Citation
America connecting

Material Information

Title:
America connecting internet utilization as a predictor of community participation
Creator:
Phipps, Antony Allan
Place of Publication:
Denver, Colo.
Publisher:
University of Colorado Denver
Publication Date:
Language:
English
Physical Description:
209 leaves : ; 28 cm

Thesis/Dissertation Information

Degree:
Doctorate ( Doctor of Philosophy)
Degree Grantor:
University of Colorado Denver
Degree Divisions:
College of Architecture and Planning, CU Denver
Degree Disciplines:
Design and Planning
Committee Chair:
Van Vliet, Willem
Committee Members:
Sancar, Fahriye
Studer, Raymond
Muller, Brian
Staeheli, Lynn A.

Subjects

Subjects / Keywords:
Internet -- Political aspects -- United States ( lcsh )
Political participation -- Computer network resources -- United States ( lcsh )
Internet -- Political aspects ( fast )
Political participation -- Computer network resources ( fast )
Politics and government -- Computer network resources ( fast )
Politics and government -- Computer network resources -- United States ( lcsh )
United States ( fast )
Genre:
bibliography ( marcgt )
theses ( marcgt )
non-fiction ( marcgt )

Notes

Bibliography:
Includes bibliographical references (leaves 197-209).
General Note:
College of Architecture and Planning
Statement of Responsibility:
by Antony Allan Phipps.

Record Information

Source Institution:
|University of Colorado Denver
Holding Location:
|Auraria Library
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
51821078 ( OCLC )
ocm51821078
Classification:
LD1190.A72 2002d .P44 ( lcc )

Full Text
AMERICA CONNECTING: INTERNET UTILIZATION
AS A PREDICTOR OF COMMUNITY PARTICIPATION
by
Antony Allan Phipps
B.A., Williams College, 1966
M.S., Massachusetts Institute of Technology, 1973
A thesis submitted to the
University of Colorado at Denver
in partial fulfillment
of the requirements for the degree of
Doctor of Philosophy
Design and Planning
2002


2002 by Antony Allan Phipps
All rights reserved.


This thesis for the Doctor of Philosophy
degree by
Antony Allan Phipps
has been approved
by
Willem Van Vliet
b

7
Z
Date


Phipps, Antony Allan (Ph.D., Design and Planning)
America Connecting: Internet Utilization as a Predictor of Community Participation
Thesis directed by Professor Willem Van Vliet
ABSTRACT
Will Americans increasing use of the Internet undermine or enhance their
involvement in community affairs? The first part of this thesis finds a strong, positive
relationship between Internet use and four composite measures of community
participation. Controlling for 22 attributes of a large nation-wide sample of
households in the Social Capital Community Benchmark Survey (SCCBS),
.multivariate regression analyses show that Internet use is a positive predictor of
political involvement, participation in civic associations, social interaction, and
charitable activities. Although a direct causal connection between Internet use and
community participation cannot be inferred, the analyses provide strong evidence that
the Internet is an enabling technology likely to strengthen, rather than weaken,
individual levels of community participation.
Using multivariate analysis of residuals, the second part of this dissertation
examines whether mean levels of Internet use at the community level predict higher-
or lower-than-expected levels of participation in 33 SCCBS communities. Social
interaction is the only form of community participation with a positive relationship to
mean levels of Internet use. Internet use does not predict higher- or lower-than-
expected levels of political, civic or charitable involvement when factors of
population density, homeownership, socio-economic status, race and segregation
levels are taken into account. Community participation is higher than expected in
communities with higher densities, more highly educated adult populations, higher
percentages of homeownership and stronger planning websites. It is lower than
expected in communities with higher median incomes, higher percentages of
professional workers and higher levels of segregation. With the exception of
charitable activities, communities with high segregation levelsparticularly white-
black and white-Hispanicare found to have much lower-than-expected community
participation scores.
This abstract accurately represents the content of the candidates thesis. I recommend
its publication.
Signed
Willem Van Vliet
IV


ACKNOWLEDGEMENT
The author is indebted to The Denver Foundation, Rose Community Foundation, The
Piton Foundation and The Community Foundation Serving Boulder County for
providing the opportunity to analyze the results of the Social Capital Community
Benchmark Survey for the communities of Denver and Boulder County. In addition,
the author gratefully acknowledges the insights and critical reviews of earlier
manuscripts provided by Professors Willem Van Vliet and Fahriye Sancar of the
College of Architecture and Planning of the University of Colorado at Denver, and
Professor Lynn Staeheli of the Department of Geography of the University of
Colorado in Boulder


DEDICATION
I dedicate this dissertation to my fiancee, Betsy, whose unflagging support kept me
going through thick and thin; to my son and daughter, Tyson and Verena, who first
encouraged me to undertake the journey in the first place; and to those many others
who kept me sane and alive in the process. Finally, I dedicate this work to one of the
wisest men I have ever known, who said...
If you want to shrink something,
you must first allow it to expand.
If you want to get rid of something,
you must first allow it to flourish.
If you want to take something,
you must first allow it to be given.
This is called the subtle perception
of the way things are.
The soft overcomes the hard.
The slow overcomes the fast.
Let your workings remain a mystery.
Just show people the results.
Lao-Tzu, Tao Te Citing
5th century B.C.E.


CONTENTS
Figures..................................................................... ix
Tables...................................................................... x
Chapter
1. Internet Use and Community Participation............................ 1
1.1 Community Participation and Planning.................................. 1
1.2 Growth Of The Internet.................................................4
1.3 The Community, Social Capital, Internet Nexus..........................5
1.4 Previous Research..................................................... 8
1.5 Plan For The Dissertation.............................................13
2. Research Question and Methods........................................ 17
2.1 Questions for Empirical Inquiry.......................................18
2.2 Sources of Data...................................................... 22
2.2.1 The Social Capital Community Benchmark Survey........................ 22
2.2.2 Secondary Sources.................................................... 25
2.3 Dependent and Independent Variables.................................. 26
2.3.1 Dependent Variables: Community Participation..........................26
2.3.2 Measuring Internet Use.............................................. 32
2.3.3 Control Variables.....................................................35
2.4 Analytical Strategy...................................................39
2.4.1 Individual-Level Analysis Model.......................................40
2.4.2 Community-Level Analysis Model........................................41
2.4.3 Model Assumptions.....................................................42
2.5 Limitations of the Analyses...........................................43
3. Predicting Individual-Level Participation.............................46
vu


3.1 OLS Regression Procedures and Assumptions.............................46
3.2 Regression Results for Community Participation Summary................48
3.3 Regression Results for Different Forms of Participation...............51
3.4 Predicting Internet Use...............................................56
3.5 Summary of Individual-Level Results...................................61
4. Predicting Community-Level Participation;.............................66
4.1 Rationale for Community-Level Analyses................................68
4.2 Application of Predictive Models to SCCBS Sites.......................70
4.2.1 Selection of SCCBS Communities for Analysis...........................73
4.2.2 Calculating Community Participation Residuals.........................79
4.2.3 Community Comparisons of Internet Use.................................85
4.2.4 Community Comparisons of Context Factors..............................89
4.2.5 Limitations of the Analysis...........................................96
4.3 Community Level Results Effects of Internet Utilization.............97
4.3.1 Community Specific Regressions...................................... 100
4.3.2 Pooled Regression with Site Dummies and Interaction Terms........... 100
4.4 Measuring the Effects of Community Context on Participation..........110
4.5 Summary and Interpretation of Results................................118
5. Implications For Planning............................................125
5.1 The Internet and Community Participation.............................126
5.2 The Internet as a Vehicle for Supporting Individual Participation....129
5.3 Barriers and Opportunities at the Community Level....................131
5.4 Contextualizing the SCCBS Data...................................... 133
5.5 The World Wide Web is Not Enough.....................................136
Appendix A: Predicting Community-Level Internet Use........................ 138
Appendix B: Supplementary Tables And Figures...............................144
Appendix C: Social Capital Community Benchmark Survey......................156
Bibliography...............................................................197
viii


FIGURES
Figure
1 Plan for the Dissertation..................................................14
2 Mean Participation Scores for Internet Users and Non-Users................ 35
3 Community Participation Summary Residuals................................. 83
4 Mean Standardized Internet Use by Community............................... 89
5 Community Participation Residuals by Internet Use......................... 98
6 Community Participation Residuals by Internet Use (Without
SFO, SEA and BLD)................f...................................... 99
B-l Mean Politics Residuals by Community......................................147
B-2 Mean Civic Residuals by Community.........................................148
B-3 Mean Social Residuals by Community........................................149
B-4 Mean Charitable Residuals by Community....................................150
IX


TABLES
Table
2-1 Communities in the Social Capital Community Benchmark Survey...........24
2-2 Community Participation Indices and Components.......................... 27
2-3 Correlations of Internet Utilization Index with Component Variables.... 31
2-4 Statistics for Community Participation Dependent Variables.............. 31
2-5 Correlations of Internet Utilization with Component Items............... 34
2- 6: Covariates in the Regression Analyses................................... 37
3- 1 Regression Results for Community Participation Summary.................. 49
3-2 Regression Results for Four Components of Community Participation....... 52
3- 3 Regression Results for Internet Use.................................... 59
4- 1 Social Capital Community Benchmark Survey Communities................... 74
4-2 Mean Percent Urban and Percent of Sample in Rural Areas Analysis
and Non-Analysis Sample Communities..................................... 76
4-3 National and Community Sample Characteristics........................... 77
4-4 Comparison of Participation and Internet Scores for National and
Community Samples....................................................... 78
4-5 Pooled Regression Results for Four Forms of Community Participation
Plus Summary............................................................ 81
4-6 Mean Community Participation Residuals by Analysis Community.......... 82
x


4-7 Correlations of Mean Community Participation Residuals................ 85
4-8 Components of Internet Use (Community Analysis Sample)................ 86
4-9 Standardized Internet Scores by Community............................. 87
4-10 Characteristics of Analysis Sample Communities........................ 95
4-11 Regression Summary for Community Participation Residuals by
Internet Use........................................................ 97
4-12 Summary Statistics for 33 Regressions by Community....................101
4-13 Pooled Regression of Internet, Site and Interaction Terms on
Community Participation Summary Residuals...........................103
4-14 Internet Use, Site Dummies and Interaction Terms for Politics, Civic,
Social and Charitable Residuals.....................................105
4-15 Comparison of Standardized Internet Use for Individual and
Community Samples...................................................109
4-16 Community-Level Predictors of Summary Participation Residuals.........113
4-17 Community-Level Predictors of Political, Civic, Social and Charitable
Participation.......................................................115
A-l Community-Level Predictors of Aggregate Internet Use................. 140
B-l Politics Factor Score Component Loadings............................. 145
B-2 Civic Engagement Factor Score Component Loadings..................... 145
B-3 Social Interaction Factor Score Component Loadings................... 146
B-4 Charitable Activities Factor Score Component Loadings................ 146
B-5 Local Civic Factor Score Component Loadings...........................146
B-6 Pooled Regression of Internet Use and Site Interactions on
Electoral Politics Residual.........................................151
xi


B-7 Pooled Regression of Internet Use and Site Interactions on
Civic Engagement Residual.............................................. 152
B-8 Pooled Regression of Internet Use and Site Interactions on
Social Interaction Residual........................................... 153
B-9 Pooled Regression of Internet Use and Site Interactions on
Charitable Activities Residual......................................... 154
B-10 Summary Scores for Community Planning Websites..........................155
xi i


CHAPTER 1
INTERNET USE AND COMMUNITY PARTICIPATION
This dissertation addresses the question of the relationship of Internet
utilization to levels of community participation in the United States. Using primary
data from the Social Capital Community Benchmark Survey (SCCBS) and other
secondary sources of information, the research examines this question at both the
individual and community levels. This chapter begins with an overview of the
relevance of community participation to urban planning issues, and in Section 1.2
summarizes the conflicting interpretations of the economic, political and social
consequences of increasing Internet utilization. Section 1.3 establishes the nexus
between community participation, social capital and Internet utilization and Section
1.4 provides a summary of relevant previous research. The chapter concludes by
laying out the research framework for the remainder of the dissertation.
1.1 Community Participation and Planning
Since the mid-1960s community participation has been an increasingly
important element of both planning theory and practice because it is a measure of the
degree to which local stakeholders have access to the planning process and have an
1


opportunity to influence the planning outcomes that affect their physical, social and
economic environment (Jacobs 1961; Davidoff 1965; Amstein 1969; Forester 1989;
Healy 1992; Ploger 2001). The forms and levels of citizen involvement with
community issues define the multiple (and often conflicting) identities of the public
interestpresumably a central client of planning practice (Staeheli 1997, Beauregard
2001). In addition, high levels of civic engagement and trust are positively related to
homeownership (DiPasquale and Glaeser 1999), nonprofit housing (Keyes et al.
1996), housing revitalization efforts (Saegert and Winkel 1998), neighborhood
stability (Temkin and Rohe 1998), local government performance (Rice 2001),
community organizing (Gittell and Vidal 1998), and local community and economic
development programs (Servon 1998; Rutnik 1999).
All of these subjects are central to community planning programs around the
country. In the absence of direct information about public participation in local
planning dialogs that is comparable across many sites, the more general measures of
community participation examined here (political, civic, social and charitable) are
taken as proxies for public dialog in local planning issues. Stakeholders in
communities with high levels of participation cannot always address planning issues
more successfully that those in low participation communities (Hibbard and Lurie
2000; Bollens 2002). Public dialogue about divisive issues such as growth
management or environmental protection may secure high levels of participation
without yielding consensus. This may be especially true with regard to issues of racial
2


and ethnic segregation (Briggs 2002). However, it seems likely that high levels of
participation do provide an enabling context for successful planning (Innes 1996;
Innes and Booher 1999), and that successful planning in turn helps to build
community participation as an aspect of social capital (Reardon 1999; Allen 2001).
Community participation is a building block of social capital, defined to
include as well norms of reciprocity, honesty and trust among the individuals
comprising a community and in civil society generally (Allen 2001; Bimber 2000;
Brehm and Rahn 1997; Edwards and Foley 1997; Fukuyama 1999; Putnam 2000;
Stone 2001; Uslaner 2000; Wellman, Haase et al. 2001) In Bowling Alone Robert D.
Putnam (2000) argues that over the last three decades Americas stock of social
capital has eroded, and demonstrates this trend with reference to declining
participation in political, civic, social and charitable organizations and activities. In
examining Americans increasing use of the Internet as a possible counter-trend to
this decline, he offers a preliminary conclusion that Very few things can yet be said
with any confidence about the connection between social capital and Internet
technology (Putnam 2000, p. 170). His analyses, therefore, leave open the question
of whether use of the Internet may help to strengthen Americans political, civic and
social participation in their communities and thereby to build social capital.
3


1.2 Growth of the Internet
American households are increasingly connected to the Internet. The
Nielsen/Netratings service estimates that over 169 million Americans had access to
the Net in July 2001 (Nielsen 2001). With 2 billion pages of information and
revenues estimated at $850 billion, the Internet doubled in size in 2000 and is
expected to interconnect over 1 billion users worldwide by 2005 (USIC 2000).
Moreover, the Internet provides a diverse array of services ranging from email to
information gathering, entertainment, decision support and economic transactions
(Howard et al. 2001). Given its rapid growth, economic power and amorphous
breadth of service, there is little question that the Net has the capacity to affect many
important aspects of our lives. Numerous researchers have explored the political,
social and economic consequences of the Internet. For example, Berman and
Weitzner (1997) underscore the emerging importance of the Internet as a platform
for democracy that encourages political participation, while Bimber (Bimber 2000)
notes the possibility that the information revolution will contribute to greater
fragmentation and pluralism in the structure of civic engagement. DiMaggio et al.
(2001) cite the expanding literature suggesting that the Internet enhances social ties
defined in many ways, often by reinforcing existing behavior patterns. However,
other research indicates certain negative social impacts among heavy users of the Net
such as reduced sociability and diminished interpersonal contacts and
communications (Nie 2001). While many researchers anticipate a beneficial
4


restructuring of economic activity (Armstrong and Hagel 1996), others warn of the
negative polarizing effects of the digital divide that appear likely to deprive specific
groups of full access to the economic opportunities of the information society
(Castells 1996, Graham 2000, Thornton 2001). Within these conflicting
interpretations a key unexplained question is whether increasing Internet usage will
undermine or enhance Americans participation in their communities. This
dissertation proposes to address this question by examining the results of the Social
Capital Community Benchmark Survey, conducted in the latter half of 2000.
1.3 The Community, Social Capital, Internet Nexus
The question of individuals participation in their communities has fueled
much sociological debate and theoretical development over the last 150 years (Paxton
1999). Wellman (1979) provides a useful framework for this debate in his dialectic
about whether community has been lost, saved or liberated. The community lost
argument begins with Tonnies concern about an implicit social transition from
Gemeinschaft (community defined by internal bonds of language, place, kinship,
custom and faith) to Gesellschaft (the larger, external and predominantly urban
society with its latent disintegrating conflicts held in check through the formal
mechanisms of convention, law, politics and public debate) (Tonnies 1955). The
Community Lost argument echoes Durkheims concern about loss of the
conscience collective and the negative social effects of an increasing division of
5


labor (Durkheim 1984); Simmels critique of individuality and differentiation (Wolff
1950); Wirths concern for urbanizations displacement of family, kinship and
neighborhood bonds (Wirth 1964); and Arendts critique of mass society and urban
anonymity (Arendt 1958). Thus, Castells (1996) structural schizophrenia between
the space of flows and the space of place, and Putnams (1995, 2000) declining-
social-capital thesis appear as modem versions of a much older tale of societys fall
from communal grace (Reitzes and Reitzes 1992).
Modem American versions of the community saved argument originate in
the nineteenth century. Alexis de Tocqueville found the willingness of Americans to
join community associations in the pursuit of self-interest rightly understood to be
both a source and an effect of democracy, and a foundational attribute of civil society
(Tocqueville 1969; Whittington 1998). More recent examinations of community
vitality and survival by Jacobs (1961), Gans (1962), Liebow (1967), Suttles (1972),
Stack (1975) and others celebrate the placed-based solidary networks that provide
assistance to individuals and mediate the exigencies of modern urban life. As
Wellman and others point out, however, these analyses are weakened by their
assumption that a communitys supportive social structures are necessarily tied to
geographically defined places like urban neighborhoods or small towns (Fischer et
all977; Wellman 1999).
As a third alternative, the community liberated perspective proposes that
community is not spatially delimited. Melvin Webber (1963) was one of the first to
6


recognize the importance of community without propinquity and to argue that
widely dispersed (non-local) communities of interest overlap geographically defined
communities of place. The essential meaning of community lies in the supportive
networks that link individuals and groups to each other, irrespective of place
(Wellman and Leighton 1979). The conceptual importance of non-place
communities, defined by networks and not by places or groups, gathered theoretical
momentum with Granovetters (1973) influential study of the strength of weak ties.
His research shows overlapping, informal networks of acquaintances as playing
important roles in the diffusion of influence and information, the creation of
opportunities for mobility, and the strength of community organization. The
importance of personal networks in generating trust and delimiting malfeasance was
further elaborated in Granovetters (1985) notion of the embeddedness of economic
actions within pre-existing social networks. The importance of social networks for
enhancing economic opportunity among disadvantaged minority youth has been
recently demonstrated in Briggs (1998) study of the impacts of housing mobility on
the social capital of black and Latino adolescents.
Today most social scientists will accept four basic notions about community:
(a) that it frequently transcends place, (b) that it reflects shared interests among
interacting individuals or groups, (c) that it is evidenced by a commitment to
common goals or purposes, and (d) that participation in community may take many
forms including political, organizational, social and charitable. Within this framework
7


what is uncertain is whether the Internet tends to support or weaken these forms of
community participation.
1.4 Previous Research
The Internet reflects a complex landscape of applications, purposes, and
users (Haythomthwaite 2001, p. 363). It is therefore not surprising that analyses of
Internet impacts on society reflect a broad diversity of interpretations (DiMaggio et
al. 2001). One debate centers on whether Internet use supports or detracts from
community participation. Many scholars regard the Internet as having a positive
effect on peoples involvement with their communities.1 James Slevin, for example,
interprets the Internet as a community-building modality of cultural transmission that
opens up new opportunities for dialogue and deliberation, empowers people to make
things happen rather than have things happen to them, and facilitates new forms of
solidarity and cooperation (Slevin 2000, p. 47). Other proponents of the Net such as
Wellman (1999), Negroponte (1995), and Dyson et al. (1998) argue that the Internet
has opened up a new world for a globally interactive society with increasing access to
information and opportunities that expand social engagement, and personal choice.
The Internet is also seen as a platform for strengthening democracy, civic
participation and civil society due to (a) its open structure, high accessibility and low
1 The issue of whether community is defined geographically or in terms of person-to-person networks
regardless of place complicates the question of the Internets influence on community participation.
See Wellman and Leighton, 1979.
8


cost, (b) its bi-directional nature that permits two-way communication with leaders
(e.g. via town meetings), and (c) its non-hierarchical structure that allows the
development of grass-roots political efforts (Berman 1997). Many observers see
citizen participation in the electoral process as a building block of social capital that is
significantly advanced within the expanding regime of teledemocracy (Dahl 1989;
Etzioni 1993; Grossman 1995; Browning 1996; Bimber 1998). Examples of recent
empirical studies that support these positive interpretations include the following:
Based on the National Geographic Society Website surveys (but using
a self-selected sample), Wellman, Haase et al. (2001) find that online
interaction supplements face-to-face and telephone communication,
and that heavy Internet use is associated with increased participation in
voluntary organizations and politics.
Based on a survey of 342 community members of a mid-sized city in
Southern California, Blanchard and Horan (2000) suggest, inter alia,
that social interaction at the local level is enhanced by virtual
interaction in cyberspaceparticularly if virtual communities develop
around physically based communities.
Kavanaughs follow-up study of the Blacksburg Electronic Village
concludes that the Internet does appear to facilitate increases in
community involvement, but that growth occurs among people who
are already poised to be active in the community (Kavanaugh
1999).
An analysis of telephone surveys by the Pew Internet and American
Life Project finds that Internet use enhances social activities,
particularly among those with greater online experience (Howard,
Rainie, and Jones 2001).
Shah et al. (2001) find that Internet use is positively related to civic
engagement among Generation-X users.
9


In another study for the Pew Internet and American Life Project,
Larsen and Rainie (2002, p. 1) emphasize the powerful positive effects
of email use by local governments, particularly among officials who
learn about constituents opinions and activities, and among local
stakeholders who are being heard and recognized thanks to email.
On the other side, however, a number of analyses emphasize the perceived
negative effect of the Internet on two aspects communitysocial interaction and
differences in access among different social and economic groups (the digital divide).
With respect to social interaction the HomeNet Project at Carnegie Mellon University
was one of the first to find small but significant declines in social interaction and
psychological well being among 169 households in Pittsburgh after the Internet was
introduced into their homes (Kraut et al. 1998). Many of these initial findings of
negative social effects, however, appear to have been reversed in the second wave of
interviews with panel participants (Kraut et al. 2001). Both Locke (1998) and
Putnam (2000) suggest that computer-mediated communications may be responsible
for at least some of the observed drops in associational membership across a wide
range of civic organizations. Likewise, an Internet-based survey at the Stanford
Institute for the Quantitative Study of Society shows that increasing Internet use
results in decreasing social engagement, increasing work hours and decreasing use of
traditional media (Nie and Erbring 2000). Similar findings are reported in the
NPR/Kaiser/Kennedy School of Government survey with 58 percent of respondents
10


saying that computers have led people to spend less time with their families and
friends (National Public Radio 2000).
With respect to the digital divide, many scholars lament the perceived
fragmentation and polarization of urban society that may be occasioned by the digital
divide (Schon, Sanyal, and Mitchell 1999; Graham 2000). If the resources and
opportunities afforded by the digital revolution are not shared equally among all
segments of the population, it seems clear that disadvantaged groups will be unable to
participate successfully in the new global economy. The split between the
technological haves and have-nots will likely generate the "truly fundamental
social cleavages of the informational age (Castells 1996, p.346). Research in the late
1990s indeed shows less participation by minorities (primarily Afro-Americans and
Hispanics), low-income persons, elderly, rural and single-parent households (NTIA
1995; 1998; 1999; Lebo 2000; Loges and Kim 2001). Geographical comparisons of
Internet access and use in different localities reassert the importance of place in
measures of technological deprivation. Low-income urban populations and rural
households show significantly lower levels of Internet utilization (Servon and Nelson
2001).
However, as access to the Internet has grown to about 107 million adults,
earlier estimations of the distributional side effects of the World Wide Wedge
appear to have been overstated. The most recent NTIA surveys (2000) show a
diminishing gap in Internet access among income, gender, age and racial groups from
11


1998 to 2000. DiMaggio and Hargittai (2001) argue that, as Internet penetration
increases toward universal access in the United States, concern for minimizing the
digital divide (measured by dichotomous measures of access and use) should be .
replaced by concern for reducing digital inequality. This requires measures of
inequality among Internet users with respect to equipment, autonomy of use,
technical skill, social support and purpose). Similarly, Jung et al. (Jung, Qiu, and Kim
2001) propose a more nuanced Internet connectedness index to account for multiple
dimensions of inequality in Internet access and utilization.
Definitions of community participation are fluid and range from involvement
in community organizations to informal social interaction to political activity to
giving and volunteering. As a result, the construct validity of the dependent variable
(i.e. community participation) is not always clearly established (Brehm 1997; Shah
2001). The usefulness of some of the research is also limited by small sample sizes,
selection bias from self-selected samples (e.g. the National Geographic Survey) and
the absence of multivariate controls for demographic effects clearly associated with
community participation. These deficiencies limit the generalizability of results and
undermine clarification of the relative contribution of household characteristics,
attitudes, location and Internet use to community participation.
12


1.5 Plan for the Dissertation
This research explores the question of the relationship of Internet utilization to
community participation at two levelsfirst at the individual level, and then at the
community level. Chapters 2 and 3 address the former, and Chapter 4 addresses the
latter. An overall schema for the dissertation is provided in Figure 1 on the next page.
Chapter 2, Research Question and Methods, Begins with an explication of the
research question to be addressed and outlines the main propositions to be explored in
the analysis. Section 2.2 then describes the Social Capital Community Benchmark
Survey, the primary data source for the research, and identifies the other secondary
data sources used in the analysis. This section includes a discussion of both the
national sample used for the individual-level analyses as well as the community
sample used for the community-level analysis. Section 2.3 then discusses the
dependent and independent/predictor variables for analysis, and includes a description
of the distribution of these characteristics among respondents of the national sample.
Section 2.4 describes the analytical methods and programs used to assess Internet use
as a predictor of community participation. Finally, Section 2.5 discusses the
limitations of the data sources.
13


re 1: Plan for the Dissertation
imm

CHAPTER 1
INTERNET USE AND
COMMUNITY
PARTICIPATION
To what extent is Internet utilization
a significant predictor of individual
and aggregate levels of community
participation?
'A

vr% l 'a* * '. . -
f -
m
&£&

CHAPTER 2 lf§ CHAPTER 3
RESEARCH K PREDICTING
QUESTION \ INDIVIDUAL-LEVEL
& METHODS / PARTICIPATION
(National Sample)
-

XX^.^J'yv ....... .., ^
CR< Summary
V^ypoiitics" :
^vtMcV' :
k Charitable
__________
falntd^neFUse,
4ir;v.
wi* Ifah CHAPTER 4 fa APPENDIX A -
PREDICTING PREDICTING
XS COMMUNITY- Rationale L-V COMMUNITY-
LEVEL Predicts e Models LEVEL
PARTICIPATION 1 nternet l sc INTERNET USE
;; (Community Sample) Context Factors
-iyi-rv
7l£17
<*
CHAPTER 5
IMPLICATIONS
FOR PLANNING
Understanding Community' Participation
Using the internet to Support Participation
Barriers and Opportunities for Enhancing
Participation
^^^**<* - ^ *£*
WiTm** V . **ri
14


Chapter 3 presents the result of the regression models used to assess the
contribution of Internet use as a predictor of overall measures of community
participation and of the four forms of participation that comprise it. It begins with an
overview of the ordinary least-squares (OLS) regression program and the specific
procedures and parameters established for the analysis. Section 3.2 then presents
regression results for the summary community participation measure and assesses the
role of Internet use as a predictor of community participation net of demographic,
socioeconomic, attitudinal and location factors. Section 3.3 then presents regression
results for the four separate components of community participationpolitical
involvement, civic engagement, social interaction and charitable activities. Section
3.4 assesses those factors predicting Internet use with particular attention to evidence
of the digital divide, and the relationships of location and community participation to
Internet use. The chapter concludes with a discussion of the individual-level results.
Chapter 4 presents the results of the community-level analyses of participation
in 33 SCCBS sites. Section 4.1 summarizes the rationale for these analyses and
recapitulates relevant research. Section 4.2 then presents the four-step analytical
strategy for applying predictive regression models to 33 SCCBS communities and
notes the particular sampling and statistical parameters that may limit the
generalizability and degree of certainty of the findings. Section 4.3 presents the
results of the analysis of community participation residuals using the mean level of
Internet use as a predictor of the degree to which communities show higher- or lower-
15


than-expected levels of political participation, civic engagement, social interaction
and charitable activities. The results vary by site with Internet use a modest predictor
of social interaction only. Given these findings, Section 4.4 examines the relative
importance of 15 context factors that appear related to levels of community
participation, and Section 4.5 offers an interpretation of analytical results.
In Chapter 5 the dissertation concludes with a discussion of the implications
of research results for community planning with respect to (a) increasing our
understanding of community participation; (b) using the Internet to support individual
participation in community affairs; and (c) identifying barriers and opportunities for
enhancing participation at the community level. Internet utilization is a significant
positive predictor of individual and, to a lesser extent, aggregate levels of community
participation. However, increased Internet use is unlikely to be sufficient by itself in
overcoming the current patterns of racial and economic segregation that divide urban
America and undermine the full participation of all its citizens.
16


CHAPTER 2
RESEARCH QUESTION AND METHODS
The perspectives offered by previous research leave considerable uncertainty
as to whether the recent explosion of Internet utilization in the United States is likely
to support or undermine community participation at the individual and aggregate
(community) levels. Putnam argues persuasively that over the last fifty years, levels
of social capital have been declining in American communities (Putnam 1995, 2000).
Community participation is a building block of social capital. The purpose of the
present research is not to confirm or confront the declining-social-capital thesis, but
rather to explore the association of Internet use with variations in four types of
individual behaviors that comprise community participationpolitical activism,
participation in civic associations, informal social interaction and charitable activities.
An overall summary measure of community participation is also analyzed. Previous
research has demonstrated the importance of individual demographic and socio-
economic characteristics and context factors such as location that influence both
community participation and Internet use (Putnam 2000; Wellman, Haase et al. 2001;
Robinson, Kestnbaum et al. 2000; Howard, Rainie et al. 2001; Shah et al., 2001a).
Quantitative analyses that that do not simultaneously control for the factors of
17


education, income, age, sex, race/ethnicity, location and other factors are likely to be
misleading in assessing the association between Internet use and community
involvement, and in measuring multiple forms of the digital divide.
2.1 Questions for Empirical Inquiry
The primary research question to be addressed in this dissertation may be
phrased as follows:
To what extent is Internet utilization a significant predictor of individual
and aggregate levels of community participation?
This question may be disaggregated into a series of four related sub questions
Sub question 1: At the individual level, what is the direction and strength of
association between Internet utilization and the four different forms of
community involvement, namely political activity, participation in civic
associations, informal social interaction, and charitable activities?
It is proposed in this research that levels of Internet use will be positively
related to all four types of community participation (and therefore to the summary
measure) after controlling for demographic, socio-economic, attitudinal and location
factors.2 Many of the diverse sources of information about the social impacts of the
Internet reported in Chapter 1 suggest a positive link between community
i
participation and Internet utilization at the individual level, although a causal
2 Putnam himself provides the null hypothesis for this research: By 1999 three independent studies
(including my own) had confirmed that once we control for the higher education levels of Internet
users, they are indistinguishable from nonusers when it comes to civic engagement. (Putnam 2000, p.
170).
18


hypothesis cannot be supported (Wellman, Haase et al. 2001; Blanchard 2000;
Kavanaugh 1999; Howard, Rainie et al. 2001; Shah 2001.)3
Sub question 2: What individual-level characteristics appear to be most
strongly associated with levels of Internet use and community participation,
and to what extent do they indicate a continuation of the digital divide?
It is proposed that many of the same demographic, socio-economic, and
attitudinal factors found previously to be associated with Internet use will be
important here. These include income, education, age (negative), sex, household size,
marital status, student status, race, hours worked per week, and hours of television
watched per day. Other predictors not extensively examined in the literature include
unemployment or disability status, language used in survey (English or Spanish)4,
citizenship, commuting time, tenure (owner/renter), length of residence in the
community, perceived quality of life, social trust, and community alienation. Finally,
the analysis also explores non-linear relationships among key continuous variables
(the square Of income, education, age, work hours, commute hours and quality of life)
in order to determine whether their relationships to Internet use and community
participation attenuate or strengthen with higher or lower values of these predictors.
Sub question 3: At the community level, after taking into account the
characteristics of the sample in each community, to what extent is the level of
Internet utilization in the community a significant predictor of four different
forms of participationnamely, political activity, civic engagement, informal
3 See Section 1.3 above for a fuller discussion of these points.
4 The language used in the survey is an important (negative) predictor of participation among Hispanic
households, and establishes the lesser relative importance of Hispanic ethnicity, and relatively greater
importance of language facility in predicting participation outcomes.
19


social interaction, and charitable activitiesand overall levels of community
participation?
It is proposed that Internet utilization is positively related to levels of
community participation only in certain communities and only for certain forms of
participation. Community participation may be relatively high in certain types of
communities (e.g. less dense and lower-income communities) that show lower levels
of Internet utilizationperhaps as a result of the digital divide or levels of racial
segregation (NTIA 1995; Schon 1999; Allen 2000; Hare 2000; Servon and Nelson
2001; Briggs 2002). Similarly, high-Intemet communities (e.g. relatively wealthy
suburban communities) may show reduced levels of community participation,
particularly political activism and participation in civic associations) (Putnam 2000,
DeFilippis 2001).
Sub question 4: To what extent are the community context factors of
population size, growth rates, density, level of urbanization, socio-economic
status (income, education, occupational status), homeownership, racial
concentrations and segregation significant predictors of community
participation and Internet utilization?
It is proposed that community participation and Internet utilization will vary
for different levels of these predictors. Although communities with higher minority
concentrations and lower socio-economic status may tend to exhibit lower levels of
community participation (Saguaro Seminar 2001, Putnam 2000), the nature of these
relationships is uncertain. Higher levels of Internet use may reflect greater social
interaction in denser, more urbanized communities (Talen 1999, Allen 2001);
20


however, the ability of the Internet to support public dialog and participation in local
issues may be offset in situations of underlying racial* segregation and income
inequality (Oliver 1999; Briggs 2002;). For a more complete discussion of these
potential relationships, see Section 4.2.4 in Chapter 4.
The intent of this research is to inform the ongoing debate about how
Americans use of the Internet is likely to enhance or diminish community
participation generally, and how it may provide a useful medium for increasing
community participation in local planning efforts. If, after accounting for the diverse
influences of demographic, socio-economic and location factors, Internet use predicts
peoples levels of community involvement at the individual level, then the Internet
may be regarded as a potential source of support for greater community participation.
To the extent that Internet utilization is significantly diminished among certain
groups, and at the same time levels of participation among these same groups are
lower than those of their counterparts, then such findings would tend to confirm the
negative association of the digital divide with the political, civic, social and charitable
forms of participation
At the community level it will be helpful to know under what conditions
(context factors) and for which types of communities the Internet appears to predict
the levels and types of community participation. Like participation itself, the Internet
is unlikely to be the deus ex machina that will intercede and cure all the ills of urban
society. However, to the extent that it helps to build connections among the members
21


of diverse communities and enhances access to relevant information, services and
personal networks implicitly useful for addressing urban problems (e.g. health,
education, jobs, housing, etc.), then use of the Internet may amplify the positive
political, civic, social and charitable interactions already taking place (Penzias 1997;
Kavanaugh 2001).
2.2 Sources of Data
The primary source of data used to address the above research question is the
Social Capital Community Benchmark Survey. Secondary sources of data, used to
derive predictor variables in the community-level analyses, are the U.S. Census of
Population and Housing (1990 and 2000), and The Lewis Mumford Center for
Comparative Urban and Regional Research for segregation indices. In addition, the
research examined community planning websites in each of the 33 analysis
communities in order to measure variations in soliciting stakeholder participation and
providing substantive information on planning issues. Each of these sources is
described briefly below.
2.2.1 The Social Capital Community Benchmark Survey.
The analyses undertaken in this research use both the national and community sample
data from the Social Capital Community Benchmark Survey conducted in the latter
half of 2000. (Saguaro Seminar 2001). Building upon Putnams work in Bowling
22



Alone, the survey was a collaborative effort on the part of the Saguaro Seminar: Civic
Engagement in America at the John F. Kennedy School of Government at Harvard
University and some three dozen community and private foundations to assess
differences in civic engagement across the United States. The survey included a
national sample of 3,003 respondents and an additional 26,230 respondents in 40
communities nationwide (across 29 states). The survey, averaging 26 minutes, was
conducted by telephone using random-digit-dialing during the period July to
November 2000. TNS Intersearch, an international survey firm, was commissioned
to conduct the interviewing, and prepare the data for analysis. The national sample (N
= 3,003) of the continental U.S. contains an over-sampling of black and Hispanic
respondents; 501 non-Hispanic blacks were surveyed and 502 Hispanics participated.
In addition, each sponsoring organization (largely community foundations)
determined the sample size and sampling geography for each community. As shown
in Table 2-1 below, the samples range in size from 500-1,500 interviews (the
exceptions are North Minneapolis with a total sample of 452, and Southeast South
Dakota with 368). For the community level analyses, the sample sizes are sufficient
to detect a difference of proportions of 4.38% in the typical site (N=500) with a 95%
degree of confidence. All analyses at the individual level are based on weighted data
to reflect accurately the sampling methods and to allow for a sufficient representation
of minorities across all sites.
23


Table 2-1: Communities in the Social Capital Community Benchmark Survey
Community Variable Name CASES
Phoenix/Maricopa Co. PHX 501
Atlanta Metro ATL 510
Baton Rouge BTR 500
Birmingham Metro BIR 500
Charlotte region/14 county CHR 1,500
Syracuse/Onondaga County SYR 541
Chicago Metro CHI 750
Cincinnati Metro CIN 1,001
East Tennessee TEN 500
Houston/Harris County HOU 500
Kanawha Valley (WV) KWV 500
Kalamazoo Co. KLZ 500
Los Angeles Co. LAX 515
St. Paul Metro STP 503
San Diego Co. SDG 504
San Francisco (city) SFO 500
Detroit Metro/7-co. DET 501
Winston-Salem/Forsyth Co. WIN 750
York (PA) YRK 500
Central Oregon ORE 500
Yakima (WA) YAK 500
Montana MTN 502
Indiana IND 1,001
Fremont/Newaygo Co. (MI) FRE 753
Cleveland/Cuyahoga Co. CLV 1,100
New Hampshire NHM 711
Greensboro/Guilford Co. GRN 752
Peninsula-Silicon Valley PSV 1,505
Lewiston-Auburn (ME) LEW 523
Bismarck (ND) BIS 506
Seattle SEA 502
Grand Rapids (city) GRP 502
Boston (city) BOS 604
Boulder (CO) BLD 500
Delaware DEL 1,383
Rochester Metro (NY) RCH 988
Minneapolis MIN 501
North Minneapolis NMN 452
Rural SE South Dakota RSD 368
Denver (city/co.) DEN 501
Total Community Sample 26,230
National Sample NTL 3,003
Total with National Sample 29,233
Source: Social Capital Community Benchmark Survey (2000)
24


2.2.2 Secondary Sources
Census data from the 1990 and 2000 Decennial Census of Housing and
Population are added to the community level database in order to facilitate analysis of
potential context factors that may affect levels of participation and Internet use. The
variables used in those analyses are
Metropolitan Area Population (millions)
Mean Population Density (persons per square mile)
Mean Percent of Population in Urbanized Area
1990-2000 Change in Population
Percent of Housing Units Owner-Occupied
Area Median Income
Percent College Graduates
Percent Professional Occupations
Percent Non-Hispanic White
Percent Non-Hispanic Black
Percent Hispanic
Percent Asian
In addition dissimilarity indices indicating levels of segregation (white-black,
white-Hispanic, and white-Asian) were obtained from the Lewis Mumford Center
(Lewis Mumford Center 2001). Section 4.2.4 discusses the construction of those
indices and their inclusion in the community-level analysis. In the case of both census
data and segregation indices, the values reported are those of the communities from
which the SCCBS samples were drawn. Therefore, the statistics for Denver and San
Francisco, for example, are those of the cities, not their metropolitan areas, so that
they may be distinguished from Boulder and Peninsula/Silicon Valley, respectively.
25


Finally, the research included an examination of the planning agency websites
in each of the 33 analysis communities. These websites were evaluated in three areas:
the substantive information provided, the solicitation of stakeholder participation, and
the method of presentation. Specific indicators in each of these areas were:
Opportunities for
Substantive Information Community Input
Attractiveness &
Ease of Use
Basic Community Information
Planning Regulations
Meeting Minutes
Comprehensive Plan
Growth/Development Issues
Links to Other Sources
Comprehensiveness of
Information (0-4)
Comments/Suggestions
Staff Email Addresses
Agency Structure
Index/Site Map
Search Function
Ease of Navigation (0-4)
Attractive Design (0-4)
Size of Website (1-4)
Unless otherwise noted in this list, each element was rated as present (1) or absent
(0), and a summary score was created as the sum of individual elements.
2.3 Dependent and Independent Variables
The following sections describe how dependent and independent variables
were constructed for analyzing the extent to which Internet use predicts community
participation.
2.3.1 Dependent Variable: Community Participation
For the purpose of this study, five composite indices of community participation are
used as dependent variables in the analysis. The first four of these measures
summarize respondent activities over the previous 12 months with respect to (1)
26


electoral politics, (2) civic engagement, (3) social interaction, and (4) charitable
activities. They were constructed from 45 items in the SCCBS questionnaire as factor
scores using principal component analysis. See Table 2-2 below.
Table 2-2: Community Participation Indices and Components
Indices of Community Participation Variables Comprising The Index
Electoral Politics [POLFS1] Factor score of: (l)Voted in 1996 presidential election, (2) currently registered to vote, (3) Interest in politics and national affairs (1-4), (4) political knowledge scale (1-4), (5) days in the past week respondent read a daily newspaper (0-7), (6) attended a political meeting or rally in the past 12 months (0-1), (7) Participate in political group (0/1).
Civic Engagement [CIVFS2] Factor score of: (1) Signed a petition in past 12 months; (2) Worked on a community project in past 12 months; Participated in...(3) demonstrations, boycotts, or marches in past 12 months; (4) sports club, league, or outdoor activity; (5) youth organization; (6) parent association or other school support group; (7) veterans group; (8) neighborhood association; (9) seniors groups; (10) charity or social welfare organization; (11) labor union; (12) professional, trade, farm or business association; (13) service or fraternal organization; (14) ethnic, nationality, or civil rights organization; (15) literary, art, or musical group; (1-6) hobby, investment, or garden club; (17) self- help program; (18) Belong to other kinds of clubs or organizations; (19) BeIonged to any group that took local action for reform; (20) Served as an officer or on a committee; (21)Number of: attended a club meeting, (22) Number of: attended public meeting discussing school or town affairs.
Social Interaction [SOCFS1] Factor Score of: frequency of: (1) Had friends over to your home, (2) hung out with friends in a public place, (3) socialized with co- workers outside of work, (4) played cards or board games with others, (5) visited with relatives; (6) How often talk with or visit immediate neighbors; (7) Number of close friends
Charitable Activities [VOLFS 1] Factor Score of: (1) $ contributed to church or religious causes, (2) $ contributed to non-religious charities, (3) Number of (combined) times volunteered; (4) volunteered for religion, (5) for needy, (6) for youth-school, (7) for neighborhood-civic, (8) for health, (9) for culture and the arts.
Community Participation Summary Index [CPSUM4] Mean of: Electoral Politics, Civic Engagement, Social Interaction and Charitable Activities
Source: Saguaro Seminar, Harvard University ( 2001) Combined Data 2 SPSS Codebook,
27


This method of index construction has the advantage of creating four standardized z-
scores (ratio scale) with a sample mean of 0 and a standard deviation of 1, without
having to worry about scaling differences among individual items.
In order to ensure that an appropriate set of groupings was derived, all 45
items were entered into the factor analysis program and between four and seven
factor groupings were tested for the national sample. Percentages of variance
explained ranged from 29.4% to 38.7%. Based on rotated factor loadings, the results
of these tests indicated that most of the 45 questionnaire items sorted themselves into
the groups that would normally be expected. That is, all seven questions relating to
politics tended to load highly on the same factor. For the most part those relating to
civic participation, social interaction and charitable activities did the same. These
results tend to confirm the face validity of the four separate measures of participation
as independent components of community participation.
The second step in the construction of community participation measures
involved using factor analysis to derive a single factor (principal component method)
for each of the four groups of variables. Thus, seven political items, 22 civic
engagement items, seven social interaction items and nine charitable activities items
were entered as groups, and the program extracted summary factor scores or indices
for each set of questions and for each respondent (four scores total for each
respondent corresponding to political, civic, social and charitable activities). The
28


factor loadings on each questionnaire item for the four community participation
component are provided in Appendix B, Table B-l through B-5.
One of the problems in the original survey is that one cannot, on the basis of
the phrasing of the questions pertaining to civic participation, distinguish what is truly
a local involvement from one that is not. For example, participation in a professional,
trade, farm or business association (Question 33J) could indicate involvement in a
very active local business group addressing local zoning issues, or it could indicate
membership in a national trade group with annual conventions in Hawaii. As a final
potential dependent variable, a strictly local factor score for civic engagement, was
extracted as a subset of nine of the original 22 civic engagement items that most
would interpret as dedicated to local issues (worked on a community project, served
as an officer on a committee, attended a public meeting discussing school or town
affairs, participated in a youth organization, participated in a service or fraternal
organization, belonged to any groups that took local action for reform, participated in
a parent association or other school support group, participated in a neighborhood
association, or participated in a seniors group). The Factor loadings on this local
version of civic engagement are provided in Appendix B, Table B-5. Unfortunately,
when this dependent variable was used in the predictive models, it did not
significantly alter the size or sign of the beta coefficients of particular predictors, but
it did reduce model fit (lower R-squares). The reason for the latter is that fewer items
in the factor score had fewer positive responses, thereby skewing the distribution
29


even more and lowering variability in the measure across respondents. In addition, the
intraclass correlation (Chronbachs alpha) dropped from 0.370 to 0.221, and the
standardized item alpha dropped from 0.793 to 0.660. For this reason, the civic
engagement model reported here includes in the dependent variable all of the 23
items referred to in Table 2-2. The predictor coefficients were approximately the
same for the local version of civic participation as for the expanded version.
Respondents answers to the 45 questionnaire items are interpreted as
indications of behaviors (not perceptions or opinions) with respect to their
community activities. Using these four indices, an overall summary index of
community participation is computed as the mean score of these indices (see bottom
of Table 2-2). The reason for using the mean (rather than the factor score) is that it
gives equal weight to each of the four components and produces more variation.
Because all component measures are standardized scores, the community
participation summary score also has a mean of zero, but its standard deviation is less
than one (0.74). As shown in Table 2-3 The consistently high Pearson correlations of
each component with the summary index (varying from 0.60 to 0.84) tend to support
the construct Validity of the summary index as an indicator of respondents overall
participation in their communities. The lower correlations among the four separate
indices (ranging from 0.084 to 0.694) suggest that each is measuring a separate aspect
of community participation.
30


Table 2-3: Pearson Correlations Among Community Participation Components
Community Participation Electoral Politics Civic Engagement Social Interaction Charitable Activities
Community Participation 1
Electoral Politics 599** 1
Civic Engagement .841** .384** 1
Social Interaction .669** .084** .450** 1
Charitable Activities .835** .345** .694** .415** 1
Total N 3,003 3,003 3,003 3,003 3,003
Source: Social Capital Community Benchmark Survey (2000)
** Correlation is significant at the 0.01 level (2-tailed).
The following table provides a summary of statistics describing the four
component measures of community participation as well as the summary score itself.
The only difficulty is that the factor scores for civic engagement, social interaction
and charitable activities are more positively skewed than would be desirable, and as a
result tend to reduce model fit.
Table 2-4: Statistics for Community Participation Dependent Variables
Descriptives Community Participation Summary Electoral Politics Factor Score Civic Engagement Factor Score Social Interaction Factor Score Charitable Activities Factor Score
N 3,003 3,003 3,003 3,003 3,003
Mean 0.000 0.000 0.000 0.000 0.000
Std. Error of Mean 0.013 0.018 0.018 0.018 0.018
Median -0.101 0.094 -0.232 -0.188 -0.338
Std. Deviation 0.739 1.000 1.000 1.000 1.000
Variance 0.547 1.000 1.000 1.000 1.000
Skewness 0.528 -0.085 1.044 0.928 0.829
Kurtosis 0.074 -0.559 0.732 0.903 -0.276
Minimum -1.552 -2.043 -1.098 -2.054 -1.110
Maximum 3.087 2.307 4.279 4.001 3.306
Internal Consistency (Standardized alpha) 0.704 0.793 0.676 0.809
Source: Social Capital Community Benchmark Survey (2000)
31


2.3.2 Measuring Internet Use
In order to estimate the association of Internet use with different forms of
community participation, seven SCCBS questions were used to construct a
continuous measure of Internet utilization:
19. How many hours do you spend using the Internet or email IN A TYPICAL
WEEK, not counting the times you do so for work? (0-25)
20. Do you have access to the Internet in your home? (0/1)
33Q. Are you involved in any group that meets only over the Internet? (0/1)
42. Do you ever telecommute; that is spend a whole day or more per week
working at home instead of going to your main place of work? (0/1)
43. In a typical 5-day workweek, how many days do you normally work at
home? (0-5)
56K. How many times in the past twelve months have you) participated in an on-
line discussion over the Internet? (0-60)
65. How many different telephone numbers does your household have, not
counting those dedicated to a fax machine or computer? (0-9)
Questions 42 and 43 were combined and included in the index because they
provide a measure of the extent to which the respondent relies upon information and
communications technologies (telephone, fax, internet, etc.) to support home-based
work, even though it is recognized that some respondents may not use any such
media. Question 65 was also included because more than one phone line is almost
essential for Internet usage on a regular basis. Given the lack of direct information on
32


how people are using the Internet (email, news, school work, entertainment, research,
shopping, etc.), all three questions are regarded as partial proxies for Internet
connectivity. For the community analysis sample as a whole, 55.8% of respondents
use the Internet at home for an average of 4.3 hours per week. Only 4% of Internet
users are involved in groups that meet over the Internet, and the typical user has had
5.8 Internet discussions over the previous 12 months. Internet users have on average
more phones and work more often at home than non-users of the Internet. Measures
of the internal consistency and reliability of the components of Internet utilization are
relatively high with an average interclass correlation of 0.497 and a standardized item
alpha of 0.496.
The Internet use measure was derived in the same manner as other indices of
community participationthat is, as the factor score of the individual questiomiaire
items using principal components analysis. In order to facilitate the interpretation of
regression results, the Internet utilization index was re-scaled from 0 to 100. The
mean score for the sample as a whole was 13.06 with a standard deviation of 15.00.
Approximately 34 percent of the sample (1028 respondents) did not use the Internet
at all, did not work at home and had no extra telephone. The implications of this
positively skewed distribution on regression results are discussed in Section 3.3
below. Pearson correlations of the Internet utilization index with the individual
component variables are shown in Table 2-5.
33


Table 2-5: Correlations of Internet Utilization Index with Component Items
Internet Utilization Components Sample Mean/Percent Pearson Correlation Coefficients (r)
Internet access at home 54.5% 0.612
Hours/week on Internet 2.64 0.666
Involved in group that meets over Internet 2.7% 0.527
Number of online Internet discussions/year 3.71 0.632
Days/week normally work at home 0.357 0.401
Number of extra telephone lines 1.24 0.405
Source: Social Capital Community Benchmark Survey (2000)
The pattern of these correlations suggests that the number of extra telephone lines
and the number of days per week the respondent works at home may diminish
somewhat the internal consistency of the summary measure of Internet utilization.
They do, however, add information to the general concept of Internet connectedness
(Jung, Qiu, and Kim 2001) and indicate the increased importance of home bases for
social interaction (Wellman, Salaff et al.1996 ). A comparison of Internet users and
non-users with respect to the five indices of community participation is provided in
the following chart. On average non-users of the Net (N=1028) report significantly
lower levels of community participation than Net users (N=1975), and these
differences are statistically significant at the .001 level of confidence. It should be
kept in mind, however, that these comparisons mask the importance of many
demographic and socioeconomic characteristics that influence both Internet use and
community participation (e.g. age, education, income, etc.). Therefore, a multivariate
34


approach such as that used in Chapter 3 is essential for disentangling the net
contribution of Internet use to community participation.
Figure 2: Mean Participation Scores for Internet Users and Non-Users
(Scaled 0-100)
Community Electoral Politics Civic Engagement Social Interaction Charitable
participaton (0- (0-100) (0-100) (0-100) Activities (0-100)
100)
Type of participation
2.3.3 Control Variables
In order to measure the independent contribution of Internet use to variations
in community participation, 22 covariates are used as control variables in the
regression analyses. Previous research shows that a broad variety of demographic and
socio-economic characteristics influence the degree of an individuals Internet use.
Alternatively, they provide correlates indicating a digital divide (NTIA 1998,1999,
2000; USIC 2000; Shah et al. 2001; Kraut et al. 2001; Uslaner 2000). Many of these
same variables have also been reported as being significantly related to peoples
community involvement and social networks (Brehm and Rahn 1997; Hampton and
35


Wellman 2000; Putnam 2000; Uslaner 2000; Wellman, Haase et al. 2001;). The
decision of which covariates to use in the models is guided both by previous research,
and empirically by one-way Pearson correlation coefficients (or chi-squares in the
case of nominal-level variables) between each characteristic and the dependent
variables. The selected covariates together with their distributional characteristics in
the national sample of 3003 respondents are provided in Table 2-4 on the following
page.5
Three attitudinal indices are included as covariatesperceived quality of life,
social trust and community alienation. A Quality of Life index was derived as a
composite summary of five survey questions describing respondents overall
happiness, state of health, rating of ones community as a place to live, perceived
impact in making the local community a better place to live, and level of satisfaction
with personal financial situation. Some research suggests that quality of life (as a
desirable outcome) is enhanced by community participation, social interaction and/or
Internet technology (McPheat 1996; White et al. 1999; Saguaro Seminar 2001).
5 Location attributes (within or outside metropolitan areas, density of residential areas, and region of
the country) have shown to be related to community participation as well (Putnam 1995, p. 664). In
testing the relative contribution of these variables, however, none was found to be statistically
significant when the other covariates were included in the equations.
36


Table 2-6: Covariates in the Regression Analyses
%of Approx. Value
Sample Characteristics Sample Min Max Mean of Mean Valid N
Total household Income 1.00 6.00 3.24 +/-$50,000 2723
Education 1.00 7.00 3.32 "some college" 2984
Age 18.00 92.00 44.63 2954
Female (0/1) 52.3 3003
Household Size 1.00 30.00 3.15 2988
Married (0/1) 58.9 3003
Student (0/1) 3.5 3003
Unemployed or disabled (0/1) 7.2 3003
[Non-Hispanic White (reference group)] 72.8 2965
Non-Hispanic Black 11.9 2965
Asian 2.0 2965
Hispanic 10.1 2965
Alaskan Native/Amer. indian/Other 3.1 2965
Interview in Spanish 4.4 3003
US Citizen 94.2 2991
Hours Worked/Week 0.00 96.00 29.80 3003
Hours Commuting/Day 0.00 4.00 0.26 3003
Length of Residence 1.00 6.00 3.67 +/- 8 years 3000
Homeowner 72.6 2979
TV Hours/Weekday 0.00 12.00 3.04 2977
Quality of Life Index -3.66 1.95 0.00 3003
Social Trust Index -4.35 2.05 0.00 3003
Community Alienation (Agree with statement "The people running my community do not really care what happens to me.") 34.3 2932
Source: Social Capital Community Benchmark Survey (2000)
Others see quality of life as a predicate for greater community involvement and social
cohesion, especially as it conditions peoples willingness to become positively
involved in community affairs (Kawachi and Kennedy 1997; Bothwell, Gindroz, and
Lang 1998). In this analysis the quality of life index is intended to approximate
Uslaners optimism and control variables found to be so important for
37


interpersonal trust and social interaction (Uslaner 2002). All responses are scaled in
the same direction (from low to high) and combined as the mean of their z-scores.6
The average measure intraclass correlation (Chronbachs alpha) for the five items is
0.563 and the standardized item alpha is 0.586.
In the case of social trust, many have argued that it is a predisposing condition
for community participation and a central element of social capital generally (Brehm
and Rahn 1997; Fukuyama 1995; Gittell and Vidal 1998; Purdue 2001; Uslaner
2000). Trust is also presumed to influence, and be influenced by, Internet use (Lebo
2001; Nie and Erbring 2000; Shah, Kwak, and Holbert 2001), although some recent
research finds that frequency of Internet use is not directly related to levels of trust.
As stated by Uslaner, Trusting people are .. .no more or less likely to go on-line than
misanthropes (Uslaner 2000, p. 22). Because of the connections between trust and
both community participation and Internet use, the social trust index in this analysis is
used as a covariate to help isolate the net association of Internet use with community
participation. It was constructed as the factor score of responses to 14 questionnaire
items covering general level of trust, trust in others (neighbors, co-workers, co-
religionists, shop clerks), trust in institutions (local police, local news media, local
and national governments), and racial group trust (whites, African Americans, Asians,
Hispanics and own racial group). In the construction of this measure the average
6 Factor analysis using the principal components method is not used in the construction of the quality-
of-life index in order to give equal weight to each of the components of quality of life.
38


intraclass correlation (Chronbachs alpha) was 0.892, and the standardized item alpha
was 0.8985.
A final variable, community alienation, was included in the analysis to
assess whether people who agreed with the statement the people running my
community do not really care what happens to me," were more or less likely to
participate in their communities.
As reported below, the size and sign of the standardized beta coefficients and
their related t-statistics in the regression models are interpreted as measures of the
relative contributions of the individual covariates and the Internet summary score to
overall variation in the five measures of community participation. Because many of
these 22 control variables are correlated with Internet utilization (but not to the extent
of jeopardizing the assumption of independence), their inclusion in the multivariate
regression models is important for deriving an accurate estimate of the independent
strength and direction of association between Internet utilization and community
participation.
2.4 Analytical Strategy
The principal question addressed in this research is the extent to which
Internet use is a significant predictor of individual and community levels of political
involvement, participation in civic associations, social interaction, and charitable
39


activities, after taking into account demographic, socioeconomic, attitudinal and
location characteristics of respondents. If a positive relationship exists, it would
support the notion of the Internet as a technology that enables greater community
participation, but does not necessarily cause it.7
2.4.1 Individual-Level Analysis Model
The overall strategy for testing the relationship of Internet use to community
participation at the individual level centers on the construction of 5 ordinary least
squares (OLS) regression models that test the contribution of Internet use to variation
in each one of the dependent variables while holding constant variations in
respondents demographic, socioeconomic, and attitudinal characteristics (quality of
life, social trust and community alienation). For a useful overview of the statistics
and assumptions of multivariable models see (Feinstein 1996).
These linear models were of the general form:
Yi5 = A + P1X1...+ PnXn+ PiXj + 6
where
7 In addition to the general caveat that correlation does not imply causation, causal effects of Internet
use on changes in community participation cannot be inferred for two reasons: (a) there is no
randomized experimental design with an appropriate control group; and (b) the SCCBS survey data are
used to derive measures of community participation and Internet use at a single point in time; therefore
change in behavior cannot be assessed. There is the further problem that individuals use of the
Internet constitutes an attribute of the individual and is, therefore not properly regarded as an
instrumental cause per se. For a useful discussion of causal inference from statistical models, see
Holland, P.W. (1986). "Statistics and Causal Inference (with discussion)." Journal of the American
Statistical Association 81, 945-970.
40


Yi..5
A
4 community participation indices, plus summary index;
Intercept (grand mean of the national sample)
Beta coefficients for the socio-economic, demographic, quality
of life and social trust covariates
Values of covariates
Beta coefficients for the respondents Internet utilization index
Value of respondents Internet utilization index;
Error term
Pi... Pn
£
The regressions were performed with SPSS version 11.01. Squared terms for
income, education, age, hours worked per week, commuting hours, quality of life and
Internet use were also included in the equations to test for non-linear effects of these
measures. Education, for example has been demonstrated to have a significant
positive relationship with both community participation and Internet use. What is not
known, however, is whether this positive relationship changes at the highest levels of
education. Similarly, Internet use may be a significant, positive predictor of various
forms of community participation. However, this positive relationship may diminish
at the highest levels of Internet use as high-level users withdraw from community
interaction. Using squared terms for these predictors (after de-meaning the scores)
provides a method for testing nonlinear relationships.
2.4.2 Community-Level Analysis Model
For the community-level assessment of Internet use as a predictor of
community participation, the analytical method is similar to the individual-level
analysis and relies on standard OLS regression models. However, in this case the
41


dependent variable is the difference between actual levels of participation and those
that are predicted based on the respondents particular set of demographic,
socioeconomic and attitudinal characteristics. Predicted community participation
scores are calculated using the same model as for the individual-level analysis, using
the unstandardized beta coefficients as weights for each attribute. The residuals
(actual minus predicted scores) provide a measure of the degree to which a
respondent has higher- or lower-than-expected community participation scores.
Standardized residuals are then used as dependent variables to test the net predictive
power of Internet use by itself at both the individual and community levels. The
advantage of this approach is twofold. First, the residual incorporates (accounts for)
all of the variance of the demographic, socioeconomic and attitudinal covariates that
need to be controlled for if an accurate assessment of Internet use as a predictor is to
be secured. Second, with only 33 cases in the community analysis sample, the
number of variables on the right-hand side of the equations needs to be strictly
reduced. This analysis-of-residuals method is an efficient way to accomplish that
objective, but its effectiveness depends, first of all, on the outcomes of individual-
level model. A more complete description of the rationale and methods used in the
community-level analysis is provided in Chapter 4.
2.4.3 Model Assumptions
Most of the five basic assumptions underlying OLS regression models are not
at issue in this analysisparticularly for the individual-level analyses where large
42


sample sizes help to assure assumptions of normality in both dependent and
independent variables, and in the error term. Standard tests in the SPSS regression
program help to assure that (a) the beta coefficients are maintained in a linear format
(Feinstein 1996); (b) that the predictors are not collineari.e. they are independent of
one another;8 (c) that the mean of the error term does not depend on the independent
(predictor) variables (parameter estimates are unbiased); (d) that the variance of the
error term is constant across levels of the predictors (homoskedastic); and (e) that the
error term is random and normally distributed with a mean of zero.
Two specific adjustments to the basic model have been made. First, the
prediction of Internet utilization uses the natural log form of that dependent variable
in order to offset problems of heteroskedasticity (Internet use is positively skewed).
Second, in the community-level analysis, backwards regression is used to minimize
the number of significant predictors and thereby, support model robustness (Western
1995). Here the small sample size (N=33) almost assures the models will be sensitive
to outliers. See Chapter 4.
2.5 Limitations of the Analysis
There are four considerations that limit the present analysis. First, the SCCBS
was developed to measure individuals levels of community participation and trust as
a national baseline against which to monitor subsequent changes over time. It
8 Tolerance statistics are examined to assure that predictors are not collinear to the extent of biasing
estimates.
43


therefore presents a national cross-section that tells little about how individuals
behaviors and attitudes may be changing over time. The strength of the survey lies in
the detailed measures of different forms of civic engagement at a given point in time,
and in the potential for analyzing the relationship of these measures to different
respondent characteristics including Internet use.
Second, the SCCBS was not designed to explore in detail how respondents are
using the Internet. Shah et al. (2001) have shown that individual level differences in
civic participation are related to the ways in which people use the Internet. In
particular, the survey does not measure the degree to which respondents are using the
Internet for email, information gathering, social interaction, commercial transactions,
entertainment or other uses. It is possible to determine, however, whether people
have access to the Internet at home, how much time they spend on the Net, how much
they are involved in on-line discussion groups, and how much they work at home.
Because respondents were not asked directly or indirectly how their use of the
Internet influenced their patterns of civic engagement, and because changes over time
are not measured, it is not possible to establish a causal model of the influence of
Internet use on community participation. However, statistical assessment of the
strength and direction of the association between Internet utilization and measures of
engagement does permit a greater understanding of the linkages between Internet use
and community involvement.
44


Third, there is considerable diversity among the types of communities
included in the SCCBS. They range from portions of inner cities (e.g. North
Minneapolis), to inner cities (e.g. Denver, San Francisco) to suburban portions of
large metropolitan areas (e.g. Peninsula-Silicon Valley), to entire metropolitan areas
(e.g. Chicago, San Diego), to large regions (Southeast South Dakota), and finally to
whole states (e.g. Montana, Delaware, New Hampshire and Indiana). While sample
sizes within particular communities are large enough for the multivariate analyses
(generally 500 respondents or greater), the range of community types and their non-
random selection for the SCCBS presents a challenge to the generalizability of the
predictive models at higher levels of aggregation. The selection of 33 particular
communities for the aggregate-level analyses and a comparison of their sample
characteristics with that of the national sample in Chapter 4 offset to a considerable
degree the issue of generalizability of the aggregate models.
Finally, several of the predictor variables have forms that are less than optimal
for the prediction models. In the SCCBS database the income, education and length-
of-residence variables are provided in ordinal formats that reduce their power as
predictors, even though they are still statistically significant. As mentioned above, the
primary predictor, Internet use, is positively skewed, and therefore, as a dependent
variable, must be transformed to its logarithmic form to minimize problems of
heteroskedasticity.
45


CHAPTER 3
PREDICTING INDIVIDUAL-LEVEL PARTICIPATION
The analysis that follows is presented in three parts. Section 3.1 summarizes
the regression procedures used in the analysis. Sections 3.2 and 3.3 focus on
regression results describing the relationship between Internet use and community
participationfirst, with respect to the global measure of participation, then with
respect to four different forms of engagement (political, civic, social and charitable).
Section 4.4 addresses the question of the extent to which respondents demographic
and socio-economic characteristics, quality of life, levels of social trust and location
factors are associated with levels of Internet use. This analysis shows that as of 2000
Internet use was still strongly conditioned by income, education, age (negative),
student status, ethnicity (negative for Hispanics only) and nonmetropolitan location
(negative). The data do not, however, support the presence of a digital divide among
other racial and ethnic minorities.
3.1 OLS Regression Procedures
In order to draw appropriate inferences about Internet use as a predictor of
community participation, all regressions are performed using weighted versions of
46


both dependent and predictor variables. A comparison of weighted versus unweighted
regression models had relatively little impact on the slopes and signs of the
predictors, or overall model fit. Weighted regressions are preferred because they take
into account the original sample design that involved an over-sampling of blacks and
Hispanics in the national sample in order to secure appropriate levels of
representation (Saguaro_Seminar 2001).
In the individual level analysis a variety of step-wise and simultaneous models
were examined. The step-wise model enters first that predictor with the highest
stochastically significant F, and at each step adds a new variable with the highest F
among the remaining predictors until there are no remaining variables with F statistics
higher than p<0.05. Because it is possible for the inclusion of new variables to change
the relative contribution and significance of predictors already in the model, at each
step the program also removes any variables with an F greater than p>0.10 (called the
F-to-remove parameter). The step-wise approach is useful in helping to identify (a)
the relative importance of individual predictors, (b) the effects of particular predictors
on changes in model fit (R-square), and (c) the most efficient combination of
significant predictors that maximize model fit. (Feinstein 1996). Overall, comparing
the step-wise and simultaneous methods, the results of a step-wise approach yield
little additional information beyond that obtained with the simultaneous model.9 In
9 Part of this result may be due to the relatively large sample size of 3,003. Smaller sample sizes tend
to make the results more model dependent where coefficients and R-square can become artifacts of
the model chosen, rather than accurate descriptors of particular outcomes.
47


the latter a complete picture of the independent effects of Internet use net of
respondent attributes is provided, and shows the relative contribution of both
significant and insignificant predictors to variations in community participation. The
step-wise approach did show, however, that the inclusion of Internet use at the last
step changed overall R-square by 2-3%.
For the community-level predictive models of participation, a backwards
regression approach was used. Chapter 4 provides a more complete discussion of that
methodology.
3.2 Regression Results for Community Participation Summary
Table 3-1 on the following page shows the results of regressing respondent
characteristics, quality of life, social trust, alienation and Internet use on overall
community participation. The size and sign of the standardized beta coefficients
indicate the percentage by which each characteristic would change with a one-
standard-deviation change in the dependent variable. For example, a one-standard-
deviation change in the community participation index would correspond to a 6.5
percent increase in a respondents total household income, or a 32.1 percent increase
in the respondents level of education.10 Squared terms for income, education age,
hours worked per week, commute hours per day, and quality of life are included in
10 The t-test statistic provides a measure of the significance of the beta coefficients; that is, the higher
the t-statistic, the less likely the individual estimates are due to chance.
48


Table 3-1: Regression Results for Community Participation Summary
Respondent A ttributes Form/1 Community Participation Standardized Coefficients Beta t-Statistic Sis.
Intercept -7.199 ***
Total household Income DM 0.065 3.519 ***
- Income squared 0.034 2.188 *
Education DM 0.321 15.105 ***
- Education squared -0.078 -4.267 ***
Age DM 0.072 3.407 **
- Age squared
Female 0/1
Household Size 0.083 4.942 ***
Married 0/1 -0.030 -1.772 +
Student 0/1 0.061 3 709 ***
Unemployed or disabled 0/1 -0.040 -2.338 *
Non-Hispanic Black 0/1
Asian 0/1 -0.046 -2.990 **
Hispanic 0/1
Alaskan Native/Amer. Indian 0/1
Interview in Spanish 0/1 -0.053 -2.624 **
US Citizen 0/1 0.073 3.984 ***
Hours Worked/Week DM 0.069 3.087 **
- Work Hours Squared 0.060 3.650 ***
Hours Commuting/Day DM -0.089 -3.309 **
- Commute Hours Squared 0.066 2.778 **
Length of Residence 0.084 5.118 ***
Homeowner 0/1
TV Hours/Weekday -0.027 ' -1.694 +
Quality of Life Index 0-100 0.196 10.748 ***
- Quality of Life Squared 0.066 4.258 ***
Social Trust Index 0-100 0.081 4.489 ***
Community Alienation 0/1 -0.074 -4.645 ***
Internet Utilization Index 0-100 0.164 6.922 ***
- Internet Utilization Squared -0.049 -2.226 *
Adjusted R-Square 35.3%
F-Test (significance) 55.532***
Sample Size 3003
Source: 2000 Social Capital Community Benchmark Survey
Note 1: DM=variable de-meaned; 0/l=dummy variable; 0-100 variable scaled to 100 points.
Note 2: *** = p<0.001; ** = pO.Ol; = p<0.05; + = p<0.10___________________
49


the equation because scatter plots revealed a potential non-linear (quadratic)
relationship of these covariates to the dependent variable. In order to assure a correct
interpretation of the beta coefficients and points of inflection (changes in slope), and
in order to reduce problems of collinearity, these variables were de-meaned before
they were squared (Kuha 1999).11 As shown at the bottom of the table, the model
explains 35.3% of the variance in community participation (R-square), and the overall
equation is highly significant with an F-statistic of 55.53. The results of this model
may be summarized as follows.
First, overall community participation (the mean of political, civic, social and
charitable activities) is positively related to respondents education, perceived quality
of life, Internet use, length of residence, household size (number of children), social
trust, citizenship status, age, hours worked per week, income, and student statusin
the order of decreasing beta coefficients. Education is nearly three times as important
as income. Quality of life and Internet utilization are nearly twice as important as
income in predicting levels of overall community participation.
Overall community participation is negatively related to hours commuting per
day, sense of community alienation, conducting the interview in Spanish, being
Asian, being unemployed or disabled, being married and spending more time
11 De-meaning a predictor simply involves subtracting the sample mean of a variable from the value
for each case, so than the mean for the revised variable becomes zero. This has the effect of de-linking
squared terms from the un-squared ones (reducing collinearity). The coefficients for un-squared terms
represent the slope of the predictor at the mean of 0, and the coefficients of the squared terms are the
slope(s) from the mean of 0 outwards.
50


watching TV. The latter two variables are statistically significant at only the 10
percent level of confidence.
Black, Hispanic and Native Americans respondents appear, on average, to be
no different from whites in their levels of participation. When the language of the
interview is not included in the equation, Hispanic ethnicity becomes a significant
negative predictor. This would suggest that facility with the English language plays a
positive role in community participation among Hispanic households.
The beta coefficients for the squared terms indicate that the positive effects of
income, hours worked and quality of life on community participation tend to get
stronger at higher levels of those variables. At the same time, the positive effect of
education and Internet use, and the negative effect of commuting time attenuate or
level off at higher levels of those variables. The highest levels of Internet use do not
necessarily correspond to equally high levels of community participation.
3.3 Internet Use And Forms of Community Participation
The above results indicate that Internet use is an important form of support forif not
a direct cause ofpeoples participation in community. This finding leads to the
question To which forms of community participation is Internet use most strongly
related? Table 3-2 provides a comparison of regression results for four different
forms of community participation: electoral politics, civic engagement, social
interaction and charitable activities. The comparisons reveal that Internet
51


Table 3-2: Regression Results for Four Components of Community Participation
Electoral Politics Civic Engagement Social Interaction Charitable Activities
Std Coefficients Std Coefficients Std Coefficients Std.Coefficients
Respondent Attributes Beta t-Stat Sig. Beta t-Stat Sig. Beta t-Stat Sig. Beta t-Stat Sig.
Intercept -9.743 *** -4.093 *** -2.777 ** -4.361 ***
Total household Income 0.062 3.095 ** 0.066 3.270 ** 0.063 3.148 **
- Income squared 0.029 1.748 + 0.031 1.842 +
Education 0.382 19.376 *** 0.296 12.975 *** 0.064 2.756 ** 0.224 9.703 ***
- Education squared -0.128 -7.552 *** -0.068 -3.420 **
Age 0.336 17.190 *** 0.082 3.619 *** -0.279 -12.159 *** 0.065 2.832**
. Age squared -0.049 -2.880 ** 0.125 6.316 ***
Female -0.076 -5.333 *** 0.066 3.964***
Household Size 0.100 5.546 *** 0.098 5.405***
Married 0.051 3.232 ** -0.038 -2.051 * -0.105 -5.680 ***
Student 0.053 3.002 ** 0.077 4.347 *** 0.038 2.140*
Unemployed/disabled -0.034 -2.091 * -0.035 -1.911 + -0.032 -1.688 +
Non-Hispanic Black 0.033 2.296 * 0.048 2.879 ** -0.087 -5.091 *** 0.048 2.823 **
Asian -0.034 -2.351 * -0.029 -1.720 + -0.037 -2.193 * -0.037 -2.220 *
Hispanic 0.035 1.661 + -0.060 -2.839 **
AL Native/Amer. Indian 0.043 2.662 ** -0.032 -1.939 +
Interview in Spanish -0.032 -1.717 + -0.037 -1.707 -0.053 -2.440 * -0.038 -1.746 +
US Citizen 0.156 9.211 *** 0.047 2.388 *
Hours Worked/Week 0.064 2.649 ** 0.082 3.372 ** 0.077 3.181 **
- Work Hours Squared 0.078 4.445' *** 0.044 2.441 0.058 3.252**
Hours Commuting/Day -0.051 -2.043 * -0.122 -4.178 *** -0.066 -2.250 *
- Commute Hours Squared 0.045 2.058 * 0.096 3.716 ***
Length of Residence 0.072 4.744 *** 0.058 3.267 ** 0.072 4.029 *** 0.044 2.463 *
Homeowner 0.038 2.396 * -0.038 -2.057 *
TV Hours/Weekday -0.061 -3.515***
Quality of Life Index 0.078 4.601 *** 0.121 6.171 *** 0.197 9.934 *** 0.171 8.637***
- Quality of Life Squared 0.027 1.885 + 0.056 3.368 ** 0.052 3.103 ** 0.056 3.317**
Social Trust Index 0.115 6.874 *** 0.046 2.346 * 0.058 2.960**
Community Alienation -0.031 -2.086 * -0.051 -2.984 ** -0.053 -3.094 ** -0.085 -4.971 ***
Internet Use Index 0.109 4.971 *** 0.166 6.515 *** 0.095 3.687 *** 0.135 5.265***
- Internet Use Squared -0.081 -3.990 *** -0.059 -2.477 *
Adjusted R-Square 44.5% 25.5% 23.3% 24.0%
F-Test (significance) 81.110 *** 35.211 *** 31.470 *** 32.641 ***
Sample Size 3003 3003 3003 3003
Source: 2000 Social Capital Community Benchmark Survey
*** = pO.OOl; ** = p<0.01; = p<0.05; + = pO.lO
52


utilization is positively and significantly related to all four forms of community
participation, but that its relative importance differs across the four types. Moreover,
the relative contributions of specific respondent characteristics and location factors
also vary across types of community participation. In comparing the four models, the
main results may be summarized as follows.
Overall, the regression models are most successful in predicting levels of
political engagement (adjusted R2 = 44.5%) and civic engagement (adjusted R2 =
25.5%); the model for social interaction explains the least proportion of variance
(23.3%). Internet utilization is positively and significantly related to all four forms of
community participation. A one-standard-deviation increase in participation
corresponds to between 9 percent (social interaction) and 17 percent (civic
engagement) increases in Internet utilization. Compared to the other predictors,
Internet utilization ranks second in importance for civic engagement, third for
charitable activities, fourth for social interaction and fifth for electoral politics.
However, for two of the four forms of participation (electoral politics and charitable
activities) these relationships are reversed at the highest levels of Internet use. The
beta coefficients for the squared terms of Internet use indicate that participation in
electoral politics and charitable activities tends to decline among the heaviest users of
the Net.
Education is the single most important positive predictor for electoral politics,
civic engagement and charitable activities (giving and volunteering). It is less
53


important for social interaction where quality of life, Internet utilization, hours
worked per week, student status, length of residence and income are more powerful
predictors. As shown by the significant negative coefficient for education squared,
involvement in electoral politics tends to diminish at the highest education levels.
Age is highly significant as a positive predictor of electoral politics, civic
engagement, and charitable activities. Peoples participation in these forms of
community tends to increase with age. However, social interaction tends to decline
with age up to about age 60, but this negative relationship is reversed after that.
Income level is positively related to civic participation, social interaction and
charitable activities, although its relationship to political engagement is insignificant.
At least for this sample, women are significantly less likely than men to be
politically engaged, and no different from men in their involvement with civic
activities or in levels of social interaction. On the other hand, they are much more
likely than men to be involved in charitable activities.
Among minority respondents, Hispanics are much less likely than non-
Hispanic whites to be socially engaged, but slightly more likely to be civically
involved once language of the interview is taken into account. Their differences with
respect to electoral politics and charitable activities are not statistically significant.
Asians also show reduced levels of engagement across all four forms of participation,
compared with their white counterparts. Non-Hispanic blacks, on the other hand,
show significantly higher levels of participation in electoral politics, civic
54


engagement and charitable activity relative to whites. They are, however,
significantly less likely to be socially engaged in their communities.
Quality of life is a consistently strong predictor of community participation
across all four forms of engagement, but particularly so for social interaction,
charitable activities and civic engagement. Out of 24 predictors it ranks first as a
predictor in the social interaction model, second in the charitable activities model,
third in civic engagement and fifth in the electoral politics model.
As expected, social trust scores show a strong positive relation to engagement
in electoral politics, social interaction and charitable activities, but they are neutral
with respect to civic engagement. Opposite effects are noted for community
alienation. Those agreeing with the statement the people running my community do
not really care what happens to me are much less likely to be involved in all four
forms of community participation. Alienation breeds lack of engagement or even
withdrawal.
In analyzing these different forms of community participation, it is clear that
they are not uniformly related to particular household characteristics. While
education, quality of life, Internet utilization, social trust, alienation and age are
consistently important predictors; the strength and direction of relationships to other
predictors depend upon the types of community participation considered. As
discussed in Chapter 4, this finding undermines somewhat the concept of an overall
measure of participation. Results are different for different forms of participation.
55


3.4 Predictors of Internet Use
The foregoing analysis leaves little doubt about the positive relationship of
Internet use to community participation overall, and to the four forms of participation
that comprise it. If the Internet acts as gateway to expanded community involvement,
then the question of who is using that gateway is important for assessing the influence
of the digital divide on levels of community participation among specific groups.
Much has been written about how various demographic and socio-economic
characteristics and context factors influence peoples use of the Internet. A sequence
of NTIA reports in the late 1990s, for example, documents both increasing use of the
Net as well as an important, but decreasing, digital divide between various
disadvantaged groups (low-income families, racial and ethnic minorities, women,
older households, and rural families) and their counterparts (NTIA 1995, 1998, 1999,
2000) . Howard et al. (2001, p. 383) report significant differences in use between
men and women, young and old, those of different races and ethnic groups, and those
of different socioeconomic status. Both the Stanford and UCLA studies confirm
these differences, although some of them are small (Nie and Erbring 2000; Lebo
2001) . The major problem with these studies is that they offer bivariate comparisons
without controlling for multiple and conflicting relationships among independent
variables simultaneously.
The following analysis is directed toward understanding how individuals
background characteristics, locations, and levels of community participation are
56


related to their use of the Internet. As Internet penetration increases, the differences in
Internet use among specific groups in the population are expected to decrease,
resulting in a narrowing of the digital divide (DiMaggio et al. 2001). Knowing how
much people use the Net is more interesting and useful than knowing whether they
are connectedhence the continuous measure of Internet utilization in this analysis.
Approximately 54.5% of the SCCBS national sample respondents (N=3,003)
said they had access to the Internet at home. If this number is used to estimate
Americans access to the Internet, then roughly 107 million adults had such access in
2000. Similarly, 49.3% of the national sample said they made some use of the
Internet on a weekly basis. Extrapolating from Census data, this means that there
were nearly 97 million adult active users of the Internet in the last half of 2000. This
number is far higher than that estimated by Cyberdialogue (roughly 72 million) in
State of the Internet 2000 (USIC 2000). These estimates are presumed to be
conservative because they do not take into account Internet access and use by young
people under 18 years of age, and because the percentages reflect households (one
respondent per household) rather than individuals. The Internet has become a two- 12
12 These estimates were derived as: Total adult population (209,128,094) percentage with telephones
(94%) percent of national sample with Internet access at home (54.5%), OR percent actually using
Internet at home on a weekly basis (49.3%) = 107,136,322 (with access) and 96,914,140 (actively
using). These estimates are slightly higher than those found in the Pew Internet & American Life
Project where 94 million Americans were estimated to have Internet access (Howard, Rainie, and
Jones 2001).
57


way source of information, communication and entertainment for nearly one out of
two households in the United States.
The multiple regression analysis results for Internet use, reported in Table 3-3
below, follow the same format as those presented previously with the exception that
the predictors of Internet use are ordered by the size of their beta coefficients.
Because the sample distribution of Internet use is positively skewed, the logarithm of
Internet use is used as the dependent variable to reduce the problem of
heteroskedasticity. Geographic characteristics (non-metropolitan locations, zip code
density and region of the country relative to the Midwest) are included as predictors
of Internet use. The summary measure of community participation from Section 3.1
above is also included as a predictor in order to test its relative contribution to
Internet utilization.13 The results tend to confirm the findings of previous research
with the exception of certain aspects of the digital divide. In order of importance,
education, income, community participation, student status, household size (number
of children), homeownership, hours worked per week, married respondent, and
residence in the Northeast and West are all highly significant, positive predictors of
Internet use. Age is the strongest negative predictor of Internet use, followed by
nonmetropolitan location, Hispanic status, the square of education, hours spent
commuting, length of residence in the community, the square of hours worked per
13 It should be emphasized that the strength and direction of association between predictors
(independent variables) and Internet use cannot be interpreted as indicating causal relationships. See
note 7 supra.
58


Table 3-3: Predictors of Internet use
Natural Logarithm of Internet Use Standardized Coefficients
Respondent A ttributes Beta t-Statistic Sig.'
Intercept 9.825 ***
Education 0.218 9.557***
Total household Income 0.161 8.417***
Community Participation Index 0.140 7.438***
Student 0.097 5.727 ***
Household Size 0.087 4.959***
Homeowner 0.072 3.959***
Hours Worked/Week 0.053 2.268 *
Married 0.051 2.860**
Western region 0.049 2.555 *
Northeast region 0.038 1.960 +
US Citizen 0.029 1.531
Population density (zip 1997) 0.025 1.510
Southern region 0.020 1.014
Asian 0.019 1.139
Unemployed or disabled 0.015 0.807
Community Alienation 0.015 0.878
- Commute Hours Squared 0.014 0.581
TV Hours/Weekday 0.013 0.753
- Age squared -0.012 -0.624
- Income squared -0.016 -1.012
- Quality of Life Squared -0.021 -1.281
Alaskan Native/Amer. Indian -0.023 -1.477
Social Trust Index -0.023 -1.226
Non-Hispanic Black -0.026 -1.567
Quality of Life Index -0.034 -1.763 +
Female -0.039 -2.439 *
Interview in Spanish -0.042 -2.032 *
- Work Hours Squared -0.045 -2.631 **
Lengtli of Residence -0.052 -2.989**
Hours Commuting/Day -0.056 -1.986*
- Education squared -0.057 -3.002**
Hispanic -0.063 -3.065 **
Non-metropolitan county -0.085 -5.208 ***
Age -0.190 -8.737***
Adjusted R-Square F-Test (significance) Sample Size 29.1% 43.396*** 3003
Source: 2000 Social Capital Community Benchmark Survey
Note 1: *** = p<0.001; ** = p<0.01; = p<0.05; + = p<0.10
59


week, Spanish interview and being female. Other minorities (blacks, Asians and
Native Americans) are not significantly different from their white counterparts in
Internet use. Commuting hours shows a distinct tradeoff with Internet use; those
spending more hours commuting are using the Net less. The negative betas for the
squared terms work hours and education indicate that Internet use attenuates at
the highest levels of those variables. Finally the data indicate that being a U.S. citizen,
population density, living in the South, being unemployed or disabled, community
alienation, social trust and quality of life are not significant predictors of Internet use.
Overall, the 34 predictors account for nearly 30 percent of the variance in Internet
use.
With respect to the digital divide, the preliminary conclusion to be drawn
from these results is that lower income, less educated, older, female, Hispanic and
rural families (still) show lower levels of Internet use (NTIA 1999, 2000). However,
utilization levels among the unemployed or disabled, non-Hispanic blacks, Asians
and Native Americans are no different from their counterparts.
Internet use is significantly and positively related to peoples levels of
community participation. Separate analyses indicate that people who are strong users
of the Net are more likely to be politically and civically involved in their
communities, while charitable and social activities show little relationship to Internet
use (data not shown). Clearly, Internet use is not merely a function of who people
60


are, but also reflects what they do with their lives. It appears unlikely that greater
Internet use causes greater community participation; it seems more likely that the
Internet supports people who are already actively involved in their communities
(Kavanaugh 1999).
3.5 Summary of Individual Level Results
Analysis of the SCCBS national sample data indicates that Internet utilization
is strongly associated with levels of community participation. This is true both for the
summary measure of participation and for the four different forms that comprise it
political, civic, social and charitable. After taking into account respondent
demographic and socio-economic characteristics, the net strength of this association
is highest for civic participation, followed by charitable activities, electoral politics
and social interaction.
In the case of electoral politics and charitable activities, there is some
evidence to support the idea that levels of community participation do decline among
the very heaviest users of the Net, based on the negative coefficients for the squared
terms of Internet use. However, contrary to the findings of some previous research,
these results challenge the notion that The Internet could be the ultimate isolating
technology that further reduces our participation in communities even more than
television did before it (Nie and Erbring 2000, p.17).
61


Peoples levels of community participation are influenced by a variety of
socio-economic and demographic characteristics. This research finds that community
participation increases with education, quality of life, length of residence, household
size, social trust, citizenship, age, hours worked, income and student status.
Community participation declines with commuting time, and among those with a
sense of alienation from community, those interviewed in Spanish, Asians, married
households and those spending more time watching television. Without taking these
factors into account it is not possible to assess the net strength or direction of the
association between Internet use and community participation.
Many of the demographic and socio-economic characteristics found in
previous research to be related to levels of Internet utilization are also found here to
be important predictors of use. These include education, income, age (negative),
student status, household size, and hours worked per week. Internet utilization
declines with length of residence in the community, among women, Hispanics and
those living in rural (nonmetropolitan) communities. Asians, Non-Hispanic blacks
and Native Americans are not significantly different from whites in their levels of
Internet use. Internet use is higher among those living in the Northeast and West.
Levels of community participation (primarily electoral politics and civic engagement)
are strong predictors of Internet use.
From these analyses a general pattern appears to be emerging. Internet use is a
strong, positive predictor of community participation, and the more people participate
62


in their communities, the more they use the Internet. Like the telephone, the Internet
is a tool that makes communication, decision-making and information retrieval more
convenient and powerful. Those who are already pre-disposed to participate in their
communities (whether place-based or not) are also those who become more active
users of the Internet because it helps them do more of what they are already doing. As
Penzias has noted, Technology acts as an enablerand... will allow us to be
whatever we are alreadyonly more so (Penzias 1997, p. 1042). The Internet is a
new gateway to community participation that allows those online to do more of what
they are already doing.
In Bowling Alone Putnam asks the question, How can we use the enormous
potential of computer-mediated communication to make our investments in social
capital more productive? (Putnam 2000, p. 180). The results of this research suggest
that Americans use of the Internet does support community participation. Even
though community participation and Internet use lag among Hispanic households, as
Internet penetration continues to expand, the potential for social isolation and
diminished access among particular socio-economic groups appears to be less
pronounced than it once was (NTIA 2000). The implications for planning, discussed
at greater length in Chapter 5, are threefold. First, it seems likely that the Internet
represents a new gateway to greater political, civic, social and charitable participation
by individuals (regardless of cause). As a result, planners, like other local decision-
makers, have an opportunity to use this medium as a vehicle for providing relevant
63


planning information to, and securing feedback from, stakeholders in the community
(Larsen and Rainie, 2002). As demonstrated in Chapter 4, some planning websites are
already well advanced in this regard.
Second, the fear that increasing Internet use will lead to interpersonal
withdrawal and reduced social interaction seems exaggerated (Nie and Erbring,
2000). Indeed, controlling for many other individual characteristics (particularly age),
increasing Internet use is associated with increasing social interaction. On the other
hand, minority groupsAfrican Americans, Hispanics and Asiansare significantly
less likely to be socially active than their white counterparts. This is of particular
concern for social capital formation, because it is through social contacts that
minorities and nonminorities alike increase their economic opportunities and life
chances (Briggs 2002). As suggested in Chapter 4, at the community level patterns of
racial and economic segregation may be partially responsible for reduced levels of
social interaction among minorities. For planners this result should at least ignite a
question as to whether current patterns of spatial separation among racial/ethnic
groups at the local level are undermining the nations achievement of equal
opportunity and social welfare goals.
Third, there is a great deal of variability in Internet use at the individual level.
Fully 44 percent of the national sample does not use the Net at all. At the other end,
the top ten percent of the sample use it for 8 or more hours per week. This wide
variation among individuals facilitates the identification of an important positive
64


relationship between Internet use and community participation, after netting out the
influence of socioeconomic, demographic and attitudinal characteristics. However,
because planning deals with aggregates (not individuals), an important question is
whether these individual patterns persist at an aggregatei.e. communitylevel. The
extent to which a communitys mean level of Internet use predicts higher- or lower-
than-expected levels of participation is the primary focus of Chapter 4.
65


CHAPTER 4
PREDICTING COMMUNITY-LEVEL PARTICIPATION
The analyses in the previous chapter clearly establish that for individuals,
Internet use is a strong positive predictor of community participation after taking into
account the multiple influences of demographic, socio-economic and attitudinal
factors related to political involvement, civic engagement, social interaction and
charitable activities. Among respondents of the national SCCBS sample, those more
actively engaged in their communities use the Internet more than those not so
engaged. This pattern is consistent across all four types of community participation
examined in these analyses. These findings support a view of the Internet as an
information and communications technology that enablesbut does not necessarily
causehigher individual levels of political, civic, social and charitable participation.
However, as noted previously, there is a great deal of variability both in
levels of community participation as well as Internet use among the 3,003
respondents comprising the SCCBS national sample. It cannot be assumed that higher
aggregate levels of Internet use at the community level will be associated with higher
aggregate levels of community participation. The relevant issue for urban planning is
66


whether these positive relationships between Internet use and community
participation persist at the community level where planning takes place. If they do,
then both planners and community organizations interested in planning outcomes may
find the Internet a useful medium for increasing community participation in local
planning.
This chapter addresses two questions about the relationship of mean levels of
community participation to levels of Internet use in SCCBS sites. They are:
1. At the community level, after taking into account the characteristics of the
sample in each community, to what extent is the level of Internet utilization a
significant predictor of overall levels of community participation and its four
different componentspolitical activity, civic engagement, social interaction,
and charitable activities?
2. To what extent do the community context factors of community size, growth
rate, population density, level of urbanization, socio-economic status (income,
education and occupation), racial concentration and segregation levels account
for differing levels of community participation and Internet utilization?
Section 4.1 summarizes the rationale for addressing these two questions with three
propositions about why planners would want to know about the community-level
relationship of Internet use to participation. Section 4.2 then presents a four-step
analytical strategy for applying predictive regression models to 33 SCCBS
communities and notes the particular sampling and statistical parameters that may
limit the generalizability and degree of certainty of the findings. Section 4.3 focuses
on the results of the analysis of community participation residuals using mean levels
of Internet use as predictors of the degree to which communities show higher- or
67


lower-than-expected levels of political participation, civic engagement, social
interaction and charitable activities. These analyses indicate that Internet use is only
modestly successful in predicting these forms of community participation. Given
these findings, Section 4.4 examines the relative importance of 14 context factors that
appear to influence levels of community participation over and above that of Internet
use. Finally, Section 4.5 offers an interpretation of the analytical results.
4.1 Rationale for Community-Level Analyses
Why would one want to know about the relationship of Internet use to
community participation at the aggregate (community) level? For planners, there are
at least three good reasons. First, planning deals with a collective concern called the
public interestan interest that presumably transcends the claims of individuals and
the domination of specialized interest groups or subcultures (Bollens 2002). Despite
its fragility and cleavages, a communitys public interest is larger than the sum of
individual selves. It moves with the participatory behaviors of its membersvoting,
voicing, moving, joining, giving, taking, planning. At its best, planning is a mirror to
both self and other, reflecting multiple interests back to each as a common vision for
a better world. That individuals vote, join, meet, give and use the Internet is
interesting, but not particularly useful. That whole communities can change for the
better as the result of greater engagement has consequence. If the Internet is one of
the connective paths to greater engagement among the many, then planners need to
68


know. Such knowledge will assist them in securing greater participation in planning
decisions, prioritizing community goals and identifying the most efficient means for
attaining them.
Second, many would argue that planning is about community-based structures
(economic, social, political, physical, ecological) that determine who gets what,
when, where and how (Fainstein 2000; Friedmann 2000). Others would argue these
structures are nothing but the virtual creations of real people who through their social
acts reinforce or modify pre-existing power relations (Giddens 1984). Planners
mediate the flows of power and information that define place (Castells 1996; Sassen
2001), and urban places are aggregates of community social structures (Morris 1996).
If the new, many-to-many medium of the Internet restructures both place-based and
relational communities by expanding and redirecting flows of power and information
(Graham and Healey 2000), then planners need to understand the potential
relationship of Internet use to community change at the city level. Such an
understanding can strengthen a pro-active restructuring of power relations in society
through broad-based information exchange, and facilitate the implementation of
planning agendas through consensus building and collaboration.
Third, planning is presumably about public process, policy and decision-
making aimed at solving problems (Innes 1999; Campbell and Marshall 1999; Healey
1992). In a post-modern, multicultural world, questions of equity and access
challenge the meaning of a just planning process (Sandercock 2000, Umemoto 2001).
69


As shown in Chapter 3, a digital divide, though shrinking, still excludes Hispanics,
rural families, older people, and to a lesser extent, women from full access to the
information and knowledge afforded by the Internet. A medium like the Internet can
both transcend and reinforce culture (Levy 2001) because it permits access and
identity to those who might otherwise be excluded from process. Planners should be
concerned if individual-level differences in Internet use and community participation
are reflected at a level where, as facilitators of process, they can and should make a
difference in who participates and who does not.
4.2 Application of Predictive Models to SCCBS Sites
The results of the individual-level analyses reported in Chapter 3 and based on
the national sample of 3,003 respondents show that there are many significant
demographic, socio-economic and attitudinal characteristics that are related to a
summary measure of community participation and to the four different forms of
participation that comprise it. All 22 independent variables (as well as six squared
terms) used in those analyses are significant predictors of one or more measures of
participation. Most of these same variables are significant predictors of Internet use at
the individual level as well. For this reason, multivariate regression analysis is used to
control for these multiple influences on participation and to estimate the direction and
size of the contribution of Internet use by itself to variations in community
participation. It is for the same reason that a community level analysis comparing
70


(unadjusted) mean levels of participation to mean levels of Internet use would be
inaccurate and misleading. A positive or negative relationship between participation
and Internet use in such comparisons would most likely be the result of many
intervening factors such as income, education, age, race, etc., and not Internet use
per se.
As described below, in order to control for these background sample
characteristics within each site, overall predictive models of participation (summary
measure and four components) are estimated using the same 22 demographic, socio-
economic and attitudinal predictors (and excluding measures of Internet use).
National sample data are excluded from these models. The critical outputs of such
models are (a) the predicted levels of community participation that would be expected
given the particular attributes of the samples in each site, and (b) the residuals
calculated for each case as the difference between actual participation scores and
predicted outcomes. So that SCCBS communities can be compared using the same
metric, the analysis relies upon a single pooled regression model for each of the 5
measures of participation (political, civic, social and charitable plus summary) with
all 33 sites combined.
Applying each models intercept (mean level of participation) plus the beta
coefficients for each attribute yields a predicted participation score for each
respondent. Using standard algebraic notation, the formula for deriving predicted
individual scores may be expressed as
71


A
A + p,X1...+ p22X22 = (Yj.,5-8)
Y,..5
where
A
Y1..5
Predicted political, civic, social and charitable participation
level, plus summary index;
Intercept (grand mean of all community samples);
Beta coefficients for the 22 socio-economic, demographic, and
attitudinal covariates;
Values of covariates for each respondent;
Actual participation scores (4 components plus summary); and
Error term or unexplained variance (residual)
A
Pi... P22
X, ...X22
Y, ..5
£
The only differences between this formula and that used in the general analytical
model specified in Chapter 2 is that the calculation of predicted scores leaves out the
error term (unexplained variance or residual) which itself is the difference between
actual and predicted scores.
The general analytical strategy for estimating the net contribution of mean
levels of Internet use to variations in community participation across sites involves a
four-step process of
1. Estimating the parameters of the pooled regression model for 22
covariates across all communities (not including the national sample);
2. Calculating the predicted levels of participation and residuals for each case
(built into the SPSS regression package);
3. Deriving mean predicted and residual scores, as well as mean levels of
Internet use within each SCCBS community; and
4. Testing the relative contribution of Internet use to variations in actual
levels of community participation, while controlling for the influence of
the 22 covariates as captured in the predicted participation scores.
72


In the last step the relevant comparison to test the predictive power of Internet use is
straightforward and may be expressed as...
Actual Participation = What would be predicted based on 22 covariates + Internet
use + error
By transposing the first term on the right side of the equals sign to the left side, the
statement may be simplified to...
Actual-Predicted scores = Community participation residual = Internet Use + Error
term."
If the community participation residuals (actual predicted scores) are standardized
with a mean of zero and a standard deviation of one, then they may be interpreted and
compared as measures of the degree to which mean levels of community participation
are higher or lower than would be expected (relative to a zero mean) given the
particular characteristics of the samples within each SCCBS community.
Before undertaking the required analyses to test this general model, it is first
necessary to select from the original 40 SCCBS sites the particular communities for
which the analyses are relevant.
4.2.1 Selection of SCCBS Communities for Analysis
Table 4-1 below provides a listing of the 40 communities participating in the
Social Capital Community Benchmark Survey. The number for each site identifies 14
14 Note that there is no intercept specified in this model because, if normal assumptions of linear OLS
regression models hold true, then the estimated residuals (unexplained variance) are unbiased and will
have a mean of zero.
73


the sequence within which participating community foundations agreed to sponsors
their portions of the survey. Final sample sizes for each site are shown in brackets.
Fortunately, community sample sizes are judged to be sufficient in most communities
to detect a difference of proportions of 4.38% in the typical site (N=500) with a 95%
degree of confidence.
Table 4-1: Social Capital Community Benchmark Survey Communities (Sample
__________Sizes in Brackets)_______________________________________
1 Phoenix/Maricopa County [501] 2 Atlanta Metro (7-county) [510]
3 Baton Rouge [500] 4 Birmingham Metro [500]
5 Charlotte Region/14 counties [1500] 6 Syracuse/Onondaga County [541]
7 Chicago Metro (Cook County) [750] 8 Cincinnati Metro [1001]
9 East Tennessee [500] 10 Houston/Harris County [500]
11 Kanawha Valley (Charlotte, WV) [500] 12 Kalamazoo County (MI) [500]
13 Los Angeles County [515] 14 St. Paul Metro [503]
15 San Diego County [504] 16 San Francisco (city) [500]
17 Detroit Metro/7-County [501] 18 Winston-Salem/Forsyth Cnty [750]
19 York (PA) [500] 20 Central Oregon [500]
21 Yakima (WA). [500] 22 Montana [502]
23 Indiana (state) [1001] 24 Fremont/Newaygo County [753]
25 Cleveland/Cuyahoga County [1100] 27 New Hampshire [711]
30 GreensbOro/Guilford Co [752] 32 Peninsula-Silicon Valley [1505]
33 Lewiston-Auburn (ME) [523] 34 Bismarck (ND) [506]
35 Seattle [5021 36 Grand Rapids (city) [502]
37 Boston (city) [604] 39 Boulder County (CO) [500]
40 Delaware [1383] 44 Rochester Metro (NY) [988]
46 Minneapolis [501] 47 North Minneapolis [452]
48 Rural S.E. South Dakota [368] 49 Denver (city/county) [501]
51 National Sample [3003] Total Weighted SCCBS sample: 29,239
Source: Social Capital Community Benchmark Survey (2000)
It is clear, however, that the communities are not readily comparable. There is
considerable diversity among the types of communities included in the SCCBS. They
range from portions of inner cities (i.e. North Minneapolis), to central cities (i.e. San
74


Francisco, Denver, Boston, Minneapolis, Seattle), to combinations of suburban and
partial metropolitan areas (i.e. Peninsula/Silicon Valley), to metropolitan areas (e.g.
Detroit, San Diego, Houston, etc.), to large regions (i.e. Southeast South Dakota, East
Tennessee) and, finally, to whole states (e.g. Montana, Delaware, New Hampshire
and Indiana). These large differences in scale make it difficult comparing cities and
metropolitan areas with large regions and whole states. For the purpose of this
analysis, the concept of community is a priori defined to include only sub-regional,
predominantly urban areas with defined geographic boundaries.15 Given these
criteria, the seven SCCBS sites excluded from the analysis are East Tennessee,
Central Oregon, Montana, Indiana, New Hampshire, Delaware and rural Southeast
South Dakota. The selection diminishes global sample size available for analysis by
18.9%, from 26, 216 to 21,270. As shown in Table 4-2 below, these areas differ from
the analysis sample primarily in terms of their mean level of urbanization (51.8%
versus 78.7% for the analysis sample) and in terms of the percentage of the sample
living in rural areas (44.7% versus 5.2% for the analysis sample).
15 A second reason for excluding these sites is that important measures of racial segregation are not
readily available at the state and/or regional level.
75


Table 4-2: Mean Percent Urban and Percent of Sample in Rural Areas in Analysis and
Non-Analysis Sample Communities.a
SCCBS Communities Number of Communities Sample N Percent Urban in Zip code (1990) Percent of sample in Rural Areas
Analysis Sample Communities 33 21,270 78.7% 5.2%
Non-Analysis Sample Communities 7 4,946 51.8% 44.7%
Total 40 26,216 73.6% 12.6%
Source: Social Capital Community Benchmark Survey (2000).
a Comparison does not include national sample.
A second issue confronting the interpretation of differences in community
participation residuals is that the selection of SCCBS sites was not random. Although
the survey respondents within each site were randomly selected, the communities
themselves were not. Rather, participation depended upon the willingness of local
community foundations to sponsor their portion of the community surveys (Saguaro
Seminar, 2001). This means that the findings of these analyses at the community
level cannot technically be generalized as representing the full population of all urban
communities within the United States. However, the following considerations offset
to a considerable degree the seriousness of this limitation of non-generalizability.
First, as shown in Table 4-3 below, the community sample (N=21,270 in 33
communities) is very similar to the national sample (N=3,003) in terms of those
characteristics used as covariates in the community participation models (income
76


Table 4-3: National and Community Sample Characteristics
Percent of Sample
National Community
Nominal Measures Sample Sample
Female (0/1) 52.30% 52.50%
Married (0/1) 58.90% 55.20%
Student (0/1) 3.50% 4.10%
Unemployed or disabled (0/1) 7.20% 6.00%
Non-Hispanic White 72.80% 68.80%
Non-Hispanic Black 11.90% 13.70%
Asian 2.00% 4.40%
Hispanic 10.10% 9.50%
AlaskanNative/Amer. Indian/Other 3.10% 3.60%
Interview in Spanish 4.40% 5.00%
US Citizen 94.20% 93.10%
Homeowner 72.60% 69.20%
Community Alienation (Agree with statement "The people running my community do not really care what happens to me.") 34.30% 34.50%
Continuous/Ordinal Measures Mean National Community Sample Sample Standard Deviation National Community Sample Sample
Total household Income (1-6) 3.24 3.27 1.44 1.41
Education (1-7) 3.32 3.38 1.88 1.74
Age (18-118) 44.63 44.47 17.24 17.29
Household Size (1-30) 3.15 3.04 2.01 1.69
Hours Worked/Week (0-96) 29.80 29.39 24.84 24.37
Hours Commuting/Day (0-4.92) 0.26 0.28 0.39 0.42
Length of Residence (1-6) 3.67 3.58 1.48 1.48
TV Hours/Weekday (0-12) 3.04 3.10 2.69 2.74
Quality of Life Index (0-100) 69.05 68.98 15.69 15.70
Social Trust Index (0-100) 68.28 69.97 15.77 17.64
Source: Social Capital Community Benchmark Survey (2000)
education, age, etc.). The community sample has slightly higher percentages of
student, black and Asian respondents, and slightly lower percentages of married,
77


white, Hispanic and home-owning respondents. Among continuous- or ordinal-level
attributes, only the variables of household size and length of residence show
statistically significant differences, with the community sample showing slightly
lower means for these predictors. Because the national sample is random (and
therefore generalizable to the U.S. population as a whole), the close similarity
between the national and community sample means suggests that the latter sample is
roughly comparable to the national sample.
As shown in Table 4-4 below, the national and community samples are not
Table 4-4: Comparison of Participation and Internet Scores for National and Community
Samples*
Community
Participation Electoral Civic Social Charitable Internet
Summary Politics Engagement Interaction Activities Index
Mean National Sample 33.82 59.20 19.98 33.92 25.14 10.65
Community Sample 33.78 49.48 17.14 33.87 23.69 9.72
Standard National Sample 16.05 26.97 17.61 16.52 22.65 12.29
Deviation Community Sample 15.32 22.43 15.25 18.25 21.85 11.40
Variance National Sample 257.53 727.18 310.13 272.75 512.81 151.09
Community Sample 234.57 503.02 232.45 333.05 477.41 130.05
Skewness National Sample 0.53 -0.46 1.09 0.93 0.83 1.79
Community Sample 0.45 -0.25 1.07 0.79 0.83 1.96
Kurtosis National Sample 0.08 -0.72 0.83 0.90 -0.28 4.54
Community Sample -0.07 -0.47 1.04 0.24 -0.30 5.38
N National Sample 3,003 3,003 3,003 3,003 3,003 3,003
Community Sample 21,270 21,270 21,270 21,270 21,270 21,270
Source: Social Capital Community Benchmark Survey (2000),
78


widely divergent in terms of the derived dependent measures of community
participation, nor in terms of the predictor Internet use.16 The table compares statistics
for the mean, standard deviation, variance, skewness and kurtosis of these variables.
The statistical similarity of the measures holds true even though these measures were
derived from two separate pools of respondents. For these measures, the community
sample approximates that of the national sample, thereby making the community
sample appear more representative than might otherwise be the case. Based on
these data, the similarities in characteristics of the two SCCBS samples suggest that
the self-selection of communities participating in the SCCBS is not likely to bias
substantially the community-level models of participation behavior or lead to findings
and conclusions different from what would otherwise obtain with a random
community selection process.
4.2.2 Calculating Community Participation Residuals
Using the general model outlined above, regression equations are estimated
for each of the four forms of community participation (political, civic, social and
charitable) and for the summary measure of community participation overall. They
are based upon the pooled community analysis sample (N=21,270) and use the same
22 predictors as the national model in Chapter 3. Regression results are shown in
16 For this comparison, all measures are scaled 0 to 100.
79


Table 4-5 on the following page. As would be expected, the R-square statistics for
these regressionsvarying from 15.6% to 38.7%are somewhat smaller than those
in the national sample models estimated in Chapter 3, and the F-statistics are much
higher due to the much larger sample sizes. A comparison of the standardized beta
coefficients of the predictors for these regressions with those of the national sample
shows that the sign of the coefficients are the same, and that the relative importance
of predictors for the model is similar. However, the absolute values of the coefficients
for about half of the attributesthat is their slopesare different (data not shown).
As outputs of these regressions, SPSS calculates predicted values for each
respondent and for each dependent variable based on the particular attributes of the
respondent. These are then subtracted from the respondents actual scores on the
dependent variables, creating individual-level residuals denoting the degree to which
the respondents actual scores are higher (a positive residual) or lower (a negative
residual) than would be expected from the predictive models. These individual-level
residuals are then standardized using a z-score transformation (a mean of zero and a
standard deviation of one for the entire sample). Standardized residuals are then
averaged within sites, thus deriving a mean standardized residual for each of the 33
analysis communities and for each of the scores.
Mean standardized residuals for each of the measures of community
participation in each of the sites are provided in Table 4-6. The analysis communities
80


Table 4-5: Pooled Regression Results for Four Forms of Community Participation Plus Summary

Respondent Attributes Community Participation Coefficients" B Std. Error Electoral Politics Coefficients" B Std. Error Civic Engagement Coefficients" B Std. Error Social Interaction Coefficients" B Std. Error Charitable Activities Coefficients" B Std. Error
(Constant) 23.128 0.586 ** 28.980 1.026 ** 7.689 0.800 ** 23.924 0.979 ** 4.874 1.137 **
Total household Income 1.169 0.078 *** 1.223 0.108 ** 0.826 0.084 ** 0.776 0.103 1.438 0.119
- Income squared 0.103 0.038 ** -0.029 0.052 0.122 0.041 ** 0.032 0.050 0.235 0.058 ***
Education 2.799 0.073 *** 4.251 0.100 *** 2.652 0.078 *** 0.501 0.095 *** 2.811 0.111 ***
- Education squared -0.403 0.030 ** -0.589 0.041 *** -0.198 0.032 *** -0.374 0.039 *** -0.328 0.045 **
Age 0.003 0.008 0.361 0.010 *** 0.048 0.008 *** -0.367 0.010 ** 0.031 0.012 *
- Age squared 0.001 0.000 *** -0.003 0.000 ** -0.001 0.000 * 0.009 0.000 -0.002 0.000 **
Female 0.046 0.182 -3.120 0.251 *** -0.069 0.195 0.148 0.239 3.146 0.278 ***
Household Size 0.482 0.060 ** -0.233 0.083 ** 0.668 0.064 *** 0.008 0.079 1.215 0.091 ***
Married -0.804 0.205 *** 0.189 0.282 -0.789 0.220 *** -2.706 0.269 *** 0.923 0.313 *
Student 2.185 0.492 ** 1.513 0.677 * 3.225 0.527 *** -0.474 0.646 3.335 0.750 ***
Unemployed or disabled -0.603 0.420 -1.366 0.578 * 0.030 0.451 -0.153 0.552 -0.975 0.641
Non-Hispanic Black 1.926 0.281 ** 2.364 0.386 ** 4.396 0.301 ** -4.308 0.369 ** 4.366 0.428 ***
Asian -5.356 0.461 ** -5.618 0.635 ** -3.246 0.495 ** -5.896 0.606 ** -4.540 0.703 ***
Hispanic -0.921 0.406 * -0.121 0.558 1.173 0.435 ** -4.570 0.533 *** 0.172 0.619
AlaskanNative/Amer. Indian 2.231 0.480 *** 1.227 0.661 + 3.214 0.515 ** 0.947 0.631 2.115 0.733 *
Interview in Spanish -2.143 0.546 ** -3.181 0.752 *** -0.752 0.586 -3.854 0.717 ** 0.099 0.833
US Citizen 7.023 0.445 ** 14.885 0.612 ** 3.773 0.477 *** 3.127 0.584 *** 4.833 0.678 **
Hours WorkedAVeek 0.013 0.005 * -0.009 0.008 0.013 0.006 * 0.015 0.007 * 0.023 0.008 **
- Work Hours Squared 0.001 0.000 *** 0.000 0.000 0.001 0.000 *** 0.001 0.000 ** 0.001 0.000 ***
Hours Commuting/Day -1.517 0.405 *** 0.130 0.557 -0.023 0.434 -3.418 0.532 *** -2.154 0.617 **
- Commute Hours Squared 0.414 0.183 * 0.004 0.252 -0.014 0.196 0.944 0.240 0.566 0.279 *
Length of Residence 1.040 0.068 *** 1.207 0.094 ** 0.579 0.073 ** 1.063 0.089 *** 0.940 0.104 ***
Homeowner 0.459 0.226 * 1.753 0.311 0.340 0.242 -1.481 0.297 *** 1.438 0.345 **
TV HoursAVeekday -0.278 0.034 *** -0.238 0.047 *** -0.355 0.037 *** 0.066 0.045 -0.465 0.052 **
Quality of Life Index 0.165 0.007 ** 0.132 0.009 *** 0.129 0.007 *** 0.111 0.009 *** 0.223 0.011 **
- Quality of Life Squared 0.001 0.000 ** 0.000 0.000 0.001 0.000 ** 0.000 0.000 0.002 0.000 **
Social Trust Index 1.308 0.107 *** 0.098 0.008 0.029 0.007 *** 0.068 0.008 * 0.081 0.009 **
Community Alienation (cat) -1.293 0.196 *** -0.579 0.270 -0.640 0.210 ** -1.118 0.257 ** -2.425 0.299 ***
Adjusted R-Square 30.55% 38.67% 19.35% 15.62% 20.64%
F-Test (significance) 335.2 *** 479.97 *** 183.23 ** 141.61 ** 198.5 ***
Sample Size 21,270 21,270 21,270 21,270 # 21,270
Source: 2000 Social Capital Community Benchmark Survey (Community Analysis Sample)
Note8: coefficients are the unstandardized beta coefficients used to compute predicted scores and residuals


are listed in descending order of their community participation summary residuals.
Figure 3 shows graphically how communities line up on the summary measure.
Table 4-6: Mean Community Participation Residuals by Analysis Community
Community Participation Electoral Civic Social Charitable
Community Summary Politics Engagement Interaction Activities
San Francisco (city)*** 0.19 0.36 0.17 0.06 -0.05
North Minneapolis* 0.16 0.24 0.16 -0.03 0.10
Seattle* 0.14 0.21 0.18 0.03 -0.02
Lewiston-Auburn (ME)+ 0.12 0.29 -0.03 0.16 -0.07
Minneapolis* 0.12 0.18 0.05 0.03 0.09
Yakima (WA) 0.10 0.07 0.07 0.12 0.00
Bismarck (ND) 0.10 0.26 -0.02 0.06 0.00
Denver (city/co.) 0.10 0.13 0.10 0.02 0.02
Baton Rouge 0.09 0.10 -0.01 0.07 0.11
Kanawha Valley (WV) 0.08 0.16 0.05 0.01 0.03
Greensboro/Guilford Co. 0.07 0.03 0.05 -0.06 0.17
Fremont/Newaygo Co. (MI) 0.05 -0.12 0.02 0.17 0.05
Grand Rapids (city) 0.05 0.01 0.05 -0.02 0.10
Boston (city) 0.05 0.34 0.03 -0.16 -0.04
Boulder County (CO) 0.05 0.12 0.13 0.02 -0.13
Kalamazoo Co. 0.04 -0.10 0.01 0.16 0.01
Los Angeles Co. 0.04 -0.03 0.03 -0.01 0.10
Phoenix/Maricopa Co. 0.03 -0.01 0.00 0.13 -0.03
Detroit Metro/7-co. 0.03 -0.01 -0.01 0.13 -0.04
Charlotte region/14 county 0.01 -0.10 0.01 -0.08 0.17
Syracuse/Onondaga County 0.00 -0.05 0.02 0.01 0.01
St. Paul Metro 0.00 0.10 -0.08 -0.04 0.03
Birmingham Metro -0.02 -0.05 0.00 -0.01 -0.01
Winston-Salem/Forsyth Co. -0.02 -0.02 -0.03 -0.11 0.11
San Diego Co. -0.04 -0.08 0.00 0.01 -0.04
Cincinnati Metro -0.06 -0.20 -0.02 0.02 0.02
Cleveland/Cuyahoga Co. -0.06 -0.04 0.01 -0.04 -0.10
Chicago Metro -0.07 -0.10 -0.05 0.03 -0.09
Atlanta Metro -0.09 -0.15 -0.07 -0.13 0.08
Houston/Harris County* -0.13 -0.08 -0.13 -0.08 -0.08
Rochester Metro (NY)* -0.13 -0.19 -0.08 -0.03 -0.08
York (PA)** -0.16 -0.34 -0.09 -0.01 -0.04
Peninsula-Silicon Valley*** -0.19 -0.08 -0.15 -0.08 -0.21
Mean for All Communities 0.02 0.03 0.01 0.01 0.00
Minimum -0.19 -0.34 -0.15 -0.16 -0.21
Maximum 0.19 0.36 0.18 0.17 0.17
N 33.00 33.00 33.00 33.00 33.00
Source: Social Capital Community Benchmark Survey (2000)
82


Figure 3: Community Participation Summary Standardized Residuals
-0.25
-0.20
-0.15
Standard Deviations from Mean of 0
-0.10 -0.05 0.00 0.05
0.10
0.15
0.20
0.25
Franasco(citV)*'
- £ *
* .
||||§|f
iSJfel

,, ,
U>
CO
m
o
o
V)
Houst<
v Roi
ic ll
Yp
Peninsula-SiliconValli
" ' -r.
v

Cjleveland/Cuy tfiqgdCp;
iKvsChicag> Metro c
Atlanta Mei'rt
on/Hams County* i~
ester Metro (\IY)* ezn
lljfJi
§||§i§j|
Aw *
#, i
Wem0k
.
Biimingharfi Metro
Winston-Salem/Fori yth Cos
San Diego G a:
i.Cincinnati Metro
ork (PA)", c
M
~Snn
nNorth Minneapolis*
attle*
Baton Rouge
Charlotte
Syracuse/Or
Boston (city)1:
Boulder Cou tty (CO):
Kalamazoo Co -
Li is Angeles Cc
PhoenixIMancopa Co
Detr >it Metro/7-co *'
egictn/14 cou tty.
Sti Paul Met o: \

ondaga Cour ly
Kanawha Valley (VW)
S3 Greensboro/Guilford Co:-;
Fremont/NeWaygo Go.- (Mj)
Grand Rapids (city)
i Lewistdn-Aubum (Ml
l Minneapolis*:
lf)+
Yakima (WA)
Bismarck (N 3)
Denver (city/ x>.)
\ ;%
, j-

(Appendix B provides similar charts for each of the four forms of community
participation). One may interpret these scores as proportions of standard deviations
away from a residual mean of zero. Thus, in the case of San Francisco (the
community with the highest mean residual on overall community participation), the
mean level of community participation is 19 percent of a standard deviation higher
than would be expected given the particular combinations of demographic, socio-
economic and attitudinal characteristics of the respondents in that community.
Similarly, the
83


Peninsula/Silicon Valley community south of San Francisco (the community with the
lowest mean residual score) has mean community participation scores 19% of a
standard deviation below what would be predicted based on respondent attributes.
Communities with asterisks (*) or plus (+) signs indicate those that are significantly
different from the mean community participation summary residualin this case the
Syracuse and St. Paul metropolitan areas.
Another outcome of the community-level predictive models is obvious in the
tabulation of residuals shown in Table 4-6. Community participation components
political, civic, social and charitablecan differ widely within and across sites. Thus,
the City of Boston, which is in the mid-range of scores on overall participation, civic
engagement and charitable activities, has the second highest mean residual for
political participation (after San Francisco), and the lowest residual score for social
interaction. Clearly, communities with high or low residuals in one form of
community participation do not necessarily have high or low residuals in another.
These differences in mean residuals among communities are confirmed by their
Pearson correlation statistics as shown in the following table. (See Table 4-7). Overall
community participation is highly correlated with electoral politics and civic
engagement, but only moderately correlated with social interaction and charitable
activities. Electoral politics and civic engagement are moderately correlated (0.581),
but neither residual is significantly correlated with social interaction or charitable
activities.
84


Neither social interaction nor charitable activities is significantly correlated with the
mean residuals of the other three forms of community participation.
Table 4-7: Correlations of Mean Community Participation Residuals
Community
Participation Electoral Civic Social Charitable
_________________________________________Summary Politics Engagement Interaction Activities
Community Participation Summary 1
- Electoral Politics .803(**) 1
- Civic Engagement .sue*) .581(**) 1
- Social Interaction .422(*) 0.101 0.211 1
- Charitable Activities .402(*) 0.026 0.269 -0.121 1
** Correlation is significant at the 0.01 level (2-tailed).
* Correlation is significant at the 0.05 level (2-tailed).
Source: Social Capital Community Benchmark Survey (2000)
For these communities at least, the correlation patterns confirm (a) that the residuals
represent measures of different phenomena; and (b) that communities with higher- or
lower-than expected scores in one form of community participation will not
necessarily show the same level on other forms.
4.2.3 Community Comparisons of Internet Use
The main point of the analysis in this chapter is to discover whether Internet
use in the 33 analysis communities predicts higher- or lower-than-expected levels of
community participation. In order to address this question, a standardized measure of
Internet utilization (the independent predictor variable) is calculated for all
respondents in the community analysis sample (N=21,270) using the same method
described for the national sample analysis in Chapter 2. As described there, factor
85


scores using a single principal component are calculated for each case based on
responses to six questionnaire items. Descriptive statistics (mean and standard
deviations) for each item are presented in Table 4-8 below. For the community
analysis sample as a whole, 55.8% of respondents use the Internet at home for an
average of 4.3 hours per week. Only 4% of Internet users are involved in groups that
meet over the Internet, and the typical user has had 5.8 Internet discussions over the
previous 12 months. Internet users have on average more phones and work more
often at home than non-users of the Internet.
Table 4-8: Components of Internet Use (Community Analysis Sample)
Internet Index Components Non-User Mean 1 Mean User Std. Deviation Total Std Mean Deviation
Internet access at home (recode) 0.00 1.000 0.558 0.497
Involved in groups that meet over the Internet 0.00 0.044 0.205 0.028 0.166
Days/week normally work at home (recode) 0.20 0.350 1.093 0.293 1.015
Number of online internet discussions (recode) 0.00 5.794 15.056 3.561 12.135
Number of phone lines (recode) 1.10 1.296 0.753 1.222 0.661
Hours/week on internet at home (recode) 0.00 4.312 5.608 2.650 4.872
N 8,198 13,072 21,270
Percent of Community Analysis Sample 38.5% 61.5% 100.0%
Source: Social Capital Community Benchmark Survey (2000)
Mean standardized Internet scores for each of the 33 communities are
presented in the following table with communities ordered from highest to lowest
scores. (See Table 4-9). The mean for the sample as a whole is 0 and is closest to that
86


Table 4-9: Standardized Internet Scores by Community
Std. Std. 95% Confidence Interval for Mean
Mean Deviation Error Lower Bound Upper Bound____________________N
San Francisco** 0.443 1.323 0.059 0.327 0.559 500
Seattle* 0.270 1.254 0.056 0.161 0.380 502
Boulder* 0.235 1.013 0.045 0.146 0.324 500
Atlanta* 0.222 1.088 0.048 0.127 0.316 510
Peninsula-Silicon Valley* 0.167 1.041 0.027 0.114 0.219 1,505
St. Paul Metro* 0.104 0.932 0.042 0.023 0.186 503
Phoenix 0.087 1.058 0.047 -0.006 0.180 501
Boston 0.072 0.931 0.038 -0.003 0.146 604
Kalamazoo 0.065 0.878 0.039 -0.012 0.143 500
Greensboro 0.061 1.060 0.039 -0.015 0.137 752
San Diego 0.058 0.954 0.042 -0.025 0.142 504
Baton Rouge 0.025 0.962 0.043 -0.060 0.109 500
Birmingham 0.024 1.083 0.048 -0.071 0.120 500
Denver 0.015 1.071 0.048 -0.079 0.109 501
Rochester 0.011 0.991 0.032 -0.051 0.073 988
Detroit 0.010 1.025 0.046 -0.079 0.100 501
Syracuse 0.003 1.000 0.043 -0.082 0.087 541
TOTAL 0.000 1.000 0.007 -0.013 0.013 21,270
Cincinnati -0.024 0.985 0.031 -0.086 0.037 1,001
Lewiston-Auburn -0.028 1.080 0.047 -0.121 0.065 523
York -0.042 0.955 0.043 -0.126 0.042 500
Winston-Salem -0.062 0.965 0.035 -0.131 0.008 750
Houston -0.075 0.924 0.041 -0.156 0.006 500
Cleveland* -0.095 1.008 0.030 -0.154 -0.035 1,100
Minneapolis* -0.096 0.899 0.040 -0.175 -0.017 501
Bismarck* -0.099 0.898 0.040 -0.177 -0.020 506
Charlotte* -0.101 0.916 0.024 -0.147 -0.054 1,500
Los Angeles* -0.103 0.972 0.043 -0.187 -0.019 515
Chicago* -0.103 0.923 0.034 -0.169 -0.037 750
Grand Rapids* .-0.163 0.870 0.039 -0.239 -0.087 502
Kanawha Valley* -0.164 0.917 0.041 -0.245 -0.084 500
North Minneapolis* -0.166 0.917 0.043 -0.251 -0.082 452
Fremont (MI)* -0.207 0.872 0.032 -0.269 -0.145 753
Yakima (WA)* -0.229 0.869 0.039 -0.305 -0.152 500
*=95% confidence interval above/below grand mean of 0
**=99% confidence interval above/below grand mean of 0.
Source: Social Capital Community Benchmark Survey (2000)
87


of Syracuse (0.003). Asterisks indicate that the 95% confidence interval for a
community sits above or below the sample mean of zero. The top five communities
with respect to levels of Internet use are San Francisco, Seattle, Boulder, Atlanta, and
Peninsula-Silicon Valley, in that order.17 Communities with the lowest levels of
Internet use are Yakima, Fremont, North Minneapolis, Kanawha Valley (Charlotte,
West Virginia) and Grand Rapids. The rankings of analysis communities according
their standardized Internet scores are shown in Figure 4 on the following page.
17 San Francisco, Peninsula-Silicon Valley (San Jose), and Seattle are also in the top five listing of the
86 most wired cities in the country, according to Yahoo! Internet Life magazine (Currier 2002).
Boston ranks fourth in the YIL study, and eighth here. In fact, among cities rated by YIL, ten of the
top 24 are also in the top 15 scores here. These comparisons enhance the validity of the Internet use
scores developed for this research.
88