20030418Staff 2nd Resp - attach I -3 Going Mobile .pdf

Telecommunications Policy ](]]]])]]]–]]]
Going mobile: substitutability between fixed and
mobile access$
Mark Rodini, Michael R.Ward*, Glenn A.Woroch
Abstract
We empirically estimate the substitutability of fixed and mobile services for telecommunications access
using a large, US household survey conducted over the period 2000–2001.Estimated cross-price elasticities
confirm that second fixed line and mobile services are substitutes for one another.The extent of fixed-
mobile substitution has important implications for policy toward fixed network unbundling, fixed-mobile
vertical separation, and universal service.
r 2003 Elsevier Science Ltd.All rights reserved.
Keywords:Telecommunications; Access; Mobile; Demand; Substitution
1. Introduction
The spectacular growth of mobile phone service is well documented, with mobile users now
exceeding one billion worldwide.1 By any measure, mobile service has reached rough parity with
fixed line service in most regions, and surpasses traditional fixed-line penetration in developed and
developing countries alike.Yet fixed lines have not been significantly displaced.While the total
number of fixed lines has been flat in recent years, with net contraction in selected pockets,
supplemental fixed lines continue to grow fueled by Internet and other data access needs.Overall,
significant changes in the composition of households’ access portfolios have taken place.
These developments invite analysis of the substitutability of fixed and mobile services as a
means of access.The two services provide many of the same basic functions (dialtone, phone
number, long distance access), and their price levels and structures have converged over time.
ARTICLE IN PRESS
$Prepared for the Conference on ‘‘Competition in Wireless: Spectrum, Service and Technology Wars,’’ organized by
the Public Utility Research Center at the University of Florida, Gainesville, February 2002.
*Corresponding author.Department of Economics, University of Texas at Arlington, 330 Business Building,
Box 19479—UTA, Arlington, TX 76013, USA.
E-mail address:[e-mail] (M.R. Ward).
1ITU (2002).
0308-5961/03/$-see front matter r 2003 Elsevier Science Ltd.All rights reserved.
doi:10.1016/S0308-5961(03)00010-7
Important differences between the two remain, however, with wireless service distinguished by its
mobility and fixed service boasting superior transmission quality and bandwidth.
This paper provides empirical evidence of the degree of substitution between fixed and mobile
telephone services.We employ a data set constructed from a US consumer panel that contains
detailed information on households’ communications purchase behavior.These data, reported
over a 2-year period, include information about a household’s subscription to fixed and mobile
telephone service as well as usage of those and other communications services.Since the data set
does not provide fixed and mobile prices, we develop a two-stage estimator of expected prices for
access and usage of each service.We then determine how consumers’ subscription decisions vary
with the (estimated) mobile and second fixed line prices to calculate own- and cross-price
elasticities of demand.
In general, we find that mobile access demand is elastic with respect to its own prices when
compared to the range of typical elasticity estimates for fixed line service.We also confirm
previous research that households’ subscription to second fixed lines are much more price elastic
than primary lines.Estimated cross-price elasticities confirm that second fixed lines and mobile
service are access substitutes.In particular, we find a significant response in mobile subscription to
fixed line rates.
These results can inform the regulatory debate over local telephone service competition.2
Increasingly, US telecommunications policy has moved away from direct government oversight of
firms’ pricing decisions to greater reliance on competitive forces.A finding that mobile service
offers a close substitute for fixed service would serve to undermine the rationale for regulation of
fixed local services, and would bolster efforts to expand wireless alternatives, possibly by relaxing
spectrum caps.Requirements of interconnection and unbundling of incumbent networks that
represent the heart of the 1996 Telecommunications Act become less justified, and possibly
harmful, when a viable, facilities-based alternative such as wireless mobile constrains fixed
network providers.A close substitute for fixed service would also bring into question the need for
safeguards against various kinds of anti-competitive conduct.The rationale for structural or
functional separation aimed at preventing ‘‘discriminatory dealing’’ between fixed and mobile
divisions of an integrated carrier begin to fade.3 Moreover, when consumers subscribe to mobile
service in place of fixed access, subsidies paid only to fixed-line carriers to promote universal
service are unjustified.In general, support targeted exclusively toward fixed-line penetration may
not be furthering the goal of increased connectivity.
This paper is the first to our knowledge that examines fixed-mobile access substitution using
household level data.The data also permit us to examine patterns over time to some extent.Our
approach to estimating prices of the two services appears to be novel yet appropriate to the
empirical questions we seek to answer.
2. Background on fixed-mobile substitution
Historically, mobile telephone service did not pose an attractive alternative to fixed service.
Given its high relative price, cellular service was truly a luxury, not a substitute for fixed line.
ARTICLE IN PRESS
2For a more detailed discussion of fixed-mobile competition and its policy implications, see Woroch (2002).
3See Mini (2001)and Reiffen, Schumann and Ward (2000).
M. Rodini et al. / Telecommunications Policy ](]]]])]]]–]]]2
Cellular technology also lagged significantly in non-price terms: transmission quality and
geographic coverage were poor by fixed-line standards; early cell phones were cumbersome and
not at all portable.These differences resulted in cellular’s limited penetration rates when first
introduced in the 1980s.Since those early days, mobile service adoption has grown at astonishing
rates as the quality of the service and the performance of mobile providers steadily improved.
The spread of mobile services comes at a time when telecommunications authorities and the
public are concerned over the lack of speed in the development of competition in local services.
Unbundling and resale of network services to facilitate entry has not produced the competition
originally envisioned by legislation such as the 1996 Telecommunications Act in the US; nor have
competitive carriers overbuilt incumbent networks to any great extent.While large business
customers in dense city centers may be able to choose between providers, most smaller businesses
and residents have few facilities-based options available.The slower-than anticipated pace at
which fixed-line incumbents have relinquished market share to competitors leaves mobile services
as their most immediate and potent competition.
The promise of fixed-mobile competition turns, in part, on whether, under prevailing market
demand and cost conditions, mobile alternatives prevent fixed-line carriers from exercising market
power.As costs of mobile telephony continue to drop, allowing prices to fall and quality to rise,
mobile becomes an increasingly attractive alternative to fixed-line service.Technically, mobile is a
substitute because users can place and receive voice calls just as they do with fixed service.The
relevant questions are whether consumers shift minutes from their fixed line(s) to their mobile
phones, and whether they meet their expanding access needs with mobile lines or even disconnect
their fixed lines and replace them with mobile phones.
So far, the incidence of users who forgo fixed service entirely and depend completely on mobile,
are few in number.The profile of the typical wireless-only user is young and single.In its last two
annual reports on the wireless industry, the FCC concluded from its review of thirdparty research
that about 3% of US users rely on mobile as their only phone service, while about 12% reported
that they purchased mobile service rather than adding a supplementary fixed line.4 Two recent
surveys of British residential consumers, conducted a year apart, both found that 5–6% of
individuals claim mobile as their only source of telephone service in their home.5
An alternative view is that fixed and mobile services are complementary.Mobile service enables
calls that were otherwise impossible, as when a user is traveling in a car or walking on the street.
These and other mobile calls typically originate or terminate on a fixed line, so that mobile usage
could stimulate fixed-line usage.6 Also, even new calling that never leaves the mobile network
could nevertheless trigger return calls from fixed lines that otherwise might not occur.7
The academic literature containing econometric analysis of the demand for mobile
communications is limited.Most of the contributions rely on data aggregated to the metropolitan
or country level.Hausman (1999, 2000)estimates the elasticity of aggregate subscription to
cellular service in the 30 largest US markets over the period 1988–1993.Ahn and Lee (1999)
ARTICLE IN PRESS
4FCC (2001, 2002).
5Oftel (2001, 2002).
6A household may make calls between its home fixed line and mobile phones carried by members while away from
home.
7See Taylor (2002).
M. Rodini et al. / Telecommunications Policy ](]]]])]]]–]]]3
estimate demand for mobile access in Korea using more recent wireless subscription data for 64
countries.
Only recently has research appeared that examines fixed-mobile substitution.Sung, Kim and
Lee (2000)find that the number of Korean mobile subscribers is positively correlated with the
number of fixed-line disconnects, but negatively related to the number of new fixed-line
connections, suggesting net substitution between the two services.This pattern occurs even while
the stock of fixed lines is positively correlated with the number of mobile subscribers, offering
evidence that the two services are complements.8
A more recent, and in some ways a more developed, body of work exists on demand for
secondary fixed lines.Eisner and Waldon (2001)and Duffy-Deno (2001)examine the
determinants of household subscription to supplemental fixed lines in the US in the late 1990s.
In an earlier study,Solvason (1997)used a 1994 Canadian household survey to estimate the choice
among three options: no fixed service, a single line and multiple lines.All of these investigations
were made possible only with access to large, detailed household surveys that recorded the
number of fixed lines as well as individual demographic information.
3. Modeling household access and usage decisions
The decision to subscribe to fixed and mobile services is part of a larger telecommunications
consumption problem.To model this problem, we take the household as the decision maker.9
Given various prices and income, the household chooses a portfolio of the two kinds of lines, and
then chooses how much of each line to use so as to maximize household utility.10 When evaluating
portfolios of lines, households derive value from the connectivity and mobility they offer.
Connectivity refers to the two-way bandwidth made possible by the line.Mobility is gauged by the
size of the region over which dial tone is available, as well as the ground speed at which the
connection can be maintained.
Utility derived from subscription to either service derives from the use of the service.We do not
separately model the usage decision, but rather include usage prices in the reduced form demand for
access.Motivation for this specification was first given in the context of a logit model by Perl (1978).
We follow the standard demand specification for subscription decisions in which demand for
these goods depends on their own price, the prices of substitutes and complements, and on
household income, plus attributes of the household and the characteristics of the various
alternatives.Mobile subscription usually involves a monthly charge, possibly over the contract
period that may run a year or two.11 Mobile usage is charged by the minute, or some fraction
ARTICLE IN PRESS
8Ahn and Lee (1999)also find evidence of complementarity also using aggregate data.
9While mobile telephony is a single-user technology, whether to subscribe to a mobile service is likely to be part of a
household’s overall budgeting problem.Note also that answers given by survey respondents may not represent
accurately the preferences of the entire household or the household’s designated decision maker–a weakness of the
survey procedure used to collect information in our data set.
10See Taylor (1994)for a review of related studies modeling both of subscription and usage decisions.
11Ideally, subscription to either service will include one-time expenditures required to initiate service.Fixed service
involves a ‘‘connection fee’’ and mobile service an ‘‘activation fee.’’ Both will require some investment in terminal
equipment.Even prepaid mobile service which eliminates subscription fees altogether requires a handset.
M. Rodini et al. / Telecommunications Policy ](]]]])]]]–]]]4
thereof, with additional usage-sensitive charges for long distance and international service and
when the user is ‘‘roaming’’ outside her home calling area.These usage fees depend crucially on
the boundaries dividing the home zone and the roaming region.
A complication of mobile pricing (and fixed service as well) that raises empirical issues is that
users select from a menu of calling plans, each one with a different monthly charge, a different
amount of included ‘‘free’’ minutes, and a different per-minute charge for usage exceeding the
allowance.Furthermore, mobile prices vary across providers within each market, and similarly for
fixed-line services as residential competition emerges.12 To simplify, we assume that consumers
make their choice of calling plan based on their expected calling volume for both services in their
respective market.The challenge is to summarize ‘‘the price’’ of mobile service along a single
dimension.
Summarizing prices as a single access and per-minute usage price oversimplifies much more
complex pricing schedules.Ideally, one would want data on the price plan details to which
the household has subscribed, in addition to the set of alternatives available at the time of the
subscription decision.Without this, an alternative approach might be an attempt to derive a set of
two-part tariffs that would vary by household usage ‘‘type.’’Miravete (2002)assumes such a
structure in determining how a monopoly supplier should design its pricing schedule.Miravete
also argues that, at any point in time, subscribers should not be expected to be consuming at the
optimum for the plan they chose ex ante; that is, exactly on the lower envelope of a continuum of
two-part tariffs.Our approach is consistent with this, and assumes that while the correct plan was
chosen, the optimum usage may not be observed on a particular bill, for a household, and out of
computational convenience, we approximate the variation in household usage and expenditures as
having derived from a single two-part tariff.Clearly, a more detailed treatment of the pricing
complexity would be preferable.
It is probable that non-price quality improvements, rather than price reductions, played a major
role in the increase in mobile adoption rates.Non-price characteristics of mobile services can vary
considerably, and they should be taken into account.Coverage area has always been a critical
aspect of mobile service–in the early days when the networks were first built, as well as today in
terms of digitalization and 3G capabilities.Moreover, local areas exist where calling volume
demanded currently is beyond the capacity of the infrastructure or where transmission is
obstructed.These ‘‘hot spots’’ and ‘‘dead zones’’ are largely determined by the number and
placement of base stations.In addition, with new digital transmission and signaling technologies,
the feature set of mobile phones has expanded considerably, the size and weight of the handset has
fallen, and the battery life has been extended.
As mentioned above, the price and availability of fixed-line alternatives will condition mobile
demand.Besides basic local and long distance services, a household may have the choice to use a
ARTICLE IN PRESS
12We also do not take account of bundled vertical services (e.g., call waiting, voice mail) which vary across plans, or
of bundled minutes which vary by peak and off-peak periods.One discrepancy stems from the fact that, in the US, both
incoming and outgoing mobile calls are charged, or at least deducted against any usage allowance, whereas only
outgoing calls are priced for fixed line service.Consequently, relative price of mobile service will tend to be
underestimated because fixed line usage is underestimated.This bias will tend to disappear as mobile service moves to
flat rates.One additional discrepancy would arise if a user’s employer paid for all or part of a respondent’s fixed or
mobile bill.We believe that it is unlikely that a household will submit a bill in response to the survey if this were the
case.
M. Rodini et al. / Telecommunications Policy ](]]]])]]]–]]]5
work phone (for both voice and Internet), cable telephony, or considerably less often, fixed
wireless access.Options that specifically deliver some degree of mobility are paging services and
public payphones.
Increasingly, demand for data transmission is driving individuals’ access and usage choices.
Today, dial-up access over traditional phone lines is the predominant means to access the
Internet.Often a household will dedicate a second fixed line for this purpose.13 The introduction
of high-speed access, especially digital subscriber line (DSL) and cable modem services, however,
likely reduces the demand for these lines.
Demographic characteristics of a household likely play a role in determining its communication
needs.The number and age distribution of household members, their education levels,
employment status and occupations all may be important.Women, for example, mention
security as an important reason for subscribing to mobile services even if they rarely use it.One
would also expect that individuals who telecommute or run a business from home would have
higher demand for fixed-line service.In contrast, workers who face long commute times are
predisposed to subscribe to mobile service.
4. Data description
The main source of data in our analysis are the Bill Harvesting data from TNS Telecoms
ReQuest Market Monitors along with its survey responses.14 This quarterly sample of US
household consumption of various telecommunications services is derived from a large national
panel.Participating households are asked to submit one set each of their original bills for local,
long distance, cable TV, cellular and Internet services.Besides summary information, the data set
extracts detailed call information from each ‘‘harvested’’ phone bill.While these data were first
collected in 1995, this paper uses data from the 2-year period January 2000–December 2001
during which a uniform sampling method and survey instrument were employed.While
constructed from a subset of a panel, the Bill Harvesting data do not themselves constitute a
panel.15 Finally, household demographic information was also collected as well as responses from
survey questions.
Table 1 provides overall summary statistics for the 2000–2001 ReQuests data.We limit our
analysis to the 294 urban areas in the US for which we have topographical and population data
we use to determine prices below, but report summary statistics for the complete sample as well.
Fig.1 depicts the subscription to mobile service, pagers and second fixed lines derived from the
ReQuest survey over this period.The percentage of households with mobile service increased
from just under 45% in 2000 up to over 56% in 2001, while second fixed line subscriptions
remained flat and pager use fell on net.These figures match closely the growth rates reported by
the CTIA (2002)and by other nationwide data sources.
Fig.2 plots responses from a ReQuest survey question asking mobile subscribers why they
subscribed to mobile service.While the fraction of the population that indicates that it was at the
ARTICLE IN PRESS
13Connection to a facsimile machine is another motivation for a second fixed line that is increasingly less common.
14See owww.tnstelecoms.com>.
15About a quarter of all survey respondents actually submitted their phone bills, and of these households, about 10%
were re-sampled during the 2000–2001 period.
M. Rodini et al. / Telecommunications Policy ](]]]])]]]–]]]6
ARTICLE IN PRESS
Table 1
Summary statistics
Variable Complete sample ‘‘Urban’’ sample
Mobile subscription 50.6% 53.0%
Second fixed-line subscription 20.0% 21.5%
Income o $25K 26.0% 23.6%
Income $20–$25 8.5% 7.9%
Income $25–$35 13.5% 13.3%
Income $35–$50 17.0% 17.2%
Income $50–$60 8.9% 9.3%
Income $60–$75 9.2% 10.1%
Income o$75 9.2% 10.1%
HH Size: single 24.8% 25.2%
HH Size: two 32.6% 32.5%
HH Size: three 16.9% 16.7%
HH Size: four 15.4% 15.5%
HH Size: five or more 10.3% 10.1%
Children present 37.4% 36.6%
Female headed 29.1% 29.8%
Married 55.6% 54.9%
Race: White 87.0% 85.8%
Race: Black 6.7% 7.4%
Race: Asian 3.8% 4.2%
Race: Other 2.4% 2.6%
Age: under 25 7.6% 7.0%
Age: 25–34 21.4% 20.9%
Age: 35–44 24.8% 24.9%
Age: 45–54 17.5% 18.2%
Age: 55–64 12.1% 12.3%
Age: 65 or Over 16.6% 16.7%
Educ HH Head: grade school 16.3% 14.6%
Educ HH Head: some high school 35.3% 33.4%
Educ HH Head: HS graduate 21.9% 23.0%
Educ HH Head: some college 17.5% 19.0%
Educ HH Head: college graduate 9.1% 10.0%
Personal computer 60.3% 62.5%
Internet access 49.9% 52.7%
Modem 38.0% 40.2%
Pager 16.0% 17.7%
Fax machine 12.4% 13.5%
Answering machine 59.5% 60.3%
Frequently brings work home 22.9% 23.8%
Home–based business 13.6% 13.3%
Telecommutes 6.8% 7.3%
Number of observations 327,920 230,684
M. Rodini et al. / Telecommunications Policy ](]]]])]]]–]]]7
expense of fixed service increases over time (rising from 7.1% to 11.2%), that fraction remains
quite small and is consistent with data reported by the FCC for the US and by OFTEL for the
UK.
While the data set contains access fee and usage charge breakdowns, these are drawn only from
what is explicitly recorded on the bill.The Bill Harvesting data set includes neither prices of the
subscribed services, nor prices for services not chosen.Moreover, since bills do not indicate the
commencement date of the service, and because companies differ in their billing conventions, it is
not possible to back out the ex ante subscription and usage charges that households face when
making their subscription and usage decisions.This is true for fixed service as well as for mobile.16
ARTICLE IN PRESS
0%
5%
10%
15%
20%
25%
30%
35%
40%
45%
50%
55%
60%
2000:1 2000:2 2000:3 2000:4 2001:1 2001:2 2001:3 2001:4
Wireless Paging Second Line
Fig.1. Percent of Household with service.
0.0%
1.0%
2.0%
3.0%
4.0%
5.0%
6.0%
7.0%
2000:1 2000:2 2000:3 2000:4 2001:1 2001:2 2001:3 2001:4
Replace Existing Home Phone Instead of Home Phone As only Phone
Fig.2. Why mobile was purchased.
16Some of the earliest local access demand studies controlled for the availability of flat and measured service to a
household.See Perl (1978).
M. Rodini et al. / Telecommunications Policy ](]]]])]]]–]]]8
Below, we spell out how expected marginal prices for these services can be estimated from the
information contained in the bills.This is done for both fixed and mobile services.
5. Specification
5.1. Fixed and mobile access and usage prices
Our primary goal is to estimate the degree to which consumers’ decisions to subscribe to mobile
service and supplementary fixed lines are determined by the prices of both services.Pricing of the
two services—especially mobile service—tend to be highly complex, with carriers offering multiple
calling plans with non-uniform prices that vary over time and across region.Furthermore, many
households will have subscribed to either service sometime in the past under a rate plan that may
not even be available to new subscribers during the sample period.
The enormous complexity of the pricing options available to households in an area necessitates
that we simplify.For this reason, we estimate prices using the empirical relationship between fixed
and mobile bills and the observed usage levels of each service.We specify a random coefficients
regression to estimate a representative two-part tariff for each geographic region, where fixed-line
prices vary by state and mobile prices vary by cellular franchise area.17
For mobile service, we regress the amount of the monthly bill submitted by households (after
deducting equipment, features and associated tax charges) on a set of shifters,zjk, which likely
affect pricing across regions, as well as on these same measures interacted with the minutes of use
reported on the bill:
Billi ¼
X
k
ðak þ ekÞzjk þ
X
k
ðdk þ mkÞzjk Minutesi;ð1Þ
where i indexes the household,j indexes the geographic area where household i resides, and k
indexes the shifter.We have included additive random error terms,ek and mk, on the coefficients,
ak and dk, resulting in a random coefficients specification.18 The dependent variable is the
observed bill net of equipment and features charges.19 The shifters include many direct and
indirect measures of cost of building and operating mobile phone systems in a geographic area, as
well as a constant term and a time trend.
Using the estimated regression coefficients, we take as the access and usage prices in market j
the fitted values for the intercept and slope of the regression in (1):
Ps
Aj ¼
X
#akzjk;Ps
Uj ¼
X #dkzjk;
ARTICLE IN PRESS
17Cellular Franchise Areas, also known as the Mobile Service Areas and Rural Service Areas are geographic
delineations assigned by the FCC to award mobile service licenses.We limit our analysis to the largest 294 Mobile
Service Areas that were designed to roughly correspond with Census SMSAs.
18We presume that these errors have mean zero and are uncorrelated with the regression error.
19Households with excessively large bills relative to recorded minutes were dropped from the wireless price
estimation, but as with fixed line, were assigned estimated prices corresponding to their region of residence for demand
estimation.
M. Rodini et al. / Telecommunications Policy ](]]]])]]]–]]]9
where s = f and m refer to fixed and mobile service, respectively.A different set of prices is
calculated for both 2000 and 2001 to be used in separate demand specifications for each year.
For fixed line prices, we follow the same basic approach as with mobile prices, with some
exceptions.First, since bill data exists for access and usage separately, we estimate access and
usage components in (1) separately.The difficulty of disentangling usage charges from access on
wireless bills for a large number of observations prohibits estimating wireless access and usage
separately.Second, we replace the set of shifters,zjk, with state dummy variables corresponding to
the residence of the subscriber.The state dummies reflects continuing statelevel regulatory
oversight of fixed local access and local toll prices.20 Thus, we assume all households in state k
face the same fixed-line access and local toll usage prices.21
For mobile prices, we include continuous supply variables that vary across geographic markets.
We chose this approach instead of simply using regional dummy variables for several reasons.
First, since mobile retail prices are not regulated, they are more likely to reflect supplyside factors
other than regulatory differences.Second, data are available by cellular franchise area that proxy
for cost differences.Third, with fewer observed mobile subscribers and more geographic areas, we
observe relatively few household bills in some geographic markets.In this case, it is more likely
that idiosyncratic household usage patterns for an outlier will affect estimated prices than
underlying cost differences as measured by the dummy variable approach.A summary of these
variables and their sources is reported in Table 2.
The choice of using supply-side measures to estimate mobile prices reflects the problem of
endogeneity of prices in the main demand estimation.Actual price variability will reflect both
demand and supply factors.Thus, to identify exogenous supply prices from consumers’
willingness to pay, we predict prices using zjk variables that are correlated with them, but that are
likely uncorrelated with household-level demand determinants.
Variables included in the zjk matrix include five land topography indicators, ranging from areas
characterized as ‘‘flat plains’’ to ‘‘mountainous,’’ where costs would be expected to be higher due
to the need for more cell towers.Additional cost determinants we include are the population
density, a measure of average marginal tax rates, and the number of months a cellular market has
been in operation.Other included variables that affect the suppliers’ prices are the number of
mobile providers in the market, a measure of industry concentration (i.e., HHI), and a survey
measure of perceived cellular quality.The regional industry structure variables reflect competitive
ARTICLE IN PRESS
20I some cases fixed local prices vary within a state but we believe that this variation is minimal compared to other
potential sources of measurement error, such as the unobserved choice of local calling plans.Furthermore, intrastate
differences are usually due to partitioning the state between an RBOC and an independent.Since RBOCs dominate
large metropolitan areas, and since our analysis is limited to urban areas, we believe that state dummy variables
adequately capture the main sources of variation.
Secondly, it is debatable whether fixed local prices reflect actual costs of provision.We take the position that the
primary determinants of price vary at the state-level, usually based on cost or regulatory conditions.Therefore, state
dummy variables should adequately capture cross-sectional price differences.We thank an anonymous referee for
raising this issue.
21Household with reported negative access fees were dropped from the fixed line regressions, although they were
assigned prices corresponding to their state of residence for access demand estimation.
M. Rodini et al. / Telecommunications Policy ](]]]])]]]–]]]10
pricing pressures, and the quality measure, derived as geographic market averages of wireless
service satisfaction rating of households, reflects the cost of providing quality service.22
Though not included in the zjk matrix, we add a third component to (1) in the estimation of
wireless prices by including indicator variables for heavy mobile use households to control for
possible additional promotional discounts to new subscribers that may be present, but
unobservable, in the bill.Thus, the non-variable component of the predicted mobile access
fee—the fitted coefficient on the intercept—includes the influence of these potential rewards for
choosing wireless access.
One concern with how we estimate prices is that they were based on observed bills rather than
on the actual choices faced by consumers at the time of the access decision.23 In reality, mobile
service plans typically include some amount of ‘‘free minutes’’ with a per-minute charge for each
additional minute beyond the allowance.In this case, our estimate of the usage price derives from
two margins.The first is the ex post probability that a household exceeds its ‘‘free minutes’’ times
the per-minute fee for additional minutes.The second relates to a household’s ex ante choice of
calling plan.Typically, consumers can choose from a menu of plans that tend to offer greater
usage allowance for a higher monthly access charge.The difference between the prices and the
expected usage for two plans represents a shadow price of usage.Miravete (2002)provides
evidence that consumers typically make correct ex ante choices of calling plan when choosing
between flat rate and measured local service, and that they learn from their mistakes.This implies
that deriving the ex ante prices based on ex post observed expenditures leads to a fair
approximation of the actual effective choice they faced in deciding between either wireless or a
second fixed line.
An implication for the estimation of (1) for mobile prices is that the access decision was likely
based on expected usage over the contract period which could differ from actual usage.As a
result, so long as a customer does not exceed his ‘‘free minutes,’’ actual usage is irrelevant to the
ARTICLE IN PRESS
Table 2
Variables and sources used in estimating mobile prices
Variable Source
Population density US Gazetteer Place and Zipcode Files (1990), zipcode table
Land surface topography USDA Economic Research Services Natural Amenities Index (2001)
Marginal tax rate State and Local Taxation of Cellular Industry,CTIA (2002)
Months of operation Cellular Telephone Industry Association, CTIA (1991)
Number of firms ReQuest Survey,TNS Telecoms (2000–2001)
HHI ReQuest Survey,TNS Telecoms (2000–2001)
Quality of service ReQuest Survey,TNS Telecoms (2000–2001)
Billing Quarter ReQuest Survey,TNS Telecoms (2000–2001)
22Other factors driven by strategic concerns such as penetration pricing, and the desire to build a national footprint
also affect wireless prices, and would be necessary components of a structural model of wireless price setting.Yet, while
we do include some structural measures that could proxy for strategic response, we do not model price setting behavior
of fixed and mobile providers.Our main requirement is to identify cross-sectional price variation independent from
factors that affect demand at the household level.
23Unfortunately, besides the fact that we do not observe the actual choices households faced, we also cannot observe
the date of initial subscription.
M. Rodini et al. / Telecommunications Policy ](]]]])]]]–]]]11
bill amount.However, the bill amount will depend on the number of bundled ‘‘free minutes’’
which in turn depends on the expected calling volume.The implicit usage price is thus a function
of consumers’ expected minutes used.In addition, since minutes demanded are a function of
mobile prices, endogeneity is likely to bias estimates of a and d.Therefore, we use a two-stage
estimator, where minutes are first regressed against a set of variables related to customer calling
volume.The resulting predicted number of minutes is both free of differences in supply conditions
and represents the rationally expected calling volume conditional on the available demographic
information.The projection of minutes on these demand shifters is then used as an instrumented
value for Minutesi in (1).The demand shifters include the demographic variables listed in Table 1.
We estimate (1) using least squares weighted by sample weights.24 Because the error structure
allows for heteroskedasticity, the estimation also makes the Huber–White correction to the
variance matrix.While the estimation exhibits is plagued by low predictive power common to
cross-sectional analyses (R2 values in the 0.30–0.36 range),F-tests always reject that the
coefficients are not jointly statistically significant.Table 3 reports the averages of the estimated
marginal prices for 2000 and 2001.
5.2. Demand specification for secondary fixed and mobile access
We specify a household’s decision whether to subscribe to a second fixed line or to mobile
service using a binary logit model, estimated separately from the decision to subscribe to the
other.The basic estimating equation for each service,s =f or m, is as follows:
ps
i ¼Pr Uiðsubscribe to sÞ >Uiðnot subscribe to sÞ½ :ð2Þ
where Ui represents household i’s utility derived from an alaternative.In this specification, the
probability of household i subscribing to service s depends on the access and usage prices for both
services, as well as on a matrix of household and market characteristics,xi.A household will
subscribe if the utility of subscription exceeds the utility of non-subscription.Under the logit
ARTICLE IN PRESS
Table 3
Estimated access and usage prices
Variable 2000 2001
Mean Std.dev. Mean Std.dev.
Mobile access price $23.16 (3.81) $24.39 (2.32)
Mobile usage price $0.11 (0.031) $0.08 (0.010)
Number of HHs 3677 5204
Fixed access price $32.73 (4.50) $34.10 (4.51)
Local toll usage price $0.08 (0.044) $0.08 (0.055)
Number of HHs 33,827 32,142
24We used sample weights provided by TNS, Inc.that are based on age, income and location of residence, and are
intended to make the survey data more representative of a random cross-section of US households.
M. Rodini et al. / Telecommunications Policy ](]]]])]]]–]]]12
specification, the probability of subscription is expressed as
log ð1 ps
i Þ=ps
i
 ¼bs
AmPm
Ai þ bs
UmPm
Ui þ bs
Af Pf
Ai þ bs
Uf Pf
Ui þ gsxi:ð3Þ
Note that although the logits specify the probability of choosing access, usage prices enter
because, as previously discussed, the access choice is dependent on expected usage.The two logits
we estimate—one for fixed service and one for mobile—could be stochastically dependent through
correlation in the error terms.Based on a check for this stochastic dependence, we treat the two
decisions are independent below.25
The explanatory variables,xi, include household income and size, the head of household’s race,
age, education, marital status and work habits, as well as the presence in the home of such
additional services like Internet access and a fax machine.Table 1 lists the household-level
controls included.Market-level controls include average daily commute times, average annual
vehicle miles, and an average measure of self-reported quality of existing mobile and fixed-line
service within a geographic area.
6. Estimation
We estimate demand for each of two years: 2000 and 2001.Given the spectacular growth in
mobile subscription and usage over this period, our presumption is that demand behavior may
likely vary across the 2 years.With growth in subscriber base, we expect that new mobile
subscribers will tend to be more price sensitive than those who have previously adopted the
technology.At the same time, it is possible that continuing subscribers have incorporated mobile
phone service more tightly into their daily routines and became more inelastic in the process.In
the end we will find that, in most cases, average demand behavior does not change even while
massive changes occur in the nature of the customer base.
6.1. Mobile access
Averages of household elasticities that were computed using the mobile access logit estimations are
given in Table 4;26 estimates of all the coefficients are reported in Table 5.We find that the likelihood
of a household subscribing to mobile service is decreasing in access charges for mobile service and
increasing in fixed line access charges.We also find own-price elasticities of mobile access demand
with respect to the monthly recurring charge of 0.43 for both years, that are both highly statistically
significant.As is typical, usage elasticities are not consistently estimated.The overall elasticity of
access and usage price increase is 0.60 for 2000, which falls in the range of other demand elasticities
estimated using aggregate data from earlier time periods and other markets.27
ARTICLE IN PRESS
25We estimated a bivariate probit for the two subscription decisions on the same sample and found negligible impact
on estimated elasticities, and a positive but small correlation between error terms of the two probits.
26Following Train (1986), we calculated these as the mean elasticity across consumers evaluated at their predicted
probability of subscribing.
27Hausman (1999)reports a price elasticity of subscription of 0.51 for cellular subscription in the 30 largest US
markets over the period 1988–1993.Ahn and Lee (1999)estimate subscription as a function of average connection,
recurring and usage changes using aggregate country-level data.While they do not report price elasticities, it is possible
to infer probit own-price elasticity of the monthly recurring charge equal to 0.36.
M. Rodini et al. / Telecommunications Policy ](]]]])]]]–]]]13
The cross-price elasticity of fixed access price on mobile demand is positive and statistically
significant: 0.18 for 2000 and 0.13 for 2001, and again these are estimated with a high degree of
precision.These results indicate that, all else held constant, consumers are more likely to subscribe
to mobile service in areas with relatively higher fixed-line rates and/or relatively lower mobile
rates.We conclude that mobile service is a substitute for fixed-line access, although this could be
due to substitution out of secondary lines.
6.2. Second fixed line access
We also estimate the demand for a second fixed line using the logit model described by (3).
Average price elasticities for second fixed line subscription are reported in Table 6; detailed logit
coefficient estimates can be found in Table 7.Table 6 shows that second fixed line subscription is
fairly elastic with demand elasticities of –0.69 and –0.65. Compared with previous elasticity
estimates for the first fixed line, reported to be much less elastic than –0.10, these results indicate
that second lines are more own-price elastic, as would be expected.However, our second-line
elasticity estimates are somewhat larger than estimates of own-price elasticities for second fixed
lines in studies found in the literature.28 We also find positive crosselastic effects from mobile
prices of 0.26–0.22, although these are not statistically significant at the usual confidence levels.
Our data are not rich enough to estimate the cross-elastic effect from wireless price changes on
the decision to subscribe to any fixed line.Only about 1–2% of our sample does not subscribe to
fixed line service.However, demand theory implies that the cross-elasticity from mobile price
changes to fixed subscription is very likely to take on the same sign of the effect from fixed
price changes to mobile subscription.29 By the Slutsky equation, budget-share weighted cross-
price elasticities should be nearly equal,
WiZijEWjZji:
ARTICLE IN PRESS
Table 4
Logit price elasticities for mobile access
Price variable 2000 2001
Mobile access 0.43*0.43*
Mobile usage 0.17* 0.04
Fixed access 0.18* 0.13*
Local toll usage 0.00 0.01
Number of HHs 95,174 96,000
Note:Elasticities are calculated as averages of household elasticities.*Significant at the 1% level.
28Using earlier Bill Harvesting data,Duffy-Deno (2001)estimates an own-price elasticity for second lines of 0.59
using a binomial logit model.Relying on the same source for subscription data,Eisner and Waldon (2001)estimate a
bivariate probit model of second fixed line and ISP subscription, and find an elasticity of 0.35. Employing a large
household sample for Canada,Solvason (1997)estimates an elasticity of secondary line demand of 0.48.
29See Deaton and Muelbauer (1980).
M. Rodini et al. / Telecommunications Policy ](]]]])]]]–]]]14
ARTICLE IN PRESS
Table 5
Determinants of wireless subscription logit regression results
2000 2001
Coef.Std.Err Coef.Std.Err
Price local fixed access 0.010* (0.002) 0.009* (0.002)
Price local toll usage 0.014 (0.237) 0.271 (0.172)
Price mobile access 0.035* (0.006)0.039* (0.009)
Price mobile usage 3.081* (0.806) 1.113 (1.839)
Income $20K–$25K 0.423* (0.031) 0.371* (0.030)
Income $25K–$35K 0.650* (0.026) 0.575* (0.026)
Income $35K–$50K 0.940* (0.026) 0.862* (0.025)
Income $50K–$60K 1.241* (0.031) 1.119* (0.032)
Income $60K–$100K 1.429* (0.032) 1.312* (0.033)
Income $100K+ 1.663* (0.034) 1.559* (0.036)
Income unknown 1.899* (0.039) 1.727* (0.040)
HH Size: two 0.083* (0.024) 0.141* (0.024)
HH Size: three 0.091* (0.033) 0.125* (0.034)
HH Size: four 0.100* (0.038) 0.185* (0.039)
HH Size: five or more 0.038 (0.042) 0.104* (0.042)
Children present 0.055 (0.026)0.007 (0.026)
Female headed 0.447* (0.023) 0.358* (0.023)
Married 0.336* (0.025) 0.239* (0.025)
Race: Black 0.317* (0.030) 0.366* (0.030)
Race: Asian 0.005 (0.038) 0.012 (0.037)
Race: Other 0.100 (0.048) 0.106 (0.045)
Age: 25–34 0.064 (0.032) 0.069 (0.031)
Age: 35–44 0.084* (0.032)0.167* (0.031)
Age: 45–54 0.161* (0.034)0.282* (0.032)
Age: 55–64 0.195* (0.036)0.418* (0.035)
Age: 65 or over 0.504* (0.036)0.743* (0.034)
Some high school 0.112* (0.024) 0.040 (0.023)
HS graduate 0.179* (0.026) 0.117* (0.026)
Some college 0.114* (0.028) 0.018 (0.029)
College graduate 0.071 (0.034)0.231* (0.035)
Personal computer 0.367* (0.025) 0.435* (0.025)
Internet access 0.356* (0.025) 0.291* (0.024)
Modem 0.103* (0.021) 0.227* (0.020)
Pager 0.797* (0.020) 0.748* (0.024)
Fax machine 0.451* (0.025) 0.484* (0.026)
Answering machine 0.162* (0.016) 0.230* (0.016)
Freq. brings work home 0.238* (0.020) 0.318* (0.021)
Home–based business 0.155* (0.024) 0.165* (0.025)
Telecommutes 0.122* (0.033) 0.209* (0.036)
Avg. commute 0.015* (0.002) 0.016* (0.002)
Miles 0.000* (0.000) 0.000* (0.000)
FIRE Empl. 0.000* (0.000) 0.000* (0.000)
LEC avg.rating 0.082 (0.044)0.001 (0.043)
Mobile avg. rating 0.440* (0.061)0.005 (0.074)
Intercept 3.558* (0.333)1.800* (0.533)
Observations 95,174 96,000
Log likelihood 53,603 52,842
Pseudo R2 0.19 0.20
M. Rodini et al. / Telecommunications Policy ](]]]])]]]–]]]15
We can use the cross-elasticity estimates from Table 6 and the ratio of average expenditures on
mobile and fixed service (0.45 in 2000, and 0.53 in 2001) to calculate the effect of mobile access
price changes on fixed line access.Applying the symmetry condition yields a crosselasticity from
mobile access price to fixed line subscription of about 0.06–0.08. As expected, this calculation,
which theoretically applies to subscription to any fixed line service, yields a somewhat smaller
estimated magnitude than that estimated specifically for secondary lineservice.
6.3. Non-price factors
Non-price factors that significantly increase the probability of subscription include income, age,
race, female head of household, married households and household size.Table 8 reports the
predicted probability of subscribing to mobile and second line service as a function of income,
age, households size, and other household characteristics.Mobile subscription exhibits a strong
positive income effect with subscription rates for members of highest income group almost 40%
higher than the lowest group, a 90–140% increase.It is perhaps not surprising that younger
consumers have been more willing to adopt this new technology.However, it is surprising that
subscription probabilities of the oldest households are 13–20% lower.
Consistent with Eisner and Waldon’s (2001)finding of relatively greater telecommunications
demand by blacks, our results suggest that black households have 8% higher subscription rates.
Consistent with the demand for security hypothesis, female headed households have 9% higher
subscription rates.Variables related to the demand for intra-household calls, such as married and
household size, tend to increase the likelihood of mobile subscription.Similarly, variables related
to communications intensity of the household—such as personal computer, internet access,
modem pager, fax machine, and working from home—all increase the probability of
subscribing.30
As for mobile service, non-price factors such as income, age, race, household size, and
education also affect the probability of secondary line subscription.Table 8 reports the percentage
ARTICLE IN PRESS
Table 6
Logit price elasticities for second fixed line access
Price variable 2000 2001
Mobile access 0.15 0.25
Mobile usage 0.11 0.03
Fixed access 0.68*0.62*
Local toll usage 0.01 0.03*
Number of HHs 92,126 94,093
Note:Elasticities are calculated as averages of household elasticities.*Significant at the 1% level.
30While these variables tend to be highly, positively correlated, multicollinearity is not a problem as confirmed by the
very low standard errors on these variables.
M. Rodini et al. / Telecommunications Policy ](]]]])]]]–]]]16
ARTICLE IN PRESS
Table 7
Determinants of second fixed-line subscription logit regression results
2000 2001
Coef.Std.Err Coef.Std.Err
Price local fixed access 0.027* (0.003)0.023* (0.003)
Price local toll usage 0.218 (0.304)0.554* (0.215)
Price mobile access 0.008 (0.008) 0.013 (0.011)
Price mobile usage 1.347 (1.014)0.448 (2.328)
Income $20K–$25K 0.021 (0.047) 0.068 (0.047)
Income $25K–$35K 0.001 (0.039) 0.044 (0.039)
Income $35K–$50K 0.130* (0.036) 0.174* (0.036)
Income $50K–$60K 0.277* (0.041) 0.350* (0.041)
Income $60K–$100K 0.452* (0.040) 0.483* (0.040)
Income $100K+ 0.624* (0.040) 0.670* (0.041)
Income unknown 1.040* (0.043) 1.065* (0.043)
HH Size: two 0.406* (0.033) 0.377* (0.033)
HH Size: three 0.843* (0.041) 0.794* (0.041)
HH Size: four 1.059* (0.046) 0.981* (0.046)
HH Size: five or more 1.264* (0.050) 1.178* (0.049)
Children present 0.339* (0.030)0.290* (0.030)
Female headed 0.106* (0.031) 0.130* (0.031)
Married 0.166* (0.032)0.139* (0.032)
Race: Black 0.587* (0.036) 0.602* (0.035)
Race: Asian 0.019 (0.046)0.010 (0.045)
Race: Other 0.188* (0.061) 0.066 (0.059)
Age: 25–34 0.162* (0.045)0.146* (0.044)
Age: 35–44 0.225* (0.045) 0.288* (0.043)
Age: 45–54 0.495* (0.045) 0.588* (0.044)
Age: 55–64 0.481* (0.049) 0.614* (0.048)
Age: 65 or over 0.208* (0.052) 0.418* (0.050)
Some high school 0.091* (0.033)0.023 (0.032)
HS graduate 0.115* (0.035)0.084 (0.034)
Some college 0.199* (0.037)0.176* (0.036)
College graduate 0.305* (0.042)0.231* (0.041)
Personal computer 0.312* (0.037) 0.459* (0.037)
Internet access 0.812* (0.033) 0.618* (0.032)
Modem 0.227* (0.025) 0.253* (0.022)
Pager 0.370* (0.022) 0.344* (0.024)
Fax machine 1.313* (0.024) 1.206* (0.023)
Answering machine 0.026 (0.020)0.030 (0.020)
Freq. brings work home 0.085* (0.022) 0.117* (0.022)
Home–based business 0.528* (0.025) 0.556* (0.024)
Telecommutes 0.558* (0.032) 0.548* (0.031)
Avg. commute 0.032* (0.003) 0.029* (0.003)
Miles 0.000 (0.000) 0.000 (0.000)
FIRE Empl. 0.000 (0.000) 0.000 (0.000)
LEC avg.rating 0.063 (0.056)0.196* (0.054)
Mobile avg. rating 0.404* (0.082) 0.419* (0.096)
Intercept 5.079* (0.430)4.923* (0.678)
Observations 92,126 94,083
Log likelihood 37,331 38,211
Pseudo R2 0.23 0.21
M. Rodini et al. / Telecommunications Policy ](]]]])]]]–]]]17
increase in the probability of subscribing to a secondary telephone line as a function of income
and age.Income plays a somewhat smaller role in second line subscription than with mobile
subscription with subscription rates for members of highest income group only 9% higher than
the lowest group, a 65–75% increase.31 Second line subscription seems to peak among middle
aged families, aged 45–65 with a household size of three or more.As expected, working from
home and having a fax machines or Internet access has a strong positive effect on second line
subscribership.
ARTICLE IN PRESS
Table 8
Predicted effects of various non-price factors
Mobile Second fixed
2000 (%) 2001 (%) 2000 (%) 2001 (%)
Actual subscription rate 44.8 52.2 18.2 18.0
Income o $20K 28.9 39.1 12.9 12.1
Income $20K–$25K 38.3 48.2 12.6 12.9
Income $25K–$35K 43.8 53.3 12.9 12.6
Income $35K–$50K 51.0 60.4 14.4 14.1
Income $50K–$60K 58.5 66.3 16.3 16.4
Income $60K–$100K 63.0 70.5 18.9 18.3
Income > $100K 68.2 75.3 21.6 21.3
Income unknown 73.1 78.3 29.5 28.6
Age: 24 or under 52.9 64.8 13.4 12.1
Age: 25–34 54.5 66.4 11.6 10.6
Age: 35–44 50.8 60.9 16.2 15.5
Age: 45–54 48.9 58.1 20.2 19.8
Age: 55–64 48.0 54.8 20.0 20.2
Age: 65 or over 40.4 46.7 16.0 17.2
HH size: one 47.8 56.3 9.7 9.7
HH size: two 49.9 59.7 13.9 13.6
HH size: three 50.1 59.3 20.0 19.3
HH size: four 50.3 60.8 23.7 22.3
HH size: five or more 48.7 58.8 27.6 25.9
No (%) Yes (%) No (%) Yes (%) No (%) Yes (%) No (%) Yes (%)
Female head 46.0 57.1 56.3 64.8 15.9 17.4 16.5 18.4
Black 48.7 56.6 58.2 66.8 15.9 25.4 15.5 25.1
Personal computer 43.7 52.8 51.8 62.4 13.5 17.5 11.9 17.6
Internet access 44.7 53.5 54.8 61.9 10.9 21.6 11.4 19.3
Modem 48.2 50.8 56.7 62.2 14.7 17.8 14.2 17.6
Pager 45.4 64.8 56.1 73.0 15.0 20.3 14.8 19.7
Fax machine 47.8 59.0 57.2 68.4 13.7 37.1 13.4 34.0
Answering machine 46.8 50.9 55.6 61.1 16.2 15.8 15.7 15.3
Freq. brings work home 47.9 53.8 57.0 64.6 15.7 16.8 15.1 16.7
Home–based business 48.8 52.7 58.3 62.3 15.0 23.0 14.5 22.9
Telecommutes 49.1 52.1 58.5 63.5 15.4 24.1 15.0 23.4
Note:This table reports the predicted probability of subscribing to each service in each year for households in each
group evaluated at the sample mean for all other variables.Actual subscription rates were computed for the estimation
sample, not the full sample.
31See Taylor (1994).
M. Rodini et al. / Telecommunications Policy ](]]]])]]]–]]]18
7. Conclusions
Substitutability between fixed and mobile telephone service impacts public policy toward
competition in both of these markets.In the US, the principal concern over competition in these
markets derives from the market power held by providers of fixed-line local telephone service.Our
estimates indicate that mobile service is a moderate substitute for fixed-line access.Evolving usage
patterns suggest that mobile and fixed service will become greater substitutes over time.It would
be premature, however, to infer from these estimates that mobile service currently constrains local
telephone service market power to any economically significant degree over the 2000–2001 period.
But while impediments to easy substitution remain—especially connection quality and number
portability—discrepancies between the two services are fading.
Substitutability may increase over time due to continued price declines and feature
improvements of mobile services outpacing those of fixed service.A small but increasing number
of households are opting for mobile service in lieu of fixed service.At some point in the near
future, it is possible that mobile telephone service will be able to significantly constrain fixed line
providers’ exercise of market power.When this does occur, it will be appropriate to modify many
of the current regulatory stances toward telecommunications provision.In the meantime, while
mobile has displaced fixed service for a select subset of the population, the two services appear to
have achieved a coexistence in the marketplace as well as in household budgeting, each providing
consumers with particular advantages.
Acknowledgements
We are grateful to TNS Telecoms for generously sharing the Bill Harvesting data and providing
technical assistance.This paper has benefited greatly from the research assistance of Fariza
Ahmad at the University of Illinois.Both Ahmad and Rodini received financial support from the
Center for Research on Telecommunications Policy of the Haas Business School, University of
California, Berkeley.We are grateful to Mark Jamison and an anonymous referee for their
comments on an earlier draft.
References
Ahn, H., & Lee, M.-H. (1999). An econometric analysis of the demand for access to mobile telephone networks.
Information Economics and Policy,11, 297–305.
Cellular Telephone Industry Association (CTIA) (1991).State and local taxation of cellular industry.
Cellular Telephone Industry Association (CTIA) (2002).Semi-annual mobile telephone industry survey.Washington,
DC: CTIA, June ohttp://www.wow-com.com/>.
Deaton, A., & Muelbauer, J. (1980).Economics and consumer behavior.New York: NY: Cambridge University Press.
Duffy-Deno, K.(2001).Demand for additional telephone lines: An empirical note.Information Economics & Policy,13,
283–299.
Eisner, J., & Waldon, T. (2001). The demand for bandwidth: Second telephone lines and on-line services.Information
Economics & Policy,13, 301–309.
Federal Communications Commission (FCC) (2001).Sixth annual report and analysis of competitive market conditions
with respect to commercial mobile services.Wireless Competition Bureau, June 20, 2001.
ARTICLE IN PRESS
M. Rodini et al. / Telecommunications Policy ](]]]])]]]–]]]19
Federal Communications Commission (FCC) (2002).Seventh annual report and analysis of competitive market
conditions with respect to commercial mobile services.Wireless Competition Bureau, July 3, 2002.
Hausman, J.(1999).Cellular telephone, new products and the CPI.Journal of Business and Economic Statistics,17(2),
188–194.
Hausman, J.(2000).Efficiency effects on the US economy from wireless taxation.National Tax Journal,53(Part 2
Special Issue), 733–742.
International Telecommunications Union (ITU) (2001).Yearbook of statistics: Telecommunication services, 1990–1999.
Geneva (www.itu.int).
Mini, F.(2001).The role of incentives for opening monopoly markets: Comparing GTE and BOC cooperation with
local entrants.Journal of Industrial Economics,49(3), 379–414.
Miravete, E.(2002).Choosing the wrong calling plan? Ignorance and learning.American Economic Review.in
preparation.
OFTEL (2001).Consumers’ use of mobile telephony: Summary of Oftel residential survey.Fifth wave of Oftel’s quarterly
residential consumer survey conducted in May 2001, July 2001.
OFTEL (2002).Consumers’ use of mobile telephony: Summary of Oftel residential survey.Seventh wave of Oftel’s
quarterly residential consumer survey conducted in November 2001, 29 January 2002.
Perl, L.(1978).Economic and social determinants of residential demand for basic telephone service.White Plains, NY:
National Economic Research Associates, March.
Reiffen, D., Schumann, L., & Ward, M. R. (2000). Discriminatory dealing with downstream competitors: Evidence
from the cellular industry.Journal of Industrial Economics,48(3), 253–286.
Solvason, D.L.(1997).Cross-sectional analysis of residential telephone subscription in Canada using 1994 Data.
Information Economics & Policy,9, 241–264.
Sung, N., Kim, C.-G., & Lee, Y.-H. (2000). Is POTS dispensable? Substitution effects between mobile and fixed
telephones in Korea.Paper presented at International Telecommunications Society biennial conference, Buenos Aires,
July 2000.
Taylor, L.(1994).Telecommunications demand in theory and practice.Dordrecht, The Netherlands: Kluwer Academic
Publishers.
Taylor, L. (2002). Demand analysis. In: Cave, M., Majumdar, S., & Vogelsang, I. (Eds.),Handbook of
telecommunications economics.Amsterdam: Elsevier Publishing.
TNS Telecoms (2000–2001).ReQuest Market Monitor s, Jenkintown, PA, USA.(www.tnstelecoms.com).
Train, K.(1986).Qualitative choice analysis: Theory, econometrics and an application to automobile demand.Cambridge:
MIT Press.
USDA (2001).Economic Research Services Natural Amenities Scale (http://www.ers.usda.gov/emphases/rural/data/
amenities).
US Gazetteer Place and Zipcode Files (1990).(http://www.census.gov/geo/www/gazetter/places.html).
Woroch, G. (2002). Local Network Competition, In: Cuve, M., Majumdar, S. & Vogelsang, I. (Eds.),Handbook of
telecommunications economics.Amsterdam: Elsevier Publishing.
ARTICLE IN PRESS
M. Rodini et al. / Telecommunications Policy ](]]]])]]]–]]]20