations from thehazard model The dependent variable equals one in the year ofderegulation and zero otherwise We include the same set ofregressors that were used in the hazard model and add state-levelxed effects This specication thus removes the cross-statevariation from the independent variables and focuses on theirmovements over time In this model the signs of the coefficientson the interest group factors are consistent with the results fromthe hazard model and the coefficients for the small bank relativecapital ratio and for the relative size of insurance where states cansell insurance are statistically signicant
In the second model we eliminate the time-series variationby averaging each of the explanatory variables over time We thushave one observation per state that had branching restrictions asof 1970 (N 5 39) The dependent variable equals the inverse of thenumber of years until deregulation which represents an estimateof the statersquos probability of deregulating in a year under theassumption that this probability is constant over time and eachyear provides an independent observation32 Although we haveonly 39 observations the coefficients on the small bank marketshare small bank relative capital ratio and small rm share
31 As further robustness check we also tried a population density variablewhich Abrams and Settle [1993] found relevant for regulatory change during the1920s and 1930s but its coefficient was not statistically signicant and did notaffect the other results
32 For instance the dependent variable would take the value of 01 for astate that deregulated in 1980 For the three states that did not deregulate thedependent variable equals 0
QUARTERLY JOURNAL OF ECONOMICS1456
variables are statistically signicant The signs of the effects ofthe insurance variables are the same as those reported in thehazard model but not statistically signicant
The hazard model and linear probability models thus producevery similar results The estimates from the latter suggest thatthe effects for the small bank market share and small rm sharevariables are driven primarily by cross-state variation the effectof the insurance variable is driven primarily by time-seriesvariation and the small bank capital effect is driven by bothcross-sectional and time-series variation
Ex Post Consequences of Deregulation Finally to check theplausibility of our interpretation of the hazard results we con-sider whether the ex post consequences of deregulation areconsistent with the ex ante lobbying positions we attribute to eachinterest group (see Kroszner and Strahan [1998] for details) Wend that small banks lose market share following deregulationand in states where banks can enter the insurance business theinsurance sector shrinks relative to the banking sector followingderegulation Borrowers also benet from deregulation throughlower average interest rates on loans These ndings support theprivate-interest interpretation of the coefficients in the hazardmodel groups that will benet lobby to speed deregulation andthose that will be harmed lobby to slow it
C Voting on Interstate Branching Deregulation in the U SHouse of Representatives
We now examine whether the forces driving intrastate branch-ing deregulation also drive interstate deregulation Financialservices interests are active contributors and lobbyists in Washing-ton Their political action committees constitute the largest groupof contributors to legislators providing nearly 20 percent of totalcongressional campaign contributions [Makinson 1992] and muchof their lobbying effort involves competition among rival interestswithin nancial services (see Kroszner and Stratmann [1998])
As noted above after virtually all states adopted intra- andinterstate branching deregulation the 1994 Riegle-Neal Act re-pealed the 1927 McFadden Act to eliminate all barriers tointerstate banking and branching by 1997 The key votes concern-ing the Riegle-Neal Act were either voice votes or extremelylopsided so we could not estimate a voting model from them Anumber of bills and amendments related to interstate branching
WHAT DRIVES DEREGULATION 1457
had been debated in Congress during the years prior to thepassage of the Riegle-Neal Act but a search of the weekly BNABanking Reporter and the Congressional Record produced onlyone roll-call vote related to interstate branching that was notlopsided This vote occurred in the House on November 14 1991and concerned an amendment sponsored by Wylie (R-OH) andNeal (D-NC) to introduce interstate banking and branchingderegulation to a nancial services reform package33 Althoughthe amendment passed by 210 to 208 the bill to which it wasattached subsequently was defeated
To check for the inuence of the factors we considered in thestate-level reforms we examine both the sponsorship of interstatebanking legislation and voting on the amendment The sponsorsof the Wylie-Neal amendment are from states with low small bankshares 004 in Ohio (Wylie) and 002 in North Carolina (Neal) Incontrast the sample mean in 1991 is 008 (median 5 007)Michigan home state of the Senatersquos sponsor of the 1994 Riegle-Neal Act also had relatively low small bank strength (small bankshare of 005)
Table IV reports estimates of a probit model where thedependent variable equals one if the legislator voted in favor ofthe amendment and zero otherwise The explanatory variablesare the same as those in the hazard model except that we use aparty affiliation indicator variable equal to one for Democrats andzero for Republicans as our proxy for political factors Thecoefficients in Table IV are the marginal effects (lsquolsquoslopesrsquorsquo) of aone-unit change of each variable on the probability that alegislator will vote for the amendment34 Note that a forcefavoring deregulation has a negative coefficient in the hazardmodel (Table III) but a positive coefficient in the probit model
Consistent with the state-level deregulation process legisla-tors are more likely to support the amendment if their states havea relatively low share of small banks As in the hazard model thefraction of small banks is the most important interest groupinuence on a legislatorrsquos voting decision A one-standard-deviation increase in small banksrsquo market share (from the mean)is associated with a decline in the probability of voting in favor of
33 The Wylie-Neal amendment also included provisions limiting certaininsurance and real estate powers of national banks [Congressional RecordNovember 14 1991 pp 10239ndash10242]
34 Since we have multiple legislators from each state we adjust the standarderrors to correct for the lack of independence among observations clustered in thesame state
QUARTERLY JOURNAL OF ECONOMICS1458
TAB
LE
IVM
AR
GIN
AL
EF
FE
CT
SF
RO
MA
PR
OB
ITM
OD
EL
OF
TH
EIN
FL
UE
NC
EO
FP
OL
ITIC
AL
EC
ON
OM
YF
AC
TO
RS
ON
TH
EH
OU
SE
VO
TE
ON
TH
EW
YL
IE-N
EA
L
AM
EN
DM
EN
TT
OP
ER
MIT
INT
ER
ST
AT
EB
AN
KB
RA
NC
HIN
G
NO
VE
MB
ER
14
1991
(1)
(2)
(3)
(4)
(5)
(6)
Smal
lban
kas
set
shar
eof
allb
anki
ng
asse
tsin
stat
e2
417
(12
4)2
454
(12
2)2
407
(12
0)2
443
(14
7)2
450
(14
9)2
490
(14
7)C
apit
alra
tio
ofsm
allb
anks
rela
tive
tola
rge
inth
est
ate
180
(37
5)2
68(3
42)
256
(37
5)1
69(3
78)
153
(38
0)2
45(4
25)
Rel
ativ
esi
zeof
insu
ranc
ein
stat
esw
here
bank
sm
ayse
llin
sura
nce
0ot
herw
ise
22
34
(10
2)2
249
(1
14)
22
33
(10
1)2
249
(1
04)
22
51
(10
2)2
154
(11
1)In
dica
tor
is1
ifba
nks
may
sell
insu
ranc
ein
the
stat
e0
79
(0
06)
074
(00
6)0
69
(0
06)
070
(00
6)0
70
(0
06)
067
(00
7)R
elat
ive
size
ofin
sura
nce
inst
ates
whe
reba
nks
may
not
sell
insu
ranc
e0
oth
erw
ise
180
(06
6)2
42
(0
72)
154
(0
67)
173
(06
6)1
72
(0
67)
203
(0
91)
Smal
lrm
shar
eof
the
num
ber
ofr
ms
inth
est
ate
21
23(0
92)
20
83(0
89)
21
62(1
07)
21
32(0
95)
21
30(1
00)
20
94(1
16)
Indi
cato
ris
1if
Rep
rese
ntat
ive
isa
Dem
ocra
t2
035
(00
7)2
035
(00
7)2
036
(00
7)2
035
(00
7)2
035
(00
7)2
036
(00
7)A
vera
geyi
eld
onba
nklo
ans
inth
est
ate
min
usF
edfu
nds
rate
mdash10
01
(51
5)mdash
mdashmdash
mdash
Ave
rage
fail
ure
rate
198
5to
1991
mdashmdash
228
(14
1)mdash
mdashmdash
Indi
cato
ris
1if
stat
eh
asun
itba
nkin
gla
wmdash
mdashmdash
004
(00
7)0
04(0
07)
20
01(0
08)
Indi
cato
ris
1if
stat
ech
ange
sba
nk
insu
ran
cepo
wer
smdash
mdashmdash
mdash2
001
(00
7)2
003
(00
7)In
clud
esre
gion
alin
dica
tors
N
oN
oN
oN
oN
oYe
sP
seud
o-R
20
170
180
170
170
170
18L
ogli
keli
hood
224
024
223
826
223
951
224
013
224
010
223
808
p-va
lue
ofx2
for
regr
essi
on
001
0
01
001
0
01
001
0
01
Th
ede
pen
den
tvar
iabl
eis
one
ifth
ele
gisl
ator
vote
sfo
rth
eam
endm
enta
nd
zero
ifag
ains
tT
he
repo
rted
coef
fici
ents
are
the
effe
cts
ofa
uni
tch
ange
ofth
ein
depe
nde
ntva
riab
le(f
rom
the
mea
n)on
the
prob
abili
tyof
voti
ng
infa
vor
ofth
eam
endm
ent
All
vari
able
sar
em
easu
red
byst
ate
N5
418
the
num
ber
ofR
epre
sent
ativ
esvo
tin
gon
the
amen
dmen
tS
mal
lban
kas
sets
har
eis
the
perc
ent
ofba
nki
ng
asse
tsin
the
stat
eh
eld
byba
nks
belo
wth
em
edia
nsi
zeof
ban
kin
each
stat
ein
each
year
Rel
ativ
eca
pita
lrat
iois
the
capi
talt
oas
sets
rati
oof
smal
lban
ksm
inu
sth
atof
larg
eba
nks
Siz
eof
insu
ran
cere
lati
veto
bank
ing
plu
sin
sura
nce
inth
est
ate
ism
easu
red
asgr
oss
stat
epr
oduc
tfro
min
sura
nce
divi
ded
bygr
oss
stat
epr
odu
ctfr
omin
sura
nce
plu
sba
nkin
gIn
dica
tor
ison
eif
stat
ela
wpe
rmit
sba
nks
toen
ter
the
insu
ran
cebu
sin
ess
Sm
all
rmsh
are
isth
epe
rcen
tofa
lles
tabl
ishm
ents
inth
est
ate
that
hav
efe
wer
than
twen
tyem
ploy
ees
Par
tyin
dica
tor
ison
eif
the
Rep
rese
nta
tive
isa
Dem
ocra
tA
vera
geyi
eld
onba
nklo
ans
equa
lsto
tali
nter
est
inco
me
onal
ldom
esti
clo
ans
mad
eby
ban
ksin
the
stat
edi
vide
dby
tota
lloa
ns
The
aver
age
bank
failu
rera
teis
the
rati
oof
asse
tsin
faile
dba
nks
duri
ng
agi
ven
stat
edi
vide
dby
tota
lban
kas
sets
inth
atst
ate
aver
aged
over
the
year
s19
85to
1991
Un
itba
nki
ng
indi
cato
req
ual
son
efo
rst
ates
wit
hun
itba
nkin
gre
stri
ctio
nsT
hech
ange
inin
sura
nce
pow
ers
indi
cato
req
ual
son
eif
the
stat
ech
ange
dth
ela
wto
perm
itba
nks
tose
llin
sura
nce
duri
ngth
esa
mpl
epe
riod
We
divi
deth
eco
unt
ryin
tofo
urre
gion
san
din
clu
dere
gion
alin
dica
tors
inco
lum
n(5
)R
obu
stst
anda
rder
rors
are
inpa
ren
thes
es
an
d
den
ote
stat
isti
cally
sign
ica
nta
tth
e1
5an
d10
perc
entl
evel
s
WHAT DRIVES DEREGULATION 1459
branching of approximately 17 percent The relative capital-to-assets ratio of small and large banks however does not have astatistically signicant effect in the probit model
The impact of rival interests outside of banking is alsoconsistent with intrastate deregulation results Where banks cansell insurance legislators from states with larger insurancesectors relative to banking are less likely to vote for the amend-ment35 A one-standard-deviation increase in the relative size ofinsurance in those states which permit banks to sell insurancedecreases the probability that a legislator will favor the amend-ment by about 13 percent Turning to consumer interests thecoefficient on small rm share is not statistically signicant butthe coefficient on interest rates in the state is A one-standard-deviation increase in the average interest rate on loans raises theprobability that a legislator will support the amendment byroughly 8 percent The coefficient estimate for the bankingdistress variable measured as the average bank failure rate from1985 to 1991 is small and statistically signicant Overall theprobit analysis of the vote on national branching deregulationsupports the private-interest theory of deregulation and providesa consistency check that the importance of interests operating onthe state legislatures are very similar to those operating at thefederal level
V WHY DID BRANCHING DEREGULATION BEGIN DURING THE
1970S SHOCKS TO THE EQUILIBRIUM SUPPORTING BANK
BRANCHING RESTRICTIONS
A complete explanation of regulatory exit should account forwhy deregulation begins in the 1970s as well as the timing of thestate-by-state reforms [Kane 1996] In this section we describebroad technological legal and economic shocks that altered thepolitical-economy equilibrium which had kept antibranching regu-lations little changed for at least 30 years
Beginning in the 1970s three innovations reduced the valueto the protected banks of local geographic monopolies (see TableV) First automatic teller machines (ATMs) helped to erode thegeographic ties between customers and banks Second checkablemoney market mutual funds and the Merrill Lynch Cash Manage-ment Account demonstrated that banking by mail and telephone
35 The positive and statistically signicant estimates on the other insurancevariables may reect insurance industry support for the amendmentrsquos provisionslimiting banksrsquo insurance powers
QUARTERLY JOURNAL OF ECONOMICS1460
provided a convenient alternative to local banks36 Third techno-logical innovation and deregulation reduced transportation andcommunication costs particularly since the 1970s thereby lower-ing the costs for customers to use distant banks By increasing theelasticity of deposits supplied to banks these innovations reducedthe value of geographical restrictions to their traditional benecia-ries and thereby reduced their incentive to ght to maintain them[Peltzman 1976]
On the lending side increasing sophistication of credit-scoring techniques following innovations in information process-ing nancial theory and the development of large credit databases began to diminish the value of knowledge that localbankers had about the risks of borrowers in the community As a
36 Regulation Q which limited the interest rates that banks could pay ondeposits may have helped to drive depositors away from banks when the gapbetween market rates and deposit ceilings grew during the 1970s The eliminationof interest rate ceilings on large denomination certicates of deposit during the1970s appears to have hurt smaller and retail-oriented banks relative to largerwholesale banks [James 1983]
TABLE VBROAD TRENDS IN COMMERCIAL BANKING 1950ndash1995
YearNumberof ATMs
Domesticbank
deposits(billions)
Moneymarketmutual
fund(billions)
Percent ofdeposits 1
moneyfunds heldby banks
Smallbanksrsquo
percent ofbankingassets
Averagenumber of
bankfailures
(1) (2) (3) (4) (5) (6)1950 0 $ 154 $ 0 100 NA 41955 0 191 0 100 NA 31960 0 228 0 100 24 21965 0 330 0 100 20 41970 0 479 0 100 18 61975 9750 775 4 99 18 61980 18500 1182 76 94 17 101985 61117 1787 242 88 14 601990 80156 2339 493 83 11 1791995 122706 2552 745 77 8 61
Column (1) ATM gures are from Bank Network News The EFT Network Data Book (New YorkFaulkner and Gray Inc) The 1975 gure was unavailable 9750 is the number of ATMs in 1978 the rst yearfor which complete data are available Columns (2)ndash(4) Banksrsquo domestic deposits are from the Reports ofIncome and Condition money market mutual funds are from the Flow of Funds Data on all bank depositsforeign plus domestic are only available beginning in 1970 The trend in banksrsquo share (column (4)) is the sameusing total deposits instead of domestic deposits Column (5) Percent of banking assets held by small bankswhere a small bank is dened as a commercial bank with less than $100 million in assets in 1994 dollarsThese data are based on the Reports of Income and Condition Data on small banks are not available before1960 Column (6) Five-year average number of bank failures where the nal year is indicated in the rstcolumn These data are from FDIC Annual Report and the Quarterly Banking Prole
WHAT DRIVES DEREGULATION 1461
result of these innovations a national market developed forresidential mortgages credit card receivables have been securi-tized and bank lending to small business now relies less on thejudgment of loan officers and more on standardized scoringmodels
These changes have increased the potential protability forlarge banks to enter what had been the core of small bankactivities Large banksrsquo incentive to increase their lobbying pres-sure to be able to expand into these markets has thus beenincreasing over time In fact small banksrsquo market share began todecline even prior to the branching deregulation (Table V) As thevalue of a local banking relationships declined small rms thatwere the main borrowers from the small banks also probablybecame more likely to favor the entry of large banks into localmarkets With the deadweight costs of preventing large bankentry rising the private-interest theory predicts that small localbanks would become less likely able to maintain the branchingrestrictions [Becker 1983] Deregulation that reduces deadweightcosts of regulation is also consistent with the public-interesttheory
Kane [1996] argues that another major shock to the oldequilibrium was the increasing public awareness of the costlinessof having government-insured but (geographically) undiversiednancial institutions In the late 1970s the failure rate of banksbegan to rise (Table V) In the 1980s the Savings and Loan crisisand taxpayer bailout further heightened public awareness aboutthe costs of restrictions that make depository institutions morelikely to require infusions of taxpayer funds The failures thusmay have heightened public support for branching deregulation(see also Abrams and Settle [1993]) As noted above we did notnd any evidence that banking failures or distress in a stateaffected the speed with which the state deregulated but distressstill may be a common factor affecting all states
These technological economic and legal shocks generatedconditions that changed the long-standing balance favoring theantibranching forces The marginal value of lobbying to repealbranching restrictions increased just as the relative value to thesmall banks of maintaining branching restrictions was decliningThese nationwide shocks are common factors across states andthey are consistent with the positive duration dependence associ-ated with state-level deregulation that we demonstrated in FigureII In addition a broad change in lsquolsquoideologyrsquorsquo against government
QUARTERLY JOURNAL OF ECONOMICS1462
regulation could explain the positive duration dependence duringthe sample period
VI CONCLUSIONS
The private-interest theory of regulation can account for thepattern of bank branching deregulation of the 1970s and 1980sBeneciaries of branching regulation had supported a coalitionfavoring geographical restrictions despite their costs to consum-ers of nancial services Innovations that began in the 1970saltered the value of the restrictions to the affected parties and theresulting competition among interest groups can explain thesubsequent deregulation While some of our results are alsoconsistent with the public interest theorymdashfor example deregula-tion occurs earlier when small banks are in a relatively weaknancial positionmdashother results particularly evidence on theimportance of rivalries between small and large banks andbetween banking and insurance are difficult to explain with thepublic-interest approach We also nd that political-institutionalfactors affect deregulation although these variables may act asproxies for unmeasured economic interests The greater success ofthe private-interest theory to explain deregulation here than inprevious work may be due to the richer cross-sectional andtime-series variation in branching deregulation that can beexploited with a hazard model Future empirical studies ofendogenous deregulation thus may be most fruitful where changehas occurred across states over time such as in franchisinginsurance and public utilities (eg Joskow [1996])
Technological and nancial innovations will continue to erodethe benets to any interest group of maintaining regulatorybarriers in nancial services These forces are likely to bringabout reforms both domestically for example through legislationthat would increase bank powers (see Kroszner [1996] Krosznerand Rajan [1994 1997] and Kroszner and Stratmann [1998]) andinternationally for example through the extension of nancialservices provisions of NAFTAto reduce geographic barriers acrosscountries [Kroszner 1997]
APPENDIX EFFECTS OF DIFFERENT TYPES OF BRANCHING
DEREGULATION ON ECONOMIC GROWTH BANKING STRUCTUREAND BANK EFFICIENCY
We measure the importance of alternative forms of deregula-tion by their consequences for the economic growth and the
WHAT DRIVES DEREGULATION 1463
structure and efficiency of the banking sector across the states(see Jayaratne and Strahan [1996 1998]) To do so we run anOLS panel regression relating indicators for each type of deregula-tion to potential consequences Our dependent variables includeproxies for economic growth (growth in personal income) bankstructure (the log of the number of banking companies and thenumber of branches per banking company) and efficiency (the logof total noninterest expenses and the ratio of loan loss provisionsto total loans) The four deregulation indicators (MBHC branch-ing through merger and acquisition unrestricted statewide branch-ing and interstate banking) are equal to one in the years afterderegulation and zero before If a state deregulated before thesample period begins in 1970 the indicator is one throughout ifthe state does not deregulate then the indicator is zero through-out All regressions include state xed effects and time effects
Table VI shows the results of including the set of four
TABLE VIMULTIVARIATE REGRESSIONS OF THE EFFECTS OF DIFFERENT TYPES OF BRANCHING
DEREGULATION ON ECONOMIC GROWTH BANKING STRUCTUREAND BANK EFFICIENCY
Dependent variable
Growthin
personalincome
Log of of
bankingcompanies
Branchesper
bankingcompany
Log ofnon-
interestexpenses
Loan lossprovisions
(loans 1leases)
Merger amp acquisitionbranchingindicator
0013(0004)
2 0162(0037)
1146(0453)
2 0107(0035)
2 0007(0001)
Unrestricted state-wide branchingindicator
2 0008(0005)
0086(0059)
2 0043(0574)
0070(0037)
0005(0002)
Interstate bankingindicator
0006(0005)
2 0052(0057)
1064(0511)
0005(0030)
2 0001(0001)
Multibank holdingcompany indicator
2 0005(0006)
2 0096(0052)
2 0364(0529)
2 0095(0036)
2 0001(0001)
Dependent variablemean [median]
0086[0082]
4881[5136]
6649[5055]
13812[13803]
0009[0006]
This table contains regressions of state-level economic growth and state-level measures of the structureand efficiency of the banking industry Each dependent variable is regressed on a set of state indicatorvariables a set of year indicator variables and the set of four deregulation indicators these deregulationindicators are equal to one after deregulation and zero before If a state was always regulated the indicator iszero throughout if it was always deregulated the indicator is one throughout Sample period is 1975 to 1991and Delaware and South Dakota are dropped so N 5 833 Noninterest expenses become available only in1984 so N 5 392 The coefficient on the deregulation indicators is reported with robust standard errors inparentheses and denote statistically signicant at the 1 5 and 10 percent levels
QUARTERLY JOURNAL OF ECONOMICS1464
deregulation indicators simultaneously The branching by mergerand acquisition indicator is statistically signicant and of theexpected sign in all ve of the equations and larger in absolutevalue than any of the other coefficients on the indicators for theother types of deregulation Since there may be multicollinearityamong our indicators we also ran univariate specications whichincluded each indicator separately and the results are verysimilar These results imply that branching by merger andacquisition is the most important type of deregulation measuredby its consequences for economic growth banking structure andbank efficiency See Kroszner and Strahan [1998] for more details
UNIVERSITY OF CHICAGO GRADUATE SCHOOL OF BUSINESS AND NATIONAL BUREAU OF
ECONOMIC RESEARCH
FEDERAL RESERVE BANK OF NEW YORK RESEARCH amp MARKET ANALYSIS GROUP
REFERENCES
Abrams Burton A and Russell F Settle lsquolsquoPressure-Group Inuence and Institu-tional Change Branch-Banking Legislation during the Great DepressionrsquorsquoPublic Choice LXXVII (1993) 687ndash705
Amel Dean lsquolsquoState Laws Affecting the Geographic Expansion of CommercialBanksrsquorsquo unpublished manuscript Board of Governors of the Federal ReserveSystem 1993
Becker Gary S lsquolsquoA Theory of Competition among Pressure Groups for PoliticalInuencersquorsquo Quarterly Journal of Economics XCVIII (1983) 371ndash400
Berger Allen Anil Kashyap Joseph Scalise lsquolsquoThe Transformation of the U SBanking Industry What a Long Strange Trip Itrsquos Beenrsquorsquo Brookings Papers onEconomic Activity 2 (1995) 55ndash218
Berger Allen Anthony Saunders Joseph Scalise and Gregory Udell lsquolsquoThe Effectsof Bank Mergers and Acquisitions on Small Business Lendingrsquorsquo Journal ofFinancial Economics L (1998) 187ndash229
Brennan Geoffrey and James Buchanan lsquolsquoTowards a Tax Constitution forLeviathanrsquorsquo Journal of Public Economics VIII (1977) 255ndash273
Brickley James A and Christopher M James lsquolsquoThe Takeover Market CorporateBoard Composition and Ownership Structure The Case of Bankingrsquorsquo Journalof Law and Economics XXX (1987) 161ndash180
Cole Rebel and John Wolken lsquolsquoFinancial Services Used by Small BusinessesEvidence from the 1993 Survey of Small Business Financesrsquorsquo Federal ReserveBulletin LXXXI (1994) 629ndash676
Dixit Avinash The Making of Economic Policy A Transactions-Cost PoliticsPerspective (Cambridge MA MIT Press 1996)
Economides Nicolas R Glenn Hubbard and Darius Palia lsquolsquoThe Political Economyof Branching Restrictions and Deposit Insurancersquorsquo Journal of Law andEconomics XXIX (1996) 667ndash704
Flannery Mark lsquolsquoThe Social Costs of Unit Banking Restrictionsrsquorsquo Journal ofMonetary Economics XIII (1984) 237ndash249
Greene William H Econometric Analysis third edition (Upper Saddle River NJPrentice-Hall 1997)
Gunther Jeffery lsquolsquoRegional Capital Imbalances and the Removal of InterstateBanking Restrictionsrsquorsquo Economics Letters XLIV (1994) 439ndash442 lsquolsquoThe Erosion of Intrastate Branch Banking Restrictionsrsquorsquo unpublishedmanuscript Federal Reserve Bank of Dallas 1996
Irwin Douglas and Randall Kroszner lsquolsquoInterests Institutions and Ideology inSecuring Policy Change The Republican Conversion to Trade Liberalizationafter Smoot-Hawleyrsquorsquo Journal of Law and Economics XLII (1999) 643ndash673
WHAT DRIVES DEREGULATION 1465
James Christopher lsquolsquoAn Analysis of Intra-Industry Differences in the Effect ofRegulation The Case of Deposit Rate Ceilingsrsquorsquo Journal of Monetary Econom-ics XII (1983) 417ndash442
Jarrell Gregg lsquolsquoChange at the Exchange The Causes and Effects of DeregulationrsquorsquoJournal of Law and Economics XXVII (1984) 273ndash312
Jayaratne Jith and Philip E Strahan lsquolsquoThe Finance-Growth Nexus Evidencefrom Bank Branch Deregulationrsquorsquo Quarterly Journal of Economics CI (1996)639ndash670
Jayaratne Jith and Philip E Strahan lsquolsquoEntry Restrictions Industry Evolutionand Dynamic Efficiency Evidence from Commercial Bankingrsquorsquo Journal of Lawand Economics XLIX (1998) 239ndash274
Joskow Paul lsquolsquoComment on lsquoPower Struggles Explaining Deregulatory Reformsin Electricity Marketsrsquo rsquorsquo Brookings Papers on Economic Activity Microeconom-ics (1996) 251ndash264
Joskow Paul and Roger Noll lsquolsquoRegulation in Theory and Practice An Overviewrsquorsquoin Studies in Public Regulation Gary Fromm ed (Cambridge MA MITPress 1981) pp 1ndash65
Kahn Matthew E and John G Matsusaka lsquolsquoDemand for Environmental GoodsEvidence from Voting Patterns on California Initiativesrsquorsquo Journal of Law andEconomics XL (1997) 137ndash173
Kalbeisch J and R Prentice The Statistical Analysis of Failure Time Data (NewYork John Wiley amp Sons 1980)
Kane Edward J lsquolsquoDe Jure Interstate Banking Why Only NowrsquorsquoJournal of MoneyCredit and Banking XXVIII (1996) 141ndash161
Kiefer Nicholas lsquolsquoEconomic Duration Data and Hazard Functionsrsquorsquo Journal ofEconomic Literature XXVI (1988) 646ndash679
Kroszner Randall S lsquolsquoThe Evolution of Universal Banking and Its Regulation inTwentieth Century Americarsquorsquo in Universal Banking Financial System DesignReconsidered Anthony Saunders and Ingo Walter eds (New York IrwinProfessional Publishers 1996) pp 70ndash99 lsquolsquoThe Political Economy of Banking and Financial Regulation in the U Srsquorsquo inThe Banking and Financial Structure in the NAFTA Countries and ChileGeorge M von Furstenberg ed (Boston Kluwer Academic Publishers 1997)pp 200ndash213
Kroszner Randall S and Raghuram G Rajan lsquolsquoIs the Glass-Steagall ActJustied A Study of the U S Experience with Universal Banking before1933rsquorsquo American Economic Review LXXXIV (1994) 810ndash832
Kroszner Randall S and Raghuram G Rajan lsquolsquoOrganization Structure andCredibility Evidence from Bank Securities Activities before the Glass-Steagall Actrsquorsquo Journal of Monetary Economics XXXIX (1997) 475ndash516
Kroszner Randall S and Philip E Strahan lsquolsquoRegulatory Incentives and the ThriftCrisis Dividends Mutual-to-Stock Conversions and Financial DistressrsquorsquoJournal of Finance LI (1996) 1285ndash1320
Kroszner Randall S and Philip E Strahan lsquolsquoWhat Drives Deregulation Econom-ics and Politics of the Relaxation of Bank Branching Restrictionsrsquorsquo NBERWorking Paper No 6637 1998
Kroszner Randall S and Thomas Stratmann lsquolsquoInterest Group Competition andthe Organization of Congress Theory and Evidence from Financial ServicesPolitical Action Committeesrsquorsquo American Economic Review LXXXVIII (1998)pp 1163ndash1187
Kroszner Randall S and Thomas Stratmann lsquolsquoDoes Political Ambiguity PayCorporate Campaign Contributions and Rewards to Legislator Reputationrsquorsquounpublishedmanuscript University of Chicago Graduate School of Business1999
Lin D Y and L J Wei lsquolsquoThe Robust Inference for the Cox Proportional HazardsModel Journal of the American Statistical Association LXXXIV (1989)1074ndash1078
Macey Jonathan R and Geoffrey P Miller Banking Law and Regulation (BostonLittle Brown and Company 1992)
Makinson Larry Open Secrets The Encyclopedia of Congressional Money andPolitics (Washington DC Congressional Quarterly 1992)
Niskanen WilliamBureaucracy and Representative Government (Chicago Aldine-Atherton 1971)
QUARTERLY JOURNAL OF ECONOMICS1466
Noll Roger lsquolsquoComment on Peltzmanrsquorsquo Brookings Papers Microeconomics SpecialIssue (1989) 48ndash58
North Douglass lsquolsquoA Transactions Cost Theory of Politicsrsquorsquo Journal of TheoreticalPolitics II (1990) 355ndash367
Peltzman Sam lsquolsquoToward a More General Theory of Regulationrsquorsquo Journal of Lawand Economics XIX (1976) 109ndash148 lsquolsquoConstituent Interest and Congressional Votingrsquorsquo Journal of Law andEconomics XXVII (1984) 181ndash210 lsquolsquoThe Economic Theory of Regulation after a Decade of DeregulationrsquorsquoBrookings Papers Microeconomics Special Issue (1989) 1ndash41
Peterson Mitchell and Raghuram Rajan lsquolsquoThe Benets of Lending RelationshipsEvidence from Small Business Datarsquorsquo Journal of Finance XLIX (1994) 3ndash38
Poole Keith and Howard Rosenthal Congress A Political-Economic History ofRoll Call Voting (Oxford Oxford University Press 1997)
Stata Stata Reference Manual Release 5 (College Station TX Stata Press 1997)Stigler George J lsquolsquoThe Theory of Economic Regulationrsquorsquo Bell Journal of Economics
and Management Science II (1971) 3ndash21Stigler George J Chicago Studies in the Political Economy (Chicago University
of Chicago Press 1988)Strahan Philip E and James Weston lsquolsquoSmall Business Lending and the Changing
Structure of the Banking Industryrsquorsquo Journal of Banking and Finance XXII(1998) 821ndash845
Sylla Richard John Legler and John Wallis lsquolsquoBanks and State Public Finance inthe New Republic The United States 1790ndash1860rsquorsquo Journal of EconomicHistory XLVII (1987) 391ndash403
White Eugene The Regulation and Reform of the American Banking System1900ndash1929 (Princeton Princeton University Press 1983)
Winston Clifford lsquolsquoEconomic Deregulation Days of Reckoning for Microecono-mistsrsquorsquo Journal of Economic Literature XXXI (1993) 1263ndash1289
WHAT DRIVES DEREGULATION 1467
subsequent timing of deregulation For instance some statesbegan the period with unit banking while other states allowedlimited branching and the formation of multibank holdingcompanies
To test whether the stringency of initial restrictions onbranching matters we rst divide the sample into states with andwithout the most extreme form of branching regulation unitbanking28 Column (4) of Table III includes an indicator variableequal to one if the state began the period with unit bankingrestrictions The coefficient on this variable is positive andstatistically signicant but the other results remain virtuallyunchanged Next we include an indicator variable that is one ifthe state permitted multibank holding companies (MBHCs) toexist at the beginning of our sample period Thirty-four statesbegan the period allowing MBHCs The coefficient estimate forthis indicator variable (not reported) is small and not statisticallysignicant and its inclusion does not change the effects of theother variables
Eleven states expand bank powers to include the sale ofinsurance products during our sample period This could proxy foromitted factors about the relative strength of the insuranceindustry in the state In column (5) we include an indicatorvariable that is one if the state changes the law during our sampleperiod to permit banks to sell insurance business and zerootherwise The coefficient estimate is small and statisticallyinsignicant and does not affect the other results29
In column (6) we introduce four regional indicator variablesNorth South Midwest and West30 The coefficients on small bankshare the relative capital-to-assets ratio and small rm sharechange little Although the coefficient estimate still suggests that
28 We classify states that prohibited branching but permitted banks toestablish facilities as unit banking states The unit banking indicator is one for thefollowing states CO AR FL IL IA KS MN MO MT NE ND OK TX WI WVand WY
29 We conducted a likelihood ratio test of whether the unit banking statesthe MBHC states and the states that changed their regulations of bank insurancepowers could be pooled with the other states In all three cases we could not rejectpooling of the data
30 Since intrastate deregulation generally preceded interstate deregulationand the latter typically took the form of regional interstate compacts thepossibility of participating in a regional interstate banking compact could haveinuenced the decision to deregulate intrastate branching Our denitions of theregions are region 1 (South) contains AL AR DC FL GA KY LA MS NC OKSC TN TX and VA region 2 (Northeast) contains CT MA MD ME NH NJ NYPA RI VT and WV region 3 (Midwest) contains IA IL IN KS MI MN MO NEND OH SD and WI region 4 (West) contains the other states
WHAT DRIVES DEREGULATION 1455
in states where banks can sell insurance a larger relative size ofthe insurance sector delays deregulation the effect is no longerstatistically signicant The coefficient on the indicator variablefor whether a state expands bank powers to permit insurancesales during the sample period however now becomes statisti-cally signicantly negative This implies that deregulation occursroughly 41 years earlier in states where the insurance lobby maybe relatively weak While the inclusion of the regional indicatorsdoes not affect the other results the results on insurance doappear to be sensitive to the specication of the hazard model31
Estimation Technique To check the robustness of using ahazard model and to investigate whether the time-series or thecross-sectional variation in each of the factors accounts for thehazard results we estimate two simple linear probability models(neither are reported) First we run a pooled time-series cross-section OLS regression using the 637 observations from thehazard model The dependent variable equals one in the year ofderegulation and zero otherwise We include the same set ofregressors that were used in the hazard model and add state-levelxed effects This specication thus removes the cross-statevariation from the independent variables and focuses on theirmovements over time In this model the signs of the coefficientson the interest group factors are consistent with the results fromthe hazard model and the coefficients for the small bank relativecapital ratio and for the relative size of insurance where states cansell insurance are statistically signicant
In the second model we eliminate the time-series variationby averaging each of the explanatory variables over time We thushave one observation per state that had branching restrictions asof 1970 (N 5 39) The dependent variable equals the inverse of thenumber of years until deregulation which represents an estimateof the statersquos probability of deregulating in a year under theassumption that this probability is constant over time and eachyear provides an independent observation32 Although we haveonly 39 observations the coefficients on the small bank marketshare small bank relative capital ratio and small rm share
31 As further robustness check we also tried a population density variablewhich Abrams and Settle [1993] found relevant for regulatory change during the1920s and 1930s but its coefficient was not statistically signicant and did notaffect the other results
32 For instance the dependent variable would take the value of 01 for astate that deregulated in 1980 For the three states that did not deregulate thedependent variable equals 0
QUARTERLY JOURNAL OF ECONOMICS1456
variables are statistically signicant The signs of the effects ofthe insurance variables are the same as those reported in thehazard model but not statistically signicant
The hazard model and linear probability models thus producevery similar results The estimates from the latter suggest thatthe effects for the small bank market share and small rm sharevariables are driven primarily by cross-state variation the effectof the insurance variable is driven primarily by time-seriesvariation and the small bank capital effect is driven by bothcross-sectional and time-series variation
Ex Post Consequences of Deregulation Finally to check theplausibility of our interpretation of the hazard results we con-sider whether the ex post consequences of deregulation areconsistent with the ex ante lobbying positions we attribute to eachinterest group (see Kroszner and Strahan [1998] for details) Wend that small banks lose market share following deregulationand in states where banks can enter the insurance business theinsurance sector shrinks relative to the banking sector followingderegulation Borrowers also benet from deregulation throughlower average interest rates on loans These ndings support theprivate-interest interpretation of the coefficients in the hazardmodel groups that will benet lobby to speed deregulation andthose that will be harmed lobby to slow it
C Voting on Interstate Branching Deregulation in the U SHouse of Representatives
We now examine whether the forces driving intrastate branch-ing deregulation also drive interstate deregulation Financialservices interests are active contributors and lobbyists in Washing-ton Their political action committees constitute the largest groupof contributors to legislators providing nearly 20 percent of totalcongressional campaign contributions [Makinson 1992] and muchof their lobbying effort involves competition among rival interestswithin nancial services (see Kroszner and Stratmann [1998])
As noted above after virtually all states adopted intra- andinterstate branching deregulation the 1994 Riegle-Neal Act re-pealed the 1927 McFadden Act to eliminate all barriers tointerstate banking and branching by 1997 The key votes concern-ing the Riegle-Neal Act were either voice votes or extremelylopsided so we could not estimate a voting model from them Anumber of bills and amendments related to interstate branching
WHAT DRIVES DEREGULATION 1457
had been debated in Congress during the years prior to thepassage of the Riegle-Neal Act but a search of the weekly BNABanking Reporter and the Congressional Record produced onlyone roll-call vote related to interstate branching that was notlopsided This vote occurred in the House on November 14 1991and concerned an amendment sponsored by Wylie (R-OH) andNeal (D-NC) to introduce interstate banking and branchingderegulation to a nancial services reform package33 Althoughthe amendment passed by 210 to 208 the bill to which it wasattached subsequently was defeated
To check for the inuence of the factors we considered in thestate-level reforms we examine both the sponsorship of interstatebanking legislation and voting on the amendment The sponsorsof the Wylie-Neal amendment are from states with low small bankshares 004 in Ohio (Wylie) and 002 in North Carolina (Neal) Incontrast the sample mean in 1991 is 008 (median 5 007)Michigan home state of the Senatersquos sponsor of the 1994 Riegle-Neal Act also had relatively low small bank strength (small bankshare of 005)
Table IV reports estimates of a probit model where thedependent variable equals one if the legislator voted in favor ofthe amendment and zero otherwise The explanatory variablesare the same as those in the hazard model except that we use aparty affiliation indicator variable equal to one for Democrats andzero for Republicans as our proxy for political factors Thecoefficients in Table IV are the marginal effects (lsquolsquoslopesrsquorsquo) of aone-unit change of each variable on the probability that alegislator will vote for the amendment34 Note that a forcefavoring deregulation has a negative coefficient in the hazardmodel (Table III) but a positive coefficient in the probit model
Consistent with the state-level deregulation process legisla-tors are more likely to support the amendment if their states havea relatively low share of small banks As in the hazard model thefraction of small banks is the most important interest groupinuence on a legislatorrsquos voting decision A one-standard-deviation increase in small banksrsquo market share (from the mean)is associated with a decline in the probability of voting in favor of
33 The Wylie-Neal amendment also included provisions limiting certaininsurance and real estate powers of national banks [Congressional RecordNovember 14 1991 pp 10239ndash10242]
34 Since we have multiple legislators from each state we adjust the standarderrors to correct for the lack of independence among observations clustered in thesame state
QUARTERLY JOURNAL OF ECONOMICS1458
TAB
LE
IVM
AR
GIN
AL
EF
FE
CT
SF
RO
MA
PR
OB
ITM
OD
EL
OF
TH
EIN
FL
UE
NC
EO
FP
OL
ITIC
AL
EC
ON
OM
YF
AC
TO
RS
ON
TH
EH
OU
SE
VO
TE
ON
TH
EW
YL
IE-N
EA
L
AM
EN
DM
EN
TT
OP
ER
MIT
INT
ER
ST
AT
EB
AN
KB
RA
NC
HIN
G
NO
VE
MB
ER
14
1991
(1)
(2)
(3)
(4)
(5)
(6)
Smal
lban
kas
set
shar
eof
allb
anki
ng
asse
tsin
stat
e2
417
(12
4)2
454
(12
2)2
407
(12
0)2
443
(14
7)2
450
(14
9)2
490
(14
7)C
apit
alra
tio
ofsm
allb
anks
rela
tive
tola
rge
inth
est
ate
180
(37
5)2
68(3
42)
256
(37
5)1
69(3
78)
153
(38
0)2
45(4
25)
Rel
ativ
esi
zeof
insu
ranc
ein
stat
esw
here
bank
sm
ayse
llin
sura
nce
0ot
herw
ise
22
34
(10
2)2
249
(1
14)
22
33
(10
1)2
249
(1
04)
22
51
(10
2)2
154
(11
1)In
dica
tor
is1
ifba
nks
may
sell
insu
ranc
ein
the
stat
e0
79
(0
06)
074
(00
6)0
69
(0
06)
070
(00
6)0
70
(0
06)
067
(00
7)R
elat
ive
size
ofin
sura
nce
inst
ates
whe
reba
nks
may
not
sell
insu
ranc
e0
oth
erw
ise
180
(06
6)2
42
(0
72)
154
(0
67)
173
(06
6)1
72
(0
67)
203
(0
91)
Smal
lrm
shar
eof
the
num
ber
ofr
ms
inth
est
ate
21
23(0
92)
20
83(0
89)
21
62(1
07)
21
32(0
95)
21
30(1
00)
20
94(1
16)
Indi
cato
ris
1if
Rep
rese
ntat
ive
isa
Dem
ocra
t2
035
(00
7)2
035
(00
7)2
036
(00
7)2
035
(00
7)2
035
(00
7)2
036
(00
7)A
vera
geyi
eld
onba
nklo
ans
inth
est
ate
min
usF
edfu
nds
rate
mdash10
01
(51
5)mdash
mdashmdash
mdash
Ave
rage
fail
ure
rate
198
5to
1991
mdashmdash
228
(14
1)mdash
mdashmdash
Indi
cato
ris
1if
stat
eh
asun
itba
nkin
gla
wmdash
mdashmdash
004
(00
7)0
04(0
07)
20
01(0
08)
Indi
cato
ris
1if
stat
ech
ange
sba
nk
insu
ran
cepo
wer
smdash
mdashmdash
mdash2
001
(00
7)2
003
(00
7)In
clud
esre
gion
alin
dica
tors
N
oN
oN
oN
oN
oYe
sP
seud
o-R
20
170
180
170
170
170
18L
ogli
keli
hood
224
024
223
826
223
951
224
013
224
010
223
808
p-va
lue
ofx2
for
regr
essi
on
001
0
01
001
0
01
001
0
01
Th
ede
pen
den
tvar
iabl
eis
one
ifth
ele
gisl
ator
vote
sfo
rth
eam
endm
enta
nd
zero
ifag
ains
tT
he
repo
rted
coef
fici
ents
are
the
effe
cts
ofa
uni
tch
ange
ofth
ein
depe
nde
ntva
riab
le(f
rom
the
mea
n)on
the
prob
abili
tyof
voti
ng
infa
vor
ofth
eam
endm
ent
All
vari
able
sar
em
easu
red
byst
ate
N5
418
the
num
ber
ofR
epre
sent
ativ
esvo
tin
gon
the
amen
dmen
tS
mal
lban
kas
sets
har
eis
the
perc
ent
ofba
nki
ng
asse
tsin
the
stat
eh
eld
byba
nks
belo
wth
em
edia
nsi
zeof
ban
kin
each
stat
ein
each
year
Rel
ativ
eca
pita
lrat
iois
the
capi
talt
oas
sets
rati
oof
smal
lban
ksm
inu
sth
atof
larg
eba
nks
Siz
eof
insu
ran
cere
lati
veto
bank
ing
plu
sin
sura
nce
inth
est
ate
ism
easu
red
asgr
oss
stat
epr
oduc
tfro
min
sura
nce
divi
ded
bygr
oss
stat
epr
odu
ctfr
omin
sura
nce
plu
sba
nkin
gIn
dica
tor
ison
eif
stat
ela
wpe
rmit
sba
nks
toen
ter
the
insu
ran
cebu
sin
ess
Sm
all
rmsh
are
isth
epe
rcen
tofa
lles
tabl
ishm
ents
inth
est
ate
that
hav
efe
wer
than
twen
tyem
ploy
ees
Par
tyin
dica
tor
ison
eif
the
Rep
rese
nta
tive
isa
Dem
ocra
tA
vera
geyi
eld
onba
nklo
ans
equa
lsto
tali
nter
est
inco
me
onal
ldom
esti
clo
ans
mad
eby
ban
ksin
the
stat
edi
vide
dby
tota
lloa
ns
The
aver
age
bank
failu
rera
teis
the
rati
oof
asse
tsin
faile
dba
nks
duri
ng
agi
ven
stat
edi
vide
dby
tota
lban
kas
sets
inth
atst
ate
aver
aged
over
the
year
s19
85to
1991
Un
itba
nki
ng
indi
cato
req
ual
son
efo
rst
ates
wit
hun
itba
nkin
gre
stri
ctio
nsT
hech
ange
inin
sura
nce
pow
ers
indi
cato
req
ual
son
eif
the
stat
ech
ange
dth
ela
wto
perm
itba
nks
tose
llin
sura
nce
duri
ngth
esa
mpl
epe
riod
We
divi
deth
eco
unt
ryin
tofo
urre
gion
san
din
clu
dere
gion
alin
dica
tors
inco
lum
n(5
)R
obu
stst
anda
rder
rors
are
inpa
ren
thes
es
an
d
den
ote
stat
isti
cally
sign
ica
nta
tth
e1
5an
d10
perc
entl
evel
s
WHAT DRIVES DEREGULATION 1459
branching of approximately 17 percent The relative capital-to-assets ratio of small and large banks however does not have astatistically signicant effect in the probit model
The impact of rival interests outside of banking is alsoconsistent with intrastate deregulation results Where banks cansell insurance legislators from states with larger insurancesectors relative to banking are less likely to vote for the amend-ment35 A one-standard-deviation increase in the relative size ofinsurance in those states which permit banks to sell insurancedecreases the probability that a legislator will favor the amend-ment by about 13 percent Turning to consumer interests thecoefficient on small rm share is not statistically signicant butthe coefficient on interest rates in the state is A one-standard-deviation increase in the average interest rate on loans raises theprobability that a legislator will support the amendment byroughly 8 percent The coefficient estimate for the bankingdistress variable measured as the average bank failure rate from1985 to 1991 is small and statistically signicant Overall theprobit analysis of the vote on national branching deregulationsupports the private-interest theory of deregulation and providesa consistency check that the importance of interests operating onthe state legislatures are very similar to those operating at thefederal level
V WHY DID BRANCHING DEREGULATION BEGIN DURING THE
1970S SHOCKS TO THE EQUILIBRIUM SUPPORTING BANK
BRANCHING RESTRICTIONS
A complete explanation of regulatory exit should account forwhy deregulation begins in the 1970s as well as the timing of thestate-by-state reforms [Kane 1996] In this section we describebroad technological legal and economic shocks that altered thepolitical-economy equilibrium which had kept antibranching regu-lations little changed for at least 30 years
Beginning in the 1970s three innovations reduced the valueto the protected banks of local geographic monopolies (see TableV) First automatic teller machines (ATMs) helped to erode thegeographic ties between customers and banks Second checkablemoney market mutual funds and the Merrill Lynch Cash Manage-ment Account demonstrated that banking by mail and telephone
35 The positive and statistically signicant estimates on the other insurancevariables may reect insurance industry support for the amendmentrsquos provisionslimiting banksrsquo insurance powers
QUARTERLY JOURNAL OF ECONOMICS1460
provided a convenient alternative to local banks36 Third techno-logical innovation and deregulation reduced transportation andcommunication costs particularly since the 1970s thereby lower-ing the costs for customers to use distant banks By increasing theelasticity of deposits supplied to banks these innovations reducedthe value of geographical restrictions to their traditional benecia-ries and thereby reduced their incentive to ght to maintain them[Peltzman 1976]
On the lending side increasing sophistication of credit-scoring techniques following innovations in information process-ing nancial theory and the development of large credit databases began to diminish the value of knowledge that localbankers had about the risks of borrowers in the community As a
36 Regulation Q which limited the interest rates that banks could pay ondeposits may have helped to drive depositors away from banks when the gapbetween market rates and deposit ceilings grew during the 1970s The eliminationof interest rate ceilings on large denomination certicates of deposit during the1970s appears to have hurt smaller and retail-oriented banks relative to largerwholesale banks [James 1983]
TABLE VBROAD TRENDS IN COMMERCIAL BANKING 1950ndash1995
YearNumberof ATMs
Domesticbank
deposits(billions)
Moneymarketmutual
fund(billions)
Percent ofdeposits 1
moneyfunds heldby banks
Smallbanksrsquo
percent ofbankingassets
Averagenumber of
bankfailures
(1) (2) (3) (4) (5) (6)1950 0 $ 154 $ 0 100 NA 41955 0 191 0 100 NA 31960 0 228 0 100 24 21965 0 330 0 100 20 41970 0 479 0 100 18 61975 9750 775 4 99 18 61980 18500 1182 76 94 17 101985 61117 1787 242 88 14 601990 80156 2339 493 83 11 1791995 122706 2552 745 77 8 61
Column (1) ATM gures are from Bank Network News The EFT Network Data Book (New YorkFaulkner and Gray Inc) The 1975 gure was unavailable 9750 is the number of ATMs in 1978 the rst yearfor which complete data are available Columns (2)ndash(4) Banksrsquo domestic deposits are from the Reports ofIncome and Condition money market mutual funds are from the Flow of Funds Data on all bank depositsforeign plus domestic are only available beginning in 1970 The trend in banksrsquo share (column (4)) is the sameusing total deposits instead of domestic deposits Column (5) Percent of banking assets held by small bankswhere a small bank is dened as a commercial bank with less than $100 million in assets in 1994 dollarsThese data are based on the Reports of Income and Condition Data on small banks are not available before1960 Column (6) Five-year average number of bank failures where the nal year is indicated in the rstcolumn These data are from FDIC Annual Report and the Quarterly Banking Prole
WHAT DRIVES DEREGULATION 1461
result of these innovations a national market developed forresidential mortgages credit card receivables have been securi-tized and bank lending to small business now relies less on thejudgment of loan officers and more on standardized scoringmodels
These changes have increased the potential protability forlarge banks to enter what had been the core of small bankactivities Large banksrsquo incentive to increase their lobbying pres-sure to be able to expand into these markets has thus beenincreasing over time In fact small banksrsquo market share began todecline even prior to the branching deregulation (Table V) As thevalue of a local banking relationships declined small rms thatwere the main borrowers from the small banks also probablybecame more likely to favor the entry of large banks into localmarkets With the deadweight costs of preventing large bankentry rising the private-interest theory predicts that small localbanks would become less likely able to maintain the branchingrestrictions [Becker 1983] Deregulation that reduces deadweightcosts of regulation is also consistent with the public-interesttheory
Kane [1996] argues that another major shock to the oldequilibrium was the increasing public awareness of the costlinessof having government-insured but (geographically) undiversiednancial institutions In the late 1970s the failure rate of banksbegan to rise (Table V) In the 1980s the Savings and Loan crisisand taxpayer bailout further heightened public awareness aboutthe costs of restrictions that make depository institutions morelikely to require infusions of taxpayer funds The failures thusmay have heightened public support for branching deregulation(see also Abrams and Settle [1993]) As noted above we did notnd any evidence that banking failures or distress in a stateaffected the speed with which the state deregulated but distressstill may be a common factor affecting all states
These technological economic and legal shocks generatedconditions that changed the long-standing balance favoring theantibranching forces The marginal value of lobbying to repealbranching restrictions increased just as the relative value to thesmall banks of maintaining branching restrictions was decliningThese nationwide shocks are common factors across states andthey are consistent with the positive duration dependence associ-ated with state-level deregulation that we demonstrated in FigureII In addition a broad change in lsquolsquoideologyrsquorsquo against government
QUARTERLY JOURNAL OF ECONOMICS1462
regulation could explain the positive duration dependence duringthe sample period
VI CONCLUSIONS
The private-interest theory of regulation can account for thepattern of bank branching deregulation of the 1970s and 1980sBeneciaries of branching regulation had supported a coalitionfavoring geographical restrictions despite their costs to consum-ers of nancial services Innovations that began in the 1970saltered the value of the restrictions to the affected parties and theresulting competition among interest groups can explain thesubsequent deregulation While some of our results are alsoconsistent with the public interest theorymdashfor example deregula-tion occurs earlier when small banks are in a relatively weaknancial positionmdashother results particularly evidence on theimportance of rivalries between small and large banks andbetween banking and insurance are difficult to explain with thepublic-interest approach We also nd that political-institutionalfactors affect deregulation although these variables may act asproxies for unmeasured economic interests The greater success ofthe private-interest theory to explain deregulation here than inprevious work may be due to the richer cross-sectional andtime-series variation in branching deregulation that can beexploited with a hazard model Future empirical studies ofendogenous deregulation thus may be most fruitful where changehas occurred across states over time such as in franchisinginsurance and public utilities (eg Joskow [1996])
Technological and nancial innovations will continue to erodethe benets to any interest group of maintaining regulatorybarriers in nancial services These forces are likely to bringabout reforms both domestically for example through legislationthat would increase bank powers (see Kroszner [1996] Krosznerand Rajan [1994 1997] and Kroszner and Stratmann [1998]) andinternationally for example through the extension of nancialservices provisions of NAFTAto reduce geographic barriers acrosscountries [Kroszner 1997]
APPENDIX EFFECTS OF DIFFERENT TYPES OF BRANCHING
DEREGULATION ON ECONOMIC GROWTH BANKING STRUCTUREAND BANK EFFICIENCY
We measure the importance of alternative forms of deregula-tion by their consequences for the economic growth and the
WHAT DRIVES DEREGULATION 1463
structure and efficiency of the banking sector across the states(see Jayaratne and Strahan [1996 1998]) To do so we run anOLS panel regression relating indicators for each type of deregula-tion to potential consequences Our dependent variables includeproxies for economic growth (growth in personal income) bankstructure (the log of the number of banking companies and thenumber of branches per banking company) and efficiency (the logof total noninterest expenses and the ratio of loan loss provisionsto total loans) The four deregulation indicators (MBHC branch-ing through merger and acquisition unrestricted statewide branch-ing and interstate banking) are equal to one in the years afterderegulation and zero before If a state deregulated before thesample period begins in 1970 the indicator is one throughout ifthe state does not deregulate then the indicator is zero through-out All regressions include state xed effects and time effects
Table VI shows the results of including the set of four
TABLE VIMULTIVARIATE REGRESSIONS OF THE EFFECTS OF DIFFERENT TYPES OF BRANCHING
DEREGULATION ON ECONOMIC GROWTH BANKING STRUCTUREAND BANK EFFICIENCY
Dependent variable
Growthin
personalincome
Log of of
bankingcompanies
Branchesper
bankingcompany
Log ofnon-
interestexpenses
Loan lossprovisions
(loans 1leases)
Merger amp acquisitionbranchingindicator
0013(0004)
2 0162(0037)
1146(0453)
2 0107(0035)
2 0007(0001)
Unrestricted state-wide branchingindicator
2 0008(0005)
0086(0059)
2 0043(0574)
0070(0037)
0005(0002)
Interstate bankingindicator
0006(0005)
2 0052(0057)
1064(0511)
0005(0030)
2 0001(0001)
Multibank holdingcompany indicator
2 0005(0006)
2 0096(0052)
2 0364(0529)
2 0095(0036)
2 0001(0001)
Dependent variablemean [median]
0086[0082]
4881[5136]
6649[5055]
13812[13803]
0009[0006]
This table contains regressions of state-level economic growth and state-level measures of the structureand efficiency of the banking industry Each dependent variable is regressed on a set of state indicatorvariables a set of year indicator variables and the set of four deregulation indicators these deregulationindicators are equal to one after deregulation and zero before If a state was always regulated the indicator iszero throughout if it was always deregulated the indicator is one throughout Sample period is 1975 to 1991and Delaware and South Dakota are dropped so N 5 833 Noninterest expenses become available only in1984 so N 5 392 The coefficient on the deregulation indicators is reported with robust standard errors inparentheses and denote statistically signicant at the 1 5 and 10 percent levels
QUARTERLY JOURNAL OF ECONOMICS1464
deregulation indicators simultaneously The branching by mergerand acquisition indicator is statistically signicant and of theexpected sign in all ve of the equations and larger in absolutevalue than any of the other coefficients on the indicators for theother types of deregulation Since there may be multicollinearityamong our indicators we also ran univariate specications whichincluded each indicator separately and the results are verysimilar These results imply that branching by merger andacquisition is the most important type of deregulation measuredby its consequences for economic growth banking structure andbank efficiency See Kroszner and Strahan [1998] for more details
UNIVERSITY OF CHICAGO GRADUATE SCHOOL OF BUSINESS AND NATIONAL BUREAU OF
ECONOMIC RESEARCH
FEDERAL RESERVE BANK OF NEW YORK RESEARCH amp MARKET ANALYSIS GROUP
REFERENCES
Abrams Burton A and Russell F Settle lsquolsquoPressure-Group Inuence and Institu-tional Change Branch-Banking Legislation during the Great DepressionrsquorsquoPublic Choice LXXVII (1993) 687ndash705
Amel Dean lsquolsquoState Laws Affecting the Geographic Expansion of CommercialBanksrsquorsquo unpublished manuscript Board of Governors of the Federal ReserveSystem 1993
Becker Gary S lsquolsquoA Theory of Competition among Pressure Groups for PoliticalInuencersquorsquo Quarterly Journal of Economics XCVIII (1983) 371ndash400
Berger Allen Anil Kashyap Joseph Scalise lsquolsquoThe Transformation of the U SBanking Industry What a Long Strange Trip Itrsquos Beenrsquorsquo Brookings Papers onEconomic Activity 2 (1995) 55ndash218
Berger Allen Anthony Saunders Joseph Scalise and Gregory Udell lsquolsquoThe Effectsof Bank Mergers and Acquisitions on Small Business Lendingrsquorsquo Journal ofFinancial Economics L (1998) 187ndash229
Brennan Geoffrey and James Buchanan lsquolsquoTowards a Tax Constitution forLeviathanrsquorsquo Journal of Public Economics VIII (1977) 255ndash273
Brickley James A and Christopher M James lsquolsquoThe Takeover Market CorporateBoard Composition and Ownership Structure The Case of Bankingrsquorsquo Journalof Law and Economics XXX (1987) 161ndash180
Cole Rebel and John Wolken lsquolsquoFinancial Services Used by Small BusinessesEvidence from the 1993 Survey of Small Business Financesrsquorsquo Federal ReserveBulletin LXXXI (1994) 629ndash676
Dixit Avinash The Making of Economic Policy A Transactions-Cost PoliticsPerspective (Cambridge MA MIT Press 1996)
Economides Nicolas R Glenn Hubbard and Darius Palia lsquolsquoThe Political Economyof Branching Restrictions and Deposit Insurancersquorsquo Journal of Law andEconomics XXIX (1996) 667ndash704
Flannery Mark lsquolsquoThe Social Costs of Unit Banking Restrictionsrsquorsquo Journal ofMonetary Economics XIII (1984) 237ndash249
Greene William H Econometric Analysis third edition (Upper Saddle River NJPrentice-Hall 1997)
Gunther Jeffery lsquolsquoRegional Capital Imbalances and the Removal of InterstateBanking Restrictionsrsquorsquo Economics Letters XLIV (1994) 439ndash442 lsquolsquoThe Erosion of Intrastate Branch Banking Restrictionsrsquorsquo unpublishedmanuscript Federal Reserve Bank of Dallas 1996
Irwin Douglas and Randall Kroszner lsquolsquoInterests Institutions and Ideology inSecuring Policy Change The Republican Conversion to Trade Liberalizationafter Smoot-Hawleyrsquorsquo Journal of Law and Economics XLII (1999) 643ndash673
WHAT DRIVES DEREGULATION 1465
James Christopher lsquolsquoAn Analysis of Intra-Industry Differences in the Effect ofRegulation The Case of Deposit Rate Ceilingsrsquorsquo Journal of Monetary Econom-ics XII (1983) 417ndash442
Jarrell Gregg lsquolsquoChange at the Exchange The Causes and Effects of DeregulationrsquorsquoJournal of Law and Economics XXVII (1984) 273ndash312
Jayaratne Jith and Philip E Strahan lsquolsquoThe Finance-Growth Nexus Evidencefrom Bank Branch Deregulationrsquorsquo Quarterly Journal of Economics CI (1996)639ndash670
Jayaratne Jith and Philip E Strahan lsquolsquoEntry Restrictions Industry Evolutionand Dynamic Efficiency Evidence from Commercial Bankingrsquorsquo Journal of Lawand Economics XLIX (1998) 239ndash274
Joskow Paul lsquolsquoComment on lsquoPower Struggles Explaining Deregulatory Reformsin Electricity Marketsrsquo rsquorsquo Brookings Papers on Economic Activity Microeconom-ics (1996) 251ndash264
Joskow Paul and Roger Noll lsquolsquoRegulation in Theory and Practice An Overviewrsquorsquoin Studies in Public Regulation Gary Fromm ed (Cambridge MA MITPress 1981) pp 1ndash65
Kahn Matthew E and John G Matsusaka lsquolsquoDemand for Environmental GoodsEvidence from Voting Patterns on California Initiativesrsquorsquo Journal of Law andEconomics XL (1997) 137ndash173
Kalbeisch J and R Prentice The Statistical Analysis of Failure Time Data (NewYork John Wiley amp Sons 1980)
Kane Edward J lsquolsquoDe Jure Interstate Banking Why Only NowrsquorsquoJournal of MoneyCredit and Banking XXVIII (1996) 141ndash161
Kiefer Nicholas lsquolsquoEconomic Duration Data and Hazard Functionsrsquorsquo Journal ofEconomic Literature XXVI (1988) 646ndash679
Kroszner Randall S lsquolsquoThe Evolution of Universal Banking and Its Regulation inTwentieth Century Americarsquorsquo in Universal Banking Financial System DesignReconsidered Anthony Saunders and Ingo Walter eds (New York IrwinProfessional Publishers 1996) pp 70ndash99 lsquolsquoThe Political Economy of Banking and Financial Regulation in the U Srsquorsquo inThe Banking and Financial Structure in the NAFTA Countries and ChileGeorge M von Furstenberg ed (Boston Kluwer Academic Publishers 1997)pp 200ndash213
Kroszner Randall S and Raghuram G Rajan lsquolsquoIs the Glass-Steagall ActJustied A Study of the U S Experience with Universal Banking before1933rsquorsquo American Economic Review LXXXIV (1994) 810ndash832
Kroszner Randall S and Raghuram G Rajan lsquolsquoOrganization Structure andCredibility Evidence from Bank Securities Activities before the Glass-Steagall Actrsquorsquo Journal of Monetary Economics XXXIX (1997) 475ndash516
Kroszner Randall S and Philip E Strahan lsquolsquoRegulatory Incentives and the ThriftCrisis Dividends Mutual-to-Stock Conversions and Financial DistressrsquorsquoJournal of Finance LI (1996) 1285ndash1320
Kroszner Randall S and Philip E Strahan lsquolsquoWhat Drives Deregulation Econom-ics and Politics of the Relaxation of Bank Branching Restrictionsrsquorsquo NBERWorking Paper No 6637 1998
Kroszner Randall S and Thomas Stratmann lsquolsquoInterest Group Competition andthe Organization of Congress Theory and Evidence from Financial ServicesPolitical Action Committeesrsquorsquo American Economic Review LXXXVIII (1998)pp 1163ndash1187
Kroszner Randall S and Thomas Stratmann lsquolsquoDoes Political Ambiguity PayCorporate Campaign Contributions and Rewards to Legislator Reputationrsquorsquounpublishedmanuscript University of Chicago Graduate School of Business1999
Lin D Y and L J Wei lsquolsquoThe Robust Inference for the Cox Proportional HazardsModel Journal of the American Statistical Association LXXXIV (1989)1074ndash1078
Macey Jonathan R and Geoffrey P Miller Banking Law and Regulation (BostonLittle Brown and Company 1992)
Makinson Larry Open Secrets The Encyclopedia of Congressional Money andPolitics (Washington DC Congressional Quarterly 1992)
Niskanen WilliamBureaucracy and Representative Government (Chicago Aldine-Atherton 1971)
QUARTERLY JOURNAL OF ECONOMICS1466
Noll Roger lsquolsquoComment on Peltzmanrsquorsquo Brookings Papers Microeconomics SpecialIssue (1989) 48ndash58
North Douglass lsquolsquoA Transactions Cost Theory of Politicsrsquorsquo Journal of TheoreticalPolitics II (1990) 355ndash367
Peltzman Sam lsquolsquoToward a More General Theory of Regulationrsquorsquo Journal of Lawand Economics XIX (1976) 109ndash148 lsquolsquoConstituent Interest and Congressional Votingrsquorsquo Journal of Law andEconomics XXVII (1984) 181ndash210 lsquolsquoThe Economic Theory of Regulation after a Decade of DeregulationrsquorsquoBrookings Papers Microeconomics Special Issue (1989) 1ndash41
Peterson Mitchell and Raghuram Rajan lsquolsquoThe Benets of Lending RelationshipsEvidence from Small Business Datarsquorsquo Journal of Finance XLIX (1994) 3ndash38
Poole Keith and Howard Rosenthal Congress A Political-Economic History ofRoll Call Voting (Oxford Oxford University Press 1997)
Stata Stata Reference Manual Release 5 (College Station TX Stata Press 1997)Stigler George J lsquolsquoThe Theory of Economic Regulationrsquorsquo Bell Journal of Economics
and Management Science II (1971) 3ndash21Stigler George J Chicago Studies in the Political Economy (Chicago University
of Chicago Press 1988)Strahan Philip E and James Weston lsquolsquoSmall Business Lending and the Changing
Structure of the Banking Industryrsquorsquo Journal of Banking and Finance XXII(1998) 821ndash845
Sylla Richard John Legler and John Wallis lsquolsquoBanks and State Public Finance inthe New Republic The United States 1790ndash1860rsquorsquo Journal of EconomicHistory XLVII (1987) 391ndash403
White Eugene The Regulation and Reform of the American Banking System1900ndash1929 (Princeton Princeton University Press 1983)
Winston Clifford lsquolsquoEconomic Deregulation Days of Reckoning for Microecono-mistsrsquorsquo Journal of Economic Literature XXXI (1993) 1263ndash1289
WHAT DRIVES DEREGULATION 1467
in states where banks can sell insurance a larger relative size ofthe insurance sector delays deregulation the effect is no longerstatistically signicant The coefficient on the indicator variablefor whether a state expands bank powers to permit insurancesales during the sample period however now becomes statisti-cally signicantly negative This implies that deregulation occursroughly 41 years earlier in states where the insurance lobby maybe relatively weak While the inclusion of the regional indicatorsdoes not affect the other results the results on insurance doappear to be sensitive to the specication of the hazard model31
Estimation Technique To check the robustness of using ahazard model and to investigate whether the time-series or thecross-sectional variation in each of the factors accounts for thehazard results we estimate two simple linear probability models(neither are reported) First we run a pooled time-series cross-section OLS regression using the 637 observations from thehazard model The dependent variable equals one in the year ofderegulation and zero otherwise We include the same set ofregressors that were used in the hazard model and add state-levelxed effects This specication thus removes the cross-statevariation from the independent variables and focuses on theirmovements over time In this model the signs of the coefficientson the interest group factors are consistent with the results fromthe hazard model and the coefficients for the small bank relativecapital ratio and for the relative size of insurance where states cansell insurance are statistically signicant
In the second model we eliminate the time-series variationby averaging each of the explanatory variables over time We thushave one observation per state that had branching restrictions asof 1970 (N 5 39) The dependent variable equals the inverse of thenumber of years until deregulation which represents an estimateof the statersquos probability of deregulating in a year under theassumption that this probability is constant over time and eachyear provides an independent observation32 Although we haveonly 39 observations the coefficients on the small bank marketshare small bank relative capital ratio and small rm share
31 As further robustness check we also tried a population density variablewhich Abrams and Settle [1993] found relevant for regulatory change during the1920s and 1930s but its coefficient was not statistically signicant and did notaffect the other results
32 For instance the dependent variable would take the value of 01 for astate that deregulated in 1980 For the three states that did not deregulate thedependent variable equals 0
QUARTERLY JOURNAL OF ECONOMICS1456
variables are statistically signicant The signs of the effects ofthe insurance variables are the same as those reported in thehazard model but not statistically signicant
The hazard model and linear probability models thus producevery similar results The estimates from the latter suggest thatthe effects for the small bank market share and small rm sharevariables are driven primarily by cross-state variation the effectof the insurance variable is driven primarily by time-seriesvariation and the small bank capital effect is driven by bothcross-sectional and time-series variation
Ex Post Consequences of Deregulation Finally to check theplausibility of our interpretation of the hazard results we con-sider whether the ex post consequences of deregulation areconsistent with the ex ante lobbying positions we attribute to eachinterest group (see Kroszner and Strahan [1998] for details) Wend that small banks lose market share following deregulationand in states where banks can enter the insurance business theinsurance sector shrinks relative to the banking sector followingderegulation Borrowers also benet from deregulation throughlower average interest rates on loans These ndings support theprivate-interest interpretation of the coefficients in the hazardmodel groups that will benet lobby to speed deregulation andthose that will be harmed lobby to slow it
C Voting on Interstate Branching Deregulation in the U SHouse of Representatives
We now examine whether the forces driving intrastate branch-ing deregulation also drive interstate deregulation Financialservices interests are active contributors and lobbyists in Washing-ton Their political action committees constitute the largest groupof contributors to legislators providing nearly 20 percent of totalcongressional campaign contributions [Makinson 1992] and muchof their lobbying effort involves competition among rival interestswithin nancial services (see Kroszner and Stratmann [1998])
As noted above after virtually all states adopted intra- andinterstate branching deregulation the 1994 Riegle-Neal Act re-pealed the 1927 McFadden Act to eliminate all barriers tointerstate banking and branching by 1997 The key votes concern-ing the Riegle-Neal Act were either voice votes or extremelylopsided so we could not estimate a voting model from them Anumber of bills and amendments related to interstate branching
WHAT DRIVES DEREGULATION 1457
had been debated in Congress during the years prior to thepassage of the Riegle-Neal Act but a search of the weekly BNABanking Reporter and the Congressional Record produced onlyone roll-call vote related to interstate branching that was notlopsided This vote occurred in the House on November 14 1991and concerned an amendment sponsored by Wylie (R-OH) andNeal (D-NC) to introduce interstate banking and branchingderegulation to a nancial services reform package33 Althoughthe amendment passed by 210 to 208 the bill to which it wasattached subsequently was defeated
To check for the inuence of the factors we considered in thestate-level reforms we examine both the sponsorship of interstatebanking legislation and voting on the amendment The sponsorsof the Wylie-Neal amendment are from states with low small bankshares 004 in Ohio (Wylie) and 002 in North Carolina (Neal) Incontrast the sample mean in 1991 is 008 (median 5 007)Michigan home state of the Senatersquos sponsor of the 1994 Riegle-Neal Act also had relatively low small bank strength (small bankshare of 005)
Table IV reports estimates of a probit model where thedependent variable equals one if the legislator voted in favor ofthe amendment and zero otherwise The explanatory variablesare the same as those in the hazard model except that we use aparty affiliation indicator variable equal to one for Democrats andzero for Republicans as our proxy for political factors Thecoefficients in Table IV are the marginal effects (lsquolsquoslopesrsquorsquo) of aone-unit change of each variable on the probability that alegislator will vote for the amendment34 Note that a forcefavoring deregulation has a negative coefficient in the hazardmodel (Table III) but a positive coefficient in the probit model
Consistent with the state-level deregulation process legisla-tors are more likely to support the amendment if their states havea relatively low share of small banks As in the hazard model thefraction of small banks is the most important interest groupinuence on a legislatorrsquos voting decision A one-standard-deviation increase in small banksrsquo market share (from the mean)is associated with a decline in the probability of voting in favor of
33 The Wylie-Neal amendment also included provisions limiting certaininsurance and real estate powers of national banks [Congressional RecordNovember 14 1991 pp 10239ndash10242]
34 Since we have multiple legislators from each state we adjust the standarderrors to correct for the lack of independence among observations clustered in thesame state
QUARTERLY JOURNAL OF ECONOMICS1458
TAB
LE
IVM
AR
GIN
AL
EF
FE
CT
SF
RO
MA
PR
OB
ITM
OD
EL
OF
TH
EIN
FL
UE
NC
EO
FP
OL
ITIC
AL
EC
ON
OM
YF
AC
TO
RS
ON
TH
EH
OU
SE
VO
TE
ON
TH
EW
YL
IE-N
EA
L
AM
EN
DM
EN
TT
OP
ER
MIT
INT
ER
ST
AT
EB
AN
KB
RA
NC
HIN
G
NO
VE
MB
ER
14
1991
(1)
(2)
(3)
(4)
(5)
(6)
Smal
lban
kas
set
shar
eof
allb
anki
ng
asse
tsin
stat
e2
417
(12
4)2
454
(12
2)2
407
(12
0)2
443
(14
7)2
450
(14
9)2
490
(14
7)C
apit
alra
tio
ofsm
allb
anks
rela
tive
tola
rge
inth
est
ate
180
(37
5)2
68(3
42)
256
(37
5)1
69(3
78)
153
(38
0)2
45(4
25)
Rel
ativ
esi
zeof
insu
ranc
ein
stat
esw
here
bank
sm
ayse
llin
sura
nce
0ot
herw
ise
22
34
(10
2)2
249
(1
14)
22
33
(10
1)2
249
(1
04)
22
51
(10
2)2
154
(11
1)In
dica
tor
is1
ifba
nks
may
sell
insu
ranc
ein
the
stat
e0
79
(0
06)
074
(00
6)0
69
(0
06)
070
(00
6)0
70
(0
06)
067
(00
7)R
elat
ive
size
ofin
sura
nce
inst
ates
whe
reba
nks
may
not
sell
insu
ranc
e0
oth
erw
ise
180
(06
6)2
42
(0
72)
154
(0
67)
173
(06
6)1
72
(0
67)
203
(0
91)
Smal
lrm
shar
eof
the
num
ber
ofr
ms
inth
est
ate
21
23(0
92)
20
83(0
89)
21
62(1
07)
21
32(0
95)
21
30(1
00)
20
94(1
16)
Indi
cato
ris
1if
Rep
rese
ntat
ive
isa
Dem
ocra
t2
035
(00
7)2
035
(00
7)2
036
(00
7)2
035
(00
7)2
035
(00
7)2
036
(00
7)A
vera
geyi
eld
onba
nklo
ans
inth
est
ate
min
usF
edfu
nds
rate
mdash10
01
(51
5)mdash
mdashmdash
mdash
Ave
rage
fail
ure
rate
198
5to
1991
mdashmdash
228
(14
1)mdash
mdashmdash
Indi
cato
ris
1if
stat
eh
asun
itba
nkin
gla
wmdash
mdashmdash
004
(00
7)0
04(0
07)
20
01(0
08)
Indi
cato
ris
1if
stat
ech
ange
sba
nk
insu
ran
cepo
wer
smdash
mdashmdash
mdash2
001
(00
7)2
003
(00
7)In
clud
esre
gion
alin
dica
tors
N
oN
oN
oN
oN
oYe
sP
seud
o-R
20
170
180
170
170
170
18L
ogli
keli
hood
224
024
223
826
223
951
224
013
224
010
223
808
p-va
lue
ofx2
for
regr
essi
on
001
0
01
001
0
01
001
0
01
Th
ede
pen
den
tvar
iabl
eis
one
ifth
ele
gisl
ator
vote
sfo
rth
eam
endm
enta
nd
zero
ifag
ains
tT
he
repo
rted
coef
fici
ents
are
the
effe
cts
ofa
uni
tch
ange
ofth
ein
depe
nde
ntva
riab
le(f
rom
the
mea
n)on
the
prob
abili
tyof
voti
ng
infa
vor
ofth
eam
endm
ent
All
vari
able
sar
em
easu
red
byst
ate
N5
418
the
num
ber
ofR
epre
sent
ativ
esvo
tin
gon
the
amen
dmen
tS
mal
lban
kas
sets
har
eis
the
perc
ent
ofba
nki
ng
asse
tsin
the
stat
eh
eld
byba
nks
belo
wth
em
edia
nsi
zeof
ban
kin
each
stat
ein
each
year
Rel
ativ
eca
pita
lrat
iois
the
capi
talt
oas
sets
rati
oof
smal
lban
ksm
inu
sth
atof
larg
eba
nks
Siz
eof
insu
ran
cere
lati
veto
bank
ing
plu
sin
sura
nce
inth
est
ate
ism
easu
red
asgr
oss
stat
epr
oduc
tfro
min
sura
nce
divi
ded
bygr
oss
stat
epr
odu
ctfr
omin
sura
nce
plu
sba
nkin
gIn
dica
tor
ison
eif
stat
ela
wpe
rmit
sba
nks
toen
ter
the
insu
ran
cebu
sin
ess
Sm
all
rmsh
are
isth
epe
rcen
tofa
lles
tabl
ishm
ents
inth
est
ate
that
hav
efe
wer
than
twen
tyem
ploy
ees
Par
tyin
dica
tor
ison
eif
the
Rep
rese
nta
tive
isa
Dem
ocra
tA
vera
geyi
eld
onba
nklo
ans
equa
lsto
tali
nter
est
inco
me
onal
ldom
esti
clo
ans
mad
eby
ban
ksin
the
stat
edi
vide
dby
tota
lloa
ns
The
aver
age
bank
failu
rera
teis
the
rati
oof
asse
tsin
faile
dba
nks
duri
ng
agi
ven
stat
edi
vide
dby
tota
lban
kas
sets
inth
atst
ate
aver
aged
over
the
year
s19
85to
1991
Un
itba
nki
ng
indi
cato
req
ual
son
efo
rst
ates
wit
hun
itba
nkin
gre
stri
ctio
nsT
hech
ange
inin
sura
nce
pow
ers
indi
cato
req
ual
son
eif
the
stat
ech
ange
dth
ela
wto
perm
itba
nks
tose
llin
sura
nce
duri
ngth
esa
mpl
epe
riod
We
divi
deth
eco
unt
ryin
tofo
urre
gion
san
din
clu
dere
gion
alin
dica
tors
inco
lum
n(5
)R
obu
stst
anda
rder
rors
are
inpa
ren
thes
es
an
d
den
ote
stat
isti
cally
sign
ica
nta
tth
e1
5an
d10
perc
entl
evel
s
WHAT DRIVES DEREGULATION 1459
branching of approximately 17 percent The relative capital-to-assets ratio of small and large banks however does not have astatistically signicant effect in the probit model
The impact of rival interests outside of banking is alsoconsistent with intrastate deregulation results Where banks cansell insurance legislators from states with larger insurancesectors relative to banking are less likely to vote for the amend-ment35 A one-standard-deviation increase in the relative size ofinsurance in those states which permit banks to sell insurancedecreases the probability that a legislator will favor the amend-ment by about 13 percent Turning to consumer interests thecoefficient on small rm share is not statistically signicant butthe coefficient on interest rates in the state is A one-standard-deviation increase in the average interest rate on loans raises theprobability that a legislator will support the amendment byroughly 8 percent The coefficient estimate for the bankingdistress variable measured as the average bank failure rate from1985 to 1991 is small and statistically signicant Overall theprobit analysis of the vote on national branching deregulationsupports the private-interest theory of deregulation and providesa consistency check that the importance of interests operating onthe state legislatures are very similar to those operating at thefederal level
V WHY DID BRANCHING DEREGULATION BEGIN DURING THE
1970S SHOCKS TO THE EQUILIBRIUM SUPPORTING BANK
BRANCHING RESTRICTIONS
A complete explanation of regulatory exit should account forwhy deregulation begins in the 1970s as well as the timing of thestate-by-state reforms [Kane 1996] In this section we describebroad technological legal and economic shocks that altered thepolitical-economy equilibrium which had kept antibranching regu-lations little changed for at least 30 years
Beginning in the 1970s three innovations reduced the valueto the protected banks of local geographic monopolies (see TableV) First automatic teller machines (ATMs) helped to erode thegeographic ties between customers and banks Second checkablemoney market mutual funds and the Merrill Lynch Cash Manage-ment Account demonstrated that banking by mail and telephone
35 The positive and statistically signicant estimates on the other insurancevariables may reect insurance industry support for the amendmentrsquos provisionslimiting banksrsquo insurance powers
QUARTERLY JOURNAL OF ECONOMICS1460
provided a convenient alternative to local banks36 Third techno-logical innovation and deregulation reduced transportation andcommunication costs particularly since the 1970s thereby lower-ing the costs for customers to use distant banks By increasing theelasticity of deposits supplied to banks these innovations reducedthe value of geographical restrictions to their traditional benecia-ries and thereby reduced their incentive to ght to maintain them[Peltzman 1976]
On the lending side increasing sophistication of credit-scoring techniques following innovations in information process-ing nancial theory and the development of large credit databases began to diminish the value of knowledge that localbankers had about the risks of borrowers in the community As a
36 Regulation Q which limited the interest rates that banks could pay ondeposits may have helped to drive depositors away from banks when the gapbetween market rates and deposit ceilings grew during the 1970s The eliminationof interest rate ceilings on large denomination certicates of deposit during the1970s appears to have hurt smaller and retail-oriented banks relative to largerwholesale banks [James 1983]
TABLE VBROAD TRENDS IN COMMERCIAL BANKING 1950ndash1995
YearNumberof ATMs
Domesticbank
deposits(billions)
Moneymarketmutual
fund(billions)
Percent ofdeposits 1
moneyfunds heldby banks
Smallbanksrsquo
percent ofbankingassets
Averagenumber of
bankfailures
(1) (2) (3) (4) (5) (6)1950 0 $ 154 $ 0 100 NA 41955 0 191 0 100 NA 31960 0 228 0 100 24 21965 0 330 0 100 20 41970 0 479 0 100 18 61975 9750 775 4 99 18 61980 18500 1182 76 94 17 101985 61117 1787 242 88 14 601990 80156 2339 493 83 11 1791995 122706 2552 745 77 8 61
Column (1) ATM gures are from Bank Network News The EFT Network Data Book (New YorkFaulkner and Gray Inc) The 1975 gure was unavailable 9750 is the number of ATMs in 1978 the rst yearfor which complete data are available Columns (2)ndash(4) Banksrsquo domestic deposits are from the Reports ofIncome and Condition money market mutual funds are from the Flow of Funds Data on all bank depositsforeign plus domestic are only available beginning in 1970 The trend in banksrsquo share (column (4)) is the sameusing total deposits instead of domestic deposits Column (5) Percent of banking assets held by small bankswhere a small bank is dened as a commercial bank with less than $100 million in assets in 1994 dollarsThese data are based on the Reports of Income and Condition Data on small banks are not available before1960 Column (6) Five-year average number of bank failures where the nal year is indicated in the rstcolumn These data are from FDIC Annual Report and the Quarterly Banking Prole
WHAT DRIVES DEREGULATION 1461
result of these innovations a national market developed forresidential mortgages credit card receivables have been securi-tized and bank lending to small business now relies less on thejudgment of loan officers and more on standardized scoringmodels
These changes have increased the potential protability forlarge banks to enter what had been the core of small bankactivities Large banksrsquo incentive to increase their lobbying pres-sure to be able to expand into these markets has thus beenincreasing over time In fact small banksrsquo market share began todecline even prior to the branching deregulation (Table V) As thevalue of a local banking relationships declined small rms thatwere the main borrowers from the small banks also probablybecame more likely to favor the entry of large banks into localmarkets With the deadweight costs of preventing large bankentry rising the private-interest theory predicts that small localbanks would become less likely able to maintain the branchingrestrictions [Becker 1983] Deregulation that reduces deadweightcosts of regulation is also consistent with the public-interesttheory
Kane [1996] argues that another major shock to the oldequilibrium was the increasing public awareness of the costlinessof having government-insured but (geographically) undiversiednancial institutions In the late 1970s the failure rate of banksbegan to rise (Table V) In the 1980s the Savings and Loan crisisand taxpayer bailout further heightened public awareness aboutthe costs of restrictions that make depository institutions morelikely to require infusions of taxpayer funds The failures thusmay have heightened public support for branching deregulation(see also Abrams and Settle [1993]) As noted above we did notnd any evidence that banking failures or distress in a stateaffected the speed with which the state deregulated but distressstill may be a common factor affecting all states
These technological economic and legal shocks generatedconditions that changed the long-standing balance favoring theantibranching forces The marginal value of lobbying to repealbranching restrictions increased just as the relative value to thesmall banks of maintaining branching restrictions was decliningThese nationwide shocks are common factors across states andthey are consistent with the positive duration dependence associ-ated with state-level deregulation that we demonstrated in FigureII In addition a broad change in lsquolsquoideologyrsquorsquo against government
QUARTERLY JOURNAL OF ECONOMICS1462
regulation could explain the positive duration dependence duringthe sample period
VI CONCLUSIONS
The private-interest theory of regulation can account for thepattern of bank branching deregulation of the 1970s and 1980sBeneciaries of branching regulation had supported a coalitionfavoring geographical restrictions despite their costs to consum-ers of nancial services Innovations that began in the 1970saltered the value of the restrictions to the affected parties and theresulting competition among interest groups can explain thesubsequent deregulation While some of our results are alsoconsistent with the public interest theorymdashfor example deregula-tion occurs earlier when small banks are in a relatively weaknancial positionmdashother results particularly evidence on theimportance of rivalries between small and large banks andbetween banking and insurance are difficult to explain with thepublic-interest approach We also nd that political-institutionalfactors affect deregulation although these variables may act asproxies for unmeasured economic interests The greater success ofthe private-interest theory to explain deregulation here than inprevious work may be due to the richer cross-sectional andtime-series variation in branching deregulation that can beexploited with a hazard model Future empirical studies ofendogenous deregulation thus may be most fruitful where changehas occurred across states over time such as in franchisinginsurance and public utilities (eg Joskow [1996])
Technological and nancial innovations will continue to erodethe benets to any interest group of maintaining regulatorybarriers in nancial services These forces are likely to bringabout reforms both domestically for example through legislationthat would increase bank powers (see Kroszner [1996] Krosznerand Rajan [1994 1997] and Kroszner and Stratmann [1998]) andinternationally for example through the extension of nancialservices provisions of NAFTAto reduce geographic barriers acrosscountries [Kroszner 1997]
APPENDIX EFFECTS OF DIFFERENT TYPES OF BRANCHING
DEREGULATION ON ECONOMIC GROWTH BANKING STRUCTUREAND BANK EFFICIENCY
We measure the importance of alternative forms of deregula-tion by their consequences for the economic growth and the
WHAT DRIVES DEREGULATION 1463
structure and efficiency of the banking sector across the states(see Jayaratne and Strahan [1996 1998]) To do so we run anOLS panel regression relating indicators for each type of deregula-tion to potential consequences Our dependent variables includeproxies for economic growth (growth in personal income) bankstructure (the log of the number of banking companies and thenumber of branches per banking company) and efficiency (the logof total noninterest expenses and the ratio of loan loss provisionsto total loans) The four deregulation indicators (MBHC branch-ing through merger and acquisition unrestricted statewide branch-ing and interstate banking) are equal to one in the years afterderegulation and zero before If a state deregulated before thesample period begins in 1970 the indicator is one throughout ifthe state does not deregulate then the indicator is zero through-out All regressions include state xed effects and time effects
Table VI shows the results of including the set of four
TABLE VIMULTIVARIATE REGRESSIONS OF THE EFFECTS OF DIFFERENT TYPES OF BRANCHING
DEREGULATION ON ECONOMIC GROWTH BANKING STRUCTUREAND BANK EFFICIENCY
Dependent variable
Growthin
personalincome
Log of of
bankingcompanies
Branchesper
bankingcompany
Log ofnon-
interestexpenses
Loan lossprovisions
(loans 1leases)
Merger amp acquisitionbranchingindicator
0013(0004)
2 0162(0037)
1146(0453)
2 0107(0035)
2 0007(0001)
Unrestricted state-wide branchingindicator
2 0008(0005)
0086(0059)
2 0043(0574)
0070(0037)
0005(0002)
Interstate bankingindicator
0006(0005)
2 0052(0057)
1064(0511)
0005(0030)
2 0001(0001)
Multibank holdingcompany indicator
2 0005(0006)
2 0096(0052)
2 0364(0529)
2 0095(0036)
2 0001(0001)
Dependent variablemean [median]
0086[0082]
4881[5136]
6649[5055]
13812[13803]
0009[0006]
This table contains regressions of state-level economic growth and state-level measures of the structureand efficiency of the banking industry Each dependent variable is regressed on a set of state indicatorvariables a set of year indicator variables and the set of four deregulation indicators these deregulationindicators are equal to one after deregulation and zero before If a state was always regulated the indicator iszero throughout if it was always deregulated the indicator is one throughout Sample period is 1975 to 1991and Delaware and South Dakota are dropped so N 5 833 Noninterest expenses become available only in1984 so N 5 392 The coefficient on the deregulation indicators is reported with robust standard errors inparentheses and denote statistically signicant at the 1 5 and 10 percent levels
QUARTERLY JOURNAL OF ECONOMICS1464
deregulation indicators simultaneously The branching by mergerand acquisition indicator is statistically signicant and of theexpected sign in all ve of the equations and larger in absolutevalue than any of the other coefficients on the indicators for theother types of deregulation Since there may be multicollinearityamong our indicators we also ran univariate specications whichincluded each indicator separately and the results are verysimilar These results imply that branching by merger andacquisition is the most important type of deregulation measuredby its consequences for economic growth banking structure andbank efficiency See Kroszner and Strahan [1998] for more details
UNIVERSITY OF CHICAGO GRADUATE SCHOOL OF BUSINESS AND NATIONAL BUREAU OF
ECONOMIC RESEARCH
FEDERAL RESERVE BANK OF NEW YORK RESEARCH amp MARKET ANALYSIS GROUP
REFERENCES
Abrams Burton A and Russell F Settle lsquolsquoPressure-Group Inuence and Institu-tional Change Branch-Banking Legislation during the Great DepressionrsquorsquoPublic Choice LXXVII (1993) 687ndash705
Amel Dean lsquolsquoState Laws Affecting the Geographic Expansion of CommercialBanksrsquorsquo unpublished manuscript Board of Governors of the Federal ReserveSystem 1993
Becker Gary S lsquolsquoA Theory of Competition among Pressure Groups for PoliticalInuencersquorsquo Quarterly Journal of Economics XCVIII (1983) 371ndash400
Berger Allen Anil Kashyap Joseph Scalise lsquolsquoThe Transformation of the U SBanking Industry What a Long Strange Trip Itrsquos Beenrsquorsquo Brookings Papers onEconomic Activity 2 (1995) 55ndash218
Berger Allen Anthony Saunders Joseph Scalise and Gregory Udell lsquolsquoThe Effectsof Bank Mergers and Acquisitions on Small Business Lendingrsquorsquo Journal ofFinancial Economics L (1998) 187ndash229
Brennan Geoffrey and James Buchanan lsquolsquoTowards a Tax Constitution forLeviathanrsquorsquo Journal of Public Economics VIII (1977) 255ndash273
Brickley James A and Christopher M James lsquolsquoThe Takeover Market CorporateBoard Composition and Ownership Structure The Case of Bankingrsquorsquo Journalof Law and Economics XXX (1987) 161ndash180
Cole Rebel and John Wolken lsquolsquoFinancial Services Used by Small BusinessesEvidence from the 1993 Survey of Small Business Financesrsquorsquo Federal ReserveBulletin LXXXI (1994) 629ndash676
Dixit Avinash The Making of Economic Policy A Transactions-Cost PoliticsPerspective (Cambridge MA MIT Press 1996)
Economides Nicolas R Glenn Hubbard and Darius Palia lsquolsquoThe Political Economyof Branching Restrictions and Deposit Insurancersquorsquo Journal of Law andEconomics XXIX (1996) 667ndash704
Flannery Mark lsquolsquoThe Social Costs of Unit Banking Restrictionsrsquorsquo Journal ofMonetary Economics XIII (1984) 237ndash249
Greene William H Econometric Analysis third edition (Upper Saddle River NJPrentice-Hall 1997)
Gunther Jeffery lsquolsquoRegional Capital Imbalances and the Removal of InterstateBanking Restrictionsrsquorsquo Economics Letters XLIV (1994) 439ndash442 lsquolsquoThe Erosion of Intrastate Branch Banking Restrictionsrsquorsquo unpublishedmanuscript Federal Reserve Bank of Dallas 1996
Irwin Douglas and Randall Kroszner lsquolsquoInterests Institutions and Ideology inSecuring Policy Change The Republican Conversion to Trade Liberalizationafter Smoot-Hawleyrsquorsquo Journal of Law and Economics XLII (1999) 643ndash673
WHAT DRIVES DEREGULATION 1465
James Christopher lsquolsquoAn Analysis of Intra-Industry Differences in the Effect ofRegulation The Case of Deposit Rate Ceilingsrsquorsquo Journal of Monetary Econom-ics XII (1983) 417ndash442
Jarrell Gregg lsquolsquoChange at the Exchange The Causes and Effects of DeregulationrsquorsquoJournal of Law and Economics XXVII (1984) 273ndash312
Jayaratne Jith and Philip E Strahan lsquolsquoThe Finance-Growth Nexus Evidencefrom Bank Branch Deregulationrsquorsquo Quarterly Journal of Economics CI (1996)639ndash670
Jayaratne Jith and Philip E Strahan lsquolsquoEntry Restrictions Industry Evolutionand Dynamic Efficiency Evidence from Commercial Bankingrsquorsquo Journal of Lawand Economics XLIX (1998) 239ndash274
Joskow Paul lsquolsquoComment on lsquoPower Struggles Explaining Deregulatory Reformsin Electricity Marketsrsquo rsquorsquo Brookings Papers on Economic Activity Microeconom-ics (1996) 251ndash264
Joskow Paul and Roger Noll lsquolsquoRegulation in Theory and Practice An Overviewrsquorsquoin Studies in Public Regulation Gary Fromm ed (Cambridge MA MITPress 1981) pp 1ndash65
Kahn Matthew E and John G Matsusaka lsquolsquoDemand for Environmental GoodsEvidence from Voting Patterns on California Initiativesrsquorsquo Journal of Law andEconomics XL (1997) 137ndash173
Kalbeisch J and R Prentice The Statistical Analysis of Failure Time Data (NewYork John Wiley amp Sons 1980)
Kane Edward J lsquolsquoDe Jure Interstate Banking Why Only NowrsquorsquoJournal of MoneyCredit and Banking XXVIII (1996) 141ndash161
Kiefer Nicholas lsquolsquoEconomic Duration Data and Hazard Functionsrsquorsquo Journal ofEconomic Literature XXVI (1988) 646ndash679
Kroszner Randall S lsquolsquoThe Evolution of Universal Banking and Its Regulation inTwentieth Century Americarsquorsquo in Universal Banking Financial System DesignReconsidered Anthony Saunders and Ingo Walter eds (New York IrwinProfessional Publishers 1996) pp 70ndash99 lsquolsquoThe Political Economy of Banking and Financial Regulation in the U Srsquorsquo inThe Banking and Financial Structure in the NAFTA Countries and ChileGeorge M von Furstenberg ed (Boston Kluwer Academic Publishers 1997)pp 200ndash213
Kroszner Randall S and Raghuram G Rajan lsquolsquoIs the Glass-Steagall ActJustied A Study of the U S Experience with Universal Banking before1933rsquorsquo American Economic Review LXXXIV (1994) 810ndash832
Kroszner Randall S and Raghuram G Rajan lsquolsquoOrganization Structure andCredibility Evidence from Bank Securities Activities before the Glass-Steagall Actrsquorsquo Journal of Monetary Economics XXXIX (1997) 475ndash516
Kroszner Randall S and Philip E Strahan lsquolsquoRegulatory Incentives and the ThriftCrisis Dividends Mutual-to-Stock Conversions and Financial DistressrsquorsquoJournal of Finance LI (1996) 1285ndash1320
Kroszner Randall S and Philip E Strahan lsquolsquoWhat Drives Deregulation Econom-ics and Politics of the Relaxation of Bank Branching Restrictionsrsquorsquo NBERWorking Paper No 6637 1998
Kroszner Randall S and Thomas Stratmann lsquolsquoInterest Group Competition andthe Organization of Congress Theory and Evidence from Financial ServicesPolitical Action Committeesrsquorsquo American Economic Review LXXXVIII (1998)pp 1163ndash1187
Kroszner Randall S and Thomas Stratmann lsquolsquoDoes Political Ambiguity PayCorporate Campaign Contributions and Rewards to Legislator Reputationrsquorsquounpublishedmanuscript University of Chicago Graduate School of Business1999
Lin D Y and L J Wei lsquolsquoThe Robust Inference for the Cox Proportional HazardsModel Journal of the American Statistical Association LXXXIV (1989)1074ndash1078
Macey Jonathan R and Geoffrey P Miller Banking Law and Regulation (BostonLittle Brown and Company 1992)
Makinson Larry Open Secrets The Encyclopedia of Congressional Money andPolitics (Washington DC Congressional Quarterly 1992)
Niskanen WilliamBureaucracy and Representative Government (Chicago Aldine-Atherton 1971)
QUARTERLY JOURNAL OF ECONOMICS1466
Noll Roger lsquolsquoComment on Peltzmanrsquorsquo Brookings Papers Microeconomics SpecialIssue (1989) 48ndash58
North Douglass lsquolsquoA Transactions Cost Theory of Politicsrsquorsquo Journal of TheoreticalPolitics II (1990) 355ndash367
Peltzman Sam lsquolsquoToward a More General Theory of Regulationrsquorsquo Journal of Lawand Economics XIX (1976) 109ndash148 lsquolsquoConstituent Interest and Congressional Votingrsquorsquo Journal of Law andEconomics XXVII (1984) 181ndash210 lsquolsquoThe Economic Theory of Regulation after a Decade of DeregulationrsquorsquoBrookings Papers Microeconomics Special Issue (1989) 1ndash41
Peterson Mitchell and Raghuram Rajan lsquolsquoThe Benets of Lending RelationshipsEvidence from Small Business Datarsquorsquo Journal of Finance XLIX (1994) 3ndash38
Poole Keith and Howard Rosenthal Congress A Political-Economic History ofRoll Call Voting (Oxford Oxford University Press 1997)
Stata Stata Reference Manual Release 5 (College Station TX Stata Press 1997)Stigler George J lsquolsquoThe Theory of Economic Regulationrsquorsquo Bell Journal of Economics
and Management Science II (1971) 3ndash21Stigler George J Chicago Studies in the Political Economy (Chicago University
of Chicago Press 1988)Strahan Philip E and James Weston lsquolsquoSmall Business Lending and the Changing
Structure of the Banking Industryrsquorsquo Journal of Banking and Finance XXII(1998) 821ndash845
Sylla Richard John Legler and John Wallis lsquolsquoBanks and State Public Finance inthe New Republic The United States 1790ndash1860rsquorsquo Journal of EconomicHistory XLVII (1987) 391ndash403
White Eugene The Regulation and Reform of the American Banking System1900ndash1929 (Princeton Princeton University Press 1983)
Winston Clifford lsquolsquoEconomic Deregulation Days of Reckoning for Microecono-mistsrsquorsquo Journal of Economic Literature XXXI (1993) 1263ndash1289
WHAT DRIVES DEREGULATION 1467
variables are statistically signicant The signs of the effects ofthe insurance variables are the same as those reported in thehazard model but not statistically signicant
The hazard model and linear probability models thus producevery similar results The estimates from the latter suggest thatthe effects for the small bank market share and small rm sharevariables are driven primarily by cross-state variation the effectof the insurance variable is driven primarily by time-seriesvariation and the small bank capital effect is driven by bothcross-sectional and time-series variation
Ex Post Consequences of Deregulation Finally to check theplausibility of our interpretation of the hazard results we con-sider whether the ex post consequences of deregulation areconsistent with the ex ante lobbying positions we attribute to eachinterest group (see Kroszner and Strahan [1998] for details) Wend that small banks lose market share following deregulationand in states where banks can enter the insurance business theinsurance sector shrinks relative to the banking sector followingderegulation Borrowers also benet from deregulation throughlower average interest rates on loans These ndings support theprivate-interest interpretation of the coefficients in the hazardmodel groups that will benet lobby to speed deregulation andthose that will be harmed lobby to slow it
C Voting on Interstate Branching Deregulation in the U SHouse of Representatives
We now examine whether the forces driving intrastate branch-ing deregulation also drive interstate deregulation Financialservices interests are active contributors and lobbyists in Washing-ton Their political action committees constitute the largest groupof contributors to legislators providing nearly 20 percent of totalcongressional campaign contributions [Makinson 1992] and muchof their lobbying effort involves competition among rival interestswithin nancial services (see Kroszner and Stratmann [1998])
As noted above after virtually all states adopted intra- andinterstate branching deregulation the 1994 Riegle-Neal Act re-pealed the 1927 McFadden Act to eliminate all barriers tointerstate banking and branching by 1997 The key votes concern-ing the Riegle-Neal Act were either voice votes or extremelylopsided so we could not estimate a voting model from them Anumber of bills and amendments related to interstate branching
WHAT DRIVES DEREGULATION 1457
had been debated in Congress during the years prior to thepassage of the Riegle-Neal Act but a search of the weekly BNABanking Reporter and the Congressional Record produced onlyone roll-call vote related to interstate branching that was notlopsided This vote occurred in the House on November 14 1991and concerned an amendment sponsored by Wylie (R-OH) andNeal (D-NC) to introduce interstate banking and branchingderegulation to a nancial services reform package33 Althoughthe amendment passed by 210 to 208 the bill to which it wasattached subsequently was defeated
To check for the inuence of the factors we considered in thestate-level reforms we examine both the sponsorship of interstatebanking legislation and voting on the amendment The sponsorsof the Wylie-Neal amendment are from states with low small bankshares 004 in Ohio (Wylie) and 002 in North Carolina (Neal) Incontrast the sample mean in 1991 is 008 (median 5 007)Michigan home state of the Senatersquos sponsor of the 1994 Riegle-Neal Act also had relatively low small bank strength (small bankshare of 005)
Table IV reports estimates of a probit model where thedependent variable equals one if the legislator voted in favor ofthe amendment and zero otherwise The explanatory variablesare the same as those in the hazard model except that we use aparty affiliation indicator variable equal to one for Democrats andzero for Republicans as our proxy for political factors Thecoefficients in Table IV are the marginal effects (lsquolsquoslopesrsquorsquo) of aone-unit change of each variable on the probability that alegislator will vote for the amendment34 Note that a forcefavoring deregulation has a negative coefficient in the hazardmodel (Table III) but a positive coefficient in the probit model
Consistent with the state-level deregulation process legisla-tors are more likely to support the amendment if their states havea relatively low share of small banks As in the hazard model thefraction of small banks is the most important interest groupinuence on a legislatorrsquos voting decision A one-standard-deviation increase in small banksrsquo market share (from the mean)is associated with a decline in the probability of voting in favor of
33 The Wylie-Neal amendment also included provisions limiting certaininsurance and real estate powers of national banks [Congressional RecordNovember 14 1991 pp 10239ndash10242]
34 Since we have multiple legislators from each state we adjust the standarderrors to correct for the lack of independence among observations clustered in thesame state
QUARTERLY JOURNAL OF ECONOMICS1458
TAB
LE
IVM
AR
GIN
AL
EF
FE
CT
SF
RO
MA
PR
OB
ITM
OD
EL
OF
TH
EIN
FL
UE
NC
EO
FP
OL
ITIC
AL
EC
ON
OM
YF
AC
TO
RS
ON
TH
EH
OU
SE
VO
TE
ON
TH
EW
YL
IE-N
EA
L
AM
EN
DM
EN
TT
OP
ER
MIT
INT
ER
ST
AT
EB
AN
KB
RA
NC
HIN
G
NO
VE
MB
ER
14
1991
(1)
(2)
(3)
(4)
(5)
(6)
Smal
lban
kas
set
shar
eof
allb
anki
ng
asse
tsin
stat
e2
417
(12
4)2
454
(12
2)2
407
(12
0)2
443
(14
7)2
450
(14
9)2
490
(14
7)C
apit
alra
tio
ofsm
allb
anks
rela
tive
tola
rge
inth
est
ate
180
(37
5)2
68(3
42)
256
(37
5)1
69(3
78)
153
(38
0)2
45(4
25)
Rel
ativ
esi
zeof
insu
ranc
ein
stat
esw
here
bank
sm
ayse
llin
sura
nce
0ot
herw
ise
22
34
(10
2)2
249
(1
14)
22
33
(10
1)2
249
(1
04)
22
51
(10
2)2
154
(11
1)In
dica
tor
is1
ifba
nks
may
sell
insu
ranc
ein
the
stat
e0
79
(0
06)
074
(00
6)0
69
(0
06)
070
(00
6)0
70
(0
06)
067
(00
7)R
elat
ive
size
ofin
sura
nce
inst
ates
whe
reba
nks
may
not
sell
insu
ranc
e0
oth
erw
ise
180
(06
6)2
42
(0
72)
154
(0
67)
173
(06
6)1
72
(0
67)
203
(0
91)
Smal
lrm
shar
eof
the
num
ber
ofr
ms
inth
est
ate
21
23(0
92)
20
83(0
89)
21
62(1
07)
21
32(0
95)
21
30(1
00)
20
94(1
16)
Indi
cato
ris
1if
Rep
rese
ntat
ive
isa
Dem
ocra
t2
035
(00
7)2
035
(00
7)2
036
(00
7)2
035
(00
7)2
035
(00
7)2
036
(00
7)A
vera
geyi
eld
onba
nklo
ans
inth
est
ate
min
usF
edfu
nds
rate
mdash10
01
(51
5)mdash
mdashmdash
mdash
Ave
rage
fail
ure
rate
198
5to
1991
mdashmdash
228
(14
1)mdash
mdashmdash
Indi
cato
ris
1if
stat
eh
asun
itba
nkin
gla
wmdash
mdashmdash
004
(00
7)0
04(0
07)
20
01(0
08)
Indi
cato
ris
1if
stat
ech
ange
sba
nk
insu
ran
cepo
wer
smdash
mdashmdash
mdash2
001
(00
7)2
003
(00
7)In
clud
esre
gion
alin
dica
tors
N
oN
oN
oN
oN
oYe
sP
seud
o-R
20
170
180
170
170
170
18L
ogli
keli
hood
224
024
223
826
223
951
224
013
224
010
223
808
p-va
lue
ofx2
for
regr
essi
on
001
0
01
001
0
01
001
0
01
Th
ede
pen
den
tvar
iabl
eis
one
ifth
ele
gisl
ator
vote
sfo
rth
eam
endm
enta
nd
zero
ifag
ains
tT
he
repo
rted
coef
fici
ents
are
the
effe
cts
ofa
uni
tch
ange
ofth
ein
depe
nde
ntva
riab
le(f
rom
the
mea
n)on
the
prob
abili
tyof
voti
ng
infa
vor
ofth
eam
endm
ent
All
vari
able
sar
em
easu
red
byst
ate
N5
418
the
num
ber
ofR
epre
sent
ativ
esvo
tin
gon
the
amen
dmen
tS
mal
lban
kas
sets
har
eis
the
perc
ent
ofba
nki
ng
asse
tsin
the
stat
eh
eld
byba
nks
belo
wth
em
edia
nsi
zeof
ban
kin
each
stat
ein
each
year
Rel
ativ
eca
pita
lrat
iois
the
capi
talt
oas
sets
rati
oof
smal
lban
ksm
inu
sth
atof
larg
eba
nks
Siz
eof
insu
ran
cere
lati
veto
bank
ing
plu
sin
sura
nce
inth
est
ate
ism
easu
red
asgr
oss
stat
epr
oduc
tfro
min
sura
nce
divi
ded
bygr
oss
stat
epr
odu
ctfr
omin
sura
nce
plu
sba
nkin
gIn
dica
tor
ison
eif
stat
ela
wpe
rmit
sba
nks
toen
ter
the
insu
ran
cebu
sin
ess
Sm
all
rmsh
are
isth
epe
rcen
tofa
lles
tabl
ishm
ents
inth
est
ate
that
hav
efe
wer
than
twen
tyem
ploy
ees
Par
tyin
dica
tor
ison
eif
the
Rep
rese
nta
tive
isa
Dem
ocra
tA
vera
geyi
eld
onba
nklo
ans
equa
lsto
tali
nter
est
inco
me
onal
ldom
esti
clo
ans
mad
eby
ban
ksin
the
stat
edi
vide
dby
tota
lloa
ns
The
aver
age
bank
failu
rera
teis
the
rati
oof
asse
tsin
faile
dba
nks
duri
ng
agi
ven
stat
edi
vide
dby
tota
lban
kas
sets
inth
atst
ate
aver
aged
over
the
year
s19
85to
1991
Un
itba
nki
ng
indi
cato
req
ual
son
efo
rst
ates
wit
hun
itba
nkin
gre
stri
ctio
nsT
hech
ange
inin
sura
nce
pow
ers
indi
cato
req
ual
son
eif
the
stat
ech
ange
dth
ela
wto
perm
itba
nks
tose
llin
sura
nce
duri
ngth
esa
mpl
epe
riod
We
divi
deth
eco
unt
ryin
tofo
urre
gion
san
din
clu
dere
gion
alin
dica
tors
inco
lum
n(5
)R
obu
stst
anda
rder
rors
are
inpa
ren
thes
es
an
d
den
ote
stat
isti
cally
sign
ica
nta
tth
e1
5an
d10
perc
entl
evel
s
WHAT DRIVES DEREGULATION 1459
branching of approximately 17 percent The relative capital-to-assets ratio of small and large banks however does not have astatistically signicant effect in the probit model
The impact of rival interests outside of banking is alsoconsistent with intrastate deregulation results Where banks cansell insurance legislators from states with larger insurancesectors relative to banking are less likely to vote for the amend-ment35 A one-standard-deviation increase in the relative size ofinsurance in those states which permit banks to sell insurancedecreases the probability that a legislator will favor the amend-ment by about 13 percent Turning to consumer interests thecoefficient on small rm share is not statistically signicant butthe coefficient on interest rates in the state is A one-standard-deviation increase in the average interest rate on loans raises theprobability that a legislator will support the amendment byroughly 8 percent The coefficient estimate for the bankingdistress variable measured as the average bank failure rate from1985 to 1991 is small and statistically signicant Overall theprobit analysis of the vote on national branching deregulationsupports the private-interest theory of deregulation and providesa consistency check that the importance of interests operating onthe state legislatures are very similar to those operating at thefederal level
V WHY DID BRANCHING DEREGULATION BEGIN DURING THE
1970S SHOCKS TO THE EQUILIBRIUM SUPPORTING BANK
BRANCHING RESTRICTIONS
A complete explanation of regulatory exit should account forwhy deregulation begins in the 1970s as well as the timing of thestate-by-state reforms [Kane 1996] In this section we describebroad technological legal and economic shocks that altered thepolitical-economy equilibrium which had kept antibranching regu-lations little changed for at least 30 years
Beginning in the 1970s three innovations reduced the valueto the protected banks of local geographic monopolies (see TableV) First automatic teller machines (ATMs) helped to erode thegeographic ties between customers and banks Second checkablemoney market mutual funds and the Merrill Lynch Cash Manage-ment Account demonstrated that banking by mail and telephone
35 The positive and statistically signicant estimates on the other insurancevariables may reect insurance industry support for the amendmentrsquos provisionslimiting banksrsquo insurance powers
QUARTERLY JOURNAL OF ECONOMICS1460
provided a convenient alternative to local banks36 Third techno-logical innovation and deregulation reduced transportation andcommunication costs particularly since the 1970s thereby lower-ing the costs for customers to use distant banks By increasing theelasticity of deposits supplied to banks these innovations reducedthe value of geographical restrictions to their traditional benecia-ries and thereby reduced their incentive to ght to maintain them[Peltzman 1976]
On the lending side increasing sophistication of credit-scoring techniques following innovations in information process-ing nancial theory and the development of large credit databases began to diminish the value of knowledge that localbankers had about the risks of borrowers in the community As a
36 Regulation Q which limited the interest rates that banks could pay ondeposits may have helped to drive depositors away from banks when the gapbetween market rates and deposit ceilings grew during the 1970s The eliminationof interest rate ceilings on large denomination certicates of deposit during the1970s appears to have hurt smaller and retail-oriented banks relative to largerwholesale banks [James 1983]
TABLE VBROAD TRENDS IN COMMERCIAL BANKING 1950ndash1995
YearNumberof ATMs
Domesticbank
deposits(billions)
Moneymarketmutual
fund(billions)
Percent ofdeposits 1
moneyfunds heldby banks
Smallbanksrsquo
percent ofbankingassets
Averagenumber of
bankfailures
(1) (2) (3) (4) (5) (6)1950 0 $ 154 $ 0 100 NA 41955 0 191 0 100 NA 31960 0 228 0 100 24 21965 0 330 0 100 20 41970 0 479 0 100 18 61975 9750 775 4 99 18 61980 18500 1182 76 94 17 101985 61117 1787 242 88 14 601990 80156 2339 493 83 11 1791995 122706 2552 745 77 8 61
Column (1) ATM gures are from Bank Network News The EFT Network Data Book (New YorkFaulkner and Gray Inc) The 1975 gure was unavailable 9750 is the number of ATMs in 1978 the rst yearfor which complete data are available Columns (2)ndash(4) Banksrsquo domestic deposits are from the Reports ofIncome and Condition money market mutual funds are from the Flow of Funds Data on all bank depositsforeign plus domestic are only available beginning in 1970 The trend in banksrsquo share (column (4)) is the sameusing total deposits instead of domestic deposits Column (5) Percent of banking assets held by small bankswhere a small bank is dened as a commercial bank with less than $100 million in assets in 1994 dollarsThese data are based on the Reports of Income and Condition Data on small banks are not available before1960 Column (6) Five-year average number of bank failures where the nal year is indicated in the rstcolumn These data are from FDIC Annual Report and the Quarterly Banking Prole
WHAT DRIVES DEREGULATION 1461
result of these innovations a national market developed forresidential mortgages credit card receivables have been securi-tized and bank lending to small business now relies less on thejudgment of loan officers and more on standardized scoringmodels
These changes have increased the potential protability forlarge banks to enter what had been the core of small bankactivities Large banksrsquo incentive to increase their lobbying pres-sure to be able to expand into these markets has thus beenincreasing over time In fact small banksrsquo market share began todecline even prior to the branching deregulation (Table V) As thevalue of a local banking relationships declined small rms thatwere the main borrowers from the small banks also probablybecame more likely to favor the entry of large banks into localmarkets With the deadweight costs of preventing large bankentry rising the private-interest theory predicts that small localbanks would become less likely able to maintain the branchingrestrictions [Becker 1983] Deregulation that reduces deadweightcosts of regulation is also consistent with the public-interesttheory
Kane [1996] argues that another major shock to the oldequilibrium was the increasing public awareness of the costlinessof having government-insured but (geographically) undiversiednancial institutions In the late 1970s the failure rate of banksbegan to rise (Table V) In the 1980s the Savings and Loan crisisand taxpayer bailout further heightened public awareness aboutthe costs of restrictions that make depository institutions morelikely to require infusions of taxpayer funds The failures thusmay have heightened public support for branching deregulation(see also Abrams and Settle [1993]) As noted above we did notnd any evidence that banking failures or distress in a stateaffected the speed with which the state deregulated but distressstill may be a common factor affecting all states
These technological economic and legal shocks generatedconditions that changed the long-standing balance favoring theantibranching forces The marginal value of lobbying to repealbranching restrictions increased just as the relative value to thesmall banks of maintaining branching restrictions was decliningThese nationwide shocks are common factors across states andthey are consistent with the positive duration dependence associ-ated with state-level deregulation that we demonstrated in FigureII In addition a broad change in lsquolsquoideologyrsquorsquo against government
QUARTERLY JOURNAL OF ECONOMICS1462
regulation could explain the positive duration dependence duringthe sample period
VI CONCLUSIONS
The private-interest theory of regulation can account for thepattern of bank branching deregulation of the 1970s and 1980sBeneciaries of branching regulation had supported a coalitionfavoring geographical restrictions despite their costs to consum-ers of nancial services Innovations that began in the 1970saltered the value of the restrictions to the affected parties and theresulting competition among interest groups can explain thesubsequent deregulation While some of our results are alsoconsistent with the public interest theorymdashfor example deregula-tion occurs earlier when small banks are in a relatively weaknancial positionmdashother results particularly evidence on theimportance of rivalries between small and large banks andbetween banking and insurance are difficult to explain with thepublic-interest approach We also nd that political-institutionalfactors affect deregulation although these variables may act asproxies for unmeasured economic interests The greater success ofthe private-interest theory to explain deregulation here than inprevious work may be due to the richer cross-sectional andtime-series variation in branching deregulation that can beexploited with a hazard model Future empirical studies ofendogenous deregulation thus may be most fruitful where changehas occurred across states over time such as in franchisinginsurance and public utilities (eg Joskow [1996])
Technological and nancial innovations will continue to erodethe benets to any interest group of maintaining regulatorybarriers in nancial services These forces are likely to bringabout reforms both domestically for example through legislationthat would increase bank powers (see Kroszner [1996] Krosznerand Rajan [1994 1997] and Kroszner and Stratmann [1998]) andinternationally for example through the extension of nancialservices provisions of NAFTAto reduce geographic barriers acrosscountries [Kroszner 1997]
APPENDIX EFFECTS OF DIFFERENT TYPES OF BRANCHING
DEREGULATION ON ECONOMIC GROWTH BANKING STRUCTUREAND BANK EFFICIENCY
We measure the importance of alternative forms of deregula-tion by their consequences for the economic growth and the
WHAT DRIVES DEREGULATION 1463
structure and efficiency of the banking sector across the states(see Jayaratne and Strahan [1996 1998]) To do so we run anOLS panel regression relating indicators for each type of deregula-tion to potential consequences Our dependent variables includeproxies for economic growth (growth in personal income) bankstructure (the log of the number of banking companies and thenumber of branches per banking company) and efficiency (the logof total noninterest expenses and the ratio of loan loss provisionsto total loans) The four deregulation indicators (MBHC branch-ing through merger and acquisition unrestricted statewide branch-ing and interstate banking) are equal to one in the years afterderegulation and zero before If a state deregulated before thesample period begins in 1970 the indicator is one throughout ifthe state does not deregulate then the indicator is zero through-out All regressions include state xed effects and time effects
Table VI shows the results of including the set of four
TABLE VIMULTIVARIATE REGRESSIONS OF THE EFFECTS OF DIFFERENT TYPES OF BRANCHING
DEREGULATION ON ECONOMIC GROWTH BANKING STRUCTUREAND BANK EFFICIENCY
Dependent variable
Growthin
personalincome
Log of of
bankingcompanies
Branchesper
bankingcompany
Log ofnon-
interestexpenses
Loan lossprovisions
(loans 1leases)
Merger amp acquisitionbranchingindicator
0013(0004)
2 0162(0037)
1146(0453)
2 0107(0035)
2 0007(0001)
Unrestricted state-wide branchingindicator
2 0008(0005)
0086(0059)
2 0043(0574)
0070(0037)
0005(0002)
Interstate bankingindicator
0006(0005)
2 0052(0057)
1064(0511)
0005(0030)
2 0001(0001)
Multibank holdingcompany indicator
2 0005(0006)
2 0096(0052)
2 0364(0529)
2 0095(0036)
2 0001(0001)
Dependent variablemean [median]
0086[0082]
4881[5136]
6649[5055]
13812[13803]
0009[0006]
This table contains regressions of state-level economic growth and state-level measures of the structureand efficiency of the banking industry Each dependent variable is regressed on a set of state indicatorvariables a set of year indicator variables and the set of four deregulation indicators these deregulationindicators are equal to one after deregulation and zero before If a state was always regulated the indicator iszero throughout if it was always deregulated the indicator is one throughout Sample period is 1975 to 1991and Delaware and South Dakota are dropped so N 5 833 Noninterest expenses become available only in1984 so N 5 392 The coefficient on the deregulation indicators is reported with robust standard errors inparentheses and denote statistically signicant at the 1 5 and 10 percent levels
QUARTERLY JOURNAL OF ECONOMICS1464
deregulation indicators simultaneously The branching by mergerand acquisition indicator is statistically signicant and of theexpected sign in all ve of the equations and larger in absolutevalue than any of the other coefficients on the indicators for theother types of deregulation Since there may be multicollinearityamong our indicators we also ran univariate specications whichincluded each indicator separately and the results are verysimilar These results imply that branching by merger andacquisition is the most important type of deregulation measuredby its consequences for economic growth banking structure andbank efficiency See Kroszner and Strahan [1998] for more details
UNIVERSITY OF CHICAGO GRADUATE SCHOOL OF BUSINESS AND NATIONAL BUREAU OF
ECONOMIC RESEARCH
FEDERAL RESERVE BANK OF NEW YORK RESEARCH amp MARKET ANALYSIS GROUP
REFERENCES
Abrams Burton A and Russell F Settle lsquolsquoPressure-Group Inuence and Institu-tional Change Branch-Banking Legislation during the Great DepressionrsquorsquoPublic Choice LXXVII (1993) 687ndash705
Amel Dean lsquolsquoState Laws Affecting the Geographic Expansion of CommercialBanksrsquorsquo unpublished manuscript Board of Governors of the Federal ReserveSystem 1993
Becker Gary S lsquolsquoA Theory of Competition among Pressure Groups for PoliticalInuencersquorsquo Quarterly Journal of Economics XCVIII (1983) 371ndash400
Berger Allen Anil Kashyap Joseph Scalise lsquolsquoThe Transformation of the U SBanking Industry What a Long Strange Trip Itrsquos Beenrsquorsquo Brookings Papers onEconomic Activity 2 (1995) 55ndash218
Berger Allen Anthony Saunders Joseph Scalise and Gregory Udell lsquolsquoThe Effectsof Bank Mergers and Acquisitions on Small Business Lendingrsquorsquo Journal ofFinancial Economics L (1998) 187ndash229
Brennan Geoffrey and James Buchanan lsquolsquoTowards a Tax Constitution forLeviathanrsquorsquo Journal of Public Economics VIII (1977) 255ndash273
Brickley James A and Christopher M James lsquolsquoThe Takeover Market CorporateBoard Composition and Ownership Structure The Case of Bankingrsquorsquo Journalof Law and Economics XXX (1987) 161ndash180
Cole Rebel and John Wolken lsquolsquoFinancial Services Used by Small BusinessesEvidence from the 1993 Survey of Small Business Financesrsquorsquo Federal ReserveBulletin LXXXI (1994) 629ndash676
Dixit Avinash The Making of Economic Policy A Transactions-Cost PoliticsPerspective (Cambridge MA MIT Press 1996)
Economides Nicolas R Glenn Hubbard and Darius Palia lsquolsquoThe Political Economyof Branching Restrictions and Deposit Insurancersquorsquo Journal of Law andEconomics XXIX (1996) 667ndash704
Flannery Mark lsquolsquoThe Social Costs of Unit Banking Restrictionsrsquorsquo Journal ofMonetary Economics XIII (1984) 237ndash249
Greene William H Econometric Analysis third edition (Upper Saddle River NJPrentice-Hall 1997)
Gunther Jeffery lsquolsquoRegional Capital Imbalances and the Removal of InterstateBanking Restrictionsrsquorsquo Economics Letters XLIV (1994) 439ndash442 lsquolsquoThe Erosion of Intrastate Branch Banking Restrictionsrsquorsquo unpublishedmanuscript Federal Reserve Bank of Dallas 1996
Irwin Douglas and Randall Kroszner lsquolsquoInterests Institutions and Ideology inSecuring Policy Change The Republican Conversion to Trade Liberalizationafter Smoot-Hawleyrsquorsquo Journal of Law and Economics XLII (1999) 643ndash673
WHAT DRIVES DEREGULATION 1465
James Christopher lsquolsquoAn Analysis of Intra-Industry Differences in the Effect ofRegulation The Case of Deposit Rate Ceilingsrsquorsquo Journal of Monetary Econom-ics XII (1983) 417ndash442
Jarrell Gregg lsquolsquoChange at the Exchange The Causes and Effects of DeregulationrsquorsquoJournal of Law and Economics XXVII (1984) 273ndash312
Jayaratne Jith and Philip E Strahan lsquolsquoThe Finance-Growth Nexus Evidencefrom Bank Branch Deregulationrsquorsquo Quarterly Journal of Economics CI (1996)639ndash670
Jayaratne Jith and Philip E Strahan lsquolsquoEntry Restrictions Industry Evolutionand Dynamic Efficiency Evidence from Commercial Bankingrsquorsquo Journal of Lawand Economics XLIX (1998) 239ndash274
Joskow Paul lsquolsquoComment on lsquoPower Struggles Explaining Deregulatory Reformsin Electricity Marketsrsquo rsquorsquo Brookings Papers on Economic Activity Microeconom-ics (1996) 251ndash264
Joskow Paul and Roger Noll lsquolsquoRegulation in Theory and Practice An Overviewrsquorsquoin Studies in Public Regulation Gary Fromm ed (Cambridge MA MITPress 1981) pp 1ndash65
Kahn Matthew E and John G Matsusaka lsquolsquoDemand for Environmental GoodsEvidence from Voting Patterns on California Initiativesrsquorsquo Journal of Law andEconomics XL (1997) 137ndash173
Kalbeisch J and R Prentice The Statistical Analysis of Failure Time Data (NewYork John Wiley amp Sons 1980)
Kane Edward J lsquolsquoDe Jure Interstate Banking Why Only NowrsquorsquoJournal of MoneyCredit and Banking XXVIII (1996) 141ndash161
Kiefer Nicholas lsquolsquoEconomic Duration Data and Hazard Functionsrsquorsquo Journal ofEconomic Literature XXVI (1988) 646ndash679
Kroszner Randall S lsquolsquoThe Evolution of Universal Banking and Its Regulation inTwentieth Century Americarsquorsquo in Universal Banking Financial System DesignReconsidered Anthony Saunders and Ingo Walter eds (New York IrwinProfessional Publishers 1996) pp 70ndash99 lsquolsquoThe Political Economy of Banking and Financial Regulation in the U Srsquorsquo inThe Banking and Financial Structure in the NAFTA Countries and ChileGeorge M von Furstenberg ed (Boston Kluwer Academic Publishers 1997)pp 200ndash213
Kroszner Randall S and Raghuram G Rajan lsquolsquoIs the Glass-Steagall ActJustied A Study of the U S Experience with Universal Banking before1933rsquorsquo American Economic Review LXXXIV (1994) 810ndash832
Kroszner Randall S and Raghuram G Rajan lsquolsquoOrganization Structure andCredibility Evidence from Bank Securities Activities before the Glass-Steagall Actrsquorsquo Journal of Monetary Economics XXXIX (1997) 475ndash516
Kroszner Randall S and Philip E Strahan lsquolsquoRegulatory Incentives and the ThriftCrisis Dividends Mutual-to-Stock Conversions and Financial DistressrsquorsquoJournal of Finance LI (1996) 1285ndash1320
Kroszner Randall S and Philip E Strahan lsquolsquoWhat Drives Deregulation Econom-ics and Politics of the Relaxation of Bank Branching Restrictionsrsquorsquo NBERWorking Paper No 6637 1998
Kroszner Randall S and Thomas Stratmann lsquolsquoInterest Group Competition andthe Organization of Congress Theory and Evidence from Financial ServicesPolitical Action Committeesrsquorsquo American Economic Review LXXXVIII (1998)pp 1163ndash1187
Kroszner Randall S and Thomas Stratmann lsquolsquoDoes Political Ambiguity PayCorporate Campaign Contributions and Rewards to Legislator Reputationrsquorsquounpublishedmanuscript University of Chicago Graduate School of Business1999
Lin D Y and L J Wei lsquolsquoThe Robust Inference for the Cox Proportional HazardsModel Journal of the American Statistical Association LXXXIV (1989)1074ndash1078
Macey Jonathan R and Geoffrey P Miller Banking Law and Regulation (BostonLittle Brown and Company 1992)
Makinson Larry Open Secrets The Encyclopedia of Congressional Money andPolitics (Washington DC Congressional Quarterly 1992)
Niskanen WilliamBureaucracy and Representative Government (Chicago Aldine-Atherton 1971)
QUARTERLY JOURNAL OF ECONOMICS1466
Noll Roger lsquolsquoComment on Peltzmanrsquorsquo Brookings Papers Microeconomics SpecialIssue (1989) 48ndash58
North Douglass lsquolsquoA Transactions Cost Theory of Politicsrsquorsquo Journal of TheoreticalPolitics II (1990) 355ndash367
Peltzman Sam lsquolsquoToward a More General Theory of Regulationrsquorsquo Journal of Lawand Economics XIX (1976) 109ndash148 lsquolsquoConstituent Interest and Congressional Votingrsquorsquo Journal of Law andEconomics XXVII (1984) 181ndash210 lsquolsquoThe Economic Theory of Regulation after a Decade of DeregulationrsquorsquoBrookings Papers Microeconomics Special Issue (1989) 1ndash41
Peterson Mitchell and Raghuram Rajan lsquolsquoThe Benets of Lending RelationshipsEvidence from Small Business Datarsquorsquo Journal of Finance XLIX (1994) 3ndash38
Poole Keith and Howard Rosenthal Congress A Political-Economic History ofRoll Call Voting (Oxford Oxford University Press 1997)
Stata Stata Reference Manual Release 5 (College Station TX Stata Press 1997)Stigler George J lsquolsquoThe Theory of Economic Regulationrsquorsquo Bell Journal of Economics
and Management Science II (1971) 3ndash21Stigler George J Chicago Studies in the Political Economy (Chicago University
of Chicago Press 1988)Strahan Philip E and James Weston lsquolsquoSmall Business Lending and the Changing
Structure of the Banking Industryrsquorsquo Journal of Banking and Finance XXII(1998) 821ndash845
Sylla Richard John Legler and John Wallis lsquolsquoBanks and State Public Finance inthe New Republic The United States 1790ndash1860rsquorsquo Journal of EconomicHistory XLVII (1987) 391ndash403
White Eugene The Regulation and Reform of the American Banking System1900ndash1929 (Princeton Princeton University Press 1983)
Winston Clifford lsquolsquoEconomic Deregulation Days of Reckoning for Microecono-mistsrsquorsquo Journal of Economic Literature XXXI (1993) 1263ndash1289
WHAT DRIVES DEREGULATION 1467
had been debated in Congress during the years prior to thepassage of the Riegle-Neal Act but a search of the weekly BNABanking Reporter and the Congressional Record produced onlyone roll-call vote related to interstate branching that was notlopsided This vote occurred in the House on November 14 1991and concerned an amendment sponsored by Wylie (R-OH) andNeal (D-NC) to introduce interstate banking and branchingderegulation to a nancial services reform package33 Althoughthe amendment passed by 210 to 208 the bill to which it wasattached subsequently was defeated
To check for the inuence of the factors we considered in thestate-level reforms we examine both the sponsorship of interstatebanking legislation and voting on the amendment The sponsorsof the Wylie-Neal amendment are from states with low small bankshares 004 in Ohio (Wylie) and 002 in North Carolina (Neal) Incontrast the sample mean in 1991 is 008 (median 5 007)Michigan home state of the Senatersquos sponsor of the 1994 Riegle-Neal Act also had relatively low small bank strength (small bankshare of 005)
Table IV reports estimates of a probit model where thedependent variable equals one if the legislator voted in favor ofthe amendment and zero otherwise The explanatory variablesare the same as those in the hazard model except that we use aparty affiliation indicator variable equal to one for Democrats andzero for Republicans as our proxy for political factors Thecoefficients in Table IV are the marginal effects (lsquolsquoslopesrsquorsquo) of aone-unit change of each variable on the probability that alegislator will vote for the amendment34 Note that a forcefavoring deregulation has a negative coefficient in the hazardmodel (Table III) but a positive coefficient in the probit model
Consistent with the state-level deregulation process legisla-tors are more likely to support the amendment if their states havea relatively low share of small banks As in the hazard model thefraction of small banks is the most important interest groupinuence on a legislatorrsquos voting decision A one-standard-deviation increase in small banksrsquo market share (from the mean)is associated with a decline in the probability of voting in favor of
33 The Wylie-Neal amendment also included provisions limiting certaininsurance and real estate powers of national banks [Congressional RecordNovember 14 1991 pp 10239ndash10242]
34 Since we have multiple legislators from each state we adjust the standarderrors to correct for the lack of independence among observations clustered in thesame state
QUARTERLY JOURNAL OF ECONOMICS1458
TAB
LE
IVM
AR
GIN
AL
EF
FE
CT
SF
RO
MA
PR
OB
ITM
OD
EL
OF
TH
EIN
FL
UE
NC
EO
FP
OL
ITIC
AL
EC
ON
OM
YF
AC
TO
RS
ON
TH
EH
OU
SE
VO
TE
ON
TH
EW
YL
IE-N
EA
L
AM
EN
DM
EN
TT
OP
ER
MIT
INT
ER
ST
AT
EB
AN
KB
RA
NC
HIN
G
NO
VE
MB
ER
14
1991
(1)
(2)
(3)
(4)
(5)
(6)
Smal
lban
kas
set
shar
eof
allb
anki
ng
asse
tsin
stat
e2
417
(12
4)2
454
(12
2)2
407
(12
0)2
443
(14
7)2
450
(14
9)2
490
(14
7)C
apit
alra
tio
ofsm
allb
anks
rela
tive
tola
rge
inth
est
ate
180
(37
5)2
68(3
42)
256
(37
5)1
69(3
78)
153
(38
0)2
45(4
25)
Rel
ativ
esi
zeof
insu
ranc
ein
stat
esw
here
bank
sm
ayse
llin
sura
nce
0ot
herw
ise
22
34
(10
2)2
249
(1
14)
22
33
(10
1)2
249
(1
04)
22
51
(10
2)2
154
(11
1)In
dica
tor
is1
ifba
nks
may
sell
insu
ranc
ein
the
stat
e0
79
(0
06)
074
(00
6)0
69
(0
06)
070
(00
6)0
70
(0
06)
067
(00
7)R
elat
ive
size
ofin
sura
nce
inst
ates
whe
reba
nks
may
not
sell
insu
ranc
e0
oth
erw
ise
180
(06
6)2
42
(0
72)
154
(0
67)
173
(06
6)1
72
(0
67)
203
(0
91)
Smal
lrm
shar
eof
the
num
ber
ofr
ms
inth
est
ate
21
23(0
92)
20
83(0
89)
21
62(1
07)
21
32(0
95)
21
30(1
00)
20
94(1
16)
Indi
cato
ris
1if
Rep
rese
ntat
ive
isa
Dem
ocra
t2
035
(00
7)2
035
(00
7)2
036
(00
7)2
035
(00
7)2
035
(00
7)2
036
(00
7)A
vera
geyi
eld
onba
nklo
ans
inth
est
ate
min
usF
edfu
nds
rate
mdash10
01
(51
5)mdash
mdashmdash
mdash
Ave
rage
fail
ure
rate
198
5to
1991
mdashmdash
228
(14
1)mdash
mdashmdash
Indi
cato
ris
1if
stat
eh
asun
itba
nkin
gla
wmdash
mdashmdash
004
(00
7)0
04(0
07)
20
01(0
08)
Indi
cato
ris
1if
stat
ech
ange
sba
nk
insu
ran
cepo
wer
smdash
mdashmdash
mdash2
001
(00
7)2
003
(00
7)In
clud
esre
gion
alin
dica
tors
N
oN
oN
oN
oN
oYe
sP
seud
o-R
20
170
180
170
170
170
18L
ogli
keli
hood
224
024
223
826
223
951
224
013
224
010
223
808
p-va
lue
ofx2
for
regr
essi
on
001
0
01
001
0
01
001
0
01
Th
ede
pen
den
tvar
iabl
eis
one
ifth
ele
gisl
ator
vote
sfo
rth
eam
endm
enta
nd
zero
ifag
ains
tT
he
repo
rted
coef
fici
ents
are
the
effe
cts
ofa
uni
tch
ange
ofth
ein
depe
nde
ntva
riab
le(f
rom
the
mea
n)on
the
prob
abili
tyof
voti
ng
infa
vor
ofth
eam
endm
ent
All
vari
able
sar
em
easu
red
byst
ate
N5
418
the
num
ber
ofR
epre
sent
ativ
esvo
tin
gon
the
amen
dmen
tS
mal
lban
kas
sets
har
eis
the
perc
ent
ofba
nki
ng
asse
tsin
the
stat
eh
eld
byba
nks
belo
wth
em
edia
nsi
zeof
ban
kin
each
stat
ein
each
year
Rel
ativ
eca
pita
lrat
iois
the
capi
talt
oas
sets
rati
oof
smal
lban
ksm
inu
sth
atof
larg
eba
nks
Siz
eof
insu
ran
cere
lati
veto
bank
ing
plu
sin
sura
nce
inth
est
ate
ism
easu
red
asgr
oss
stat
epr
oduc
tfro
min
sura
nce
divi
ded
bygr
oss
stat
epr
odu
ctfr
omin
sura
nce
plu
sba
nkin
gIn
dica
tor
ison
eif
stat
ela
wpe
rmit
sba
nks
toen
ter
the
insu
ran
cebu
sin
ess
Sm
all
rmsh
are
isth
epe
rcen
tofa
lles
tabl
ishm
ents
inth
est
ate
that
hav
efe
wer
than
twen
tyem
ploy
ees
Par
tyin
dica
tor
ison
eif
the
Rep
rese
nta
tive
isa
Dem
ocra
tA
vera
geyi
eld
onba
nklo
ans
equa
lsto
tali
nter
est
inco
me
onal
ldom
esti
clo
ans
mad
eby
ban
ksin
the
stat
edi
vide
dby
tota
lloa
ns
The
aver
age
bank
failu
rera
teis
the
rati
oof
asse
tsin
faile
dba
nks
duri
ng
agi
ven
stat
edi
vide
dby
tota
lban
kas
sets
inth
atst
ate
aver
aged
over
the
year
s19
85to
1991
Un
itba
nki
ng
indi
cato
req
ual
son
efo
rst
ates
wit
hun
itba
nkin
gre
stri
ctio
nsT
hech
ange
inin
sura
nce
pow
ers
indi
cato
req
ual
son
eif
the
stat
ech
ange
dth
ela
wto
perm
itba
nks
tose
llin
sura
nce
duri
ngth
esa
mpl
epe
riod
We
divi
deth
eco
unt
ryin
tofo
urre
gion
san
din
clu
dere
gion
alin
dica
tors
inco
lum
n(5
)R
obu
stst
anda
rder
rors
are
inpa
ren
thes
es
an
d
den
ote
stat
isti
cally
sign
ica
nta
tth
e1
5an
d10
perc
entl
evel
s
WHAT DRIVES DEREGULATION 1459
branching of approximately 17 percent The relative capital-to-assets ratio of small and large banks however does not have astatistically signicant effect in the probit model
The impact of rival interests outside of banking is alsoconsistent with intrastate deregulation results Where banks cansell insurance legislators from states with larger insurancesectors relative to banking are less likely to vote for the amend-ment35 A one-standard-deviation increase in the relative size ofinsurance in those states which permit banks to sell insurancedecreases the probability that a legislator will favor the amend-ment by about 13 percent Turning to consumer interests thecoefficient on small rm share is not statistically signicant butthe coefficient on interest rates in the state is A one-standard-deviation increase in the average interest rate on loans raises theprobability that a legislator will support the amendment byroughly 8 percent The coefficient estimate for the bankingdistress variable measured as the average bank failure rate from1985 to 1991 is small and statistically signicant Overall theprobit analysis of the vote on national branching deregulationsupports the private-interest theory of deregulation and providesa consistency check that the importance of interests operating onthe state legislatures are very similar to those operating at thefederal level
V WHY DID BRANCHING DEREGULATION BEGIN DURING THE
1970S SHOCKS TO THE EQUILIBRIUM SUPPORTING BANK
BRANCHING RESTRICTIONS
A complete explanation of regulatory exit should account forwhy deregulation begins in the 1970s as well as the timing of thestate-by-state reforms [Kane 1996] In this section we describebroad technological legal and economic shocks that altered thepolitical-economy equilibrium which had kept antibranching regu-lations little changed for at least 30 years
Beginning in the 1970s three innovations reduced the valueto the protected banks of local geographic monopolies (see TableV) First automatic teller machines (ATMs) helped to erode thegeographic ties between customers and banks Second checkablemoney market mutual funds and the Merrill Lynch Cash Manage-ment Account demonstrated that banking by mail and telephone
35 The positive and statistically signicant estimates on the other insurancevariables may reect insurance industry support for the amendmentrsquos provisionslimiting banksrsquo insurance powers
QUARTERLY JOURNAL OF ECONOMICS1460
provided a convenient alternative to local banks36 Third techno-logical innovation and deregulation reduced transportation andcommunication costs particularly since the 1970s thereby lower-ing the costs for customers to use distant banks By increasing theelasticity of deposits supplied to banks these innovations reducedthe value of geographical restrictions to their traditional benecia-ries and thereby reduced their incentive to ght to maintain them[Peltzman 1976]
On the lending side increasing sophistication of credit-scoring techniques following innovations in information process-ing nancial theory and the development of large credit databases began to diminish the value of knowledge that localbankers had about the risks of borrowers in the community As a
36 Regulation Q which limited the interest rates that banks could pay ondeposits may have helped to drive depositors away from banks when the gapbetween market rates and deposit ceilings grew during the 1970s The eliminationof interest rate ceilings on large denomination certicates of deposit during the1970s appears to have hurt smaller and retail-oriented banks relative to largerwholesale banks [James 1983]
TABLE VBROAD TRENDS IN COMMERCIAL BANKING 1950ndash1995
YearNumberof ATMs
Domesticbank
deposits(billions)
Moneymarketmutual
fund(billions)
Percent ofdeposits 1
moneyfunds heldby banks
Smallbanksrsquo
percent ofbankingassets
Averagenumber of
bankfailures
(1) (2) (3) (4) (5) (6)1950 0 $ 154 $ 0 100 NA 41955 0 191 0 100 NA 31960 0 228 0 100 24 21965 0 330 0 100 20 41970 0 479 0 100 18 61975 9750 775 4 99 18 61980 18500 1182 76 94 17 101985 61117 1787 242 88 14 601990 80156 2339 493 83 11 1791995 122706 2552 745 77 8 61
Column (1) ATM gures are from Bank Network News The EFT Network Data Book (New YorkFaulkner and Gray Inc) The 1975 gure was unavailable 9750 is the number of ATMs in 1978 the rst yearfor which complete data are available Columns (2)ndash(4) Banksrsquo domestic deposits are from the Reports ofIncome and Condition money market mutual funds are from the Flow of Funds Data on all bank depositsforeign plus domestic are only available beginning in 1970 The trend in banksrsquo share (column (4)) is the sameusing total deposits instead of domestic deposits Column (5) Percent of banking assets held by small bankswhere a small bank is dened as a commercial bank with less than $100 million in assets in 1994 dollarsThese data are based on the Reports of Income and Condition Data on small banks are not available before1960 Column (6) Five-year average number of bank failures where the nal year is indicated in the rstcolumn These data are from FDIC Annual Report and the Quarterly Banking Prole
WHAT DRIVES DEREGULATION 1461
result of these innovations a national market developed forresidential mortgages credit card receivables have been securi-tized and bank lending to small business now relies less on thejudgment of loan officers and more on standardized scoringmodels
These changes have increased the potential protability forlarge banks to enter what had been the core of small bankactivities Large banksrsquo incentive to increase their lobbying pres-sure to be able to expand into these markets has thus beenincreasing over time In fact small banksrsquo market share began todecline even prior to the branching deregulation (Table V) As thevalue of a local banking relationships declined small rms thatwere the main borrowers from the small banks also probablybecame more likely to favor the entry of large banks into localmarkets With the deadweight costs of preventing large bankentry rising the private-interest theory predicts that small localbanks would become less likely able to maintain the branchingrestrictions [Becker 1983] Deregulation that reduces deadweightcosts of regulation is also consistent with the public-interesttheory
Kane [1996] argues that another major shock to the oldequilibrium was the increasing public awareness of the costlinessof having government-insured but (geographically) undiversiednancial institutions In the late 1970s the failure rate of banksbegan to rise (Table V) In the 1980s the Savings and Loan crisisand taxpayer bailout further heightened public awareness aboutthe costs of restrictions that make depository institutions morelikely to require infusions of taxpayer funds The failures thusmay have heightened public support for branching deregulation(see also Abrams and Settle [1993]) As noted above we did notnd any evidence that banking failures or distress in a stateaffected the speed with which the state deregulated but distressstill may be a common factor affecting all states
These technological economic and legal shocks generatedconditions that changed the long-standing balance favoring theantibranching forces The marginal value of lobbying to repealbranching restrictions increased just as the relative value to thesmall banks of maintaining branching restrictions was decliningThese nationwide shocks are common factors across states andthey are consistent with the positive duration dependence associ-ated with state-level deregulation that we demonstrated in FigureII In addition a broad change in lsquolsquoideologyrsquorsquo against government
QUARTERLY JOURNAL OF ECONOMICS1462
regulation could explain the positive duration dependence duringthe sample period
VI CONCLUSIONS
The private-interest theory of regulation can account for thepattern of bank branching deregulation of the 1970s and 1980sBeneciaries of branching regulation had supported a coalitionfavoring geographical restrictions despite their costs to consum-ers of nancial services Innovations that began in the 1970saltered the value of the restrictions to the affected parties and theresulting competition among interest groups can explain thesubsequent deregulation While some of our results are alsoconsistent with the public interest theorymdashfor example deregula-tion occurs earlier when small banks are in a relatively weaknancial positionmdashother results particularly evidence on theimportance of rivalries between small and large banks andbetween banking and insurance are difficult to explain with thepublic-interest approach We also nd that political-institutionalfactors affect deregulation although these variables may act asproxies for unmeasured economic interests The greater success ofthe private-interest theory to explain deregulation here than inprevious work may be due to the richer cross-sectional andtime-series variation in branching deregulation that can beexploited with a hazard model Future empirical studies ofendogenous deregulation thus may be most fruitful where changehas occurred across states over time such as in franchisinginsurance and public utilities (eg Joskow [1996])
Technological and nancial innovations will continue to erodethe benets to any interest group of maintaining regulatorybarriers in nancial services These forces are likely to bringabout reforms both domestically for example through legislationthat would increase bank powers (see Kroszner [1996] Krosznerand Rajan [1994 1997] and Kroszner and Stratmann [1998]) andinternationally for example through the extension of nancialservices provisions of NAFTAto reduce geographic barriers acrosscountries [Kroszner 1997]
APPENDIX EFFECTS OF DIFFERENT TYPES OF BRANCHING
DEREGULATION ON ECONOMIC GROWTH BANKING STRUCTUREAND BANK EFFICIENCY
We measure the importance of alternative forms of deregula-tion by their consequences for the economic growth and the
WHAT DRIVES DEREGULATION 1463
structure and efficiency of the banking sector across the states(see Jayaratne and Strahan [1996 1998]) To do so we run anOLS panel regression relating indicators for each type of deregula-tion to potential consequences Our dependent variables includeproxies for economic growth (growth in personal income) bankstructure (the log of the number of banking companies and thenumber of branches per banking company) and efficiency (the logof total noninterest expenses and the ratio of loan loss provisionsto total loans) The four deregulation indicators (MBHC branch-ing through merger and acquisition unrestricted statewide branch-ing and interstate banking) are equal to one in the years afterderegulation and zero before If a state deregulated before thesample period begins in 1970 the indicator is one throughout ifthe state does not deregulate then the indicator is zero through-out All regressions include state xed effects and time effects
Table VI shows the results of including the set of four
TABLE VIMULTIVARIATE REGRESSIONS OF THE EFFECTS OF DIFFERENT TYPES OF BRANCHING
DEREGULATION ON ECONOMIC GROWTH BANKING STRUCTUREAND BANK EFFICIENCY
Dependent variable
Growthin
personalincome
Log of of
bankingcompanies
Branchesper
bankingcompany
Log ofnon-
interestexpenses
Loan lossprovisions
(loans 1leases)
Merger amp acquisitionbranchingindicator
0013(0004)
2 0162(0037)
1146(0453)
2 0107(0035)
2 0007(0001)
Unrestricted state-wide branchingindicator
2 0008(0005)
0086(0059)
2 0043(0574)
0070(0037)
0005(0002)
Interstate bankingindicator
0006(0005)
2 0052(0057)
1064(0511)
0005(0030)
2 0001(0001)
Multibank holdingcompany indicator
2 0005(0006)
2 0096(0052)
2 0364(0529)
2 0095(0036)
2 0001(0001)
Dependent variablemean [median]
0086[0082]
4881[5136]
6649[5055]
13812[13803]
0009[0006]
This table contains regressions of state-level economic growth and state-level measures of the structureand efficiency of the banking industry Each dependent variable is regressed on a set of state indicatorvariables a set of year indicator variables and the set of four deregulation indicators these deregulationindicators are equal to one after deregulation and zero before If a state was always regulated the indicator iszero throughout if it was always deregulated the indicator is one throughout Sample period is 1975 to 1991and Delaware and South Dakota are dropped so N 5 833 Noninterest expenses become available only in1984 so N 5 392 The coefficient on the deregulation indicators is reported with robust standard errors inparentheses and denote statistically signicant at the 1 5 and 10 percent levels
QUARTERLY JOURNAL OF ECONOMICS1464
deregulation indicators simultaneously The branching by mergerand acquisition indicator is statistically signicant and of theexpected sign in all ve of the equations and larger in absolutevalue than any of the other coefficients on the indicators for theother types of deregulation Since there may be multicollinearityamong our indicators we also ran univariate specications whichincluded each indicator separately and the results are verysimilar These results imply that branching by merger andacquisition is the most important type of deregulation measuredby its consequences for economic growth banking structure andbank efficiency See Kroszner and Strahan [1998] for more details
UNIVERSITY OF CHICAGO GRADUATE SCHOOL OF BUSINESS AND NATIONAL BUREAU OF
ECONOMIC RESEARCH
FEDERAL RESERVE BANK OF NEW YORK RESEARCH amp MARKET ANALYSIS GROUP
REFERENCES
Abrams Burton A and Russell F Settle lsquolsquoPressure-Group Inuence and Institu-tional Change Branch-Banking Legislation during the Great DepressionrsquorsquoPublic Choice LXXVII (1993) 687ndash705
Amel Dean lsquolsquoState Laws Affecting the Geographic Expansion of CommercialBanksrsquorsquo unpublished manuscript Board of Governors of the Federal ReserveSystem 1993
Becker Gary S lsquolsquoA Theory of Competition among Pressure Groups for PoliticalInuencersquorsquo Quarterly Journal of Economics XCVIII (1983) 371ndash400
Berger Allen Anil Kashyap Joseph Scalise lsquolsquoThe Transformation of the U SBanking Industry What a Long Strange Trip Itrsquos Beenrsquorsquo Brookings Papers onEconomic Activity 2 (1995) 55ndash218
Berger Allen Anthony Saunders Joseph Scalise and Gregory Udell lsquolsquoThe Effectsof Bank Mergers and Acquisitions on Small Business Lendingrsquorsquo Journal ofFinancial Economics L (1998) 187ndash229
Brennan Geoffrey and James Buchanan lsquolsquoTowards a Tax Constitution forLeviathanrsquorsquo Journal of Public Economics VIII (1977) 255ndash273
Brickley James A and Christopher M James lsquolsquoThe Takeover Market CorporateBoard Composition and Ownership Structure The Case of Bankingrsquorsquo Journalof Law and Economics XXX (1987) 161ndash180
Cole Rebel and John Wolken lsquolsquoFinancial Services Used by Small BusinessesEvidence from the 1993 Survey of Small Business Financesrsquorsquo Federal ReserveBulletin LXXXI (1994) 629ndash676
Dixit Avinash The Making of Economic Policy A Transactions-Cost PoliticsPerspective (Cambridge MA MIT Press 1996)
Economides Nicolas R Glenn Hubbard and Darius Palia lsquolsquoThe Political Economyof Branching Restrictions and Deposit Insurancersquorsquo Journal of Law andEconomics XXIX (1996) 667ndash704
Flannery Mark lsquolsquoThe Social Costs of Unit Banking Restrictionsrsquorsquo Journal ofMonetary Economics XIII (1984) 237ndash249
Greene William H Econometric Analysis third edition (Upper Saddle River NJPrentice-Hall 1997)
Gunther Jeffery lsquolsquoRegional Capital Imbalances and the Removal of InterstateBanking Restrictionsrsquorsquo Economics Letters XLIV (1994) 439ndash442 lsquolsquoThe Erosion of Intrastate Branch Banking Restrictionsrsquorsquo unpublishedmanuscript Federal Reserve Bank of Dallas 1996
Irwin Douglas and Randall Kroszner lsquolsquoInterests Institutions and Ideology inSecuring Policy Change The Republican Conversion to Trade Liberalizationafter Smoot-Hawleyrsquorsquo Journal of Law and Economics XLII (1999) 643ndash673
WHAT DRIVES DEREGULATION 1465
James Christopher lsquolsquoAn Analysis of Intra-Industry Differences in the Effect ofRegulation The Case of Deposit Rate Ceilingsrsquorsquo Journal of Monetary Econom-ics XII (1983) 417ndash442
Jarrell Gregg lsquolsquoChange at the Exchange The Causes and Effects of DeregulationrsquorsquoJournal of Law and Economics XXVII (1984) 273ndash312
Jayaratne Jith and Philip E Strahan lsquolsquoThe Finance-Growth Nexus Evidencefrom Bank Branch Deregulationrsquorsquo Quarterly Journal of Economics CI (1996)639ndash670
Jayaratne Jith and Philip E Strahan lsquolsquoEntry Restrictions Industry Evolutionand Dynamic Efficiency Evidence from Commercial Bankingrsquorsquo Journal of Lawand Economics XLIX (1998) 239ndash274
Joskow Paul lsquolsquoComment on lsquoPower Struggles Explaining Deregulatory Reformsin Electricity Marketsrsquo rsquorsquo Brookings Papers on Economic Activity Microeconom-ics (1996) 251ndash264
Joskow Paul and Roger Noll lsquolsquoRegulation in Theory and Practice An Overviewrsquorsquoin Studies in Public Regulation Gary Fromm ed (Cambridge MA MITPress 1981) pp 1ndash65
Kahn Matthew E and John G Matsusaka lsquolsquoDemand for Environmental GoodsEvidence from Voting Patterns on California Initiativesrsquorsquo Journal of Law andEconomics XL (1997) 137ndash173
Kalbeisch J and R Prentice The Statistical Analysis of Failure Time Data (NewYork John Wiley amp Sons 1980)
Kane Edward J lsquolsquoDe Jure Interstate Banking Why Only NowrsquorsquoJournal of MoneyCredit and Banking XXVIII (1996) 141ndash161
Kiefer Nicholas lsquolsquoEconomic Duration Data and Hazard Functionsrsquorsquo Journal ofEconomic Literature XXVI (1988) 646ndash679
Kroszner Randall S lsquolsquoThe Evolution of Universal Banking and Its Regulation inTwentieth Century Americarsquorsquo in Universal Banking Financial System DesignReconsidered Anthony Saunders and Ingo Walter eds (New York IrwinProfessional Publishers 1996) pp 70ndash99 lsquolsquoThe Political Economy of Banking and Financial Regulation in the U Srsquorsquo inThe Banking and Financial Structure in the NAFTA Countries and ChileGeorge M von Furstenberg ed (Boston Kluwer Academic Publishers 1997)pp 200ndash213
Kroszner Randall S and Raghuram G Rajan lsquolsquoIs the Glass-Steagall ActJustied A Study of the U S Experience with Universal Banking before1933rsquorsquo American Economic Review LXXXIV (1994) 810ndash832
Kroszner Randall S and Raghuram G Rajan lsquolsquoOrganization Structure andCredibility Evidence from Bank Securities Activities before the Glass-Steagall Actrsquorsquo Journal of Monetary Economics XXXIX (1997) 475ndash516
Kroszner Randall S and Philip E Strahan lsquolsquoRegulatory Incentives and the ThriftCrisis Dividends Mutual-to-Stock Conversions and Financial DistressrsquorsquoJournal of Finance LI (1996) 1285ndash1320
Kroszner Randall S and Philip E Strahan lsquolsquoWhat Drives Deregulation Econom-ics and Politics of the Relaxation of Bank Branching Restrictionsrsquorsquo NBERWorking Paper No 6637 1998
Kroszner Randall S and Thomas Stratmann lsquolsquoInterest Group Competition andthe Organization of Congress Theory and Evidence from Financial ServicesPolitical Action Committeesrsquorsquo American Economic Review LXXXVIII (1998)pp 1163ndash1187
Kroszner Randall S and Thomas Stratmann lsquolsquoDoes Political Ambiguity PayCorporate Campaign Contributions and Rewards to Legislator Reputationrsquorsquounpublishedmanuscript University of Chicago Graduate School of Business1999
Lin D Y and L J Wei lsquolsquoThe Robust Inference for the Cox Proportional HazardsModel Journal of the American Statistical Association LXXXIV (1989)1074ndash1078
Macey Jonathan R and Geoffrey P Miller Banking Law and Regulation (BostonLittle Brown and Company 1992)
Makinson Larry Open Secrets The Encyclopedia of Congressional Money andPolitics (Washington DC Congressional Quarterly 1992)
Niskanen WilliamBureaucracy and Representative Government (Chicago Aldine-Atherton 1971)
QUARTERLY JOURNAL OF ECONOMICS1466
Noll Roger lsquolsquoComment on Peltzmanrsquorsquo Brookings Papers Microeconomics SpecialIssue (1989) 48ndash58
North Douglass lsquolsquoA Transactions Cost Theory of Politicsrsquorsquo Journal of TheoreticalPolitics II (1990) 355ndash367
Peltzman Sam lsquolsquoToward a More General Theory of Regulationrsquorsquo Journal of Lawand Economics XIX (1976) 109ndash148 lsquolsquoConstituent Interest and Congressional Votingrsquorsquo Journal of Law andEconomics XXVII (1984) 181ndash210 lsquolsquoThe Economic Theory of Regulation after a Decade of DeregulationrsquorsquoBrookings Papers Microeconomics Special Issue (1989) 1ndash41
Peterson Mitchell and Raghuram Rajan lsquolsquoThe Benets of Lending RelationshipsEvidence from Small Business Datarsquorsquo Journal of Finance XLIX (1994) 3ndash38
Poole Keith and Howard Rosenthal Congress A Political-Economic History ofRoll Call Voting (Oxford Oxford University Press 1997)
Stata Stata Reference Manual Release 5 (College Station TX Stata Press 1997)Stigler George J lsquolsquoThe Theory of Economic Regulationrsquorsquo Bell Journal of Economics
and Management Science II (1971) 3ndash21Stigler George J Chicago Studies in the Political Economy (Chicago University
of Chicago Press 1988)Strahan Philip E and James Weston lsquolsquoSmall Business Lending and the Changing
Structure of the Banking Industryrsquorsquo Journal of Banking and Finance XXII(1998) 821ndash845
Sylla Richard John Legler and John Wallis lsquolsquoBanks and State Public Finance inthe New Republic The United States 1790ndash1860rsquorsquo Journal of EconomicHistory XLVII (1987) 391ndash403
White Eugene The Regulation and Reform of the American Banking System1900ndash1929 (Princeton Princeton University Press 1983)
Winston Clifford lsquolsquoEconomic Deregulation Days of Reckoning for Microecono-mistsrsquorsquo Journal of Economic Literature XXXI (1993) 1263ndash1289
WHAT DRIVES DEREGULATION 1467
TAB
LE
IVM
AR
GIN
AL
EF
FE
CT
SF
RO
MA
PR
OB
ITM
OD
EL
OF
TH
EIN
FL
UE
NC
EO
FP
OL
ITIC
AL
EC
ON
OM
YF
AC
TO
RS
ON
TH
EH
OU
SE
VO
TE
ON
TH
EW
YL
IE-N
EA
L
AM
EN
DM
EN
TT
OP
ER
MIT
INT
ER
ST
AT
EB
AN
KB
RA
NC
HIN
G
NO
VE
MB
ER
14
1991
(1)
(2)
(3)
(4)
(5)
(6)
Smal
lban
kas
set
shar
eof
allb
anki
ng
asse
tsin
stat
e2
417
(12
4)2
454
(12
2)2
407
(12
0)2
443
(14
7)2
450
(14
9)2
490
(14
7)C
apit
alra
tio
ofsm
allb
anks
rela
tive
tola
rge
inth
est
ate
180
(37
5)2
68(3
42)
256
(37
5)1
69(3
78)
153
(38
0)2
45(4
25)
Rel
ativ
esi
zeof
insu
ranc
ein
stat
esw
here
bank
sm
ayse
llin
sura
nce
0ot
herw
ise
22
34
(10
2)2
249
(1
14)
22
33
(10
1)2
249
(1
04)
22
51
(10
2)2
154
(11
1)In
dica
tor
is1
ifba
nks
may
sell
insu
ranc
ein
the
stat
e0
79
(0
06)
074
(00
6)0
69
(0
06)
070
(00
6)0
70
(0
06)
067
(00
7)R
elat
ive
size
ofin
sura
nce
inst
ates
whe
reba
nks
may
not
sell
insu
ranc
e0
oth
erw
ise
180
(06
6)2
42
(0
72)
154
(0
67)
173
(06
6)1
72
(0
67)
203
(0
91)
Smal
lrm
shar
eof
the
num
ber
ofr
ms
inth
est
ate
21
23(0
92)
20
83(0
89)
21
62(1
07)
21
32(0
95)
21
30(1
00)
20
94(1
16)
Indi
cato
ris
1if
Rep
rese
ntat
ive
isa
Dem
ocra
t2
035
(00
7)2
035
(00
7)2
036
(00
7)2
035
(00
7)2
035
(00
7)2
036
(00
7)A
vera
geyi
eld
onba
nklo
ans
inth
est
ate
min
usF
edfu
nds
rate
mdash10
01
(51
5)mdash
mdashmdash
mdash
Ave
rage
fail
ure
rate
198
5to
1991
mdashmdash
228
(14
1)mdash
mdashmdash
Indi
cato
ris
1if
stat
eh
asun
itba
nkin
gla
wmdash
mdashmdash
004
(00
7)0
04(0
07)
20
01(0
08)
Indi
cato
ris
1if
stat
ech
ange
sba
nk
insu
ran
cepo
wer
smdash
mdashmdash
mdash2
001
(00
7)2
003
(00
7)In
clud
esre
gion
alin
dica
tors
N
oN
oN
oN
oN
oYe
sP
seud
o-R
20
170
180
170
170
170
18L
ogli
keli
hood
224
024
223
826
223
951
224
013
224
010
223
808
p-va
lue
ofx2
for
regr
essi
on
001
0
01
001
0
01
001
0
01
Th
ede
pen
den
tvar
iabl
eis
one
ifth
ele
gisl
ator
vote
sfo
rth
eam
endm
enta
nd
zero
ifag
ains
tT
he
repo
rted
coef
fici
ents
are
the
effe
cts
ofa
uni
tch
ange
ofth
ein
depe
nde
ntva
riab
le(f
rom
the
mea
n)on
the
prob
abili
tyof
voti
ng
infa
vor
ofth
eam
endm
ent
All
vari
able
sar
em
easu
red
byst
ate
N5
418
the
num
ber
ofR
epre
sent
ativ
esvo
tin
gon
the
amen
dmen
tS
mal
lban
kas
sets
har
eis
the
perc
ent
ofba
nki
ng
asse
tsin
the
stat
eh
eld
byba
nks
belo
wth
em
edia
nsi
zeof
ban
kin
each
stat
ein
each
year
Rel
ativ
eca
pita
lrat
iois
the
capi
talt
oas
sets
rati
oof
smal
lban
ksm
inu
sth
atof
larg
eba
nks
Siz
eof
insu
ran
cere
lati
veto
bank
ing
plu
sin
sura
nce
inth
est
ate
ism
easu
red
asgr
oss
stat
epr
oduc
tfro
min
sura
nce
divi
ded
bygr
oss
stat
epr
odu
ctfr
omin
sura
nce
plu
sba
nkin
gIn
dica
tor
ison
eif
stat
ela
wpe
rmit
sba
nks
toen
ter
the
insu
ran
cebu
sin
ess
Sm
all
rmsh
are
isth
epe
rcen
tofa
lles
tabl
ishm
ents
inth
est
ate
that
hav
efe
wer
than
twen
tyem
ploy
ees
Par
tyin
dica
tor
ison
eif
the
Rep
rese
nta
tive
isa
Dem
ocra
tA
vera
geyi
eld
onba
nklo
ans
equa
lsto
tali
nter
est
inco
me
onal
ldom
esti
clo
ans
mad
eby
ban
ksin
the
stat
edi
vide
dby
tota
lloa
ns
The
aver
age
bank
failu
rera
teis
the
rati
oof
asse
tsin
faile
dba
nks
duri
ng
agi
ven
stat
edi
vide
dby
tota
lban
kas
sets
inth
atst
ate
aver
aged
over
the
year
s19
85to
1991
Un
itba
nki
ng
indi
cato
req
ual
son
efo
rst
ates
wit
hun
itba
nkin
gre
stri
ctio
nsT
hech
ange
inin
sura
nce
pow
ers
indi
cato
req
ual
son
eif
the
stat
ech
ange
dth
ela
wto
perm
itba
nks
tose
llin
sura
nce
duri
ngth
esa
mpl
epe
riod
We
divi
deth
eco
unt
ryin
tofo
urre
gion
san
din
clu
dere
gion
alin
dica
tors
inco
lum
n(5
)R
obu
stst
anda
rder
rors
are
inpa
ren
thes
es
an
d
den
ote
stat
isti
cally
sign
ica
nta
tth
e1
5an
d10
perc
entl
evel
s
WHAT DRIVES DEREGULATION 1459
branching of approximately 17 percent The relative capital-to-assets ratio of small and large banks however does not have astatistically signicant effect in the probit model
The impact of rival interests outside of banking is alsoconsistent with intrastate deregulation results Where banks cansell insurance legislators from states with larger insurancesectors relative to banking are less likely to vote for the amend-ment35 A one-standard-deviation increase in the relative size ofinsurance in those states which permit banks to sell insurancedecreases the probability that a legislator will favor the amend-ment by about 13 percent Turning to consumer interests thecoefficient on small rm share is not statistically signicant butthe coefficient on interest rates in the state is A one-standard-deviation increase in the average interest rate on loans raises theprobability that a legislator will support the amendment byroughly 8 percent The coefficient estimate for the bankingdistress variable measured as the average bank failure rate from1985 to 1991 is small and statistically signicant Overall theprobit analysis of the vote on national branching deregulationsupports the private-interest theory of deregulation and providesa consistency check that the importance of interests operating onthe state legislatures are very similar to those operating at thefederal level
V WHY DID BRANCHING DEREGULATION BEGIN DURING THE
1970S SHOCKS TO THE EQUILIBRIUM SUPPORTING BANK
BRANCHING RESTRICTIONS
A complete explanation of regulatory exit should account forwhy deregulation begins in the 1970s as well as the timing of thestate-by-state reforms [Kane 1996] In this section we describebroad technological legal and economic shocks that altered thepolitical-economy equilibrium which had kept antibranching regu-lations little changed for at least 30 years
Beginning in the 1970s three innovations reduced the valueto the protected banks of local geographic monopolies (see TableV) First automatic teller machines (ATMs) helped to erode thegeographic ties between customers and banks Second checkablemoney market mutual funds and the Merrill Lynch Cash Manage-ment Account demonstrated that banking by mail and telephone
35 The positive and statistically signicant estimates on the other insurancevariables may reect insurance industry support for the amendmentrsquos provisionslimiting banksrsquo insurance powers
QUARTERLY JOURNAL OF ECONOMICS1460
provided a convenient alternative to local banks36 Third techno-logical innovation and deregulation reduced transportation andcommunication costs particularly since the 1970s thereby lower-ing the costs for customers to use distant banks By increasing theelasticity of deposits supplied to banks these innovations reducedthe value of geographical restrictions to their traditional benecia-ries and thereby reduced their incentive to ght to maintain them[Peltzman 1976]
On the lending side increasing sophistication of credit-scoring techniques following innovations in information process-ing nancial theory and the development of large credit databases began to diminish the value of knowledge that localbankers had about the risks of borrowers in the community As a
36 Regulation Q which limited the interest rates that banks could pay ondeposits may have helped to drive depositors away from banks when the gapbetween market rates and deposit ceilings grew during the 1970s The eliminationof interest rate ceilings on large denomination certicates of deposit during the1970s appears to have hurt smaller and retail-oriented banks relative to largerwholesale banks [James 1983]
TABLE VBROAD TRENDS IN COMMERCIAL BANKING 1950ndash1995
YearNumberof ATMs
Domesticbank
deposits(billions)
Moneymarketmutual
fund(billions)
Percent ofdeposits 1
moneyfunds heldby banks
Smallbanksrsquo
percent ofbankingassets
Averagenumber of
bankfailures
(1) (2) (3) (4) (5) (6)1950 0 $ 154 $ 0 100 NA 41955 0 191 0 100 NA 31960 0 228 0 100 24 21965 0 330 0 100 20 41970 0 479 0 100 18 61975 9750 775 4 99 18 61980 18500 1182 76 94 17 101985 61117 1787 242 88 14 601990 80156 2339 493 83 11 1791995 122706 2552 745 77 8 61
Column (1) ATM gures are from Bank Network News The EFT Network Data Book (New YorkFaulkner and Gray Inc) The 1975 gure was unavailable 9750 is the number of ATMs in 1978 the rst yearfor which complete data are available Columns (2)ndash(4) Banksrsquo domestic deposits are from the Reports ofIncome and Condition money market mutual funds are from the Flow of Funds Data on all bank depositsforeign plus domestic are only available beginning in 1970 The trend in banksrsquo share (column (4)) is the sameusing total deposits instead of domestic deposits Column (5) Percent of banking assets held by small bankswhere a small bank is dened as a commercial bank with less than $100 million in assets in 1994 dollarsThese data are based on the Reports of Income and Condition Data on small banks are not available before1960 Column (6) Five-year average number of bank failures where the nal year is indicated in the rstcolumn These data are from FDIC Annual Report and the Quarterly Banking Prole
WHAT DRIVES DEREGULATION 1461
result of these innovations a national market developed forresidential mortgages credit card receivables have been securi-tized and bank lending to small business now relies less on thejudgment of loan officers and more on standardized scoringmodels
These changes have increased the potential protability forlarge banks to enter what had been the core of small bankactivities Large banksrsquo incentive to increase their lobbying pres-sure to be able to expand into these markets has thus beenincreasing over time In fact small banksrsquo market share began todecline even prior to the branching deregulation (Table V) As thevalue of a local banking relationships declined small rms thatwere the main borrowers from the small banks also probablybecame more likely to favor the entry of large banks into localmarkets With the deadweight costs of preventing large bankentry rising the private-interest theory predicts that small localbanks would become less likely able to maintain the branchingrestrictions [Becker 1983] Deregulation that reduces deadweightcosts of regulation is also consistent with the public-interesttheory
Kane [1996] argues that another major shock to the oldequilibrium was the increasing public awareness of the costlinessof having government-insured but (geographically) undiversiednancial institutions In the late 1970s the failure rate of banksbegan to rise (Table V) In the 1980s the Savings and Loan crisisand taxpayer bailout further heightened public awareness aboutthe costs of restrictions that make depository institutions morelikely to require infusions of taxpayer funds The failures thusmay have heightened public support for branching deregulation(see also Abrams and Settle [1993]) As noted above we did notnd any evidence that banking failures or distress in a stateaffected the speed with which the state deregulated but distressstill may be a common factor affecting all states
These technological economic and legal shocks generatedconditions that changed the long-standing balance favoring theantibranching forces The marginal value of lobbying to repealbranching restrictions increased just as the relative value to thesmall banks of maintaining branching restrictions was decliningThese nationwide shocks are common factors across states andthey are consistent with the positive duration dependence associ-ated with state-level deregulation that we demonstrated in FigureII In addition a broad change in lsquolsquoideologyrsquorsquo against government
QUARTERLY JOURNAL OF ECONOMICS1462
regulation could explain the positive duration dependence duringthe sample period
VI CONCLUSIONS
The private-interest theory of regulation can account for thepattern of bank branching deregulation of the 1970s and 1980sBeneciaries of branching regulation had supported a coalitionfavoring geographical restrictions despite their costs to consum-ers of nancial services Innovations that began in the 1970saltered the value of the restrictions to the affected parties and theresulting competition among interest groups can explain thesubsequent deregulation While some of our results are alsoconsistent with the public interest theorymdashfor example deregula-tion occurs earlier when small banks are in a relatively weaknancial positionmdashother results particularly evidence on theimportance of rivalries between small and large banks andbetween banking and insurance are difficult to explain with thepublic-interest approach We also nd that political-institutionalfactors affect deregulation although these variables may act asproxies for unmeasured economic interests The greater success ofthe private-interest theory to explain deregulation here than inprevious work may be due to the richer cross-sectional andtime-series variation in branching deregulation that can beexploited with a hazard model Future empirical studies ofendogenous deregulation thus may be most fruitful where changehas occurred across states over time such as in franchisinginsurance and public utilities (eg Joskow [1996])
Technological and nancial innovations will continue to erodethe benets to any interest group of maintaining regulatorybarriers in nancial services These forces are likely to bringabout reforms both domestically for example through legislationthat would increase bank powers (see Kroszner [1996] Krosznerand Rajan [1994 1997] and Kroszner and Stratmann [1998]) andinternationally for example through the extension of nancialservices provisions of NAFTAto reduce geographic barriers acrosscountries [Kroszner 1997]
APPENDIX EFFECTS OF DIFFERENT TYPES OF BRANCHING
DEREGULATION ON ECONOMIC GROWTH BANKING STRUCTUREAND BANK EFFICIENCY
We measure the importance of alternative forms of deregula-tion by their consequences for the economic growth and the
WHAT DRIVES DEREGULATION 1463
structure and efficiency of the banking sector across the states(see Jayaratne and Strahan [1996 1998]) To do so we run anOLS panel regression relating indicators for each type of deregula-tion to potential consequences Our dependent variables includeproxies for economic growth (growth in personal income) bankstructure (the log of the number of banking companies and thenumber of branches per banking company) and efficiency (the logof total noninterest expenses and the ratio of loan loss provisionsto total loans) The four deregulation indicators (MBHC branch-ing through merger and acquisition unrestricted statewide branch-ing and interstate banking) are equal to one in the years afterderegulation and zero before If a state deregulated before thesample period begins in 1970 the indicator is one throughout ifthe state does not deregulate then the indicator is zero through-out All regressions include state xed effects and time effects
Table VI shows the results of including the set of four
TABLE VIMULTIVARIATE REGRESSIONS OF THE EFFECTS OF DIFFERENT TYPES OF BRANCHING
DEREGULATION ON ECONOMIC GROWTH BANKING STRUCTUREAND BANK EFFICIENCY
Dependent variable
Growthin
personalincome
Log of of
bankingcompanies
Branchesper
bankingcompany
Log ofnon-
interestexpenses
Loan lossprovisions
(loans 1leases)
Merger amp acquisitionbranchingindicator
0013(0004)
2 0162(0037)
1146(0453)
2 0107(0035)
2 0007(0001)
Unrestricted state-wide branchingindicator
2 0008(0005)
0086(0059)
2 0043(0574)
0070(0037)
0005(0002)
Interstate bankingindicator
0006(0005)
2 0052(0057)
1064(0511)
0005(0030)
2 0001(0001)
Multibank holdingcompany indicator
2 0005(0006)
2 0096(0052)
2 0364(0529)
2 0095(0036)
2 0001(0001)
Dependent variablemean [median]
0086[0082]
4881[5136]
6649[5055]
13812[13803]
0009[0006]
This table contains regressions of state-level economic growth and state-level measures of the structureand efficiency of the banking industry Each dependent variable is regressed on a set of state indicatorvariables a set of year indicator variables and the set of four deregulation indicators these deregulationindicators are equal to one after deregulation and zero before If a state was always regulated the indicator iszero throughout if it was always deregulated the indicator is one throughout Sample period is 1975 to 1991and Delaware and South Dakota are dropped so N 5 833 Noninterest expenses become available only in1984 so N 5 392 The coefficient on the deregulation indicators is reported with robust standard errors inparentheses and denote statistically signicant at the 1 5 and 10 percent levels
QUARTERLY JOURNAL OF ECONOMICS1464
deregulation indicators simultaneously The branching by mergerand acquisition indicator is statistically signicant and of theexpected sign in all ve of the equations and larger in absolutevalue than any of the other coefficients on the indicators for theother types of deregulation Since there may be multicollinearityamong our indicators we also ran univariate specications whichincluded each indicator separately and the results are verysimilar These results imply that branching by merger andacquisition is the most important type of deregulation measuredby its consequences for economic growth banking structure andbank efficiency See Kroszner and Strahan [1998] for more details
UNIVERSITY OF CHICAGO GRADUATE SCHOOL OF BUSINESS AND NATIONAL BUREAU OF
ECONOMIC RESEARCH
FEDERAL RESERVE BANK OF NEW YORK RESEARCH amp MARKET ANALYSIS GROUP
REFERENCES
Abrams Burton A and Russell F Settle lsquolsquoPressure-Group Inuence and Institu-tional Change Branch-Banking Legislation during the Great DepressionrsquorsquoPublic Choice LXXVII (1993) 687ndash705
Amel Dean lsquolsquoState Laws Affecting the Geographic Expansion of CommercialBanksrsquorsquo unpublished manuscript Board of Governors of the Federal ReserveSystem 1993
Becker Gary S lsquolsquoA Theory of Competition among Pressure Groups for PoliticalInuencersquorsquo Quarterly Journal of Economics XCVIII (1983) 371ndash400
Berger Allen Anil Kashyap Joseph Scalise lsquolsquoThe Transformation of the U SBanking Industry What a Long Strange Trip Itrsquos Beenrsquorsquo Brookings Papers onEconomic Activity 2 (1995) 55ndash218
Berger Allen Anthony Saunders Joseph Scalise and Gregory Udell lsquolsquoThe Effectsof Bank Mergers and Acquisitions on Small Business Lendingrsquorsquo Journal ofFinancial Economics L (1998) 187ndash229
Brennan Geoffrey and James Buchanan lsquolsquoTowards a Tax Constitution forLeviathanrsquorsquo Journal of Public Economics VIII (1977) 255ndash273
Brickley James A and Christopher M James lsquolsquoThe Takeover Market CorporateBoard Composition and Ownership Structure The Case of Bankingrsquorsquo Journalof Law and Economics XXX (1987) 161ndash180
Cole Rebel and John Wolken lsquolsquoFinancial Services Used by Small BusinessesEvidence from the 1993 Survey of Small Business Financesrsquorsquo Federal ReserveBulletin LXXXI (1994) 629ndash676
Dixit Avinash The Making of Economic Policy A Transactions-Cost PoliticsPerspective (Cambridge MA MIT Press 1996)
Economides Nicolas R Glenn Hubbard and Darius Palia lsquolsquoThe Political Economyof Branching Restrictions and Deposit Insurancersquorsquo Journal of Law andEconomics XXIX (1996) 667ndash704
Flannery Mark lsquolsquoThe Social Costs of Unit Banking Restrictionsrsquorsquo Journal ofMonetary Economics XIII (1984) 237ndash249
Greene William H Econometric Analysis third edition (Upper Saddle River NJPrentice-Hall 1997)
Gunther Jeffery lsquolsquoRegional Capital Imbalances and the Removal of InterstateBanking Restrictionsrsquorsquo Economics Letters XLIV (1994) 439ndash442 lsquolsquoThe Erosion of Intrastate Branch Banking Restrictionsrsquorsquo unpublishedmanuscript Federal Reserve Bank of Dallas 1996
Irwin Douglas and Randall Kroszner lsquolsquoInterests Institutions and Ideology inSecuring Policy Change The Republican Conversion to Trade Liberalizationafter Smoot-Hawleyrsquorsquo Journal of Law and Economics XLII (1999) 643ndash673
WHAT DRIVES DEREGULATION 1465
James Christopher lsquolsquoAn Analysis of Intra-Industry Differences in the Effect ofRegulation The Case of Deposit Rate Ceilingsrsquorsquo Journal of Monetary Econom-ics XII (1983) 417ndash442
Jarrell Gregg lsquolsquoChange at the Exchange The Causes and Effects of DeregulationrsquorsquoJournal of Law and Economics XXVII (1984) 273ndash312
Jayaratne Jith and Philip E Strahan lsquolsquoThe Finance-Growth Nexus Evidencefrom Bank Branch Deregulationrsquorsquo Quarterly Journal of Economics CI (1996)639ndash670
Jayaratne Jith and Philip E Strahan lsquolsquoEntry Restrictions Industry Evolutionand Dynamic Efficiency Evidence from Commercial Bankingrsquorsquo Journal of Lawand Economics XLIX (1998) 239ndash274
Joskow Paul lsquolsquoComment on lsquoPower Struggles Explaining Deregulatory Reformsin Electricity Marketsrsquo rsquorsquo Brookings Papers on Economic Activity Microeconom-ics (1996) 251ndash264
Joskow Paul and Roger Noll lsquolsquoRegulation in Theory and Practice An Overviewrsquorsquoin Studies in Public Regulation Gary Fromm ed (Cambridge MA MITPress 1981) pp 1ndash65
Kahn Matthew E and John G Matsusaka lsquolsquoDemand for Environmental GoodsEvidence from Voting Patterns on California Initiativesrsquorsquo Journal of Law andEconomics XL (1997) 137ndash173
Kalbeisch J and R Prentice The Statistical Analysis of Failure Time Data (NewYork John Wiley amp Sons 1980)
Kane Edward J lsquolsquoDe Jure Interstate Banking Why Only NowrsquorsquoJournal of MoneyCredit and Banking XXVIII (1996) 141ndash161
Kiefer Nicholas lsquolsquoEconomic Duration Data and Hazard Functionsrsquorsquo Journal ofEconomic Literature XXVI (1988) 646ndash679
Kroszner Randall S lsquolsquoThe Evolution of Universal Banking and Its Regulation inTwentieth Century Americarsquorsquo in Universal Banking Financial System DesignReconsidered Anthony Saunders and Ingo Walter eds (New York IrwinProfessional Publishers 1996) pp 70ndash99 lsquolsquoThe Political Economy of Banking and Financial Regulation in the U Srsquorsquo inThe Banking and Financial Structure in the NAFTA Countries and ChileGeorge M von Furstenberg ed (Boston Kluwer Academic Publishers 1997)pp 200ndash213
Kroszner Randall S and Raghuram G Rajan lsquolsquoIs the Glass-Steagall ActJustied A Study of the U S Experience with Universal Banking before1933rsquorsquo American Economic Review LXXXIV (1994) 810ndash832
Kroszner Randall S and Raghuram G Rajan lsquolsquoOrganization Structure andCredibility Evidence from Bank Securities Activities before the Glass-Steagall Actrsquorsquo Journal of Monetary Economics XXXIX (1997) 475ndash516
Kroszner Randall S and Philip E Strahan lsquolsquoRegulatory Incentives and the ThriftCrisis Dividends Mutual-to-Stock Conversions and Financial DistressrsquorsquoJournal of Finance LI (1996) 1285ndash1320
Kroszner Randall S and Philip E Strahan lsquolsquoWhat Drives Deregulation Econom-ics and Politics of the Relaxation of Bank Branching Restrictionsrsquorsquo NBERWorking Paper No 6637 1998
Kroszner Randall S and Thomas Stratmann lsquolsquoInterest Group Competition andthe Organization of Congress Theory and Evidence from Financial ServicesPolitical Action Committeesrsquorsquo American Economic Review LXXXVIII (1998)pp 1163ndash1187
Kroszner Randall S and Thomas Stratmann lsquolsquoDoes Political Ambiguity PayCorporate Campaign Contributions and Rewards to Legislator Reputationrsquorsquounpublishedmanuscript University of Chicago Graduate School of Business1999
Lin D Y and L J Wei lsquolsquoThe Robust Inference for the Cox Proportional HazardsModel Journal of the American Statistical Association LXXXIV (1989)1074ndash1078
Macey Jonathan R and Geoffrey P Miller Banking Law and Regulation (BostonLittle Brown and Company 1992)
Makinson Larry Open Secrets The Encyclopedia of Congressional Money andPolitics (Washington DC Congressional Quarterly 1992)
Niskanen WilliamBureaucracy and Representative Government (Chicago Aldine-Atherton 1971)
QUARTERLY JOURNAL OF ECONOMICS1466
Noll Roger lsquolsquoComment on Peltzmanrsquorsquo Brookings Papers Microeconomics SpecialIssue (1989) 48ndash58
North Douglass lsquolsquoA Transactions Cost Theory of Politicsrsquorsquo Journal of TheoreticalPolitics II (1990) 355ndash367
Peltzman Sam lsquolsquoToward a More General Theory of Regulationrsquorsquo Journal of Lawand Economics XIX (1976) 109ndash148 lsquolsquoConstituent Interest and Congressional Votingrsquorsquo Journal of Law andEconomics XXVII (1984) 181ndash210 lsquolsquoThe Economic Theory of Regulation after a Decade of DeregulationrsquorsquoBrookings Papers Microeconomics Special Issue (1989) 1ndash41
Peterson Mitchell and Raghuram Rajan lsquolsquoThe Benets of Lending RelationshipsEvidence from Small Business Datarsquorsquo Journal of Finance XLIX (1994) 3ndash38
Poole Keith and Howard Rosenthal Congress A Political-Economic History ofRoll Call Voting (Oxford Oxford University Press 1997)
Stata Stata Reference Manual Release 5 (College Station TX Stata Press 1997)Stigler George J lsquolsquoThe Theory of Economic Regulationrsquorsquo Bell Journal of Economics
and Management Science II (1971) 3ndash21Stigler George J Chicago Studies in the Political Economy (Chicago University
of Chicago Press 1988)Strahan Philip E and James Weston lsquolsquoSmall Business Lending and the Changing
Structure of the Banking Industryrsquorsquo Journal of Banking and Finance XXII(1998) 821ndash845
Sylla Richard John Legler and John Wallis lsquolsquoBanks and State Public Finance inthe New Republic The United States 1790ndash1860rsquorsquo Journal of EconomicHistory XLVII (1987) 391ndash403
White Eugene The Regulation and Reform of the American Banking System1900ndash1929 (Princeton Princeton University Press 1983)
Winston Clifford lsquolsquoEconomic Deregulation Days of Reckoning for Microecono-mistsrsquorsquo Journal of Economic Literature XXXI (1993) 1263ndash1289
WHAT DRIVES DEREGULATION 1467
branching of approximately 17 percent The relative capital-to-assets ratio of small and large banks however does not have astatistically signicant effect in the probit model
The impact of rival interests outside of banking is alsoconsistent with intrastate deregulation results Where banks cansell insurance legislators from states with larger insurancesectors relative to banking are less likely to vote for the amend-ment35 A one-standard-deviation increase in the relative size ofinsurance in those states which permit banks to sell insurancedecreases the probability that a legislator will favor the amend-ment by about 13 percent Turning to consumer interests thecoefficient on small rm share is not statistically signicant butthe coefficient on interest rates in the state is A one-standard-deviation increase in the average interest rate on loans raises theprobability that a legislator will support the amendment byroughly 8 percent The coefficient estimate for the bankingdistress variable measured as the average bank failure rate from1985 to 1991 is small and statistically signicant Overall theprobit analysis of the vote on national branching deregulationsupports the private-interest theory of deregulation and providesa consistency check that the importance of interests operating onthe state legislatures are very similar to those operating at thefederal level
V WHY DID BRANCHING DEREGULATION BEGIN DURING THE
1970S SHOCKS TO THE EQUILIBRIUM SUPPORTING BANK
BRANCHING RESTRICTIONS
A complete explanation of regulatory exit should account forwhy deregulation begins in the 1970s as well as the timing of thestate-by-state reforms [Kane 1996] In this section we describebroad technological legal and economic shocks that altered thepolitical-economy equilibrium which had kept antibranching regu-lations little changed for at least 30 years
Beginning in the 1970s three innovations reduced the valueto the protected banks of local geographic monopolies (see TableV) First automatic teller machines (ATMs) helped to erode thegeographic ties between customers and banks Second checkablemoney market mutual funds and the Merrill Lynch Cash Manage-ment Account demonstrated that banking by mail and telephone
35 The positive and statistically signicant estimates on the other insurancevariables may reect insurance industry support for the amendmentrsquos provisionslimiting banksrsquo insurance powers
QUARTERLY JOURNAL OF ECONOMICS1460
provided a convenient alternative to local banks36 Third techno-logical innovation and deregulation reduced transportation andcommunication costs particularly since the 1970s thereby lower-ing the costs for customers to use distant banks By increasing theelasticity of deposits supplied to banks these innovations reducedthe value of geographical restrictions to their traditional benecia-ries and thereby reduced their incentive to ght to maintain them[Peltzman 1976]
On the lending side increasing sophistication of credit-scoring techniques following innovations in information process-ing nancial theory and the development of large credit databases began to diminish the value of knowledge that localbankers had about the risks of borrowers in the community As a
36 Regulation Q which limited the interest rates that banks could pay ondeposits may have helped to drive depositors away from banks when the gapbetween market rates and deposit ceilings grew during the 1970s The eliminationof interest rate ceilings on large denomination certicates of deposit during the1970s appears to have hurt smaller and retail-oriented banks relative to largerwholesale banks [James 1983]
TABLE VBROAD TRENDS IN COMMERCIAL BANKING 1950ndash1995
YearNumberof ATMs
Domesticbank
deposits(billions)
Moneymarketmutual
fund(billions)
Percent ofdeposits 1
moneyfunds heldby banks
Smallbanksrsquo
percent ofbankingassets
Averagenumber of
bankfailures
(1) (2) (3) (4) (5) (6)1950 0 $ 154 $ 0 100 NA 41955 0 191 0 100 NA 31960 0 228 0 100 24 21965 0 330 0 100 20 41970 0 479 0 100 18 61975 9750 775 4 99 18 61980 18500 1182 76 94 17 101985 61117 1787 242 88 14 601990 80156 2339 493 83 11 1791995 122706 2552 745 77 8 61
Column (1) ATM gures are from Bank Network News The EFT Network Data Book (New YorkFaulkner and Gray Inc) The 1975 gure was unavailable 9750 is the number of ATMs in 1978 the rst yearfor which complete data are available Columns (2)ndash(4) Banksrsquo domestic deposits are from the Reports ofIncome and Condition money market mutual funds are from the Flow of Funds Data on all bank depositsforeign plus domestic are only available beginning in 1970 The trend in banksrsquo share (column (4)) is the sameusing total deposits instead of domestic deposits Column (5) Percent of banking assets held by small bankswhere a small bank is dened as a commercial bank with less than $100 million in assets in 1994 dollarsThese data are based on the Reports of Income and Condition Data on small banks are not available before1960 Column (6) Five-year average number of bank failures where the nal year is indicated in the rstcolumn These data are from FDIC Annual Report and the Quarterly Banking Prole
WHAT DRIVES DEREGULATION 1461
result of these innovations a national market developed forresidential mortgages credit card receivables have been securi-tized and bank lending to small business now relies less on thejudgment of loan officers and more on standardized scoringmodels
These changes have increased the potential protability forlarge banks to enter what had been the core of small bankactivities Large banksrsquo incentive to increase their lobbying pres-sure to be able to expand into these markets has thus beenincreasing over time In fact small banksrsquo market share began todecline even prior to the branching deregulation (Table V) As thevalue of a local banking relationships declined small rms thatwere the main borrowers from the small banks also probablybecame more likely to favor the entry of large banks into localmarkets With the deadweight costs of preventing large bankentry rising the private-interest theory predicts that small localbanks would become less likely able to maintain the branchingrestrictions [Becker 1983] Deregulation that reduces deadweightcosts of regulation is also consistent with the public-interesttheory
Kane [1996] argues that another major shock to the oldequilibrium was the increasing public awareness of the costlinessof having government-insured but (geographically) undiversiednancial institutions In the late 1970s the failure rate of banksbegan to rise (Table V) In the 1980s the Savings and Loan crisisand taxpayer bailout further heightened public awareness aboutthe costs of restrictions that make depository institutions morelikely to require infusions of taxpayer funds The failures thusmay have heightened public support for branching deregulation(see also Abrams and Settle [1993]) As noted above we did notnd any evidence that banking failures or distress in a stateaffected the speed with which the state deregulated but distressstill may be a common factor affecting all states
These technological economic and legal shocks generatedconditions that changed the long-standing balance favoring theantibranching forces The marginal value of lobbying to repealbranching restrictions increased just as the relative value to thesmall banks of maintaining branching restrictions was decliningThese nationwide shocks are common factors across states andthey are consistent with the positive duration dependence associ-ated with state-level deregulation that we demonstrated in FigureII In addition a broad change in lsquolsquoideologyrsquorsquo against government
QUARTERLY JOURNAL OF ECONOMICS1462
regulation could explain the positive duration dependence duringthe sample period
VI CONCLUSIONS
The private-interest theory of regulation can account for thepattern of bank branching deregulation of the 1970s and 1980sBeneciaries of branching regulation had supported a coalitionfavoring geographical restrictions despite their costs to consum-ers of nancial services Innovations that began in the 1970saltered the value of the restrictions to the affected parties and theresulting competition among interest groups can explain thesubsequent deregulation While some of our results are alsoconsistent with the public interest theorymdashfor example deregula-tion occurs earlier when small banks are in a relatively weaknancial positionmdashother results particularly evidence on theimportance of rivalries between small and large banks andbetween banking and insurance are difficult to explain with thepublic-interest approach We also nd that political-institutionalfactors affect deregulation although these variables may act asproxies for unmeasured economic interests The greater success ofthe private-interest theory to explain deregulation here than inprevious work may be due to the richer cross-sectional andtime-series variation in branching deregulation that can beexploited with a hazard model Future empirical studies ofendogenous deregulation thus may be most fruitful where changehas occurred across states over time such as in franchisinginsurance and public utilities (eg Joskow [1996])
Technological and nancial innovations will continue to erodethe benets to any interest group of maintaining regulatorybarriers in nancial services These forces are likely to bringabout reforms both domestically for example through legislationthat would increase bank powers (see Kroszner [1996] Krosznerand Rajan [1994 1997] and Kroszner and Stratmann [1998]) andinternationally for example through the extension of nancialservices provisions of NAFTAto reduce geographic barriers acrosscountries [Kroszner 1997]
APPENDIX EFFECTS OF DIFFERENT TYPES OF BRANCHING
DEREGULATION ON ECONOMIC GROWTH BANKING STRUCTUREAND BANK EFFICIENCY
We measure the importance of alternative forms of deregula-tion by their consequences for the economic growth and the
WHAT DRIVES DEREGULATION 1463
structure and efficiency of the banking sector across the states(see Jayaratne and Strahan [1996 1998]) To do so we run anOLS panel regression relating indicators for each type of deregula-tion to potential consequences Our dependent variables includeproxies for economic growth (growth in personal income) bankstructure (the log of the number of banking companies and thenumber of branches per banking company) and efficiency (the logof total noninterest expenses and the ratio of loan loss provisionsto total loans) The four deregulation indicators (MBHC branch-ing through merger and acquisition unrestricted statewide branch-ing and interstate banking) are equal to one in the years afterderegulation and zero before If a state deregulated before thesample period begins in 1970 the indicator is one throughout ifthe state does not deregulate then the indicator is zero through-out All regressions include state xed effects and time effects
Table VI shows the results of including the set of four
TABLE VIMULTIVARIATE REGRESSIONS OF THE EFFECTS OF DIFFERENT TYPES OF BRANCHING
DEREGULATION ON ECONOMIC GROWTH BANKING STRUCTUREAND BANK EFFICIENCY
Dependent variable
Growthin
personalincome
Log of of
bankingcompanies
Branchesper
bankingcompany
Log ofnon-
interestexpenses
Loan lossprovisions
(loans 1leases)
Merger amp acquisitionbranchingindicator
0013(0004)
2 0162(0037)
1146(0453)
2 0107(0035)
2 0007(0001)
Unrestricted state-wide branchingindicator
2 0008(0005)
0086(0059)
2 0043(0574)
0070(0037)
0005(0002)
Interstate bankingindicator
0006(0005)
2 0052(0057)
1064(0511)
0005(0030)
2 0001(0001)
Multibank holdingcompany indicator
2 0005(0006)
2 0096(0052)
2 0364(0529)
2 0095(0036)
2 0001(0001)
Dependent variablemean [median]
0086[0082]
4881[5136]
6649[5055]
13812[13803]
0009[0006]
This table contains regressions of state-level economic growth and state-level measures of the structureand efficiency of the banking industry Each dependent variable is regressed on a set of state indicatorvariables a set of year indicator variables and the set of four deregulation indicators these deregulationindicators are equal to one after deregulation and zero before If a state was always regulated the indicator iszero throughout if it was always deregulated the indicator is one throughout Sample period is 1975 to 1991and Delaware and South Dakota are dropped so N 5 833 Noninterest expenses become available only in1984 so N 5 392 The coefficient on the deregulation indicators is reported with robust standard errors inparentheses and denote statistically signicant at the 1 5 and 10 percent levels
QUARTERLY JOURNAL OF ECONOMICS1464
deregulation indicators simultaneously The branching by mergerand acquisition indicator is statistically signicant and of theexpected sign in all ve of the equations and larger in absolutevalue than any of the other coefficients on the indicators for theother types of deregulation Since there may be multicollinearityamong our indicators we also ran univariate specications whichincluded each indicator separately and the results are verysimilar These results imply that branching by merger andacquisition is the most important type of deregulation measuredby its consequences for economic growth banking structure andbank efficiency See Kroszner and Strahan [1998] for more details
UNIVERSITY OF CHICAGO GRADUATE SCHOOL OF BUSINESS AND NATIONAL BUREAU OF
ECONOMIC RESEARCH
FEDERAL RESERVE BANK OF NEW YORK RESEARCH amp MARKET ANALYSIS GROUP
REFERENCES
Abrams Burton A and Russell F Settle lsquolsquoPressure-Group Inuence and Institu-tional Change Branch-Banking Legislation during the Great DepressionrsquorsquoPublic Choice LXXVII (1993) 687ndash705
Amel Dean lsquolsquoState Laws Affecting the Geographic Expansion of CommercialBanksrsquorsquo unpublished manuscript Board of Governors of the Federal ReserveSystem 1993
Becker Gary S lsquolsquoA Theory of Competition among Pressure Groups for PoliticalInuencersquorsquo Quarterly Journal of Economics XCVIII (1983) 371ndash400
Berger Allen Anil Kashyap Joseph Scalise lsquolsquoThe Transformation of the U SBanking Industry What a Long Strange Trip Itrsquos Beenrsquorsquo Brookings Papers onEconomic Activity 2 (1995) 55ndash218
Berger Allen Anthony Saunders Joseph Scalise and Gregory Udell lsquolsquoThe Effectsof Bank Mergers and Acquisitions on Small Business Lendingrsquorsquo Journal ofFinancial Economics L (1998) 187ndash229
Brennan Geoffrey and James Buchanan lsquolsquoTowards a Tax Constitution forLeviathanrsquorsquo Journal of Public Economics VIII (1977) 255ndash273
Brickley James A and Christopher M James lsquolsquoThe Takeover Market CorporateBoard Composition and Ownership Structure The Case of Bankingrsquorsquo Journalof Law and Economics XXX (1987) 161ndash180
Cole Rebel and John Wolken lsquolsquoFinancial Services Used by Small BusinessesEvidence from the 1993 Survey of Small Business Financesrsquorsquo Federal ReserveBulletin LXXXI (1994) 629ndash676
Dixit Avinash The Making of Economic Policy A Transactions-Cost PoliticsPerspective (Cambridge MA MIT Press 1996)
Economides Nicolas R Glenn Hubbard and Darius Palia lsquolsquoThe Political Economyof Branching Restrictions and Deposit Insurancersquorsquo Journal of Law andEconomics XXIX (1996) 667ndash704
Flannery Mark lsquolsquoThe Social Costs of Unit Banking Restrictionsrsquorsquo Journal ofMonetary Economics XIII (1984) 237ndash249
Greene William H Econometric Analysis third edition (Upper Saddle River NJPrentice-Hall 1997)
Gunther Jeffery lsquolsquoRegional Capital Imbalances and the Removal of InterstateBanking Restrictionsrsquorsquo Economics Letters XLIV (1994) 439ndash442 lsquolsquoThe Erosion of Intrastate Branch Banking Restrictionsrsquorsquo unpublishedmanuscript Federal Reserve Bank of Dallas 1996
Irwin Douglas and Randall Kroszner lsquolsquoInterests Institutions and Ideology inSecuring Policy Change The Republican Conversion to Trade Liberalizationafter Smoot-Hawleyrsquorsquo Journal of Law and Economics XLII (1999) 643ndash673
WHAT DRIVES DEREGULATION 1465
James Christopher lsquolsquoAn Analysis of Intra-Industry Differences in the Effect ofRegulation The Case of Deposit Rate Ceilingsrsquorsquo Journal of Monetary Econom-ics XII (1983) 417ndash442
Jarrell Gregg lsquolsquoChange at the Exchange The Causes and Effects of DeregulationrsquorsquoJournal of Law and Economics XXVII (1984) 273ndash312
Jayaratne Jith and Philip E Strahan lsquolsquoThe Finance-Growth Nexus Evidencefrom Bank Branch Deregulationrsquorsquo Quarterly Journal of Economics CI (1996)639ndash670
Jayaratne Jith and Philip E Strahan lsquolsquoEntry Restrictions Industry Evolutionand Dynamic Efficiency Evidence from Commercial Bankingrsquorsquo Journal of Lawand Economics XLIX (1998) 239ndash274
Joskow Paul lsquolsquoComment on lsquoPower Struggles Explaining Deregulatory Reformsin Electricity Marketsrsquo rsquorsquo Brookings Papers on Economic Activity Microeconom-ics (1996) 251ndash264
Joskow Paul and Roger Noll lsquolsquoRegulation in Theory and Practice An Overviewrsquorsquoin Studies in Public Regulation Gary Fromm ed (Cambridge MA MITPress 1981) pp 1ndash65
Kahn Matthew E and John G Matsusaka lsquolsquoDemand for Environmental GoodsEvidence from Voting Patterns on California Initiativesrsquorsquo Journal of Law andEconomics XL (1997) 137ndash173
Kalbeisch J and R Prentice The Statistical Analysis of Failure Time Data (NewYork John Wiley amp Sons 1980)
Kane Edward J lsquolsquoDe Jure Interstate Banking Why Only NowrsquorsquoJournal of MoneyCredit and Banking XXVIII (1996) 141ndash161
Kiefer Nicholas lsquolsquoEconomic Duration Data and Hazard Functionsrsquorsquo Journal ofEconomic Literature XXVI (1988) 646ndash679
Kroszner Randall S lsquolsquoThe Evolution of Universal Banking and Its Regulation inTwentieth Century Americarsquorsquo in Universal Banking Financial System DesignReconsidered Anthony Saunders and Ingo Walter eds (New York IrwinProfessional Publishers 1996) pp 70ndash99 lsquolsquoThe Political Economy of Banking and Financial Regulation in the U Srsquorsquo inThe Banking and Financial Structure in the NAFTA Countries and ChileGeorge M von Furstenberg ed (Boston Kluwer Academic Publishers 1997)pp 200ndash213
Kroszner Randall S and Raghuram G Rajan lsquolsquoIs the Glass-Steagall ActJustied A Study of the U S Experience with Universal Banking before1933rsquorsquo American Economic Review LXXXIV (1994) 810ndash832
Kroszner Randall S and Raghuram G Rajan lsquolsquoOrganization Structure andCredibility Evidence from Bank Securities Activities before the Glass-Steagall Actrsquorsquo Journal of Monetary Economics XXXIX (1997) 475ndash516
Kroszner Randall S and Philip E Strahan lsquolsquoRegulatory Incentives and the ThriftCrisis Dividends Mutual-to-Stock Conversions and Financial DistressrsquorsquoJournal of Finance LI (1996) 1285ndash1320
Kroszner Randall S and Philip E Strahan lsquolsquoWhat Drives Deregulation Econom-ics and Politics of the Relaxation of Bank Branching Restrictionsrsquorsquo NBERWorking Paper No 6637 1998
Kroszner Randall S and Thomas Stratmann lsquolsquoInterest Group Competition andthe Organization of Congress Theory and Evidence from Financial ServicesPolitical Action Committeesrsquorsquo American Economic Review LXXXVIII (1998)pp 1163ndash1187
Kroszner Randall S and Thomas Stratmann lsquolsquoDoes Political Ambiguity PayCorporate Campaign Contributions and Rewards to Legislator Reputationrsquorsquounpublishedmanuscript University of Chicago Graduate School of Business1999
Lin D Y and L J Wei lsquolsquoThe Robust Inference for the Cox Proportional HazardsModel Journal of the American Statistical Association LXXXIV (1989)1074ndash1078
Macey Jonathan R and Geoffrey P Miller Banking Law and Regulation (BostonLittle Brown and Company 1992)
Makinson Larry Open Secrets The Encyclopedia of Congressional Money andPolitics (Washington DC Congressional Quarterly 1992)
Niskanen WilliamBureaucracy and Representative Government (Chicago Aldine-Atherton 1971)
QUARTERLY JOURNAL OF ECONOMICS1466
Noll Roger lsquolsquoComment on Peltzmanrsquorsquo Brookings Papers Microeconomics SpecialIssue (1989) 48ndash58
North Douglass lsquolsquoA Transactions Cost Theory of Politicsrsquorsquo Journal of TheoreticalPolitics II (1990) 355ndash367
Peltzman Sam lsquolsquoToward a More General Theory of Regulationrsquorsquo Journal of Lawand Economics XIX (1976) 109ndash148 lsquolsquoConstituent Interest and Congressional Votingrsquorsquo Journal of Law andEconomics XXVII (1984) 181ndash210 lsquolsquoThe Economic Theory of Regulation after a Decade of DeregulationrsquorsquoBrookings Papers Microeconomics Special Issue (1989) 1ndash41
Peterson Mitchell and Raghuram Rajan lsquolsquoThe Benets of Lending RelationshipsEvidence from Small Business Datarsquorsquo Journal of Finance XLIX (1994) 3ndash38
Poole Keith and Howard Rosenthal Congress A Political-Economic History ofRoll Call Voting (Oxford Oxford University Press 1997)
Stata Stata Reference Manual Release 5 (College Station TX Stata Press 1997)Stigler George J lsquolsquoThe Theory of Economic Regulationrsquorsquo Bell Journal of Economics
and Management Science II (1971) 3ndash21Stigler George J Chicago Studies in the Political Economy (Chicago University
of Chicago Press 1988)Strahan Philip E and James Weston lsquolsquoSmall Business Lending and the Changing
Structure of the Banking Industryrsquorsquo Journal of Banking and Finance XXII(1998) 821ndash845
Sylla Richard John Legler and John Wallis lsquolsquoBanks and State Public Finance inthe New Republic The United States 1790ndash1860rsquorsquo Journal of EconomicHistory XLVII (1987) 391ndash403
White Eugene The Regulation and Reform of the American Banking System1900ndash1929 (Princeton Princeton University Press 1983)
Winston Clifford lsquolsquoEconomic Deregulation Days of Reckoning for Microecono-mistsrsquorsquo Journal of Economic Literature XXXI (1993) 1263ndash1289
WHAT DRIVES DEREGULATION 1467
provided a convenient alternative to local banks36 Third techno-logical innovation and deregulation reduced transportation andcommunication costs particularly since the 1970s thereby lower-ing the costs for customers to use distant banks By increasing theelasticity of deposits supplied to banks these innovations reducedthe value of geographical restrictions to their traditional benecia-ries and thereby reduced their incentive to ght to maintain them[Peltzman 1976]
On the lending side increasing sophistication of credit-scoring techniques following innovations in information process-ing nancial theory and the development of large credit databases began to diminish the value of knowledge that localbankers had about the risks of borrowers in the community As a
36 Regulation Q which limited the interest rates that banks could pay ondeposits may have helped to drive depositors away from banks when the gapbetween market rates and deposit ceilings grew during the 1970s The eliminationof interest rate ceilings on large denomination certicates of deposit during the1970s appears to have hurt smaller and retail-oriented banks relative to largerwholesale banks [James 1983]
TABLE VBROAD TRENDS IN COMMERCIAL BANKING 1950ndash1995
YearNumberof ATMs
Domesticbank
deposits(billions)
Moneymarketmutual
fund(billions)
Percent ofdeposits 1
moneyfunds heldby banks
Smallbanksrsquo
percent ofbankingassets
Averagenumber of
bankfailures
(1) (2) (3) (4) (5) (6)1950 0 $ 154 $ 0 100 NA 41955 0 191 0 100 NA 31960 0 228 0 100 24 21965 0 330 0 100 20 41970 0 479 0 100 18 61975 9750 775 4 99 18 61980 18500 1182 76 94 17 101985 61117 1787 242 88 14 601990 80156 2339 493 83 11 1791995 122706 2552 745 77 8 61
Column (1) ATM gures are from Bank Network News The EFT Network Data Book (New YorkFaulkner and Gray Inc) The 1975 gure was unavailable 9750 is the number of ATMs in 1978 the rst yearfor which complete data are available Columns (2)ndash(4) Banksrsquo domestic deposits are from the Reports ofIncome and Condition money market mutual funds are from the Flow of Funds Data on all bank depositsforeign plus domestic are only available beginning in 1970 The trend in banksrsquo share (column (4)) is the sameusing total deposits instead of domestic deposits Column (5) Percent of banking assets held by small bankswhere a small bank is dened as a commercial bank with less than $100 million in assets in 1994 dollarsThese data are based on the Reports of Income and Condition Data on small banks are not available before1960 Column (6) Five-year average number of bank failures where the nal year is indicated in the rstcolumn These data are from FDIC Annual Report and the Quarterly Banking Prole
WHAT DRIVES DEREGULATION 1461
result of these innovations a national market developed forresidential mortgages credit card receivables have been securi-tized and bank lending to small business now relies less on thejudgment of loan officers and more on standardized scoringmodels
These changes have increased the potential protability forlarge banks to enter what had been the core of small bankactivities Large banksrsquo incentive to increase their lobbying pres-sure to be able to expand into these markets has thus beenincreasing over time In fact small banksrsquo market share began todecline even prior to the branching deregulation (Table V) As thevalue of a local banking relationships declined small rms thatwere the main borrowers from the small banks also probablybecame more likely to favor the entry of large banks into localmarkets With the deadweight costs of preventing large bankentry rising the private-interest theory predicts that small localbanks would become less likely able to maintain the branchingrestrictions [Becker 1983] Deregulation that reduces deadweightcosts of regulation is also consistent with the public-interesttheory
Kane [1996] argues that another major shock to the oldequilibrium was the increasing public awareness of the costlinessof having government-insured but (geographically) undiversiednancial institutions In the late 1970s the failure rate of banksbegan to rise (Table V) In the 1980s the Savings and Loan crisisand taxpayer bailout further heightened public awareness aboutthe costs of restrictions that make depository institutions morelikely to require infusions of taxpayer funds The failures thusmay have heightened public support for branching deregulation(see also Abrams and Settle [1993]) As noted above we did notnd any evidence that banking failures or distress in a stateaffected the speed with which the state deregulated but distressstill may be a common factor affecting all states
These technological economic and legal shocks generatedconditions that changed the long-standing balance favoring theantibranching forces The marginal value of lobbying to repealbranching restrictions increased just as the relative value to thesmall banks of maintaining branching restrictions was decliningThese nationwide shocks are common factors across states andthey are consistent with the positive duration dependence associ-ated with state-level deregulation that we demonstrated in FigureII In addition a broad change in lsquolsquoideologyrsquorsquo against government
QUARTERLY JOURNAL OF ECONOMICS1462
regulation could explain the positive duration dependence duringthe sample period
VI CONCLUSIONS
The private-interest theory of regulation can account for thepattern of bank branching deregulation of the 1970s and 1980sBeneciaries of branching regulation had supported a coalitionfavoring geographical restrictions despite their costs to consum-ers of nancial services Innovations that began in the 1970saltered the value of the restrictions to the affected parties and theresulting competition among interest groups can explain thesubsequent deregulation While some of our results are alsoconsistent with the public interest theorymdashfor example deregula-tion occurs earlier when small banks are in a relatively weaknancial positionmdashother results particularly evidence on theimportance of rivalries between small and large banks andbetween banking and insurance are difficult to explain with thepublic-interest approach We also nd that political-institutionalfactors affect deregulation although these variables may act asproxies for unmeasured economic interests The greater success ofthe private-interest theory to explain deregulation here than inprevious work may be due to the richer cross-sectional andtime-series variation in branching deregulation that can beexploited with a hazard model Future empirical studies ofendogenous deregulation thus may be most fruitful where changehas occurred across states over time such as in franchisinginsurance and public utilities (eg Joskow [1996])
Technological and nancial innovations will continue to erodethe benets to any interest group of maintaining regulatorybarriers in nancial services These forces are likely to bringabout reforms both domestically for example through legislationthat would increase bank powers (see Kroszner [1996] Krosznerand Rajan [1994 1997] and Kroszner and Stratmann [1998]) andinternationally for example through the extension of nancialservices provisions of NAFTAto reduce geographic barriers acrosscountries [Kroszner 1997]
APPENDIX EFFECTS OF DIFFERENT TYPES OF BRANCHING
DEREGULATION ON ECONOMIC GROWTH BANKING STRUCTUREAND BANK EFFICIENCY
We measure the importance of alternative forms of deregula-tion by their consequences for the economic growth and the
WHAT DRIVES DEREGULATION 1463
structure and efficiency of the banking sector across the states(see Jayaratne and Strahan [1996 1998]) To do so we run anOLS panel regression relating indicators for each type of deregula-tion to potential consequences Our dependent variables includeproxies for economic growth (growth in personal income) bankstructure (the log of the number of banking companies and thenumber of branches per banking company) and efficiency (the logof total noninterest expenses and the ratio of loan loss provisionsto total loans) The four deregulation indicators (MBHC branch-ing through merger and acquisition unrestricted statewide branch-ing and interstate banking) are equal to one in the years afterderegulation and zero before If a state deregulated before thesample period begins in 1970 the indicator is one throughout ifthe state does not deregulate then the indicator is zero through-out All regressions include state xed effects and time effects
Table VI shows the results of including the set of four
TABLE VIMULTIVARIATE REGRESSIONS OF THE EFFECTS OF DIFFERENT TYPES OF BRANCHING
DEREGULATION ON ECONOMIC GROWTH BANKING STRUCTUREAND BANK EFFICIENCY
Dependent variable
Growthin
personalincome
Log of of
bankingcompanies
Branchesper
bankingcompany
Log ofnon-
interestexpenses
Loan lossprovisions
(loans 1leases)
Merger amp acquisitionbranchingindicator
0013(0004)
2 0162(0037)
1146(0453)
2 0107(0035)
2 0007(0001)
Unrestricted state-wide branchingindicator
2 0008(0005)
0086(0059)
2 0043(0574)
0070(0037)
0005(0002)
Interstate bankingindicator
0006(0005)
2 0052(0057)
1064(0511)
0005(0030)
2 0001(0001)
Multibank holdingcompany indicator
2 0005(0006)
2 0096(0052)
2 0364(0529)
2 0095(0036)
2 0001(0001)
Dependent variablemean [median]
0086[0082]
4881[5136]
6649[5055]
13812[13803]
0009[0006]
This table contains regressions of state-level economic growth and state-level measures of the structureand efficiency of the banking industry Each dependent variable is regressed on a set of state indicatorvariables a set of year indicator variables and the set of four deregulation indicators these deregulationindicators are equal to one after deregulation and zero before If a state was always regulated the indicator iszero throughout if it was always deregulated the indicator is one throughout Sample period is 1975 to 1991and Delaware and South Dakota are dropped so N 5 833 Noninterest expenses become available only in1984 so N 5 392 The coefficient on the deregulation indicators is reported with robust standard errors inparentheses and denote statistically signicant at the 1 5 and 10 percent levels
QUARTERLY JOURNAL OF ECONOMICS1464
deregulation indicators simultaneously The branching by mergerand acquisition indicator is statistically signicant and of theexpected sign in all ve of the equations and larger in absolutevalue than any of the other coefficients on the indicators for theother types of deregulation Since there may be multicollinearityamong our indicators we also ran univariate specications whichincluded each indicator separately and the results are verysimilar These results imply that branching by merger andacquisition is the most important type of deregulation measuredby its consequences for economic growth banking structure andbank efficiency See Kroszner and Strahan [1998] for more details
UNIVERSITY OF CHICAGO GRADUATE SCHOOL OF BUSINESS AND NATIONAL BUREAU OF
ECONOMIC RESEARCH
FEDERAL RESERVE BANK OF NEW YORK RESEARCH amp MARKET ANALYSIS GROUP
REFERENCES
Abrams Burton A and Russell F Settle lsquolsquoPressure-Group Inuence and Institu-tional Change Branch-Banking Legislation during the Great DepressionrsquorsquoPublic Choice LXXVII (1993) 687ndash705
Amel Dean lsquolsquoState Laws Affecting the Geographic Expansion of CommercialBanksrsquorsquo unpublished manuscript Board of Governors of the Federal ReserveSystem 1993
Becker Gary S lsquolsquoA Theory of Competition among Pressure Groups for PoliticalInuencersquorsquo Quarterly Journal of Economics XCVIII (1983) 371ndash400
Berger Allen Anil Kashyap Joseph Scalise lsquolsquoThe Transformation of the U SBanking Industry What a Long Strange Trip Itrsquos Beenrsquorsquo Brookings Papers onEconomic Activity 2 (1995) 55ndash218
Berger Allen Anthony Saunders Joseph Scalise and Gregory Udell lsquolsquoThe Effectsof Bank Mergers and Acquisitions on Small Business Lendingrsquorsquo Journal ofFinancial Economics L (1998) 187ndash229
Brennan Geoffrey and James Buchanan lsquolsquoTowards a Tax Constitution forLeviathanrsquorsquo Journal of Public Economics VIII (1977) 255ndash273
Brickley James A and Christopher M James lsquolsquoThe Takeover Market CorporateBoard Composition and Ownership Structure The Case of Bankingrsquorsquo Journalof Law and Economics XXX (1987) 161ndash180
Cole Rebel and John Wolken lsquolsquoFinancial Services Used by Small BusinessesEvidence from the 1993 Survey of Small Business Financesrsquorsquo Federal ReserveBulletin LXXXI (1994) 629ndash676
Dixit Avinash The Making of Economic Policy A Transactions-Cost PoliticsPerspective (Cambridge MA MIT Press 1996)
Economides Nicolas R Glenn Hubbard and Darius Palia lsquolsquoThe Political Economyof Branching Restrictions and Deposit Insurancersquorsquo Journal of Law andEconomics XXIX (1996) 667ndash704
Flannery Mark lsquolsquoThe Social Costs of Unit Banking Restrictionsrsquorsquo Journal ofMonetary Economics XIII (1984) 237ndash249
Greene William H Econometric Analysis third edition (Upper Saddle River NJPrentice-Hall 1997)
Gunther Jeffery lsquolsquoRegional Capital Imbalances and the Removal of InterstateBanking Restrictionsrsquorsquo Economics Letters XLIV (1994) 439ndash442 lsquolsquoThe Erosion of Intrastate Branch Banking Restrictionsrsquorsquo unpublishedmanuscript Federal Reserve Bank of Dallas 1996
Irwin Douglas and Randall Kroszner lsquolsquoInterests Institutions and Ideology inSecuring Policy Change The Republican Conversion to Trade Liberalizationafter Smoot-Hawleyrsquorsquo Journal of Law and Economics XLII (1999) 643ndash673
WHAT DRIVES DEREGULATION 1465
James Christopher lsquolsquoAn Analysis of Intra-Industry Differences in the Effect ofRegulation The Case of Deposit Rate Ceilingsrsquorsquo Journal of Monetary Econom-ics XII (1983) 417ndash442
Jarrell Gregg lsquolsquoChange at the Exchange The Causes and Effects of DeregulationrsquorsquoJournal of Law and Economics XXVII (1984) 273ndash312
Jayaratne Jith and Philip E Strahan lsquolsquoThe Finance-Growth Nexus Evidencefrom Bank Branch Deregulationrsquorsquo Quarterly Journal of Economics CI (1996)639ndash670
Jayaratne Jith and Philip E Strahan lsquolsquoEntry Restrictions Industry Evolutionand Dynamic Efficiency Evidence from Commercial Bankingrsquorsquo Journal of Lawand Economics XLIX (1998) 239ndash274
Joskow Paul lsquolsquoComment on lsquoPower Struggles Explaining Deregulatory Reformsin Electricity Marketsrsquo rsquorsquo Brookings Papers on Economic Activity Microeconom-ics (1996) 251ndash264
Joskow Paul and Roger Noll lsquolsquoRegulation in Theory and Practice An Overviewrsquorsquoin Studies in Public Regulation Gary Fromm ed (Cambridge MA MITPress 1981) pp 1ndash65
Kahn Matthew E and John G Matsusaka lsquolsquoDemand for Environmental GoodsEvidence from Voting Patterns on California Initiativesrsquorsquo Journal of Law andEconomics XL (1997) 137ndash173
Kalbeisch J and R Prentice The Statistical Analysis of Failure Time Data (NewYork John Wiley amp Sons 1980)
Kane Edward J lsquolsquoDe Jure Interstate Banking Why Only NowrsquorsquoJournal of MoneyCredit and Banking XXVIII (1996) 141ndash161
Kiefer Nicholas lsquolsquoEconomic Duration Data and Hazard Functionsrsquorsquo Journal ofEconomic Literature XXVI (1988) 646ndash679
Kroszner Randall S lsquolsquoThe Evolution of Universal Banking and Its Regulation inTwentieth Century Americarsquorsquo in Universal Banking Financial System DesignReconsidered Anthony Saunders and Ingo Walter eds (New York IrwinProfessional Publishers 1996) pp 70ndash99 lsquolsquoThe Political Economy of Banking and Financial Regulation in the U Srsquorsquo inThe Banking and Financial Structure in the NAFTA Countries and ChileGeorge M von Furstenberg ed (Boston Kluwer Academic Publishers 1997)pp 200ndash213
Kroszner Randall S and Raghuram G Rajan lsquolsquoIs the Glass-Steagall ActJustied A Study of the U S Experience with Universal Banking before1933rsquorsquo American Economic Review LXXXIV (1994) 810ndash832
Kroszner Randall S and Raghuram G Rajan lsquolsquoOrganization Structure andCredibility Evidence from Bank Securities Activities before the Glass-Steagall Actrsquorsquo Journal of Monetary Economics XXXIX (1997) 475ndash516
Kroszner Randall S and Philip E Strahan lsquolsquoRegulatory Incentives and the ThriftCrisis Dividends Mutual-to-Stock Conversions and Financial DistressrsquorsquoJournal of Finance LI (1996) 1285ndash1320
Kroszner Randall S and Philip E Strahan lsquolsquoWhat Drives Deregulation Econom-ics and Politics of the Relaxation of Bank Branching Restrictionsrsquorsquo NBERWorking Paper No 6637 1998
Kroszner Randall S and Thomas Stratmann lsquolsquoInterest Group Competition andthe Organization of Congress Theory and Evidence from Financial ServicesPolitical Action Committeesrsquorsquo American Economic Review LXXXVIII (1998)pp 1163ndash1187
Kroszner Randall S and Thomas Stratmann lsquolsquoDoes Political Ambiguity PayCorporate Campaign Contributions and Rewards to Legislator Reputationrsquorsquounpublishedmanuscript University of Chicago Graduate School of Business1999
Lin D Y and L J Wei lsquolsquoThe Robust Inference for the Cox Proportional HazardsModel Journal of the American Statistical Association LXXXIV (1989)1074ndash1078
Macey Jonathan R and Geoffrey P Miller Banking Law and Regulation (BostonLittle Brown and Company 1992)
Makinson Larry Open Secrets The Encyclopedia of Congressional Money andPolitics (Washington DC Congressional Quarterly 1992)
Niskanen WilliamBureaucracy and Representative Government (Chicago Aldine-Atherton 1971)
QUARTERLY JOURNAL OF ECONOMICS1466
Noll Roger lsquolsquoComment on Peltzmanrsquorsquo Brookings Papers Microeconomics SpecialIssue (1989) 48ndash58
North Douglass lsquolsquoA Transactions Cost Theory of Politicsrsquorsquo Journal of TheoreticalPolitics II (1990) 355ndash367
Peltzman Sam lsquolsquoToward a More General Theory of Regulationrsquorsquo Journal of Lawand Economics XIX (1976) 109ndash148 lsquolsquoConstituent Interest and Congressional Votingrsquorsquo Journal of Law andEconomics XXVII (1984) 181ndash210 lsquolsquoThe Economic Theory of Regulation after a Decade of DeregulationrsquorsquoBrookings Papers Microeconomics Special Issue (1989) 1ndash41
Peterson Mitchell and Raghuram Rajan lsquolsquoThe Benets of Lending RelationshipsEvidence from Small Business Datarsquorsquo Journal of Finance XLIX (1994) 3ndash38
Poole Keith and Howard Rosenthal Congress A Political-Economic History ofRoll Call Voting (Oxford Oxford University Press 1997)
Stata Stata Reference Manual Release 5 (College Station TX Stata Press 1997)Stigler George J lsquolsquoThe Theory of Economic Regulationrsquorsquo Bell Journal of Economics
and Management Science II (1971) 3ndash21Stigler George J Chicago Studies in the Political Economy (Chicago University
of Chicago Press 1988)Strahan Philip E and James Weston lsquolsquoSmall Business Lending and the Changing
Structure of the Banking Industryrsquorsquo Journal of Banking and Finance XXII(1998) 821ndash845
Sylla Richard John Legler and John Wallis lsquolsquoBanks and State Public Finance inthe New Republic The United States 1790ndash1860rsquorsquo Journal of EconomicHistory XLVII (1987) 391ndash403
White Eugene The Regulation and Reform of the American Banking System1900ndash1929 (Princeton Princeton University Press 1983)
Winston Clifford lsquolsquoEconomic Deregulation Days of Reckoning for Microecono-mistsrsquorsquo Journal of Economic Literature XXXI (1993) 1263ndash1289
WHAT DRIVES DEREGULATION 1467
result of these innovations a national market developed forresidential mortgages credit card receivables have been securi-tized and bank lending to small business now relies less on thejudgment of loan officers and more on standardized scoringmodels
These changes have increased the potential protability forlarge banks to enter what had been the core of small bankactivities Large banksrsquo incentive to increase their lobbying pres-sure to be able to expand into these markets has thus beenincreasing over time In fact small banksrsquo market share began todecline even prior to the branching deregulation (Table V) As thevalue of a local banking relationships declined small rms thatwere the main borrowers from the small banks also probablybecame more likely to favor the entry of large banks into localmarkets With the deadweight costs of preventing large bankentry rising the private-interest theory predicts that small localbanks would become less likely able to maintain the branchingrestrictions [Becker 1983] Deregulation that reduces deadweightcosts of regulation is also consistent with the public-interesttheory
Kane [1996] argues that another major shock to the oldequilibrium was the increasing public awareness of the costlinessof having government-insured but (geographically) undiversiednancial institutions In the late 1970s the failure rate of banksbegan to rise (Table V) In the 1980s the Savings and Loan crisisand taxpayer bailout further heightened public awareness aboutthe costs of restrictions that make depository institutions morelikely to require infusions of taxpayer funds The failures thusmay have heightened public support for branching deregulation(see also Abrams and Settle [1993]) As noted above we did notnd any evidence that banking failures or distress in a stateaffected the speed with which the state deregulated but distressstill may be a common factor affecting all states
These technological economic and legal shocks generatedconditions that changed the long-standing balance favoring theantibranching forces The marginal value of lobbying to repealbranching restrictions increased just as the relative value to thesmall banks of maintaining branching restrictions was decliningThese nationwide shocks are common factors across states andthey are consistent with the positive duration dependence associ-ated with state-level deregulation that we demonstrated in FigureII In addition a broad change in lsquolsquoideologyrsquorsquo against government
QUARTERLY JOURNAL OF ECONOMICS1462
regulation could explain the positive duration dependence duringthe sample period
VI CONCLUSIONS
The private-interest theory of regulation can account for thepattern of bank branching deregulation of the 1970s and 1980sBeneciaries of branching regulation had supported a coalitionfavoring geographical restrictions despite their costs to consum-ers of nancial services Innovations that began in the 1970saltered the value of the restrictions to the affected parties and theresulting competition among interest groups can explain thesubsequent deregulation While some of our results are alsoconsistent with the public interest theorymdashfor example deregula-tion occurs earlier when small banks are in a relatively weaknancial positionmdashother results particularly evidence on theimportance of rivalries between small and large banks andbetween banking and insurance are difficult to explain with thepublic-interest approach We also nd that political-institutionalfactors affect deregulation although these variables may act asproxies for unmeasured economic interests The greater success ofthe private-interest theory to explain deregulation here than inprevious work may be due to the richer cross-sectional andtime-series variation in branching deregulation that can beexploited with a hazard model Future empirical studies ofendogenous deregulation thus may be most fruitful where changehas occurred across states over time such as in franchisinginsurance and public utilities (eg Joskow [1996])
Technological and nancial innovations will continue to erodethe benets to any interest group of maintaining regulatorybarriers in nancial services These forces are likely to bringabout reforms both domestically for example through legislationthat would increase bank powers (see Kroszner [1996] Krosznerand Rajan [1994 1997] and Kroszner and Stratmann [1998]) andinternationally for example through the extension of nancialservices provisions of NAFTAto reduce geographic barriers acrosscountries [Kroszner 1997]
APPENDIX EFFECTS OF DIFFERENT TYPES OF BRANCHING
DEREGULATION ON ECONOMIC GROWTH BANKING STRUCTUREAND BANK EFFICIENCY
We measure the importance of alternative forms of deregula-tion by their consequences for the economic growth and the
WHAT DRIVES DEREGULATION 1463
structure and efficiency of the banking sector across the states(see Jayaratne and Strahan [1996 1998]) To do so we run anOLS panel regression relating indicators for each type of deregula-tion to potential consequences Our dependent variables includeproxies for economic growth (growth in personal income) bankstructure (the log of the number of banking companies and thenumber of branches per banking company) and efficiency (the logof total noninterest expenses and the ratio of loan loss provisionsto total loans) The four deregulation indicators (MBHC branch-ing through merger and acquisition unrestricted statewide branch-ing and interstate banking) are equal to one in the years afterderegulation and zero before If a state deregulated before thesample period begins in 1970 the indicator is one throughout ifthe state does not deregulate then the indicator is zero through-out All regressions include state xed effects and time effects
Table VI shows the results of including the set of four
TABLE VIMULTIVARIATE REGRESSIONS OF THE EFFECTS OF DIFFERENT TYPES OF BRANCHING
DEREGULATION ON ECONOMIC GROWTH BANKING STRUCTUREAND BANK EFFICIENCY
Dependent variable
Growthin
personalincome
Log of of
bankingcompanies
Branchesper
bankingcompany
Log ofnon-
interestexpenses
Loan lossprovisions
(loans 1leases)
Merger amp acquisitionbranchingindicator
0013(0004)
2 0162(0037)
1146(0453)
2 0107(0035)
2 0007(0001)
Unrestricted state-wide branchingindicator
2 0008(0005)
0086(0059)
2 0043(0574)
0070(0037)
0005(0002)
Interstate bankingindicator
0006(0005)
2 0052(0057)
1064(0511)
0005(0030)
2 0001(0001)
Multibank holdingcompany indicator
2 0005(0006)
2 0096(0052)
2 0364(0529)
2 0095(0036)
2 0001(0001)
Dependent variablemean [median]
0086[0082]
4881[5136]
6649[5055]
13812[13803]
0009[0006]
This table contains regressions of state-level economic growth and state-level measures of the structureand efficiency of the banking industry Each dependent variable is regressed on a set of state indicatorvariables a set of year indicator variables and the set of four deregulation indicators these deregulationindicators are equal to one after deregulation and zero before If a state was always regulated the indicator iszero throughout if it was always deregulated the indicator is one throughout Sample period is 1975 to 1991and Delaware and South Dakota are dropped so N 5 833 Noninterest expenses become available only in1984 so N 5 392 The coefficient on the deregulation indicators is reported with robust standard errors inparentheses and denote statistically signicant at the 1 5 and 10 percent levels
QUARTERLY JOURNAL OF ECONOMICS1464
deregulation indicators simultaneously The branching by mergerand acquisition indicator is statistically signicant and of theexpected sign in all ve of the equations and larger in absolutevalue than any of the other coefficients on the indicators for theother types of deregulation Since there may be multicollinearityamong our indicators we also ran univariate specications whichincluded each indicator separately and the results are verysimilar These results imply that branching by merger andacquisition is the most important type of deregulation measuredby its consequences for economic growth banking structure andbank efficiency See Kroszner and Strahan [1998] for more details
UNIVERSITY OF CHICAGO GRADUATE SCHOOL OF BUSINESS AND NATIONAL BUREAU OF
ECONOMIC RESEARCH
FEDERAL RESERVE BANK OF NEW YORK RESEARCH amp MARKET ANALYSIS GROUP
REFERENCES
Abrams Burton A and Russell F Settle lsquolsquoPressure-Group Inuence and Institu-tional Change Branch-Banking Legislation during the Great DepressionrsquorsquoPublic Choice LXXVII (1993) 687ndash705
Amel Dean lsquolsquoState Laws Affecting the Geographic Expansion of CommercialBanksrsquorsquo unpublished manuscript Board of Governors of the Federal ReserveSystem 1993
Becker Gary S lsquolsquoA Theory of Competition among Pressure Groups for PoliticalInuencersquorsquo Quarterly Journal of Economics XCVIII (1983) 371ndash400
Berger Allen Anil Kashyap Joseph Scalise lsquolsquoThe Transformation of the U SBanking Industry What a Long Strange Trip Itrsquos Beenrsquorsquo Brookings Papers onEconomic Activity 2 (1995) 55ndash218
Berger Allen Anthony Saunders Joseph Scalise and Gregory Udell lsquolsquoThe Effectsof Bank Mergers and Acquisitions on Small Business Lendingrsquorsquo Journal ofFinancial Economics L (1998) 187ndash229
Brennan Geoffrey and James Buchanan lsquolsquoTowards a Tax Constitution forLeviathanrsquorsquo Journal of Public Economics VIII (1977) 255ndash273
Brickley James A and Christopher M James lsquolsquoThe Takeover Market CorporateBoard Composition and Ownership Structure The Case of Bankingrsquorsquo Journalof Law and Economics XXX (1987) 161ndash180
Cole Rebel and John Wolken lsquolsquoFinancial Services Used by Small BusinessesEvidence from the 1993 Survey of Small Business Financesrsquorsquo Federal ReserveBulletin LXXXI (1994) 629ndash676
Dixit Avinash The Making of Economic Policy A Transactions-Cost PoliticsPerspective (Cambridge MA MIT Press 1996)
Economides Nicolas R Glenn Hubbard and Darius Palia lsquolsquoThe Political Economyof Branching Restrictions and Deposit Insurancersquorsquo Journal of Law andEconomics XXIX (1996) 667ndash704
Flannery Mark lsquolsquoThe Social Costs of Unit Banking Restrictionsrsquorsquo Journal ofMonetary Economics XIII (1984) 237ndash249
Greene William H Econometric Analysis third edition (Upper Saddle River NJPrentice-Hall 1997)
Gunther Jeffery lsquolsquoRegional Capital Imbalances and the Removal of InterstateBanking Restrictionsrsquorsquo Economics Letters XLIV (1994) 439ndash442 lsquolsquoThe Erosion of Intrastate Branch Banking Restrictionsrsquorsquo unpublishedmanuscript Federal Reserve Bank of Dallas 1996
Irwin Douglas and Randall Kroszner lsquolsquoInterests Institutions and Ideology inSecuring Policy Change The Republican Conversion to Trade Liberalizationafter Smoot-Hawleyrsquorsquo Journal of Law and Economics XLII (1999) 643ndash673
WHAT DRIVES DEREGULATION 1465
James Christopher lsquolsquoAn Analysis of Intra-Industry Differences in the Effect ofRegulation The Case of Deposit Rate Ceilingsrsquorsquo Journal of Monetary Econom-ics XII (1983) 417ndash442
Jarrell Gregg lsquolsquoChange at the Exchange The Causes and Effects of DeregulationrsquorsquoJournal of Law and Economics XXVII (1984) 273ndash312
Jayaratne Jith and Philip E Strahan lsquolsquoThe Finance-Growth Nexus Evidencefrom Bank Branch Deregulationrsquorsquo Quarterly Journal of Economics CI (1996)639ndash670
Jayaratne Jith and Philip E Strahan lsquolsquoEntry Restrictions Industry Evolutionand Dynamic Efficiency Evidence from Commercial Bankingrsquorsquo Journal of Lawand Economics XLIX (1998) 239ndash274
Joskow Paul lsquolsquoComment on lsquoPower Struggles Explaining Deregulatory Reformsin Electricity Marketsrsquo rsquorsquo Brookings Papers on Economic Activity Microeconom-ics (1996) 251ndash264
Joskow Paul and Roger Noll lsquolsquoRegulation in Theory and Practice An Overviewrsquorsquoin Studies in Public Regulation Gary Fromm ed (Cambridge MA MITPress 1981) pp 1ndash65
Kahn Matthew E and John G Matsusaka lsquolsquoDemand for Environmental GoodsEvidence from Voting Patterns on California Initiativesrsquorsquo Journal of Law andEconomics XL (1997) 137ndash173
Kalbeisch J and R Prentice The Statistical Analysis of Failure Time Data (NewYork John Wiley amp Sons 1980)
Kane Edward J lsquolsquoDe Jure Interstate Banking Why Only NowrsquorsquoJournal of MoneyCredit and Banking XXVIII (1996) 141ndash161
Kiefer Nicholas lsquolsquoEconomic Duration Data and Hazard Functionsrsquorsquo Journal ofEconomic Literature XXVI (1988) 646ndash679
Kroszner Randall S lsquolsquoThe Evolution of Universal Banking and Its Regulation inTwentieth Century Americarsquorsquo in Universal Banking Financial System DesignReconsidered Anthony Saunders and Ingo Walter eds (New York IrwinProfessional Publishers 1996) pp 70ndash99 lsquolsquoThe Political Economy of Banking and Financial Regulation in the U Srsquorsquo inThe Banking and Financial Structure in the NAFTA Countries and ChileGeorge M von Furstenberg ed (Boston Kluwer Academic Publishers 1997)pp 200ndash213
Kroszner Randall S and Raghuram G Rajan lsquolsquoIs the Glass-Steagall ActJustied A Study of the U S Experience with Universal Banking before1933rsquorsquo American Economic Review LXXXIV (1994) 810ndash832
Kroszner Randall S and Raghuram G Rajan lsquolsquoOrganization Structure andCredibility Evidence from Bank Securities Activities before the Glass-Steagall Actrsquorsquo Journal of Monetary Economics XXXIX (1997) 475ndash516
Kroszner Randall S and Philip E Strahan lsquolsquoRegulatory Incentives and the ThriftCrisis Dividends Mutual-to-Stock Conversions and Financial DistressrsquorsquoJournal of Finance LI (1996) 1285ndash1320
Kroszner Randall S and Philip E Strahan lsquolsquoWhat Drives Deregulation Econom-ics and Politics of the Relaxation of Bank Branching Restrictionsrsquorsquo NBERWorking Paper No 6637 1998
Kroszner Randall S and Thomas Stratmann lsquolsquoInterest Group Competition andthe Organization of Congress Theory and Evidence from Financial ServicesPolitical Action Committeesrsquorsquo American Economic Review LXXXVIII (1998)pp 1163ndash1187
Kroszner Randall S and Thomas Stratmann lsquolsquoDoes Political Ambiguity PayCorporate Campaign Contributions and Rewards to Legislator Reputationrsquorsquounpublishedmanuscript University of Chicago Graduate School of Business1999
Lin D Y and L J Wei lsquolsquoThe Robust Inference for the Cox Proportional HazardsModel Journal of the American Statistical Association LXXXIV (1989)1074ndash1078
Macey Jonathan R and Geoffrey P Miller Banking Law and Regulation (BostonLittle Brown and Company 1992)
Makinson Larry Open Secrets The Encyclopedia of Congressional Money andPolitics (Washington DC Congressional Quarterly 1992)
Niskanen WilliamBureaucracy and Representative Government (Chicago Aldine-Atherton 1971)
QUARTERLY JOURNAL OF ECONOMICS1466
Noll Roger lsquolsquoComment on Peltzmanrsquorsquo Brookings Papers Microeconomics SpecialIssue (1989) 48ndash58
North Douglass lsquolsquoA Transactions Cost Theory of Politicsrsquorsquo Journal of TheoreticalPolitics II (1990) 355ndash367
Peltzman Sam lsquolsquoToward a More General Theory of Regulationrsquorsquo Journal of Lawand Economics XIX (1976) 109ndash148 lsquolsquoConstituent Interest and Congressional Votingrsquorsquo Journal of Law andEconomics XXVII (1984) 181ndash210 lsquolsquoThe Economic Theory of Regulation after a Decade of DeregulationrsquorsquoBrookings Papers Microeconomics Special Issue (1989) 1ndash41
Peterson Mitchell and Raghuram Rajan lsquolsquoThe Benets of Lending RelationshipsEvidence from Small Business Datarsquorsquo Journal of Finance XLIX (1994) 3ndash38
Poole Keith and Howard Rosenthal Congress A Political-Economic History ofRoll Call Voting (Oxford Oxford University Press 1997)
Stata Stata Reference Manual Release 5 (College Station TX Stata Press 1997)Stigler George J lsquolsquoThe Theory of Economic Regulationrsquorsquo Bell Journal of Economics
and Management Science II (1971) 3ndash21Stigler George J Chicago Studies in the Political Economy (Chicago University
of Chicago Press 1988)Strahan Philip E and James Weston lsquolsquoSmall Business Lending and the Changing
Structure of the Banking Industryrsquorsquo Journal of Banking and Finance XXII(1998) 821ndash845
Sylla Richard John Legler and John Wallis lsquolsquoBanks and State Public Finance inthe New Republic The United States 1790ndash1860rsquorsquo Journal of EconomicHistory XLVII (1987) 391ndash403
White Eugene The Regulation and Reform of the American Banking System1900ndash1929 (Princeton Princeton University Press 1983)
Winston Clifford lsquolsquoEconomic Deregulation Days of Reckoning for Microecono-mistsrsquorsquo Journal of Economic Literature XXXI (1993) 1263ndash1289
WHAT DRIVES DEREGULATION 1467
regulation could explain the positive duration dependence duringthe sample period
VI CONCLUSIONS
The private-interest theory of regulation can account for thepattern of bank branching deregulation of the 1970s and 1980sBeneciaries of branching regulation had supported a coalitionfavoring geographical restrictions despite their costs to consum-ers of nancial services Innovations that began in the 1970saltered the value of the restrictions to the affected parties and theresulting competition among interest groups can explain thesubsequent deregulation While some of our results are alsoconsistent with the public interest theorymdashfor example deregula-tion occurs earlier when small banks are in a relatively weaknancial positionmdashother results particularly evidence on theimportance of rivalries between small and large banks andbetween banking and insurance are difficult to explain with thepublic-interest approach We also nd that political-institutionalfactors affect deregulation although these variables may act asproxies for unmeasured economic interests The greater success ofthe private-interest theory to explain deregulation here than inprevious work may be due to the richer cross-sectional andtime-series variation in branching deregulation that can beexploited with a hazard model Future empirical studies ofendogenous deregulation thus may be most fruitful where changehas occurred across states over time such as in franchisinginsurance and public utilities (eg Joskow [1996])
Technological and nancial innovations will continue to erodethe benets to any interest group of maintaining regulatorybarriers in nancial services These forces are likely to bringabout reforms both domestically for example through legislationthat would increase bank powers (see Kroszner [1996] Krosznerand Rajan [1994 1997] and Kroszner and Stratmann [1998]) andinternationally for example through the extension of nancialservices provisions of NAFTAto reduce geographic barriers acrosscountries [Kroszner 1997]
APPENDIX EFFECTS OF DIFFERENT TYPES OF BRANCHING
DEREGULATION ON ECONOMIC GROWTH BANKING STRUCTUREAND BANK EFFICIENCY
We measure the importance of alternative forms of deregula-tion by their consequences for the economic growth and the
WHAT DRIVES DEREGULATION 1463
structure and efficiency of the banking sector across the states(see Jayaratne and Strahan [1996 1998]) To do so we run anOLS panel regression relating indicators for each type of deregula-tion to potential consequences Our dependent variables includeproxies for economic growth (growth in personal income) bankstructure (the log of the number of banking companies and thenumber of branches per banking company) and efficiency (the logof total noninterest expenses and the ratio of loan loss provisionsto total loans) The four deregulation indicators (MBHC branch-ing through merger and acquisition unrestricted statewide branch-ing and interstate banking) are equal to one in the years afterderegulation and zero before If a state deregulated before thesample period begins in 1970 the indicator is one throughout ifthe state does not deregulate then the indicator is zero through-out All regressions include state xed effects and time effects
Table VI shows the results of including the set of four
TABLE VIMULTIVARIATE REGRESSIONS OF THE EFFECTS OF DIFFERENT TYPES OF BRANCHING
DEREGULATION ON ECONOMIC GROWTH BANKING STRUCTUREAND BANK EFFICIENCY
Dependent variable
Growthin
personalincome
Log of of
bankingcompanies
Branchesper
bankingcompany
Log ofnon-
interestexpenses
Loan lossprovisions
(loans 1leases)
Merger amp acquisitionbranchingindicator
0013(0004)
2 0162(0037)
1146(0453)
2 0107(0035)
2 0007(0001)
Unrestricted state-wide branchingindicator
2 0008(0005)
0086(0059)
2 0043(0574)
0070(0037)
0005(0002)
Interstate bankingindicator
0006(0005)
2 0052(0057)
1064(0511)
0005(0030)
2 0001(0001)
Multibank holdingcompany indicator
2 0005(0006)
2 0096(0052)
2 0364(0529)
2 0095(0036)
2 0001(0001)
Dependent variablemean [median]
0086[0082]
4881[5136]
6649[5055]
13812[13803]
0009[0006]
This table contains regressions of state-level economic growth and state-level measures of the structureand efficiency of the banking industry Each dependent variable is regressed on a set of state indicatorvariables a set of year indicator variables and the set of four deregulation indicators these deregulationindicators are equal to one after deregulation and zero before If a state was always regulated the indicator iszero throughout if it was always deregulated the indicator is one throughout Sample period is 1975 to 1991and Delaware and South Dakota are dropped so N 5 833 Noninterest expenses become available only in1984 so N 5 392 The coefficient on the deregulation indicators is reported with robust standard errors inparentheses and denote statistically signicant at the 1 5 and 10 percent levels
QUARTERLY JOURNAL OF ECONOMICS1464
deregulation indicators simultaneously The branching by mergerand acquisition indicator is statistically signicant and of theexpected sign in all ve of the equations and larger in absolutevalue than any of the other coefficients on the indicators for theother types of deregulation Since there may be multicollinearityamong our indicators we also ran univariate specications whichincluded each indicator separately and the results are verysimilar These results imply that branching by merger andacquisition is the most important type of deregulation measuredby its consequences for economic growth banking structure andbank efficiency See Kroszner and Strahan [1998] for more details
UNIVERSITY OF CHICAGO GRADUATE SCHOOL OF BUSINESS AND NATIONAL BUREAU OF
ECONOMIC RESEARCH
FEDERAL RESERVE BANK OF NEW YORK RESEARCH amp MARKET ANALYSIS GROUP
REFERENCES
Abrams Burton A and Russell F Settle lsquolsquoPressure-Group Inuence and Institu-tional Change Branch-Banking Legislation during the Great DepressionrsquorsquoPublic Choice LXXVII (1993) 687ndash705
Amel Dean lsquolsquoState Laws Affecting the Geographic Expansion of CommercialBanksrsquorsquo unpublished manuscript Board of Governors of the Federal ReserveSystem 1993
Becker Gary S lsquolsquoA Theory of Competition among Pressure Groups for PoliticalInuencersquorsquo Quarterly Journal of Economics XCVIII (1983) 371ndash400
Berger Allen Anil Kashyap Joseph Scalise lsquolsquoThe Transformation of the U SBanking Industry What a Long Strange Trip Itrsquos Beenrsquorsquo Brookings Papers onEconomic Activity 2 (1995) 55ndash218
Berger Allen Anthony Saunders Joseph Scalise and Gregory Udell lsquolsquoThe Effectsof Bank Mergers and Acquisitions on Small Business Lendingrsquorsquo Journal ofFinancial Economics L (1998) 187ndash229
Brennan Geoffrey and James Buchanan lsquolsquoTowards a Tax Constitution forLeviathanrsquorsquo Journal of Public Economics VIII (1977) 255ndash273
Brickley James A and Christopher M James lsquolsquoThe Takeover Market CorporateBoard Composition and Ownership Structure The Case of Bankingrsquorsquo Journalof Law and Economics XXX (1987) 161ndash180
Cole Rebel and John Wolken lsquolsquoFinancial Services Used by Small BusinessesEvidence from the 1993 Survey of Small Business Financesrsquorsquo Federal ReserveBulletin LXXXI (1994) 629ndash676
Dixit Avinash The Making of Economic Policy A Transactions-Cost PoliticsPerspective (Cambridge MA MIT Press 1996)
Economides Nicolas R Glenn Hubbard and Darius Palia lsquolsquoThe Political Economyof Branching Restrictions and Deposit Insurancersquorsquo Journal of Law andEconomics XXIX (1996) 667ndash704
Flannery Mark lsquolsquoThe Social Costs of Unit Banking Restrictionsrsquorsquo Journal ofMonetary Economics XIII (1984) 237ndash249
Greene William H Econometric Analysis third edition (Upper Saddle River NJPrentice-Hall 1997)
Gunther Jeffery lsquolsquoRegional Capital Imbalances and the Removal of InterstateBanking Restrictionsrsquorsquo Economics Letters XLIV (1994) 439ndash442 lsquolsquoThe Erosion of Intrastate Branch Banking Restrictionsrsquorsquo unpublishedmanuscript Federal Reserve Bank of Dallas 1996
Irwin Douglas and Randall Kroszner lsquolsquoInterests Institutions and Ideology inSecuring Policy Change The Republican Conversion to Trade Liberalizationafter Smoot-Hawleyrsquorsquo Journal of Law and Economics XLII (1999) 643ndash673
WHAT DRIVES DEREGULATION 1465
James Christopher lsquolsquoAn Analysis of Intra-Industry Differences in the Effect ofRegulation The Case of Deposit Rate Ceilingsrsquorsquo Journal of Monetary Econom-ics XII (1983) 417ndash442
Jarrell Gregg lsquolsquoChange at the Exchange The Causes and Effects of DeregulationrsquorsquoJournal of Law and Economics XXVII (1984) 273ndash312
Jayaratne Jith and Philip E Strahan lsquolsquoThe Finance-Growth Nexus Evidencefrom Bank Branch Deregulationrsquorsquo Quarterly Journal of Economics CI (1996)639ndash670
Jayaratne Jith and Philip E Strahan lsquolsquoEntry Restrictions Industry Evolutionand Dynamic Efficiency Evidence from Commercial Bankingrsquorsquo Journal of Lawand Economics XLIX (1998) 239ndash274
Joskow Paul lsquolsquoComment on lsquoPower Struggles Explaining Deregulatory Reformsin Electricity Marketsrsquo rsquorsquo Brookings Papers on Economic Activity Microeconom-ics (1996) 251ndash264
Joskow Paul and Roger Noll lsquolsquoRegulation in Theory and Practice An Overviewrsquorsquoin Studies in Public Regulation Gary Fromm ed (Cambridge MA MITPress 1981) pp 1ndash65
Kahn Matthew E and John G Matsusaka lsquolsquoDemand for Environmental GoodsEvidence from Voting Patterns on California Initiativesrsquorsquo Journal of Law andEconomics XL (1997) 137ndash173
Kalbeisch J and R Prentice The Statistical Analysis of Failure Time Data (NewYork John Wiley amp Sons 1980)
Kane Edward J lsquolsquoDe Jure Interstate Banking Why Only NowrsquorsquoJournal of MoneyCredit and Banking XXVIII (1996) 141ndash161
Kiefer Nicholas lsquolsquoEconomic Duration Data and Hazard Functionsrsquorsquo Journal ofEconomic Literature XXVI (1988) 646ndash679
Kroszner Randall S lsquolsquoThe Evolution of Universal Banking and Its Regulation inTwentieth Century Americarsquorsquo in Universal Banking Financial System DesignReconsidered Anthony Saunders and Ingo Walter eds (New York IrwinProfessional Publishers 1996) pp 70ndash99 lsquolsquoThe Political Economy of Banking and Financial Regulation in the U Srsquorsquo inThe Banking and Financial Structure in the NAFTA Countries and ChileGeorge M von Furstenberg ed (Boston Kluwer Academic Publishers 1997)pp 200ndash213
Kroszner Randall S and Raghuram G Rajan lsquolsquoIs the Glass-Steagall ActJustied A Study of the U S Experience with Universal Banking before1933rsquorsquo American Economic Review LXXXIV (1994) 810ndash832
Kroszner Randall S and Raghuram G Rajan lsquolsquoOrganization Structure andCredibility Evidence from Bank Securities Activities before the Glass-Steagall Actrsquorsquo Journal of Monetary Economics XXXIX (1997) 475ndash516
Kroszner Randall S and Philip E Strahan lsquolsquoRegulatory Incentives and the ThriftCrisis Dividends Mutual-to-Stock Conversions and Financial DistressrsquorsquoJournal of Finance LI (1996) 1285ndash1320
Kroszner Randall S and Philip E Strahan lsquolsquoWhat Drives Deregulation Econom-ics and Politics of the Relaxation of Bank Branching Restrictionsrsquorsquo NBERWorking Paper No 6637 1998
Kroszner Randall S and Thomas Stratmann lsquolsquoInterest Group Competition andthe Organization of Congress Theory and Evidence from Financial ServicesPolitical Action Committeesrsquorsquo American Economic Review LXXXVIII (1998)pp 1163ndash1187
Kroszner Randall S and Thomas Stratmann lsquolsquoDoes Political Ambiguity PayCorporate Campaign Contributions and Rewards to Legislator Reputationrsquorsquounpublishedmanuscript University of Chicago Graduate School of Business1999
Lin D Y and L J Wei lsquolsquoThe Robust Inference for the Cox Proportional HazardsModel Journal of the American Statistical Association LXXXIV (1989)1074ndash1078
Macey Jonathan R and Geoffrey P Miller Banking Law and Regulation (BostonLittle Brown and Company 1992)
Makinson Larry Open Secrets The Encyclopedia of Congressional Money andPolitics (Washington DC Congressional Quarterly 1992)
Niskanen WilliamBureaucracy and Representative Government (Chicago Aldine-Atherton 1971)
QUARTERLY JOURNAL OF ECONOMICS1466
Noll Roger lsquolsquoComment on Peltzmanrsquorsquo Brookings Papers Microeconomics SpecialIssue (1989) 48ndash58
North Douglass lsquolsquoA Transactions Cost Theory of Politicsrsquorsquo Journal of TheoreticalPolitics II (1990) 355ndash367
Peltzman Sam lsquolsquoToward a More General Theory of Regulationrsquorsquo Journal of Lawand Economics XIX (1976) 109ndash148 lsquolsquoConstituent Interest and Congressional Votingrsquorsquo Journal of Law andEconomics XXVII (1984) 181ndash210 lsquolsquoThe Economic Theory of Regulation after a Decade of DeregulationrsquorsquoBrookings Papers Microeconomics Special Issue (1989) 1ndash41
Peterson Mitchell and Raghuram Rajan lsquolsquoThe Benets of Lending RelationshipsEvidence from Small Business Datarsquorsquo Journal of Finance XLIX (1994) 3ndash38
Poole Keith and Howard Rosenthal Congress A Political-Economic History ofRoll Call Voting (Oxford Oxford University Press 1997)
Stata Stata Reference Manual Release 5 (College Station TX Stata Press 1997)Stigler George J lsquolsquoThe Theory of Economic Regulationrsquorsquo Bell Journal of Economics
and Management Science II (1971) 3ndash21Stigler George J Chicago Studies in the Political Economy (Chicago University
of Chicago Press 1988)Strahan Philip E and James Weston lsquolsquoSmall Business Lending and the Changing
Structure of the Banking Industryrsquorsquo Journal of Banking and Finance XXII(1998) 821ndash845
Sylla Richard John Legler and John Wallis lsquolsquoBanks and State Public Finance inthe New Republic The United States 1790ndash1860rsquorsquo Journal of EconomicHistory XLVII (1987) 391ndash403
White Eugene The Regulation and Reform of the American Banking System1900ndash1929 (Princeton Princeton University Press 1983)
Winston Clifford lsquolsquoEconomic Deregulation Days of Reckoning for Microecono-mistsrsquorsquo Journal of Economic Literature XXXI (1993) 1263ndash1289
WHAT DRIVES DEREGULATION 1467
structure and efficiency of the banking sector across the states(see Jayaratne and Strahan [1996 1998]) To do so we run anOLS panel regression relating indicators for each type of deregula-tion to potential consequences Our dependent variables includeproxies for economic growth (growth in personal income) bankstructure (the log of the number of banking companies and thenumber of branches per banking company) and efficiency (the logof total noninterest expenses and the ratio of loan loss provisionsto total loans) The four deregulation indicators (MBHC branch-ing through merger and acquisition unrestricted statewide branch-ing and interstate banking) are equal to one in the years afterderegulation and zero before If a state deregulated before thesample period begins in 1970 the indicator is one throughout ifthe state does not deregulate then the indicator is zero through-out All regressions include state xed effects and time effects
Table VI shows the results of including the set of four
TABLE VIMULTIVARIATE REGRESSIONS OF THE EFFECTS OF DIFFERENT TYPES OF BRANCHING
DEREGULATION ON ECONOMIC GROWTH BANKING STRUCTUREAND BANK EFFICIENCY
Dependent variable
Growthin
personalincome
Log of of
bankingcompanies
Branchesper
bankingcompany
Log ofnon-
interestexpenses
Loan lossprovisions
(loans 1leases)
Merger amp acquisitionbranchingindicator
0013(0004)
2 0162(0037)
1146(0453)
2 0107(0035)
2 0007(0001)
Unrestricted state-wide branchingindicator
2 0008(0005)
0086(0059)
2 0043(0574)
0070(0037)
0005(0002)
Interstate bankingindicator
0006(0005)
2 0052(0057)
1064(0511)
0005(0030)
2 0001(0001)
Multibank holdingcompany indicator
2 0005(0006)
2 0096(0052)
2 0364(0529)
2 0095(0036)
2 0001(0001)
Dependent variablemean [median]
0086[0082]
4881[5136]
6649[5055]
13812[13803]
0009[0006]
This table contains regressions of state-level economic growth and state-level measures of the structureand efficiency of the banking industry Each dependent variable is regressed on a set of state indicatorvariables a set of year indicator variables and the set of four deregulation indicators these deregulationindicators are equal to one after deregulation and zero before If a state was always regulated the indicator iszero throughout if it was always deregulated the indicator is one throughout Sample period is 1975 to 1991and Delaware and South Dakota are dropped so N 5 833 Noninterest expenses become available only in1984 so N 5 392 The coefficient on the deregulation indicators is reported with robust standard errors inparentheses and denote statistically signicant at the 1 5 and 10 percent levels
QUARTERLY JOURNAL OF ECONOMICS1464
deregulation indicators simultaneously The branching by mergerand acquisition indicator is statistically signicant and of theexpected sign in all ve of the equations and larger in absolutevalue than any of the other coefficients on the indicators for theother types of deregulation Since there may be multicollinearityamong our indicators we also ran univariate specications whichincluded each indicator separately and the results are verysimilar These results imply that branching by merger andacquisition is the most important type of deregulation measuredby its consequences for economic growth banking structure andbank efficiency See Kroszner and Strahan [1998] for more details
UNIVERSITY OF CHICAGO GRADUATE SCHOOL OF BUSINESS AND NATIONAL BUREAU OF
ECONOMIC RESEARCH
FEDERAL RESERVE BANK OF NEW YORK RESEARCH amp MARKET ANALYSIS GROUP
REFERENCES
Abrams Burton A and Russell F Settle lsquolsquoPressure-Group Inuence and Institu-tional Change Branch-Banking Legislation during the Great DepressionrsquorsquoPublic Choice LXXVII (1993) 687ndash705
Amel Dean lsquolsquoState Laws Affecting the Geographic Expansion of CommercialBanksrsquorsquo unpublished manuscript Board of Governors of the Federal ReserveSystem 1993
Becker Gary S lsquolsquoA Theory of Competition among Pressure Groups for PoliticalInuencersquorsquo Quarterly Journal of Economics XCVIII (1983) 371ndash400
Berger Allen Anil Kashyap Joseph Scalise lsquolsquoThe Transformation of the U SBanking Industry What a Long Strange Trip Itrsquos Beenrsquorsquo Brookings Papers onEconomic Activity 2 (1995) 55ndash218
Berger Allen Anthony Saunders Joseph Scalise and Gregory Udell lsquolsquoThe Effectsof Bank Mergers and Acquisitions on Small Business Lendingrsquorsquo Journal ofFinancial Economics L (1998) 187ndash229
Brennan Geoffrey and James Buchanan lsquolsquoTowards a Tax Constitution forLeviathanrsquorsquo Journal of Public Economics VIII (1977) 255ndash273
Brickley James A and Christopher M James lsquolsquoThe Takeover Market CorporateBoard Composition and Ownership Structure The Case of Bankingrsquorsquo Journalof Law and Economics XXX (1987) 161ndash180
Cole Rebel and John Wolken lsquolsquoFinancial Services Used by Small BusinessesEvidence from the 1993 Survey of Small Business Financesrsquorsquo Federal ReserveBulletin LXXXI (1994) 629ndash676
Dixit Avinash The Making of Economic Policy A Transactions-Cost PoliticsPerspective (Cambridge MA MIT Press 1996)
Economides Nicolas R Glenn Hubbard and Darius Palia lsquolsquoThe Political Economyof Branching Restrictions and Deposit Insurancersquorsquo Journal of Law andEconomics XXIX (1996) 667ndash704
Flannery Mark lsquolsquoThe Social Costs of Unit Banking Restrictionsrsquorsquo Journal ofMonetary Economics XIII (1984) 237ndash249
Greene William H Econometric Analysis third edition (Upper Saddle River NJPrentice-Hall 1997)
Gunther Jeffery lsquolsquoRegional Capital Imbalances and the Removal of InterstateBanking Restrictionsrsquorsquo Economics Letters XLIV (1994) 439ndash442 lsquolsquoThe Erosion of Intrastate Branch Banking Restrictionsrsquorsquo unpublishedmanuscript Federal Reserve Bank of Dallas 1996
Irwin Douglas and Randall Kroszner lsquolsquoInterests Institutions and Ideology inSecuring Policy Change The Republican Conversion to Trade Liberalizationafter Smoot-Hawleyrsquorsquo Journal of Law and Economics XLII (1999) 643ndash673
WHAT DRIVES DEREGULATION 1465
James Christopher lsquolsquoAn Analysis of Intra-Industry Differences in the Effect ofRegulation The Case of Deposit Rate Ceilingsrsquorsquo Journal of Monetary Econom-ics XII (1983) 417ndash442
Jarrell Gregg lsquolsquoChange at the Exchange The Causes and Effects of DeregulationrsquorsquoJournal of Law and Economics XXVII (1984) 273ndash312
Jayaratne Jith and Philip E Strahan lsquolsquoThe Finance-Growth Nexus Evidencefrom Bank Branch Deregulationrsquorsquo Quarterly Journal of Economics CI (1996)639ndash670
Jayaratne Jith and Philip E Strahan lsquolsquoEntry Restrictions Industry Evolutionand Dynamic Efficiency Evidence from Commercial Bankingrsquorsquo Journal of Lawand Economics XLIX (1998) 239ndash274
Joskow Paul lsquolsquoComment on lsquoPower Struggles Explaining Deregulatory Reformsin Electricity Marketsrsquo rsquorsquo Brookings Papers on Economic Activity Microeconom-ics (1996) 251ndash264
Joskow Paul and Roger Noll lsquolsquoRegulation in Theory and Practice An Overviewrsquorsquoin Studies in Public Regulation Gary Fromm ed (Cambridge MA MITPress 1981) pp 1ndash65
Kahn Matthew E and John G Matsusaka lsquolsquoDemand for Environmental GoodsEvidence from Voting Patterns on California Initiativesrsquorsquo Journal of Law andEconomics XL (1997) 137ndash173
Kalbeisch J and R Prentice The Statistical Analysis of Failure Time Data (NewYork John Wiley amp Sons 1980)
Kane Edward J lsquolsquoDe Jure Interstate Banking Why Only NowrsquorsquoJournal of MoneyCredit and Banking XXVIII (1996) 141ndash161
Kiefer Nicholas lsquolsquoEconomic Duration Data and Hazard Functionsrsquorsquo Journal ofEconomic Literature XXVI (1988) 646ndash679
Kroszner Randall S lsquolsquoThe Evolution of Universal Banking and Its Regulation inTwentieth Century Americarsquorsquo in Universal Banking Financial System DesignReconsidered Anthony Saunders and Ingo Walter eds (New York IrwinProfessional Publishers 1996) pp 70ndash99 lsquolsquoThe Political Economy of Banking and Financial Regulation in the U Srsquorsquo inThe Banking and Financial Structure in the NAFTA Countries and ChileGeorge M von Furstenberg ed (Boston Kluwer Academic Publishers 1997)pp 200ndash213
Kroszner Randall S and Raghuram G Rajan lsquolsquoIs the Glass-Steagall ActJustied A Study of the U S Experience with Universal Banking before1933rsquorsquo American Economic Review LXXXIV (1994) 810ndash832
Kroszner Randall S and Raghuram G Rajan lsquolsquoOrganization Structure andCredibility Evidence from Bank Securities Activities before the Glass-Steagall Actrsquorsquo Journal of Monetary Economics XXXIX (1997) 475ndash516
Kroszner Randall S and Philip E Strahan lsquolsquoRegulatory Incentives and the ThriftCrisis Dividends Mutual-to-Stock Conversions and Financial DistressrsquorsquoJournal of Finance LI (1996) 1285ndash1320
Kroszner Randall S and Philip E Strahan lsquolsquoWhat Drives Deregulation Econom-ics and Politics of the Relaxation of Bank Branching Restrictionsrsquorsquo NBERWorking Paper No 6637 1998
Kroszner Randall S and Thomas Stratmann lsquolsquoInterest Group Competition andthe Organization of Congress Theory and Evidence from Financial ServicesPolitical Action Committeesrsquorsquo American Economic Review LXXXVIII (1998)pp 1163ndash1187
Kroszner Randall S and Thomas Stratmann lsquolsquoDoes Political Ambiguity PayCorporate Campaign Contributions and Rewards to Legislator Reputationrsquorsquounpublishedmanuscript University of Chicago Graduate School of Business1999
Lin D Y and L J Wei lsquolsquoThe Robust Inference for the Cox Proportional HazardsModel Journal of the American Statistical Association LXXXIV (1989)1074ndash1078
Macey Jonathan R and Geoffrey P Miller Banking Law and Regulation (BostonLittle Brown and Company 1992)
Makinson Larry Open Secrets The Encyclopedia of Congressional Money andPolitics (Washington DC Congressional Quarterly 1992)
Niskanen WilliamBureaucracy and Representative Government (Chicago Aldine-Atherton 1971)
QUARTERLY JOURNAL OF ECONOMICS1466
Noll Roger lsquolsquoComment on Peltzmanrsquorsquo Brookings Papers Microeconomics SpecialIssue (1989) 48ndash58
North Douglass lsquolsquoA Transactions Cost Theory of Politicsrsquorsquo Journal of TheoreticalPolitics II (1990) 355ndash367
Peltzman Sam lsquolsquoToward a More General Theory of Regulationrsquorsquo Journal of Lawand Economics XIX (1976) 109ndash148 lsquolsquoConstituent Interest and Congressional Votingrsquorsquo Journal of Law andEconomics XXVII (1984) 181ndash210 lsquolsquoThe Economic Theory of Regulation after a Decade of DeregulationrsquorsquoBrookings Papers Microeconomics Special Issue (1989) 1ndash41
Peterson Mitchell and Raghuram Rajan lsquolsquoThe Benets of Lending RelationshipsEvidence from Small Business Datarsquorsquo Journal of Finance XLIX (1994) 3ndash38
Poole Keith and Howard Rosenthal Congress A Political-Economic History ofRoll Call Voting (Oxford Oxford University Press 1997)
Stata Stata Reference Manual Release 5 (College Station TX Stata Press 1997)Stigler George J lsquolsquoThe Theory of Economic Regulationrsquorsquo Bell Journal of Economics
and Management Science II (1971) 3ndash21Stigler George J Chicago Studies in the Political Economy (Chicago University
of Chicago Press 1988)Strahan Philip E and James Weston lsquolsquoSmall Business Lending and the Changing
Structure of the Banking Industryrsquorsquo Journal of Banking and Finance XXII(1998) 821ndash845
Sylla Richard John Legler and John Wallis lsquolsquoBanks and State Public Finance inthe New Republic The United States 1790ndash1860rsquorsquo Journal of EconomicHistory XLVII (1987) 391ndash403
White Eugene The Regulation and Reform of the American Banking System1900ndash1929 (Princeton Princeton University Press 1983)
Winston Clifford lsquolsquoEconomic Deregulation Days of Reckoning for Microecono-mistsrsquorsquo Journal of Economic Literature XXXI (1993) 1263ndash1289
WHAT DRIVES DEREGULATION 1467
deregulation indicators simultaneously The branching by mergerand acquisition indicator is statistically signicant and of theexpected sign in all ve of the equations and larger in absolutevalue than any of the other coefficients on the indicators for theother types of deregulation Since there may be multicollinearityamong our indicators we also ran univariate specications whichincluded each indicator separately and the results are verysimilar These results imply that branching by merger andacquisition is the most important type of deregulation measuredby its consequences for economic growth banking structure andbank efficiency See Kroszner and Strahan [1998] for more details
UNIVERSITY OF CHICAGO GRADUATE SCHOOL OF BUSINESS AND NATIONAL BUREAU OF
ECONOMIC RESEARCH
FEDERAL RESERVE BANK OF NEW YORK RESEARCH amp MARKET ANALYSIS GROUP
REFERENCES
Abrams Burton A and Russell F Settle lsquolsquoPressure-Group Inuence and Institu-tional Change Branch-Banking Legislation during the Great DepressionrsquorsquoPublic Choice LXXVII (1993) 687ndash705
Amel Dean lsquolsquoState Laws Affecting the Geographic Expansion of CommercialBanksrsquorsquo unpublished manuscript Board of Governors of the Federal ReserveSystem 1993
Becker Gary S lsquolsquoA Theory of Competition among Pressure Groups for PoliticalInuencersquorsquo Quarterly Journal of Economics XCVIII (1983) 371ndash400
Berger Allen Anil Kashyap Joseph Scalise lsquolsquoThe Transformation of the U SBanking Industry What a Long Strange Trip Itrsquos Beenrsquorsquo Brookings Papers onEconomic Activity 2 (1995) 55ndash218
Berger Allen Anthony Saunders Joseph Scalise and Gregory Udell lsquolsquoThe Effectsof Bank Mergers and Acquisitions on Small Business Lendingrsquorsquo Journal ofFinancial Economics L (1998) 187ndash229
Brennan Geoffrey and James Buchanan lsquolsquoTowards a Tax Constitution forLeviathanrsquorsquo Journal of Public Economics VIII (1977) 255ndash273
Brickley James A and Christopher M James lsquolsquoThe Takeover Market CorporateBoard Composition and Ownership Structure The Case of Bankingrsquorsquo Journalof Law and Economics XXX (1987) 161ndash180
Cole Rebel and John Wolken lsquolsquoFinancial Services Used by Small BusinessesEvidence from the 1993 Survey of Small Business Financesrsquorsquo Federal ReserveBulletin LXXXI (1994) 629ndash676
Dixit Avinash The Making of Economic Policy A Transactions-Cost PoliticsPerspective (Cambridge MA MIT Press 1996)
Economides Nicolas R Glenn Hubbard and Darius Palia lsquolsquoThe Political Economyof Branching Restrictions and Deposit Insurancersquorsquo Journal of Law andEconomics XXIX (1996) 667ndash704
Flannery Mark lsquolsquoThe Social Costs of Unit Banking Restrictionsrsquorsquo Journal ofMonetary Economics XIII (1984) 237ndash249
Greene William H Econometric Analysis third edition (Upper Saddle River NJPrentice-Hall 1997)
Gunther Jeffery lsquolsquoRegional Capital Imbalances and the Removal of InterstateBanking Restrictionsrsquorsquo Economics Letters XLIV (1994) 439ndash442 lsquolsquoThe Erosion of Intrastate Branch Banking Restrictionsrsquorsquo unpublishedmanuscript Federal Reserve Bank of Dallas 1996
Irwin Douglas and Randall Kroszner lsquolsquoInterests Institutions and Ideology inSecuring Policy Change The Republican Conversion to Trade Liberalizationafter Smoot-Hawleyrsquorsquo Journal of Law and Economics XLII (1999) 643ndash673
WHAT DRIVES DEREGULATION 1465
James Christopher lsquolsquoAn Analysis of Intra-Industry Differences in the Effect ofRegulation The Case of Deposit Rate Ceilingsrsquorsquo Journal of Monetary Econom-ics XII (1983) 417ndash442
Jarrell Gregg lsquolsquoChange at the Exchange The Causes and Effects of DeregulationrsquorsquoJournal of Law and Economics XXVII (1984) 273ndash312
Jayaratne Jith and Philip E Strahan lsquolsquoThe Finance-Growth Nexus Evidencefrom Bank Branch Deregulationrsquorsquo Quarterly Journal of Economics CI (1996)639ndash670
Jayaratne Jith and Philip E Strahan lsquolsquoEntry Restrictions Industry Evolutionand Dynamic Efficiency Evidence from Commercial Bankingrsquorsquo Journal of Lawand Economics XLIX (1998) 239ndash274
Joskow Paul lsquolsquoComment on lsquoPower Struggles Explaining Deregulatory Reformsin Electricity Marketsrsquo rsquorsquo Brookings Papers on Economic Activity Microeconom-ics (1996) 251ndash264
Joskow Paul and Roger Noll lsquolsquoRegulation in Theory and Practice An Overviewrsquorsquoin Studies in Public Regulation Gary Fromm ed (Cambridge MA MITPress 1981) pp 1ndash65
Kahn Matthew E and John G Matsusaka lsquolsquoDemand for Environmental GoodsEvidence from Voting Patterns on California Initiativesrsquorsquo Journal of Law andEconomics XL (1997) 137ndash173
Kalbeisch J and R Prentice The Statistical Analysis of Failure Time Data (NewYork John Wiley amp Sons 1980)
Kane Edward J lsquolsquoDe Jure Interstate Banking Why Only NowrsquorsquoJournal of MoneyCredit and Banking XXVIII (1996) 141ndash161
Kiefer Nicholas lsquolsquoEconomic Duration Data and Hazard Functionsrsquorsquo Journal ofEconomic Literature XXVI (1988) 646ndash679
Kroszner Randall S lsquolsquoThe Evolution of Universal Banking and Its Regulation inTwentieth Century Americarsquorsquo in Universal Banking Financial System DesignReconsidered Anthony Saunders and Ingo Walter eds (New York IrwinProfessional Publishers 1996) pp 70ndash99 lsquolsquoThe Political Economy of Banking and Financial Regulation in the U Srsquorsquo inThe Banking and Financial Structure in the NAFTA Countries and ChileGeorge M von Furstenberg ed (Boston Kluwer Academic Publishers 1997)pp 200ndash213
Kroszner Randall S and Raghuram G Rajan lsquolsquoIs the Glass-Steagall ActJustied A Study of the U S Experience with Universal Banking before1933rsquorsquo American Economic Review LXXXIV (1994) 810ndash832
Kroszner Randall S and Raghuram G Rajan lsquolsquoOrganization Structure andCredibility Evidence from Bank Securities Activities before the Glass-Steagall Actrsquorsquo Journal of Monetary Economics XXXIX (1997) 475ndash516
Kroszner Randall S and Philip E Strahan lsquolsquoRegulatory Incentives and the ThriftCrisis Dividends Mutual-to-Stock Conversions and Financial DistressrsquorsquoJournal of Finance LI (1996) 1285ndash1320
Kroszner Randall S and Philip E Strahan lsquolsquoWhat Drives Deregulation Econom-ics and Politics of the Relaxation of Bank Branching Restrictionsrsquorsquo NBERWorking Paper No 6637 1998
Kroszner Randall S and Thomas Stratmann lsquolsquoInterest Group Competition andthe Organization of Congress Theory and Evidence from Financial ServicesPolitical Action Committeesrsquorsquo American Economic Review LXXXVIII (1998)pp 1163ndash1187
Kroszner Randall S and Thomas Stratmann lsquolsquoDoes Political Ambiguity PayCorporate Campaign Contributions and Rewards to Legislator Reputationrsquorsquounpublishedmanuscript University of Chicago Graduate School of Business1999
Lin D Y and L J Wei lsquolsquoThe Robust Inference for the Cox Proportional HazardsModel Journal of the American Statistical Association LXXXIV (1989)1074ndash1078
Macey Jonathan R and Geoffrey P Miller Banking Law and Regulation (BostonLittle Brown and Company 1992)
Makinson Larry Open Secrets The Encyclopedia of Congressional Money andPolitics (Washington DC Congressional Quarterly 1992)
Niskanen WilliamBureaucracy and Representative Government (Chicago Aldine-Atherton 1971)
QUARTERLY JOURNAL OF ECONOMICS1466
Noll Roger lsquolsquoComment on Peltzmanrsquorsquo Brookings Papers Microeconomics SpecialIssue (1989) 48ndash58
North Douglass lsquolsquoA Transactions Cost Theory of Politicsrsquorsquo Journal of TheoreticalPolitics II (1990) 355ndash367
Peltzman Sam lsquolsquoToward a More General Theory of Regulationrsquorsquo Journal of Lawand Economics XIX (1976) 109ndash148 lsquolsquoConstituent Interest and Congressional Votingrsquorsquo Journal of Law andEconomics XXVII (1984) 181ndash210 lsquolsquoThe Economic Theory of Regulation after a Decade of DeregulationrsquorsquoBrookings Papers Microeconomics Special Issue (1989) 1ndash41
Peterson Mitchell and Raghuram Rajan lsquolsquoThe Benets of Lending RelationshipsEvidence from Small Business Datarsquorsquo Journal of Finance XLIX (1994) 3ndash38
Poole Keith and Howard Rosenthal Congress A Political-Economic History ofRoll Call Voting (Oxford Oxford University Press 1997)
Stata Stata Reference Manual Release 5 (College Station TX Stata Press 1997)Stigler George J lsquolsquoThe Theory of Economic Regulationrsquorsquo Bell Journal of Economics
and Management Science II (1971) 3ndash21Stigler George J Chicago Studies in the Political Economy (Chicago University
of Chicago Press 1988)Strahan Philip E and James Weston lsquolsquoSmall Business Lending and the Changing
Structure of the Banking Industryrsquorsquo Journal of Banking and Finance XXII(1998) 821ndash845
Sylla Richard John Legler and John Wallis lsquolsquoBanks and State Public Finance inthe New Republic The United States 1790ndash1860rsquorsquo Journal of EconomicHistory XLVII (1987) 391ndash403
White Eugene The Regulation and Reform of the American Banking System1900ndash1929 (Princeton Princeton University Press 1983)
Winston Clifford lsquolsquoEconomic Deregulation Days of Reckoning for Microecono-mistsrsquorsquo Journal of Economic Literature XXXI (1993) 1263ndash1289
WHAT DRIVES DEREGULATION 1467
James Christopher lsquolsquoAn Analysis of Intra-Industry Differences in the Effect ofRegulation The Case of Deposit Rate Ceilingsrsquorsquo Journal of Monetary Econom-ics XII (1983) 417ndash442
Jarrell Gregg lsquolsquoChange at the Exchange The Causes and Effects of DeregulationrsquorsquoJournal of Law and Economics XXVII (1984) 273ndash312
Jayaratne Jith and Philip E Strahan lsquolsquoThe Finance-Growth Nexus Evidencefrom Bank Branch Deregulationrsquorsquo Quarterly Journal of Economics CI (1996)639ndash670
Jayaratne Jith and Philip E Strahan lsquolsquoEntry Restrictions Industry Evolutionand Dynamic Efficiency Evidence from Commercial Bankingrsquorsquo Journal of Lawand Economics XLIX (1998) 239ndash274
Joskow Paul lsquolsquoComment on lsquoPower Struggles Explaining Deregulatory Reformsin Electricity Marketsrsquo rsquorsquo Brookings Papers on Economic Activity Microeconom-ics (1996) 251ndash264
Joskow Paul and Roger Noll lsquolsquoRegulation in Theory and Practice An Overviewrsquorsquoin Studies in Public Regulation Gary Fromm ed (Cambridge MA MITPress 1981) pp 1ndash65
Kahn Matthew E and John G Matsusaka lsquolsquoDemand for Environmental GoodsEvidence from Voting Patterns on California Initiativesrsquorsquo Journal of Law andEconomics XL (1997) 137ndash173
Kalbeisch J and R Prentice The Statistical Analysis of Failure Time Data (NewYork John Wiley amp Sons 1980)
Kane Edward J lsquolsquoDe Jure Interstate Banking Why Only NowrsquorsquoJournal of MoneyCredit and Banking XXVIII (1996) 141ndash161
Kiefer Nicholas lsquolsquoEconomic Duration Data and Hazard Functionsrsquorsquo Journal ofEconomic Literature XXVI (1988) 646ndash679
Kroszner Randall S lsquolsquoThe Evolution of Universal Banking and Its Regulation inTwentieth Century Americarsquorsquo in Universal Banking Financial System DesignReconsidered Anthony Saunders and Ingo Walter eds (New York IrwinProfessional Publishers 1996) pp 70ndash99 lsquolsquoThe Political Economy of Banking and Financial Regulation in the U Srsquorsquo inThe Banking and Financial Structure in the NAFTA Countries and ChileGeorge M von Furstenberg ed (Boston Kluwer Academic Publishers 1997)pp 200ndash213
Kroszner Randall S and Raghuram G Rajan lsquolsquoIs the Glass-Steagall ActJustied A Study of the U S Experience with Universal Banking before1933rsquorsquo American Economic Review LXXXIV (1994) 810ndash832
Kroszner Randall S and Raghuram G Rajan lsquolsquoOrganization Structure andCredibility Evidence from Bank Securities Activities before the Glass-Steagall Actrsquorsquo Journal of Monetary Economics XXXIX (1997) 475ndash516
Kroszner Randall S and Philip E Strahan lsquolsquoRegulatory Incentives and the ThriftCrisis Dividends Mutual-to-Stock Conversions and Financial DistressrsquorsquoJournal of Finance LI (1996) 1285ndash1320
Kroszner Randall S and Philip E Strahan lsquolsquoWhat Drives Deregulation Econom-ics and Politics of the Relaxation of Bank Branching Restrictionsrsquorsquo NBERWorking Paper No 6637 1998
Kroszner Randall S and Thomas Stratmann lsquolsquoInterest Group Competition andthe Organization of Congress Theory and Evidence from Financial ServicesPolitical Action Committeesrsquorsquo American Economic Review LXXXVIII (1998)pp 1163ndash1187
Kroszner Randall S and Thomas Stratmann lsquolsquoDoes Political Ambiguity PayCorporate Campaign Contributions and Rewards to Legislator Reputationrsquorsquounpublishedmanuscript University of Chicago Graduate School of Business1999
Lin D Y and L J Wei lsquolsquoThe Robust Inference for the Cox Proportional HazardsModel Journal of the American Statistical Association LXXXIV (1989)1074ndash1078
Macey Jonathan R and Geoffrey P Miller Banking Law and Regulation (BostonLittle Brown and Company 1992)
Makinson Larry Open Secrets The Encyclopedia of Congressional Money andPolitics (Washington DC Congressional Quarterly 1992)
Niskanen WilliamBureaucracy and Representative Government (Chicago Aldine-Atherton 1971)
QUARTERLY JOURNAL OF ECONOMICS1466
Noll Roger lsquolsquoComment on Peltzmanrsquorsquo Brookings Papers Microeconomics SpecialIssue (1989) 48ndash58
North Douglass lsquolsquoA Transactions Cost Theory of Politicsrsquorsquo Journal of TheoreticalPolitics II (1990) 355ndash367
Peltzman Sam lsquolsquoToward a More General Theory of Regulationrsquorsquo Journal of Lawand Economics XIX (1976) 109ndash148 lsquolsquoConstituent Interest and Congressional Votingrsquorsquo Journal of Law andEconomics XXVII (1984) 181ndash210 lsquolsquoThe Economic Theory of Regulation after a Decade of DeregulationrsquorsquoBrookings Papers Microeconomics Special Issue (1989) 1ndash41
Peterson Mitchell and Raghuram Rajan lsquolsquoThe Benets of Lending RelationshipsEvidence from Small Business Datarsquorsquo Journal of Finance XLIX (1994) 3ndash38
Poole Keith and Howard Rosenthal Congress A Political-Economic History ofRoll Call Voting (Oxford Oxford University Press 1997)
Stata Stata Reference Manual Release 5 (College Station TX Stata Press 1997)Stigler George J lsquolsquoThe Theory of Economic Regulationrsquorsquo Bell Journal of Economics
and Management Science II (1971) 3ndash21Stigler George J Chicago Studies in the Political Economy (Chicago University
of Chicago Press 1988)Strahan Philip E and James Weston lsquolsquoSmall Business Lending and the Changing
Structure of the Banking Industryrsquorsquo Journal of Banking and Finance XXII(1998) 821ndash845
Sylla Richard John Legler and John Wallis lsquolsquoBanks and State Public Finance inthe New Republic The United States 1790ndash1860rsquorsquo Journal of EconomicHistory XLVII (1987) 391ndash403
White Eugene The Regulation and Reform of the American Banking System1900ndash1929 (Princeton Princeton University Press 1983)
Winston Clifford lsquolsquoEconomic Deregulation Days of Reckoning for Microecono-mistsrsquorsquo Journal of Economic Literature XXXI (1993) 1263ndash1289
WHAT DRIVES DEREGULATION 1467
Noll Roger lsquolsquoComment on Peltzmanrsquorsquo Brookings Papers Microeconomics SpecialIssue (1989) 48ndash58
North Douglass lsquolsquoA Transactions Cost Theory of Politicsrsquorsquo Journal of TheoreticalPolitics II (1990) 355ndash367
Peltzman Sam lsquolsquoToward a More General Theory of Regulationrsquorsquo Journal of Lawand Economics XIX (1976) 109ndash148 lsquolsquoConstituent Interest and Congressional Votingrsquorsquo Journal of Law andEconomics XXVII (1984) 181ndash210 lsquolsquoThe Economic Theory of Regulation after a Decade of DeregulationrsquorsquoBrookings Papers Microeconomics Special Issue (1989) 1ndash41
Peterson Mitchell and Raghuram Rajan lsquolsquoThe Benets of Lending RelationshipsEvidence from Small Business Datarsquorsquo Journal of Finance XLIX (1994) 3ndash38
Poole Keith and Howard Rosenthal Congress A Political-Economic History ofRoll Call Voting (Oxford Oxford University Press 1997)
Stata Stata Reference Manual Release 5 (College Station TX Stata Press 1997)Stigler George J lsquolsquoThe Theory of Economic Regulationrsquorsquo Bell Journal of Economics
and Management Science II (1971) 3ndash21Stigler George J Chicago Studies in the Political Economy (Chicago University
of Chicago Press 1988)Strahan Philip E and James Weston lsquolsquoSmall Business Lending and the Changing
Structure of the Banking Industryrsquorsquo Journal of Banking and Finance XXII(1998) 821ndash845
Sylla Richard John Legler and John Wallis lsquolsquoBanks and State Public Finance inthe New Republic The United States 1790ndash1860rsquorsquo Journal of EconomicHistory XLVII (1987) 391ndash403
White Eugene The Regulation and Reform of the American Banking System1900ndash1929 (Princeton Princeton University Press 1983)
Winston Clifford lsquolsquoEconomic Deregulation Days of Reckoning for Microecono-mistsrsquorsquo Journal of Economic Literature XXXI (1993) 1263ndash1289
WHAT DRIVES DEREGULATION 1467