The Internet, Public Libraries, and the Digital Divide

Abstract

Virtually every public library in the United States provides public access Internet computers as a role central to its mission. This article addresses the issue of why the Internet matters for public libraries, both in terms of impacts on the general public, particularly the digitally disadvantaged (part one of the article) and impacts on libraries themselves (part two of the article). Each part of this report begins with an extensive literature review, followed by a data analysis section. In part one, the author uses the 2000 United States Census dataset to evaluate library efforts to bridge the digital divide, by analyzing differences in the growth of public terminals in library systems serving counties with different levels of household income, households in poverty, non-white households, and non-English-speaking households. The analysis finds no disparity in the number of public computers available in areas with high and low incomes but finds a significant—and widening—disparity in the number of computers available in areas with a higher versus lower percentage of non-white and non-English-speaking households. In part two of the report, the author uses a random effects linear regression model to estimate the effects of Internet access on library usage. This analysis finds that having Internet terminals (versus having no Internet terminals) has a significant positive effect on a library's visits and reference transactions, but not on a library's circulation. An increase in the number of Internet terminals has no significant effect, positive or negative, on visits, reference transactions, or circulation.

KEYWORDS:

• digital divide
• public libraries
• internet
• computers

The Internet has had a profound impact on public libraries. Virtually every public library in the United States provides public access Internet computers, and libraries and the public increasingly see Internet provision as a role central to the mission of the public library. This article addresses the issue of why the Internet matters for public libraries, both in terms of impacts on the general public, particularly the digitally disadvantaged or “information have nots” (part one of the article), and on libraries themselves (part two). Part two further investigates the findings from the literature review by analyzing statistical data from two primary sources: the national Public Libraries Survey and the 2000 United States Census. ¹ For analyses in the first report, the longitudinal dataset used covers the years 1998–2006.

To this dataset is added 2000 county-level Census data on the percent of households living below the poverty line; the percent of the population of races other than white; the percent of household whose main language is other than English; and the median household income in dollars. Library service areas align variously with cities, counties, and school districts and some special library districts span several of these jurisdictions, so the match to Census data is imperfect. Despite this limitation, county-level measures are accurate enough predictors of library service area characteristics for the purposes of this analysis. ²

In part one, the Census dataset is used to evaluate library efforts to bridge the digital divide, by analyzing differences in the growth of public terminals in library systems serving counties with different levels of household income, households in poverty, non-white households, and non-English-speaking households. This analysis finds no disparity in the number of public computers available in areas with high and low incomes, but finds a significant, and widening, disparity in the number of computers available in areas with a higher versus lower percentage of non-white and non-English-speaking households.

In part two, a random-effects linear regression model is used to estimate the effects of Internet access on library usage. This analysis finds that having Internet terminals (versus having no Internet terminals) has a significant positive effect on a library's visits and reference transactions, but not on a library's circulation. An increase in the number of Internet terminals has no significant effect, positive or negative, on visits, reference transactions, or circulation.

PUBLIC LIBRARIES AND THE DIGITAL DIVIDE

The idea of a “digital divide” that separates “information haves” from “have nots” was introduced by the National Telecommunications and Infrastructure Administration (NTIA) of the United States Department of Commerce (USDC) in a series of reports entitled Falling Through the Net (USDC 1995; 1998; 1999; 2000). These reports, released from 1995–2000, extend the concept of “universal service” from telephone service to computer and Internet access. The NTIA reports and many other contemporaneous empirical studies demonstrated wide disparities in computer and Internet access along numerous demographic lines, including income, race, education, and geographic region. Although some of these disparities have lessened or disappeared since the mid-1990s,and others have shifted, much evidence demonstrates that many disparities remain, and these disparities tend to magnify existing levels of disadvantage.

Public libraries were identified early on as important players in the task of equalizing computer and Internet access. In 1993, Vice President Al Gore suggested that public libraries could serve as a “safety net” in providing Internet access, and President Bill Clinton, in his 1994 State of the Union address, declared his intention “to connect every classroom, every clinic, every library, and every hospital in America to a national information superhighway by the year 2000” (McClure, Bertot, and Zweizig 1994, p. 1). The NTIA Falling Through the Net reports (USDC 1995; 1998; 1999; 2000) (themselves a product of the Clinton Administration) identify public libraries as important alternative access points.

Part one of this article explains the ways in which libraries have responded to the problem of unequal access to computers and the Internet. A review of the utilization of public library computers and Internet capabilities demonstrates that people who belong to groups characterized as digitally disadvantaged do tend to use library computers more frequently than those who are more advantaged, and they are more likely to depend on the library for access, although in some cases these differences are minimal. Wide disparities still remain between library systems, and, even when available, access alone is not sufficient to remedy broad inequities among various demographic groups.

Analysis of National Center for Education Statistics (NCES)/Institute of Museum and Library Services (IMLS) data on the number of Internet terminals in library systems—categorized by 2000 county-level Census data on household income, poverty level, race, and language spoken at home—shows that libraries have made strides in providing Internet access in high-poverty areas. They have been much less successful, however, in providing access in areas with higher than average levels of non-white and non-English-speaking households.

Disparities of Access in the United States

Numerous empirical studies have shown persistent disparities separating economic, racial, ethnic, and language groups in their access to computers and the Internet. While people in all categories of income, race, ethnicity, and language have seen an increase in access to technology over the past two decades, these increases have not affected everyone in the same way. Multiple digital divides persist, and new divides open as new technologies are introduced.

Falling through the net

In 1995, the NTIA released the first of a series of four Falling Through the Net reports (USDC 1995; 1998; 1999; 2000), which examined disparities of access to computers and the Internet throughout the United States. These reports were released through 2000, after which they were superseded by a new series entitled A Nation Online (USDC 2002; 2004). These reports demonstrate a continued increase in home access to computers and the Internet with persistent disparities among advantaged and disadvantaged groups.

The 1995 report (USDC 1995), which was based on data gathered by a November 1994 United States Census Current Population Survey, called for an extension of the concept of “universal service” from telephone penetration to computer and modem access. The report documented significant disparities of access based on demographic characteristics, and identified five categories of “information have nots”: the urban and rural poor, urban and rural racial minorities, young (younger than age 25 years) and old (age 55 years and older), less-educated in central cities, and residents of Northeast central cities and the South (USDC 1995).

The second Falling Through the Net report (USDC 1998), released in 1998 and drawing from 1997 data from the Census Current Population Survey, introduced the term “digital divide”. This report found that, in 3 years, computer and Internet access had increased substantially for United States households in general, but that significant disparities remained. In some cases, disparities were increasing. In particular, the gap in computer ownership and Internet access had grown between those with higher and lower incomes and between racial minorities and whites (USDC 1998).

The 1998 (USDC 1998) report identified four categories that were the “least connected,” which differed somewhat from the “have nots” of 1995 (USDC 1995): rural poor, rural and central city racial minorities, young households, and female-headed households. The third Falling Through the Net report (USDC 1999) was released in 1999 and drew from 1998 data from the Census Current Population Survey. This report showed that computer and Internet access had continued to increase, but that the digital divide was still widening. The 1999 report identified significant disparities of access between urban and rural households, lower- and upper-income households, racial and ethnic groups, and groups with different education levels.

The 2000 Falling Through the Net report (USDC 2000), which focused on a new concept of “digital inclusion” (xiii), highlighted continued increase in Internet access for all groups of respondents, and narrowing or stabilizing of some of the disparities identified in earlier reports. However, divides were still evident among households with different levels of income and education, different racial and ethnic groups, old and young, single and dual-parent families, and those with and without disabilities.

In 2002, NTIA released its fifth report (USDC 2002) summarizing Census data on Internet and computer use. This report (now entitled A Nation Online, rather than Falling Through the Net) is striking in its complete avoidance of the term digital divide. The 2002 report emphasized the fact that disadvantaged groups were increasing their computer and Internet use, but said little about persistent disparities. Advances in computer and Internet use were significant—for the nation as a whole and for disadvantaged groups. Relying on a Gini coefficient to measure inequality of distribution, the report identified a lessening of inequality of computer ownership and Internet use, measured against income, education, family type, and race/Hispanic origin. The report did note that inequalities still remained, particularly with regard to broadband access. The report did not take into account disparities in “intensity of use” (USDC 2002, 86), noting that “a person who occasionally uses the Internet at the library is counted the same as someone using a broadband connection for hours a day” (86).

In 2004, NTIA released its sixth report (USDC 2004) on Internet and computer use, and its second report bearing the title A Nation Online. This report, which drew from 2003 data from the Census Current Population Survey, focused on Americans' access to broadband Internet. 19.9% of Americans (33.9% of Internet users) had broadband Internet at home, up from 9.1% the previous year. Rural households were much less likely to have broadband access, with 22.1% of non-broadband rural households reporting that it was unavailable (compared with only 4.7% of non-broadband urban households). In addition, Internet users with broadband at home were more likely to engage in more online activities.

A number of authors (e.g., Bertot 2003; Fairlie 2005) have criticized the NTIA for shifting its focus from the “digital divide” to “digital inclusion,” arguing that the persistent digital divide is not getting the attention it deserves. Although Falling Through the Net and A Nation Online reports rhetoric merits such criticism, the reports' data sets provide substantial evidence of persistent disparities of computer and Internet access across multiple demographic categories, with groups that tend to be disadvantaged in other ways (e.g., minorities, the poor, and the lower-educated) also having the lowest levels of connection.

Many persistent disparities

Other studies have found similar evidence of persistent disparities, although some disagree about the relative magnitude or importance of the multiple divides. Most research is in agreement that significant divides persist along lines of income, race, education, and language. A gender divide existed early on, but disappeared fairly quickly (Liu 1996; United States General Accounting Office [USGAO] 2001; Rainie, Fox, Horrigan, Fallows, Lenhart, et al. 2005).

Hoffman and Novak (1998) found no significant difference between blacks and whites in home computer access after controlling for income. Likewise, a 2000 Pew Internet and American Life study (Lenhart, Rainie, Fox, Horrigan, and Spooner 2000) found significant disparities in Internet access between whites and non-whites, but concluded that most of this difference could be explained by income: 50% of whites had access compared with 36% of blacks and 44% of Hispanics; 78% of whites in households earning more than $75,000 were online compared with 79% of blacks and 69% of Hispanics in similar economic circumstances. Likewise, 68% of whites in households earning less than $30,000 were not online compared with 75% of blacks and 74% of Hispanics in similar households. Significant regional disparities in Internet access also appear to reflect differences in income and education levels (Spooner, Meredith, and Rainie 2003). These studies suggest that economic disparities may be more important to remedy than racial or geographic disparities.

However, Fairlie's (2005) analysis of the United States Census Current Population Survey data found that income accounted for some, but not all, of the racial disparities in home computer and Internet access. Hoffman and Novak (1998) also found that white students were much more likely than black students to access the Internet outside of home, work, or school, even after controlling for income. This line of research suggests that in terms of providing alternative access points, race may matter.

A 2001 USGAO report (USGAO 2001) found significant disparities of Internet access by race, education, and income levels, but noted that new technologies often reach individuals with higher levels of income and education, and urban areas, first, and suggested that disparities in Internet access could even out over time. Likewise, in 2001, the Pew Internet and American Life Project reported that “the Internet population looks more and more like the overall population of the United States,” but observed persistent disparities of access in the categories of income and age (Rainie and Packel 2001).

In 2000, the Pew Project found that 57% of Americans without Internet access said they did not plan to get online. Older Americans were the most likely to articulate this position (Lenhart, Rainie, Fox, Horrigan, and Spooner 2000). A 2003 Pew study also found that while Internet adoption had increased in all demographic categories, older Americans, minorities, those with modest amounts of income and education, those without jobs, non-parents, rural Americans, and Americans with disabilities were all significantly less likely than their counterparts to use the Internet (Madden and Rainie 2003).

D'Elia, Jörgensen, Woelfel, and Rodger (2002) found that Hispanic Americans and blacks used the Internet less than whites; lower-educated and lower-income people also used the Internet less than their higher-educated and higher-income counterparts. Similarly, Pew reported in 2000 that blacks used the Internet less in a typical day, but blacks who did use the Internet were more likely than whites to research jobs, find places to live, seek entertainment, get health care information, and pursue hobbies (Spooner, Rainie, Fox, Horrigan, and Lenhart 2000).

A digital language divide appears to exist as well. For example, Spanish-speaking Hispanics are much less likely than English-speaking Hispanics or non-Hispanic whites to own computers or access the Internet at home, even after controlling for education, family income, and immigrant status (Fairlie 2005). In 2007, only 32% of Spanish-speaking Hispanics used the Internet, compared with 78% of English-speaking Hispanics and 71% of non-Hispanic whites (Fox and Livingston 2007). English-speaking Asian Americans are the heaviest users of the Internet (Spooner, Rainie, and Meredith 2001).

Internet access for people with disabilities is also uneven. Although the Internet improves information access for people with disabilities, it also poses barriers to access (Liu 1996). Of Americans with disabilities, 38% go online compared with 58% of all Americans (Lenhart, Horrigan, Rainie, Allen, Boyce, et al. 2003).

Children and students serve as a notable case because they often have a multitude of access points. However, their access reflects disparities similar to those of the population as a whole. Research by the Public Access Computing Project (PACP) found that children ages 12–18 years have access to the Internet at an average of 4.2 different sites, and many strategized their use of multiple facilities for computer access (Gordon, Moore, Gordon, and Heuertz 2003). Of those surveyed by PACP, 90% had access at home and 50% had access at a public library. In a 2007 survey, 100% of youth reported having Internet access at one or more locations, and 85.8% had access at home. Black youth, youth in urban areas, youth whose parents were less educated, and youth who were not performing well in school were less likely to have Internet access at home (D'Elia, Abbas, and Bishop 2007).

New technologies, new divides

As new Internet technologies develop, digital divides that seem to be closing reappear. For example, in the mid-2000s, as more and more Americans in all demographic categories were beginning to have at least some level of Internet access, evidence began to mount of large divides in broadband access (considered as 1.5 Mbps or greater connection speed).

In 2005, Americans age 65 years and older, blacks, and those with less education were less likely to go online than others and also less likely to have high-speed access (Fox 2005). By 2008, 55% of Americans had broadband Internet connections at home; however, only 25% of low-income Americans and 43% of blacks had broadband Internet at home. These two groups also saw slower growth rates in home broadband Internet connections, possibly suggesting a widening divide. Older Americans, lower-middle-income Americans, and rural Americans had lower rates of broadband access at home; however, these groups showed faster-than-average growth from 2007 to 2008, possibly suggesting a narrowing divide (Horrigan 2008a).

In 2007, the Pew Internet and American Life Project began investigating mobile access to the Internet, digital data, and tools. A 2007 study found that 58% of adults had used a cell phone or personal digital assistant for mobile non-voice data activities, and 41% had used the Internet away from home or work with a wireless laptop or a handheld device. English-speaking Hispanics were the most frequent “untethered” users. The majority of untethered users also had access at home, usually broadband (Horrigan 2008b).

In 2008, a panel of technology experts surveyed by Pew predicted that mobile devices would be the primary means of Internet connection by 2020 (Anderson and Rainie 2008). Horrigan (2009) identifies personal mobility as the key factor in “sophisticated” (98) use of information and communication technologies, and notes:

As a large portion of the online population gravitates to wireless and mobile access to supplement their home high-speed connection, this increases the supply of and demand for online content. Institutions—whether they are governments or news organizations—have greater incentives to optimize their services to be consumed online. More people have greater opportunity to share their advice, creativity, and observations online. This makes exclusion from the network of people and information found online more costly than in the past. (99)

The most persistent digital divides appear to separate Americans of differing income and education levels, race, and language. Additional divides exist along lines of region, age, and disability. Furthermore, while digital divides seem to narrow over time, new gulfs open up as new technologies are introduced. There has not been one single digital divide, but rather a series of divides that attend each new technology: first computers, then dial-up Internet access, then broadband access, and now mobile access. In addition, divides appear to exist in terms of quality of access, although these are less frequently measured.

Benefits of Computer Access

Disparities of access to computers and the Internet are important for numerous reasons. A wide range of studies have shown at least moderate benefits of computer access and information technology skills in several categories, including economic and educational advancement, community participation, access to government services, and access to health information (USDC 2000; 2002; Fairlie 2005).

Access to home computers appears to provide some measurable educational benefit to students. Students with home computers perform higher on standardized tests; however, benefits of having a computer at home are greater for boys, white students, and students of higher socioeconomic status (Attewell and Battle 1999). Children with home computers perform slightly (but significantly) better than those without computers on several cognitive tests (Attewell, Suazo-Garcia, and Battle 2003). Teenagers with home computers are 6 to 8 percentage points more likely to graduate from high school, after controlling for individual, parental, and family characteristics including measures of education and income. Home computers also reduce negative activities such as school suspension and crime (Fairlie 2005).

In an experimental study in which college students were given free computers, Fairlie and London (2009) found small positive effects on educational outcomes for the students who received the computers. Minority students and students living farthest from campus saw the most educational benefit, in part because of increased flexibility of access. However, students often used the computers for non-educational purposes (e.g., games and social networking) and the students who spent more time on these activities saw less educational benefit.

Internet access can provide economic benefits, both indirectly, through development of marketable technology skills, and directly, through e-commerce. Goss and Phillips (2002) found that Internet skills positively affect wages, and Morton, Zettelmeyer, and Silva-Russo (2003) observed that online shopping can smooth out disparities between prices paid by whites and minorities; minorities can gain an economic advantage by shopping online.

The Internet can also benefit people in everyday life, by helping people find information to make major decisions (Horrigan and Rainie 2006), and (potentially) to increase social capital, although it can also negatively impact socialization. Informational uses of the Internet can have positive effects on individual production of social capital, although social-recreational uses can have negative effects (Shah, Kwak, and Holbert 2001). Internet use can reduce time spent in social interactions (Nie and Erbring 2002; Nie and Hillygus 2002).

Insofar as there are benefits to Internet access in general, these benefits increase with broadband access. For one reason, broadband users participate in more online activities (Horrigan and Rainie 2002). The USDC, in a 2008 report on broadband access in the United States, identified numerous benefits of broadband access:

By making it possible to access, use, and share information, news, and entertainment with ever increasing speed, broadband technology knits geographically-distant individuals and businesses more closely together, increases productivity, and enriches the quality of life. In so doing, it catalyzes economic growth and job creation that, in turn, provide unparalleled new opportunities for our nation's citizens. (USDC 2008, i)

Finally, as information technologies advance and greater numbers of people join telecommunications networks, there is a danger that the increasingly small numbers of “information have nots” will suffer even greater disadvantages. Tongia and Wilson (2007) argue that the cost of exclusion from a network rises exponentially as fewer people are excluded.

Criticism of the Digital Divide Concept

Not everyone agrees with the picture of the digital divide painted by the United States government and others following its lead. The primary criticism of this approach is of its generally narrow focus on access, to the exclusion of other factors that influence inequalities, including technology use, information literacy, and broader social conditions such as economic and class inequalities.

Gurstein (2003) argues that with its narrow focus on access to computers and the Internet, the digital divide “is little more than a marketing campaign for Internet service providers” (Abstract). He proposes an alternative approach called “effective use,” which considers not just access but “the capacity and opportunity to successfully integrate ICTs [information and communication technologies] into the accomplishment of self or collaboratively identified goals” (§ 5 [“Effective Use”]).

One substantial problem with a narrow focus on access is the possibility of increasing access without actually decreasing inequalities. For example, the NTIA reports released by the Bush Administration (i.e., those from 2000 onward) began to paint a rosy picture of a closing divide, with much government research during this time period lessening its focus on quite substantial remaining inequalities (Bertot 2003).

Stevenson (2009) has criticized the entire Falling Through the Net series (USDC 1995; 1998; 1999; 2000), as well as the 2002 A Nation Online report (USDC 2002; 2004) for framing the policy debate around the digital divide as a problem of access rather than one of class struggle related to economic modes of production, arguing that the government's depiction of the digital divide as an individual rather than a social problem served as a vehicle for claiming the divide had closed “as more and more Americans purchased computers” (Stevenson 2009, 11).

Warschauer (2003) also criticizes the digital divide concept for ignoring broader social problems with too narrow a focus on improving access, arguing that technology use is as important as technology access:

Technology does not exist as an external variable to be injected from the outside to bring about certain results. It is woven into social systems and processes. And from a policy standpoint, the goal of bringing technology to marginalized groups is not merely to overcome a technological divide but instead to further a process of social inclusion. Realizing this objective involves not only providing computers and Internet links or shifting to online platforms but also developing relevant content in diverse languages, promoting literacy and education, and mobilizing community and institutional support toward achieving community goals. Technology then becomes a means, and often a powerful one, rather than an end in itself. (23)

The concept of the digital divide has been focused largely on access to computers and the Internet. While access is an important measure, it is not necessarily the most appropriate indicator of social or technological equality. The digital divide is best treated as a multifaceted concept, encompassing not just access to computers and technology but telecommunications infrastructure, economic conditions, information access, and information literacy (Bertot 2003).

Libraries and the Digital Divide

Library provision of free public Internet access does appear to reach those who are digitally disadvantaged. Members of disadvantaged groups tend to use library computers more frequently and are more likely to rely on the library as their only source of Internet access. However, great disparities exist in the level of access among library systems, and even among branches within a system. Furthermore, provision of Internet access is only one part of bridging the digital divide, which is a function of broader social and technological inequities.

Libraries bridge the digital divide

Public libraries have long been recognized as important players in providing equal access to computers and the Internet, at least since the Clinton Administration's advocacy for library provision of Internet access, and both empirical and anecdotal evidence supports the observation that the digitally disadvantaged are heavy users of library computers. Library patrons engage in a great variety of activities on the Internet.

Libraries as public access points for the digitally disadvantaged

Libraries have often advocated public computer access as a means of narrowing information gaps. Lago (1993) argued that public libraries should be lobbying for funding to provide public access, without which:

information will surely become a pay-per-use commodity, effectively denying the majority of Americans the ability to reach their full potential and creating two distinct classes within our society: a small cadre of those in the know and those left at their mercy (66).

Echoing this sentiment, Liu (1996) suggested that Internet provision could especially help libraries bridge gaps in information access between rural and metropolitan areas, and Hendrix (1997) promoted the public library as “the bus service … on the [Information] Superhighway” (4).

The NTIA Falling Through the Net reports (USDC 1995; 1998; 1999; 2000) emphasize the importance of public access points such as public libraries in providing Internet access, particularly to disadvantaged groups. The 1998 report, for example, concludes, “Because it may take time before these groups become connected at home, it is still essential that schools, libraries, and other community access centers … provide computer access in order to connect significant portions of our population” (USDC 1998, § 4 [“Policy Implications”]).

The 1999 report found that disadvantaged groups were much more likely to use the Internet at a public library than the national average, in particular unemployed persons (three times more likely), and those “not in the labor force” (twice as likely), as well as those earning less than $25,000, those with less than a high-school education, those in female-headed households, and American Indians/Eskimos/Aleuts, blacks, and Hispanics (USDC 1999). The 2000 NTIA report (USDC 2000) suggested that public libraries remained more important as Internet access points for disadvantaged groups; blacks, Asian Americans, and Pacific Islanders were more likely to use the Internet at the public library than whites or Hispanics, and unemployed individuals were more likely to use the Internet at the public library than employed individuals. By 2001, public libraries were still more important for disadvantaged groups: 8.6% of whites who used the Internet accessed it at public libraries, compared to 11.6% of Asian Americans/Pacific Islanders, 13.8% of Hispanics, and 18.7% of blacks. However, most library Internet users of all races also used the Internet at other locations (USDC 2002).

In addition to the NTIA reports, an abundance of evidence supports the finding that, while only a small proportion of the total population uses library computers, those who do are more likely to be digitally disadvantaged. A 2002 Census Current Population Survey found that, compared with the national average of households that used public libraries for computer and Internet access (8.9%), households with certain demographic characteristics were more or less likely to have used libraries for this purpose. The most likely households to use libraries were:

• households in the Midwest (9.9%) and West (9.9%);

• households in city centers (10%);

• black, non-Hispanic (12.9%) and mixed race/ethnicity households (12.3%);

• households with higher levels of education (10%–11.1% for households with at least one member who had attended college);

• households below the poverty line (10.4%); households with children, particularly with students (17.9% of households with a high school student); and

• households with a job-seeker (17.1%) (Glander and Dam 2006).

PACP research found that, while computer use varies widely across income, education, and employment levels, library computer use is nearly equal across all levels in all three categories (Heuertz, Gordon, Moore, and Gordon 2002; Moore, Gordon, Gordon, and Heuertz 2002). Even more striking, trends of computer use and library computer use by race move in the opposite direction from one another: Asian American and white respondents were most likely to be computer users but least likely to be library computer users, while black, Hispanic, and Native American respondents were less likely to be computer users but more likely to be library computer users. Similarly, among computer users, library computer use is more prevalent among, and more important to, respondents in the lowest income bracket, those with less than a high school education, the unemployed and those who are working part time, and blacks, Hispanics, and Native Americans. Additionally, low-income library computer users were less likely to have computers at home, or to have access anywhere outside the public library (Heuertz, Gordon, Moore, and Gordon 2002; Moore, Gordon, Gordon, and Heuertz 2002).

Black students are more likely than white students to use library computers for schoolwork, and they use library computers more (Gordon, Moore, Gordon, and Heuertz 2003). For youth without access to the Internet at home, the library is the least frequently used alternate source of access; however, Hispanic, black, Asian American and American Indian youth, youth in urban and suburban areas, and youth whose primary language at home is a language other than English were more likely to access the Internet at the public library (D'Elia, Abbas, and Bishop 2007).

In addition to being more frequent library computer users, members of disadvantaged groups are more likely to rely on library computers as their only point of Internet access. A 1998 evaluation of the “OnLine at PA Libraries” grant program found that 35% of library Internet users had no other means of accessing the Internet (Oder 1998). A PACP study (Heuertz, Gordon, Moore, and Gordon 2002) found that 20% of people who use library computers said they had no other computer access and 30% said they had no other Internet access; patrons with library-only Internet access were more likely to have lower incomes and lower education levels. A federally funded study of computer use in Colorado libraries in 2002 found that 67% of all patrons and 85% of patrons older than age 55 years had no other access to the Internet except through library computers (Moe 2004). A 2004 Pew study (Harwood and Rainie 2004) found that many of those who use the Internet from “third places” (1) such as libraries are “anywhere, anytime users of the Internet” (1), who go online in multiple locations. A small number (3% of the total United States population) depend on “third places” as their only point of access; these users tend to be from poor, rural households, and to be newcomers to the Internet. In 2008, 69% of library computer users reported that they did not have Internet access at home, particularly people living in poorer communities and seniors (Barber and Wallace 2008).

Anecdotal evidence also points to heavy use of library computers—in particular e-mail—by the homeless, although no statistics support this finding, and librarians cannot always tell whether a patron is homeless simply by observation (Rogers 1999; Wylie 1999; Weiss 2000; Westneat 2009).

How library computers are used

Discussions of whether library provision of access to computers resolves technological inequalities are greatly informed by an understanding of how people actually use the computers in the library.

The US Impact study—a research project at the University of Washington Information School examining the impact of free access to computers and the Internet on the well-being of individuals, families, and communities ³ —is the most expansive study to date on how people use public library computers and how this use impacts their lives, but several smaller-scale descriptive studies have previously investigated the question of use (American Library Association [ALA] 2007; Bertot, Jaeger, Langa, and McClure 2006a; 2006b; Curry 2002; Gordon, Moore, and Gordon 2003; Gordon, Moore, Gordon, & Heuertz 2003; Heuertz, Gordon, Moore, & Gordon 2002 Gross, Drezang, and Holt 2004; Mantell, 2008; Moe 2004; Slone 2007). While not all agree on the most frequent uses, the consensus seems to be that library computers are used for a great variety of activities—not all of them anticipated by librarians. As Schuyler (1995) put it:

We started out with this Big-Internet thing thinking we could provide some valuable information to the public free of extra charge … Teledemocracy would be proven, and, not just coincidentally, the public library would be sitting in the middle of this whole thing, survival guaranteed by the essential nature of the service. Right. But the fact is, I want access to multi-user dungeons, Internet relay chat, alt.sex.bondage, and the Anarchist's Cookbook, so please get out of my way (“please” optional) (43).

Most studies agree that some of the most frequent activities that library patrons engage in on library computers are e-mail; ⁴ general web surfing; education, research, and information-seeking; and entertainment. Other common uses include job seeking (ALA 2007), social networking (Curry 2002), e-commerce/e-business (Heuertz, Gordon, Moore, & Gordon 2002), and e-government (Bertot, Jaeger, Langa, & McClure 2006a; 2006b; Bertot McClure, Wright, Jensen, and Thomas 2008). The United States Impact study is investigating use in terms of the following seven domains: civic engagement, e-commerce, education, e-government, health, employment, and social inclusion (Becker et al. 2010; Research Overview 2009).

Several other noteworthy uses of the Internet in libraries, which may not be captured by these six domains, are mentioned in recent library literature. These include word processing and spreadsheets (Heuertz, Gordon, Moore, & Gordon 2002), learning basic computer skills (Moe 2004; Gordon, Moore, Gordon, & Heuertz 2003), printing (McClure, Bertot, & Zweizig 1994; Heuertz, Gordon, Moore, & Gordon 2002), and pornography (Burt 2000). ⁵

Several studies have identified differences in use among various demographic groups. PACP studies found some differences in use between users with and without other ways of accessing the Internet: users with no other access were more likely to use library computers for e-mail while those with other access were more likely to use the computers for education, word processing, business, games, stocks and bonds, and spreadsheets (Heuertz, Gordon, Moore, & Gordon 2002). PACP also observed that low-income patrons were more likely to use library computers for employment-related functions such as job seeking and resume preparation; economic downturns have brought greater attention to this role (Gordon, Moore, & Gordon 2003; Brustein 2009; Gronowska 2009; Saulny and Cullotta 2009; Yates 2009). Low-income adults are also more likely to use library computers for chatting and games than higher-income adults (Gordon, Moore, and Gordon 2003). A 2002 study of Colorado library users (Moe 2004) found that younger users, minorities, users with lower education levels, and poorer users were significantly more likely to use library computers to develop basic technology skills such as searching the Internet, sending e-mail, and using word processing programs, and that users older than age 55 years, Asian Americans/Pacific Islanders, lower-income users, and more highly educated users were more likely to seek staff assistance or enroll in computer classes.

But disparities still remain

Despite attempts by libraries and policymakers to lessen inequities through computer and Internet provision, great disparities in access still remain, in many cases along the same lines that separate individuals and households in their computer and Internet access. Furthermore, provision of access by public libraries is not sufficient to remedy broad social inequities.

Disparities of access among public libraries

A 1994 Public Libraries and the Internet (PLI) survey (McClure, Bertot, and Zweizig 1994)—the first in a series of ongoing longitudinal studies of public libraries' Internet connectivity and use—found disparities in public library connection to the Internet by region, population of legal service area, urban/rural designation, and operating/material expenditures. Public libraries in the West reported the greatest percentage of Internet connections (28.2%), and public libraries in the Midwest reported the least (15.4%). Libraries serving a population of 1 million or more were nearly six times more likely to be connected to the Internet (77.0%) than libraries serving a population of 5,000 or less (13.3%). Urban libraries were nearly five times more likely to be connected to the Internet (78.9%) than rural libraries (16.8%). The number of libraries providing public access to the Internet (12.7%) did not vary significantly by population or urban/rural designation, although it did vary by region; libraries connected to the Internet in the West and South were more likely to provide public Internet access terminals than those in the Midwest and Northeast.

Rural libraries have faced particular barriers to Internet development, including insufficient funding, inadequate staff and staff training, and lack of access to infrastructure (Liu 1996; Weingarten, Bolt, Bard, and Windhausen 2007). PACP research (Heuertz, Gordon, Gordon, and Moore 2003) found that rural libraries were slower to implement public access computing than their urban counterparts, but that rural librarians reported public access computing services to be less of a burden than urban and suburban librarians. Rural libraries struggle more with resources, especially technology training and support. As one state librarian put it, there are “no geeks in the farm belt” (Heuertz, Gordon, Gordon, and Moore 2003, 68). Many small libraries have no dedicated information technology staff: “Staff at small, independent libraries say it is not unusual for them to call on library users (especially teens), friends, and spouses for help” (Barber and Wallace 2008, 53). By 2004, the PLI study (Bertot, McClure, and Jaeger 2004) found a significant disparity between rural and urban libraries in terms of Internet access and broadband connectivity, as well as continuing regional disparities. Comparison of the 2004 and 2006 studies shows disparities of access among states (Bertot, McClune, and Jaeger 2004; Bertot, McClure, Jaeger, and Regan 2006; Jaeger, Bertot, McClure, and Rodriguez 2007). Some of the regional disparities of access can be explained by variations in the cost of connection. For example, the cost of broadband access varies greatly from region to region, state to state, and even within a state. Furthermore, in parts of the country broadband is simply not available (Weingarten, Bolt, Bard, and Windhausen 2007).

Libraries may not provide sufficient access to the Internet for patrons with disabilities. In 2000, only 28.8% of public libraries provided special software or hardware for people with disabilities (Bertot and McClure 2000), and in 2004, only 16.9% offered Internet training specifically targeted toward individuals with disabilities (Bertot, McClure, and Jaeger 2004). In addition, computer software is generally only provided in English (and sometimes Spanish), despite the multitude of languages spoken by library patrons (Gordon, Gordon, and Moore 2001).

In addition to disparities among library systems, great disparities can exist among branches within a library system, and these disparities can exacerbate existing inequalities. Library resources can be unevenly distributed among communities within the system due to differing levels of community advocacy; private residents in wealthier communities can augment library-provided resources; and commercial entities may be more likely to support communities with more economic power. Agosto (2005) and Hall (2007) give anecdotal evidence of intra-system disparities; Jue, Koontz, Magpantay, Lance, and Seidl (1999) offer empirical evidence of income disparities within library service areas.

Access Is Not Equity

Even if libraries can provide greater equality of access to computers and the Internet, access alone is not enough to ensure technological equity. Public access to the Internet is not necessarily of the same quality as home access. Access does not ensure effective use, and the digital divide is a function of broader social conditions for which access to computers and the Internet is not a sufficient remedy.

Public access simply may not be able to provide the same benefits that home access can. The NTIA 2002 A Nation Online report (USDC 2002) argues that “home Internet access may be thought of as a higher quality type of access because it is available (theoretically) 24 hours a day, seven days a week, while school or library access periods are limited to specific hours and often with time limits per session” (38). Families who own technology (e.g., computers, cell phones, and Internet connections) use it to stay connected with each other throughout the day, a use that is infeasible with library computers (Kennedy, Smith, Wells, and Wellman 2008).

In addition, social inequalities may make public computer use less beneficial to those who are most disadvantaged. For example, compared to white students, significantly more black students report that “it is too difficult to focus at the library computer stations; that there isn't enough time on the library computers; that they don't feel comfortable in libraries; and they don't have a library card” (Gordon, Moore, Gordon, and Heuertz 2003).

Perhaps as a result of the problems mentioned above, public library computers do not necessarily reach those users with the least computer access. While disadvantaged groups tend to rely more than others on libraries as Internet access points, the number of the most needy people using library computers has remained quite small. In 1999, 1.4% of the population used library computers, according to NTIA (USDC 1999), and this number had jumped to 5.4% by 2001 and 6.0% by 2004 (Harwood and Rainie 2004; USDC 2002). These numbers are significant but reflect only a small portion of the “information have nots.” These “have nots” may have other reasons for not using the Internet than simply a lack of access points: a 2003 Pew study (Lenhart, Horrigan, Rainie, Allen, Boyce, et al. 2003) found that 60% of Internet non-users know of a place in their community where the Internet is available, most often a public library.

Evidence seems to suggest that many library computer users have other places to access computers and the Internet, and do not depend on the library for access (D'Elia, Jörgensen, Woelfel, and Rodger 2002; Heuertz, Gordon, Moore, and Gordon 2002; Oder 1998; USDC 2002). Barber and Wallace (2008) found that the majority of library computer users had computer access, but not Internet access, at home. Children in particular tend to have multiple access points, and many strategize their use of multiple facilities for computer access (Gordon, Moore, Gordon, and Heuertz 2003).

Furthermore, library computers may not lend themselves to the kind of use that would make the most difference in remedying technological inequities. Gordon, Moore, Gordon, and Heuertz (2003) note:

As students begin to use computers, they are primarily consumers of computer resources, such as information and recreation. Learning to use computers as producers of digital products and information—such as creating web pages, programming computers or other creative uses such as video editing, requires more access than libraries are currently able to provide. Thus, the digital divide still keeps the disadvantaged on the consumer side of computer use, making the leap to the provider side extremely challenging (7).

Likewise, Gurstein (2003) also criticizes policies that focus on improving access to information and communication technologies without sufficient attention to their use. He argues that technology access as construed by government and private philanthropic efforts often means access “to consume and receive rather than produce and distribute” (§4 [“Access and beyond”]).

Aside from the technological difficulties in managing access and use of library computers, a larger and more confounding question is whether access is the most appropriate tool for remedying a digital divide that is a “multi-dimensional phenomenon” (Tongia and Wilson 2007, 3) encompassing not just access to technology but broader social and economic issues such as access to financial resources, knowledge, and training.

A 1998 Benton Foundation report (Benton Foundation 1998) on technological inequalities argued that information technology gaps between the rich and poor are not only the result of a lack of resources in poor communities, but also due to technological equality's low priority for most people, the ambivalence of poor people about the benefits of technology, and poor communities' lack of political clout.

Technology contributes to the potential deconcentration of jobs away from old urban manufacturing centers, which disproportionately affects the urban poor (United States Congress, Office of Technology Assessment 1995). The digital divide can be viewed as a function of labor conditions in the knowledge economy, not just the level of access to technology (Stevenson 2009). As Liu (1996) noted:

Social equity issues are more complicated than just providing equal access to the Internet or networked information systems … The federal government may be able to provide equal access to the national information network for everyone, but equal access does not mean equal participation in the process of telecommunications and Internet education. There are certainly no guarantees of equal outcomes. If equal access cannot lead to equal participation and outcomes, it may not be meaningful (35).

Lower income library users are more likely to use library computers, and are more likely to report that libraries are their only means of access to computers or the Internet, but lower income people are also less likely to have library cards, to use libraries, or to feel “comfortable” in their libraries. Additionally, lower income people are more likely to say that they do not have enough computer access at their libraries (Gordon, Moore, and Gordon 2003, 5), and there tend to be fewer library outlets in high-poverty areas (Jue, Koontz, Magpantay, Lance, and Seidl 1999).

As Gordon, Moore, and Gordon (2003) note, provision of computers, while helpful, may be inadequate at dealing with larger policy problems:

Since some logical policy implications—creating more libraries in low income neighborhoods, helping the less educated to get more education, giving poor people more money, and/or enabling people to move to higher-income neighborhoods—are not politically feasible at the moment, the findings from these analyses raise important challenges for libraries with regard to outreach if those with the greatest apparent need are to be served (18).

Data Analysis

To examine one aspect of public libraries' attempts to bridge digital divides, the growth of public Internet access workstations was compared in libraries serving areas with different demographic characteristics. Using 2000 county-level Census data, libraries were compared across four categories: 1) median household income, 2) percent of households below the poverty line, 3) percent of non-white residents, and 4) percent of households whose primary language is other than English. Findings show that libraries have made great strides in remedying divides related to income, but that widening disparities exist in the provision of Internet access in areas above and below the median of non-white and non-English-speaking households.

This analysis is subject to the same criticisms that Stevenson (2009) lodges against the PACP studies: it relies on descriptive statistics, which can show correlations but not causation; it identifies the digitally disadvantaged in terms of specific demographic variables; and it relies on the definition of the digital divide as a function of access to the Internet. Bearing these important caveats in mind, this analysis still provides some important information on the changing disparities of Internet access in public libraries over four demographic variables from 1998–2006. As Gurstein (2003), an outspoken advocate for expanding the digital divide concept beyond mere access, concedes, “Of course, ‘access’ is fundamental and basic to all other developments and uses of ICT [information and communication technologies] technology. Without ‘access’ little else is possible” (§ 4 [“Access and beyond”]).

Descriptive statistics

Minimum, median, and maximum levels of the four demographic variables used in this analysis are shown in Table 1. Of all the counties served by the library systems in the NCES/IMLS data, the lowest median annual household income was $17,524 (Pulaski County, KY) and the highest was $98,245 (Emporia County, VA). The lowest percentage of county households below the poverty line was 2.5% (Mercer County, NJ) and the highest was 46.7% (Marshall County, SD). The lowest percentage of county non-white households was 0.3% (Grant County, NE) and the highest was 94.2% (Wade Hampton, AK). The lowest percentage of county non-English-speaking households was 0.5% (Hooker County, NE) and the highest was 92.1% (Maverick County, TX).

TABLE 1 County-Level Census Data of Library Service Areas

Demographic variable	Minimum	Median	Maximum
Median annual household income ($)	$17,524	$42,006	$98,245
Households below poverty line (%)	2.5%	11.8%	46.7%
Non-white households below poverty line (%)	0.3%	6.0%	94.2%
Non-English-speaking households below poverty line (%)	0.5%	5.9%	92.1%

The statistics in Table 1 make clear that the demographic makeup of the counties served by public libraries varies considerably. The next sections examine the library response to digital inequities, by showing how libraries serving counties above and below the median in each of the four categories have experienced different growth rates in their public Internet access.

Differences between counties above and below medians

Analysis of the NCES/IMLS data shows that counties above and below each of the four medians experienced very different levels of Internet access in their public libraries. Results suggest that, at least as measured by the number of terminals per population, county-level disparities of public library Internet provision by household income and poverty level have disappeared, while disparities by race and language are widening.

Table 2 lists the mean number of public library Internet terminals per 1000 residents in counties with annual median incomes of above and below the national median of $42,006. In 1998, library systems serving poorer counties (as measured by median household income) had a moderately, but significantly, higher number of Internet workstations. However, this disparity disappeared after 1998, and from 2002 on, libraries in “poorer” counties actually had a higher per capita number of workstations, and this trend appears to be growing. Figure 1 graphically illustrates these differences in growth between libraries serving “poorer” and “richer” counties.

TABLE 2 Mean Number of Public Library Internet Terminals by County Annual Household Income

	Mean number of internet terminals per 1000 residents
Year	County annual household income < national median	County annual household income > national median
1998	.3585	.4135 *
1999	.7393	.7274
2000	1.0001	.9160 *
2001	1.1972	1.1221
2002	1.5117	1.3374 *
2003	1.7932	1.5885 *
2004	1.9226	1.6034 **
2005	2.1186	1.7486 **
2006	2.2023	1.8536 **
Note *Statistically significant difference (p < 0.05); **statistically significant difference (p < 0.01).

FIGURE 1 Growth of Public Internet Terminals by Household Income.

Table 3 and Figure 2 show a similar trend by comparing the difference in Internet terminal growth in libraries serving counties in which more or less than 10% of the households are below the federal poverty line. Table 3 lists the mean number of public library Internet terminals per 1000 residents in counties with poverty rates of greater than and less than 10%. ⁶ In 1998, libraries in poorer counties (as measured by percent of households below the poverty line) had significantly fewer workstations per 1000, but after 1998 there was no significant difference.

TABLE 3 Mean Number of Public Library Internet Terminals by County Poverty Level

	Mean number of internet terminals per 1000 residents
Year	County with households below poverty line ≤ 10%	County with households below poverty line > 10%
1998	.4476	.3540 *
1999	.6998	.7489
2000	.8993	.9863
2001	1.0938	1.1915
2002	1.3070	1.4813
2003	1.6195	1.7258
2004	1.6223	1.8316
2005	1.7490	2.0234
2006	1.8942	2.0939
Note *Statistically significant difference (p < 0.01).

FIGURE 2 Growth of Public Internet Terminals by Poverty Level.

Measuring growth of Internet workstations in library systems serving richer or poorer counties demonstrates the strides libraries have made in providing Internet access to the digitally disadvantaged. By this measure, federal programs such as E-rate (Bertot, McClure, and Fletcher 1997) and private gifts such as those made by the Gates Foundation would appear to have been successful in narrowing, or perhaps even eliminating, the digital divide.

However, comparing library systems by the racial and linguistic makeup of the counties they serve tells a different story. Table 4 and Figures 3 and 4 illustrate the different growth rates in library systems serving counties that have a percentage of non-white and non-English-speaking households above and below the median for all library systems (6.0% and 5.9%, respectively). The differing growth rates in both of these comparisons show widening disparities in public library Internet access across racial and linguistic lines. In all years, libraries in more racially and linguistically diverse counties (as measured by percent of non-white and non-English-speaking households) had significantly fewer terminals per 1000, and the mean difference increased over the 9 years of observation.

TABLE 4 Mean Number of Public Library Internet Terminals by Percentage Non-White and Non-English-Speaking Households in County

	Mean number of Internet terminals per 1000 residents
	Counties with percentage of non-white households:		Counties with percentage of non-English-speaking households:
Year	Below national median	Above national median	Below national median	Above national median
1998	.4691	.2942 **	.4125	.3600 *
1999	.8619	.6002 **	.7946	.6724 **
2000	1.1246	.7909 **	1.0644	.8558 **
2001	1.3561	.9622 **	1.2744	1.0491 **
2002	1.6596	1.1889 **	1.5539	1.3003 **
2003	1.9314	1.4539 **	1.8404	1.5490 **
2004	2.0525	1.4780 **	1.9524	1.5845 **
2005	2.2610	1.6126 **	2.1586	1.7221 **
2006	2.3941	1.6679 **	2.2785	1.7925 **
Note *Statistically significant difference (p < 0.05); **statistically significant difference (p < 0.01).

FIGURE 3 Growth of Public Internet Terminals by Race.

FIGURE 4 Growth of Public Internet Terminals by Language.

Figures 5 and 6 tell this story in a different way, by combining the categories of poverty (percent of county residents below the federal poverty line) with race and language. Figure 5 illustrates different growth rates for libraries serving counties in four combinations of the categories measuring poverty (more or less than 10% of county households below the poverty line) and percentage of non-white households (more or less than 6.0% of county households non-white).

FIGURE 5 Growth of Public Internet Terminals by Poverty Level and Race.

FIGURE 6 Growth of Public Internet Terminals by Poverty Level and Language.

This comparison shows that libraries serving counties with a higher percentage of white households have more Internet terminals, regardless of poverty level. But for libraries serving counties with fewer white households, poverty level is significant. Libraries that serve poor counties with more white households have significantly fewer Internet terminals than those that serve rich or poor counties with more white households.

Similarly, libraries serving rich or poor counties with a higher percentage of English-speaking households (as illustrated in Figure 6) have more Internet terminals than those serving counties with greater numbers of non-English-speaking households. In addition, libraries serving counties with more English-speaking households appear to buck the trend of more terminals in higher-poverty counties. For libraries serving counties with less than the median (5.9%) percentage of non-English-speaking households, libraries in wealthier counties (as measured by fewer than 10% of households below the poverty line) have more Internet terminals than libraries in poorer, whiter areas.

Discussion

This analysis demonstrates that, while libraries have made some substantial strides in providing access to computers and the Internet for some disadvantaged groups—in particular, those living in areas with lower than average household incomes and high poverty levels—they still have a long way to go in order to provide access for all disadvantaged groups commensurate with access available to those residing in higher than average non-white and non-English-speaking households.

In addition, this analysis suggests a further caution in declaring the digital divide closed. While the use of county-level demographic data paints disparities in broad strokes, it shows distinct differences in Internet access among library systems serving different areas. However, the fact that the county data sources used cannot show user-level data means that it likely overestimates the level of access for disadvantaged individuals within counties. For example, a poor person who happens to live in a low-poverty area may have a significantly lower level of public library Internet access than a poor person who lives in a high-poverty area.

The digital divide is a fuzzy concept, and one that has received less attention in recent years as more and more people gain some form of access to the Internet. However, as newer technologies develop, divides in access persist. Libraries provide an important alternative means of access for disadvantaged people, although the neediest people are not necessarily the ones who use library computers. Substantial disparities among libraries also keep libraries from providing access that comes close to matching access for members of family in higher-income households.

Finally, as numerous authors have pointed out, access to computers and the Internet is only one measure of technological correspondence, and it may not be the most important. It is likely that people who are disadvantaged in terms of access are disadvantaged in other ways as well. While libraries may not be able to remedy all social inequities, they should be aware that providing access is only part of the story. If libraries are to address the digital divides that still remain, they should focus on other services beyond access, such as outreach and training, that will help to address inequities at a broader level.

THE INTERNET AND PUBLIC LIBRARIES

Opinion and analysis about the effects of computers and the Internet on public library services has generally taken two tacks: First, the effects of the Internet and computer use in general on public library services, exemplified by D'Elia, Jörgensen, Woelfel, and Rodger (2002); second, the effects of access to computers and the Internet provided by the library on library services, exemplified by Heuertz, Gordon, Gordon, and Moore (2003). These two approaches are not necessarily mutually exclusive, and in many cases the effects attributed to libraries' public access computers would appear to extend to computers and the Internet generally.

The Internet has had broad impacts on public libraries in ways not directly related to the provision of public access, including effects on acquisitions, cataloging, reference services, interlibrary loan, electronic publishing, digitization and electronic document delivery, library and information science education, and library administration (Liu 1996; 2001; McKenna 1994). These effects are beyond the scope of this article; however, they are mentioned as a way of placing library provision of Internet access in a wider context of public libraries and the Internet.

Part two of this article investigates the impact of the Internet and public provision of Internet access on libraries themselves. The discussion examines the effects of the Internet on library service roles, highlighting three particular ways in which libraries have distinguished themselves as relevant in relation to the Internet: as guides to the Internet, more than the Internet, and providers of Internet access. The article then digs a little deeper into the effects of Internet provision on library use and policies. In the data analysis section, a random-effects regression model is used to estimate the impact of public Internet terminals on library visits, circulation, and reference transactions.

Effects on Library Service Roles

From the beginnings of Internet adoption in libraries, librarians have wrestled with defining Internet provision in terms of traditional library roles, and with reconceptualizing traditional roles in terms of Internet access. Numerous authors have suggested new or changed library roles as a result of the Internet; these roles illustrate the variety of ways in which libraries have conceptualized Internet provision. Some library roles still appear to be in the process of development. Some of the suggested new library roles, such as screen monitors and emergency responders, are completely new, while others, such as providing training and assistance, are extensions of libraries' traditional roles. Some, such as “anyplace anywhere anytime individualized information provider” (McClure and Jaeger 2009, 51) are supposed to supersede traditional roles, and others, such as providing job-seeking resources, are seen as adding on to or complementing existing roles.

McClure, Ryan, and Moen (1993, 25–26) note that the standard Planning and Role-Setting for Public Libraries, which identified eight “key service roles” for libraries to use in planning, inhibited public libraries' technology development in its inattention to technology planning and its ignorance of any roles related to electronic networking. Focus groups suggested that the role-setting guide be expanded to include a new service role such as “Electronic Networked Public Library.”

Other roles identified for public libraries and librarians in relation to the Internet have been broad and wide-ranging. They include Internet access provider; teacher/trainer; information service provider (including health care information, financial information, education, job seeking, community participation, gaming, and “anyplace anywhere anytime individualized information provider” [McClure and Jaeger 2009, 51–51] ); information evaluator; writer/content producer; web designer; e-government service provider; economic developer; marketer/promoter; “compliance officer”/ screen monitor; policy maker; unblocker; judger of purpose/denier of access; emergency responder; youth educational support provider; connector of friends, families, and others; digital library manager; virtual, seamless, and endless electronic resources provider; digital workplace/space; and digital ombudsperson (McClure and Jaeger 2009, pp. 50–51; McNeely 1999; Ratzan 2004; Valauskas 1995; Weingarten, Bolt, Bard, and Windhausen 2007, 3–5).

There is a lack of agreement in the library literature on the relationship between new roles and traditional library roles, and on the effects that new roles have had on traditional roles. For example, arguments and opinions in United States v. American Library Association, the 2003 Supreme Court case upholding the Children's Internet Protection Act, attributed 11 roles to librarians, only three of which librarians agreed with. Librarians took issue with Supreme Court characterization of their roles as protectors, “compliance officers”/screen monitors, policy makers, unblockers, and judgers of purpose/deniers of access, among others (Ratzan 2004). Bertot (2009) found that for some libraries, increases in public access technology services come at the expense of other library services; likewise, McClure and Jaeger (2009) argue that public libraries have deemphasized traditional roles in favor of new “Internet-enabled” (1–2) service roles and responses without a clear vision for these roles. They argue that without reconciliation of new roles with traditional roles, public libraries risk losing community support, becoming victims of federal technology policies, altering the meaning of the library profession, and failing to integrate service planning and technology planning.

In some cases, computers and traditional library services can be complementary. For example, job seekers may consult library books on resume writing and use library computers to write their resumes (Mantell 2008). Similarly, patrons may ask librarians to help them evaluate Internet sources, and to answer more complex reference questions, since they can find answers to simple questions on the Internet (Putnam, Feldstein, and Cohen 2003). Training patrons in using the Internet may help them use other information sources and develop more general information literacy skills (Brandt 2001). PACP found that libraries' traditional duties had stayed the same or increased while they had added additional roles, particularly providing one-on-one computer assistance (Heuertz, Gordon, Moore, & Gordon 2002). Librarians reported increased stress and workloads as a result of new computers, due to perceptions of changing roles and the need to police computer use (Gordon, Gordon, and Moore 2001). Regardless of the specific effects, it is clear that the Internet has considerably expanded the scope of library roles and services.

Library Relevance in the Age of the Internet

This article has identified three major themes that relate to how libraries define their value and relevance in relation to the Internet.

A great deal of writing about libraries and the Internet focuses on the question of libraries' relevance as information providers when so much is available and accessible online. The Internet is often described as a digital library (Dowlin 1999; Li 2008; Rainie Fox, Horrigan, Fallows, Lenhart, et al. 2005; Stefik 1996), and the Internet is perceived by many to be a superior source of information access which displaces the library (De Rosa, Cantrell, Cellentani, Hawk, Jenkins, et al. 2005; Li 2008; McClure, Ryan, and Moen 1993). Provision of Internet access in public libraries, coupled with related services, has had broad impacts on how libraries position themselves and define their value, and throughout the development of the Internet and its implementation in libraries, the library community has expressed mixed feelings about how the Internet affects public library relevance, and how libraries should adapt or respond. Libraries have taken three general approaches to explaining their relevance with regard to Internet provision: librarians as guides to the Internet, librarians as more than the Internet, and librarians as providers of Internet access.

Guides to the internet

The public view of the Internet as a universal information source is widely criticized by librarians as a false belief (Li 2008, 166). Services that libraries have offered to remedy this misperception include education and training, guidance and navigation, and even Internet organization.

Provision of training has been a ubiquitous service of public libraries since the beginnings of Internet access: In 1994, when only 20.9% of public libraries were connected to the Internet, 17.0% of public libraries offered Internet training services, more than the number of libraries which provided public Internet access (12.7%) (McClure, Bertot, and Zweizig 1994). By 2008, only 26.6% of libraries reported not offering information technology training services, and libraries reported that the impacts of public training included providing information literacy skills, offering technology training opportunities to those who would otherwise not have any, helping students with their school assignments and school work, providing general technology skills, helping patrons complete job applications, and helping users access and use e-government services and resources (Bertot, McClure, and Jaeger 2008). Porter and King (2008) add that librarians help users make use of social media and construct better searches on search engines, and Kirkpatrick (2007) imagines “librarians of the future” who help library users upload and classify online media.

Numerous authors have identified the need for Internet training by librarians. Brandt (2001) argues for Internet training for library patrons and librarians because of the Internet's rapidly changing nature and because public Internet skills do not meet the expectations placed on individuals by the information society. Li (2008) identifies five themes in librarians' perceptions of the Internet as an information source: it provides digital information; it is of enormous size; it keeps expanding; it keeps changing; it provides transient, perishable information; and it has little quality control. All these features of the Internet point to a need for librarians as information guardians, guides, and educators. Cooper (2007) argues that reference staff should offer to help find information on the Internet when patrons sign to use Internet terminals. He argues that providing access is not enough: “merely having access to a computer no more makes one expert in its use than having access to a kitchen makes one a chef” (18). He suggests that librarians should keep abreast of online resources so they can help patrons use them.

Most libraries report heavy demand for training services. In 1996, a rural Utah librarian reported overwhelming attendance at Internet training classes: “People expect us to train them, so if they come to the library and don't know how to use the Internet, we have to take time to show them” (Chepesiuk 1996, 44). In 2004, only 5.5% of libraries surveyed by the PLI study reported that patrons had not expressed interest in receiving training (Bertot, McClure, and Jaeger 2004).

Even when patrons do not expect to need Internet assistance, librarians often provide “‘at-the-point-of-need’ training service” (Bertot, Jaeger, Langa, and McClure 2006b, § 3 [“Public libraries and e-government”]). Of Colorado library users surveyed in 2002, 19% reported that they relied on staff assistance to learn new technology skills (Moe 2004). Likewise, PACP research found that 17% of library patrons said a librarian showed them how to use a computer, and 30% said they learned how to use a computer from a librarian. 67% of computer-using patrons ask librarians for help when they have a problem with a library computer (Heuertz, Gordon, Moore, and Gordon 2002). Not all libraries have been able to respond to patron demand; however, in 38.9% of public libraries provided public training “when patrons request it”; 26.7% provided training “when library staff members have time to provide it”; and 31.2% did not have “sufficient resources staff, or space to provide training to patrons” (Bertot, McClure, and Jaeger 2004, 30).

Julien and Hoffman (2008) found that most (but not all) public library staff viewed provision of online information literacy training as an important role of the contemporary public library, but that public libraries face numerous challenges in providing information literacy instruction to library patrons, including societal challenges, related to the pervasiveness of “infotainment” (33) and the digital divide; institutional challenges within libraries, related to marketing, funding, staffing, or planning; infrastructural problems, related to connectivity and physical space constraints; and pedagogical challenges, related to staff training, “ennui … with mundane technological questions” (34–35), and low levels of patron skill. However, despite the recognition by staff of the importance of information literacy training, and libraries' success in providing Internet access, “people are mostly training themselves, developing their [information literacy] skills through personal experience and seeking help from informal personal sources, such as friends and family, rather than through formal library training” (39).

Internet training in libraries is not unidirectional. A 2003 PACP survey found evidence that “intergenerational teaching of adults by youth” was common: 80% of children ages 12–18 said they had taught adults about computers, and many adults mentioned seeing “tech savvy kids” (Gordon, Moore, Gordon, and Heuertz 2003, 5) teaching other children and adults about computers in the library. Librarians in small libraries also report receiving technology training from their users (Barber and Wallace 2008, 53).

Another approach that librarians have taken to show their value in the Internet age is through applying their traditional skills toward organizing or improving the Internet itself. Valauskas (1995) suggests that librarians could be the “Lewises and Clarks” of the Internet (182), and compares the task of online information organization to the challenge faced by early library cataloguers with the onslaught of new printing technologies in the nineteenth century. Web sites founded by librarians in the mid-1990s, such as the Librarians' Internet Index (available at lii.org) and the Internet Public Library (available at ipl.org) aimed to evaluate and organize online information resources. Librarians have applied collection development skills to “select” online materials for reference (Hogan 1999; Shuman 2001, 145–178). More recently, librarians have used their skills to contribute to Wikipedia (Chillingworth 2008; Pressley and McCallum 2008), and projects such as Hakia (available at hakia.com) and ReferenceExtract (available at referencextract.org) aim to use librarians' expertise to increase the credibility of search engine results (Hadro 2008; Anderson 2008).

Dowlin (1999) links librarians' value to their ability to guide Internet users, stating, “The public will support institutions and professions that help navigate, or even reduce Data Smog” (13). He forecasts that while the Internet will decrease the need for traditional factual reference questions, librarians will increasingly be called upon to train patrons in using technology, develop web sites with local information, serve as “a human interface to the Internet in the library,” and create pathfinders and navigation tools (12).

‘Many librarians have noted that provision of Internet and information literacy training has been one of the biggest changes in terms of how they conceive of their jobs. Hull (2001), in a typical formulation of the statement, writes: “In the 21st century, Librarians as custodians will be less relevant than Librarians as facilitators [sic]” (37). Similarly, a public library reference manager reported in 2008 that the Internet was “changing how we define reference work. We're about teaching people” (Barber and Wallace 2008, 53).

More than the internet

A second approach that libraries have taken in defining their value with regard to the Internet has been to emphasize the multifaceted services that they provide beyond what the Internet can offer. The Internet is only one of many services that libraries provide, and some authors have suggested that libraries should place more emphasis on these other services instead of trying to compete with the Internet.

The Internet may affect the way in which people use libraries, or at least the services they seek there. D'Elia, Jörgensen, Woelfel, and Rodger (2002) found that the Internet did not appear to have an effect on why or how often people use the library, but 33% of U.S. respondents to a 2005 survey conducted by the Online Computer Library Center (OCLC) said they had used libraries less since they started using the Internet (De Rosa Cantrell, Cellentani, Hawk, Jenkins, et al. 2005).

People who use both the Internet and the library use them for different purposes and see them as beneficial in different ways. Respondents to a 2002 survey (D'Elia, Jörgensen, Woelfel, and Rodger 2002) who used both the Internet and the library preferred the Internet to obtain information about consumer products, business, community services, government, and jobs and careers; to do research about their jobs; to do research for personal projects; to read newspapers and magazines; and to do kids' schoolwork. Respondents' preferred the library to obtain information about local history and genealogy and for children's reading. Survey respondents rated the Internet higher for ease of getting there, time to get there, availability (hours of access), range of resources, expectation of finding what is sought, ability to act immediately on the information obtained, currency of the information, fun, enjoyability of browsing, and the ability to work alone (versus being among people at the library). They rated the library higher for ease of use, cost, availability of paper copy (versus digital copy), accuracy of information, helpfulness of librarians, (versus Net helplines), and protection of user privacy . Respondents to the 2005 OCLC survey (De Rosa, Jörgensen, Woelfel, and Rodger 2005) also tended to rate libraries higher than search engines as credible and accurate sources of information, and Janes (2004) argues from a similar standpoint that librarians “are superior to search engines” (9).

Studies such as these demonstrating the increasing role of the Internet in solving everyday information needs have led some authors to question whether information services are as important as libraries' other services. Vavrek (2000) cautions that the Internet is encroaching on the library's role as a provider of reader's advisory services, online databases, and current information, but that still unique to libraries are programming activities for adults and children, outreach, and face-to-face contact. Rodger, D'Elia and Jörgensen (2001) argue that

Information service is only a part of the portfolio of public libraries. Libraries also offer books to lend, programs for adults and children, literacy services, classes, safe havens, and a variety of other services. Perhaps the time has come to reflect on the extent to which a significant part of the information-service function can be performed on the Internet, and to concentrate on other parts of our portfolio … To put all our eggs in the basket of information services for the long term would be unwise (61).

Providers of internet access

The third major approach that libraries have taken in defining their role in relation to the Internet has been to emphasize their position as providers of Internet access. (Of course, libraries' guides-to-the-Internet and more-than-the-Internet roles are in many ways dependent on their provision of Internet access, but the focus in this section is on the effects that are more strictly a result of Internet provision.)

In some cases, direct service provision—versus information provision—has been a new role for libraries. While providing direct services is not entirely new for libraries (libraries have traditionally provided such services as public programs and literacy services [Rodger, D'Elia and Jörgensen 2001]), with Internet provision libraries have begun to play new direct service roles, particularly in regard to e-government services (Burgess 2006; Bertot, Jaeger, Langa, and McClure 2006a; 2006b). Taking on the role of Internet provider has helped libraries maintain their relevance in the eyes of the public—in the words of Putnam, Feldstein, and Cohen (2003), “The computer and Internet developments that some believed would kill public libraries actually provide new reasons for going to them” (47). Internet provision has had noteworthy impacts on public library roles, particularly as a result of increased use of the library and resulting policy changes.

New users

Multiple studies have found substantial increases in library use as a result of libraries' provision of Internet access. Evaluation of the “OnLine at PA Libraries” grant project, which provided Internet access to Pennsylvania libraries beginning in 1996, found that 26% of library Internet users were new library users (Oder 1998). A 2004 Gates Foundation report (Bill and Melinda Gates Foundation 2004) concludes, from interpretation of NCES/IMLS data, that increased library visits are “partially attributable to the availability of computers with Internet access” (4). PACP research (Gordon, Gordon, and Moore 2001) found that libraries participating in the first round of Gates grants witnessed average increases of 36% in patron traffic as a result. Many patrons surveyed by the ALA's Library Funding and Technology Access Study in 2008 (Barber and Wallace 2008) reported that “computers are the primary reason they come to the library” (54). In addition, many adults see libraries as “overwhelmed” with children using computers, particularly during after-school hours (Gordon, Moore, Gordon, and Heuertz 2003, 3).

Despite the increased workload that results, provision of Internet access and the attendant increase in library use benefits public libraries as a tool for demonstrating their relevance. Rural libraries in particular have reported multiple benefits from adopting public access computers, including increased visits, better access to information, enhanced library visibility and reputation, and greater job satisfaction (Garofolo 1995; Heuertz, Gordon, Gordon, and Moore 2003). As one Alabama librarian whose public library was one of the first recipients of the Gates Library Foundation grants reported in 1998, residents felt that they had “moved up” and were no longer “behind the times” (St. Lifer 1998, 5).

Policy changes

As providers of Internet access, libraries have had to evaluate and refine some traditional policies, such as protection of patron privacy, and have had to implement some new ones, such as computer time limits and Internet filtering. Many librarians report that the management of Internet policies poses greater challenges than those posed by the computers themselves (Bertot 2009).

The ALA has taken a stand on protecting patron policy since 1939, when its Code of Ethics for Librarians stated “It is the librarian's obligation to treat as confidential any private information obtained through contact with library patrons” (ALA 2005b, § 1 [“Privacy policies and ALA”]). Provision of Internet access has affected the kinds of records that libraries have the ability to keep, and the ways in which libraries formulate and implement privacy policies. In 2000, an ALA task force concluded that libraries should be concerned with protecting patrons' Internet privacy, including user tracking by remote sites, users giving out personal information online, logs or caches of user activity on library computers, and screen view privacy (ALA 2000).

The USA PATRIOT Act granted more expansive powers to the federal government to collect data from library records, including library circulation records and records of Internet use (ALA 2005a). Some libraries responded by changing retention policies or deleting Internet record and computer sign-up sheets (Orecklin, Ressner, and Thigpen 2003; Reader revolt 2003).

Libraries have also had to craft policies to limit the amount of time an individual patron can spend using public computers. A majority of libraries report insufficient numbers of public access computers to meet patron demand at least some of the time (ALA 2009). In order to manage demand, nearly all libraries implement time limits on Internet use (Heuertz, Gordon, Moore, and Gordon 2002). Libraries have struggled with providing appropriate limits and developing effective policies for access.

For example, the Rochester (MN) Public Library tried several unsuccessful methods for managing Internet sign-up and time limits before implementing an automated reservation system (Betcher 2000), and the San Francisco Public Library (San Francisco, CA) found that the honor system was unsuccessful at ensuring that patrons adhered to the library's 30-minute Internet time limit. Lee (2000) considered this time limit necessary because of patrons' frivolous Internet use:

[S]ince email and electronic chat is free if you have an Internet connection, patrons have grown obsessively dependent on the need to use a library's Internet Access Workstation. Libraries will never have enough workstations to satisfy the public's need for web surfing. Since the “need” is so strong, any plan, electronic or otherwise, that allows the public to police themselves will fail (41).

However, Internet time limits can affect what patrons are able to do on library computers. Tasks such as job seeking, completing homework assignments, and performing complex searches, for example, can take longer than time limits permit (Heuertz, Gordon, Moore, and Gordon 2002; Slone 2007). Time limits may be more stringent than necessary: Heuertz, Gordon, Moore, and Gordon (2002) observed that 40% of public libraries maintain their time limits for computer use even if no one is waiting.

Libraries have struggled to create effective policies regarding patron privacy and computer time limits but certainly the biggest policy headache for libraries with regard to providing Internet access has been the issue of Internet filtering. While a full review of literature related to Internet filtering in public libraries is beyond the scope of this article, the following discussion is a sketch here a few of the ways in which the issue of Internet filtering has affected the roles of librarians.

From the earliest days of public Internet access, libraries have had to make decisions about what level of access to provide, and to balance the interests and demands of their communities with traditional library values and policies related to information access and intellectual freedom. Bastian (1997) casts the decision by libraries about how to limit the level of Internet access in the terms of traditional library roles, arguing that libraries must retain responsibility for their offerings but also protect First Amendment rights.

When Maryland's state library system began offering public Internet access in 1994, gopher-based services were free, but library patrons could pay extra for e-mail, FTP (file transfer protocol), and telnet services (Flagg 1994). In 1995, the Kitsap (WA) Regional Library System offered limited Internet access, which included e-mail and gopher but not FTP (Schuyler 1995). Likewise, Scully (1998) argued that while providing access to “electronic publications,” (i.e., informational content available via the Internet) was a “core service” of Australian public libraries and should be provided for free, other online “non-information” services, such as e-mail, training, and in-depth reference service, were “value-added” and could be provided on a pay basis (§ 7 [“Value adding the Internet]”).

Schneider (1999) claims that libraries originally balked at offering e-mail because it was “interactive” (132). Minkel (2000) justifies the decision by many public libraries not to allow chat or e-mail on public computers because “Libraries don't offer a bank of telephones to make free long distance calls, do they?” (33). In 2002, 23% did not allow patrons to use e-mail, and 62% did not allow patrons to use chat rooms (Heuertz, Gordon, Moore, and Gordon 2002). However, many other online services require e-mail accounts for use, and library patrons may need to use library computers to set up e-mail accounts (Schneider 1999; Barber and Wallace 2008; Mantell 2008).

Schuyler (1995) argued that limiting library users' access to the sites they want was impractical and possibly inappropriate, and in 1998, only 10.7% of libraries with public Internet access had filters on some or all of their computers, although 71.1% had an “acceptable use policy” in place governing use of the public Internet terminals (Bertot and McClure 1998). By 2004, 40% of libraries had some level of filtering on their public Internet computers (Bertot, McClure, and Jaeger 2004).

The dangers of unfiltered Internet access, in particular the ability of children to access pornography, and of any users to access child pornography, have caused concern among some library constituents. Burt (2000) summarizes this view:

The anonymous environment of the public library offers the ideal place to access [a] sea of pornography. Children who want to avoid supervised access to the Internet at home and school, men who do not wish to risk their pornography habits being discovered by their wives and children, transients without any other access to Internet pornography, pedophiles wishing to download illegal child pornography, and sexual perpetrators wishing to expose others to pornography would all be attracted to a public library to obtain free access to the Internet. (2).

Some librarians agree with this view (Burt 2000; McCabe 2003); however, the ALA has downplayed these fears (Burt 2000), and most adults generally view libraries as “safe places” for children to use computers (Gordon, Moore, Gordon, and Heuertz 2003, 5). When Judith Krug, longtime director of the ALA's Office of Intellectual Freedom and anti-filtering advocate, died in 2009, the New York Times explained to its readers that Internet filtering was “an issue taken up by libraries because many people with no computers at home use library computers” (Martin 2009).

Legislation at both the state and federal level has influenced libraries' filtering decisions (Bailey-Hainer 2002; Herb 1999; Michigan library 1999). Of particular consequence has been the 2000 Children's Internet Protection Act (CIPA), which mandated that libraries receiving E-rate funds must “protect against access by adults and minors to visual depictions that are obscene, child pornography, or—with respect to use of computers with Internet access by minors—harmful to minors” (Universal Service Administrative Company [USAC] 2009, § 2 [“Requirements of CIPA”]).

Many libraries have opted to stop applying for the federal program E-rate funds as a result of the filtering requirement, either because of objections to the requirement or simply because E-rate funds do not provide enough to offset the costs of purchasing, maintaining, and training staff in the use of Internet filters (Bertot, McClure, and Jaeger 2004; San José 2009). Because CIPA ties federal aid to Internet filtering, wealthier libraries are better able to refuse funding and allow unfiltered access, leading to greater impacts in low-income areas (Heins, Cho, and Feldman 2006).

Although librarians disagree with the role of “compliance officers”/screen monitors mentioned by the Supreme Court in United States v. American Library Association, the 2003 case that upheld the constitutionality of CIPA, Wyatt (2006) argues that librarians must engage in at least minimal monitoring of patron Internet usage “to ensure that patrons are not engaging in illegal activities, such as downloading child pornography, and that children are not inadvertently exposed to pornography” (78).

Data Analysis

It is clear from the preceding literature review that the Internet in general, and public provision of the Internet specifically, have had wide-ranging impacts on public libraries. In this section, analysis of NCES/IMLS data is used to investigate the effects of having public Internet terminals, and of increasing the number of public Internet terminals, on three measures of library use: 1) number of annual visits, 2) total circulation, and 3) number of reference transactions. Using a random-effects regression model to account for correlations between the multiple observations of each library system, findings show that the impacts of increased Internet workstations on libraries may not be as significant as some have claimed (Oder 1998; Bill and Melinda Gates Foundation 2004; Gordon, Gordon, and Moore 2001; Barber and Wallace 2008). ⁷

Descriptive statistics

As shown in Figure 7, both total visits and total circulation grew significantly during the years analyzed. The average public library system received approximately 29,000 more visits and circulated 38,000 more items in 2006 than in 1998. ⁸ Reference transactions were not significantly different between 1998 and 2006. In 1998, the average visitor checked out 4.8 items and made 0.7 reference transactions per visit. By 2006, these ratios had both decreased significantly: the average visitor checked out 1.4 items and made 0.6 reference transactions per visit.

FIGURE 7 Growth of Circulation, Visits, and Reference Transaction 1998–2006.

Model specification

Figure 8 illustrates the conceptual model used to analyze the effects of public Internet terminals on library outputs (visits, circulation, and reference transactions) while accounting for other factors that also influence those outputs. Library visits are an important measure of volume of library use, which was expected to be positively influenced by an increase in public Internet terminals. Circulation and reference transactions measure how the library is used. These measures were also expected these to be influenced by public Internet terminals. This analysis tests the relationship of these outputs to terminals when other factors are held constant.

FIGURE 8 Model of the Impact of Public Internet Provision on Library Outputs.

The hypothesis was that two additional sets of factors contribute to library outputs: time and other library characteristics. Time is included as a factor because this author assumed that, independently of public Internet terminals and other library characteristics, library outputs change in a predictable way from year to year, due to changing conditions not otherwise included in the model. The group of other library characteristics includes library collection, population of service area, and number of outlets (branch and central locations). These three characteristics can all be reasonably assumed to affect all three outputs: a library with a larger collection will have more circulation transactions than a library with a smaller collection, independently of any effects of Internet terminals, for example.

Random effects model

A random effects regression model was used to estimate the effects of computers on library usage over time. A random effects model is a statistical method that allows for correlated observations within subjects (in this case, library systems). Random effects models are commonly used to analyze longitudinal data, in which repeated measurements are made of the same subjects over time, because they allow for correlation between multiple observations of a single subject (Argue, Johnson, and White 1999; Hedeker 2004; Johnson 1995; Laird and Ware 1982; Peugh and Enders 2005; SPSS 2009).

Random effects models are useful for this analysis for several reasons. First, repeated measurements are made of each library system over multiple years. A standard linear regression would not be appropriate because the multiple observations of a single library system can be assumed to demonstrate some degree of correlation. Second, many library systems do not have observations of all variables in every year (some library systems did not even exist during all nine years of the data). Unlike linear regression models, in which cases with missing data are eliminated from the analysis, random effects models allow all cases with at least one observation to be included, which increases the statistical power of the analysis. The MIXED command in SPSS was used to perform random effects modeling, following the methods discussed in Peugh and Enders (2005) and SPSS (2009).

Dependent variables

The three dependent variables considered, as shown in Figure 8, are total annual visits, total annual circulation, and total annual reference transactions. The natural log of each of these three variables was used, which gives them a more normal distribution and also allows the parameter estimates of the Internet impact and control variables to represent percent changes in the dependent variables.

Internet impact variables

The primary explanatory variable in the model is the number of public Internet workstations. Because of extreme skewness in the distribution of this variable, it was recoded into two different binary variables and used in two series of regressions. The first variable (labeled in Table 5 as Terminals) represents a non-zero number of public Internet terminals, compared with zero terminals. This variable is used in the A series of regressions. The second recoded explanatory variable (labeled in Table 5 as Increase) represents an increase in the number of Internet terminals from the previous year, as compared to a decrease or no change in the number of terminals from the previous year. This variable was used in the B series of regressions. Because this calculation requires data from a previous year, observations from the year 1998 are excluded from this series of models.

TABLE 5 Summary of Findings from Random Effects Models

Model	1A	1B	2A	2B	3A	3B
Dependent Variable	Ln (Visits)	Ln (Visits)	(Ln) Circulation	(Ln) Circulation	(Ln) Reference Transactions	(Ln) Reference Transactions
Estimates of Fixed Effects
Intercept	2.65 *	2.57 *	4.29 *	4.10 *	−0.45 *	−0.52 *
Connected (A series)	0.04 *	—	0.00	—	0.08 *	—
Increase (B series)	—	0.00	—	0.00	—	0.00
Year	0.02 *	0.03 *	0.01 *	0.01 *	−0.01 *	−0.01 *
Ln (Collection)	0.23 *	0.22 *	0.19 *	0.18 *	0.25 *	0.24 *
Ln (Population)	0.60 *	0.63 *	0.52 *	0.55 *	0.69 *	0.72 *
Single Outlet	−0.21 *	−0.19 *	−0.28 *	−0.27 *	−0.27 *	−0.25 *
Estimates of Covariance Parameters
Residual	0.07 *	0.06 *	0.04 *	0.04 *	0.23 *	0.21 *
Intercept + Year
(subject = Library System ID)
UN (1,1) (intercept variance)	0.60 *	0.66 *	0.69 *	0.69 *	1.44 *	1.64 *
UN (2,1) (covariance between slopes and intercepts)	−0.03 *	−0.04 *	−0.02 *	−0.02 *	−0.08 *	−0.11 *
UN (2,2) (slope variance)	0.01 *	0.01 *	0.01 *	0.01 *	0.02 *	0.03 *
Information Criterion
Bayesian Information criterion (BIC)	71353.69	58008.17	40189.21	31190.88	156864.24200	134543.59
Note *p < 0.01.

Control variables

Four additional control variables are included in the model. Year is a measure of the year of the observation, recoded as a continuous variable with integer values between 0 (1998) and 8 (2006), which allows the intercept to be interpreted as the expected value of the outcome variable in 1998 and facilitates analysis of cases that do not have data for all variables at all time points (Peugh and Enders 2005; Hedeker 2004). Year was expected to have a significant positive effect on visits and circulation, with the assumption that these outputs increase over time as a result of other factors not included in the model. The presumed independent effect of time on reference transactions was uncertain, as this outcome did not change significantly over the years included in the data.

The other three variables represent library and service area characteristics that would reasonably be assumed to influence library usage. Collection represents a library system's total expenditures on the library collection, including print, electronic, and other materials. While there are other variables available in the NCES/IMLS dataset that also measure the collection (including number of books and serial volumes), expenditures are the most easily commensurate measure of both size and quality of the collection, both factors which positively influence library use (De Rosa, Cantrell, Cellentani, Hawk, Jenkins, et al. 2005). Population is a measure of the unduplicated population of a library system's service area. ⁹ Collection and Population are both logged, for the same reasons as the outcome variables. Finally, Single Outlet is a binary variable representing a library having only one outlet (either a central or branch location), as compared to library systems with two or more outlets. Measures of population and single/multi-outlet are included to account for the wide variation in the size of library systems in the dataset, with the expectation that population will have a significant positive effect on all three dependent variables, while being in the single outlet category will have a significant negative effect on all dependent variables.

Using the SPSS MIXED command, all independent variables were included as fixed effects. In addition, the intercept term and Year were included as random effects. This allows both intercepts (expected values of the outcome variables in 1998) and slopes (change in the dependent variables over time) to vary randomly by library system.

Findings

Table 2 summarizes the findings from the random effects analyses. In the A series of models, which estimate the effect of providing public Internet workstations on the three outcome variables, all control variables have significant effects on all three dependent variables. As expected, Ln (Collection) and Ln (Population) have significant positive effects on all three dependent variables, suggesting that increases in collection and population correspond to increases in library visits, circulation, and reference transactions. Single Outlet has a significant negative effect on all three outcomes, which suggests that, even adjusting for collection expenditures and population of library service area, single outlet library systems have less use than multi-outlet systems. Year has a significant positive effect on visits and circulation, suggesting an increase in these two dependent variables over time, independently of all the other factors in the model. Interestingly, Year has a significant negative effect on reference transactions, suggesting that with the other factors in the model held constant, reference transactions decreased over the years included in the data.

The A-series explanatory variable, Connected, has a significant positive effect on visits and reference, and no significant effect on circulation. This suggests that, holding collection expenditures, population, and single/multi-outlet status constant, providing public access Internet computers does increase library visits as well as reference transactions, but appears to have no independent effect on circulation. In the B series of models, which replace the explanatory variable Connected with Increase to measure the effect of an increase in public terminals on the three outcome variables, the effects of the four control variables are similar to the A series; all are significant in the expected direction. However, Increase is not significant for any of the three outcomes, which suggests that when other contributing factors are accounted for, an increase in public Internet terminals has no significant effect on library usage. For all three outcomes the Bayesian information criterion, a measure of model fit, is lower for the B series, which suggests that these models are a better fit than the A series.

In all six models, all covariance components are significant, which confirms the appropriateness of including the intercept term and Year as random effects. The significantly negative covariance between slopes and intercepts suggests that, for all models, libraries with lower initial outcomes increased at a greater rate (or, in the case of reference transactions, decreased at a lesser rate).

Discussion

There is little doubt that the Internet—and, specifically, public provision of access to the Internet—has had a dramatic effect on library roles, services, and usage. As McClure and Jaeger (2009) proclaim: “the Internet is more significant to libraries than has been any technology since the book—changing social roles, expectations, and impacts of libraries in the eyes of patrons, communities, and governments” (vi). As Internet provision has become ubiquitous, libraries have added a considerable number of new services to their portfolio. While many of these services are natural extensions of libraries' traditional roles, in many cases they require libraries to refashion, or at least rearticulate, these roles. In many respects the impact of computers on libraries has been positive; many librarians see the addition of public computers as a substantial contribution to their value in their communities. However, many also see the addition of new computers as placing a strain on libraries, because while they place new demands on libraries, they do not alleviate any of the existing ones.

This analysis of the NCES/IMLS data suggests that some of the claims that computers bring hordes of new patrons may be weaker than originally thought. In particular, the insignificance of an increase in public workstations as a contributing factor to any of the three outcome variables suggests that while connecting to the Internet originally may have increased libraries' usage, the addition of new computers does not necessarily have the same effect.

However, there are several important limitations to this analysis. First, quantitative measurement of the outcome variables can say nothing about the qualitative changes that public access computers have brought to libraries. For example, a “visit” by a person who only comes in the building to check out a book would place a much smaller strain on library resources than a visit by a person with no technological skills who comes to the library to learn how to write a resume, find jobs, and apply to them online. Likewise, a simple factual reference question and a lengthy Internet training session may both be recorded in the NCES/IMLS statistics as equivalent “reference transactions.”

Second, the simple measures of Internet provision used in this analysis ignore the wide varieties of quality of Internet access in libraries. Adding one new Internet terminal to an already taxed-to-the-limit narrowband network would presumably have a much smaller impact than an upgrade of a library's infrastructure from dialup to broadband without increasing the number of terminals. Additionally, while the inclusion of time as a random effect allows the model to control for changes in the effects of adding new computers in 1999 versus 2006, it does not necessarily capture the changes in user demand or needs for types of Internet services or connection speeds during the years of the data.

Furthermore, the insignificance of new terminals on library usage should also be interpreted to mean that adding new public computers does not decrease the use of other library services. Additional computers may place all the qualitative strains on libraries mentioned above, but they do not appear to lessen demand for libraries' other services; at least as measured by visits, circulation, and reference transactions.

CONCLUSION

As Barber and Wallace (2008) explain, “There has been a not-so-quiet revolution in libraries, but it is far from over” (55). Technology, the Internet, and the digital divide are constantly changing and advancing, for better or worse. Libraries have been remarkably quick to identify community needs and respond with appropriate technologies, often in the face of severe resource constraints. There is no sign of this stopping. New and ongoing issues, such as uneven broadband deployment, ever-increasing demand for bandwidth, and advances in mobile Internet access, coupled with steady or shrinking technology budgets and persistent myths about library irrelevance in relation to new technologies, promise to present libraries with continuing challenges. This article concludes with a summary of key findings, questions, and recommendations for libraries to consider in planning to meet these challenges.

Key Findings

• Public libraries, supported by both governmental and private funding initiatives, have played a key role in providing free public access to computers and the Internet for the digitally disadvantaged. Free public access to computers and the Internet have helped to bridge the digital divide, but that gap still exists. The digital divide is a moving target, and new divides open up with new technologies. Libraries help bridge some inequities more than others, and substantial disparities exist among library systems.

• Access alone is not enough to remedy technological inequities. Public access is not of the same quality as home access; access does not ensure effective use; and broader social and economic issues hamper the ability of access to remedy social ills. Libraries currently go beyond access by providing training, outreach, and content development, but many do not have the resources to provide these services at a high level. For libraries to successfully address technological inequities, they will need to continue to look beyond access.

• Internet assistance and training provided by staff are very important: they are a natural extension of libraries' information literacy role, they are how libraries stand out from other providers of Internet access, and they will continue to be relevant as technology changes. There is much that libraries can do to improve their training role, including providing better training for staff, more time for staff to help computer users, and more outreach to library patrons about the training possibilities available. Provision of training is resource-intensive, however, and improving training services requires a greater investment than is currently available in many libraries.

• The Internet is used in libraries in a wide variety of ways. Not all uses are seen as equally valuable by librarians or the public, but these value judgments are not always grounded in a coherent vision. E-mail, for example, is often seen as a less important use of library computers than job seeking; however, employers often require applicants to have e-mail accounts and may correspond with applicants by e-mail. Library policies, such as privacy policies, time limits, and filtering, affect patrons' use of library computers.

• As mobile devices become cheaper and more widely used, more people have access to the Internet, but increased demand for online services makes the cost to those not connected much higher. Most libraries have become providers of wireless Internet access, but this places an increasing strain on library bandwidth and infrastructure. The full impact of mobile computing on libraries is unclear.

• Computers and Internet access have substantially impacted library roles and services. However, Internet provision is still only one of many roles and services that libraries perform, and Internet services may both complement other library services and compete with other services for resources and attention.

Recommendations

This report concludes with two recommendations for librarians, library administrators, researchers, and policymakers. First, improve the understanding, for individual libraries and the library profession in general, of the value of access to computers and the Internet. And second, look beyond access to a broader range of technological services and roles. Providing Internet access has had an enormous impact on public libraries, and libraries have played a key role in narrowing digital divides. It is important for librarians to maintain and improve their provision of public Internet access, and to convince their communities and policymakers of the importance of this role. However, it is just as important for libraries to think beyond the provision of access if they are to continue to play this role. Access alone is not enough to resolve digital inequities, and the value of access to Internet terminals, relative to other offerings such as wireless access and Internet training services, may become less clear if mobile computing and universal broadband provision continue to increase.

Understanding the value of access

Providing free access to computers and the Internet is a central component of most libraries' missions and activities, and clearly provides great value to the public and to libraries themselves. However, it also provides great challenges to library resources. While external funding initiatives have supported libraries' public technology investments, overall library funding has not kept pace with the growth of patron demands or library service provision. As with other public library services, demand for technology access increases in economic downturns, just as libraries find their funding the least stable. In addition, lingering perceptions of public library irrelevance with the ascendancy of the Internet, while utterly unsupported by any quantitative or qualitative evidence, threaten libraries' ability to maintain their public technology access services in addition to their other services. It is important for libraries to develop a better understanding of the impacts and value of providing access to computers and the Internet, as compared to, but also in interaction with, other library services.

Provision of Internet access is not a stand-alone service. It integrates and interacts with other library services, including reference service (for example, when computer users ask for help in using computers or finding information online), circulation (when computer users supplement their computer use with library materials), and instruction and training (when library staff provide computer and Internet training). Even patrons who come to the library for the sole purpose of using a computer may take advantage of other services during their visit, and vice versa. As a result, public computer services are affected not just by technology funding, but also by funding for staffing, collections, and library open hours. Libraries should take a holistic view of all the services they offer to determine how Internet provision fits in to the big picture.

Understanding and communicating the value of access to computers and the Internet will be useful for both decision making and advocacy. It will help libraries plan for new and continuing services, and also help communicate these plans, and the role of public technology access with them, to funders, policymakers, and their communities.

Moving beyond access

Libraries have done a remarkable job of responding to the access needs of the disadvantaged, but free access to computers is only part of the story. Libraries should pay attention to ways in which they can leverage their provision of technology access with their other services and roles to move beyond access and address broader technological inequalities. Some libraries already accomplish this goal by providing computer and information literacy training, outreach, and content development, and they will be even more effective at fostering social equality if they continue to focus on services beyond public access.

Consider, for example, the argument that public access is inferior to home access. Home computer use can give some users a richer, more varied online experience than they can typically get through public use, and home computer users may benefit more from their computer use than do public users. But some libraries have offered circulating computers for use outside the library, which could be one way to address this challenge. Or take the troubling observation that people who are disadvantaged in other ways tend to have fewer public resources, such as library computers, available to them as a result of weaker political clout and poor advocacy skills. Libraries in disadvantaged communities could respond to these problems by being more vocal advocates for their communities and by working with partners or developing training programs to help build community capacity and advocacy skills.

There will of course always be social ills that libraries cannot remedy, at least not directly (for example, “giving poor people more money,” [Gordon, Moore, and Gordon 2003, 18]). However, by taking a wider view and thinking of solutions that go beyond access provision, libraries can make a greater impact in developing healthy communities. It may in fact be non-technical roles of public libraries—e.g., literacy, training, and community building—that give public technology its greatest value. It is a mistake to think that libraries are relevant only because they provide computer and Internet access. Rather, technology is imbued with greater power to help individuals and communities precisely because it is provided in a public library.

This article is derived from the author's degree project for a Master of Public Administration degree at the University of Washington. In that venture, the author acknowledges the assistance of the following individuals: Rachel Kleit (Associate Professor, Evans School of Public Affairs, University of Washington) provided significant support and review of the degree project Samantha Becker (US Impact Study Project Manager, Information School, University of Washington) collaborated on the data preparation; and Anita Rocha (Statistical Programmer and Data Manager, Center for Studies in Demography and Ecology, University of Washington) provided support and advice with statistical analysis and random-effects modeling.

Notes

1. The Public Libraries Survey (PLS) is an annual survey, administered to the United States public library systems by the National Center for Education Statistics (NCES) from 1989–2005 and by the Institute of Museum and Library Services (IMLS) from 2006 to date. PLS data files are available from IMLS for the years 1992–2006 (IMLS 1998–2006). PLS began collecting data on Internet terminals in 1997, but a substantial number of libraries did not provide this data until 1999.

2. The data were originally prepared by Samantha Becker and the author for use in a previous report.

3. An earlier version of this article was written as a report to the United States Impact Study Project Team (Kinney 2009).

4. Although a substantial number of libraries do not allow patrons to use email (Schneider 1999; Minkel 2000; Heuertz, Gordon, Moore, and Gordon 2002).

5. For more on the impacts of pornography access in public libraries, see “Filtering” section of this article.

6. This article uses 10% as the dividing line for poverty level, rather than the median of 11.8%, because this is the level that the Gates Foundation used in determining eligibility for its library technology grants. Analysis based on an 11.8% dividing line would look very similar.

7. See the “New Users” section of this article for a summary of observations that Internet terminals have led to an increase in library use.

8. Statistically significant at a p < 0.05 level.

9. NCES/IMLS calculates the unduplicated population of a library's legal service area by prorating the library's population of legal service area to the state's total population of legal service area. This calculation is made to account for overlapping legal service areas, for example, a county library and city library that serve the same population (USDE 2003).

United States v. American Library Association, Inc., 539 U.S. 194 (2003)