Referendums, Trust, and Tax Evasion∗

Simon Hug†and Franziska SporriCIS, IPZ, Universitat Zurich

Paper prepared for presentation at the ECPR JointSessions of Workshops, Helsinki, May 7 - 12 2007

First preliminary version: April 20, 2007

Abstract

Countries engaged in transitions from Communist rule to democracyand market economies face a considerable challenge to foster tax moraleand to reduce tax evasion. Tax evasion, however, severly limits the capac-ities of these reformed states, which reduces in a vicious circle the citizens’incentives to pay taxes. The question then arises how this vicious circlemay be broken. Several authors suggest that political institutions may dothe trick. Based on insights that institutions allowing a direct say on policyissues (especially tax questions) may increase tax morale, we propose tostudy comparatively how the link between trust and tax morale is affectedby these institutions. We find that allowing for referendums strengthensthe link between trust and tax morale, but that this effect depends on thenature of the institutions.

∗ This paper emanates from the research project “Civil society, trust and democratisationin Russia” (with Nicolas Hayoz) financed by the Swiss National Science Foundation (SCOPESGrant IB7310-110958). Initial thoughts were articulated at the Conference “Tax Evasion, Trust,and State Capacities” (St. Gallen 2004) financed by the Swiss National Science Foundation(Grant IB7110-103155) the Swiss Agency for Development and Cooperation ( 7F-00527.04.25),and the University of St. Gallen. The support by these various founding agencies is gratefullyappreciated.

† Center for Comparative and International Studies; Institut fur Politikwissenschaft; Uni-versitat Zurich; Hirschengraben 56; 8001 Zurich; Switzerland; phone +41 (0)44 634 50 90/1;fax: +41 (0)44 634 5098; email: simon.hug@access.unizh.ch

1

1 Introduction

For countries engaged in transition processes from Communist rule to democ-

racy and market economies, a major problem relates to tax evasion (see for

instance Martinez-Vazquez and McNab, 1997; Easter, 2002; Martinaz-Vazguez

and Alm, 2002; Torgler, 2003a; Hayoz and Hug, 2006). While loosely speaking

taxes failed to be levied by communist states, the newly democratized states had

to find resources to offer the basic public services of states. Obviously, various

privatization strategies (e.g., Brada, 1996) allowed at least initially to contribute

to the budget of the new democracies. In the longer term, however, tax receipts

have to be the major component to finance state activities.1

In the absence of sufficient resources a state’s capacities are severely dimin-

ished, and it faces considerable difficulties to offer the basic public goods. These

lacking state capacities and the ensuing underprovision of public goods, however,

also affect the tax morale of citizens. Why should a citizen pay taxes, if the

state fails to offer basic services? Hence, several authors consider this interac-

tion between state capacities and tax evasion as a vicious circle (see for instance

Easter, 2002; Easter, 2006; Uslaner, 2006). The question then becomes how this

vicious circle can be broken or even turned into a virtuous circle.

Several scholars have emphasized the important role that trust plays in ex-

plaining tax morale and tax evasion (e.g. Feld and Frey, 2002a; Torgler, 2003c;

Torgler, 2006). For the policy question how tax morale may be improved, this

explanation only pushes the question up one level, since now a state needs to

know how trust may be fostered among its citizens. In the literature on trust

(for an overview see Hardin, 2006) an animated debate turns around the question

whether political institutions may foster such trust, or whether this is not possi-

ble (see for instance Rothstein and Stolle, 2002; Uslaner, 2006). When it comes

to tax evasion and tax morale, however, there seems to be some theoretical and

empirical evidence that institutional elements play a role. More precisely, insti-

tutions used to determine taxes seem to matter. Pommerehne, Hart and Frey

(1994), Pommerehne and Weck-Hannemann (1996), Feld and Frey (2002b), and

Feld and Frey (2006) suggest that at least in Switzerland, direct involvement in

decisions on taxes and the rules of the “tax-man” significantly affect tax evasion.

1Obviously this does not necessarily hold for states, like Russia (Easter, 2006), with consid-erable natural resources who can profit, for instance, from increasing oil prices.

2

Given these results the question arises whether this impact of institutional

arrangements also holds comparatively when looking at the differences in terms

of tax morale across countries. More precisely, do institutions allowing citizens to

directly affect policies influence tax morale? This question is all the more relevant,

since many of the newly democratized and formerly communist countries have

introduced in part wide-ranging institutions for referendums in their constitution

(Brady and Kaplan, 1994; White and Hill, 1996; Auer and Butzer, 2001; Hug and

Tsebelis, 2002; Hug, 2005).

Consequently we wish to address the question how institutions, more specif-

ically institutions allowing citizens a direct say on policy decisions, affect tax

morale. Building up on the existing comparative work on tax morale we wish

to assess how the important link between trust and tax morale is affected by

institutions allowing for referendums. To do so we rely on combining survey data

from various sources with detailed information on the institutional setup of the

countries considered. We find, that the possibility of holding referendums mostly

strengthens the link between trust and tax morale.

In the next section we briefly review the literature on tax evasion as it is linked

to the question of trust. Section three is devoted to a discussion of the institutions

allowing for referendums, and how they might influence the effect of trust on tax

morale. In section four we present our empirical results which rely on two-step

estimations of a simple multilevel model (e.g., Achen, 2005; Franzese, 2005; Lewis

and Linzer, 2005). Section five concludes.

2 Tax evasion, trust, and institutions

As mentioned above, the acceptance of tax evasion is especially related to peo-

ple’s trust in political institution. Uslaner (2006) suggests that if people viewed

taxation as some sort of collective good, generalized trust might matter for tax

compliance. The evidence of Uslaner’s analysis suggests that people do not see

taxes as an obligation to their fellow citizens as much as they do as a contract

with the state. Citizens are more likely to justify tax evasion by judging the lead-

ers than the fellow citizens. Therefore, people look to public officials to determine

whether the elites are behaving honestly.

Having a strong legal system matters. But confidence in the legal system

only seems to matter when there is an effective and fair judiciary (e.g. in Roma-

3

nia, where the legal system is weaker, perceptions that the courts and police are

fair don’t matter). Generally, the political institutions shape the perception of

procedural fairness. Therefore, the fairness of the legal system is more strongly

related to tax compliance than is simple confidence in the legal system. People’s

tax compliance seems to depend on how well the entire government works, not

just the judiciary. Uslaner finds very modest correlations between tax compli-

ance and generalized trust. The author stresses as well the problem that it might

be possible to construct better institutions but it is extremely difficult to con-

struct better functioning institutions. Uslaner doesn’t find a correlation between

democratization measures and corruption and concludes: ”Institutions are not

nearly as sticky as corruption - and structural change does not track the level of

transparency.”

Feld and Frey’s (2006)emphasize as well on the contractual dimension of pay-

ing taxes as on the importance of fairness. Their argument is that a psychological

tax contract which establishes a fiscal exchange between the state and the citizens

shapes tax compliance to a large extent. This contractual relationship implies

duties and rights for each contract partner. Income redistribution is the more

accepted by affluent citizens the more the political process is perceived to be fair

and the more policy outcomes are considered as legitimate. Additionally, it is

important how the tax office treats the taxpayer (hierarchical relationship treat-

ing taxpayers as inferiors is contra productive). In accordance with the Crowding

Theory (Frey, 1997b; Feld and Frey, 2002a) a ’respectful’ relationship of the tax

authorities to the taxpayers crowds in tax morale while an ’authoritarian’ rela-

tionship using instruments of deterrence reduces the incentives to evade taxes

(higher probability of being punished) on the one hand, and on the other hand

crowds out the intrinsic motivation to pay taxes. Only well functioning instru-

ments of deterrence may avoid a reduction in tax revenue caused by the decline

of tax moral and solely a strong state disposes of such capacities.

The link between direct democratic institutions and tax evasion has been

studied in several surveys, especially for the case of the Swiss cantons. Ac-

cording to these studies (e.g., Pommerehne and Frey, 1992; Pommerehne and

Weck-Hannemann, 1996; Frey, 1997a; Feld and Frey, 2002a), the more direct de-

mocratic the political institutions are, the lower is tax evasion. Feld and Frey

(2006) conclude that the treatment of the taxpayer in the case of no tax decla-

4

ration is correlated with higher tax evasion but depends on the extent of direct

democratic institutions: The respectful treatment has a significant negative im-

pact on tax evasion in more directly democratic cantons. In contrast, it increases

tax evasion in more representative democratic cantons. Taken these findings into

account, more direct democratic institutions improve the legitimacy of a state by

giving people a larger say in politics. In particular, the possibility of voting on

the proposal of a fine, leads to a higher tax morale as Feld and Tyran (2002) find

as a main result.

In a series of studies Torgler (2003a, 2003a, 2003b, 2005) has explored the

determinants of tax morale. Based on survey evidence from the World Val-

ues Survey and the European Values Survey (European Values Study Group and

Association, 2006) he finds that tax morale in the Central/Eastern European

Countries (CEE) is significantly higher than in countries of the Former Soviet

Union (FSU) (Torgler, 2006). In explaining both interpersonal and cross-country

differences he finds that all trust variables are highly significant (trust in legal sys-

tem, trust in government, satisfaction with national officers, and evaluation of the

political system) and correlate positively with a higher tax moral, whereas cor-

ruption affects tax morale negatively. It is also important to focus on inequality

because it underlies trust and,indirectly, is the foundation of corruption. Corrup-

tion leads to more inequality, what results in a vicious circle (inequality produces

low trust, low trust causes corruption and corruption increases inequality).

Transition countries are faced with many problems that have an impact on

people’s trust in institutions: In times of uncertainty as during the transition

process from Communist rule to democracy and market economies, social capital

in terms of mutual trust and honesty is an important factor to conclude deregula-

tions and privatization successfully (Frey, 2002). Large shadow economy reflects

weak and inefficient external institutions and have a negative impact on the de-

gree of tax morale (Alm and Torgler, 2006). Torgler (2003a, 2003a, 2003b, 2005)

concludes that differences in the reform procedures may influence the higher tax

morale in the CEE countries compared to the FSU countries: Smaller size of the

shadow economy, better established property rights as well as more stable in-

stitutions in general have reduced uncertainty and fostered trust in the political

institutions in the CEE countries. Countries in negotiation with the EU had an

additional motivation and pressure for rapid reforms (Schneider, 2002). Stable

5

and transparent institutions create reliability, reduce uncertainty and therefore

help the government to produce trust. Taxes might be seen as a price for gov-

ernment’s positive actions. If citizens perceive institutions as fair and trust the

public officials and the political institutions, their tax morale might increase.

3 Referendums and tax evasion

While the effect of good functioning political institutions through trust and low

political corruption appears to affect positively tax morale, these results raise

the question, however, whether particular institutions may heighten the levels of

trust and lower political corruption. In the literature on political institutions and

trust, this question is still hotly debated and no common conclusion has emerged

(see for instance Rothstein and Stolle, 2002; Uslaner, 2006). In research on taxes,

however, one particular type of institution often is referred to, namely referen-

dums. This link between taxes and referendums is certainly in part due to the

well known tax-revolt referendums in the United States (e.g. Smith, 1998), which,

starting from California, swept across several states and reduced property taxes

considerably (see also Kiewiet and Szakaly, 1996; Gerber, Lupia, McCubbins and

Roderick Kiewiet, 2000).2

Referendums may, however, not only affect the levying of taxes, but also

tax morale. Studies by Pommerehne and Frey (1992), Pommerehne and Weck-

Hannemann (1996), Frey (1997a), Feld and Frey (2002a) suggest that while ref-

erendums may affect the tax level, they also affect tax morale. Pommerehne and

Weck-Hannemann (1996) find that if voters can decide on tax levels, the extent

of tax evasion is lower. Feld and Frey’s (2006) study shows that the way in

which tax officials interact with tax contributors also affect tax morale. More to

the point, Hug (2005) finds that in Central and East European countries with

referendums, satisfaction with the way in which democracy works is higher.

This suggests that the influence of institutions allowing citizens to have a

direct say on policies might also affect tax morale. In order to assess whether

this is true, we first need an assessment of what kind of possibilities exist for

citizens to have a direct say on policies. Hug and Tsebelis (2002) and Hug

(2004) offer two related typologies of such institutions, which are tightly linked

2Interesting to note in this context is that early writing on Swiss referendums (e.g., Rappard,1923) discussed how initiatives for higher taxes failed.

6

to Tsebelis’s (2002) veto-player theory, respectively game theoretic models of

different institutional provisions discussed in Hug (2004). The first starts off

by distinguishing between required and nonrequired referendums. Among the

nonrequired referendums one has to distinguish between referendums triggered

by veto-players (veto-player referendums) or not. Among the latter category

two types of referendums are possible, namely those where the proposal to be

voted on has been adopted by veto-players (popular veto) and those where the

proposal is written by another actor (popular initiative). Table 1 reports for the

countries covered in our empirical analyses the availability of these four types of

referendums.3

Table 1: Referendum institutionsa

nation required Veto-player referendum Popular veto Popular initiativeArgentina 0 0 0 0Armenia 0 0 0 1Australia 1 0 0 0Azerbaijan 0 0 0 1Belarus 0 0 1 1Brazil 0 0 0 0Bulgaria 0 0 0 0Chile 0 0 0 1Croatia 1 0 0 1Estonia 0 0 0 1Finland 0 0 0 0Georgia 0 0 1 1Germany 0 0 0 0India 0 0 0 0Japan 1 0 0 0Latvia 1 0 1 0Lithuania 1 1 1 1Macedonia 1 0 0 1Mexico 0 0 0 0Moldova 1 1 1 1Nigeria 0 0 0 0Norway 0 0 0 0Peru 0 0 0 0Philippines 1 1 1 1Poland 0 0 0 0Russia 0 0 0 1Slovenia 0 1 1 1Spain 1 0 0 1Sweden 0 0 0 1Switzerland 1 1 1 0Taiwan 0 0 0 0Ukraine 1 0 1 1United States 0 0 0 0Uruguay 1 1 1 0Venezuela 1 0 0 0

aSources: Hug and Tsebelis (2002) and Hug (2004).

3Strictly speaking, given that we are interested in the effect of referendums in the realm oftax policies, we should have information on whether these institutions can be used for decisionson taxes. So far we have not collected this information, which is quite difficult to generate.In addition, most often the constitutions are rather unspecific regarding the topics covered byrules allowing for referendums, and decisions having implications on taxes can often not beruled out.

7

The second typology builds in part on a set of dichotomous indicators pro-

posed by Suksi (1993). The first one, as above, distinguishes between required

(type I) and nonrequired referendums. Among the latter a distinction is made

between passive (type IV) and active ones. In the passive ones, no actor outside

government is involved in triggering or drafting a referendum proposal, while in

active ones, an outside actor is involved in at least one of the two activities.

When a non-governmental actor only triggers the referendum vote, the vote is on

a government policy (type II), while if this actor also proposes a new policy, the

voters decide on a opposition policy (type III). Table 2 provides for the same list

of countries as above the information what institutions are available.

Table 2: Referendum institutionsa

type I type II type III type IVrequired non-required/active- non-required/active- type IV: non-required/

Nation vote on government policy vote on opposition policy passiveArgentina 0 0 0 0Armenia 0 0 0 1Australia 1 0 0 0Azerbaijan 0 0 0 1Belarus 0 0 1 1Brazil 0 0 0 0Bulgaria 0 0 0 0Chile 0 0 0 1Croatia 1 0 0 1Estonia 0 0 0 1Finland 0 0 0 0Georgia 0 0 1 1Germany 0 0 0 0India 0 0 0 0Japan 1 0 0 0Latvia 1 0 1 0Lithuania 1 1 1 1Macedonia 1 0 0 1Mexico 0 0 0 0Moldova 1 1 1 1Nigeria 0 0 0 0Norway 0 0 0 0Peru 0 0 0 0Philippines 1 1 1 1Poland 0 0 0 0Russia 0 0 0 1Slovenia 0 1 1 1Spain 1 0 0 1Sweden 0 0 0 1Switzerland 1 1 1 0Taiwan 0 0 0 0Ukraine 1 0 1 1United States 0 0 0 0Uruguay 1 1 1 0Venezuela 1 0 0 0

aSources: Hug and Tsebelis (2002) and Hug (2004).

4 Referendums, trust, and tax morale

The starting point of our analysis is Torgler’s (2006) study of how tax morale

is affected by individual level characteristics and the level of trust. Given that

8

we are interested in the effect of institutions at the national level, we are in a

classic case of a multilevel model. The individual level effects are nested in the

national contexts, and we presume that the national contexts, at least in part

are dependent on the presence of institutions allowing for referendums. Such

multilevel models experience increasing popularity (see for instance Jones and

Steenbergen, 2002; Raudenbusch and Anthony, 2002; Kedar and Shively, 2005).

In our context, however, where at the first level the number of observations is very

large (i.e., around 1000 individuals per country) the leverage of a joint estimation

is rather minute. For this reason two level estimations, where in a first regression

the individual level model is estimated for each unit of the second level (i.e.,

countries) and the estimated coefficients then regressed on explanatory variables,

is preferable (see most recently Achen, 2005; Beck, 2005; Franzese, 2005).4

This being said, the question arises how institutions allowing for referendums

should affect the individual level relationship linking individual characteristics to

tax morale. The first and simplest hypothesis is that having the possibility to vote

on particular policies simply raises the average level of tax morale. In that case we

should find differences in the intercepts estimated for the various countries at the

individual level, and these differences should relate to the presence of institutions

allowing for referendums.

A second and slightly more complicated hypothesis would suggest that the

effect of specific variables, more precisely variables measuring individuals’ trust

in political institutions, should depend on the presence of institutions allowing

citizens a direct say on policies.

The data we use comes from the third wave of the World Values Survey

(European Values Study Group and Association, 2006).

Our main dependent variable at the individual level comes from the responses

to the following question:

F116.- Please tell me for each of the following statements whether

you think it can always be justified, never be justified, or something

4We also follow Lewis and Linzer (2005) suggestion against using weighted least squares(WLS) in the second stage, given that the standard errors for the estimated coefficients bothreflect variability due to characterists at the second level and measurement issues at the firstlevel. Hence, assuming, as in a WLS-analysis, that all variation is due to the first element mayresult in considerable biases. We also estimated our models with a generalized-linear modelframework (Rabe-Hesketh, Skrondal and Pickles, 2001) and obtained largely similar results.

9

in between, using this card. (Read out statements. Code one answer

for each statement).

Cheating on taxes if you have a chance

1 Never justifiable

2, 3, 4, 5, 6, 7, 8, 9,

10 Always justifiable

To measure individuals’ trust in institutions we employ three items from a

battery of question dealing with confidence and two questions relating to the po-

litical system in general and the current office holders in particular. The questions

dealing with confidence in institutions dealt with the legal system, the national

government and parliament and had the following format:

E075.- I am going to name a number of organizations. For each one,

could you tell me how much confidence you have in them: is it a great

deal of confidence, quite a lot of confidence, not very much confidence

or none at all?

Parliament

o 1 A great deal o 2 Quite a lot o 3 Not very much o 4 None at all

Given the way in which this variable is coded we would expect a positive

relationship. Persons with little confidence (high values) are also more likely to

find tax evasion justifyable.

The question eliciting a rating of the political system was posed in the follow-

ing way:

E111.- People have different views about the system for governing this

country. Here is a scale for rating how well things are going: 1 means

very bad; 10 means very good.

Where on this scale would you put the political system as it is today?

1 Bad

2, 3, 4, 5, 6, 7, 8, 9,

10 Very good

10

For this variable we would expect a negative relationship, since positive ratings

(high values) should lead individuals to find tax evasion less justifyable.

Finally, the question on the incumbents in office was asked in the following

way:

E125.- How satisfied are you with the way the people now in national

office are handling the country’s affairs? Would you say you are very

satisfied, fairly satisfied, fairly dissatisfied or very dissatisfied?

o 1 Very satisfied o 2 Fairly satisfied o 3 Fairly dissatisfied o 4 Very

dissatisfied

Given that higher values for this variable suggest less satisfaction, we expect

a positive relationship with the variable tax morale.

Based on Torgler’s (2006) study we estimate an individual level model which

almost completely corresponds to his specification. We employ the same inde-

pendent variables and only refrain from recoding the dependent variable5 The

main independent variables are obviously those that measure the trust in politi-

cal institutions.6

The results from a pooled analysis covering all 35 countries we consider appear

in table 4. These results should only be considered as illustrative, since they

obviously do not reflect the possible heterogeneity in the effects of the various

independent variables.

Table 4 reports the results of the second stage. Estimating the models re-

ported in 4 for each of the 35 countries separately yields for our five trust measures

35 coefficients. These five sets of coefficients form the dependent variables for our

second level. As explanatory variable we use four indicator variables measuring

what kind of referendum institutions are available in a country.

5Given the skewed distribution of the dependent variable Torgler (2006) groups together thehigher values. We refrain from doing this, since it only results in throwing away information.

6Here we also follow Torgler (2006), who considers these ordinal variables as continuous.This renders our hierarchical estimation procedure simpler.

11

Table 3: Explaining tax morale: pooled analysisordered probit

bvariable (s.e.)a

female -0.088(0.029)

married -0.033(0.037)

living together 0.084(0.049)

divorced 0.093(0.055)

separated -0.028(0.054)

widow 0.037(0.046)

part-time employed -0.092(0.048)

self employed -0.057(0.082)

retired -0.068(0.049)

at home -0.201(0.078)

student -0.102(0.059)

un employed 0.026(0.049)

other -0.003(0.070)

middle aged -0.137(0.026)

older -0.304(0.044)

old -0.452(0.067)

confidence legal system (v137) 0.078(0.025)

confidence national government (v142) 0.019(0.025)

confidence parliament (v144) 0.014(0.024)

rating political system today (v152) 0.010(0.011)

incumbent satisfaction (v165) 0.074(0.029)

cut1 0.395(0.132)

cut2 0.670(0.132)

cut3 0.918(0.131)

cut4 1.081(0.130)

cut5 1.384(0.128)

cut6 1.539(0.128)

cut7 1.715(0.131)

cut8 1.932(0.133)

cut9 2.080(0.136)

Log pseudo-likelihood -61539.10738292

aThe standard errors are clustered by country.

12

Concerning the results reported in table 4 we find that the confidence both

in the legal system and in the national government leads to higher tax morale.

Similarly, we find a positive and statistically significant relationship between sat-

isfaction with the incumbent and tax morale, while the effect for confidence in

the parliament fails to reach statistical significance.7 We also fail to find a sta-

tistically significant effect for the rating of the political system, even tough we

expected a negative relationship.8

Obviously, the analyses reported in table 4 are based on the presumption that

the effects of the various independent variables are identical in all countries, and

in addition, that there are no level differences in tax morale across the set of

countries. Torgler (2006), however, finds in his analyses some systematic country

differences, especially between countries from Central and Eastern Europe, and

successor states of the former Soviet-Union. In order to allow for these differences,

we estimated the same model for each of the 35 countries individually, and the

estimated coefficients become our dependent variable in the second stage. Table

4 reports the estimated results.9

We use as dependent variable the estimated coefficients for our five measures

of trust in political institutions, as well as the value of the first threshold from

the ordered probit. Given that we are interested whether referendum institutions

explain differences in these coefficients, F-tests of the whole set of dichotomous

indicators provide the most relevant information. Table 4 suggests that only the

relationship between confidence in the national government and tax morale is

statistically significantly affected by referendum institutions. In addition, this

effect only appears with the typology based on Hug (2004). When looking more

closely at the estimated coefficients one notes that the effect of confidence in

the national government on tax morale is especially strong if an opposition may

launch a referendum on its own policy. The effect is considerably weakened if

an opposition may trigger a referendum on a government policy. The two other

types of referendums appear not to modify the relationship compared to the one

7Since all four of these independent variables are coded such that higher values reflect lessconfidence, and the dependent variable has the same polarity, a positive relationship is expected.

8Higher values for this variable indicate a higher rating.9Given that Torgler (2006) finds a significant difference between Central and East European

countries and the successor states of the Soviet-Union. For this reason we also controlled forthe length of the democratic regime. This variable, however, failed to a have an effect on ourdependent variables and also did not affect our other results.

13

appearing in countries with no referendums.

If we look at the estimated coefficients for the other typology we find in

terms of the effects a rather similar picture. Having the possibility to trigger a

popular initiative strengthens the effect of confidence in the national government,

while the other institutions have negative effects. These effects are, however,

statistically not significant and not terribly large.

Moving on to the other coefficients we find largely similar patterns, but as

mentioned before, we cannot reject the null hypothesis that there are no dif-

ferences in these effects. Interesting to note are the differences in the constant

terms10 across countries. We find lower average values of tax morale in countries

allowing for referendums on opposition policies triggered by the opposition. On

the other hand it appears that in countries allowing for other types of nonrequired

referendums, the tax morale is on average higher. Required referendums, on the

other hand appear not to make a difference.

10Given that the ordered probit model was estimated without intercept, the constant termis equivalent to the value of the first cutoff threshold.

14

Tab

le4:

Expla

inin

gth

eeff

ect

oftr

ust

onta

xm

oral

e:se

cond

leve

lan

alysi

sconst

ant

legalsy

stem

nati

onalgov.

parl

iam

ent

pol.

syst

.nat.

off.

bb

bb

bb

bb

bb

bb

vari

able

(s.e

.)(s

.e.)

(s.e

.)(s

.e.)

(s.e

.)(s

.e.)

(s.e

.)(s

.e.)

(s.e

.)(s

.e.)

(s.e

.)(s

.e.)

type

I:re

quir

ed

0.0

33

-0.0

39

0.0

33

0.0

26

-0.0

34

-0.0

51

-0.0

25

-0.0

32

0.0

06

0.0

06

-0.0

01

0.0

03

(0.1

97)

(0.1

96)

(0.0

29)

(0.0

30)

(0.0

36)

(0.0

39)

(0.0

34)

(0.0

34)

(0.0

12)

(0.0

12)

(0.0

43)

(0.0

43)

type

II:non-r

equir

ed/acti

ve-v

ote

on

govern

ment

policy

-0.2

31

-0.0

47

-0.1

35

-0.0

50

-0.0

26

0.0

48

(0.3

28)

(0.0

49)

(0.0

59)

(0.0

56)

(0.0

20)

(0.0

72)

type

III:

non-r

equir

ed/acti

ve-v

ote

on

opposi

tion

policy

0.1

83

0.0

44

0.1

26

0.0

73

0.0

17

-0.0

00

(0.2

81)

(0.0

42)

(0.0

51)

(0.0

48)

(0.0

17)

(0.0

62)

type

IV:non-r

equir

ed/pass

ive

-0.2

86

0.0

25

-0.0

38

0.0

02

-0.0

02

0.0

06

(0.1

78)

(0.0

27)

(0.0

32)

(0.0

31)

(0.0

11)

(0.0

39)

veto

-pla

yer

refe

rendum

-0.3

01

0.0

08

-0.0

23

-0.0

04

0.0

04

0.0

16

(0.1

77)

(0.0

27)

(0.0

35)

(0.0

31)

(0.0

10)

(0.0

39)

popula

rveto

-0.1

21

0.0

26

-0.0

27

0.0

06

-0.0

25

0.0

42

(0.1

93)

(0.0

30)

(0.0

38)

(0.0

34)

(0.0

11)

(0.0

42)

popula

rin

itia

tive

0.1

26

0.0

14

0.0

59

0.0

46

0.0

11

0.0

04

(0.2

27)

(0.0

35)

(0.0

45)

(0.0

40)

(0.0

13)

(0.0

50)

const

ant

0.4

68

0.5

31

0.0

34

0.0

35

0.0

82

0.0

88

0.0

24

0.0

27

0.0

00

0.0

04

0.0

33

0.0

18

(0.1

27)

(0.1

38)

(0.0

19)

(0.0

21)

(0.0

23)

(0.0

27)

(0.0

22)

(0.0

24)

(0.0

08)

(0.0

08)

(0.0

28)

(0.0

30)

F0.7

60

0.9

00

1.2

20

0.9

50

2.1

50

0.7

40

0.6

70

0.4

80

0.4

60

1.2

40

0.2

50

0.4

50

p(F

)0.5

59

0.4

76

0.3

22

0.4

48

0.0

99

0.5

74

0.6

15

0.7

53

0.7

67

0.3

15

0.9

10

0.7

71

rmse

0.4

97

0.4

93

0.0

74

0.0

75

0.0

90

0.0

98

0.0

86

0.0

87

0.0

30

0.0

29

0.1

10

0.1

08

n35

35

35

35

35

35

35

35

35

35

35

35

15

5 Conclusion

Newly democratized countries, especially if they have to reform their economies

at the same time, face an important challenge when it comes to tax collection.

Hence, it can hardly surprise that in democratizing former communist countries

tax morale (Torgler, 2006) is low and the shadow economy (Schneider, 2002).

Much of this, as Torgler (2006) nicely shows, is dependent on a lack in trust in

these countries.

As studies both at the theoretical and empirical level have shown, however,

there are ways how trust and confidence in a specific type of institution, namely

the “tax-man” may be increased. More precisely, it appears that a direct say on

policies, especially tax policies, increase the tax morale.

In the present study we took this as starting point to assess whether cross-

nationally we would find such an effect. Our empirical results suggest that at

least the effect of confidence in the national government on tax morale is affected

by the degree to which citizens may participate in decisions through referendums.

The results, however, are still clearly tentative. First of all, our indicators for

the presence or absence of referendums do not cover tax decisions specifically.11

Second, given that we rely on the third wave of the World Values Survey, our

geographical coverage is a bit arbitrary. Thirdly, the exact theoretical foundations

for the effects we expect are tenuous to say the least. Hence, future research will

have to try to assess whether our results remain robust and whether a solid

theoretical basis for them can be derived.

11A cursory glance at the constitutions suggests, however, that few prohibit referendums ontax issues.

16

Appendix

In tables 5 and 6 we provide the descriptive statistics of the variables used in the

individual level, respectively aggregated level analyses.

Table 5: Descriptive statistics individual levelVariable Min Mean Max Std. Dev. Obstax morale 1 2.728 10 2.539 37383female 0 0.503 1 0.500 37383married 0 0.605 1 0.489 37383living together 0 0.056 1 0.230 37383divorced 0 0.042 1 0.201 37383separated 0 0.019 1 0.137 37383widowed 0 0.061 1 0.240 37383part time 0 0.087 1 0.281 37383selfemployed 0 0.093 1 0.291 37383retired 0 0.146 1 0.353 37383at home 0 0.101 1 0.302 37383student 0 0.064 1 0.245 37383unemployed 0 0.088 1 0.283 37383other 0 0.021 1 0.142 37383middle aged 0 0.432 1 0.495 37383older 0 0.182 1 0.386 37383old 0 0.104 1 0.305 37383confidence legal system (v137) 1 2.575 4 0.865 37383confidence national government (v142) 1 2.696 4 0.890 37383confidence parliament (v144) 1 2.799 4 0.854 37383rating political system today (v152) 1 4.323 10 2.270 37383incumbent satisfaction (v165) 1 2.925 4 0.817 37383type I: required 0 0.379 1 0.485 37383type II: non-required/active-vote on government policy 0 0.147 1 0.354 37383type III: non-required/active-vote on opposition policy 0 0.276 1 0.447 37383type IV: non-required/passive 0 0.464 1 0.499 37383veto-player referendum 0 0.471 1 0.499 37383popular veto 0 0.293 1 0.455 37383popular initiative 0 0.276 1 0.447 37383

Table 6: Descriptive statistics aggregated levelVariable Min Mean Max Std. Dev. Obsconstant -0.410 0.362 1.810 0.490 35coefficient “legal system” -0.092 0.062 0.232 0.075 35coefficient “national government” -0.151 0.065 0.224 0.096 35coefficient “parliament” -0.110 0.028 0.266 0.084 35coefficient “political system” -0.053 0.002 0.103 0.029 35coefficient “national incumbent” -0.179 0.044 0.234 0.105 35type I: required 0 0.371 1 0.490 35type II: non-required/active-vote on government policy 0 0.171 1 0.382 35type III: non-required/active-vote on opposition policy 0 0.286 1 0.458 35type IV: non-required/passive 0 0.457 1 0.505 35veto-player referendum 0 0.486 1 0.507 35popular veto 0 0.371 1 0.490 35popular initiative 0 0.286 1 0.458 35

17

References

Achen, Christopher H. 2005. “Two-Step Hierarchical Estimation: Beyond Re-gression Analysis.” Political Analysis 13(4):447–456.

Alm, James and Benno Torgler. 2006. “Culture Differences and Tax Morale in theUnited States and in Europe.” Journal of Economic Psychology 27(2):224–246.

Auer, Andreas and Michael Butzer, eds. 2001. Direct Democracy: The Easternand Central European Experience. Aldershot: Ashgate.

Beck, Nathaniel. 2005. “Multilevel Analyses of Comparative Data: A Comment.”Political Analysis 13(4):457–458.

Brada, Josef C. 1996. “Privatization is Transition - Or Is It?” Journal of Eco-nomic Perspectives 10(2):67–86.

Brady, Henry E. and Cynthia S. Kaplan. 1994. Eastern Europe and the For-mer Soviet Union. In Referendums Around the World, ed. David Butler andAustin Ranney. London: MacMillan pp. 174–217.

Easter, Gerald M. 2002. “Politics of Revenue Extraction in Post-CommunistStates: Poland and Russia Compared.” Politics & Society 30(4):599–627.

Easter, Gerald M. 2006. Taxation and State Re-Formation in Russia: Policy, Ca-pacity, Compliance. In Tax Evasion, Trust, and State Capacities, ed. NicolasHayoz and Simon Hug. Bern: Peter Lang pp. 227–267.

European Values Study Group and World Values Survey Association. 2006.“EUROPEAN AND WORLD VALUES SURVEYS FOUR-WAVE INTE-GRATED DATA FILE, 1981-2004, V.20060423, 2006. Aggregate File.” Pro-ducers: Analisis Sociologicos Economicos y Polıticos (ASEP) and JD Sys-tems (JDS), Madrid, Spain/Tilburg University, Tilburg, The Netherlands.Data Files Suppliers: Analisis Sociologicos Economicos y Politicos (ASEP)and JD Systems (JDS), Madrid, Spain/Tillburg University, Tillburg, TheNetherlands/ Zentralarchiv fur Empirische Sozialforschung (ZA), Cologne,Germany:) Aggregate File Distributors: Analisis Sociologicos Economicos yPolıticos (ASEP) and JD Systems (JDS), Madrid, Spain/Tillburg Univer-sity, Tilburg, The Netherlands/Zentralarchiv fur Empirische Sozialforschung(ZA) Cologne, Germany.

Feld, Lars P. and Bruno S. Frey. 2002a. “Deterrence and Morale in Taxation: AnEmpirical Analysis.” CESifo Working Paper No. 760, August 2002.

Feld, Lars P. and Bruno S. Frey. 2002b. “Trust Breeds Trust: How Taxpayers areTreated.” Economics of Governance 3(2):87–99.

Feld, Lars P. and Bruno S. Frey. 2006. Tax Evasion in Switzerland: The Rolesof Deterrence and Tax Morale. In Tax Evasion, Trust, and State Capacities,ed. Nicolas Hayoz and Simon Hug. Bern: Peter Lang pp. 123–153.

Feld, Lars P. and Jean-Robert Tyran. 2002. “Tax Evasion and Voting: An Ex-perimental Analysis.” Kyklos 55(2):197–222.

18

Franzese, Robert J. Jr. 2005. “Empirical Strategies for Various Manifestations ofMultilevel Data.” Political Analysis 13(4):430–446.

Frey, Bruno S. 1997a. “A Constitution for Knaves Crowds Out the Civic Virtues.”Economic journal 107:1043–1053.

Frey, Bruno S. 1997b. Not Just for The Money. An Economic Theory of PersonalMotivation. Cheltenham: Edward Elgar.

Frey, Bruno S. 2002. “Direct Democracy for Transition Economies.” Paper forthe Collegium Budapest, Institute for Advanced Study.

Gerber, Elisabeth R., Arthur Lupia, Mathew D. McCubbins and D Roder-ick Kiewiet. 2000. Stealing the Initiative: How State Government Respondsto Direct Democracy. Upper Saddle River: Prentice Hall.

Hardin, Russell. 2006. Trust. Cambridge: Polity Press.

Hart, Albert, Lars P. Feld and Werner W. Pommerehne. 1997. “Steuerhin-terziehung und ihre Kontrolle in unterschiedlichen politischen Systemen.”Homo oeconomicus 14(4):469 – 487.

Hayoz, Nicolas and Simon Hug, eds. 2006. Tax Evasion, Trust, and State Capac-ities. Bern: Peter Lang.

Hug, Simon. 2004. “Occurrence and Policy Consequences of Referendums. ATheoretical Model and Empirical Evidence.” Journal of Theoretical Politics16(3):321–356.

Hug, Simon. 2005. “The Political Effects of Referendums. An Analysis of In-stitutional Innovations in Eastern and Central Europe.” Communist andPost-Communist Studies 38(4):475–499.

Hug, Simon and George Tsebelis. 2002. “Veto Players and Referendums aroundthe World.” Journal of Theoretical Politics 14(4 (October)):465–516.

Jones, Bradford S. and Marco R. Steenbergen. 2002. “Modeling Multilevel DataStructures.” American Journal of Political Science 46(1):218–237.

Kedar, Orit and W. Phillips Shively. 2005. “Introduction to the Special Issue.”Political Analysis 13(4):297–300.

Kiewiet, D. Roderick and Kristin Szakaly. 1996. “Constitutional Limitations onBorrowing: An Analysis of State Bonded Indebtedness.” Journal of Law,Economics and Organization 12(1):62–97.

Lewis, Jeffrey B. and Drew A. Linzer. 2005. “Estimating Regression Models inWhich the Dependent Variable Is Based on Estimates.” Political Analysis13(4):345–364.

Martinaz-Vazguez, Jorge and James Alm. 2002. Institutions, Paradigms, andTax Evasion in Developing and Transition Countries. In Public Finance inDeveloping and Transition Countries, ed. James Alm and Jorge Martinez-Vazquez. Edgar Elgar.

19

Martinez-Vazquez, J. and R. M. McNab. 1997. “Tax Systems in TransitionEconomies.” Working Paper 971, Georgia State University, Atlanta.

Pommerehne, Werner W., Albert Hart and Bruno S. Frey. 1994. “Tax Morale,Tax Evasion and the Choice of Policy Instruments in Different PoliticalSystems.” Public Finance 49:52–69.

Pommerehne, Werner W. and Bruno S. Frey. 1992. “The Effects of Tax Admin-istration on Tax Morale.” Unpublished Manuscript, University of Zurich.

Pommerehne, Werner W. and Hannelore Weck-Hannemann. 1996. “Tax Rates,Tax Administration and Income Tax Evasion in Switzerland.” Public Choice88(1/2):161–170.

Rabe-Hesketh, Sophia, Anders Skrondal and Andrew Pickles. 2001. “GLLAMMManual.” Technical Report 2001/01, Department of Biostatistics and Com-puting, Institute of Psychiatry, King’s College, London.

Rappard, William E. 1923. “Democracy vs. Demagogy I+II.” Political ScienceQuarterly 38(2 June):290–306, 361–389.

Raudenbusch, Stephen W. and Bryk S. Anthony. 2002. Hierarchical Linear Mod-els. Applications and Data Analysis Methods (Second Edition). ThousandsOaks: Sage Publications.

Rothstein, Bo and Dietlind Stolle. 2002. “How Political Institutions Create andDestroy Social Capital: An Institutional Theory of Generalized Trust.” Pa-per prepared for delivery at the 2002 Annual Meeting of the American Po-litical Science Association, Boston, August 29-September 1, 2002.

Schneider, Friedrich. 2002. “The Size and Development of the Shadow Economiesof 22 Transition and 21 OECD Countries,.” Discussion Paper No. 514, June,IZA Bonn.

Smith, Daniel A. 1998. Tax Crusaders and the Politics of Direct Democracy. NewYork: Routledge.

Suksi, Markku. 1993. Bringing in the People: A Comparison of ConstitutionalForms and Practices of the Referendum. Dordrecht: Martinus Nijhoff.

Torgler, Benno. 2003a. “Tax Morale in Transition Countries.” Post-CommunistEconomies 15:357–381.

Torgler, Benno. 2003b. “Tax Morale, Rule Governed Behaviour and Trust.” Con-stitutional Political Economy 14(2):119–140.

Torgler, Benno. 2003c. “To Evade Taxes or Not: That Is the Question.” Journalof Socio-Economics. 32(3):283–302.

Torgler, Benno. 2005. “Tax Morale and Direct Democracy.” European Journal ofPolitical Economy 21:525–531.

Torgler, Benno. 2006. Tax Morale in Central and Eastern European Countries.In Tax Evasion, Trust, and State Capacities, ed. Nicolas Hayoz and SimonHug. Bern: Peter Lang pp. 155–185.

20

Tsebelis, George. 2002. Veto Players: How Political Institutions Work. Princeton:Princeton University Press.

Uslaner, Eric M. 2006. Tax Evasion, Trust, and the Strong Arm of the Law.In Tax Evasion, Trust, and State Capacities, ed. Nicolas Hayoz and SimonHug. Bern: Peter Lang pp. 17–50.

White, Stephen and Ronald J. Hill. 1996. Russia, the former Soviet Union andEastern Europe: The Referendum as a Flexible Political Instrument. In TheReferendum Experience in Europe, ed. Michael Gallagher and Pier VincenzoUleri. London: MacMillan pp. 153–170.

rtte/simon/ April 20, 2007

21