See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/320679215

Measuring social trust and trusting the measure

Article in The Social Science Journal · October 2017

DOI: 10.1016/j.soscij.2017.10.001

CITATIONS READS

7 1,706

3 authors, including:

Ryan SEAN Bakker Jeffrey D. Berejikian

University of Essex University of Georgia
33 PUBLICATIONS 1,764 CITATIONS 23 PUBLICATIONS 569 CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Expert Surveys of Political Parties View project

All content following this page was uploaded by Jeffrey D. Berejikian on 23 November 2018.

The user has requested enhancement of the downloaded file.

G Model
SOCSCI-1425; No. of Pages 11 ARTICLE IN PRESS
The Social Science Journal xxx (2017) xxx–xxx

Contents lists available at ScienceDirect

The Social Science Journal

journal homepage: www.elsevier.com/locate/soscij

Measuring social trust and trusting the measure

Florian Justwan a,∗ , Ryan Bakker b , Jeffrey D. Berejikian b
a
Department of Politics and Philosophy, University of Idaho, 875 Perimeter Drive, MS 3165, Moscow, ID 83844-3165, USA
b
School of Public and International Affairs, University of Georgia, USA

a r t i c l e i n f o a b s t r a c t

Article history: Decades of rigorous quantitative scholarship have generated a wealth of knowledge regard-
Received 31 March 2017 ing the causes and consequences of crossnational variations in social trust. However, while
Received in revised form 15 August 2017 some social science disciplines have made significant contributions to this conversation,
Accepted 3 October 2017
others have largely failed to do so. The field of international relations, for example, has
Available online xxx
lagged behind in producing aggregate-level scholarship on social trust. This is surprising
given that (1) trust influences public opinion and thereby the incentive structure for political
Keywords:
leaders and (2) many peacebuilding efforts directly target the levels of trust in post-conflict
Social trust
settings. Country-level trust scholarship in international relations and the social sciences
Latent variable
Bayesian statistics more generally is hampered by data scarcity. The main purpose of this article is to present
a new publicly available data set on aggregate levels of social trust. Relying on a set of
19 widely accepted correlates, we construct a new cross-sectional measure of the concept
that covers all countries from 1946 to 2010. We then perform a series of empirical tests
establishing the validity of our measure. Finally, we offer a number of bivariate analyses to
demonstrate the broad utility of our new variable for scholars in the social sciences.
© 2017 Western Social Science Association. Published by Elsevier Inc. All rights reserved.

1. Introduction significant contributions to this conversation, others have

The main purpose of this article is to present a new pub-
licly available data set on aggregate levels of social trust.
“Social trust” refers to an individual’s beliefs about the
general trustworthiness of others and it is part of a per-
son’s worldview regarding the benevolence of other human
beings. Aggregated up to the state-level, the variable con-
stitutes an important dimension of a country’s political
culture. Decades of rigorous quantitative scholarship have
generated a wealth of knowledge regarding the causes and
consequences of cross-national variations in social trust
(Knack & Zak, 2001; Rothstein & Uslaner, 2005; You, 2012).
However, while some social science disciplines have made
∗ Corresponding author.
E-mail addresses: fjustwan@uidaho.edu (F.
rbakker@uga.edu (R. Bakker), jberejik@uga.edu (J.D. Berejikian).
largely failed to do so. The field of international relations,
for example, has lagged behind in producing aggregate-
level scholarship on social trust. This is a shortcoming for
two reasons.
First, trust influences individual-level foreign policy
preferences. Citizens who tend to put faith in strangers
have a different outlook on global politics than individu-
als who harbor a very negative outlook on human nature
(Fair, Kaltenthaler, & Miller, 2013; Rathbun, 2011a). Since
public opinion affects the incentive structure for political
leaders (Bueno de Mesquita, 2003; Putnam, 1988), system-
atic cross-national variations in trust are therefore likely to
have major effects on country-level behavior.
Second, at the conceptual level, trust has been identi-
fied as an important dependent variable for international
relations scholars. At the end of civil conflict, restoring
trust in a given society is imperative in order to pre-
Justwan),
vent renewed outbreak of fighting and in order to rebuild

https://doi.org/10.1016/j.soscij.2017.10.001
0362-3319/© 2017 Western Social Science Association. Published by Elsevier Inc. All rights reserved.

Please cite this article in press as: Justwan, F., et al. Measuring social trust and trusting the measure. The Social Science
Journal (2017), https://doi.org/10.1016/j.soscij.2017.10.001
G Model
SOCSCI-1425; No. of Pages 11 ARTICLE IN PRESS
2 F. Justwan et al. / The Social Science Journal xxx (2017) xxx–xxx
political and economic institutions (Zak & Knack, 2001).
Doyle and Sambanis (2006, p. 47) thus “see the problem of
rebuilding a war-torn state as one of rebuilding social trust”
and many peacebuilding efforts (such as dialogue pro-
grams and truth commissions) directly target the levels of
trust in post-conflict settings (Rettig, 2008; Schaap, 2005;
Svensson & Brouneus, 2013). Nevertheless, even though
social trust is an important variable in the theories of many
international relations scholars, empirical work in this area
has not adequately assessed its causes and consequences.
The quantitative literature on post-conflict reconciliation,
for instance, has largely focused on other outcomes such
as conflict relapse, democratization, or changes in human
rights practices (DeRouen et al., 2010; Meernik, Nichols, &
King, 2010).
Aggregate-level trust scholarship in international rela-
tions (and in the social sciences more generally) is
hampered by data scarcity. Existing country-level mea-
sures of the concept are based on aggregated survey
responses and they only generate a very limited number
of observations. This prevents scholars from conducting
large-n statistical tests. In this article, we provide a rem-
edy to this problem. We first assert that existing measures
of trust are best understood as observable but “noisy” man-
ifestations of a latent concept. We then adopt a latent
variable approach which has the advantage of reducing
the potential for measurement error. More specifically, we
deploy a Bayesian measurement model to construct a new
social trust variable that offers significantly greater cov-
erage (more countries and years) than existing measures
of trust. This new measure will allow scholars (in inter-
national relations and all other social science subfields)
to answer a wide range of novel research questions such
as: (1) do trusting societies pursue different kinds of for-
eign policies than untrusting societies or (2) what types
of peacebuilding initiatives create social trust in war-torn
societies?
This article has six sections. First, we provide a short
overview of existing individual-level research on social
trust and we use this scholarship to motivate the construc-
tion of a new country-level measure. In the second section,
we present existing measures of social trust and discuss
some important weaknesses. In section three, we describe
our Bayesian measurement model and the latent variable
resulting from this procedure. In the fourth section, we con-
duct a series of validity checks for our latent measure. Fifth,
we discuss a number of applications of our new latent trust
measure. The sixth section concludes this article.
2. Theoretical motivation: the importance of trust
in the social sciences
Social trust is an individual-level trait that shapes a per-
son’s worldview about the benevolence of other human
beings. Trusting individuals have been found to be more
optimistic about the prospects of cooperation and less
concerned about exploitation than their untrusting coun-
terparts (Rathbun, 2011a). In addition, high-trust survey
respondents are consistently more likely to empathize
with people from a different background (Uslaner, 2002).
This, in turn, allows social trusters to build more expan-
Please cite this article in press as: Justwan, F., et al. Measuring social trust and trusting the measure. The Social Science
Journal (2017), https://doi.org/10.1016/j.soscij.2017.10.001
sive moral communities and to create cooperative bonds
across racial and ethnic boundaries (Rathbun, 2011a).
Given these characteristics, it is not surprising that social
trusters are more integrated and active in their commu-
nities, more likely to engage in pro-social behavior (i.e.
in costly actions that benefit the society as a whole), and
more likely to have good/very good levels of self-reported
health and life-satisfaction (Bjørnskov, 2003; Poortinga,
2006; Sønderskov, 2011).
More relevant for our present purposes, social trust has
also been demonstrated to influence a wide range of polit-
ical attitudes. Domestically, people who score high on this
trait are more likely to approve of re-distributive policies
that benefit cultural minorities, and they harbor more pos-
itive feelings towards immigrants (Crepaz, 2008). In the
words of Herreros and Criado (2009, p. 339) “social trust
implies an expectation that strangers are trustworthy and
[. . .] that this expectation should be extended to all peo-
ple, including immigrants [. . .].” Furthermore, trust is an
important predictor for micro-level preferences about gov-
ernment intervention. As Aghion, Algan, Cahuc, & Shleifer
(2010) show, cynics have a bigger base propensity to favor a
strongly regulated economy than social trusters. Similarly,
they are more likely to agree that “competition is harmful
[and] brings out the worst in people” (p. 1030).
More recently, scholars have started to investigate the
effect of social trust on foreign policy preferences. At
the most general level, the variable makes people more
supportive of internationalist foreign policies and less
likely to endorse the principle of isolationism (Brewer &
Steenbergen, 2002). Similarly, trusting survey respondents
have been found to endorse broadly cooperative foreign
policies and they are less prone to seeing the realm of
global politics as a competition of “all against all” (Binning,
2007). Moving on to more specific policy issues, recent
research suggests that social trust increases individual-
level support for foreign aid, free trade, and participation
in international institutions (Bayram, 2016; Fair, et al.,
2013; Kaltenthaler & Miller, 2013; Rathbun, 2011a). Since
“trusters have a dispositional tendency to believe that the
risks of opportunism are [low]” (Rathbun, 2011b, p. 8),
bilateral assistance programs, close economic ties, and par-
ticipation in intergovernmental organizations appear more
attractive (Fair, et al., 2013). Finally, available evidence
suggests that trust affects attitudes in the context of bor-
der disputes. Justwan (2017) finds that citizen in claimant
states are more likely to favor adjudication by an inter-
national court if they are high in social trust. Cynics, by
contrast, are less willing to let a third party determine
the terms of a settlement. Subsequent survey research by
Justwan and Fisher (2017) has expanded on these findings
by showing that trust also increases support for territorial
concessions.
A summary of these research findings is presented in
Table 1. The individual-level dynamics summarized here
have important theoretical implications. It is well estab-
lished that mass attitudes change the incentive structure
for political leaders (Bueno de Mesquita, 2003; Putnam,
1988). Since trust influences public policy preferences
(about world politics and other issues), there are strong
theoretical reasons to believe that high-trust countries
G Model
SOCSCI-1425; No. of Pages 11 ARTICLE IN PRESS
F. Justwan et al. / The Social Science Journal xxx (2017) xxx–xxx
Table 1
Theoretical motivation: known individual-level consequences of social trust.
Field Dependent variable
Int. relations Support for foreign aid
Int. relations Support for participation in internat. organizations
Int. relations Support for adjudication in border disputes
Int. relations Support for territorial concessions
Int. relations Support for “internationalism” (vs. “isolationism”)
Int. relations Support for “cooperative foreign policies”
Int. relations/economics Support for free trade
Economics Opposition to government intervention in economy
Comp. politics/sociology Attitudes towards immigrants
Sociology Memberships in civil society organizations
Public health Self-reported health
pursue different types of policies than low-trust states.
Some studies provide evidence for the link between mass
attitudes and policy output. For example, Rothstein and
Uslaner (2005) show that trust (at the aggregate level) is
positively related to the size of the welfare state. Like-
wise, trusting societies have more “immigrant friendly”
integration policies than other states (Justwan, 2015).
Nevertheless, as explained below, testing country-level
implications of existing individual-level theories has been
hampered by data scarcity.
3. Survey-based measures of trust
Existing studies largely rely on survey data to measure
the overall level of trust in a society. For example, the
General Trust Question (GTQ) asks respondents whether
“most people can be trusted” or whether “you can’t be too
careful when dealing with people.” High-trust societies are
then defined as polities in which most citizens answer this
question positively. For cross-national comparisons, the
World Values Survey (WVS) is the most widely utilized data
source because it collects responses to the GTQ across the
largest number of countries and over the longest period of
time (Rothstein & Uslaner, 2005; You, 2012).
However, survey-based measures like the WVS are
not unproblematic. First, researchers have found that
differential item functionality (DIF) often impedes the
cross-national comparability of survey responses. DIF
refers to the idea that individuals in two or more coun-
tries sometimes understand an identical survey question
in vastly different ways (King, Murray, Salomon, & Tandon,
2004). There are reasons to believe that DIF is an issue with
regards to social trust. For example, it is not clear to what
type of interaction the concept of trust applies and what
the expression “most people” refers to. As Delhey, Newton,
and Welzel (2011) demonstrate, respondents in Asian soci-
eties tend to have very different groups of people in mind
than individuals in Western countries when they are asked
to assess the trustworthiness of strangers.1 As a result,
cross-national trust research based on the GTQ sometimes
produces conflicting results. Economic development, for
example, has been shown to be causally related to social
trust (Baliamoune-Lutz, 2011). However, while trust is pos-
1
In particular, the authors argue that the “trust radius” in these societies
is narrower.
Please cite this article in press as: Justwan, F., et al. Measuring social trust and trusting the measure. The Social Science
Journal (2017), https://doi.org/10.1016/j.soscij.2017.10.001
3
Citations
Paxton and Knack (2011) and Bayram (2016)
Rathbun (2011a, 2012) and Fair et al. (2013)
Justwan (2017)
Justwan and Fisher (2017)
Brewer and Steenbergen (2002) and Brewer et al. (2004)
Binning (2007)
Kaltenthaler and Miller (2013)
Aghion et al. (2010)
Herreros and Criado (2009)
Sønderskov (2011)
Poortinga (2006)
itively correlated with GDP/capita at the global level, there
is no relationship between both variables in Asia.2
A second (and arguably even bigger) problem with
survey-based measures of social trust is related to cover-
age. Although the WVS provides data for a large number
of countries at various time points, the overall number
of country-years available is limited (n = 232). At most
researchers have data for only five time points per country,
and these are spaced at roughly five-year intervals extend-
ing back only to 1981. Various surveys such as the Euro-,
Latino-, and Afro-barometer supplement the WVS and pro-
vide GTQ data. However, the addition of these “Barometers”
offers little improvement as the total number of added
country-years is modest (n = 371). In addition, the sever-
ity of the data coverage problem varies across geographic
regions. For many non-Western developing countries, trust
data are particularly scarce.
4. A Bayesian latent variable model
4.1. Model description
Given the above, we develop a measure of aggregate
social trust levels that provides better time and space cov-
erage than do any of the existing survey-based measures.
Here, we adopt a Bayesian measurement model approach
and treat trust as a latent concept. We chose this approach
because, as a variable that is hard to operationalize (see
above), relying on any single observable indicator for trust
measurement would be inadequate. Although the GTQ
found in the WVS seems to capture much of what schol-
ars mean by “trust,” a latent variable constructed with a
broader set of correlates (listed below) provides “cleaner”
and more reliable estimates.
Latent models treat latent concepts as explanatory vari-
ables, while the observable indicators of a concept are
treated as outcome variables. Given the nature of cross-
national data, there are varying degrees of missingness
across countries on the observable indicators for trust.3 A
2
The correlation between GDP/capita (data obtained from World Bank)
and social trust (operationalized with the WVS) at the global level is 0.58
(n = 222). For countries in Asia, the correlation is only 0.03 (n = 46).
3
An additional advantage of the Bayesian approach is that these mod-
els automatically provide measures of uncertainty for the latent concept.
These measures of uncertainty are closely related to the amount of miss-
ingness in the observed indicators. That is, country/years with fewer
G Model
SOCSCI-1425; No. of Pages 11 ARTICLE IN PRESS
4 F. Justwan et al. / The Social Science Journal xxx (2017) xxx–xxx

Bayesian approach to latent variable estimation accounts
for such variation because missing values on the depen-
dent variable are automatically imputed by conditioning
on observed data.4 Formally, the model is:
 
yi,j ∼N i,j , j2
i,j = ˇj Xi,j εi,j
Xi,j ∼N (0, 1)
εi,j ∼N (0, 1)
j2 ∼IG (0.1, 0.1)
ˇj ∼N (0, 10)
Above, i indexes country/year and j the observed indica-
tor. X is the latent measure of trust. The ys are observed
indicators. Finally, ˇ is the factor loading.
4.2. Observable indicators of trust
To identify a set of observable indicators we canvased
trust research across the disciplines of political science,
sociology, psychology, and neuro-economics. We found 19
variables that scholars consistently associate with trust.
Descriptive statistics and data sources are provided in
Table 2.
Our model combines survey-based indicators of trust,
as well as institutional, social-psychological, and bio-
logical/environmental correlates of the concept. For the
survey-based indicators, we rely on data from the (1) World
Values Survey and (2) the Euro-, Afro-, Latino-, and Asi-
abarometers. For each available country–year, we add the
proportion of “social trusters” as an observation to our
dataset.
We add nine institutional correlates of trust: (3) a coun-
try’s score on the Polity IV index (ranging from −10 to
+10) as well as (4) the squared value of that number to
account for non-linearity (Marshall & Jaggers, 2002); (5)
an indicator capturing the quality of a country’s bureau-
cracy (range: 0.04–1.00), which is obtained from the
International Country Risk Guide (PRS Group, 2015); (6)
an ordinal variable, taken from the CIRI Human Rights
Project (Cingranelli, Richards, & Clay, 2014), which mea-
sures judicial independence (range 0–2); (7) the ratio of
non-currency money relative to the total money supply in
a given country–year (range: 0–100; source: Clague, Keefer,
Knack, & Olson, 1999); (8) the degree of corruption as mea-
sured by Transparency International’s (2017) Corruption
Perceptions Index (range: 0.4–10); (9) a country’s score
on the Freedom House (2017) Civil Liberties Index (range:
1–7); (10) the size of a state’s informal sector relative to
its total GDP (range: 0.08–1.13; Source: Elgin & Oztunali,
observable indicators of trust will yield estimates with greater uncer-
tainty.
4
See Treier and Jackman (2008) for more information on Bayesian
latent variable models.
2012); and (11) a binary indicator (self-coded) for former
Soviet states.
Next, we add three social-psychological correlates: (12)
the proportion of Protestants in a country’s population
(range: 0–97.8; Source: La Porta, López-de-Silanes, Shleifer,
& Vishny, 1999); (13) a society’s degree of income inequal-
ity as measured by the GINI Index (range: 15.90–73.90;
Source: World Bank, 2015); and (14) a state’s level of ethnic
fractionalization (range: 0.01–1.00), as provided by Fearon
and Laitin (2003).
Finally, building on recent research in the field of
neuro-economics (Zak & Fakhar, 2006), we introduce five
biological/environmental correlates of trust into the mea-
surement model: (15) fertility rate, measured by total
births per woman (range: 1.08–9.22); (16) protected areas
(such as national parks) as percent of a country’s total ter-
ritory (range: 0–98.11); (17) the proportion of a state’s
female population (range: 24.26–55.85); (18) the overall
prevalence of malnourishment (range: 5.00–70.00); and
(19) the degree of water pollution in a country, measured
as water pollutants in kilograms per day per worker (range
0.09–0.45). All data for these indicators were obtained from
the World Bank (2015).
The online Appendix contains the precise theoretical
logic through which each indicator is linked to social
trust. Furthermore, Appendix Table A1 provides a detailed
description of the operationalization for all 19 indicators.
4.3. Estimation and results
We estimated the above model using MCMC in Win-
BUGS with 10,000 iterations and 2 separate chains. The first
2,500 iterations were discarded as a burn-in. Below, we
summarize the remaining 7,500 iterations.5 Prior to esti-
mating the model, we standardized the indicators (i.e. we
rescaled all variables to have a mean of zero and a standard
deviation of one). This step is essential in order to make the
factor loadings comparable.
In Fig. 1 we present the factor loadings. For each
indicator’s factor loading, we report the mean values of
the posterior distribution and associated 95% credible
intervals. The graph shows that all factor loadings are sta-
tistically distinguishable from zero. As expected, higher
values in survey-based measures of trust are positively
related to our latent variable (the factor loadings for the
World Values Survey and the Barometers are 0.57 and
0.53 respectively). The indicators with the highest abso-
lute factor loadings are the ones capturing (1) the degree
of corruption in a society (fl: 0.96), (2) the extent to
which governments grant civil liberties to their citizens (fl:
−0.91), and (3) the magnitude of political rights in a country
(fl: 0.79).6
The resulting trust variable ranges from −1.92 in Equa-
torial Guinea (1987) to 2.18 in Finland (2004) and Denmark
5
There are strong signs of convergence according to a number of tests
presented in the online Appendix.
6
It is worth noting that higher values on the corruption and civil lib-
erties indicators indicate lower levels of corruption and civil liberties. In
other words, Fig. 1 suggests that more corrupt societies have lower levels
of trust as do countries in which civil liberties are restricted.

Please cite this article in press as: Justwan, F., et al. Measuring social trust and trusting the measure. The Social Science
Journal (2017), https://doi.org/10.1016/j.soscij.2017.10.001
G Model
SOCSCI-1425; No. of Pages 11 ARTICLE IN PRESS
F. Justwan et al. / The Social Science Journal xxx (2017) xxx–xxx 5

Table 2
Indicators for measurement model.

Indicator type and name Data source Value range Exp.a Indicator relationship to
social trust suggested by

Survey-based measures
General Trust Question World values survey 0.03–0.74 +
General Trust Question Barometers 0.02–0.75 +

Institutional correlates
Political rights Polity IV (Marshall & Jaggers, 2002) −10 to 10 + Delhey and Newton (2005)
Political rights (non-linear effect) Polity IV (squared) 0–100 + You (2012)
Corruption Transparency International (2017) 0.40–10 + You (2012)
Bureaucratic quality ICRG (PRS Group, 2015) 0.04–1.00 + Rothstein and Stolle (2008)
Judicial independence Cingranelli et al. (2014) 0–2 + Rothstein and Stolle (2008)
Non-currency money/total money supply Clague et al. (1999) 0.03–0.99 + Baliamoune-Lutz (2011)
Civil liberties Freedom House (2017) 1–7 − Delhey and Newton (2005)
Institutional effectiveness Elgin and Oztunali (2012) 0.08–1.13 − Rothstein and Stolle (2008)
Former soviet country hand-coded 0–1 − Bjørnskov (2006)

Social-psychological correlates
Protestantism La Porta et al. (1999) 0–97.80 + Delhey and Newton (2005)
Income inequality Gini index (World Bank, 2015) 15.90–73.90 − Rothstein and Uslaner (2005)
Ethnic fractionalization Fearon and Laitin (2003) 0.01–1.00 − Delhey and Newton (2005)

Biological/environmental correlates
Fertility rate World Bank (2015) 1.08–9.22 + Zak and Fakhar (2006)
Environmentally protected areas World Bank (2015) 0–98.11 + Zak and Fakhar (2006)
Size of female population World Bank (2015) 24.26–55.85 + Zak and Fakhar (2006)
Prevalence of malnourishment World Bank (2015) 5.00–70.00 − Zak and Fakhar (2006)
Water pollution World Bank (2015) 0.09–0.45 − Zak and Fakhar (2006)
a
Expectation for the relationship between a given indicator and the estimated latent trust variable.

Fig. 1. Factor loadings.

(2004). The variable has a mean of 0 and a variance of
approximately 1. The units of this variable are not directly
interpretable, but we can say that higher values indicate
higher societal trust levels in a given country/year. Given
that the countries on the high end of our new variable all
tend to be from Northern Europe and those on the low end
from sub-Saharan Africa and the Middle East, we feel there
is a good amount of face validity with this variable.
In Fig. 2, we plot the mean trust scores and associated
uncertainty estimates for a small sample of countries in the

Please cite this article in press as: Justwan, F., et al. Measuring social trust and trusting the measure. The Social Science
Journal (2017), https://doi.org/10.1016/j.soscij.2017.10.001
G Model
SOCSCI-1425; No. of Pages 11 ARTICLE IN PRESS
6 F. Justwan et al. / The Social Science Journal xxx (2017) xxx–xxx

Fig. 2. Mean trust scores for a sample of countries.

dataset (covering all geographic regions). Appendix Fig. A1 are classified as “low-trusting” in our dataset. This find-
shows the mean trust scores for all countries in the inter- ing is hardly surprising. According to previous research,
national system. Appendix Fig. A2 graphs average levels of the results provided by the WVS are considered to be as
societal trust by year and region. In the section that follows, unreliable for these same seven states (Bjørnskov, 2006;
we provide tests of the validity of this new variable.7 Delhey et al., 2011; Dinesen, 2013). The reasons for this vary
by country. In some non-democratic settings (like Saudi
Arabia and Iraq), conducting survey research is extremely
5. Validity checks
challenging and the obtained samples are therefore prob-
ably not reflective of true population parameters. In other
5.1. Descriptive evidence
states (like China), WVS scores likely suffer from above-
noted differential item functionality (Delhey et al., 2011).
In Fig. 3 we plot available trust scores in the WVS
The fact that our measure places these countries at the low
against the estimated values on our latent measure. The
end of the spectrum should therefore be seen as evidence
graph reveals several interesting findings. First, and most
in favor of our latent variable.
importantly, our latent trust measure and the WVS both
Finally, Fig. 3 also demonstrates that some countries
rank a large majority of countries in a similar fashion.
(Cyprus, Slovenia, Portugal, and Chile) receive higher val-
For example, Norway, Denmark, and Sweden are high-
ues on our latent measure than in the WVS. Here too, there
trust societies according to both operationalizations of
are strong reasons to doubt the reliability of the available
the concept. Conversely, many African countries (such as
WVS data since other survey projects provide drastically
Zimbabwe, Rwanda, and Nigeria) can be found at the low
different scores for these societies. According to the Inter-
end on both measures. This general picture provides con-
national Social Survey Programme (ISSP), for example, over
siderable face validity for our variable.
58% of respondents in Cyprus were social trusters in 2007
The second important insight contained in Fig. 3 is that
which is significantly higher than the number provided
our measure generates values for a (small) set of coun-
by the WVS in 2006 (12.8%). Similarly, the ISSP scores for
tries that are quite different than the WVS. For example,
Slovenia in 2007 (38.8%) and Portugal in 2004 (30.0%) are
Saudi Arabia, Iraq, Iran, Vietnam, China, Indonesia and
considerably higher than the values recorded by the WVS
Thailand are consistently placed as high trusting societies
for Slovenia in 2005 (18.1%) and Portugal in 1999 (12.3%).8
in traditional (survey-based) studies of trust, however they

7 8
Some of the following validity checks are also discussed in Justwan Source: International Social Survey Programme: Leisure Time and
(2017) and Justwan and Fisher (2017). Sports. GESIS Data Archive, Cologne. ZA4850 Data file Version 2.0.0.

Please cite this article in press as: Justwan, F., et al. Measuring social trust and trusting the measure. The Social Science
Journal (2017), https://doi.org/10.1016/j.soscij.2017.10.001
G Model
SOCSCI-1425; No. of Pages 11 ARTICLE IN PRESS
F. Justwan et al. / The Social Science Journal xxx (2017) xxx–xxx
Fig. 3. Comparison of available country–years.
Lastly, according to a 2006 survey by the Latin America
Public Opinion Project, more than half of all respondents in
Chile have “high interpersonal trust” which is substantially
different from the WVS score for 2005 (12.1%) which is
represented in Fig. 3.9 This sizable amount of variation sug-
gests that existing survey-based trust measures for these
polities are inconclusive.
In order to assess the robustness of the measurement
model, we estimated a number of alternative versions in
which we removed certain indicators. The results (pre-
sented in the online Appendix) suggest that the latent
construct is very stable and its distribution does not depend
on the inclusion/exclusion of certain correlates.
The correlation between our construct and the WVS
is considerable (r = 0.46). When we examine states that
receive scores above “0” (thus excluding the outliers
described above), the association increases even further
(r = 0.75). In the online Appendix (Fig. A3) we plot all
survey-based observations of trust (thus going beyond the
World Values Survey and also including data from the
Euro-, Latino-, Afro-, and Asiabarometer, ISSP, European
Social Survey and General Social Survey) against our latent
measure. The correlation between our measure and these
surveys (n = 781) is r = 0.60. When we exclude the low-end
outliers (identified above) we observe a correlation of 0.74
(n = 644).
Perfect correlation between our measure and the WVS
is neither possible nor desirable. As mentioned earlier,
survey-based measures of trust are plagued by many prob-
lems and the data provided through them have to be taken
with a grain of salt. A comparison of two survey-based mea-
sures of trust is instructive: the available data points from
the World Values Survey and the “Barometers” correlate
at roughly 0.34 (n = 25). This reminds us that even survey-
9
Source: The AmericasBarometer by the Latin American Public Opinion
Project (LAPOP), www.LapopSurveys.org (accessed, July 12, 2017).
Please cite this article in press as: Justwan, F., et al. Measuring social trust and trusting the measure. The Social Science
Journal (2017), https://doi.org/10.1016/j.soscij.2017.10.001
7
based measures of trust sometimes produce inconclusive
values for any given country–year.
5.2. Internal validity
Next, we calculated the correlations between our latent
variable and other aggregate-level indicators that have
been found to be consequences of social trust. We then
compared these correlation coefficients with the corre-
sponding values obtained with the WVS measure. The point
of this test is to assess whether both trust measures pro-
duce similar correlations. Our latent variable is “internally
valid” if the magnitude and direction of Pearson’s r is
roughly similar for both trust measures.
Previous scholarship has found a robust connection
between social trust and (1) well-being/life satisfaction
(Bjørnskov, 2003), (2) attitudes towards out-group mem-
bers (Herreros & Criado, 2009), (3) pro-social behaviors
(Putnam, 2000), and (4) support for democracy (Putnam,
2000). In Table 3 we present ten pairs of correlations
between our latent construct, the WVS measure of trust,
and a number of aggregate-level variables tapping into
these four dimensions.
The results provide strong evidence for the validity of
our measure. Both operationalizations of trust are positively
related to country-levels of life satisfaction, self-reported
health, political tolerance, and a variable capturing the
prevalence of the norm of “caring for others.” Similarly,
for both measures, “trusting” societies have higher lev-
els of democratic system support (as captured by two
separate variables). By contrast, both trust measures are
negatively related to support for stricter immigration laws,
unfavorable attitudes towards immigrants, and anti-social
behavior such as fare evasion on public transportation or
cheating on taxes. Importantly, Table 3 does not only pro-
vide important insights about correlation directions but
also about correlation strengths. In particular, our analyses
reveal that indicators that are only weakly related to the
WVS measure of trust (like the proportion of people who
G Model
SOCSCI-1425; No. of Pages 11 ARTICLE IN PRESS
8 F. Justwan et al. / The Social Science Journal xxx (2017) xxx–xxx

Table 3
Internal validity tests (World Values Survey).

Variable (number of Operationalization Corr. with WVS trust Corr. with latent trust
observations)

Well-being and life satisfaction

Life satisfaction (n = 138) Country–average: life satisfaction r = 0.37 r = 0.60
Self-reported health (n = 138) Country–average: self-reported health r = 0.21 r = 0.27

Attitudes towards out-groups

Favor stricter immigration % of people who favor stricter immigration r = −0.33 r = −0.28
laws (n = 115) policies
Reject jobs for immigrants % of people who believe jobs should go to r = −0.56 r = −0.56
(n = 133) nationals over immigrants.
Political tolerance (n = 43) % of people who think that their “least liked” r = 0.48 r = 0.47
social group should be allowed to hold public
demonstrations

Pro-social behavior
Fare evasion on public % of people who think it is justifiable to evade r = −0.30 r = −0.23
transport (n = 135) fares
Cheating on taxes (n = 139) % of people who think it is justifiable to cheat r = −0.15 r = −0.20
on taxes
Importance of helping Country–average: self-reported importance of r = 0.20 r = 0.24
strangers (n = 42) “caring for others”

Attitudes towards democracy

Democratic system support % of people who agree that democracy is best r = 0.21 r = 0.23
(n = 77) form of government
Importance of living in % of people who find it important to live in a r = 0.34 r = 0.35
democracy (n = 44) democracy

think it is justifiable to cheat on taxes) are also only weakly
associated with our latent measure. Other variables (like
the proportion of people who believe that jobs should go
to nationals over immigrants) demonstrate strong correla-
tions with both measures of trust.
In addition to the tests summarized in Table 3, we
also compared the correlation coefficients for both trust
measures and various social development indicators. Fur-
thermore, we replicated two articles on aggregate-level
effects of social trust.10 The results of these analyses can
be found in the online Appendix (Tables A2–A4). Again, our
latent variable produces correlations and regression results
that are near-identical to the ones obtained with the WVS
measure. Taken together, the results discussed in this sec-
tion strongly suggest that our latent variable is internally
valid.
6. Potential applications
6.1. Trust as an independent variable in the social
sciences
Our new measure can be used to predict a wide range
of outcomes within the social sciences. In particular, the
latent variable presented here should be useful to all
researchers who are interested in studying aggregate-level
phenomena for which social trust shapes mass attitudes
and behaviors (see Table 1). Below, we discuss potential
applications in political science, economics and sociology.
10
More specifically, we replicated Justwan’s (2015) analysis of immi-
grant voting rights and Zak and Knack’s (2001) study on the economic
effects of society-wide trust levels.
Our political science applications specifically come from
the subfield of international relations. As mentioned ear-
lier, social trust levels do not feature prominently as an
independent variable in this literature. We believe that this
is a major shortcoming since trust influences societal for-
eign policy preferences and thereby the incentive structure
for political leaders (Bayram, 2016; Brewer & Steenbergen,
2002; Fair et al., 2013; Justwan, 2017; Kertzer & Brutger,
2016). As a result, there are strong reasons to believe that
“trusting countries” pursue different foreign policies than
“untrusting” societies. Preliminary statistical analyses sug-
gest that this is, in fact, the case. In Table 4, we analyze
the relationship between social trust and the probability
that (1) two states form a bilateral military alliance, and (2)
a state initiates a militarized interstate dispute in a dyad.
We chose these outcomes since research has found that
trust predicts individual-level attitudes towards participa-
tion in international institutions and the use of military
force (Brewer, Gross, Aday, & Willnat, 2004; Fair et al.,
2013). Results show that there is a statistically significant
(bivariate) relationship between these variables and social
trust: countries with more trusting populations appear to
be more likely to form bilateral alliances and less likely to
initiate militarized interstate disputes.
Our variable may also help to predict economic phe-
nomena. We know from previous scholarship that trusting
societies have faster economic growth rates than other
countries. So far however, this relationship has only been
explored for country–years after 1969 (Zak & Knack, 2011).
Our variable allows us to extend the temporal domain of the
analysis. Bivariate regression models reveal that high-trust
societies do not only grow faster in the period from 1970
to 2010 but also between 1946 and 1969. For both sub-
samples, the trust-coefficient is positive and statistically

Please cite this article in press as: Justwan, F., et al. Measuring social trust and trusting the measure. The Social Science
Journal (2017), https://doi.org/10.1016/j.soscij.2017.10.001
G Model
SOCSCI-1425; No. of Pages 11 ARTICLE IN PRESS
F. Justwan et al. / The Social Science Journal xxx (2017) xxx–xxx 9

Table 4
Latent trust as an independent variable.

IV Dependent variable ˇ (std. error) p-Value Operationalization of Data source Reduction of

(number of obs.) DV missingness (rel. to
WVS)

Latent trust Alliance onset in 0.65a (0.02) 0.01 Onset of bilateral Crescenzi, 73.8%
dyad (n = 1,219,668) military alliance in Kathman,
dyad Kleinberg, and
Wood (2012)
Latent trust MID initiation −0.16b (0.05) 0.01 MID initiation in a Weeks (2012) 93.6%
(n = 997,479) given year
Latent trust Trade restrictions −6.54c (2.00) 0.01 Import duties (as % of Rose (2004) 96.0%
(n = 2,288) imports)
c
Latent trust GDP growth 0.27 (0.13) 0.04 Annual GDP growth World Bank 96.5%
[1970–2010] rate in percent
(n = 6,238)
Latent trust GDP growth 0.74c (0.31) 0.02 Annual GDP growth World Bank 100%
[1946–1969] rate in percent
(n = 748)
Latent trust Associational 0.17c (0.05) 0.01 Average number of WVS 3.6%
memberships memberships in civil
(n = 168) society organizations
per citizen
a
Logit regression. Std. errors clustered around dyads. Trust operationalized as “weak link” in dyad (value of less trusting state).
b
Logit regression. Std. errors clustered around directed dyads.
c
OLS regression. Std. errors clustered around countries.

significant, which suggests that existing research findings
also apply to earlier time periods. Next, we use our variable
to examine the effect of trust on country-level trade restric-
tions (Rose 2004). We anticipate a negative relationship
between both variables based on existing individual-level
research connecting these two concepts (Kaltenthaler &
Miller, 2013). This expectation is supported by the data.
According to our analysis, high-trust countries tend to
levy lower import duties than their low-trust counterparts
(p < 0.01). In other words: there is evidence that the demon-
strated individual-level effect of trust on trade preferences
translates to tangible country-level outcomes.
Finally, we assess the relationship between our latent
construct and associational membership. Previous survey
research in sociology has shown that trust has a posi-
tive effect on participation in civil society organizations
(Sønderskov, 2011). Thus, there are reasons to believe that
our measure might help to explain country-level variations
on this dimension. Here too, our theoretical expectations
are confirmed. As the final row in Table 4 shows, countries
with higher trust-levels have more civically engaged pop-
ulations than states with distrustful citizenries (p < 0.01).
It is now up to future research to determine whether the
bivariate relationships described in this section hold after
the inclusion of appropriate control variables.
It is important to note that our trust variable reduces
missingness on the independent variable for each of the
examples above.11 For instance, the WVS measure of trust
allowed us to analyze trade policies for a total of 92
country–years. Our latent construct increases this number
to 2,288. This amounts to a reduction of missingness of over
11
In Table 4, “reduction of missingness” is calculated as the percentage-
wise increase in non-missing observations on the independent variable
(social trust) relative to available observations on the dependent variable.
Formally, the equation is: [N (latent trust) − N (WVS trust)]/N (dependent
variable).
96% and it demonstrates once again that our measure sig-
nificantly outperforms exiting trust variables in terms of
data coverage. In some cases, this will allow researchers
to study entirely new phenomena (like the relationship
between trust and conflict onset). In some other cases, our
measure will enable researchers to expand the temporal
domain of previously analyzed relationships (like the effect
of trust on GDP growth).
6.2. Trust as a dependent variable
Finally, our measure can also provide important insights
if used as a dependent variable. As mentioned earlier, many
peacebuilding efforts (such as dialogue programs and truth
commissions) explicitly try to increase the levels of social
trust in post-conflict settings (Rettig, 2008; Schaap, 2005;
Svensson & Brouneus, 2013). However, due to the lack of
data on aggregate trust levels, the effectiveness of many
post-conflict initiatives has never been tested directly. The
variable presented in this article allows researchers to rem-
edy this problem. Using data from Olsen, Payne, and Reiter
(2010), we conducted a preliminary statistical test regard-
ing the effect of various transitional justice mechanisms on
social trust in the aftermath of civil war. We estimated four
OLS models in which the unit of analysis is the civil war
and the dependent variable is the change of social trust in
a country within ten years after the end of a given internal
conflict. We then separately assessed the effect of trials,
truth commissions, amnesties, and lustration policies on
the outcome variable.12 Our results (Table 5) suggest that
three of these mechanisms do increase social trust: the
presence of trials, truth commissions and lustration poli-
cies is positively (and significantly) related to changes in
trust in post-conflict settings. This is therefore the first
12
See Olsen et al. (2010) for a definition of these terms.

Please cite this article in press as: Justwan, F., et al. Measuring social trust and trusting the measure. The Social Science
Journal (2017), https://doi.org/10.1016/j.soscij.2017.10.001
G Model
SOCSCI-1425; No. of Pages 11 ARTICLE IN PRESS
10 F. Justwan et al. / The Social Science Journal xxx (2017) xxx–xxx
Table 5
Latent trust as a dependent variable.
IV Dependent variable ˇ (std. error)
(number of observations)
Trial 0.37 (0.16)
Change in
latent trust
Truth commission 0.25 (0.10)
(n = 99)
Amnesty 0.07 (0.09)
Lustration 0.45 (0.10)
All models are OLS regressions with standard errors clustered around countries.
quantitative analysis to provide evidence that certain tran-
sitional justice programs may reach their self-defined goal.
Amnesties however, do not seem to have the same effect.
The coefficient for this variable is positive but it fails to
reach statistical significance. We now leave it up to future
research to determine whether these findings hold up in
the context of multiple regression analyses.
7. Conclusion
In this article, we have developed a new measure of
social trust. We have done so using a Bayesian latent
variable model that allows us to recover estimates for
country–years that previous measures of trust did not
provide. Social scientists can use our measure as both inde-
pendent and dependent variable.
We believe that researchers in the field of international
relations may find our variable particularly useful. In addi-
tion to topics related to the effectiveness of transitional
justice, we can now begin to analyze the impact of social
trust on a wide range of country-level and dyadic outcomes.
We hope that scholars will find the possibilities created by
our new measure intriguing and start to ask and answer
questions that have previously been limited by lack of data.
Appendix A. Supplementary data
Supplementary data associated with this arti-
cle can be found, in the online version, at
https://doi.org/10.1016/j.soscij.2017.10.001.
References
Aghion, P., Algan, Y., Cahuc, P., & Shleifer, A. (2010). Regulation and dis-
trust. Quarterly Journal of Economics, 125, 1015–1049.
Baliamoune-Lutz, M. (2011). Trust-based social capital, institutions, and
development. Journal of Socio-Economics, 40, 335–346.
Bayram, A. B. (2016). Generalized trust and the moral basis of public sup-
port for foreign development aid. Foreign Policy Analysis, 13, 133–153.
Binning, K. R. (2007). It’s us against the world: How distrust in Americans
versus people-in-general shapes competitive foreign policy prefer-
ences. Political Psychology, 28, 777–799.
Bjørnskov, C. (2003). The happy few: Cross-country evidence on social
capital and life satisfaction. Kyklos, 56, 1–16.
Bjørnskov, C. (2006). Determinants of generalized trust: A cross-country
comparison. Public Choice, 130, 1–21.
Brewer, P. R., Gross, K., Aday, S., & Willnat, L. (2004). International trust
and public opinion about world affairs. American Journal of Political
Science, 48, 93–109.
Brewer, P. R., & Steenbergen, M. R. (2002). All against all: How beliefs about
human nature shape foreign policy opinions. Political Psychology, 23,
39–58.
Bueno de Mesquita, B. (2003). The logic of political survival. Cambridge,
Mass: MIT Press.
Please cite this article in press as: Justwan, F., et al. Measuring social trust and trusting the measure. The Social Science
Journal (2017), https://doi.org/10.1016/j.soscij.2017.10.001
p-Value Operationalization of Data source
dependent variable
0.02 Change in
social trust Olsen et al.
0.01 within 10 years (2010)
0.46 after the end of
0.01 Civil War
Cingranelli, D. L., Richards, D. L., & Clay, K. C. (2014). The CIRI Human Rights
Dataset. In (2014.04.14 ed.).
Clague, C., Keefer, P., Knack, S., & Olson, M. (1999). Contract-intensive
money: Contract enforcement, property rights, and economic perfor-
mance. Journal of Economic Growth, 4, 185–211.
Crepaz, M. (2008). Trust beyond borders. Immigration, the welfare state, and
identity in modern societies. Ann Arbor: University of Michigan Press.
Crescenzi, M. J. C., Kathman, J. D., Kleinberg, K. B., & Wood, R. M. (2012).
Reliability, reputation, and alliance formation. International Studies
Quarterly, 56, 259–274.
Delhey, J., & Newton, K. (2005). Predicting cross-national levels of social
trust: Global pattern or nordic exceptionalism? European Sociological
Review, 21, 311–327.
Delhey, J., Newton, K., & Welzel, C. (2011). How general is trust in most
people? Solving the radius of trust problem. American Sociological
Review, 76, 786–807.
DeRouen, K., Ferguson, M. J., Norton, S., Park, Y. H., Lea, J., & Streat-Bartlett,
A. (2010). Civil war peace agreement implementation and state capac-
ity. Journal of Peace Research, 47, 333–346.
Dinesen, P. T. (2013). Where you come from or where you live? Examining
the cultural and institutional explanation of generalized trust using
migration as a natural experiment. European Sociological Review, 29,
114–128.
Doyle, M. W., & Sambanis, N. (2006). Making war and building peace: United
Nations peace operations. Princeton, N.J: Princeton University Press.
Elgin, C., & Oztunali, O. (2012). . pp. 1–48. Shadow economies around the
world: model based estimates (5) Bogazici University Department of
Economics Working Papers.
Fair, C. C., Kaltenthaler, K., & Miller, W. (2013). Busting out. Iranian public
opinion toward the NPT. The Nonproliferation Review, 20, 123–136.
Fearon, J., & Laitin, D. (2003). Ethnicity, insurgency and civil war. American
Political Science Review, 97, 75–90.
Freedom House. (2017). Freedom in the world reports.. Available online.
http://freedomhouse.org/reports
Herreros, F., & Criado, H. (2009). Social trust, social capital and perceptions
of immigration. Political Studies, 57, 337–355.
Justwan, F. (2015). Disenfranchised minorities: Trust, definitions of cit-
izenship and non-citizen voting rights in developed democracies.
International Political Science Review, 36, 373–392.
Justwan, F. (2017). Trusting publics: The impact of generalized social trust
on the decision to pursue binding conflict management. Journal of
Conflict Resolution, 61, 590–614.
Justwan, F., & Fisher, S. (2017). Generalized social trust and international
dispute settlement. International Interactions, 43, 717–743.
Kaltenthaler, K., & Miller, W. J. (2013). Social psychology and public
support for trade liberalization. International Studies Quarterly, 57,
784–790.
Kertzer, J. D., & Brutger, R. (2016). Decomposing audience costs: Bring-
ing the audience back into audience cost theory. American Journal of
Political Science, 60, 234–249.
King, G., Murray, C. J. L., Salomon, J. A., & Tandon, A. (2004). Enhancing the
validity and cross-cultural comparability of measurement in survey
research. American Political Science Review, 98, 191–207.
Zak, P., & Knack, S. (2001). Trust and growth. The Economic Journal, 111,
295–321.
La Porta, R., López-de-Silanes, F., Shleifer, A., & Vishny, R. (1999). The
quality of government. Journal of Law, Economics and Organization, 15,
222–279.
Marshall, M. G., & Jaggers, K. (2002). Polity IV project: Political
regime characteristics and transitions, 1800–2002. Available online:
http://www.systemicpeace.org/polity/polity4.htm.
Meernik, J., Nichols, A., & King, K. L. (2010). The impact of international
tribunals and domestic trials on peace and human rights after civil
war. International Studies Perspectives, 11, 309–334.
G Model
SOCSCI-1425; No. of Pages 11 ARTICLE IN PRESS
F. Justwan et al. / The Social Science Journal xxx (2017) xxx–xxx
Olsen, T. D., Payne, L. A., & Reiter, A. G. (2010). Transitional justice in
the world, 1970–2007: Insights from a new dataset. Journal of Peace
Research, 47, 803–809.
Paxton, P., & Knack, S. (2011). Individual and country-level factors affect-
ing support for foreign aid. International Political Science Review, 33,
171–192.
Poortinga, W. (2006). Social capital: An individual or collective resource
for health? Social Science & Medicine, 62, 292–302.
PRS Group. (2015). International country risk guide.. Available online.
http://epub.prsgroup.com/products/international-country-risk-guide-icrg
Putnam, R. (1988). Diplomacy and domestic politics: The logic of two level
games. International Organization, 42, 427–460.
Putnam, R. (2000). Bowling alone: The collapse and revival of American
community. New York: Simon & Schuster.
Rathbun, B. (2011a). Before hegemony: Generalized trust and the cre-
ation and design of international security organizations. International
Organization, 65, 243–273.
Rathbun, B. (2011b). The ‘magnificent fraud’: Trust, international cooper-
ation and the hidden domestic politics of American multilateralism
after world war II. International Studies Quarterly, 55, 1–21.
Rathbun, B. (2012). From vicious to virtuous circle: Moralistic trust, dif-
fuse reciprocity, and the American security commitment to Europe.
European Journal of International Relations, 18, 323–344.
Rettig, M. (2008). Gacaca: Truth, justice, and reconciliation in postconflict
Rwanda? African Studies Review, 51, 25–50.
Rose, A. K. (2004). Do WTO members have more liberal trade policy?
Journal of International Economics, 63, 209–235.
Please cite this article in press as: Justwan, F., et al. Measuring social trust and trusting the measure. The Social Science
Journal (2017), https://doi.org/10.1016/j.soscij.2017.10.001
View publication stats
11
Rothstein, B., & Stolle, D. (2008). The state and social capital: An institu-
tional theory of generalized trust. Comparative Politics, 40, 441–459.
Rothstein, B., & Uslaner, E. (2005). All for all: Equality, corruption and social
trust. World Politics, 58, 41–72.
Schaap, A. (2005). Political reconciliation. London; New York: Routledge.
Sønderskov, K. M. (2011). Does generalized social trust lead to associ-
ational membership? Unravelling a bowl of well-tossed spaghetti.
European Sociological Review, 27, 419–434.
Svensson, I., & Brouneus, K. (2013). Dialogue and interethnic trust: A ran-
domized field trial of ‘sustained dialogue’ in Ethiopia. Journal of Peace
Research, 50, 561–575.
Transparency International. (2017). Corruption perceptions index.. Avail-
able online. https://www.transparency.org/
Treier, S., & Jackman, S. (2008). Democracy as a latent variable. American
Journal of Political Science, 52, 201–217.
Uslaner, E. (2002). The moral foundations of trust. Cambridge: Cambridge
University Press.
Weeks, J. (2012). Strongmen and straw men: Authoritarian regimes
and the initiation of international conflict. American Political Science
Review, 106, 326–347.
World Bank. (2015). World development indicators.. Available online.
http://data.worldbank.org/data-catalog/world-development-indicators
You, J.-S. (2012). Social trust: Fairness matters more than homogeneity.
Political Psychology, 33, 701–721.
Zak, P., & Fakhar, A. (2006). Neuroactive hormones and interpersonal trust:
International evidence. Economics and Human Biology, 4, 412–429.