Content uploaded by David Gabel
Author content
All content in this area was uploaded by David Gabel on May 10, 2014
Content may be subject to copyright.
DRAFT
Gabel and Gideon 8/31/2006 1
Disconnecting: Universal Service on the Decline
David Gabel Carolyn Gideon
Abstract
Universal service in voice telephony is generally taken for granted in the United States.
Once a policy for extending the monopoly network to include all citizens, modifications continue
to be made to achieve the goal of full penetration in a competitive environment. The debate on
universal service policy reform has included discussions of general versus targeted subsidies,
support for high-cost areas versus low-income households, and source of contributions.
Universality of voice telephone service seemed to have benefited from the combination of the
end of a ubiquitous monopolist and the corresponding universal service policies. Penetration
increased from 91.4% in 1983 to 95.5% in March 2003. More recently, however, telephone
penetration has dropped significantly. Recent data from the FCC shows a significant decline in
the number of U.S. households that have a telephone of any kind (including mobile), with
telephone penetration down to 92.9% in November 2005 (FCC 2006). This decline is both
statistically significant and meaningful, as approximately 2.6% of US households that could
reach 911 for emergencies in 2003 no longer can. Since the current census questionnaire asks the
respondent if they are using a wireline, wireless, or other type of telephone, the decline is not due
to substitution away from wireline. While most states show overall declines in telephone
penetration, several states experienced either gains or statistically insignificant declines. The
trends are not highly correlated with which RBOC is serving the state.
What has changed in this period to cause this reduction in telephone penetration? In this
paper we explore the policies and conditions of the different states to determine what is driving
this decline in universal service. Using regression analysis of state-level data, we determine the
extent to which recent changes in penetration are due to changes in economic and demographic
conditions or in policies to promote targeted programs such as Lifeline and Link-Up, and
consumer protection policies.
Previous studies have established that monthly price is a poor indicator of penetration, as
price elasticity is very low. The cost of connection and the predictability of charges are more
important to household subscription and disconnection. Based on these previous findings, low-
income households are more likely to disconnect when pay-for-use services with unpredictable
charges are used, resulting in unpaid bills. Without strong consumer protection rules, such
households confront service disconnection and significant obstacles to reconnection, even
through targeted assistance programs like Lifeline and Link-Up. For this reason we emphasize
the role of consumer protection laws as determinants of penetration changes. The writing of
these rules are especially challenging today because consumers can run up debt using non-
regulated services and this can lead to termination of their regulated service.
We find that the recent decline in universal service in the U.S. is driven by an increase
black and recent immigrant populations, telephone service provider marketing and sales
practices, inadequate consumer protection laws, and increases in wireless telephones per capita.
Lifeline effectiveness does not appear to mitigate the decline in penetration, while Link-Up
effectiveness may have a limited effect.
DRAFT
Gabel and Gideon 8/31/2006 2
Disconnecting: Universal Service on the Decline
David Gabel Carolyn Gideon
Introduction
Universal service in voice telephony is generally taken for granted in the United States.
Once a policy for extending the monopoly network to include all population, modifications
continue to be made to achieve the goal of full penetration in a competitive environment1.When
telephone competition was first introduced in the US many expressed concern that universal
service would not be sustainable in a competitive market. Universal service policies had
allegedly relied on cross-subsidies in the monopolist’s prices, and the ability to maintain these
subsidies would disappear as competition developed.2 However, universality of voice telephone
service recently seemed to have benefited from the combination of the end of a ubiquitous
monopolist and the corresponding universal service policy debate. As shown in Figure 1,
penetration increased from 91.4% in 1983 to 95.5% in March 2003. Since then, however,
telephone penetration has dropped significantly. Recent data from the FCC shows a significant
decline in the number of U.S. households that have a telephone of any kind, including mobile,
with telephone penetration down to 92.9% in November 2005 (FCC 2006). This decline is both
statistically significant and meaningful, as approximately 2.6% (approximately 3 million) of U.S.
households that could reach 911 for emergencies in 2003 no longer can. While most states show
overall declines in telephone penetration, several states experienced either gains or statistically
insignificant declines. What has changed in this period to cause this reduction in telephone
penetration? In this paper we will explore the policies and conditions of the different states to
determine what is driving this decline in universal service in the United States.
Universal Service Policy and Trends
Historically, universal service policy involved maintaining low rates for local residential
telephone service (relative to cost) while increasing the price well above costs for business
customers and other services considered more discretionary. Local telephone service prices were
kept low for all subscribers, even the very wealthy. The effectiveness of such policy has been
challenged extensively in the literature, where monthly price for local residential telephone
service has been found to have little if any impact on penetration. Several papers have estimated
the elasticity of connection with respect to monthly price and have consistently found this
elasticity to be insignificant (Crandall and Waverman 2000; Garbacz and Thompson 2005;
Rosston and Wimmer 2000). Since price elasticity of residential telephone subscription is very
low, there are, at best, very small gains in penetration when monthly prices are lowered
(Kaserman, Mayo, Flynn 1990). Further, other services, like long distance, for which prices were
increased to subsidize local residential telephone service, were found to have much higher
elasticity with respect to monthly price. Thus maintaining low local residential prices through
cross subsidies not only was ineffective for increasing penetration, but also causing distortions in
the consumption of other services. Even if artificially low prices did help penetration, they were
1 See Mueller, 1997 for thorough history of use of term universal Service and related policies
2 We ignore discussions of expanding the scope of services to be included in universal service policy as this paper is
concerned simply with voice telephony.
DRAFT
Gabel and Gideon 8/31/2006 3
being provided far beyond the relevant population, to all households even though most would
subscribe to telephone service without the inducement of this benefit.
Figure 1: U.S. Telephone Penetration 1983-2005
•Today rate same as 20 years ago
89%
90%
91%
92%
93%
94%
95%
96%
1983
1985
1986
1987
1989
1990
1991
1993
1994
1995
1997
1998
1999
2001
2002
2003
2005
July 1986
Households with telephones (%)
89%
90%
91%
92%
93%
94%
95%
96%
1983
1985
1986
1987
1989
1990
1991
1993
1994
1995
1997
1998
1999
2001
2002
2003
2005
July 1986
Households with telephones (%)
Source: FCC
In addition to being ineffective, the traditional cross-subsidy-based funding arrangement
was becoming less sustainable with increased competition in telephone since the AT&T breakup
in 1983. Yet penetration increased significantly after 1983, as shown in Figure 1. Policy shifts to
alternative means of funding and providing universal service benefits appear to have been
effective during this period. The FCC recently attributed the high penetration rate in the US to
these programs (FCC 2004). In particular, Lifeline and Link-Up programs were introduced in
1984 to target assistance with payment to low-income subscribers to keep them connected.
Lifeline provides discounted monthly service fee for a single telephone line in primary residence.
Link-Up provides a discounted fee for initial installation of telephone service. Lifeline and Link-
Up are federal programs. Some states, however, have chosen to establish their own Lifeline and
Link-Up programs, providing additional support for low-income consumers. Otherwise, a state
defaults to the Federal criteria (known as a federal default state).
The purpose of these programs is to target assistance to low-income households,
correcting the inefficiencies of past universal service programs that kept residential local rates
artificially low to all households. Optimal policy provides assistance only to those who otherwise
would not subscribe. Yet the true effectiveness of these programs is currently limited, as only
one third of the households eligible for Lifeline or Link-Up programs are actually enrolled
(Burton and Mayo 2005). Variation across states in Lifeline and Link-Up take-up rates has
enabled studies to find the determinants of Lifeline and Link-Up enrollment. These studies
provide conflicting results. According to an FCC study, the quality of outreach is an important
element to success of Lifeline and Link-Up programs in a state (FCC 2004). However, Burton
DRAFT
Gabel and Gideon 8/31/2006 4
and Mayo (2005) find outreach efforts to be insignificant, while eligibility criteria, restriction of
access to additional lines or vertical services while in Lifeline program, easy enrollment
procedures, and total average household spending on telephone services are the biggest drivers of
Lifeline participation.
Despite the apparent earlier success of these programs, penetration rates have dropped
significantly since their peak at 95.5% in March 2003. The most recent measure of average US
penetration is 92.9% as of November 2005. Most, but not all, states have declines in penetration
in these recent years. The changes in penetration vary sufficiently across states that analysis of
changes in penetration and its drivers can increase our understanding of its cause. To better
understand what is causing the recent declines in telephone penetration we first discuss the
potential drivers of telephone subscribership.
Figure 2: U.S. Penetration and Major Policy Events
Explaining Universal Service Decline
Prices
There are several possible explanations for the recent decline in universal service shown
above. First, it is possible that increased local residential rates from rebalancing rates in the face
of competition has caused an increase in disconnects. Many anticipated declines in universal
service as competition inevitably drove prices closer to costs for local residential service.
Increased competition in long distance brought prices down, reducing access fees that funded
low local service prices. Local competition is similarly forcing a rebalancing of rates between
business and residential service, causing residential prices to rise (Knittel 2004). This, many
feared, could reduce penetration and jeopardize universal service objectives. However, the low
elasticity of subscribership with respect to monthly fee found in multiple studies (Crandall and
Waverman 2000; Garbacz and Thompson 2005; Rosston and WImmer 2000) is suggestive that
the recent drop is not due to increased monthly prices for local residential service.
The connection charge has been found to be more important to subscribership than
monthly fees. Crandall and Waverman (2000) found that the installation charge is a significant
predictor of penetration in the U.S. Thus it is possible that installation fee increases have caused
the sudden and dramatic drop in penetration since 2002. However, as shown in Table 1, there has
been little change in average installation fee, which has remained around $36 from 2002 through
2005. Table 1: Average U.S. Telephone Connection Charges 2002-2005
Year Obs Mean Std. dev. Max
2002 42 36.0455 11.887 61.55303
2003 42 36.36827 12.22901 60.29684
2004 41 36.80802 10.7012 58.98367
2005 41 35.62457 10.35201 57.21831
DRAFT
Gabel and Gideon 8/31/2006 5
Income
Studies of telephone penetration consistently find income to be an important driver
(Garbacz and Thompson 1997 and 2003; Crandall and Waverman 2000). In the FCC’s most
recent report on telephone subscribership, November 2005 penetration for households with
incomes below $5000 was 79.4% while penetration for households with income between
$100,000 and $149,999 was 97.7% (FCC 2006). If penetration is driven by income, and not
price, then best policies to promote universality would be those addressing low income
households who might require assistance to remain connected. Findings discussed above imply
there will be greater benefit from assistance for installation fees than for monthly rate payments,
that is, greater benefit from Link-Up than from Lifeline. We would expect income, or more
specifically, lower income or increased incidences of poverty, to be a possible explanation for
the recent penetration decline.
Between 2002 and 2004 there was a small increase in percent of households with income
levels less than the poverty level. This small increase is not sufficient to explain the large
decline in the household penetration rate.
Table 2: Average Percentage of U.S. Households in Poverty 2002 – 2004
Year Obs Mean Dev. Min Max
2002 51 11.79412 3.181975 5.8 19.8
2003 51 11.85098 3.011204 5.8 18.1
2004 51 12.07843 2.983777 5.4 18.6
Unexpected Charges
Unpredictability of bills and inability to control them are also problems with telephone
service for low income households, which could be contributing to its high correlation with
penetration. When charges for long distance and value added or discretionary services vary with
use, especially when such use is not fully controllable by the head of household, bills may be
unexpectedly high, leading to outstanding unpaid balances. Mueller and Schement (1996) found
such unpredictability led to outstanding balances, and ultimately disconnection from the
network. When unpaid balances remain outstanding, they are often obstacles to reconnection as
phone companies may require they be paid first. Similarly, an FCC study finds outstanding debt
for accumulated toll service fees is a significant negative predictor of subscribership (FCC 2004).
Disconnects due to unexpected charges increase with new advanced offerings, and users
cannot reconnect with the resulting debt of old unpaid bills outstanding and poor credit history
(Mueller and Schement). It may be that increases in unexpected charges result from increased
marketing and promotion efforts of the telephone service providers. As local telephone service
providers have increasingly become providers of multiple telecommunications services,
including long distance, mobile, and broadband, they may be increasingly cross selling services
that are charged by use, increasing the unpredictability of subscribers’ monthly bills. Bills for
wireless telephone service can vary unpredictably, especially when subscribers go beyond the
minute allocation of their calling plan. However, the recent practice of bundling different
DRAFT
Gabel and Gideon 8/31/2006 6
services together may decrease this unpredictability. Bundling usually involves a fixed monthly
fee and therefore greater predictability If the effect of increased marketing and promotion efforts
on low-income households is greater than the effect of selling bundled services, this could lead to
an increase in unexpected large increases in monthly bills for low income households, forcing
them to disconnect their telephone service.
Inter-modal Competition
Some respond to the drop in telephone penetration as though it was expected, as new
communication technologies compete with traditional wireline voice. However, Figure 1 shows
the decline in the number of people with access to a telephone of any type, including mobile
telephones. The Current Population Study (CPS) question used to measure penetration is: “Does
this house, apartment, or mobile home have telephone service from which you can both make
and receive calls? Please include cell phones, regular phones, and any other type of telephone,”
(FCC 2006). This question has been used since December 2004 (FCC 2006). Prior to that,
mobile phones may have been excluded from reported penetration levels. Inclusion of mobile
telephones based on this wording change should increase penetration, not decrease it, if there has
been substitution from wireline to mobile telephony. Thus substitution of traditional landline
telephone service with wireless or other new technologies for voice service cannot explain the
recent decline in universal service.
Data showing that wireline connections are declining while wireless connections are
increasing are often used as evidence of substitution3 . However, this does not explain the decline
in households with access to any form of telephone (wireline or wireless) shown above. Studies
conducted by the Bureau of Labor Statistics and the Center for Disease Controls have tried to
discern the relationship between increasing mobile telephone penetration and decreasing wireline
penetration to determine how much substitution has occurred. As seen in Figure 3, some of the
increase in wireless is for customers with both. More recent data obtained by a Center for
Disease Control survey shows the percentage of adults in households with a landline telephone
decreasing by 3.4% from January 2003 to December 2004, while the percentage of adults in
households with only a wireless telephone increased by 2.7% and the percentage of adults in
households with no telephone increased by 0.6% during this period (Blumberg, Luke, Cynamon
2005). Thus it is clear that not all of the subscribers disconnecting from wireline telephone
remain connected through wireless or other forms of telephony.
Figure 3: Estimates of percentage of households reporting telephone service between 1994
and the first quarter of 2003, by type of service
3 See, for example, Taylor 2005.
DRAFT
Gabel and Gideon 8/31/2006 7
0
5
10
15
20
25
30
35
40
45
50
19941 19951 19961 19971 19981 19991 20001 20011 20021 20031
No phone bill
Cell bill
Cell bill only
Cell and landline
Source: Consumer Expenditure interview Survey, Bureau of Labor Statistics (Tucker et al 2005)
The decline in penetration may have a more indirect connection to inter-modal
competition. We do see from the data discussed above that mobile penetration is increasing. This
may in fact be causing the beginning of a decrease in landline subscription, as consumers
substitute mobile for landline more fully. Low income households in particular may be likely to
make this substitution complete and discontinue their landline in favor of mobile if they think the
total cost of service will be lower. However, the subscribers may experience
unpredictably/uncontrollably high bills (partly from paying for minutes on received calls as well
as outgoing, and also confusing details of calling plans) and must disconnect mobile service, now
unable to reconnect landline service due to outstanding balances or poor credit history. Low
income households are also more likely to substitute mobile for landline if they have outstanding
balances for their landline service that they are unable to pay. Such consumers may have
discontinued their landline in favor of mobile, found their mobile bills even less in their control
and more volatile and are forced to disconnect this as well, perhaps adding additional unpaid
balances for the mobile phone and service.
The Bureau of Labor Statistics survey shows that beginning in 2001, there has been some
substitution of cellular for landline telephones, with the beginning of an increase in the number
of households with cellular telephone service only (Tucker et. al. 2005). This is consistent with
the above story that such substitution can lead to disconnection from all telephone service. Prior
to the 1st quarter of 2001, less than 1% of households used only mobile telephone. This then
increases to 4% in the 1st quarter of 2003. These households with only cellular telephone service
are most likely to be students, renters, single-person households, urban, and households that are
not in the highest income quartile (Tucker et al 2005). This is not inconsistent with the story that
lower income households may be making this substitution and then finding themselves
DRAFT
Gabel and Gideon 8/31/2006 8
disconnected completely. A later survey by the Center for Disease Control shows that by
December 2004, 5.5% of households had cellular telephone only (Blumberg, Luke, Cynamon
2005). Using survey data through 2003, prior to the decline in U.S. penetration,Tucker et. al.
note that as the number of households with mobile only increases, the number of households
without any phone is decreasing. This suggests that from 2001 to 2003, unconnected households
were making connections with mobile phones. However, the sharp decrease in telephone
penetration occurs after this survey. Are these incremental disconnected households coming from
the mobile-only category?
Lifeline, Link-Up and Universal Service
The Lifeline and Link-Up programs may be helpful in understanding the cause of the
recent decline in universality of voice telephony. There is a federal minimum level of support
that must be provided to low-income households through these programs. Some states have
established their own Lifeline and Link-Up programs, providing additional support for low-
income consumers. Only one third of the eligible households currently subscribe to Lifeline and
Link-Up (FCC 2004). The FCC responded by expanding eligibility, adopting outreach
guidelines. (FCC 2004), and subsequently, “Lifeline Across America,” a national outreach
program (FCC 2005). As the Lifeline and Link-Up programs are specifically targeted at low-
income households, those consumers most likely to drop telephone service, changes in these
policies, their implementation, or their ability to address the dynamics of low-income telephone
use, could impact penetration. Thus, it is important to understand how these programs address
the difficulties of low-income households in obtaining and maintaining their telephone service,
and how these differ across states.
Some existing studies provide indications of how the Lifeline and Link-Up programs
assist low-income households. Based on FCC Report and Order and Further Proposed
Rulemaking (Docket 03-109, 4/29/2004), eligibility criteria are important. Originally the federal
requirements for eligibility were linked to participation in other government assistance programs
that are based on income eligibility. However, this may prevent some low-income consumers in
need of Lifeline and Link-Up from being eligible. This is particularly affected by recent public
assistance legislative changes that places time limits on participation, even for households who
otherwise remain eligible for such assistance based on sufficiently low income. So states with an
income eligibility requirement as well as program participation eligibility requirements may
achieve greater penetration through greater use of Lifeline and Link-Up. For those states with
income eligibility, the higher the multiple of Federal Poverty Guidelines used as income criteria,
the higher the Lifeline subscription rates.
The FCC finds that outreach regarding Lifeline and Link-Up programs impacts uptake of
benefits by those eligible: “According to an August 2000 report by Telecommunications
Industries Analysis Project, the Lifeline/Link-Up take rate almost tripled from 13.1% to 39.6%
when states implemented outreach initiatives designed to increase telephone penetration and
participation.” Outreach programs make a difference, but there is no data on outreach
expenditures by state. (Report and Order and FPR, paragraph 42). Burton and Mayo (2005) use a
binary variable to reflect a state’s outreach efforts based on FCC identification of states engaged
DRAFT
Gabel and Gideon 8/31/2006 9
in special outreach efforts4. They find outreach is not a significant predictor of Lifeline
enrollment. Rather, enrollment is enhanced by higher benefits and easy enrollment procedures.
Enrollment is dampened by restrictions on vertical services and additional lines and the
remaining average household expenditure for telephone services after Lifeline benefits. Federal
default states, which use federal rules and criteria, may be more or less successful in achieving
penetration through use of Lifeline and Link-Up than states with their own programs and rules
and criteria, depending on relative effectiveness of rules and criteria.
While the differences in Lifeline and Link-Up programs may help explain the difference
in penetration between states, its impact on the recent decline in penetration is less clear. Only
changes in these programs, and their effectiveness in helping low income households that would
otherwise disconnect stay connected, would cause the recent decline in penetration.
Telephone Provider
It is possible that the practices of the telephone providers are linked to changes in
telephone penetration. A recent PURC study found that Lifeline participation rates were lower
for Verizon, Alltel and small telephone companies than for BellSouth, Sprint, Qwest, and SBC
(Holt and Jamison 2006). This may be due to the telephone companies’ different efforts in
outreach or providing general information to consumers particularly in collections and other
interactions potentially indicating financial difficulty. Telephone company strategies in
marketing toll and value-added services and/or bundled services may also impact subscribership,
as discussed above. Table 3 shows that the recent change in penetration is not highly correlated
with the RBOC for more telephone service providers. Qwest has a higher correlation of .46,
showing increases in penetration. The Verizon correlation of -.33 also is worth noting, showing a
decrease in penetration. Once we correct for other factors in our analysis below, this may be a
meaningful driver of change.
Table 3: Correlation of RBOCs with Change in Penetration from 2003-2005
bellso~h Sbc verizon Qwest ch_pe~05
bellsouth 1
sbc -0.2393 1
verizon -0.2916 -0.375 1
qwest -0.2393 -0.3077 -0.375 1
ch_penetr~05 -0.0126 -0.1002 -0.3269 0.4571 1
Inadequate Consumer Protection Laws
Consumer protection may be an important part of understanding the decline in
penetration. Since disconnections are often caused by unpredictable charges, and large fees,
deposits and/or outstanding unpaid balances are similarly obstacles to reconnection, consumer
protection laws may help households remain connected, or reconnect. In fact, consumer
protection laws may determine how the other potential drivers of penetration have contributed to
4 They obtained this from FCC Docket 96-45, Adopted 3/27/03.
DRAFT
Gabel and Gideon 8/31/2006 10
the recent penetration decline. If income has decreased for many households in the country,
states with strong consumer protection laws can dampen the effect on penetration. Similarly,
fewer disconnects (and resulting barriers to reconnection) from unexpected charges, from
increased bundled sales by telephone companies, aggressive marketing of advanced pay-by-use
services, or from substitution to mobile, will occur when there is a higher level of consumer
protection.
If our hypothesis is correct, then we would expect to find differences in the consumer
protection policies of the states which experienced declines in penetration and those that did not.
While one possibility for the decline is that there was a change in consumer protection laws,
another is that there was a change in other behavior by households in their consumption of
telephone service and some consumer protection laws were better able to preserve penetration
under these circumstances than others.
A Model of Penetration Change by State
Data In this section we develop an empirical model of changes in penetration by state in order
to test our hypothesis of the importance of strong consumer protection to maintain penetration
and thus universal service in times of innovative new products and the inevitable instability of
experimentation and adoption by consumers. The dataset includes differences for years 2003 and
2005 for most variables, and for years 2002 and 2004 for variables where data for 2005 is
unavailable.
Based on the discussion above, we have created a regression model to predict the change
in state penetration between 2003 and 2005. Table 4 provides a list and description of the
explanatory variables contained in this model.
Table 4: Variable Definitions
Ch_Penetration_03_05 Change in percentage households with telephone service in the housing
unit from 2003 to 2005
Ch_immigrant_02_04 Change in percent of foreign-born residents arrived in US 2000 or later
from 2002 to 2004
Ch_Poverty_02_04 Change in percent of population with income less than 135% of federal
poverty level from 2002 to 2004
Ch_Uncollectibles_03_05 Change in percent of ILEC retail uncollectibles from 2003 to 2005
Ch_Black_03_05 Change in percentage of households that are Black from 2003 to 2005
Ch_Wireless_03_04 Change in percent of wireless phones per capita from 2003 to 2004
Ch_Wireless_03_05 Change in percent of wireless phones per capita from 2003 to 2005
Bellsouth Dummy variable equal to 1 in states where BellSouth provides local
service; 0 otherwise
SBC Dummy variable equal to 1 in states where SBC provides local service; 0
otherwise
Verizon Dummy variable equal to 1 in states where Verizon provides local service;
0 otherwise
Qwest Dummy variable equal to 1 in states where Qwest provides local service; 0
otherwise
DRAFT
Gabel and Gideon 8/31/2006 11
Ch_Lifeline_take_02_04 Change in percentage of eligible consumers enrolled in Lifeline from 2002
to 2004
Ch_Linkup_take_02_04 Change in percentage of eligible consumers enrolled in Link-Up from
2002 to 2004
State_rule Dummy variable equal to 1 in states where no deposit is required for
telephone connection if customer agrees to toll blocking; 0 otherwise
State_rule_uncollectibles State_rule * Uncollectibles
Disconnect_nonbasic Dummy variable equal to 1 in states where customer can be disconnected
for non-payment of non-basic services; 0 if customer can only be
disconnected for non-payment of basic service
Disconnect_uncollectibles Disconnect_nonbasic * Uncollectibles
The variables used in this model of state telephone penetration correspond with the
factors of penetration discussed in the previous section. As the price of monthly service has
consistently been found insignificant in past studies of penetration, it is not included in this
model. As income is consistently found to be a significant and important predictor of penetration
it is captured by change in the percent of households below 135% of the federal poverty level.
Change in real income per capita and change in percentage of households in bankruptcy are not
included as they did not improve the model’s explanatory power. We expect change in poverty to
be significant and negatively related to penetration. The uncollectibles variable is also an
indicator of households facing uncertain and/or unmanageable expenses in general.
Uncollectibles should also be negatively related to penetration. Income is also reflected in
demographic variables often associated with lower income households, such as the change in
Black households and the change in recent immigrant households.
Inter-modal competition is captured by the wireless variables. As the penetration variable
is based on the CPS question described above, it includes wireless as well as wireline. Therefore,
we do not expect this variable to be significant. If it is significant and positive, it may reflect
disconnected households reconnecting with wireless phones, thus increasing penetration. An
insignificant coefficient for change in wireless will indicate that wireless is not contributing or
dampening penetration. This may be because the added wireless phones complement wireline, or
because they substitute for wireline in connected households. A significant negative coefficient
would indicate that increased use of wireless phones results in an increase of disconnects from
any telephone. This may be possible as wireless is our only variable that reflects unpredictability
of monthly bills, known to decrease penetration as discussed above.
Note that wireless is treated as an exogenous variable. One might argue that household
wireline and wireless decisions are made simultaneously. However, most wireless use is
complementary to wireline. Also, our wireless variable increases when a new phone is purchased
for one member of a household where there is an existing wireline telephone. Since most
wireless decisions are independent of wireline decisions, the change in wireless is exogenous to
the change in penetration.
Lifeline and Link-Up program effectiveness by state is captured by the change in Lifeline
and change in Link-Up variables, which capture the change in enrollment for these programs as a
percentage of eligible households. This would capture the end result of the states’ outreach
efforts, enrollment procedures, and other aspects of the programs that help them keep low-
income households connected. A significant positive coefficient for these variables would
DRAFT
Gabel and Gideon 8/31/2006 12
indicate that the programs are effectively contributing to universal service, mitigating the recent
decline in penetration. This would indicate that the Lifeline and Link-Up users would otherwise
not be connected to the network and the programs are truly targeting the marginal users.
Insignificant coefficients for these programs would indicate that even as they become more
effective, they are not preventing the recent increase in disconnections, perhaps not reaching the
true marginal consumer who would not subscribe without the benefits of these programs.
The effect of the telephone providers’ practices on penetration is captured by the dummy
variables representing the different telephone companies. Any significant coefficients for these
dummy variables will indicate that the company practices impact telephone penetration, and the
sign indicates if the effect is to increase penetration or decrease it.
The effect of state consumer protection laws is captured by the two dummy variables for
requiring connection without a deposit to hook up service for a customer who agrees to block toll
calls (State_rule), and for allowing a consumer to be disconnected for non-payment of any non-
basic service (Disconnect_nonbasic). These two laws appear to be the most critical in low-
income households obtaining and maintaining their telephone service. A significant positive
coefficient for State_rule would indicate that this law helps more households connect to the
telephone network. A significant negative coefficient for Disconnect_nonbasic would indicate
that forbidding disconnection for non-payment of nonbasic service helps low-income households
stay connected to the network. The effect of these laws on how damaging uncollectible balances
are is captured by the interactive variables. If the interactive variable multiplying State_rule by
Uncollectibles is significant and positive, then the law helps dampen the effect of increased
uncollectible balances on penetration. A significant and negative coefficient on the interactive
variable multiplying Disconnect_nonbasic by Uncollectibles would indicate that allowing such
disconnections increases the negative effect of unpaid balances on penetration.
Results of Econometric Model
The results of the econometric estimation of state penetration are shown in Tables 5 and 6.
Table 5: Regression Results for Model 1: Concurrent Change in Wireless
Source | SS df MS Number of obs = 47
-------------+------------------------------ F( 10, 36) = 6.54
Model | 86.6452096 10 8.66452096 Prob > F = 0.0000
Residual | 47.6641438 36 1.32400399 R-squared = 0.6451
-------------+------------------------------ Adj R-squared = 0.5465
Total | 134.309353 46 2.91976855 Root MSE = 1.1507
------------------------------------------------------------------------------
ch_penetr~05 | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
ch_immigr~04 | -.5527075 .4148025 -1.33 0.191 -1.393966 .2885509
ch_povert~04 | .092177 .1090415 0.85 0.404 -.1289695 .3133235
ch_wirele~05 | -.1491937 .0404032 -3.69 0.001 -.2311352 -.0672521
ch_linkup~04 | -.0513482 .0434556 -1.18 0.245 -.1394803 .0367839
ch_black_~05 | -1.237413 .3995454 -3.10 0.004 -2.047728 -.4270972
qwest | .7477564 .4933796 1.52 0.138 -.2528638 1.748377
DRAFT
Gabel and Gideon 8/31/2006 13
state_rul~le | 179.2377 111.9818 1.60 0.118 -47.87192 406.3473
disconnect~s | -104.9107 44.66264 -2.35 0.024 -195.4908 -14.3307
state_rule | 2.060589 1.689921 1.22 0.231 -1.36673 5.487909
disconnect~c | -1.492556 .6887996 -2.17 0.037 -2.889507 -.0956061
_cons | 1.004697 .7390475 1.36 0.182 -.4941611 2.503555
Model 1 shows effective prediction of state penetration with adjusted R2=.55. The change
in percent of retail uncollectibles, per cap income, and bankruptcy are not included. Poverty is
not significant, but the change in Black is significant and the change in recent immigrant
population is significant at the 10% level.5 The coefficients for both of these variables are
negative, indicating an increase in Black or recent immigrant households is associated with a
decrease in state penetration. Thus, while increases in poverty are not driving the penetration
decline, increases in traditionally lower-income households are contributing to the decline.
The importance of the telephone service provider is unclear, as only the dummy variable
for Qwest is significant. The coefficient for Qwest is positive, indicating that Qwest practices
may be more consistent with the goals of universal service than those of other basic telephone
service providers.
The universal service policy variables apparently are not driving the recent change in
penetration. The Lifeline variable is not included as it is not significant, indicating the program
does not explain the recent decline in penetration. This may be an indication of the program
reaching customers after they already subscribe to telephone service, rather than reaching
households that are not connected. The Link-Up variable is included but is not significant,
indicating that increases in effectiveness of this program (% eligible who are enrolled) do not
help explain the recent change in penetration, and thus universal service.
The consumer protection laws do appear to have an impact on the change in state
penetration. The effect of allowing telephone service providers to disconnect service for
nonpayment of non-basic services is clear. Both the dummy and interactive variables for this law
are significant and negative. This means that allowing such disconnections contributes to
declines in penetration, all else equal, and that uncollectibles will increase penetration declines
more in states with this law. The effect of state rules requiring service be connected without a
deposit if the subscriber agrees to toll blocking is less significant but still potentially important.
The coefficient for the dummy variable for this law is significant at the 15% level and positive,
indicating it is possible that this law contributes to increases in state penetration. The interactive
variable for this law is significant at the 10% level and has a positive coefficient, indicating that
uncollectible balances contribute to penetration declines less for states with this law.
The effect of the change in wireless on penetration change indicated by this model is
difficult to understand. The change in wireless is significant with a negative coefficient,
indicating increases in wireless telephones result in decreases in penetration. This is puzzling, as
penetration includes wireless telephones. If the increase in wireless telephones was
predominantly for households substituting for wireline telephones, increases in wireless would
result in no change or an increase in penetration, as some of these households may have had no
5 Significance is determined based on one-tailed tests.
DRAFT
Gabel and Gideon 8/31/2006 14
telephone before. If the wireless increase is complementary to wireline use, then there should be
no impact on penetration.
We offer three possible explanations for the relationship between wireless change and
penetration change found in Model 1. The first possible explanation is that increased wireless use
indicates increased purchase of services that are charged by use, increasing the unpredictability
of the telephone bill, leading to disconnection. This may be increasingly common as telephone
service providers pursue more aggressive marketing and promotion efforts. Consumers who
purchase wireline and wireless from the same provider can be disconnected from both for
nonpayment of their wireless bill in states that do not have a law preventing this (i.e.
Disconnect_nonbasic=0). Second, the increase in wireless may be substitution by low income
households, who then find their telephone bills are not lower and/or are less predictable and must
discontinue service. Third, it is possible that, despite the specific mention of mobile phones in
the survey question, respondents misunderstand the question and answer based on wireline
telephone only. However, we do not think such misunderstanding would vary systematically by
state, and thus think this third explanation, while possible, is not likely.
To further explore the possibility that the increase in wireless substitution later leads to
disconnection, we created Model 2, replacing the change in wireless variable used in Model 1
with one that indicates the prior change. Table 6 shows the results of Model 2, where the change
in wireless variable is included as lagged variable, reflecting the change from 2003 to 2004. Note
that the explanatory power of this model is the same as for Model 1. The change in recent
immigrant population is now not significant, suggesting some relationship between this variable
and the change in wireless. The change in black households remains significant with a negative
coefficient. Qwest is now significant at the 5% level, still with a positive coefficient. In this
model change in Link-Up effectiveness is now significant at the 10% level. Surprisingly, though,
the coefficient of Link-Up is negative, indicating that increases in Link-Up enrollment come with
decreases in penetration. The effect of allowing disconnection of service for nonpayment of non-
basic services remains a significant and negative driver of penetration change and of how
uncollectibles impact penetration. The state rule requiring connection without deposit with toll
blocking is now not significant, through its effect on how uncollectibles impact penetration
change is significant at the 15% level.
Table 6: Regression Results for Model 2: Lagged Change in Wireless
Source | SS df MS Number of obs = 46
-------------+------------------------------ F( 10, 35) = 6.41
Model | 86.7629957 10 8.67629957 Prob > F = 0.0000
Residual | 47.3613482 35 1.35318138 R-squared = 0.6469
-------------+------------------------------ Adj R-squared = 0.5460
Total | 134.124344 45 2.98054097 Root MSE = 1.1633
------------------------------------------------------------------------------
ch_penetr~05 | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
ch_immigr~04 | -.4144566 .4356334 -0.95 0.348 -1.298839 .4699261
ch_povert~04 | .0894517 .1102044 0.81 0.422 -.134275 .3131785
ch_wirele~04 | -.1796928 .0514457 -3.49 0.001 -.2841331 -.0752526
ch_linkup~04 | -.0605156 .0443799 -1.36 0.181 -.1506116 .0295804
ch_black_~05 | -.9373659 .3668317 -2.56 0.015 -1.682074 -.192658
qwest | 1.222593 .4820546 2.54 0.016 .24397 2.201216
DRAFT
Gabel and Gideon 8/31/2006 15
state_rul~le | 145.9389 113.6705 1.28 0.208 -84.82449 376.7023
disconnect~s | -125.4627 43.54829 -2.88 0.007 -213.8705 -37.05501
state_rule | 1.608809 1.720849 0.93 0.356 -1.8847 5.102317
disconnect~c | -1.832856 .6816894 -2.69 0.011 -3.216759 -.4489534
_cons | .3237723 .6043539 0.54 0.596 -.9031314 1.550676
------------------------------------------------------------------------------
The lagged change in wireless is significant with a negative coefficient, providing
evidence that the second explanation of the wireless relationship above is possible. A closer look
at wireless use in the U.S. population provides additional support for this explanation. Table 7
shows 2004 U.S. wireless subscription by age. Wireless penetration is 85% in 2004 for ages 20
through 49. Thus, any additional wireless penetration growth is from the very young, those older
than 50, or the more marginal customers age 20 through 49. These customers are more likely to
have lower incomes, find themselves unable to continue to pay for the service, and, if they are
using wireless telephony as a substitute for wireline, lose their connection to any telephone
network.
Table 7: U.S. Wireless Penetration by Age
U.S. Wireless Industry Subscriber Forecast (in thousands)
Source: CTIA; Bear, Stearns & Co. Inc. estimates.
These preliminary results suggest that the recent decline in US penetration is driven by an
increase in Black and recent immigrant populations, telephone service provider marketing and
sales practices, inadequate consumer protection laws, and increases in wireless telephones per
capita. Lifeline does not appear to have any effect on the change in penetration, while Link-Up
may have a limited effect. More in-depth study of state telephone use and disconnection, both
wireless and wireline, is necessary to fully understand this unexpected relationship between
wireless telephones and penetration.
Conclusion
Telephone penetration has been declining in the United States since 2003. What we find
most remarkable about this trend is that it has received so little attention. Acknowledgements of
the decline seem to be limited to two reactions: (1) this is expected as people increasingly
DRAFT
Gabel and Gideon 8/31/2006 16
substitute wireless for wireline telephone service, and (2) Lifeline and Link-Up effectiveness
should be improved. As this study shows, both responses are misguided.
The first reaction is based on careless disregard of the definition of penetration. Wireless
substitution should not lead to penetration declines, as wireless telephone use is included in
measures of penetration. If penetration declines are in fact the result of wireless substitution, then
policymakers must remember that the disconnected households no longer have use of a wireless
phone. This reaction indicates lack of concern for a decline in universal service that may cause
serious hardships for households with no ability to access emergency services or use a telephone
to obtain employment.
Efforts to improve the effectiveness of Lifeline and Link-Up, such as the recent “Lifeline
Across America” initiative, are being implemented without evidence that they are likely to
address the recent decline in penetration. In fact, this study shows that the effectiveness of
Lifeline does not explain the recent decline in penetration, meaning greater success in enrolling
Lifeline-eligible consumers into the program has not improved a states’ recent penetration
decline. The effect of Link-Up effectiveness on the recent penetration decline is not clear. It is
necessary to first understand the cause of the decline in universal service in order to develop an
effective remedy. This study shows that improved consumer protection laws are more likely to
reverse the universal decline than improvements to the take-up of Lifeline and Link-Up
programs.
DRAFT
Gabel and Gideon 8/31/2006 17
REFERENCES
Blumberg, Stephen, Julian Luke, and Marcie Cynamon. 2005. “The Prevalence and Impact of
Wireless Substitution: Updated Data from the 2004 National Health Interview Survey.”
Department of Health and Human Services, Centers for Disease Control and Prevention. May 14.
Burton, Mark and John Mayo. 2005. “Understanding Participation in Social Programs: Why
Don’t Households Pick up the Lifeline?” Presented at CRRI.
Crandall, Robert and Leonard Waverman. 2000. Who Pays for Universal Service? When
Telephone Subsidies Become Transparent. Washington DC: Brookings Institute.
FCC, 2006. “Telephone Subscribership in the US (Data through November 2005).” May.
FCC, 2005. “FCC and NARUC Launch ‘Lifeline Across America’ to Raise Awareness of
Lifeline and Link-up Programs.” Press Release. July 26.
FCC 2004. “In the Matter of Lifeline and Link-Up. Report and Order and further NPR.” WC
Docket No. 03-109. April 29.
Garbacz and Thompson. 1997. “Assessing the Impact of FCC Lifeline and Link-Up Programs on
Telephone Penetration.” Journal of Regulatory Economics. 11:67-78.
Garbacz, Christopher and Herbert Thompson. 2005. “Universal telecommunication service: A
world perspective.” Information Economics and Policy 17:495-512.
Holt, Lynne and Mark A. Jamison. 2006. “Making Telephone Service Affordable for Low-
Income Households: An Analysis of Lifeline and Link-Up Telephone programs in Florida.”
PURC.
Kaserman, Mayo and Flynn.
Knittel, Christopher. 2004. “Regulatory Restructuring and Incumbent Price Dynamics: The Case
of Local Telephone Markets.” Review of Economics and Statistics 86 (2).
Mueller, Milton. 1997. Universal Service. Cambridge, MA: MIT Press.
Mueller, Milton and Jorge Schement.
Rosston, Gregory and Bradley Wimmer. “The ‘state of universal service.” Information
Economics and Policy 12:261-283.
Taylor, William E. 2005. Testimony on behalf of Verizon Communications Inc. and MCI, Inc.
Before the Washington Utilities and Transportation Commission. Docket No. UT-050814. In the
Matter of Verizon Communications, Inc. and MCI, Inc. for Approval of Agreement and plan of
Merger. June 28.
DRAFT
Gabel and Gideon 8/31/2006 18
Tucker, Clyde, J. Micahel Brick, Brian Meekins, and David Morganstein. 2005. “Household
Telephone Service and Usage Patterns in the U.S. in 2004.” Bureau of Labor and Statistics.
