the effects of electric utility decoupling on energy efficiency

64
The Effects of Electric Utility Decoupling on Energy Efficiency Jenya Kahn-Lang 1 In partial fulfillment of a Bachelor of Arts Degree in Economics and Environmental Analysis, 2011-12 academic year, Pomona College, Claremont, California 1 I would like to thank Professor Bowman Cutter for his encouragement and feedback throughout the course of this project. His constant support and direction made this project a possibility. I am extremely appreciative to professors John Jurewitz and Pierangelo De Pace for sharing their expertise in their respective fields with me. I would also like to thank Professors Gary Smith, Eleanor Brown, and Char Miller for their indispensible assistance and enthusiasm. Finally, I could not possibly thank my parents enough for all of their help over the years and with this project.

Upload: dangdieu

Post on 02-Jan-2017

214 views

Category:

Documents


0 download

TRANSCRIPT

TheEffectsofElectricUtilityDecouplingonEnergyEfficiency

JenyaKahn­Lang1

InpartialfulfillmentofaBachelorofArtsDegreeinEconomicsand

EnvironmentalAnalysis,

2011­12academicyear,PomonaCollege,Claremont,California

1IwouldliketothankProfessorBowmanCutterforhisencouragementandfeedbackthroughoutthecourseofthisproject.Hisconstantsupportanddirectionmadethisprojectapossibility.IamextremelyappreciativetoprofessorsJohnJurewitzandPierangeloDePaceforsharingtheirexpertiseintheirrespectivefieldswithme.IwouldalsoliketothankProfessorsGarySmith,EleanorBrown,andCharMillerfortheirindispensibleassistanceandenthusiasm.Finally,Icouldnotpossiblythankmyparentsenoughforalloftheirhelpovertheyearsandwiththisproject.

2

AbstractTraditionalutilityregulationexacerbatesenergyinefficiencybycreatinganincentivetoencourageoverconsumptionofenergy.Revenuedecouplingmechanismseliminatethisthroughputincentive,butcriticsarguethatdecoupledutilitieswillstillnotactivelypromoteenergyefficiency.ThispapermodelstheinterplaybetweenutilityDSMinvestmentandregulatorincentivesandstandards.Itexploresthehypothesisthatdecouplingmechanismsdecreaseresidentialelectricityconsumption,butonlythroughtheireffectonthelevel,directdemand‐reducingeffects,andindirectdemand‐reducingeffectsthroughpriceofDSMinvestment.Itthenempiricallyanalyzestheeffectsofdecouplingmechanismsonresidentialper‐customerelectricityuse.Thefindingssuggestthatdecouplingmechanismsdosignificantlydecreaseresidentialelectricityconsumptionoverall,butthatthissignificancedisappearswhentheeffectsofdecouplingmechanismsonretailelectricitypriceandDSMinvestmentareproperlyaccountedfor.Resultsindicatethatdecouplingmechanismsincreasethedirectandindirectdemand‐reducingeffectsofDSMinvestment.FindingsconcerningtherelationbetweendecouplingmechanismsandlevelofDSMspendingsuggestapositivecorrelation,buttheyareinconclusiveduetolimitedsamplesize.

3

TableofContents• Introduction_______________________________________________________________________4• LiteratureReview_________________________________________________________________7

o EnergyInefficiency________________________________________________________7o TraditionalRegulation____________________________________________________8o SuggestedSolutionsConcerningtheThroughputIncentive

andEnergyEfficiency____________________________________________________10o DecouplingMechanisms–Overview____________________________________12o ArgumentsfortheImplementationofDecouplingMechanisms_____14o ArgumentsAgainsttheImplementationofDecoupling

Mechanisms_______________________________________________________________16o BriefHistoryofDecouplingAdoptionintheU.S.______________________20o MeasuringtheEffectsofDSMandDecouplingonEnergyEfficiency_21

• ModelingDSMInvestmentIncentives__________________________________________23o Introduction______________________________________________________________23o PartA–TraditionalRegulation_________________________________________26o PartB–Decoupling______________________________________________________28o Conclusions_______________________________________________________________29

• TheoryandMethod_______________________________________________________________31o HypothesestobeTested_________________________________________________31o DataandMethods________________________________________________________33

• Results_____________________________________________________________________________38o OverallEffectsofDecouplingonConsumptionandPrice_____________38o EffectofDecouplingMechanismsonLevelofDSMSpending________40o EffectofDecouplingMechanismsontheDirectDemand‐Reducing

EffectsofDSMSpending_________________________________________________41o EffectofDecouplingMechanismsontheEfficacyofDSMInvestment

duetoitsIndirectEffectonConsumptionThroughPrice_____________42• RobustnessTests_________________________________________________________________43

o UseofanAnnualInstrumentalVariable________________________________43o Pre‐ExistingTrends______________________________________________________44o PotentialEndogeneityoftheDecouplingVariable____________________45

• Conclusions_______________________________________________________________________46• References________________________________________________________________________50• AppendixA________________________________________________________________________57• AppendixB________________________________________________________________________61

4

Introduction

Overconsumptionofenergyisthesourceoftwoofthelargestissuesfacing

today’sworld:globalclimatechange(Stern2008)andenergysecurity(Brookings

2011).Energydemandisincreasingworldwideandisprojectedtocontinue

increasing,atannualaverageratesofabout3.2%and1.1%foremergingand

matureeconomiesrespectively(AsifandMuneer2005).Asaresult,scientists

projectacontinuationofthetrendsofincreasingairpollution,especiallycarbon

dioxide(CO2)andotherGreenhouseGas(GHG)pollutionrelatedtoclimatechange,

andadecreasingstockofnon‐renewableenergysources(Smil2010).

Inefficientenergyuseisexacerbatingglobalclimatechange.Fossilfuel

combustionistheprimarycauseofthe80%increaseinCO2emissionsbetween

1970and2004(IPCC2007).Coal‐firedpowerplantsemitabout1tonofCO2for

everyMegawattofelectricityprovidedtocustomers(Hurleyetal.2008).Whilethe

exactcostsofglobalclimatechangearecurrentlyindebate,withnumerous

estimatesaround3‐5%ofgrossdomesticproduct(GDP)peryear,climatechangeis

arguablythelargestmarketfailurepresentintoday’sworld.Duetoenvironmental

andhealthexternalitiesassociatedwithGHGemissions,thecostsassociatedwith

failingtoadequatelymitigateclimatechangeconsiderablyexceedthemitigation

costs(Stern2006,AckermanandStanton2008).

Sincethemajorityoftheworld’senergycomesfromfinite,rapidlydepleting

sources(Smil2010),whicharelocatedinonlyafewspecificcountries,manypeople

worryaboutthesustainedattainabilityofenergy.TheUS,likemanyothercountries

includingChina,India,andUK,reliesheavilyonenergyimportationtofulfillits

growingdemand.USoilreservescouldsustainthecountry’sdemandforlessthan4

years(AsifandMuneer2005).Worldwide,fossilfuelsaregettingincreasingly

difficultandlesscost‐effectivetoobtainandconvertintousefulformsofenergy.

Inevitably,societywillneedtotransitiontoapost‐fossilfuelworld,butthereisalot

ofuncertaintyregardingthetiminganddetailsofthisprocess(Smil2010).

Whileproduction‐basedtechnologicaladvanceswillprobablybepartofthe

solution,increasedenergyefficiencyanddecreasedenergyconsumptionisessential

5

foreffectiveclimatechangemitigationandforimprovementstoenergysecurity.

Duetothescaleofnecessarycarbonstorage,itisunlikelythatpeoplecouldcapture

andstoreanadequatefractionofourcurrentcarbonemissionsthroughcarbon

sequestrationmethods(Bryce2010).Duringthecontinuedperiodofheavyreliance

onfossilfuels,therefore,asubstantialreductioninemissionsmustbearesultof

lowerdemandforenergyproduction.Furthermore,whiletherearemanypotential

sourcesofenergyinaninevitablepost‐fossilfuelenergyregime,theyarealllimited

inthepotentialusefulenergytheycouldprovidesociety.Expertsagreethatthe

futureenergyregimewillmostlikelyneedtocopewithlowerenergyandpower

densitiesbydecreasingenergyconsumption(Smil2010).Accordingtothe

EnvironmentalProtectionAgency(EPA)(2007),increasingenergyefficiencyin

residential,business,andindustrialsectorsisoneofthemostcost‐effectivewaysto

addresstheissuesofenergysecurity,airpollution,andclimatechange.

Therecurrentlyexistmanycost‐effectiveandnegativecostopportunitiesin

energyefficiency,but,forthemostpart,peopledonotseemtobecapitalizingon

theseopportunities(Gillingham,Newell,andPalmer2009).Forexample,many

consumershaveyettoreplacetheirinefficientincandescentlightbulbswith

efficientcompact‐fluorescentlightbulbs(CFL’s)orlight‐emittingdiode(LED)

fixtures,eventhoughdoingsowouldsavetheconsumermoneyanddecrease

energyuse(McKinsey2007;EnergyStar2012).Economistshavesuggesteda

multitudeofreasonsforthisabundanceofseeminglyirrationaldecision‐making.

Themainimplicationofthesestudiesisthatconsumersneedexternalassistancein

makingefficientenergydecisions.

Utilitypromotionofenergyefficiencyisvaluable.Utilitiespromoteenergy

efficiencythroughvariousdemand‐side‐management(DSM)programsthataimto

alterthelevelandtimingofconsumerenergyconsumption.Theseprogramsinclude

assistingconsumerspracticallyandfinanciallyininstallingenergyefficient

technologies,educatingconsumersaboutenergyconsumptionmanagement,and

institutingpricingschemesintendedtotransformconsumerincentives.DSM

programsexistforindustrial,commercial,andresidentialcustomers(Loughranand

6

Kulick2004).Gillingham,NewellandPalmer(2006)foundthatutility‐based

demand‐side‐management(DSM)programsproducedgreaterenergyreductions

thangovernmentorthird‐partyDSMefforts.TrainandStebel(1987)foundthat

peopleareskepticalthatspecificsupposedlyenergy‐efficientdeviceswillbecost‐

effectiveandthatthisskepticismisreducedwhenutilitiesofferrebates.

Undertraditionalenergyutilityregulation,however,utilitiesareincentivized

toencourageinefficientenergyuse.Thisphenomenonisknownasthethroughput

incentiveofelectricandnaturalgasutilities.Asitisasubstantialbarriertoutility

promotionofenergyefficiency,itisimportanttoaddressthisincentivestructure.

Onemechanismthatcanbeimplementedtoaddressthethroughput

incentiveissueandtherebyencourageenergyefficiencyisdecoupling,inwhich

revenuesarepartiallyorfullydetachedfromandunaffectedbysalesquantities.

Whileadecouplingmechanismeliminatesautility’saversiontodemandside

management(DSM),thereissomedebateoverwhetherthesemechanisms

encourageutilitiestoactivelyinvestinDSMandtopromoteenergyefficiency

(NARUC2007).

Thispaperaimstoaddressthequestionofwhetherdecouplingmechanisms

promoteenergy‐efficientpracticesandcost‐effectivechangesinenergy

consumptioninpracticethroughanalysisofhistoricaltrendsinenergy

consumption.Itbeginsbyreviewingtheexistingliteratureonthesubjectand

performingeconomicmodelingtoexploretheinterplaybetweenutilityDSM

investmentandregulatorincentivesandstandards.Econometrictechniquesare

thenutilizedtoanalyzetheempiricalrelationbetweendecouplingmechanismsand

residentialelectricityconsumption.Finally,thispaperaddressessomepotential

concernswiththeunderlyingassumptionsmadeinthisanalysisandpresentssome

robustnesstestsontheresults.

7

LiteratureReview

EnergyInefficiency

HeightenedawarenessabouttheharmfuleffectsofGHGemissionsand

widespreadconcernsaboutenergysecurityhavesparkedincreasedresearchon

energyefficiencyoverthepastfewdecades.Numerousacademicshaveattempted

touncoverthestructuralandbehavioralfactorsthataffectenergyconsumptionand

energyefficiencyinvestment(GlaeserandKahn2008,Baumann2008).Othershave

focusedonunderstandingthereasonsthatsomepeoplesupportenergyefficiency

measuresandchoosetoparticipateinenergyefficiencyprogramswhileothersdo

not(Musti,Kortum,andKockelman2011,KotchenandMoore2007,CostaandKahn

2010).

Despiteincreasedawarenessabouttheneedforenergyefficiencyand

improvedunderstandingofenergyefficientpractices,homeownersarenot

investinginenergyefficiencyasmuchastheyshould(Brennan2009b).Historically,

consumershaveactedasiftheyhadabout25%discountrateintheirchoicesof

investinginconservation,whichissubstantiallyhigherthanmostestimatesofthe

prevailingsocialdiscountrate(Hausman1979).Galluppollsthroughoutthepast

decadeshowthatAmericansbelievethatconservation,asopposedtoincreased

fossilfuelproduction,isthekeytoaddressingthenation’senergyproblems(Saad

2011).Brennan(2009b)attributesthisdiscrepancybetweenbeliefsandactionsto

consumerchoicefailure.Gillingham,Newell,andPalmer(2009)identifyheuristic

decision‐makingandboundedrationalityaspotentialcausesofconsumerchoice

failureregardingenergyefficiency.Inotherwords,decisionsconsumersmakeabout

energyefficiencyareoftenirrationalduetolackofavailableinformation,human

cognitivelimitations,thecognitiveburdenofdecision‐makingandthetimeittakes

tomakedecisions.Gillingham,Newell,andPalmer(2009)alsorecognizeprospect

theoryasapotentialbehavioralfailure.Prospecttheorystatesthatpeopleare

hesitanttochangetheirbehavior,suchasinvestinginenergyefficiencyiftheyare

unsureofthepotentialcostsandbenefitsbecausetheyaredisproportionately

aversetopotentiallossesrelativetopotentialgains.Gillingham,Newell,andPalmer

8

(2009)foundthatconsumersoftenignoreenergyefficiencyinpurchasingdecisions

becausetheycannotdistinguishbetweenproductsthataremoreenergyefficient

andproductsthatarelessenergyefficientsinceallproductsarebeingadvertisedas

beingenergyefficient.

Anothersourceofinefficienciesinenergyusestemsfromaconceptknownas

theprincipal‐agentproblem.Peoplewhouseenergyinabuildingyetdonotpay

utilitybills,suchasmanyrenters,havenoincentivetominimizetheirenergyuseor

toinvestinenergyefficiency.Theresidentsmakethedecisionabouthowmuch

energytoconsumeandhowmuchmoneytoinvestinfutureenergysavings,but

theydonotexperienceanyoftheprivatecostsoftheirchosenactions.Asaresult,

theytendtoover‐consumeandunder‐investcomparedtotheefficientsetofactions

(Gillingham,Newell,&Palmer2009).IncreasedutilityDSMprogramsmay

somewhatincreasetheenergyefficiencyofsuchresidentsduetoenvironmental

consciousness,butneitherDSMprogramsnorutilitypricestructurestacklethecore

inefficiencyoftheprincipal‐agentissue.

TraditionalRegulation

Undertraditionalregulation,theratesofanelectricutilityareestablishedby

anexternalpartyduringageneralratecase,whichusuallyoccurevery3‐6years,or,

insomestatesonlyattherequestoftheutility(OhioPUCO2011,WisconsinPSC

2001).

Inaratecase,thethirdpartyexaminesalloftheutility’sexpensesduringa

pastorfuturetestyear,hearstestimoniesfromallinterestedparties,and

determinesarevenuerequirement,ortherevenuenecessarytocovertheutility’s

expensesandtaxesandtogiveafairrateofreturnoninvestment(Lazer,Weston,

andShirley2011,OhioPUCO2011,WisconsinPSC2001).Fixedcostsandcosts

associatedwithdepreciationareincludedinthedeterminationofutilityexpenses

(EPA2007).Thethirdpartyregulatoroftenholdsformalhearingtohear

testimoniesofallinterestedparties,includingutilityrepresentatives,customers,

andinvestors.Inmanystates,theutilitiescanapplyforarehearingiftheyare

9

dissatisfiedwiththeregulator’sdecision.Insomeextremecases,utilitiescaneven

bringratecasestothestatesupremecourt(OhioPUCO2011).

Oncearevenuerequirementisfinalized,theallowedrevenueisthendivided

bythenumberofunitsofelectricitysold(orexpectedtobesold)duringthetest

periodinordertodeterminetheutility’saverageelectricityrateforthenextperiod.

Theaverageratesareoftencomputedseparatelyforeachcustomerclass2.These

averageratesareessentiallyconstantuntiltheutilitynextundergoesaratecase.

(Lazer,Weston,andShirley2011)

Onenotableaspectofanelectricutility’scoststructureisthattheutility

incursvirtuallynonetmarginalcostsperunitofelectricityproduced.Thepotential

marginalcostsofautilitywouldbelabor,distribution,andfuelcosts.However,

laboranddistributioncostsvarywithnumberandlocationofcustomersandare

essentiallynotaffectedbyper‐customerenergyuse.Theonlycostsrelatedto

energyusearefuelcosts,butthesearepass­throughs,orcoststhataredirectly

transferredtocustomers.Variationinfuelcostsusuallyappearsasamonthlytariff

orcreditoncustomerbills,oftenreferredtoasfueladjustmentclausesorpowercost

adjustments(EnerStar2012).Eachmonth,theutilityrecalculatesthefuelcost

associatedwitheachcustomerbasedonthemonth’sfuelpricesandthecustomer’s

electricityconsumptionandchargesthecustomerexactlythatamount.

Anothernotableaspectofatraditionallyregulatedutility’sratestructureis

thatifactualrevenueduringthetimeperiodbetweenratecasesdiffersfrom

expectedrevenueforanyreasonnotcoveredbyadjustmentclauses,theutilitygains

aprofitorsuffersaloss.Utilitieshavesubstantialfixedcoststhatrequireadequate

revenuetofinance.Theseincludesunkcapitalcostsassociatedwithincreasing

generationcapacity,buildingtransmissionlines,andDSMinvestments.Ifrevenue

doesnotreachthethresholdvalueofcoveringfixedcosts,theutilitysuffersaloss.

(CSI2008)

2Manyutilitiesvaryrateschedulesbycustomerclass.Customerclassesgroupcustomerswithsimilardemandcharacteristicstogether(ex‐small,residentialcustomers).

10

Underthetraditionalstructureofregulation,utilitiesaredeterredfrom

investinginenergyefficiencyandevenhaveincentivestoactivelylobbyagainst

energyefficiencyinfavorofdecreasingefficiency.Brennan(2008)arguesthat

utilitiesonlyhaveincentivestopromoteenergyefficiencyandotherdemand‐

reductionsifpricecapscausepricestofallbelowmarginalcostsbutthatwhen

pricesaregreaterthanmarginalcosts,utilitieshaveincentivestoincrease

throughputi.e.electricityusage.Inalmostallinstances,regulatedretailpricesdo

surmountmarginalcosts(EIA1997).Thus,intheperiodsbetweenratecases,the

utilityhasanincentivetoincreaserevenuesabovewhathadbeenexpectedbythe

ratessetbyincreasingelectricitysold.Thisisknownasathroughputincentiveand

isamajorbarriertotheutilitypromotionofenergyefficiency(EPA2007).

Increasedenergyefficiencydecreasesthroughput,sotheutilitieshaveanincentive

toactivelyopposeenergyefficiencyprograms.Ifutilitiesneverthelessdocreate

programstopromoteenergyefficiency,notonlydotheyloserevenuesfrom

electricitysales,buttheyalsohavetocoverspendingonprogramimplementation.

Thiscanleadtolosses(knownaslostmarginrecovery).AninvestigationbyIdaho

PowerCompanyfoundthat,inpractice,revenuelossfromenergyefficiency

programsdoesindeeddecreaseautility’srecoveryoffixedcosts,whichproduces

barrierstoutilityinvestmentindemand‐reducingenergyefficiencyprograms(EPA

2007).

SuggestedSolutionsConcerningtheThroughputIncentiveandEnergyEfficiency

Thissectionbrieflydiscussesthemeritsanddrawbacksoffourprominent

proposalsforaddressingthethroughputincentiveproblem:governmentmonetary

incentives,othertypesofthirdpartypromotionmechanisms,straightfixed‐variable

ratedesigns,andreal‐timepricing.

Oneproposalisthatthegovernmentshouldprovidemonetaryincentivesto

utilitiesforinvestinginenergyefficiency.Whilethiswouldassistincoveringthe

costsofutility‐sponsoredDSMprograms,monetaryincentivesalonewould

probablynotbeeffective,astheincentivemustbesufficienttocoverthecostsof

11

programimplementationandcompensateforlostmargins.Suchincentives,ifnot

designedwithadequatecare,mayalsocauseutilitiestoinvestalotofmoneyin

energyefficiencyprogramsthatarenotparticularlycost‐effective(Lazer,Weston,

andShirley2011).

Moregenerally,manypromotersofenergyefficiencyarguethatathirdparty,

andnotutilities,shouldberesponsibleforencouragingenergyefficiency.They

contendthatutilitiesareinnatelynotinthebestpositiontopromoteenergy

efficiency.Utilitieshaveanadvantageoverthird‐partyenergy‐efficiencyproviders,

however,astheyhavemorecompleteinformationandrecordsaboutenergyuse.

Moreimportantly,athird‐partyenergy‐efficiencyproviderwouldbemosteffective

iftheutility’sthroughputincentivecouldbeeliminated(bydecoupling,for

instance),asthiswoulderadicatelobbyingagainstthethirdpartyandwould

increasethelikelihoodofrelevantinformationdisclosure.InVermontandOregon,

third‐partyenergy‐efficiencyprovidersfounddecouplingtobeahelpfuladditionto

theirownencouragementofenergyefficiency(Lazer,Weston,andShirley2011).

Anotherproposalisimplementationofastraightfixed‐variable(SFV)rate

design(Lesh2009).UnderSFV,thereisalargemandatory,non‐volume‐based

chargeforallcustomersthatcoversallormostoftheutility’sfixedcosts,rendering

variablecostsclosertothetruemarginalcostofproducingelectricity.This

decreasesandcaneveneliminatethethroughputincentive.However,SFVrate

designssimultaneouslydiscourageenergyefficiencybydecreasingtheincentivefor

consumerstoreducetheirenergyusethroughlowerusage(marginal)costs.In

addition,SFVisregressive(EPA2007).Sincelow‐incomehouseholdstendto

consumelessenergy,alargerfixedchargewilldisproportionatelyburdenlow‐

incomehouseholds.Boonin(2009)promotesaSFVdesigncombinedwitha

Revenue‐NeutralEnergyEfficiencyFeebate(REEF),whichtheoreticallycounteracts

theregressivenatureoftheSFVdesignandreestablishescustomerincentivesto

decreaseenergyuse.Unfortunately,itmaybedifficultandadministrativelycostlyto

constructafeebatesystemthataccuratelydeterminesindividualenergyefficiency

andappropriatelycounteractstheregressivenatureofSFV.

12

Realtimepricing(RTP)ofelectricityhasbeensuggestedasamethodof

increasingenergyefficiency,limitingmarketpowerofutilities,andreducing

electricitydemandvariance(HollandandMansur2008).WithRTP,utilitiesset

pricesasafunctionofactualdemandatthegiventime.Accordingtoeconomic

theory,thiscausesadisincentivetouseelectricityduringhoursofpeakdemand,

which,duetovariabilityinefficiencyofgenerationequipment,shouldhavea

disproportionatepositiveeffectontheenvironment(HollandandMansur2008).In

contrast,Ito(2010)foundthatactualhousehold‐electricity‐usedatasuggeststhat

consumersrespondtoaveragepricesinsteadofmarginalprices,whichlowers

projectedestimatesofdecreasesinpeakdemandduetoRTP.

DecouplingMechanisms­Overview

Decouplingmechanismsaddressthethroughputincentivebyfullyor

partiallydetachingutilityrevenuesfromsales.Underadecouplingmechanism,the

allowedrevenuesaresetduringageneralratecase,andratesareadjustedduring

regular,periodictrue­upsthatkeeplongrunrevenueequaltoallowedrevenueby

changingprices.Iftheutilitymakeslessrevenuethanexpectedinagivenperiod,

thenratesareincreasedorasurchargeisaddedinthefollowingperiod.Conversely,

ifthereisasurplusinrevenue,themoneyisrefundedtothecustomersintheform

oflowerrates,lumpsumtariffs,orotherreconciliationmechanisms(EPA2007).

Notethatnoreassessmentsofcostsorallowedrevenueareperformedduringtrue‐

ups(Lazer,Weston,andShirley2011).

Thereareavarietyofdecouplingmechanisms.Onedifferencebetweenthem

istheeffectofincreasingtheutility’scustomerbase.Understraightrevenue

decoupling,anincreaseincustomersdoesnotchangeautility’sallowedrevenues

and,asaresult,tendstohurtratherthanhelptheutility.Revenuepercustomer

(RPC)decoupling,however,fixesallowedrevenuepercustomer.Thusallowed

revenueduringanygiventimeperiodisdeterminedbythefollowingformula:

AllowedRevenue=(AllowedRevenuesperCustomer)*(#ofCustomers)

13

RPCdecouplingcoulddiscourageutilitiesfromincreasingtheircustomerbaseif

newcustomerswerehigherenergyusers.SomeRPCdecouplingmechanisms

addressthisissuebyincludingdifferentallowedRPC’sfornewandexisting

customers.(Lazer,Weston,andShirley2011)

Decouplingmechanismsalsodifferwithregardtothefrequencyoftrue‐ups.

Utilitieswithaccrualdecouplingmechanismsonlyundergoannualtrue‐ups,while

thosewithcurrentdecouplingexperiencetrue‐upseverybillingperiod.True‐upsfor

otherdecoupledutilitiesmayoccuronaquarter‐orsemi‐annualbasis(Lazer,

Weston,andShirley2011).

Anotherwaydecouplingschemesdifferisbythereasonsallowedfor

revenueadjustmentsduringtrue‐ups.Afullydecoupledutilityadjustsforany

deviationofactualrevenuesfromallowedrevenues,regardlessofthecauseofthese

deviations.Autilitywithpartialdecouplingadjustsforacertain,fixedpercentageof

thedepartureofactualrevenuefromallowedrevenuesduringtrue‐ups.Partial

decouplingmechanismsalsodonotdiscriminatebasedoncauseofrevenue

deviations.Partialdecouplingreducesthethroughputincentivewithoutcompletely

eliminatingit.Duringatrue‐upforautilitywithlimiteddecoupling,disparities

betweenactualrevenuesandallowedrevenuesmayormaynotleadtoadjustments

dependingonthecauseofthedisparities.Onecommonexampleoflimited

decouplingisutilitiesthathaveonlylost­marginmechanisms,whereadjustmentsfor

revenuedecreasesareonlymadeifthedecreasescanbeexplicitlyidentifiedasthe

resultofDSMprograms.Ifexecutedcorrectly,limiteddecouplingeliminatesthe

throughputdisincentiveassociatedwithenergyefficiencyprogramswithout

eliminatingutilityresponsibilitytoprovidecoverageandassumeriskduetoother

causes,suchaseconomicfluctuations.Thelargestdrawbackoflimiteddecouplingis

itsrelativelyhighadministrativeandmonitoringcostscomparedtootherformsof

decoupling(Lazer,Weston,andShirley2011).

Otherdecouplingmechanismsincludeattritionadjustments,inflationminus

productivityadjustments,andKfactoradjustments.Attritionadjustmentsare

14

essentiallyannualabbreviatedratecaseswhenallowedrevenuesarereassessed

andalteredslightlyinresponsetoknownandmeasurablechangesinautility’s

costs.Thismechanismdecreasestheneedforfrequentfullratecasesifautility

doesnotundergoalargestructuralchange.Inflationminusproductivityadjustments

alterallowedrevenuesinresponsetoshiftsininflationorproductivity.This

mechanism,alongwithregulartrue‐ups,alsodecreasestheneedforfrequentfull

ratecases.Anothermechanismtochangeallowedrevenuesbetweenratecasesis

knownasK­factoradjustment.AKfactoradjustmentisapredetermined,usually

annual,rateincreaseordecreasebetweenratecases.Forexample,autilitymight

receivea1%annualincreaseinallowedrevenuesbetweenratecases.TheKfactor

itselfcanbedesignedtochangeovertimeifgrowthisnotpredictedtobelinear,or

canbeappliedtoallowedrevenuespercustomerinsteadoftotalallowedrevenues.

Althoughallthreeofthesemechanismsaimtoreducetheneedforfrequentrate

cases,occasionalratecasesarestillnecessarywithallthreeofthesedecoupling

mechanisms.

ArgumentsfortheImplementationofDecouplingMechanisms

Decouplingmechanismsareattractiveforseveralreasons.Theyare

beneficialtotheenvironment(GraniereandCooley1994)andaremorepolitically

feasiblethanmanymechanismstopromoteenergyefficiency.Theyencourage

energyefficiencybyeliminatingapoliticalbarriertoenergyefficiencypromotion,

removinganeconomicbarriertoutilityenergyefficiencyinvestment,andpartially

shiftingtheresponsibilityofmakingenergyefficiencydecisionsfromconsumersto

suppliers.Decouplingmechanismsarepoliticallyfeasible,astheyinvolverelatively

lowadministrativecostsandarestructuredinawaythatfostersutilitysupport.

Fromanenergy‐efficiencystandpoint,thedecoupling‐causedeliminationor

reductionintheutilitythroughputincentiveisessentialforutilityorthird‐party

promotionenergyefficiency.Forone,theremovalofathroughputdisincentive

decreasesoreliminatesutilities’politicaloppositiontopublicenergyefficiency

programsandgovernmentmeasures(Brennan2009a).Moreover,thedecoupling

15

incentivestructurealsorendersutilitiesmorewillingtotakeactionsthemselves

thatencourageenergyefficiency.Forinstance,underdecouplingmechanisms,

utilitiesaremorewillingtoinstituteratedesignsthatencourageenergyefficiency

(Lazer,Weston,andShirley2011).

Brennan(2008,2009a)arguesthatdecouplingisfundamentallyabout

transferringresponsibilityforenergyefficiencydecisionsfromconsumersto

suppliers.Asdiscussedabove,manyconsumersdonothavethetoolstomake

rationaldecisionsregardingenergyefficiency,bututilityandgovernmentDSM

programscanreducesomeofthecognitivebarriersbyusingconsumerfinancial

incentivestopromotemoreefficientlevelsofenergyconsumptionandbysupplying

informationaboutspecificenergyefficiencymeasuresandtheircalculatedeffects.

Fullandpartialdecouplingmechanismstendtobepoliticallyfeasible,

partiallybecausetheyrequirerelativelylowadministrativecost.Unlikesome

alternativeenergy‐efficiencypromotionmethods,decouplingmechanismsdonot

requireregulatorstoevaluatetheefficacyofenergyefficiencyprograms.

Furthermore,decouplingmechanismsreducetheneedforfrequentratecasesand

avoidtheassociatedcosts(EPA2007).Theonlypubliccostofadecoupling

mechanismisthetimeittakesstateemployeestolearnaboutthenewsystemand

establishproperregulations(CSI2008).

Anotherreasonforthepoliticalfeasibilityofdecouplingmechanismsistheir

tendencytoattainutilitysupport.Utilitiestendtosupportdecouplingbecauseit

stabilizestheirrevenuesandthusreducestheirrisksassociatedwithrevenue

volatility.Thisdecreasestheircostsofcapital,bothintermsofinterestratesand

equityreturnsrequired(Lazer,Weston,andShirley2011).Inaddition,utilities

oftenpreferdecouplingtotraditionalregulationbecausetheybelievethatthe

traditionalratecasesdonottakeallreal‐worldfactorsintoaccountintheir

predictionsandcalculationsofraterequirements,leadingtoearningsthatarelower

thanauthorizedearningsafewyearsaftertheratecase(Costello2011).

16

ArgumentsAgainsttheImplementationofDecouplingMechanisms

Opponentsofdecouplinghavevoicedseveralconcernsaboutdecoupling

mechanisms.Theyareskepticalabouttheefficacyofdecouplingmechanismsin

promotingenergyefficiency.Thissectiondiscussesthecurrentdebatessurrounding

thesepotentialdrawbacksofdecouplingmechanismsandincludespolicy

suggestionstomitigatethesedrawbacks,whenapplicable.

Onecategoryofcritiquesofdecouplingmechanismsisthattheyfailto

provideanincentiveforenergyefficiencyinvestment.Manyexpertsarguethat,

althoughdecouplingmechanismswilleliminatetheincentivetoactivelyincrease

energyconsumption,utilitieswillstillfailtopursueenergyefficiencybecause

supply‐sideinvestmentswillcontinuetobemoreattractivetothemthanDSM

investments(Kihm2009).Historically,utilitieshavegainedfromsupply‐side

investmentsbecausetraditionalratesettinghasdefactoallowedmorethana

normalrateofreturnontheseinvestments.IfinvestmentsinDSMdonotprovide

comparablereturns,forinstancebecausetheycausedecreasedfuturedemandfor

utilityservice,itwouldpresumablybemoredifficultforautilitytoattractfunding

forDSMinvestmentsthanforsupply‐sideinvestments(EPA2007).

Kihm(2009)arguesthatevenifDSMperformanceincentivesforutilities

werelargeenoughtoprovidecomparableratesofreturnforDSMandsupply‐side

investments,manyutilitieswouldcontinuetoprefersupply‐sideinvestments

becausetheyarelargerinscale.Utilitieshavehistoricallypreferredlargerscale

investmentstosmallerscaleinvestments,aconceptknownastheAverch‐Johnson

(A‐J)hypothesis.Thispreferenceoftenremainsevenwithchoicesinvolvingalarge‐

scaleinvestmentwithalowerrateofreturnthananalternativesmall‐scale

investmentsuchasDSMinvestments.Kihmarguesthatdecouplingonlyworksfor

utilitiesnotsubjecttotheA‐Jeffect,orutilitieswithallowedratesofreturncloseto

thecostofcapital,anditisunlikelythatmostutilitiesfallintothiscategory.Finally,

utilitymanagersandregulatorshavehistoricallybackedexpansionsinfuture

electricitygenerationcapacitybecausetheriskstotheutilityoffutureinadequate

electricitysuppliesarehigherthantheriskofhavingexcesscapacity.(Kihm2009)

17

Ontheotherhand,thereareseveralreasonsthatadecoupledutilitymay

preferDSMinvestmentstosupply‐sideinvestments,evenwithoutthepresenceof

performanceincentives.Fromacustomerservicestandpoint,DSMinvestments

improvethepublic’sviewoftheutility,enablemorepositiveutility‐customer

interactionthansupply‐sideinvestments,andgeneratelesspoliticaloppositionthan

supply‐sideinvestments.Supply‐sideinvestmentsusuallyinvolvebuilding

pollution‐emittingpowerplantsnearspecificcommunities,whomayresistthe

projectorrequestcompensationforitsundesirableeffects(Bloomquist1974).

Conversely,DSMinvestmentsdonotseverelyadverselyimpactaconcentrated

groupofpeople;theyusuallysubstantiallyhelpaconcentratednumberofpeople–

thecustomerswhochoosetoparticipateintheDSMprograms–ataminutecostto

many(ECW1997).Asaresult,DSMinvestmentswilloftengeneratelesspolitical

oppositionthansupply‐sideinvestments.Theutilitycanevencapitalizeonthis

distributionofcostsandbenefitsandimproveitspublicratingsbypublicizingthe

savingsofthecustomerswhobenefitthemost.

Moreover,withtheuncertaintyofthefutureenergyregime,DSM

investmentsmayactuallyattractinvestmentmorethansupply‐sideinvestments

duetotherelativeeffectsonriskandenergysecurity.Reducedenergyconsumption

anddecreaseddependenceonenergyfromout‐of‐stateoroverseasourcesrenders

autilitylesssusceptibletounexpectedfuturechanges.Theseincludeunanticipated

capitalcostincreases,increasesinenergyregulationstringency,transportationcost

increases,andpoliticallyoreconomically‐inducedfossilfuelsupplyshortages.

(Hurleyetal.2008,EPA2007)

Opponentsofdecouplingarealsoskepticalofitsefficacyinpromoting

energyefficiencybecauseitcreatesadisincentiveforconsumerenergyefficiency

investments.Theyarguethatdecouplingmechanismsdiscouragecustomersto

decreaseelectricityusebecausedoingsowillcauseafutureincreaseinelectricity

rates(GraniereandCooley1994).However,eachindividual’seffectonthepriceis

trivial.Thus,fortheindividual,decreasedconsumptionoutweighsanyeffectoftheir

ownbehavioronrates(NARUC2007).Themonetarybenefitsofreducing

18

consumptionareveryconcentrated,whilethecorrespondingcostsfromthe

associatedrateincreasearewidespread.Asaresult,eachcustomernotonlyhasan

incentivetoreduceconsumption,butalsohasanincentivetoreduceconsumption

evenmoresubstantiallythantheothercustomers(ECW1997).

Finally,decouplingmechanismshavelimitedefficacybecausetheyfailto

addresssomeimportantcausesofenergyinefficiency.Theprincipal‐agentproblem

stillexistsinthepresenceofdecouplingmechanismssinceconsumerswhodonot

payfortheirenergyusestilllackincentivestodecreasethisuseunderdecoupling

(Gillingham,Newell,andPalmer2009).Similarly,theadoptionofdecouplingdoes

notsolveconsumerchoicefailuresandmanyconsumerswhowouldbenefitfrom

energyefficiencychangeswillnotmakethosechanges,whetherornottheirutility

possessesadecouplingmechanism(Brennan2008).

Inaddition,opponentsofdecouplingworrythatthemechanismwillincrease

thevolatilityofelectricityratesandcausefrequent,significantrateincreases(Lesh

2009).GraniereandCooley(1994)didanumericalstudyonthisissueand

concludedthat,“ratepayersmighthavetodealwithsubstantialvolatility.”Onthe

otherhand,Eto,Stoft,andBelden(1994)foundthatERAM,California’decoupling

mechanism,reducedratevolatilityandhadanegligibleeffectonratelevels.

Similarly,Lesh(2009)foundthatdecouplingadjustmentstendtobeverysmall

relativetoresidentialretailrates,withnosignificantupwardordownwardtrendin

rates.Comparedtootheradjustmentfactorsonbills,suchasfuelandpurchased‐

poweradjustmentclauses,volume‐balancingdecouplingadjustmentstendtoalter

rateslessdrastically(Lazer,Weston,andShirley2011).

Moreover,evenifelectricityratesdoincrease,thereisabeneficialaspectto

thisinthatitadjustsforthenegativeexternalitiesontheenvironment.Michelfelder

(1993)foundthatelectricityratesareconsistentlybelowmarginalsocialcostsof

producingelectricity.Thus,increasingelectricitypricesactuallybringsthemcloser

tothelong‐runmarginalsocialcostofelectricityproduction,whichultimately

benefitsratepayers(GraniereandCooley1994).Furthermore,Bholeetal.(2011)

foundsignificantpositivecorrelationbetweenelectricitypricesandstateenergy

19

efficiencyexpenditures,especiallyaboveathresholdpricelevel.Heproposedthat

higherelectricitypricesincreasethecost‐effectivenessofDSMprograms.In

addition,Linn(2008)contendsthathigherenergypricesincreasetheprobability

thatplantsadoptnewenergyefficiencytechnologies,especiallyfornewplants,

althoughKahnandMansur(2011)suggestthatnewplantsinenergy‐intensive

industrieswouldfactorelectricitypricesintotheirlocationdecisionandspecifically

chooseageographicallocationwhereelectricitypriceswerelower.TheCenterfor

StateInnovation(2008)arguesthat,whiledecouplingmechanismsmayincrease

energycostsforsomecustomersintheshortrun,thediminishedneedfornew

powerplantsduetoreducedenergydemandwillultimatelylowerratestoall

customers.

Opponentsofdecouplingareparticularlyconcernedaboutdecoupling

mechanismsshiftingweatherandeconomicrisksfromutilitiestoconsumersinthe

formofratevolatility(Meneken2007).Theyarguethatutilitiesbearnoneofthe

costsassociatedwithextremeweatheroreconomicrecessionsunderdecoupling.

TheypointtoMaine’sbriefexperiencewithadecouplingmechanismasanexample.

ShortlyafterCentralMainePoweradoptedafulldecouplingmechanismin1991,

Mainefellintoarecession.Requiredrevenuesweretoohigh,astheyhadbeenset

basedonatrialperiodduringathrivingeconomy.Thus,pricesincreaseddrastically

asenergyconsumptionfell,andcustomersborethefullburdenoftherecession

whileCentralMainePoweremergedunscathed.Thedecouplingmechanismwas

ultimatelyterminatedduetoanimositytowardsthemechanismproducedbythis

incident.Manycustomersbelievedthetruepurposeofthemechanismwastoshift

riskfromutilitiestocustomers.(EPA2007)

AreportfromtheEPA(2007)conteststhatthisriskshiftisinherentto

decouplingandarguesthattheexistenceandmagnitudeofthisriskshiftvarieswith

thedecouplingdesign.Lazer,Weston,andShirley(2011)completelydismissthe

argumentthatdecouplingshiftsweatherriskstoconsumersunderfullandpartial

decouplingmechanismsandarguethatdecouplingdecreasestheriskforutilities

andconsumers.Ifanextremeweathereventcausespeopletoconsumemore

20

electricity,decouplingcausesratestodecrease,makingtheenergyconsumption

moreaffordableforconsumers.Lazer,Weston,andShirley(2011)doadmitthat

decouplingexacerbateseffectsofrecessions,buttheyarguethatgeneralratecases

dosoaswell.Theysuggestcombiningdecouplingwithacaponrateincreasesto

preventsituationsliketheonethatoccurredinMaine.

Someopponentsofdecouplingmayarguethatdecouplingisregressive.Low‐

incomecustomerstendtohavelesselectricityreductionpotentialthanhigher‐

incomeusers.Whiletheydonotreapmanyofthebenefitsofreducedelectricityuse

duetoenergyefficiencyprograms,theyarestillsubjecttothehigherratescaused

bytheoverallreduceddemand(Lesh2009).Toaddressthepotentialconcernof

increasedshortruncostsforlow‐incomecustomers,mostdecoupledutilitiesoffer

lowerratestolow‐incomecustomersanddofundingandsurchargingonacustomer

class,orrateschedule,basis(Lesh2009).

Otherconcernsregardingdecouplingincludethelackofincentiveforutilities

torestoreserviceafterastormandincreasedregulatorycostsduetofrequenttrue‐

ups.Thesearebothfoundedyetnotparticularlycompellingconcerns.Thefirstissue

caneasilybeaddressedthroughadditionalrequirementsorincentivesinthedesign

ofthedecouplingmechanism(Lazer,Weston,andShirley2011).Periodictrue‐ups

doproduceadministrativeandregulatorycosts,buttrue‐upsareless

administrativelycostlythanratecases.Theincreasedregulatorycostduetotrue‐

upsisoffsettosomedegreebyareductionintheneedforfrequentratecasesand

theirassociatedcosts(EPA2007).

BriefHistoryofDecouplingAdoptionintheU.S.

DecouplingmechanismshavebeenoperationalintheUnitedStatessince

1978,whenPacificGas&Electricdecouplednaturalgassales(AGA2007).Four

yearslater,PG&EalsoimplementedadecouplingmechanismknownastheElectric

RateAdjustmentMechanism(ERAM),forelectricitysales(McCarthy2009).Other

Californianutilitiesadoptedsimilarmechanismsshortlythereafter.Intheearly

1990’s,Maine,NewYork,andWashingtonalsoadopteddecouplingmechanisms,

21

althoughMainediscontinueditsmechanismafterabrieftrialperiod(Etoetal.1994,

EPA2007).BySeptember2007,atleastonenaturalgasorelectricutilityin16

stateshadimplementedadecouplingmechanism,anddecouplingmechanismswere

pendingin12morestates(EPA2007).AsofJune2011,utilitiesinabout28states

hadadoptedadecouplingmechanism,anddecouplingmechanismimplementation

decisionswerependingin12morestates(Lazer,Weston,andShirley2011,ACEEE

2011).

MeasuringtheEffectsofDSMandDecouplingonEnergyEfficiency

Thereexistsmuchdebateintheliteratureabouthowtomeasuretheefficacy

ofDSMprograms.Thetrueprogrameffectscannotbemeasuredbysolelyanalyzing

theobservedchangeinenergyconsumptionofcustomerswhoparticipateinthe

DSMprograms.Ononehand,JoskowandMarron(1992)arguethatsavingsofutility

DSMprogramstendtobeoverstatedduetooverlyoptimisticequipmentlifetimes

andtheexistenceoffreeriders,orprogramparticipantswhowouldhaveinvestedin

energyefficiencyevenwithoututilityinvolvement.Ontheotherhand,manyDSM

programanalystsworryabouttheopposite:understatingthebenefitsofDSM

programs.Programparticipantsmaydiscussenergyefficientbehaviorand

technologywithnon‐participants,whoconsequentlydecreasetheirenergyuse

withoutanyadditionaldirectutilityinvestment.Forexample,acustomermight

experiencehugereductionsinheatingandcoolingcostsafterusingautilityrebate

toaddatticinsulationinherhome,raveaboutthisexperiencetoafriendwhoisnot

eligiblefortherebate,andconvincethefriendtoinstallatticinsulationaswell.

Thesenon‐programparticipantsarecommonlyreferredtoasfreedrivers

(Gillinghametal.2009).Duemostlikelytomeasurementdifferences,Parformak

andLave(1996)foundthat99%ofutility‐reportedestimatesofsavingsfromDSM

arefairlyaccurateifonecontrolsforprice,weather,&economicactivity,while

LoughranandKulick(2004)foundthatutilityestimatesofelectricitysavingsdueto

DSMinvestmentaremuchlargerthanthetruesavings.Gowever,Auffhammeretal.

22

(2008)pointedoutflawsinLoughranandKulick’smethods.Thus,thereisno

consensusonthereliabilityofutilitysavingsestimatesfromDSM.

Thecurrentliteraturelacksexpostempiricalanalysisthatusesrecentdatato

measuretheeffectsofdecouplingmechanismsonenergydemand.Paststudieson

theimpactsofdecouplingmechanismsonenergydemandhavebeenalmost

exclusivelybasedonexanteeconomicmodeling.PaststudiesofDSMprograms,

however,haveincludedbothexanteeconomicmodelingandexpostempirical

studiesoftheimpactofDSMprogramsonenergyefficiency.Probablyduetothe

uncertaintiessurroundingmeasuringsavingsfromenergyefficiency,thereappears

tobeadiscrepancyindemandreductionestimatesbetweentheseexpostandex

antestudies.Forexample,studiesofthecost‐effectivenessofutilityDSMprograms

thatuseexpostempiricalanalysistendtofindprogramslesscost‐effectivethan

thosethatrelyprimarilyonexantemethods(Arimuraetal.2011).Intermsof

researchontheimpactofdecouplingmechanismsonenergyefficiency,thissuggests

thatresultsfromexantemodelsmaydiffersignificantlyfromthetrueeffects.Thus,

thereisaneedforexpostempiricalstudiesontheimpactofdecouplingmechanisms

onenergydemand.

Factorsthatarecommonlybelievedtoaffectresidentialelectricitydemand

includeDSMspendingincurrentandpreviousyears,economicactivitylevel,energy

prices,weatherconditions,andenvironmentalsensitivity,orthegenerallevelof

concernofcustomersabouttheenvironment(Delmasetal.2005,Arimuraetal.

2011).JaffeandStavins(1995)developedafunctionalformfortherelationbetween

DSMspendingandelectricitydemandthattakesintoaccountthenonlinearrelation.

Arimuraetal.(2009)pointedoutapossibleendogeneityissueinmodelsof

electricitydemandwithDSMspendingconsideredanexogenousvariable.They

wereconcernedthatlevelofDSMspendingmaybecorrelatedwithunexplained

fluctuationsindemand.TheyattemptedtouseTwo‐StageLeastSquares(2SLS)

estimationandinstrumentforDSMspending.However,theirresultsdonotsupport

thehypothesisthatDSMspendingisendogenous(Arimuraetal.2011).

23

Arimuraetal.(2011)conductedafairlythoroughanalysisonthecost‐

effectivenessofelectricityenergyefficiencyprogramsandsomeofitsdeterminants,

includingdecouplingmechanisms.Theresultsofthisstudysuggestthatdecoupling

mechanismsmaystrengthenthedemand‐reducingeffectofDSMspending.

Unfortunately,duemostlikelytoinsufficientdatafromdecoupledelectricutilities,

thesefindingswereinsignificant.

Thepresentstudyaimstobuildonthepreviousliteratureandthemethods

ofArimuraetal.(2011)tomorethoroughlyexploretherelationbetween

decouplingmechanismsandelectricitydemand.Itfirstmodelstheeconomic

incentivesbehindDSMinvestmentundertraditionalregulationandundera

decouplingmechanismandthenempiricallyanalyzestheeffectsofdecoupling

mechanismsonenergyconsumptionpercustomer.Thisstudydisregardsthe

potentialissueofendogeneityduetocorrelationofDSMspendingwithunexplained

demand,asArimuraetal.(2011)’sresultsdonotsuggestthatthisissueof

endogeneityexists.Finally,robustnesschecksareperformedontheresultsofthe

empiricalanalyses.

ModelingDSMInvestmentIncentives

Introduction

Thesituationinwhichtheregulatorsetsarequiredpriceorrevenueforthe

utilityandtheutilitydecideshowmuchandhoweffectivelytoinvestinDSM

programscanberepresentedasaprincipal‐agentgame.Theregulatoristhe

principal,andtheutilityistheagent.Asinallprincipal‐agentgames,theprincipal

hasanobjectivefunction,butdoesnothavealltheinformationitneedstobest

maximizethisfunction.Theagenthastheinformationnecessarytobettermaximize

thisfunction,butitactsonlyinitsownself‐interest.Duetothisasymmetric

information,theprincipalneedstheagentinordertoreachitsobjective,butitruns

theriskthattheagent’sobjectiveswilldifferfromtheoptimalattainmentofthe

principal’sobjective.Theregulatormustdesignthegameinsuchawaythatthe

incentivesofthetwopartiesalign.

24

Inthisgame,theregulator(principal)wantstomaximizeasocialwelfare

function,whiletheutility(agent)isonlyconcernedwithitsprofits.Theutilityhas

moreinformationthantheregulatordoesabouttheDSMoptionsavailableand

abouttheprojectedefficacyofeachoption.Theutilitycanusethisinformation

asymmetrytoincreaseitsprofitsbyactingstrategically.

Inasimpleversionofaprincipal‐agentgameonDSMprograminvestment

undertraditionalregulation,theregulatorwantstomaximizesocialwelfare.Social

welfareisdefinedastheintegralofdemand,ormarginalsocialbenefitofconsuming

oneunitofelectricity,lessthemarginalsocialcostofproducingandconsumingthat

electricity.Itisafunctionoftheexogenous,regulator‐determinedprice(p),quantity

ofelectricityconsumed(q),andthemarginalnetsocialbenefitelectricity

consumptionschedule.Itisassumedthatmarginalsocialcostofelectricity

consumptionisgreaterthanmarginalprivatecost,sincetheGHGemissions,air

pollution,andotherenvironmentalcostsassociatedwithelectricityarenot

internalizedintheprivatecostofelectricity.Forthesamereasons,itisalso

assumedthatthequantityofelectricityconsumedisabovetheefficientquantity.

Theregulator,therefore,wantstodecreasequantityofelectricitysoldby

encouragingDSMspending.Theregulatorcandosoby:

A. Requiring$XofDSMspending

B. Adjustingtherateofreturnoninvestmentviaaspecificprice(p)orrevenue

(R)basedontheutility’sreportedDSMspendinginthepreviousperiod

or

C. AdjustingtherateofreturnoninvestmentviaaspecificporRbasedon

perceivedDSMinvestmentefficacy.

Forthepurposeofthisexercise,itisassumedthattheregulatorcanchoose

oneoftwoallowedratesofreturn,ahighrewardrate(associatedwithpHorRH)and

alowpunishmentrate(associatedwithpLorRL).Strategy(B)assumesthatthe

rewardpriceissetiftheutilityinvestsatleast$XinDSMandthatthepunishment

25

priceissetotherwise.Strategy(C)assumesthattheregulatoruseselectricity

consumptioninthepreviousperiod(i.e.sincethelastratecase)todeterminewhich

p(orR)tosetinthecurrentperiod.Foranygivenperiod,theregulatorcantryto

encouragetheutilitytoinvestinDSMbypromisingtosetitspriceinafutureperiod

pt+1(orRt+1)asafunctionofestimateddemand‐reducingeffectofDSMinvestment

inthepreviousperiod.ThevalueofthisdemandreductionduetoDSMishenceforth

denotedEDSM.

Understrategy(C),whiletheregulatorcannotpreciselymeasuretheefficacy

ofaDSMprogramrelativetootherpotentialprograms,itcanestimatetheefficacy

byobservingenergyconsumptionpercapita.Theregulatorcanexaminemarket

trendsandrelevantfactorstoestimateprojectedconsumption,qp,tinthegiven

period.ItcannotconfidentlyattributechangesinrealizeddemandtoDSM

investmentorlackthereof;yetitknowsthatactualenergyconsumption(qa,t)equals

qp,t–EDSMt+ε t,whereε tcapturesunexpectedvariationandhasmeanzero.The

regulatorcan,therefore,setprice(pt+1)orrevenue(Rt+1)inthenextperiodasa

functionofqp,t‐qa,tsoastoalterprojectedrevenuesforperiodt+1asaresultof

actualquantityofelectricityconsumedinthepreviousperiod(dpt+1qt+1/dqa,t<0).

Forsimplicity,itisassumedthattheregulatordeterminesacutoffqc,,tsuchthatqc,,t

≤qc,,t.Foranygivenperiodt,theregulatorsetstherateaspH,torRH,tiftheprevious

periodresultedinaqa,t­1≤qc,t­1andsetstherateaspL,torRL,totherwise.Whenthe

regulatorchoosestosetprices,itsetstheratesbasedonprojectedrevenuesand

associatedquantities(qHandqLrespectively),suchthatpHqH>pLqL.Thus,the

regulatortakesthedemand‐reducingeffectofDSMintoaccountwhensettingthese

rates.

Thisisacrediblestrategyforsomerewardrate(pHorRH),punishmentrate

(pLorRL),andqcbecausetheprospectofincreasedfutureconsumersurplusdueto

decreasedconsumptionoutweighsthepotentialdecreaseinconsumersurplusasa

resultofahigherprice,forsomesufficiently‐largeconsumptionreductionandsome

sufficiently‐smallpriceincrease.Notethatthismodeldoesnotattempttouncover

thespecificvalueofX,therewardrate,thepunishmentrate,ortheqc.Itsimply

26

demonstratesthedifferencesineffectsofregulator‐orthirdparty‐imposedDSM

standardsandincentivesonutilityDSMinvestmentundertraditionalregulation

versusdecouplingmechanisms.

PartA–TraditionalRegulation

Undertraditionalregulation,theutility’sobjectivefunctionisdenoted:

Maxπ t=ptqt–ft–St

whereftisthefixedcostsassociatedwithsupplyingelectricitytothecurrent

customersandStistheamountthattheutilityspendsonDSMinperiodt.Notethat

thismodelassumesnomarginalcostsofelectricitydistribution.Foranygivenp,set

bytheregulator,theutilitywantstomaximizeq.Thefirstorderderivativeofthe

aboveequationwithrespecttoDSMspendingisthefollowing:

idπidS

= p idq

idS−1

withdqt/dSt≡EDSMt<0.

IftheregulatordoesnotinstituteanyrulesorincentivesconcerningDSM

andsetspricesexogenously,itisevidentfromtheaboveequationthattheutility

willmaximizeqandwillnotinvestinDSMatall.Iftheregulatorchoosesstrategy

(A)andrequirestheutilitytospend$XonDSM,theutilityhasanincentivetosteer

awayfromeffectiveDSMprograms.Theutilitywouldprefertoburnthemoneythan

spenditonDSM.Similarly,iftheregulatorchoosesstrategy(B)andbasespriceson

theamountofDSMspendinginthepreviousperiod,theutilityhasanincentiveto

spend$XonDSMifthepresentvalueofthebenefitsofdoingsoisgreaterthanthe

costs,butitwillchoosetodosoasineffectivelyaspossible.Iftheutilitycanspend

$XonDSMwithoutactuallyaffectingenergyconsumptionatall(dQ/dS=0),itwill

choosetodosoifδ(pHqH–pLqL)>Xwhereδ istheutility’sdiscountfactor(0≤δ≤1).

27

TheutilitymustbebribedorpunishedmoreiftheDSMspendingwereunavoidably

slightlyeffective.

Iftheregulatorusessignalsotherthanspendingtoestimatetheutility’s

effectiveinvestmentinDSMandsetspricesaccordingly(strategyC),theutilitywill

effectively3investanamount,X,inDSMinagivenperiodifthepresentvalueofthe

benefitsofthisinvestmentisgreaterthantheassociatedcost.Thisisrepresentedby

thefollowinginequality:

[θ1(pHqH)+(1­θ1)(pLqL)–θ2(pLqL)–(1­θ2)(pHqH)]*δ ≥X+pt*EDSM,t

whereθ1istheprobabilityoflowdemandbeingrealized(qa<qc)iftheutility

invests$XinDSM,θ2istheprobabilityofhighdemandbeingrealized(qa>qc)ifthe

utilitydoesnotinvestinDSM,,andptisthepriceinthecurrentperiod.

Assumingthattheexpectedvalueofqaisgreateriftheutilitydoesnotinvest

thanifitdoes,(i.e.θ1+θ2>1),thisinequalitycanberearrangedtothefollowing:

(pHqH­pLqL)≥[X+ptEDSM,t]*[1/(δ(θ1+θ2­1))]

Theutilitywillonlyeffectivelyinvest$Xinenergyefficiencyifthe

punishmentpriceisassociatedwitharevenuesufficientlylessthantherevenue

associatedwiththerewardprice.Thesufficientdifferencedecreasesifthedirect

costoftheDSMinvestmentdecreases,ifthepriceinthecurrentperiodissmaller,if

theexpecteddemand‐reducingefficacyoftheinvestmentissmaller,iftheutility

valuesthefuturerelativelymore(δ isgreater),oriftheprobabilityofthedemand

realizationcorrectlyreflectingtheutility’sactionsisgreater(i.e.thevariabilityofε

islower).

3Forsimplicity,efficacyisconstrainedheretotakingonevalue(1)iftheinvestmentiseffectiveand0otherwise.

28

PartB–Decoupling

Underafulldecouplingmechanism,theregulatorchoosestheutility’s

requiredrevenue(R).TheutilitystillchooseswhethertoinvestinDSMandwhether

todosoeffectively.Inthiscase,however,theutility’sprofitisonlyafunctionof

requiredrevenues(andanychangesinfixedcosts).Anyshort‐termexcessor

shortageofrevenueisquicklyoffsetbypricetrue‐upsinthenextperiod.The

utility’sobjectivefunctionisrepresentedas:

Maxπ =R­f–S

Notethatquantityofenergysolddoesnotinfluencetheutility’sprofitsadverselyor

favorably(dR/dS=0).

WithnoregulationsorincentivesforDSMinvestment,thedecoupledutility

hasanincentivenottoinvestinDSM(dπ/dS=‐1).Iftheregulatorchoosesstrategy

(A)andrequirestheutilitytospend$XonDSM,theutilityisindifferenttoefficacy

oftheinvestment.Similarly,iftheregulatoroffersahigherallowedrevenueifthe

utilityspendsatleast$XinDSM(B),theutilitywilldosoifthepresentvalueofthe

benefitsoutweighsthecosts,δ(RH­RL)>S,andwillagainbeindifferenttotheefficacy

oftheinvestment.Theutilityhasnoprivateincentivetoinvestmorethanthe

requiredamountineitherofthesecases.

Theregulatorcantrystrategy(C)andencouragetheutilitytoinvestinDSM

inarepeatedgamebysettingarequiredrevenueeachperiod,RHorRL(with

RH>RL),asafunctionofactualenergyconsumptionqainthepreviousperiod.The

utilitywillchoosetoefficientlyinvest$XinDSMifandonlyifthepresentvalueof

thebenefitsofinvesting$XinDSMoutweighthecosts.Inthecaseofadecoupled

utility,thissituationisrepresentedbythefollowinginequality:

[θ1(RH)+(1­θ1)(RL)–θ2(RL)–(1­θ2)(RH)]*δ ≥X

29

whereδ ,θ1,andθ2arespecifiedastheywereinpartA.Underadecoupling

mechanism,theutilityexperiencesnocostofdecreasedconsumptionduetoDSM

investment.AssumingtheexpectedvalueofqaisloweriftheutilityinvestsinDSM,

thesituationadherestotheaboveinequalityifandonlyif:

(RH–RL)≥X*[1/(δ(θ1+θ2­1))]

Theonlyfactorsthatdeterminewhetherornottheutilitywillinvest$Xin

DSMaretherelativeeffectsofdoingsoonfuturerevenues,theutility’sdiscount

factor,andtheprobabilityoftheregulatorcorrectlyassessingtheutility’sactions.

Onceagain,themoretheregulatorwantstheutilitytoinvestinDSM,thelargerthe

rewardrevenuemustberelativetothepunishmentrevenue.Thisrevenuedisparity

canbesmalleriftheutilityvaluesthefuturerelativelymoreorifthevarianceofεis

low.

Conclusions

ComparingParts(A)and(B)suggeststhefollowingconclusionsaboutthe

relativelevelsandefficaciesofDSMinvestmentunderdecouplingmechanismsand

traditionalregulationinthespecifiedscenarios:

1. Withnoregulatorincentivesorstandards,neitherdecouplednor

traditionallyregulatedutilitieswillinvestinDSM

2. WithregulatorstandardsonrequiredDSMspending,decoupledutilitieswill

investmoreeffectivelyinDSMthantraditionallyregulatedutilities

3. WithregulatorincentivesbasedonamountofDSMspendingonly,decoupled

utilitieswillbemorelikelytoinvestinDSMandwilldosomoreeffectively

thantraditionallyregulatedutilities

4. WithregulatorincentivesbasedonperceivedDSMinvestmentefficacyas

inferredfromenergydemand,decoupledutilitieswillproducemoreeffective

DSMinvestmentthantraditionallyregulatedutilities

30

Thefirstthreeconclusionshavebeenpreviouslydiscussed,butthefourthisnot

immediatelyapparent.Understrategy(C),theregulatormustofferalarger

reward/punishmentdiscrepancytoautilitythatisregulatedtraditionallythanit

mustoffertoadecoupledutilityinordertoencouragethesameamountofeffective

DSMinvestment.Thisisapparentbycomparingtheinequalitiesthatrepresentthe

sufficientconditionsforregularutilityinvestment$XinDSMundereachtypeof

regulation.SincepHqHisessentiallyanalogoustoallowedrevenues,RH,thesufficient

conditionis:

(RH­RL)>[X+p0EDSM]*[1/(δ(θ1+θ2­1))]

undertraditionalregulation.Underafulldecouplingmechanism,thesufficient

conditionis:

(RH­RL)≥X*[1/(δ(θ1+θ2­1))]

SincepHandEDSMtakeonlypositivevalues,itisevidentthat:

[X+p0EDSM]*[1/(δ(θ1+θ2­1))]>X*[1/(δ(θ1+θ2­1))]

Thus,thesufficientreward/punishmentstructuretoencourageDSMinvestmentof

$Xisgreaterundertraditionalregulationthanunderadecouplingmechanism.

Furthermore,itisreasonabletoassumethatincreasingRH‐RLiscostlyforthe

regulator,asitisboundedbyaminimumrevenuethatwillcausetheutilitytoshut

downandwouldcausedecreasedconsumersurplusduetohigherenergypriceson

theotherextreme.Asaresult,regulatorswillprobablynotofferreward/

punishmentstructuresthatincentivizecomparableDSMinvestmentsunderthetwo

typesofregulation.Thereward/punishmentstructureassociatedwiththe

decoupledutilitywillprobablyinducemoreDSMinvestment.Thus,utilitieswill

probablyinvestmoreinDSMunderdecouplingthanundertraditionalregulation.

31

Itisimportanttonotethatthismodelisnotaperfectrepresentationofthe

interactionsbetweenutilitiesandregulatorsconcerningDSMinvestment.

Furthermore,theseincentivemodelsonlycompareincentivesunderdecoupling

mechanismstothoseundertraditionalregulation.Utilitiesunderregulationtypes

thatareneithertraditionalregulationnordecouplingmechanismsmaypossess

differentincentivesthantraditionallyregulatedordecoupledutilities.Nonetheless,

thispapercautiouslyextendstheimplicationsofthismodeltoformhypotheses

abouthowdecouplingmechanismsaffectutilityDSMinvestmentsandelectricity

consumptioncomparedgenerallytoallotherutilities.

TheoryandMethods

HypothesestobeTested

ThepreviousmodelssuggestthattheeffectofdecouplingonDSMcanbe

brokendownintoatleasttwoeffects:itseffectonthelevelofDSMexpenditureand

itseffectontheefficacyofDSMexpenditure,givenafixedlevelofinvestment.

Accordingtothemodels,eachofthetwoeffectsshouldreduceenergyconsumption

ifdecouplingiscombinedwithDSMincentivesorenergy‐efficiencystandards.

Furthermore,underdecouplingmechanisms,reducedconsumptioninherently

affectsprice.Combiningthiswithdownward‐slopingdemandproducesathird

pathwaybywhichdecouplingandDSMreducedemand.

Therefore,thisstudyinvestigatesthehypothesisthatdecoupling

mechanismsdecreaseresidentialelectricityconsumption,butonlythroughtheir

effectonthelevel,directdemand‐reducingeffects,andindirectdemand‐reducing

effectsthroughpriceofDSMinvestment.Thishypothesisisrepresentedgraphically

32

below:

Specifically,totestthisoverallhypothesis,thispaperexploresthefollowing

subordinatehypotheses:

1.Decouplingmechanismsonlyaffectresidentialelectricity

consumptionandretailpricethroughtheireffectsonDSM

investment.

2.DecouplingmechanismsareassociatedwithincreasedutilityDSM

expenditure.

3.Decouplingmechanismsincreasethedirectdemand‐reducingeffectof

residentialDSMprograms.

4.DecouplingmechanismsincreasetheefficacyofDSMinvestment

throughtheindirecteffectofDSMonconsumptionthroughprice.

Analysesofthesesubordinatehypothesesallowforabetterunderstandingofthe

natureandpotentialmagnitudesoftheeffectsofdecouplingmechanismsonenergy

efficiency.Duetothelimitedtimeframeofthedata,onlyveryshort‐term

realizationsofthesehypothesescanbeanalyzedinthisstudy.Asdiscussedlater,

33

however,itisreasonabletobelieveveryshortruneffects(1‐4yrs)willindeed

translateintoslightlylonger‐termeffects(≈5‐10yrs).

DataandMethods

ThispaperbuildsonpreviousanalysesofenergyefficiencyandDSM

investmenttomodeltheeffectsofdecouplingmechanismsonelectricity

consumption.LoughranandKulick(2004)conductedatime‐seriescross‐

section/time‐seriesanalysisontheeffectofDSMinvestmentonelectricity

consumption,andArimuraetal.(2011)similarlymodeledthecost‐effectivenessof

electricityenergyefficiencyprograms.Contrarytothesepreviousstudies,the

presentpaperincludespost‐2006data4,recognizestheendogeneityissuesof

includingutility‐levelelectricitypricesasaright‐hand‐sidevariable,andincludesa

vectoroftimefixedeffectstobettercontrolfornationaltimetrendsandstructural

breaksinelectricityconsumption.

DuetodiscrepanciesinfrequencyofreportinginEIAforms,twoprimary

datasetswereassembledforthisanalysis:onewithmonthlyobservationsandone

withannualobservations.Theformerisusedforthemajorityoftheanalysis,while

thelatterisusedforanalyzingthesecondsubordinatehypothesis–theeffectof

decouplingmechanismsonlevelofDSMspending–only.Theentirestudyfocuses

ontheresidentialsectorandusesdatafromJanuary2001toDecember2010.

Forthemonthlydataset,dataonmonthlyelectricityconsumptionbysector

foreachUSelectricutilitywereretrievedfromtheEnergyInformation

Administrationdatabase(FormEIA‐826).Thesedatawerecombinedwithannual

customerdatabysectorforeachutility(FormEIA‐861)toobtainelectricity

consumptionpercustomerbysectorforeachutility.Utilitiesthatservecustomers

inmultiplestateswereseparatedintomultipleutilities,oneperstate,because

utilitydecouplingregulationlawsvarybystate.

Fourhundredfourutilities,eachwith50‐120monthlyobservations,were

includedintheanalysesonmonthlydata.Seventyutilitieswithfewerthan50

4ThevastmajorityofdecoupledUSutilitiesdidsopost‐2006.

34

monthsofdatawereexcludedfromthemonthlyanalysis.Extremeobservations,or

consumptionobservationsthatvariedfromthepriormonthbyafactoroffiveor

greaterwereexaminedandalteredorexcludedfromthedata,asappropriate.These

observationsarelikelyresultsofmergersoracquisitions,temporaryutility

shutdowns,humanerrorinreporting,ornaturaldisasterssuchashurricane

Katrina.Inthecaseofsupposedmergersoracquisitions,utilitieswerelabeledas

multiple,uniqueutilitiesbeforeandaftertheevent.Afewobservationsthat

portrayednegativesaleswerealsodiscardedfromtheanalysis.Intotal,eleven

observationswerediscardedduetonegativesalesorsuspicionofutilityshutdowns,

humanerror,ornaturaldisasters.

DSMspendingdataisonlyavailableontheannuallevelfromFormEIA‐861.

Therefore,anannualdatasetwascreatedtoanalyzeDSMspending.Inthisanalysis,

utilitieswithfewerthanfiveyearsofdatawerediscarded.Datafromutilitiesthat

residedinmultiplestateswerealsoexcludedfromtheanalysissincedistinguishing

betweenspendingineachstatewasnotfeasiblefortheseutilities.Itisnotablethat

manyutilitiesarecompletelyexcludedfromFormEIA‐861.Only161utilitiesand

1,502annualobservationswereincludedinthispartoftheanalysis,primarilydue

tolackofDSMspendingdata.

Somethird‐partyDSMspendingthroughutilitiesisalsoreportedbyutilities

onEIAForm861.Asshownintheliterature,thisisreasonable,astheexistenceofa

decouplingmechanismaffectsathird‐partyenergyefficiencyprovider‘sabilityto

effectivelyinvestinDSM(Lazer,Weston,andShirley2011).Theinclusionof

ratepayer‐funded“third‐party”DSMspendingreducesbiasresultingfromthe

geographicalheterogeneityinstructureandexistenceofregulatoryinstitutions

responsibleforpromotingDSM.

Thispaperreplicatesmanyofthecontrolvariablesandusesseveralofthe

datasourcesfrompreviousstudies.Asadeparturefromsomeoftheprevious

35

literature,thispapercontrolsfortheexistenceofastateenergyefficiencyresource

standard(EERS),statequarterlypersonalincome,andtimefixedeffects.5

OLSand2SLScross‐section/time‐seriesanalysesareusedtoinvestigatethe

statedhypotheses.First,theeffectofadecouplingmechanismonresidential,per‐

customerelectricityconsumptionisfirstestimatedusingthefollowingOLSmodel:

Equation1:

Electusmt=B0+B1Decouplingusmt+B2Res.Customersust+B3NatGasPricesmt+

B4FuelOilPricemt+B5GDPst+B6Popst+B7Incomest+B8EERSst+

B9EnvSensitivityst+B10CDDsmt+B11HDDsmt+αu+γsm+εusmt

whereElect=thenaturallogarithmofresidential,per‐customerelectricity

consumption,EERS=adummyvariableforwhetherornotthestatehadanEnergy

EfficiencyResourceStandardinplace,EnvSensitivity=anstateenvironmental

sensitivitymeasure,CDD=coolingdegreedays,HDD=heatingdegreedays,u=

utility,s=state,m=month,andt=year.αandγarevectorsofutility‐levelandtime

fixedeffects,respectively.SeeAppendixBforfulldescriptionsofthevariables.The

naturallogarithmofresidentialelectricityconsumptionisusedasadependent

variabletocapturepercentage,asopposedtoabsolute,changesinconsumption.

Byincludingavectorofutility‐levelfixedeffects,oradummyindicator

variableforeachutility,theanalysiscontrolsforallsystematicdifferencesbetween

utilities,suchassize,thatdonotvarywithinthedecade.Theinclusionofavectorof

timefixedeffectsnormalizesoutanynationwidetimetrendsandtime‐specific

changesthatwouldaffectthedependentvariable.Utilitycustomerbaseandstate‐

andtime‐specificindicatorsofweather,economicgrowth,populationgrowth,

5IncontrasttoArimuraetal.’sstudy,thispaperdoesnotattempttomodeltherelationbetweenbuildingcodes,housingstarts,andenergyconsumption.Arimuraetal.neglecttoaccountforthetimelaginimportanceofbuildingcodes.Withoutamechanismtocapturethiscomplicatedrelation,thesevariablesseemtopickupotherpoliticalandenvironmentalsensitivityfactorsandmayactuallyconfoundresults.

36

income,environmentalstandards,anddemonstratedenvironmentalconcernare

exogenouscontrolvariables.Withallofthesecontrols,theonlyremainingvariation

fortheexistenceofdecouplingandotherpotentialomittedvariablestoexplainis

utility‐specificchangesinconsumptionpercustomerovertimethatarenotaresult

ofchangesinsizeofthecustomersbase.

Thefullmodeloftheeffectofdecouplingmechanismsonresidential,per‐

customerelectricityconsumptionincludesDSMspendingandaninteraction

betweenDSMspendingandthedecouplingdummyvariable.This2SLSmodelalso

controlsforaninstrumentedutility‐levelretailelectricityprice:

Equation2:

Electusmt=B0+B1Decouplingusmt+B2Res.Customersust+B3ElectricityPricemt+

B4NatGasPricesmt+B5FuelOilPricemt+B6GDPst+B7Popst+B8Incomest

+B9EERSst+B10EnvSensitivityst+B11CDDsmt+B12HDDsmt+B13DSMust+

B14DecouplingxDSMusmt+αu+γsm+εusmt

withalltermsdefinedastheywereinthepreviousmodel.IntheRobustnessTests

section,thisstudyexploresthepotentialissueofendogeneityofdecouplinginthis

equation.

DespitetheresearchofJaffeandStavins(1995)regardingthelingering

effectsofDSMinvestmentonconsumption,thisanalysisfindsthatalinear

relationshipbetweenthelogarithmsofthisyear’sDSMinvestmentand

consumptionisthebest‐fittingmodeloftherelationbetweenrecentDSM

investmentandthisyear’sconsumption.LaggedvaluesofDSMinvestmentarenot

significant.

Toaddresstheendogeneityofelectricitypricethrough2SLS,2005state‐

leveldataonthepercentofelectricitygenerationfromoil,naturalgas,andcoalare

interactedwithmonthlycitygate(wholesale)naturalgasprice,monthlyUSoil

refinercosts,andtheaverageannualpriceofcoal,respectively.State‐levelpercent

37

generationfromrenewablesourcesisalsointeractedwithUScoalprice6.Thereis

noreasontobelievethatthesevariableswouldimpactelectricitydemandexcept

throughtheireffectonelectricitypriceandpotentiallythroughtheireffectonretail

naturalgasprice.Atestofjointsignificanceonthecoefficientsoftheseinstruments

inthefirst‐stageregressionproducesanF‐statisticof45.8,whichisassociatedwith

ap‐valueoflessthan.0001.Thisconfirmsthevalidityoftheseinstruments.

Thefollowingcross‐section/time‐seriesmodeloftotalutilityannualDSM

spendingperresidentialcustomerisusedtoexploretheeffectofdecouplingon

levelofDSMspending:

Equation3:

DSMust=B0+B1Decouplingust+B2Res.Customersust+B3GDPst+B4Popst+B5Incomest

+B6EERSst+B7EnvSensitivityst+B8CDDst+B9HDDst+αu+γt+εust

whereDSM=thenaturallogarithmofannualDSMspending,u=utility,s=state,

andt=year.αandγarevectorsofutility‐levelandtimefixedeffects,respectively.

Standarderrors,clusteredonutility,wereusedinallregressionstoaccount

forheteroskedasticityandautocorrelation.Evenwhenavectoroftimefixedeffects

isincluded,aWoodridgetestforautocorrelationinpaneldatageneratedan

F(1,403)statisticof355,whichisassociatedwithalessthan.1%chancethatthe

datadoesnotsufferfromautocorrelation.Theclusteredstandarderrorsaremuch

morerobusttoissuesofautocorrelationandheteroskedasticitythanthetraditional

standarderrors.Thetrueautocorrelationstructureofthedatacouldnotbe

captured,sothisapproachisthebestavailableoption.Notethatthepresenceof

autocorrelationsolelyaffectsthepredictedstandarderrorsofpointestimates,while

inaccurateeffortstocapturetheautocorrelationstructurewouldcauseendogeneity

issuesandbiasthepointestimates.6Interactiontermsbetweenpercentrenewablegenerationandotherfuelpricesaswellastheinteractionbetweenmonthlyuraniumpriceandpercentgenerationfromnuclearenergywereexcludedbecausetheywerefoundtobeirrelevantinthefirststage.

38

Results

OverallEffectsofDecouplingonConsumptionandPrice

Consistentwithsubordinatehypothesis#1,thissectionshowsthat

decouplingmechanismsarecorrelatedwithelectricityconsumptionandpriceonly

throughtheirrelationwithDSMinvestment.Thesefindingsarerobusttoexclusion

oftheenvironmentalsensitivityindicatorvariableandthedummyvariableof

existenceofastateenergyefficiencyresourcestandard.

Thefirstmodel(Equation1)includesdecouplingbutnotprice,DSMorits

interactionwithDSM.Thus,thecoefficientondecouplinginthismodelcanbe

understoodasthefulleffectofDSMthroughitsvariouschannels.AsColumn1of

Table1indicates(SeeAppendixA),thecoefficientofdecoupling(B1)inthismodelis

‐.05.Thiscoefficientrepresentsthedifferenceintheexpectedmeansofthenatural

logarithmofconsumptionbetweendecoupledutilitiesandnon‐decoupledutilities.

Alternatively,theratioofthemeanofconsumptionfordecoupledutilitiesovernon‐

decoupledutilitiesise‐.05,orabout.95.Allelseequal,ifautilityswitchestohavinga

decouplingmechanismfromhavingnone,expectedresidential,per‐customer

electricityconsumptiondecreasesbyabout5%.Thesignofthisresultissignificant

atthe2%levelforatestofthenullhypothesisthatdecouplingmechanismshasno

effectonresidentialelectricityconsumption.

Whentheeffectsofdecouplingonconsumptionworkingthroughpriceand

DSMinvestmentareeliminated,however,thissignificancedisappears.Totestthe

effectsofeliminatingthesepotentialcausalavenues,Equation2(Column2Table1–

AppendixA)addsascontrolsresidentialelectricityprice,utilitylogannualDSM

spending,andaninteractiontermofDSMspendingandadummyforadecoupling

mechanismtothemodeldescribedbyEquation1.Becauseelectricitypricemaybe

endogenous,thisequationwasestimatedwith2SLS.Thefirststageregressionis

giveninColumn1ofTable3(AppendixA).

Withtheinclusionofthesevariables,thecoefficientonthedecoupling

dummyshrinksto‐.007andisnotsignificantlydifferentfromzero.Sincethemodel

39

controlsforlevelofDSMinvestmentandfortheincreaseddemand‐reducingeffects

ofDSMinvestmentassociatedwithdecoupledutilities,thiscoefficientestimatesthe

effectofdecouplingmechanismsnotrelatedtoDSMinvestmentorpriceon

electricitydemand.FordecoupledutilitiesthatdonotparticipateinanyDSM

investment,themodelfindsnosignificanteffectofhavingdecoupledonelectricity

consumption.

Thisresultisrobusttoexclusionoftheenvironmentalsensitivityrating

variableandthevariableforexistenceofanenergyefficiencyresourcestandard

(Column3Table1).Becauseenvironmentalsensitivityratingsandtheexistenceof

anenergyefficiencyresourcestandardarecorrelatedwiththepresenceofa

decouplingmechanism(r=.26and.30,respectively),onemightbeconcernedwith

thepresenceofthesecontrolsinthemodel.Whenthesevariablesareexcludedfrom

theanalysis,decouplingstillshowsnosignificantrelationtoconsumptionother

thanthroughitseffectonDSMinvestmentandprice.Thepointestimateand95%

confidenceintervalontheisolateddecouplingcoefficientbarelychange.

Consistentwiththeliteratureandcontrarytothefearofmanyopponentsof

decoupling,decouplingmechanismsalonedonotappeartoinherentlysignificantly

increaseprices.Column2ofTable3rerunsthefirststageanalysiswithoutthe

variablesofDSMinvestmentanditsinteractionwithdecoupling.Withoutholding

thesevariablesconstant,decouplingdoesappeartoincreaseprices.Specifically,

decouplingisassociatedwitha1%increaseinpricewhenDSMinvestmentis

ignored,andthisrelationissignificantatthe10%level.

IftheeffectsofDSMinvestmentareproperlyaccountedfor,however,these

resultsdisappear(Column1Table3–AppendixA).HoldingDSMinvestment

constant,ifautilityhasadecouplingmechanismbutnoDSMinvestment,the

decouplingmechanismispredictedtohavenosignificanteffectonprice.The

coefficientondecouplingis.004andisnotsignificantatanyconventionallevel.This

insignificanceremainswhentheenvironmentalsensitivityandEERScontrol

variablesareexcluded(Column3–Table3).Therefore,thisstudyconcludesthat

decouplingmechanismsonlyaffectpricethroughtheireffectonDSMspending.

40

Thestrikingdifferenceinmagnitudeandsignificanceofthecoefficient

associatedwithdecouplingbetweenmodelsofelectricityconsumptionthatinclude

DSMspendingandthosethatdonotisexploredinthefollowingsections.

EffectofDecouplingMechanismsonLevelofDSMSpending

Toprovidemoreinsightintotherelationbetweendecouplingmechanisms

andelectricityconsumption,thisstudyanalyzesthreepotentialimpactsthat

decouplingmechanismsmayhaveonDSMinvestment.Thissectioninvestigatesthe

secondsubordinatehypothesis:theeffectofdecouplingmechanismsonthelevelof

DSMspending.Contrarytothehypothesis,anestimateoftheOLSmodelofDSM

spendingoutlinedinEquation3(Column1Table2–AppendixA)suggeststhat

decouplingandlevelofDSMspendingarenotsignificantlycorrelated.

BecauseitisimpossibletoengageinnegativeDSMspendingandbecause

therearemanyutilitieswith$0DSMspending,atobitmodelmaybemore

appropriatethananOLSmodelinthisinstance.Evenwiththismodelspecification,

however,thecoefficientondecouplingisnotstatisticallydifferentthanzero

(Column2Table2–AppendixA).Thisprovidesfurtherevidencethatdecoupling

mechanismsandlevelofDSMspendingareuncorrelated.

Itisimportant,however,toconsiderthedistributionofDSMspending.The

distributionisnotnormalandhasaskewnessof6.99.Therearemanyobservations

withzeroorverylowvaluesofDSMspendingandafewobservationswithvery

extremehighvalues.Toavoidassuminganormaldistributionandtolimitthe

effectsofoutliers,aquantileregressionwasused.Theresultsofthismodelshowa

positiveeffectofdecouplingonlevelofutilityDSMinvestmentthatissignificantat

the10%level(Column2Table2‐AppendixA).Thisregressionsuggeststhat

decouplingdoesencouragehigherutilityDSMspendingandthatthenormal

distributionalassumptionmayhavecausedinaccurateestimatesintheOLSand

tobitmodels.Thequantilemodelestimatesthat,allelseequal,thepossessionofa

decouplingmechanismincreasesautility’spredictedDSMspendingby41%.

41

Becauseofthefailureofthequantilemodeltocontrolforheteroskedasticity

andautocorrelation7,however,cautionshouldbeexercisedwheninterpretingthe

significanceoftheseresults.Furthermore,recallthatsinceDSMinvestmentdatais

onlyavailableannually,thisanalysishasfarfewerobservationsthantheanalysisof

Equation1.Astheremaybesomebiasduetotherelativelysmallsamplesizeofthis

dataset,thisisanotherreasonthatthepointestimategivenbythismodelshouldbe

treatedwithcaution.

EffectofDecouplingMechanismsontheDirectDemand­ReducingEffectsofDSM

Spending

Thethirdsubordinatehypothesisinthispaperisthat,controllingforthe

levelofDSMinvestment,decouplingmechanismsincreasethedemand‐reducing

effectsofDSMspending.Totestthishypothesis,the2SLSmodelofEquation2

(Column2Table1–AppendixA)isagainused.ThismodelcontrolsforlevelofDSM

investmentandfortheinteractionbetweendecouplingandDSMinvestment.The

coefficientonthisinteractiontermrepresentstheincreaseinefficacyofDSM

investmentassociatedwithdecouplingmechanisms.Thiscoefficientisestimatedto

be‐.004andissignificantatthe5%level.Thissupportsthehypothesisthat

decouplingmechanismsincreasedemand‐reducingeffectofutilityDSMinvestment

foragivenlevelofinvestment.GivenafixedlevelofinitialDSMinvestment,this

modelestimatesthata10%increaseinDSMinvestmentofadecoupledutility

decreasesestimatedresidential,per‐customerelectricityconsumptionby.04%

morethana10%increaseinDSMinvestmentofanon‐decoupledutilitywould

decreaseestimatedconsumption.

7Duetoinsufficientobservationsclusteredstandarderrorscouldnotbecomputedinthispartoftheanalysis.

42

EffectofDecouplingMechanismsontheEfficacyofDSMInvestmentduetoitsIndirect

EffectonConsumptionThroughPrice

Besidesincreasingthedemand‐reducingeffectsofDSMinvestment,

decouplingmechanismsmayalsoincreasetheefficacyofDSMinvestmentat

reducingconsumptionthroughtheeffectofDSMinvestmentonprice.Inutilities

withdecoupling,DSMinvestmentispredictedtoincreaseprices,andhigherprices

arepredictedtodecreaseconsumption(Hypothesis4).Thissectionexploresthis

subordinatehypothesisbyconsideringtheestimationoftheeffectofDSM

investmentofdecoupledutilitiesonprice,contrastingthiseffectwiththatofnon‐

decoupledutilitiesthatinvestinDSM,andverifyingthepredictedeffectofincreased

pricesonconsumption.

Aspredicted,resultssuggestthatDSMinvestmentincreasesresidential

electricitypriceunderdecouplingmechanismsandhasnoeffectonpriceunder

othertypesofregulation.Inthefirststageofthe2SLSmodeldepictedinEquation2,

thecoefficientonthevariabledepictingDSMinvestmentisessentiallyzeroand

associatedwithat‐statisticof.17(Column1Table3–AppendixA).Thisindicates

that,fornon‐decoupledutilities,DSMinvestmentdoesnotsignificantlyaffectprice.

Amongonlydecoupledutilities,however,theredoesappeartobeapositiverelation

betweenDSMinvestmentandprice.Theinteractiontermbetweendecouplingand

DSMinvestment(Column1Table3–AppendixA)ispositiveandsignificantatthe

10%level.ThisweaklysupportsthehypothesisthatDSMinvestmentincreases

residentialelectricityprice,atleastintheshortrun,underdecouplingmechanisms.

Consistentwitheconomictheory,thisstudyalsofindsevidencethathigher

pricesdecreaseelectricitydemand.Thecoefficientonpriceinthesecondstageof

this2SLSmodelis‐1.47,whichisstatisticallysignificantatthe.001level(Column2

Table1–AppendixA).Apriceincreaseof$.01/kWhisassociatedwitha1.5%

decreaseindemand.Atameanpriceofabout$.10/kWh,thisimpliesaprice

elasticityofabout.15,whichiswithinthe‐2.001to‐.004range(median=‐.28)of

recentpriceelasticityestimatesofshortrundemandintheliterature(Espeyand

Espey2004).

43

RobustnessTests

ThissectionaddressessomepotentialconcernsabouttheestimatesinTable

1.Itfirstteststhesensitivityoftheresultstoincludingamonthlyratherthanannual

coalpriceasaninstrumentalvariabletopredictelectricityretailprice.Itthen

exploreswhetherthecorrelationbetweenpossessionofadecouplingmechanisms

andelectricityconsumptionispartiallyduetothecontinuationofpre‐existing

trends.Finally,thissectionusesprobitanalysistoexplorethepossibilitythat

electricityconsumptionaffectsautility’sdecisiontodecoupleanddiscussesthe

potentialbiascausedbythepresenceofthisendogeneityissue.

UseofanAnnualInstrumentalVariable

Somereadersmayfindtheuseofannualcoalpriceasopposedtomonthly

coalpriceasaninstrumentformonthlyelectricitypricetroubling.Specifically,they

mayworrythattheuseofareplicatedannualvariablemightartificiallyincreasethe

power,ortheperceivedrelevance,oftheinstruments.MonthlyUScoalpriceswere

notavailable.TochecktherobustnessoftheannualUScoalpriceestimate,models

wereestimatedusingamonthlycoalpriceindex,Australianthermalcoal,inplaceof

annualUScoalprice.ThismonthlyindexisfoundtobehighlycorrelatedwithUS

coalprice,butitisfoundtobeaninferiorinstrumentbecauseitisslightlyless

powerfulinexplainingUSretailelectricityprices.

Thereplacementdoesnotaltertherejectionofthenullhypothesisthatthe

instrumentsarerelevant.Itdoeslowerthe95%intervalforthesecondstageeffects

ofa$.01increaseelectricityretailpriceonelectricityconsumption,fromadecrease

ofbetween.6%and2.3%toadecreaseofbetween.2%and1.7%.Whenbothcoal

priceindicatorsareincluded,theeffectoftheAustraliancoalpriceindexon

residentialelectricitypricelosessignificance,andthe95%confidenceintervalof

thedecreaseinconsumptionassociatedwitha$.01priceincreasebecomes.4%to

.9%.Whilethedifferencesintheseintervalsarenontrivial,theoverlapofthe

confidenceintervalsisnotable.Furthermore,theuseofthemonthlyindicator

44

variableinsteadoftheannualUSpricevariablehasanegligibleeffectonthepoint

estimatesandsignificanceoftheotherexogenousvariables.Thisprovidesevidence

forrobustnessoftheanalysesinthisstudytoatheoreticalchangeinannualversus

monthlyUScoalprice.

Pre­ExistingTrends

Becauseofthetime‐seriesnatureofthedata,itisimportanttoconsiderthe

possibilitythatpercustomerelectricityconsumptionwasdecreasingpre‐

decouplingfortheutilitiesthatlaterdecoupled.Becausealloftheutilitiesinthe

datasetthatdecoupledbetweentheyearsof2001and2010didsoafter2005,pre‐

existingtrendscanbecapturedbyanalyzingdatafrom2001‐2005.Utilitiesthat

adopteddecouplingmechanismspriorto2001wereexcludedfromthispartofthe

analysis.

Column4ofTable1(SeeAppendixA)showstheestimationofamodelusing

only2001‐2005dataofresidentialper‐customerelectricityconsumptiononayear

trend,aninteractiontermbetweenayeartimetrend8andwhethertheutility

decoupledbetween2006and2010,controlvariables,andutility‐fixedeffects.The

interactiontimetrendshouldcaptureanysystematicdifferencesinenergyuse

trendsbetweenutilitiesthatlateradopteddecouplingmechanismsandthosethat

didnotdoso.Ifper‐customerenergyconsumptiondecreasedrelativelymoreor

increasedrelativelylessinutilitiesthatlaterdecoupled,thecoefficientofthis

variableshouldbenegative.Infact,asseeninColumn4ofTable1,thecoefficientof

thisinteractiontermisactuallypositiveandsignificantatthe1%level.Controlling

forthevariablesinthepreviousmodel,pre‐2006energyconsumptionwasactually

increasingmoreforutilitiesthatdecoupledduring2006‐2010thanforutilitiesthat

didnotdecoupleduringthistimeperiod.Thus,theperceivedeffectsofdecoupling

mechanismsreducingenergyconsumptiondonotseemtosimplybethe

continuationofpre‐existingutility‐specifictrends.

8Note–thisyeartimetrendisnotusedinmodelswithtimefixedeffectsduetoitsinsignificance.

45

PotentialEndogeneityoftheDecouplingVariable

Theupwardtrendinconsumptionforutilitiesthatlaterdecoupledsuggests

thatincreasesindemandmighthavecausedtheadoptionofdecoupling

mechanisms.Inresponsetothisconcern,thisstudyusesprobitanalysisto

investigatewhatcausesutilitiestodecouple.Column1ofTable4(AppendixA)

showstheresultsofamodelthatpredictswhetherautilitydecoupledbetween

2006and2010asafunctionofJanuary2001valuesofvariousindependent

variablesforthe384utilitieswhohadnotdecoupledby2006.Theseresultsindicate

thatutilitieswithalotofcustomerswhoeachusealotofenergythatarelocatedin

wealthy,environmentally‐consciousstatesaremostlikelytodecouple.Controlling

fortheseJanuary2001characteristics,theresultsinColumn2ofTable4estimate

howtrendsinvariablesbetweenJanuary2001andDecember2005affected

utilities’decisionstodecouple.Theseresultssuggestthattheonly2001‐2005trend

thatissignificantinpredictingdecouplingadoptionwastheinstitutionofastate

energyefficiencyresourcestandard.

Furthermore,thet‐statisticassociatedwiththeincreaseinenergyuseper

customerisabout1,whichsuggeststhat,ifanything,growingenergyuseper

customermayalsoincreasethechancesofautilitydecoupling,althoughthis

relationisnotstatisticallysignificant.Thisfindingisnotsimplydueto

multicollinearitybetweentrendvariables.Thesignandtheinsignificanceofthe

coefficientpersistswhenpercustomerconsumptionistheonlytimetrendincluded

(Column3Table4–AppendixA).

Theresultsoftheprobitmodelssuggestthatthedecouplingdummyvariable

mayormaynotbeendogenousintheper‐customerelectricityconsumptionmodels.

Thefactthatutilitieswithhighper‐customerconsumptiontendtoadoptdecoupling

mechanismsshouldnotcauseendogeneityissuesbecauseutility‐fixedeffectsare

46

includedintheconsumptionmodels.9Incontrast,thesuggestionthattrendsinper‐

customerconsumptionmayaffectautility’sdecisiontodecouple,albeitstatistically

insignificant,provokesconcernaboutpotentialendogeneityissuesinthemodelsof

electricityconsumptionofTable1.Ifthiscausalrelationdoesexist,however,the

resultsoftheTable4probitmodelssuggestthatincreasingper‐customerelectricity

consumptionincreasesthechancesofautilityhavingadecouplingmechanism.This

impliesthatendogeneitywouldbiasthecoefficientofthedecouplingvariableinthe

positivedirection.Thus,thecoefficientsofdecouplingintheelectricityconsumption

modelsareeitherunbiasedorclosertozerothanthetruecoefficient.Thisfurther

strengthenstheargumentthatdecouplingmechanismsreduceper‐customer

residentialelectricityconsumption.

LongRunImplications

Thefindingsinthispaperrepresentshortrunrelationships(<5years)

betweendecouplingmechanismsandefficiencyofresidentialelectricity

consumption.Thereissomereasontobelievethattheserelationshipspersist,in

somemagnitude,foranextendedperiodoftime.AsJaffeandStavins(1995)found,

DSMinvestmentsaffectenergyconsumptionforyears.

ThefindingthatDSMinvestmentofutilitieswithdecouplingmechanisms

causeshigherprices(andthathigherpricesimpactdemand),however,isinnatelya

shortrunresult.ThelinkbetweenDSMinvestmentsofdecoupledutilitiesandprice

isthat,underadecouplingmechanism,reduceddemandwillcausepricetoincrease

withthenextscheduledtrue‐up.Onceanewratecaseoccurs,however,new

revenuetargetsandcorrespondingpricesaresetbyaregulatorandareless

dependentonpreviousconsumption.Decreasedconsumptionitselfmayormaynot

affecttheregulator’spricedecision.Thus,increaseinpricemayonlypersistuntil

thenextratecase(≈3‐6yrs).

9Theerrorterminthesemodelsonlycapturesutilityandtime‐specificchangesinconsumptionandwouldnot,therefore,becorrelatedwithdifferencesinabaselinemonth’sper‐customerelectricityconsumptionacrossutilities.

47

Nonetheless,eveniftheregulatordecidestorestorepricestotheirpre‐true‐

uplevels,consumptionisunlikelytoreturntothe“efficient”marketlevelatthis

price.Energyefficiencyinvestmentsandbehaviorchangescausehysteresis,or

history‐dependent,effects.Ifendusersrespondtohighelectricitypricesbybuying

energyefficientappliancesormakingbehavioralchangestowastelessenergy,

lowerenergypricesmaynotcausethemtogetridoftheirnew,efficientappliances

ortochoosetoswitchbacktowastefulbehaviors.Therefore,theeffectondemand

ofhigherpricesduringtrue‐upsmayverylikelypersistthroughsubsequentrate

cases.

Conclusions

Withwidespreadconcernaboutglobalclimatechangeandenergysecurity,

theefficacyofmechanismstoimproveenergyefficiencyisofconsiderableinterest.

Recentliteratureonthedeterminantsofenergyinefficiency,includingconsumer

choicefailureandutilitythroughputincentives,hasespeciallyincreasedinterestin

promotingutilityinvolvementinDSM.Decouplingmechanismsareanincreasingly

populartechniquetoencourageutilityinvolvementinenergyefficiencyissues,but

littleempiricalanalysishadbeenconductedintotheeffectsanddeterminantsof

decouplingmechanismsonelectricitydemand.Thispaperexploresthisrelation

usingrecentutility‐leveldata.

ModelsoftheinterplaybetweenutilityDSMinvestmentandregulator

incentiveshaveimportantimplicationsontheexpectedeffectsofdecoupling

mechanismsonenergyefficiency.Thesemodelsindicatethatdecoupling

mechanismsonlyaffectinvestmentinDSMinvestmentifadditionalDSMstandards

orincentivesexist.Moreover,inthepresenceofstandardsonDSMspending,

decoupledutilitieswillinvestmoreeffectivelyinDSMthantraditionallyregulated

utilities.Similarly,inthepresenceofincentivesbasedonlyonlevelofDSM

spending,decoupledutilitieswillbemorelikelytoinvestinDSMandwilldoso

moreeffectivelythantraditionallyregulatedutilities.Finally,withregulator

incentivesbasedonperceivedDSMinvestmentefficacy,asinferredfromenergy

48

demand,decoupledutilitieswillproducemoreeffectiveDSMinvestmentthan

traditionallyregulatedutilities

Basedonthesemodels,thispaperexploresthehypothesesthatdecoupling

mechanismsdecreaseresidentialelectricitydemandandthatthisisaccomplished

throughencouragingagreaterlevelofDSMinvestment,alargerdemand‐reducing

effectofanygivenlevelDSMinvestment,andapriceincreasethatfurther

discouragesconsumption.Thispapertheorizesthattheeffectofdecoupling

mechanismsisprimarilyduetoitseffectsonutilityDSMinvestmentanditseffects

onpricethroughDSMinvestment.

Consistentwiththehypotheses,empiricalresultssuggestthatdecoupling

mechanismsarecorrelatedwithelectricityconsumptionandpriceonlythrough

theirrelationwithDSMinvestment.NotcontrollingfortheeffectsofpriceandDSM

investment,pointestimatesofthedecreaseinconsumptionassociatedwith

decouplingmechanismsarequitelargeandsignificantatthe5%level.Whenthe

effectsofdecouplingonconsumptionworkingthroughpriceandDSMinvestment

areeliminated,however,thissignificancedisappears.

Asaresultofasmallsamplesize,thispaperneitheraffirmsnorrejectsthe

hypothesisthatdecouplingmechanismsincreasethelevelofutilityDSMspending.

ThelackofsignificanceinOLSandtobitmodelssuggeststhatdecoupling

mechanismsarenotcorrelatedwithlevelofDSMinvestment.However,theresults

ofaquantileregression,whichaccountsfortheskeweddistributionofDSM

spending,doessuggestthatdecouplingmechanismsmayindeedincreaselevelof

DSMspeding.Moreresearchisneededtothoroughlyexplorethishypothesis.

GivenafixedlevelofDSMspending,resultsstronglysuggestthatdecoupling

mechanismsincreasethedirectdemand‐reducingeffectsofDSMspending.Alarge

estimatedeffectofincreasedDSMinvestmentofdecoupledutilitiesrelativetothat

ofotherutilitiesisfoundtobesignificantatthe5%level.Thissubstantiatesthe

resultofthetheoreticalmodelconcerningefficacyofDSMinvestment.Specifically,

themodelindicatesthatDSMinvestmentofdecoupledutilitiesismoreeffective

thanthatoftraditionally‐regulatedutilitieswhenregulatorsdesignDSMinvestment

49

incentivemechanismsandregulationsthatarebasedprimarilyonlevelofDSM

spending.

Thispaperalsofindsweakevidencethatdecouplingmechanismsincrease

theefficacyofDSMinvestmentinreducingconsumptionthroughtheeffectofDSM

investmentonprice.DSMinvestmentbydecoupledutilitiesissignificantly

negativelycorrelatedwithpriceatthe10%level.Higherpriceswerealsofoundto

beveryhighlycorrelatedwithdecreasedelectricitydemand.

Testsonpre‐2006datademonstratethattheperceivedeffectofdecoupling

mechanismsonelectricityconsumptionisnotduetothecontinuationofpre‐

existingutility‐specifictrendsinelectricityconsumption.Probitanalysessuggest

thatthedecouplingdummyvariablemayormaynotbeendogenousintheper‐

customerelectricityconsumptionmodels.However,thesignofthepointestimates

oftherelationbetweendecouplingandDSMsuggeststhat,ifanything,thetrue

impactofdecouplingonelectricityconsumptionisactuallymorenegativethan

estimatedinthispaper.Thisstrengthenstheargumentthatdecouplingmechanisms

reduceper‐customerresidentialelectricityconsumption.

Whileitisreasonablyclearthatdecouplingmechanismspromotemore

efficientelectricityuseintheresidentialsectorthroughmoreeffectiveutilityDSM

investment,thelinkbetweendecouplingmechanismsandlevelofDSMexpenditure

remainsrelativelyambiguous.CollectingDSMspendingdataformoreutilities

wouldimprovetheestimationoftheeffectofdecouplingmechanismsonlevelof

DSMspending.Morein‐depthresearchofthedeterminantsofDSMinvestmentand

thespecificregulatorpoliciesconcerningDSMinvestmentwouldalsoassistin

understandingtheinterplaybetweenincentivestructure,levelofDSMinvestment,

andefficacyofDSMinvestment.Withthisadditionalresearchwillcomeaneven

greaterunderstandingoftheeffectsofdecouplingmechanismsonenergyefficiency.

50

References

ACEEE(AmericanCouncilforanEnergy‐EfficientEconomy)(2011).“StateEnergyEfficientPolicyDatabase”.AmericanCouncilforanEnergy‐EfficientEconomy.Availableat:http://www.aceee.org/sector/state‐policy

Ackerman,FrankandElizabethA.Stanton(2008).“WhatWe’llPayifGlobal

WarmingContinuesUnchecked.”NaturalResourceDefenseCouncilReportMay2008.Availableat:http://www.nrdc.org/globalwarming/cost/cost.pdf

AGA(AmericanGasAssociation)(2007).“UpdateonRevenueDecoupling

Mechanisms.”NaturalGasRateRound‐UpApril2007.Availableat:www.epa.state.il.us/air/climatechange/documents/subgroups/power‐energy/aga‐update‐on‐revenue‐decoupling‐mechanisms.pdf

Arimura,ToshiH.,ShanjunLi,RichardG.Newell,andKarenPalmer(2011).“Cost‐

EffectivenessofElectricityEnergyEfficiencyPrograms.”ResourcesfortheFutureDiscussionPaperRFFDP09‐48‐REVrevisedApril2011.Availableat:http://www.rff.org/RFF/Documents/RFF‐DP‐09‐48‐REV.pdf

Asif,M.andT.Muneer(2007).“Energysupply,itsdemandandsecurityissuesfor

developedandemergingeconomies.”RenewableandSustainableEnergyReviews11(7):1388‐1413

Auffhammer,Maximilian,CarlBlumstein,andMeredithFowlie(2008).“Demand

SideManagementandEnergyEfficiencyRevisited.”EnergyJournal29(3):91–104

Baumann,David(2008).“TheEffectsofGovernmentalEnergyEfficiencyActivities

onElectricityConsumption–AnEconometricAnalysis.”LBJSchoolofPublicAffairs

Bhole,BhuratandSunitaSurana(2011).“ElectricityPricesandStateCommitment

toEnergyEfficiencyintheU.S.”EnergyEfficiency4:9‐16Blomquist,Glenn(1974).“TheEffectofElectricUtilityPowerPlantLocationonArea

PropertyValue.”LandEconomics50(1):97‐100Boonin,DavidM.(2009).“ARateDesigntoIncreaseEfficiencyandReduceRevenue

Requirements”ElectricityJournal22(4):68‐78Brennan,Tim(2009).“TheChallengesofClimateforEnergyMarkets.”RFF

DiscussionPaperNo.09‐32.

51

Brennan,Tim(2008)."’NightoftheLivingDead’or‘BacktotheFuture’?ElectricUtilityDecoupling,RevivingRate‐of‐ReturnRegulation,andEnergyEfficiency.”RFFDiscussionPaperNo.08‐27.

Brennan,Tim(2009)."OptimalEnergyEfficiencyPoliciesandRegulatoryDemand‐

SideManagementTests:HowWellDoTheyMatch?"RFFDiscussionPaperNo.08‐46

Brookings(2011).“EnergySecurityInitiative.”AccessedNov2011.Availableat:

http://www.brookings.edu/projects/energy‐security.aspxBryce,Robert(2010).“ABadBetOnCarbon.”TheNewYorkTimes.Availableat:

http://www.nytimes.com/2010/05/13/opinion/13bryce.htmlBureauofEconomicAnalysis(2011).“RegionalData:GDP&PersonalIncome.”U.S.

DepartmentofCommerceaccessedNov2011.Availableat:http://www.bea.gov/iTable/iTable.cfm?reqid=70&step=1&isuri=1&acrdn=1

Costello,Ken(2011).“SomeAdvicetoRegulatorsonFormulaRatePlans.”

ElectricityJournal24(2):44‐54CSI(CenterforStateInnovation)(2008).“UtilityRateDecoupling:Conserving

Energy:APolicyBriefbytheCenterforStateInnovation”CenterforStateInnovation.Availableat:http://www.stateinnovation.org/Publications/All‐Publications/Utility‐Rate/UtilityRates.aspx

Costa,DoraL,andMathewE.Kahn(2010).“WhyHasCalifornia’sResidentialElectricityConsumptionBeenSoFlatSincethe1980’s?AMicroeconometricApproach.”NationalBureauofEconomicResearchWorkingPaper15978.

Edelstein,PaulandLutzKilian(2009).“Howsensitiveareconsumerexpendituresto

retailenergyprices?”JournalofMonetaryEconomics56(6):766‐779ECW(EnergyCenterofWisconsin)(1997).“Market‐BasedUtilityDSMPrograms.”

Report,Aug1997.Availableat: www.ecw.org/ecwresults/162‐1.pdf

EIA(U.S.EnergyInformationAdministration)(1997).“ElectricityPricingina

CompetitiveEnvironment:MarginalCostPricingofGenerationServicesandFinancialStatusofElectricUtilities.”EnergyInformationAdministration–OfficeofIntegratedAnalysisandForecasting.Availableat:ftp://ftp.eia.doe.gov/electricity/0614.pdf

52

EIA(U.S.EnergyInformationAdministration)(2011).“FormEIA‐826DataMonthlyElectricUtilitySalesandRevenueData.”DepartmentofEnergy/EnergyInformationAdministration.Availableat:http://www.eia.gov/cneaf/electricity/page/eia826.html

EIA(U.S.EnergyInformationAdministration)(2011).“FormEIA‐861FinalDataFile

for2010.”DepartmentofEnergy/EnergyInformationAdministration.Availableat:http://www.eia.gov/cneaf/electricity/page/eia861.html

EnerStar(2012).“PowerCostAdjustment–FrequentlyAskedQuestions.”Accessed

Apr2012.Availableat:http://www.enerstar.com/ms_power_cost_adj.html

EnergyStar(2012).“LightBulbsforConsumers.”AccessedFeb2012.Availableat:

http://www.energystar.gov/index.cfm?fuseaction=find_a_product.showProductGroup&pgw_code=LB

EPA(EnvironmentalProtectionAgency)(2007).“AligningUtilityIncentiveswith

InvestmentinEnergyEfficiency:AResourceoftheNationalActionPlanforEnergyEfficiency.”EnvironmentalProtectionAgency.Availableat:http://www.epa.gov/cleanenergy/documents/suca/incentives.pdf

Espey,JamesA.,Espey,Molly,(2004)“TurningontheLights:AMeta‐analysisof

ResidentialElectricityDemandElasticities,”JournalofAgriculturalandAppliedEconomics,36(1):pp.65‐81

Eto,Joseph,StevenStoft,andTimothyBelden(1994).“TheTheoryandPracticeof

Decoupling.”LawrenceBerkeleyLaboratory.Availableat:http://escholarship.org/uc/item/9gp741dm;jsessionid=5E68872CEE40922AE7945995A9724D59#page‐1

GIllingham,Kenneth,RichardNewell,andKarenPalmer(2009).“EnergyEfficiency

EconomicsandPolicy.”ResourcesfortheFutureDiscussionPaper09‐13April2009

GIllingham,Kenneth,RichardNewell,andKarenPalmer(2006).“EnergyEfficiency

Policies:ARetrospectiveExamination.”AnnualReviewEnvironmentalResources31:161‐92

Glaeser,EdwardL,andMathewE.Kahn(2008).“TheGreennessofCities:Carbon

DioxideEmissionsandUrbanDevelopment.”WorkingPaper14238NationalBureauofEconomicResearchAug2008

53

Graniere,RobertJ,Cooley,Anthony(1994).“DecouplingandPublicUtilityRegulation.”TheNationalRegulatoryResearchInstituteReportNRRI94‐14Aug1994.Availableat:http://nrri.org/pubs/electricity/94‐14.pdf

Hausman,Jerry(1979).“IndividualDiscountRatesandthePurchaseand

UtilizationofEnergy‐UsingDurables.”BellJournalofEconomics,10(1):33–54

Hansen,D.G.(2007).“AReviewofNaturalGasDecouplingMechanismsand

AlternativeMethodsforAddressingUtilityDisincentivestoPromoteConservation.UtahDepartmentofPublicUtilities,DocketNo.05‐057‐T01,DPUExhNo.6.1(DGH‐A.1)

Holland,S.P.,Mansur,E.T.(2008).“IsReal‐TimePricingGreen?:TheEnvironmental

ImpactsofElectricityDemandVariance.”ReviewofEconomicsandStatistics90(3):550‐561

Hurley,Doug,KenjiTakahashi,BruceBlewald,JenniferKallay,andRobinMaslowski

(2008).“CostsandBenefitsofElectricUtilityEnergyEfficiencyinMassachusetts.”SynapseEnergyEconomics,IncFinalReport.August2008.Availableat:http://www.synapse‐energy.com/Downloads/SynapseReport.2008‐08.0.MA‐Electric‐Utility‐Energy‐Efficiency.08‐075.pdf

IPCC(IntergovernmentalPanelonClimateChange)(2007).“ClimateChange2007:

SynthesisReport.”IPCCFourthAssessmentReport:ClimateChange2007.Availableat:http://www.ipcc.ch/publications_and_data/ar4/syr/en/mains2‐1.html

IMF(InternationalMonetaryFund)(2012).WorldEconomicOutlookdata.Accessed

2012.Ito,Koichiro(2010).“Doconsumersrespondtomarginaloraverageprice?Evidence

fromnonlinearelectricitypricing.”UniversityofCalifornia,BerkeleyDraftPaper.Availableat:http://ecnr.berkeley.edu/vfs/PPs/ItoKoi/web/JMP_Koichiro_Ito_UC_Berkeley_2010_1122.pdf

Jaffe,AdamB.,andRobertN.Stavins(1995).“DynamicIncentivesofEnvironmental

Regulations:TheEffectsofAlternativePolicyInstrumentsonTechnologyDiffusion.”JournalofEnvironmentalEconomicsandManagement29(3):43–63

54

Joskow,Paul,andDonaldMarron(1992).“WhatDoesaNegawattReallyCost?EvidencefromUtilityConservationPrograms.”EnergyJournal13(4):41–74

Kahn,MathewE.andErinT.Mansur(2011).“DoLocalEnergyPricesandRegulation

AffecttheGeographicConcentrationofEmployment?ABorderPairsApproach.”NationalBureauofEconomicResearchWorkingPaper16538Jun2011.Availableat:ei.haas.berkeley.edu/pdf/working_papers/WP209.pdf

Kihm,Steven(2009).“WhenRevenueDecouplingWillWork…AndWhenItWon’t.”

ElectricityJournal22(8):19‐28Kotchen,MathewJandMichaelR.Moore(2007).“Privateprovisionof

environmentalpublicgoods:Householdparticipationingreen‐electricityprograms.”JournalofEnvironmentalEconomicsandManagement53:1‐16

Lazer,Jim,FrederickWeston,andWayneShirley(2011).“RevenueRegulationand

Decoupling:AGuidetoTheoryandApplication.”TheRegulatoryAssistanceProject(RAP)Jun2011.Availableat:www.raponline.org/document/download/id/902

LeagueofConservationofVoters(2010).“NationalEnvironmentalScorecard09:

SecondSession111thCongress,”LeagueofConservationofVoters.Availableat:http://www.lcv.org/scorecard/past‐scorecards/pdf/scorecare‐2010.pdf

LeagueofConservationofVoters(2011).“PastScorecards.”LeagueofConservation

ofVoters.Availableat:http://www.lcv.org/scorecard/past‐scorecards/

Lesh,PamelaG.(2009).“RateImpactsandKeyDesignElementsofGasandElectric

UtilityDecoupling:AComprehensiveReview.”ElectricityJournal22(8):65‐71

Linn,Joshua(2008).“EnergyPricesandtheAdoptionofEnergy‐SavingTechnology.”

TheEconomicJournal118(533):1986‐2012Loughran,DavidS.,andJonathanKulick(2004).“Demand‐SideManagementand

EnergyEfficiencyintheUnitedStates”.EnergyJournal25(1):19–41MagaliDelmas,MichaelV.Russo,MariaJ.Montes‐Sancho(2005)."Deregulationand

ResourceReconfigurationInTheElectricUtilityIndustry"UniversityofCaliforniaEnergyInstituteApril2005.Availableat:http://www.ucei.berkeley.edu/PDF/EPE_012.pdf

55

McCarthy,KevinE.(2009).“ElectricRateDecouplinginOtherStates.”OLRResearchReportJan2009.Availableat:http://www.cga.ct.gov/2009/rpt/2009‐R‐0026.htm

McKinsey&Company(2007).“ReducingU.S.GreenhouseGasEmissions:HowMuch

atWhatCost?”US.GreenhouseGasAbatementMappingInitiativeExecutiveReport

MenekenH.L.(2007).“RevenueDecoupling:APolicyBriefoftheElectricity

ConsumersResourceCouncil.”TheElectricityConsumersResourceCouncilJan2007

Michelfelder,Richard(1993).“EvolvingElectricUtilityRegulatoryPolicy:

InternalizingtheSocialCostsofProduction.”TheAmericanJournalofEconomicsandSociology52(2)227‐239.Availableat:http://www.jstor.org/stable/3487062

Moore,ThomasG.(1970).“TheEffectivenessofRegulationofElectricUtilityPrices.”

SouthernEconomicJournal36(4):365‐375Musti,Sashank,KatherineKortum,andKaraM.Kockelman(2011).“Household

energyuseandtravel:Opportunitiesforbehavioralchange.”TransportationResearchPartD1649‐56.]

NARUC(NationalAssociationofRegulatoryUtilityCommissioners)(2007).

“DecouplingofElectric&GasUtilities:FrequentlyAskedQuestions(FAQ).”NationalAssociationofRegulatoryUtilityCommissioners2007.Availableat:http://www.naruc.org/Publications/NARUCDecouplingFAQ9_07.pdf

NationalWeatherServiceClimatePredictionCenter(2012).“DegreeDaysStatistics.”

AccessedFeb2012.Availableat:http://www.cpc.ncep.noaa.gov/products/analysis_monitoring/cdus/degree_days/

Parformak,PaulW.,andLesterB.Lave(1996).“HowManyKilowattsAreina

Negawatt?Verifying‘ExPost’EstimatesofUtilityConservationImpactsattheRegionalLevel.”EnergyJournal17(4):59–88

PSC(DelawarePublicServiceCommission)(2007).“BeforethePublicService

CommissionoftheStateofDelaware,intheMatteroftheInvestigationofthePublicServiceCommissionintoRevenueDecouplingMechanismsforPotentialAdoptionandImplementationbyElectricandNaturalGasUtilitiesSubjecttotheJurisdictionofthePublicServiceCommission”.PSCRegulationDocketNo.59Mar2007

56

PSC(PublicServiceCommissionofWisconsin)(2001).“TypicalStepsinaMajorRateCase.”PublicServiceCommissionofWisconsin.Availableat:psc.wi.gov/thelibrary/publications/general/general04.pdf

PUCO(PublicUtilitiesCommissionofOhio)(2011).“RateCaseProcess.”Public

UtilitiesCommissionofOhioaccessedOct2011.Availableat:http://www.puco.ohio.gov/puco/index.cfm/consumer‐information/consumer‐topics/rate‐case‐process/

Saad,Lydia(2011).“InU.S.,ExpandingEnergyOutputStillTrumpsGreenConcerns.”

Gallup.Mar2011.Availableat:http://www.gallup.com/poll/146651/Expanding‐Energy‐Output‐Trumps‐Green‐Concerns.aspx

Smil, Vaclav (2011), Energy Transitions: History, Requirements, Prospects. Praeger. Stern,N.(2008).“TheEconomicsofClimateChange.”AmericanEconomicReview

98(2):1‐37Train,KennethE.andJudiE.Strebel(1987).“EnergyConservationandRebatesin

CommercialFoodEnterprises.”AmericanAgriculturalEconomicsAssociation69(1):106‐114

57

AppendixA–ModelResultsTable1:

Regressions of Log Residential Energy Consumption Per Customer (1) (2) (3) (4)

OLS 2SLS 2SLS -

Robustness Pre-existing

Trend Decoupling -0.05** -0.01 -0.01 x10 (-2.509) (-0.550) (-0.600) DSM x 0.3411** 0.3711** 0.1012 (2.283) (2.390) (0.964) Decoupling*DSM x -0.3911** -0.4111** x (-1.981) (-2.005) Res. Customers -1.00*** -1.00*** -1.00*** -1.00*** (-405.390) (-407.717) (-407.849) (-743.815) Electricity Price x -1.47*** -1.42*** -0.11 (-3.185) (-3.039) (-0.090)

Natural Gas Price -0.1211 0.0811 0.0611 0.1311

(-0.712) (0.444) (0.355) (0.992) GDP 0.25** 0.26*** 0.24*** -0.13 (2.576) (2.785) (2.788) (-0.679) Pop 0.09 0.07 0.07 0.34* (0.941) (0.716) (0.744) (1.913) Personal Income -0.2711 -0.2211 -0.2211 -0.9411*** (-1.610) (-1.296) (-1.219) (-6.909) EERS 0.01 0.02 x 0.01 (0.794) (1.320) (0.786) Env. Sensitivity 0.0311 0.02 x -0.02 (1.208) (0.928) (-1.209) CDD 0.1111*** 0.1111*** 0.1111*** 0.1211*** (13.443) (13.338) (13.368) (27.607) HDD 0.0211*** 0.0211*** 0.0211*** 0.0411*** (5.463) (5.248) (5.243) (28.215) Year x x x 0.02** (2.394) Year*Decoupling x x x 0.01** (2.082) √ √ √ √

Controlled for Utility

√ √ √ x

Controlled for Time

2.70**

Constant (2.465)

N 44,718 44,656 44,656 21,263 R2 0.986 0.977 0.977 0.983 t-statistics clustered on utility in parentheses *** p<0.01, ** p<0.05, * p<0.1

10Omittedbecauseofcollinearity11Coefficientmultipliedby100

58

Table2:Regressions of Log Total Annual Utility DSM

Expenditures (1) (2) (3) OLS Tobit Quantile Decoupling 0.04 0.05 0.34* (0.074) (0.090) (1.855) Res. Customers -0.01 -0.00 -0.00 (-0.147) (-0.084) (-0.169) GDP 1.86 2.51 -0.11 (0.581) (0.668) (-0.204) Pop 1.04 1.39 1.78* (0.250) (0.303) (1.836) Personal Income -0.09 -0.15 0.00 (-0.578) (-0.788) (0.000) EERS 1.96 2.02 0.72*** (1.415) (1.423) (2.804) Env. Sensitivity 0.01 0.01 0.00 (1.004) (1.194) (0.315) CDD 0.00 0.00 -0.00 (0.379) (0.415) (-0.000) HDD -0.00 -0.00 -0.00 (-0.157) (-0.139) (-0.000) Controlled for Utility √ √

Controlled for Time √ √

Constant -13.98 -21.38 9.01 (-0.413) (-0.533) (1.569) N 1,502 1,502 1,501 (Pseudo) R2 0.901 0.291 0.727 t-statistics in parentheses (clustered on utility for OLS and Tobit) *** p<0.01, ** p<0.05, * p<0.1

59

Table3:First-Stage Regressions of Residential Electricity Price (1) (2) (3) Decoupling 0.00 0.01* 0.00 (0.772) (1.849) (0.718) DSM -0.00 x 0.00 (-0.167) (0.050) Decoupling*DSM 0.1012* x 0.1012 (1.699) (1.648) Res. Customers -0.00 -0.00 -0.00 (-0.516) (-0.463) (-0.559) Natural Gas Price 0.0412*** 0.0412*** 0.0412*** (3.761) (3.608) (3.531) GDP 0.00 0.00 -0.00 (0.082) (0.176) (-0.129) Pop -0.02 -0.02 -0.02 (-0.991) (-1.081) (-1.009) Personal Income 0.0312* 0.0312* 0.0312* (1.881) (1.850) (1.949) EERS 0.00 0.00 x (1.141) (1.128) Env. Sensitivity 0.0112 0.0112 x (1.537) (1.512) CDD 0.00 0.00 0.00 (0.287) (0.295) (3.531) HDD -0.0112*** -0.0112*** -0.0112*** (-5.585) (-5.581) (-5.622) 0.0213** 0.0213** 0.0213**

% Gen Coal*Price Coal (2.459) (2.485) (-3.585)

0.0413*** 0.0413*** 0.0413***

% Gen Gas*Price Gas (9.317) (9.316) (2.394)

-0.0213*** -0.0213*** -0.0213*** % Gen Oil*Price Oil (-4.942) (-4.979) (9.319) -0.0313*** -0.0313*** -0.0313***

% Gen Renew*Price Coal (-3.551) (-3.689) (-4.926)

√ √ √

Controlled for Utility

√ √ √

Controlled for Time

0.13 0.12 0.16

Constant (0.996) (0.925) (1.211)

N 44,656 44,656 44,656 R2 0.900 0.900 0.900 t-statistics clustered on utility in parentheses *** p<0.01, ** p<0.05, * p<0.1

12Coefficientmultipliedby10013Coefficientmultipliedby10,000

60

Table4:

Probit Regressions of Probability of Decoupling (1) (2) (3) Res. Sales per Customer - 2001 0.28*** 0.23***

0.23*** (3.097) (3.333) (2.929)

Res. Customers - 2001 0.18*** 0.20*** 0.19*** (2.672) (2.627) (2.668) GDP per capita - 2001 2.28** 2.80*** 2.29** (2.569) (2.690) (2.575) Personal Income - 2001 -0.15 -0.11 -0.14 (-1.138) (-0.712) (-1.094) EERS - 2001 -0.38 0.59 -0.38 (-1.148) (1.203) (-1.130) Env. Sensitivity - 2001 0.01** 0.01** 0.01** (2.141) (2.087) (2.153) CDD - 2001 0.1414 0.1214 0.1314 (1.265) (0.929) (1.269) HDD - 2001 -0.0114 -0.0114 -0.0114 (-0.330) (-0.136) (-0.319) Electricity Price - 2001 5.07 3.20 5.01 (1.220) (0.653) (1.199) % Gen from Coal - 2001 -0.01 -0.01 -0.01 (-1.581) (-1.254) (-1.591) % Gen from Gas - 2001 -0.01 -0.02* -0.01 (-1.314) (-1.724) (-1.310) % Gen from Renewables - 2001 -0.01 -0.01

-0.01 (-0.822) (-0.718) (-0.839)

Δ15 Res. Sales per Customer 0.56 0.02

x (0.995) (0.288)

Δ Res. Customers x 0.56 x (0.978) Δ GDP per capita x 5.21 x (1.522) Δ EERS x 1.67** x (2.313) Δ Env. Sensitivity x 0.01 x (.721) Δ Electricity Price x 9.01 x (1.247)

Constant -26.24*** -32.68*** -26.42***

(-2.783) (-2.898) (-2.793) N 384 384 384 Pseudo R2 0.347 0.421 0.348 z-statistics in parentheses *** p<0.01, ** p<0.05, * p<0.1

14Coefficientmultipliedby10015Δ=value2005–value2001

61

AppendixB–VariableDescriptionsResidentialelectricityconsumptionpercustomer~Inaccordancewiththe

techniquesofArimuraetal.(2011)andLoughranandKulick(2004),variableson

monthlyresidentialenergyconsumptionpercustomerforeachutilityandyear

werecalculatedfrommonthlyenergyconsumptiondatafromFormEIA‐826and

annualenergycustomerbysectordatafromFormEIA‐861.Foreachutility,year,

andsector,monthlyconsumptiondataweredividedbythecorrespondingnumber

ofcustomersforthatyeartoattainestimatedelectricityconsumptionpercustomer.

Consumptionismeasuredinmegawatthours(MWh).Thisanalysisusesthenatural

logarithmoftheresidentialelectricityconsumptionpercustomer.Thisstudyuses

nominalconsumption,althoughtheuseofrealconsumptionwouldnotalterthe

resultsduetolinearityofthelogarithmicfunction16.

DSM~DataonannualtotalutilityspendingonDSMprograms(measuredin

dollars)wereextractedfromFormEIA‐861.Thisvariableshouldreflectallofa

utility’snominalspending–directandindirect–inagivenyearonDSMprograms.

Tokeeputilitieswith$0DSMexpenditureinthedata,thevariableusedinthe

analysisisthenaturallogarithmof1+nominalDSMspending.

Intheory,theDSMvariableshouldcaptureonlyspendingonresidentialDSM

andshouldbemeasuredperresidentialcustomer.Unfortunately,DSMspending

wasnotbrokendownbysectorformostoftheyearsinthisstudy.Dataonthe

breakdownofDSMexpendituresbyutilityin2010,however,areavailable.In2010,

lessthanhalfofDSMexpenditureswereusedforprogramsintheresidentialsector.

ThisintroducesconsiderablemeasurementerrorinthemeasureofDSM.

DSMspendingdataismissingforsomeutilitiesinsomeyearsinEIAForm

EIA‐861.Thisispresumablybecausemanyutilitieschosetowithholdinformation

ontheirDSMspendingfromtheEIAduringyearswhentheirspendingislowor16log(realconsumption)=log(nominalconsumption/priceindex)=log(nominalconsumption)–log(priceindex)=log(nominalconsumption)–c1.Iflog(realconsumption)=BX+c2,thenlog(realconsumption)=BX+(c1+c2)=BX+c3wherec’sdenoteconstants.Thus,onlytheconstanttermisaffectedbythischoice.

62

nonexistent.Inthispaper,weassumethatutilitieswhodonotsubmitinformation

ontheirDSMspendinghadnoDSMspending.

Preliminaryanalysisexperimentedwiththenaturallogarithmsof1+total

DSMexpenditureand1+DSMexpenditurepercustomer.Bothhaveobviousflaws.

WhichisbetterdependsonthenatureofchangesinutilityDSMspendingovertime.

TheresultswhenDSMisthedependentvariablearesimilarregardlessofchoiceof

measure.WhenDSMisusedasanexplanatoryvariable,preliminaryanalysis

suggestedthattheformermeasurehadgreaterexplanatorypowerandistherefore

thefocusofthisanalysis.

Decoupling~Adummydecouplingvariablewascreatedusinginformationfrom

ACEEEandfromvariousonlinedockets.Thisvariabletakesthevalue1ifthe

specifiedutilityhadanytypeofdecouplingmechanisminplaceinthegivenmonth

andyear,andittakesthevalue0otherwise.

ResidentialCustomers~Annualdataonnumberofresidentialcustomersper

utilitywereretrievedfromFormEIA‐861.Thenaturallogarithmofnumberof

customerswasusedinthisanalysis.

ElectricityPrice~Monthlyutilityelectricityresidentialretailpriceswere

retrievedfromFormEIA‐826.Pricesareinunitsof$/kWh.

RetailNaturalGasPrice~Followingtheliterature,naturalgaspriceisincludedin

themodelsbecauseitisapotentialsubstituteforelectricity.Monthlyresidential

naturalgaspricesbystatewereacquiredfromtheEIANaturalGasPrices

ResidentialPriceDataSeries.Naturalgaspricesareindollarspercubicfeet.

GDP~InkeepingwithArimuraetal.(2011)andLoughranandKulick(2004),data

onannualnominalGDPpercapitabystatewereretrievedfromtheBureauof

EconomicAnalysisfortheyears2001‐2010.GDPismeasuredinthousandsof

63

currentdollars.ThenaturallogarithmofGDPwasusedintheanalysis.Eachannual

valuewasusedforallmonthlyobservationsinagivenstateandyear.

QuarterlyPersonalIncome~Dataonstatequarterlypersonalincomeinmillions

ofnominaldollarswereobtainedfromtheBureauofEconomicAnalysis.The

naturallogarithmofpersonalincomewasusedinthisanalysis.

Pop~AvariableforannualpopulationwascomputedusingannualstateGDPand

annualper‐capitastateGDPfromtheBureauofEconomicAnalysis.Thenatural

logarithmpopulationwasusedintheanalysis.Eachannualvaluewasusedforall

monthlyobservationsofconsumptioninagivenstateandyear.

EERS~AdummyvariableforwhetherornotthestatehadanEnergyEfficiency

ResourceStandardinplaceduringthegivenmonthandyearwasgeneratedbased

ondatafromACEEEandvariousdockets.

EnvironmentalSensitivity~Usingsimilarmethodstothoseofpreviousstudieson

electricityconsumption(Delmasetal.2005,Arimuraetal.2011),anannual

environmentalsensitivityratingwascalculatedforeachstatebasedonscoresfrom

theLeagueofConservationVoter’sNationalEnvironmentalScorecards.Thisrating

averagestheLeagueofConservationVoter‘smeanscoresofthepoliticiansinthe

houseandinthesenateofeachstate.Theratingsareonascaleof0to100andare

basedonhowthepoliticiansvoteonkeyenvironmentallegislation.

CDDandHDD~Population‐weightedmonthlyheatingandcoolingdegreedaysby

statewerecompiledfromtheNationalWeatherServiceClimatePredictionCenter

archiveddegreedaysstatistics.Experimentswithvariousclimateindicatorsand

functionalformsofCDDandHDDsuggestthattherespectivelinearmodelsbest

capturetherelationbetweenclimateandpercapitaenergyconsumption.

64

%GenerationfromOil/Coal/Gas/Renewables~Thesevariablesrepresentthe

percentageofelectricityinthegivenstatethatwasgeneratedin2005usingeachof

oil,coal,naturalgas,andenergyfromvariousrenewableenergysources,

respectively.ThesedatawereretrievedfromtheEIAwebsite.

USOil/Gas/CoalPrices~Theseseriesreflecttheinputpricesfacedbyproducers

ofelectricity.TheserieswereprocuredfromtheEIAwebsite.USoilrefinerprices

andnaturalgascitygatepriceswereavailableonamonthlylevel.UScoalprices

wereonlyavailableontheannuallevel.

AustralianThermalCoalPrice~MonthlydataonAustralianthermalcoalprice,a

prominentcoalpriceindex,weretakenfromtheInternationalMonetaryFund.

PricesareinU.S.dollarspermetricton.

i.Time~Eachobservationofthevariable,time,representsthemonth(ex‐February

2001)whenthecorrespondingconsumptionoccurred.Thisvariablewas

partitionedinto119variables,oneforeachofthemonthsexceptforJanuary2001.

Together,thesevariablescaptureallnationwidetimeandseasonaldiscrepancies.

i.UtilityID~EachutilityintheU.S.hasauniqueidentificationnumber.Dataonthe

variable,UtilityID,wereretrievedfromFormsEIA‐826andEIA‐861.Itdistinguishes

eachutilityinthedatasetsfromotherutilities.Additional,unofficialID’swere

createdforutilitiesthatextendedtomultiplestatestoenabletheseentitiestobe

treatedasuniqueutilities.Onedummyvariablewascreatedforeachutility.For

eachobservation,thevalueofoneofthesevariablesisoneandthevaluesoftherest

arezero.