Content uploaded by Farah Mallah
Author content
All content in this area was uploaded by Farah Mallah on Nov 12, 2018
Content may be subject to copyright.
!
!
Confidence in the State and Individual Agency (working paper)
1
!
Confidence in the State Shapes Beliefs of Individual Agency
Diwakar Kishore
Farah Mallah
Beliefs shape the culture in any community which impacts the State.
However, States also exercise the power of shaping beliefs. In this paper we
examined the relationship between individual’s confidence in the State and
their beliefs in hard work, agency and their justification of crime. Using,
multi-level models we find a positive association between perception of hard
work and agency. This association is stronger among those with lower
income levels. We also find a negative association between justification of
crime and confidence in the state, but only among non-religious groups.
A body of literature on “Equity Theory” indicates that a person’s motivation and beliefs are often
based on their perception of the fairness of a system they operate in (Siegel, 2007). In that regard, this
research paper is an attempt at finding the association between the confidence in state (the police, the courts
and the central government), and the perception of ones agency, controlling for other variables.
Further, in this research paper, we try and evaluate the association between the confidence in state
and individuals’ perception of whether petty crimes are justified. Sah (2006) presented a model to predict
corruption levels based on the theory that people’s perception of corruption and their experience with it
influences the likelihood of them engaging in corrupt behavior. While, in a recent study by Gachter &
Schulz (2016) they find that individual’s “intrinsic honesty” is lower in countries with higher “prevalence
of rule violation”. Both papers suggest that how people perceive institutions is likely to affect their
behavior. In this paper we explore three main questions:
RQ1: How does an individual’s perception of justifying petty crimes relate to confidence in state?
RQ2: How does an individual’s perception of whether hard work brings success relate to
confidence in state?
RQ3: How does an individual’s perception of freedom of choice and control over their life
outcomes relates to confidence in state?
!
!
Confidence in the State and Individual Agency (working paper)
2
!
I. Description of data
Data source
We are using person level data from WVS-6 (2011-2014) data-set made available by the World
Value Survey, a global network of social scientists studying changing values and their impact on social and
political life. The World Values Survey consists of surveys conducted in almost 100 countries around the
world, using a common questionnaire. For some of the country level variables, we use the QoG Standard
Data made available by the Quality of Government Institute, an independent research institute within the
Department of Political Science at the University of Gothenburg. Additionally, we also use country level
data GDP per capita from the World Bank. The original WVS dataset had 90350 observations from 60
countries. However, we had to eliminate Taiwan and Palestine since we were not able to find country level
information on these countries in other datasets. Additionally, in order to balance the regression results, we
only kept observations which had no missing dependent or independent variables of interest. After cleaning
the data, for final analysis, our dataset includes data from 46 countries, in which there are 48927 individual
level observations.
Level 1 variables
In order to answer the research questions, three relevant outcome variables were used from the
WVS. (1) justifying petty crime; this is based on the composite of three survey questions that capture how
likely individuals are to justify crime. (2) Hard work brings success; captures an individual’s perception of
whether hard work brings success. This variable was standardized; it was originally on a scale from 1 to 10
where 1 indicates completely agreeing with the statement “in the long run, hard work usually brings a better
life” and 10 indicates complement agreeing with the statement “hard work doesn’t generally bring success
– it’s more a matter of luck and connections” (WVS, 2011-2014). We reversed the scale for ease of
interpretation. (3) Control over life outcomes; this captures individual’s perception of freedom of choice
and control over their own lives (this variable was standardized); it was originally on a scale from 1 to 10
!
!
Confidence in the State and Individual Agency (working paper)
3
!
where 1 is “no choice at all” and 10 is “a great deal of choice” to “indicate how much freedom of choice
and control you feel you have over the way your life turns out” (WVS, 2011-2014).
Prior to conducting the principal component analysis to create the first composite indicator for
justification of crime, we calculated the Cronbach’s alpha of the responses of the individuals to four survey
questions: (1) perception of how justifiable it is to claim government benefits to which one is not entitled,
(2) how justifiable it is to avoid a fare in public transport, (3) how justifiable it is to steal property, and (4)
how justifiable it is to cheat on taxes if one has the chance. The Cronbach alpha was estimated to be 0.78.
Cronbach’s alpha allows us to estimate correlation between two sets of observed scores from the replication
of this measurement procedure. Hence, 0.78 is the expected value of the correlations between average scores
and average scores of a replication, in this case replication of items. Alternatively, it can be viewed as the
“observed score variance” that is accounted for by the “true score variance”. Further, looking at the
eigenvectors in the first principal component, it suggests that the indicator variables were similarly weighted
(range 0.47 to 0.52). Additionally, with an Eigenvalue of 2.41, this component explains most of the
variation of all the components, accounting for 60% of total variation. Figures 2, 3 and 4 in appendix E
show case the variation in the main outcome variables by country: (1) Justification of crime, (2) perception
of hard work, and (3) control over ones life.
We controlled for the following level 1 variables: self-reported social class (ranging from 1 “lower
class” to 5 “upper class”; individuals gender; individual’s age; highest level of education as self-reported
(ranging from 1 “no formal education” to 9 “university level education, with degree”); self-reported
religiosity level of an individual (ranging from 1 “atheist” to 3 “religious”); the religious denomination an
individual belongs to (religion - there are 39 religious denomination)1; individual’s perception of equal
opportunity.
The main level 1 predictor variable, Confidence in State, is a composite indicator based on the first
principal component from a principal component analysis of a set of observations of correlated variables,
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
1 We did not treat religious denomination as an outcome variable as to not lose many degrees of freedom, instead we
allowed the intercept to vary by religious denomination as well as the association between religiosity and the
outcome variables.
!
!
Confidence in the State and Individual Agency (working paper)
4
!
recording individual’s confidence in the police, the courts and the central government. The Cronbach’s
alpha of the responses of the individual’s confidence in police, courts and central government is estimated
to be 0.79. Hence, the “observed score variance” that is accounted for by the “true score variance” in this
case is 0.79. Further, looking at the eigenvectors in the first principal component, it suggests that the
indicator variables were similarly weighted (ranging from 0.54 to 0.60). Additionally, with an Eigenvalue
of 2.13, this component explains most of the variation of all the components, accounting for 71% of total
variation.
Level 2 variables
The level 2 predictor variables used for this analysis are: regime type in the countries (which
included 1-Central government was formed based on a democratic election; 2-form of government with a
monarch as a head of state, 3-government that is administrated by military forces, 4-one political party has
the right to form the government, 5-multiple political parties across the political spectrum run for national
election, and all have the capacity to gain control of government offices but the government is not
democratically elected. and 6-others). We also include country’s GDP per capita.2
I. Methodology
We use a linear random effects model to answer the research questions, taking into account both
within and between country variation. People’s perception of agency and justification of crime is dependent
on both country level and individual level factors.
A multilevel model (random effects model in particular) allows us to ask some questions about the
variance, i.e., how much do outcomes vary across countries; hypothesizing two levels of variance: variance
at the individual level (how much do individual viewpoints differ from one another e.g. because of
individual-level covariates like social status, level of education etc.), and variance at the country level (how
much do countries differ from one another e.g. because of country-level covariates like GDP per capita?).
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
2!We scaled the GPD data we got from the World Bank by dividing it by 1000. The average GDP in the dataset is
12.04.!
!
!
Confidence in the State and Individual Agency (working paper)
5
!
It also takes into consideration the variation in the number of observations in each cluster. The
number of observations in a country range from 303 and 2786. The random effects model takes into account
this variation in its estimate. Similarly, in our data set we found 39 different religious denominations, with
observations varying from 1 to 10,000 in each group. Instead of including them as fixed effects, we allowed
the intercept and slope of religiosity to vary by religion.
The Figure 1 below shows substantial variability in confidence in state distribution between
countries. This can be seen by comparing the between-country median score differences. Substantial within-
country variation, meanwhile, is displayed by the spread of scores within each box. Given the clustering of
individuals within country, and apparent relevance to confidence in country, a linear regression approach
ignoring clustering would likely lead to violations of residual independence and normality. We could
address clustering through fixed effects model. However, a random effects approach has several advantages
if key assumptions hold; first it draws upon information in between-country variance to generate more
precise parameter estimates. Second, it is more efficient, avoiding the loss of degrees of freedom required
to estimate fixed country intercepts. Third, it can explicitly model country-level predictors. Based on the
unconditional model (Model 1) in Tables 1-3 in Appendix C, we find that between country variation
accounts for between 16.3% and 8% of the total variation in outcomes. While within country variation
accounts for between 80% and 90% of the total variation in outcomes.
Figure 1: Confidence in State by Country
!
!
Confidence in the State and Individual Agency (working paper)
6
!
We also find no evidence of a non-linear association between our outcomes of interest (control over
life outcomes, hard work brings success and perception of justice) and the dependent variable of interest
“confidence in state”. As part of our preliminary analysis we aggregated the variables of interest at the
country level and examined the trends between them as shown in Appendix A. We also plotted the trend
between the different outcome variables and the confidence in state variable using the pooled data (ignoring
the clustering of the data). The plots are shown in Appendix A and B.
We decided to treat the three outcomes of interest as separate outcomes given that the correlation
between them is not strong as shown in figure 2 below. This suggests that each is capturing a different
construct.
Figure 2: Correlation between the three outcomes variables
Hard work: Belief hard work brings success (and not luck or other variables); Control: belief of
control over own life outcomes. Justify crime: petty crime (bribery, stealing and not paying
taxes) is justified.
!
Hard!work!
Control!
Justify!crime!
!
!
Confidence in the State and Individual Agency (working paper)
7
!
III. Models
We progressively add onto our unconditional model the covariate of interest (confidence in state)
in Model 2, followed by individual level controls in Model 3 and country level controls in Model 4 and
interaction terms in Model 5. We follow the same steps examining each of the three outcome variables of
interest in Tables 1-3. .The equation below summarizes the regressions run. X’ is a matrix of individual
level control and Y’ are country level controls. In Tables 1-3 we did not allow the association between
confidence in state and religiosity to vary by country and religion, respectively. In Table 4 we allow them
to vary (as shown in the equation below). We find that in allowing the association between confidence in
state and the various outcome variable to vary by country the fit of each of the outcomes of interest in
models improve (𝑋"= 2352.6, p < 0.01; 𝑋"= 215.9, p < 0.01; 𝑋"= 45.9, p < 0.01).
We find that Model 4, in each of the taxonomies (Table 1-3) best fits the model, with the lowest
AIC. Therefore, in Table 4 we build on Model 3 (which includes level 2 covariates) and Model 4 estimates
(which include the interaction terms), while additionally allowing the slopes to vary.
IV. Results
We find no statistically significant association between the level of confidence in state an
individual has and the degree of justification of crime in Model 1, Table 4, controlling for individual and
country level characteristics (t = -0.10, p = 0.91). In Model 2, we find a negative association between
confidence in state and the degree of justification of crime among those who self-identify as atheists, (t= -
!
!
Confidence in the State and Individual Agency (working paper)
8
!
3.28, p < 0.05) such that one standard deviation higher level of confidence in state is associated with a 0.07
standard deviation lower degree of justification of crime. We find no statistically significant association
between confidence in state and justification of crime among those who identify as religious and non-
religious (t = 0.77, p = .26; t = 0.73, p = .27). Figure 5 presents the association between confidence in state
and justification of crime for those who identify as atheists compared to those who identify as religious.3
Figure 5: Association between justifying crime and confidence in state by religiosity
The association between confidence in state and justification of crime also varies across countries.
We find that 95% of the observations are estimated to lie between +/- 0.20 points from a typical country’s
association between confidence in state and justification of crime (in standard deviations) as shown in
figure 4 below.4 We observe that some countries seem to have a positive association. This may explain
the low typical association between justifying crime and confidence in state.
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
3 The trend shown is for a typical male in the population. We set the control variables to the average of each.
4 The trend shown is for a typical male in the population. We set the control variables to the average of each.
Self;identified!
religios ity:!
;2!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! !!!!!!!0!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!2!!!!!
!
!
Confidence in the State and Individual Agency (working paper)
9
!
Table 4: Fitted random effects models - final regressions of three outcomes with random slopes
GDP per capita was scaled so one unit is equivalent to 1000. Only interactions terms that are significantly different from 0 were kept in the final
regressions. All used random effects models. The association between religiosity and the various outcomes was allowed to vary by religion. Model
1 and 2: Justifying Crime; Model 3 and 4: Hard work brings success; Model 5 and 6: Control over life outcomes.
!
!
Confidence in the State and Individual Agency (working paper)
10
!
Figure 6: Random slopes and intercepts by country - Justification of crime
On the other hand, we find a positive association between confidence in state and both the belief
hard work leads to success (and it is not luck or connections) (Model 3 and 4 in Table 4). We also find a
positive association between confidence in state and belief one has control over their life outcomes (Model
5 and 6 in Table 4). In Model 3, Table 4, we find one standard deviation higher level of confidence in state
is associated with a 0.039 standard deviation stronger belief that hard work leads to success (t= 3.87, p <
0.001). Similarly, we find that a one standard deviation higher level of confidence in state is associated
with a 0.045 standard deviation stronger belief of control over life outcomes (t= 7.25, p < 0.001). Both
associations vary by social group; They are stronger the lower the socioeconomic group an individual self-
identifies in. We find that a one standard deviation higher level in confidence in state is associated with a
0.06 and 0.01 standard deviation higher belief in hard work among those who identify as “lower class”
compared to those who identify as “upper class”, respectively. In terms of control over life outcomes, we
find that one standard deviation higher level of confidence in state is associated with a 0.058 and 0.026
standard deviation higher perception of control over life outcomes, among those who identify as lower class
!
!
Confidence in the State and Individual Agency (working paper)
11
!
compared to those who identify as upper class, respectively. This association is presented in figure 7 and
8.5
Figure 7: Association between belief in hard work and confidence in state
Figure 8: Association between control over life outcomes and confidence in state
The association between confidence in state and belief in hard work also vary across countries,
though to a less extent compared to the justification of crime. We find that 95% of the observations are
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
5 The trend shown is for a typical male in the population. We set the control variables to the average of each.
Self;identified!
social!class:!
Self;identified! s ocial!
class:!
!
!
Confidence in the State and Individual Agency (working paper)
12
!
estimated to lie between +/- 0.12 points from a typical individuals’ association between confidence in
state and belief in hard work (in standard deviations) as shown in figure 9 below.6 Similarly, We find that
95% of the observations are estimated to lie between by +/- 0.09 points from a typical (male?)
individuals’ association between confidence in state and belief in hard work (in standard deviations) as
shown in figure 10 below.7
Figure 9: Random slopes and intercepts by country - Belief in hard work
Figure 10: Random slopes and intercepts by country - Control over one's life outcomes
!
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
6 The trend shown is for a typical male in the population. We set the control variables to the average of each.
7 The trend shown is for a typical male in the population. We set the control variables to the average of each. Figure
9 is based on regression estimates of Model 3, Table 4 and Figure 10 is based on Model 5, Table 4.
!
!
Confidence in the State and Individual Agency (working paper)
13
!
Sensitivity check
Outcome variable of both “belief in hard work” and “control over one’s own life outcomes” are
ordinal variables. As such, a linear regression may not be the most fitting, though it would allow us to better
compare the results with the justification of crime variable which is continuous. In order to check if the
results are robust, we ran a cumulative logistic mixed methods regression for each of the outcomes. Tables
5 and 6 in the appendix compare the results of the cumulative logistic model (Model 2) with the linear
random effects models (Model 1). We find that the sign and significance of the estimates remain the same,
suggesting that our results accurately represent the data. Given the estimates of Model 2, in Table 5 and 6,
cumulative logistic functions, we find that one standard deviation increase in confidence in state is
associated with around 0.11 logits stronger perception that hard work leads to success as well as control
over life outcomes among those who self-identify at “lower income”, holding all other covariates constant.
V. Limitations and Suggestions for Future Research
In this research paper we were able to shed light on the association between individual’s perception
of state and their day-to-day perception of agency and justification of crime. That said, the generalizability
of this association is limited by a number of factors discussed below.
1.!Study design limitation:
We were constrained to use the variables available in the WVS dataset. While having data on
individual’s perception was informative, having data on their behavior would have been even more accurate
as perception may not transfer to actions. For example, having country level data on crime-rate would have
been helpful. It is also important to consider that there may be a latent factor such as “optimism” which is
influencing an individual’s response to both the confidence in state survey questions and the outcome
variables. Further, in most countries, WVS only surveyed about 1400 people. These people might not be
representative of the population of the country which raises the question of external validity.
2.!Double barreled question
!
!
Confidence in the State and Individual Agency (working paper)
14
!
One of the outcome variables (individual’s perception of freedom of choice) was asked to the survey
respondents in the following way: “Some people feel they have completely free choice and control over
their lives, while other people feel that what they do has no real effect on what happens to them. Please use
this scale where 1 means "no choice at all" and 10 means "a great deal of choice" to indicate how much
freedom of choice and control you feel you have over the way your life turns out (code one number):” This
seems like a double barreled question where respondents are being asked two questions: one regarding
“free choice” and the other regarding “control over their lives”. These two questions might invoke different
responses from the respondents if asked separately. However, the survey only allows for one response.
This, potentially, limits the confidence while extrapolating from our research outputs. Additionally, the
WVS dataset has different scales on measurements for different questions, which is not ideal.
3.!Few clusters
While we opted for a multilevel model for our research, we only have information for 46 countries.
This means that the level 2 variables might have been limited by the fact that there were very few level 2
clusters, which may explain why besides Equal opportunities variable, other level 2 variables were not
significant.
4.!Heteroscedasticity of residuals
We were not able to correct for the potential heteroscedasticity of residuals for the justification of crime
variable, as the lmer model specification does not have that option. Using a different model would have
allowed us to use robust standard errors to account for potential heteroscedasticity, given the strong right
skew of the justification of crime variable. Further, we may be observing a ceiling effect in the case of the
justification of crime variable, given that most individuals stated that each of the petty crimes are never
justified (25% of the observations score -1.15).
VI. Implications:
Our findings imply that, among other things (social class, age and religiosity), people’s confidence
in state machineries is associated with people’s perception of control over their lives; how they perceive
the results of hard work; and their thinking of what petty crimes are justified (though only among
!
!
Confidence in the State and Individual Agency (working paper)
15
!
individuals who identify as atheists). One explanation for why that may be is is a that the state is the only
lawmaker and enforcer in a country. If an individual resides in a highly regressive and/or corrupt state, such
individual might not feel that they have freedom to choose and exercise control over their lives. For
example, if individuals have no confidence in the judiciary, it might be because the judiciary is weak, and
people feel that justice is not served to the individuals. Alternately, other than being lawmaker and enforcer,
States are often the biggest employer and service provider. In such a situation, in a country where people
have very little confidence in a state or its institution, they might not believe that they are getting sufficiently
rewarded for their hard work. Thirdly, most states run on the taxes it collects from the people. If individuals
have low confidence in the state machinery, and they believe that state does not serve them well, it might
lead to people holding beliefs that cheating on taxes is not completely unjustified.
Finally, it is interesting and intuitive to note that social class and age impact people’s perception of
control over their lives; how they perceive the results of hard work; and impacts their thinking of what petty
crimes are justified. While social class impacts various aspects of control over one’s own life and
opportunities of success, it is fascinating to see that it also impacts the justification of petty crimes. More
research needs to go into this aspect. However, with regards to age, it seems fair that it impacts all moral
and value based decision making of an individual.
Further analysis shows that gender and religiosity were statistically significant predictor variables
for all the models we built for predicting the outcome variables individual’s perception of freedom of choice
and control over their life; individual’s perception of hard work bringing success. This suggest that,
controlling for the other variables, male and female categorically differ on these two questions. Given the
amount of discrimination women have experienced in several centuries, it is not surprising that gender of
the respondent impacts their perception of control over life outcomes and perception of hard work bringing
success. However, it is in fact interesting to note that more religious people are more likely to believe that
they have higher control over their life outcomes and that hard work brings success than less religious
people.
!
!
Confidence in the State and Individual Agency (working paper)
16
!
VII. REFERENCES
●!Gächter, S., & Schulz, J. F. (2016). Intrinsic honesty and the prevalence of rule violations across
societies. Nature, 531(7595), 496.
●!Inglehart, R., C. Haerpfer, A. Moreno, C. Welzel, K. Kizilova, J. Diez-Medrano, M. Lagos, P.
Norris, E. Ponarin & B. Puranen et al. (eds.). 2014. World Values Survey: Round Six - Country-
Pooled Datafile Version: http://www.worldvaluessurvey.org/WVSDocumentationWV6.jsp.
Madrid: JD Systems Institute.
●!Teorell, Jan, Stefan Dahlberg, Sören Holmberg, Bo Rothstein, Natalia Alvarado Pachon &
Richard Svensson. 2018. The Quality of Government Standard Dataset, version Jan18. University
of Gothenburg: The Quality of Government Institute, http://www.qog.pol.gu.se
doi:10.18157/QoGStdJan18
●!Siegel, P. H., Schraeder, M., & Morrison, R. (2007). A taxonomy of equity factors. Journal of
Applied Social Psychology, 38(1), 61-75.
●!Sah, R. (2007). Corruption across countries and regions: Some consequences of local
osmosis. Journal of Economic Dynamics and Control, 31(8), 2573-2598.
●!Economides, S., Himmrich, J. (2016) What price autonomy? Brexit’s effect on Britain’s soft
power, trade deals and European security. LSE. Retrived from:
http://blogs.lse.ac.uk/brexit/2016/05/23/what-price-autonomy-brexits-effect-on-britains-soft-
power-trade-deals-and-european-security/
!
!
!
Confidence in the State and Individual Agency (working paper)
17
!
Appendix A: Big Hack – Aggregate data results
!
!
Confidence in the State and Individual Agency (working paper)
18
!
!
!
!
Confidence in the State and Individual Agency (working paper)
19
!
Appendix B: Pooled data – without regard to clustering
!
!
Confidence in the State and Individual Agency (working paper)
20
!
!
!
Confidence in the State and Individual Agency (working paper)
21
!
Appendix C: Fitted Regression Results
Table 1: Fitted Random Effects Model for Justification of Crime
GDP per capital was scaled so one unit is equivalent to 1000. Only interactions terms that are significantly different from 0 were
kept in the final regressions.
!
!
Confidence in the State and Individual Agency (working paper)
22
!
Table 2: Fitted Random Effects Model for Belief Hard Work Brings Success
GDP per capital was scaled so one unit is equivalent to 1000. Only interactions terms that are significantly different from 0 were
kept in the final regressions.
!
!
Confidence in the State and Individual Agency (working paper)
23
!
Table 3: Fitted random effects model for belief one has control over one's own life outcomes
GDP per capital was scaled so one unit is equivalent to 1000. Only interactions terms that are significantly different from 0 were
kept in the final regressions.
Table 4: Fitted random effects models - final regressions of three outcomes with random slopes
!
!
Confidence in the State and Individual Agency (working paper)
24
!
GDP per capital was scaled so one unit is equivalent to 1000. Only interactions terms that are significantly different from 0 were
kept in the final regressions. All used random effects models. The association between religiosity and the various outcomes was
allowed to vary by religion. Model 1 and 2: Justifying Crime; Model 3 and 4: Hard work brings success; Model 5 and 6: Control
over life outcomes.
Table 5: Sensitivity check using cumulative logistic mixed methods - Belief working hard brings success
!
!
Confidence in the State and Individual Agency (working paper)
25
!
Model 2: cumulative logistic model regression estimates. Model 1: linear random effects regression estimates.
Table 5: Sensitivity check using cumulative logistic mixed methods - Belief in control over life outcomes
!
!
Confidence in the State and Individual Agency (working paper)
26
!
Model 2: cumulative logistic model regression estimates. Model 1: linear random effects regression estimates.
Appendix D: Principal Component Analysis outputs
!
!
Confidence in the State and Individual Agency (working paper)
27
!
Justifying Crime
Confidence in state:
!
!
!
Confidence in the State and Individual Agency (working paper)
28
!
Appendix E: Descriptive graphs for outcome variables and independent variable of interest
Figure 1: Confidence in State by Country
Figure 2. Justification of crime, by country
!
!
Confidence in the State and Individual Agency (working paper)
29
!
Figure 3. Perception of hard work, by country
Figure 4. Control over ones life, by country
!
!
Confidence in the State and Individual Agency (working paper)
30
!
Appendix F: Country level controls descriptive statistics
!
!
Confidence in the State and Individual Agency (working paper)
31
!
!
!
!
Confidence in the State and Individual Agency (working paper)
32
!
Appendix G: Individual level control descriptive statistics (Gender, Education, relative social class, age,
religiosity)
!
!
Confidence in the State and Individual Agency (working paper)
33
!
!
!
Confidence in the State and Individual Agency (working paper)
34
!
!
!
Confidence in the State and Individual Agency (working paper)
35
!
!
!
Confidence in the State and Individual Agency (working paper)
36
!
!
!
!
Confidence in the State and Individual Agency (working paper)
37
!
Appendix H: Observed variation in slopes prior to controlling for other variables (unpooled)
!
!
Confidence in the State and Individual Agency (working paper)
38
!
!