UNIVERSITAT

ROVIRA I VIRGILI
DEPARTAMENT D’ECONOMIA

WORKING PAPERS

Col·lecció “DOCUMENTS DE TREBALL DEL

DEPARTAMENT D’ECONOMIA - CREIP”

Education and tax morale

David Rodríguez Justicia

Bernd Theilen

Document de treball n.06- 2017

DEPARTAMENT D’ECONOMIA – CREIP

Facultat d’Economia i Empresa
 UNIVERSITAT
ROVIRA I VIRGILI
DEPARTAMENT D’ECONOMIA

Edita:
Departament d’Economia CREIP
www.fcee.urv.es/departaments/economia/publi www.urv.cat/creip
c_html/index.html Universitat Rovira i Virgili
Universitat Rovira i Virgili Departament d’Economia
Facultat d’Economia i Empresa Av. de la Universitat, 1
Av. de la Universitat, 1 43204 Reus
43204 Reus Tel.: +34 977 758 936
Tel.: +34 977 759 811 Email: creip@urv.cat
Fax: +34 977 758 907
Email: sde@urv.cat

Adreçar comentaris al Departament d’Economia / CREIP

ISSN edició en paper: 1576 - 3382

ISSN edició electrònica: 1988 - 0820

DEPARTAMENT D’ECONOMIA – CREIP

Facultat d’Economia i Empresa
 Education and tax morale

David Rodrı́guez-Justiciaa ∗ and Bernd Theilenb∗

January 24, 2017

Abstract
While the determinants of tax morale have been widely studied in the literature, surprisingly,
the fundamental influence of education on tax morale has yet to be investigated. Given the
insights in the psychological and political science literature about the role of education in the
formation of social values, in this paper, we analyze two channels through which education
shapes tax morale. We find that while the tax morale of individuals that are net receivers of
welfare state benefits increases with their educational level, it decreases with educational level
among those who are net contributors. Furthermore, our results indicate that the more highly
educated, who have been shown to be better able to assess information in the media on public
affairs, exhibit higher levels of tax morale in countries that have better quality public services,
a fairer tax system and more transparent institutions.

JEL classification: H26; H52; I25

Key words: Tax morale; Tax compliance; Education; Welfare state benefits; Trust in public
institutions

a
Corresponding author. Departament d’Economia and CREIP, Universitat Rovira i Virgili, Spain, E-mail:
david.rodriguez@urv.es
b
Departament d’Economia and CREIP, Universitat Rovira i Virgili, Spain, E-mail: bernd.theilen@urv.es
∗
The authors declare that they have no relevant or material financial interests that relate to the research
described in this work.

1
1 Introduction
After the financial crisis in 2008 and the following sovereign debt crisis, many countries
have experienced a substantial fall in their tax revenues and have been obliged to reduce
the size of the public sector and to cut welfare benefits. In such a context, reducing tax
evasion and increasing the tax morale of citizens has become a major objective of public
fiscal policy. In a report for the European Parliament, Murphy (2012) estimates that in
2009 the tax revenue loss due to tax evasion in the EU exceeded e860 billion. This quan-
tity is similar to total healthcare expenditure in the EU. Unsurprisingly, the fight against
tax evasion has become one of the EU’s principle concerns (European Commission, 2016).

The literature agrees that tax evasion is a complex phenomenon which, beyond the
traditional cost-benefit analysis, is influenced by several other factors that together make
up so-called tax morale and which consists of personal values, social norms and attitudes
towards public institutions. While the relationship between an individual’s tax morale
and some of its determinants is already well understood (e.g., tax behavior of the individ-
ual’s reference groups, age, gender, religion), the fundamental influence of education on
tax morale has to date been ignored. This is all the more surprising as the psychology and
political sciences literature tells us that education is an important factor in channeling
individual behavior regarding social values, political attitudes and the general assessment
of public affairs. On the basis of these insights, in this article we use data from the
European Values Study (EVS) to explore the role of education as an indirect channel in
shaping individuals’ tax morale.

The results in the aforementioned literature indicate that there are two main reasons
why educational level should influence tax morale. First, it is an indicator of the higher
cognitive abilities that are necessary to understand the relationship between tax pay-
ments and many of the indirect benefits obtained individually in modern welfare states.
Therefore, we expect individuals who are the main beneficiaries of welfare state benefits
to exhibit higher tax morale. Moreover, this effect should be more pronounced among
the more highly educated because they are more aware of the connection between tax
payments and the benefits received from the state. Indeed, our result indicate that for
unemployed and retired individuals tax morale and the educational level are positively
related. By contrast, for the self-employed, who obtain comparatively lower benefits in
exchange for their tax payments, tax morale and the educational level are negatively re-
lated. The second reason why we expect education to have an influence on tax morale is
that more highly educated citizens are better informed and enabled to process information
from the media. This affects their relationship with public institutions and, consequently,
their tax morale. Our results confirm this conjecture. We find that the more highly ed-
ucated exhibit higher levels of tax morale in countries that have better services, a fairer
tax system and more transparent institutions.

The remainder of this paper is organized as follows. Section 2 relates our study to the
literature. In Section 3 we put forward two hypotheses regarding the role of education
in determining tax morale. Section 4 describes the data and the empirical model. In

2
 Section 5 we discuss the results of our empirical model and provide some robustness
checks. Section 6 concludes.

2 Literature review
As tax compliance is an important topic with a vast amount of literature devoted to
it, this section gives only a brief review of the literature most related to our study.1 We
distinguish between the economic literature based on the traditional tax evasion model and
the multi-discipline literature on tax morale. Furthermore, we comment more specifically
on the literature that has analyzed the impact of education on tax compliance.

2.1 The traditional tax evasion model

The individual motives or incentives for tax evasion have been studied in the framework
of the classical tax evasion model developed by Allingham and Sandmo (1972). This
model analyzes a tax payer’s decision to evade taxes as an expected-utility maximization
problem where effective tax evasion depends, on the one hand, on tax enforcement by the
tax authority (the probability of detection and the severity of fines and sanctions for tax
evaders), and on the other hand, on the benefits from tax evasion (evaded tax payments
by under-declaration). However, as this model has failed to yield good predictions of
the observed tax payers’ behavior in empirical studies or in laboratory experiments, the
recent literature has focused on other factors to explain tax compliance. These are related
to voluntary tax compliance and the concept of tax morale as opposed to enforced tax
compliance on which the traditional tax evasion model relies.

2.2 Tax morale

The concept of tax morale is based on voluntary tax compliance and stems from the psy-
chology and political science literature which has used the concepts of personal and social
norms, trust in authorities and socio-demographic controls to explain tax compliance. In
line with the literature we categorize the associated studies into three branches: personal
and social norms, trust in authorities and further socio-demographic controls.

Personal and social norms. The psychology literature has distinguished between
personal and social norms (Wenzel, 2004; Hofmann et al., 2008). Personal norms comprise
personal values, ethical reasoning, inequality aversion and religious beliefs and deal with
what is generally perceived as good or bad. For example, the religious convictions of
individuals have been proved to be an important factor for voluntary tax compliance
(Grasmick et al., 1991; Stack and Kposowa, 2006; Torgler, 2006). Social norms are socially
shared beliefs about how members of a group should behave and, according to Sigala
et al. (1999), are one of the most important predictors of tax compliance. They find that

1
For more extensive literature reviews on tax evasion and tax morale see, for example, Andreoni et al.
(1998), Hofmann et al. (2008), Kirchler et al. (2008), Pickhardt and Prinz (2014).

3
 a taxpayer’s compliance crucially depends on the perceived tax evasion in her reference
group (friends, neighbours, or colleagues). Finally, Konrad and Qari (2012) find that
patriotic persons exhibit higher levels of tax morale.

Trust in public institutions. As another important factor for tax morale the litera-
ture has identified citizens’ trust in public sector institutions where perceptions regarding
the fairness and efficiency of the welfare state play a prominant role. According to Feld
and Frey (2002), the relationship between taxpayers and authorities can be understood
as an implicit or ‘psychological’ contract. Taxpayers expect that the government provides
goods and services in exchange for their tax payments. As a result, tax compliance is
higher (lower) in situations in which citizens are satisfied (discontent) with the indirect
benefits they receive through the quality and quantity of public provision (e.g., Alm and
Jackson, 1993; Barone and Mocetti, 2011). Regarding the general quality of public provi-
sion it has been shown that ‘trust’ in political leadership and in the public administration
leads to more voluntary tax compliance (e.g., Torgler, 2004, 2005b; Alm et al., 2006).2
In the literature, the quality of public institutions and of service provision has been
measured by the effectiveness of tax deterrence, the treatment of taxpayers by the tax
authority, ethnic fractionalization, institutional transparency (corruption), and income
inequality. Regarding the effectiveness of tax evasion deterrence, a clear relationship be-
tween the intensity of control and the severity of sanctions, on the one hand, and tax
compliance, on the other hand, cannot be established. Following Feld and Frey (2007),
these ambiguous effects of tax deterrence can be explained by the fact that while more
audits reduce tax evasion they can also create an atmosphere of mistrust that reduces tax
compliance (Pommerehne and Frey, 1992). With respect to the treatment of taxpayers
by tax authorities, Frey and Feld (2002) and Feld and Frey (2002) show for Switzerland
that an increased dialogue between tax payers and tax authorities contributes to raising
tax morale. This is particularly the case in cantons that use referendums and initiatives
in political decision making, whereas in cantons with a predilection for representative
decision making a more authoritarian approach is found to be more effective.3 Ethnic
fractionalization is shown to have a negative impact on tax compliance by Lago-Peñas
and Lago-Peñas (2010). Moreover, Torgler (2006) finds that a higher level of perceived
corruption (less institutional transparency) lowers tax morale.4 Finally, Doerrenberg and
Peichl (2013) find that individuals in countries with a more progressive tax rate system
are more likely to exhibit a higher general tax morale whereas, however, this effect de-
creases with the individual income level.

2
Different measures of trust based on individual perceptions have been used in these studies (e.g., trust
in government, trust in the president, trust in the legal system, trust in officials), showing a positive
relationship with tax morale
3
Regarding trust in tax authorities (i.e., the relationship between taxpayers and the tax office), Kirchler
et al. (2008) suggest the ‘slippery slope’ framework for tax compliance in which both the power of tax
authorities (tax enforcement) and trust in the tax authorities are relevant dimensions for understanding
enforced and voluntary compliance.
4
Friedman et al. (2000) show empirically in a cross-country study that corruption and the size of the
shadow economy are positively correlated.

4
 Socio-demographic control variables. In addition to the two aforementioned groups
of variables, most studies include a large number of socio-demographic variables such as
age, gender, occupational status, marital status, income level, and educational level. Re-
garding the impact of these variables on tax morale it has been found that elder (e.g.,
Torgler, 2005b; Martinez-Vazquez and Torgler, 2009), women (e.g., Torgler and Murphy,
2004; Alm and Torgler, 2006; Torgler and Valev, 2010), retired persons (e.g., Torgler,
2005a, 2006; Konrad and Qari, 2012) and married individuals (e.g., Torgler, 2005b; Alm
and Torgler, 2006) exhibit higher levels of tax morale, while the self-employed (e.g., Tor-
gler, 2004; Alm and Torgler, 2006) manifest lower levels of tax morale. Finally, with
respect to the effect of income on tax morale, the results are less clear.5 For example,
Torgler (2006), Alm et al. (2006) and Doerrenberg and Peichl (2013) find a negative re-
lationship between income and tax morale, while Konrad and Qari (2012) and Torgler
et al. (2008) do not find that income has a significant impact on tax morale.

2.3 Education and tax compliance

Despite the large number of studies that analyze the determinants of tax compliance and
tax morale, none of them has comprehensively focused on the role of education in shaping
tax morale. What is known about the influence of education on tax compliance stems
from student questionnaires, country survey data or studies that include education as a
further control. The limitations of these studies are that they have either been based on
a limited number of student questionnaires or on surveys for specific countries. Further-
more, the studies that have included education merely as a further socio-economic control
have obtained no unanimous result for its impact on tax morale.

Regarding the results obtained from studies based on student questionnaires, Chan
et al. (2000) analyze responses from 157 students from two universities, one in the U.S.
and the other in Hong Kong. They observe a negative relationship between educational
level and tax compliance. McGee and Ross (2012) compare student surveys form six
countries and obtain mixed results regarding the relationship between education and tax
compliance. In Brazil, Russia and China the most opposed to tax evasion are individuals
with a low level of education, and in India and the U.S. the more highly educated exhibit
higher tax morale. In contrast, in Germany, it is those with a medium level of education
who exhibit the lowest levels of tax compliance. Finally, Ahmed and Braithwaite (2005)
survey 447 Australian graduates and find that the way in which tertiary education is
financed influences the subsequent tax morale of the more highly educated.

The influence of education on tax morale has also been analyzed using country survey
data. For the Netherlands, Groot and van den Brink (2010) examine a survey data set

5
Notice, that from the theoretical tax evasion models by Allingham and Sandmo (1972) and Yitzhaki
(1974) the predicted influence of income on tax compliance is also ambiguous.

5
from 1996 on criminal behaviour to analyse the effects of education on offences and crimes
committed. Among other results, they obtain that among the 2951 respondents of the
survey the probability of committing tax fraud increases with years of education. The op-
posite result is obtained by Alarcón-Garcı́a et al. (2012) who use Spanish data from a 2007
survey based on 1329 observations to analyze the relationship between gender and tax
morale. They find that the level of education in general and knowledge of fiscal norms in
particular are important determinants of the individual’s declared attitude towards fraud.
In the case of education, they obtain that the individual attitude against fraud increases
with the educational level.

Finally, education has been used as a socio-economic control in numerous studies to

explain tax morale. Studies that identify a positive relationship between education and
tax morale are, for example: Torgler (2005a), who analyzes the relationship between direct
democracy and tax morale for Switzerland; Torgler (2005b), who analyzes the determi-
nants of tax morale in Latin American countries; Konrad and Qari (2012), who explores
the relationship between patriotism and tax morale; Torgler et al. (2008), who investigate
the effect of tax morale on tax compliance for the U.S. and Turkey; and Torgler (2012),
who explores differences in tax morale in 10 eastern European countries. Examples of
studies that find a negative relationship are: Torgler (2006), who analyzes the role of
religiosity on tax morale; Frey and Torgler (2007), who study how perceptions of other
taxpayers’ behaviour influence an individual’s tax morale; Lago-Peñas and Lago-Peñas
(2010), who explore the determinants of tax morale in European countries; and Doerren-
berg and Peichl (2013), who investigate the effect of tax progressivity on tax morale.

3 Hypotheses: Education and tax morale

The previous literature review shows that while educational level has been included in
different studies as an explanatory control variable, its fundamental influence on tax com-
pliance behaviour has been completely ignored. This is all the more surprising as we know
from numerous studies in the fields of psychological and political sciences that the level
of education is an important factor that channels individuals’ behaviour regarding social
values, political attitudes and the general assessment of public affairs. There are two main
reasons why educational level should have a substantial influence on tax compliance be-
haviour. First, a higher educational level is an indicator of higher cognitive abilities Arrow
(1973) which are essential for understanding the functioning of modern welfare states and
the relationship between its general (indirect) benefits and individual tax compliance. For
example, Lewis (1978, 1982) and Niemirowski et al. (2003) find that having obtained a
certain educational level is a necessary condition for understanding tax laws, which in turn
is important for generating trust in authorities. Second, information from media is a key
factor in shaping citizens’ valuation of government performance (e.g., Keohane and Nye,
1998; Newton and Norris, 2000; Norris, 2000) and education is paramount for processing
this information. As formulated by Sniderman et al. (1993), “Citizens acquired through
formal schooling not simply relevant information they required to reason about political

6
choices, but more fundamentally the ability to manipulate information efficiently and to
gather it effectively after they had left school”. Thus, the knowledge gap theory proposed
by Tichenor et al. (1970) states that a higher educational level leads to a greater acquisi-
tion of knowledge from news, which entails that more highly educated citizens are better
informed even when all citizens are exposed to exactly the same information. Eveland
and Scheufele (2000) show that this knowledge gap between the low and high educated
becomes even bigger among light media users. In the same vein, Price and Zaller (1993)
argued that prior knowledge is a key factor for assessing new information. Furthermore,
it has been shown that the more highly educated pay more attention to political mass
media (e.g., McCombs and Shaw, 1972; Freedman and Goldstein, 1999) and are generally
less inclined towards holding a passive attitude to mass media. For example, Johnson and
Kaye (2003) use survey data gathered in the context of the US presidential elections in
1996 and 2000 and find that the amount of time in seeking political information online is
positively associated with the educational level. Finally, education also motivates general
interest in public affairs and civic engagement to the extent that the more highly educated
are more prone to be politically active (Dalton, 2005).

Summarizing both arguments, more highly educated citizens are better able to un-
derstand the functioning of modern welfare states and are more informed about the
performance of governments and public administrations. Both aspects are essential for
shaping individuals’ tax compliance decisions. The decision to voluntarily comply with
tax obligations is a complex and multifaceted issue where individuals, when assessing
the performance of the public sector, have to consider both the personal (direct) benefits
from public service delivery and general (indirect) benefits from a well-organized welfare
state (general quality of public provision, effective deterrence, transparency of public in-
stitutions, income redistribution, etc.). Therefore, educational level should influence an
individual’s tax compliance behavior because it affects both the quality of information
obtained about government and public sector performance, and the understanding of the
relationship between individual compliance and the quality of general public services. In
the following we formulate two hypotheses regarding the influence of educational level on
tax morale.

Our first hypothesis regarding the influence of education on tax morale considers the
link between tax morale and individual direct benefits from tax compliance. According
to the definition by Feld and Frey (2007), tax morale can be understood as the individ-
ual’s intrinsic motivation to pay taxes which is the result of a ‘psychological tax contract’
between citizens and the state where citizens receive goods and services in exchange for
their tax payments. However, the amount of goods and services that an individual re-
ceives from the state is not the same for all. Thus, individuals with children benefit
from public education, the retired from public pensions, and the unemployed from public
unemployment benefits. Therefore, tax morale should vary across individuals according
to their personal situation. Moreover, these considerations should lead us to expect that
education is an important channel that makes citizens conscious of the link between tax
payments and individual (direct) benefits from the tax system. Accordingly, the first
hypothesis we formulate is:

7
 Hypothesis 1: Larger direct benefits from the welfare state positively affect an individ-
ual’s tax morale and the effect increases with the educational level.

The second hypothesis considers the more complex relationship between tax payments
and the general benefits that citizens obtain from a well-organized welfare state. Exam-
ples of these indirect benefits are the general quality of public services, the fairness of the
tax system and the transparency of public institutions. Again, education plays a crucial
role in assessing these indirect benefits which ultimately affect an individual’s intrinsic
motivation to pay taxes. For example, given that assessing public sector performance
requires individuals to pay attention to political mass media and to process the infor-
mation received, more highly educated citizens will be better able to make a less-biased
evaluation. Furthermore, given that evaluating the indirect benefits from tax payments
requires higher cognitive abilities and that these are correlated with educational level, we
also expect this to be a key factor in shaping tax morale. On the basis of these consider-
ations, the second hypothesis we formulate is:

Hypothesis 2: The educational level positively (negatively) affects tax morale when the
indirect benefits from the welfare state are large (small).

Figure 1 displays the general structure of the determinants of tax compliance, as

discussed in Section 2, and summarizes our hypotheses regarding the channels through
which education shapes individuals’ tax morale.

Insert Figure 1 about here.

4 Empirical approach
4.1 Data
The micro-level data is from the 2008 wave of the EVS, which is a commonly used database
in the tax morale literature.6 The EVS particularly suites our aims as it enables the study
of a representative group of individuals for a large set of relatively homogeneous countries.
Out of 47 European countries included in the survey, 29 were finally included in the anal-
ysis, namely, Austria, Belgium, Bulgaria, Croatia, Cyprus, Czech Republic, Denmark,
Estonia, Finland, France, Germany, Greece, Hungary, Ireland, Italy, Latvia, Lithuania,
Luxembourg, Norway, Poland, Portugal, Romania, Slovakia, Slovenia, Spain, Sweden,
Switzerland, Turkey and the United Kingdom.7 Our macro-level data stems from differ-

6
We limit the analysis to the 2008 wave as some of our variables are only available for this year.
7
The country selection criterion responds to data availability. Albania, Armenia, Azerbaijan, Belarus,
Bosnia and Herzegovina, Georgia, Kosovo, Macedonia, Moldova, Montenegro, the Netherlands, Northern
Ireland, the Russian Federation, Serbia and Ukraine have not been included because of either lack of
observations (in most of the cases) or missing data for some of our country-level variables (this is the
case of the Netherlands and Malta).

8
 ent sources which can be consulted in Table 1. Table 2 and Table 3 display the descriptive
statistics for our dichotomous and continuous variables, and for our categorical variables,
respectively.

4.1.1 Dependent variable: tax morale

Our dependent variable (tax morale) stems from the EVS and has been constructed on
the basis of the respondents’ answers to the following question:

Please tell me for each of the following whether you think it can always be justified,
never be justified, or something in between, using this card: ‘Cheating on tax if you have
the chance’

Respondents were asked to assess this issue on a ten-point scale, from 1 (never) to 10
(always). As is common in the literature, the answers were recoded into a four-point scale
where we used the following criterion: responses 7 through 10 were combined into a value
0 (low tax morale), while the remaining responses were combined in groups of two (1 and
2 into 3; 3 and 4 into 2; and 5 and 6 into 1). However, to check the robustness of our
results from the chosen categorization, in Section 5.2 we also use the original ten-point
scale.
The question of whether the responses to this question really provide unbiased data
has been widely discussed in the literature. Following Doerrenberg and Peichl (2013), the
general conclusions of the literature are that such a bias exists (Elffers et al., 1987); that
this bias, however, is lower if one asks about tax morale instead of tax compliance (Frey
and Torgler, 2007); and that for developed countries tax morale (as obtained from the
EVS) and actual tax compliance are highly correlated (Richardson, 2006; Torgler et al.,
2008). Consequently, we consider that this question allows us to obtain an appropriate
measure of tax morale.

4.1.2 Explanatory variables

Our main explanatory variable is education which is separated into three categories: low,
medium and high.8 Low education indicates that an individual has completed (compul-
sory) elementary education or has inadequately completed it. Medium education indicates
that an individual has some sort of secondary school education. Finally, high education
captures those individuals with university studies (completed or uncompleted). The re-
maining variables are grouped into four classes: personal and social norms, direct benefits,
indirect benefits and control variables.

Personal and social norms. As variables that indicate personal and social norms we
consider whether an individual is religious, whether she is patriotic, and whether she is

8
The details on the measurement and definition of the variables and the sources from which the data has
been retrieved are in Table 1.

9
 inequality averse. All these variables are defined as dummy variables were 1 indicates a
positive attitude (religious, patriotic, inequality averse) and 0 a negative attitude (non-
religious, non-patriotic, inequality non-averse). As discussed in Section 2.2, we expect
these variables to be positively related to tax morale.

Direct benefits. As mentioned before, an individual’s intrinsic motivation for volun-

tary tax compliance stems from a psychological tax contract between citizens and the
state where citizens counterbalance their tax obligations with the goods received in ex-
change. As formulated in Hypothesis 1, we expect that an individual that directly benefits
more from public goods provision will have a stronger belief in a moral obligation for tax
compliance. To test this hypothesis, as direct benefits we consider the following variables:
the number of children, employed, unemployed, self-employed, retired, and other. As pub-
lic goods and transfers benefit individuals with children, unemployed individuals, and
retired individuals more than proportionally, we expect both variables to have a positive
relationship with tax morale. In contrast, employed and self-employed individuals are
expected to exhibit lower tax morale as they obtain comparatively lower benefits from
the state in exchange for their tax payments.

Indirect benefits. To test Hypothesis 2, we consider the ‘indirect benefits’ that mod-
ern welfare states offer and that are expected to influence an individual’s tax morale. As
indicators of these indirect benefits we consider public social expenditure, ethnic fraction-
alization, transparency, income inequality, relative redistribution, and deterrence. Public
social expenditure is measured as the share of the sum of public health, education and
social benefits expenditure in GDP.9 Ethnic fractionalization (from Alesina et al., 2003)
is used as an indicator of the match of public goods provided by the state and the prefer-
ences for public goods of citizens (Alesina et al., 1997). Thus, a higher fractionalization
indicates a larger mismatch which should lower tax morale. Transparency measures “the
perceived levels of public-sector corruption in a given country on a scale from zero (highly
corrupt) to ten (highly clean)” (Transparency International, 2008). Income inequality is
measured as the pre-tax Gini index of inequality in equivalized household market income,
which has been found, as mentioned before, to be negatively related to tax morale in some
studies. Relative redistribution is the percentage reduction in market-income inequality
due to taxes and transfers and, therefore, indicates the effectiveness of the public sector
in reducing income inequality. Finally, Deterrence measures the effort and effectiveness
of tax authorities in fighting tax evasion. According to Hypothesis 2 we expect that the
more highly educated, who are supposed to be better informed about government perfor-
mance, to exhibit higher tax morale as a response to positive performance (more social
expenditure, low income inequality, high relative redistribution, low fractionalization, high
transparency, effective deterrence) and vice versa.

9
Missing 2008 values on public education expenditure for Greece, Luxembourg, Romania, and Turkey are
imputed with data from 2005, 2007, 2007 and 2006, respectively. The missing 2008 value on public health
expenditure for Croatia is imputed with the 2012 value.

10
Controls. Finally, as further control variables we include variables commonly used in
the literature, namely Income, Age, Gender, Marital status, and Natural-born citizen. The
expected impact of these variables on tax morale has already been extensively commented
in Section 2.

4.2 Empirical model

As is common in the literature, we use an ordered probit model to account for the ranking
information of our four-point scale dependent variable (tax morale). Due to the non-linear
form of the ordered probit estimation the size of the coefficients should not be directly
interpreted, but the focus should be on the sign and the significance of the estimates. In
all specifications the model includes individual and country level variables (with Germany
as the reference country). Clustered standard errors by country are reported to avoid an
underestimation of standard errors because of intra-group error correlation. Accordingly,
our estimation model is:
∗
yi,c = β 0 xi,c + εi,c (4.1)
where y ∗ is a latent variable (tax morale of individual i in country c), and x is a vec-
∗
tor of explanatory variables. The latent variable yi,c is only observable when it crosses
thresholds:
∗
yi,c =j if αj < j ≤ αj+1 , j = 0, 1, 2, 3,
∗
and the probability that yi,c = j is:
∗
P (yi,c = j) = F (αj+1 − β 0 xi,c ) − F (αj − β 0 xi,c )

where F denotes the standard normal cumulative distribution function. The four cate-
gories for our tax morale variable y ∗ are: low (j = 0), medium low (j = 1), medium high
(j = 2), and high (j = 3).

To test our first hypothesis, we include interactions between the educational level
and the direct benefits variables. Country dummies are used to account for unobserved
country effects. More specifically, our first hypothesis is estimated with the following
model (Models I and II):

β 0 xi,c = Edui,c + P Si,c + DBi,c + DBi,c × Edui,c + CLi,c + νc + ηj + εi,c (4.2)

where Edu indicates educational level dummies (medium, high) with their corresponding
coefficients; P S is a vector of personal and social norms dummies (Religious, Patriotic,
Inequality averse) with their corresponding coefficients; DB is a vector of direct ben-
efits dummies (Number of children, Unemployed, Self-employed, Retired, Other) with
their corresponding coefficients; CL is a vector of control variables (two income dummies
(medium, high), Age, Gender, Married, Widowed, Divorced, Natural-born citizen); νc is
a vector of country dummies; and ηj is a vector of three intercepts for each tax morale
category (middle low, middle high, high as compared to the base category of low). The
individual-level variables used in our empirical models can be considered as uncorrelated.

11
 Notice, that this is also case for education and income whose correlation coefficient is
0.30.10

To test the second hypothesis of whether the educational level acts as an indirect
channel in shaping tax morale when individuals assess the indirect benefits of the welfare
state, we substitute the country-fixed effects in equation 4.2 with different country-level
variables which we interact with the educational level. That is, we estimate the following
models (Models III-VIII):

β 0 xi,c = Edui,c + P Si,c + DBi,c + DBi,c × Edui,c + CLi,c + IBi,c + IBi,c × Edui,c + ηj + εi,c
(4.3)
where IB are the indirect benefit variables (Public social expenditure (Model III), Ethnic
fractionalization (Model IV), Transparency (Model V), Income inequality (Model VI),
Relative redistribution (Model VII), and Deterrence (Model VIII)). The reason for in-
cluding these country-specific variables in alternative model specifications is that some
them are highly correlated with each other.

As mentioned before, we test our first hypothesis by including interactions between

the variables that account for the direct benefits from tax compliance (Number of chil-
dren, Unemployed, Self-employed and Retired ) and dummies indicating the respondents’
educational levels. Thus, hypothesis 1 is tested with 15 (13) different estimates in Specifi-
cation I (II).11 The problem of testing hypothesis 1 is that with 15 independent variables
the chance of finding at least one significative effect is arround 80%.12 There are different
methodologies to adjust the p-values allowing to reduce the chance of Type I errors due
to multiple hypothesis testing (see Newson (2003) for a thorough description of these
methods). Following Elkins et al. (2017), we use a variety of these methods and consider
the impact of a variable as statistically significant if it remains significant for the majority
of the employed adjustment methods whose results are reported in the Appendix. For
our main results, variable estimates that are significant after adjusting the p-value for
multiple hypothesis testing are highlighted in bold Table 4.

10
As can be expected this is not the case for the correlation coefficients of some of control variables Age
with Retired (0.70), or Age with Widowed (0.43). The exclusion of one of these variables, however, does
not change our results qualitatively. More details on correlations can be found in the correlation matrix
which is in the supplementary material to this article.
11
Notice, that multiple hypothesis testing in case of hypothesis 2 is a minor problem as there are only three
interactions in each specification.
12
The probability of having at least one significant effect with a significance level of α = 0.10 and m = 15
variables is 1–(1 − α)m .

12
5 Results
5.1 Regression results
The estimation results are displayed in Table 4. As the interpretation of the estimated
coefficients in the ordered probit estimation model is not straight forward, in the dis-
cussion of the results we concentrate on the significance and the sign of the estimated
coefficients. Specifications I and II contain country fixed effects while specifications III-
VIII include different country contextual-level variables that allow us to test Hypothesis
2. By contrast, as cross-country differences are best accounted for in specifications I and
II we consider these to be most suitable for testing Hypothesis 1 which is related to the
individual characteristics of the respondents.

Specifications I and II differ with respect to the inclusion of education interaction

terms with the variable Number of children. As can be seen from Table 4, in specifi-
cation I these interactions turn out to be non-significant. However, in specification II
without these interactions, the number of children has a significant positive impact on
tax morale. The direct (non-interacted) effect of education on tax morale turns out to be
non-linear. Thus, while individuals with a medium level of education exhibit lower levels
of tax morale, there is no significant difference between individuals with a low level of
education and those with a high level of education. With respect to the influence of per-
sonal and social norms, our results are in line with what has been found in the literature.
Religious, Patriotic and Inequality averse individuals are more likely to manifest higher
tax morale.

Insert Table 4 about here.

We check our first hypothesis by examining the interaction between educational levels
and the variables indicating the direct benefits from tax compliance. As mentioned before,
with regard to the number of children education shows no distinguishable influence on
tax morale. In contrast, for the Unemployed, Self-employed, and Retired variables, educa-
tional level is an important channel for assessing the individual beneficiary status in the
context of the psychological tax contract. Thus, unemployed individuals with a medium
or a high educational level are more likely to exhibit higher tax morale than those with a
low educational level. We take this as evidence for the fact that the more highly educated
are more conscious of the benefits they receive from general tax compliance. The same
is true for retired individuals. In line with this argument, self-employed individuals, who
generally obtain comparably fewer benefits from the state, exhibit lower tax morale when
their educational level is medium or high. Considering these results together, we accept
Hypothesis 1 that education plays an important role in shaping individuals’ tax morale
according to their beneficiary status in the welfare state.

With respect to the controls used in models I and II, generally, the sign and significance
of the estimates are in line with the previous empirical studies which are summarized in

13
 Section 2. For the income level dummies, as in Konrad and Qari (2012), we find no signifi-
cant influence on tax morale.13 Notice, that Natural-born citizen, which to our knowledge
has not been used in previous studies as a determinant of tax morale, is negatively re-
lated to tax morale. According to the ‘psychological tax contract’, one reason for this
may be that citizens who are nationals by birth expect to receive more and better goods
and services from the state in exchange for their tax payments than citizens who are not
nationals by birth.

As mentioned before, we use specifications III-VIII to test the validity of Hypothesis

2. We expect that the individuals with a higher level of education, who are supposed
to be better informed about general public sector performance and the indirect benefits
from the welfare state, will increase (lessen) their tax compliance in response to good
(bad) performance. We use Public social expenditure to carry out our first measurement
of this performance in Specification III. We find that higher public social expenditure
leads to higher tax morale among the medium and highly educated in comparison to the
low educated. Ethnic fractionalization, as a measure of the mismatch between the public
goods received and desired by citizens, causes tax morale to fall as level of education
increases. For the impact of our third and fourth measure, Transparency and Income
inequality, respectively, we find no significant differences between the various educational
levels. Our fifth measure of general public sector performance, Relative redistribution,
has a generally negative impact on tax morale, but tax morale increases with educational
level. Finally, for the impact of Deterrence on tax morale we find no significant differences
between the various educational levels. Taken together, these results lead us to accept
Hypothesis 2; that is, the educational level of individuals affects the extent to which their
tax morale is impacted by the indirect benefits that they obtain from the welfare state.
Good performance (more social expenditure, higher transparency, better relative income
redistribution) increases the tax morale of individuals with a high level of education more
than that of individuals with a low level of education, while bad performance (more frac-
tionalization) decreases the tax morale of individuals with a high level of education more
than that of individuals with a low level of education.

5.2 Robustness checks

To check the robustness of our results we perform three alternative estimations. Firstly,
to test whether the results are sensitive to the categorization of our dependent variable
tax morale, we estimate models I-VIII with the original ten-point scale from the question-
naire. The results are displayed in Table 5, which shows that the results do not change
qualitatively.

Secondly, we use the generalized linear model (GLM) as an alternative estimation

method. For this purpose, we rescale our dependent variable from the four-point scale

13
See also Section 5.2, where the relationship between income and tax morale is further discussed.

14
 to take values between 0 and 1. As shown in Table 6, this also does not lead to any
substantial changes to the results we obtained before.

Finally, even if income and education are only weakly correlated, one might argue that
some of the effects of education on tax morale described in Section 5 stem from income
and not from education.14 To analyze this question, we perform two additional robustness
checks in which income level dummies replace our educational level dummies as interac-
tion terms. In the first one, we replace all education interactions with income interaction,
while in the second one, we substitute only education interactions with country-level vari-
ables. The results are displayed in Table 7 and Table 8, respectively. As can be observed,
at the individual level, none of the interaction coefficients are significant. With respect
to the interactions between income and the country-level variables, although some of the
interaction coefficients are significant (Ethnic fractionalization with high income, and Rel-
ative redistribution with medium income), we observe that the explanatory power of these
models is below that of the corresponding models in Section 5. In conclusion, we take
these results as evidence that the aforementioned impacts on the relationship between
direct and indirect benefits and tax morale are mainly channeled through education and
not income.

6 Conclusions
This study analyzes the role of education in shaping tax morale, a fundamental question
that has been totally ignored by the existing literature. Given the results in the psy-
chological and political science literature, where it is well understood that education is
an important factor in channeling individual behavior regarding social values, and given
that an individual’s intrinsic motivation to pay taxes is the result of a psychological tax
contract, we derive two hypotheses. First, we expect that individuals who obtain higher
direct benefits from the state exhibit higher tax morale and that this effect is more pro-
nounced for the more highly educated because they are more aware of the connection of
between tax payments and benefits received from the state. Our results indicate that
education, indeed, has an important impact on tax morale for those individuals that are
beneficiaries of the welfare state (i.e., the unemployed, the retired, etc.). Second, as the
more highly educated are better informed, we expect that educational level positively
(negatively) affects tax morale when the indirect benefits that citizens obtain from the
welfare state are large (small). Our results confirm this hypothesis. We find that the
more highly educated exhibit higher levels of tax morale in countries with better quality
services, a fairer tax system and more transparent institutions.

14
The income level variable has been introduced as the harmonized household income level categorized
into 3 intervals where the data is directly provided by the EVS. Alternatively, we have used the original
twelve-point scale variable for two additional robustness checks in which Income has been measured either
by dummy variables or as a continuous variable. The results obtained from these two robustness checks
(not reported) do not differ from those obtained using the harmonized variable provided by the EVS.

15
 Some important policy implications can be derived from these findings. First, as some
of the influence of education on tax morale is channeled through better information about
public affairs, it is particularly important to increase information about direct and indirect
benefits of a tax-financed welfare state, especially in the case of the less well educated.
Second, increasing the educational level of the population would be a good instrument
for increasing tax morale and reducing tax evasion. However, this is only the case when
individuals perceive that what they receive in exchange for their tax payments from the
state is of high quality. Otherwise, increasing the educational level of the population
would have just the contrary effect and reduce tax morale. Therefore, the impact on tax
morale of the observed tendency of a steadily increasing mean educational level among
the populations of many European countries should be assessed in the light of the cuts
in social benefits introduced in many of these countries since the financial crisis of 2008.
The next EVS wave might allow us to assess how both tendencies have affected overall
tax morale in European countries.

Acknowledgements
Financial support from the Spanish Ministerio de Ciencia e Innovación under project
ECO2013-42884-P is gratefully acknowledged. We thank Helmut Herwartz, Óscar Martı́nez
Ibáñez, Jordi Sardà Pons and the participants of the GRODE research seminar for helpful
suggestions and comments.

Appendix
Insert Table 1 here.

Insert Table 2 here.

Insert Table 3 here.

Insert Table 5 here.

Insert Table 6 here.

Insert Table 7 here.

Insert Table 8 here.

Insert Table 9 here.

Insert Table 10 here.

16
References
Ahmed, E. and Braithwaite, V. (2005). A need for emotionally intelligent policy: Link-
ing tax evasion with higher education funding. Legal and Criminological Psychology,
10:291–308.

Alarcón-Garcı́a, G., Beyaert, A., and de Pablos, L. (2012). Fiscal awareness : A study of
the female attitude towards fraud in Spain. In M. Pickhardt & A. Prinz (Eds.), Tax
Evasion and the Shadow Economy (pp. 61–86). Cheltenham, UK, Northampton, MA,
USA: Edward Elgar.

Alesina, A., Baqir, R., and Easterly, W. (1997). Public goods and ethnic divisions.

Alesina, A., Devleeschauwer, A., Easterly, W., Kurlat, S., and Wacziarg, R. (2003). Frac-
tionalization. Journal of Economic Growth, 8:155–194.

Allingham, M. G. and Sandmo, A. (1972). Income tax evasion: A theoretical analysis.

Journal of Public Economics, 1:323–338.

Alm, J. and Jackson, B. R. (1993). Fiscal exchange, collective decision institutions, and
tax compliance. Journal of Economic Behavior and Organization, 22:285–303.

Alm, J., Martinez-Vazque, J., and Torgler, B. (2006). Russian attitudes toward pay-
ing taxes – before, during, and after the transition. International Journal of Social
Economics, 33(12):832–857.

Alm, J. and Torgler, B. (2006). Culture differences and tax morale in the United States
and in Europe. Journal of Economic Psychology, 27(2):224–246.

Andreoni, J., Erard, B., and Feinstein, J. (1998). Tax compliance. Journal of Economic
Literature, 36(2):818–860.

Arrow, K. J. (1973). Higher education as a filter. Journal of Public Economics, 2(3):193–

216.

Barone, G. and Mocetti, S. (2011). Tax morale and public spending inefficiency. Inter-
national Tax and Public Finance, 18(6):724–749.

Chan, C. W., Troutman, C. S., and O’Bryan, D. (2000). An expanded model of taxpayer
compliance: Empirical evidence from the United States and Hong Kong. Journal of
International Accounting, Auditing and Taxation, 9(2):83–103.

Dalton, R. J. (2005). The social transformation of trust in government. International

Review of Sociology, 15(1):133–154.

Doerrenberg, P. and Peichl, A. (2013). Progressive taxation and tax morale. Public
Choice, 155(3-4):293–316.

Elffers, H., Weigel, R. H., and Hessing, D. J. (1987). The consequences of different strate-
gies for measuring tax evasion behavior. Journal of Economic Psychology, 8(3):311–337.

17
Elkins, R. K., Kassenboehmer, S. C., and Schurer, S. (2017). The stability of personality
traits in adolescence and young adulthood. Journal of Economic Psychology. forth-
coming.

European Commission (2016). European Commission website (2016). Avail-

able at: http://ec.europa.eu/taxation customs/fight-against-tax-fraud-tax-evasion/
a-huge-problem en. Retrieved in October, 2016.

EUROSTAT (2016). Eurostat database. Available at: http://ec.europa.eu/eurostat/

data/database. Retrieved in August, 2016.

Eveland, W. P. and Scheufele, D. A. (2000). Connecting news media use with gaps in
knowledge and participation. Political Communication, 17(3):215–237.

EVS (2011). European Values Study 2008 (release 3, 2011), 4th wave, integrated dataset.
GESIS data archive, Cologne, Germany, ZA4800 data file version 3.0.0 (2011-11-20).
Available at: http://zacat.gesis.org. Retrieved in December, 2015.

Feld, L. P. and Frey, B. S. (2002). Trust breeds trust: How taxpayers are treated. Eco-
nomics of governance, 3(2):87–99.

Feld, L. P. and Frey, B. S. (2007). Tax compliance as the result of a psychological tax
contract: The role of incentives and responsive regulation. Law and Policy, 29(1):102–
120.

Freedman, P. and Goldstein, K. (1999). Measuring media exposure and the effects of
negative campaign ads. American Journal of Political Science, 43(4):1189–1208.

Frey, B. S. and Feld, L. P. (2002). Deterrence and morale in taxation: An empirical

analysis.

Frey, B. S. and Torgler, B. (2007). Tax morale and conditional cooperation. Journal of
Comparative Economics, 35(1):136–159.

Friedman, E., Johnson, S., Kaufmann, D., and Zoido-Lobaton, P. (2000). Dodging the
grabbing hand: the determinants of unofficial activity in 69 countries. Journal of Public
Economics, 76(3):459–493.

Grasmick, H. G., Bursik, R. J., and Cochran, J. K. (1991). “Render unto Caesar what is
Caesar’s”: Religiosity and taxpayers’ inclinations to cheat. The Sociological Quarterly,
32(2):251–266.

Groot, W. and van den Brink, H. M. (2010). The effects of education on crime. Applied
Economics, 42(3):279–289.

Hofmann, E., Hoelzl, E., and Kirchler, E. (2008). Preconditions of voluntary tax compli-
ance. Zeitschrift für Psychologie / Journal of Psychology, 216(4):209–217.

18
Johnson, T. J. and Kaye, B. K. (2003). A boost or bust for democracy?: How the web
influenced political attitudes and behaviors in the 1996 and 2000 presidential elections.
The International Journal of Press/Politics, 8(3):9–34.

Keohane, R. O. and Nye, J. J. S. (1998). Power and interdependence in the information

age. Foreign Affairs, pages 81–94.

Kirchler, E., Hoelzl, E., and Wahl, I. (2008). Enforced versus voluntary tax compliance:
The ”slippery slope” framework. Journal of Economic Psychology, 29(2):210–225.

Konrad, K. A. and Qari, S. (2012). The last refuge of a scoundrel? Patriotism and tax
compliance. Economica, 79(315):516–533.

Lago-Peñas, I. and Lago-Peñas, S. (2010). The determinants of tax morale in compar-

ative perspective: Evidence from European countries. European Journal of Political
Economy, 26(4):441–453.

Lewis, A. (1978). Perceptions of tax rates. British Tax Review, 6:358–366.

Lewis, A. (1982). The psychology of taxation. Oxford: Martin Robertson.

Martinez-Vazquez, J. and Torgler, B. (2009). The evolution of tax morale in modern

Spain. Journal of Economic Issues, 43(1):1–28.

McCombs, M. E. and Shaw, D. L. (1972). The agenda-setting function of mass media.

Public Opinion Quarterly, 36(2):176–1987.

McGee, R. W. and Ross, A. M. (2012). Education level and ethical attitude toward
tax evasion: A six-country study. Journal of Legal, Ethical and Regulatory Issues,
15(2):93–138.

Murphy, R. (2012). Closing the European tax gap. A report for group of the Progressive
Alliance of Socialists & Democrats in the European Parliament. Technical report, Tax
research LLP, Norfolk.

Newson, R. &. A. S. T. (2003). Multiple-test procedures and smile plots. Stata Journal,
3(2):109–132.

Newton, K. and Norris, P. (2000). Confidence in public institutions: Faith, culture or

performance. In Disaffected democracies. Princeton University Press.

Niemirowski, P., Baldwin, S., and Wearing, A. J. (2003). The influence of tax related
behaviours, beliefs, attitudes, and values on Australian taxpayer compliance. Is tax
avoidance intentional and how serious an offence is it? Journal of Australian Taxation,
6(1):132–165.

Norris, P. (2000). A virtuous circle: Political communications in postindustrial societies.

Cambridge University Press.

19
OECD (2009). Tax administration in OECD and selected non-OECD countries: Compar-
ative information series (2008). Technical report, Centre for Tax Policy and Administra-
tion. Available at: http://www.oecd.org/ctp/administration/CIS-2008.pdf. Retrieved
in March, 2015.

OECD (2011). Tax administration in OECD and selected non-OECD countries: Compar-
ative information series (2010). Technical report, Centre for Tax Policy and Administra-
tion. Available at: http://www.oecd.org/ctp/administration/CIS-2010.pdf. Retrieved
in March, 2015.

OECD (2016). Health spending indicator. OECD Database. Available at: http://data.
oecd.org. Retrieved in August, 2015.

Pickhardt, M. and Prinz, A. (2014). Behavioral dynamics of tax evasion - A survey.

Journal of Economic Psychology, 40:1–19.

Pommerehne, W. W. and Frey, B. S. (1992). The effects of tax administration on tax

morale. ((No. Serie II, Nr. 191). Universität Konstanz, Juristische Fakultät.).

Price, V. and Zaller, J. (1993). Who gets the news? Alternative measures of news
reception and their implications for research. Public Opinion Quarterly, 57(2):133–164.

Richardson, G. (2006). Determinants of tax evasion: A cross-country investigation. Jour-

nal of International Accounting, Auditing and Taxation, 15(2):150–169.

Sigala, M., Burgoyne, C. B., and Webley, P. (1999). Tax communication and social
influence : Evidence from a British sample. 241(June 1998).

Sniderman, P. M., Brody, R. A., and Tetlock, P. E. (1993). Resoning and choice: Explo-
rations in political psychology. Cambridge University Press.

Solt, F. (2014). The Standardized World Income Inequality Database. Available at:
http://dataverse.harvard.edu. Retrieved in December, 2015.

Stack, S. and Kposowa, A. (2006). The effect of religiosity on tax fraud acceptability : A
cross-national analysis. Jounal for the Scientific Study of Religion, 45:325–351.

Tichenor, P. J., Donohue, G. A., and Olien, C. N. (1970). Mass media flow and differential
growth in knowledge. Public Opinion Quarterly, 34(2):159–170.

Torgler, B. (2004). Tax morale in Asian countries. Journal of Asian Economics, 15(2):237–
266.

Torgler, B. (2005a). Tax morale and direct democracy. European Journal of Political
Economy, 21:525–531.

Torgler, B. (2005b). Tax morale in Latin America. Public Choice, 122:133–157.

Torgler, B. (2006). The importance of faith: Tax morale and religiosity. Journal of
Economic Behavior and Organization, 61(1):81–109.

20
Torgler, B. (2012). Tax morale, Eastern Europe and European enlargement. Communist
and Post-Communist Studies, 45(1-2):11–25.

Torgler, B., Demir, I. C., Macintyre, A., and Schaffner, M. (2008). Causes and con-
sequences of tax morale: An empirical investigation. Economic Analysis and Policy,
38(2):313–339.

Torgler, B. and Murphy, K. (2004). Tax morale in Australia: What shapes it and has it
changed over time? Journal of Australian Taxation, 7:298.

Torgler, B. and Valev, N. T. (2010). Gender and public attitudes toward corruption and
tax evasion. Contemporary Economic Policy, 28(4):554–568.

Transparency International (2008). Corruption Perceptions Index (2008). Available at:

http://www.transparency.org/. Retrieved in December, 2015.

Wenzel, M. (2004). An analysis of norm processes in tax compliance. Journal of Economic

Psychology, 25(2):213–228.

World Bank (2016). World Development Indicators. Available at: http://data.worldbank.

org/data-catalog/world-development-indicators. Retrieved in February, 2016.

Yitzhaki, S. (1974). A note on income tax evasion: A theoretical analysis. Journal of

Public Economics, 3:201–202.

21
Tax enforcement Benefits from tax evasion

- Probability of detection - Evaded tax payments

- Fines and sanctions

The traditional tax evasion model

(Enforced tax compliance)

Tax compliance

Tax morale
(Voluntary tax compliance)

Benefits from tax compliance

Personal and social norms (Trust in public institutions)

- Religious
- Patriotic
ed
- Inequality averse
i on uc
at at
uc io
n
ed

Personal (direct) benefits from General (indirect) benefits from

welfare state welfare state
(Public institutions micro-performance) (Public institutions macro-performance)

- Number of children - Public social expenditure

- Employed - Ethnic fractionalization
- Unemployed - Transparency
- Self-employed - Income inequality
- Retired - Relative redistribution
- Deterrence

Figure 1: Education and tax compliance.

22
Variable Definition Measurement Source

Tax morale Respondents’ tax morale Rescaled into a four-point scale. Responses 7 EVS (2011).
through 10 were combined into a value 0 (low tax
morale), while the remaining responses were com-
bined in groups of two (5 and 6 into 1; 3 and 4
into 2; 1 and 2 into 3).

Education Respondents’ educational Three dummy variables (low, medium and high) EVS (2011).
level accounting for whether the respondent has ade-
quately or inadequately completed primary (com-
pulsory), secondary or tertiary education respec-
tively.

Religious Respondents’ religious be- Dichotomous variable taking value 1 if the respon- EVS (2011).
liefs dent declares to be a religious person and 0 if oth-
erwise (not religious or convinced atheist).

Patriotic Respondents’ patriotic Dichotomous variable taking vale 1 if the respon- EVS (2011).
feelings dent declares to be very or quite proud of being a
citizen of the country and 0 otherwise (not very or
not at all proud).

Inequality Respondents’ inequality- Dummy variable taking value 1 for the first two EVS (2011).
averse aversion inclination points in a ten-point scaled answer in which the
respondents indicate their views on income equal-
ity (where value 1 stands for “incomes should be
made more equal”, and 10 “there should be greater
incentives for individual effort”).

Number of Respondents’ number of Continuous variable accounting for the individu- EVS (2011).
children children at home als’ number of children at home.

Unemployed Respondents’ employment Dichotomous variable taking value 1 if the respon- EVS (2011).
status dent is currently unemployed and 0 if otherwise.

Self- Respondents’ employment Dichotomous variable taking value 1 if the respon- EVS (2011).
employed status dent is currently self-employed and 0 if otherwise.

Retired Respondents’ employment Dichotomous variable taking value 1 if the respon- EVS (2011).
status dent is retired/pensioned and 0 if otherwise.

Other Respondents’ employment Dichotomous variable taking value 1 if the respon- EVS (2011).
status dent is (military service, housewife not otherwise
employed, student, not working because of disabil-
ity, other reasons) and 0 if otherwise.

Income Respondents’ income level Three dummy variables (low, medium and high) EVS (2011).
accounting for the respondent’s income level.

Age Respondents’ age Respondent’s age calculated using the year of EVS (2011).
birth.

Gender Respondents’ gender Dichotomous variable taking value 1 for female EVS (2011).
and 0 for male.

Table 1: Data definitions and sources.

23
Variable Definition Measurement Source

Married Respondents’ marital sta- Dichotomous variable taking value 1 if the respon- EVS (2011).
tus as married dent is currently married or in a partnership and
0 if otherwise.

Widowed Respondents’ marital sta- Dichotomous variable taking value 1 if the respon- EVS (2011).
tus as widowed dent is currently widowed and 0 if otherwise.

Divorced Respondents’ marital sta- Dichotomous variable taking value 1 if the respon- EVS (2011).
tus as divorced dent is currently divorced or separated and 0 if
otherwise.

Natural- Country citizenship ob- Dichotomous variable taking value 1 if the respon- EVS (2011).
born tained by birth dent obtained the country citizenship by birth and
citizen 0 if otherwise.

Public so- Public healthcare, educa- Measured as the share of public health, education EUROSTAT (2016).
cial expen- tion and social expenditure and social benefits expenditure in GDP. Health expenditure data
diture for AUT, GRE, IRL, ITA,
TUR, UK is from OECD
(2016).

Ethnic Fractionalization index of F ractionalizationj = 1 − Σn 2

i=1 sij with sij being Alesina et al. (2003).
fractional- ethnic groups by country the share of group i in country j
ization

Transparency Corruption Perceptions In- Measures the perceived levels of public-sector cor- Transparency Interna-
dex ruption in a given country on a scale from zero tional (2008).
(highly corrupt) to ten (highly clean).

Income in- Pre-taxes Gini index Estimate of Gini index of inequality in equivalized Solt (2014).
equality (square root scale) household market (pre-tax and
pre-transfer) income

Relative The percentage reduction The difference between the post-tax gini and pre Own construction using
redistribu- in market-income inequal- tax gini, divided by pre-tax gini, multiplied by 100 data from Solt (2014).
tion ity due to taxes and trans-
fers

Deterrence Deterrence power of the Number of tax administration staff as a proportion Own construction using
administration of the total labor force multiplied by the value of data from OECD (2009,
completed audits as a proportion of total net col- 2011). Total labor force
lections has been taken from the
World Bank (2016).

Table 1: Continued. Data definitions and sources.

24
Variable Mean Std. Dev. Min Max

Individual-level variables
Religious 0.653 0.476 0 1
Patriotic 0.886 0.318 0 1
Inequality averse 0.232 0.422 0 1
Natural-born citizen 0.958 0.201 0 1
Number of children 0.926 1.349 0 13
Age 48.505 17.557 16 108
Gender 0.548 0.498 0 1

Country-level variables
Public social expenditure 32.718 7.892 18.760 46.590
Ethnic fractionalization 0.255 0.162 0.047 0.587
Transparency 6.252 1.718 3.600 9.300
Income inequality 46.685 5.105 33.752 58.393
Relative redistribution 35.398 11.697 3.626 48.858
Deterrence 0.839 0.732 0.007 3.045

Table 2: Descriptive statistics of dummy and continuous variables

25
Variable Value Frequency Percent

Tax morale Low (0) 1,710 5.96

Medium Low (1) 2,543 8.86
Medium High (2) 3,932 13.69
High (3) 20,530 71.50
Total 28,715 100

Educational level Low 4,334 15.09

Medium 17,924 62.42
High 6,457 22.49
Total 28,715 100

Occupational status Employed 13,677 47.63

Unemployed 1,734 6.04
Self-employed 1,514 5.27
Retired 7,345 25.58
Other 4,445 15.48
Total 28,715 100

Income level Low 9,522 33.16

Medium 10,641 37.06
High 8,552 29.78
Total 28,715 100

Marital status Never married 6,665 23.21

Married / partnership 16,160 56.28
Widowed 3,118 10.86
Divorced / separated 2,772 9.65
Total 28,715 100

Table 3: Descriptive statistics of categorical variables.

26
 I II III IV V VI VII VIII
(H1) (H1) (H2) (H2) (H2) (H2) (H2) (H2)

Medium education (Ref.: low) -0.132** -0.107** -1.061** -0.044 -0.680 -1.212 -0.662** -0.121
(0.055) (0.048) (0.361) (0.104) (0.428) (0.798) (0.143) (0.079)
High education (Ref.: low) 0.004 0.013 -1.296** 0.226** -0.930** -0.927 -0.784** -0.028
(0.064) (0.051) (0.410) (0.112) (0.451) (0.798) (0.216) (0.099)

Personal and social norms

Religious (Ref.: non religious) 0.160** 0.160** 0.107** 0.118** 0.120** 0.120** 0.081* 0.098*
(0.031) (0.031) (0.034) (0.049) (0.041) (0.045) (0.036) (0.045)

Patriotic (Ref.: non patriotic) 0.259** 0.259** 0.280** 0.259** 0.275** 0.269** 0.288** 0.285**
(0.037) (0.037) (0.051) (0.049) (0.050) (0.042) (0.053) (0.051)

Inequality averse 0.092** 0.092** 0.123* 0.109* 0.125** 0.118** 0.127** 0.116*
(0.034) (0.034) (0.048) (0.046) (0.047) (0.045) (0.047) (0.049)

Direct benefits
Number of children 0.009 0.023** 0.023** 0.026** 0.026** 0.025** 0.023** 0.029**
(0.013) (0.006) (0.006) (0.007) (0.006) (0.006) (0.006) (0.006)
x medium education 0.022
(0.017)
x high education 0.006
(0.022)

Unemployed (Ref.: employed) -0.280* -0.286** -0.093 0.029 -0.012 -0.070 -0.189 0.001
(0.111) (0.107) (0.146) (0.187) (0.165) (0.156) (0.127) (0.169)
x medium education 0.309** 0.314** 0.199 0.082 0.124 0.181 0.272* 0.094
(0.113) (0.110) (0.125) (0.145) (0.130) (0.128) (0.117) (0.129)
x high education 0.337* 0.346* 0.217 0.086 0.153 0.188 0.293* 0.106
(0.141) (0.137) (0.151) (0.157) (0.151) (0.151) (0.148) (0.150)

Self-employed (Ref.: employed) 0.146 0.142 0.227* 0.334* 0.273* 0.247 0.136 0.295
(0.100) (0.103) (0.110) (0.168) (0.133) (0.133) (0.103) (0.161)
x medium education -0.322** -0.317** -0.353** -0.477** -0.397** -0.378** -0.277** -0.432**
(0.108) (0.110) (0.099) (0.147) (0.115) (0.120) (0.107) (0.149)
x high education -0.431** -0.428** -0.461** -0.607** -0.487** -0.476** -0.371** -0.555**
(0.130) (0.133) (0.127) (0.169) (0.143) (0.137) (0.131) (0.170)

Retired (Ref.: employed) -0.070 -0.062 -0.101 -0.120 -0.132 -0.142 -0.108 -0.137
(0.063) (0.061) (0.098) (0.090) (0.087) (0.088) (0.090) (0.088)
x medium education 0.216** 0.203** 0.269* 0.313** 0.325** 0.309** 0.251** 0.313**
(0.073) (0.071) (0.109) (0.113) (0.114) (0.103) (0.093) (0.108)
x high education 0.184 0.180 0.282* 0.325* 0.338* 0.338** 0.269* 0.323**
(0.097) (0.094) (0.134) (0.135) (0.135) (0.128) (0.119) (0.125)

Other (Ref.: employed) -0.065 -0.066 0.088 0.221 0.181 0.103 -0.009 0.194
(0.069) (0.069) (0.073) (0.151) (0.119) (0.104) (0.068) (0.145)
x medium education 0.137 0.137 -0.008 -0.144 -0.101 -0.025 0.084 -0.117
(0.075) (0.076) (0.066) (0.127) (0.096) (0.093) (0.075) (0.122)
x high education -0.015 -0.011 -0.109 -0.259* -0.203* -0.148 -0.034 -0.242*
(0.087) (0.088) (0.092) (0.120) (0.093) (0.094) (0.095) (0.116)

Controls
Medium income (Ref.: low) -0.001 -0.000 0.008 0.023 0.013 0.010 0.000 0.020
(0.030) (0.030) (0.039) (0.036) (0.041) (0.040) (0.037) (0.041)
High income (Ref.: low) -0.033 -0.033 -0.037 -0.027 -0.031 -0.038 -0.047 -0.030
(0.050) (0.050) (0.052) (0.053) (0.054) (0.056) (0.051) (0.054)

Age 0.009** 0.009** 0.008** 0.007** 0.007** 0.007** 0.009** 0.008**

(0.001) (0.001) (0.001) (0.002) (0.002) (0.002) (0.001) (0.002)

Gender (Ref.: male) 0.185** 0.186** 0.165** 0.165** 0.162** 0.166** 0.170** 0.162**
(0.027) (0.027) (0.033) (0.037) (0.037) (0.032) (0.028) (0.036)

Married (Ref.: never married) 0.018 0.020 0.020 0.036 0.028 0.021 0.000 0.016
(0.022) (0.023) (0.026) (0.029) (0.027) (0.027) (0.028) (0.031)
Widowed (Ref.: never married) -0.047 -0.045 -0.080* -0.035 -0.062 -0.067 -0.105* -0.070
(0.037) (0.038) (0.033) (0.047) (0.041) (0.047) (0.043) (0.055)
Divorced (Ref.: never married) -0.121** -0.120** -0.131** -0.113** -0.125** -0.117** -0.140** -0.130**
(0.033) (0.034) (0.040) (0.037) (0.042) (0.037) (0.042) (0.045)

Natural-born citizen -0.166** -0.166** -0.131* -0.145* -0.122* -0.127* -0.151** -0.154**
(0.048) (0.048) (0.056) (0.059) (0.053) (0.055) (0.056) (0.055)

27
 I II III IV V VI VII VIII
(H1) (H1) (H2) (H2) (H2) (H2) (H2) (H2)

Indirect benefits
Public social expenditure -0.027*
(0.014)
x medium education 0.027*
(0.011)
x high education 0.036**
(0.013)

Ethnic fractionalization 0.184

(0.501)
x medium education -0.697
(0.406)
x high education -1.321**
(0.434)

Transparency -0.068
(0.073)
x medium education 0.075
(0.060)
x high education 0.128
(0.066)

Income inequality -0.036

(0.019)
x medium education 0.020
(0.016)
x high education 0.017
(0.016)

Relative redistribution -0.025**

(0.005)
x medium education 0.015**
(0.004)
x high education 0.021**
(0.006)

Deterrence 0.241
(0.144)
x medium education -0.132
(0.129)
x high education -0.094
(0.131)

Constant cut1 -1.202** -1.179** -1.922** -1.058** -1.477** -2.794** -1.859** -0.888**
(0.115) (0.107) (0.496) (0.153) (0.526) (0.949) (0.213) (0.110)
Constant cut2 -0.644** -0.622** -1.389** -0.522** -0.944 -2.260* -1.324** -0.355**
(0.114) (0.107) (0.496) (0.153) (0.527) (0.951) (0.212) (0.112)
Constant cut3 -0.124 -0.102 -0.893 -0.024 -0.449 -1.763 -0.826** 0.142
(0.118) (0.111) (0.500) (0.160) (0.533) (0.953) (0.213) (0.119)

Observations 28,715 28,715 28,715 28,715 28,715 28,715 28,715 28,715

Countries FE YES YES NO NO NO NO NO NO
Pseudo R2 0.0598 0.0597 0.0277 0.0304 0.0268 0.0291 0.0313 0.0285

Table 4: Estimation results. Ordered probit with clustered standard errors by country
(29 clusters). Robust standard errors in parentheses where ** indicate p < 0.01, and *
p < 0.05, respectively. Estimates in bold denote significant effects after adjusting the p-
value for multiple hypothesis testing by different methods using 0.1 level as false discovery
rate (see Table 10).

28
 I II III IV V VI VII VIII
(H1) (H1) (H2) (H2) (H2) (H2) (H2) (H2)

Medium education (Ref.: low) -0.073 -0.053 -1.066** 0.009 -0.636 -1.147 -0.566** -0.119+
(0.059) (0.055) (0.302) (0.091) (0.400) (0.714) (0.130) (0.071)
High education (Ref.: low) 0.040 0.045 -1.235** 0.204* -0.834* -0.995 -0.671** -0.043
(0.067) (0.058) (0.352) (0.095) (0.418) (0.718) (0.211) (0.090)

Personal and social norms

Religious (Ref.: non religious) 0.150** 0.151** 0.117** 0.128** 0.129** 0.128** 0.092** 0.107*
(0.027) (0.027) (0.032) (0.047) (0.039) (0.044) (0.033) (0.042)

Patriotic (Ref.: non patriotic) 0.243** 0.243** 0.266** 0.246** 0.262** 0.257** 0.274** 0.273**
(0.033) (0.033) (0.049) (0.047) (0.047) (0.040) (0.050) (0.048)

Inequality averse 0.123** 0.123** 0.152** 0.139** 0.154** 0.149** 0.157** 0.145**
(0.029) (0.029) (0.043) (0.041) (0.041) (0.039) (0.042) (0.043)

Direct benefits
Number of children 0.009 0.019** 0.018** 0.023** 0.023** 0.021** 0.019** 0.025**
(0.013) (0.006) (0.005) (0.006) (0.006) (0.005) (0.005) (0.006)
x medium education 0.017
(0.016)
x high education 0.002
(0.019)

Unemployed (Ref.: employed) -0.179 -0.183 -0.023 0.103 0.065 0.015 -0.099 0.083
(0.127) (0.123) (0.156) (0.193) (0.174) (0.167) (0.141) (0.179)
x medium education 0.198+ 0.202+ 0.119 -0.004 0.035 0.085 0.172 -0.002
(0.119) (0.115) (0.127) (0.145) (0.132) (0.131) (0.121) (0.133)
x high education 0.244+ 0.252+ 0.157 0.030 0.079 0.115 0.213 0.035
(0.137) (0.133) (0.146) (0.156) (0.147) (0.148) (0.141) (0.146)

Self-employed (Ref.: employed) 0.209* 0.207* 0.289* 0.394* 0.335* 0.317* 0.208+ 0.359*
(0.103) (0.103) (0.123) (0.181) (0.151) (0.150) (0.116) (0.174)
x medium education -0.354** -0.350** -0.381** -0.503** -0.426** -0.414** -0.315** -0.465**
(0.103) (0.104) (0.099) (0.157) (0.129) (0.131) (0.108) (0.159)
x high education -0.476** -0.474** -0.500** -0.636** -0.529** -0.526** -0.423** -0.599**
(0.122) (0.123) (0.122) (0.170) (0.150) (0.143) (0.128) (0.172)

Retired (Ref.: employed) 0.015 0.020 -0.022 -0.044 -0.055 -0.061 -0.027 -0.060
(0.065) (0.063) (0.093) (0.077) (0.072) (0.080) (0.087) (0.072)
x medium education 0.128+ 0.117+ 0.190* 0.238** 0.249** 0.232** 0.172* 0.237**
(0.067) (0.066) (0.094) (0.092) (0.091) (0.087) (0.082) (0.086)
x high education 0.126 0.125 0.229+ 0.277* 0.287* 0.281* 0.214* 0.272**
(0.092) (0.090) (0.119) (0.115) (0.113) (0.111) (0.107) (0.104)

Other (Ref.: employed) -0.039 -0.040 0.112+ 0.247+ 0.211* 0.148 0.028 0.225+
(0.059) (0.059) (0.067) (0.135) (0.104) (0.096) (0.064) (0.126)
x medium education 0.098 0.098 -0.036 -0.176 -0.136+ -0.074 0.041 -0.154
(0.062) (0.062) (0.056) (0.109) (0.077) (0.082) (0.064) (0.100)
x high education -0.013 -0.010 -0.105 -0.253* -0.205* -0.163+ -0.045 -0.244*
(0.086) (0.086) (0.093) (0.114) (0.088) (0.096) (0.094) (0.108)

Controls
Medium income (Ref.: low) -0.027 -0.027 -0.014 0.002 -0.009 -0.011 -0.022 0.000
(0.028) (0.028) (0.035) (0.034) (0.037) (0.036) (0.034) (0.035)
High income (Ref.: low) -0.071 -0.071 -0.069 -0.058 -0.063 -0.068 -0.079+ -0.060
(0.044) (0.044) (0.046) (0.047) (0.048) (0.050) (0.045) (0.047)

Age 0.008** 0.008** 0.007** 0.006* 0.006** 0.007** 0.008** 0.007**

(0.001) (0.001) (0.002) (0.003) (0.002) (0.002) (0.001) (0.002)

Gender (Ref.: male) 0.174** 0.174** 0.153** 0.153** 0.149** 0.153** 0.157** 0.151**
(0.025) (0.025) (0.032) (0.036) (0.035) (0.031) (0.027) (0.035)

Married (Ref.: never married) 0.034+ 0.036+ 0.037 0.053+ 0.045+ 0.039 0.019 0.034
(0.020) (0.020) (0.025) (0.028) (0.026) (0.027) (0.028) (0.030)
Widowed (Ref.: never married) -0.050 -0.049 -0.082* -0.038 -0.064 -0.070 -0.107* -0.070
(0.035) (0.035) (0.035) (0.046) (0.040) (0.044) (0.043) (0.052)
Divorced (Ref.: never married) -0.094** -0.093** -0.102** -0.084* -0.096* -0.091** -0.111** -0.099*
(0.028) (0.029) (0.037) (0.034) (0.039) (0.035) (0.039) (0.042)

Natural-born citizen -0.156** -0.156** -0.119* -0.131* -0.112* -0.114* -0.138** -0.145**
(0.045) (0.045) (0.051) (0.056) (0.047) (0.051) (0.052) (0.051)

29
 I II III IV V VI VII VIII
(H1) (H1) (H2) (H2) (H2) (H2) (H2) (H2)

Indirect benefits
Public social expenditure -0.028*
(0.012)
x medium education 0.028**
(0.009)
x high education 0.035**
(0.011)

Ethnic fractionalization 0.259

(0.486)
x medium education -0.795*
(0.386)
x high education -1.250**
(0.407)

Transparency -0.067
(0.069)
x medium education 0.073
(0.057)
x high education 0.114+
(0.062)

Income inequality -0.033+

(0.017)
x medium education 0.020
(0.015)
x high education 0.018
(0.015)

Relative redistribution -0.023**

(0.004)
x medium education 0.013**
(0.003)
x high education 0.018**
(0.005)

Deterrence 0.221+
(0.131)
x medium education -0.094
(0.115)
x high education -0.066
(0.118)

Constant cut1 -1.754** -1.737** -2.533** -1.641** -2.067** -3.234** -2.379** -1.504**
(0.120) (0.120) (0.443) (0.146) (0.514) (0.872) (0.205) (0.101)
Constant cut2 -1.623** -1.606** -2.405** -1.513** -1.939** -3.106** -2.251** -1.376**
(0.114) (0.113) (0.438) (0.142) (0.509) (0.869) (0.198) (0.099)
Constant cut3 -1.368** -1.350** -2.158** -1.264** -1.691** -2.859** -2.004** -1.128**
(0.106) (0.103) (0.433) (0.139) (0.503) (0.866) (0.189) (0.097)
Constant cut4 -1.146** -1.129** -1.944** -1.049** -1.478** -2.646** -1.791** -0.914**
(0.105) (0.102) (0.432) (0.141) (0.503) (0.865) (0.186) (0.101)
Constant cut5 -0.944** -0.927** -1.750** -0.854** -1.284* -2.451** -1.596** -0.720**
(0.107) (0.104) (0.434) (0.142) (0.505) (0.865) (0.187) (0.103)
Constant cut6 -0.591** -0.574** -1.412** -0.514** -0.946+ -2.113* -1.257** -0.381**
(0.103) (0.100) (0.432) (0.139) (0.504) (0.867) (0.185) (0.102)
Constant cut7 -0.389** -0.372** -1.219** -0.320* -0.753 -1.919* -1.063** -0.188+
(0.105) (0.102) (0.434) (0.142) (0.506) (0.867) (0.185) (0.105)
Constant cut8 -0.072 -0.055 -0.916* -0.016 -0.451 -1.616+ -0.759** 0.115
(0.105) (0.102) (0.435) (0.145) (0.509) (0.869) (0.184) (0.108)
Constant cut9 0.311** 0.328** -0.549 0.352* -0.084 -1.248 -0.390* 0.483**
(0.108) (0.104) (0.433) (0.149) (0.507) (0.867) (0.181) (0.109)

Observations 28,715 28,715 28,715 28,715 28,715 28,715 28,715 28,715

Countries FE YES YES NO NO NO NO NO NO
Pseudo R2 0.0421 0.0421 0.0206 0.0220 0.0196 0.0209 0.0230 0.0215

Table 5: Estimation results. Ordered probit with clustered standard errors by country (29
countries). Robust standard errors in parentheses where ** indicate p < 0.01, * p < 0.05
and + p < 0.1 respectively. The independent variable is tax morale in a ten-point scale.

30
 I II III IV V VI VII VIII
(H1) (H1) (H2) (H2) (H2) (H2) (H2) (H2)

Medium education (Ref.: low) -0.108* -0.079 -0.908** -0.017 -0.568 -0.980 -0.588** -0.090
(0.048) (0.043) (0.332) (0.093) (0.388) (0.746) (0.134) (0.073)
High education (Ref.: low) 0.028 0.040 -1.130** 0.238* -0.801* -0.666 -0.681** 0.003
(0.057) (0.046) (0.374) (0.101) (0.409) (0.741) (0.191) (0.092)

Personal and social norms

Religious (Ref.: non religious) 0.144** 0.144** 0.091** 0.101* 0.105** 0.101* 0.067* 0.082*
(0.028) (0.029) (0.031) (0.045) (0.037) (0.041) (0.033) (0.041)

Patriotic (Ref.: non patriotic) 0.253** 0.253** 0.272** 0.252** 0.267** 0.262** 0.279** 0.277**
(0.035) (0.035) (0.048) (0.045) (0.046) (0.039) (0.049) (0.047)

Inequality averse 0.060* 0.060* 0.088 0.075 0.090* 0.084 0.091* 0.082
(0.030) (0.030) (0.045) (0.043) (0.044) (0.043) (0.045) (0.046)

Direct benefits
Number of children 0.003 0.020** 0.020** 0.023** 0.023** 0.023** 0.020** 0.026**
(0.012) (0.006) (0.006) (0.006) (0.006) (0.006) (0.006) (0.006)
x medium education 0.026
(0.016)
x high education 0.008
(0.020)

Unemployed (Ref.: employed) -0.250** -0.256** -0.077 0.037 -0.000 -0.054 -0.165 0.006
(0.092) (0.088) (0.126) (0.159) (0.142) (0.132) (0.109) (0.143)
x medium education 0.275** 0.281** 0.171 0.061 0.100 0.152 0.240* 0.076
(0.094) (0.090) (0.109) (0.122) (0.111) (0.108) (0.101) (0.108)
x high education 0.264* 0.275* 0.151 0.030 0.089 0.124 0.223 0.053
(0.124) (0.121) (0.135) (0.136) (0.133) (0.134) (0.133) (0.132)

Self-employed (Ref.: employed) 0.133 0.128 0.210 0.303 0.254 0.230 0.129 0.266
(0.098) (0.099) (0.114) (0.164) (0.138) (0.135) (0.105) (0.155)
x medium education -0.295** -0.288** -0.327** -0.437** -0.369** -0.352** -0.260** -0.394**
(0.099) (0.101) (0.096) (0.140) (0.117) (0.117) (0.100) (0.139)
x high education -0.404** -0.400** -0.438** -0.569** -0.462** -0.455** -0.358** -0.517**
(0.126) (0.127) (0.126) (0.163) (0.145) (0.137) (0.127) (0.162)

Retired (Ref.: employed) -0.046 -0.037 -0.078 -0.095 -0.103 -0.113 -0.083 -0.110
(0.060) (0.057) (0.090) (0.083) (0.080) (0.081) (0.083) (0.081)
x medium education 0.182** 0.166* 0.232* 0.270** 0.281** 0.266** 0.215* 0.270**
(0.067) (0.065) (0.100) (0.104) (0.104) (0.095) (0.085) (0.098)
x high education 0.145 0.140 0.239 0.277* 0.288* 0.287* 0.226* 0.276*
(0.091) (0.088) (0.124) (0.125) (0.125) (0.120) (0.112) (0.117)

Other (Ref.: employed) -0.050 -0.051 0.076 0.194 0.158 0.092 -0.011 0.165
(0.068) (0.068) (0.072) (0.142) (0.112) (0.100) (0.068) (0.134)
x medium education 0.116 0.116 -0.006 -0.127 -0.089 -0.023 0.077 -0.097
(0.073) (0.073) (0.066) (0.122) (0.092) (0.091) (0.075) (0.115)
x high education -0.048 -0.044 -0.123 -0.255* -0.207* -0.165 -0.057 -0.237*
(0.079) (0.080) (0.083) (0.112) (0.083) (0.088) (0.089) (0.106)

Controls
Medium income (Ref.: low) 0.005 0.005 0.013 0.028 0.017 0.013 0.006 0.024
(0.029) (0.029) (0.038) (0.035) (0.039) (0.038) (0.036) (0.038)
High income (Ref.: low) -0.023 -0.024 -0.026 -0.015 -0.020 -0.028 -0.035 -0.020
(0.046) (0.046) (0.049) (0.049) (0.051) (0.052) (0.048) (0.050)

Age 0.008** 0.008** 0.007** 0.006** 0.006** 0.007** 0.008** 0.007**

(0.001) (0.001) (0.001) (0.002) (0.002) (0.002) (0.001) (0.002)

Gender (Ref.: male) 0.175** 0.176** 0.160** 0.161** 0.158** 0.162** 0.164** 0.159**
(0.026) (0.026) (0.030) (0.034) (0.033) (0.029) (0.026) (0.033)

Married (Ref.: never married) 0.013 0.016 0.014 0.029 0.022 0.015 -0.004 0.010
(0.021) (0.022) (0.024) (0.028) (0.026) (0.026) (0.027) (0.030)
Widowed (Ref.: never married) -0.044 -0.042 -0.075* -0.034 -0.057 -0.064 -0.098* -0.067
(0.035) (0.036) (0.032) (0.044) (0.039) (0.044) (0.041) (0.051)
Divorced (Ref.: never married) -0.122** -0.121** -0.129** -0.113** -0.124** -0.118** -0.138** -0.130**
(0.030) (0.031) (0.036) (0.034) (0.038) (0.033) (0.038) (0.041)

Natural-born citizen -0.144** -0.144** -0.112* -0.127* -0.103* -0.110* -0.131* -0.134**
(0.044) (0.044) (0.053) (0.055) (0.049) (0.051) (0.052) (0.052)

31
 I II III IV V VI VII VIII
(H1) (H1) (H2) (H2) (H2) (H2) (H2) (H2)

Indirect benefits -0.024

Public social expenditure (0.013)
0.023*
x medium education (0.010)
0.033**
x high education (0.012)

Ethnic fractionalization 0.124

(0.455)
x medium education -0.622
(0.363)
x high education -1.192**
(0.391)

Transparency -0.055
(0.067)
x medium education 0.064
(0.055)
x high education 0.114
(0.060)

Income inequality -0.031

(0.018)
x medium education 0.016
(0.015)
x high education 0.012
(0.015)

Relative redistribution -0.022**

(0.005)
x medium education 0.014**
(0.003)
x high education 0.019**
(0.005)

Deterrence 0.216
(0.135)
x medium education -0.116
(0.119)
x high education -0.083
(0.123)

Constant 0.599** 0.573** 1.199** 0.452** 0.786 1.919* 1.157** 0.293**

(0.108) (0.102) (0.455) (0.148) (0.479) (0.884) (0.206) (0.108)

Observations 28,715 28,715 28,715 28,715 28,715 28,715 28,715 28,715

Countries FE YES YES NO NO NO NO NO NO
AIC .734568 .7345047 .7686598 .7653929 .7694225 .7672753 .7653959 .7678534
BIC -278892 -278910.4 -277913.1 -278006.9 -277891.2 -277952.8 -278006.8 -277936.2

Table 6: Estimation results. GLM with clustered standard errors by country (29 clusters).
Robust standard errors in parentheses where ** indicate p < 0.01, and * p < 0.05,
respectively.

32
 I II III IV V VI VII VIII
(H1) (H1) (H2) (H2) (H2) (H2) (H2) (H2)

Medium income (Ref.: low) -0.031 -0.017 0.094 0.078 0.110 0.498 0.175 -0.043
(0.044) (0.040) (0.196) (0.061) (0.150) (0.311) (0.091) (0.056)
High income (Ref.: low) -0.064 -0.054 -0.262 0.104 -0.124 0.371 0.060 -0.107
(0.069) (0.059) (0.226) (0.104) (0.181) (0.485) (0.098) (0.068)

Personal and social norms

Religious (Ref.: non religious) 0.161** 0.161** 0.108** 0.118* 0.124** 0.125** 0.079* 0.102*
(0.031) (0.031) (0.035) (0.050) (0.043) (0.046) (0.037) (0.046)

Patriotic (Ref.: non patriotic) 0.261** 0.260** 0.286** 0.261** 0.278** 0.270** 0.295** 0.289**
(0.037) (0.037) (0.054) (0.050) (0.051) (0.042) (0.053) (0.051)

Inequality averse 0.093** 0.093** 0.129** 0.112* 0.129** 0.122** 0.128** 0.118*
(0.035) (0.035) (0.050) (0.048) (0.048) (0.046) (0.048) (0.050)

Direct benefits
Number of children 0.012 0.022** 0.030** 0.028** 0.031** 0.027** 0.029** 0.030**
(0.013) (0.007) (0.006) (0.008) (0.007) (0.007) (0.006) (0.007)
x medium income 0.018
(0.015)
x high income 0.013
(0.017)

Unemployed (Ref.: employed) -0.011 -0.013 0.116 0.125 0.129 0.121 0.098 0.112
(0.059) (0.059) (0.084) (0.090) (0.091) (0.088) (0.077) (0.090)
x medium income 0.018 0.022 -0.069 -0.064 -0.075 -0.063 -0.080 -0.064
(0.101) (0.100) (0.093) (0.088) (0.096) (0.091) (0.089) (0.087)
x high income -0.086 -0.084 -0.080 -0.098 -0.082 -0.119 -0.132 -0.092
(0.099) (0.100) (0.119) (0.118) (0.118) (0.113) (0.107) (0.114)

Self-employed (Ref.: employed) -0.166 -0.169 -0.066 -0.072 -0.058 -0.070 -0.089 -0.084
(0.090) (0.089) (0.106) (0.115) (0.110) (0.106) (0.092) (0.115)
x medium income -0.038 -0.034 -0.075 -0.091 -0.084 -0.082 -0.081 -0.072
(0.086) (0.086) (0.081) (0.086) (0.084) (0.081) (0.076) (0.089)
x high income 0.009 0.013 -0.039 -0.052 -0.039 -0.044 -0.038 -0.043
(0.109) (0.110) (0.109) (0.115) (0.111) (0.113) (0.104) (0.114)

Retired (Ref.: employed) 0.069 0.073 0.055 0.066 0.066 0.054 0.052 0.052
(0.045) (0.044) (0.045) (0.048) (0.045) (0.047) (0.046) (0.051)
x medium income 0.064 0.057 0.092 0.086 0.085 0.088 0.081 0.085
(0.069) (0.068) (0.074) (0.075) (0.074) (0.075) (0.073) (0.075)
x high income 0.042 0.037 0.082 0.063 0.081 0.078 0.063 0.062
(0.068) (0.066) (0.085) (0.082) (0.084) (0.077) (0.075) (0.080)

Other (Ref.: employed) 0.024 0.023 0.058 0.058 0.060 0.053 0.053 0.059
(0.072) (0.071) (0.082) (0.085) (0.086) (0.079) (0.075) (0.083)
x medium income 0.004 0.006 0.039 0.062 0.048 0.046 0.013 0.045
(0.087) (0.087) (0.093) (0.096) (0.095) (0.088) (0.087) (0.091)
x high income 0.047 0.048 0.098 0.102 0.100 0.074 0.051 0.076
(0.085) (0.086) (0.093) (0.096) (0.095) (0.091) (0.087) (0.092)

Controls
Medium education (Ref.: low) 0.023 0.024 -0.126 -0.133 -0.140 -0.137 -0.054 -0.135
(0.051) (0.051) (0.117) (0.148) (0.135) (0.122) (0.057) (0.133)
High education (Ref.: low) 0.116 0.117* -0.030 -0.042 -0.053 -0.033 0.051 -0.032
(0.059) (0.059) (0.110) (0.145) (0.127) (0.121) (0.066) (0.132)

Age 0.009** 0.009** 0.008** 0.007** 0.007** 0.008** 0.009** 0.008**

(0.001) (0.001) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002)

Gender (Ref.: male) 0.187** 0.187** 0.164** 0.166** 0.163** 0.169** 0.169** 0.166**
(0.027) (0.027) (0.033) (0.036) (0.035) (0.031) (0.028) (0.034)

Married (Ref.: never married) 0.021 0.022 0.032 0.046 0.043 0.035 0.016 0.029
(0.024) (0.024) (0.027) (0.034) (0.031) (0.032) (0.028) (0.036)
Widowed (Ref.: never married) -0.044 -0.042 -0.068* -0.033 -0.045 -0.055 -0.093* -0.061
(0.037) (0.037) (0.034) (0.049) (0.044) (0.050) (0.043) (0.055)
Divorced (Ref.: never married) -0.119** -0.119** -0.125** -0.107** -0.119** -0.108** -0.133** -0.123**
(0.033) (0.033) (0.041) (0.039) (0.043) (0.040) (0.041) (0.046)

Natural-born citizen -0.168** -0.167** -0.123* -0.133* -0.112* -0.124* -0.145* -0.146**
(0.048) (0.048) (0.058) (0.062) (0.055) (0.056) (0.057) (0.057)

33
 I II III IV V VI VII VIII
(H1) (H1) (H2) (H2) (H2) (H2) (H2) (H2)

Indirect benefits
Public social expenditure -0.003
(0.010)
x medium income -0.003
(0.005)
x high income 0.006
(0.007)

Ethnic fractionalization -0.253

(0.331)
x medium income -0.327
(0.182)
x high income -0.652*
(0.285)

Transparency 0.018
(0.035)
x medium income -0.020
(0.021)
x high income 0.008
(0.026)

Income inequality -0.013

(0.011)
x medium income -0.011
(0.007)
x high income -0.009
(0.011)

Relative redistribution -0.009

(0.008)
x medium income -0.005*
(0.003)
x high income -0.004
(0.003)

Deterrence 0.093
(0.082)
x medium income 0.049
(0.053)
x high income 0.058
(0.070)

Constant cut1 -1.077** -1.068** -1.068* -1.079** -0.857** -1.579** -1.212** -0.904**
(0.107) (0.103) (0.433) (0.183) (0.318) (0.593) (0.347) (0.174)
Constant cut2 -0.520** -0.511** -0.537 -0.544** -0.326 -1.046 -0.679* -0.372*
(0.102) (0.098) (0.435) (0.183) (0.321) (0.596) (0.345) (0.174)
Constant cut3 -0.000 0.009 -0.044 -0.048 0.168 -0.550 -0.182 0.124
(0.108) (0.103) (0.444) (0.192) (0.330) (0.600) (0.349) (0.182)

Observations 28,715 28,715 28,715 28,715 28,715 28,715 28,715 28,715

Countries FE YES YES NO NO NO NO NO NO
Pseudo R2 0.0591 0.0591 0.0242 0.0277 0.0241 0.0275 0.0288 0.0266

Table 7: Estimation results. Ordered probit with clustered standard errors by country
(29 clusters). Robust standard errors in parentheses where ** indicate p < 0.01, and *
p < 0.05, respectively.

34
 I II III IV V VI VII VIII
(H1) (H1) (H2) (H2) (H2) (H2) (H2) (H2)

Medium education (Ref.: low) -0.132* -0.107* -0.214* -0.202 -0.223 -0.246* -0.159* -0.222*
(0.055) (0.048) (0.108) (0.132) (0.119) (0.107) (0.066) (0.111)
High education (Ref.: low) 0.004 0.013 -0.104 -0.096 -0.121 -0.127 -0.036 -0.101
(0.064) (0.051) (0.106) (0.132) (0.118) (0.109) (0.077) (0.113)

Personal and social norms

Religious (Ref.: non religious) 0.160** 0.160** 0.107** 0.115* 0.122** 0.121** 0.077* 0.098*
(0.031) (0.031) (0.034) (0.049) (0.042) (0.045) (0.036) (0.045)

Patriotic (Ref.: non patriotic) 0.259** 0.259** 0.282** 0.258** 0.275** 0.268** 0.292** 0.286**
(0.037) (0.037) (0.052) (0.049) (0.050) (0.042) (0.053) (0.051)

Inequality averse 0.092** 0.092** 0.127** 0.109* 0.127** 0.120** 0.126** 0.116*
(0.034) (0.034) (0.049) (0.047) (0.047) (0.045) (0.047) (0.049)

Direct benefits
Number of children 0.009 0.023** 0.029** 0.027** 0.030** 0.027** 0.029** 0.029**
(0.013) (0.006) (0.006) (0.007) (0.006) (0.006) (0.006) (0.006)
x medium education 0.022
(0.017)
x high education 0.006
(0.022)

Unemployed (Ref.: employed) -0.280* -0.286** 0.023 0.057 0.040 -0.014 -0.041 0.020
(0.111) (0.107) (0.179) (0.182) (0.190) (0.188) (0.158) (0.187)
x medium education 0.309** 0.314** 0.080 0.050 0.073 0.122 0.119 0.074
(0.113) (0.110) (0.141) (0.136) (0.149) (0.152) (0.137) (0.144)
x high education 0.337* 0.346* 0.096 0.068 0.085 0.126 0.121 0.089
(0.141) (0.137) (0.161) (0.150) (0.165) (0.173) (0.158) (0.160)

Self-employed (Ref.: employed) 0.146 0.142 0.326* 0.348 0.344* 0.285 0.250* 0.316
(0.100) (0.103) (0.151) (0.183) (0.162) (0.155) (0.118) (0.172)
x medium education -0.322** -0.317** -0.452** -0.493** -0.466** -0.418** -0.393** -0.455**
(0.108) (0.110) (0.134) (0.161) (0.144) (0.140) (0.115) (0.157)
x high education -0.431** -0.428** -0.551** -0.600** -0.566** -0.514** -0.493** -0.574**
(0.130) (0.133) (0.154) (0.180) (0.166) (0.157) (0.135) (0.178)

Retired (Ref.: employed) -0.070 -0.062 -0.118 -0.106 -0.112 -0.139 -0.118 -0.133
(0.063) (0.061) (0.091) (0.097) (0.084) (0.088) (0.087) (0.089)
x medium education 0.216** 0.203** 0.304** 0.303* 0.310** 0.314** 0.276** 0.309**
(0.073) (0.071) (0.112) (0.122) (0.113) (0.105) (0.095) (0.110)
x high education 0.184 0.180 0.323* 0.314* 0.326* 0.339** 0.295* 0.318*
(0.097) (0.094) (0.136) (0.142) (0.134) (0.129) (0.121) (0.129)

Other (Ref.: employed) -0.065 -0.066 0.219 0.264 0.238 0.159 0.122 0.212
(0.069) (0.069) (0.134) (0.170) (0.159) (0.143) (0.085) (0.161)
x medium education 0.137 0.137 -0.140 -0.185 -0.158 -0.083 -0.052 -0.136
(0.075) (0.076) (0.110) (0.142) (0.133) (0.126) (0.080) (0.137)
x high education -0.015 -0.011 -0.261* -0.299* -0.278* -0.205 -0.193* -0.262*
(0.087) (0.088) (0.119) (0.134) (0.125) (0.116) (0.098) (0.128)

Controls
Medium income (Ref.: low) -0.001 -0.000 0.141 0.092 0.130 0.564 0.185 -0.017
(0.030) (0.030) (0.187) (0.064) (0.143) (0.325) (0.097) (0.047)

High income (Ref.: low) -0.033 -0.033 -0.173 0.124 -0.074 0.456 0.093 -0.079
(0.050) (0.050) (0.227) (0.102) (0.177) (0.480) (0.101) (0.066)

Age 0.009** 0.009** 0.007** 0.007** 0.007** 0.007** 0.008** 0.007**

(0.001) (0.001) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002)

Gender (Ref.: male) 0.185** 0.186** 0.159** 0.161** 0.158** 0.164** 0.165** 0.162**
(0.027) (0.027) (0.036) (0.039) (0.037) (0.033) (0.029) (0.036)

Married (Ref.: never married) 0.018 0.020 0.022 0.034 0.032 0.026 0.009 0.018
(0.022) (0.023) (0.026) (0.029) (0.028) (0.028) (0.028) (0.032)
Widowed (Ref.: never married) -0.047 -0.045 -0.073* -0.040 -0.053 -0.064 -0.099* -0.070
(0.037) (0.038) (0.035) (0.047) (0.043) (0.047) (0.043) (0.054)
Divorced (Ref.: never married) -0.121** -0.120** -0.130** -0.113** -0.126** -0.115** -0.137** -0.130**
(0.033) (0.034) (0.041) (0.038) (0.042) (0.038) (0.042) (0.045)

Natural-born citizen -0.166** -0.166** -0.128* -0.141* -0.117* -0.128* -0.148** -0.153**
(0.048) (0.048) (0.057) (0.060) (0.054) (0.055) (0.057) (0.056)

35
 I II III IV V VI VII VIII
(H1) (H1) (H2) (H2) (H2) (H2) (H2) (H2)

Indirect benefits
Public social expenditure -0.001
(0.009)
x medium income -0.004
(0.005)
x high income 0.004
(0.007)

Ethnic fractionalization -0.306

(0.316)
x medium income -0.281
(0.180)
x high income -0.601*
(0.277)

Transparency 0.020
(0.035)
x medium income -0.019
(0.021)
x high income 0.006
(0.027)

Income inequality -0.011

(0.011)
x medium income -0.012
(0.007)
x high income -0.011
(0.010)

Relative redistribution -0.008

(0.007)
x medium income -0.005
(0.003)
x high income -0.004
(0.003)

Deterrence 0.094
(0.082)
x medium income 0.044
(0.051)
x high income 0.058
(0.069)

Constant cut1 -1.202** -1.179** -1.108** -1.177** -0.938** -1.617** -1.298** -1.003**
(0.115) (0.107) (0.386) (0.180) (0.284) (0.558) (0.324) (0.159)
Constant cut2 -0.644** -0.622** -0.576 -0.641** -0.406 -1.083 -0.764* -0.469**
(0.114) (0.107) (0.389) (0.182) (0.288) (0.561) (0.323) (0.162)
Constant cut3 -0.124 -0.102 -0.081 -0.144 0.088 -0.586 -0.266 0.027
(0.118) (0.111) (0.397) (0.191) (0.297) (0.564) (0.327) (0.169)

Observations 28,715 28,715 28,715 28,715 28,715 28,715 28,715 28,715

Countries FE YES YES NO NO NO NO NO NO
Pseudo R2 0.0598 0.0597 0.0259 0.0295 0.0259 0.0290 0.0300 0.0284

Table 8: Estimation results. Ordered probit with clustered standard errors by country
(29 clusters). Robust standard errors in parentheses where ** indicate p < 0.01, and *
p < 0.05, respectively.

36
 1 2 3 4 5 6 7 8 9 10 11 12

1 Tax morale 1.00

2 Education -0.05 1.00
3 Religious 0.08 -0.15 1.00
4 Patriotic 0.09 -0.07 0.11 1.00
5 Inequality averse 0.04 -0.13 0.03 -0.02 1.00
6 Number of children 0.03 -0.06 0.09 0.03 0.02 1.00
7 Unemployed 0.00 -0.07 -0.01 -0.04 0.06 0.05 1.00
8 Self-employed -0.04 0.03 0.00 0.01 -0.02 0.05 -0.06 1.00
9 Retired 0.10 -0.18 0.10 0.04 0.05 -0.20 -0.15 -0.14 1.00
10 Other 0.01 -0.17 0.06 0.01 0.03 0.06 -0.11 -0.10 -0.25 1.00
11 Income -0.05 0.30 -0.08 0.00 -0.10 0.12 -0.14 0.07 -0.28 -0.06 1.00
12 Age 0.13 -0.21 0.13 0.05 0.03 -0.08 -0.13 -0.04 0.70 -0.23 -0.25 1.00
13 Gender 0.08 -0.04 0.14 0.02 0.02 0.06 -0.03 -0.11 0.00 0.20 -0.08 0.02
14 Married 0.06 -0.01 0.10 0.05 0.01 0.32 -0.05 0.06 -0.01 -0.04 0.27 0.13
15 Widowed 0.06 -0.18 0.10 0.03 0.03 -0.09 -0.06 -0.06 0.39 -0.06 -0.28 0.43
16 Divorced -0.03 0.05 -0.06 -0.05 -0.01 -0.03 0.02 0.00 -0.03 -0.06 -0.12 0.05
17 Natural-born citizen -0.02 -0.03 -0.02 0.03 0.00 -0.03 -0.02 0.01 0.00 0.01 0.02 -0.02
18 Public social expenditure -0.02 0.18 -0.19 0.01 -0.03 -0.11 -0.06 0.01 0.02 -0.12 0.01 0.06
19 Ethnic fractionalization -0.09 0.00 0.01 -0.06 -0.06 0.00 0.01 -0.04 -0.01 0.02 0.02 -0.02
20 Transparency 0.00 0.20 -0.23 0.01 -0.05 -0.08 -0.07 -0.02 -0.03 -0.08 0.03 0.04
21 Income inequality -0.08 0.06 0.01 -0.04 -0.03 -0.07 -0.01 -0.01 0.01 -0.05 0.00 0.02
22 Relative redistribution -0.09 0.24 -0.17 -0.01 -0.03 -0.11 -0.08 -0.03 0.03 -0.15 0.01 0.06
23 Deterrence 0.07 -0.07 0.07 -0.01 0.05 0.03 0.04 0.03 0.01 0.04 -0.05 -0.02

13 14 15 16 17 18 19 20 21 22 23

1 Tax morale
2 Education
3 Religious
4 Patriotic
5 Inequality averse
6 Number of children
7 Unemployed
8 Self-employed
9 Retired
10 Other
11 Income
12 Age
13 Gender 1.00
14 Married -0.07 1.00
15 Widowed 0.17 -0.40 1.00
16 Divorced 0.05 -0.37 -0.11 1.00
17 Natural-born citizen -0.02 -0.02 0.00 -0.02 1.00
18 Public social expenditure -0.05 -0.04 -0.07 0.05 -0.04 1.00
19 Ethnic fractionalization 0.03 -0.01 0.04 0.01 -0.04 -0.48 1.00
20 Transparency -0.04 -0.05 -0.08 0.08 -0.06 0.74 -0.24 1.00
21 Income inequality 0.02 -0.05 0.01 0.06 -0.02 0.25 -0.19 0.17 1.00
22 Relative redistribution -0.02 -0.07 -0.02 0.07 -0.05 0.66 -0.34 0.56 0.49 1.00
23 Deterrence -0.01 0.03 -0.01 -0.02 0.06 -0.01 -0.23 -0.31 0.13 -0.01 1.00

Table 9: Correlation matrix

37
 Specification I Specification II Specification III Specification IV Specification V Specification VI Specification VII Specification VIII
(# p-values: 15) (# p-values: 13) (# p-values: 3) (# p-values: 3) (# p-values: 3) (# p-values: 3) (# p-values: 3) (# p-values: 3)

# Reject p-value # Reject p-value # Reject p-value # Reject p-value # Reject p-value # Reject p-value # Reject p-value # Reject p-value

Bonferroni (1) 4 0.00667 6 0.00769 2 0.03333 1 0.03333 0 0.03333 0 0.03333 3 0.03333 0 0.03333
Sidak (1) 4 0.00700 6 0.00807 2 0.03451 1 0.03451 0 0.03451 0 0.03451 3 0.03451 0 0.03451
Holm (2) 4 0.00909 7 0.01667 3 0.1 1 0.05000 0 0.03333 0 0.03333 3 0.1 0 0.03333
Holland (2) 4 0.00953 7 0.01741 3 0.1 1 0.05132 0 0.03451 0 0.03451 3 0.1 0 0.03451
Liu 1 (2) 5 0.01612 7 0.03988 3 0.3 1 0.07805 0 0.03451 0 0.03451 3 0.3 0 0.03451
Liu 2 (2) 5 0.015 7 0.03611 3 0.3 1 0.07500 0 0.03333 0 0.03333 3 0.3 0 0.03333
38

Hochberg (3) 4 0.00833 7 0.01429 3 0.1 1 0.03333 0 0.03333 0 0.03333 3 0.1 0 0.03333
Rom (3) 4 0.00874 7 0.01492 3 0.1 1 0.03417 0 0.03417 0 0.03417 3 0.1 0 0.03417
Simes (3) 6 0.04 8 0.06154 3 0.1 1 0.03333 0 0.03333 0 0.03333 3 0.1 0 0.03333
Yekutieli (3) 4 0.00804 7 0.01693 3 0.05455 1 0.01818 0 0.01818 0 0.01818 3 0.05455 0 0.01818
Krieger (3) 8 0.08081 10 0.18182 3 0.09091 2 0.09091 0 0.03030 0 0.03030 3 0.09091 0 0.03030

Table 10: Robustness check for multiple hypothesis testing with 11 different methods of p-value adjustment. The numbers in the
first column denote the existing three different approaches: (1) one-step (2) step-down and (3) step-up. The p-value adjusted
for a False Discovery Rate or a Family Wise Error Rate is 0.1.