You are on page 1of 18

460057

State Politics & Policy QuarterlyHawes et al.


SPAXXX10.1177/1532440012460057

State Politics & Policy Quarterly

Social Capital in the 50 States:


13(1) 121­–138
© The Author(s) 2012
Reprints and permission:
Measuring State-Level Social sagepub.com/journalsPermissions.nav
DOI: 10.1177/1532440012460057
Capital, 1986–2004 http://sppq.sagepub.com

Daniel Hawes1, Rene Rocha2,


and Kenneth Meier3

Abstract
The idea of social capital has had a powerful impact on the study of politics, policy, and
social science at large. Much of what we know about the causes and effects of social
capital, however, is limited by the nature of data used regularly by scholars working in
this area. Current data sets allow researchers to study changes in social capital over
time at the national level and static differences in the distribution of social capital
across the states.The inability of scholars to know how social capital varies over time
and across space limits the kinds of questions that can be asked. In this article, we use
sample data of more than 20,000 individuals conducted biannually in the continuous
48 states by the marketing research firm MediaMark Research, Inc. to create a unique
measure of social capital that varies across time and space. These data are available at
the state level from 1986 through 2004.

Keywords
Social Capital, Civic Engagement, Trust

The idea of social capital has had a powerful impact on the study of politics, policy,
and social science at large. In recent years, scholars have paid increasing attention
toward understanding what social capital is and exploring the direct and indirect con-
sequences of living in high social capital environments. Much of what we know about
the causes and effects of social capital, however, is limited by the nature of data used
regularly by scholars working in this area.

1
Kent State University, Kent, OH, USA
2
University of Iowa, Iowa City, IA, USA
3
Texas A&M University, College Station, TX, USA

Corresponding Author:
Rene Rocha, University of Iowa, 341 Schaeffer Hall, Iowa City, IA 52242, USA
Email: rene-rocha@uiowa.edu
122 State Politics & Policy Quarterly 13(1)

Although the concept of social capital has been applied globally (Callahan 2005; Letki
2006; Putnam 1993; Tavits 2006; Warde et al. 2003), several works have focused on study-
ing its role within the context of the state and local politics in the United States (Carden,
Courtemanche, and Meiners 2009; Hero 2003; 2007; Knack 2002; Putnam 2000; Rice
2001; Tavits 2006). Current data sets allow researchers to study changes in social capital
over time at the national level (Keele 2005) and static differences in the distribution of
social capital across the states (Putnam 2000). The inability of scholars to know how social
capital varies over time and across space limits the kinds of questions that can be asked.
Static measures also pose a problem because many rival explanations of policy change,
such as racial/ethnic diversity and political culture, are highly correlated with social capital.
Temporal shifts in racial/ethnic diversity do not always correspond with changes in social
capital (see Hawes and Rocha 2011), allowing scholars the opportunity to establish the
extent to which social connectedness alone creates shifts in political outcomes.
In this article, we offer a unique measure of social capital that varies across time
and space. We rely on sample data of more than 20,000 individuals conducted biannu-
ally in the contiguous 48 states by MediaMark Research, Inc. (MRI), a marketing
research firm. Although limited in some respects, the MediaMark data are able to
construct measures that tap the nonattitudinal components of Putnam’s comprehensive
social capital index. These data are available at the state level from 1986 through 2004.

Social Capital: What Is It? What Is It Good for?


In its simplest form, social capital can be said to refer to “connections among indi-
viduals” or in other words, “social networks and the norms of reciprocity and trust-
worthiness that arise from them” (Putnam 2000, 19). Putnam’s particular definition of
social capital can be further broken down into five components: community and
organizational life, engagement in public affairs, community volunteerism, informal
sociability, and social trust. There is no shortage of claims regarding the positive ben-
efits for individuals holding social capital or even for those lacking in social capital
but living in social capital-rich environments. In addition to bettering health, happi-
ness, and safety, social capital appears to improve government performance by mak-
ing it easier for the public to make policy demands on government; thanks to higher
levels of political sophistication and cooperation within society (Claibourn and Martin
2007; Putnam 2000; Tavits 2006).
Although many normatively desirable outcomes are known, recent research also
suggests that any positive externalities associated with social capital may not be shared
by members of all groups in society. Studies finding that the benefits of social capital
are limited (e.g., Hero 2003; 2007) have left scholars scrambling to better understand
the conditions under which social capital has effects that are positive or negative as
well as large or small.
One possible reason for divergent empirical results is that different components of
the concept carry different implications. Although often treated as a unified concept,
Hawes et al. 123

several authors suggest that social or generalized trust is distinct from organizational
membership, civic engagement, and volunteerism. Uslaner (2004, 501) argues that any
connection between social capital and inequality in American society is absent when
social capital is viewed as generalized trust, which he defines as “trust in people who
are likely to be different from ourselves, rather than trust in people like ourselves.”
This distinction is well supported on conceptual and empirical grounds (Knack 2002;
Knack and Keefer 1997; Paxton 1999; Stolle and Rochon 1998). Some scholars, for
example, find that civic engagement is unrelated to government performance (Knack
2002; Rice 2001), while social trust displays a consistent effect.
This conclusion implies that researchers should be careful to separate out the effects
of trust and civic engagement; however, doing so is difficult. In clarifying the theoreti-
cal distinction between trust and engagement, Uslaner and Brown (2005) concede that
the two are highly related empirically. Their research suggests that high levels of trust
cause individuals to take part in communal activities.1 Brehm and Rahn (1997, 999)
make a different causal argument, but nevertheless stress the close tie between engage-
ment and trust, stating “civic engagement and interpersonal trust are in a tight recipro-
cal relationship, where the connection is stronger from participation to interpersonal
trust rather than the reverse.”

Social Capital across Time and Space


For reasons we discuss below, our concern is primarily with the nonattitudinal pieces
of social capital or those elements that do not concern generalized or social trust.
Developing a measure of social trust that varies across time for each of the U.S. states
is beyond the scope of available data. Indeed, current measures, including those
employed by Putnam (2000), often use General Social Survey (GSS) data even though
the primarily sampling units used by the GSS are not states (i.e., the samples are not
representative at the state level). Such measures do seem to produce reliable results
(Brace et al. 2002), and Uslaner and Brown (2005) have used this approach to develop
a measure of trust that varies by decade. Developing annual estimates, however,
remains unfeasible. Our MediaMark data set does allow for the constructions of
annual estimates of civic engagement, organizational membership, and volunteerism
across the states.
Social capital is but one of many organizational typologies that have been used to
classify states. Others include political culture, racial/ethnic diversity, and policy lib-
eralism. These typologies are almost always considered in isolation, and the presenta-
tion of empirical evidence that accounts for multiple ways of categorizing states is
rare. Hero (2003) offers a notable exception with his attempt to unite these “multiple
traditions” within American politics within a single study, but he is unfortunately con-
strained by the cross-sectional nature of available data on social capital in the states.
This poses a problem given the high degree of correlation between various classifica-
tion systems, making it difficult to parse out the individual effects of each theory.
124 State Politics & Policy Quarterly 13(1)

An overtime approach allows for a greater degree of variation between states and
aids scholars in sorting out the particular effects of social capital for social, economic,
and political outcomes. The need for an overtime measure also follows from the clear
dynamic element of social capital stressed by Putnam (2000) and others. Looking at
changes in social capital over time at the subnational level also enables scholars to
determine whether the causes of variation in social capital across space are the same
as the causes of variation in social capital across time.
In short, our understanding of how social capital affects politics and policy is
severely restricted by studies that can only account for cross-sectional variation across
space and temporal variation in the United States as a whole. The measure we intro-
duce in this article will allow us to move beyond these limitations and better compre-
hend the causes and consequences of social capital in America.

Constructing Social Capital


Data and Measures

Robert Putnam operationalizes state-level social capital by factor analyzing 14 state-


level items spanning several decades. Table 1 lists the items, source, and years used
by Putnam’s comprehensive social capital index. As indicated in the table, not all
items are available for all states; in addition, most of the items are averages of values
over multiple years. Putnam’s index is divided into five categories. Using MediaMark
data, we are able to capture three of these categories: community organizational life,
engagement in public affairs, and community volunteerism. Although our measure
does not contain measures of informal sociability or social trust, it does capture the
nonattitudinal aspects of social capital.
MRI is a marketing research firm that conducts The Survey of the American
Consumer, which consists of personal interviews with more than 20,000 individuals
biannually in the contiguous 48 states.2 Households are selected using a stratified sam-
ple composed of metropolitan areas, nonmetropolitan counties, and the 10 major U.S.
markets. The responses are then weighted and balanced based on a variety of factors,
including region, race, household size, income, and education.3 There is an overall
yearly average of 728 unweighted responses per state when these data are aggregated
by state.4 Although all states are surveyed, a number of states are clustered based on
geographic proximity (see Table 2). For example, North Carolina and South Carolina
are clustered, as are the Dakotas.5 Most of the clusters do not appear to be problematic
because the clustered states tend to be similar to one another in terms of demographic
characteristics.6 To create state-level values, these clusters were disaggregated to indi-
vidual states. Thus, some states share the same values for the MRI data.
Although the MRI survey has numerous commercial uses, it also includes a number
of items pertaining to public engagement, organizational membership, and volun-
teerism similar to those used by Putnam in his comprehensive index. Table 3 presents
the 22 items included in our index. The first set of items pertains to community orga-
nizational life. In particular, nearly every year,7 MRI asked respondents whether they
Hawes et al. 125

Table 1. Putnam’s Comprehensive Social Capital Index


Item Source Years No. of states

Measures of community organizational life


  Served on committee of some local organization in the last year (%) Roper 1974–94 43
  Served as officer of some club or organization in the last year (%) Roper 1974–94 43
  Number of civic and social organizations per 1,000 population USDOC 1977–92 50
  Average number of club meetings attended in the last year DDB-NLS 1975–98 48
  Average number of group memberships GSS 1974–94 40
Measures of engagement in public affairs
  Turnout in presidential elections SAUS 1988, 1992 50
  Attendance at any public meeting on town or school affairs in the last Roper 1974–94 43
year (%)
Measures of community volunteerism
  Number of nonprofit 501(c)(3) organizations per 1,000 population NPA 1989 50
  Average number of times worked on community project in the last year DDB-NLS 1975–98 48
  Average number of times volunteered in the last year DDB-NLS 1975–98 48
Measures of informal sociability
  Agree that “I spend a lot of time visiting friends” DDB-NLS 1975–98 48
  Average number of times entertained at home in the last year DDB-NLS 1975–98 48
Measures of social trust
  Agree that “Most people can be trusted” GSS 1974–96 41
  Agree that “Most people are honest” DDB-NLS 1975–98 48

Note: Roper = Roper Social and Political Trends archives; USDOC = U.S. Department of Commerce; DDB-NLS = DDB
Needham Life Styles archive; GSS = General Social Survey; SAUS = Statistical Abstract of the United States; NPA =
Nonprofit Almanac.

belong to the following organizations: a fraternal order, a religious club, a civic club,
a veteran club, a body of local government, or a country club. The weighted responses
were then converted to percentages of the adult population for each state. As can be
seen in Table 3, there is considerable variation in group membership.
The second set of items capture engagement in public affairs. Using the MRI data,
we include the percentage of the population that reported engaging in the following
activities: (a) voted in a federal, state, or local election; (b) wrote to an elected official
about a matter of public business; (c) wrote to an editor of a magazine or newspaper;
(d) wrote or telephoned a radio or television station; (e) wrote something that has been
published; (f) addressed a public meeting; (g) visited an elected official to express a
point of view; (h) actively worked for a political party or candidate; and (i) engaged in
fund-raising. In addition to these nine items, we include state-level voter turnout for
national elections (congressional and presidential).8
Finally, we include six items that capture community volunteerism. First, we include
a measure of nonprofit activity in each state. Based on internal revenue service (IRS)
filings, the National Center for Charitable Statistics (NCCS) has collected data on the
total number of nonprofit organizations (NPOs) in each state from 1995 onward. NPOs
can broadly be divided into two types: those required to file a Form 990 and those that
are not. In general, all federally tax-exempt nonprofits with incomes greater than
$25,000 and all 501(c)(3) private foundations (regardless of income) are required to file
Form 990. Nonprofits with incomes lesser than $25,000 and most faith-based
126 State Politics & Policy Quarterly 13(1)

Table 2. MRI State Clusters


 1 Alabama/Mississippi
 2 Arkansas/Louisiana/Oklahoma
 3 California
 4 Delaware/Maryland/West Virginia
 5 Florida
 6 Georgia
 7 Illinois
 8 Indiana
 9 Kentucky
10 Maine/New Hampshire/Vermont
11 Massachusetts
12 Michigan
13 Minnesota/Iowa
14 Missouri
15 Montana/Idaho/Wyoming/Colorado
16 Nebraska/Kansas
17 New Jersey
18 New Mexico/Arizona/Utah/Nevada
19 New York
20 North Carolina/South Carolina
21 North Dakota/South Dakota
22 Ohio
23 Pennsylvania
24 Rhode Island/Connecticut
25 Tennessee
26 Texas
27 Virginia
28 Washington/Oregon
29 Wisconsin
Note: MRI = MediaMark Research, Inc.

organizations are not required to file Form 990. For our measure of nonprofits, we
elected to use the former category of nonprofits. We do this because, given the filing
requirements, this provides some sense of comparability in nonprofit type and, to some
degree, size across the states. We use this count to create a per capita measure of non-
profit presence in each state.
A related measure is charitable giving. Beginning in 1995, the Catalogue for
Philanthropy has developed and published the “Generosity Index”—a state ranking of
charitable giving. Using IRS data on adjusted gross income and itemized charitable
contributions, the Generosity Index ranks states on their charitable contributions, given
their level of income.9 We use this index as one of our items in our social capital index.
Hawes et al. 127

Table 3. Descriptive Statistics of Social Capital Items


Source Years missing M SD Minimum Maximum

Community organizational life


  Fraternal order MRI 1991 5.47 2.42 0.34 18.27
  Religious club MRI 1988, 1991 7.04 3.17 0.00 18.38
  Civic club MRI 1991 3.40 1.61 0.00 11.76
  Veteran club MRI 1989, 1991 4.22 2.64 0.35 24.75
  Country club MRI 1986, 1991 2.43 1.43 0.00 12.86
  Body of local government MRI 1991 1.37 1.06 0.00 12.43
Engagement in public affairs
  Voted federal, state, or local election MRI None 49.43 7.72 23.39 72.66
  Wrote to an editor of a magazine or newspaper MRI None 4.60 1.89 0.00 10.95
  Wrote or telephoned a radio or television station MRI None 7.33 2.69 0.00 18.11
  Wrote to an elected official about a matter of public MRI None 7.06 2.75 0.08 16.63
business
  Wrote something that has been published MRI None 3.10 1.55 0.00 8.45
  Addressed a public meeting MRI None 5.94 2.52 0.26 16.56
  Visited an elected official to express a point of view MRI None 3.59 1.70 0.00 10.92
  Actively worked for a political party or candidate MRI None 2.55 1.34 0.00 9.66
  Engaged in fund-raising MRI None 10.67 3.30 0.65 27.61
  Voter turnout SAUS None 45.19 11.49 3.50 76.70
Community volunteerism
  NPOs per 1,000 population NCCS 1986-94 2.09 1.25 0.93 10.89
  Generosity Index CFP 1986-94 25.50 14.45 1.00 50.00
  Contributed to public television MRI None 8.54 3.50 0.56 23.16
  Public television contributions (composite score) MRI None 0.00 1.00 −2.28 3.97
  Actively worked as a volunteer (nonpolitical) MRI None 16.75 5.06 1.62 33.43
  Took an active part in local civic issue MRI None 6.13 2.45 0.44 16.92

Note: MRI = MediaMark Research, Inc., Survey of the American Consumer (http://www.gfkmri.com/Products/
TheSurveyoftheAmericanConsumer.aspx); SAUS = Statistical Abstract of the United States (http://www.census.gov/
compendia/statab/); NPOs = nonprofit organizations; NCCS = National Center for Charitable Statistics (http://nccs.
urban.org); CFP = Catalogue for Philanthropy (http://www.catalogueforphilanthropy.org/natl/generosity_index/2006.
html).

In addition, the MRI data include a measure of the percentage of the population that
contributed to public television (Public Broadcasting Service [PBS]). These data also
break down contributors into three groups based on the amount of the contribution.
This allows us to consider not only how many individuals within a state contributed
but also how generous the contributions generally were. From 1986 through 2000,
these groupings were those who contributed less than $25, those who contributed
between $25 and $49, and those who contributed more than $50. From 2001 through
2004, the contribution categories were those who contributed less than $50, those who
contributed between $50 and $100, and those who contributed more than $100. Given
this discrepancy in contribution amounts, we use factor analysis on these items to cre-
ate a single composite score of charitable giving to public television.
The fifth item used to measure community volunteerism is the percentage of the adult
population that took an “active part in some local civic issue” in the past 12 months.
128 State Politics & Policy Quarterly 13(1)

Table 4. Factor Loadings from Full Factor Analysis


Factor 1 Factor 2 Factor 3 Factor 4 Uniqueness

Eigenvalues 7.57 1.81 1.68 1.30  


Proportion of variance 0.344 0.082 0.076 0.059  
Community organizational life
  Fraternal order 0.3603 0.5656 0.0031 0.2445 0.4905
  Religious club 0.0287 0.792 0.0451 −0.0495 0.3675
  Civic club 0.1319 0.5139 0.3757 −0.059 0.5739
  Veteran club 0.0769 0.5953 0.2998 0.34 0.4342
  Country club 0.1614 0.3998 0.2289 −0.418 0.5870
  Body of local government −0.0607 0.1948 0.6475 0.3675 0.4041
Engagement in public affairs
  Voted federal, state, or local election 0.4643 0.4283 0.1445 0.351 0.4569
  Wrote to an editor of a magazine or newspaper 0.7023 0.1436 0.1392 0.0632 0.4628
  Wrote or telephoned a radio or television station 0.6556 0.1461 −0.0233 −0.2494 0.4860
  Wrote to an elected official about a matter of public business 0.6192 0.2108 0.1854 0.2593 0.4705
  Wrote something that has been published 0.641 0.1767 0.0406 0.1562 0.5318
  Addressed a public meeting 0.4761 0.3073 0.4666 0.3061 0.3675
  Visited an elected official to express a point of view 0.1888 0.1874 0.694 0.1154 0.4343
  Actively worked for a political party or candidate 0.2387 0.0075 0.7595 −0.0922 0.3576
  Engaged in fund-raising 0.3926 0.4281 0.3989 0.2563 0.4377
  State voter turnout 0.0333 0.0549 0.1095 0.736 0.4423
Community volunteerism
  NPOs per 1,000 population 0.2218 0.1331 0.1122 0.7809 0.3108
  Generosity Index 0.4965 −0.2191 0.0139 0.4284 0.5217
  Contributed to public television 0.7538 0.0369 0.306 0.1034 0.3262
  Public television contributions (composite score) 0.5115 −0.1103 0.4068 0.2262 0.5096
  Actively worked as a volunteer (nonpolitical) 0.4926 0.4808 0.2182 0.4649 0.2624
  Took an active part in local civic issue 0.4917 0.3049 0.4406 0.2429 0.4122

Note: NPOs = nonprofit organizations. Cronbach’s  Alpha coefficient; standardized = .90, unstandardized = .78. Principal
components factor analysis. Rotated (orthogonal) factor loadings. The shaded area indicated the primary factor loading
for each item.

Finally, we also include a measure of the percentage of the state adult population that
reported having actively worked as a volunteer in a nonpolitical setting in the past
12 months.

Scale Analysis
To create a dynamic state-level measure of social capital, we factor analyze the
22 items. These items display considerable internal consistency producing a standard-
ized Cronbach’s coefficient alpha of .90.10 Table 4 presents the rotated factor load-
ings11 from the full factor analysis. Four factors were retained; however, the
eigenvalues sharply drop off after the first factor, suggesting the presence of one
dominant factor (DeVellis 2003). The first factor explains about 34% of the variation
in these items, compared with about 8% for the second factor. The first three factors
appear to capture three unique concepts: namely, general engagement in political and
Hawes et al. 129

civic issues, group membership, and political activism. The first factor, for example,
captures volunteerism (e.g., charitable giving, volunteering) as well as general interest
and engagement in current events and local civic matters (e.g., writing a letter to the
editor of a newspaper or magazine).
Most of the community organizational life items primarily load on the second fac-
tor. Membership in a country club loads positively on this factor, but is slightly more
strongly (negatively) associated with the fourth factor. The third factor appears to be
more explicitly political. This includes membership in a body of local government and
active engagement in political campaigns. Engagement in fund-raising loads rather
evenly on first three factors. Conceptually, this makes sense as fund-raising could
relate to either volunteerism or political activism. Furthermore, organizations rou-
tinely conduct fund-raising activities; thus, group membership is likely to be associ-
ated with increased likelihood of fund-raising.
Given the drop off in the eigenvalues after the first factor, and as the primary
objective is to create a single index of social capital, we reran the factor analysis and
only retained the first factor. From this, we created a single composite score, which
gives us a social capital values for each state for each year from 1986 through 2004.12
Table 5 presents the factor loadings for the composite index. For our final index, we
take a weighted moving average of the factor score to smooth year-to-year fluctua-
tions in the measure.13

Validating the Measure


The next step is to validate our dynamic measure of social capital. Given that the
measure is conceptually based on Putnam’s construct, we argue that it is sound in
terms of face and content validity at least to the extent that Putnam’s index is valid.14
One way to examine construct validity in our measure of social capital is to compare
it with Putnam’s comprehensive index. Table 6 presents the correlations of the indi-
vidual items used in our index and the index itself. The correlations range from a low
of .20 (religious club membership) to a high of .83 (addressed a public meeting). The
average correlation is .57. The third column in Table 6 presents the correlation
between the items in our measure with Putnam’s index. These range from .06 (country
club membership) to a high of .729 (NPOs per 1,000 population). The average item
correlation with the Putnam index is .36. Although this seems rather low, the overall
correlation between the two indices is .68.
Temporal variation. In Bowling Alone: The Collapse and Revival of American Com-
munity, Putnam dedicates a considerable number of pages to discussing how social
capital has been declining in the United States. Our measure, then, should mimic this
pattern at the national level. By aggregating the dynamic measure to the national level,
we obtain a single national value for each year. Figure 1 presents the overtime change
in aggregated social capital. Similar to Putnam’s conclusions, we also find a general
decline in social capital at the national level, although the trend is not monotonic.15
130 State Politics & Policy Quarterly 13(1)

Table 5. Final Factor Loadings from Dynamic Social Capital Index

Factor 1 Uniqueness
Community organizational life
  Fraternal order 0.584 0.659
  Religious club 0.372 0.861
  Civic club 0.469 0.780
  Veteran club 0.589 0.654
  Country club 0.228 0.948
  Body of local government 0.495 0.755
Engagement in public affairs
  Voted federal, state, or local election 0.699 0.511
  Wrote to an editor of a magazine or newspaper 0.614 0.623
  Wrote or telephoned a radio or television station 0.386 0.851
  Wrote to an elected official about a matter of public business 0.688 0.527
  Wrote something that has been published 0.578 0.666
  Addressed a public meeting 0.787 0.380
  Visited an elected official to express a point of view 0.579 0.665
  Actively worked for a political party or candidate 0.482 0.768
  Engaged in fund-raising 0.735 0.460
  Voter turnout (presidential and midterm elections) 0.385 0.852
Community volunteerism
  NPOs per 1,000 population 0.562 0.685
  Generosity Index 0.401 0.839
  Contributed to public television 0.694 0.518
  Public television contributions (composite score) 0.566 0.680
  Actively worked as a volunteer (nonpolitical) 0.820 0.327
  Took an active part in local civic issue 0.760 0.423
Note: NPOs = nonprofit organizations.

Furthermore, as our measure is dynamic, the correlation between the two indices
varies by year. Column 2 in Table 7 presents these correlations over time, which range
from .62 (1988) to .82 (2003). Table 7 also examines the correlation between our index
and five other variables, providing an additional test of construct validity for our mea-
sure. Based on previous research (e.g., Putnam 2000), social capital should be nega-
tively related to all five variables. To provide a baseline, the first row provides the
correlation between Putnam’s index and each of the five variables. The first variable is
Daniel Elazar’s (1972) 7-point political culture index that measures states political
cultures in terms of how moralistic, individualistic, or traditionalistic they are. This
index ranges from 1 to 8, where 1 represents moralistic states and 8 represents
traditional-moralistic states. Unlike the others measures, however, the political culture
measure is time invariant; thus, states are coded the same for all years. Theoretically,
traditionalistic states should fare worse in terms of social capital, whereas moralistic
Hawes et al. 131

Table 6. Correlations between Dynamic Items and Putnam’s Social Capital Indices
Putnam’s index

Dynamic social All


Correlation with capital index years 1999

Community organizational life


  Fraternal order .622 .409 .372
  Religious club .197 .251 .362
  Civic club .375 .237 .177
  Veteran club .485 .480 .631
  Country club .316 .058 .052
  Body of local government .484 .364 .216
Engagement in public affairs
  Voted federal, state, or local election .527 .384 .505
  Wrote to an editor of a magazine or newspaper .699 .355 .432
  Wrote or telephoned a radio or television station .577 .186 .186
  Wrote to an elected official about a matter of public business .783 .405 .227
  Wrote something that has been published .608 .364 .511
  Addressed a public meeting .828 .422 .644
  Visited an elected official to express a point of view .681 .360 .464
  Actively worked for a political party or candidate .534 .134 .271
  Engaged in fund-raising .660 .333 .470
  Voter turnout (presidential and midterm elections) .414 .608 .796
Community volunteerism
  NPOs per 1,000 population .562 .729 .769
  Generosity Index .401 .220 .204
  Contributed to public television .659 .323 .099
  Public television contributions (composite score) .582 .272 .142
  Actively worked as a volunteer (nonpolitical) .765 .598 .566
  Took an active part in local civic issue .773 .401 .419

Note: NPOs = nonprofit organizations.

states—which emphasize communitarian values—should be characterized by rela-


tively high levels of social capital. Thus, we expect our index of social capital to be
negatively correlated with Elazar’s political culture index. Next, we examine the rela-
tionship between social capital and heavy television viewership, a relationship we
expect to be negative (Putnam 2000, 303). We also examine two measures of crime:
the murder rate and the prison population as a percentage of a state’s total adult popu-
lation. The final column presents the correlations between social capital and the Kids
Count Index—a multi-indicator index on child well-being used by Putnam (2000,
297–98). This index includes indicators on health, poverty, criminal activity, and edu-
cation for children and teens (see appendix). As expected, our index is negatively and
strongly correlated with all five variables for all years.
Changes in spatial variation. As our social capital measure varies over time, another way
we can compare Putnam’s measure with our data is to average the values within each state
over the 19 years. Indeed, this is the technique Putnam used in creating his comprehen-
sive state-level index. To do this, we average the values of the individual items for
132 State Politics & Policy Quarterly 13(1)

Figure 1. Nationally aggregated social capital over time

Table 7. Annual Correlations with Social Capital


Putnam’s Political Heavy television Murder Prison Kids Count
Year index culture viewership rate population Index

Correlation with Putnam: −.769 −.388 −.720 −.319 −.746


1986 .623 −.661 −.262 −.397 −.414 —
1987 .621 −.657 −.137 −.480 −.406 —
1988 .620 −.650 −.073 −.487 −.426 —
1989 .680 −.676 −.450 −.649 −.495 —
1990 .700 −.711 −.357 −.689 −.518 −.635
1991 .774 −.772 −.470 −.731 −.612 —
1992 .799 −.793 −.299 −.709 −.654 —
1993 .768 −.802 −.329 −.673 −.609 —
1994 .757 −.815 −.756 −.632 −.583 −.717
1995 .717 −.817 −.713 −.662 −.607 —
1996 .710 −.820 −.486 −.631 −.601 −.718
1997 .763 −.816 −.584 −.710 −.634 −.766
1998 .788 −.797 −.536 −.732 −.579 −.707
1999 .806 −.783 −.615 −.727 −.603 −.713
2000 .727 −.693 −.585 −.671 −.531 −.651
2001 .746 −.708 −.473 −.738 −.628 −.797
2002 .736 −.682 −.707 −.671 −.567 −.772
2003 .818 −.706 −.632 −.712 −.606 −.775
2004 .765 −.729 −.683 −.710 −.589 −.735
Overall .677 −.696 −.442 −.505 −.535 −.677
correlation

Note: See appendix for data sources.


Hawes et al. 133

Figure 2. Correlation between averaged social capital and Putnam’s index


Note: Overall correlation = .80.

each state over all available years. This provides us a single value for each item for
each state. We then factor analyze these items, using the same process described
above.16 The overall correlation between this averaged measure of social capital and
Putnam’s measure is .824. Figure 2 presents the relationship between the two indices.
Our dynamic index, however, gives us the ability to examine state-level changes in
social capital over time. As noted above, Putnam makes a persuasive case that social
capital has been declining in the United States. Although this does appear in the trend
at a national level, as indicated by Figure 1, this is not universally the case at the state
level. In all, 10 states had higher levels of social capital in 2004 than in 1986.17 Thus,
although social capital may generally be declining in the United States as a whole, that
is not necessarily the case for all states. Figure 3 presents an example of states that
appear to have witnessed significant changes in social capital over time, both posi-
tively and negatively. The first graph in Figure 3 depicts apparent declines in 3 states:
Oregon, Arizona, and Idaho. Alternatively, the second graph shows that South
Carolina, Virginia, and Mississippi all seem to have seen increases in social capital,
particularly since the mid-1990s.18

Discussion
Although we believe our dynamic measure of social capital will prove to be a useful
contribution to the study of state politics, it is not without its flaws. Perhaps the most
134 State Politics & Policy Quarterly 13(1)

Figure 3. Divergent state trends in social capital

problematic aspect of our measure is the state clusters in the MRI data, the primary
source of data for our measure. Although most of the clusters are likely not problematic,
the New Mexico/Arizona/Utah/Nevada cluster poses some potential problems. Indeed,
the largest outlier in the correlation between the dynamic and Putnam indices is Nevada
(e.g., see Figure 1). We believe the clustering of these states attenuates Utah’s scores and
inflates Nevada’s scores. One potential remedy is to add additional non-MRI data or
weigh the non-MRI data more heavily, at least for these states. Although a better option,
the former is also more formidable as it is difficult to find state-level data over time.
Thus, for now, caution should be used when using these scores, particularly for these
problematic states.
Despite its flaws, this measure offers potential in addressing questions related to
causes and effects of social capital. This measure also allows researchers to explore
state-specific dynamics. As noted above, trends in social capital do not appear to be
uniform across the states. This measure can help shed light on what factors (e.g., eco-
nomic, immigration, diversity) contribute to the decline (or enhancement) of social
capital at the state level. We hope that scholars will use this measure to more fully
explore the effects of social capital on public policy, an area where earlier research has
found some startling correlations. However, these relationships have only been tested
using cross-sectional data, which makes such causal statements problematic. Our hope
is that this project will allow researchers to better assess these and other questions.
Hawes et al. 135

Appendix
Data Sources for Table 7

  Putnam’s index: www.bowlingalone.com


  Political culture: Moralistic, traditionalistic, individualistic,
  Television viewership: Percentage of population in top quintile of TV viewing
(MediaMark Research, Inc. [MRI] data)
  Murder rate:  Practical Researcher Data Set (State Politics & Policy
Quarterly (SPPQ),
  http://academic.udayton.edu/SPPQ-TPR/index.htm)
  Prison population: Total prison population divided by adult population
  (Department of Justice Bureau, Office of Justice Statistics,
   http://bjs.ojp.usdoj.gov/index.cfm)
  Kids Count Index: Annie E. Casey Foundation, KIDS COUNT Data book
  (http://datacenter.kidscount.org/)
Items in Kids Count Index
Percentage of low-birth-weight babies
Infant mortality rate (deaths per 1,000 live births)
Child deaths (deaths per 100,000 children ages 1–14)
Teen deaths per 100,000 from accident, homicide, and suicide (ages 15–19)
Teen birthrate (births per 1,000 females ages 15–17)
Teens ages 16 to 19 not in school and not high school graduates
Teens ages 16 to 19 not attending school and not working
Percentage of children living in families where no parent has full-time, year-round employment
Percentage of children in poverty
Percentage of children in single-parent families

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship,
and/or publication of this article.

Funding
The authors received no financial support for the research, authorship, and/or publication of this
article.

Notes
  1. Specifically, Uslaner and Brown (2005) use General Social Survey (GSS) data to show that
states where respondents exhibit high levels of trust in others are characterized by higher voter
turnout, participation in political meetings, and an increased likelihood of petition signing.
136 State Politics & Policy Quarterly 13(1)

  2. In 2005, GfK Group AG purchased MediaMark Research, Inc. (MRI) and changed its name
to MediaMark Research & Intelligence, LLC.
  3. A more detailed account of the weighting and balancing procedures is available at www
.gfkmri.com
  4. Although MediaMark collects individual-level data, we were only able to purchase data in
its aggregated form due to funding limitations.
  5. A total of 31 states are grouped in 12 clusters.
  6. The primary exception is the Utah, Arizona, New Mexico, and Nevada cluster. This cluster
includes a state with an above-average social capital score (Utah) as well as the lowest
ranking state (Nevada) on Putnam’s scale. This may potentially inflate Nevada’s scores or
deflate Utah’s scores.
  7. Some items in our index are not available for every year. All items are available from 1995
to 2004; for the previous years, only the items that are available are used. Those scores then
replace the missing values in the overall measure.
  8. For nonelection years, we use the weighted average of the values from the previous and
succeeding elections.
  9. The Generosity Index is generated by subtracting a state’s charitable contribution rank from
their adjusted gross income rank (i.e., Adjusted Gross Income (AGI) rank − Itemized Chari-
table Contributions (ICC) rank = Gap; Rank of Gap = Generosity Index).
10. The alpha for the unstandardized items is .78. The alpha coefficients for Putnam’s index are
.79 (unstandardized) and .95 (standardized). If we follow Putnam in averaging our items
across all the years (i.e., create a static cross-sectional index), the alphas increase to .80
(unstandardized) and .96 (standardized).
11. We use orthogonal factor rotation as our ultimate goal is to extract a single factor. That said,
the individual items load on the same factors when oblique rotation is used.
12. All 22 items are analyzed in the full analysis, which includes data from 1995 to 2004. For
the previous years, all items that are available are used (see Table 3). The final measure,
then, combines the full factor (1995–2004) with the factor scores from previous years.
Thus, a total of four factor scores were created and combined.
13. For each state, we include values for the two previous and succeeding values. Formally, the
weighting formula is as follows: (1/11) × [1 × SC(t−2) + 2 × SC(t−1) + 5 × SC(t) + 2 × SC(t+1)
+ 1 × SC(t+2)], where SC is the social capital and t is the year.
14. We understand that there is considerable debate as to whether Putnam’s index of social capi-
tal is an appropriate measure. Although we agree that his operationalization of social capital
has problems, our goal for this project is to provide an overtime measure of social capital as
conceptualized and operationalized by Putnam. We are not arguing that this is the only or
best way to measure social capital.
15. We should note that our aggregate data do not allow us to say with certainty that this trend
exists for all individuals. A decline in social capital could be the result of all individuals
becoming less involved or it may be that some individuals are somewhat more involved
while others are much less involved. Only individual-level data would allow researchers to
understand temporal variation in such detail.
Hawes et al. 137

16. The first factor has an eigenvalue of 13.21 and accounts for more than 60% of the variation
in the items.
17. The 10 states are Georgia, New York, Connecticut, Rhode Island, Alabama, Mississippi,
New Jersey, Virginia, South Carolina, and North Carolina.
18. We should caution that these examples are not representative of all states and are states that
display the most significant change. Furthermore, the trends in these data are not neces-
sarily systematic. Stationarity tests indicate that some of the state temporal variation may
actually be random walks, although this varies by state and test specification. Hence, we are
not claiming that social capital has increased/declined in any particular state, per se. Rather,
future research should explore these state-level temporal dynamics in more detail.

References
Brace, Paul, Kellie Sims-Butler, Kevin Arceneaux, and Martin Johnson. 2002. “Public Opinion
in the American States: New Perspectives Using National Data.” American Journal of Politi-
cal Science 46:173–89.
Brehm, John, and Wendy M. Rahn. 1997. “Individual-Level Evidence for the Causes and Con-
sequences of Social Capital.” American Journal of Political Science 41:999–1023.
Callahan, William A. 2005. “Social Capital and Corruption: Vote Buying and the Politics of
Reform in Thailand.” Perspectives on Politics 3:495–508.
Carden, Art, Charles Courtemanche, and Jeremy Meiners. 2009. “Does Wal-Mart Reduce Social
Capital?” Public Choice 138:109–36.
Claibourn, Michele P., and Paul S. Martin. 2007. “The Third Face of Social Capital: How
Membership in Voluntary Associations Improves Policy Accountability.” Political Research
Quarterly 60:192–201.
DeVellis, Robert F. 2003. Scale Development: Theory and Applications. 2nd ed. Thousand
Oaks, CA: Sage.
Elazar, Daniel. 1972. American Federalism: A View from the States. New York: Crowell.
Hawes, Daniel P., and Rene R. Rocha. 2011. “Culture, Capital, and Diversity: Assessing the
Determinants of Civic and Economic Equity in the American States.” Political Research
Quarterly 64:924–37.
Hero, Rodney E. 2003. “Multiple Theoretical Traditions in American Politics and Racial Policy
Inequality.” Political Research Quarterly 56:401–408.
Hero, Rodney E. 2007. Racial Diversity and Social Capital. New York: Cambridge University
Press.
Keele, Luke. 2005. “Macro Measures and Mechanics of Social Capital.” Political Analysis
13:139–56.
Knack, Stephen. 2002. “Social Capital and the Quality of Government: Evidence from the
States.” American Journal of Political Science 46:772–85.
Knack, Stephen, and Philip Keefer. 1997. “Does Social Capital Have an Economic Payoff? A
Cross-Country Investigation.” Quarterly Journal of Economics 112:1251–88.
Letki, Natalia. 2006. “Investigating the Roots of Civic Morality: Trust, Social Capital, and Insti-
tutional Performance.” Political Behavior 28:305–25.
138 State Politics & Policy Quarterly 13(1)

Paxton, Pamela. 1999. “Is Social Capital Declining in the United States? A Multiple Indicator
Assessment.” American Journal of Sociology 105:88–127.
Putnam, Robert D. 1993. Making Democracy Work: Civic Traditions in Modern Italy. Princeton,
NJ: Princeton University Press.
Putnam, Robert D. 2000. Bowling Alone: The Collapse and Revival of American Community.
New York: Simon & Schuster.
Rice, Tom. 2001. “Social Capital and Government Performance in Iowa Communities.” Journal
of Urban Affairs 23:375–90.
Stolle, Dietlind, and Thomas Rochon. 1998. “Are All Associations Alike? Member Diversity,
Associational Type and the Creation of Social Capital.” American Behavioral Scientist
42:47–65.
Tavits, Margit. 2006. “Making Democracy Work More? Exploring the Linkage between Social
Capital and Government Performance.” Political Research Quarterly 59:211–25.
Uslaner, Eric M. 2004. “Trust and Social Bonds: Faith in Others and Policy Outcomes Recon-
sidered.” Political Research Quarterly 57:501–507.
Uslaner, Eric M., and Mitchell Brown. 2005. “Inequality, Trust, and Civic Engagement.” American
Politics Research 33:868–94.
Warde, Alan, Gindo Tampubolon, Brian Longhurst, Kathryn Ray, Mike Savage, and Mark
Tomlinson. 2003. “Trends in Social Capital: Membership Associations in Great Britain,
1991–98.” British Journal of Political Science 33:515–34.

Bios
Daniel Hawes is Assistant Professor of Political Science at Kent State University. His research
interests include public administration, public policy, and race and politics.

Rene Rocha is Associate Professor of Political Science at the University of Iowa. His research
interests include race and politics, immigration, and representation.

Kenneth Meier is the Charles H. Gregory Chair in Liberal Arts at Texas A&M University. His
research interests include public management, public administration, and swine aviation.

You might also like