Available online at www.sciencedirect.

com

Procedia - Social and Behavioral Sciences 72 (2013) 73 90

European Framework for Measuring Progress (E-Frame): Proceedings of the Expert Meeting
on Social Capital

Social capital in 2009: An index for the Netherlands

Jacqueline van Beuningena,*, Hans Schmeetsa,b
a
Statistics Netherlands, CBS-weg 11, 6412 EX, Heerlen, The Netherlands
b
Department of Political Science, Maastricht University, Grote Gracht, 90-92,6211 SZ, Maastricht, The Netherlands

Abstract

This paper integrates a set of independent social capital indicators into one index by using structural equation modeling
(SEM) based on partial least squares estimation (PLS). The social capital index consists of two dimensions: participation and
trust. In each of the two dimensions, three levels are distinguished: social (micro), organizational (meso), and political
(macro). The main objectives of the index are to: (1) provide a coherent overview of social capital in the Netherlands; (2)
monitor social capital over time; and (3) compare subpopulations.
A broad spectrum of indicators is included, however, these are only weakly correlated and consequently treated as distinct.
Therefore, traditional index development methods such as factor analysis and reflective modeling cannot be applied.
Consequently, formative modeling in which the indicators are specified as causes rather than as effects is used. We employ
the Permanent Survey on Living Conditions 2009 (POLS), administered face-to-face to 7,560 respondents in the Netherlands.
We find that the index predicts well-being and health, which demonstrates nomological validity. Subsequently, bootstrapping
is conducted to test the robustness of the index.

2013
2013 The
The Authors.
Authors. Published
Published by
by Elsevier
Elsevier Ltd.
Ltd. Open access under CC BY-NC-ND license.
Selection and/or peer-review under responsibility of Hans Schmeets/Rik Linssen, Department of Political Science, Faculty
Selection and/or
of Arts and peer-review
Social under responsibility
Sciences, Maastricht ofthe
University, Hans Schmeets / Rik Linssen, Department of Political Science, Faculty
Netherlands.
of Arts and Social Sciences, Maastricht University, the Netherlands

Keywords: social capital index; structural equation modeling; formative indicators; partial least squares

* Corresponding author. Tel.: +31 45 570 6539

E-mail address: jwa.vanbeuningen@cbs.nl

1877-0428 2013 The Authors. Published by Elsevier Ltd. Open access under CC BY-NC-ND license.
Selection and/or peer-review under responsibility of Hans Schmeets/Rik Linssen, Department of Political Science, Faculty
of Arts and Social Sciences, Maastricht University, the Netherlands.
doi:10.1016/j.sbspro.2013.02.007
74 Jacqueline van Beuningen and Hans Schmeets / Procedia - Social and Behavioral Sciences 72 (2013) 73 90

1. Introduction

Various national governments and the EU as a whole have been interested in measuring the social capital of
their citizens. Concerns about recent developments in society, such as decreasing solidarity, increasing
individualism, and social isolation have all contributed to the popularity of this subject (Chan, To, and Chan
2006; Forrest and Kearns 2001; Jeannotte 2000). It has been proposed that when traditional relationships are
under pressure and peoples trust in others wanes, social capital in a country should decline. This could, in turn,
result in increased violence and detachment (Ultee, Arts, and Flap 1996). Therefore, monitoring social capital is
relevant. Until now Statistics Netherlands has monitored various social capital aspects over the last 15 years
(Schmeets 2010). Some of these aspects develop differently over time, which inhibits overall conclusions about
the possible decline or rise of social capital in the Netherlands. Therefore, this paper describes the development
of an index that synergizes a range of social capital indicators in an effort to monitor capital over time and help
policy makers identify subgroups that lag behind.

Social capital has been referred to as the resources the individual has access to by using their connections
(Paxton 1999; Putnam 1995), while others have focused primarily on social norms and obligations (Coleman
1988). We refer to Neira, Vzquez, and Portela (2009) for an overview of social capital definitions. We do not
propose a single definition, but want to bring the key social capital components together into one index. Trust and
participation have been identified as key components of social capital although studies often include only one of
the two as social capital proxy (Brehm and Rahn 1997; Coleman 1990; Guillen, Coromina, and Saris 2011; Neira
et al. 2009).

Paxton (1999) points at two problems related to the measurement of social capital: (a) the distinction between
indicators and outcomes of social capital is blurred; (b) most studies have used single indicators for the two key
components trust and participation where multiple indicators should be used to capture the multidimensional
social capital concept correctly. Finally, the concept social capital is used to describe populations and refers to a
community property, while it is typically measured at the individual level (Brehm and Rahn 1997; Guillen et al.
2011). Thus, the level of analysis does not correspond with the social capital conceptualization. In this paper, the
focus is on the second measurement issue of social capital, namely developing a more complete measure of social
capital.

Previous research has found that the two key components of social capital, participation and trust, are not
highly related (Newton 2001; Smerli and Newton 2008; Guillen et al. 2011). This complicates the development
of an index that integrates both components. Specifically, traditional analyses methods such as confirmatory
factor analyses are based on maximizing correlations. Without high correlations between indicators these
methods are not useful. Therefore, we specify indicators as causes of the dimensions rather than as effects. This
distinction is known as formative versus reflective modeling (Edwards and Bagozzi 2000; Jarvis, Mackenzie, and
Podsakoff 2003). Putnam (2000, p. 137) has claimed the causal relationships between indicators and dimensions
of social capital are as tangled as well-tossed spaghetti. On the one hand, trust should increase participation,
because it increases connectedness. On the other hand, participation in organizations and politics can increase
institutional trust. Therefore, there seems to be a recursive relationship (Brehm and Rahn 1997). Because we are
not interested in the relationship between participation and trust but in the effect of these dimensions combined
on social capital, we treat participation and trust as separate components of social capital.
 Jacqueline van Beuningen and Hans Schmeets / Procedia - Social and Behavioral Sciences 72 (2013) 73 90 75

The main contribution of this paper is an empirical illustration of the development of a social capital index.
First, we develop an index by determining indicator weights empirically in a predominantly formative
measurement model. The index scores for various subgroups are also reported. Second, we test the validity and
stability of the index.

2. Previous social capital conceptualizations and measurement

Several social capital definitions can be found in the literature (e.g. Bernard 1999; De Hart et al. 2002). These
include various levels of participation that generate a feeling of community: social (informal social relations),
civic (in organizations) and political participation (on a national level). The concept of social capital is also
defined as social networks and trust often referring to the work of Bourdieu (1983), Coleman (1988; 1990) and
Putnam (1995; 2000). The description adopted by the OECD is as follows: networks together with shared
norms, values and understandings that facilitate co-operation within or among groups, which has been widely
approved by various National Statistical Institutes (e.g. Australian Bureau of Statistics 2004; Beauvais and
Jenson 2002; Berger-Schmitt 2002; Bernard 1999; Coutts et al. 2007; Harper and Kelly 2003; Iisakka 2006;
Jeannotte 2000; Spellerberg 2001). In the presented definitions various forms of participation and trust play a
central role. Even though levels of participation and trust can be aggregated to represent a community property,
these can only be measured at the individual level.

Some social capital research is based on micro (individual) data (e.g. Paxton 1999), whereas most research
reports national capital averages (e.g. Putnam 2000). Other research has included aggregated scores of separate
social capital indicators that have been identified by factor analysis (e.g., Hooghe and Vanhoutte 2011; Nieminen
et al. 2008) or have used single-item proxies of social capital (e.g., Neira et al. 2009). However, only few studies
have developed overall social capital measures using multiple indicators (e.g. Paxton 1999; Guillen et al. 2011).
Although these approaches are valid, all rely on a priori determination of indicator weights. This is feasible when
the importance of social capital aspects is known. In contrast, empirically determining weights does not impose
the indicator weights, which ensures the social capital specification reflects the actual importance of indicators as
perceived by respondents.

3. Content of the social capital index: participation and trust

Considering the remarks discussed in paragraph 2 as to the concept and measurement of social capital, we
employ an index of social capital on a micro level. The concept is based on a framework of social capital which
includes the two main dimensions of social capital participation and trust (Schmeets and Te Riele 2010). The
index consists of six subdimensions that are part of the two main dimensions. Both dimensions refer to three
levels (for participation see also Eliasoph 1998; Van der Meer 2009): social (micro, the individual),
organizational (meso: the organizations), and political (macro: in the sphere of the state) (Figure 1).
Consequently, the two main dimensions refer to the following subdimensions: social participation, organizational
participation, and political participation (participation) and social trust, organizational trust, and political trust
(trust). Next, the subdimensions are discussed in more detail.

Social participation is conceptualized as having social contacts with relevant others. Social contacts facilitate
information flows and the extent to which people can receive or give support and information (Flap and Vlker
2004).
76 Jacqueline van Beuningen and Hans Schmeets / Procedia - Social and Behavioral Sciences 72 (2013) 73 90

Organizational participation is concerned with memberships of organizations and the attendance of events,
but also with participation in labor or education. Here too, people can maintain their social contacts and meet new
people as well as contribute to society.

Besides these, Putnam (1995) treats political participation as part of social capital. We include this type of
participation in the index, because political involvement is important for governing a country and consolidating
the democracy by increasing tolerance, seeking compromise, and training future leaders (Paxton 2002). In
addition, civilians can change governmental policies by being politically active. Voting, memberships of political
parties or taking political action are all part of political participation. Although this may not build social capital
directly, participating in and developing a countys democratic system creates boundary conditions within which
civilians have freedom to make their own choices. Therefore, voting is not conceptualized as an outcome of
social capital, but as part of it. A strong democratic system, in turn, facilitates the formation of voluntary
associations (Paxton 2002). These relationships or memberships are only useful if they are trusting and positive
(Paxton 1999).

The second component necessary to create generalized reciprocity within a society, is trust. Trust can exist at
various levels: between individuals termed specific trust, from the individual to others in general termed
generalized trust, or from the individual to an organization termed institutional trust. However, since we are
interested in trust in general in society, specific trust is in line with Paxton (1999) excluded from the index.
Generalized trust is a condition for forming positive, reciprocal ties with others and increases the willingness
to act in favor of the community, which is important for building social capital (Putnam, Leonardi, and Nanetti
1993). In principle this type of trust refers to trust in the average person. This is included in the index and
referred to as social trust. Organizational trust refers to trust in general institutions such as the police,
jurisprudence, and the press. Trust in organizations is necessary for an aggregated social capital measure (Paxton
1999). This trust may vary due to an individuals membership of an organization. Additionally, political trust
refers to political institutions in particular, such as the Dutch parliament or coalition. This ensures that civilians
can function effectively in society. For example, trust in the government is necessary for people to pay taxes,
which can then be reallocated to create a more equal dispersion of wealth.

The index consists of both behavioral as well as attitudinal indicators (Appendix A). Social participation was
measured as the frequency of contacts with family, friends, and neighbors. Participation in organizations was
measured as active participation in club activities at least once a month and having a paid job for at least 12 hours
a week. Political participation was operationalized as voting during the previous national elections and executing
political actions in the last five years. Social trust was measured by generalized trust by asking whether the
respondent trusts others. Institutional trust was measured as trust in the army, police, judges, civil servants, press
and large companies. Finally, political trust was measured as trust in the Dutch parliament. The various social as
well as organizational participation items were measured since 1997. The political participation items are based
on the Dutch Parliamentary Election Study (DPES). Likewise, the trust items are based on the DPES as well. The
correlation matrix with the indicators can be found in Appendix B. It is notable that correlations between the
indicators range from -0.08 to 0.50, which is rather low. Kowalski (1972) notes that correlations can be biased
due to non-normal distributions, which is the case for most indicators in this study. Thus, the correlation
coefficient may not be an informative statistic of association. However, this non-normality does not violate the
models assumptions, since the PLS method is able to cope with highly skewed variables.
 Jacqueline van Beuningen and Hans Schmeets / Procedia - Social and Behavioral Sciences 72 (2013) 73 90 77

Social Participation

Organizational
Participation Participation

Political
Participation
Social Capital
Social
Trust

Organizational
Trust Trust

Political Trust

Figure 1. Conceptual model of social capital index including two main dimensions and six subdimensions

4. Data

We use the Permanent Survey on Living Conditions 2009 (POLS 2009), administered in the Netherlands to
7,560 respondents. The data-collection mode is computer assisted personal interviewing (CAPI). The response
rate is 61%. The data collection was conducted from January 1st to December 31st. Statistics Netherlands (SN)
started POLS in 1997 as the survey to collect data on a broad spectrum of quality of life indicators, including
health, well-being, safety issues, and social participation. In 2009 a module was implemented to serve the
research programme on social cohesion. Various trust items were included as well as political participation
indicators. Each interview lasted approximately 45 minutes.

4.1. Index development

A hierarchical model with two main dimensions and six subdimensions related to the three main themes was
constructed. Dimensions are treated as unobserved or latent constructs. The dimensions consist of one or more
indicators. The direction of causality in the relationships between indicators and dimensions can differ (Blalock
1964; Bollen and Lennox 1991; Diamantopoulos and Winklhofer 2001; Edwards and Bagozzi 2000; Fayers and
Hand 2002). Indicators can form the construct, therefore, the direction of causality is from the indicator to the
construct (i.e., formative or causal). Alternatively, the construct can determine the indicators in which case the
direction of causality is from the construct to the indicator (i.e., reflective or effect). In Structural Equation
Modeling the standard is reflective modeling, although the direction of a relationship should be determined
conceptually first. There are two criteria formulated by Jarvis et al. (2003) to determine whether a relationship
78 Jacqueline van Beuningen and Hans Schmeets / Procedia - Social and Behavioral Sciences 72 (2013) 73 90

should be modeled in a formatively as opposed to reflectively: (1) there is no reason to expect the indicators will
be correlated and (2) excluding an indicator may alter the meaning of the dimension.

Using these two criteria we can determine the direction of the relationship of the indicators with the six social
capital dimensions. Indicators within one dimension are not expected to correlate. For example, people who have
frequent contact with family members do not necessarily interact a lot with friends. If anything, high family
contact could result in low friend contact due to time constraints. The same goes for other behaviors, such as
participating in organizational activities, working, voting or taking political action. Further, trusting one
institution does not imply trusting another. As Paxton (1999, p. 117) notes: That is, scandals in a particular year,
related to a particular institution, could drive responses about trust in that institution in that year. This implies
that the trust rating of one institution does not necessarily relate to trust scores of other institutions. Second, all
indicators are key components of social capital, excluding one would change the interpretation of the concept.
One exception is the generalized social trust indicator of the social trust dimension. In this case the indicator is a
direct representation of the construct and should be modeled reflectively.

Previous empirical research has modeled the social participation component of social capital formatively
(Guillen et al. 2011; Paxton 1999). We extend this research by modeling both participation and trust formatively,
incorporating a broader set of indicators, and determining indicator weights empirically. We do not use
summated scales, because we do not assume that all indicators are equally important.

The hierarchical model or path model is composed using Structural Equation Modeling (Chin 1998). This
way, collinearity between dimensions is accounted for. A partial least squares model (PLS), which is a
component-based estimation method, with mostly formative indicators was created (Ringle et al. 2009). This
way, the linear combination of the indicators can be maximized to construct the index instead of maximizing
covariances between indicators. The exact linear combination of indicators, or weights, is used to maximize the
explained variance of the respective construct. These constructs, in this case the six subdimensions, consist of
weighted averages of the indicators and are determined empirically. There are several advantages to using this
estimation method instead of the traditionally used Maximum Likelihood estimation (ML) (Chin 1998; Ringle et
al. 2009):

1. PLS has less strict distributional assumptions than ML. Therefore, PLS can cope with non-normally
distributed data and multicollinearity between indicators. In the case of social capital, most indicators, such
as those referring to social contacts and institutional trust, are highly skewed.
2. Measurement models can be formative, reflective, or both in PLS, whereas ML only selects one of the two.
Because the social capital index includes both types of directional relationships with the social trust
indicator being reflective, PLS is appropriate.
3. PLS is suggested for explorative studies aimed at making predictions. In this case the selection of
indicators and relationships between indicators and dimensions has not been fully explored. Therefore, a
confirmative analysis strategy using ML would not be advisable.

Four analytical steps have been taken to prepare the dataset for estimation. First, all indicators were recoded if
necessary to reflect more participation or trust with a higher score. Dichotomous indicators have been dummy
coded, where 0 reflects No or Low and 1 reflects Yes or High. Second, interval variables have been
tested to check whether these can be treated as continuous variables. For each variable the answer categories
were included in the model as dummy variables (number of dummies = number of categories - 1). The linear
effects of the coefficients of the categories were checked against the category ranking. This is the case for all
interval variables except for political trust. Because of model consistency and parsimony and the fact that
differences are very small, all interval-scaled variables are treated as continuous variables instead of a set of
 Jacqueline van Beuningen and Hans Schmeets / Procedia - Social and Behavioral Sciences 72 (2013) 73 90 79

dummies. These variables have been standardized to rule out any effects of different scaling. Third, outliers have
been inspected by means of the Mahalanobis distance statistic. This identifies multivariate outliers and can show
diverging answer patterns. In an alternative model extreme outliers above the critical value were excluded to
check if the outliers affect results (D2(15) > 37,70; N = 6,115 instead of N = 7,560). The weights of the
indicators in the models with and without outliers do not differ significantly (t(28) = 0,93; ns). Therefore, outliers
were not excluded. Mean imputation was used to ensure all observations could be included, since no instrumental
information was available for imputation. Imputation is also important because missing data could result in
indicator weight instability, which leads to misspecification of the model (Cenfetelli and Bassellier 2009). This
imputation method does not affect the size of the indicator weights.

The model was identified by Multiple Table Analysis, also called a hierarchical components model
(Tenenhaus 2004; Lohmller 1989). Additional reflective indicators are usually included to identify the model,
because scaling of the latent variable is necessary. However, in this case no reflective indicators were available,
because respondents cannot indicate their level of social capital themselves. Therefore, the indicators are not only
used to predict the subdimensions (i.e., primary loadings), but also the corresponding main dimension and the
overall social capital index construct (i.e., secondary loadings). Thus, the indicators are connected to several
constructs; e.g. family contacts affects social participation as well as participation and social capital as a whole.
This approach is used instead of adding outcome variables when estimating the indicator loadings, because for
example adding well-being would optimize the loadings towards maximizing well-being. The available outcome
variables are used as a validity check after estimation of the initial model.

The model is estimated in R using the PLSPM package. A path weighting scheme was used so weights can be
interpreted as regression coefficients. The initial model was estimated with an iterative algorithm with 100
iterations during which weights were optimized. Subsequently, the initial model was used as the input of the
bootstrapping analysis. Bootstrapping was employed to make sure weights are stable. Figure 2 shows the
resulting structural model with the estimated indicator and dimension weights. The weights of the indicators on
the main dimensions and social capital are left out of figure 2 to ensure readability.

The indicator weights of the six dimensions vary from 0.18 to 0.82, except for the two single item indicators
with 1.00 weights. All indicators and dimensions in the model contribute significantly to the index. The weights
can be used to show the relative contribution of each indicator or dimension. Some weights seem low, but this
does not imply that these indicators are irrelevant. Weights in formative measurement models are often lower
than reflective factor loadings. Weights do not necessarily explain high variance to contribute to the meaning of
the index (Gtz, Liehr-Gobbers and Kraft 2010). In addition, weights are calculated from the first-order on the
two second-order constructs participation and trust. These, in turn, contribute independently to the social capital
index. All six subdimensions are relevant as they contribute significantly to the index. Further, it is shown that
the three weights of the subdimensions on participation only differ slightly indicating that all three domains
contribute equally to participation. Likewise, the weights on trust were calculated. The results demonstrate that in
particular organizational trust in army, police, judges, civil servants, press and big companies contributes to
the trust score. The two other aspects social and political trust, both measured with a single indicator are less
important for the assessment of the trust construct. Furthermore, the weights from the second-order to the third-
order construct show that trust contributes more to social capital than participation does. This implies that trust
outweighs participation when forming a social capital index for the Netherlands.
80 Jacqueline van Beuningen and Hans Schmeets / Procedia - Social and Behavioral Sciences 72 (2013) 73 90

4.2. Index validity and robustness

To evaluate the index we take several analytical steps. First, we examine the quality of the measurement
model by analyzing the Goodness-of-Fit statistic. Second, bootstrapping was used to examine the stability of the
indicator and dimension weights. Third, the quality of the structural model was checked by investigating the
nomological validity of the model by including two well-known outcomes of social capital, namely well-being
and health (Helliwell and Putnam 2004; Kawachi, Kennedy, and Glass 1999; Lindn-Bostrm, Persson, and
Eriksson 2010). Finally, we test whether alternative dimensional structures perform worse than the chosen
structure in terms of model fit.

4.2.1. Index evaluation

Internal consistency reliability is not an appropriate criterion to evaluate formative constructs. For formative
measurement models there are only a limited number of fit statistics. One of these is the Goodness-of-Fit reported
(GoF) which indicates how well the model describes the data. This statistic is different from fit statistics for
reflective models in that it uses the square root of the average communality multiplied with the average explained
variance.

Indicators First-order Second-order Third-order

Family constructs constructs construct
contacts 0.41*
Friends 0.75* Social
contacts
Participation 0.50*
Neighbors
contacts 0.34*
Organizational 0.82* Organizational
activities
Participation
0.47*
Participation
Work 0.64*

Voting 0.67* Political 0.55*

0.48*
Political
0.66*
Participation
actions

Generalized 1.00*
trust Social Social
Trust in army 0.18*
Trust
0.37*
Capital
Trust in police 0.23*
0.76*
Trust in judges 0.42*
0.33* Organizational 0.67*
Trust in civil
servants Trust Trust
0.20*
Trust in press

Trust in large
companies 0.18*
0.24*

Trust in 1.00* Political

parliament Trust

* p < 0.05
 Jacqueline van Beuningen and Hans Schmeets / Procedia - Social and Behavioral Sciences 72 (2013) 73 90 81

Figure 2 Structural model of the social capital index (standardized data)

Therefore, it does not take the number of parameters in the model into account. The model has an absolute GoF
of 0.71 and a relative GoF of 0.49, which is higher than the threshold value of 0.36 for a good fit (Wetzels,
Odekerken-Schrder, and van Oppen 2009). Thus, we can conclude that the model represents the data
appropriately.

4.2.2. Index weights robustness

Next, the indicator weights are examined separately. Resampling is an appropriate strategy to validate a model
empirically and bootstrapping is a resampling with replacement method. The procedure selects samples
consisting of the same number of observations as the original sample (Tenenhaus 2005). The default 100 re-
samples of 100 cases were used. The bootstrap shows that none of the original weights differs significantly from
the average weights generated by the bootstrap, since all original weights are within the confidence intervals of
the bootstrapped average weights. This shows that the estimated weights are robust. The average weights and
95% confidence intervals can be found in Appendix D. The effects of the dimensions on the index, or the
structural model, also do not differ significantly from the bootstrapped effects.

4.2.3. Outcomes of social capital

Further, the social capital index affects both well-being ( = 0.18; p < 0.05) and health ( = 0.20; p < 0.05)
positively and significantly. The model with these outcomes controls for age, because this is an important
predictor of well-being and health. Furthermore, health and well-being are related so the effect of health on well-
being is included. The explained variance for both well-being and health (R2 = 0.13) is reasonable given the fact
that many other important predictors for well-being and health, such as income and education, have been
excluded. In addition, the model with outcome variables is tested using two indices (the scores on the
participation and trust dimensions) instead of one composite index. This shows that the absolute GoF for this
model (GoF = 0.29) is lower than the original model (GoF = 0.36).

5. Results for subpopulations

The index scores are standardized with a mean of 0 and a standard deviation of 1 making the group
classifications comparable. Index scores of groups are visualized in Figure 3. Based on previous research on
participation and trust from 1997 onwards we anticipate particularly low ratings on the index for low educated
people, elderly, non-western ethnic minorities, and people belonging to the Islam or the Catholic Church
(Coumans and Te Riele 2010; Kloosterman and Schmeets 2010; Schmeets 2010). Groups are distinguished on
the following characteristics: age, ethnicity, education, and denomination. Index scores are shown in Appendix
C. The group rankings are stable: when the index is re-estimated using only half of the sample that is randomly
selected the ranking of the groups remains the same. As expected, the low educated and non-western minorities
score lowest on the index compared to highly educated and Dutch natives. Overall, social capital decreases with
age for people of 45 or older. Protestants and Conservative reformed people also score higher than people
without denomination, Catholics, Islamites, and Dutch reformed.

The dimension scores of the aforementioned groups are analyzed to check the extent to which the index scores
are consistent with the underlying dimensions (Appendix C shows participation and trust scores). These rankings
are mostly consistent with the index rankings. However, there are some differences in rankings on participation
and trust. For example, non-western minorities have higher trust than Dutch natives and western minorities
although this is the opposite for the overall index. Furthermore, the youngest age group scores somewhat lower
than older age groups on organizational and political participation while scoring highest on the index
82 Jacqueline van Beuningen and Hans Schmeets / Procedia - Social and Behavioral Sciences 72 (2013) 73 90

-1 0 1

18-24
25-44

Age
45-64
65+

Native
Ethnicity
Western immigrant
Non-western immigrant

Higher
Vocational
Education

Secundary
Lower
Primary

Protestant
Conservative reformed
Denomination

None
Catholic
Islamic
Dutch reformed

Figure 3. Visual representation of the social capital index

 Jacqueline van Beuningen and Hans Schmeets / Procedia - Social and Behavioral Sciences 72 (2013) 73 90 83

6. Discussion and conclusion

This paper describes the development of a social capital index consisting of two main dimensions, namely
participation and trust. Advantages of a composite index include: the ability to summarize complex or
multidimensional issues, provide the big picture, help attract public interest, and help reduce the list of necessary
indicators by focusing on the key components (Saltelli 2007). Especially when varying aspects of the concept
contradict each other or develop differently, it can be useful to summarize these into an overall statistic. In this
case, the added value of the index is the combination of distinct aspects into one statistic because it provides a
way to draw overall conclusions about relative group performance. However, an index like this one is
complementary to the separate social capital components which should be included when index results are
discussed.

The index explains significant variance in both well-being and health. This corresponds with previous research
and is an indication of the validity of the index. Even though there could be other relevant aspects of social
capital, it demonstrates, at least partially, the validity of the index. In this index we find that trust is a relatively
more important part of social capital than participation.

The purpose of the index is to compare groups to show differences in social capital. Groups that score high on
the index are: adolescents, highly educated, Protestants, and conservative Reformed people. People who are 65 or
older or are uneducated have a relatively low score.

Indicator weights are context dependent (Hardin et al. 2011). Therefore, these weights cannot be generalized
to other countries. The aim of this paper is to show an empirical illustration of a social capital index. For theory
testing weights must be held constant across datasets by implementing fixed-weight composites. Different
samples would be necessary to determine the optimal indicator weights. However, we are confident that this set
of indicators and dimensions is conceptually meaningful in other settings as well.

Acknowledgements

The authors thank Saskia te Riele, Rianne Kloosterman, Godelief Mars, Koos Arts, Karolijne van der Houwen, Mark van der
Loo, and Abby Israels from Statistics Netherlands, and Martin Wetzels from Maastricht University for their help and advice.
The authors thank the reviewers and participants of the Maastricht expert meeting on social capital for their useful comments.

About the Authors

Jacqueline van Beuningen is statistical researcher at Statistics Netherlands in the fields of social cohesion, well-being, and
mobility.
Hans Schmeets is professor in 'social statistics' at the department of political science at Maastricht University, the
Netherlands, and programme manager at Statistics Netherlands in the fields of social cohesion, well-being, and public opinion
research.

References

Australian Bureau of Statistics (2004). Measuring social capital. An Australian framework and indicators. Canberra: ABS.
Arts, K., & Te Riele, S. (2010). Vrijwilligerswerk. In: Schmeets, H (Ed.), Sociale samenhang: Participatie, vertrouwen en
integratie, 53-70. Den Haag/Heerlen: CBS.
84 Jacqueline van Beuningen and Hans Schmeets / Procedia - Social and Behavioral Sciences 72 (2013) 73 90

Beauvais, C., & Jenson, J. (2002). Social cohesion: Updating the state of the research. Ottawa: Canadian Policy Research
Networks.
Berger-Schmitt, R. (2002). Considering social cohesion in quality of life assessments: Concept and measurement. Social
Indicators Research, 58, 403-428.
Bernard, P. (1999). Social cohesion: A critique. Ottawa: Canadian Policy Research Networks.
Blalock, H. M. (1964). Causal inferences in nonexperimental research. Chapel Hill: University of North Carolina Press.
Bollen, K., & Lennox, R. (1991). Conventional wisdom on measurement: A structural equation perspective. Psychological
Bulletin, 110 (2), 305-314.
Brehm, J., & Rahn, W. (1997). Individual-level evidence for the causes and consequences of social capital. American Journal
of Political Science, 41(3), 999-1023.
Cenfetelli, R. T., & Bassellier, G. (2009). Interpretation of formative measurement in informations systems research. MIS
Quarterly, 33(4), 689-707.
Chan, J., To, H.P., & Chan, E. (2006). Reconsidering social cohesion: Developing a definition and analytic framework for
empirical research. Social Indicators Research, 75, 273-302.
Chin, W. W. (1998). Issues and opinion on structural equation modeling. MIS Quarterly, 22(March), 7-16.
Cohen J., & Cohen, P. (1983). Applied multiple regression/correlation analysis for the behavioural sciences. Hillsdale, NJ:
Erlbaum.
Coleman, J. S. (1988). Social capital in the creation of human capital. American Journal of Sociology. 94, 95-120.
Coleman, J. S. (1990). Foundations of social theory. Cambridge: Belknap Press.
Cot, S., & Healy. T. (2001). The well-being of nations. The role of human and social capital. Paris: OECD.
Coumans, M., & Te Riele, S. (2010). Verschillen in sociale en maatschappelijke participatie. Bevolkingstrends (Quarter 1),
39-45.
Coutts, A., Ramos Pinto, P., Cave, B. & Kawachi, I. (2007). Social capital indicators in the UK. London: Commission for
Racial Equality.
De Hart, J., Knol, F., Maas-de Waal, C. & Roes, T. (2002). Zekere banden: Sociale cohesie, leefbaarheid en veiligheid. SCP:
Den Haag.
Diamantopoulos, A., & Winklhofer, H. M. (2001). Index construction with formative indicators: An alternative to scale
development. Journal of Marketing Research, 38(May), 290-277.
Edwards, J. R., & Bagozzi, R. P. (2000). On the nature and direction of relationships between constructs and measures.
Psychological Methods, 5(2), 155-174.
Eliasoph, L. (1998). Avoiding politics: How Americans produce apathy in everyday life. Cambridge: Cambridge University
Press.
Fayers, P. M., & Hand, D. J. (2002). Causal variables, indicator variables and measurement scales: An example from quality
of life. Journal of the Royal Statistical Society: Series A (Statistics in Society), 165(2), 233-253.
Flap, H. D., & Vlker, B. (2004). Social networks and performance at work. In: Flap, H. D., & Vlker, B.
(Eds.), Creation and returns of social capital (pp. 172-196). London: Routledge,
Forrest, R., & Kearns, A. (2001). Social cohesion, Social capital and the neighbourhood. Urban Studies, 38(12), 2125-2143.
Fukuyama, F. (2001). Social capital, civil society and development. Third World Quarterly, 22(1), 7-20.
Gtz, O., Liehr-Gobbers, K, & Kraft, M. (2010). Evaluation of structural equation models using the partial least squares
(PLS) Approach. In: Vinci et al. (Eds.), Handbook of partial least squares: Concepts, methods, and applications (pp.
691-712). Berlin: Springer.
Guillen, L., Coromina, L., & Saris, W. (2011). Measurement of social participation and its place in social capital theory.
Social Indicators Research, 100, 331-350.
Hardin, A. M., Chang, J. J. C. J., Fuller, M. A., & Torkzadeh, G. (2011). Formative measurement and academic research: In
search of measurement theory. Educational and Psychological Measurement, 71(April), 281-305.
Harper, R., & Kelly, M. (2003). Measuring social capital in the United Kingdom. London: Office for National Statistics.
Helliwell, J., & Putnam, R. D. (2004). The social context of well-being. Philosophical Transactions of the Royal Society of
London, 359, 1435-1446.
Hooghe, M., & Vanhoutte, B. (2011). Subjective well-being and social capital in Belgian communities: The impact of
community characteristics on subjective well-being indicators in Belgium. Social Indicators Research, 100, 17-36.
Iisakka, L. (ed.) (2006). Social capital in Finland. Statistical Review. Helsinki: Statistics Finland.
Jarvis, C. B., MacKenzie, S. B., & Podsakoff, P. M. (2003). A critical review of construct indicators and measurement model
misspecification in marketing and consumer research. The Journal of Consumer Research, 30(2), 199-218.
Jeannotte, S. (2000). Social cohesion around the world: An international comparison of definitions and issues. Hull: Strategic
Research and Analysis Directorate.
Kawachi, I., Kennedy, B. P., & Glass, R. (1999). Social capital and self-rated health: A contextual analysis. American Journal
of Public Health, 89(8), 1187-1193.
 Jacqueline van Beuningen and Hans Schmeets / Procedia - Social and Behavioral Sciences 72 (2013) 73 90 85

Kowalski, C. J. (1972). On the effects of non-normality on the distribution of the sample product-moment correlation
coefficient. Journal of the Royal Statistical Society, Series C (Applied Statistics), 21(1), 1-12.
Kloosterman, R., & Schmeets, H. (2010). Vertrouwen in medemens en instituties toegenomen aan het begin van de 21e eeuw.
Bevolkingstrends (Quarter 2), 43-49.
Lindn-Bostrm, M., Persson C., & Eriksson, C. (2010). Neighbourhood characteristics, social capital and self-rated health
A population-based survey in Sweden. BMC Public Health, 10(628), 1-15.
Lohmller, J.-B. (1989). Latent variable path modeling with partial least squares. New York: Springer-Verlag.
Neira, I., Vzquez, E., & Portela, M. (2009). An empirical analysis of social capital and economic growth in Europe (1980-
2000). Social Indicators Research, 92, 111-129.
Newton, K. (2001). Trust, social capital, civil society, and democracy. International Political Science Review, 22(2), 201-214.
Nieminen, T., Martelin, T., Koskinen, S., Simpura, J., Alanen, E., Hrknen, T.,, & Aromaa, A. (2008). Measurement and
socio-demographic variation of social capital in a large population-based survey. Social Indicators Research, 85, 405-
423.
Paxton, P. (1999). Is social capital declining in the United States? A multiple indicator assessment. American Journal of
Sociology, 105(1), 88-127.
Paxton, P. (2002). Social capital and democracy: An interdependent relationship. American Sociological Review, 67(2), 254-
277.
Portes, A. (1998). Social capital. Its origins and applications in modern sociology. Annual Review of Sociology, 24, 1-24.
Putnam, R., Leonardi, R., & Nanetti, R.Y. (1993). Making democracy work. Princeton: Princeton University Press.
Putnam, R. (1995). Tuning in, turning out: The strange disappearance of social capital in America. Political Science and
Politics, 28, 664683.
Putnam, R. (2000). Bowling alone. The collapse and revival of American community. New York: Simon & Schuster.
Rajulton, F., Ravanera, Z. R., & Beaujot, R. (2007). Measuring social cohesion: An experiment using the Canadian national
survey of giving, volunteering, and participating. Social Indicators Research, 80, 461-492.
Ringle, C. M., Gtz, O., Wetzels, M., & Wilson, B. (2009). On the use of formative measurement specifications in structural
equation modeling: A monte carlo simulation study to compare covariance-based and partial least squares model
estimation methodologies. Research Memoranda from Maastricht, 14, 1-41.
Saltelli, A. (2007). Composite indicators between analysis and advocacy, Social Indicators Research, 81, 65-77.
Schmeets, H. (2010). Sociale samenhang: Participatie, vertrouwen en integratie. Den Haag/Heerlen: CBS.
Schmeets, H. (2010). Religie en sociale samenhang. In: H. Schmeets (Ed.), Sociale samenhang: Participatie, vertrouwen en
integratie (pp. 165-182). Den Haag/Heerlen: CBS.
Schmeets, H., & Te Riele, S. (2010). Sociale Samenhang in het drieluik van participatie, vertrouwen en integratie. In:
Schmeets, h. (Ed.), Sociale Samenhang: Participatie, vertrouwen en integratie (pp. 7-16). Den Haag/Heerlen: CBS.
Schmeets, H., & Te Riele, S. (2012). Declining participation and trust in the Netherlands? Trends and correlations, 1989-
2010. Paper presented at the Authors Conference Declining Political Trust, Disentchantment with Politics and Methods
of Political Participation, 26-27 January 2012, Cologne, Germany.
Smerli, S., & Newton, K. (2008). Social trust and attitudes toward democracy. Public Opinion Quarterly, 72(4), 706-724.
Spellerberg, A. (2001). Framework for the measurement of social capital in New Zealand. Wellington: Statistics New
Zealand.
Tenenhaus, M. (2004). PLS regression and PLS path modeling for multiple table analysis. COMPSTAT, 1-12.
Tenenhaus, M., Vinzi, V. E., Chatelin, Y. M., & Lauro, C. (2005). PLS path modeling. Computational Statistics & Data
Analysis, 48, 159-205.
Van der Meer, T. (2009). States of freely associating citizens: Comparative studies into the impact of state institutions on
social, civic, and political participation. ICS-dissertation. Nijmegen: Radboud Universiteit Nijmegen.
Van der Houwen, K. (2010). Informele hulp en mantelzorg. In: Schmeets, H. (Ed.), Sociale samenhang: Participatie,
vertrouwen en integratie (pp.31-40). Den Haag/Heerlen: CBS.
Ultee, W., Arts, W., & Flap, H. (1996). Sociologie: Vragen, uitspraken, bevindingen. Groningen: Wolters-Noordhoff.
Vinzi, V. E., Trinchera, L., & Amato, S. (2010). PLS path modeling: From foundations to recent developments and open
issues for model assessment and improvement. In: Vinci et al. (Eds.), Handbook of partial least squares: Concepts,
methods, and applications (pp. 47-82). Berlin: Springer.
Wetzels, M., Odekerken-Schrder, G., & Van Oppen, C. (2009). Using PLS path modeling for assessing hierarchical
construct models: Guidelines and empirical illustration. MIS Quarterly, 33(March), 177-195.
86 Jacqueline van Beuningen and Hans Schmeets / Procedia - Social and Behavioral Sciences 72 (2013) 73 90

Appendix A.

Table A.1 Study measures

1. Social participation

Family contacts (Rarely or never At least once a week)

How often do you have contact with family members outside the home?

Contacts with friends (Rarely or never At least once a week)

How often do you have contact with friends or close acquaintances?

Contacts with neighbors (Rarely or never At least once a week)

How often do you have contact with neighbors?

2. Organizational participation

Organizational activities (0 = No, 1 = Yes)

Do you participate in organizational activities at least once a month?

Work (0 = Less than twelve hours, 1 = Twelve hours or more)

How many hours per week do you spend on a paid job?

3. Political participation

Political action (0 = No (9), 1 = Yes (1-8))

In the last five years, I have participated in political actions or influenced politicians/ civil servants via: (1) radio, tv, newspaper, (2) political
parties or organizations, (3) government organized meeting, (4) personal contact with politician or civil servant, (5) protest group, (6) public
protest or demonstration, (7) internet, e-mail, SMS, (8) otherwise or (9) none of these.

Voting (0 = No, 1 = Yes)

Have you voted in the most recent national elections?

4. Social trust

Generalized social trust (0 = One cant be too careful, 1 = Most people can be trusted)
Indicate which statement you agree with most: (1) most people can be trusted, (2) one cant be too careful.
 Jacqueline van Beuningen and Hans Schmeets / Procedia - Social and Behavioral Sciences 72 (2013) 73 90 87

Table A.1 Study measures (continued)

5. Organizational trust

Trust in army (1 = Not at all, 4 = A lot)

Indicate to which extent you trust the army.

Trust in judges (1 = Not at all, 4 = A lot)

Indicate to which extent you trust the judges.

Trust in civil servants (1 = Not at all, 4 = A lot)

Indicate to which extent you trust civil servants.

Trust in press (1 = Not at all, 4 = A lot)

Indicate to which extent you trust the press.

Trust in police (1 = Not at all, 4 = A lot)

Indicate to which extent you trust the police.

Trust in large companies (1 = Not at all, 4 = A lot)

Indicate to which extent you trust the large companies.

6. Political trust

Trust in parliament (1 = Not at all, 4 = A lot)

Indicate to which extent you trust the Dutch parliament.
 88
Appendix B.

Table B.1. Descriptives and correlation matrixa

Indicators M (SD) 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
1 Contacts with family 4.73 (0.76) 1
2 Contacts with friends 4.63 (0.87) 0.09 1
3 Contacts with neighbors 4.17 (1.36) 0.12 0.11 1
4 Organizational activities - 0.04 0.15 0.06 1
5 Work - 0.06 0.08 0.01 -0.08 1
6 Voting - 0.07 0.04 0.05 0.11 -0.01 1
7 Political actions - 0.01 0.07 0.02 0.07 0.16 0.13 1
8 Generalized social trust - 0.02 0.10 0.03 0.11 0.13 0.14 0.15 1
9 Trust in army 2.61 (0.83) 0.05 0.07 0.01 0.08 0.03 0.11 0.02 0.09 1
10 Trust in police 2.78 (0.74) 0.05 0.04 0.04 0.07 0.01 0.08 0.02 0.18 0.28 1
11 Trust in judges 2.76 (0.79) 0.03 0.09 0.00 0.07 0.14 0.11 0.11 0.26 0.31 0.42 1
12 Trust in civil servants 2.51 (0.72) 0.03 0.08 0.03 0.08 0.03 0.11 0.03 0.17 0.26 0.39 0.38 1
13 Trust in press 2.24 (0.71) 0.01 0.05 -0.00 0.03 0.01 0.06 0.03 0.15 0.18 0.26 0.28 0.25 1
14 Trust in large companies 2.47 (0.71) 0.05 0.09 -0.01 0.06 0.05 0.02 -0.01 0.11 0.25 0.23 0.25 0.30 0.25 1
15 Trust in parliament 2.34 (0.77) 0.02 0.10 0.01 0.11 0.05 0.15 0.07 0.23 0.29 0.41 0.44 0.50 0.31 0.30 1

a
No descriptives are available for dichotomous indicators. Correlations of 0.03 and higher are significant (p < 0.01).
Jacqueline van Beuningen and Hans Schmeets / Procedia - Social and Behavioral Sciences 72 (2013) 73 90
 Jacqueline van Beuningen and Hans Schmeets / Procedia - Social and Behavioral Sciences 72 (2013) 73 90 89

Appendix C.

Table C.1. Average group scores on index

Population Index (SD) Ranking Participation (SD) Trust (SD)

Age
12-24 years 0.22 (0.74) 1 -0.02 (0.71) 0.31 (0.80)
25-44 years 0.19 (0.98) 1 0.19 (0.98) 0.12 (0.98)
45-64 years -0.06 (1.04) 2 0.11 (1.03) -0.14 (1.04)
65 and older -0.41 (1.04) 3 -0.46 (1.06) -0.26 (1.03)
Ethnicity
Dutch native 0.03 (1.00) 1 0.05 (0.98) 0.00 (1.00)
Western minority -0.15 (1.07) 2 -0.25 (1.11) -0.03 (1.02)
Non-western minority -0.15 (0.93) 2 -0.35 (1.06) 0.02 (0.97)
Education
Higher professional, university 0.53 (0.83) 1 0.48 (0.89) 0.40 (0.86)
Vocational 0.08 (0.92) 2 0.12 (0.92) 0.03 (0.97)
Secondary -0.12 (0.96) 3 -0.10 (0.91) -0.10 (1.03)
Lower -0.39 (0.97) 4 -0.30 (0.97) -0.33 (1.01)
Primary -0.43 (1.01) 4 -0.51 (1.00) -0.24 (1.01)
Denomination
Protestant 0.30 (0.83) 1 0.21 (0.91) 0.26 (0.84)
Conservative reformed 0.29 (0.88) 1 0.18 (0.87) 0.26 (0.92)
None 0.02 (1.02) 2 0.03 (0.99) 0.02 (1.02)
Catholic -0.07 (1.02) 3 0.02 (1.01) -0.11 (1.01)
Islamic -0.10 (0.84) 3 -0.33 (0.96) 0.08 (0.92)
Dutch reformed -0.15 (0.97) 3 -0.19 (1.03) -0.09 (0.98)
90 Jacqueline van Beuningen and Hans Schmeets / Procedia - Social and Behavioral Sciences 72 (2013) 73 90

Appendix D.

Table D.1. Bootstrap analysis results (initial latent variable scores have been used as input values for the bootstrap analysis)b

Dimension/Indicator Average weight 95% Confidence Interval

Participation 0.48 0.474 0.489
Social participation 0.50 0.494 0.512
Family contacts 0.40 0.380 0.425
Contacts with friends 0.77 0.752 0.784
Contacts with neighbors 0.32 0.306 0.350
Organizational participation 0.47 0.464 0.480
Organizational activities 0.82 0.809 - 0.837
Work 0.64 0.623 0.655
Political participation 0.55 0.534 0.556
Voting 0.67 0.650 0.691
Political actions 0.66 0.644 0.685
Trust 0.76 0.755 0.769
Social trust 0.37 0.363 0.373
Generalized social trust 1.00 -
Organizational trust 0.67 0.669 0.680
Trust in army 0.17 0.157 0.180
Trust in police 0.22 0.213 0.232
Trust in judges 0.43 0.418 0.446
Trust in civil servants 0.33 0.324 0.346
Trust in press 0.20 0.187 0.205
Trust in large companies 0.17 0.160 0.184
Political trust 0.24 0.232 0.239
Trust in parliament 1.00 -

b
Scores on the dimensions are path coefficients, weights only apply to indicators.