Академический Документы
Профессиональный Документы
Культура Документы
By Jed Kolko
SUMMARY
California Economic
Policy is a series
analyzing and discussing
policy issues affecting
the California economy.
California Economic Policy
Broadband for All?
C
increase profits, or raise productivity? Does mak-
alifornia policymakers both locally and at
ing broadband more available to public workers
the state level are undertaking numerous
lower crime rates or improve emergency response?
initiatives to raise the level of residential
The answers to these essential questions are largely
broadband Internet adoption in the state. Through
unknown, and academic research is only beginning
regulation, subsidies, and direct provision, state
to approach them.
and local governments seek to make broadband
Nonetheless, there are several cogent argu-
more widely available and, where available, to raise
ments for the proposition that government should
adoption rates among groups less likely to have
be involved in raising broadband availability and
access. However, in trying to overcome these gaps
adoption and that—in the absence of government
in availability and adoption, policymakers lack clear
involvement—broadband use could be below a
information about who in California has access to
socially optimal level. On the supply side, broad-
broadband and who can get it. In fact, the only
band provision involves high fixed costs, especially
comprehensive measure of availability widely used
in rural areas; if providers were to spread the fixed
by policymakers is flawed and certainly overstates
cost of provision among subscribers, prices would
the level of broadband availability in the state.
be above marginal cost and too few people would
This report uses an alternative measure to
adopt. Furthermore, broadband is most efficiently
assess the extent of broadband availability, adop-
provided using publicly owned resources: Digital
tion, and the digital divide within California. It
subscriber lines (DSL) and cable lines follow exist-
seeks to answer the following questions:
ing rights-of-way, and wireless networks involve
• Does California lead or lag the country in
siting antennae on public property. On the demand
broadband adoption?1
side, there might be positive externalities in broad-
• Are there inequalities in broadband availabil-
band adoption, so that the benefits to society of
ity within California?
someone adopting broadband exceed individual
• Are there inequalities in adoption within Cali-
benefit. Also, broadband adoption encourages some
fornia and, if so, are such inequalities more
online behaviors, such as looking up medical infor-
pronounced for broadband adoption or com-
mation, leading to better health outcomes—some-
puter ownership?
thing most societies consider to be a public benefit.
Why should overcoming a broadband digital Finally, businesses may be drawn to places where
divide, or raising the level of broadband adoption, broadband is more widely available, both for bet-
be a policy goal? No one argues ter infrastructure and for a workforce that is more
that government should boost technologically literate, so governments consider
Through regulation, ownership of other technologies broadband to be an economic development tool. 2
subsidies, and direct such as DVD players and digital The first section of this report reviews the dif-
cameras. The difference is that ferent aspects of the term “digital divide.” The
provision, state and local
broadband access (and Internet second section explains the economics of broad-
governments seek to access generally) is believed to band and the technical features that could lead to
make broadband more give social or economic benefits geographic differences in availability. The third
widely available and, that are in the public interest. But section outlines current broadband policy at the
does making broadband more federal, state, and local levels. The fourth explains
where available, to raise
available to residents improve an alternative approach to measuring broadband
adoption rates among health outcomes, lower unem- availability: This involves inferring availability
groups less likely to have ployment, or improve job qual- from adoption patterns found in a particularly rich
access. ity? Does making broadband dataset, and it overcomes important shortcomings
2 P u b l ic P o l ic y I n s tit u te o f C a l if o r n ia
California Economic Policy
Broadband for All?
in other measures of broadband availability tradi- ties in how much benefit indi-
tionally used in policy analysis. The fifth section viduals get from information Why should overcoming
presents findings about broadband availability and technology. Furthermore, the a broadband digital
adoption in California. The final section draws digital divide can refer to a wide
divide, or raising the level
conclusions and suggests courses of action. range of information technolo-
gies. This report focuses on the of broadband adoption,
digital divide in broadband, and be a policy goal?
Gaps in Broadband Availability this section describes how avail-
ability, adoption, and complementary skills con-
and Adoption tribute to the broadband digital divide. This report
A
general definition of the digital divide is also considers the digital divides in Internet access
that it is “the gap separating those indi- generally, of which broadband is one aspect, and
viduals who have access to new forms of computer ownership, which is for most people a
information technology from those who do not.”4 prerequisite for adopting broadband.
The digital divide encompasses disparities in avail- The first divide—that of availability—means
ability, in adoption, and in complementary skills, that technologies are available for some people
all of which can ultimately contribute to dispari- and not others. We discuss below why broadband
P u b l ic P o l ic y I n s tit u te o f C a l if o r n ia 3
California Economic Policy
Broadband for All?
could be more widely available in urban areas than information—one might know which sites offer
in rural areas, and in richer areas than in poorer reliable and trustworthy medical advice but the
areas. other, searching for the same information, might
The second digital divide refers to levels of adop- click on the first visible sponsored link and wind
tion, which can also differ across groups. Richer up in the hands of a quack. Such gaps in comple-
people have higher rates of broadband adoption mentary skills might or might not manifest them-
than poorer people do. That the rich have more is selves in adoption levels. It could be that people
not surprising, but there are two important related who are less technology literate have lower demand
research questions about broadband adoption that for broadband as a consequence. Alternatively, it
remain to be answered. First, do race and ethnicity could be that people lacking complementary skills
influence technology adoption, are no less likely to adopt broadband but benefit
after differences in income and less from their broadband access than those with
People have different other factors are controlled for? better skills.
If so, perhaps that reflects racial
levels of knowledge and
inequalities in technology liter-
comfort with technolo- acy that policy could help over-
gies, and so even giving come. Moreover, research sug- Broadband Economics
T
away broadband and gests that disadvantaged groups he Internet’s infrastructure consists of
computers would not can benefit disproportionately transmission routes, which include fiber-
from Internet access, so targeting optic cable, coaxial cable, copper wiring,
make the benefits of
broadband policy to raise adop- and wireless links; the infrastructure also includes
information technology tion rates among these groups connection points, where data are handed off from
accessible to all. could be especially desirable. 5 one route to another. An analogy with roads is use-
Second, how much does income ful: The Internet backbone is a network of high-
matter for broadband adoption if computer owner- capacity fiber-optic cables (like interstate high-
ship is held constant? If income affects computer ways), which connect to lower-capacity routes (like
ownership more than it affects broadband adop- smaller highways), which in turn connect to last-
tion, then making broadband less expensive and mile networks (like local roads) that lead directly
more widely available would have little effect on to residences.
its adoption unless there were also efforts to raise The Federal Communications Commission
computer ownership levels among lower-income (FCC) defines broadband as an Internet service
people. that offers speeds of at least 200 kilobits per second
The third digital divide concerns gaps in skills (kbps) in at least one direction. Most residential
complementary to information technologies; this broadband services today offer speeds significantly
is often referred to as technology literacy or flu- faster than this, typically in the range of 1.5 mega-
ency. People have different levels of knowledge bits per second (Mbps: A megabit equals 1,000
and comfort with technologies, and so even giving kbps) to 6.0 Mbps downstream (i.e., data flow-
away broadband and computers would not make ing from the Internet to an end-user, like a music
the benefits of information technology accessible download). Upstream (i.e., data flowing from an
to all if the recipients lack knowledge and familiar- end-user to the Internet, such as a sent email or a
ity with it. Complementary skills are not limited to search request) speeds are typically slower, in the
technical knowledge and comfort levels with hard- range of 384 kbps to 1.5 Mbps. By comparison,
ware. Two people equally familiar with technology top downstream speeds over a dial-up modem are
might not reap the same benefits from an Internet 56 kbps—only 1/100th as fast as the top of the
connection if they have different abilities to filter range for broadband.6
4 P u b l ic P o l ic y I n s tit u te o f C a l if o r n ia
California Economic Policy
Broadband for All?
The two primary residential broadband tech- miles from a central office.10 AT&T and Verizon
nologies are DSL and cable.7 Both are last-mile are the main providers of DSL service in the United
technologies, which means that they connect resi- States, with AT&T dominant in California.
dences to the larger Internet network.8 Both DSL Cable infrastructure consists of a head-end,
and cable rely on existing infrastructure to provide which forwards local Internet traffic to the wider
Internet services: DSL shares the copper wiring used Internet and serves thousands of homes, and opti-
for telephone service, and cable shares a hybrid of cal nodes, which are connected to the head-end
fiber and coaxial cable (HFC) used for cable televi- with fiber-optic cable and to residences with coax-
sion service. To offer broadband, DSL and cable ial cable. The coaxial cable is the same infrastruc-
providers must upgrade this existing infrastruc- ture that delivers cable television service. To pro-
ture. Upgrading involves high initial fixed costs, vide broadband, cable television providers upgrade
and the technologies themselves have limitations their networks by adding nodes and moving them
on their deployment (described below). Because closer to residences, which in effect replaces some
of high fixed costs and technological limitations, of the coaxial cable in the network with fiber.
some cities or neighborhoods are more profitable Cable companies also have to install equipment
for broadband providers to serve than others. (analogous to the telephone DSLAMs) that route
Upgrading infrastructures—and the effects and switch digital data, and they install amplifi-
on geographic availability—are different for DSL ers that improve the upstream data transmission.11
and cable.9 DSL’s use of existing telephone cop- Unlike DSL, cable infrastructure is shared: The
per wiring means that it is a dedicated service—it coaxial cable connects residences to the optical
runs directly from the residence to the service pro- node in a loop, so that the bandwidth any resi-
vider and is not shared with any other residences. dence receives depends in part on the number of
Because DSL technology can transmit data even residences sharing the node. Thus, the local fixed
when the line is also in use for a telephone call, costs of upgrading an area’s infrastructure to make
DSL provides an always-on Internet connection. cable broadband available can involve (1) moving
To offer DSL, telephone companies must install optical nodes closer to residences, (2) building new
DSLAM (digital subscriber line access multiplexer) optical nodes, (3) upgrading the upstream path,
equipment that aggregates Internet data from the and (4) installing equipment at the head-end.12
service area and forwards it to the larger high- Comcast and Time Warner are the main providers
ways of the Internet. DSL works only within three of cable broadband service in California.
miles of a telephone company central office. Where These fixed costs mean that cable and DSL
providers offer multiple tiers of DSL service (such are more profitable in areas where the costs can
as 1.5, 3.0, and 6.0 Mbps), the faster tiers might be spread over more subscrib-
be available only to residences closer to the cen- ers. Areas that are higher den-
tral office. The prevalence of telephone company sity or higher income or both Because of high fixed
central offices depends on population density. tend to be more profitable. costs and technological
Large cities have multiple central offices and less- In a higher-income neighbor- limitations, some cities
dense areas have far fewer per square mile; this hood, more residents are likely
alone makes DSL more widely available in higher- to adopt broadband, so provid-
or neighborhoods
density areas. Alternatively, DSL architecture can ers seeking to make broadband are more profitable for
consist of a copper connection from a residence available in the most profitable broadband providers
to an intermediate node, called a street cabinet, areas first would choose higher- to serve than others.
which is connected with fiber to the central office. income areas; this effect is mag-
By installing street cabinets, telephone providers nified for cable providers, since upgrading their
can offer DSL service to areas farther than three networks offers new potential revenue from both
P u b l ic P o l ic y I n s tit u te o f C a l if o r n ia 5
California Economic Policy
Broadband for All?
F
for DSL service, they also can be served at lower ederal, state, and local governments all play
installation costs per subscriber.13 a role in shaping the availability and adop-
For all of these reasons, broadband availabil- tion of broadband. The federal government,
ity should differ geographically according to aver- through Congress, the executive, and the Supreme
age income and density. There might also be geo- Court, makes the most important regulatory deci-
graphic differences in availability that are unique sions, whereas state and local governments play a
to a particular provider: Most areas in the United larger role in subsidizing and in some cases directly
States are served by a dominant telephone pro- providing broadband services.
vider and a dominant cable provider, and each can At the federal level, the FCC regulates tele-
make different strategic decisions about when to communications. The 1996 Telecommunications
introduce broadband service to their regions. The Act, the 2005 Supreme Court decision in National
age and physical condition of existing telephone Cable & Telecommunications Association et al. v.
or cable infrastructure can also affect the cost of Brand X Internet Services et al., and related FCC
introducing broadband. rulings have created the regulatory framework that
In addition to these two wireline technologies, exists today. Under this framework, telephone,
two wireless broadband technologies are begin- wireless, television, and Internet providers are
ning to be used.14 One is satellite, which although able to “compete in any market against any other”
available nearly everywhere in (FCC web site), but broadband providers are not
the United States, offers a slower required to give competitors wholesale access to
Federal, state, and speed and lower reliability for their infrastructures so that competitors can resell
local governments all a higher monthly price than services to consumers.17 The FCC also shapes
either cable or DSL.15 In prac- broadband policy by placing conditions on merg-
play a role in shaping ers between broadband providers and by allocating
tice, satellite broadband appeals
the availability and to consumers only where DSL wireless spectrum.18 Finally, the FCC administers
adoption of broadband. and cable are unavailable and so the universal service requirement, which guaran-
does not actually compete with tees that even the most remote areas have telephone
them. The other wireless broadband technology
16
service.19
is Wi-Fi (the technical term is 802.11x), which States and local governments are still left with
offers high speed within a very short distance of a important elements of broadband regulation, one
base station. Wi-Fi is commonly used in conjunc- being control of the physical development of infra-
tion with DSL or cable to make broadband access structure. Broadband networks typically follow
available wirelessly within a home, office, café, or public rights-of-way such as roads or rail tracks,
public space. The cost of setting up base stations and deploying broadband infrastructure costs
and antennas to provide Wi-Fi coverage is much providers less when it occurs in tandem with pub-
lower than upgrading or building wireline infra- lic works projects or when done simultaneously
structures such as DSL, cable, or fiber-to-the-home by multiple broadband providers. In California,
(FTTH). In a handful of locations, Wi-Fi offers improving rights-of-way access for broadband
citywide public access to the Internet and actually deployment is the most prominent element of
competes with cable and DSL. This new wave of Governor Arnold Schwarzenegger’s 2006 Execu-
municipal Wi-Fi initiatives is discussed below as an tive Order on “Expanding Broadband Access and
example of broadband policy. Usage in California,” which created a California
6 P u b l ic P o l ic y I n s tit u te o f C a l if o r n ia
California Economic Policy
Broadband for All?
Text Box 2. Will New Internet Access Technologies Overcome the Digital Divide?
The economics of DSL and cable result in greater geographic is a strong possibility of a future, persistent digital divide
differences in availability than those of dial-up access do. in FTTH availability; at least as important is that only some
Will the next generation of technologies have the same telecom companies are considering FTTH. In fact, a study
effect or will they widen or narrow the digital divide? prepared for San Francisco’s evaluation of the feasibility
The most promising next-wave broadband technol- of a municipally built, owned, and operated fiber-optic
ogies are FTTH and WiMax. 20 FTTH extends fiber-optic network argues that the city is already on the losing side
cable—which already connects the Internet all the way to of the FTTH digital divide because Verizon, the company
the telephone company central offices and the cable pro- deploying most large FTTH projects in the country, is not
viders’ optical nodes—the rest of the way to homes. Fiber the dominant telephone provider in San Francisco.25
would, in effect, replace the copper wiring and coaxial The other promising next-wave technology is WiMax,
cable in use today. Because the capacity of fiber is far a wireless technology. WiMax offers Internet connectiv-
greater than that of either copper or coaxial cable, users ity over a range of one to 30 miles from a transmission
would access much higher speeds downstream and tower; this compares to the hundreds of feet that are Wi-
upstream—potentially into the gigabit-per-second (gbps) Fi’s limit. The downstream and upstream bandwidth of
range, hundreds of times faster than today’s fastest cable WiMax depends on the number of simultaneous users,
or DSL services. 21 but speeds could rival those of DSL and cable. WiMax is
Recent policy reforms at the state and federal levels not yet being used for broad-based Internet access. Theo-
have reduced some of the regulatory challenges to deliv- retically, the fixed costs of WiMax deployment should be
ering cable television and these give telephone companies much lower per subscriber than costs for wireline tech-
a stronger incentive to build FTTH networks.22 Telephone nologies, because the infrastructure consists of widely
companies are interested in FTTH as much for its abil- spaced antennas, not extensive wiring. The wide range
ity to deliver television service as to deliver high-speed of WiMax signals could also bring high-speed service to
Internet service; this is so that they can compete fully harder-to-reach rural areas.
with cable providers, who can offer television, Internet, Next-wave technologies are no guarantee of over-
and telephony (using voice-over-IP [Internet protocol]) coming today’s digital divide. In fact, the high fixed costs
over their existing networks.23 However, current adoption of FTTH make it likely that some areas will receive ser-
in California is minimal. SureWest, a regional telecom vice long after others do. Furthermore, with the devel-
service provider, has over 20,000 FTTH subscribers in opment of faster access technologies, expectations
the Sacramento area. 24 AT&T, the dominant local phone about adequate service ratchet upward. This is not only
provider, has not announced plans for any major FTTH because the digital divide refers to relative differences,
deployment. not absolute levels. It is also because online applications
Fiber costs more than DSL or cable, and it requires are designed for users’ current bandwidth; as typical resi-
replacement of the existing connections to customers’ dential bandwidth increases, online applications incorpo-
homes, so its roll-out proceeds neighborhood by neighbor- rate more bandwidth-hungry content (such as video and
hood. Multiunit dwellings, dense areas, and new develop- interactivity), and access technologies that were once
ments cost less to wire with fiber than other areas, so there adequate cease to be so.
P u b l ic P o l ic y I n s tit u te o f C a l if o r n ia 7
California Economic Policy
Broadband for All?
Broadband Task Force composed of public and Silicon Valley. Major initiatives are under negotia-
private stakeholders to coordinate efforts to raise tion for San Francisco, for the greater Sacramento
broadband adoption and identify ways to fund new region, and for a Silicon Valley–wide network; Los
technology investments. 26 Angeles has also announced a citywide initiative. 31
States and localities also play a large role Most of these call for low-cost or free wireless
through subsidizing and directly providing broad- access, provided by partners such as Google, Earth-
band. California’s primary broadband subsidy Link, and MetroFi, supported by subscriptions or
program is the Teleconnect Fund, which pays half advertising, with little financial investment by the
the cost of Internet access for qualified schools, public sector. Of the 58 initiatives, all but four are
libraries, community organizations, and other in densely populated areas of the Bay Area, South-
nonprofits. Funded from a statewide fee on tele- ern California, or greater Sacramento. Text box 3
phone service, the fund’s 2006–2007 fiscal year provides more detail on the justifications and chal-
budget is $22 million. 27 In addition, the Califor- lenges of municipal wireless.
nia Public Utilities Commission (CPUC) recently California’s broadband policy can be summed
created the California Emerging Technology Fund, up as follows: First, subsidies focus more on insti-
an independent nonprofit foundation to be funded tutional access to broadband than residential access
with $60 million over five years from AT&T and to broadband. Second, municipal Wi-Fi initiatives
Verizon as conditions of their respective mergers are widespread and are concentrated in the densely
with SBC and MCI. The fund’s mission is “achiev- populated parts of the state. Third, policy is geared
ing ubiquitous access to broadband and advanced toward broadband access, not computer ownership.
services in California, particularly in underserved
communities through the use of existing and emerg-
ing technologies,” although specific strategies have
not yet been selected. 28 Measuring Broadband
T
Whereas state governments have focused on rying to measure the extent of the digital
subsidies to encourage adoption, more and more divide is challenging. Publicly available
localities are attempting to provide broadband household surveys on broadband adop-
directly, both by themselves and in partnership tion, Internet access, and computer ownership are
with private companies. In the inadequate for studying recent trends in Califor-
late 1990s, a few localities across nia. For example, the federal Current Population
Whereas state govern- the country built fiber-optic net- Survey last included technology questions in 2003,
works. 29 These early projects and there are no plans to do so again. The Pew
ments have focused on
often involved public owner- Internet & American Life Project surveys house-
subsidies to encourage ship of networks and were in holds about technology adoption annually or
adoption, more and more direct response to the perceived more frequently, but with only 4,000 respondents
localities are attempting lack of service provision by the nationally, the sample is too small to draw con-
to provide broadband phone and cable companies. 30 In clusions about California. 32 A proprietary survey,
the past couple of years, many the Technographics Benchmark conducted by For-
directly, both by them- localities have turned to Wi-Fi as rester, a technology research and consulting firm,
selves and in partnership a wireless standard and a way to is used for this analysis. Forrester annually surveys
with private companies. bring broadband service at low 60,000–100,000 households about their technol-
or no cost to a wide area. In Cal- ogy adoption and behaviors. 33
ifornia, 58 localities have Wi-Fi initiatives under Measuring the divide in broadband availabil-
way. Service is operational in Anaheim, the San ity is more challenging than measuring broadband
Diego County tribal nations, and several cities in adoption. Broadband providers treat service avail-
8 P u b l ic P o l ic y I n s tit u te o f C a l if o r n ia
California Economic Policy
Broadband for All?
P u b l ic P o l ic y I n s tit u te o f C a l if o r n ia 9
California Economic Policy
Broadband for All?
ability maps as sensitive competitive information, where broadband was available (see Table 1). In the
and regulators have not required that provid- most sparsely populated tenth of zip codes, 96.2
ers make these maps public. Although individual percent of the population lived in zip codes where
consumers can check availability at an individual broadband was available, and in the poorest tenth
address on most providers’ web sites, anecdotal of zip codes, 99.4 percent of the population lived in
evidence suggests that this information is often zip codes where broadband was available.
inaccurate or incomplete.43 Nevertheless, these FCC data are the basis
The only widely used data on broadband avail- for most policy studies and academic work about
ability come from FCC surveys of broadband pro- broadband availability. In its most recent report on
viders. The FCC publishes a count of the number broadband deployment in the United States in 2004,
of providers with at least one subscriber in each the FCC concludes that broadband is nearly univer-
zip code nationally.44 This method overstates the sally available in urban areas and “significant prog-
level of availability because broadband may be ress is being made towards ubiquitous availability of
available in only some parts of a zip code. Using advanced services in rural areas.”45 The California
this measure could also under- Public Utilities Commission (2006) also relies on
state geographic differences in the FCC’s zip code data for its maps of broadband
Rather than rely on FCC broadband availability if provid- availability, which show at least one broadband pro-
ers make broadband available vider offering service in every zip code in Califor-
data, this analysis infers
only in richer or denser portions nia.46 Several academic studies rely on the FCC data
broadband availability by of a zip code. Recent FCC data as well, although one notes that “the FCC count of
examining the relation- suggest implausibly high levels high speed line providers within a zip code may seri-
ship between location of broadband availability, ren- ously misrepresent competitive options available to
dering the data misleading for the totality of residents within that zip code … [but]
and broadband adoption
assessing the digital divide. The there is no practical alternative to using the FCC
at the individual level, FCC reports that in December data in assessing broadband availability.”47
controlling for numerous 2005, 99.8 percent of the U.S. Rather than rely on FCC data, this analy-
individual characteristics. population lived in a zip code sis infers broadband availability by examining
10 P u b l ic P o l ic y I n s tit u te o f C a l if o r n ia
California Economic Policy
Broadband for All?
Percentage
California
code and a middle-income rural zip code.49 Because 25 23% 28%
O
ur first research question is whether Cali- income levels make it more prof- 68 percent of U.S. house-
fornia leads or lags the nation in broad- itable to offer broadband here
holds with Internet
band adoption. Nationally, residential than elsewhere in the country. A
broadband adoption has grown rapidly, from 7 third possibility is that another access (broadband or
percent of households at the end of 2000 to 39 per- California-specific factor—such dial-up) at home, more
cent in 2005 (see Figure 1). 53 Looking just at the as state policies or decisions by than half have broadband.
68 percent of U.S. households with Internet access AT&T, the dominant local tele-
(broadband or dial-up) at home, more than half phone provider in California—contributed to faster
have broadband. During 2005, broadband adop- deployment and therefore higher adoption. These
tion grew from 28 percent to 39 percent, so the three possibilities are not mutually exclusive.
P u b l ic P o l ic y I n s tit u te o f C a l if o r n ia 11
California Economic Policy
Broadband for All?
To assess the importance of each, regression anal- 73 percent and 83 percent. More than half the dif-
ysis can identify the separate relationships of individ- ferences in broadband adoption between regions
ual demographics and location characteristics such remain even after controlling for individual demo-
as neighborhood income and density on broadband graphics, suggesting that inequalities in broadband
adoption. The results of this analysis suggest that availability across regions account for much of the
about half of the gap between broadband adoption differences in broadband adoption. In contrast,
in California and that in the rest the differences between regions in overall Internet
of the nation is due to California’s access and computer ownership are due to differ-
The results of this favorable individual demograph- ences in individual demographic characteristics
analysis suggest that ics.54 Location characteristics that across those regions.60
about half of the affect providers’ decisions to make The third research goal is to understand the
gap between broadband broadband available—neighbor- inequalities in broadband adoption in California
hood income and population den- apart from availability. Although it is obvious (as
adoption in California
sity—account for the other half we confirm below) that broadband adoption rates
and that in the rest of of the gap.55 Because individual are higher among richer households, two impor-
the nation is due to demographics and location char- tant questions about income and broadband adop-
California’s favorable acteristics account for the entire tion whose answers are not obvious are (1) do race
difference between broadband and ethnicity affect broadband adoption, holding
individual demographics.
adoption in California and that in income constant, and (2) does income affect broad-
the rest of the country, we infer that other factors band adoption because income also affects com-
unique to California did not collectively change the puter ownership, which is essentially a prerequisite
level of broadband adoption.56 for broadband adoption?61
The second research question is whether there Broadband adoption, Internet adoption, and
are inequalities in broadband availability within computer ownership all vary by income in Cali-
California. As noted above, this report infers avail- fornia. Among households with incomes over
ability from the relationship between location $100,000, 68 percent have broadband, compared
characteristics and broadband adoption. Another to 49 percent of households with incomes between
regression analysis reveals that broadband adop- $50,000 and $75,000 and compared to 24 percent
tion in California is significantly higher in higher- of households with incomes under $25,000 (see
income and denser zip codes, even after controlling Table 3). The gaps for Internet adoption (broad-
for individual demographics. 57 That is, people of band and dial-up combined) and computer owner-
the same age, race, income, and so on are more ship across income groups are also considerable.
likely to adopt broadband if they live in a richer Looking across racial and ethnic groups, it is
or denser area. 58 The effect of location on avail- important to keep in mind that Forrester conducted
ability means that broadband adoption differs con- this survey in English only. These data therefore
siderably across regions within California, some represent people with very high English proficiency
of which are much richer and denser than others. and exclude a significant share of California’s
Broadband adoption ranges from under 30 percent Hispanic population.62 Disparities in broadband
in the north of the state and in the Sierras to just adoption across racial and ethnic groups are less
over 50 percent in the Bay Area and greater Los wide than across income groups but they are still
Angeles (see Table 2). 59 The differences in Inter- apparent, with 46 percent of both non-Hispanic
net access (broadband and dial-up combined) and whites and English-proficient Hispanics having
computer ownership across regions are smaller: broadband, compared with 63 percent of Asian
Internet access ranges from 61 percent to 76 per- Americans and only 36 percent of African Ameri-
cent and computer ownership ranges only between cans.63 Since income among non-Hispanic whites is
12 P u b l ic P o l ic y I n s tit u te o f C a l if o r n ia
California Economic Policy
Broadband for All?
higher than income among Hispanics, and income English and Spanish, broadband adoption is 50
and broadband adoption are strongly related, why percent among Hispanics who speak English as
do both groups have the same level of broadband much or more than Spanish at
adoption in California? Relative to non-Hispanic home and responded to the sur-
whites, Hispanics on average have larger house- vey in English. Broadband adop- Broadband adoption
holds and are younger, both of which are positively tion is only 20 percent among ranges from under
correlated with broadband adoption. Hispanics who speak Spanish 30 percent in the north
Other research shows that Hispanics in Cali- more than English at home or
of the state and in the
fornia who prefer speaking Spanish at home have responded to the survey in Span-
much lower broadband adoption than English- ish. Broadband adoption among
Sierras to just over
speaking Hispanics. According to the March 2007 all California Hispanics is 29 50 percent in the Bay Area
PPIC Statewide Survey, conducted by telephone in percent, so looking only at His- and greater Los Angeles.
P u b l ic P o l ic y I n s tit u te o f C a l if o r n ia 13
California Economic Policy
Broadband for All?
panics with high English proficiency overstates the likelihood of broadband adoption by 10 percent.68
level of broadband adoption among Hispanics.64 Both relationships are statistically significant: The
Holding other individual demographics con- effect of income on computer ownership means
stant, there is a statistically significant difference that there is a digital divide in computer owner-
for broadband adoption among Hispanics and ship, but the effect of income on broadband among
African Americans relative to whites. English- computer owners means that, even among com-
preferring Hispanics are 6 percent less likely puter owners, there is a digital divide in broadband
to have broadband at home than non-Hispanic adoption.
whites; the difference for African Americans is 10
percent.65 Relative to California’s overall broad-
band adoption of 47 percent, these gaps are large. Conclusions and
Moreover, for Hispanics this is the gap between
adoption among non-Hispanic whites and Hispan- Recommendations
W
ics with high English proficiency. The racial gaps in ithin California, there are digital
broadband adoption, however, divides in both broadband availability
cannot be explained by provid- and broadband adoption. Broadband
Holding other individual ers’ failing to offer service in is more widely available in higher-density residen-
minority neighborhoods because tial areas and urban regions than in lower-density
demographics constant . . .
there was no statistically signifi- areas and rural regions; the level of neighborhood
English-preferring cant effect of the racial composi- income also affects availability. These inequali-
Hispanics are 6 percent tion of a zip code on broadband ties arise, in part, from the technical features and
less likely to have adoption. economic realities of broadband provision. The
broadband at home than The other important fac- digital divide in adoption encompasses not only
tor about the digital divide in broadband but computer ownership as well, and
non-Hispanic whites;
broadband adoption is the role income and race and ethnicity both affect broad-
the difference for African of computer ownership. A home band adoption in California.
Americans is 10 percent. computer is for the most part Our findings have important implications for
a prerequisite for broadband broadband policy. If closing gaps in broadband
adoption.66 If the relationship between income and availability is a policy goal, raising availability in
broadband adoption reflects lower computer own- rural areas should be the top priority. The Califor-
ership by poorer households, then making broad- nia Emerging Technology Fund should focus first
band more widely available at low cost might have on deployment in rural areas, and the Broadband
little effect on broadband adoption. Looking across Task Force could help identify clearly the barriers
income groups, computer ownership ranges from to providers’ offering service in rural areas. Eas-
58 percent among households with income under ing the access to rights-of-way might facilitate
$25,000 to 93 percent among households with rural broadband deployment, but the economics of
income $100,000 and above. Regression analysis broadband make rural areas costly to serve even
helps assess how much this gap in computer own- in the absence of any regulatory or permitting fac-
ership contributes to the digital divide in broad- tors, so the state could consider direct subsidies to
band adoption. Controlling for other demograph- providers serving rural areas. Although broadband
ics, raising log income by one standard deviation availability lags in lower-income areas as well, den-
increases the likelihood of computer ownership by sity has a greater effect on availability; furthermore,
7 percent.67 Then, looking only at computer own- there is no evidence that broadband availability is
ers and again controlling for other demographics, lower in neighborhoods with higher concentrations
raising income by the same amount increases the of Hispanics and African Americans.
14 P u b l ic P o l ic y I n s tit u te o f C a l if o r n ia
California Economic Policy
Broadband for All?
To close gaps in broadband adoption, the focus New technologies, such as FTTH, are also likely
should be broadened to include racial and ethnic to benefit urban areas more than rural areas, so
and income groups with lower adoption rates, the urban-rural gap is likely to persist even if policy
including those living in neighborhoods where efforts successfully raise the level of rural broad-
broadband service is already available. Improving band availability.
technology literacy may also be essential. Munici- Two broader questions remain to be researched
pal Wi-Fi initiatives will help meet these goals so and answered. First, is it more desirable from
long as they provide broadband at lower cost than a cost-benefit perspective to raise the degree of
DSL and cable providers do. Recent initiatives sug- broadband adoption for those
gest that they will: Google’s proposal to provide with less access than it is to
free, ad-supported service in San Francisco, and overcome other inequalities that If closing gaps in
the prevailing rate of about $20 per month that Californians face? On one hand, broadband availability
EarthLink charges for its Wi-Fi service, are two the economic and social benefits
is a policy goal, raising
examples. Municipalities need not resort to pub- of broadband are unknown and
lic ownership or operation of Wi-Fi networks to are largely taken as an article availability in rural areas
achieve these public benefits. Public involvement of faith; on the other, the costs should be the top priority.
should be reserved for the investments that the pri- to the public sector of raising To close gaps in broad-
vate sector is less likely to make, such as helping broadband adoption can be low band adoption, the focus
raise technology literacy and improving computer if the private sector bears the
should be broadened to
access among disadvantaged residents, both of cost, as is the case with most
which are among the “digital inclusion” goals of municipal Wi-Fi initiatives. Sec- include racial and ethnic
San Francisco’s Wi-Fi initiative.69 ond, will the development of and income groups with
Ironically, because the Wi-Fi networks that future Internet access technolo- lower adoption rates.
are operational or under consideration are over- gies such as WiMax or FTTH
whelmingly in the urban parts of the state, they are mitigate or aggravate today’s inequalities in tech-
likely to widen rather than close the gap in avail- nology access? Both of these questions should be
ability between urban and rural areas. Policymak- considered—even if they cannot be answered—
ers, therefore, should think less in terms of closing as California’s state and local governments weigh
the gap between urban and rural availability, lest various strategies for raising broadband adoption
they hold back urban initiatives, and more in terms and closing the digital divide. v
of raising the absolute level of rural availability.
P u b l ic P o l ic y I n s tit u te o f C a l if o r n ia 15
California Economic Policy
Broadband for All?
Notes 10
The website www.dslreports.com/prequal/distance offers
estimates of distance between a residential address and the
1
The California Public Utilities Commission (2005) cites nearest central office and maps of central office locations.
California’s #14 ranking among states (according to the
2003 TechNet survey) as an important indicator of wheth- 11
Traditional cable television is a one-way communica-
er the state is “maintaining its lead in broadband usage.” tion: Video is broadcast from the cable head-end to the
In the California Public Utilities Code, the main broad- residence. Internet service, as well as digital cable service,
band policy objectives are assuring “continued affordabili- is two-way communication, which requires a different
ty and widespread availability” of broadband and promot- “upstream” infrastructure.
ing “economic growth, job creation, and substantial social
benefits.” It is hard to see how the “substantial social ben- 12
Cable providers might have to make additional infra-
efits” depend on California’s broadband adoption relative structure investments as broadband adoption rises, con-
to that of other states, rather than the absolute level, but it gestion increases, and bandwidth speeds fall.
is plausible that the relative ranking could affect economic
growth and job creation if businesses consider broadband 13
High residential density is one reason why Korea, Hong
availability and adoption a factor in deciding where to Kong, and Japan have higher broadband adoption and higher
open, expand, or move their operations. speeds of service available than the United States does.
2
Only limited research has been completed on the effects In addition to satellite and Wi-Fi, there is also WiMax,
14
4
Gunkel (2003). 16
See “With a Dish, Broadband Goes Rural” (2006).
5
See Morton, Zettelmeyer, and Silva-Risso (2003). 17
The laws and rulings that make up the federal regula-
tion of broadband have involved the sometimes competing
6
Dial-up Internet access, in contrast to cable and DSL, goals of (1) maximizing competition among broadband
is nearly ubiquitous in the United States and never had a providers, (2) not reducing the incentives that broadband
long period of geographic disparities in availability. Mak- providers have to invest in infrastructure and new applica-
ing dial-up available does not involve a high fixed cost: tions, and (3) treating DSL and cable broadband services
Dial-up Internet service providers (ISPs) need to establish evenly despite the fact that telephone companies as “com-
a point-of-presence (POP) in an area by making a local mon carriers” have traditionally been subject to much
phone number available as a dial-in number, and although stricter federal regulation than cable companies have.
this involves adding some switching equipment in the tele-
phone company central office, the infrastructure invest- 18
In 2005, the FCC required that AT&T offer “naked
ment is minimal. Downes and Greenstein (2002) docu- DSL,” under which consumers can subscribe to AT&T’s
ment that dial-up Internet service spread quickly to even DSL service without subscribing to AT&T’s local telephone
the most rural counties: In 1997, 99 percent of the U.S. service, as a condition of its merger with SBC. Then, in late
population lived in counties with at least one ISP, and 92 2006, the FCC required specific pricing for naked DSL as
percent lived in counties with seven or more ISPs. There a condition of its merger with BellSouth. See “AT&T to
are scale economies in having a POP, but the costs to Offer ‘Naked DSL’ for Far Less Than Before” (2007).
upgrade infrastructure for broadband service are much
larger (Greenstein and Prince, 2006). 19
A short explanation of universal service is available
at the FCC’s website, www.fcc.gov/cgb/consumerfacts/
7
DSL refers to a family of technologies, technically de- universalservice.html. Universal service applies only to
scribed as xDSL. Residential DSL is most often ADSL telephone, not Internet, service. However, because DSL
(asymmetric DSL), and higher-capacity variations include uses the telephone network, the cost of providing DSL in
SDSL and VDSL. remote areas is lower than if the telephone network did
not cover those areas. In contrast, cable companies face no
8
Many smaller businesses subscribe to DSL as well. Larger universal service requirement. As a result, DSL adoption is
businesses rely on higher-bandwidth connections such as higher than cable broadband adoption in very low-density
T1 and T3 lines. Cable broadband is almost exclusively a rural areas.
residential service.
20
Fiber-to-the-home is one technology in a class of FTTx
9
The description of Internet infrastructure and the process technologies: They differ in whether the fiber network
for upgrading networks relies on Federal Communications extends all the way to the individual home, or to the curb
Commission (2000) and Corning (2001, 2005). (FTTC), or to a multiunit building (FTTB). Fiber networks
include those that have dedicated fiber running between the
16 P u b l ic P o l ic y I n s tit u te o f C a l if o r n ia
California Economic Policy
Broadband for All?
P u b l ic P o l ic y I n s tit u te o f C a l if o r n ia 17
California Economic Policy
Broadband for All?
42
See “What If They Built an Urban Wireless Network location characteristics strongly influence the choice of
and Hardly Anyone Used It?” (2006). The article cites the cable versus DSL, which is much more plausibly due to
competition from Wi-Fi connections, mobile phone data availability differences than to location-driven differences
services, and fixed-wire broadband as a major reason for in underlying demand for one broadband technology ver-
the lack of interest in paying for Taipei’s municipal Wi-Fi sus the other. Still, it is important to keep in mind that this
service. approach measures availability as the relationship between
location characteristics and adoption, which might also be
43
A personal anecdote: Over a two-day period in August capturing some effects in addition to availability. Further-
2006 of checking the AT&T website and calling several more, this measure does not capture reasons other than zip
customer service departments to sign up for new DSL ser- code density and zip code income for variation in availabil-
vice, the author was told by different people that (1) DSL ity, such as zip-code-level topographical differences that
was unavailable at the address, (2) only a slower DSL ser- might make some zip codes more expensive to serve.
vice (up to 1.5 Mbps) was available at the address, (3) all
speeds of DSL service were available at the address, and (4) 51
This approach implicitly captures geographic differences
the address was not a valid address. in broadband prices. If there is more competition among
broadband providers in higher-density areas, this could
44
These counts include both providers who own the broad- result in lower prices and therefore higher adoption. The
band pipes (such as the dominant telephone and cable notion of “availability” used here is not just a binary mea-
companies) and resellers. Although residential custom- sure but rather a continuous measure that captures dimen-
ers account for most of the dominant telephone and cable sions of availability such as price and quality of service.
companies’ subscribers, some resellers serve primarily
business customers. 52
DiMaggio et al. (2004) review the academic literature on
the digital divide in complementary skills and provide more
45
FCC (2004), p. 38. The FCC uses “advanced services” detail on the challenges researchers face in studying it.
as a synonym for broadband. Two of the five FCC com-
missioners dissented from the upbeat conclusions of the 53
Forrester’s measure of broadband refers to “high-speed
report, and both noted the inadequacy of using the one- Internet access” and mentions DSL and cable modem as
subscriber-in-a-zip-code measure of availability. Still, the examples. In other questions in Forrester’s surveys that
main concern of both dissenters was the lower level of ask about specific technologies, broadband includes DSL,
broadband adoption in the United States relative to that cable, fixed wireless, and satellite.
in Korea, Japan, and other countries, not the inequality of
broadband availability within the United States. Further- 54
That is, the coefficient on the California dummy vari-
more, not only do global broadband leaders have a higher able in a probit regression where broadband adoption is
percentage of households with broadband, average broad- the dependent variable falls by about half when individual
band speeds are also much higher. demographics are included.
46
The CPUC emphasizes the inequality in the number of 55
That is, the coefficient on the California dummy variable
providers in zip codes, rather than using the data as evidence in a probit regression where broadband adoption is the
of ubiquitous access to broadband, and notes that “this rep- dependent variable is not significantly different from zero
resentation does not depict the availability of broadband for when individual demographics and zip code characteris-
every resident in each identified zip code area.” tics (log median income and log density) are included.
47
Flamm (2006). Grubesic (2006), and Prieger (2003) also 56
Because there could be many factors unique to Califor-
use FCC broadband data. nia that affect broadband adoption, one cannot conclude
that they are irrelevant individually. For instance, if, hypo-
48
This section of the report draws heavily on Kolko (2007). thetically, AT&T had been unusually aggressive in rolling
Details on the methodology for inferring broadband avail- out broadband in California, but state policy made deploy-
ability can be found there. ment more difficult than in other states, these two effects
could cancel each other out. Thus, we emphasize that fac-
49
Unlike the CPS or Pew, Forrester includes respondents’ tors unique to California did not collectively have an effect
zip codes. on broadband adoption.
50
This approach raises two methodological concerns. First, 57
The web-only appendix www.ppic.org/content/other/
location characteristics could be capturing unobserved indi- 707JKEP_web_appendix.pdf presents regression results
vidual characteristics. Second, location itself could affect for California and also for the United States. Higher-income
the demand for broadband if, for instance, people with less and higher-density zip codes have higher broadband adop-
access to doctors or stores have greater demand for online tion, controlling for individual characteristics, both for
health advice or online shopping. Kolko (2007) finds that California and the United States. The effect of income is
location characteristics have a much smaller or insignifi- smaller for California than for the United States, although
cant effect on technologies other than broadband and that statistically significant in both cases. The effect of density
is similar for California and the United States.
18 P u b l ic P o l ic y I n s tit u te o f C a l if o r n ia
California Economic Policy
Broadband for All?
58
These significant and positive relationships do not hold level of broadband adoption in California rises from 55
when Internet access (broadband or dial-up) is the depen- percent among all Californians to 65 percent when these
dent variable, or when computer ownership is the depen- Hispanics are excluded, suggesting that the Forrester figure
dent variable. This is further evidence that the relation- for English-speaking Californians would fall by up to 10
ship between zip code characteristics and adoption is due percentage points if non-English-speakers were also sur-
to availability, since the availability of dial-up access and veyed. These figures are higher than the Forrester figures
computer ownership do not vary geographically. Relative for overall adoption in part because the PPIC survey was
to the overall broadband adoption level of 47 percent, liv- conducted 15 months after the Forrester survey. Also, the
ing in a zip code with a log density one standard deviation PPIC data, unlike Forrester data, show a gap in adoption
(1.72) above the mean raises the likelihood of broadband between Hispanics preferring to speak English and whites,
adoption by six percentage points—an amount not only in part because a higher level of English proficiency would
statistically significant but large in magnitude as well. be required to complete Forrester’s mail survey in English
than to participate in PPIC’s telephone survey in English.
59
Table 2 shows the nine regions defined by the Economic An earlier survey, the October 2003 Current Population
Strategy Panel, but we have split the San Joaquin Valley into Survey, reports that broadband adoption is 18 percent
northern and southern regions, and we have split the South- for California Hispanics who are not Spanish-only versus
ern California region into Greater Los Angeles and Inland only 2 percent for Spanish-only Hispanics in California.
Empire. The distribution of Forrester’s respondents across Data from PPIC and the Current Population Survey are
these regions is similar to the population distribution. weighted using standard demographics to represent the
population more accurately, but the figures reported here
60
That is, an F-test does not reject the hypothesis that the do not further adjust for demographics using a regression
regional dummies all equal zero in a regression of Inter- framework.
net adoption (and, separately, computer ownership) on
individual characteristics and regional dummies. This test 65
The regression underlying this finding also controls for
does reject the hypothesis of insignificant regional dum- zip code income and density.
mies when broadband adoption is the dependent variable.
66
Why might a household have broadband without a com-
61
Although this section focuses on race, ethnicity, and in- puter? Some videogame consoles, digital video record-
come, it is also a goal of California policy to consider the ers, and smartphones can connect directly to the Internet
digital divide between people with disabilities and people using cable or DSL. Households could also have broad-
without. Forrester’s surveys do not ask about disability band access for use with an employer-provided computer.
status. In California, 80 percent of households have a computer.
Among computer-owning households, 58 percent have
62
The Forrester Technographics survey is long, detailed, broadband at home. Among households with broadband,
and conducted by mail, so respondents need considerable 98 percent own a computer.
English proficiency to complete it. Furthermore, both His-
panics and African Americans are underrepresented in 67
Raising log income by one standard deviation at the
Forrester’s survey relative to their shares in the population mean of log income is equivalent to raising income from
reported by the Census. However, the income distribu- $48,000 to $114,000. The standard deviation of log income
tions for these groups in Forrester’s survey are quite close is .875.
to those in the March 2005 Current Population Survey,
so Forrester’s sample appears to be a reasonable enough 68
These results are from (1) a probit regression of computer
representation for us to report results. ownership on individual characteristics and (2) a probit
regression of broadband adoption on individual and zip
63
Some non-Hispanic whites, African Americans, and code characteristics, conditional on computer ownership.
Asian Americans also lack English proficiency, but the
English proficiency requirement should affect the repre- 69
Another strategy for increasing computer ownership is
sentativeness of the Hispanic sample most. manufacturing and distributing extremely low-cost com-
puters, which is the goal of the nonprofit One Laptop Per
64
Because Hispanics preferring to speak Spanish are a Child foundation, www.laptop.org. Its focus is raising
considerable share of California’s population, the overall computer use among children in developing countries.
P u b l ic P o l ic y I n s tit u te o f C a l if o r n ia 19
California Economic Policy
Broadband for All?
Brown, Jeffery, and Austan Goolsbee, “Does the Internet Federal Communications Commission, Availability of
Make Markets More Competitive? Evidence from the Life Advanced Telecommunications Capacity in the United
Insurance Industry,” Journal of Political Economy, Vol. States: Fourth Report to Congress, Washington, D.C.,
110, No. 3, 2002, pp. 481–507. 2004.
California Public Utilities Commission, “Broadband Federal Communications Commission, High Speed Ser-
Deployment in California,” San Francisco, California, vices for Internet Access: Status as of December 31, 2005,
2005. Washington, D.C., July 2006.
California Public Utilities Commission, “Broadband Flamm, Kenneth, “Diagnosing the Disconnected: Where
Deployment in California,” San Francisco, California, and Why Is Broadband Access Unavailable in the U.S.?”
2006. unpublished draft, 2006.
“Cities Deliver Broadband for Less,” Wired News, Gillett, Sharon, William Lehr, and Carlos Osorio, “Local
March 7, 2003, available at www.wired.com/news/ Government Broadband Initiatives,” Telecommunications
business/0,1367,57927,00.html. Policy, Vol. 28, 2004, pp. 537–558.
Clarke, Richard, Kevin Hassett, Zoya Ivanova, and Lau- “Google Gives City Free Wi-Fi,” San Francisco Chronicle,
rence Kotlikoff, “Assessing the Gains from Telecom Com- August 16, 2006.
petition,” NBER Working Paper #10482, Cambridge,
Massachusetts, May 2004. Goolsbee, Austan, “In a World Without Borders: The
Impact of Taxes on Internet Commerce,” Quarterly Jour-
Clay, Karen, Ramayya Krishnan, Eric Wolff, and Danny nal of Economics, Vol. 115, No. 2, 2000, pp. 561–576.
Fernandes, “Retail Strategies on the Web: Price and
Non-Price Competition in the Online Book Industry,” Goolsbee, Austan, and Peter Klenow, “Evidence on Learn-
Journal of Industrial Economics, Vol. 50, No. 3, 2002, ing and Network Externalities in the Diffusion of Home
pp. 351–367. Computers,” Journal of Law and Economics, Vol. 45,
2002, pp. 317–343.
Columbia Telecommunications Corporation, “Fiber
Optics for Government and Public Broadband: A Feasibil- Greenstein, Shane, and Jeff Prince, “The Diffusion of the
ity Study,” prepared for the City and County of San Fran- Internet and the Geography of the Digital Divide in the
cisco, Columbia, Maryland, 2007. United States,” NBER Working Paper #12182, Cambridge,
Massachusetts, 2006.
“Companies That Fought Cities on Wi-Fi, Now Rush to
Join In,” Wall Street Journal, March 20, 2006. Grubesic, Tony, “A Spatial Taxonomy of Broadband
Regions in the United States,” Information Economics
Corning, Inc.,“Bridging the Last Mile: Access Network and Policy, Vol. 18, 2006, pp. 423–448.
Wireline Architectures,” White Paper WP 6300, Corning,
New York, 2001. Gunkel, David, “Second Thoughts: Toward a Critique of
the Digital Divide,” New Media & Society, Vol. 5, 2003,
Corning, Inc., “Broadband Technology Overview,” White pp. 499–522.
Paper WP 6321, Corning, New York, 2005.
Kolko, Jed, “Why Should Governments Support Broadband
Criterion Economics, “The Effects of Ubiquitous Broad- Adoption?” Public Policy Institute of California, Working
band Adoption on Investment, Jobs, and the U.S. Econ- Paper #2007.01, San Francisco, California, 2007.
omy,” Washington, D.C., 2003.
Kuhn, Peter, and Mikal Skuterud, “Internet Job Search
DiMaggio, Paul, et al., “From Unequal Access to Differen- and Unemployment Durations,” American Economic
tiated Use,” in Kathryn Neckerman, ed., Social Inequality, Review, Vol. 94, No. 1, March 2004, pp. 218–232.
Russell Sage Foundation, New York, New York, 2004.
Morton, Fiona Scott, Florian Zettelmeyer, and Jorge Silva-
Downes, Tom, and Shane Greenstein, “Understanding Risso, “Internet Car Retailing,” Journal of Industrial
Why Universal Service Obligations May Be Unnecessary: Economics, Vol. 49, No. 4, 2001, pp. 501–519.
The Private Development of Local Internet Access Mar-
kets,” unpublished draft, 2006.
20 P u b l ic P o l ic y I n s tit u te o f C a l if o r n ia
California Economic Policy
Broadband for All?
Morton, Fiona Scott, Florian Zettelmeyer, and Jorge Silva- “S.F. Picks Google Wi-Fi Team,” San Francisco Chronicle,
Risso, “Consumer Information and Discrimination: Does April 6, 2006.
the Internet Affect the Pricing of New Cars to Women and
Minorities?” Quantitative Marketing and Economics, “Some Worries as San Francisco Goes Wireless,” New
Vol. 1, 2003, pp. 65–92. York Times, April 10, 2006.
Prieger, James, “The Supply Side of the Digital Divide: Is “Verizon’s Fios Services Build Momentun,” CNET, August
There Equal Availability in the Broadband Internet Access 1, 2006, available at http://news.com.com/Verizons+Fios+
Market?” Economic Inquiry, Vol. 41, No. 2, April 2003, services+build+momentum/2100-1034_3-6101038.html.
pp. 346–363.
“What If They Built an Urban Wireless Network and Hardly
Prince, Jeff, “Measuring the Digital Divide: Structural Anyone Used It?” New York Times, June 26, 2006.
Estimation of Demand for Personal Computers,” unpub-
lished manuscript, 2003. “Wi-Pie in the Sky,” The Economist, March 9, 2006.
“Public, Private Collaboration To Design Silicon Valley “With a Dish, Broadband Goes Rural,” New York Times,
Wide Wireless Network,” Joint Venture Silicon Valley November 14, 2006.
press release, January 26, San Jose, California, 2006.
P u b l ic P o l ic y I n s tit u te o f C a l if o r n ia 21
California Economic Policy
Broadband for All?