Docstoc

Universal Service Reform _ Convergence

Document Sample
Universal Service Reform _ Convergence Powered By Docstoc
					                  Universal Service Reform & Convergence
                               USF Policy for the 21st Century

                                   Presented at the
  34th Research Conference on Communication, Information, and Internet Policy (TPRC)
                            September 29 - October 1, 2006
                                    Arlington, VA

                                          S. Derek Turner
                                         Research Director
                                             Free Press
                                         September 2006
                                   Revised September 30th 2006



                                              Abstract
Congress is currently considering overhauling several key components of the 1996
Telecommunications Act, in the attempt to bring the law in line with recent advances in market
competition and technology. One major area under consideration for reform is the Universal Service
Fund. Many observers believe that the current universal service funding mechanism is both
inefficient and unsustainable. Furthermore, the access charge regime -- an implicit universal service
subsidy mechanism -- may also be in need of reform in order to face the realities of a converged
marketplace. Despite the fact that broadband is viewed by many as the communications technology
of the 21st century, under current regulations, only schools, libraries, and rural health care facilities
are eligible to receive universal service support explicitly for broadband services. However, several
legislative proposals currently under consideration in Congress would extend universal service to
broadband. The impetus to expand USF to broadband is driven in part by recent data that show the
U.S. falling behind other developed nations in measures of broadband penetration, speed, and price.
However, a satisfactory explanation of this “broadband lag” is needed in order for members of
Congress and the FCC to craft policies that adequately move the U.S. towards the goal of universal,
affordable broadband access. This report employs comparative statistical methods to characterize
and understand the differences in broadband performance between the U.S. and other Organization
for Economic Cooperation and Development (OECD) nations. A key finding is that the U.S.’s
unusually high poverty rate may be a major factor contributing to this country’s increasingly poor
international broadband performance. This finding suggests that extending USF to broadband may
help ameliorate this disturbing trend. However, any proposal to reform USF must ensure that
contributions are collected and distributed in a more efficient and equitable manner (compared to
the current system), while at the same time adhering to the statutory goals of the program. This report
examines the costs and benefits of several contribution reform proposals, and concludes that a
numbers-connectivity contribution system would be far more efficient than the current revenue-based
approach, and that most consumers would benefit under a properly designed numbers-connectivity
system. The report also finds that one of the major economic justifications for universal service --
network externalities -- is most relevant when applying universal service to broadband. The report
recommends gradually transitioning the universal service system away from the archaic voice
standard, and towards a “bits” metric, where in the future all recipients of universal service support
will be required to offer a broadband capable connection.
Executive Summary

Congress is currently considering overhauling several key components of the 1996 Telecommunications Act,
in the attempt to bring the law in line with recent advances in market competition and technology. One
major area under consideration for reform is the Universal Service Fund. Many observers believe that the
current universal service funding mechanism is both inefficient and unsustainable. Furthermore, the access
charge regime -- an implicit universal service subsidy mechanism -- may also be in need of reform in order
to face the realities of a converged marketplace.

The Federal Universal Service Fund now collects and distributes approximately $7 billion per year in
subsidies. Nearly 60 percent of this is distributed to telecommunications carriers operating in “high-cost”
areas -- locations that are typically rural and thus more costly to serve on a per-line basis. Funding for the
high-cost subsidy (and the other three Federal USF programs -- Low-Income, E-Rate, and Rural Health Care)
is collected via a tax on the interstate portion of telecommunications carriers’ revenues. This tax, called the
“contribution factor”, now stands at nearly 10.6%, having almost doubled in the past five years. Structural
problems with the way USF funds are administered to some rural carriers, combined with the general
consumer trend away from the use of long distance services (substituting emails for voice calls) has created
a situation where the contribution factor will likely continue to rise unless the program is reformed.

Economic analysis reveals that the current manner of USF collection, which places the burden on elastic
long distance services, creates substantial inefficiencies, on the order of several billion dollars per year. The
financial burden of the universal service program, which is designed to benefit the rural consumer, actually
falls hardest on rural consumers. This is because interstate and intrastate long distance charges are far
higher in rural states, as these markups are used to offset the high per line costs in those states. Thus the
rural customer who makes large amounts of long distance calls bears the burden of subsidizing other rural
users who make few long distance calls. Further analysis demonstrates that the average user in most states
would benefit under a system of cost-based pricing. The burden of such a change would fall mostly on the
poor rural customers who make few long distance calls, but the magnitude of this increased burden would
be small, no more than a few dollars per month in additional charges for the average low-income customer.

One of the original economic justifications for universal service is to capture network externalities.
However the realities of today’s telephone markets and the essential nature of this technology to everyday
lives means that in the absence of universal service policies, total household telephone penetration would
only fall by one to three percentage points. However, the network externality framework is most apt when
applied to emerging networks. This report finds that subsidizing broadband deployment would likely
produce benefits that exceed costs, a result of network externalities and other surplus gains realized in
unserved areas with pent-up demand.

Despite the fact that broadband is viewed by many as the communications technology of the 21st century,
under current regulations, only schools, libraries, and rural health care facilities are eligible to receive
universal service support for broadband services. However, several legislative proposals currently under
consideration in Congress would extend universal service to broadband. The impetus to expand USF to
broadband is driven in part by recent data that show the U.S. falling behind other developed nations in
measures of broadband penetration, speed, and price.

This report employs comparative statistical methods to characterize and understand the differences in
broadband performance between the U.S. and other Organization for Economic Cooperation and
Development (OECD) nations. A key finding is that the U.S.’s unusually high poverty rate may be a major
factor contributing to this country’s increasingly meager international broadband performance. This finding
suggests that extending the USF Low-Income program to broadband use may help ameliorate this
disturbing trend. However, given that the largest barrier to low-income broadband adoption is not the cost
of the connection, but the cost of the computer equipment, this subsidy would have limited utility if not
administered in conjunction with programs that distribute low-cost computers to those recipients of the
USF-broadband subsidy.

                                                                                                                2
Comparative statistical methods were also used to investigate the differences in broadband performance
between U.S. states. Even after controlling for income, education, and poverty, states with higher
proportions of rural populations had significantly lower broadband penetration levels. This suggests that
extending USF high-cost subsidies to broadband networks may help close the gap in broadband penetration
between urban and rural areas.

This report evaluates several USF reform alternatives that make changes to how funds are collected and
distributed, and evaluates these alternatives against the following criteria.

•   Does the alternative maximize the availability, affordability, and adoption of telecommunications
    services and advanced information services?
•   Does the alternative minimize surplus losses?
•   Are the costs and benefits of the change allocated in an equitable manner?
•   Does the alternative minimize concentration and control of content access?
•   Is the alternative politically feasible?

After considering four contribution alternatives and four distribution alternatives, this report recommends
the following:

Contributions
•   The current system of contributions assessment should be abandoned in favor of a “numbers and
    capacity” flat monthly fee charged to end-users. This method would be able to maintain current USF
    funding levels by charging each residential user no more than $1 additional per month, an amount
    which would likely be partially or fully offset by a corresponding reduction in long distance and cellular
    charges. Under this alternative, large business would not be able to shift their contribution burden
    onto residential subscribers, as they would be assessed a fee based indirectly on their network usage, via
    a fixed charge based on capacity.

Distributions
•   Any recipient of universal service funds should, within a period of 5 years, be capable of providing end-
    users with a broadband capable connection.
•   The method in which a carrier’s costs are calculated should be uniform across all carriers, and should
    rely on each carrier’s own forward-looking long-run incremental costs.
•   The Commission should explore in a few test markets a system of reverse auctions, awarding USF
    monies to all bidders that fall within 15% of the lowest bid. If this auction process is deemed to be
    successful, the Commission should expand its use into other high-cost areas.
•   The Low-Income program should be expanded to include a $15 per month broadband Internet
    connection subsidy. In conjunction with this program, the Commission should also explore ways of
    coordinating with community groups and non-profits in order to increase the availability of low-cost
    and “recycled” computer equipment to low-income households.
•   The $2.25 billion cap on the Schools and Libraries program should be maintained, and funds should be
    targeted more towards low-income communities -- by lowering the 20% top-end subsidy for the higher-
    income schools to 5%, and increasing the subsidy for low-income schools to 100%.
•   Any carrier that receives universal service support should adhere to the FCC principle of network
    neutrality.




                                                                                                              3
Defining Universal Service - History and Rationale

The current Federal Universal Service program traces its roots to AT&T’s former nationwide monopoly on
telephony services. As the first patents held by the Bell Company began to expire at the turn of the 20th
century, many local telephone markets began to see new entrants and competition. Some rural areas that
the Bell monopoly had previously refused to serve got their first exposure to telephony using crude systems
set-up and operated by community cooperatives.1 Though prices dropped as a result of this new
competition, the nation’s telecommunications system was in disarray. The Bell companies refused to
interconnect with many of its competitors (and vise versa), creating a system where customers had to be on
the same network as those they wished to call (and receive calls from). AT&T (the parent company of local
Bell exchanges) began to pursue a “tipping point” strategy, dramatically expanding its national reach by
building new exchanges and acquiring smaller independent local companies. It is in this context that the
concept of “universal service” arose. 2

This aggressive move towards a tipping point (and the ongoing interconnection problems) resulted in the
1913 Kingsbury Commitment between the Justice Department and AT&T. This agreement required that
AT&T sell its stake in Western Electric, cease acquisition of independent exchanges, and interconnect its long
distance network with other local exchanges. However by 1921 many in Congress had begun to view
telephony as a natural monopoly, and with the passage of the Willis-Graham Act, moved towards granting
AT&T that status. Three years after the Acts passage, the Interstate Commerce Commission approved AT&T’s
acquisition of 223 of the 234 remaining independent exchange companies. 3 Though Willis-Graham went a
long way towards establishing Vail’s vision of “universal service”, this legislation bore no resemblance to
modern universal service policy. For example, though artificially high business rates are currently levied as a
method for cross-subsidizing residential service, the Willis-Graham Act specifically prohibited this practice.

1934 to 1996: Monopolies and Implicit Cross-Subsidies 4

The Communications Act of 1934 contains the first example of federal universal telecommunications
service policy. Though not mentioned specifically, the 1934 Act did create and direct the FCC “to make
available, so far as possible, to all the people of the United States a rapid, efficient... wire and radio
communication service with adequate facilities at reasonable charges”.5 But the newly created Commission
initially did little to enforce this vision, just intervening to regulate the rates AT&T charged in certain
markets. It was not until the 1950’s that the Commission began to allocate an arbitrarily high amount of
AT&T’s costs to the “interstate” jurisdiction, effectively creating a system where overpriced long distance
service was used to subsidize underpriced local service (at rates set by the Commission). 6

But this cross subsidy, along with advances in microwave technology, created arbitrage opportunities,
opening the door to the demise of the “natural monopoly” view of AT&T’s system. By the mid 1970’s MCI
had gained regulatory approval (albeit begrudgingly) to compete with AT&T in certain segments of the long

1
  Some of these systems were just multi-party lines operating on barbed wire. However, after the markets opened to competition, rural areas
exceeded urban areas in telephone penetration, a trend that continued until the great depression. This suggests that though economies of
density are important in network industries, that rural users highly valued the time saved by the telephone. For a detailed history of the early
telephone industry see Claude S. Fischer, “America Calling: A Social History of the Telephone to 1940”.
2
  In 1907, AT&T’s president Theodore Vail used the term to describe his company’s business plan to establish a single telephone system that
served all customers . See Milton Mueller, “Universal Service in Telephone History: A Reconstruction”, 17, Telecommunications Policy, 352-356,
1993.
3
  Mark Lloyd, “Whatever Happened to Antitrust?” Center for American Progress, April 5th 2006. Archived at
http://www.freepress.net/news/14851
4
  The term “cross-subsidy” used here is informal, and generally means that some set of services are priced below their long run incremental
cost, offset by some other service priced above cost. This is not necessarily same thing as the more rigorous economic definition put forth by
Faulhaber. See Faulhaber, “Cross-subsidization: Pricing in public enterprises”, American Economic Review, 65, 966-977, 1975.
5
  The Communications Act of 1934, as enacted. 47USC § 151.
6
  This shifting of cost burden to the interstate jurisdiction began in the 1950’s, but was not explicitly intended as a method of increasing local
subscribership. It was not until 1971, with the implementation of the “Ozark Plan” that the Commission explicitly stated this was the goal of
their rate plan. See “Prescription of Procedures for Separating and Allocating Plant Investment, Operating Expenses, Taxes and Reserves
Between the Intrastate and the Interstate Operations of Telephone Companies”, Report and Order, 16 F.C.C.2d 317, 1969.

                                                                                                                                                4
distance market. Also around this time the Justice Department filed its antitrust suit against AT&T, seeking
to break up its half-century old protected monopoly. AT&T’s response to these legal and competitive
pressures was to modernize and formalize the definition of “universal service”.

Put simply, AT&T’s view was that any threat to its monopoly status (and the implicit system of cross-
subsidies), would destroy its ability to provide service in all areas of the country, to any consumer that
requested it, at prices comparable to those charged in other areas of the country. AT&T believed that
competition would cause the nation’s telephone penetration level, which was at that time was around 91%,
to fall dramatically. Though AT&T lost the battle (being divested in 1984), this notion of universal service
was permanently ingrained in the regulatory paradigm.

The 1982 consent decree established competitive markets in long distance and special access services (e.g..
business to business dedicated lines), but maintained monopolies in local residential services. AT&T
became a long distance company, spinning off its local exchanges into seven independent regional Bell
operating companies (RBOC’s), each with protected regulated monopoly status.7 The separation of the local
and long distance markets, and the presence of long distance competition meant that the old system of
implicit subsides (where AT&T just “balanced the books” with high long distance charges) was no longer
viable. A new subsidy had to be created in order to maintain a “universal service” system of low-cost,
geographically averaged local rates (that is, generally low-cost, as rates were/are actually below cost in some
areas, while above cost in others).

To address this issue, the Commission established a system of “access charges” paid by long distance carriers
to the local exchange companies that originated and terminated calls. These access charges artificially
elevated the cost of long distance, and allowed local companies to remain solvent in the face of local rates
set (in many cases) below cost by the Commission.

Any system of cross-subsidies designed to offset the cost of providing universal service is problematic from
an efficiency standpoint, even under a monopoly regime like AT&T’s pre-divestiture operation. This type of
pricing artificially inflates demand for some services while depressing it for others. For example, AT&T
would levy 100% markups on business lines, even though the cost to provide this and residential service
were essentially identical (this process, though weakened in the post-1996 Act era of competition, persists
somewhat today).

However problematic and costly, the system of cross-subsidies (via long distance access charges and
geographic rate averaging) was sustainable in the local monopoly environment, because captive customers
had no other options. But if competition were allowed at the local level, the entire system would collapse.
This was the precise burden that Congress faced as it sought to “deregulate” the telecommunications sector
in the mid-1990’s.

At the time, Congress believed that changing technologies would eventually kill any last vestige of the need
to view local telephony as a natural monopoly. Members wanted to open up local markets to competition,
giving new “facilities-based” providers (i.e. those who would extend services to residents and businesses
using their own infrastructure), and non-facilities-based providers (i.e. those leasing capacity from the local
incumbent at wholesale rates) the right to compete with the incumbents. Non-facilities providers were
given access rights because Congress recognized that rollout of completely new networks would be cost-
prohibitive, and that temporary wholesale access would help get new competitors off the ground. 8

7
  This is somewhat of a simplification, as there remained many small local exchanges (mostly rural) that had existed all throughout AT&T’s
reign as a monopoly.
8
  This process of wholesale access is known as “unbundling” and is a key aspect of the 1996 Act. Unbundling regulatory policies (used in
other network industries such as electricity distribution) are aimed at encouraging competition in markets built up under monopoly
protection. Telecommunications regulators all throughout the developed world use these so-called “access-based” policies (as opposed to
solely facilities based) as a part of their liberalization efforts. However, these provisions have traditionally been applied only to
telecommunications services (like telephony) and not information services (like cable-TV). Their use (and effectiveness) in encouraging
broadband deployment is surrounded by controversy. Since 2001, the FCC has issued a steady stream of rulings that gradually weakened

                                                                                                                                             5
But if local markets were open to competition, it would be impossible for the incumbents (or new entrants)
to provide below cost service in certain areas. Under full competition, local access rates would undergo a
natural rebalancing, where on average, rural rates rise as urban rates drop.9 No one in Congress was willing
to “deregulate” to such a degree. Thus, in order to keep local rates low, Congress created an explicit subsidy
system known as the “Universal Service Fund”.

Universal Service and the Telecommunications Act of 1996

A principal goal of the 1996 Act was to foster the creation of competitive markets in all sectors of the
telecommunications industry. The Act was envisioned as a way to transition to this vision without shocking
the industry, or allowing the previously protected local monopolies (the RBOC’s and other incumbent local
exchange carriers, or ILEC’s) to abuse their market power. The Act allowed new competitors at the local level
(the so-called competitive local exchange carriers, or CLEC’s), but the ILEC’s were temporarily barred from
participation in markets other than local telephony service. Once an ILEC’s local market was deemed
sufficiently competitive (by a state board), they were then free to enter other markets, such as long distance
service.

As indicated above, maintaining universal service in a competitive market was an inherently difficult
problem for Congress to solve as it overhauled the 1934 Act. But this was made even more complex by two
arbitrary distinctions left over from the AT&T monopoly era -- distinctions that remain to this day.

The first distinction stems from how federal and state regulators have historically divided up the costs of the
“local loop” between intrastate (state) and interstate (federal) jurisdictions. 10 For the purposes of universal
service cost recovery, 25% of the loop’s costs are allocated to the interstate jurisdiction, with the remainder
falling under the intrastate jurisdiction. On the federal side, the FCC generally allows the service provider to
recoup the interstate portion of its costs through access charges levied on long distance carriers, and by
imposition of monthly subscriber-line charges on end users. The intrastate portion of costs is recovered
through intrastate access charges, fees on caller ID and call waiting, and monthly rates for basic local service.
While the majority of these charges are implicit “taxes”, the flat-rate subscriber charges are often listed on a
consumer’s bill as a “regulatory recovery fee”. This artificial separation of costs is problematic, as it bears no
actual resemblance to how an individual loop is used. Furthermore, new generation telephony that is
carried in part (or in full) over the Internet has allowed some carriers to disguise where a call is originating
from – a quasi-legal practice that has created new arbitrage opportunities that frustrate collection of
universal service revenues.

The second distinction arises in the different regulatory treatment of RBOC ILEC’s and rural ILEC’s. Rural
ILEC’s (as measured by size of customer base, not geography) are subject to rate-of-return regulation while
large ILEC’s are subject to price cap regulation. In addition, the provision of universal support to rural
ILEC’s is much more generous than that provided to the Bell’s. This distinction creates problems for
universal service, as rate-of-return carriers have little incentive to hold down costs or innovate (more on this
point below). Congress could have jettisoned these regulatory artifacts and created a universal service
mechanism that better reflected marketplace realities. However, the entrenched interests of certain players
as well as the path-dependent nature of telecom regulatory policy translated into Congress’ paradoxical
attempt to make big changes while not changing too much at all.


such policies as applied to DSL service. As of August 2006, DSL service is no longer subject to any common carrier regulations stemming
from the previous Computer Inquiry proceedings at the FCC.
9
  This is simply a result of the economies of density involved in deploying telecommunications infrastructure, an industry with high fixed
costs and low marginal costs. Deploying to rural areas is often far more expensive on a per line basis than deploying in urban areas.
However, the full result of rate rebalancing is not quite so clear. Remember that long-distance rates are held artificially high even in the
presence of competition by the imposition of access charges. Thus, it is very likely that a rural customer who makes significant amounts of
long distance calls would fare better under full rate rebalancing. This will be discussed further in the analysis section of the report.
10
   The local loop is the portion of the public switched telephone network (PSTN) that runs from the central switching office to the customer’s
premises. This portion of the network is generally regarded as the “bottleneck” of the system, due to its natural monopoly features. The
local loop is sometimes referred to as the “last mile”.

                                                                                                                                             6
Competition and Universal Service – Congress Moves to Explicit Subsidies

Section 254 of the 1996 Act (47U.S.C. § 254) established the current universal service system. In this
section, Congress outlined seven principles of universal service, some containing elements of the post-1974
notion of universal service, and some embodying new goals.11 These are:

1) Quality and Rates – Congress directed that “quality services... be available at just, reasonable and
   affordable rates”.
2) Access to advanced services – Congress established the principle that “access to advanced
   telecommunications and information services should be provided in all regions of the Nation”. This is
   important, because this principle embodies not just traditional telephony, but “information services”
   such as high-speed Internet.
3) Access in rural and high cost areas – This principle embodies the decades-old process of providing service
   in rural and other high-cost areas that is “reasonably comparable to those services provided in urban
   areas and that are available at rates that are reasonably comparable” to rates charged in urban areas.
   This principle also maintains the notion that “low-income consumers” should also have access to these
   services, effectively embracing the FCC’s practice of subsidizing poor customers, which began in the
   mid-1980’s.
4) Equitable and nondiscriminatory contributions – This principle makes it explicit that universal service will
   be paid for by “all providers of telecommunications services” in an “equitable and nondiscriminatory”
   manner. The language is important, as “telecommunications services” does not include companies that
   just provide information services, such as Internet Service Providers (ISP’s).
5) Specific and predictable support mechanisms – This principle simply embodies the notion that whatever
   mechanism for support the Federal-State Joint Board on universal service chose, it should not inhibit
   any business’ ability to fiscally plan for the future.
6) Access to advanced telecommunications services for schools, health care, and libraries – This principle is
   completely new in universal service policy. No longer would universal service be just a program that
   kept local rates commensurate across the country, but now universal service would also subsidize
   telecommunications for a very specific segment of the population – schools (elementary and secondary
   – not colleges), public libraries, and health care facilities.
7) Additional principles – The Joint Federal-State Commission was given the freedom to determine other
   principles that were “necessary and appropriate for the protection of the public interest”.

Section 254 goes onto fully define certain terms, and provides guidance for the new recipients of universal
service subsidies – schools, libraries, and health care providers. However, Congress left the implementation
details to the Joint Board and the Commission. Principles (1) and (3) precluded any move towards fully
rebalancing rates, and set the stage for the creation of subsidies to support high-cost providers. However,
though most commentators agree that Congress wanted to move to a system of explicit subsidies, nowhere
in the Act is this intent made clear. 12

The 1996 Act was signed into law on February 8th 1996. Fifteen months later the FCC released its final
implementation rules for § 254, adopting virtually all of the recommendations offered by the Joint
Board six months earlier. 13 The Commission created four disparate programs to implement the Act’s vision
of universal service.

•    High Cost - This program ensures that all consumers have access to and pay rates for
     telecommunications services that are reasonably comparable to those in urban areas. The High Cost
     fund received approximately 59% of all universal service fund disbursements in 2005.



11
   P.L. 104-104, § 254 (b).
12
   Congressional intent for explicit subsidies is mentioned in the conference report. See, H.R. Conf. Rep. No. 104-458, 1996.
13
   Report and Order in the Matter of Federal-State Joint Board on Universal Service, CC Docket 96- 45, May 8, 1997. For the Joint Board’s
recommendations see Federal-State Joint Board on Universal Service, CC Docket No. 96-45, Recommended Decision, November 8th 1996.

                                                                                                                                            7
•    Low Income - This program provides discounts that make basic, local telephone service affordable for
     more than 7 million low-income consumers. It consists of 3 components, Lifeline, Link Up, and Toll
     Limitation Service. Lifeline support reduces eligible consumers' monthly charges for basic telephone
     service. Link Up support reduces the cost of initiating new telephone service. Toll Limitation Service
     support allows eligible consumers to subscribe to toll blocking or toll control at no cost. The Low
     Income fund received approximately 12% of all universal service fund disbursements in 2005.

•    Rural Health Care - This program provides reduced rates to rural health care providers for
     telecommunications and Internet services, bringing their costs in line with their urban counterparts for
     similar telecommunications services. The Rural Health Care fund received approximately 0.4% of all
     universal service fund disbursements in 2005.

•    Schools and Libraries - This program (also known as E-rate) provides affordable telecommunications
     services and Internet access to schools and libraries. This support goes to service providers that provide
     discounts on eligible services to eligible schools, school districts, and libraries. The Schools and
     Libraries fund received approximately 29% of all universal service fund disbursements in 2005. Though
     very successful in achieving its stated aim, the fund has been plagued with accusations of waste, fraud,
     and abuse. 14

Current Status of Universal Service and Impetus for Reform

As mentioned, the old method of universal service was unsustainable in a competitive market. This is
because new market entrants can “cream-skim” or “cherry-pick” low-cost customers -- those living in areas
cheaper to serve. This in turn lowers the total pool of funds available to the ILEC for subsidizing the high-
cost, universal service-qualifying customers.
The Commission’s implementation of § 254 of the Act attempts to deal with the potential cream-skimming
problem with the creation of the high-cost fund. ILEC’s are usually the recipient of subsidies from the high-
cost fund, as they are usually the “carrier of last resort”.15 However, these funds are available to any carrier
that is willing to serve all customers (within a defined area), and who is also designated as an “Eligible
Telecommunications Carrier” (ETC) by the state regulatory agency. 16 ETC’s can include both wireless and
competitive local exchange carriers, who can ultimately compete head-to-head with the ILEC’s for low-cost
customers. Therefore, the high-cost subsidy is portable. 17

This attempt to encourage competition in local markets comes with a trade-off. An increase in competition
translates into the need for increased funds to subsidize the ETC and reimburse the ILEC for its revenue loss.
This is because as the ILEC’s customer base shrinks in the face of competition, it must recover its fixed costs
from fewer lines. This increases the ILEC’s overall per line cost. In turn, this translates into a higher per-line
subsidy, which is also available to the ETC competitor (because its subsidy is based on the incumbent’s cost
structure, another design flaw of the USF system). Further exacerbating the problem is the fact that a single
customer can subscribe to both wireline and wireless service, each from a carrier receiving the high-cost
subsidy. Not surprisingly, both the amount of funds going to ETC’s and the total size of the program have
increased significantly since its inception (see Figure 1).




14
   “Fraud charges cloud plan for 'wired' classrooms”, Randy Dotinga, Christian Science Monitor, June 17th 2004.
15
   “Carrier of last resort” (or COLR) is a regulatory distinction granted to certain telecommunications providers who agree to provide service at
affordable rates to any customer requesting it, and to also advertise the availability of these services. In exchange for assuming COLR status,
the carrier is allowed to earn a “reasonable rate of return” on its overall investment, something not guaranteed to new entrants or long
distance providers.
16
   See P.L. 104-104, section 214, subsection (e), for a full explanation of this designation.
17
   A subsidy is considered “portable” if it is paid to any firm that provides services. The need for portable subsidies stems from the fact that in
some areas, the retail service price is held (by regulators) below actual costs. If a new market entrant were only as efficient as the incumbent,
then competition would not be possible. The portable subsidy covers the deficit between cost and price, though the subsidy is currently
based on the incumbent’s, not the competitor’s cost -- a very problematic distinction that will be discussed further.

                                                                                                                                                 8
                          Figure 1: ETC Support & Telephone Penetration – Top 5 ETC States
                                                                                                                          Telephone
                                                                ETC Support                        Telephone
                                           ETC Support                                                                    Penetration
                               State                              Ranking                          Penetration
                                              2005                                                                         Ranking
                                                              (50 States & DC)                        2004
                                                                                                                       (50 States & DC)
                           Mississippi      $   59,102,764                             1             89.57%                    50
                           Arkansas         $   41,642,095                             2             88.60%                    51
                           Wisconsin        $   29,948,395                             3             95.47%                    16
                           Iowa             $   29,439,124                             4             95.43%                    17
                           Kansas           $   27,483,412                             5             94.77%                    24

                          Source: 2005 Universal Service Monitoring Report

The calls for USF reform center around the
growth in the overall fund and the apparent                                                Figure 2: USF Contribution Base vs. Cont. Factor
shrinking of the contribution base. The
majority of universal support funds come
from carriers who are operating in the
most competitive sectors of the market




                                                                                                                                              Contribution Factor
                                                                   Contribution Base
(wireline long distance service and
wireless telephony). Though
contributions to the fund are made in a
predictable and nondiscriminatory
manner (as per the Act), the way in which
the contribution burden is distributed
among the different sectors of the industry
(and in turn, consumers) may raise equity
concerns. Furthermore, taxing services that
consumers are most likely to substitute away
from is very problematic from an efficiency                                      Source: 2005 Universal Service Monitoring Report
standpoint (more on these issues in the next
section).

 Currently, the amount telecom carriers pay into the fund is determined by a “contribution factor” assessed
on their total interstate and international revenues. Each quarter, the Universal Service Administering
Corporation (USAC) calculates this contribution factor based on the expected needs of the fund and the
expected revenues of contributors. Since 2001 (after the collection methodology was retooled following a
court decision that limited the total pool of funds) the contribution factor has grown while the base of
contributions had dropped (see Figure 2).18

These trends are likely due to several factors. First, the total size of universal service fund disbursements
increased from $1.7 billion in 1999 to an estimated $4.2 billion for 2006.19 Second, the available pool of
funds (contribution base) has decreased as consumers substitute away from wireline long distance and
paging services towards email, wireless long distance, and Internet telephony (so-called voice over Internet
protocol, or VoIP). Third, there has been an increase in so-called “phantom traffic” or calls whose
originating location cannot be identified, and thus cannot be adequately assessed as inter or intrastate
traffic. Fourth, while wireless/cellular use has increased over this time period, wireless companies do not
contribute in the same manner as traditional long distance exchange carriers. These companies use the FCC



18
   The FCC initially based contributions for the schools and libraries and rural health care programs on interstate, international, and intrastate
end-user telecommunications revenues, while contributions for high-cost and low-income support mechanisms were based on interstate
and international end-user telecommunications revenues. However, this method was contested in court, and the intrastate portion was
ruled invalid by the United States Court of Appeals for the Fifth Circuit, The Commission then established a single contribution base for all
universal service support mechanisms based on interstate and international revenues. See “Federal State Joint Board on Universal Service,
Access Charge Reform, Sixteenth Order on Reconsideration and Eighth Report and Order”, CC Docket No. 96-45; and “Sixth Report and Order
“, CC Docket No. 96-262, 1999.
19
   The 2006 estimate can be found at http://www.universalservice.org/about/universal-service/fund-facts.aspx

                                                                                                                                                                    9
created “safe harbor”, which allows them to arbitrarily allocate a portion of their revenues to the interstate
jurisdiction, regardless of the actual amount of interstate calls conducted.

Voice-over-Internet-protocol telephony (VoIP) threatens to change the entire industry. Services like Vonage
and Skype have made significant gains in market share over the past several years, and use of this alternative
platform is expected to rise exponentially. If this occurs, it will lead to further erosion of the USF
contribution base, which in turn will require significant increases in the contribution factor levied on
traditional carriers (that is, unless VoIP providers are required to contribute to USF, a possibility under
serious consideration at the Commission, and evaluated later in this report). Thus it appears if present
trends continue, and the total size of the fund is not capped, the contribution factor will continue to rise --
perhaps to unsustainable levels.

The current problems with USF can principally be attributed to two design aspects of the system -- the
continued reliance on implicit rather than explicit subsidies, and the fact that most of the burden of
universal service contributions are placed on services that consumers are most likely to substitute away from
(towards new technologies) or use less of when prices are high.

Reforming the program in a manner that addresses these concerns, focusing on both economic efficiency
and distributional concerns should be a priority. But political realities may make this an unrealistic
constraint. Politicians favor implicit subsidies over explicit taxes for obvious reasons. Furthermore, the
political power of rural Senators is quite high, with the Chairmanship of the Commerce Committee
currently held by Ted Stevens of Alaska, a very vocal supporter of universal service. 20 Stevens, like many
other rural state Senators, has a strong belief that the current universal service policy design unambiguously
benefits their constituents, and an initial look at the data seems to confirm this (see Figure 3). However,
there is empirical evidence to the contrary, suggesting that the average citizen in rural states may fare better
under a more efficiently designed system (see alternatives section).

                                      Figure 3: USF & The Commerce Committee – 2004 Data




source: 2005 Monitoring Report; Census Data




20
     For example see “Stevens Says Resolving USF is Pathway to Telecom Bill”, Communications Daily, March 8th 2006.

                                                                                                                      10
Universal Service and Advanced Telecommunication & Information Services

The problems outlined above have fueled calls for contribution reform both in the Congress and at the
Commission. However, the manner in which USF money is distributed is also under consideration, and
both will ultimately need to be handled in concert. This section of the report introduces the issues
comprising the debate about distribution reform and change, in particular the question of further extending
USF subsidies to broadband services.

The phenomenon of convergence is shifting the old paradigms of telecommunications policy, creating
practical pressures on the old regulatory structure. Whereas just twenty years ago it seemed that the separate
titles of the 1934 Communications Act were quite appropriate in their separation of technologies into
“bins” (i.e. Title II for telephony, Title III for broadcasting, Title IV for cable, and Title I for ancillary
services), the digital age has eroded these once sensible boundaries. Advanced telecommunications and
information services -- in particular broadband Internet technologies -- are driving this movement towards
regulatory obsolescence. The Internet makes it possible for telephony, television, and data services to be
delivered via twisted copper pair (of the traditional telephone), coaxial cable (of traditional cable
television), and broadcast airwaves.

It is unclear if Congress anticipated convergence when they crafted the 1996 Act. However, it is clear that
they did anticipate the proliferation and importance of advanced services, and enacted legislation that was
meant to be flexible in its ability to encourage growth and adoption of these technologies. This is made
very clear in § 254 which states, “[u]niversal service is an evolving level of telecommunications services that
the Commission shall establish periodically under this section, taking into account advances in
telecommunications and information technologies and services”. 21 Though Congress did not at the time
choose to explicitly mandate general universal service for advanced information services, they did create two
new programs that specifically support advanced services for schools, libraries, and rural health care
centers. 22 Thus, these two programs explicitly provide subsidies for broadband services, albeit in a very
targeted manner.23 The Schools and Libraries program had by 2001 brought broadband service to nearly
90% of schools, and 95% of libraries. This program is viewed by many of its Congressional supporters as
critical, as it is often the only method of broadband access offered to some rural populations. Furthermore,
there is a clear need for efforts in this area, as a recent Organization for Economic Co-operation and
Development (OECD) study demonstrated that the U.S. has the 4th highest level in the OECD of 15-year
old students who have never accessed a computer. 24

While the Act recognized the immediate importance of broadband for schools, libraries, and rural health
care centers, it clearly took a wait and see attitude on whether broadband should also receive high-cost and
low-income universal service support. Congress established criteria (though arguably vague) governing how
the Joint Board and the Commission determine if advanced services like broadband qualify for universal
service support. Section 254 (c), states that

           Universal service is an evolving level of telecommunications services that the Commission shall
           establish periodically under this section, taking into account advances in telecommunications and
           information technologies and services. The Joint Board in recommending, and the Commission in


21
   § 254 (c)(1)
22
   § 254 (h)(2)(A) specifically directs the Commission to enhance “access to advanced telecommunications and information services for all
public and nonprofit elementary and secondary school classrooms, health care providers, and libraries”. This clause clearly goes beyond the
criteria for determining what services qualify for universal support found in § 254 (c)(1)(A-D).
23
   Of course, nowhere in this section of Act is “broadband” mentioned, but the Commission, acting on the recommendations of the Joint
Board, interpreted § 254 (h)(2)(A) as including “high-speed services” of greater than 1.544 Mbps, at the time the speed of a T-1 connection (a
very expensive dedicated symmetrical access line, whose download speeds are commonly matched or exceeded by far less expensive DSL
and cable modem connections).
24
   “Are students ready for a technology-rich world?”, OECD, January 2006. See Appendix 1 for a complete discussion of results from this
study.

                                                                                                                                            11
          establishing, the definition of the services that are supported by Federal universal service support
          mechanisms shall consider the extent to which such telecommunications services-
                    (A) are essential to education, public health, or public safety;
                    (B) have, through the operation of market choices by customers, been subscribed
                    to by a substantial majority of residential customers;
                    (C) are being deployed in public telecommunications networks by
                    telecommunications carriers; and
                    (D) are consistent with the public interest, convenience, and necessity.

The language of a “substantial majority of residential customers” seems to currently preclude broadband, as
only approximately 34% of residential households currently subscribe to broadband services. 25 Yet it should
be noted that broadband capable networks are already supported by universal service funds. Many of the
local exchange carriers in rural and non-rural high-cost areas have built converged networks that carry both
voice and broadband data, as it is more efficient to do so when investing in network upgrades. The fixed
costs incurred constructing and maintaining these networks are currently offset by universal service funds.

Putting aside for the moment the question of whether USF should now (or ever) be extended to broadband
(and the economics surrounding this question), there is a concern that Congress may not have anticipated.
That certain market structures and market failures may lead to the “substantial majority” threshold never
being met, or reached far after the benefits of USF extension have passed. Indeed, the impending “crisis” in
the USF program and the need for reform is viewed by some members of Congress as a window of
opportunity to explicitly extend universal service to broadband now. Lawmakers from rural states that have
historically been underserved by broadband providers are leading this push, with several bills from both
parties introduced in the past year. Each of these bills attempts to deal with the contributions problem,
while also incorporating broadband under the universal service umbrella.

The Broadband Problem – The Motivation Behind Efforts to Expand USF

There is almost universal consensus among U.S. policymakers that widespread availability and adoption of
broadband technology is vital to the nation’s continued economic growth and its competitiveness in the
developing global “information economy”. President Bush has articulated a goal of “universal, affordable”
broadband access by 2007. 26 In the Telecommunications Act of 1996, Congress directed the FCC to
“encourage the deployment, on a reasonable and timely basis, of advanced telecommunications capability
to all Americans”, defining “advanced telecommunications” as “broadband” technology that “enables users
to originate and receive high-quality voice, data, graphics, and video telecommunications using any
technology” (italics added). 27

Despite an initial rapid uptake of broadband services, recent data suggests broadband adoption in the U.S.
is slowing. 28 This trend combined with the apparent overall slowing of household Internet adoption may
be cause for concern. Some observers believe if broadband fails to gain widespread adoption across all
socioeconomic groups, the future of the “e-conomy” is jeopardized.




25
   This figure is based upon several surveys, and sources of data. The most recent, and arguably authoritative, comes from the FCC. As of
June 2005, the Commission reports that there were 38.5 million residential high-speed lines. Given the current census estimate of
approximately 114 million U.S. households, and assuming one line per household, this equals about 33.8% of households. See “High-Speed
Services for Internet Access: Status as of June 30, 2005”, Industry Analysis and Technology Division, Wireline Competition Bureau, April 2006.
This data is based on Form 477 filings of all high-speed providers, regardless of technology or subscriber base.
26
   Remarks by President Bush on April 26th 2004, at the American Association of Community Colleges annual convention. Available at
http://www.whitehouse.gov/news/releases/2004/04’20040426-6.html.
27
   Public Law 104-104, Section 706, “Advanced Telecommunications Incentives.”
28
   John B. Horrigan, “Broadband Adoption at Home in the United States: Growing but Slowing,” Pew Internet & American Life Project,
September 24 2005.

                                                                                                                                            12
Describing The Broadband Problem

The Internet is one of the most important technological innovations of the past century, having radically
transformed the way we communicate, conduct business, and perform research. The rate of Internet
adoption has been quite impressive, with the portion of U.S. households connected to the Internet
increasing from 5 percent in 1994, to nearly 70 percent in 2004. 29 Similarly, the rate of growth in adoption
of “high-speed” or “broadband” connections is also impressive when compared to other technologies. 30
The first commercial deployments of broadband occurred in the late 1990’s, but by 2005, roughly one-third
of U.S. households were broadband subscribers, with an overwhelming majority connecting through the
DSL or cable-modem platforms. 31
While this growth is impressive when viewed in isolation, it looks average compared to what has occurred
in other developed nations. Much of the concern in Congress over broadband stems from this international
comparison, and what it may mean for the future of the U.S.’s technology sector, as well as overall
economic productivity and growth.

The U.S.’s ranking in measures of global broadband performance continues to fall behind that of other
developed nations, despite the fact that this country was an early adopter of the technology. Since 2001, the
U.S. has fallen from 4th to 15th in the world in broadband penetration. 32 Over the same time period, the
U.S. fell from 4th to 12th in broadband penetration among the 30 member nations of the Organization of
Economic Cooperation and Development. 33

Other countries also seem to be surpassing the U.S. in terms of broadband speed and price. The typical U.S.
residential broadband connection offers download speeds between 1-3 Mbps (million “bits” per second)
and costs on average $35 per month. 34 However, in France some consumers are offered service that is 10
times faster than the typical U.S. connection, bundled with 100 TV channels and unlimited telephone
service, all for only $38 USD per month. 35 A 26 Mbps connection in Japan retails for $22 per month, and
this is considered somewhat of a “low-speed” connection there, as fiber optic connections carrying data at
100 Mbps are common in Tokyo. 36

In addition to falling behind in terms of adoption, speed, and price, the U.S. continues to struggle with the
so-called “digital divide”. According to the latest available U.S. Census Current Population Survey, more

29
   “A Nation Online: Entering the Broadband Age”, National Telecommunications & Information Administration (NTIA), September 2004.
30
   There is no universally accepted definition of what characterizes a “broadband” or “high-speed” Internet connection. “Dial-up” or “narrow-
band” Internet connections are constrained by physical limitations, which create an upper limit on data transfer speed of 56 kbps (kilo bits
per second, or thousand bits per second) in both directions (upload and download). Generally, broadband technologies are considered to
be any connection to the Internet that is “always on” (i.e., does not require dialing in), and faster than dial-up. The two most common types
of broadband connections are DSL (digital subscriber line) and cable-modem. The typical maximum download speeds of these platforms are
1000 kbps and 3000 kbps respectively. The FCC defines a “high-speed” connection as one capable of transmitting at or greater than 200
kbps in either direction. The OECD puts the threshold for broadband at 256 kbps, while some nations like Canada set the threshold
definition much higher, at 1500 kbps.
31
   "Broadband Statistics," Organization for Economic Cooperation and Development, June 2005 (available at http://oecd.org); and “The
Internet of Things”, International Telecommunications Union, November 2005 (available at http://itu.int). Total household Internet
penetration in the U.S. is approximately 70%, with broadband households making up slightly less than half of the total Internet households.
“Broadband penetration” is defined as number of subscribers per 100 inhabitants, while “household broadband penetration” is defined as
number of subscribers per 100 households.
32
   International Telecommunications Union, Broadband Statistics 2005, available at http://itu.int.
33
   "Broadband Statistics," Organization for Economic Cooperation and Development, December 2005, released April 2006. The non-OECD
nations leading the U.S. in the ITU data set are Israel, Taiwan, and Hong Kong.
34
   “Broadband Reality Check”, S. Derek Turner, Free Press, Consumers Union, Consumer Federation of America, August 2005. The data on
speed, price, and availability was culled from various sources, including leading company offers. The recent FCC Form 477 report now
includes a section on transfer rates, though the categories chosen are somewhat odd. The April report indicated about 53% of the total
connections in the U.S. (home and business) were between 2.5 Mbps and 10 Mbps, the majority of these being cable connections.
Approximately 30% fell into the category of between 200 kbps and 2.5 Mbps, consisting mostly of ADSL lines. Given that the average ADSL
offer from leading companies is between 768 kbps and 1.5 Mbps, and that the typical cable offer is for a 3 Mbps connection, with some areas
having access to 4 or 6 Mbps, it is probably reasonable to put the “average” connection as between 1 and 3 Mbps. Furthermore, the cable
speeds are rarely reached, as these lines are often shared within neighborhoods.
35
   “Competition Brings French 15 Mbps-18 Mbps Broadband,” Online Reporter. October 30 2004.
36
   Thomas Bleha, “Down to the Wire,” Foreign Affairs, May/June 2005.

                                                                                                                                           13
than half of households with annual incomes above $100,000 have broadband, while less than one-third of
households with incomes below $100,000 do, and the gap is even wider at the extreme ends of the income
scale.37 Also, urban and suburban populations have broadband adoption rates far above that in rural areas,
and the gap continues to widen. 38

There is also evidence that broadband remains unavailable and/or unaffordable to a significant percentage
of U.S. households. Recent data from the Yankee Group demonstrates that high prices are the number one
reason reported for the decision to not subscribe to broadband.39Also, while the number of households
reporting more than one available broadband provider grew over the past several years, the number
reporting no service available remained constant at 19% (for the period of 2000 to 2003). 40 Recent Pew
Survey data showed that 27% of rural dial-up users report no available broadband service. 41 The latest FCC
data (from June 2005) reports that “as a nationwide average... high-speed DSL connections were available
to 76% of the households to whom ILEC’s could provide local telephone service, and that high-speed cable
modem service was available to 91% of the households to whom cable system operators could provide
cable TV service”. 42

Some observers argue that America’s “broadband problem” is due in part to the lack of a national
broadband policy, which has led to a stagnant, non-competitive duopoly market. 43 Critics also contend that
broadband provision is controlled primarily by companies who have incentive to underserve based upon
concerns of cannibalizing their primary markets. 44

There does appear to be merit to the duopoly claim. The top ten broadband providers account for nearly
85% of the entire U.S. market, with Comcast and AT&T alone controlling over one-third of the market. 45

Together these data form a potentially troublesome picture. Though the U.S. was an early adopter of
broadband, it appears that other countries are leaving it behind. And even more bothersome, as broadband
technology diffuses throughout American society, the gap between haves and have-nots appears to be
widening.

Understanding the Broadband Problem

In the policy debate surrounding broadband deployment and adoption, it is often assumed that the
difference in penetration between the U.S. and other countries is a function of geographic factors, namely
population density. 46 This notion is attractive, as it is usually true that the overall cost per deployed
connection is lower in more densely populated areas (all other things being equal).47 But the question
remains, do geographic factors explain the observed differences in broadband penetration rates among

37
   “Current Population Survey Supplements October 2003”, U.S. Census Bureau, March 2004.
38
   “Rural Broadband Internet Use”, John B. Horrigan and Katherine Murray, Pew Internet and American Life Project, February 2006.
39
   Yankee Group Research Inc., 2006. Available at http://www.emarketer.com/Article.aspx?1003833
40
   Presentation by Kevin J. Martin at the 22nd Annual Institute on Telecommunications Policy & Regulation, December 3, 2004.
41
   See footnote 41. In this survey, 35% or rural dial-up users did not know if broadband was available, while 38% said it was.
42
   “High-Speed Services for Internet Access: Status as of June 30, 2005”, Industry Analysis and Technology Division, Wireline Competition
Bureau, April 2006. This is the first year the FCC collected this specific type of data, so a historical comparison is not possible.
43
   “The Broadband Problem: Anatomy of a Market Failure and a Policy Dilemma”, Charles H. Ferguson, Brookings Institution Press, 2004.
44
   The argument here is that cable companies (who have large states in content providing enterprises) will want to keep download and
particularly upload speeds below the 10-20 Mbps range, speeds where high-quality video competition (and piracy) could threaten their core
business. Telecom companies likewise are concerned about VoIP cannibalizing their switched network business line (self-cannibalization is
commonly thought to be behind the LEC’s slow entry into the DSL market – they did not want to lose the margins earned on second-
residential lines used for dialup connections). Telecom companies appear to be poised to enter the video markets, pushing Congress to
enact national franchise legislation that will make this entry much easier. Likewise, cable companies are deploying VoIP in certain areas, in
direct competition with the switched network incumbents. But it remains to be seen exactly how the phone companies will be able to offer
video services. New generation DSL technologies like ADSL2+ and VDSL can reach speeds needed for video, but require the customer be
located close to the central office, or for fiber to be pushed deeper into neighborhoods, connecting with remote DSLAM terminals.
45
   If the proposed merger of AT&T (formerly SBC) and BellSouth is approved, the top two broadband providers will have a combined market
share of over 40%.
46
   See Kevin Martin, “United States of Broadband,” Wall Street Journal, July 7, 2005.
47
   For a comprehensive review of this subject see “The Essential Guide to Telecommunications”, Annabel Z. Dodd, 2005.

                                                                                                                                           14
OECD nations? What other factors are contributing to the poor U.S. performance relative to other OECD
countries?

To answer these questions, a cross-sectional econometric analysis of the 30 nations of the OECD was
undertaken. The approach is somewhat similar to that of other researchers, but differences in methodology
may lead, in some instances, to divergent conclusions. 48 The results of this analysis have important
implications for the universal service regime, as well as other policies aimed at increasing broadband access
and adoption. These implications are discussed fully in the next section of this report. Detailed
explanation of this analysis and its conclusions are discussed in the appendix to this report.

Also, a cross-sectional econometric analysis of the 50 states and the District of Columbia was undertaken
using state-level broadband penetration data from the FCC, along with recent Census Bureau demographic
data. Detailed explanation of this analysis and its conclusions are discussed in the appendix to this report.

Key findings include:

     •     The factors most important for predicting broadband penetration are median household income,
           and poverty rate. 49 The data suggest that the high poverty rate in the U.S. (relative to other OECD
           nations) may partly explain the poor performance relative to other OECD nations. This is likely
           due to the fact that high poverty leads to lower overall demand for broadband.

     •     The difference in population density between OECD nations does not adequately explain the
           differences in broadband penetration, nor does the percent of population living in urban areas.

     •     The U.S. has an unusually high level of 15-year old students who report having never used a
           computer, with only Slovakia, Mexico, and Turkey reporting higher levels. The data suggest that
           this high percentage is in part due to the U.S.’s high poverty rate. This result is important, as a
           recent OECD study demonstrates the importance of both home and school computer access on
           measures of educational performance.

     •     Within the U.S., the factors most important for predicting broadband penetration are median
           household income and the percentage of a state’s population living in rural areas.

It is important to note that while the findings from the international comparison help explain the
differences between OECD countries, they do not explain differences within the U.S. itself. Other
researchers have devoted significant effort to answering similar questions about differences between
broadband deployment and adoption in the U.S., at the state-level and lower aggregate levels. Most notable
is the recent work of Kenneth Flam from the University of Texas at Austin. Flamm investigated the
determinants of broadband penetration in the United States, concluding that income and wealth are among
the most important factors. 50 He also determined that absolute market size, not population density, is a key
determinant of broadband penetration.




48
   See Kim, Bauer, and Wildman, “Broadband uptake in OECD Countries: Policy lessons from comparative statistical analysis”, August 29th
2003, for another econometric approach to evaluating the differences between OECD nations. See Maldoom, Marsden, Sidak, and Singer,
“Broadband in Europe: How Brussels Can Wire the Information Society”, 2005, for analysis of EU member nations. These studies suffer from
some of the same limitations present in this analysis; the fact that the use of cross-sectional data brings the assumption that the relationship
between the dependant and independent variables are the same in each country, and the inherent ecological fallacies involved in country-
level data. A time-series approach would remedy this weakness, and future work will attempt to do so.
49
   The OECD data used for this analysis defines poverty rates as “the share of individuals with equivalised disposable income less than 50% of
the median for the entire population”. See “Income Distribution and Poverty in OECD Countries in the Second Half of the 1990s”, Michael
Forster and Marco Mira d'Ercole, OECD Social, Employment, and Migration Working Paper #22, March 10 2005.
50
   “The Role of Economics, Demographics, and State Policy in Broadband Availability”, Kenneth Flamm, Presented at the PURC/London
Business School Conference on “The Future of Broadband: Wired and Wireless, 2005”, February 24, 2005.

                                                                                                                                              15
Should Universal Service be Extended to Broadband?

Before evaluating alternatives for reforming general USF contribution and distribution mechanisms, I will
address the question of whether USF subsidies should be extended to broadband services. Specifically, what
are the economic justifications for subsidizing broadband universal service, and do the benefits of
subsidization exceed the costs? Also, any broadband universal service program will have to resolve the
questions present in the current system (aside from the contribution mechanism problem). What types of
providers should be eligible to receive support (i.e. should the subsidy be technology neutral)? What
minimum quality standards should be required? Should in-market competition be allowed (i.e. should
more than one provider in a given geographic area be eligible for the subsidy)? It’s important to explore
these issues, for each possible policy design comes with its own unique costs and benefits, as well as other
potential trade-offs.

Market Failure - Duopoly

Given that the cable and DSL platforms dominate the residential market (accounting for 98.9% of
subscriptions), and that these services are usually each provided by a single dominant regional provider, it is
likely that the overall broadband market exists as a tight oligopoly with regional duopoly characteristics. 51
Indeed, though there has been an increasing level of competition between the DSL and cable modem
platforms, prices have largely held constant over the past several years, and appear to be on the rise – a
telltale sign of a duopoly market. 52

In this oligopoly, the market price for broadband services is above marginal cost (which is near zero for
homes already wired), and consequently there exists some inefficiency in the form of deadweight loss.
However, in industries with high fixed costs and economies of scale, it is impractical to price at marginal
cost. In this situation, the efficient price is known as the Ramsey price. 53

Under Ramsey pricing, the markup above marginal cost should be inverse to the price elasticity of
demand. 54 That is, the more flexible demand for the product, the smaller the price markup. Viewed in the
context of traditional telecommunications markets, if there is no straightforward method to allocate costs
among different services, the most efficient solution is to recover cost through services that the customer
would be reluctant to drop, and not through optional services that customers would easily give up at the
margins.

This concept of inefficiencies created by taxing services that are relatively more elastic is very important to
the current problems surrounding universal service and will be discussed further in the alternatives section
(recall that USF is largely funded from long distance services and business lines, which are much more
elastic than basic connectivity). Let it be noted for now that as the pricing structure for plain old telephone
service (POTS) clearly diverges from optimal Ramsey prices, so do pricing structures for broadband services.
Examination of broadband markets reveals a complex system of cross-subsidies and bundling, which
deviate significantly from the efficient Ramsey price.55



51
   FCC form 477 report, April 2006.
52
   Pew survey data shows the nationwide average price for broadband was $38 per month in 2002 and $39 per month in2004. The
Telecommunications Industry Association (TIA) predicts that DSL prices will rise slightly in 2006. See
http://www.lightreading.com/document.asp?doc_id=89108.
53
   For a comprehensive discussion of telecommunications network economics, see “Competition in Telecommunications”, Laffont and Tirole,
2000.
54
   This standard Ramsey rule actually requires some modification in traditional telephony markets, because there are actually separate prices
for access (monthly flat charge) and usage (i.e. per minute charges). In U.S. broadband markets, there is no “usage” charge per se, as there is
no long-distance component, or per bit download/upload charge (the exception to this currently is in some satellite markets). For a
comprehensive treatment of the modified Ramsey rule in two-part tariff markets, see Michael H. Riordan, “Universal Residential Telephone
Service”, in “Handbook of Telecommunications Economics”, Cave et. al. editors, 2002.
55
   “Digital Crossroads”, Nuechterlein and Weiser, 2005.

                                                                                                                                             16
However, it is uncertain what role universal service could play in correcting the duopoly/oligopoly market
problems. If USF funds were available to broadband providers in high-cost areas, on a technology neutral
and portable basis, then it is likely that third platform providers would enter these markets. However if
subsidies were distributed in an alternative (and perhaps more efficient) manner, such as reverse auctions,
then it is possible that the large incumbents would win these bids, leaving the duopoly problem in place.
These trade-offs will be examined further in the distributions alternatives section. Regardless, duopoly is a
problem in urban areas as well, and USF is unlikely to affect this segment of the market.

Market Failure – Network Externalities

In telecommunication markets, the benefit one derives from network use is a positive function of the
number of other people who also consume the service. No individual user fully internalizes the amount of
this benefit when deciding whether to join or leave the network. Furthermore, no network provider fully
internalizes this externality when deciding what size network to build. 56 This phenomenon is known as a
direct positive network externality. There are also possible indirect network externalities in
telecommunications markets. If a subscriber base becomes large, it can create new markets for numerous
network-based transactions, such as online bill payment. This type of indirect network externality is not
that important for the PSTN, but is very relevant to the nascent broadband network.

Classic network externalities are a primary (though not often cited) justification for universal service
policies. The practice of high usage rates offsetting low access rates is designed to attract marginal users,
who themselves would not originate many calls, to join the network -- benefiting other subscribers who
wish to call these marginal customers. However, it is possible that the marginal customers who would not
originate many calls are also likely to not receive many calls. If this is the case, then the value of adding
them to the network is small, and probably does not justify the distortion in access and usage prices.
The key question to address when considering a universal service subsidy scheme designed to capture
network externalities is, does the benefit of capturing the network externality from subsidizing the marginal
consumers exceed the loss created by the inefficiencies of the financing scheme? When considering the very
mature, high penetration telephone network, and the current financing scheme, the answer is likely no. As
shown in the alternatives section below, full rate rebalancing would likely lead to few customers dropping
out of the network. Thus it would be more efficient to target the subsidies at the marginal customers who
would drop service under efficient access pricing.

However, while there is debate amongst economists on the appropriateness of network externality
arguments surrounding the PSTN, there is consensus that these externalities are important in the emerging
Internet market. Services like email and instant messaging are obvious candidates for classic network
externality treatment, as they are new forms of two-way communication.57 But the Internet (especially
broadband) enables another type of communication -- broadcasting. Jacques Crémer constructed a
convincing theoretical argument that network externalities may be quite strong for broadband Internet
services. He examined the shape of network externalities for broadcasting services that can be offered via
broadband, concluding that the marginal value of an added user could be quite large “as the percentage of
the population that is connected approaches 100%”. 58

56
   Laffont and Tirole (2000) argue that operators of mature networks (i.e. the PSTN) do “to a large extent” internalize these positive network
externalities. However, this assertion is based on the premise that the average subscriber is indifferent to being able to connect to the
marginal subscriber. Even if this value is extremely small, it becomes large when multiplied by the millions of marginal subscribers who may
not join the network without universal service subsidies.
57
   Email is also subject to another type of positive externality -- call externality. A call externality occurs when some of the benefits of the
communication accrue to the recipient, and are not internalized by the originator of the communication. For example, I email you a
message that contains hyperlinks in the text. You then click on a link and it takes you to a website that you find quite valuable. You may also
feel that a colleague of yours would also value the website, and you send her a message.
58
   Jacques Crémer, “Network externalities and universal service obligation in the Internet”, European Economic Review, 44, 1021-1031, 2000.
Crémer concludes that subsidies should be targeted at the originators of the broadcast services. However, it is likely that he did not imagine
the proliferation of “new media” offered via amateur broadcasters (blogs, podcasts). If his conclusion is correct, then subsidies should be
targeted at providers/subscribers who offer/purchase connections with faster upload speeds. However, in the U.S. upload speeds of most
residential connections are very slow, and not suitable for large-scale content distribution.

                                                                                                                                              17
Network externalities are particularly relevant to the rural broadband market, where despite declining costs
in network equipment, and a clear presence of unmet demand, traditional network providers are hesitant to
offer service. A 2001 study by Austan Goolsbe demonstrated that subsidies targeted to these unserved
markets would bring hundred of millions of dollars in consumer surplus gains. 59 He showed that subsidies
to investment would be more efficient than direct subsidies to users. This is a reasonable conclusion from a
theoretical standpoint -- a subsidy that lowers the price in existing markets will attract marginal users who
likely place a low value on use of the product (they’re at the bottom of the demand curve). Using subsidies
to bring service to unserved markets gives broadband access to consumers who attach a high value to the
service (they’re at the top of the demand curve – i.e. businesses).

Goolsbe’s conclusion that the consumer gains from useage subsides “are significantly lower than their cost
at the time that the subsidies are introduced” relies on the premise that network externalities do not offset
the cost of the subsidies. Goolsbe considered this, and calculated that to “justify such a subsidy would
require a positive externality equal to at least 20 percent of the direct consumer surplus”. However, it
should be noted that Goolsbe’s methodology relied on demand curves calculated from willingness-to-pay
survey data collected on 70 markets (U.S. cities) in 1999. At the time, the typical price elasticity of demand
at $40 per month was between -2.1 and -3.7. However, over time these elasticities have likely fallen as
broadband has become a more critical part of many peoples lives. In 2002 Paul Rappaport used survey data
to estimate price elasticities for cable modem service, putting the value at -0.75 (at $40 per month). He also
looked at DSL service and calculated an elasticity of -1.5 (also at $40 per month).60 Crandall, Sidak, and
Singer (2003), using methods similar to Rappoport, obtain price elasticities for both services equal to -1.2
(at $40 per month). 61

What this means is that now, in 2006, these elasticities are likely far lower than what Goolsbe used in his
2001 research. Thus, if there remain differences between the value of subsidizing usage versus infrastructure
deployment, they are likely far smaller than in 2001. If so, then the size of the positive network externality
needed to justify subsiding use is likely lower than the 20% of direct consumer surplus estimated by
Goolsbe. Likewise, the losses from a market not being served have probably declined somewhat, as the
growth in demand since 1999 likely overcame some of the fixed cost market entry issues.

Universal Service for Broadband is Justified – But Program Design Matters

Taken together as a whole, there does appear to be a valid economic justification for subsidies that increase
broadband use. The 2003 Crandall study mentioned above put the direct consumer surplus in the U.S.
from broadband use at approximately $10 billion per year (in 2003, when household broadband
penetration was approximately 20%) and estimated that at household penetration levels of 50% and 95%,
consumer surpluses would rise to $100 billion and $350 billion respectively. Clearly the cost of allowing
broadband deployment and adoption to languish is high.

Extending universal service funding to broadband (as opposed to some other type of subsidy) may be a cost-
effective way of increasing subscribership, but there are some serious equity concerns of doing so. The
benefits of subsidizing new broadband deployment in high-cost areas will accrue to those with the highest
willingness to pay for these services. This category is comprised of individuals (and businesses) who have
the most ability to pay for these services.

The low-income subsidies of the current universal service system are a somewhat effective means of
maintaining marginal (often low-income) customers’ subscriptions to the network (see next section). Given
that price is one of the main reasons cited for not subscribing to broadband it seems reasonable to assume

59
   “Subsidies, the Value of Broadband, and the Importance of Fixed Costs”, Austan Goolsbee, University of Chicago, GSB, December 5 2001.
60
   See Rappaport et. al. "Residential Demand for Access to the Internet," Temple University, 2002.
61
   See Crandall et. al. "The Effect of Ubiquitous Broadband Adoption on Investment, Jobs, and the U.S. Economy," Criterion Economics, L.L.C.,
2003.

                                                                                                                                            18
that usage subsidies targeted to low-income users would increase overall broadband penetration. However,
unlike telephone service, broadband use requires the purchase of expensive computer equipment, which
can more than double the effective monthly cost of broadband.62 Thus the actual effectiveness of a low-
income broadband use subsidy is likely to be far lower than that seen in the case of telephony (which itself
is somewhat small). To maximize the value of this subsidization approach, it should be administered in
conjunction with programs that provide low-cost or free computer hardware to the low-income customer.

Thus, the extension of universal service to broadband comes with an equity trade-off: telecommunications
consumers in urban areas will subsidize broadband for consumers in rural and insular areas -- who are
likely to be on the average more wealthy than the urban telecom users. Another approach would be to
increase funding of other rural deployment incentive programs, such as the Rural Utilities Service loan
program, which are funded from general tax revenues.63 This would be more efficient than the generation of
funds from telecom taxes, and may be a slight equity improvement given the somewhat progressive nature
of the income tax structure.

A final concern is how exactly would the funds for this expansion for USF be collected? General USF
contribution alternatives are discussed below, but it is worth noting that since demand for most broadband
services remains somewhat elastic (recall the 2002-2003 estimates of between -0.75 and -1.5), any taxes on
these services could reduce demand at the margins. To investigate this, let’s assume the following. Imagine
that demand for broadband is linear, and that elasticity is -0.5 at the current market equilibrium of 43
million lines at $30 per line. 64 Now let’s say that a 10% USF charge is levied on each of these lines as a flat
subscriber-line fee. This brings the actual price up to $33, and reduces consumption by 2.15 million lines,
resulting in 40.85 million subscriptions. This tax has an associated deadweight loss of $3.23 million (per
month) and raises $122.55 million in revenues per month ($1.471 billion per year). If all of this revenue is
redirected in the form of a 100% monthly subsidy to low-income consumers, it could fund approximately
4.09 million new customers, resulting in approximately 1.94 million net customers added (see figure 4,
scenario one). If the revenue were redirected in the form of a $15 per month subsidy to low-income
consumers (approximately the current cost of dial-up service), it could result in a net addition of up to 6
million customers, assuming these consumers would subscribe at $15/month -- which is doubtful (see
figure 4, scenario two).
However, if all the above conditions were the same, but the demand elasticity was -1.0, the same 10% tax
would result in higher deadweight loss ($6.45 million per month), lower tax revenue ($116 million per
month) and an overall loss of 430,000 customers (after the subsidy was distributed at the 100% coverage
level). Under the $15 per month redistribution, a maximum of 7.7 million low-income subscribers could

62
   Though Moore’s law effects and economies of scope have significantly and rapidly brought down the retail costs of personal computers,
they still remain expensive to many low-income consumers. Assuming a retail price of $500 for a basic desktop PC is amortized over three
years, this results in an additional $13.88 per month. However, most low-income consumers cannot afford the lump sum payment for the
PC, and will choose either to forgo use, or purchase with financing – an option that could potentially triple the cost of the PC. Having said all
this, there are many programs designed at recycling old computers for use by low-income households. These programs, administered in
conjunction with broadband use subsidies would likely be effective at increasing subscribership, though it is uncertain by exactly how much.
Also, the deployment of 3G wireless broadband networks for use in handheld portable phones may be a cheaper alternative than purchase
of a PC. PCS companies often offer these phones at prices far below cost in exchange for long-term service commitments. But many poor
persons lack the necessary credit to qualify for these purchases.
63
   This program is funded through the Department of Agriculture. See 7 CFR § 1738, “Rural Broadband Access Loan and Loan Guarantee
Program” as authorized by “The Farm Security and Rural Investment Act of 2002“, P.L. 101–171. This program is however flawed and is itself
in need of reform in order for it to be more effective. The program has been allocated nearly $3 billion, but spent less than half of that since
its inception. The application standards may be to blame for this, as “applicants that have not been profitable for at least two years are
rejected if they do not have enough cash on hand to cover a full year's operating expenses.” (See Vikas Bajaj, “Money Is There to Aid Rural
Internet, but Loans Are Hard to Get”, New York Times, November 29 2005). The program also seems to exacerbate regional monopoly
problems, as 70 percent of the funds have been awarded to LEC’s. The program has also been criticized for allocating funds to wealthy
exurban and suburban areas. The inspector general of the Agriculture Department found in a recent audit that $103 million in loans were
awarded to suburban areas, including $45.6 million that went to 19 wealthy subdivisions outside of Houston Texas. Furthermore, the fund
severely restricts the type of entities who qualify for the loans. For example, municipalities cannot receive funds if another carrier operates in
the area.
64
   Obviously this is a vast oversimplification, as the shape of the demand curve is likely to be very quadratic at the extremes (i.e. some
customers, such as businesses, will subscribe at very high prices). Also, for simplicity’s sake, in this model I am assuming that price equals
marginal cost, and that the supply curve is flat.

                                                                                                                                               19
                                                                                 Figure 4: Effects of Broadband USF Tax
be added, resulting in a possible maximum net                              (if levied on BB & used for low-income program)
addition of 3.4 million customers (see figure                                                Scenario One Scenario Two Scenario Three Scenario Four
4, scenarios three and four).                                       Demand Elasticity                   -0.5         -0.5            -1           -1
                                                                         Initial Price                 $30          $30            $30          $30
                                                                    Initial Subscribers          43,000,000   43,000,000    43,000,000    43,000,000
                                                                           Tax rate                    10%          10%            10%          10%
It is worth noting that in the first example,                             New Price                    $33          $33            $33          $33
where demand elasticity was -0.5 and the                          Initial Subscribers Lost        2,150,000    2,150,000      4,300,000    4,300,000
                                                                       Subsidy Level                  100%          50%           100%          50%
subsidy was 100%, the tax transferred per net                    Low-Income Customers
                                                                                4,085,000                      8,170,000      3,870,000       7,740,000
                                                                            Added
additional customer is $63.33 per month.                                        1,935,000
                                                                     Net Lines Added                           6,020,000        -430,000       3,440,000
Thus this option to redistribute income to                                     $3,225,000
                                                                       DWL ($/mo.)                            $3,225,000     $6,450,000      $6,450,000
                                                                             $122,550,000
                                                                   Tax Revenue ($/mo.)                      $122,550,000   $116,100,000    $116,100,000
poor consumers may be more costly than a                           DWL per low-income
                                                                                    $0.79                          $0.39          $1.67           $0.83
                                                                 customer added ($/mo.)
direct subsidy funded from general revenues                        DWL per net addition
                                                                                    $1.67                          $0.54          N/A             $1.88
that are taxed in a more efficient manner.                                 ($/mo.)
                                                                    Tax per low-income
                                                                                   $30.00                         $15.00           $30           $15.00
Also, demand elasticities may be highest in                      customer added ($/mo.)
                                                                   Tax per net addition
the areas where the program ostensibly would                               ($/mo.)
                                                                                   $63.33                        $20.36           N/A            $33.75

like to add customers -- rural areas. Thus, it’s Source: Author’s calculations
possible that this type of redistribution
scheme could produce a net loss in rural household broadband penetration.

It is likely that any extension of USF to broadband would be initiated along with other general USF
contribution reforms. An option discussed in the next section would change the contribution method from
a tax on interstate carrier revenues, to a fixed charge per telephone number (including all VoIP numbers,
and possibly all IP addresses). This method collects USF revenues in a more efficient manner, as demand
elasticities for telephone access are lower than use. Thus, this collection method could raise the current
total USF expenditures for less per customer than is currently charged, on average. This would leave room
to raise additional funds for broadband while holding the average customer harmless. However, there are
distributional concerns surrounding this alternative that will be addressed in the next section.

In summary, broadband subsidies do seem to be justified by network externalities (not to mention social
equity and regional planning considerations, as well as other non-network positive externalities). In fact, as
we will see in the next section, there is good reason to make broadband-capable networks the only service
eligible for any kind of universal service support. But subsidizing broadband provision in high-cost areas
via the current USF model may be more costly (in terms of DWL) than doing so through other direct
methods such as the Agriculture Department’s Rural Broadband Access Loan and Loan Guarantee
Program. 65

Although research from 2001 suggest that subsidizing broadband use may be more costly from an efficiency
standpoint than subsidizing network construction, this depends largely on the demand for broadband in a
given area and the size of the associated network externalities. Today it is likely that the differences between
the costs (and benefits) of usage versus deployment subsidies are much smaller than they were in 2001, and
vary from market to market.

Subsidizing low-income consumers via a tax on broadband services could result in a net loss of customers;
therefore any new revenue for this purpose should come from a larger tax base. Finally, the cost of a
personal computer may be a prohibitive barrier to broadband adoption by low-income consumers,
regardless of the existence of a connection subsidy. Therefore the effectiveness of any low-income
broadband USF program would be in doubt if not administered in concert with programs that lower the
cost of computer hardware.


65
  See Jerry Hausman and Howard Shelanski, “Economic Welfare and Telecommunications Regulation: The E-Rate Policy for Universal Service
Subsidies”, Yale Journal on Regulation, 16, 19, 1999; and Hausman, “Taxation by Telecommunications Regulation: The Economics of the E-
Rate”, AEI Press, 1998. USF fees may cost the economy an additional $0.64 to $1.47 for each dollar in revenue they produce (though these
1999 estimates probably overstate the current economic cost as long-distance access fees have declined since then). These shadow costs
are higher than is the case of general funds, as the economic losses attributed to general federal taxes are estimated to range between $0.25
and $0.40 for each additional dollar collected.

                                                                                                                                                           20
Evaluation of USF Reform Alternatives

The criteria used to evaluate USF reform alternatives are largely based on the “public interest” intentions
contained within the 1934 Communications Act, and confirmed and expanded by the 1996 Act, which
states in part the intent “to make available, so far as possible... a rapid, efficient, Nationwide... wire and
radio communication service with adequate facilities at reasonable charges”.66 Also, “[a]ccess to advanced
telecommunications and information services should be provided in all regions of the Nation”67; and
“[c]onsumers in all regions of the Nation, including low-income consumers and those in rural, insular, and
high cost areas, should have access to telecommunications and information services, including
interexchange services and advanced telecommunications and information services, that are reasonably
comparable to those services provided in urban areas and that are available at rates that are reasonably
comparable to rates charged for similar services in urban areas”.68 In addition to these aspects of the “public
interest”, the FCC and other federal agencies have a long history of including economic efficiency in
evaluations of how policies serve the public interest. Courts have upheld this inclusion, as long as the
consideration of efficiency is not in conflict with Congressional statutory goals. 69 Therefore alternatives will
be measured against the following set of criteria:

•    Maximize the availability, affordability, and adoption of telecommunications services and advanced
     information services
•    Minimize deadweight loss and surplus losses
•    Allocate costs and benefits in an equitable manner
•    Minimize concentration and control of access to content by any provider
•    Is politically feasible

The Status Quo: Who Pays & Who Receives?

The nearly 7.2 billion dollar per year Federal Universal Service Fund is financed via a quarterly assessment
on all telecommunication carriers’ interstate and international revenues. These carriers in turn collect this
revenue by price markups on their services.
                                                              Figure 5: The Federal Universal Service System
The majority of universal service funding
is used to finance the deficit operations of
carriers that serve high-cost areas.
However, the monies from the high-cost
fund are not adequate enough to cover
the full 25% interstate portion of the
high-cost carriers operating costs. The
remainder is collected via a flat monthly
subscriber line charge (SLC), and by
access charges levied on long distance
carriers who use the local companies
network to receive and terminate long
distance charges. The FCC sets these
access and subscriber line charges. And
though access charges have steadily
dropped, and SLC’s have risen since the
implementation of the 1996 Act, they are
still considered to be above cost and thus
an implicit subsidy to the local companies.

66
   47 USC §151
67
   47 USC §254 (b) (2)
68
   47 USC §254 (b) (3)
69
   NAACP v. Federal Power Commission, 425 U.S. 662, 669, 1976; also Business Roundtable v. SEC, 905 F.2d 406, 413, 1990.

                                                                                                                           21
One important regulatory distinction in the access charge regime pertains to rural carriers. The 1996 Act
allowed these carriers to be treated differently for the purposes of implementing the pro-competitive aspects
of the 1996 Act. These carriers are largely exempt from the competition provisions contained in § 251 and
§ 252 of the Act, and as a result, generally charge much higher access rates. Under § 254 (g), the long
distance carriers cannot pass these charges back onto the rural customers, but must include them in
nationally averaged long distance rates. Thus this is another implicit subsidy paid ostensibly by urban
customers to benefit rural customers.

The remaining 75% of a carriers’ operating cost is
                                                                      Figure 6: Average Monthly Telephone Expenditures
collected from intrastate traffic, and falls under the
state regulatory jurisdiction. Some of the
intrastate operating deficit is collected via a fixed
monthly subscriber line charge, but state
regulators often use above-cost intrastate access
charges and geographic rate averaging as a subsidy
method.

On average, it appears that the amount spent per
month on telephone service increases with higher
levels of household income. But the poorest
households on average do seem to spend a
significant percentage of their monthly telephone
outlay on long-distance service. Thus at first
glance it appears that above-cost long distance                          Source: Crandall & Waverman, 2000
access charges (passed onto end-users in the form of
higher per minute charges and monthly fixed rates)
may be just as, if not more, costly to consumers than                       Figure 7: Expenditures by Decile of Spending
higher subscriber line charges (see Figures 6 and 7                                      <$10,00    $10,000- $20,000- $40,000- >$75,00
                                                                            Decile*
for data on household telephony expenditures).                                              0       $20,000 $40,000 $75,000       0
                                                                         Lowest             $0.00     $0.00 $0.00 $0.18          $0.63
                                                                         Second             $0.00     $0.36    $1.01 $2.90       $4.93
This is precisely the question that Robert W. Crandall                   Third              $0.86     $2.14 $3.66 $7.22          $9.68
and Leonard Waverman sought to answer in their                           Fourth             $3.13     $5.33 $7.27 $11.83 $15.26
work on the subject, “Who Pays for Universal                             Fifth              $5.80     $8.96 $11.32 $16.74 $21.06
Service”, published in 2000. 70 This is by far the most                  Sixth              $9.35    $13.10 $15.71 $22.37 $27.75
                                                                         Seventh          $14.56     $18.21 $21.76 $29.50 $37.28
comprehensive treatment of the subject, and selected
                                                                         Eighth           $20.83     $25.26 $29.34 $39.01 $49.37
data and conclusions are discussed here. However, it                     Ninth             $31.91    $37.90 $42.81 $54.60 $67.36
is worth noting that this work is now over 6 years                       Tenth            $83.72     $87.26 $88.85 $118.18 $130.92
old, and some of the data they relied on is itself                       Average          $17.02     $19.84 $22.16 $30.25 $36.42
nearly a decade old. Since this time the FCC has, via    *Level of monthly LD spending

its CALLS proceeding, lowered intercarrier access         Source: Crandall & Waverman, 2000
charges, and offset these lost revenues with an
increase in the subscriber line charge. 71 However, some of the reforms implemented by the FCC did not
apply in the same manner to rural carriers, and thus many of the cross-subsidies mentioned throughout this
report are still in place. Thus while Crandall and Waverman’s exact calculations may not be valid today, the
general conclusions are still relevant.




70
   Robert W. Crandall and Leonard Waverman, “Who Pays for Universal Service”, Brookings Institution Press, Washington DC, 2000. Some of
the data from this work that is presented here came from a research survey conducted in 1996 by PNR & Associates, entitled “Bill Harvesting
Survey III”.
71
   See http://www.fcc.gov/cgb/consumerfacts/Calls2.html for more information on the CALLS (Coalition of Affordable Local and Long
Distance Service) plan and intercarrier compensation/access charge reform.

                                                                                                                                          22
                                                                               Figure 8: Net Monthly Consumer Gains from
Crandall and Waverman’s basic conclusion is that                                  Cost-Based Rebalancing ($/subscriber)
current U.S. universal service policies are inefficient,                                          Annual Household Income (1996)
unnecessary, and that the system should be                                            Cost           $10,000- $20,000- $40,000-
                                                                            State          <$10,000                              >$75,000
scrapped, which at the time they estimated would                                     Model           $20,000 $40,000 $75,000
                                                                          CA        BCPM      -$1.97    -$0.90  -$0.08    $3.31     $5.56
save the U.S. economy $7 billion dollars per year.                        CA        CI        -$1.25    -$0.89  -$0.44     $1.12    $2.25
They estimated the change in residential phone                            CA        HCPM       $0.15    -$0.43   -$0.21   $0.23     $0.73
charges if all rates were priced at cost, determined by                   MS        BCPM      -$1.18     $4.76    $3.14  $20.38    $32.89
                                                                          MS        CI       -$2.46     -$0.21  -$0.32    $7.78    $14.62
three different cost proxy models. 72                                     MS        HCPM      -$1.24     $3.19    $2.50  $18.34    $30.98
                                                                          NJ        BCPM      -$1.95    -$0.34    $0.02   $4.04     $5.31
They concluded that under cost-based pricing,                             NJ        CI        -$1.61    -$0.71  -$0.48    $1.90     $2.74
                                                                          NJ        HCPM      -$0.71    -$0.45  -$0.42    $0.85     $1.28
average monthly U.S. local telephone rates would                          WY        BCPM     -$3.68      $1.17    $0.19  $10.96    $19.42
either decrease by as much as $0.76, or increase by                       WY        CI        -$3.17    -$0.62  -$0.95    $4.53     $9.20
                                                                          WY        HCPM     -$3.83      $0.63  -$0.09    $9.08    $16.79
as much as $13.62 (depending on the model used;
the FCC’s model put the value at a $5.03 increase).                Source: Crandall & Waverman, 2000
Depending on the state, the numbers were quite
different. For example, in New Jersey, a state with                       Figure 9: Annual Overall Effects of USF policies
almost no rural areas, rates rose by $4.26 in the FCC’s                                          (as of 2000)
model. In California, a state with large urban areas
                                                                                       Annual Welfare Gain by
and a small rural population, rates rose by $1.20. And                                     Income Category
in Wyoming, a very rural state, rates increased sharply
                                                                                                                    Cost to LD
in the FCC model, up $18.31 per month.                               Cost Model $0 to $20,000 >$20,000
                                                                                                                    Companies
The authors calculated that telephone penetration                   BCPM             $435 Million -$4.1 Billion $1.9 to $3.4 Billion
rates would rise slightly in urban areas, remain largely            CI               $433 Million -$1.5 Billion $2.3 to $3.1 Billion
unchanged in suburban areas, and decrease by a few                  HCPM             $343 Million -$1.3 Billion $1.5 to $2.1 Billion
percentage points in rural areas. They then                           Source: Crandall & Waverman, 2000
demonstrate that on a nationwide average, the welfare
losses from these dropped subscribers
                                                Figure 10: Avg. Mo. Consumer Gains From Cost-Based Rebalancing
are more than offset by increased
                                                                  Net Gain per Subscriber                       Net Gain per Subscriber
use of long distance services. This                                  by Model ($/mo.)                             by Model ($/mo.)
was true for every cost-proxy model      Sample of % Rural                                   Sample % Rural
used. This is expected, as the             Urban Population BCPM              CI    HCPM of Rural Population BCPM         CI    HCPM
                                           States       (2004)                                States   (2004)
demand elasticities (at that time)           CA             5.56%   $1.64     $0.36    $0.11   VT        61.82% $11.29    $2.68    $6.55
were very different for local and            NJ             5.65%   $2.31     $0.90    $0.35 ME          59.77%  $9.47    $2.64    $5.52
long distance services (local               NV              8.49%   $3.75     $1.53    $2.43 WV          53.95%  $7.31    $0.93    $3.78
                                            MA              8.63%   $0.47     $0.02    $0.18   MS        51.23%  $9.16    $2.46    $8.04
elasticity values used by the authors        RI             9.08%   $1.38     $0.24    $0.14   SD        48.14%  $5.18    $1.80    $1.31
were between -0.05 and -0.001                FL           10.71%    $2.88     $0.91    $0.40 MT          45.92%  $7.59    $2.10    $2.49
                                             AZ           11.83%    $3.01     $0.72    $0.15 ND          44.10%  $4.61    $1.78    $0.80
moving from lowest income                    CT           12.26%    $2.55     $0.73    $0.48    IA       38.92%  $4.44    $1.37    $1.42
category to highest; long distance          NY            12.51%    $0.30     $0.09    $0.45 WY          34.92%  $5.14    $1.55    $4.10
elasticity values used were between -       MD            13.93%    $0.87     $0.08    $0.12   ID        33.58%  $5.88    $5.55    $2.13
1 and -0.6 moving from low to high         Source: Crandall & Waverman, 2000; U.S. Census Estimates 2004
income categories).

However the welfare gains varied significantly at the state level. In California, due to its small rural
population, lower long distance rates were not enough to offset the increased local monthly fee. However,
in states like Wyoming with a high percentage of rural households, the current policy of below-cost local
rates is offset by very high long distance rates, thus under cost-based pricing these states see gains in

72
  Cost-based pricing means that each subscriber pays what these models say is the actual cost to serve them. Under this scenario, most
urban rates would change little, while rural rates would rise sharply. However, these local rate changes in theory are offset somewhat by
corresponding drops in long distance bills. These models produce estimates that in some circumstances vary significantly from one another.
Crandall and Waverman took this into account. The models used were the Hatfield 5.0 (favored by long distance companies), the Benchmark
Cost Proxy Model 3.1 (BCPM, favored by the ILEC’s) and the Hybrid Cost Proxy Model (HCPM, developed and used by the FCC). They also
used a compromise model, which they called “CI”. These models are used to estimate forward-looking long-run incremental costs. It is
worth noting that the authors did not use the Hatfield model for most of their calculation, for it implied that all current regulatory pricing for
telephony was above cost-- even for rural areas.

                                                                                                                                                23
consumer welfare. And what’s striking, these gains in the rural states are distributed amongst all three area
types within the state – urban, suburban, and rural. For example, in Mississippi, under the FCC cost model,
urban areas see a net gain of $31.75 per month per subscriber; suburban areas see a net gain of $32.52 per
month per subscriber; and rural areas experience a net gain of $3.28 per month per subscriber.

Given that high-income consumers spend more on average per month for long distance, these consumers
reap more benefits than lower income consumers under cost-based pricing (see Figure 8). But Figure 6
showed that within an income category, the amount of long distance spending varied considerably. Thus
the regressive pattern shown in figure 28 misses the fact that though on average lower income consumers
lose welfare under cost-based pricing, some low-income consumers actually benefit significantly. Also, even
on average the benefit of the current system of non-cost based rates actually brings only a small benefit to
the average poor customer. Take for example Mississippi, where the average household earning below
$10,000 per year saves $1.24 under the current system, while the well-off households lose nearly $31 per
month.

Crandall and Waverman calculated that the total national welfare gain that could be realized through a shift
to cost-based pricing would be between $2.5 and $7 billion a year, depending on the proxy model used.
What’s more, they conclude that the overall gains by average low-income consumers (who do not use long
distance services very much) from the current system come at a very high cost to middle-income and
wealthy consumers, as well as long distance companies (from lost producer surplus; see Figure 9).

Because the amount of rate distortion is directly correlated with the percentage of rural households in a
given state, the states where average consumers would reap the most benefit from cost-based pricing are the
states with the largest share of rural subscribers (see Figure 10).
                                                                                                  Figure 11: Percentage of
                                                                                                Households Benefiting from
These findings are quite surprising given the strong level of support                          Cost Based Rebalancing (1996)
for universal service policies among rural politicians and                                                            % of HH Benefiting
regulators. In rural states, the residents who make significant                                                        from Rebalancing
                                                                                                                           by Model
amounts of long distance calls, and/or who also reside in urban                                  State       Area      BCPM     HCPM
areas within these states heavily subsidize the rural households.                                          Urban          50.4%   100.0%
But in urban states, because they have so few rural customers to                                           Suburban       36.4%    17.4%
                                                                                                  CA
                                                                                                           Rural           7.1%     0.0%
support, the burden is not nearly as great as the one that falls on                                        Total          41.7%    51.3%
urban customers in rural states. Clearly this is an example of the                                         Urban          71.6%   100.0%
“median voter” theory in action. Indeed, Crandall and Waverman                                             Suburban       78.7%   100.0%
                                                                                                  MS
                                                                                                           Rural          34.3%    34.3%
recognized that because the use of long distance services is highly                                        Total          41.1%    45.2%
skewed, the benefits of cost-based pricing will also likely be skewed                                      Urban          61.9%   100.0%
on a household basis. Though the efficiency gains are large on                                             Suburban       37.6%    36.3%
                                                                                                  NJ
                                                                                                           Rural            NA       NA
average, in most states rebalancing creates benefits for between                                           Total          40.2%    43.1%
40% and 51% of households statewide (see Figure 11).                                                       Urban          54.0%    66.0%
                                                                                                           Suburban       52.0%    81.0%
                                                                                                  WY
                                                                                                           Rural          36.8%    29.8%
A final finding by Crandall and Waverman, one partially confirmed                                          Total          42.1%    47.2%
by other researchers, is worth illustrating. Using 1990 Census data       Source: Crandall & Waverman, 2000
(by town) matched against pricing data provided to the authors by
LEC’s, they found that the Lifeline monthly subsidy program had no significant effect on penetration. They
also found that increases in installation charges lowered penetration. 73 Other researchers have concluded
that Lifeline and Linkup do increase penetration, but that the effect size is small relative to the cost. 74


73
   They also found however, that Linkup was negatively associated with penetration, a finding that makes little sense, and is contrary to their
other finding of decreases in penetration with increasing installation charges, and is also contrary to other studies on this matter. Given that
in their model only two states (DE and IL) lacked Linkup programs (because regulators in those states decided against implementing them as
penetration was already high), it is likely that this bizarre result is due to endogeneity bias.
74
   See Garbacz et. al. “Assessing the impact of FCC Lifeline and Linkup programs on telephone penetration”, Journal of Regulatory Economics,
11, 67-78, 1997; also Eriksson et. al. “Targeted and untargeted subsidy schemes: Evidence from post-divestiture efforts to promote universal
service”, Journal of Law and Economics, 41, 477-502.

                                                                                                                                              24
What conclusions can we draw from Crandall and Waverman’s work, which is based on 10-year old data?
Since this study was published long distance access rates have fallen from about 4 cents per minute (on
both ends of the call) to about 2 cents per minute. Also over this time, subscriber line charges have risen
from $3.50 to about $6 per month for residential primary lines. The SLC is used by LEC’s to recoup costs,
and this, along with increases in efficiency, has allowed the FCC to shift some of the burden away from long
distance access charges. However, over this same time period the size of the federal high-cost fund has more
than doubled, reaching nearly $4 billion last year. So while long distance access charges have decreased,
contributions by long distance carriers have increased, and these contributions are financed by markups on
elastic long distance and cellular services. So while Crandall and Waverman’s exact figures may not be
accurate (and some clearly are not, as rural populations within many states has shifted somewhat since they
conducted their work), their overall conclusion that financing USF via markups on elastic services is
fundamentally inefficient remains very valid.


USF Contribution Reform Alternatives

The following contribution reform alternatives will be assessed against the previously outlined criteria:

        •    Eliminate USF – full rate rebalancing
        •    Status quo – let present trends continue
        •    Numbers method
        •    Numbers and capacity

Contribution Option 1: Eliminate USF and Fully Rebalance Rates

The consequences of this option are essentially outlined in the previous exploration of Crandall and
Waverman’s work. While it may be the best option from an efficiency standpoint, total telephone
penetration will drop, and this will disproportionately affect the poor. Though it should be noted that since
Crandall and Waverman’s work, there has been a marked shift away from toll telephony towards fixed
monthly fees. This will obviously diminish the magnitude of the efficiency gains.

The costs of this option will be born primarily by the poor on average, and the benefits will be reaped by the
most wealthy, primarily those in rural states.

This alternative does not maximize the adoption of telecommunications services, as it will lead to a
potential loss of rural subscribers on the order of several percentage points. Availability will likely decrease
in rural areas, as previously subsidized ETC’s find it uneconomical to serve in high-cost areas, even with the
freedom to charge cost-based prices. This alternative certainly does not maximize the affordability of local
telephone service in rural areas, if affordability is defined as prices equal to current levels. On the other hand,
this option does increase the affordability of long distance services in all areas of the country. However,
because a customer must first subscribe to local service prior to using long distance service, those who are
most poor and place a moderate amount of long distance calls will not see an increase in affordability.
Furthermore, given that under rate rebalancing cellular companies will no longer be required to contribute
to USF, their rates are likely to decrease, on average. Increasingly, even low-income consumers are adopting
cellular service in addition to wireline service, and may choose to “cut the cord” and switch to exclusive use
of wireless telephony.

The effect this alternative will have on advanced information services is unclear. It is possible that in some
markets ILEC’s subsidize below-cost telephone service with above cost DSL charges. However, recent
comments by an officer of SBC (now AT&T) seem to indicate that ILEC’s are using forced bundles of DSL
with POTS to increase the use of the latter service. 75 Most likely, in high-cost rural areas, where USF monies
are currently used to maintain and upgrade networks to make them broadband capable, the cost of DSL will

75
     Comments of SBC CFO Rick Lindler. See http://www.thestreet.com/pf/tech/scottmoritz/10226542.html

                                                                                                                 25
increase substantially in the absence of high-cost support. In some urban markets DSL costs may drop as a
result of rate balancing. Thus the net effect may be a widening of the digital divide.

This option does minimize deadweight loss. Since there are no more “taxes” (implicit or explicit) the
market will presumably reach an efficient equilibrium. On the whole it minimizes consumer surplus loss,
but if markets are examined at local levels, some rural customers will experience a loss of consumer welfare.
Likewise, at the national level producers will experience surplus gains, especially long distance carriers.
However, rural carriers who lose subsidies and then increase prices will likely experience losses in producer
surplus.

The costs of this alternative fall heavily on the rural poor, who are the least likely to use interstate
telecommunications services, and thus are the least likely to gain from rate rebalancing. The benefits of this
option accrue primarily to those residents of rural states who make significant use of interstate
telecommunications services -- primarily the wealthiest households living in the cities of rural states.

The effects of rate rebalancing on concentration of content access are ambiguous. On the one hand, cost-
based rates may encourage market entry by new efficient firms using the latest advancements in
communications technology such as VoIP or WiMax. On the other hand, in some areas the fixed cost of
deployment may be too high for competitive entry, even under increased rates. This would leave control
over communication lines in the hands of incumbents. Cable firms offering VoIP telephony would likely
gain significant market shares under rate rebalancing, and thus reverse the trend of increased DSL market
shares seen in the past several years.

This option is completely unfeasible in the current political climate, and is only seriously discussed by
economists and advocates of antitrust-based regulatory policies.

Contribution Option 2: Status Quo – Let Present Trends Continue

If nothing is done to change the current USF system, then it is very likely that the contribution factor
assessed on interstate telecommunications carriers will continue to require quarterly increases. This is
primarily due to the high likelihood that consumers will substitute away from traditional long distance
services, adopting alternative methods of communication such as VoIP, email, and instant messaging –
none of which is subject to USF contribution.

Under the status quo there will be little change in the current nationwide telephone penetration level.
However, it is very likely that long distance charges and cellular charges will continue to grow as a result of
these carriers’ increased contribution burden. Higher prices means marginal consumers will choose to use
less, or none of these services. However, gradually increasing long distance and cellular rates will make
VoIP offerings by cable companies a more attractive alternative, both to adopt and deploy.

Letting present trends continue will likely lead to an increased urban-rural broadband adoption gap. Unless
demand elasticities for broadband decrease significantly in rural areas, most providers will still consider
these markets too unprofitable to operate in. Wireless carriers, who could offer a third platform broadband
alternative, will see their average costs rise as a result of the increasing contribution burden. Consequently,
their wireless broadband services will likely remain too expensive for many consumers.

Facing little competition from LEC’s in video markets, cable broadband providers will likely keep prices
constant as they attempt to capture LEC marketshare with bundled VoIP offerings. This competition will
take place in currently served markets, and little change will be seen in rural areas without current access to
advanced information services

As demonstrated in this report, the current structure of USF results in large inefficiencies by placing
contribution burdens on services with high demand elasticities. Crandall and Waverman’s assessment of $2
to $7 billion a year in gains from rate rebalancing seems like a reasonable estimate. Jerry Hausman

                                                                                                                  26
estimated that in 1999, every $1 billion raised through above-cost long distance fees resulted in $1.25
billion in efficiency losses. 76 Given the current fund size of $7.2 billion, that’s approximately $9 billion per
year in deadweight losses.

Under the current system the burden of universal service falls disproportionately on those who are heavy
long distance users. On average, these are businesses and wealthy consumers. However, there are large
variances of long distance use within income brackets. Furthermore, some large businesses are switching to
VoIP service or dedicated access lines, shifting the contribution burden back on individual customers. Also,
it is likely that businesses pass on higher communications costs to customers. The largest beneficiaries
under the current system are those rural customers who do not use significant amounts of long distance
services.

Under the status quo the pattern of recent industry consolidation and vertical integration is likely to
continue. There is no reason to believe that the duopoly stronghold on access will reverse, and in all
likelihood it is possible that cable companies will regain some of their lost marketshare from LEC’s as they
begin to offer VoIP-TV-Internet “triple-play” services.

Nothing is more politically feasible in Washington than “doing nothing”. Members will continue to
introduce reform legislation, but much of it has little chance of passing. There is a fundamental difference
in attitude towards USF in the relevant committees in the House and Senate. If a compromise bill is
reached this year, it will likely be another vaguely worded statute (much like the 1996 Act), and result in
legal challenges that stall any implementation of reform alternatives.

Contribution Option 3: Numbers Based Method

This option would eliminate the collection of USF revenue via the contribution factor assessment on
interstate carriers, and replace it with a flat monthly fee on each telephone number, and possibly each IP
address. 77 This essentially shifts the contribution burden from usage to access, something that is very
desirable from an efficiency standpoint, because demand for access to a basic line is far less price-elastic
than that for long distance calls. This alternative is favored by FCC Commissioner Kevin Martin, and is not
a priori opposed by Senate Commerce Committee Chairman Ted Stevens, or by House Energy & Commerce
Committee Chairman Joe Barton. However, this alternative does come with some distributional concerns.
Namely, it will shift some of the contribution burden away from heavy long distance users onto light users.
This means businesses will pay less under this alternative.

How much extra per month will this alternative cost? Given recent data showing approximately 500
million telephone number assignments, the current $7.2 billion dollar USF budget would result in an
additional $1.20 per month in the form of a number charge. 78 However, some of this increase would likely
be offset by decreases in long distance and wireless bills.

Incorporating IP addresses may be troublesome. IP addresses are assigned to each computer connected to
the Internet. Most home computers are not assigned a fixed IP address, but a different addresses each time
they connect to the network. Also, each website is associated with a unique IP addresses, assigned by the
site’s host. Exact numbers are difficult to obtain, but it appears that there are at least 320 million IP




76
   Jerry Hausman and Howard Shelanski, “Economic Welfare and Telecommunications Regulation: The E-Rate Policy for Universal Service
Subsidies”, Yale Journal on Regulation, 16, 19, 1999.
77
   IP addresses identify a unique connection to the Internet. Given that most cable and DSL subscriptions are not associated with a “fixed” IP
address, the provider would have to bill each customer as if they did have a unique IP address. However, this plan would likely need to
provide a “safe harbor” for home users that have both a broadband connection and a VoIP telephone number, as well as a safe harbor for
dial-up ISP’s. The impetus for these safe harbors would be to not “double charge” customers, while still preventing arbitrage by customers
that use IP to IP telephony.
78
   FCC, “Number Resource Utilization in the United States as of December 31, 2003”, May 2004.

                                                                                                                                            27
addresses assigned in the U.S. as of 2005. 79 Imposing a tax on IP addresses that is equal to that levied on
phone numbers may not be wise, as it would create disincentives in the growing Internet market.

This alternative would have almost no negative impact on overall telephone penetration, especially if
recipients of Lifeline and Linkup support were not subject to the new fee. In fact, given the presumed lower
charges on long distance and wireless service (per minute charges), adoption and use of telephony could
increase. Rural LEC’s would experience no loss of high-cost support. Most wireless carriers would benefit,
but there would be some who incurred losses as a result of the higher fixed monthly fee (i.e. some
customers who only occasionally use the wireless phone might not subscribe due to the higher monthly fee
-- a fee not offset by lower long distance rates; some carriers such as Metro PCS who cater to intrastate
customers would probably be net losers under this alternative).

Alone, it is unclear what effect this alternative would have on advanced information services. However, a
wider contribution base means there may be new funding available to subsidize general broadband
provision.

Shifting the contribution burden to inelastic fixed access charges and away from elastic usage-based charges
will result in efficiency gains. Again, these will likely be on the order of billions in deadweight loss
reduction, billions in consumer surplus gains, and billions in producer surplus gains.

A major concern with this alternative is the fact that it shifts the burden of contribution more onto
consumers and away from businesses, but the magnitude of this shift may not be large (see discussion of
CBO analysis below). In the long run, as we saw with the status quo option, businesses may adopt
methods that allow them to shift this burden onto consumers, regardless of the USF contribution system.
This alternative does confer benefits on the customers who are the highest users of long distance services,
which again tends on average to be businesses and the wealthy. The low-income low-volume long distance
user will see an increase in their monthly bill, unless these consumers are exempted from contribution
requirements -- something I strongly recommend. Overall this method places the support burden evenly
among high-cost and low-cost rural and urban consumers. This is clearly an improvement from the current
system, which places a disproportionate burden on rural state customers that live in low-cost urban cities.

It is possible that this alternative will lower the overall burden on the wireless industry, encouraging further
deployment of 3G broadband services. A numbers based contribution mechanism would by definition
capture VoIP customers, and could discourage use of these services at the margin – though they are currently
at or below the cost of traditional wireline services. Thus the net effect could still leave VoIP the less
expensive alternative. But overall there is likely to be little change in the concentration of access by the
leading service providers.

This option is politically feasible, and may not need Congressional approval prior to implementation. A
reading of the 1996 Act indicates that the Commission already possesses the authority to collect funds in
such a manner (though the language is ambiguous enough that courts could overturn the move to a
numbers system, unless Congress enacts new legislative language granting this scheme expressed approval).
As mentioned, there is currently little opposition to this alternative, both within the Congress and the
Commission. Most industry groups are behind it, with the exception of some small rural LEC’s who benefit
disproportionately under current rules that treats them differently from large LEC’s and competitive ETC’s.

Contribution Option 4: Numbers and Capacity Based Method

This option eliminates the contribution factor assessment on interstate telecommunications carriers, and
replaces it with a flat monthly fee levied on each phone number (as in option 3), but also raises revenues by

79
  There are 4 billion possible IP addresses, and approximately 8% of these have been assigned to ARIN, the American Registry for Internet
numbers. But this figure does not account for historical assignments prior to ARIN’s establishment. See
http://www.apnic.net/community/presentations/docs/icann/apnic-gac-capetown-20041130.ppt

                                                                                                                                            28
assessing a fixed fee based on capacity. What this means is that if a business has a large data conduit to the
Internet (such as an OC-3 or T-3) which is not associated with any telephone numbers, there will be a USF
fee based on the size of this connection. Similarly, if a household customer has a DSL connection, a fee
based on the size (capacity) of that connection would be charged. An important proponent of this option
(CTIA, the wireless industry group) proposes a “no double counting rule”. That is, if the DSL line has an
associated VoIP phone number, the capacity charge would be waived. In an interview, the CTIA’s VP of
Regulatory Affairs related that the idea behind adding capacity is to keep the level of business contributions
very close to what they currently are.80 If a business were currently paying $650 per month in USF fees, then
under the numbers-capacity system they would still pay about $650 per month. This USF assessment
essentially lessens the burden placed on consumers that would exist under a pure numbers-based approach.

Most of the conclusions concerning the numbers based approach are similar under this enhanced
alternative. However, there is one potentially troubling difference. This option potentially introduces a tax
on advanced information services, separate from any association with telephony. Taxing broadband may
inhibit adoption at the margins, given the fact that demand for high-speed connections remains somewhat
elastic in many areas of the country. Thus it is my recommendation that this option be paired with a
subsidy aimed at increasing overall broadband availability and adoption.

This option could possibly lower deadweight loss and surplus losses relative to the status quo. However,
the capacity portion of this alternative places taxes on services that are somewhat elastic. It’s unclear what
the net effect would be. It is likely that the demand elasticities for long distance service are higher than the
demand elasticity (by businesses) for data capacity. If so, this option would be an improvement from an
efficiency standpoint when compared to the status quo, but not as much as a pure numbers approach.

The conclusions concerning equity presented above for the numbers-based alternative are similar for this
hybrid option. The main difference is that residential consumers would not be forced to take on as large of
a burden, as business contributions would largely remain the same.

There is likely to be little change in the concentration of access by the leading service providers under this
option.

This option, like the numbers only alternative, is very feasible. However, there may be pushback from large
businesses who would prefer a purely numbers based approach. But if the overall per line increase on
residential customers is less under this option, lawmakers are likely to prefer it. Also, the support of CTIA
and its clout on the Hill increases the likelihood of this option being adopted. The caveat outlined above
concerning the Commission’s legal ability to purse this option without prior legislative approval still
applies. The capacity aspect introduces a further legal complication. It is not clear that the Commission has
the legal authority to extend USF obligations based upon a user’s capacity. This is because the
Commission’s jurisdiction over advanced information services falls under its Title I ancillary authority.
Some experts feel that the Commission’s ancillary authority under Title I would permit it to assess USF fees
on information services, but if it did so it would surely face legal challenges.

Congressional Budget Office Assessment of Contribution Alternatives

In March of 2005 the Congressional Budget Office released a report, “Financing Universal Telephone
Service”, which evaluated contribution alternatives for USF. The CBO used a FCC model along with data up
to 2001 to predict the various distributional consequences of alternative contribution policies. Their results
are presented below.




80
     Personal communication with the author, March 31 st 2006.

                                                                                                                   29
                        Figure 32: Distributional Effects of Alternative USF Contributions Policies

                                      Share of Total Contributions              Average Monthly   Share of Total Contributions
        Contributions
                                           Long-Distance             Wireless     Charge per      Residential      Business
           Policy            ILEC's
                                              Carriers               Carriers     Household       Customers       Consumers
        Current System
                               28%               51%                   22%          $2.09            43%              57%
          (as of 2003)
        Current System
                               31%               37%                   31%          $2.26            44%              56%
           (est. 2007)
        Numbers Only
                               55%               13%                   32%          $2.47            46%              54%
           (est. 2007)
        Capacity Only
                               43%               22%                   33%          $2.28            45%              55%
           (est. 2007)
         Source: CBO, 2005

These results indicate that under a number-only or capacity-only approach that the average household USF
contribution will change little from the status quo. Also, the distributional burden between residential and
business customers would change little under either plan compared to the status quo. Under a numbers-
only system, the contribution burden does shift towards ILEC’s and away from long-distance carriers.
However, the industry consolidation of recent years is erasing the distinction between ILEC’s, long distance
carriers, and wireless providers. The two largest ILEC’s, AT&T and Verizon, are themselves two of the largest
long distance and wireless providers. Thus it’s easy to see why the number-capacity enjoys widespread
support amongst large carriers, but is opposed by small rural LEC’s -- even though these rural LEC’s pass
through their contribution burden to their customers as a “regulatory recovery fee”.


USF Distribution Reform Alternatives

The following distribution reform alternatives will be assessed against the previously outlined criteria:

    •   Eliminate USF – full rate rebalancing
    •   Status quo – let present trends continue
    •   Include broadband providers along with all current ETC’s
    •   Only distribute funds to broadband capable carriers

Given that the efficiency criterion mainly applies to contribution alternatives, it will not be used to evaluate
distribution options.

Distribution Option 1: Eliminate USF and Fully Rebalance Rates

The consequences of this option are discussed above in the contributions alternatives section, as this option
makes changes to both contribution and distribution methods. However, the effects of this option on the
Schools and Libraries and Rural Health Care programs would be large – as they would lose all funding.

In addition to the assessment of this option presented above, eliminating the E-Rate and RHC programs
would profoundly impact access to telecommunications and advanced services. Given that the RHC
program uses little USF monies, the effects here would be on the poorest clinics operating in the highest
cost areas. The impact on schools and libraries would be profound, likely leading to the complete loss of
Internet services in the poorest schools and library districts. This option is completely unfeasible in the
current political environment.
Distribution Option 2: Status Quo – Let Present Trends Continue

The consequences of this option are discussed above in the contributions alternatives section.

Distribution Option 3: Include Broadband Service Providers



                                                                                                                                 30
Subsidizing broadband is justified from an efficiency standpoint, based on the associated network
externalities. However, the expansion of the USF program may or may not be the best method for achieving
increased broadband adoption.

Obviously a new broadband subsidy would allow previously unserved customers to have access to this
advanced information service. The key questions are how big would the subsidy be, who would qualify for
it, and how would the extra revenue be collected? First, it should be noted that nearly 80% of ILEC lines are
already broadband capable. Therefore any high-cost broadband subsidy should only go towards efforts to
upgrade the remaining lines. But subsidizing a single technology is antithetical to the nature of
convergence. Cable companies currently reach 95-97% of U.S. households, and 93% of these lines are
cable-modem ready. It doesn’t seem unreasonable to use USF funds to reach the remaining 7 percent. Also,
municipalities are currently leading the efforts to deploy broadband in unserved areas; they too should
qualify for these funds. Therefore any extension of USF to broadband should toss aside the old regulatory
emphasis on local exchange carriers, and be distributed in a technology neutral manner.

However, extending broadband to currently unserved areas will be expensive (there’s a reason these areas
are not currently served) and the cost per customer added could be quite high. If the per line deployment
cost were $1000 (a potentially low estimate), a $500 million per year fund would add a maximum of
500,000 lines -- increasing the nation’s penetration level by about 1 percent. This $500 million could be
raised via the numbers alternative by an additional $0.08 per month fixed charge (on top of the $1.20 used
to offset the current needs of the fund, if traditional wireline voice service were to remain supported). 81

The costs would depend upon the associated contributions methodology. The equity of the benefits of
subsidizing broadband is somewhat of a mixed bag, and depends on the program design. As noted above,
Goolsbe demonstrated that the highest gains in surplus are realized by subsidizing the construction of new
networks. But the benefits of these networks are conferred almost exclusively on those who value
broadband the most -- the wealthy rural consumer. This is a major critique of the current USF system -- that
it benefits wealthy rural households. However, if a wealthy rural household makes a large volume of long
distance calls, they themselves actually bear a large portion of the USF burden.

This distributional concern could be partially offset by a Lifeline/Linkup style low-income subsidy for
broadband use. But again, the Linkup portion of the program would need to help some qualifying
consumers “linkup” to a personal computer. This would expand the needed size of the fund significantly,
which is highly unlikely given the current political environment (current Senate legislation only allocates a
maximum of $500 million per year for construction of broadband networks in unserved areas). Thus in
order to achieve broadband adoption by low-income consumers, community groups and non-profits
supplying donated computers will need to be part of the coordinated efforts.

Assuming the low-income user has access to a home computer, the cost of monthly broadband service will
likely need to fall to the $10 to $15 per month range. A one-billon dollar per year subsidy could fund
nearly 5.5 million low-income consumer’s broadband connections (assuming full take-up with a $15 per
month subsidy). This is a very significant number, and would certainly help close the economic digital
divide.
It’s unclear what this new subsidy would mean for access and content control. LEC’s (including their
wireless divisions) and cable companies are likely to be the largest recipients of the subsidy (if they were
portable or user-designated), but if the funds were allocated via a bidding process (see below), then it’s
possible that new private entrants or municipalities could receive a sizeable share of the funds.

Some form of broadband inclusion in USF is very feasible. The Stevens bill voted out of Committee
included a $500 million unserved areas subsidy. The House however is another matter. Chairman Barton
is somewhat hostile USF, and thus some sort of compromise will need to be hammered out.

81
  Assuming there are 500 million telephone numbers in the U.S., $500 million could be raised via an annual subscriber line charge of $1.00,
or about 8 cents per month.

                                                                                                                                              31
Distribution Option 4: Only Broadband Capable Providers Would Qualify for USF

This option was essentially the centerpiece of the USF title in the first draft of the Stevens’ bill. As
mentioned, nearly 20% of all ILEC loops are not broadband capable, and it is likely the majority of these
are located in high-cost areas. In order for those loops to continue to receive high-cost funds, the LEC
would need to upgrade them. This will raise the overall per-line cost, requiring higher levels of high-cost
funding. What’s more, though cellular companies are poised to rollout 3G broadband services (though the
initial offerings seems to only be a complement to, not a substitute for wireline broadband), many of the
smaller rural wireless ETC’s are not, and would lose funding under a broadband-only requirement.

This option would likely result in a net increase in the availability of telecommunications and advanced
information services. However, if small rural LEC’s and ETC’s could not secure the capital needed to
upgrade their networks, it is possible that rural areas would see rebalanced local charges, without the
corresponding drop in long distance charges (assuming the current contributions models stays in place).
This in turn could result in a net loss of rural subscribers. But it is likely that most LEC’s and ETC’s would
invest in network upgrades, as would cable VoIP providers.

The economic costs of this option would depend on the contributions model. As in the above alternative,
the benefits of broadband expansion would largely accrue to middle-class, upper-middle-class, and wealthy
customers, unless a low-income broadband subsidy was a part of the new program.

As stated above, the net effect on content access depends on the subsidy allocation method.

This option is somewhat politically feasible, though not unless there is a long transition period.
Furthermore, such an option would be unlikely to gain widespread Congressional support without
provisions protecting rural-LECs.

The Question of Competition – Portable Subsidies and Reverse Auctions

One of the underlying structural problems facing the universal service fund is the presence of portable
subsidies, which are both a blessing and a burden. Encouraging competition by providers for basic service
subsidies also produces competition in secondary services such as voice mail or bundled long distance
calling plans. This increased competition in vertical market segments benefits consumers by lowering the
prices for such services. However, each ILEC customer defection to an ETC causes the per-line subsidy to
increase, ballooning the total size of the fund. Basing the subsidy on each carriers’ own cost is an obvious
partial resolution to this problem. But this presents informational difficulties for both federal and state
regulators, who may struggle to accurately estimate each carriers embedded or forward looking costs.

Portable subsidies also complicate matters by allowing a single customer to receive service from multiple
universal service-supported carriers. Most commonly, a household will have a single residential fixed line
provided by the ILEC, and one (or several) cellular phones provided by wireless ETC’s.

One possible solution to the problem of multiple carriers is to allow each customer to designate a “primary
line”, thus converting the USF support into a voucher. The carrier providing that line would then receive
that customer’s per-line share of universal support, while the company (or companies) that provides the
customer’s secondary line would not receive support for their services. During the debate over the 1996 Act,
Senator John McCain (R-AZ) introduced an amendment that would establish primary line subsidies, but it
was defeated. 82




82
 S. Rep. No. 1276, 141 Cong. Rec. 8266, 1995 (text of amendment); No. 251, 104th Cong., 1st Sess., 141 Cong. Rec. D719, 1995 (vote on
amendment).

                                                                                                                                        32
In 2004, the Federal-State Joint Board proposed limiting USF support to just a single customer-designated
primary line. 83 However, the FCC never acted on this recommendation. The Joint Board members who
dissented in the 2004 decision were concerned that the designation of a single USF eligible line would
discourage investment in rural areas, as well as be administratively difficult to implement. 84 Prior to issuing
the recommendation, the Joint-Board received a letter from Senators of both parties stating that a primary
line designation policy would be “a major step backward that would thwart the essential purpose of
universal service”. 85 Thus it appears that a voucher system for controlling the growth of ETC-related costs
may not be politically feasible. However, it may be possible to implement a system that allows limited
competition while also reducing information-asymmetry related costs. This could be possible under a
system of “reverse auctions”.

The idea of using reverse auctions to determine universal service fund allocations has been under
consideration ever since the 1996 Act became law. The California Public Utilities Commission undertook a
study of the proposal, concluding, “an auction mechanism appears to be the most efficient mechanism for
reviewing the subsidy amounts in the future”.86 Though the FCC has never formally considered this option,
Chairman Martin has indicated interest in its potential use.87

An auction for universal service support could be designed to maintain some level of competition within a
given geographical area. Instead of awarding support to the lowest bidder, the Commission could design a
process that would award subsidies to the lowest bidder plus all other bidders within a certain range of that
figure. 88 This built-in ex ante competition comes with a trade-off: the benefits of in-market competition and
the costs associated with giving subsidies to less efficient bidders.

One of the advantages of reverse auctions is that they are technology-neutral. Thus new innovative
technologies are rewarded, and outdated technologies are rightly abandoned. If a cellular provider and a
WiMax VoIP provider can offer telephony service for far less than an ILEC, they would be the only recipients
of subsidies. Consumers would still have choices amongst providers and platforms, and the costs to
taxpayers would likely be lower.

Given these potential benefits, the Commission should seriously evaluate using reverse auctions. However,
while auctions may seem like a good idea in theory, there are many practical difficulties that will need to be
addressed. First is the deciding the geographical size of the subsidy area, which comes with trade-offs. If
the area size is small, it reduces opportunities for cream skimming and enables entry on the part of small
competitors. As mentioned, telecommunications markets are characterized by economies of scope, thus
“the smaller the area size the higher the cost synergies across areas”.89 A second serious practical concern is
the presence of historical asymmetries. ILEC’s who have already constructed networks have low short-term
marginal costs. New competitors may find it hard to compete in auctions against this historical sunk-cost
advantage.




83
   “Recommended Decision”, Federal-State Joint Board on Universal Service, 19, FCC Rcd 4257, 2004.
84
   See dissenting statement of Comms. Adelstein, Thompson, and Rowe, in “Recommended Decision”.
85
   Ibid. referencing a December 18th 2003 letter from Senators Dorgan, Burns, Snowe, Johnson, Baucus, Lincoln and Daschle.
86
   See California Public Utilities Commission, “Rulemaking on the Commission's Own Motion into Universal Service and to Comply with the
Mandates of Assembly Bill 3643; Motion of the Officie of Ratepayer Advocates for a Ruling Requiring Expidited Review of the Cost Proxy
Model Results”, R.95-01-020, January 24th 1995.
87
   “FCC's Martin Likes 'Reverse Auction' for Universal Service”, State Telephone Regulation Report, April 07, 2006.
88
   A proposal by GTE to the CPUC in 1997 designed this as “if at least one bid does not exceed the lowest bid by more than 15 percent of the
sum of the lowest bid and the basic service price, then all bids within that range will be accepted; if no competing bid is within the range
described above, but one is within 25 percent, then the two lowest bids will be accepted.” See “Comments of GTE submitted to the
California Public Utilities Commission, Auction Proposals for Universal Service”, 1997.
89
   Laffont and Tirole, p. 245.

                                                                                                                                           33
Appendix

Comparative Statistical Analysis of Broadband in the OECD
Cross-sectional econometric analysis methods were employed in order to better understand the differences
between the broadband performances of the 30 OECD nations. The results presented below are preliminary
observations intended to serve as the basis for further study. All data presented in this appendix (unless
otherwise noted) comes from the OECD, in particular the 2005 “OECD Communications Outlook”. This
appendix begins with an examination of the individual correlations between certain factors and broadband
penetration. I then construct a single model based on a full set of predictors. OECD data on student
computer use is examined using similar methods. Interpretation of data gathered at such a large aggregate
level is certain to raise concerns about the “ecological fallacy”. Thus, the conclusions here are merely
suggestive.

What factors are correlated with broadband penetration?

A starting point in the attempt to characterize and understand America’s “broadband problem”90 is to
determine what factors are correlated with broadband penetration.

Likely factors include:

•    Median household income
•    Population density
•    Percentage of population living in urban areas
•    Education attainment (measured as years of formal education)
•    Poverty rate
•    Broadband price

Other factors that could possibly be important include regulatory conditions, market competition, and
government investment in, and incentives for, infrastructure development.

To begin, I examine the correlation between broadband penetration and the geographic factors of
population density and “urbanicity” (the percentage of the population living in urban areas). At first glance
the significance of population density seems likely. Several of the leading countries in terms of broadband
penetration are also the two most densely populated nations in the OECD. However, Iceland is one of the
most sparsely populated OECD nations, yet it has the highest level of broadband penetration.

But a close examination of the relationship between population density and broadband penetration reveals
no significant correlation between these two variables (r=0.26, p=0.16). This result is not surprising, as the
theoretical basis for why population density would effect a country’s broadband adoption level is weak.
Countries like Iceland and Canada have vast stretches of unpopulated land, but also have high percentages
of their population living in urban areas. This suggests the relevant factor is not number of persons in a
given total area, but may be the proportion of the population living in close proximity.

Indeed, the percentage of a country’s population living in urban areas does appear to play a more significant
role than population density, but the relationship is very weak. Approximately 12% of the observed
variation in broadband penetration is explained by urbanicity, and this relationship is weakly statistically
significant (r=0.34, p=0.065).




90
  This phrase was popularized by Charles H. Ferguson; see “The Broadband Problem: Anatomy of a Market Failure and a Policy Dilemma”,
2002.

                                                                                                                                       34
Median household income is significantly correlated with broadband penetration. In this bivariate linear
model, median household income explains 32% of the observed variation in broadband penetration, and
the result is highly statistically significant (r=0.57, p=0.001)

There is also a significant negative relationship between poverty and broadband penetration. 91 In this
bivariate model, the poverty rate explains 29% of the observed variation in broadband penetration, and the
result is highly statistically significant (r=0.54, p=0.004).

Education is also moderately correlated with broadband penetration. Approximately 27% of the observed
variance in broadband penetration is explained by differences in (years of formal) education in this
bivariate comparison. This relationship is highly statistically significant (r=0.52, p=0.001).

Also, broadband penetration is significantly negatively correlated with price, in units of U.S. purchasing
power parity (r=-0.45, p=0.015; with Turkey dropped as an outlier). Due to data limitations and the low
presence of cable modem in Europe, DSL prices were used.92

Modeling Broadband Penetration

To better understand how the U.S. is performing relative to the other countries of the OECD, I construct an
econometric model based on the predictive factors discussed above.

The model is specified as:

           penetrationi = β0 + β1 (mhhinc)+ β2 (poverty)+ β3 (urban) + β4 (yreduc) + β5-7 (access policy) + εi

Where:

penetration = total broadband penetration as of June 2005
mhhinc = median household income
poverty = the share of individuals with equivalised disposable income less than 50% of the median for the
entire population
urban = percentage of the population living in urban areas
yreduc = years of formal education
access policy = dummy variables for local loop unbundling, subloop unbundling, and Bitstream access
policies

Results are presented below.




91
   The OECD data used for this analysis defines poverty rates as “the share of individuals with equivalised disposable income less than 50% of
the median for the entire population”. See “Income Distribution and Poverty in OECD Countries in the Second Half of the 1990s”, Michael
Forster and Marco Mira d'Ercole, OECD Social, Employment, and Migration Working Paper #22, March 10 2005.
92
   Prices are for mid-tier DSL service from leading service provider within each country, not including taxes, as of March 2006.

                                                                                                                                            35
                       Figure A1 – Results of Regression Model – Broadband Penetration

                        OLS: Total BB Penetration (OECD)
                                                Coefficient      Coefficient     Coefficient     Coefficient
                                                (t-statistic)    (t-statistic)   (t-statistic)   (t-statistic)
                        Median HH                0.251***        0.230***         0.230***        0.213***
                        Income
                                                  (3.59)          (3.97)           (3.74)          (2.97)
                        (thousands US$)
                                                                 -0.722***        -0.722***       -0.659**
                        Percent in Poverty
                                                                   (-3.05)         (-2.98)         (-2.37)
                        Percent Urban                                               -0.001          -0.009
                        Population                                                 (-0.02)         (-0.11)
                                                                                                     0.45
                        Years of Education
                                                                                                    (0.48)
                                                    4.76         12.05***          12.14*            7.38
                        constant
                                                   (2.89)          (3.20)           (1.74)          (0.60)
                                               R2 = 0.316       R2 = 0.571       R2 = 0.571      R2 = 0.576
                                               N = 30           N = 27           N = 27          N = 27
                        * = significant at 10% level; ** = significant at 5% level; *** = significant at 1%




          Figure A2 – Results of Regression Model with policy variables– Broadband & DSL Penetration

                          OLS: Total BB Pen. (OECD)                                     DSL Penetration
                                               Coefficient            Coefficient         Coefficient
                                               (t-statistic)          (t-statistic)       (t-statistic)
                          Median HH Income           0.213***           0.236**               0.195***
                          (thousands US$)             (2.97)             (2.46)                (3.43)
                                                      -0.659**         -0.589**            -0.502***
                          Percent in Poverty
                                                       (-2.37)          (-2.05)              (-2.94)
                          Percent Urban                -0.009            -0.013              -0.0213
                          Population                   (-0.11)          (-0.14)              (-0.40)
                                                        0.45            0.0008                -0.691
                          Years of Education
                                                       (0.48)             0.00               (-1.13)
                          Local Loop                                      4.63                 2.75
                          Unbundling                                    (1.37)                (1.37)
                          Subloop                                         -4.03              -4.21**
                          Unbundling                                    (-1.30)              (-2.28)
                                                                          1.17               3.16**
                          Bit Stream Access
                                                                        (0.46)                (2.09)
                                                        7.38              9.46                12.45
                          constant
                                                       (0.60)           (0.74)                (1.64)
                                                   R2 = 0.576       R2 = 0.635          R2 = 0.747
                                                   N = 27           N = 27              N = 27
                          * = significant at 10% level; ** = significant at 5% level; *** = significant at 1%




These models suggest that a country’s average median household income and proportion of population
living below the poverty level are two important determinants of broadband penetration for countries in the
OECD. The effects of the policy variables are somewhat ambiguous, and given the nature of the difficulty in
capturing the implementation of these policies with dummy variables, caution is warranted.

Regression diagnostics were performed on both the linear model and the log-log model. Both models
appear to be homoskedastic according to the results of hettests. Variance infllation factors (VIF’s) indicate
that there is a low possibility of multicolinearity bias. Results of linktest and ovtest diagnostics seem to
indicate that there are no omitted variables.



                                                                                                                 36
These models seem to indicate that the abnormally high poverty rate in the U.S. (relative to other OECD
nations) may be contributing to the overall poor broadband penetration level, given the U.S.’s high median
income level. Also, it seems that the geographic proximity factor of urbanicity is not a significant
determinate of broadband penetration at the national level. These results largely agree with other
investigations of broadband performance at the national level.


Student Access to Information Technologies

A recent study by the OECD based on data from the “Programme for International Student Assessment”
explored the relationship between student (15-year olds) academic performance and access to computers,
both at home and at school. 93 The results of this study highlight the importance of computers and
information technology in adequately preparing students for the future. Highlights of the study include:

        •    Students who have only limited access to computers performed below the OECD average, on
             measures of academic performance.
        •    Students without access to computers at home are, on average, one proficiency level below the
             OECD average, even after accounting for students’ socio-economic background.
        •    Students with the shortest experience of using computers scored poorly on average. Those with less
             than a year’s experience can typically perform only the simplest mathematics tasks.


A striking finding is the level of U.S. students reporting never having used a computer. The U.S. has the
fourth highest level of students who have never used a computer, trailed only by Turkey, Slovakia, and
Mexico.

To investigate the relationship between poverty and student home computer access, I constructed a model
similar to that used for broadband penetration.

The model is specified as:

                             homeacessi = β0 + β1(mhhinc) + β2(urban) + β3(yreduc) + β4 poverty + εi


Where homeaccess equals percentage of students reporting access to a home computer.

Results are presented below.

                           Figure A3 – Results of Regression Model – Student Home Computer Access

                                               N=22, R2 = 0.79 Coefficient    t       p
                                               med HH inc.            0.61     4.64   0.000
                                               poverty               -1.75    -3.66   0.002
                                               urban                  0.20     1.37   0.189
                                               avg. yrs. education    0.60     0.37   0.716
                                               constant              58.86     2.89   0.010

This result, like the one discussed above, seem to indicate that the U.S.’s high poverty rate is a significant
factor for explaining the high percentage of students reporting no home computer access. Taken together,
these results suggest that policies aim to close the “digital divide”, such as USF for broadband, may help
improve the U.S.’s standing relative to the other countries of the OECD.



93
     “Are students ready for a technology-rich world?”, OECD, January 2006.

                                                                                                                 37
Comparative Statistical Analysis of Broadband in the U.S. at the State-Level
Cross-sectional econometric analysis methods were employed in order to better understand the differences
between the broadband performances of the 50 U.S. states and the District of Columbia. The results
presented below are preliminary observations intended to serve as the basis for further study. All data
presented in this appendix come from the FCC (April 2006 Form 477 report, for data as of June 2005), and
the U.S. Census Bureau. This appendix begins with an examination of the individual correlations between
certain factors and broadband penetration. I then construct a single model based on a full set of predictors.
As was the case in the OECD comparison, interpretation of data gathered at such a large aggregate level is
certain to raise concerns about the “ecological fallacy”. Thus, the conclusions here are merely suggestive,
though they do conform to the results of past studies that investigated broadband at lower aggregate levels
(Flamm et.al. using census-block data 94) and at the individual level (GAO and Pew using survey data).

What factors are correlated with broadband penetration?

To begin, I examine the correlation between residential household broadband penetration (the number of
residential lines per 100 households) and the geographic factors of household density and “urbanicity” (the
percentage of the population living in urban areas). A state’s household density and urbanicity are
significantly correlated with residential broadband penetration (r=0.55, p<0.001 for density; r=.71, p<0.001
for urbanicity; note that for the test of correlation between household density and penetration, DC was
excluded from the data set, as its density is extremely high relative to the other states).

Median household income is significantly correlated with broadband penetration. In this bivariate linear
model, median household income explains 55% of the observed variation in broadband penetration, and
the result is highly statistically significant (r=0.74, p< 0.001).

Residential household penetration is also significantly correlated with poverty. Furthermore, penetration is
significantly correlated with the proportion of a state’s population living below 1.5, 2, 3, 4, and 5 times the
poverty level (see figure A4).

                                                             Figure A4




The percentage of a state’s population possessing a bachelor’s degree is also correlated with broadband
penetration. Approximately 41% of the observed variance in broadband penetration is explained by
differences in (years of formal) education in this bivariate comparison. This relationship is highly
statistically significant (r=0.64, p <0.001). The high level of correlation between certain variables, especially


94
  Kenneth Flam & Anindya Chaudhuri,. “An Analysis of the Determinants of Broadband Access”, Presented at the 33rd Research Conference
on Communication, Information, and Internet Policy (TPRC), September 23-25, 2005. Also, Kenneth Flamm. “The Role of Economics,
Demographics, and State Policy in Broadband Availability”, Presented at the PURC/London Business School Conference on “The Future of
Broadband: Wired and Wireless, 2005”, February 24, 2005.



                                                                                                                                    38
median household income, suggests that in a full regression model that there may be potential
multicolinearity issues.

Modeling Broadband Penetration

To better understand the variation in broadband penetration between U.S. states, I construct three
econometric models based on the predictive factors discussed above.

The first model is specified as:

         penetrationi = β0 + β1 (mhhinc) + β2 (poverty)+ β3 (urban) + β4 (educ) + β5-6 (age variables) + εi

The second model is specified as:

              penetrationi = β0 + β1 (ratio5x1xpov) + β 2 (urban) + β3 (educ) + β4-5 (age variables) + εi


Where:

penetration = total residential household broadband penetration as of June 2005
mhhinc = median household income
poverty = the share of households with incomes below state poverty threshold
urban = percentage of the population living in urban areas
educ = percentage of the population with a bachelor’s degree
age variables = median household age, and percent of households in state where median age is above 65
ratio5x1xpov = the ratio of the percentage of households with incomes above 5 times the poverty threshold,
divided by the percentage of households with incomes below the poverty threshold
.

Results are presented below.

            Figure A5 – Results of Regression Modes – Residential Household Broadband Penetration

                OLS: Residential BB Pen. (US States)                    OLS: Residential BB Pen. (US States)
                                     Coefficient Variance                                    Coefficient Variance
                                                   Inflation                                               Inflation
                                     (t-statistic)   Factor                                  (t-statistic)   Factor
                Median HH Income             0.176*
                                                             9.33       Ratio of pop over
                (thousands US$)              (1.78)                     5X pov/under 1X               0.597**   2.2
                                        0.069                           pov                            (2.05)
                Percent in Poverty                           3..56
                                       (0.52)
                Percent Urban         0.102***                          Percent Urban         0.114***
                                                             2.47                                               1.94
                Population             (4.70)                           Population              (5.92)
                Percent w/              0.079                           Percent w/            0.126**
                                                             3.29                                               2.21
                Bachelors Degree       (1.14)                           Bachelors Degree        (2.21)
                                     0.750***                                                0.688***
                Median Age                                   2.54       Median Age                              2.93
                                       (4.27)                                                   (3.65)
                Percent of             -0.150                           Percent of             -0.204*
                                                             2.88                                               2.5
                Population over 65     (-1.19)                          Population over 65     (-1.74)
                                       -29.09                                                   -20.45
                constant                                                constant
                                       (-4.33)                                                  (-3.71)
                                   R2 = 0.783                                              R2 = 0.778
                                   N = 50                                                  N = 50
                * = significant at 10% level; ** = significant at 5% level; *** = significant at 1%




                                                                                                                       39
Results from the first model show that urbanicity, median age, and to a lesser extent, median income are
significant predictors of household broadband penetration. However, the variance inflation factor for
median household income is above nine, indicating that the coefficient estimates may be unreliable due to
multicolinearity.

In the second model, median household income and poverty are replaced with a variable corresponding to
the ratio of the portion of population with household income above five times the poverty threshold to the
portion of population with household income below the poverty threshold. This variable serves as a proxy
for both median income and poverty, and captures the distribution of income and wealth. In the second
model, this ratio, urbanicity, education, and age are all significant explanatory factors of residential
broadband penetration at the state level.

Taken together, these models seem to indicate that rural states face significant challenges in increasing
broadband deployment and adoption. These results along with others (such as the 2006 GAO report) seem
to indicate that economies of density are an important consideration for broadband deployment within the
U.S. This suggests that policies aimed at lowering barriers to entry in rural areas (such as universal service)
may help to increase the supply of broadband. Likewise, policies aimed at lowering the barriers to purchase
(such as low-income universal service, or other policies that increase competition) may increase broadband
adoption in states with lower average income and higher poverty rates.




                                                                                                             40