Download - Wesley W. Wilson University of Oregon CONSUMER TELECOMMUNICATION CHOICES: EVOLUTION OF CELL PHONES
Wesley W. WilsonUniversity of Oregon
CONSUMER TELECOMMUNICATIO
N CHOICES: EVOLUTION OF CELL
PHONES
In the early 1990s, a consumer with a cell phone was rare. Today, it is rare that a consumer does not have a cell phone.
Major Issues: Are cell phones “killing” landlines? Are cell phones a substitute or a complement to
landlines? Problem:
Consumers make a telecommunications choice. They can choose to not have a phone, to have a cell phone (only), to have a landline phone (only), or both.
In this research, I use the Consumer Expenditure Survey to model these choices over an extended time period (1993-2010) as a function of individual and household characteristics.
Outline Background Model Data Results
PROBLEM AND PURPOSE
Early mobile phones were a combination of radio and telephone that connected users to the telephone network. Major quality issues – dropped calls, poor sound quality
Major events: 1973: Dr. Martin Cooper made the first
mobile call from a handheld device 1983: FCC gave permission for commercial
cellular service, and Ameritech entered the market, but very expensive.
1992: First text message sent 2002: 3G network (by 2009 most was
using 3G), much faster than 2G systems. Faster internet access.
BACKGROUND
1942: Walkie Talkie
PICTURES
1942: Walkie Talkie
PICTURES
1946: 1st Commercial Mobile Telephone (Bell)
PICTURES
1973 DynaTAC – Martin Cooper Makes First Private mobile call.
PICTURES
1983 Motorola DynaTac 1 st public
PICTURES
1989: Motorola’s MicroTAC, First Flip Phone
PICTURES
1992: Motorola International 3200. First hand size 2G phone.
PICTURES
1992: Motorola International 3200. First hand size 2G phone.
PICTURES
1999: Nokia 7110 160million sold.
PICTURES
2007: iPhone
PICTURES
Choice model based on four options: No phones (cell or land) Cell phone only Landline phone only Both cell and Landline
Explained in terms of a set of individual and household characteristics
Estimated by a multinomial logit
MODEL
Consumer Expenditure Series (Collected by BLS) Quarterly Interview Series-Contains
information on individuals and households Diary Survey-Contains information on
expendituresThese data were combined from 1983-
2010. The fi rst cellphone expenditure was observed in 1993. 158K observations 18 year time period
DATA
Table 1: Choices over time
Year NoPhone Land Cell Both Total
1993 5.26 93.04 0.04 1.66 100
1994 4.72 90.91 0.06 4.31 100
1995 6.67 87.8 0.1 5.43 100
1996 4.98 84.72 0.07 10.23 100
1997 4.93 82.42 0.11 12.54 100
1998 5.68 78.02 0.2 16.11 100
1999 5.86 74.87 0.24 19.04 100
2000 6.03 67.76 0.53 25.69 100
2001 6.22 50.81 1.8 41.16 100
2002 6.5 44.93 3.88 44.69 100
2003 7.91 44.99 5.5 41.6 100
2004 8.24 43.36 7.78 40.61 100
2005 7.68 38.07 10.05 44.2 100
2006 8.34 33.8 13.5 44.36 100
2007 7.57 27.73 18.28 46.42 100
2008 7.87 26.77 21.23 44.12 100
2009 8.06 22.75 25.56 43.64 100
2010 9.08 21.02 31.7 38.2 100
Overall 6.85 54.1 7.88 31.17 100
Type
Family
SizeReal
IncomeColleg
eMarrie
dUrba
nMale
White Age
None 1.93 20,380 0.75 0.22 0.92 0.52 0.79 36
Land 2.50 35,825 0.68 0.48 0.91 0.51 0.82 49
Cell 2.23 35,568 0.87 0.30 0.96 0.53 0.78 35Both 2.87 64,020 0.91 0.65 0.93 0.51 0.83 47
VARIABLES AND SUMMARY
The model was estimated (using landline only as the base) from 1993-2010 Estimated by year: Most parameters were
relatively stable with the exception of year, which followed a trend like pattern roughly comparable to the results with a straight trend.
Estimated over short time periods (1999-2010). But, this misses some of the important parts of the evolution, yet still roughly comparable to the entirety of the data.
Final model as an annual time trend. Virtually all coeffi cient are statistically
significant Pseudo R2 is .2428 The model is statistically significant with a
chi-square statistic that is very large (83000). Coeffi cient estimates do not imply direction of
a change in variable in a multinomial logit so marginal are typically presented (and are here as well)
ESTIMATION
PARAMETER ESTIMATES
OptionVARIABLES No Phone Cell only Both fam_size -0.237*** -0.160*** 0.0496***
rinc -1.58e-05*** 2.11e-06*** 8.76e-06***
college -0.326*** 0.136*** 0.855***
married -0.426*** -0.368*** 0.417***
urban -0.145*** 0.353*** 0.0838***
male 0.281*** 0.377*** 0.0129
white 0.0218 0.0324 0.0634***
age_ref -0.0532*** -0.0725*** -0.0105***
year 0.191*** 0.528*** 0.216***
Constant -380.7*** -1,058*** -433.6***
Observations 158,868 158,868 158,868
MARGINAL EFFECTS
Table 4: Marginal Effects
No Phone LandOnly CellOnly BothVariable dy/dx dy/dx dy/dx dy/dxfam_size -0.0173 0.0045 -0.0102 0.0230
rinc 0.0000 0.0000 0.0000 0.0000
college -0.0320 -0.1124 -0.0153 0.1597
married -0.0257 -0.0382 -0.0367 0.1006
urban -0.0090 -0.0287 0.0256 0.0121
male 0.0128 -0.0233 0.0238 -0.0133
white 0.0041 -0.0140 0.0002 0.0096
age_ref -0.0024 0.0052 -0.0043 0.0015
year 0.0025 -0.0439 0.0281 0.0133
Three diff erent groups are defined and analyzed over time. Consumer type 1 is a white, male, without
a college degree, unmarried, urban, of twenty years old with an income of $7500 and a household size of 1.
Consumer type 2 is a white, male, college degree holder, married, urban, forty-five year old with an income of $86,000, and a household of four people.
Consumer type 3 is a white, female, college degree holder, unmarried, 75 year olds, with an income of $19,000
Two diff erent years are used (defined) to compare age diff erences in choices.
CONSUMER GROUPS
TRENDS
0.2
.4.6
.8
1990 1995 2000 2005 2010year
No Phone LandCell Both
Twenty Year OldsFigure 1a
0.2
.4.6
.8
1990 1995 2000 2005 2010year
No Phone LandCell Both
Forty-Five Year OldsFigure 1b
0.2
.4.6
.81
1990 1995 2000 2005 2010year
No Phone LandCell Both
Seventy-Five Year OldsFigure 1c
0.2
.4.6
.81
1990 1995 2000 2005 2010year
Twenty Forty-FiveSeventy-Five
Probability of a Cell and AgeFigure 2
0.2
.4.6
.81
1990 1995 2000 2005 2010year
Twenty Forty-FiveSeventy-Five
Probability of a Landline and AgeFigure 3
0.2
.4.6
.81
20 30 40 50 60 70age_ref
No Phone LandCell Both
Probability of Types in 1993Figure 4a
0.2
.4.6
.8
20 30 40 50 60 70age_ref
No Phone LandCell Both
Probability of Types in 2010Figure 4b
0.2
.4.6
.81
20 40 60 80age_ref
1993 2010
Probability of a Cell 1993 and 2010Figure 5
.2.4
.6.8
1
20 40 60 80age_ref
1993 2010
Probability of Landline 1993 and 2010Figure 6
SYNOPSIS
Research points to tremendous diff erences over time in telecommunication decisions of consumers. In 1993, virtually no cell phones in the data In 2010, most have cell phones.
Major fi ndings: Land line phones have followed a secular
decline over all time periods Cell phone only and individuals with both a
cell and land are growing Land lines are falling but primarily for younger
people. Other:
If family sizes grow, more likely to have a landline or both cell and landline, and less likely not to have no phone or cell phones only.
More education leads consumers to choose to have both cell and land line phones
Married people tend to have both a landline and cell phone.