THE CAUSES AND CONSEQUENCES

OF
CORRUPTION
 
 
 

A THESIS PRESENTED IN FULFILMENT OF THE

REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY IN ECONOMICS
 

March 2011

BIN DONG
 
 
 
 

 
School of Economics and Finance
Faculty of Business
Queensland University of Technology
Gardens Point Campus
Brisbane Australia

 
 

Acknowledgements

I would like to express my gratitude to my principal supervisor Professor Benno Torgler, for
walking me through the journey of PhD study and for being there at every step as a source of
inspiration, motivation and moral support. Professor Torgler’s excellent supervision,
invaluable guidance, suggestions, corrections and empirical skills have helped shape much of
this thesis. I would also like to extend deepest appreciation to my associate supervisor
Professor Uwe Dulleck for his invaluable guidance and encouragement throughout this study.
I extend my thanks to Dr. David Johnston for offering invaluable comments and suggestions
for empirical analysis. I also extend special thanks to the members of the administrative staff
of the School of Economics and Finance.
I am extremely grateful to China Scholarship Council and Queensland University of
technology for jointly providing me the financial support for my PhD study.
I am deeply indebted to my wife Jin Kang, who has been the motivational force in my life,
and thank her for her patience, understanding and invaluable support during the PhD study. I
am grateful to my mother, brother for their selfless and unreserved support over the years.
Finally I would like to thank Ms Ying Zhou, Mr Tony Beatton and Mr Markus Schaffner
for their assistance and support.

i 
 
 

Abstract

This thesis comprehensively studies the causes and consequences of corruption in both cross-
country and within-country contexts, mainly focusing on China.
The thesis commences by extensively investigating the causes of corruption. Using the
standard economic approach, this study finds that in China regions with more anti-corruption
efforts, higher education attainment, Anglo-American historic influence, higher openness,
more access to media, higher relative wages of government employees, and a greater
representation of women in legislature are markedly less corrupt; while the social
heterogeneity, deregulation and abundance of resources, substantially breed regional
corruption. Moreover, fiscal decentralization is discovered to depress corruption significantly.
This study also observes a positive relationship between corruption and the economic
development in current China that is mainly driven by the transition to a market economy.
Focusing on the influence of political institutions on corruption, the thesis then provides
evidence that a high level of political interest helps to reduce corruption within a society,
while the effect of democracy upon corruption depends on property rights protection and
income distribution. With the social economic approach, however, the thesis presents both
cross-country and within-country evidence that the social interaction plays an important role
in determining corruption.
The thesis then continues by comprehensively evaluating the consequences of corruption
in China. The study provides evidence that corruption can simultaneously have both positive
and negative effects on economic development. And it also observes that corruption
considerably increases the income inequality in China. Furthermore this study finds that
corruption in China significantly distorts public expenditures. Local corruption is also
observed to substantially reduce FDI in Chinese regions. Finally the study documents that
corruption substantially aggravates pollution probably through a loosening of the
environmental regulation, and that it also modifies the effects of trade openness and FDI on
the stringency of environmental policy.
Overall, this thesis adds to the current literature by a number of novel findings concerning
both the causes and the consequences of corruption.

Key Words: corruption, causes, consequences, China, democracy, social interaction, political
interest, economic development.

iii 
 
 

Table of Contents

Acknowledgements ................................................................................................................................. i 
Statement of Original Authorship ......................................................................................................... ii 
Abstract ................................................................................................................................................. iii 
Table of Contents .................................................................................................................................. iv 
List of Tables ......................................................................................................................................... vii 
List of Figures ......................................................................................................................................... ix 
Chapter One  Introduction .................................................................................................................. 1 
1.1 Motivation of Thesis ..................................................................................................................... 1 
1.2 Content of Thesis .......................................................................................................................... 2 
1.2.1 Causes of Corruption .............................................................................................................. 2 
1.2.2 Consequences of Corruption ................................................................................................. 7 
1.2.3 Methodology issues ............................................................................................................... 8 
1.3 Structure of Thesis ....................................................................................................................... 10 
Chapter Two  Economic Determinants of Corruption: Chinese Evidence ....................................... 12 
2.1 Introduction ................................................................................................................................ 12 
2.2 Determinants of Corruption ........................................................................................................ 15 
2.3 Empirical Analysis ........................................................................................................................ 18 
2.3.1 Province‐level Analysis ......................................................................................................... 20 
2.3.2 City‐level Analysis ................................................................................................................. 35 
2.4 Conclusion ................................................................................................................................... 38 
Appendix ........................................................................................................................................... 41 
Chapter Three  Political Interest and Corruption: Cross‐country Evidence .................................. 44 
3.1 Introduction ................................................................................................................................ 44 
3.2 Political Interest .......................................................................................................................... 45 
3.2.1 Theoretical Considerations .................................................................................................. 45 
3.2.2 A Simple Model .................................................................................................................... 46 
3.3 Data ............................................................................................................................................. 48 
3.3.1 Dependent Variables ............................................................................................................ 48 
3.3.2 Measuring Political Interest ................................................................................................. 50 
3.3.3 Further Independent Variables ............................................................................................ 51 
3.4 Empirical Evidence ...................................................................................................................... 56 
3.4.1 International Evidence ......................................................................................................... 57 

iv 
 
 

3.4.2 Within‐Country Evidence ..................................................................................................... 70 
3.5 Conclusion ................................................................................................................................... 71 
Chapter Four  Democracy and Corruption: Cross‐country Evidence ............................................... 73 
4.1 Introduction ................................................................................................................................ 73 
4.2 Theoretical Model ....................................................................................................................... 75 
4.2.1 Model ................................................................................................................................... 76 
4.2.2 Economic Equilibrium .......................................................................................................... 77 
4.2.3 Political Equilibrium ............................................................................................................. 78 
4.3 Empirical Evidence ...................................................................................................................... 80 
4.3.1 Methodology and Data ........................................................................................................ 81 
4.3.2 Results .................................................................................................................................. 85 
4.4 Conclusion ................................................................................................................................... 92 
Appendix ........................................................................................................................................... 93 
Chapter Five  Social interaction and Corruption: Cross‐country Evidence ..................................... 94 
5.1 Introduction ................................................................................................................................. 94 
5.2 Theoretical Foundation ................................................................................................................ 96 
5.2.1 Background of the Model .................................................................................................... 98 
5.2.2 A Simple Game ................................................................................................................... 100 
5.2.3 Dynamics ............................................................................................................................ 102 
5.2.4 Conditional Corruption – Discussion and Extensions ........................................................ 102 
5.3 Data and Methodological Approach ......................................................................................... 103 
5.3.1 Micro Analysis .................................................................................................................... 103 
5.3.2 Macro Analysis ................................................................................................................... 107 
5.4 Results ....................................................................................................................................... 107 
5.4.1 Micro Level using the EVS .................................................................................................. 107 
5.4.2 Micro Level using the WVS ................................................................................................ 117 
5.4.3 Macro Level Using a Large Panel Data Set ......................................................................... 121 
5.5 Conclusion ................................................................................................................................. 125 
Chapter Six  Social interaction and Corruption: Within‐country Evidence ................................. 127 
6.1 Introduction .............................................................................................................................. 127 
6.2 Theoretical Model ..................................................................................................................... 129 
6.3 Empirical Work .......................................................................................................................... 132 
6.3.1 Data and Methodology ...................................................................................................... 132 
6.3.2 Results ................................................................................................................................ 136 

v 
 
 

6.4 Conclusion ................................................................................................................................. 139 
Appendix ......................................................................................................................................... 141 
Chapter Seven  Consequences of Corruption: Chinese Evidence ................................................ 142 
7.1 Introduction .............................................................................................................................. 142 
7.2 Literature Review ...................................................................................................................... 145 
7.3 Empirical Analysis ...................................................................................................................... 149 
7.3.1 Data and Methodology ...................................................................................................... 149 
7.3.2 Corruption, Economic Growth and Income Distribution ................................................... 153 
7.3.3 Corruption and Foreign Direct Investment ........................................................................ 160 
7.3.4 Corruption and Public Expenditures .................................................................................. 162 
7.3.5 Corruption and the Environment ....................................................................................... 164 
7.4 Conclusion ................................................................................................................................. 169 
Appendix ......................................................................................................................................... 171 
Chapter Eight  Conclusion ............................................................................................................ 172 
8.1 Summary of Findings ................................................................................................................. 172 
8.2 Policy Implications .................................................................................................................... 174 
8.3 Further Research ....................................................................................................................... 175 
References .......................................................................................................................................... 177 

vi 
 
 

List of Tables

Table 2‐1 GDP (PPP) per capita of Chinese regions in 2008 (Intl. $) ..................................................... 14 
Table 2‐2 Average annual registered cases on corruption across regions in China (1998‐2007) ......... 20 
Table 2‐3 Corruption and its determinants in China: pooled OLS estimation ...................................... 32 
Table 2‐4 Corruption and its determinants in China: fixed effects OLS estimation ............................. 33 
Table 2‐5 Corruption and its determinants in China: fixed effects 2SLS estimation ............................ 34 
Table 2‐6 Corruption and its determinants in Chinese cities ................................................................ 37 
Table 2‐7 Data description .................................................................................................................... 41 
Table 2‐8 Pairwise correlation coefficients between explanatory variables in the province‐level 
analysis .................................................................................................................................................. 42 
Table 2‐9 Pairwise correlation coefficients between explanatory variables in the city‐level analysis . 42 
Table 2‐10 First‐Stage Regressions Based on Table 2‐5 ........................................................................ 43 
Table 3‐1 Justifiability of corruption and political discussion ............................................................... 59 
Table 3‐2 Perceived corruption and political discussion ...................................................................... 60 
Table 3‐3 Justifiability of corruption and interest in politics ................................................................ 61 
Table 3‐4 Perceived corruption and political interest .......................................................................... 62 
Table 3‐5 Justifiability of corruption and important of politics in life .................................................. 63 
Table 3‐6 Perceived corruption and importance of politics in life ....................................................... 64 
Table 3‐7 2SLS results ........................................................................................................................... 67 
Table 3‐8 Justifiability of corruption in Switzerland ............................................................................. 68 
Table 3‐9 Perceived corruption in Switzerland ..................................................................................... 69 
Table 4‐1  Descriptive Statistics ............................................................................................................ 84 
Table 4‐2 Effect of democracy on corruption: review and implication (fixed effects results) ............. 86 
Table 4‐3 Effect of democracy on corruption: fixed effect results ....................................................... 87 
Table 4‐4 Effect of democracy on corruption: alternative measure of democracy.............................. 88 
Table 4‐5 Effect of democracy on corruption: IV results ...................................................................... 89 
Table 4‐6 Marginal effect of democracy on corruption ........................................................................ 91 
Table 4‐7 Validity of instrument: Muslim ............................................................................................. 93 
Table 5‐1 Influence of conditional corruption (EVS) ........................................................................... 110 
Table 5‐2 Robustness test and the influence of conditional corruption using micro and macro proxies 
(EVS) .................................................................................................................................................... 113 
Table 5‐3 2SLS results (EVS) ................................................................................................................ 115 
Table 5‐4 Causality discussion (Filtering) ............................................................................................ 116 
Table 5‐5 Conditional corruption using WVS ...................................................................................... 119 
Table 5‐6 2SLS results (WVS) .............................................................................................................. 120 
Table 5‐7 Causality discussion filtering with WVS Data ...................................................................... 121 
Table 5‐8 Evidence at the macro level ................................................................................................ 125 
Table 6‐1 Literature summary ............................................................................................................ 129 
Table 6‐2 Variables description, 1998—2007 ..................................................................................... 134 
Table 6‐3 Corruption and social interaction ....................................................................................... 138 
Table 6‐4 Average annual registered cases on corruption per capita across regions in China (1998‐
2007) ................................................................................................................................................... 141 

vii 
 
 

Table 6‐5 Pairwise correlation coefficients between variables .......................................................... 141 
Table 7‐1 GDP (PPP) per capita of Chinese regions in 2008 (Intl. $) ................................................... 144 
Table 7‐2 Average annual registered cases on corruption per capita across regions in China (1998‐
2007) ................................................................................................................................................... 150 
Table 7‐3 Effect of corruption on economic growth: cross‐province evidence .................................. 154 
Table 7‐4 Effect of corruption on economic growth: cross‐city evidence .......................................... 158 
Table 7‐5 Effects of corruption on income inequality ........................................................................ 159 
Table 7‐6 Effect of corruption rate on inbound FDI ............................................................................ 161 
Table 7‐7 Effects of corruption on public expenditures ..................................................................... 163 
Table 7‐8 Effect of corruption and the environment .......................................................................... 165 
Table 7‐9 Effect of corruption on environment policy ....................................................................... 168 
Table 7‐10 Data description ................................................................................................................ 171 

viii 
 
 

List of Figures

Figure 1‐1 Thesis Structure ................................................................................................................... 11 
Figure 2‐1 Determinants of corruption in China ................................................................................... 38 
Figure 4‐1 Relationship between democracy and corruption .............................................................. 85 
Figure 5‐1 Correlation between Justifiability of Corruption and Perceived Corruption ....................... 96 
Figure 5‐2 Description of the Corruption Game ................................................................................. 101 
Figure 5‐3 Correlation between Justifiability of Corruption and Control of Corruption .................... 106 
Figure 7‐1 Marginal effects of trade openness on environmental stringency conditional on corruption
 ............................................................................................................................................................ 166 
 

ix 
 
 

Chapter One Introduction

1.1 Motivation of Thesis

Corruption, understood as ‘‘abuse of public office for private gain’’ is a persistent feature in
human societies throughout time and space. Contemporaneous corruption scandals not only
occur in developing countries such as Nigeria, India, and China where corruption is regarded
as a norm, but also in developed economies such as France, Britain and America. The sale of
parliamentary seats in ‘rotten boroughs’ in England before the Reform Act of 1832 1 and
‘machine politics’ in larger cities in America in the late 19th and early 20th century2 are two
famous historical examples. Even in Scandinavian countries, like Sweden and Norway, which
are supposedly free-from-corruption, managers of state owned companies have been found to
take bribes.
Corruption in the public sector is viewed as the major obstacle to economic development
(Kaufmann, 1997). Solid evidence (for example, Mauro 1995, and World Bank, 1997)
demonstrates the pernicious effects of corruption upon, among other things, investment,
economic growth, environmental quality and therefore social welfare. In effect, a country is
adversely affected by the existence of corruption, and therefore anti-corruption policies are
important.
Reducing corruption requires a precise understanding of its causes and consequences. The
development of effective anti-corruption policies is based on a thorough investigation of
corruption within and across countries. However, in current research the causes and
consequences of corruption remain poorly understood and are broadly disputed. As a result, it
is difficult for governments to design coherent policies to control corruption.
This study provides new insight into the causes and consequences of corruption. We
explore the discussed factors in a within country environment to provide evidence outside the
US and in a more controlled environment, and also provide within-country and cross-country
evidence at the micro level to explore new theories in the area of corruption such as
conditional corruption. In detail the thesis first empirically examines the theoretical causes of
corruption suggested in literature in both cross-country and within-country contexts. The

                                                            
1
 Pearce, R. and Stearn, R., 2000. Access to History, Government and Reform: Britain 1815-1918 (Second
Edition). Hodder & Stoughton.
2
 Clifford, T. P., 1975. The Political Machine: An American Institution. Vantage Press.

1 
 
 

author robustly identifies the effects of its economic, political and social determinants on
corruption, employing solid statistical tools dealing with causal relationship between
observed factors. Secondly, this study comprehensively investigates the various
consequences of corruption, focusing on China, the largest developing country central to
world economy. According to my knowledge, there are few studies on the consequences of
Chinese corruption. In summary, this study is expected to make a substantial contribution to
the research of the causes and consequences of corruption, and therefore to add to effective
policy guidelines to curb corruption.

1.2 Content of Thesis

This section provides a succinct portrait of the thesis. Initially we briefly review previous
literature to build up the logical framework of the thesis (specific literature reviews are
provided in each of the individual chapters that follow). Then based on the framework that
has been introduced, the main findings of the thesis are then presented.
1.2.1 Causes of Corruption
There has been a wave of empirical studies on the causes and consequences of corruption in
recent years. With respect to the causes of corruption, this study, similar to Bardhan (2006),
points out that there are generally two different approaches to research the causes of
corruption, namely the standard economic approach and also the social economic approach.
The standard economic approach emphasizes incentives and punishments in corrupt acts
following Becker’s analytical framework (1968). According to this approach, there are three
prerequisites necessary for the incidence of corruption (Jain, 2001). First, bureaucrats have
discretionary power. Second, this power is associated with economic rents. Finally, the
deterrence to corruption, as a function of the probability of being caught and the penalty for
the corrupt act, is adequately low. The first two preconditions determine the benefit of
corruption, while the last precondition influences the cost of corruption.
Many studies adopting this approach concentrate upon economic conditions and policies
influencing the cost and/or benefit of corruption. Literature shows that regulation and
decentralization are the main determinants of the discretionary power of a government.
Economic rents, on the other hand, increase with natural resource abundance, but decrease
with economic competition proxied by trade openness. All of these factors are observed to
substantially affect the benefit of corruption (e.g. Ades and Di Tella, 1999; Fisman and Gatti,
2002a, b).

2 
 
 

The deterrence of corruption is a joint function of the possibility of being detected and the
punishment once caught. High levels of economic development, education attainment and
media access have been documented to reduce corruption by raising the possibility that
corrupt acts are detected (for example, Treisman, 2000). Historical influence also plays an
important role in corruption (see also, Treisman, 2000). Furthermore corruption has also been
found to be negatively correlated with female representation in politics, possibly because
women may feel a larger probability of being caught in an act of corruption (e.g. Dollar et al.,
2001). Social and economic heterogeneity is also an indirect determinant of the probability of
corrupt acts being caught. For example, ethnical fractionalization is believed to promote
corruption since corrupt officials may be protected by their own ethnic groups for political
reasons (see also Treisman, 2000). Finally, the relatively high wage of the public sector
implies a high opportunity cost when officials are ousted due to corruption. As a proxy for
the punishment, the (relative) wage of the public sector is found to be negatively associated
with the corruption level (e.g. Van Rijkeghem and Weder, 2001)
Studies on the causes of corruption by and large perform cross-national analyses using
subjective survey data (for example, Treisman, 2000; Fisman and Gatti, 2002a). This kind of
study, although fruitful, cannot circumvent two problems. Firstly, subjective survey data
might be biased as Treisman (2007) argues: “the data do not measure corruption itself but
opinion about its prevalence” (p. 215). Secondly, cross-country analysis often suffers from
omitted variable bias. Substantial unobservable or unmeasurable differences in institution and
culture between countries make cross-country results problematic. The disadvantages
experienced by cross-national studies can be avoided if we use within-country objective data,
since objective data do not suffer the bias of subjective data. Furthermore, homogeneity
within a country also mitigates the omitted variable bias troubling the cross-country analysis.
However, current within-country data are only proxies for corruption since corruption is
actually secretive and hence difficult to measure directly. Goel and Nelson (1998), Fisman
and Gatti (2002b), and Glaeser and Saks (2006) utilize the objective data: the number of
public officials convicted for abuse of public office as an indicator of the actual levels of
corruption in American states. However, this indicator may also reflect the anti-corruption
efforts of local judiciary. As Lambsdorff (2005) point out “the appropriateness of such data
as a proxy for corruption has thus been widely disputed” (p.1). Therefore the ideal strategy
might be to investigate the causes of corruption with both the cross-country analysis using
subjective data and the within-country analysis using objective data to get complementary
results, which also makes empirical findings robust.
3 
 
 

While there have been enormous cross-country studies, papers on the causes of corruption
using within-country data are few, and most of them are working with US data. This thesis
hence initially contributes to literature with a study on the causes of corruption in China. A
study of China has a unique advantage. Firstly, it is helpful to understand corruption in
developing and transitional economies where it is one of the central issues. Secondly, China
is well-suited for a study on corruption. On the one hand, China is fairly homogenous in
institutions, culture, and social structure. This helps us to mitigate the omitted variable bias in
empirical analysis. On the other hand, there are great economic differences between the
eastern and western provinces in China, which might make findings of corruption in China
more generalizable on a global level.
Chapter 2 in the thesis adopts a standard economic approach to explore the causes of
corruption in China using two different data sets, namely a province-level data set and a city-
level data set, to obtain robust results. China is a key player in the world economy and will
gain even further importance in the future. This study examines almost all cross-country
findings in a Chinese context using the regional number of registered cases on corruption as a
measure of corruption. Besides confirming most cross-country findings in a more controlled
setting, this study adds to literature in several ways. Firstly, this chapter use behavioural
variables3 rather than attitudinal variables (perceptions of corruption) to proxy for corruption.
Secondly anti-corruption efforts are always controlled in this study to isolate the component
of anticorruption efforts in our corruption measure though many studies including the small
amount of studies using a behavioural proxy for corruption have neglected this. Thirdly, this
study identifies a positive relationship between corruption and economic development (and
marketization) in China due to the transition process in China. Fourthly, this study provides
some novel within-country evidence such as the negative effect of British historic influence,
the positive effect of natural resource abundance, and the negative effect of female
representation in politics on corruption. Fifthly, this study presents solid evidence that even in
a nondemocratic country the access to controlled media still checks corruption.
There are a number of studies concentrating on the influence of political institutions on
corruption using the standard economic approach. Good political institutions help to control
and monitor the government and therefore reduce corruption. There are actually two kinds of
political institutions: Formal institutions such as democracy and informal ones such as
                                                            
3
 The registered cases on corruption in procurator’s offices of provinces, and the average ratios of the travel and
entertainment costs relative to the sales of investigated firms in Chinese cities in the survey conducted by World
Bank in 2005. 

4 
 
 

political interest. Econometrically, the effect of formal institutions like democracy on

corruption can only be analysed at the macro level in the cross-country context, while
political interest is useful to analyse (informal) institutions at the micro/individual level.
Democracy is theoretically supposed to reduce corruption mainly because political
competition may provide checks against corruption. “In democratic systems, competitors for
office have an incentive to discover and publicize the incumbent’s misuse of office whenever
an election beckons.” (Treisman, 2000, p. 404) This therefore raises the possibility that
corrupt acts can be detected. However, the relationship between corruption and democracy is
empirically found to be complex. Besides the linear relationship mentioned above, a
quadratic relationship between these factors, is also supported by several theoretical and
empirical articles (e.g. Mohtadi and Roe, 2003, Rock, 2007). Moreover, some scholars such
as Treisman (2000) suggest that it may take a long time for democracy to substantially reduce
corruption. Further evidence is clearly necessary. It is worth noting however that variation in
political institution within a country is not large enough in many cases for economists to
identify the relationship between democracy and corruption. This thesis therefore mainly
utilizes cross-country data sets to examine the influence of political institutions on corruption.
The thesis, for the first time ever, investigates in Chapter 3 the relationship between
political interests: an informal aspect of political institution and corruption since citizens’
political interest contributes to the probability of their being involved in the political process
(Verba et al., 1995). Innovatively, this study uses the micro-level data from the World Values
Survey to explore the impact of political interest represented by three different proxies on
both the perception of corruption and the justifiability of corruption reflecting the social norm
of corruption. It is worth noting that unlike the macro-level analysis which is popular in the
corruption study, the micro-level study is able to measure the individual characteristics and
induce robust relationships due the large amount of observations. Furthermore, as can be seen
below, it allows researchers to explore new theories such as conditional corruption.
Specifically, this study first performs a cross-country analysis with a huge data set, and then
runs a within-country analysis focusing on Switzerland to check the robustness of cross-
country results. Both analyses clearly show that a high level of political interest helps to
reduce the level of corruption within a society.
In Chapter 4 the thesis shed new light on the relationship between democracy and
corruption. To disentangle the actual linkage from previously mixed evidence, this study first
establishes a new political economy model demonstrating that the effect of democracy on
corruption is conditional on income distribution and property rights protection. Then with a
5 
 
 

cross-national panel data, this study clearly shows that the previous empirical findings lose
significance when considering the interactions between corruption, property rights protection
and income distribution. The effect of democracy upon corruption is empirically observed to
depend on the protection of property rights and income inequality. This thesis hence provides
new theoretical and empirical evidence concerning the effect of democracy on corruption.
    The social economic approach however insists that corruption to some extent arises from
social norms. It emphasizes the role of group dynamics as well as culture and history, in
determining corruption. Put simply, corrupt people will feel less guilt (moral cost) if they find
many others engaged in similar activities, and vice versa. Culture here coordinates the
expectation on others’ behaviour in a society, while history provides the initial condition.
Multiple equilibria are often expected in this circumstance. A society with an initially high
corruption level may get ‘‘locked in’’ until a ‘big push’ similar to what happened in Hong
Kong takes place (Aidt, 2003). Economists however have not focused on this approach
heretofore. Only Goel and Nelson (2007) have studied the contagion of corruption in
America.
This thesis however attempts to fill the gap with both cross-country and within-country
analyses. The study presented in Chapter 5, according to my knowledge, is the first cross-
country analysis studying the role of social interactions or social norms in corruption. This
study builds first a behavioural model to innovatively argue that engaging in corruption
results in a disutility of guilt. Guilt itself depends on a (current and past) perceived prevalence
of corruption within a society. As a novelty the empirical section presents a large amount of
evidence about the role of social interactions with two large micro level data sets and a large
macro level panel data sets covering almost 20 years. The results clearly indicate that a
willingness to be corrupt is influenced by the perceived activities of others, and the past level
of corruption. The findings above therefore underscore the relevance of social interactions on
the area of corruption. Furthermore the results also complement a large set of laboratory
experimental studies (for example, Falk, Fischbacher and Gächter, 2003) that have studied
conditional cooperation by providing evidence outside of a lab setting.
As mentioned before, the study above may suffer both omitted variable bias and
subjective data bias. Chapter 6 of this thesis then, from a different angle, turns to examine the
role of social interaction on corruption within a Chinese context using province-level panel
data. As a novelty, this study, unlike Goel and Nelson (2007), simultaneously investigates the
impacts of both the corruption level of neighbours and the corruption level in the past upon
contemporary corruption. Robust evidence is also presented that social interaction plays an
6 
 
 

important role in determining corruption rates in China. The thesis therefore contributes to
literature both by cross-country and by within-country evidence on the relevance of social
interaction in corruption.
1.2.2 Consequences of Corruption
Corruption is believed to have a detrimental effect on economic development and hence
social welfare. Many studies examine the relationship between corruption and economic
growth since there is indeed a debate on the effect of corruption on economic growth. Some
scholars (for example, Leff, 1964, and Huntington, 1968) argue that corruption may improve
efficiency and hence promote economic growth by allowing enterprisers to circumvent
cumbersome regulations with bribes especially in developing countries. However, the
majority of literature insists that corruption lowers economic growth because it may reduce
the incentive of private investment (Bardhan, 1997), distort public investment decisions
(Tanzi and Davoodi, 1997), and induce talented people into rent-seeking activities (Murphy,
Shleifer and Vishny, 1991). Most empirical studies indeed support the fact that corruption
impedes economic growth mainly through channels of investment, openness and political
instability (for example, Mauro, 1995, Mo, 2001).
Specifically, corruption is found to reduce foreign direct investment (e.g. Wei, 2000a)
because high corruption in host countries may imply high expropriation risk. Moreover,
Fredriksson et al. (2003) show that corruption may influence FDI through another channel:
environmental regulation. On the other hand, corruption may distort public investment.
According to Mauro (1998), corrupt politicians may increase public expenditure easy to
collect bribes, while decreasing expenditure providing fewer bribery opportunities.
Furthermore he empirically observes that corruption significantly reduces public expenditure
on education.
Corruption also substantially affects income distribution. Gupta, Davoodi and Alonso-
Terme (2002) find that corruption significantly increases income inequality, while Li et al.
(2000) observe that corruption influences income inequality in a reversed U-shaped manner.
The adverse effects of corruption on the environment are also documented in literature.
Welsch (2004) found that corruption aggravates pollution especially in developing countries,
while Cole (2007) provides seemingly contradicting evidence. More investigation is therefore
needed. Pellegrini and Gerlagh (2006a, b) however provide solid evidence that corruption has
a substantially negative effect on the environment policy stringency, which may imply that
corruption affects pollution mainly through environment policy making. Furthermore, both

7 
 
 

theoretical and empirical evidence has shown that corruption not only reduces the stringency
of environmental policy but also modifies the effects of other determinants of environment
policy (Fredriksson and Svensson, 2003, Damania et al., 2003, and Cole et al., 2006).
In Chapter 7 this thesis, for the first time, comprehensively investigates consequences of
corruption with complementary Chinese data sets and alternative corruption measures. The
study contributes to existing literature in several ways. Firstly, it suggests a novel perspective
of the influence of corruption on economic growth. Specifically, this study empirically finds
that corruption has simultaneously both positive and negative effects on the economic growth
of China. The overall impact of corruption might be the balance of the two effects, both of
which may depend on institutional environments. Secondly, this study provides, novel
within-country evidence that corruption increases income inequality, decreases FDI and
distorts public spending in China. Thirdly, this study documents that corruption substantially
aggravates pollution probably through loosening environmental regulation in China, and that
it modifies the effects of trade openness and FDI on the stringency of environmental policy in
a similar manner to that observed in literature to date.
There are still two important issues to be addressed. First, one may notice there is an
imbalance in regards to the number of papers on the causes of corruption compared to the
consequences of corruption. The current research project actually starts with the study about
the causes of Chinese corruption since a better understanding of the causes of corruption is
first of all required before one is able to analyse the consequences of corruption. The author
will however complete the project with more papers studying the consequences of Chinese
corruption (for example, from the micro/individual perspective) in the near future. Second, it
is worth noting that chapters 2 to 7 in this thesis are respectively built on independent papers
submitted to or published in academic journals. In order to retain the completeness of the
research project in each chapter, some overlap between chapters especially chapters
concerning corruption in China is allowed. Furthermore, the writing styles of chapters in the
thesis are a bit different due to the different requirements of journals the chapters submitted
to.
1.2.3 Methodology issues
This thesis attempts to reliably identify the causality between corruption and other relevant
factors. The important methodology issues which are the key to identification are briefly
discussed here. Detailed methodology will be described in the following chapters.
The first issue we need to address is how to measure corruption. Indices of perceived

8 
 
 

corruption such as the Corruption Perceptions Index published by Transparency International

and a corruption rating constructed by the Political Risk Services have been often used to
measure corruption in many cross-country studies. These indices are actually based on the
subjective assessments of experts or survey respondents of the extent of corruption in various
countries. The subjective indices are indeed closely correlated with each other although they
are complied by different organizations with different methodologies, suggesting “that these
different spyglasses are aimed at a common target” (Treisman 2007, p.216). Furthermore, the
perceptions indices are proved to be highly correlated with a range of generally believed
corruption determinants, indicating that they are “a helpful contribution to the understanding
of real levels of corruption” (Lambsdorff 2004, p.6). However, as Treisman (2007) argue,
corruption perception data actually reflect impressions of corruption intensity rather than
corruption itself, meaning that the data are actually correlated with survey respondents’
beliefs and other social and economic conditions (see also, Knack 2006). Such data, therefore,
cannot be convincingly used as dependent variables because their measurement error is
associated with many other background characteristics that are affected by explanatory
variables (Bertrand and Mullainathan 2001). The objective data set, usually within-country,
however can eliminate this kind of subjective data bias. For example, using the number of
public officials convicted for abuse of public office in American states as a proxy for
corruption, Glaeser and Saks (2006) examine the causes of and consequences of corruption in
America. However Lambsdorff (2004) argue that this kind of objective data may reflect the
quality of the judiciary rather than the actual corruption level. The reasonable strategy in the
research of corruption is therefore complementally using subjective and objective data to
measure corruption.
Corruption is econometrically a messy environment to analyse. According to Leamer
(1983), sensitivity analysis is therefore required here to ensure the credibility of the current
study. The detailed strategy that this thesis adopts is conducting a lot of robustness tests, 
exploring the issue with different data sets, proxies for corruption, and specifications, and
using micro and macro, within country and cross-country evidence, to show a robust picture.
Besides, the endogeneity issue needs to be addressed when identifying the causality in the
study. Two strategies are utilized to remove the potential endogeneity bias. The first strategy
is, whenever possible, to control for unobserved individual characteristics influencing both
corruption and relevant factors by including individual fixed effects in our panel regressions.
As Mo (2001) points out, “Corruption is commonly considered to be an institutional problem
that lasts for a long period” (p. 70). Fixed effect regressions therefore are suitable for the
9 
 
 

investigation of the relationship between corruption and other factors since the major source
of potential bias in our regressions might be time-invariant historical factors.
However, fixed effect regressions do not necessarily identify the causality between
corruption and other relevant factors. Fixed effects regressions cannot guarantee the causality,
since there might be time-varying omitted factors affecting both corruption and the relevant
factors. Fixed effects are not a substitute for instrumental variables. The second strategy to
address the endogeneity problem therefore is to adopt the instrumental variable approach to
identify the causality between corruption and relevant factors in our (fixed effects)
regressions. The key issue of the IV approach, the selection of instrument variables, will be
discussed in detail as the analysis proceeds.

1.3 Structure of Thesis

As mentioned before, this thesis mainly focuses on corruption in China, the largest
developing country in the world. According to the logic framework set up above, this study is
organised as follows. The first part of the thesis is Chapter 1, where an introduction of the
study is presented. The second part of the thesis is composed of Chapters 2, 3, 4, 5, and 6,
where the causes of corruption especially in China are extensively investigated. The standard
economic approach is first adopted in Chapter 2, 3 and 4. Chapter 2 examines the effects of
the economic determinants of corruption suggested by cross-country studies in China with
two different data sets. Chapter 3 and 4 analyse the political determinants of corruption with
cross-country data sets. Specifically, Chapter 3 investigates the relationship between political
interest and corruption, while Chapter 4 explores the effect of democracy upon corruption.
Findings of these two chapters may help to predict the influence of democratisation on
corruption in China although China now is an authoritarian country. Chapter 5 and 6,
however, follow the social economic approach to examine the role of social factors in
determining corruption. Both cross-country analyses and within-country analyses (China) are
performed here to make the findings solid since few studies have been previously undertaken
in this area. The third part of the thesis, namely Chapter 7, comprehensively investigates the
adverse effects of corruption on the economic development in China at a macro level. Finally
this thesis finishes with the fourth part, Chapter 8, which provides a summary of all the
findings above and some concluding remarks.
The detailed structure of this thesis is shown in Figure 1-1.

10 
 
  

Figure 1-1 Thesis Structure

INTRODUCTION Introduction Chapter 1

 

Economic determinants of corruption

Standard Economic Approach Economic determinants of corruption: Chinese evidence Chapter 2

Political determinants of corruption

CAUSES OF CORRUPTION Political interest and corruption: cross-country evidence Chapter 3

Democracy and corruption: cross-country evidence Chapter 4

Social Economic Approach Social determinants of corruption

Social interaction and corruption: cross-country evidence Chapter 5

Social interaction and corruption: Chinese evidence Chapter 6

CONSEQUENCES OF CORRUPTION Consequences of corruption: Chinese evidence Chapter 7

CONCLUSION Conclusion Chapter 8

11 
 
  

Chapter Two Economic Determinants of Corruption: Chinese Evidence4

2.1 Introduction

Since the establishment of the People’s Republic of China in 1949, corruption has vexed the
national leadership, which prior to 1978 attempted to control it primarily through mass
movements but occasionally with severe deterrents like the 1952 execution of two senior
officials, Qingshan Liu and Zishan Zhang. Since the 1978 launch of economic reform,
however, corruption has become even more widespread and according to Liu (1983), exists at
every level of China’s political system. As a result, the 1989 market price of coal, for
example, was 674 percent of the subsidized price, other producer goods sell at prices
substantially higher than those fixed by the state, and payoffs to ensure the supply of
products at state prices are very common (Rose-Ackerman 1999). Corruption in the form of
applicant bribery is also widespread in the area of enterprise licensing because industrial and
commercial enterprises in China must obtain government authorization to operate (Manion
1996). Liu (1983) thus differentiates between three types of corruption: “corrupt acts such as
embezzlement and bribes, which are common place among nations having a political system
to speak of; … appropriation of public goods, illegal trade, and housing irregularity, [which
result] from a breakdown in the central allocation system and [are] commonplace among
socialist nations … [and the] rather peculiarly Chinese Communist [practices of] illegitimate
feasting, feudal rites, false models, and illegal imprisonment and torture” (p. 603). Even the
Chinese government has admitted that corruption “is now worse than during any other period
since New China was founded in 1949. It has spread into the Party, into Government
administration and into every part of society, including politics, economy, ideology and
culture” (Liang 1994, p. 122). The seriousness of this problem is exemplified by the recent
charges against two members of the Politburo, Xitong Chen and Liangyu Chen, who accepted
huge bribes.
Not surprisingly, such rampant corruption, which seems to be a distinct feature of
contemporary China, the largest transitional and developing country, has generated much
literature, especially in sociology and political science (e.g, White 1996; Wedeman 2004;
Gong 2006). From an economics perspective, Yao (2002) argues that corruption in China is
generated by the Chinese political system, which grants and protects privileges, and Cai et al.
                                                            
4
 This chapter is under revision for resubmission to the Public Choice. 

12 
 
 

(2009), using an innovative measure of corruption in Chinese firms, find that corruption
significantly reduces firm productivity. Nevertheless, no empirical study yet exists that
comprehensively analyses the economic underpinnings of corruption in China. Rather, the
majority of extant studies on the causes of corruption are cross-national investigations that
use subjective survey data. For instance, Treisman’s (2000) comprehensive cross-country
study employs several indices of perceived corruption to explore the causes of corruption.
These studies, although fruitful, are subject to the problems of subjective bias and omitted
variable bias. First, as Treisman (2007) admits, corruption perception data actually reflect
impressions of corruption intensity rather than corruption itself, meaning that the data are
actually correlated with survey respondents’ beliefs and other social and economic conditions
(see also, Knack 2006). Such data, therefore, cannot be convincingly used as dependent
variables because their measurement error is associated with many other background
characteristics that are affected by explanatory variables (Bertrand and Mullainathan 2001).
Second, the substantial number of unobservable or unmeasurable differences in institutions
and cultures between countries makes it difficult for cross-country analyses to solve the
omitted variable bias. Admittedly, some cross-country analysts have attempted to bypass this
bias by using fixed-effect regressions, however, as Treisman (2007) and Knack (2006) point
out, the appropriateness of using some subjective corruption indices in longitudinal analyses
remains questionable.
Such disadvantages in cross-national research can certainly be mitigated by the use of an
objective within-country data set that eliminates the subjective data bias and, despite some
regional differences, provides a higher level of homogeneity that moderates the omitted
variable bias to which cross-country analyses are subject. In this respect, studies of China
have a unique advantage: China is a centralized country with unified legal and administrative
systems and a fairly homogenous society dominated in most areas by Han ethnicity and
Confucian values. This high degree of legal and social homogeneity helps to efficiently
mitigate the omitted variable bias in any empirical analysis. On the other hand, as shown in
Table 2-1, great economic differences exist between China’s rich eastern and poor western
provinces. For example, in 2008, the GDP (PPP) per capita of Shanghai, which approximates
that of Hungary, was about nine times higher than that of Guizhou province, which resembles
that of Cameroon. Thus, a study of China can provide valuable insights into the causes of
corruption in developing and transitional economies in which corruption is a central issue.
Surprisingly, however, few studies on the causes of corruption employ within-country
data and most that do are working with U.S. data. For instance, Goel and Nelson (1998)
13 
 
 

investigate the effect of government size on corruption using an American annual state-level
data set, while Fisman and Gatti (2002b) use information on the mismatch between revenue
generation and expenditure in American states to test the relationship between
decentralization and corruption. More recent studies by Leeson and Sobel (2007) and Boettke
et al. (2008) show that those American states struck most frequently by natural disasters
attract more disaster relief, creating new opportunities for political corruption comparable to
resource windfalls and therefore setting in motion rent-seeking activities. Boettke et al. (2008)
stress that the Federal Emergency Management Agency relief “is especially corrosive in
terms of corruption because of the chaotic atmosphere in which it is unavoidably deployed. In
the case of a major disaster, the combination of billions of dollars of relief being dumped
onto one location in only a short period of time, along with the confused and difficult-to-
monitor environment in which these windfalls are dispensed, create incredible temptation for
public officials to abuse their positions of power by corruptly appropriating relief funds” (p.
367).
Internationally, Svensson (2003) use firm-level data from Uganda to explore the
determinants of firm bribery payments, while Cai et al. (2009) employ firm-level data to
examine the “micro” causes of corruption in China.

                       Table 2-1 GDP (PPP) per capita of Chinese regions in 2008 (Intl. $)
Beijing 16577 Anhui 3810 Chongqing 4741
Tianjin 14590 Fujian 7922 Sichuan 4044
Hebei 6112 Jiangxi 3887 Guizhou 2321
Shanxi 5365 Shandong 8701 Yunnan 3310
Inner Mongolia 8472 Henan 5153 Tibet 3646
Liaoning 8221 Hubei 5223 Shaanxi 4799
Jilin 6184 Hunan 4608 Gansu 3185
Heilongjiang 5714 Guangdong 9886 Qinghai 4573
Shanghai 19232 Guangxi 3936 Ningxia 4706
Jiangsu 10421 Hainan 4517 Xinjiang 5232
Zhejiang 11102

In this chapter, we adopt both fixed-effect and instrumental variable (IV) approaches to
identify the causes of corruption in China using different regional data sets. Besides
confirming most cross-country findings in a more controlled setting, our study makes three
important contributions to the literature. First, we identify a positive relationship between
corruption and economic development in China, one that stems from the current transition
process. Second, we obtain novel within-country evidence on the depressive effect of the

14 
 
 

Anglo-American colonial heritage, the contributory effect of abundant natural resources, and
the depressive effect of female representation in politics on corruption. Third, we find that
even in a non-democratic country, access to controlled media keeps corruption in check.
The chapter is organized as follows: Section 2.2 reviews previous research on the causes
of corruption, Section 2.3 empirically determines the causes of corruption in China, and
Section 2.4 presents our concluding remarks.

2.2 Determinants of Corruption

Previous research has identified several possible causes of corruption, including political
institutions, the judicial system and the cultural environment; however, as these factors are
homogenous among Chinese regions, we focus here on other determinants. According to Jain
(2001), there are three prerequisites for corruption: bureaucratic discretionary power, the
association of this power with economic rents, and deterrence as a function of the probability
of being caught and penalized. Whereas the first two preconditions determine the benefit of
corruption, the third influences the cost of corruption; therefore, the regional characteristics
that affect these preconditions determine its local incidence (Becker, 1968).
Bureaucratic discretionary power over the allocation of resources is particularly
important to the existence of corruption and, according to Rose-Ackerman (1978), frequently
arises during the enforcement of regulations. That is, because bureaucrats can assign
themselves the discretion to distribute resources when setting and implementing regulations,
more regulations means more discretionary power and thus more incidences of corruption. In
contrast, levels of corruption can be expected to decrease if controlled economies become
more marketized. Governmental discretionary power can also be influenced by
decentralization, although the relationship between decentralization and corruption is still
being debated. According to Brennan and Buchanan (1980) and Weingast (1995),
decentralization introduces competition between local governments, thereby reducing
bureaucratic profits from corruption. For example, the mechanism of entry and exit in U.S.
federal states provides a strong incentive to produce public services in accordance with
individual preferences (Hirschman 1970) and can be a method of government control, as
when exits threaten firms with higher mobility (Rose-Ackerman 1999). Nevertheless, because
federalism and local autonomy combine with innovation, federalism can also serve as a
laboratory for effective policy inventions (Oates 1999). On the other hand, Shleifer and
Vishny (1993) argue that since decentralization causes the dispersion of government power,
bureaucrats that are not coordinated will over-extract rents from firms. Likewise, smallness

15 
 
 

and intimacy of local jurisdictions with patronage-ridden governments promote corrupt

relationships (Rose-Ackerman 1999). In fact, Treisman (2000), using a dummy variable that
reflects whether a state is federal, finds that federal states are seen as more corrupt. Fisman
and Gatti (2002a), however, provide cross-country evidence that fiscal decentralization in
government expenditure is significantly correlated with lower corruption. Using American
data, they also identify a positive relationship between corruption and the proportion of a
state’s expenditure derived from federal transfers (Fisman and Gatti 2002b).
Obviously, rational individuals pay bribes only if they can reap a higher marginal benefit
from doing so. Hence, economic rents related to discretionary powers are a necessary
condition for corruption, but corruption is unlikely to be generated by discretionary powers
without related rents. Indeed, Ades and Di Tella (1999) show that countries in which firms
have higher rents tend to be more corrupt. One concentrated and easily expropriable activity
of particularly high rents is natural resource exploitation (Sachs and Warner 2001), which
echoes Leite and Weidmann’s (1999) empirical finding that the incidence of corruption
depends significantly on natural resource abundance. Treisman (2000), on the other hand,
finds no strong evidence that fuel and mineral exports are positively correlated with
corruption level, although, intriguingly, Leeson and Sobel (2008) report that resource
windfalls generated by disaster relief in frequently affected American states raise public
corruption in much the same manner as rich natural resource endowments. Another source of
economic rents is lack of competition: economic rents decrease when economic activities are
marked by intensive competition. For instance, Ades and Di Tella (1996, 1999), using a
country’s openness – measured by share of imports in the GDP – to indicate firms’ external
competition, find that economic openness is negatively correlated with levels of corruption.
Treisman (2000) also provides evidence of a negative association between the share of
imports in GDP and corruption levels, a relationship similar to that recently identified by
Gerring and Thacker (2005) between corruption and trade openness.
Deterrence of corruption is a joint function of the probability of detection and
punishment once caught, a probability affected by several factors. First, higher income levels
accelerate the spread of education and democratic institutions and therefore enhance
individuals’ political involvement. They consequently enable private individuals to better
identify corrupt behaviours and punish official malfeasance. Hence, regions with richer and
more educated citizens are assumedly less corrupt. In fact, according to Treisman (2007), the
negative relationship between the incidence of corruption and income level is the strongest
and most consistent finding of empirical studies on corruption (see, e.g., La Porta et al. 1999;
16 
 
 

Ades and Di Tella 1999; Treisman 2000). The probability of being caught also depends on
the effectiveness of the country’s legal system. For instance, La Porta et al. (1999) argue that
the common law systems in Britain and its former colonies are more effective in protecting
property rights and enforcement than civil law systems, which would imply that the
probabilities of corruption being exposed are higher in common law countries. Treisman
(2000) does indeed find that, as expected, Britain and its former colonies have substantially
lower levels of corruption than other countries; however, Pellegrini and Gerlagh (2008) find
no such linkage.
Economic and social heterogeneity may also be an indirect determinant of the probability
of detection and thereby affect corruption. For example, You and Khagram (2005) argue that
“the poor are more vulnerable to extortion and less able to monitor and hold the rich and
powerful accountable as income inequality increases” (p. 136). Thus, income inequality
enables the latter to abuse their power for private gain and, as the authors confirm through
cross-country analysis, promotes higher levels of corruption. Husted (1999), on the other
hand, finds no such relationship between income inequality and corruption. One social
heterogeneity factor with the potential to promote corruption is ethnic fractionalization,
which may lead to corrupt officials being protected for political reasons by their own ethnic
groups. Nevertheless, neither Treisman (2000) nor Pellegrini and Gerlagh (2007) find strong
evidence of such a linkage, although Glaeser and Saks’ (2006) results do show a positive
correlation between corruption levels and racial division in U.S. states.
Press freedom also plays an important role in corruption detection because independent
journalists have incentives to investigate its presence or absence. Therefore, as a particular
mechanism of external control, press freedom appears to reduce corruption: firms and
individuals can reveal corrupt behaviour to a journalist and this possibility of media reporting
increases the costs of corruption for bureaucrats (i.e., increases the probability of detection).
In other words, the media can be seen as a platform for voicing complaints (Brunetti and
Weder 2003), and, as Adsera et al. (2003) show, even the “free circulation of a daily
newspaper” (their interaction term between a democratic measure and newspaper circulation)
is negatively correlated with corruption (p. 455). Likewise, Brunetti and Weder (2003) show
empirically that a high level of press freedom is associated with a low incidence of corruption,
Chowdhury (2004) emphasizes that press freedom controls corruption via the channel of
democracy, and Freille et al. (2007), using a modified extreme bounds analysis, provide
evidence that the greater the press freedom, the lower the level of corruption.

17 
 
 

Several interesting papers also address the effect of gender on corruption. For example,
Dollar et al. (2001) and Swamy et al. (2001) observe that countries with a larger share of
women in parliament or government tend to have less corruption, and Swamy et al. (2001)
provide micro evidence that women are less tolerant of bribe acceptance. In fact, according to
Paternoster and Simpson (1996), women may be more risk averse and hence perceive a larger
probability of being caught in corruption. Sung (2003), on the other hand, argues that “the
observed association between gender and corruption is spurious and mainly caused by its
context, liberal democracy – a political system that promotes gender equality and better
governance” (p. 703). Torgler and Valev (2010) thus explore whether gender matters and
whether greater equality of status and opportunities reduces gender differences. Using data
from an almost 20-year period, they find evidence of strong gender differences: even when
different time periods are investigated and opportunity factors are controlled for, women are
significantly less likely to agree that corruption is justified.
The effect of punishment on corruption levels is difficult to test because punishment is
not always comparable across countries. However, higher wages imply higher opportunity
costs when officials are ousted due to corruption, and Van Rijkeghem and Weder (2001) do
in fact find that in developing countries, a higher ratio of civil service wages to
manufacturing wages is significantly correlated with a lower level of corruption. Treisman
(2000), however, finds no clear evidence that higher government wages depress corruption.
Interestingly, even in Ancient Egypt, the pharaohs searched for ways to reduce the corruption
of their tax collectors (called scribes), who were paid high salaries to reduce their incentives
to enrich themselves by cheating taxpayers. Scribes working in the field were also controlled
by a group of special scribes in a “head office” (see Adams, 1993). However, as Rose-
Ackerman (1999) stresses, “[p]ay increases may indeed be necessary for good performance,
but only if the increases are tied to productivity and are accompanied by a reduction in the
overall level of public sector employment” (p. 87).

2.3 Empirical Analysis

Administratively, China is divided into provincial regions, prefectural regions and counties or
districts. We therefore explore the causes of corruption in China using two different regional
data sets: a province-level data set comprised of information from all 31 provincial areas in
mainland China during the 1998 to 2007 period and a city-level data set taken primarily from
a 2005 World Bank and Enterprise Survey Organization of China survey on the investment
climate of Chinese prefecture-level cities (World Bank 2006).

18 
 
 

The first data set covers 22 provinces, 5 autonomous regions and 4 municipalities but
excludes data from Hongkong, Macao and Taiwan because of the substantial differences in
their political and legal systems. We derive the corruption measure for this data set – the
number of regional registered cases of corruption per 100,000 people each year – by
collecting the number of annual registered cases of corruption in the procurator’s office by
region (listed in the China Procuratorial Yearbooks) and then dividing this number by
regional population.5 To ensure comparability of the corruption data, we use only corruption
data from 1998 to 2007 because until the Fifth Session of the Eighth National People's
Congress of China passed the 1997 Criminal Law (which includes Chapter VIII: Crimes of
Embezzlement and Bribery), the definition of corruption-related crimes fluctuated. The
resulting average regional corruption data are listed in Table 2-2. We choose conviction data6
specifically because they offer a less subjective measure of corruption, enable us to work
with longer time spans, and are not subject to problems of sampling error and/or survey non-
response (Glaeser and Saks, 2006). On the other hand, the conviction rate is driven by the
quality of the detection process; this weakness, however, may not be important for our
current study because the quality of local judicial systems in China is basically homogeneous
and our regressions control for local anti-corruption efforts.
The survey from which we take our city-level data set evaluated the investment climate
of 120 cities from almost all the Chinese provinces by sampling 100 industrial firms in each
city except for four municipalities in which it sampled 200 industrial firms (World Bank
2006). As Cai et al. (2009) argue in their firm-level analysis, travel and entertainment costs
(ETC) can be an efficient measure of corruption in Chinese firms because “Chinese managers
commonly use the ETC accounting category to reimburse expenditures used to bribe
government officials, to entertain clients and suppliers, or to accommodate managerial excess”
(p. 2). They also provide strong evidence that a “firm’s ETC consists of a mix that includes

                                                            
5
 We also replicate our provincial analysis using an alternative corruption measure: the provincial number of
officials investigated in registered cases on corruption per 100,000 population. Since this corruption measure is
only available in the 2003 to 2007 period, we do not report the estimation results here; however, notably, the
results are quite similar.  
6
 Theoretically conviction rate and the number of registered cases of corruption are different. However in China
they are actually the same. In most cases in China suspect officials are first investigated by the discipline
inspection commission of the Chinese Communist Party and its local branches. Only after they have obtained
enough evidence, the discipline inspection commission and its local branches will refer corrupt cases to the
procuratorates, and procuratorates then register the cases. Furthermore the courts and the procuratorates are both
controlled by the Chinese government. Therefore in few circumstances the courts will reject public prosecutions
against corrupt cases.
 

19 
 
 

expenditures on government officials both as ‘grease money’ and ‘protection money’,

implicit CEO pay, and managerial excesses” (p. 22). We therefore use the average travel and
entertainment costs relative to the sales of investigated firms in a city (hereafter, ETC) as a
proxy for city corruption levels on the assumption that ETC derived from representative firms
in Chinese industrialized cities indirectly reflects their overall corruption levels. Statistical
details of this variable can be found in World Bank (2006); detailed descriptions of our other
variables are given in the Appendix.

Table 2-2 Average annual registered cases on corruption across regions in China (1998-2007)
Average annual Average annual Average annual
Region registered cases per Region registered cases Region registered cases per
100,000 Pop. per 100,000 Pop. 100,000 Pop.

Tianjin 5.01 Shaanxi 3.15 Yunnan 2.61

Heilongjiang 4.77 Qinghai 3.08 Hunan 2.59
Jilin 4.5 Ningxia 3.08 Hainan 2.59
Liaoning 4.12 Hubei 3.05 Beijing 2.59
Shanxi 3.83 Guizhou 2.95 Chongqing 2.49
Hebei 3.67 Zhejiang 2.9 Anhui 2.36
Shandong 3.61 Inner Mongolia 2.77 Sichuan 2.35
Xinjiang 3.41 Shanghai 2.77 Gansu 2.05
Fujian 3.4 Jiangsu 2.71 Guangdong 2.05
Henan 3.35 Guangxi 2.64 Tibet 1.77
Jiangxi 3.29

2.3.1 Province-level Analysis

To effectively compare corruption in China (the largest developing country) to that in
America (the largest developed country), our baseline specification for investigating the
causes of corruption resembles that of Glaeser and Saks (2006)7:

Registered Cases on Corruption = α + γ•Income + δ•Education + η•Anti-corruption

+ β • Other regional characteristics + Error term

Consistent with Glaeser and Saks (2006), this specification measures regional income,
expressed as the logarithm of real gross regional product (GRP) per capita, and educational
levels, the proportion of the regional population over 6 that has completed a college degree.
We also control for anti-corruption efforts because provincial registered cases of corruption
may reflect local government efforts to fight corruption (Treisman, 2007); that is, since the
law systems in different Chinese regions are the same, any differences in anti-corruption
                                                            
7
A description of the variables is presented in Table 2-7 in the Appendix.  

20 
 
 

efforts across regions stem from the individual regions’ legal enforcement. Like Goel and
Nelson (1998), we proxy regional anti-corruption efforts by the real per capita expenditure by
local government on police, procuratorate, court and judiciary. We also control for the
differences between North and South China by adding a geographic dummy (North) into the
pooled OLS regressions and whenever possible, use one-year lagged values of the
explanatory variables in our panel regressions to account for the fact that registered cases of
corruption in one year actually reflect the number of corrupt acts done previously but
detected in that year.
To identify the determinants of corruption in China, we must first address two potential
problems in our estimations: multicollinearity (see the correlation matrix in Table 2-88) and
the endogeneity problem. To alleviate the multicollinearity in our regressions, we first adopt
a basic specification that in our panel analysis controls the condition numbers and variance
inflation factors of regressions lower than 100 and 10, respectively. We therefore assume that,
as a rule of thumb, there is no serious multicollinearity in these regressions. In regressions
using the full specification in which severe multicollinearity is detected, we control
multicollinearity by dropping one or more collinear variables 9 while still retaining the
baseline specification to minimize specification bias. Admittedly, it may be argued that
dropping offending variables may produce a specification bias that is potentially worse than
multicollinearity; however, other methods for dealing with multicollinearity – for example,
model re-specification, ridge regression or data reduction techniques like principal
component analysis – are not applicable in our case. More important, our results indicate that
removing select collinear variables from the regressions has little effect on the remaining
parameter estimates, suggesting that it produces no significant specification bias and is thus a
justifiable method for mitigating multicollinearity.
To address the more important problem of endogeneity, which is key to our identification
of corruption determinants in China, we use two strategies: first, we control for unobserved
regional characteristics that influence both corruption and its determinants by including
regional fixed effects in our panel regressions and second, we adopt an IV approach
whenever necessary to identify the causal effects of the corruption determinants. As regards
the first, because Mo (2001, p.70) describes a corruption problem as “an institutional problem
that lasts for a long period” and since the major source of potential bias in our regressions
                                                            
8
Table 2-9 presents the correlation matrix for the city-level analysis.  
9
 We here retain the condition numbers of those regressions lower than 300, which, as a rule of thumb, indicates
only moderate multicollinearity. 

21 
 
 

may be time-invariant historical factors, we choose fixed-effect regressions as the most

suitable tool for investigating the relationship between corruption and its determinants. Some
corruption determinants, however, as well as several instruments used for them, are
themselves historical factors and hence time invariant, so standard fixed-effect regressions do
not work in these cases. We thus develop a unified framework for identifying the causality
between corruption and its determinants. Specifically, following Li and Hou (2003),10 we
categorize Chinese provinces into eight groups (Northeastern China, Northern Coastal Areas
of Seaboard of China, Eastern Coastal Areas of China, Southern Coastal Areas of China,
Middle Reaches of the Yellow River, Middle Reaches of the Yangtze River, Southwestern
China and Northwestern China) according to their geographic similarity, natural resource
endowments, economic development, social structure and cultural traditions. We then
perform pooled OLS regressions, which include both regional fixed effects (eight areas) and
year fixed effects, to identify the causes of corruption in China.11 Because Chinese provinces,
especially neighbouring provinces, are to a large extent homogeneous, regional fixed effects
can capture most provincial characteristics while still enabling investigation of the time-
invariant determinants of corruption in China. In fact, the consistency between the results for
our standard fixed-effect regressions and those for our regional effects measure (applied
when the explanatory variables are all time variant) validates our approach.12 Fixed-effect
regressions, however, do not necessarily identify the causal effects on corruption of its
determinants; that is, they cannot guarantee causality when omitted time-variant factors affect
both corruption and its determinants. Therefore, as discussed in more detail below, we adopt
an IV approach whenever necessary to identify the causal effects of the corruption
determinants.
We first investigate the effects of determinants of corruption in China using a baseline
specification that allows comparison with the findings of Glaeser and Saks (2006), as well as
assessment of specification robustness and the importance of alternative factors in its
extension. We then further test the specification’s robustness with joint estimates that use the

                                                            
10
 This document is actually an official research report of the Development Research Centre of the State Council,
a leading comprehensive policy research and consulting institution operating directly under the State Council of
the Central Government of the People’s Republic of China. 
11
 For convenience, we refer to this method as the fixed-effect approach throughout the chapter (although it is
not the standard fixed-effect approach).  
12 
To save space, we do not report the results of our standard fixed-effect regressions, but these findings are
available upon request. 

22 
 
 

full specification.13 It is worth noting, however, that the multicollinearity problem becomes
more serious in our 2SLS estimation (i.e., the standard errors of the estimates become large
and there is more inconsistency in the impact of some explanatory variables). Therefore, we
again gradually exclude several collinear variables from the full specification to mitigate
multicollinearity while retaining important explanatory variables to minimize specification
error. Specifically, we first remove the regulation variable which, as shown in Table 2-8, has
the highest correlation coefficient with income of all the explanatory variables. Next, we
exclude openness, which is highly correlated with several other explanatory variables, and
finally we drop education, which is also highly correlated with some other independent
variables. Even such a reduced specification, however, is still among the richest in the causes
of corruption literature.
2.3.1.1 Deterrence and Corruption
To empirically test the effect of deterrence on corruption, we first examine the effect of
income on corruption. The results for the pooled OLS, fixed-effect OLS and fixed-effect
2SLS are presented in Tables 2-3, 2-4 and 2-5, respectively14. In determining the effect of
income, to address the potential omitted variable bias and the reverse causality between
income and corruption identified by Treisman (2000) and Glaeser and Saks (2006), we must
find an instrumental variable for regional income level. Like Hall and Jones (1999) and
Rappaport and Sachs (2003), we instrument provincial income level using a geographic
characteristic, the provincial capital’s latitude (latitude), primarily because historically
China’s developed areas – namely, the middle and lower reaches of the Yellow River since
the Han Dynasty, the middle and lower reaches of the Yangtze River since the Song Dynasty,
and Northeastern provinces since the end of the Qing Dynasty – are in relatively high
latitudes. Because our results provide no evidence that our instrumental variable affects
corruption directly or through channels other than economic development, we believe that
this choice of instruments is valid. In fact, the results from different estimations and different
specifications indicate that income significantly increases corruption in Chinese provinces15
and the effect is fairly large. For example, according to the IV estimation using the baseline

                                                            
13
 We do not include the Gini coefficient income distribution variable in our full specification because the Gini
coefficients for Chinese provinces are only available before 2001, so including them would sharply reduce our
sample size. 
14
 The first-stage regressions are presented in Table 2-10 in the Appendix.  
15
 Only in two 2SLS regressions does the income variable lose its significance, but, because of multicollinearity,
it still retains its sign. 

23 
 
 

specification (Column (1) in Table 2-5), in terms of elasticity, a 1% increase in the GRP
logarithm raises regional registered cases of corruption per 100,000 people by about 11.02%.
At first, this positive effect of income on corruption seems puzzling because it contradicts
previous literature. However, such a result might be driven by the transitional nature of
Chinese society; that is, because “the simultaneous processes of developing a market
economy, designing new political and social institutions and redistributing social assets
[creates] fertile ground for corruption” (World Bank 2000, p. vii), countries making the
transition to a market economy often experience unprecedented corruption (Levin and
Satarov 2000; Paldam and Svendsen 2000). China specifically began its transitional process
when economic reform loosened up its economy; however, political reform has lagged
behind. Therefore, in the absence of institutional and legal constraints, government continues
to play an extensive role in China’s economic environment. One unavoidable consequence of
such involvement is corruption, a type of corruption that becomes more pervasive when
government power is widened through increased economic activity. As a result, regions with
higher income levels may be more corrupt.
On the other hand, Basu and Li (2000) predict theoretically that corruption levels in
transitional countries will decrease after these nations establish or improve institutions,
effectively checking the abuse of public office as the transition processes move on. Income
level would then have a depressing influence on corruption. This prediction has been
confirmed in several Eastern European countries (World Bank 2006). We therefore
conjecture that, with the establishment of good institutions, this phenomenon may diminish
once the Chinese transition is completed. To examine this conjecture, we split the
investigation period into two sub-periods – 1998 to 2002 and 2003 to 2007 – and create a
dummy variable (period) to represent the second sub-period (2003–2007). In the IV
regression using the rich specification (see Column (16) in Table 2-5), the interaction term
between income level and the period dummy has a significantly negative effect on corruption,
which indicates that the positive effect of income on corruption during the 2003–2007 period
is substantially lower than that during the 1998–2002 period. This decrease in the positive
effect of income on corruption in China may imply that a negative effect of income on
corruption may emerge in the future.
To further determine the direction of the causality between income and corruption, we
perform a Hausman test. Like Treisman (2007) and Gundlach and Paldam (2009), we find
that the main direction in the circular relationship between income and corruption in China is
from income to corruption, and, since the Hausman tests (see Columns (1) and (13) in Table
24 
 
 

2-5) show that the IV estimates do not differ systematically from the corresponding OLS
estimates, there is no significant upward bias caused by reverse causality. This result actually
supports our explanation of the positive relationship between income and corruption.
As the tables show, we also investigate the effect of education on corruption and, contrary
to the literature, find that education has a seemingly positive effect on corruption in both the
pooled OLS and fixed-effect regressions. We then instrument provincial education attainment
by the middle-school student enrolment per 100 persons in 1952 (middle1952), which is
certainly exogenous and directly affects current education level. Our IV results (see Table 2-5)
provide moderate evidence that education lowers corruption in Chinese provinces once
income, anticorruption efforts and other variables are controlled for.16 In effect, according to
Column (1) in Table 2-5, a 1% increase in the proportion of the population over 6 that has
completed college is associated with a 0.39% decrease in registered cases of corruption per
100,000 population, a finding that is both reasonable and consistent with the literature (e.g.,
Glaeser and Saks 2006).
Interestingly, as shown in Tables 2-3, 2-4 and 2-5, our proxy for provincial anti-
corruption efforts is, as expected, negatively correlated with corruption, in most cases,
significantly and robustly. This finding suggests that we may have effectively isolated the
anti-corruption effort component from our corruption measure. It is also worth noting that our
proxy is most probably exogenous because in China, public security organs, procuratorial
organs and people’s provincial courts are all directly under central government control.
Hence, because China is typically a centralized country, the budgetary expenditures on these
organs must be approved by the central government even though most are paid out by the
provincial governments. As a result, despite obvious differences across provinces, provincial
expenditures on police, procuratorate, court and judiciary per capita are mainly exogenously
decided by the central government.
In terms of the causal effect of media on corruption, mainstream media in China,
including newspaper, radio and television, are fully controlled by the state, raising the
question of whether media can be expected to act effectively as an external control of
corruption in China (cf. the importance of a free press in Brunetti and Weder 2003).
Following Treisman (2007) and Pellegrini and Gerlagh (2008), therefore, we investigate the
effects of controlled media on corruption in China using the regional newspaper circulation
per capita as a measure. Our results, given in Tables 2-3, 2-4 and 2-5, seem encouraging. We
                                                            
16
 In several IV regressions, education loses its significance and even its sign (see Table 2-5) due primarily to
the high correlation between education and certain other explanatory variables. 

25 
 
 

find consistent evidence in both the pooled OLS, fixed-effect and IV regressions that media
in China significantly depress corruption, although the media variable does lose significance
in the IV regression using the full specification, which is not free of multicollinearity issues.
However, if we drop the variables that are highly correlated with income – namely,
marketization (deregulation), education and trade openness – the media exerts a significantly
negative effect on corruption (Column (15) in Table 2-5). As before, the removal of collinear
variables from Regression (15) has little impact on the remaining parameter estimates,
suggesting the absence of any significant specification bias. We hence conclude that media in
China, even if controlled by the Chinese government, do exert negative influences on
corruption. Thus, in contrast to Adsera et al. (2003) and Chowdhury’s (2004)  emphasis on the
key role of democracy in media’s effect on corruption, our findings indicate that even
government-controlled media can effectively depress corruption. Our results also suggest that
anti-corruption efforts, as long as they are held within limits, can also be in the interest of
non-democratic governments. In terms of media size and its relation to corruption, according
to Regression (2) in Table 2-5, a 1% increase in annual newspaper circulation per capita
decreases corruption by about 0.04%.
Before exploring the influence of Western colonialism, and specifically the Anglo-
American church university heritage, on corruption in China, a brief historical overview is
warranted of China’s modern history. After China was beaten in a sequence of wars against
foreign colonial powers during the last period of the Qing Dynasty (1840–1911), these
powers, headed by Britain, carved spheres of influence in China through a series of unequal
treaties. Thus, based on their own values, they influenced or even reconstructed local
administration systems, including the legal system, infrastructure and education, within the
domains they controlled. The persistence of these influences may be responsible for some of
the regional differences in current China. Education particularly, as instituted by the British
and Americans in China, may be a key point in this historical influence. For example, in 1890,
F.L. Hawks Pott, president of Saint John’s University, which was founded by the American
Episcopal Church, stated that “in our school, we trained China's future teachers and
propagators, making them the leaders and comperes in the future and casting the greatest
influences on the future China” (Yang and Ye 1993, p. 289). Likewise, Soochow University,
founded by the American Methodist Missionaries in 1900, was renowned in China for its
education in Anglo-American law. We therefore proxy this historical influence by the number
of universities founded by British and American churches in each province before 1922
(Anglo-American). According to the 1918–1921 general survey by the Special Committee on
26 
 
 

Survey and Occupation’s China Continuation Committee,17 there were 14 such universities in
China before 1922, distributed among Beijing, Shanghai, Shandong Jiangsu, Fujian, Hubei,
Hunan, Guangdong and Sichuan provinces. Although forcibly incorporated into public
universities in 1952 by the Chinese government, their influence does indeed persist through
their faculty members and graduates.
All baseline specification results in 2-3, 2-4 and 2-5 show that, in line with Treisman’s
(2000) cross-country findings for British influence, provinces influenced by the Anglo-
American church university heritage have lower levels of corruption than other provinces in
China. Specifically, according to Regression (3) in Table 2-5, such provinces have 0.16 point
fewer registered cases of corruption per 100,000 people on average. However, the Anglo-
American variable, although it retains its sign, loses its significance in the fixed-effect and IV
regressions using the full specification. Besides the methodological issues discussed earlier,
one plausible interpretation for such a result is that the Anglo-American church university
heritage has such a wide influence on various aspects of Chinese provinces that in addition to
its direct influence on corruption, it also affects corruption indirectly through other corruption
determinants. Whichever the case, this variable does indeed affect corruption substantially.
We therefore provide the first within-country evidence for the effect of this Anglo-American
educational influence on corruption.
We then investigate the role of social heterogeneity in Chinese corruption. First, as shown
in the fourth columns of Tables 2-3 and 2-4, we find that higher income inequality measured
by Gini coefficients (Meng et al., 2005) significantly raises the incidence of corruption in
China. This finding is clearly supported by the results for our IV approach, reported in
Column (4) of Table 2-5, which addresses potential reverse causation by using as its
instrumental variable a lagged Gini coefficient (gini86) that represents China’s 1986
initiation of comprehensive economic reform 18 . Overall, our results indicate that a 1%
increase in Gini coefficient increases provincial corruption by about 1.62%. However, as
previously mentioned, we do not test the relationship between income inequality and
corruption in the full specification because of the sharp decrease in sample size. 19
Nevertheless, the results provide primary evidence that income equality substantially
increases corruption in China, which is in line with prior cross-country findings.
                                                            
17
 The China Continuation Committee (2007) originally published its general survey of the numerical strength
and geographic distribution of Christian forces in China in Shanghai in 1922. 
18
  Li et al. (2000) have provided both theoretical and empirical evidence that corruption substantially affects
income distribution in a cross-country context. 
19
 Gini coefficient data before 2001 are only available for 30 provinces other than Tibet.  

27 
 
 

Secondly, we test the impact of racial fractionalization on corruption in China, which is

home to 56 recognized ethnic groups. 20 Given this diversity, even though Han ethnicity
accounted for 91.6% of China’s total population in the 2000 national census, we expect racial
fractionalization to have a significant influence on corruption. Accordingly, following Glaser
and Saks (2005), we measure the ethnic fractionalization in China using a dissimilarity index,
calculated as one minus a Herfindahl concentration index, 1-∑si2, where si is the population
share of ethnicity i from the national census in 2000. The fixed effects result in Column (8)
supports our prediction: the dissimilarity index is positively associated with local corruption
level, indicating that ethnic fractionalization significantly promotes corruption in Chinese
provinces. Since the ethnic composition of the population in Chinese provinces has emerged
historically, the provincial dissimilarity index is likely to be exogenous, meaning that our
fixed effects result captures the causal effect of ethnic fractionalization without any potential
endogeneity bias. Nevertheless, the effect, although significant, seems quantitatively small in
terms of elasticity: a 1% increase in the dissimilarity index raises corruption by only 0.11%,
which may reflect the fact that the Han ethnic majority predominates in the majority of
Chinese provinces. We therefore explain this finding of ethnic fractionalization’s
contributory effect on corruption to two factors not addressed in prevailing interpretations.
On the one hand, cadres belonging to ethnic minorities are selected mainly based on their
loyalty to the country and ability to control their own ethnic groups rather than their
incorruptibility. On the other, corrupt officials may be supported by their own ethnic groups
if they can defend group interests. This type of problem prevails especially in the five
autonomous regions and at least three provinces – Yunnan, Guizhou and Qinghai – in which
the Han nationality does not predominate.21
As regards the effect of the public sector’s relative wage, which can be seen as a proxy
for public officials’ opportunity cost for corrupt behaviour, official’s standard wages in the
Chinese provinces sampled here are set by the central government based on inter-provincial
differences in living standards and work conditions.. Therefore, our relative wage variable is
basically exogenous. Based on the results given in Tables 2-3, 2-4 and 2-5, our evidence for
the previous theoretical prediction that the relative wage is negatively correlated with the
incidence of corruption is weak. In terms of magnitude, however, as shown in columns (7)
                                                            
20
In addition to the Han ethnic majority, 55 ethnic groups in China are recognized as ethnic minorities. Of these,
minority groups, which like the Han majority can be found throughout the country, 18 had a population of over
one million according to the 2000 national census.
21
 Other Chinese provinces such as Sichuan, Hunan, Hubei, Gansu, Jilin, Guangdong and Heilongjiang also
contain autonomous areas of ethnic minorities and therefore to some extent experience the same issue.  

28 
 
 

and (13) of Table 2-5, a 1% increase in the public sector relative wage decreases corruption
by about 0.29%. It may be that in addition to their standard wage, Chinese officials receive
various subsidies from the government offices that employ them, subsidies that, being a type
of grey income, are not made public. Our relative wage ratios would therefore only partially
reflect the officials’ opportunity cost for corrupt behaviour and consequently would have a
weaker effect on corruption than the actual relative wage ratios in Chinese provinces.
Next, following Dollar et al. (2001), we investigate the influence on corruption of the
representation of women in the National People’s Congress (NPC), the only legislative house
in China. Using the share of female NPC members in each province as the main explanatory
variable, we find robust evidence that, consistent with the literature, Chinese provinces with a
greater representation of women in the legislature tend to have lower levels of corruption. In
terms of elasticity, a 1% increase in the representation of women in the NPC reduces
corruption by about 0.22% (Column (6) of Table 2-5). Since China’s centralized structure
means little difference in liberal democracy across Chinese provinces, this finding may
provide within-country evidence of a gender effect on corruption that is not driven by the
concerns raised by Sung (2003, p. 703) discussed beforehand. Not only are differences, if any,
controlled for by the inclusion of regional fixed effects, but the social status of women, the
main determinant of female representation in the NPC, has been established historically in
China. Therefore, our fixed-effect regressions, by controlling for economic development,
educational attainment and regional fixed effects, can identify the causal effect of gender on
corruption without any apparent time-variant factor influencing either corruption or female
representation in the NPC. In any case, corrupt officials do not necessarily discriminate
against women.
2.3.1.2 Discretionary Power, Economic Rents and Corruption
Lord Acton, the British 19th century historian, remarked that “Power tends to corrupt, and
absolute power corrupts absolutely.”22 In this subsection, therefore, we explore the influence
of discretionary power and related economic rents on corruption in China. First, to examine
the effect of fiscal decentralization on regional corruption levels, we measure fiscal
decentralization in China using the ratio of per capita provincial consolidated spending to per
capita central consolidated spending (Zhang and Zou 1998).23 As Tables 2-3, 2-4 and 2-5
show, like Fisman and Gatti (2002a, b), we provide robust evidence that fiscal

                                                            
22
 Lord Acton, 1887, Letter to Bishop Mandell Creighton. 
23
 For a detailed description of public finance in Chinese provinces, please refer to Zhang and Zou (1998).  

29 
 
 

decentralization significantly decreases provincial corruption levels in China. In fact,

according to Regression 8 in Table 2-5, a 1% increase in fiscal decentralization reduces
corruption in Chinese provinces by about 0.05%. This finding is seemingly in line with
Montinola et al.’s (1995) contention that China has a unique market-preserving federal
system in which local governments have strong fiscal incentives to support local economic
development and hence restrict their own predation on local enterprises. It may as easily be
explained, however, by Bai et al.’s (2008) argument that “fiscal decentralization, one of the
centrepieces of reform policies introduced since 1979, has unleashed strong incentives on the
part of the local governments to pursue economic development. However, without proper
mechanisms to redistribute the gains from specialization and trade across regions, the same
incentives for economic development may lead to protectionist policies favouring the local
firms and industries” (p. 318). The empirical evidence of Jin et al. (2005) and Li and Zhou
(2005) also imply that local officials in China implement the protectionist policies mainly for
local business development rather than rent seeking. It is therefore plausible that fiscal
decentralization in China generally depresses local corruption. Shleifer and Vishny’s (1993)
observation that decentralization results in “excess” rent extraction because of a lack of
coordination among officials may not apply in China.
To measure the effect of government regulation on local corruption levels, we use the
relationship between the market and the government, one of the sub-indices (regulation) of
Fan et al.’s (2010) Marketilization Index of China, constructed to measure the degree of
government regulation in Chinese provinces. In effect, this index covers five main aspects of
Chinese marketization: the relationship between the market and the government; the growth
of the non-state economy; the development of the product market; the development of the
factor market; and the market environment, including intermediaries and institutional and
jurisdictional arrangements. Based on Fan et al.’s (2010) use of the sub-index on
market/government relationship to measure provincial progress on deregulation, we assume
this sub-index to be negatively correlated with the degree of government regulation in
Chinese provinces. In fact, both the pooled OLS (Table 2-3) and fixed effects (Table 2-4)
evidence shows that the (de)regulation variable is negatively associated with corruption in
Chinese provinces.
We then employ the IV approach to deal with potential endogeneity problems. Because
provinces with a larger proportion of the state-owned sector always have more regulations
and tend to make less progress on deregulation, we use the provincial employment share of
the state-owned sector in 1978 (just prior to Chinese economic reform) to instrument the
30 
 
 

regulation variable. Our IV regressions provide strong evidence that deregulation reduces
provincial corruption. Specifically, as shown in Column (9) of Table 2-5, a 1% rise in the
marketilization index decreases corruption by approximately 2.72%.
To assess the role of economic rents, we test the associations between corruption level
and several factors that influence the scale of economic rents in China. The first, abundance
of natural resources, is supposed to be an inducement to corruption because of the associated
economic rents. However, since the export share of natural resources used by Ades and Di
Tella (1999) is unsuitable for measuring the regional abundance of natural resource in a
within-country analysis, we follow Xu and Wang (2006) and use the fraction of employment
in the mining and quarrying sector as a proxy for natural resource abundance in Chinese
provinces. As Tables 2-3, 2-4 and 2-5 show, like previous studies, we provide robust
evidence that the abundance of natural resources significantly increases corruption in Chinese
provinces. In Regression (9) in Table 2-5, particularly, a 1% increase in the fraction of
employment in the mining and quarrying sector is associated with about a 0.11% rise in
provincial corruption.
We then investigate the effect of trade openness, measured by the share of imports in
GRP, on corruption levels in Chinese provinces. As the fixed-effect regression results in
Table 2-4 show, trade openness significantly lowers the provincial corruption level.
Moreover, because regional fixed effects capture economic development, educational
attainment and cultural traditions, there is apparently no unobserved factor influencing
corruption and population simultaneously and the provincial population is exogenous. 24
Nevertheless, to address this potential simultaneity bias (Treisman 2000), we also perform IV
regressions that use population as the instrumental variable for openness (see Table 2-5). A
large population means a large market size and is therefore valued by foreign businesses. All
else being equal, import may thus be positively correlated with population size. Our IV
finding clearly supports the fixed effects result: openness substantially reduces corruption in
China. In fact, according to the instrumental variables regression using the baseline
specification (see Table 2-5), a 1% increase in trade openness reduces regional corruption by
nearly 0.35%.

                                                            
24
 We note that the one-child policy, which significantly influences regional populations in China, differs across
provinces. However, this inter-provincial policy difference is exogenously determined by the central
government and is mostly captured by our regional fixed-effect regressions. 

31 
 
 

Table 2-3 Corruption and its determinants in China: pooled OLS estimation
Annual Cases (1998-2007)
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
Income 0.61*** 0.54*** 0.75*** 0.98*** 0.57*** 0.55*** 0.59*** 0.60*** 1.00*** 0.64*** 0.62*** 1.02***
(0.11) (0.11) (0.16) (0.20) (0.13) (0.13) (0.11) (0.11) (0.16) (0.11) (0.11) (0.19)
Education 0.045** 0.092*** 0.049*** 0.019 0.046** 0.047** 0.062*** 0.046** 0.074*** 0.050*** 0.049*** 0.13***
(0.018) (0.021) (0.018) (0.060) (0.019) (0.018) (0.018) (0.018) (0.022) (0.017) (0.019) (0.025)
Anticorruption -0.0070*** -0.0071*** -0.0076*** -0.0077* -0.0069*** -0.0069*** -0.0072*** -0.0067*** -0.0095*** -0.0065*** -0.0069*** -0.010***
(0.00073) (0.00069) (0.00088) (0.0043) (0.00072) (0.00075) (0.00076) (0.00080) (0.0012) (0.00075) (0.00080) (0.0013)
North 0.72*** 0.69*** 0.60*** 0.80*** 0.72*** 0.75*** 0.64*** 0.72*** 0.46*** 0.49*** 0.71*** 0.11
(0.097) (0.096) (0.11) (0.15) (0.098) (0.11) (0.098) (0.097) (0.096) (0.13) (0.10) (0.15)
Media -0.0023*** -0.0022**
(0.00059) (0.00086)
Anglo-American -0.15* -0.16**
(0.077) (0.079)
Gini 7.62***
(2.55)
Dissimilarity -0.16 -0.28
(0.20) (0.23)
Wage 0.43 0.41
(0.39) (0.34)
Female -4.27*** -2.86**
(1.30) (1.40)
Decentralization -0.015 -0.029***
(0.010) (0.0046)
Regulation -0.14*** -0.19***
(0.030) (0.033)
Resource 0.050** 0.045**
(0.020) (0.019)
Openness -0.0013 0.011**
(0.0036) (0.0043)
Constant -2.30** -1.80* -3.46*** -7.29*** -1.92* -2.32** -1.29 -2.22** -4.77*** -2.81*** -2.40** -4.49***
(0.98) (0.96) (1.32) (1.89) (1.13) (0.99) (0.91) (0.98) (1.24) (0.98) (0.96) (1.45)
R-squared 0.33 0.35 0.34 0.37 0.33 0.33 0.36 0.33 0.37 0.35 0.33 0.46
Observations 310 308 310 120 310 310 310 310 307 310 310 305
Notes: a. Robust standard errors in parentheses; ***, ** and * denote significance at 1%, 5% and 10%, respectively. b. Unless otherwise noted, hereafter we use
the one-year lagged values of the explanatory variables in the panel regressions.

32 
 
 

Table 2-4 Corruption and its determinants in China: fixed effects OLS estimation
Annual Cases (1998-2007)
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
Income 0.56** 0.55*** 0.69*** 0.68* 0.64*** 0.56*** 0.39** 0.60*** 0.62*** 0.59*** 0.73*** 0.76***
(0.22) (0.20) (0.24) (0.35) (0.21) (0.21) (0.18) (0.22) (0.21) (0.21) (0.22) (0.17)
Education 0.029 0.10*** 0.029 -0.062 0.0073 0.027 0.046*** 0.029 0.057** 0.023 0.065*** 0.088***
(0.020) (0.017) (0.019) (0.057) (0.021) (0.020) (0.017) (0.020) (0.023) (0.019) (0.019) (0.023)
Anticorruption -0.0053*** -0.0044*** -0.0056*** -0.0011 -0.0047*** -0.0052*** -0.0048*** -0.0049*** -0.0070*** -0.0043*** -0.0029*** -0.0030**
(0.00081) (0.00061) (0.00080) (0.0047) (0.00079) (0.00079) (0.00084) (0.00081) (0.0011) (0.00081) (0.00097) (0.0014)
Newspaper -0.0048*** -0.0017**
(0.00063) (0.00081)
Anglo-American -0.21*** -0.066
(0.078) (0.073)
Gini 7.12**
(3.57)
Dissimilarity 0.82*** 0.54**
(0.23) (0.26)
Wage -0.54* -0.43
(0.32) (0.32)
Female -5.66*** -4.66***
(1.32) (1.29)
Decentralization -0.019*** -0.025***
(0.0058) (0.0047)
Regulation -0.12*** -0.090**
(0.035) (0.039)
Resource 0.054*** 0.055***
(0.013) (0.013)
Openness -0.020*** -0.0097**
(0.0037) (0.0048)
Constant -2.17 -2.62 -3.36 -4.51 -3.29* -1.58 0.28 -2.51 -1.96 -2.92 -4.34** -3.21**
(1.94) (1.79) (2.10) (2.94) (1.93) (1.95) (1.64) (1.95) (1.82) (1.89) (2.01) (1.58)
R-squared 0.56 0.62 0.57 0.60 0.57 0.56 0.60 0.56 0.57 0.58 0.58 0.68
Observations 310 308 310 120 310 310 310 310 307 310 310 305
Note: Robust standard errors in parentheses; ***, ** and * denote significance at 1%, 5% and 10%, respectively.

33 
 
 

Table 2-5 Corruption and its determinants in China: fixed effects 2SLS estimation
Annual Cases (1998-2007)
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15) (16)
Income 3.72*** 2.50** 4.14*** 0.77 3.50*** 3.11*** 4.18*** 3.44*** 3.14* 2.29*** 3.52*** 0.10 1.62** 1.11* 0.85*** 2.56**
(1.02) (1.07) (1.07) (1.02) (0.98) (1.04) (1.16) (0.91) (1.78) (0.71) (0.99) (0.58) (0.74) (0.65) (0.31) (1.03)
Income*period -1.53**
(0.61)
Education -0.22*** -0.083 -0.21*** -0.27* -0.22*** -0.17** -0.26*** -0.20*** 0.18* -0.15*** -0.013 0.17** -0.039 -0.031 -0.12
(0.075) (0.090) (0.071) (0.15) (0.074) (0.074) (0.085) (0.065) (0.10) (0.055) (0.068) (0.072) (0.066) (0.059) (0.11)
Anticorruption -0.0046*** -0.0044*** -0.0063*** 0.013 -0.0040*** -0.0046*** -0.0046** -0.0040** -0.024*** -0.0025* 0.0042 -0.0096 0.0029* -0.0017* -0.0019** 0.0050*
(0.0018) (0.0011) (0.0017) (0.010) (0.0015) (0.0015) (0.0019) (0.0016) (0.0088) (0.0014) (0.0030) (0.0060) (0.0017) (0.00097) (0.00083) (0.0028)
Media -0.0029*** -0.0044 0.0023 -0.00066 -0.0011** -0.00054
(0.0011) (0.0029) (0.0014) (0.00099) (0.00054) (0.0022)
Anglo-American -0.48*** -0.11 -0.031 -0.10 -0.084 -0.046
(0.13) (0.12) (0.084) (0.082) (0.071) (0.11)
Gini 19.3***
(7.08)
Dissimilarity 1.81*** -0.0087 1.23*** 1.12*** 0.99*** 1.57***
(0.49) (0.53) (0.39) (0.34) (0.21) (0.56)
Female -3.04* -2.61 -7.19*** -5.02*** -4.91*** -5.51***
(1.59) (2.79) (1.68) (1.29) (1.24) (1.73)
Wage -0.78 -0.10 -0.78** -0.59* -0.53* 0.13
(0.64) (0.39) (0.34) (0.33) (0.30) (0.61)
Decentralization -0.034* -0.029*** -0.031*** -0.023*** -0.021*** -0.044*
(0.019) (0.0084) (0.011) (0.0070) (0.0046) (0.025)
Regulation -1.25*** -0.33**
(0.44) (0.15)
Resource 0.066*** 0.037* 0.081*** 0.067*** 0.063*** 0.098***
(0.015) (0.021) (0.016) (0.015) (0.012) (0.021)
Openness -0.075*** 0.018 -0.034*** -0.012
(0.018) (0.025) (0.0093) (0.013)
Constant -30.1*** -19.7** -33.8*** -8.20 -28.9*** -24.2*** -33.3*** -27.6*** -17.0 -18.0*** -30.5*** 4.33 -11.4* -6.60 -4.34 -6.94
(9.05) (9.36) (9.47) (8.03) (8.77) (9.38) (10.0) (8.00) (13.4) (6.35) (8.96) (5.54) (6.62) (5.78) (2.89) (8.90)
Hausman test 16.37 1.59
[0.63] [0.99]
Observations 310 308 310 120 310 310 310 310 307 310 310 305 308 308 308 308
Note: Robust standard errors in parentheses; p-values in brackets; ***, ** and * denote significance at 1%, 5% and 10%, respectively.

34 
 
 

Overall, the empirical findings of our detailed evaluation of a substantial set of

corruption determinants in Chinese provinces are robust to a variety of specifications. One
important question does remain, however: whether our findings remain robust when applied
to a different data set. To answer this question, we replace our province-level data set with a
city-level data set.
2.3.2 City-level Analysis
To check the robustness of the above findings, we replicate the analysis using a city-level
data set and a different corruption measure, the average ratio of travel and entertainment costs
to the sales of firms in Chinese cities (ETC), which prior literature (World Bank 2006; Cai et
al. 2009) shows to be an appropriate indirect measure for corruption in Chinese cities. Such a
city-level data analysis, however, does have certain limitations. First, survey data can be
noisy and may make our estimation less precise. Second, because the 2005 World Bank
survey that originally measured ETC focuses on manufacturing enterprises, our corruption
measure may not be as comprehensive as the measure in our provincial analysis. Third, fewer
data are available at the city level, which reduces our ability to replicate all the previous
findings.
As in the province-level analysis, our specification is

ETC = α + γ•log (Income) + δ•Education + η•Anti-corruption

+ β•Other city characteristics + Error term

where the income level of a city is measured by its logarithm of GRP per capita in 2003.
Because of the unavailability of direct data, for general education level, we use a city’s local
library collections per capita in 2003. To proxy local anti-corruption efforts, because
impartial and efficient courts effectively prevent expropriation and significantly promote anti-
corruption, we use the average confidence in the courts of the local firms investigated in the
World Bank survey. We then include in our specification all the other determinants of
corruption confirmed in the provincial analysis that are available at a city level. As Table 2-9
shows, the city-level analysis raises little concern about multicollinearity.
Because city-level data on newspaper circulation is unavailable, we use the Internet
terminals per capita in a city in 2003 to measure the effect of media on corruption in Chinese
cities. This choice is particularly appropriate given the many recent cases of Internet users in
China revealing official malfeasance and corruption to force officials out of office (Freedom
House, 2009) despite government censorship of the network. As in the provincial analysis,

35 
 
 

we use the number of church universities founded by Britain and America in each city before
1922 to proxy for the Anglo-American historical influence. However, our sample cities are
not at the same administrative level. There are 4 municipalities and 5 cities specifically
designated in the state plan 25 in our sample which share their revenues and expenditures
directly with the central government, while other cities are regular prefecture-level ones
which share their revenues and expenditures mainly with provincial governments. Rather
than using the ratio of per capita provincial consolidated spending to per capita central
consolidated spending to measure fiscal decentralization as in the provincial analysis, like
Fisman and Gatti (2002b), we measure fiscal decentralization in Chinese cities by the
mismatch between government budgetary revenue and budgetary expenditure, which
indicates the dependence of city-level finance on higher-level governments. Since private
enterprises can only grow quickly in cities with less regulation, we proxy government
deregulation in Chinese cities by the employment shares of the private sector in these cities in
2003. To measure their trade openness, we use the share of imports in GRP. Finally, we
include the dummy north to indicate cities in North China.
As in the provincial analysis, to identify corruption determinants at the city level, we
obtain primary results using OLS regressions and then apply an IV approach to address
potential endogeneity problems. We here still use latitude to instrument the economic
development in Chinese cities. We cannot, however, use city population as an instrument for
openness because foreign businesses in China always target province-level markets, meaning
that the market size of cities cannot be a determinant of their imports. Rather, because special
economic zones26 have more import autonomy (which reduces import cost), we instrument
trade openness in Chinese cities using a dummy variable indicating special economic zones
(econzone), with which trade openness is of course positively correlated. Because data
limitations make it difficult to find valid instruments for education and regulation at the city
level, we lag these variables by two years to minimize the endogeneity bias.
As Table 2-6 shows, in line with our provincial findings, economic development
significantly increases corruption in Chinese cities, whereas fiscal decentralization
substantially lowers it. Anti-corruption efforts also strongly deter corruption in Chinese cities,
which implies that our regressions effectively control for these efforts. Trade openness,
                                                            
25
 A city specifically designated in the state plan is a prefecture-level city that is ruled by a province, but enjoys
provincial-level economic authority since it is under direct economic guidance of the central government. The
mayor of the city is equal in status to a vice-governor of a province. 
26
 There are actually five special economic zones in China: Shenzhen, Zhuhai, Xiamen, Shantou and Hainan
capital of which is Haikou. 

36 
 
 

deregulation, media and the Anglo-American church university heritage, however, perhaps
because of data noise, have only insignificantly negative effects on local corruption.
Overall, the evidence generated using city-level data supports the findings obtained with
province-level data, which thus enables a comprehensive profile of the causes of corruption
in China.

Table 2-6 Corruption and its determinants in Chinese cities 

ETC (2005)
OLS 2SLS
(1) (2)
Income -0.023 0.93**
(0.10) (0.48)
Education 0.0012 0.0000066
(0.0014) (0.0021)
Anticorruption -0.012*** -0.013***
(0.0024) (0.0033)
Media -0.0020 -0.12
(0.036) (0.085)
Anglo-American -0.0093 -0.057
(0.14) (0.19)
Decentralization 0.0024 0.024***
(0.0032) (0.011)
Regulation -0.043 -0.37
(0.33) (0.50)
Openness -0.0025** -0.0016
(0.0013) (0.0059)
North -0.11 -0.24
(0.087) (0.15)
Constant 2.07* -7.36
(1.05) (4.69)
R-squared 0.28
First Stage Regressions
Instruments
Latitude 0.028**
(0.011)
Econzone 48.05**
(18.59)
F test of excluded instruments
Latitude 4.97
[0.01]
Econzone 4.85
[0.01]
Anderson canon. corr. LM statistic 9.52
[0.00]
Observations 118 118
Notes: a. Standard errors in parentheses; p-values in brackets; ***, ** and * denote
significance at 1%, 5% and 10%, respectively. b. We adopt values in 2003 or 2004 for the
explanatory variables (i.e., lagged by a year or two because of data availability).

37 
 
  

2.4 Conclusion

A comprehensive case study of a representative country may prove a helpful supplement to

cross-country analyses on corruption using subjective survey data because these latter suffer
from a number of biases. In this chapter, therefore, we present a detailed investigation of the
causes of corruption in China, which uses two different sets of objective data across various
Chinese regions. Because almost no extant studies on the causes of corruption use within-
country data and those that do employ U.S. data, our study complements previous cross-
country research by isolating possible cultural and institutional differences within a specific
country. It also complements U.S. studies such as Glaeser and Saks (2006) by using panel
data, addressing omitted variable biases with an IV approach, and exploring a larger set of
independent factors. Our use of two different data sets and different measures of corruption
also allows a robust investigation of the causes of corruption.

Figure 2-1 Determinants of corruption in China

+
Income
—
Fiscal decentralization
—
Education

— +
Regulations
Anglo-American
tradition
Benefit
Cost

+
Social heterogeneity Corruption
  —
Openness
—
Relative wage

— +
Media Resource abundance

—
Female representation in legislature

As shown in Figure 2-1, we find that with the exception of the positive relationship
between income and corruption, the Chinese case generally retains the patterns of corruption
identified in cross-country analyses. Nevertheless, we also find evidence that in China, the
negative effect of income on corruption frequently suggested by literature may be temporarily
overwhelmed during the transition process and may only perhaps emerge after the transition
38 
 
 

is completed. Our results also show that, in line with Treisman (2007) and Gundlach and
Paldam (2009), the long-run causation between income and corruption runs mainly from
income to corruption. We also obtain clear evidence that, taking into account regional income
levels, educational attainment reduces corruption in Chinese provinces, which is also
significantly depressed by anti-corruption efforts. One novel result is the relatively strong
evidence that media, even under government control, does indeed act as a control instrument
for corruption in China. Another rather unique finding is that a greater representation of
women in the legislature depresses corruption in Chinese regions. As with prior cross-country
evidence on the influence of the British colonial tradition, we find that the Anglo-American
church university heritage adversely affects corruption. Our findings also support the
hypothesis that various social heterogeneities breed corrupt practices; however, we find only
weak evidence to support the theory that relatively high relative wages within the public
sector deter officials from corruption. We also provide concrete evidence that regions with a
higher degree of deregulation appear to be less corrupt in transitional China and that fiscal
decentralization tends to discourage local corruption (cf. Fisman and Gatti 2002a, b). Finally,
we provide clear evidence that openness substantially suppresses the incidence of corruption,
whereas abundance of resources acts as a breeding ground for corruption.
Our empirical findings have significant policy implications. First, during China’s
transitional process, economic development appears to increase corruption. However, given
that economic development substantially accelerates the transitional process, it should be
encouraged. Economic development coulc depress corruption after the transition is completed.
At the same time, because the negative effects of trade openness, deregulation and fiscal
decentralization on corruption suggest that competition, either between enterprises or
governments, plays an important role in corruption control, policies that induce more
competition in either commercial or political markets could help to reduce corruption. Much
attention should also be paid to anti-corruption efforts in provinces rich in resources because
such abundance provides a breeding ground for corrupt practices. The fight against
corruption should also take into account important historical influences. Even in China,
media, although government controlled, can act as a watchdog over corruption; however,
given Brunetti and Weder’s (2003) observation that a freer press controls corruption more
effectively, more press freedom would be beneficial. Additionally, although high salary
seemingly deters officials from corruption in China, as Rose-Ackerman (1999) argues, high
public sector salaries can be justified only if sector productivity increases while its size
decreases. Otherwise, high pay in the public sector, although it may decrease individual acts
39 
 
 

of corruption, is itself a form of corruption. In general, our results imply that a more educated
population and a society characterized by more income equality, gender equality and racial
equality equates with lower levels of corruption. Finally, historical influences should be
considered when fighting against corruption as they may have substantial effects on the legal
system and the moral norms (Treisman, 2000).

40 
 
  
Appendix
Table 2-7 Data description
Variable Description Source Mean S. D.
Cases Provincial registered cases on corruption in China Procuratorial Yearbooks 3.09 0.94
procurator’s office per 100,000 population
ETC Average travel and entertainment costs relative World Bank (2006) 1.13 0.45
to sales of industrial firms investigated in a city
Anti-corruption Regional expenditure on police, procuratorate, China Statistical Yearbooks 112.43 103.41
court and judiciary per capita World Bank (2006)
Confidence in courts of firms investigated in a city 63.75 16.81
Income Logarithm of per capita real GRP — province China Statistical Yearbooks 9.15 0.63
— city China City Statistical Yearbook 9.47 0.66
Education Fraction of the population over 6 with college completed China Statistical Yearbooks 5.44 4.31
Library collections per capita China City Statistical Yearbook 46.57 56.61
Wage Ratio of average government employee’ wage China Statistical Yearbooks 1.13 0.13
to the regional average wage
Openness Ratio of import to gross regional product Provincial Statistical Yearbooks 14.45 22.62
China City Statistical Yearbook 21.57 43.79
Newspaper Annual newspapers circulation per capita Provincial Statistical Yearbooks 41.38 88.07
Internet Internet terminals per capita in a city China City Statistical Yearbook 0.82 1.58
Dissimilarity The dissimilarity index The Fifth National Census of China 0.18 0.20
Gini Gini coefficients Meng, Gregory R. and Wang (2005) 0.26 0.03
Resource Employment share of the mining and quarrying sector China Statistical Yearbooks 4.93 3.75
Anglo-American Regional number of church universities founded by Britain China Continuation Committee
and America before 1922 (2007)
Decentralization Ratio of per capita provincial consolidated spending to per China Finance Yearbooks 38.20 19.52
capita central consolidated spending
(Government expenditure – government revenue ) / China City Statistical Yearbook 37.27 18.19
government expenditure
Regulation Relationship between the market and the government Fan, Wang, and Zhu (2010) 6.72 2.04
Share of employment of the private firms China City Statistical Yearbook 0.39 0.12
Resource The fraction of employment in the mining sector in 2003 China City Statistical Yearbook
Female Female representation in the National People’s Congress Documents of National People’s 0.22 0.041
Congress
Population Logarithm of regional population China Statistical Yearbooks 8.02 0.89
North Dummy indicating regions in North China China Statistical Yearbooks

41 
 
 

Table 2-8 Pairwise correlation coefficients between explanatory variables in the province-level analysis
Income Education Anticorruption Media Openness Resource Decentralization Wage Dissimilarity AngloAmerican Female Regulation North
Income 1.00
Education 0.76 1.00
Anticorruption 0.78 0.79 1.00
Media 0.49 0.76 0.56 1.00
Openness 0.73 0.80 0.74 0.75 1.00
Resource -0.30 -0.21 -0.37 -0.26 -0.43 1.00
Decentralization -0.45 -0.33 -0.15 -0.34 -0.54 0.21 1.00
Wage 0.38 0.14 0.23 0.068 0.26 -0.31 -0.29 1.00
Dissimilarity -0.36 -0.12 -0.13 -0.19 -0.29 0.011 0.51 -0.20 1.00
Anglo-American 0.47 0.24 0.25 0.22 0.46 -0.44 -0.58 0.38 -0.41 1.00
Female 0.15 0.24 0.17 0.47 0.27 -0.12 -0.35 0.17 -0.10 0.32 1.00
Regulation 0.80 0.48 0.56 0.23 0.57 -0.35 -0.58 0.44 -0.44 0.51 0.061 1.00
North 0.090 0.26 0.0061 0.16 -0.023 0.58 0.16 -0.28 0.086 -0.41 -0.078 -0.24 1.00

Table 2-9 Pairwise correlation coefficients between explanatory variables in the city-level analysis
Income Education Anticorruption Internet Decentralization Openness Regulation AngloAmerican North
Income 1.00
Education 0.50 1.00
Anticorruption 0.029 -0.18 1.00
Media 0.51 0.58 -0.018 1.00
Decentralization -0.73 -0.39 -0.086 -0.31 1.00
Openness 0.63 0.52 0.038 0.61 -0.40 1.00
Regulation 0.029 0.0089 -0.056 0.10 0.046 0.0085 1.00
Anglo-American 0.35 0.74 -0.016 0.21 -0.33 0.33 -0.044 1.00
North -0.071 -0.032 -0.181 -0.11 0.19 -0.17 -0.33 -0.15 1.00

42 
 
 

Table 2-10 First-Stage Regressions Based on Table 2-5

Instruments Coefficients of corresponding instrumental variables in first stage regressions
Latitude 0.037*** 0.037*** 0.035*** 0.035*** 0.037*** 0.036*** 0.038*** 0.037*** 0.025*** 0.045*** 0.037*** 0.027*** 0.043*** 0.043*** 0.045*** 0.043***
(0.0029) (0.0028) (0.0029) (0.0060) (0.0030) (0.0028) [0.0030] (0.0029) (0.0034) (0.0031) (0.0030) (0.0044) (0.0040) (0.0037) (0.0036) (0.0042)
Middle1952 3.76*** 3.28*** 3.89*** 2.40*** 3.71*** 3.91*** 3.83*** 3.80*** 3.58*** 3.69*** 3.74*** 3.32*** 3.29*** 3.29*** 3.31***
(0.37) (0.45) (0.37) (0.49) (0.35) (0.40) (0.39) (0.38) (0.40) (0.38) (0.37) (0.56) (0.45) (0.45) (0.45)
Gini86 0.35***
(0.11)
Employ1978 -6.77*** -8.25***
(1.40) (1.60)

Population 4.06*** 6.56*** 6.85*** 6.87***

(0.98) (0.94) (0.97) (0.96)
Latitude*period 0.018***
(0.0036)
Instruments F-test of excluded instrumental variables
Latitude 163.81 122.92 131.70 12.59 139.81 167.51 164.20 157.77 104.27 195.13 109.05 48.75 54.51 80.34 153.24 40.90
[0.00] [0.00] [0.00] [0.00] [0.00] [0.00] [0.00] [0.00] [0.00] [0.00] [0.00] [0.00] [0.00] [0.00] [0.00] [0.00]
Middle1952 81.02 53.49 86.54 20.79 115.50 79.47 80.47 80.73 44.83 83.67 59.53 34.14 59.08 83.26 44.91
[0.00] [0.00] [0.00] [0.00] [0.00] [0.00] [0.00] [0.00] [0.00] [0.00] [0.00] [0.00] [0.00] [0.00] [0.00]
Gini86 10.12
[0.00]
Employ1978 13.32 30.72
[0.00] [0.00]
Population 19.44 16.38 18.89 14.64
[0.00] [0.00] [0.00] [0.00]
Latitude*period 9.60
[0.00]
Instruments Partial R-squared of excluded instrumental variables
Latitude 0.32 0.29 0.28 0.23 0.32 0.34 0.32 0.32 0.38 0.37 0.32 0.33 0.28 0.28 0.26 0.28
Middle1952 0.62 0.48 0.63 0.37 0.65 0.62 0.62 0.62 0.61 0.62 0.63 0.44 0.46 0.46 0.46
Gini86 0.24
Employ1978 0.21 0.45
Population 0.30 0.26 0.24 0.25
Latitude*period 0.14
Anderson canon. 19.35 9.33 18.20 17.87 21.53 14.16 20.30 22.23 9.08 28.97 20.73 10.57 17.89 17.58 80.93 13.92
corr. LM statistic [0.00] [0.00] [0.00] [0.00] [0.00] [0.00] [0.00] [0.00] [0.00] [0.00] [0.00] [0.00] [0.00] [0.00] [0.00] [0.00]
Note: Robust standard errors in parentheses; p-values in brackets; ***, ** and * denote significance at 1%, 5% and 10%, respectively.

43 
 
  

Chapter Three Political Interest and Corruption: Cross-country Evidence27

3.1 Introduction

Research into the determinants of corruption has recently intensified, with an increasing
number of studies devoted to exploring the causes and consequences of corruption at the
international level. However, most of these studies explore corruption at the macro level
while only a limited number of studies have investigated the determinants of corruption at the
individual level (see, e.g., Mocan 2004, Swamy et al. 2001, Torgler and Valev 2006). This
empirical study aims to identify the determinants of corruption at the micro level by working
with a set of individual level data covering a large number of countries. We analyze a cross-
section of individuals from the World Values Survey wave III (1995-1997) employing the
perceived corruption and the justifiability of corruption as dependent variables. .An
additional objective in this chapter is to investigate whether political interest affects
corruption by working with several different proxies of political interest, i.e. discussion
intensity, interest in politics and importance of politics in life. Despite the increasing interest
of economists in the determinants of corruption, the link between political interest and
corruption has not yet come under intense empirical investigation.
We anticipate that the use of micro-data sets will afford more insights into the corruption
literature. However, there are benefits and drawbacks of such a data set. One of the major
advantages lies in the ability to investigate a broad set of countries. On the other hand,
drawing conclusions from such a large data file could prove problematic since institutional
and cultural frameworks in certain countries might influence corruption, and such features
cannot always be controlled in a satisfactory manner. To this end, we also provide within-
country evidence focusing on Switzerland, since the institutions in this country are not
homogenous. Analyzing Swiss data is interesting because the degree of institutionalized
political participation rights varies strongly between the 26 Swiss cantons.
There are five innovative aspects to this chapter: 1) it explores the relationship between
political interest and corruption using three different proxies of political interest. Previous
studies have not explored this link, but rather have discussed the impact of formal education
without considering the impact of political interest or informal education. 2) While we
observe a large number of studies at the macro-level, we observe only a limited number of
                                                            
27
 This chapter is published in the Journal of Interdisciplinary Economics 2009, 21, 295-325. 

44 
 
 

micro-level studies. Mocan (2004) suggests a possible cause of this deficiency in previous
studies: “because corruption data are available only at the aggregate (country) level, existing
research has focused on explaining the cross-country variation in corruption. Two exceptions
are Swamy et al. (2001) and Svensson (2003)” (p. 2). 3) Most studies at the macro level focus
on the perceived level of corruption without considering the willingness to bribe (justifiability
of corruption). This study explores both aspects in detail. It is interesting to note that by
measuring the willingness to accept corruption we are able to investigate the social norms of
compliance in a society. 4) We not only provide cross-country evidence at the micro level,
but also explore the robustness of this evidence by focusing on a country that has a certain
level of institutional variation (i.e. Switzerland). 5) We explore additional interesting factors
such as trust in institutions, voice and accountability and democratic participation rights.
Before considering these findings in detail, however, Section 3.2 aims to outline our
theoretical approach. Section 3.3 presents the data and Section 3.4 the empirical findings.
Section 3.5 finishes then with some concluding remarks.

3.2 Political Interest

3.2.1 Theoretical Considerations

Political interest influences the extent to which individuals go about collecting, processing,
and interpreting political matters. A government could operate with impunity if no-one is
motivated to analyze the information available regarding its activities (Rose-Ackerman 1999).
To a certain extent, political interest leads to better supervision and scrutiny of the
administration and governance performance and may contribute to a stronger sense of civic
awareness among citizens. This increased knowledge possibly augments the ability to acquire
political information at lower costs which in turn increases the individual incentive to be
informed and to discuss political issues. Hence this process acts as a sort of “multiplier
effect”. Rose-Ackerman (1997) states that corruption can be limited “by outside pressure
from the public” (p. 143). A politically interested population could demand a higher level of
transparency and may be able to better monitor and control politicians, potentially reducing
the perceived complexity of the political process
Politically interested citizens will associate with one another and engage in discussion.
Exchange of arguments and face-to-face interaction enhances group identification. and gives
citizens the opportunity to identify others’ preferences. As others’ preferences become visible,
the moral costs of free-riding or behaving illegally increase, reducing the justifiability of
corruption. If political discussion is common in a society, citizens are confronted with

45 
 
 

arguments from both sides, those favouring and those opposing a certain political outcome,
and this increases the overall level of information. In addition, once citizens become involved
in an issue, they feel responsible for the result which may create a sense of civic duty and a
higher willingness to comply. Thus, discussion provides the opportunity to clarify benefits
and costs of political issues and thus increases co-operation among group members. This
increases the human capital involved in and developed by political matters. Mocan (2004)
stresses that a higher stock of human capital reduces the tolerance of corruption.
Several studies have found that political interest contributes to the probability of their
being involved in the political process (Verba, Schlozman, and Brady 1995). Political interest
becomes an important explanatory factor in models of political behaviors from political
sophistication (Carpini and Keeter 1996) to voting (Verba, Scholzman, and Brady 1995).
Looking at tax compliance, Alm, McClelland, and Schulze (1999) argue that there is a social
norm of compliance affecting individual reporting decisions. Their findings indicate that
communication combined with the vote influences tax compliance, so that paying taxes
becomes the accepted mode of behaviour. Discussion gives the chance to clarify benefits and
costs of a topic and thus increases co-operation among group members. In general, Alm
(1996) argues, after surveying experimental findings that “I believe that the cheap talk in
combination with vote allows individuals to change the social norms, in this case to
demonstrate that evasion will not be accepted”.
Kuenzi (2006) has empirically demonstrated that civic education (non-formal education)
has a significant positive impact on political participation. This kind of education is the result
of an informal process that is not necessarily a part of an individual’s formal education.
Nevertheless, individuals certainly invest energy, time and money on this informal education.
The expenses involved in being politically interested (represented by the costs of informal
education) may outweigh the benefits (represented by increased accountability and
transparency of the administration). In our case, we can argue that people balance the cost of
maintaining a political interest with the benefit derived from controlling and reducing
corruption, (keeping in mind the consequences of corruption). To demonstrate this aspect we
first employ a simple model that allows illustration of the relationship between political
interest and the level of corruption.
3.2.2 A Simple Model
Following the spirit of Ades and Di Tella (1997), we build a simple model to elaborate the
effect of political interest on corruption. We focus on the decision making of the risk-neutral

46 
 
 

bureaucrats. We assume that the bureaucrats have ongoing wages w, and can take bribes b.
However, the probability that their corrupt acts are detected is p. This probability is mainly
determined by available information of a corrupt act. It is natural that the political interested
citizens can collect more and exploit information better than politically disinterested people.
∂p
Thus we have: ∂I 0, ∂I ∂PI 0, where I denotes information and PI represents

political interest level of citizens. If a corrupt bureaucrat is detected, he will be ousted by

citizens with a loss f. Political interest may translate into stronger actions against corruption
by identifying illegal treatment and reducing the willingness to accept bribes even when the
costs of appealing are very high or the formal mechanisms of internal and external control are
not functioning effectively. Politically disinterested individuals may surrender more easily to
extortion, as they will not take into account the consequences and issues associated with
corruption. Thus, political interest may substantially reduce the costs of fighting extortive
corruption. Politically interested persons may find channels to reveal corrupt behavior or at
least raise the costs of illegal behavior by demonstrating their higher willingness to use
instruments for voicing complaints and threatening to undermine the political support for a
government. In addition, the politically interested citizens’ process of informal education will
uncover political information, helping them understand what is expected of a legitimate
government. Such understanding reduces the constraints on potential complaints and puts
pressure on the government and the bureaucrats to act in the public interest. This is especially
important in countries where other means of constraining bureaucrats and politicians may be
lacking. Rose-Ackerman (1999) points out that groups and individuals have effective avenues
for challenging official actions. Although policies that enhance accountability and openness
“are likely to be more acceptable to democratically elected leaders, these reforms can also
have an effect in undemocratic systems whose leaders nevertheless need public support to
retain power” (p. 144). However, it is possible that the government could stonewall any
process until the protest groups have exhausted their energy and resources (Rose-Ackerman
1999). Bureaucrats could ignore the threats of lobbyists and protestors in expectation of using
such a strategy.
Every bureaucrat compares his income from corruption with one from incorruption to
decide whether to be corrupt.
The corrupt income is y 1 p PI w b p PI w f 3 1
The incorrupt income is y w 3 2
He will be corrupt if y y . It means that the bureaucrat will corrupt if
47 
 
 

p PI
b b f 3 3
1 p PI
Assume that b is distributed according to the cumulative function F(.). Then the fraction
of bureaucrats that are corrupt is given by
P 1 F b b 1 F b p PI , f 3 4
The equation above shows that the level of corruption can be a function of the political
interest level of citizens and it is obvious that P is decreasing in PI. This model provides a
direct theoretical framework to examine the effect of political interest on corruption.

3.3 Data

The data used in the present study came from World Values Survey wave III. The World
Values Survey is a worldwide investigation of socio-cultural and political change. These
surveys have assessed the basic values and beliefs of people in many countries. The World
Values Survey was first carried out in 1981-1983, with subsequent surveys being carried out
in 1990-1993, 1995-1997 and 1999-200128. We work with the third wave, as the question
referring to individual perceived corruption has only been asked in this wave29.
For the researchers who conduct and administer the World Values Survey (WVS) in their
respective countries, it is a requirement that they follow the methodological requirements of
the World Values Association. For example, surveys in the World Values Survey set are
generally based on nationally representative samples of at least 1000 individuals of 18 years
and above (although sometimes people under the age of 18 participate). The samples are
selected using probability random methods, and the questions contained within the surveys
generally do not deviate far from the original official questionnaire (for a sample of a typical
World Values Survey see www.worldvaluessurvey.org).
3.3.1 Dependent Variables
Our dependent variables are perceived corruption, and the justifiability of corruption. To
assess the level of perceived corruption from the WVS, we use the following question:
How widespread do you think bribe taking and corruption is in this country?
Almost no public officials are engaged in it (1)
A few public officials are engaged in it (2)
                                                            
28
 Data from the 1999-2001 wave became available after our study was completed.
29
 Countries in the sample: Armenia, Australia, Azerbaijan, Bangladesh, Belarus, Bosnia-Hercegovina, Brazil,
Bulgaria, Chile, Croatia, Estonia, Finland, India, Latvia, Lithuania, Macedonia, Mexico, Moldova, Nigeria,
Norway, Peru, Philippines, Russia, Serbia, Slovenia, Spain, Switzerland, Taiwan, Ukraine, Uruguay, USA,
Venezuela, Germany (differentiating between East and West Germany).

48 
 
 

Most public officials are engaged in it (3)

Almost all public officials are engaged in it (4)
The justifiability of corruption is measured with the following variable:
Please tell me for each of the following statements whether you think it can always be
justified, never be justified, or something in between: (...) someone accepting a bribe
in the course of their duties (1=always justified, 10= never justified).
The interpretation of this question is that a higher value leads to lower justifiability of
corruption. This variable can be seen as a proxy for social norms of compliance (see Torgler,
2007).
The two dependent variables used are not free from biases and problems. Use of
‘perceived corruption’ is in line with other indexes that employ measures of perceptions
(such as the Transparency International index). However, perceptions are not objective, nor
are they quantitative measures of the actual degree of corruption. Perceptions are rather an
indirect way of measuring corruption (Tanzi 2002). However, when analyzing the
Transparency International index, Treisman (2000, pp. 410-411) presents valid arguments as
to why data based on perceptions should be taken seriously. Components of the surveys and
ratings are highly correlated among themselves, even though they have been conducted with
different methodologies, different inputs and in different time periods. Such consistency
allows us to conclude that factors are almost free of biases such as a “temporal mood” or
guesses. There is also a consistency in the Transparency International over time, although the
construction of the index varies over time. Finally, the index is strongly correlated with other
corruption indexes such as the ICRG, the BI or the Gallup International.
A practical method by which we can test whether the World Values Survey question
about the perceived corruption is through the use of a useful proxy to check whether the
variable is correlated with other well-known indexes on corruption. Thus, we compare our
variable with the corruption indexes TI (Transparency International), International Country
Risk Guide (ICRG) and Quality of Government (Control of Corruption) developed by
Kaufmann, Kraay, and Mastruzzi (2003). The World Values Survey Corruption ratings are
highly correlated with the TI (r=-0.88), the ICRG (r=-0.68) and the Quality of Government
rating (r=-0.83)30.
The validity of the justifiability of corruption variable can be criticized as it reports a self-
reported and hypothetical choice (see Swamy et al. 2001). It can also be argued that an
                                                            
30
The sign is negative because for all three ratings used (TI, ICRG and Quality of Government), a higher score
corresponds to a lower corruption.

49 
 
 

individual who has engaged in corruption in the past will tend to cover up such behavior by
declaring a low justifiability of corruption in the survey. Furthermore, cross cultural
comparisons should be treated with some caution. In countries where corruption is
widespread and delays in transactions are long, additional payments to “speed up” the process
may be justifiable and a normal part of the administration process. The necessity of additional
payments is so pervasive in some countries that the bureaucratic mechanism does not operate
without them. De Soto (1989) and his research team conducted an experiment in which they
set up a small garment factory in Lima with the aim of complying with the bureaucratic
procedures and thus behaving in accordance with the law. He reports that 10 times they were
asked for a bribe to speed up the process and on two occasions, payment of the bribe was the
only way to continue the experiment. However, a side effect from higher justifiability of
corruption due to the ubiquitous nature of this behaviour is that the bureaucrats have a
stronger incentive to delay the transactions in order to extract further payments. Justifiability
is also correlated with most other corruption measurements: it is statistically significant at the
0.05 level but with lower r values compared to perceived corruption (TI (r=0.36), the ICRG
(r=0.19, not statistically significant), the Quality of Government rating (r=0.38), and
perceived corruption (r=-0.42)).
We have not analyzed the entire World Value Survey data set. Countries below 750
observations have not been included in the estimations to reduce possible biases due to a lack
of representativeness 31 . Furthermore, some countries do not have information on the
dependent variables or some of the independent variables. These countries are therefore not
considered. Furthermore, not all countries have information regarding the dependent and
independent variables integrated in the estimations32. For example, Sweden could not be
included as one of the control variables (education) has been coded differently.
3.3.2 Measuring Political Interest
We will use several proxies of political interest to investigate this main hypothesis, thereby
checking the robustness of the results. First of all we focus on the intensity of political
discussion by using the responses to the following survey question:
When you get together with your friends, would you say you discuss political
matters frequently (value 3), occasionally (value 2) or never (value 1)?
The second variable focuses on the interest in politics itself:
                                                            
31
Thus, Montenegro and the Dominican Republic have been omitted.
32
For the estimations with the dependent variable perceived corruption: Japan, South Africa, Puerto Rico, China,
Columbia. Estimation with justifiability of corruption: Japan, South Africa, Puerto Rico, Turkey and Columbia.

50 
 
 

How interested would you say you are in politics? Very interested (value 3),
somewhat interested (2), not very interested (1).
The third variable measures the importance of politics in a person’s life with the following
question:
How important is politics in your life? very (4), rather( 3), not very (2), not at
all (1).
The advantages of using three different proxies are demonstrated by our ability to conduct
a robustness test while measuring different dimensions of political interest.
3.3.3 Further Independent Variables
The correlation between political interest and corruption could be influenced by other
variables that affect corruption, complicating our efforts to isolate the impact of political
interest. Thus, we control for the education level, the marital status, political trust,
institutional conditions, religion, risk attitudes, the economic situation and the employment
status.
a) Education
The variable education 33 (continuous variable, 1=low, 9=high education) is related to
citizens’ knowledge about corruption. To observe the relative importance of political interest,
it necessary to control for formal education, as it is assumed that better educated individuals
are more aware of government’s activities and thus would be in a better position to assess the
degree of corruption. This may have a positive or a negative impact on the justifiability of
corruption and the perceived corruption, depending on the actions of the government. On the
other hand, they may be more strongly involved in corruption, understanding better the
opportunities of corruption. Thus, the effect of education is not clear and there is a lack of
empirical studies that investigate the correlation between education and corruption. Swamy et
al. (2001), for example, disregard the variable. Mocan (2004) found that a higher level of

                                                            
33
 What is the highest educational level that you have attained?
1. No formal education
2. Incomplete primary school
3. Completed primary school
4. Incomplete secondary school: technical/vocational type
5. Complete secondary school: technical/vocational type
6. Incomplete secondary: university-preparatory type
7. Complete secondary: university-preparatory type
8. Some university-level education, without degree
9. University-level education, with degree
 

51 
 
 

education leads to a higher probability of being targeted for bribes, yet a more educated
population is expected to be less tolerant of corruption.
b) Age
A limited number of studies have included age in their estimations. Swamy et al. (2001)
consider age as a control variable in their estimations of the justifiability of corruption and
find a positive but non-linear effect. The authors focused on gender differences and did not
comment on this result. Mocan (2004) also uses micro data to show an effect of age on
corruption: individuals at the age of 20 to 54 are more likely to be asked for a bribe compared
to the reference group (younger than 20). Torgler and Valev (2006) explore the impact of age
on corruption, differentiating between the same cohorts over time (age effect) as well as the
same age groups in different time periods (cohort effect). The chapter observes a consistent
age effect, while a cohort effect is less obvious. There are two major concepts that explain
the correlation between age and crime: the traditional desistance theory and the age theory.
The desistance theory asserts that the decline in crime results from factors associated with age
that reduce or change the actors’ criminality, and older people are restrained from offending
due to changes in their status or the exposure to anti-criminal institutions. On the other hand,
the age theory is based on the idea that the biological aging of the organism itself has an
impact on individuals’ criminal behaviour (for an overview see Torgler and Valev 2006).
Instead of using age as a continuous variable, we have formed four classes: Age <30, Age 30-
49, Age 50-64, Age 65+, with Age <30 as reference group, to better investigate the impact of
age.
c) Gender
Research in social psychology suggests that women are more compliant and less self-
reliant than men (e.g., Tittle 1980). In the past decade, experimental research findings have
shown that gender may influence aspects of behavior such as charitable giving, bargaining,
and household decision making (see Andreoni and Vesterlund 2001, Eckel and Grossman
2001). Evidence from the tax compliance literature shows there is a tendency for men to be
less compliant and have lower tax morale than women (see Torgler 2007). Further evidence
regarding gender differences can also be found in helping behavior (see, e.g., Eagly and
Crowley 1986) or ethical decision making (Ford et al. 1994, Glover et al. 1997 and Reiss and
Mitra 1998).
The criminology literature provides one of the best sources for observing possible gender
differences. Mears et al. (2000) report that men commit more offenses than women age “at
every age, within all racial or ethnic groups examined to date, and for all but a handful of
52 
 
 

offense types that are peculiarly female… sex differences in delinquency are independently
corroborated by self-report, victimization, and police data, and they appear to hold cross-
culturally as well as historically” (p. 143). Torgler and Valev (2007) find strong evidence that
women report a lower justifiability of committing illegal activities than men. The results
remain robust after investigating different time periods and extending the specification with
several opportunity factors such as education, employment status or income.
d) Marital status
Marital status is a further control variable (dummy variable, value 1 if the respondent is
married). Married people may be more compliant than others, especially compared to singles
because they are more constrained by their social network (Tittle 1980). It is also argued that
marriage alters public behaviour (Swamy et al. 2001). Tittle (1980) found significant
differences between the different marital statuses, with the greatest evidence for the singles,
followed by the separated or divorced. However, controlling for age, the results show that the
association between deviance and marital status was a reflection of age difference, as older
persons are more likely to be married or widowed and age was a strong predictor concerning
the deviance. Gottfredson and Hirschi (1990) also point out that the literature on crime finds
that marital status does not seem to have an impact on the likelihood of crime.
e) Economic situation
As a proxy for income we use the economic situation of an individual (dummies upper
class, middle and lower class are in the reference group). Using the exact income would
produce biases, because of difficulties comparing this variable across different countries.
Individuals with a higher income are more likely to be asked for a bribe, as are those with a
better education. Individuals with a lower income might have lower social “stakes” or
restrictions but are less in a position to take risks, because of a high marginal utility loss
(wealth reduction) if they are caught and penalized.
f) Occupation status
Another variable is the occupation status as it affects whether the respondent is in a
position to benefit from corruption (see Swamy et al. 2001). We will use a dummy variable
for self-employed individuals as they might be in the best position to invest in bribing and
benefit from corruption. Such a status may have an impact on the norms regarding bribery.

53 
 
 

g) Risk attitudes
We include a dummy variable that measures risk aversion34 as individual willingness to
behave illegally could also be a function of risk attitudes. It is interesting that few prior
survey studies have controlled for risk attitudes, since risk aversion reduces the incentive to
act illegally. Furthermore, controlling for risk attitudes affords better insights regarding the
variables of age, gender, or economic situation. It could be argued that the observed
difference between women and men or between different age groups is influenced by
different risk attitudes functions.
h) Urbanization
Mocan (2004) stresses that in larger cities the extent of bribery may be higher since the
scale of economic activities is larger and more varied in scope, resulting in increased contact
with the government. Moreover, government officials may be less personal compared to
those in smaller cities: this may reduce the opportunity costs of bribing. We use town size as
a proxy for urbanization.35
i) Religion
Religion might influence people’s habits and act as a restriction on engaging in illegal
activities (Torgler 2006). We take the frequency of church attendance (Church Attendance36)
as the religious variable, showing approximately how much time individuals devote to

                                                            
34
 Now I would like to ask you something about the things which would seem to you personally, most important
if you were looking a job. Here are some of the things many people take into account in relation to their work.
Regardless of whether you’re actually looking for a job, which one would you, personally, place first if you
were looking for a job?
1. A good income so that you do not have any worries about money
2. A safe job with no risk of closing down or unemployment
3. Working with people you like
4. Doing an important job which gives you a feeling of accomplishment
And what would be your second choice?
A dummy variable was built with the value 1, if someone has chosen 2 as first or as second choice.
35
V232. Size of town:
1. Under 2,000
2. 2,000 - 5,000
3. 5 - 10,000
4. 10 - 20,000
5. 20 - 50,000
6. 50 - 100,000
7. 100 - 500,000
8. 500,000 and more.
36
Apart from weddings, funerals, and christenings, about how often do you attend religious services these days?
More than once a week, once a week, once a month, only on special holy days, once a year, less often, never or
practically never. (7 = more than once a week to 1 = never or practically never).

54 
 
 

religion. It is anticipated that this variable tells more about behaviour than self-reported
religious attitudes.
j) Political trust
Economists have recently started to pay attention to the determinants of trust through the
literature on compliance (e.g. Torgler 2007). Trust in the state might influence citizens’
positive attitudes and commitment to the rules of a society, which ultimately has a negative
effect on illegal activities. Those institutions perceived by citizens as trustworthy, fair and
efficient could act as constraints on corruption. We are exploring several different dimensions
of trust, namely trust in the legal system 37 , trust in the government 38 , and trust in the
parliament39. The analysis will therefore cover trust at the constitutional and current politico-
economic level. Controlling for this variable will better check the impact of political interest
since individuals with a lower level of political trust might be frustrated and therefore less
interested in following politics.
k) Voice, Accountability and Democratic Rights
We also control for institutional conditions. In particular, it is important to control for the
citizens’ opportunity to translate their political interest into political actions; i.e. whether they
have a meaningful ‘voice’ in influencing the state (e.g., through voting processes). Holding
such institutional conditions constant allows analysis of how strong political interest can affect
corruption. In general, the greater the ‘voice’ of citizens, the less we expect to observe
corruption, all other things being equal. A progressive government can attempt to increase or
initiate co-operation and generate trust by developing functioning institutions. Furthermore,
co-operation is enhanced when citizens are satisfied with the way they are treated. On the
other hand, if certain sectors of the government are not benevolent, strong institutional
instruments have the potential to control politicians’ discretionary power. Voter power helps
limit the abuse of political power by selfish politicians especially since citizens cannot
completely foresee the incumbents’ preferences. The elements of direct democracy can also
empower citizens with an instrument for controlling the government. Such control has an ex
ante effect on policy formulation by elected incumbents in that they must always take into

                                                            
37
Could you tell me how much confidence you have in the legal system: is it a great deal of confidence, quite a
lot of confidence, not very much confidence or none at all? (4= a great deal to 1=none at all).
38
Could you tell me how much confidence you have in the government in your capital: is it a great deal of
confidence, quite a lot of confidence, not very much confidence or none at all? (4= a great deal to 1=none at all).
39
Could you tell me how much confidence you have in parliament: Do you have a great deal of confidence,
quite a lot of confidence, not very much confidence or no confidence at all? (4=a great deal of confidence to
1=no confidence at all).

55 
 
 

account possible voter intervention. Levi (1988) points out that a possible consequence of
creating or maintaining compliance is to provide reassurance by the government. A
government that precommits itself with democratic rules imposes self-restraints on its own
power and thus sends a signal that taxpayers are seen as responsible persons. Furthermore,
direct democratic rules signal that citizens are not ignorant or uncomprehending voters, which
might create or maintain a certain social capital stock that should also affect the justifiability of
corruption.
In the cross-country study we use Kaufmann et al. (2003) variable Voice And
Accountability for the year 1996. The variable measures the political process, civil liberties,
and political rights of a country. We are going to use an index of the degree of direct
democracy developed by Stutzer (1999) and applied in papers such as Frey and Stutzer (2000,
2002), Frey and Feld (2002), Torgler (2005), Schaltegger and Torgler (2007) when exploring
Switzerland. The index reflects the extent of direct democratic participation (1= lowest and 6=
highest degree of participation) at the cantonal level.
l) Regions
We will also control for regional differences considering the dummies CEE and FSU
(Central Eastern and Former Soviet Union countries), Latin America, Asia and Africa40. The
reference group consists of Western Europe + USA + Australia. It can be assumed that there
are regional differences in the perceived corruption and justifiability of corruption. We expect
a lower perceived corruption in the reference group countries, based on a historically high
standard of rule of law and accountable systems of governance. Furthermore, it is possible
that a higher justifiability of corruption exists in countries where these important factors are
lacking.

3.4 Empirical Evidence

We will use a weighted ordered probit estimation to correct the samples and thus to get a
reflection of the national distribution. In estimations where several countries are pooled we
have integrated an additional weighting variable. To obtain an equal number of weighted
observations (around 1500) for each survey, the original weight variable was multiplied by a
constant for each country 41 . The ordered probit models are relevant in such an analysis
insofar as they help to analyse the ranking information of the scaled dependent variable.
However, since equations in the ordered probit estimation have a nonlinear form, only the
                                                            
40
 Only one country represents Africa (Nigeria). 
41
 The World Values Survey provides the weighting variables.

56 
 
 

sign of the coefficient can be directly interpreted and not its size. Calculating the marginal
effects is therefore a method to find the quantitative effect of a variable on our dependent
variable. The marginal effect indicates the change in the share of individuals (or the
probability of) belonging to a specific perceived corruption (justifiability) level, when the
independent variable increases by one unit. In all estimations the marginal effects are
presented only for the highest value. Furthermore, it should be noticed that answers such as
“don’t know” and missing values have been eliminated in all estimations.
3.4.1 International Evidence
Tables 3-1 to 3-6 present the first results. Tables 3- 1, 3-3 and 3-5 explore the justifiability of
corruption as dependent variable, while Tables 3-2, 3-4, and 3-6 analyse the perceived
corruption. Tables 3-1 and 3-2 investigate the impact of political discussion. Tables 3-3 and
3-4 take a look at the interest in politics and Tables 3-5 and 3-6 report the findings focusing
on the importance of politics. In all tables we present four specifications. This provides a
robustness check of our main variable while taking into account that the number of
observations decreases from one estimation to the other. The baseline specification is
presented in the first regression. In a next step we add variables that measure individuals’
economic situation. In a third regression we include also the three variables on political trust.
Finally, we report a fourth regression that controls for institutional conditions within a
country, focusing on voice and accountability. The results clearly indicate that political
interest matters: in 19 out of 20 regressions the coefficient is statistically significant. We
observe that a higher level of political interest leads to a lower justifiability of corruption and
also to a lower perceived level of corruption. The marginal effects vary between 0.4
percentage points to 3.7 percentage points. Focusing on the justifiability of corruption, we
were not able to observe a decrease in the impact of political interest when controlling for
political trust and voice and accountability. On the contrary, we observe an increase in the
marginal effects. For example, in Table 3-1 we observe that an increase in the political
discussion level by one unit increases the probability of stating that corruption is never
justifiable by 1.3 percentage points. Looking at perceived corruption, we observe a decrease
in the marginal effects. However, the results still indicate that the effects are not at all
negligible. For example, specification (8) in Table 3-2 indicates that an increase in the
political discussion scale by one unit reduces the probability of reporting the highest level of
corruption by 1.6 percentage points.

57 
 
 

While we observe that political interest matters, we cannot observe a statistically

significant correlation between education and our two dependent variables (showing a
negative sign in both cases). Thus, informal education seems to be much more important than
education. This finding suggests that it is important to generate “political human capital”
rather than just generalized human capital.
Interestingly, we observe that voice and accountability reduces the justifiability of
corruption and the perceived level of corruption. The coefficient is highly statistically
significant in all specifications while also reporting large marginal effects. Thus, the findings
indicate that a more legitimate and responsive state is an essential factor for a lower level of
corruption. Similarly, political trust has a negative impact on the justifiability of corruption
and the perceived level of corruption. The joint role played by political trust can be
investigated using a Wald-test for coefficient restrictions to test for joint significance. In all
cases we can observe that the null hypothesis is rejected, meaning that the political trust
variables play a significant role in the determination of countries’ corruption level. Trust in
the legal system provides the most consistent result in all the tables. Thus, trust at the
constitutional level seems to be more important than trust at the current politico-economic
level. The marginal effects are quite substantial, particularly for the perceived corruption
regressions.
Looking at the other variables we observe that all age groups from 30 to 65+ have a
significantly lower justifiability of corruption than the reference group below 30.
Interestingly, we can observe that the marginal effects increase consistently with an increase
of the age group.

58 
 
 

Table 3-1 Justifiability of corruption and political discussion

Weighted Ordered Probit Coeff z-Stat. Marg. Coeff. z-Stat. Marg. Coeff. z-Stat. Marg. Coeff. z-Stat. Marg.
(1) (2) (3) (4)
Politicial Discussion 0.026** 2.29 0.008 0.031*** 2.62 0.010 0.035*** 2.91 0.011 0.052*** 3.28 0.017
Formal Education 0.001 0.20 0.0002 0.002 0.64 0.001 0.006* 1.68 0.002 0.005 0.93 0.001
Age 30-49 0.19*** 10.06 0.059 0.19*** 9.86 0.059 0.19*** 9.70 0.060 0.17*** 6.67 0.054
Age 50-64 0.38*** 15.96 0.11 0.39*** 15.82 0.11 0.39*** 15.27 0.11 0.36*** 10.76 0.11
Age 65+ 0.50*** 15.44 0.14 0.52*** 15.50 0.14 0.53*** 14.97 0.14 0.52*** 11.30 0.14
Female 0.14*** 9.67 0.044 0.14*** 9.57 0.045 0.14*** 9.36 0.045 0.16*** 7.79 0.050
Married 0.12*** 6.46 0.039 0.12*** 6.16 0.038 0.12*** 6.16 0.040 0.12*** 4.49 0.038
Widowed 0.15*** 4.32 0.047 0.15*** 3.96 0.044 0.14*** 3.59 0.042 0.10** 2.09 0.031
Divorced 0.016 0.42 0.005 0.006 0.16 0.002 0.008 0.21 0.003 0.013 0.27 0.004
Separated 0.068 1.29 0.021 0.076 1.43 0.024 0.095* 1.72 0.029 0.082 1.15 0.025
Upper Class -0.19*** -3.57 -0.065 -0.18*** -3.20 -0.061 -0.14** -2.00 -0.048
Upper Middle Class -0.019 -0.97 -0.006 -0.032 -1.58 -0.010 -0.026 -0.93 -0.008
Selfemployed -0.062** -2.37 -0.020 -0.068** -2.52 -0.022 -0.057** -2.06 -0.019 -0.096*** -2.76 -0.031
Risk Averse 0.077*** 4.99 0.024 0.073*** 4.57 0.023 0.073*** 4.46 0.023 0.077*** 3.47 0.024
Urbanization -0.007** -2.53 -0.002 -0.006** -2.24 -0.002 -0.005 -1.55 -0.001 -0.009** -2.34 -0.003
Church Attendance 0.012*** 3.09 0.004 0.009** 2.21 0.003 0.006 1.49 0.002 -0.001 -0.13 0.000
Legal System 0.040*** 4.07 0.013 0.071*** 5.46 0.023
Government -0.007 -0.58 -0.002 0.041*** 2.78 0.013
Parliament 0.022* 1.87 0.007 0.015 1.01 0.005
Voice And Accountability. 0.11*** 7.43 0.036
CEE And FSU -0.38*** -20.50 -0.12 -0.38*** -19.87 -0.12 -0.38*** -18.88 -0.12
Latin America -0.43*** -17.92 -0.15 -0.43*** -17.50 -0.15 -0.42*** -16.56 -0.14 -0.26*** -5.82 -0.084
Asia 0.20*** 6.31 0.061 0.35*** 9.64 0.098 0.36*** 9.28 0.10 -0.030 -0.68 -0.009
Africa -0.23*** -3.91 -0.078 -0.19*** -3.17 -0.065 -0.23*** -3.60 -0.077 0.61*** 11.26 0.16
Wald-test joint sign. polit. trust 30.66
Pseudo R2 0.025 0.027 0.027 0.034
Number of observations 41714 39669 36726 20373
Prob > chi2 0.00 0.00 0.00 0.00
Notes: In the reference group are Age<30, Man, Single, Lower Middle And Lower Class, Other Employment Status, Risk Taker, Western Europe + USA + Australia. ***, **
and * denote significance at 1%, 5% and 10%, respectively. Marginal effect = highest score (10, never justifiable). The higher the value the lower the justifiability. CEE: Central
Eastern European Countries, FSU: Former Soviet Union Countries.

59 
 
  

Table 3-2 Perceived corruption and political discussion

Weighted Ordered Probit Coeff z-Stat. Marg. Coeff. z-Stat. Marg. Coeff. z-Stat. Marg. Coeff. z-Stat. Marg.
(5) (6) (7) (8)
Political Discussion -0.030*** -2.94 -0.009 -0.030*** -2.89 -0.009 -0.024** -2.27 -0.007 -0.028* -1.95 -0.008
Formal Education -0.009*** -3.19 -0.003 0.000 0.06 0.0001 -0.003 -0.91 -0.001 -0.011** -2.56 -0.003
Age 30-49 -0.039** -2.29 -0.012 -0.049*** -2.82 -0.015 -0.064*** -3.55 -0.019 -0.024 -1.03 -0.007
Age 50-64 -0.092*** -4.35 -0.028 -0.095*** -4.40 -0.029 -0.088*** -3.90 -0.026 -0.025 -0.84 -0.007
Age 65+ -0.16*** -5.94 -0.048 -0.16*** -5.76 -0.047 -0.13*** -4.37 -0.037 -0.083** -2.07 -0.023
Female 0.020 1.54 0.006 0.027** 2.03 0.008 0.015 1.10 0.005 0.001 0.05 0.000
Married 0.011 0.65 0.004 0.011 0.60 0.003 0.026 1.39 0.008 -0.024 -0.93 -0.007
Widowed -0.028 -0.92 -0.009 -0.041 -1.33 -0.013 -0.012 -0.36 -0.004 -0.10** -2.43 -0.028
Divorced 0.069** 2.09 0.022 0.061* 1.78 0.019 0.058 1.64 0.018 0.096** 2.03 0.028
Separated 0.054 1.17 0.017 0.047 1.00 0.015 0.052 1.05 0.016 0.029 0.45 0.008
Upper Class -0.009 -0.17 -0.003 0.046 0.86 0.014 -0.12* -1.93 -0.034
Upper Middle Class -0.24*** -13.63 -0.070 -0.19*** -10.58 -0.055 -0.17*** -6.50 -0.045
Selfemployed 0.037 1.51 0.012 0.052** 2.09 0.016 0.019 0.73 0.006 0.044 1.41 0.013
Risk Averse 0.024* 1.75 0.007 0.011 0.80 0.003 0.006 0.40 0.002 0.011 0.55 0.003
Urbanization 0.034*** 13.54 0.011 0.035*** 13.53 0.011 0.024*** 9.06 0.007 0.005 1.31 0.001
Church Attendance 0.003 0.88 0.001 0.005 1.29 0.001 0.019*** 5.21 0.006 0.012** 2.27 0.003
Legal System -0.15*** -16.84 -0.046 -0.16*** -13.20 -0.046
Government -0.12*** -12.23 -0.038 -0.13*** -9.71 -0.037
Parliament -0.17*** -16.20 -0.053 -0.16*** -11.77 -0.047
Voice And Accountability -0.16*** -11.240 -0.045
CEE And FSU 0.94*** 58.66 0.29 0.89*** 53.98 0.28 0.92*** 53.36 0.28 0.98*** 24.51 0.29
Latin America 0.68*** 30.80 0.24 0.64*** 28.18 0.22 0.63*** 27.17 0.21 0.59*** 14.88 0.18
Asia 0.54*** 20.52 0.19 0.52*** 19.37 0.19 0.76*** 25.67 0.27 0.89*** 20.39 0.31
Africa 1.28*** 21.64 0.48 1.25*** 20.26 0.47 1.32*** 19.67 0.49
Wald-test joint sign. polit. trust 1867.92
Pseudo R2 0.025 0.027 0.027 0.11
Number of observations 41714 39669 36726 18942
Prob > chi2 0.00 0.00 0.00 0.00
Notes: In the reference group are Age<30, Man, Single, Lower Middle And Lower Class, Other Employment Status, Risk Taker, Western Europe + USA + Australia. ***, ** and
* denote significance at 1%, 5% and 10%, respectively. Marginal effect = highest score (4). The higher the value the lower the justifiability. CEE: Central Eastern European
Countries, FSU: Former Soviet Union Countries.

60 
 
  

Table 3-3 Justifiability of corruption and interest in politics

Weighted Ordered Probit Coeff z-Stat. Marg. Coeff. z-Stat. Marg. Coeff. z-Stat. Marg. Coeff. z-Stat. Marg.
(9) (10) (11) (12)
Interest Politics 0.013 1.57 0.004 0.018** 2.22 0.006 0.016* 1.87 0.005 0.041*** 3.58 0.013
Formal Education 0.001 0.24 0.0003 0.002 0.66 0.001 0.007* 1.82 0.002 0.004 0.73 0.001
Age 30-49 0.19*** 10.30 0.060 0.19*** 10.05 0.060 0.19*** 9.91 0.061 0.18*** 6.96 0.056
Age 50-64 0.39*** 16.32 0.11 0.40*** 16.19 0.12 0.40*** 15.64 0.12 0.37*** 11.13 0.11
Age 65+ 0.50*** 15.54 0.14 0.53*** 15.66 0.14 0.53*** 15.10 0.14 0.53*** 11.40 0.14
Female 0.14*** 9.50 0.043 0.14*** 9.47 0.044 0.14*** 9.06 0.044 0.15*** 7.56 0.048
Married 0.12*** 6.50 0.039 0.12*** 6.16 0.038 0.12*** 6.24 0.040 0.12*** 4.52 0.038
Widowed 0.15*** 4.11 0.044 0.14*** 3.70 0.041 0.13*** 3.45 0.040 0.088* 1.81 0.027
Divorced 0.016 0.44 0.005 0.006 0.17 0.002 0.012 0.32 0.004 0.015 0.29 0.005
Separated 0.068 1.28 0.021 0.075 1.40 0.023 0.097* 1.76 0.030 0.083 1.15 0.025
Upper Class -0.20*** -3.82 -0.069 -0.19*** -3.46 -0.065 -0.16** -2.22 -0.052
Upper Middle Class -0.021 -1.04 -0.007 -0.033 -1.62 -0.011 -0.033 -1.15 -0.010
Selfemployed -0.065** -2.48 -0.021 -0.070*** -2.62 -0.023 -0.063** -2.27 -0.020 -0.098*** -2.82 -0.032
Risk Averse 0.077*** 4.97 0.024 0.072*** 4.53 0.023 0.074*** 4.48 0.023 0.078*** 3.52 0.024
Urbanization -0.006** -2.29 -0.002 -0.006** -2.05 -0.002 -0.004 -1.48 -0.001 -0.008* -1.94 -0.002
Church Attendance 0.010*** 2.67 0.003 0.007* 1.72 0.002 0.005 1.19 0.002 -0.001 -0.22 0.000
Legal System 0.039*** 3.94 0.012 0.069*** 5.34 0.022
Government -0.009 -0.78 -0.003 0.036** 2.46 0.011
Parliament 0.021* 1.76 0.007 0.014 0.92 0.004
Voice And Accountability -0.38*** -18.84 -0.12 0.12*** 7.71 0.037
CEE And FSU -0.38*** -20.45 -0.12 -0.38*** -19.78 -0.12 -0.42*** -16.47 -0.14 -0.25*** -5.58 -0.080
Latin America -0.43*** -17.86 -0.15 -0.43*** -17.29 -0.15 0.37*** 9.55 0.10 -0.020 -0.47 -0.006
Asia 0.21*** 6.57 0.063 0.36*** 9.96 0.10 -0.21*** -3.40 -0.072 0.62*** 11.51 0.16
Africa -0.23*** -3.90 -0.077 -0.18*** -3.08 -0.062
Wald-test joint sign. polit. trust 26.80***
Pseudo R2 0.025 0.027 0.027 0.034
Number of observations 42056 40002 37018 20576
Prob > chi2 0.00 0.00 0.00 0.00
Notes: In the reference group are AGE<30, Man, Single, Lower Middle And Lower Class, Other Employment Status, Risk Taker, Western Europe + USA + Australia. ***, **
and * denote significance at 1%, 5% and 10%, respectively. Marginal effect = highest score (10, never justifiable). The higher the value the lower the justifiability. CEE: Central
Eastern European Countries, FSU: Former Soviet Union Countries.

61 
 
  

Table 3-4 Perceived corruption and political interest

Weighted Ordered Probit Coeff z-Stat. Marg. Coeff. z-Stat. Marg. Coeff. z-Stat. Marg. Coeff. z-Stat. Marg.
(13) (14) (15) (16)
Interest In Politics -0.090*** -12.31 -0.028 -0.087*** -11.80 -0.027 -0.053*** -6.80 -0.016 -0.055*** -5.28 -0.016
Formal Education -0.003 -0.89 -0.001 0.006** 2.13 0.002 0.0003 0.09 0.0001 -0.009** -2.06 -0.003
Age 30-49 -0.027 -1.58 -0.008 -0.037** -2.17 -0.012 -0.058*** -3.22 -0.018 -0.021 -0.90 -0.006
Age 50-64 -0.068*** -3.27 -0.021 -0.073*** -3.43 -0.023 -0.076*** -3.38 -0.023 -0.018 -0.60 -0.005
Age 65+ -0.14*** -5.09 -0.041 -0.14*** -4.97 -0.041 -0.12*** -4.03 -0.034 -0.076* -1.90 -0.021
Female 0.002 0.13 0.001 0.009 0.67 0.003 0.006 0.43 0.002 -0.007 -0.36 -0.002
Married 0.012 0.69 0.004 0.012 0.65 0.004 0.026 1.40 0.008 -0.023 -0.91 -0.007
Widowed -0.022 -0.73 -0.007 -0.036 -1.17 -0.011 -0.008 -0.25 -0.002 -0.096** -2.31 -0.027
Divorced 0.076** 2.31 0.024 0.069** 2.02 0.022 0.068* 1.94 0.021 0.11** 2.21 0.031
Separated 0.075 1.63 0.024 0.070 1.49 0.022 0.075 1.55 0.023 0.050 0.79 0.015
Upper Class -0.018 -0.36 -0.006 0.033 0.63 0.010 -0.14** -2.21 -0.038
Upper Middle Class -0.23*** -13.16 -0.067 -0.19*** -10.40 -0.054 -0.16*** -6.40 -0.044
Selfemployed 0.037 1.51 0.012 0.050** 2.05 0.016 0.016 0.61 0.005 0.039 1.27 0.011
Risk Averse 0.017 1.25 0.005 0.005 0.36 0.002 0.001 0.07 0.000 0.007 0.35 0.002
Urbanization 0.034*** 13.61 0.011 0.035*** 13.56 0.011 0.025*** 9.16 0.007 0.005 1.45 0.001
Church Attendance 0.003 0.85 0.001 0.004 1.27 0.001 0.019*** 5.08 0.006 0.010** 1.99 0.003
Legal System -0.15*** -17.05 -0.047 -0.16*** -13.15 -0.046
Government -0.12*** -12.17 -0.037 -0.13*** -9.81 -0.037
Parliament -0.17*** -15.59 -0.050 -0.15*** -11.07 -0.044
Voice And Accountability. -0.16*** -11.480 -0.047
CEE And FSU 0.92*** 57.45 0.29 0.88*** 53.05 0.28 0.91*** 52.68 0.28 0.97*** 24.21 0.29
Latin America 0.64*** 28.98 0.223 0.60*** 26.60 0.21 0.62*** 26.29 0.21 0.57*** 14.46 0.18
Asia 0.55*** 21.00 0.20 0.53*** 19.84 0.19 0.77*** 26.22 0.28 0.90*** 20.71 0.32
Africa 1.27*** 21.93 0.47 1.25*** 20.53 0.47 1.32*** 19.86 0.49
Wald-test joint sign. polit. trust 1801.40
Pseudo R2 0.058 0.059 0.090 0.11
Number of observations 38646 37245 34752 19136
Prob > chi2 0.00 0.00 0.00 0.00
Notes: In the reference group are Age<30, Man, Single, Lower Middle And Lower Class, Other Employment Status, Risk Taker, Western Europe + USA + Australia. ***, ** and
* denote significance at 1%, 5% and 10%, respectively. Marginal effect = highest score (4). The higher the value the lower the justifiability. CEE: Central Eastern European
Countries, FSU: Former Soviet Union Countries.

62 
 
  

Table 3-5 Justifiability of corruption and important of politics in life

Weighted Ordered Probit Coeff z-Stat. Marg. Coeff. z-Stat. Marg. Coeff. z-Stat. Marg. Coeff. z-Stat. Marg.
(17) (18) (19) (20)
Importance Of Politics 0.023*** 2.95 0.007 0.027*** 3.33 0.008 0.026*** 3.10 0.008 0.055*** 4.90 0.017
Formal Education 0.001 0.22 0.0002 0.002 0.71 0.001 0.006* 1.79 0.002 0.004 0.84 0.001
Age 30-49 0.18*** 9.98 0.058 0.19*** 9.83 0.058 0.192*** 9.82 0.060 0.17*** 6.67 0.054
Age 50-64 0.38*** 16.00 0.11 0.39*** 15.99 0.11 0.396*** 15.55 0.115 0.36*** 10.77 0.11
Age 65+ 0.51*** 15.53 0.14 0.53*** 15.59 0.14 0.531*** 15.08 0.143 0.53*** 11.32 0.14
Female 0.14*** 9.45 0.043 0.14*** 9.31 0.043 0.136*** 8.97 0.043 0.15*** 7.49 0.047
Married 0.13*** 6.74 0.041 0.12*** 6.30 0.039 0.126*** 6.31 0.040 0.12*** 4.65 0.039
Widowed 0.15*** 4.26 0.046 0.14*** 3.75 0.042 0.134*** 3.47 0.041 0.097*** 1.99 0.030
Divorced 0.017 0.46 0.005 0.006 0.15 0.002 0.009 0.22 0.003 0.012 0.25 0.004
Separated 0.068 1.29 0.021 0.075 1.40 0.023 0.084 1.54 0.026 0.064 0.90 0.020
Upper Class -0.20*** -3.70 -0.067 -0.186*** -3.32 -0.063 -0.15** -2.04 -0.048
Upper Middle Class -0.019*** -0.93 -0.006 -0.031 -1.51 -0.010 -0.028 -0.97 -0.009
Selfemployed -0.060*** -2.28 -0.019 -0.064** -2.38 -0.021 -0.055** -2.00 -0.018 -0.089** -2.55 -0.029
Risk Averse 0.075*** 4.81 0.023 0.071*** 4.47 0.022 0.073*** 4.42 0.023 0.077*** 3.48 0.024
Urbanization -0.006** -2.30 -0.002 -0.006** -2.05 -0.002 -0.004 -1.47 -0.001 -0.007* -1.84 -0.002
Church Attendance 0.011*** 2.80 0.003 0.008** 1.97 0.002 0.006 1.38 0.002 -0.001 -0.12 0.000
Legal System 0.040*** 4.09 0.013 0.070*** 5.32 0.022
Government -0.009 -0.80 -0.003 0.035** 2.36 0.011
Parliament 0.018 1.51 0.006 0.008 0.54 0.003
Voice And Accountability. -0.38*** -19.72 -0.12 -0.373*** -18.65 -0.119 0.13*** 8.56 0.042
CEE And FSU -0.38*** -20.48 -0.12 -0.44*** -17.73 -0.15 -0.425*** -16.85 -0.146 -0.23*** -4.98 -0.072
Latin America -0.44*** -18.23 -0.15 0.35*** 9.83 0.10 0.367*** 9.50 0.103 -0.027 -0.62 -0.009
Asia 0.21*** 6.56 0.063 -0.22*** -3.60 -0.073 -0.245*** -3.91 -0.084 0.63*** 11.67 0.16
Africa -0.25*** -4.28 -0.085
Wald-test joint sign. polit. trust 26.16***
Pseudo R2 0.025 0.027 0.027 0.034
Number of observations 41631 39614 36720 20410
Prob > chi2 0.00 0.00 0.00 0.00
Notes: In the reference group are Age<30, Man, Single, Lower Middle And Lower Class, Other Employment Status, Risk Taker, Western Europe + USA + Australia. ***, ** and *
denote significance at 1%, 5% and 10%, respectively. Marginal effect = highest score (10, never justifiable). The higher the value the lower the justifiability. CEE: Central Eastern
European Countries, FSU: Former Soviet Union Countries.

63 
 
  

Table 3-6 Perceived corruption and importance of politics in life

Weighted Ordered Probit Coeff z-Stat. Marg. Coeff. z-Stat. Marg. Coeff. z-Stat. Marg. Coeff. z-Stat. Marg.
(21) (22) (23) (24)
Importance Of Politics -0.047*** -6.59 -0.015 -0.045*** -6.23 -0.014 -0.001 -0.15 -0.0003 -0.031*** -3.01 -0.009
Formal Education -0.007** -2.52 -0.002 0.002 0.70 0.001 -0.004 -1.27 -0.001 -0.011** -2.54 -0.003
Age 30-49 -0.036** -2.16 -0.011 -0.048*** -2.77 -0.015 -0.067*** -3.71 -0.020 -0.023 -0.99 -0.007
Age 50-64 -0.085*** -4.04 -0.026 -0.090*** -4.19 -0.028 -0.091*** -4.08 -0.027 -0.021 -0.71 -0.006
Age 65+ -0.15*** -5.62 -0.045 -0.15*** -5.52 -0.046 -0.13*** -4.52 -0.038 -0.078* -1.94 -0.022
Female 0.018 1.38 0.006 0.024* 1.84 0.008 0.020 1.45 0.006 0.002 0.08 0.000
Married 0.011 0.62 0.003 0.011 0.61 0.003 0.026 1.36 0.008 -0.026 -1.02 -0.007
Widowed -0.033 -1.11 -0.010 -0.045 -1.45 -0.014 -0.016 -0.48 -0.005 -0.11*** -2.64 -0.030
Divorced 0.076** 2.30 0.024 0.068** 2.02 0.022 0.067* 1.89 0.021 0.101** 2.12 0.030
Separated 0.060 1.29 0.019 0.054 1.13 0.017 0.061 1.24 0.019 0.031 0.49 0.009
Upper Class -0.026 -0.51 -0.008 0.022 0.43 0.007 -0.16** -2.52 -0.042
Upper Middle Class -0.24*** -13.75 -0.070 -0.20*** -10.88 -0.056 -0.17*** -6.72 -0.046
Selfemployed 0.032 1.31 0.010 0.047* 1.91 0.015 0.015 0.58 0.005 0.036 1.17 0.011
Risk Averse 0.018 1.31 0.006 0.005 0.37 0.002 0.003 0.22 0.001 0.005 0.28 0.002
Urbanization 0.034*** 13.43 0.011 0.035*** 13.43 0.011 0.024*** 8.97 0.007 0.005 1.40 0.001
Church Attendance 0.004 1.02 0.001 0.005 1.42 0.002 0.019*** 5.12 0.006 0.012** 2.27 0.003
Legal System -0.15*** -16.99 -0.047 -0.16*** -13.13 -0.046
Government -0.12*** -12.18 -0.037 -0.13*** -9.70 -0.037
Parliament -0.17*** -16.03 -0.052 -0.16*** -11.15 -0.045
Voice And Accountability. -0.16*** -11.480 -0.047
CEE And FSU 0.93*** 57.43 0.29 0.88*** 52.91 0.28 0.92*** 52.73 0.28 0.97*** 24.02 0.29
Latin America 0.68*** 31.10 0.24 0.64*** 28.49 0.22 0.64*** 27.69 0.22 0.60*** 15.13 0.19
Asia 0.58*** 21.99 0.21 0.56*** 20.87 0.20 0.79*** 27.02 0.29 0.91*** 21.09 0.32
Africa 1.31*** 22.13 0.49 1.28*** 20.78 0.48 1.33*** 19.87 0.49
Wald-test joint sign. polit. trust 1828.52
Pseudo R2 0.057 0.058 0.090 0.11
Number of observations 38277 36899 34476 18979
Prob > chi2 0.00 0.00 0.00 0.00
Notes: In the reference group are Age<30, Man, Single, Lower Middle And Lower Class, Other Employment Status, Risk Taker, Western Europe + USA + Australia. ***, ** and *
denote significance at 1%, 5% and 10%, respectively. Marginal effect = highest score (4). The higher the value the lower the justifiability. CEE: Central Eastern European Countries,
FSU: Former Soviet Union Countries.

64 
 
 

However, looking at the variable perceived corruption, the coefficient is negative and
statistically significant with marginal effects varying between 2.2 and 4.9 percentage points
and increasing with age. Thus, the level of perceived corruption decreases with an increase in
age. Furthermore, the results also indicate that there are gender differences. Being female
rather than male increases the probability of a person stating that accepting a bribe is never
justifiable. This result indicates that women’s norms regarding bribery differs from the norms
held by men. However, the perceived corruption coefficient is positive and statistically
significant, indicating that women perceive corruption to be more widespread than men.
Moreover, married people are more sensitive to the social norm regarding bribery than
individuals with any other marital status. However, the coefficient is only statistically
significant for the estimations using justifiability of corruption as the dependent variable. We
observe that being in a higher income class leads to a lower justifiability of corruption and
surprisingly, we also observe a negative correlation when focusing on perceived corruption.
Self-employed people are more tolerant towards corruption and perceive corruption to be
more common. Being risk averse is correlated with a lower justifiability of corruption. The
coefficient is statistically significant in all the regressions. On the other hand, we don’t
observe a statistically significant relationship between perceived corruption and political
interest. In line with our expectations we also observe a negative relationship between
urbanization and justifiability of corruption and a negative between urbanization and
perceived corruption. The results also show that church attendance is enforcing the norm of
compliance. The correlation between church attendance and justifiability of corruption is
positive, although the coefficient is not always statistically significant and the marginal
effects are not that large. Not surprisingly, we find strong regional differences. Moreover,
inhabitants of CEE and FSU, Latin America and Africa 42 countries report a higher
justifiability of bribing when compared to the reference group. Thus, the findings show that
the social norm regarding bribery is unambiguously higher in Western Europe, USA and
Australia. We also observe that the reference group has the lowest perceived level of
corruption.
In sum, the estimation results presented in Table 3-1 to 3-6 suggest that political interest
matters, controlling in a multivariate analysis for additional factors. This is consistent with
the theoretical argument developed in Section II. It is interesting to observe the importance of
political trust and voice and accountability in this context.
                                                            
42
 As mentioned, Africa only covers the country Nigeria. This explains why in some regressions Africa is no
longer reported (variable not collected this survey).  

65 
 
 

It is reasonable to question the direction of causality in the results, and therefore our main
hypothesis can be criticized. One can argue that a higher level of perceived corruption may
lead to frustration with the lack of representative administration, and therefore to a lower
willingness to invest in the maintenance of political interest. Similarly, a higher justifiability
of corruption may induce individuals to be less interested in what happens in politics,
although the causality problem may be more severe when focusing on individuals’ perceived
level of corruption. Thus, to evaluate the direct effect of political interest on corruption it is
useful to investigate any potential causality problems through use of an instrumental variable
technique. We present in Table 3-7 six 2SLS estimations providing also detailed diagnostic
tests to check the robustness of the results. For simplicity (and due to less causality problems)
we will work with the second regression in the previous tables. The results remain robust
when considering a broader specification. In the first three specifications we focus on the
justifiability of corruption and the last three on the perceived corruption. The results indicate
that all three political interest proxies are statistically significant with a positive sign.
Political interest is instrumented through an index that measures the importance of private
interests43. We report the first-stage regression results of the instrumental variables and the F-
tests of the exclusion of the instruments. Overall, the instrument used is effective in
explaining political interest. The instrument is always statistically significant at the 1% level,
as are the F-tests for the instrument exclusion set in the first-stage regressions. On the other
hand, the variable is not correlated with our dependent variable. We also report the Anderson
canonical correlations LR test for the relevance of the instruments. A rejection of the null
hypothesis indicates that the model is identified and that the instruments are relevant (see
Hall, Rudebusch and Wilcox 1996). Moreover, we also report the Anderson-Rubin test that
the endogenous variables are jointly statistically significant. The test has the advantage of
being robust to the presence of weak instruments. Table 3-7 reports that in all cases the
Anderson canonical correlations LR test shows rejection of the null hypothesis, which
indicates that the models are identified and that the instruments are relevant. The Anderson-
Rubin test is also statistically significant. In all the cases, this test fails to reject the null
hypothesis that our instruments are valid. Thus, the 2SLS specifications also provide support
that political interest matters.

                                                            
43
Mean value of the following three questions: Please say, for each of the following, how important it is in your
life: family, friends, leisure (very 4), (rather 3), not very (2), not at all (1).

66 
 
  

Table 3-7 2SLS results

Weighted Ordered Probit Justifiability Of Corruption Perceived Corruption
Coeff. t-stat. Coeff. t-stat. Coeff. t-stat. Coeff. t-stat. Coeff. t-stat. Coeff. t-stat.
(25) (26) (27) (28) (29) (30)
Political Discussion 1.52*** 4.33 -0.35*** -2.74
Interest In Politics 1.00** 4.32 -0.23*** -2.69
Importance Of Politics 0.32*** 4.67 -0.080*** -2.76
Formal Education -0.077*** -3.95 -0.086*** -4.00 -0.021*** -2.90 0.017** 2.40 0.019** 2.42 0.005 1.53
Age 30-49 0.098* 1.87 0.13*** 2.63 0.25*** 7.86 -0.0001 0.00 -0.003 -0.17 -0.033*** -2.67
Age 50-64 0.22*** 2.82 0.25*** 3.31 0.45*** 11.98 -0.008 -0.26 -0.008 -0.28 -0.057*** -3.54
Age 65+ 0.42*** 7.01 0.35*** 4.77 0.52*** 11.97 -0.083*** -3.30 -0.060* -1.96 -0.11*** -5.13
Female 0.45*** 6.71 0.42*** 6.85 0.22*** 9.36 -0.039 -1.62 -0.035 -1.55 0.010 0.94
Married 0.098*** 2.58 0.12*** 3.26 0.16*** 5.04 0.027* 1.88 0.020 1.50 0.013 1.02
Widowed 0.18*** 3.55 0.14*** 2.76 0.17*** 3.84 -0.024 -1.02 -0.016 -0.70 -0.026 -1.17
Divorced -0.043 -0.66 -0.005 -0.08 0.024 0.42 0.061** 2.40 0.058** 2.33 0.051** 2.14
Separated 0.12 1.33 0.080 0.92 0.15* 1.93 0.038 1.11 0.057 1.68 0.034 1.01
Upper Class -0.31*** -3.02 -0.35*** -3.52 -0.30*** -3.30 0.012 0.31 0.009 0.23 -0.006 -0.15
Upper Middle Class -0.10*** -2.80 -0.14*** -3.35 -0.037 -1.32 -0.15*** -10.73 -0.14*** -8.80 -0.17*** -13.26
Selfemployed -0.057 -1.23 -0.063 -1.38 -0.055 -1.31 0.034* 1.87 0.032* 1.79 0.033* 1.89
Risk Averse 0.10*** 4.20 0.13*** 5.21 0.11*** 5.01 0.000 -0.01 -0.008 -0.78 -0.003 -0.26
Urbanization -0.010** -2.09 -0.004 -0.92 -0.006 -1.50 0.025*** 13.12 0.024*** 12.81 0.024*** 13.23
Church Attendance 0.016*** 2.65 0.005 0.88 0.007 1.26 0.002 0.82 0.004 1.63 0.004* 1.69
CEE And FSU -0.39*** -14.01 -0.22*** -4.01 -0.36*** -12.53 0.64*** 54.03 0.60*** 28.09 0.63*** 47.63
Latin America -0.32*** -3.24 -0.19 -1.49 -0.65*** -16.68 0.38*** 10.82 0.35*** 7.66 0.46*** 28.02
Asia 0.39*** 7.87 0.42*** 7.86 0.25*** 8.30 0.35*** 15.60 0.36*** 15.08 0.41*** 21.15
Africa -0.10 -1.08 -0.13 -1.32 -0.34*** -4.10 0.85*** 21.08 0.86*** 21.78 0.91*** 23.21
First stage regressions:
Political Interest
Private Interests 0.079*** 9.87 0.12*** 10.34 0.36*** 30.15 0.083*** 10.03 0.12*** 10.23 0.36*** 29.39
F-Test of excluded instruments 97.46*** 107.02*** 909.31*** 100.60*** 104.73*** 863.69***
Anderson canon. corr. likelihood ratio stat. 121.05*** 134.29*** 1166.46*** 213.87*** 131.41*** 1104.95***
Anderson-Rubin test 24.02*** 23.09*** 22.29*** 7.92*** 7.46*** 7.64***
Number of observations 38888 39212 39008 36232 36530 36354
Prob > F 0.00 0.00 0.00 0.00 0.00 0.00
Notes: In the reference group are Age<30, Man, Single, Lower Middle And Lower Class, Other Employment Status, Risk Taker, Western Europe + USA + Australia. ***, **
and * denote significance at 1%, 5% and 10%, respectively. CEE: Central Eastern European Countries, FSU: Former Soviet Union Countries.

67 
 
 

Table 3-8 Justifiability of corruption in Switzerland

Weighted Ordered Probit Coeff z-Stat. Marg. Coeff. z-Stat. Marg. Coeff. z-Stat. Marg.
(31) (32) (33)
Political Discussion 0.177** 2.32 0.049
Interest In Politics 0.12* 1.92 0.032
Importance Of Politics 0.20*** 3.30 0.053
Formal Education -0.039 -1.28 -0.011 -0.040 -1.30 -0.011 -0.041 -1.31 -0.011
Age 30-49 0.300** 2.26 0.082 0.30** 2.26 0.083 0.27** 2.04 0.073
Age 50-64 0.43** 2.48 0.11 0.42** 2.42 0.10 0.45*** 2.61 0.11
Age 65+ 0.59*** 2.90 0.14 0.57*** 2.79 0.13 0.55*** 2.69 0.13
Female 0.53*** 5.49 0.15 0.53*** 5.48 0.15 0.52*** 5.42 0.14
Married 0.26* 1.96 0.073 0.27** 2.00 0.075 0.29** 2.14 0.080
Widowed -0.066 -0.28 -0.019 -0.070 -0.30 -0.020 -0.057 -0.24 -0.016
Divorced -0.092 -0.46 -0.026 -0.063 -0.32 -0.018 -0.001 0.00 0.000
Separated -0.089 -0.25 -0.026 -0.053 -0.15 -0.015 -0.024 -0.06 -0.007
Selfemployed -0.051 -0.29 -0.015 -0.037 -0.21 -0.010 -0.045 -0.24 -0.012
Risk Averse 0.11 1.07 0.030 0.11 1.05 0.029 0.10 0.99 0.027
Urbanization -0.005 -0.20 -0.001 -0.005 -0.20 -0.001 -0.013 -0.53 -0.004
Church Attendance -0.008 -0.30 -0.002 -0.008 -0.29 -0.002 -0.022 -0.80 -0.006
Legal System 0.12* 1.69 0.032 0.11 1.58 0.030 0.13* 1.83 0.034
Democracy 0.030 0.79 0.008 0.027 0.72 0.008 0.013 0.35 0.004
Pseudo R2 0.049 0.048 0.055
Number of observations 1086 1086 1075
Prob > chi2 0.00 0.00 0.00
Notes: In the reference group are Age<30, Man, Single, Other Employment Status, Risk Taker, ***, ** and * denote significance at 1%, 5% and
10%, respectively.

68 
 
 

                       Table 3-9 Perceived corruption in Switzerland

Weighted Ordered Probit Coeff z-Stat. Marg. Coeff. z-Stat. Marg. Coeff. z-Stat. Marg.
(34) (35) (36)
Political Discussion -0.067 -0.97 -0.008
Interest In Politics -0.114** -2.19 -0.014
Importance Of Politics -0.10** -2.00 -0.012
Formal Education -0.057** -2.30 -0.007 -0.049* -1.94 -0.006 -0.053** -2.12 -0.007
Age 30-49 0.066 0.52 0.008 0.078 0.62 0.010 0.096 0.75 0.012
Age 50-64 -0.050 -0.33 -0.006 -0.022 -0.15 -0.003 -0.032 -0.21 -0.004
Age 65+ -0.25 -1.42 -0.027 -0.21 -1.23 -0.024 -0.20 -1.12 -0.022
Female -0.25*** -2.88 -0.031 -0.27*** -3.00 -0.033 -0.25*** -2.88 -0.031
Married 0.025 0.21 0.003 0.016 0.14 0.002 0.007 0.06 0.001
Widowed -0.45* -1.91 -0.042 -0.46* -1.92 -0.042 -0.46* -1.92 -0.041
Divorced -0.098 -0.44 -0.011 -0.13 -0.58 -0.015 -0.12 -0.52 -0.013
Separated -0.27 -0.62 -0.028 -0.27 -0.61 -0.027 -0.31 -0.68 -0.030
Selfemployed -0.010 -0.07 -0.001 -0.014 -0.09 -0.002 -0.057 -0.40 -0.007
Risk Averse -0.18* -1.90 -0.021 -0.18*** -1.97 -0.022 -0.16* -1.71 -0.018
Urbanization -0.036 -1.54 -0.005 -0.034 -1.43 -0.004 -0.037 -1.56 -0.005
Church Attendance -0.051** -2.21 -0.006 -0.049** -2.13 -0.006 -0.047** -2.00 -0.006
Legal System -0.48*** -7.30 -0.060 -0.48*** -7.31 -0.059 -0.47*** -7.11 -0.057
Direct Democracy -0.10*** -3.00 -0.013 -0.098*** -2.87 -0.012 -0.099*** -2.88 -0.012
Pseudo R2 0.077 0.080 0.077
Number of observations 1019 1018 1008
Prob > chi2 0.00 0.00 0.00
Notes: In the reference group are Age<30, Man, Single, Other Employment Status, Risk Taker, ***, ** and * denote significance at 1%, 5% and
10%, respectively.

69 
 
 

3.4.2 Within-Country Evidence

In general, drawing conclusions from cross-cultural comparisons is difficult because not all
features specific to a country can always be controlled in a satisfactory manner. Thus, we
extend our study, focusing on within-country data from Switzerland at the state (cantonal)
level to investigate the impact of tax morale and institutional quality. As mentioned
previously, analyses of Swiss data are interesting because Switzerland’s institutions are not
homogeneous. The degree of institutionalized political participation rights varies strongly
between the 26 Swiss cantons. In line with the previous regressions, we are going to
investigate the third wave. This is the latest available data set for Switzerland as the country
did not participate in the fourth wave. Table 3-8 and 3- 9 present the results. We make one
small change to the specification structure: instead of voice and accountability we are going
to include a democracy index 44 measured at the cantonal level. The degree of direct
democratic participation rights is measured with an index developed by Stutzer (1999). To
maximize the number of available observations we first run regressions without the variable
income45 as this variable would reduce the number of observations by almost 200 subjects.
However, in a second step we are going to discuss the results of regressions where we include
income as a control variable.
We observe that political interest also matters for Switzerland, and the quantitative effects
are quite substantial. For example, increase in the political discussion scale by one unit raises
the probability of stating that corruption is never justifiable by 4.9 percentage points. The
effect is even more relevant in further specifications. For example, if we include income in
the regression, we observe the coefficient for interest in politics in Table 3-8 is statistically
                                                            
44
 It should be noticed that the Swiss World Value Survey was not random-random but quota-random, based on
a random sample of communes and then on quotas in terms of sex, age, etc. in the selected communes. Thus, the
smallest cantons are not necessarily represented (not represented are: Appenzell a. Rh., Glarus, Jura, Nidwalden,
Uri, and Zug). On the other hand, the ISSP data set contains all 26 cantons.
45
Here is a scale of incomes (1-10). We would like to know in what group your household is, counting all
wages, salaries, pensions and other incomes that come in. Just give the letter of the group your household falls
into, before taxes and other deductions.
1. Less than 20’000 Swiss Francs
2. 20’000-26’999
3. 27’000-31’999
4. 32’000-37’999
5. 38’000-44’999
6. 45’000-51’999
7. 52’000-59’999
8. 60’000-69’999
9. 70’000-89’999
10. More than 90’000

70 
 
 

significant at the 1% level (t-value=2.15). Interestingly, we observe that a higher level of

direct democracy is positively correlated with a lower justifiability of corruption. We also
observe the tendency that trust in the legal system matters, particularly when focusing on the
perceived level of corruption. We have only included this political trust variable in the
specification as it had the strongest impact on corruption in the previous six tables. In
addition, it allows us to avoid a decrease in the number of observations. As in the previous
approach, we also observe that age, gender and marital status (being married) matter for
justifiability of corruption. On the other hand, risk attitudes are relevant when focusing on the
perceived corruption rather than on the justifiability of corruption. Similarly, urbanization
and self-employment status are not relevant at all. Moreover, religion is only relevant when
focusing on perceived corruption. It is also worthwhile to note that we did not find a
significant relationship between income and political interest in Switzerland. Finally, in line
with the previous findings we observe that overall, formal education is less relevant than
informal education or political interest. The coefficient is only statistically significant in
Table 3-9 and the marginal effects are below the values found for political interest. Thus,
here we find additional support that human capital is mainly relevant in a specialized form.

3.5 Conclusion

In recent years the topic of corruption has attracted a great deal of attention. However, there
is still a lack of empirical evidence about the determinants of corruption at the micro level.
Moreover, there are still interesting variables that have not been investigated in the past. This
empirical study analyses a cross-section of individuals using data from the World Values
Survey, investigating the determinants of corruption with two dependent variables: perceived
corruption and the justifiability of corruption. Both variables are strongly correlated with
other commonly used measurements of corruption such as the Transparency International
Corruption Perception Index, the International Country Risk Guide Index or the Quality of
Government Corruption Index. The major aim in the chapter was to investigate whether
political interest matters. Despite economists’ increasing interest in the determinants of
corruption, this factor has been widely neglected in the literature. Thus, it was highly relevant
that we investigated empirically the possible connections between political interest and
corruption.
To check the robustness we explored the relationship between political interest and
corruption using three different proxies of political interest. The results clearly indicate that
use of an education variable does not reflect the accumulation and stock of human capital. A

  71 
 

further strength of the chapter is to focus not only on the perceived level of corruption, (as is
commonly the case in the current literature), but also to consider the justifiability of bribery.
Moreover, we have provided cross-country and within-country evidence at the micro level,
controlling for the state of relevant institutional conditions. In this study, our focus on
political interest required that we control for voice and accountability and direct democratic
rights.
The econometric estimates also suggest that strength of social norms regarding bribery is
higher and the perceived level of corruption lower in the reference group (region Western
Europe, USA and Austria) compared to CEE and FSU countries, Latin America, Asia and
Africa.
All in all, the results suggest some interesting political implications. Increasing the level
of interest in politics may help to reduce the level of corruption in a society. The results also
suggest that it may be important to place more emphasis on institutions that enhance voice
and accountability and democratic participation rights. This helps to increase individuals’
social norm and perception of compliance. Thus, the results presented in this chapter mirror
those in previous studies and underscore the importance of accountability as an essential
aspect for the efficient functioning of a government and the existing institutional architecture.
However, understanding how corruption can be reduced and how government can foster
political interest remains a fruitful field for further research.
 

  72 
 

Chapter Four Democracy and Corruption: Cross-country Evidence46

4.1 Introduction

Rose-Ackerman (1999, p.127) stresses: “Democracies based on strong legal foundation

provide a stable framework for economic activity. For this framework to operate efficiently,
however, politicians must seek reelection and must feel insecure about their prospects, but
not too insecure. This leads to a ‘paradox of stability’. Too much security of tenure can
further corrupt arrangements. Too much insecurity can have the same effect.” Interestingly,
prior literature on the relationship between democracy and corruption provides mixed
evidence. Ades and Di Tella (1999) find that fewer political rights are correlated to low
corruption levels. Ades and Di Tella (1997) and Fisman and Gatti (2002), however, fail to
find any substantial effects on corruption of political rights and civil liberties respectively.
Triesman (2000) does not find a significant direct effect of democracy on corruption either,
however he documents that the duration of democracy significantly reduces corruption. On
the other hand, Goldsmith (1999) reports that political democratization is associated with a
lower degree of political corruption. Chowdhury (2004) also finds that the presence of
democracy can reduce the level of corruption significantly. From a slightly different angle,
Bohara, Mitchell and Mittendorff (2004) highlight that citizens’ participation in competitive
elections increases the control of corruption. Recently, Goel and Nelson (2005) provide
empirical evidence that less democratic countries always have a higher incidence of
corruption. Emerson (2006) also shows that more political rights have a depressing effect on
corruption. Billger and Goel (2009), however, document in their quantile regressions that
democracy significantly reduces corruption only in the most corrupt countries.
Instead of exclusively testing the linear democracy-corruption association, Montinola and
Jackman (2002) provide evidence that the effect of political competition on corruption is
nonlinear. Corruption is lower in dictatorial countries than in ones partially democratized. It
will, however, decline with the democratic level after a threshold. Sung (2003), on the other
hand, reports that the cubic function best fits the cross-national data on the relationship
between democracy and corruption.
Inconsistent empirical results suggest the demand of theoretical guidance. Unfortunately
there is little theoretical evidence on the relationship except Mohatdi and Roe (2003). They
                                                            
46
 This chapter has been submitted to Journal of Development Economics 

  73 
 

build a two-sector endogenous growth model where agents switch between rent seeking and
production. A reversed-U effect of democratization on corruption is then derived: rents and
hence levels of corruption increase in the young democracies but decrease in the mature
democracies. However, it is questionable whether their longitudinal section mechanism,
though enlightening, is able to explain existing cross-section evidence of the nexus between
democracy and corruption.
Currently, two empirical articles more related to our study have emerged. Rock (2007)
utilizes the instrument variable approach to empirically show an inverted U relationship
between democracy and corruption. Saha et al. (2009), however, perform fixed-effect
regressions to find that economic freedom always reduces corruption, while democracy
increases corruption under weak economic freedom and decreases corruption under strong
economic freedom. Both papers document the complexity of the nexus between democracy
and corruption. However, the corruption indices that both papers use are actually not suitable
for panel analysis. This will be discussed later. Furthermore the economic freedom index in
Saha et al. (2009) contains eight components ranging from micro business freedom to macro
monetary freedom. This has the disadvantage that it is difficult to identify a clear mechanism
for the interactions between democracy, economic freedom and corruption from their results.
Indeed Goel and Nelson (2005) have found that different components of economic freedom
influence corruption in different ways.
This chapter attempts to clarify miscellaneousness in past research with a uniform
framework, and therefore contribute to the literature on the linkage between democracy and
corruption. We first develop a theoretical model that incorporates the effects of property
rights protection and income distribution into the relationship between democracy and
corruption. The final effect of democracy on corruption depends on the combination of
property rights protection and income distribution in a country. For example, Uslaner (2008)
stresses that the transition to democracy and a market economy in Eastern Europe brought
great instability and rising levels of inequality. Then we utilize a large panel sample including
about 108 countries from 1995-2006 to examine the conclusion of the theory. With two
alternative measures of democracy, our empirical analysis adopts the fixed-effect approach
and then the instrumental variable approach to validate important aspects of the theoretical
model. We find in our empirical analysis that the effect of democracy on corruption
obviously depends on the level of property rights protection and income equality. The
democracy’s effect is different under different property right protection and income equality
conditions. The finding is particularly robust for property rights protection. We therefore

  74 
 

provide an insightful mechanism for the nexus between democracy and corruption, both
theoretically and empirically.
The rest of the chapter is organised as follows. Section 4.2 presents the theoretical model.
Section 4.3 provides the corresponding empirical analysis. Section 4.4 concludes with some
comments.

4.2 Theoretical Model

Current conflicting linear and nonlinear evidence of the linkage between democracy and
corruption may imply that further factors need to be taken into account in order to thoroughly
understand the relationship. From our point of view, the impact of democracy on corruption
is conditional on income distribution and property rights protection, which can be seen in the
model that follows.
According to our knowledge, there are only two recent theoretical studies related to ours.
As discussed above, Mohatdi and Roe (2003) modelled the association between democracy
and corruption. Alesina and Angeletos (2005), with a non-overlapping-generation model,
document the existence of multiple steady states in the levels of inequality, redistribution and
corruption. It seems that no work, however, has explicitly explored the nexus between
democracy, income distribution, property rights and corruption. We are therefore attempting
to fill the void.
Mohatdi and Roe (2003) assume that “democracy is about the flow of information and
access to the government” (p. 450). We, however, follow Dahl (1974) to stress that from a
constitutional perspective the essence of democracy is electoral participation and political
competition. In line with Murphy et al. (1993) and Alesina and Angeletos (2005), we treat
corruption as a rent-seeking activity.
The political economy mechanism provided here is closely related to Persson and
Tabellini (1994). The pivotal voter in a country determines the redistribution policy. The
redistributive decision therefore hinges on the difference between the income of the pivotal
voter and the average income in the society. Unequal societies where the income of the
pivotal voter is lower than the average income consequently have more redistribution from
the rich to the poor than equal ones. Rent-seeking activities and hence corruption emerges in
the allocation of the redistributive tax revenue. Furthermore, in the absence of property rights
protection, the rich are likely to gain more from appropriation of the redistributive tax
revenue than the poor though all have the access to the appropriation (rent-seeking)
technology (Gradstein, 2007). Redistribution thus cannot mitigate income inequality as

  75 
 

expected. As a result, high levels of corruption and income inequality might be self-
sustaining in democracies with unsecure property rights. Oligarchies, however, may avoid
this situation because their “pivotal agents” are often richer than the average. The situation
can also be mitigated or even eliminated in democracies with equal income distribution
and/or secure property rights. In summary, it can be seen that democracy in our setting may
breed corruption due to intensive redistribution, especially in countries that lack income
equality and property rights protection. Below we will discuss this in detail. It is worth noting
that unlike some prior studies, we treat democracy here as an exogenous variable in order to
focus on studying the relationship between democracy and corruption.
4.2.1 Model
We consider a non-overlapping-generations model where altruistic individuals with
monotonic preferences live only for one period. Each generation comprises a large number of
individuals distributed uniformly over 0,1 . Similar to Gradstein (2007), each member in
generation has the following utility:
4‐1
where is his own consumption, is the income in next period accrued to his child. The
budget constraint is,
4‐2
where is the income of individual from his parent, is his productive capital and
is his unproductive capital in rent seeking. For convenience we further assume
4‐3
where is the average income of generation . , and its distribution therefore
indicates the degree of income inequality in the model economy. The production function
without government is
, 0, 0 1. 4‐4
which exhibits diminishing returns to scale.
Following Alesina and Angeletos (2005), we assume that the government levies a flat tax
on individual capital to fund a lump-sum transfer across all individuals. The tax rate is
which has been decided by a prior process of voting. Then the sum of the transfer
is . However, the distribution of the transfer among individuals is determined by
rent seeking activities. Similar to Sonin (2003) and Gradstein (2007), we suppose that the
share of the transfer which the individual can grab is

  76 
 

4 5

which implies that given the level of public property rights protection 0), the share of
transfer an individual can get increases with his own input and decreases with the competing
inputs of the others. According to Alesina and Angeletos (2005), the corruption level is
plausibly assumed to increases with the amount of transfer.
Then the net capital endowment of individual after redistribution is given by productive
and unproductive investments:

1 4 6

And the output produced by individual in period t+1 is

1 4 7

Therefore the utility of individual can be expressed as following,

1 4 8

Similar to Sonin (2003) and Gradstein (2007), we assume that and

are exogenous to any individual since there are numerous individuals in each generation.
4.2.2 Economic Equilibrium
Given the policy, each individual makes his optimal decision47.

, ,

. .
In the economic equilibrium, individual hence has
1 1
4 9
2 1
1
4 10
2 1
1 1
4 11
2 1
                                                            
47
 Following the spirit of Gradstein (2007),  we do not include taxation into the budget constraint because the
government does not consume any in our simple model. And taxation here is a component of the technology 4-7
that agents employ. There are actually two stages in the technology. In the first stage agents obtain their net
capital endowment by paying taxes and rent seeking. Agents then produce output with their net capital
endowment. 

  77 
 

where we let

Here we can easily find that and decrease while increases when rises,
which is consistent with prior findings. The optimal utility of individual hence is
1
2 1 ln 4 12

where denotes the constant term,

2 1 ln 2 1 2
4.2.3 Political Equilibrium
The optimal tax rate to individual must satisfy

So is solved as
1
4 13
1

If the tax rate is determined by a majority vote in the society, the pivotal voter will
eventually decide the tax rate.
1
4 14
1

Not losing generality, we suppose that , where is the average income of

individuals in time and hence indicates income inequality in the economy. Then
1
4 15
e 1

As in Alesina and Angeletos (2005), the tax rate in our model indicates the corruption
level because the increase in tax rate leads to the enhancement of rent-seeking for tax revenue
and hence the rise of corruption levels, or vice versa. Therefore, we first can conclude from
equation 4-15 that the effect of democracy on corruption depends on property rights
protection, income inequality and economic development of a society. Under some
circumstances democracy may even increase corruption. However, one should note that we
use a simplistic notion of democratization where the process is mainly the delegation of
power by a pivotal voter in the previous enfranchised group (the elite) to another citizen, who

  78 
 

turns out to be the pivotal voter in the extended enfranchised group 48 . The extended
enfranchised group is therefore, as historical evidence suggests, often poorer than the
previous enfranchised group on average. The income level of the pivotal voter, namely the
median voter, in the extended enfranchised group hence is lower than that of the previous
pivotal voter, which means that democratization tends to make a poorer citizen be the pivotal
voter. According to 4-15, democratization may raise the tax rate due to more demand for
redistribution and eventually results in corruption, other things equal. The effect of
democracy will be aggravated in countries with income inequality and inefficient protection
of property. This finding is similar in spirit to Cervellati et al. (2008) who contend that under
high income inequality, democracy causes social conflict while oligarchy can avoid it.
The second conclusion deduced from 4-15 is that ceteris paribus, income inequality
causes corruption. When there is high income inequality in a country: 0, the pivotal
voter’s income is below the average and he therefore tends to choose a high tax rate for more
redistribution, which in turn results in more corruption. When income distribution is,
however, more equal in a society: 0, the pivotal voter with his income above average
will select a low tax rate due to his reluctance in redistribution, which therefore reduces
corruption. In summary, income inequality is a fertile ground for corruption, especially in
democracies. Glaeser et al. (2003) and Sonin (2003) have even claimed that unequal income
distribution is a hotbed of poor governance.
The conclusions above are in a similar spirit to the “tyranny of the majority”, which
Tocqueville (1835) warned may occur in democracies. Tyranny of the majority refers to the
circumstance where the majority might use its strength in a democracy to place its interests
above those of the minority. Specifically, if income distribution is unequal in a country, the
democratic system providing more political equality might lead to excessive redistribution or
even public expropriation, which can weaken property rights protection and cause corruption.
This danger, however, will not appear in the ideal state with prefect equality and freedom as
depicted by Tocqueville.
                                                            
48
 For example, the model disregards positive externalities derived in a direct-democratic environment via
referenda and initiative. Being able to renegotiate and shape the political environments can lead to an increase in
civic virtues. The more citizens can participate in political decision making by popular rights, the more the
“political contract” is based on trust between state/public officials and the citizens which may promote civic
duty. Citizens are in this case treated as “citizens” with extensive rights and obligations (Frey, 2003). The voting
procedure, especially public discussions prior to votes, creates a sense of civic duty, as citizens become aware of
the importance to follow the endogenously generated rules. The possibility to vote promotes social norms of
compliance and therefore may reduce corruption.

  79 
 

It is obviously that property rights protection is negatively correlated with corruption in

our model. This is reasonable since secure property rights limit expropriation (Besley and
Ghatak, 2009). In 4-15 decreases when increases, which suggests that property
rights protection depresses corruption. We can confirm this result from below

0, 4 16
⁄

As we know, is negatively correlated with the level of property rights protection

since rational individuals will invest more on expropriation under weaker protection of
property rights, other things being equal. ⁄ therefore reflects the security of property
rights. Inequality 4-16 shows that secure property rights check corruption, which coincides
with the above finding. We can further deduce from equation 4-15 that property rights
protection may act as a multiplier of democracy and income inequality in terms of influence
over corruption. In addition, democracy might influence property rights protection via the tax
rate. However, the mechanism here is somewhat indirect: voting decides the tax rate
hence affects and , and through the channel it finally influences property rights
protection ⁄ .
It is worth noting that as Cervellati et al. (2008) argue, democracy is neither sufficient nor
necessary for the protection of property rights, although it has often been found to promote
property rights protection. Actually Glaeser et al. (2004) have observed secure property rights
in oligarchies.
Last but not least, based on the inequality derived from 4-15: ⁄ 0, we can
demonstrate the basic fact that the average income level, namely economic development,
controls corruption, which has been verified by most empirical studies (Treisman, 2007).
To summarize, our theoretical finding is that corruption level is jointly determined by
democracy, property rights protection, income and income inequality. Specifically, the effect
of political democracy on corruption depends on other social and economic conditions:
property rights protection, income and income inequality.

4.3 Empirical Evidence

Our empirical analysis employs data from a large sample of 108 countries during the period
1995–2006. We first discuss the methodology and data used and report the results afterwards.

  80 
 

4.3.1 Methodology and Data

Our empirical analysis aims to examine our theoretical predictions to shed new light on the
causal effect of democracy on corruption. We are aware of the potential bias in OLS
identification due to the endogeneity problem that omitted historical factors might influence
both corruption and democracy. We therefore employ two strategies to address the
endogeneity problem. Our first strategy is to use fixed effects regressions to remove the
potential bias since the omitted variables here affecting both corruption and democracy are
generally country-specific, institutional factors which are approximately time-invariant.
However, the conventional fixed effects approach is not applicable in the current situation.
As seen below, key explanatory variables such as democracy and property rights protection
in our regressions rarely changes and hence are nearly time-invariant. Standard fixed effects
regression is inefficient in estimating the effect of minutely varying variables. According to
Plumper and Troeger (2007, p.125), “An inefficient estimation is not merely a nuisance
leading to somewhat higher SEs. Inefficiency leads to highly unreliable point estimates and
may thus cause wrong inferences in the same way a biased estimator could.” We therefore
adopt a common strategy utilized by economists and political scientists in their empirical
analyses concerning institutions. Specifically, we group countries and areas in our study by
the UN standard country and area codes classifications to generate regional dummies49. Then
we use OLS estimation including regional fixed effect dummies and time fixed effect
dummies to identify the impact of democracy on corruption. With this approach we can raise
the efficiency of our estimation by increasing the “within variance” while still controlling for
most of the omitted country-specific factors which may affect both corruption and democracy
since countries in the same category are to a large extent homogenous. For convenience, the
approach is still called a fixed-effects approach in the current chapter although it is not the
conventional one.
The fixed effects approach, however, is not a substitute for the instrumental variables
approach indeed. Our second strategy therefore is to use the instrumental variables approach
to identify the causal effect of democracy on corruption in case that there are still some time-
                                                            
49
 There are originally 22 categories in the UN standard country and area codes classifications: Eastern Africa,
Middle Africa, Northern Africa, Southern Africa, Western Africa, Caribbean, Central America, South America,
Northern America, Central Asia, Eastern Asia, Southern Asia, South-Eastern Asia, Western Asia, Eastern
Europe, Northern Europe, Southern Europe, Western Europe, Australia and New Zealand, Melanesia,
Micronesia, Polynesia. We however treat Israel, the only Jewish state in the world, as an independent category
since it is obviously different from neighbouring Arabic countries (see, Anderson, Seibert, and Wagner 2006).
Therefore we actually categorize countries into 23 groups.

  81 
 

variant omitted country characteristics influencing both democracy and income, which cannot
be controlled for by the fixed effects. Previous studies have made great efforts to address the
endogeneity problem with instruments. Hall and Jones (1999) use the distance from the
equator and the share of population speaking major European languages as instruments for
institutional quality. Acemoglou et al. (2001), however, suggest that European settler
mortality and aboriginal population density in 1500 can be employed as instruments for
current institutions in ex-colonies. When studying the effect of democracy on corruption,
Rock (2007) uses the population fraction of protestant and the latitude of a country’s capital
as instruments for democracy. All of these instruments are intended to capture the western
influence on current institutional quality. However, because the western influence is manifold
and correlated with many aspects of institution, it is difficult to declare what specifically
these variables instrument for. As we know, corruption level reflects institution quality. If we
use these instruments for democracy in our case, they may influence corruption through not
only the channel of democracy but also many other institutional channels. This would violate
the exclusion restriction. We therefore need to choose a more specific instrument for
democracy to guarantee the validity of our IV approach. Following Mobarak (2005), we
construct a dummy indicating any country with the largest proportion of population
practicing Islam (CIA, the World Factbook 200050) as an instrument for democracy. The
distribution of Muslims is of course exogenous, especially in our time horizon. Huntington
(1991, p. 307) argued: “To the extent that government legitimacy and policy flow from
religious doctrine and religious expertise, Islamic concepts of politics differ from and
contradict the premises of democratic politics.” Treisman (2000) and Paldam (2001), on the
other hand, have found that the direct effect of Islam on corruption is insignificant. We
therefore can plausibly suppose that the Islamic religion influences the corruption level only
through the channel of democracy. This validates our instrumental variable.
We now describe the data we use in our empirical analysis. To secure robustness, we
attempt to employ alternative measures of key variables in our regressions. However, we
cannot find more than one corruption measures suitable for panel analysis. Many researchers
such as Treisman (2007) have pointed out that two corruption indices often used in research:
Corruption Perception Index (CPI) compiled by Transparency International (TI) and Control
of Corruption Indicator (CC) constructed by the World Bank51, are actually inappropriate for
                                                            
50
 http://www.umsl.edu/services/govdocs/wofact2000/index.html 
51
 Saha et al. (2009) and Rock (2007) use the two indices respectively. 

  82 
 

panel analysis. Both indices have altered their constructing methodologies and data sources
over the years. Annual variations of both indices may reflect changes in the methodologies
and data sources rather than actual corruption perceptions. Kaufmann and Kraay (2002)
acknowledge that about 50% of the across-time variance of the CC index originates in
changes in data sources and weights assigned to each data source. We hence need to choose
another common corruption measure: the rating of corruption in the International Country
Risk Guide (ICRG), to perform our panel regressions. The ICRG corruption index, ranging
from 0 (most corrupt) to 6 (least corrupt), measures the degree of corruption within the
political system (e.g., demand of special payments, bribes connected with import and export
licenses, exchange controls, tax assessment, police protection, or loans) prevailing in
countries on the basis of the experts’ assessment. The ICRG index provides comparable
corruption data over time and across countries, and hence is the only corruption data set
available for the panel analysis. To obtain a proxy for corruption rather than the lack of
corruption we use negative values of the ICRG index in our estimations (-ICRG).
In line with the role democracy plays in our theoretical model, our first measure of
democracy is the Polity Regime Index: Polity2, the difference between the Polity Democracy
Indicator and the Polity Autocracy Indicator in Polity IV database. It is coded on an
evaluation of the competitiveness of political participation, the competitiveness and openness
of executive recruitment and constraints on the chief executive and ranges from -10 (full
autocracy) to 10 (full democracy). The second measure of democracy we use is the Political
Rights Index from Freedom House. This index measures the degree to which citizens in a
country have control over governors by a checklist of 10 questions about electoral process,
political pluralism and participation and functioning of government. It ranges from 1 (highest
political rights) to 7 (lowest political rights). We use this index mainly as the robustness
check since one of its ten checklist questions assesses national corruption levels, which, as
Rock (2007) points out, might lead to the problem regressing corruption on itself. To keep the
consistency with the Polity2, we use negative values of the Political Rights Index in our
regressions.
We use the property rights rating in the Index of Economic Freedom produced by the
Heritage Foundation and the Wall Street Journal, commonly utilized in previous research, to
measure the security of property rights in our analysis. This index, ranging from 0 (no
protection of property rights) to 100 (full protection of property rights), mainly assesses the
degree to which the laws of a country protect private property rights and the degree to which
its government enforces those laws. According to our knowledge, no further data source of

  83 
 

property rights protection covers our investigating period. The Index of Protection of
Property Rights compiled by the Fraser Institute, used in some priori studies, only covers
2000-2006 hence cannot be employed in our analysis.
We do not construct the actual Gini coefficient series to measure income inequality over
time and across countries since there is not a complete time series in many countries in the
World Income Inequality Database. More importantly Gini coefficients available in WIID
often seem incomparable across countries and/or over time due to their differences in survey
base (income/expenditure), income/expenditure concept, population/area coverage as well as
several other aspects. As income equality in most countries does not change dramatically in
the time perspective of a decade, we instead construct a dummy to indicate income equality
in countries by reviewing all available WIID data in our investigating period. The dummy
equals to 0 if a country passes the international warning line for the Gini coefficient: 0.40 and
1 otherwise. In this simple way, we reduce measurement error of income equality data and
make them comparable. It is worth noting that we only consider income-based Gini
coefficients in the dummy construction. Following Deininger and Squire's (1996), we add 6.6
to the expenditure-based Gini coefficients if there are no income-based ones available in a
country.
We also include other determinants of corruption identified by previous research in our
regressions. GDP per capita, population, openness proxied by import volume in percent of
GDP and natural resource abundance proxied by fuel exports in percent of merchandise
exports are all derived from the World Development Indicators. Adult literacy rates,
reflecting education attainments of countries, are gathered from the UN Human Development
Reports (1998-2009). Ethnolinguistic fractionalization data come from Alesina et al. (2003).
The detailed description of our data is provided in Table 4-1.

Table 4-1 Descriptive Statistics

Variable Observations Mean Standard Deviation
Corruption (ICRG) 1299 3.05 1.25
Democracy (Polity IV) 1510 4.51 5.96
Democracy (Freedom House) 1560 3.21 2.03
Property Rights (Index) 1476 52.40 22.89
Income Equality (Dummy) 1560 0.45 0.50
GDP per capita 1541 6.38 9.90
Literacy Rate 1534 82.06 27.16
Resource abundance (Fuel export / Commodity export) 1289 12.16 21.74
Openness (Import / GDP) 1513 44.76 25.83
Ethnic fractionalization 1560 0.42 0.25
Population 1548 0.45 1.45

  84 
 

4.3.2 Results
To obtain intuitions directing econometric analysis, we first plot the relationship between
democracy and corruption in Figure 4-1. It seems that both the linear negative effect and
quadratic effect of democracy on corruption documented in the literature are reasonable.
Such a descriptive analysis however only gives us information about the raw effects and not
the partial effects. We then test the relationship between democracy and corruption in a
multivariate analysis.

Figure 4-1 Relationship between democracy and corruption

We first briefly examine previous findings with (regional) fixed-effects panel regressions.
Column (1) and (2) in Table 4-2 successfully replicate previous results, supporting the
findings that democracy reduces corruption (see Goldsmith 1999 and Rock 2009). When we
take into account effects of property rights protection and income inequality on corruption,
the linear effect of democracy in column (3) loses its significance, which appears to be in line
with Ades and Di Tella (1997), and Fisman and Gatti (2002). The quadratic effect of
democracy in column (4), however, remains statistically significant. It seems that the
nonlinear effect of democracy is robust, as Rock (2009) argues. However, in column (5) the
quadratic term of democracy lose its significance when we consider the interactions between
property rights protection, income distribution and democracy. The results actually indicate
that the effect of democracy on corruption is modified by property rights protection and
income distribution as can be seen looking at the interaction terms. The overall effect of
democracy on corruption therefore depends on the combination of income equality and

  85 
 

security of property rights in a country, as predicted by our theoretical model. The positive
coefficient of democracy can be interpreted as the effect of democracy of corruption when
property rights and income equality are 0. Thus, in countries with no property rights and a
very high income inequality democracy induces corruption. The results of the two interaction
terms also indicate that a high level of property rights and income equality promote the
controlling effect of democracy on corruption.

Table 4-2 Effect of democracy on corruption: review and implication (fixed effects results)
Corruption level (-ICRG)
Fixed effects
(1) (2) (3) (4) (5)
Democracy -0.020*** -0.0071 -0.0077 0.0017 0.047***
(0.0060) (0.0068) (0.0067) (0.0073) (0.016)
Democracy2 -0.0049*** -0.0038*** -0.0018
(0.0013) (0.0014) (0.0015)
GDP per capita -0.062*** -0.058*** -0.044*** -0.042*** -0.039***
(0.0052) (0.0055) (0.0059) (0.0060) (0.0062)
Literacy rate -0.0024*** -0.0022*** -0.0026*** -0.0025*** -0.0024**
(0.00076) (0.00080) (0.00089) (0.00092) (0.00010)
Resource abundance 0.0055*** 0.0048*** 0.0031*** 0.0027** 0.0030***
(0.0011) (0.0011) (0.0012) (0.0012) (0.0011)
Openness -0.0054*** -0.0060*** -0.0035** -0.0041*** -0.0045***
(0.0013) (0.0014) (0.0014) (0.0015) (0.0016)
Ethnic fractionalization -0.15 -0.21 -0.16 -0.20 -0.26*
(0.15) (0.15) (0.15) (0.14) (0.15)
Population 0.015 0.027 0.0023 0.012 0.018
(0.018) (0.018) (0.017) (0.017) (0.018)
Property rights -0.013*** -0.012*** -0.0076***
(0.0021) (0.0021) (0.0028)
Income equality -0.28*** -0.26*** -0.18***
(0.063) (0.063) (0.064)
Democracy* Property rights -0.00086***
(0.00033)
Democracy* Income equality -0.037***
(0.013)
Constant -2.90*** -2.66*** -2.13*** -1.99*** -2.21***
(0.16) (0.17) (0.19) (0.19) (0.21)
R-squared 0.65 0.65 0.66 0.66 0.66
Observations 1107 1107 1089 1089 1089
Notes: Robust standard errors in parentheses, ***, ** and * denote significance at 1%, 5%
and 10% respectively. Regional and time fixed effects are controlled for in all regressions.

It makes sense to provide more evidence to consolidate our new finding. Table 4-3
supplies fixed-effect results as expected. We first get preliminary results in column (1) and (2)
with pooled-OLS regressions. Then we use regressions including regional and time fixed
effects to obtain further results in column (3) and (4). Protection of property rights
substantially reduces corruption as expected, while income inequality is an important source

  86 
 

of corruption, which also coincides with our prediction. The coefficients are statistically
significant in all four estimations. More importantly, the effect of democracy on corruption is
observed to be significantly modified by property rights protection and income distribution.
Both interaction terms are statistically significant with a negative sign. As to other controls,
income level and education attainment are observed to significantly reduce corruption, which
is in line with prior studies. Also consistent with literature, openness is found to decrease
corruption, while resource abundance is observed to increase corruption.

Table 4-3 Effect of democracy on corruption: fixed effect results

Corruption level (-ICRG)
Pooled OLS Fixed Effects
(1) (2) (3) (4)
Democracy 0.034*** 0.031** 0.038*** 0.050***
(0.013) (0.015) (0.014) (0.015)
Property rights -0.023*** -0.020*** -0.017*** -0.0072***
(0.0021) (0.0025) (0.0023) (0.0027)
Income equality -0.23*** -0.17** -0.23*** -0.18***
(0.065) (0.070) (0.062) (0.063)
Democracy* Property rights -0.00095*** -0.00062** -0.00065** -0.0010***
(0.00024) (0.00028) (0.00028) (0.00030)
Democracy* Income equality -0.034*** -0.032*** -0.046*** -0.040***
(0.0092) (0.010) (0.012) (0.013)
GDP per capita -0.023*** -0.040***
(0.0053) (0.0061)
Literacy rate -0.0038*** -0.0024**
(0.0012) (0.00099)
Resource abundance 0.0012 0.0032***
(0.0013) (0.0011)
Openness 0.0019 -0.0043***
(0.0013) (0.0016)
Ethnic fractionalization -0.062 -0.26*
(0.12) (0.15)
Population 0.069*** 0.014
(0.015) (0.018)
Constant -1.39*** -1.34*** -2.41*** -2.29***
(0.10) (0.19) (0.15) (0.20)
R-squared 0.50 0.54 0.63 0.66
Observations 1232 1089 1232 1089
Notes: Robust standard errors in parentheses, ***, ** and * denote significance
at 1%, 5% and 10% respectively.

As a further robustness test, we rerun the regressions in Table 4-4 with an alternative
measure of democracy, namely the Political Rights Index. Results in Table 4-4 generally
support those in Table 4-3, only the interaction term between democracy and income equality
in Table 4-4 loses statistical significance in the fixed-effect regressions showing, however,
the expected sign. Moreover, the results for the control variables remain robust.

  87 
 

Table 4-4 Effect of democracy on corruption: alternative measure of democracy

Corruption level (-ICRG)
Pooled OLS Fixed Effects
(1) (2) (3) (4)
Democracy 0.11*** 0.10** 0.055 0.14***
(0.038) (0.048) (0.041) (0.051)
Property rights -0.038*** -0.030*** -0.027*** -0.024***
(0.0024) (0.0033) (0.0030) (0.0032)
Income equality -0.50*** -0.44*** -0.28 -0.38**
(0.11) (0.12) (0.17) (0.18)
Democracy* Property rights -0.0036*** -0.0028*** -0.0021*** -0.0039***
(0.00070) (0.00088) (0.00081) (0.00092)
Democracy* Income equality -0.064** -0.058* -0.018 -0.046
(0.028) (0.031) (0.037) (0.040)
GDP per capita -0.022*** -0.037***
(0.0044) (0.0051)
Literacy rate -0.0034*** -0.0024***
(0.0012) (0.00093)
Resource abundance 0.0011 0.0031***
(0.0013) (0.0011)
Openness 0.0012 -0.0046***
(0.0011) (0.0012)
Ethnic fractionalization -0.083 -0.26*
(0.11) (0.14)
Population 0.059*** 0.016
(0.015) (0.017)
Constant -0.95*** -0.91*** -2.17*** -1.68***
(0.15) (0.23) (0.22) (0.26)
R-squared 0.52 0.55 0.63 0.67
Observations 1269 1119 1269 1119
Notes: Robust standard errors in parentheses, ***, ** and * denote significance
at 1%, 5% and 10% respectively.

The previous regressions do not necessarily identify the causal effect of democracy on
corruption since both democracy and corruption might be influenced by an omitted time-
varying factor. We therefore introduce IV regressions to deal with potential endogeneity
problems. As discussed before, we utilize the Muslim dummy to instrument democracy in our
2SLS regressions. The results obtained in Table 4-5 generally support our prior results. The
interaction term between democracy and property rights protection remains statistically
significant reporting the same sign. However, the product of democracy and income equality
loses its statistical significance while retaining its expected sign. This result reflects the fact
that the interaction between democracy and income equality is a bit weaker than the one
between democracy and property rights protection, as our theoretical model reveals.
In general, the effect of democracy on corruption remains conditional on the protection
level of property rights and on the income distribution in IV regressions. The overall effect of
democracy on corruption stays positive in countries with insecure property rights and unequal
income distribution, while it turns negative in countries with secure property rights and equal

  88 
 

income distribution. Control variables in IV regressions also have similar signs and
significances to those in the previous regressions.

Table 4-5 Effect of democracy on corruption: IV results

Corruption level (-ICRG)
(1) (2) (3) (4)
Democracy 0.13*** 0.090**
(Polity IV) (0.040) (0.044)
Democracy 0.30*** 0.19*
(Political Rights Index) (0.099) (0.11)
Property rights -0.0031 -0.000097 -0.052*** -0.032***
(0.0044) (0.0048) (0.0068) (0.0075)
Income equality -0.20** -0.19** -0.039 -0.23
(0.087) (0.088) (0.26) (0.22)
Democracy* Property rights -0.0036*** -0.0026***
(Polity IV) (0.00085) (0.00092)
Democracy* Income equality -0.0089 -0.011
(Polity IV) (0.020) (0.020)
Democracy* Property rights -0.0098*** -0.0063***
(Political Rights Index) (0.0019) (0.0021)
Democracy* Income equality 0.029 -0.016
(Political Rights Index) (0.062) (0.057)
GDP per capita -0.031*** -0.029***
(0.0061) (0.0051)
Literacy rate -0.0023* -0.0025**
(0.0013) (0.0010)
Resource abundance 0.00073 0.00041
(0.0012) (0.0013)
Openness -0.0045** -0.0038**
(0.0020) (0.0015)
Ethnic fractionalization -0.28** -0.19**
(0.11) (0.097)
Population 0.057*** 0.048***
(0.017) (0.016)
Constant -2.72*** -2.47*** -1.22*** -1.55***
(0.22) (0.24) (0.41) (0.48)
First stage regressions
F test of excluded IVs
Democracy 102.77[0.00] 65.92[0.00] 142.89[0.00] 108.21[0.00]
Democracy* Property rights 157.27[0.00] 80.25[0.00] 284.95[0.00] 165.40[0.00]
Democracy* Income equality 183.99[0.00] 162.47[0.00] 324.67[0.00] 280.51[0.00]
Anderson canon. corr. LM statistic 97.62[0.00] 84.79[0.00] 116.78[0.00] 100.36[0.00]
R-squared 0.52 0.60 0.56 0.63
Observations 1232 1089 1269 1119
Notes: Robust standard errors in parentheses, p-value in brackets, ***, ** and * denote
significance at 1%, 5% and 10% respectively.

IV regressions can be justified only if the instrumental variable is valid. We therefore

need to check the validity of our instrument. On the bottom of Table 4-5 we can see that the
Muslim dummy satisfies the relevance condition. We then run original regressions explicitly
including the Muslim dummy. We find that the coefficient of the Muslim dummy is small
and statistically insignificant in Table 4-7 in the Appendix. This suggests that the Muslim

  89 
 

dummy satisfies the exogeneity condition: it influences corruption only through the channel
of democracy. The validity of our instrumental variable is thus confirmed.
In all regressions above, economic development significantly depresses corruption, which
is consistent with our theoretical prediction and previous empirical results. We are not
specifically concerned with the potential simultaneity between corruption and economic
development because Treisman (2000) and Gundlach and Paldam (2009) have documented
with the IV approach that the causality in the cross-country analysis is generally from
economic development to corruption. As for the effects of other controls in our regressions,
education attainment and trade openness significantly and robustly decrease corruption as
contended in prior literature, while resource abundance and country size (population) appear
to increase corruption, which is also in line with earlier research. Ethnic diversity, however,
counter- intuitively reduces corruption in our regressions. Treisman (2000) has observed that
the originally positive effect of ethnic division on corruption becomes negative and
insignificant when controlling for economic development in cross-country regressions. His
interpretation that ethnic diversity only influences corruption indirectly by reducing
development seems irrelevant to our negative and significant results. Our theoretical model,
however, provides us with a plausible explanation. There is always at least one ethnic group
dominant in politics in an ethnically divided country. The pivotal voter in this kind of country
therefore belongs to the dominant ethnic group(s). In practice, these ethnic groups are often
richer than others due to the power in their hands. Ceteris paribus, the pivotal voter in an
ethnically divided country, according to our model, will choose a relatively low tax rate since
he is comparatively rich. This country then will have a low corruption level according to our
mechanism. In other words the seemingly counterfactual effect of ethnic division in our
regressions actually provides a substantial support to our theoretical modeling.
To show the overall effect of democracy on corruption, we calculate the marginal effect
of democracy on corruption. In our specification, the marginal effect of democracy can be
expressed as

4 18

where , , and are corruption, democracy, property rights protection and income
distribution indicators of country in period t respectively, while , and
are the coefficients of democracy, the interaction term between democracy and
property rights protection, and the interaction term between democracy and income equality.

  90 
 

The average marginal effects of democracy computed from regressions with prior
specification have been listed in Table 4-6. We find that the overall effect of democracy on
corruption is significantly negative. This supports the notion that democracy reduces
corruption. However, we also observe in Table 4-6 that both marginal effects of democracy
in IV regressions are, though not by much, obviously smaller than those in fixed-effect
regressions. The most plausible explanation for this is that there might be an unobserved
time-variant variable causing: , 0( is the error term in equation 4-16).
Fixed effects estimators hence are biased upwards. If this is the case, we can treat our fixed
effects results as upper bounds of the causal effect of democracy on corruption as Acemoglu
et al. (2008) suggest.

Table 4-6 Marginal effect of democracy on corruption

Democracy Measures Polity IV Index Political Rights Index
FE IV FE IV
Marginal Effects -0.021*** -0.053*** -0.083*** -0.15***
(0.0080) (0.013) (0.023) (0.037)
Observations 1089 1089 1119 1119
Notes: Standard errors in parentheses, ***, ** and * denote significance at
1%, 5% and 10% respectively.

More importantly, with marginal effects in Table 4-6 we can reinterpret previous
empirical research from a new angle. Most previous studies adopt the linear-additive model
without interaction terms to study the relationship between democracy and corruption. The
coefficient on democracy in their linear-additive models actually represents the (weighted)
average marginal effect of democracy in our interaction model (Brambor et al. 2006). The
results obtained in Table 4-6 hence are consistent with most previous empirical papers which
support a negative linear effect of democracy on corruption. Our results also show that
corruption is, in fact, a nonlinear function of several variables including democracy, property
rights and income inequality. As in Sung (2004) and Rock (2009), higher degree terms of
democracy such as the quadratic or cubic term in regression can partially reflect the actual
nonlinear relationship between democracy and corruption and hence might be significant in
some cases though this kind of polynomial approximation is not very appropriate. In
summary, without considering the interactions between democracy, property rights protection
and income distribution, previous studies only partially capture the actual effect of
democracy on corruption.

  91 
 

4.4 Conclusion

Democracy is believed to have many beneficial effects on countries. However, does it

prevent corruption? Rose-Ackerman (1999, p. 142) stresses: “Democratic elections are not
invariably a cure for corruption. Instead, some electoral systems are more vulnerable to
special influence than others. When narrow groups wield power, some use legal means, and
others are corrupt”. Previous literature provides mixed evidence, which still leaves the
problem open. In this study we find strong evidence that the effect of democracy on
corruption depends upon other variables such as property rights or income inequality. In
particular, we provide a theoretical and empirical investigation of the causal nexus between
democracy and corruption. The theoretical model offers a mechanism via which democracy
influences corruption. It extends previous models implementing property rights and income
distribution into the theoretical framework. Our empirical results are consistent with the
theoretical model. The effect of democracy on corruption is conditional on income
distribution and property rights protection. The findings indicate that democracy is able to
work better as a control of corruption if the property right system works and if there is a low
level of income inequality. On the other hand if property rights are not secured and there is
strong income inequality, democracy may even lead to an increase of corruption.

  92 
 

Appendix
Table 4-7 Validity of instrument: Muslim
Corruption level (-ICRG)
(1) (2)
Democracy 0.030**
(Polity IV) (0.014)
Democracy 0.11**
(Political Rights Index) (0.044)
Property rights -0.0098*** -0.025***
(0.0025) (0.0030)
Income equality -0.18*** -0.38***
(0.063) (0.10)
Democracy* Property rights -0.00084***
(Polity IV) (0.00027)
Democracy* Income equality -0.028***
(Polity IV) (0.0090)
Democracy* Property rights -0.0039***
(Political Rights Index) (0.00082)
Democracy* Income equality -0.046*
(Political Rights Index) (0.027)
GDP per capita -0.037*** -0.032***
(0.0052) (0.0042)
Literacy rate -0.0027*** -0.0027***
(0.0010) (0.00095)
Resource abundance 0.00043 0.00024
(0.0011) (0.0011)
Openness -0.0023* -0.0025**
(0.0013) (0.0011)
Ethnic fractionalization -0.16 -0.17*
(0.10) (0.096)
Population 0.062*** 0.052***
(0.014) (0.013)
Muslim dummy 0.094 0.041
(0.065) (0.060)
Constant -2.23*** -1.80***
(0.21) (0.23)
R-squared 0.62 0.63
Observations 1089 1119
Notes: Robust standard errors in parentheses, ***, ** and * denote significance at
1%, 5% and 10% respectively.

  93 
  

Chapter Five Social interaction and Corruption: Cross-country Evidence52

5.1 Introduction

Corruption is seen by many as one of the main impediments of the development of an

efficient government system. Corruption can be seen as a “symptom that something has gone
wrong in the management of the state” (Rose-Ackerman 1999, p. 9). It is a phenomenon that
is apparent throughout human history. For example, pharaohs in Ancient Egypt provided high
salaries to the tax collectors (called scribes) to increase the opportunity costs of enriching
themselves by cheating taxpayers (Adam 1993). According to Jain (2001), existence of
corruption requires three preconditions: discretionary power related to regulations (see also
Rose-Ackerman 1978), economic rents linked to the power and sufficiently low punishment.
The growing interest in institutional issues such as the transformation process of socialist
economies has led to an increase in the number of studies devoted to exploring the detailed
causes of corruption at the international level. Research has shown that a political economic
approach stressing the importance of institutions has proved to be a powerful tool in
understanding corruption (Bardhan 1997, Rose-Ackerman 1997, Abed and Gupta 2002). For
example, Kunicova and Rose-Ackerman (2005) found that electoral rules and constitutional
structures substantially affect the corruption level. Countries tend to achieve an equilibrium
position that is driven by the balance of political forces and institutions (Bird, Martinez-
Vazquez and Torgler 2006, 2008). However, most of these studies explore corruption at the
macro level while only a limited number of studies have investigated the determinants of
corruption at the individual level (see, e.g., Mocan 2008, Swamy et al. 2001, Torgler and
Valev 2006, Dong and Torgler 2009). Among the individual studies, Gordon (2009)
intriguingly observed partisan bias in federal public corruption prosecutions in America
which influences the punishment for corruption. Such studies mainly take into account the
vertical interrelationship between citizen and the state. Our study, on the other hand,
implements a horizontal perspective using a behavioral economics approach hypothesizing
that social interactions matter. In particular we stress that the own willingness be corrupt
depends on the corruption level of other individuals in a society and that current corruption is
affected by the past corruption levels. In other words, we are going to explore whether
conditional corruption matters and whether corruption is contagious. The willingness to be
                                                            
52
 This chapter has been submitted to the American Political Science Review. 

  94 
 

corrupt is influenced by the perceived activities of peers and other individuals. Thus, a
person’s willingness to be corrupt depends on the pro-social behavior of other citizens. The
more that others are perceived to be corrupt, the higher the willingness to be corrupt. The
study therefore emphasizes the relevance of social context in understanding corruption.
Theories of contagion or pro-social behavior, which take the impact of behavior or the
preferences of others into account, are used as a starting point in the theoretical approach.
Contagion effects have been observed in other illegal activities such as assassinations,
hijackings, kidnappings, and serial murders as referred to by Bikhchandi, Hirshleifer and
Welch (1998). The relevance of social interaction and crime is explored by Glaeser,
Sacerdote and Scheinkman (1996) who focus on the United States in their analysis both
across cities and across precincts in New York. The results indicate that social interaction
models provide a framework for understanding variances of cross-city crime rates.
Individuals are more likely to commit crimes when those around them do. Frey and Torgler
(2007) have found empirical evidence of conditional cooperation in the area of tax
compliance. Kahan (1998) suggests that the decision to commit crimes is highly
interdependent, based on the perceived behavior of others: “When they perceive that many of
their peers are committing crimes, individuals infer that the odds of escaping punishment are
high and the stigma of criminality is low. To the extent that many persons simultaneously
draw these inferences and act on them, moreover, their perceptions become a self-fulfilling
reality” (p. 394). As a consequence, individuals’ beliefs about crime is altered, suggesting
that social influence affects criminality and the propensity to commit crimes.
Figure 5-1 illustrates our argument: the higher the levels of perceived corruption in a
society, the more citizens see it as justified. We first provide a simple theoretical explanation
of this observation and then check whether a more thorough study of the data supports the
initial evidence suggested by the figure.
To our knowledge, our study provides findings not yet discussed in the corruption
literature. There are not many studies that investigate the relevance of conditional
cooperation and a contagion effect in regards to corruption. We use the notion of “conditional
corruption” for these effects. In particular, there is a lack of empirical evidence at the
international level. The chapter also complements a large set of laboratory experimental
studies that have studied conditional cooperation by providing evidence outside of a lab
setting. We will first conduct a micro analysis using data from two wide-ranging surveys,
namely the European Values Survey and the World Values Survey. Despite the increasing
interest of economists in the determinants of corruption, research at the micro level has not

  95 
 

yet come under intense empirical investigation. The micro analysis is complemented by a
macro analysis working with a large international data set that covers almost 20 years. Before
considering these findings in detail, however, Section 5.2 presents a theoretical model.
Section 5.3 introduces the empirical part discussing the data and Section 5.4 provides the
empirical findings. Section 5.5 finishes with some concluding remarks.

Figure 5-1 Correlation between Justifiability of Corruption and Perceived Corruption

3

BGD

URYUSA
AUS ARG
NOR POL
FIN YUG
ESP
KOREST MKD
IND RUS
CHE
2.5

BIH PER
CHL
DEU
SVN
TWN
UKR
GEO LTU
BGRBLR NGA
LVAVEN
HRV ARM
MDA
MEX
2

AZE

BRA
1.5

PHL

2 2.5 3 3.5
perceived corruption (PLC)

95% CI Fitted values

justifiability of corruption 0-3, 3=never justified)

Note: Pearson r = - 0.42. More countries than in the regression analysis.

5.2 Theoretical Foundation

In this section, we theoretically investigate conditional cooperation in a corruption

framework. We therefore use the notion of “conditional corruption” instead of conditional
cooperation. Individuals condition their corruption on the behaviour of other individuals. An
individual is prone to be corrupt if there are a sufficient number of corrupt individuals around
him. The reason put forward in the literature is one that can be called reinforcing corruption
Aidt (2003), i.e. the incentive structures of a society together with the presence of corrupt
individuals determines whether corruption is worthwhile for an individual. Either the direct
potential benefits or the direct cost of engaging in corrupt activity are lower if the number of
corrupt people in an individual’s environment or society is larger. Andvig and Moene (1990),
Murphy et al. (1991), Acemoglu (1995) and Sah (2007) stress that the return to corruption for
an individual depends on the number of individuals expected to be corrupt in the same
organisation or society. Sah (1988, 2007) also mentions learning - a corrupt official continues

  96 
 

to be corrupt if she has interacted with a sufficient number of corrupt officials in the past as
she knows that looking for bribes will be worthwhile. At the same time the cost to be caught
in a corrupt act are lower because corruption is more difficult to detect in societies where it is
more common (Lui 1986, Cadot 1987). Tirole (1996) assumes that individuals are members
of different groups and also argues with the cost of detection, namely the potential exclusion
from a group. He allows for a society with corrupt and non-corrupt individuals, where the
latter have long term advantages. Group membership acts as a signal. The cost of engaging in
corrupt activities decreases for an individual if more members of that group are corrupt as the
threat of exclusion from the group becomes weaker as group membership loses its function to
signal non-corrupt behavior To summarize, all these theories use pure income maximization
as the motive to engage in corruption. The presence of corrupt individuals in a society,
together with its institutions provides the incentives that determine individuals’ behaviour.
Several theories have been put forward to explain what constitutes conditional
cooperation in the area of behavioural economics. Most papers in the literature (Rabin, 1998
and Falk and Fehr, 2002) explain conditional cooperation in terms of reciprocity. In a
corruption context, reciprocity means, that if corruption within a society is very prevalent,
citizens feel less guilt when engaging in extra-legal activities, and are likely to act
accordingly. Several laboratory experimental studies (mainly public good experiments)
provide evidence on pro-social behaviour (for an overview, see Gächter, 2006). For example,
Fischbacher, Gächter and Fehr (2001) find that 50 percent of the subjects were conditionally
cooperative. Falk, Fischbacher and Gächter (2003) create a laboratory situation in which each
subject is a member of two economically identical groups, where only the group members are
different. They observe that the same subjects contribute different amounts, depending on the
behavior of the group. Contributions are larger when group cooperation is higher.
Alternatively, the concept of conformity (Henrich, 2004) has been used to explain conditional
cooperation. Conformity means that the motivation of behaving in a conditionally
cooperative way may be influenced by the people’s wish to fulfill the social norm of not
being corrupt and behaving according to society’s rules. While several early studies provide
evidence of conditional cooperation within a laboratory setting, an increasing number of
studies have been conducted to check the validity of such studies outside of a laboratory
setting using, for example, field experiments (see Frey and Meier, 2004a; 2004b; Heldt, 2005;
Shang and Croson, 2005, Martin and Randal 2005). The study of a contagion effect and pro-
social behavior resulting from a perceived level of corruption is an area that has largely been

  97 
 

ignored in the empirically oriented corruption literature, despite its potential to affect the
level of corruption even further.

5.2.1 Background of the Model

While the existing theoretical literature concentrates on the direct costs and benefits of
engaging in corrupt activities, we want to introduce moral costs that affect the decision to
engage in corruption. We assume that these moral costs are independent of the number of
corrupt individuals. Even under this assumption one can show that the decision to engage in
corruption depends on the attitudes of possible partners in corruption. This is what we refer to
as conditional corruption.
Since Akerlof (1980) emphasized the persistent effect of social norms on human
behaviour, interesting approaches based on the concept social norms have been developed
that help explain conditional cooperation in human activities. One way to accommodate such
behaviour is to allow for a guilt disutility if a citizen engages in an activity that is
contradicting social norms. The social customs literature provides a motive for the reason
why there can be a utility loss by the act of evading taxes (see Naylor 1989). The essence of
this approach is that violation of social norms will bring forth moral cost. In a tax compliance
framework, Gordon (1989) modifies the standard economics of crime model by including
non-pecuniary costs of evasion. Non-pecuniary or psychic cost increases as evasion increases.
The model he developed can explain why some taxpayers refuse a favorable evasion gamble.
Furthermore, dishonesty is endogenized as reputation cost. Non-pecuniary costs have a
dynamic component, varying inversely with the number of individuals having evaded in the
previous period. Other researchers such as Myles and Naylor (1996) criticize this approach,
stating that the level of evasion or non-compliance is irrelevant. Once a social custom is
broken, all utility from it is lost. In line with this argument we will also assume that once a
person breaks the social norm, the fact of doing so is what counts, not the magnitude of the
transgression. We assume in our model that incorruption is regarded as a prevailing norm in
societies (consistent with long-held moral standards), while corruption violates this social
norm. Corruption and therefore the violation of this social norm generate guilt or shame
(Elster 1989). The sentiments of guilt and shame may influence compliance behavior,
reducing the perceived benefits of corruption. According to Lewis (1982), guilt arises when
individuals realize that they have acted irresponsibly and in violation of an internalized rule
or social norm. Since incorruption is an accepted social norm, it makes sense that individuals
who choose to be corrupt feel guilty or ashamed. According to Spichtig and Traxler (2007),

  98 
 

this internal sanction against violation will be more powerful if more individuals stick to the
norm. In our case this means that when corruption is rare in a society, individuals tend to be
incorrupt since the cost of violating social norm is very high. When corruption becomes more
prevalent, more individuals become corrupt since the cost to infringe the norm declines.
Rege (2004) stresses that a social norm for cooperation can generate conditional
cooperative behaviour since the sanction for a norm deviation will force people to stick to the
norm. Andvig and Moene (1990) also incorporate moral cost associated with social norms
into traditional cost-benefit analysis developed by Becker (1976), deducing multiple self-
fulfilling equilibria. The authors stress that the probability of corruption is related to its
established frequency. On the other hand, Mishra’s (2006) model uses an evolutionary game
framework deriving multiple evolutionary stable states in corruption level. However, many
previous corruption models have the limitation that they only consider the behavioral
implications in regards to bureaucrats instead of focusing on the entire society53. Both parties,
bureaucrats and citizens, are players in the corruption game. Only analysing the behaviour of
one side is not enough to explain any phenomenon related to corruption, especially in our
context that focuses on conditional corruption.
Social norms consist of a pattern of behaviour that must be shared by other people and
sustained by their approval and disapproval (Elster 1989). Coleman (1990) stresses that social
norms are rules of conduct enforced by external or internal sanction. Polinsky and Shavell
(2000), who present a survey of the economic theory of public enforcement of law,
emphasize the aspect of social norms for future research. Social norms can be seen as a
general alternative to law enforcement in channeling individuals’ behavior. The violation of
social norms has consequences like internal sanctions (guilt, remorse) or external legal and
social sanctions as gossip and ostracism. As Polinsky and Shavel (2000) state there is an
expanding literature on social norms because of the influence social norms have on behavior,
their role as a substitute for and supplement to formal laws and the possibility that laws
themselves can influence social norms54. Fehr and Gächter (1997) define social norm as:
“behaviour regularity that is based on a socially shared belief how one ought to behave which
triggers the enforcement of the prescribed behaviour by informal social sanctions” (p. 12).

                                                            
53
Mauro (2004) employed two models to analyze the behavior of bureaucrats and citizens respectively. His
results, however, are not convincible since he did not investigate behavior of both sides simultaneously.
54
Posner (1997, pp. 365-366) looks at the incentives for obeying norms. He finds four: (i) norms that are self-
enforcing because obedience confers private benefits, (ii) norms that are enforced by emotions, iii) milder
sanctions by expressions of disapproval or ridicule and (iv) internalized norms, out of a sense of guilt or shame. 

  99 
 

5.2.2 A Simple Game

We consider a simple model of corruption. A citizen C (he) can attempt to bribe bureaucrat B
(she) to earn an extra-legal profit . C and B have utility functions depending on (expected)
income and the potential disutility if they engage in corruption. Let R be the income from any
type of activity and c an index that is equal to 1 if this activity involves corruption (0
otherwise) and  the disutility felt when engaged in corruption. The utility function is then
given as:

U(R,c) = R - c

We concentrate on a citizen’s decision whether to engage in projects that involve corruption.

Such projects generate an additional profit of  over the best non-corrupt project. This could
be, for example, a project that would have not received a public license according to the law
or it may extend a project past the limits set by the law, or it may simply speed up the normal
process. In all cases the gain  represents an additional profit over the profit earned under
normal proceedings. If this activity does not take place, all parties involved in a potential
corruptive activity receive a default payoff that is normalized to 0. For simplicity of
presentation, we assume that B accepts the bribe if indifferent and C bribes if indifferent.
Bribing and accepting a bribe incurs a cost of c for C and b for B. These costs represent
guilt felt when engaging in bribing. Thus a low value of  represents a high tendency to be
corrupt. Before the game starts  is drawn from a distribution function F and C and B draw a
value of c, b; we denote by c, b respectively the random variables and assume that both
are distributed independently according to an identical distribution function G. We assume
that when interacting, C knows B’s guilt parameter, i.e. at least b is common knowledge
between the two players. The timing of the game is as follows: 1. Nature draws , c,b; 2. C
decides whether to attempt a bribe or not and if he bribes he chooses an amount b; 3. B
decides whether to accept the bribe. Figure 5-2 summarizes the description of this game.
It has a unique sub-game perfect Nash-Equilibrium in pure strategies: B accepts all bribes
with b  b. C bribes B if   c+b paying a bribe of b=b as this bribe is accepted by B; if 
< c+b then C does not bribe.

   
100
 

Figure 5-2 Description of the Corruption Game

Nature 
c, b, 

Citizen  Do not bribe
(0, 0)
b Bribe
 Bureaucrat 

Reject  Accept

(0‐c, 0)        (‐b‐c, b‐b) 

Given the equilibrium the conditional probability for C to engage in corruption depends
on his guilt parameter and the expected value of the guilt parameter of a bureaucrat, he may
interact with: Pr(c+bc). This probability is decreasing in E(b). If we believe that
people tend to justify their behaviour then this conditional probability will determine how
justifiable C would judge participation in corruption or bribery. Corruption is less justifiable
for C the higher c and the higher E(b). In this respect we observe conditional corruption – a
low guilt cost in a society will induce even citizens with relatively high guilt cost to engage in
corrupt activities. In the empirical section we use a large survey data set to explore this
justifiability of corruption.

Hypothesis 1: Citizens find corruption less justifiable if they perceive their society less
corrupt.

We intentionally did not discuss . In the present model  is exogenous. It may depend
on the opportunities that open up only by bribery, hence a society with better institutions
should allow most beneficial activities without bribes, hence  may be lower, while a society
with weak institutions would have higher  as profitable activities are accessible only when
one bribes bureaucrats.
Note, this model’s innovation is to emphasize the importance of guilt that people
experience engaging in corrupt activities, arguing that guilt levels depend on the perceived
prevalence of corruption within a society. From a modelling perspective the model is similar
to a model where c and b represent expected costs of giving and receiving bribes, for
example because ex post investigations can reveal bribes as well as attempted bribes. We do
believe that guilt has indeed a similar effect as an expected monetary punishment (with low

   
101
 

probabilities of detection) may have. The important difference is that this guilt parameters
can and will vary across different people and that an individual’s expectation to find person
with a high propensity to engage in corruption will determine how active this individual will
become in looking for opportunities to corrupt.
5.2.3 Dynamics
The second issue we address is the question how a tendency towards corrupt or non-corrupt
behaviour. To do so we assume a simple overlapping generation model with a constant
population model. We denote by an upper index t a generation’s respective values. Each
generation has one offspring and this offspring’s guilt parameter t+1 decreases if the parent
was involved in an act of corruption and it increases if the parent was not involved in an act
of corruption. We assume that on average the absolute change in the parameter in both
directions is of the same magnitude. If this is the case then if more (less) than half of the
population is involved in corruption the median and the average  increases (decreases) with
the next generation:
Pr(2t) 1/2  0.5t+1 0.5t and E(t+1) E(t).

Hypothesis 2 Corruption is contagious: A society experiencing a high (low) level of

corruption will have increased (decreased) levels of corruption in the future.

Again, the arguments with respect to  hold, a higher  - which may be due to weak
institutions – will increase the spread of corruption over time. Strong institutions that allow
all profitable activities to be undertaken legally will decrease corruption and hence over time
will increase the guilt felt by citizens if they engage in corruption.
5.2.4 Conditional Corruption – Discussion and Extensions
The conditionality of corruption in our model comes from the fact that low guilt costs in
society make it more likely that a citizen engages in corruption. The proposed dynamics
imply that this leads to lower guilt costs within a society.
An alternative model giving rise to similar hypotheses is one where the individual’s guilt
cost depends on the frequency of corruption within a society. In this case  =f(Pr(2E())
with f’(.)<0. In this case an equilibrium can be defined as a fixed-point of this function and,
depending on the explicit assumption on the distribution function, could give rise to two or
more equilibria where some are characterized by self-confirming on average high values of 
and others by self-confirming low values of . Spichtig and Traxler (2007) provide a model
for conditional cooperation in this spirit.

   
102
 

5.3 Data and Methodological Approach

5.3.1 Micro Analysis

In the first stage we are going to work with survey data at the micro level to explore our first
hypothesis. This allows us to work with a representative set of individuals, which is not often
the case in previous (experimental) studies that have explored conditional cooperation
primarily by using students as participants55. We are going to use two micro data sets. First
we are going to work with the European Values Survey (EVS) 1999/2000, which is a
European-wide investigation of socio-cultural and political change. Next, we are going to
explore the World Values Survey (WVS), a worldwide data set that investigates socio-
cultural and political change. The WVS was first carried out in 1981-83, with subsequent
surveys being carried out in 1990-1993, 1995-1997 and 1999-2001.
5.3.1.1 European Values Survey
The EVS assesses the basic values and beliefs of people throughout Europe. The EVS
was first carried out from 1981 to 1983, then in 1990 to 1991 and again in 1999 through 2001,
with an increasing number of countries participating over time. The EVS methodological
approach is explained in detail in the European Values Survey (1999) source book, which
provides information on response rates, the stages of sampling procedures, the translation of
the questionnaire, and field work, along with measures of coding reliability, and data checks.
All country surveys were carried out by experienced professional survey organizations, (with
the exception of the one in Greece), and were performed through face-to-face interviews
among samples of adult citizens aged 18 years and older. Tilburg University coordinated the
project and provided the guidelines to guarantee the use of standardized information in the
surveys and the national representativeness of the data. To avoid framing biases, the
questions were asked in the prescribed order. The response rate varies from one country to
another; in general, the average response rate was around 60%.
Because the EVS asks an identical set of questions to people in various European
countries, the survey provides a unique opportunity to examine whether conditional
corruption matters. Our study considers representative national samples of at least 1000
individuals in each country.
                                                            
55
 Fehr et al. (2003) report that the problem with using students is that they have a higher level of education and
a higher IQ than average citizens. In addition, they often come from families with a higher than average income
and their age range is limited.  

   
103
 

Our dependent variable is justifiability of corruption assessed using the following

question:
Please tell me for each of the following statements whether you think it can always be
justified, never be justified, or something in between: (…) someone accepting a bribe
in the course of their duties (1=always justified, 10= never justified).
The interpretation of this question is that higher values are in line with a lower justifiability
of corruption. This variable can be seen as a proxy for social norms of compliance (see
Torgler 2007). We use the following question to investigate the influence of conditional
corruption:
“According to you, how many of your compatriots do the following: Accepting a
bribe in the course of their duties?” (4=almost all, 1=almost none)

5.3.1.2 World Values Survey

We are going to work with the third WVS wave as the question referring to individual
conditional corruption has only been asked in this wave. For the researchers who conduct and
administer the World Values Survey (WVS) in their respective countries, it is a requirement
that they follow the methodological requirements of the World Values Association. For
example, surveys in the World Values Survey set are generally based on nationally
representative samples of at least 1000 individuals of 18 years and above (although
sometimes people under the age of 18 participate). The samples are selected using probability
random methods, and the questions contained within the surveys generally do not deviate far
from the original official questionnaire (for a sample of a typical World Values Survey see
www.worldvaluessurvey.org). We have not analysed the entire World Value Survey data set:
countries below 750 observations have not been included in the estimations to reduce
possible biases due to a lack of representativeness56. Furthermore, some countries do not have
information on the dependent variables or some of the independent variables. These countries
are therefore not considered.57.
We use the same dependent variable as previously, namely the justifiability of corruption.
On the other hand, we assess the relevance of conditional corruption using an alternative
question that measures perceived corruption:

                                                            
56
 Thus, Montenegro and the Dominican Republic have been omitted.  
57
 Japan, South Africa, Puerto Rico, Turkey and Columbia. Moreover, Sweden could not be included as one of
the control variables (education) has been coded differently.  

   
104
 

How widespread do you think bribe taking and corruption is in this country?
Almost no public officials are engaged in it (1)
A few public officials are engaged in it (2)
Most public officials are engaged in it (3)
Almost all public officials are engaged in it (4)
5.3.1.3 Addressing the Limitations of Survey Data
Two main limitations of survey data are often raised: a self-reporting bias and cognitive
biases. We address these issues in turns.
First, the validity of the justifiability of corruption variable can be criticized as it reports a
self-reported and hypothetical choice (see Swamy et al. 2001). It can also be argued that an
individual who has engaged in corruption in the past will tend to cover up such behaviour by
declaring a low justifiability of corruption in the survey. Furthermore, cross cultural
comparisons should be treated with some caution. In countries where corruption is
widespread and delays in transactions are long, additional payments to “speed up” the
process may be justifiable and a normal part of the administration process. The necessity of
additional payments is so pervasive in some countries that the bureaucratic mechanism does
not operate without them (De Soto 1989). However Torgler and Valev (2006), show that
survey data on corruption is highly correlated with other available proxies of corruption
(Transparency International, Kaufmann, Kraay, and Mastruzzi, 2004 (KKM) and ICRG
(Knack, 1999)).
Moreover, one should note that the survey contains no information on the perceived
activities of peers (e.g., friends, work colleagues or neighbours). One can argue that a
reference group has a stronger effect on our behaviour than the overall population.
Another aspect of the self-reporting bias is that cultural differences play a role. To control
for this we extend our study by also adding a country’s overall corruption values into the
micro data set. We use the control of corruption variable developed by KKM due to the large
number of countries included in this data set. The proxy measure is driven by the traditional
notion of corruption namely “the exercise of public power for private gain” covering a
variety of aspects ranging from the frequency of “additional payments to get things done” to
the effects on the business environment (p. 8). The values lie between –2.5 and 2.5, with
higher scores corresponding to a lower level of corruption. Figure 5-3 illustrates our
argument with respect to the self-reporting bias. It shows that justifiability of corruption is
highly correlated with other more frequently used measures of corruption.

   
105
 

Figure 5-3 Correlation between Justifiability of Corruption and Control of Corruption

3
BGD
CHN URY USA AUS
ARG NOR
YUG POL ESP FIN
MKD KOR EST
RUS IND
2.5 CHE
BIH PER
SVN CHL DEU
TWN
UKR
GEO
NGA VEN BGR LTU
BLR LVA
ARM HRV
MDA
MEX
2

AZE

BRA
1.5

PHL

-1 0 1 2 3
control of corruption (KKM)

95% CI Fitted values

justifiability of corruption 0-3, 3=never justified)

Note: Pearson r= 0.38. More countries than in the regression analysis.

Second, Bertrand and Mullainathan (2001) argue that cognitive problems arise – the
experimental literature has shown that manipulations (e.g., order of the question, wording or
scales) can affect how people process and interpret questions. The problem is that
“respondents may make little mental effort in answering the question, such as by not
attempting to recall all the relevant information or by not reading through the whole list of
alternative responses” (Bertrand and Mullainathan, p. 68). To control for such problems, we
explore the correlation between two similar questions asked in the EVS (WVS) in different
parts of the interview: How interested would you say you are in politics? (IP) Very interested
(value 1), somewhat interested (2), not very interested (3), not at all interested). How
important is politics in your life? (INP) Very (1), (rather 2), not very (3), not at all (4). The
correlation at the micro level is 0.614 (0.544). Moreover, we also explore the correlation with
the following question: When you get together with your friends, would you say you discuss
political matters frequently (value 3), occasionally (value 2) or never (value 1)? (DP). The
correlation between INP and DP is 0.45 (0.53) and between IP and DP 0.56 (0.38). Thus, the
variables are highly correlated. Face-to-face interviews may also help to guarantee that
subjects are aware of the whole list of alternative responses. The EVS (WVS) has also the
advantage of being a wide-ranging survey covering a large amount of different topics. Thus,
our corruption question was only part of a larger survey, which may reduce framing biases.

   
106
 

5.3.2 Macro Analysis

To address our second hypothesis, the analysis will be complemented by use of a large
international panel macro data set: the ICRG data (see Knack 1999) covering 18 years (1986
till 2003). We use ICRG as KKM has only been collected for a limited number of years
(1996, 1998, 2000, 2002), and while ICRG allows us to study fewer countries, it provides
panel data for a longer time period. The political risk rating provided by ICRG aims to assess
the political stability of the included countries. The corruption variable is an assessment of
corruption within the political system. The measure is concerned with actual or potential
corruption in the form of excessive patronage, nepotism, job reservations, “favor-for-favors”,
secret party funding, and suspiciously close ties between politics and business. The macro
data set has the great advantage of being able to explore the importance of a contagion effect
over time.

5.4 Results

5.4.1 Micro Level using the EVS

The micro analysis will allow us to explore our first hypothesis. One can argue that the
potential conditional corruption effect could be influenced by other variables that affect
corruption. Thus, we control in our multivariate analysis for variables such as education
level58, political interest59, religion60, risk attitudes61, the economic situation62, urbanization63
                                                            
58
 EVS: Formal education: At what age did you complete or will you complete your full time education, either at
school or at an institution of higher education? Please exclude apprenticeships.
WVS: What is the highest educational level that you have attained?
10. No formal education
11. Incomplete primary school
12. Completed primary school
13. Incomplete secondary school: technical/vocational type
14. Complete secondary school: technical/vocational type
15. Incomplete secondary: university-preparatory type
16. Complete secondary: university-preparatory type
17. Some university-level education, without degree
18. University-level education, with degree
59
EVS/WVS: How important is politics in your life? very (4), (rather 3), not very (2), not at all (1).
60
EVS: Apart from weddings, funerals and christenings, how often do you attend religious services these days?
More than once a week, once a week, once a month, only on special holy days, once a year, less often,
practically never or never(8= more than once a week to 1=practically never or never).
WVS: Apart from weddings, funerals, and christenings, about how often do you attend religious services these
days? More than once a week, once a week, once a month, only on special holy days, once a year, less often,
never or practically never. (7 = more than once a week to 1 = never or practically never).
61
EVS: Here are some aspects of a job that people say are important. Please look at them and tell me which
ones you personally think are important in a job? (15 items). Risk aversion: Good job security (1=mentioned).

   
107
 

and the employment and marital status. Previous tax compliance and corruption studies
demonstrate the relevance of considering these factors (see, e.g., Torgler 2007, Dong and
Torgler 2009).
Table 5-1 presents the first results obtained by working with the European Values Survey.
In the first specification we recode the original dependent variable into a four-point scale (0,
1, 2, 3), with the value 3 standing for “never justified”. Responses 4 through 10 were
combined into a value 0 due to a lack of variance among them. This approach is consistent
with previous studies (see, e.g., Torgler and Valev 2006, Torgler 2007). In the second
specification we use the original 10-point scale. In both cases we are going to use an ordered
probit model. The ordered probit models are relevant in such an analysis insofar as they help
analyze the ranking information of the scaled dependent variable. The data structure indicates
that we have a natural cut-off point. A large amount of respondents assert that corruption is
“never justified” (71 percent of the cases). Our dependent variable therefore takes in the
following specifications: the value 1 if the respondent says that bribing is “never justified”
and 0 otherwise. This requires the use of a probit model in most of the specifications (see also
Table 5-3). We also use weighted (ordered) probit estimations to correct the samples and thus
to get a reflection of the national distribution. Moreover, since equations have a nonlinear
form, only the sign of the coefficient can be directly interpreted and not its size. We therefore
also calculate the marginal effects to find the quantitative effect of a variable on our
                                                                                                                                                                                         
WVS: Now I would like to ask you something about the things which would seem to you personally, most
important if you were looking a job. Here are some of the things many people take into account in relation to
their work. Regardless of whether you’re actually looking for a job, which one would you, personally, place first
if you were looking for a job?
5. A good income so that you do not have any worries about money
6. A safe job with no risk of closing down or unemployment
7. Working with people you like
8. Doing an important job which gives you a feeling of accomplishment
And what would be your second choice?
A dummy variable was built with the value 1, if someone has chosen 2 as first or as second choice.
62
EVS and WVS: Here is a scale of incomes and we would like to know in what group your household is,
counting all wages, salaries, pensions and other incomes that come in. Just give the letter of the group your
household falls into, after taxes and other deductions (scale from 1 to 10)..
63
EVS and WVS: Size of town:
1. Under 2,000
2. 2,000 - 5,000
3. 5 - 10,000
4. 10 - 20,000
5. 20 - 50,000
6. 50 - 100,000
7. 100 - 500,000
8. 500,000 and more.

   
108
 

dependent variable. The marginal effect indicates the change in the share of individuals (or
the probability of) belonging to a specific justifiability level, when the independent variable
increases by one unit64. In all estimations the marginal effects are presented only for the
highest value. Furthermore, it should be noted that answers such as “don’t know” and
missing values have been eliminated in all estimations.
First, we differentiate between Western and Eastern Europe as the reform process in the
transition countries has resulted in disorientation and a heavy economic burden (Kasper and
Streit (1999) and Gërxhani (2002)). The rapid collapse of institutional structures produced a
vacuum in many countries that led to large social costs, especially in terms of worsening
income inequality and poverty rates and bad institutional conditions based on uncertainty and
high transaction costs. Alm, Martinez-Vazquez and Torgler (2006) report that governments
faced difficult policy choices in this new era regarding the role of the public sector in general
and the structure of the tax system in particular. Furthermore, Kornai (1990) and Martinez-
Vazquez and McNab (2000) report that citizens in many transition countries were not used to
paying taxes at the beginning of the transition process. Thus, taxpayers may have reacted
strongly to the tax policy changes necessary for the transition from a centrally controlled
economy to a market economy. Moreover, rather than using a dummy variable to
differentiate between Western and Eastern Europe, we consider also consider country fixed
effects in specification (4).
Table 5-1 shows that the higher is the perceived corruption of other persons, the higher is
the justifiability of corruption. The coefficient is always statistically significant at the 1%
level and the size of the effect is substantial; if perceived corruption rises by one unit, the
percentage of persons reporting that corruption is never justified falls between 3.8 and 5.1
percentage points. Thus, we find support that conditional corruption matters.
                                                            
64
Again, it should be noted that higher values are connected to a lower justifiability of corruption.  

   
109
  

Table 5-1 Influence of conditional corruption (EVS)

Dependent Variable: Justifiability Coeff Z-Stat. Marg. Coeff. Z-Stat. Marg. Coeff. Z-Stat. Marg. Coeff. Z-Stat. Marg. Coeff. Z-Stat. Marg.
Of Corruption (1) (2) (3) (4) (5)
(Highest Value = Never Justified)
Models Weighted Ordered Probit Weighted Ordered Probit Weighted Probit Weighted Probit Weighted Probit
(0-3)A (1-10)A (9,10=1), Else 0
Conditional Corruption
Perceived Share Of Compatriots -0.13*** -8.61 -0.046 -0.14*** -9.29 -0.049 -0.110*** -6.85 -0.038 -0.151*** -8.79 -0.051 -0.184*** -9.62 -0.045
Accepting A Bribe (SAB)
Formal And Informal Education
Formal 0.006*** 3.07 0.002 0.007*** 3.24 0.002 0.006*** 2.95 0.002 0.004* 1.67 0.001 0.008*** 2.77 0.002
Political Interest 0.063*** 5.11 0.022 0.061*** 5.07 0.021 0.050*** 3.95 0.017 0.045*** 3.38 0.015 0.075*** 4.92 0.018
Demographic Factors
Age 30-39 0.20*** 5.85 0.066 0.19*** 5.87 0.065 0.20*** 5.55 0.066 0.22*** 5.99 0.073 0.24*** 5.84 0.053
Age 40-49 0.30*** 8.16 0.098 0.29*** 8.01 0.095 0.30*** 7.78 0.098 0.36*** 8.83 0.11 0.36*** 8.16 0.078
Age 50-59 0.34*** 8.49 0.11 0.34*** 8.56 0.11 0.34*** 8.17 0.11 0.43*** 9.79 0.13 0.43*** 8.84 0.088
Age 60-69 0.50*** 11.9 0.15 0.50*** 11.76 0.15 0.50*** 11.22 0.15 0.61*** 13.17 0.18 0.60*** 11.90 0.12
Age 70+ 0.59*** 11.37 0.17 0.57*** 11.04 0.17 0.57*** 10.43 0.17 0.71*** 12.64 0.19 0.75*** 11.59 0.13
Female 0.13*** 5.99 0.046 0.13*** 6.15 0.046 0.14*** 6.01 0.048 0.16*** 6.82 0.055 0.15*** 5.48 0.035
Marital Status
Married 0.13*** 4.09 0.044 0.13*** 4.2 0.045 0.15*** 4.43 0.051 0.12*** 3.51 0.041 0.097*** 2.60 0.024
Widowed 0.15*** 2.78 0.049 0.15*** 2.84 0.049 0.16*** 2.8 0.052 0.11* 1.83 0.035 0.11* 1.79 0.026
Divorced 0.024 0.52 0.008 0.018 0.39 0.006 0.049 1.01 0.017 0.036 0.73 0.012 0.015 0.27 0.004
Separated 0.0002 0 0.0001 -0.052 -0.52 -0.018 0.066 0.67 0.022 0.001 0.01 0 -0.11 -1.00 -0.027
Employment Status
Selfemployed 0.031 0.69 0.01 0.032 0.75 0.011 0.036 0.77 0.012 -0.12** -2.51 -0.043 -0.12** -2.13 -0.030
Risk Attitudes
Risk Averse 0.24*** 10.56 0.083 0.23*** 10.61 0.081 0.24*** 10.12 0.084 0.090*** 3.58 0.031 0.069** 2.49 0.017
Urbanization
Urbanization -0.009** -2.09 -0.003 -0.008* -1.91 -0.003 -0.010** -2.06 -0.003 -0.009* -1.74 -0.003 -0.009* -1.69 -0.002
Religiosity
Church Attendance 0.042*** 9.45 0.015 0.042*** 9.54 0.014 0.041*** 8.7 0.014 0.013** 2.48 0.005 0.024*** 3.94 0.006
Geographic Region
Western Europe 0.17*** 7.49 0.058 0.17*** 7.82 0.059 0.16*** 6.83 0.056
Country Fixed Effects NO NO NO YES YES
Pseudo R2 0.036 0.031 0.047 0.099 0.11
Number Of Observations 18168 18168 18168 18168 18168
Prob > Chi2 0.00 0.00 0.00 0.00 0.00
Notes: The reference group consists of Age<30, Man, Single/Living Together, Other Employment Status. ***, ** and * denote significance at 1%, 5% and 10%, respectively. a marginal
effects for the highest value reported (never justified). Robust standard errors.

    110
 

Looking at the other variables we find support for results from previous studies in the
literature. In line with Dong and Torgler (2009) we observe that political interest is
negatively correlated with the justifiability of corruption. An increase in the political interest
scale by one unit increases the probability of stating that taking bribes is never justified by
around 1.5 percentage points. There is also a positive correlation between education and the
social norm of compliance. However, the effect is less strong (see, e.g., specification (4)).
Moreover, we also observe that older people and women exhibit a higher willingness to
comply. These results support previous findings that explored in detail a potential age65 and
gender effect66. Married and widowed people report the lowest justifiability of corruption.
The coefficients are statistically significant in relation to the control group (singles). On, the
other hand, we do not observe that the employment status matters. However, religion is
correlated with corruption. The church as an institution induces behavioural norms and moral
constraints among their community (Torgler 2006). Religiosity seems to affect the degree of
rule breaking. Religiosity can thus be a restriction on engaging in corrupt activities.
Interestingly, we also observe that risk aversion matters. Risk averse people are less likely to
justify corruption which is consistent with suggestions in the compliance literature that risk
aversion reduces the incentive to act illegally. In our model it can be explained by
introducing risk aversion and some uncertainty of C with respect to b. Controlling for risk
attitudes allows for better insights regarding the variables of age, gender, or economic
situation. For example, it could be argued that the obtained difference between women and
men or between different age groups is influenced by different risk attitudes functions.
Hartog et al. (2002), e.g., conducted an empirical survey analysis and found that an increase
in income reduces risk aversion. The estimated coefficient for the Western Europe dummy
suggests that the institutional crisis in many transition countries in Eastern Europe after the
collapse of communism tended to have a positive effect on citizens’ justifiability of
corruption. The marginal effects indicate that being a citizen of a Western European country
rather than an Eastern European country increases the probability of responding that
corruption is never justified by more than 5 percentage points. Finally, we also explore
                                                            
65
 For example, Torgler and Valev (2006) investigate the willingness of being corrupt of the same cohorts over
time (age effect) as well as and the same age groups in different time periods (cohort effect). All in all they
observe a consistent age effect. On the other hand, a cohort effect is less obvious.
66
Torgler and Valev (2007) explored whether gender matters and whether a decrease of gender differences with
greater equality of status and better opportunities affects their willingness to comply. They find evidence for
strong gender differences. Women are significantly less likely to agree that corruption and cheating on taxes can

111 
 
 

whether urbanization matters. It has been argued that corruption may be higher in larger cities
due to the fact that the scale of economic activities is larger and more varied in scope
resulting in a higher level of government contacts. Moreover, government officials may be
less personal compared to those in smaller cities which may reduce the opportunity costs of
bribing (Mocan 2008). Table 5-1 shows that the coefficient is statistically significant at the 5
or 10% percent level with a marginal effect of 0.3 percentage points. Thus, this relationship
cannot be rejected although it should be noted that the effect is smaller in relation to other
factors.
In the last specification in Table 5-1 we go beyond the original probit model (1=never
justified) as the answer to the question might be biased by experimenter demand. It is
obvious that the “socially correct” answer would be “never justified”. Such a concern arises if
a large number people who think that bribing is justified were to instead claim that bribing is
never justified. Thus, in other words, if the respondent wants to give the “socially acceptable”
answer he would say “1” and if not he would answer truthfully. In this latter case, an answer
of “0” might be indicative of a much higher social norm than an answer of “1”. In this case
we would have a problem that respondents want to avoid looking bad in front of the
interviewer (Bertrand and Mullainathan 2001). It would also indicate that we would observe
systematic biases rather than just random errors. We therefore try a different cut-off point.
We report a probit model where we convert the values 1 and 2 to 1 (all the other values = 0).
The results in Table 5-1 indicate that conditional corruption matters showing similar
quantitative effects.
In Table 5-2 we conduct further robustness checks. We report only the findings using a
probit model as Table 5-1 has shown that the probit model provides higher Pseudo R2 values.
First, we try to better isolate a conditional cooperative effect by adding Generalized Trust as
a variable.67 Specification (5) shows that the trust coefficient is not statistically significant.
On the other hand, our conditional corruption variable (SAB) remains highly statistically
significant with a marginal effect of 5 percentage points. In a next step we add income68 as a
further variable. We have added the variable sequentially in the specification as the number
of observations decreases once you control for household income. Also here we
                                                                                                                                                                                         
be justified. The results remain robust after investigating different time periods and extending the specification
with several opportunity factors.  
67
 Generally speaking, would you say that most people can be trusted or that you can’t be too careful in your
dealings with people? (1=most people can be trusted, 0=can’t be too careful.).
68
As discussed this is a ten-point income scale from 1 to 10 (10-quantiles).  

   
112
  

Table 5-2 Robustness test and the influence of conditional corruption using micro and macro proxies (EVS)
Dependent Variable: Coeff z-Stat. Marg. Coeff. z-Stat. Marg. Coeff. z-Stat. Marg. Coeff. z-Stat. Marg. Coeff. z-Stat. Marg.
Justifiability Of Corruption (6) (7) (8) (9) (10)
Probit Clustering on countries Clustering on countries Clustering on countries
Conditional Corruption
SAB -0.15*** -8.39 -0.050 -0.16*** -8.56 -0.054 -0.16*** -6.65 -0.054
Control of Corruption 0.21*** 31.59 0.069 0.22*** 41.70 0.073 0.24*** 52.11 0.080
Formal And Informal Education
Formal 0.004 1.60 0.001 0.004 1.41 0.001 0.004 0.89 0.001 0.001 0.47 0.000 0.003 0.87 0.001
Political Interest 0.047*** 3.49 0.016 0.056*** 3.85 0.019 0.056*** 5.36 0.019 0.041*** 4.34 0.014 0.032*** 3.05 0.011
Demographic Factors
Age 30-39 0.21*** 5.60 0.069 0.21*** 5.12 0.067 0.21*** 6.05 0.067 0.18*** 6.35 0.059 0.19*** 6.01 0.060
Age 40-49 0.35*** 8.57 0.11 0.36*** 8.20 0.11 0.36*** 8.74 0.11 0.27*** 7.23 0.085 0.28*** 7.99 0.086
Age 50-59 0.42*** 9.40 0.13 0.43*** 9.09 0.13 0.43*** 9.12 0.13 0.38*** 9.21 0.12 0.36*** 8.38 0.11
Age 60-69 0.61*** 13.01 0.18 0.61*** 11.98 0.17 0.61*** 10.09 0.17 0.51*** 10.71 0.15 0.51*** 10.52 0.15
Age 70+ 0.72*** 12.68 0.19 0.71*** 11.63 0.19 0.71*** 8.33 0.19 0.62*** 10.01 0.17 0.62*** 10.49 0.17
Female 0.16*** 6.68 0.054 0.15*** 5.70 0.049 0.15*** 7.78 0.049 0.16*** 10.05 0.054 0.18*** 12.45 0.058
Marital Status
Married 0.12*** 3.41 0.041 0.13*** 3.37 0.043 0.13*** 3.30 0.043 0.14*** 4.67 0.047 0.14*** 5.12 0.045
Widowed 0.10* 1.73 0.033 0.079 1.28 0.026 0.079 1.61 0.026 0.13*** 3.27 0.041 0.12*** 3.09 0.038
Divorced 0.027 0.54 0.009 0.029 0.53 0.009 0.029 0.47 0.009 0.029 0.74 0.009 0.024 0.65 0.008
Separated 0.000 0.00 0.000 -0.044 -0.41 -0.015 -0.044 -0.58 -0.015 0.035 0.49 0.012 0.059 1.01 0.019
Employment Status
Selfemployed -0.13** -2.54 -0.044 -0.10* -1.78 -0.034 -0.10 -1.25 -0.034 -0.092* -1.79 -0.031 -0.12*** -2.72 -0.039
Risk Attitudes
Risk Averse 0.090*** 3.49 0.031 0.12*** 4.42 0.041 0.12*** 3.49 0.041 0.087*** 3.01 0.029 0.071*** 2.60 0.024
Urbanization
Urbanization -0.008 -1.55 -0.003 -0.001 -0.27 0.000 -0.001 -0.20 0.000 -0.010* -1.72 -0.003 -0.012** -2.27 -0.004
Religiosity
Church Attendance 0.014** 2.50 0.005 0.015** 2.53 0.005 0.015** 1.98 0.005 0.011* 1.91 0.003 0.011** 2.06 0.004
Trust
Generalized Trust -0.024 -0.90 -0.008 -0.025 -0.89 -0.008 -0.025 -0.78 -0.008 -0.019 -0.78 -0.006 -0.024 -1.01 -0.008
Economic Situation
Income -0.021*** -3.65 -0.007 -0.021** -2.40 -0.007 -0.012 -1.58 -0.004
Geographic Region
Country Fixed Effects YES YES YES YES YES
Pseudo R2 0.10 0.11 0.11 0.10 0.098
Number of observations 17537 15395 15395 28989 34475
Prob > chi2 0.00 0.00 0.00 0.00 0.00
Notes: The reference group consists of Age<30, Man, Single/Living Together, Other Employment Status. ***, ** and * denote significance at 1%, 5% and 10%, respectively. Robust
standard errors.
    113
 

observe a robust result. The variable SAB is statistically significant at the 1% level and the
quantitative effect even increases. Income is also statistically significant with a negative sign.
In specifications (8) to (10) we provide an interesting extension. We introduce a further
corruption variable into the specification. However, compared to SAB, Control of Corruption
measures the perceived level of corruption at the macro level69. As it can be criticized that
including an aggregated variable in a micro data set may produce downward biased standard
errors, we provide estimations with standard errors adjusted to clustering on countries.
Specification (8) shows that both corruption variables are statistically significant at the 1%
level with high marginal effects. An increase in the Control of Corruption scale by one unit
increases the probability of reporting that corruption is never justified by 6.9 percentage
points. On the other hand, the marginal effect for the SAB is consistent with the previous
findings. In specification (9) we do not include SAB to maximize the number of available
countries in the data set as SAB has not been collected in all the countries that participated in
the EVS 70 . In specification (10) we also neglect Income, to increase the number of
observations. The results are robust and the marginal effects are even higher (between 7.3
and 8.0 percentage points).
Causality remains an issue because one’s own justifiability of corruption may lead to the
expectation that others behave in the same way. However, results from strategy method
experiments done by Fischbacher et al. (2001) and Fischbacher and Gächter (2006) that
carefully investigate the causality problem suggest that causality goes from beliefs about
others’ cheating to one’s own behaviour rather than vice versa. In our empirical work, we
also present several two-stage least squares (2SLS) estimations with different instruments and
include several diagnostic tests to deal with the causality problem. However, we test for the
relevance and validity of the instruments and the overidentifying restrictions. Moreover, we
try to filter out a possible systematic bias in our conditional corruption effect by correcting
for differences between what an individual thinks and what that individual projects on others.
This provides a possible way of correcting parts of such a potential bias.

                                                            
69
 Again, higher values are in line with a lower level of corruption.
70
This allows to move from 15 to 30 countries.  

114 
 
  

Table 5-3 2SLS results (EVS)

Dependent Variable: Coeff t-Stat. Coeff. t-Stat. Coeff. t-Stat. Coeff. t-Stat.
Justifiability Of Corruption (11) (12) (13) (14)
Conditional Corruption
Perceived Share Of Compatriots -0.093*** -6.87 -0.091*** -6.63 -0.094*** -6.58 -0.036** -2.24
Accepting A Bribe (SAB)
Formal And Informal Education
Formal 0.001 1.18 0.001 1.12 0.001 0.78 0.001 1.49
Political Interest 0.014*** 3.24 0.014*** 3.25 0.017*** 3.73 0.018*** 3.95
Demographic Factors
Age 30-39 0.064*** 4.81 0.060*** 4.37 0.058*** 4.03 0.062*** 4.40
Age 40-49 0.11*** 7.81 0.11*** 7.51 0.11*** 7.17 0.11*** 7.91
Age 50-59 0.12*** 7.97 0.12*** 7.49 0.12*** 7.42 0.13*** 8.28
Age 60-69 0.18*** 11.56 0.18*** 11.41 0.17*** 10.46 0.18*** 11.53
Age 70+ 0.20*** 10.98 0.20*** 10.89 0.19*** 9.88 0.21*** 11.13
Female 0.055*** 6.87 0.056*** 6.81 0.050*** 5.83 0.050*** 6.01
Marital Status
Married 0.053*** 4.25 0.053*** 4.23 0.059*** 4.35 0.052*** 3.95
Widowed 0.042** 2.23 0.042** 2.18 0.040** 1.98 0.037* 1.92
Divorced 0.030* 1.67 0.030 1.64 0.034* 1.78 0.017 0.91
Separated 0.025 0.71 0.026 0.71 0.018 0.48 -0.002 -0.06
Employment Status
Selfemployed -0.037** -2.31 -0.037** -2.27 -0.027 -1.53 -0.032* -1.90
Risk Attitudes
Risk Averse 0.036*** 4.02 0.038*** 4.17 0.050*** 5.07 0.047*** 4.95
Urbanization
Urbanization -0.003 -1.64 -0.003 -1.52 -0.001 -0.35 -0.002 -0.95
Religiosity
Church Attendance 0.006*** 3.46 0.006*** 3.48 0.006*** 3.31 0.006*** 3.33
Economic Situation
Income -0.006*** -3.01 -0.006*** -3.25
Geographic Region
Country Fixed Effects YES YES YES YES
First stage regressions:
Index of Perceived Honesty I 0.74*** 52.91 0.73*** 51.63 0.73*** 48.67
Index of Perceived Honesty II 0.54*** 38.89
Generalized Trust -0.054*** -4.54 -0.058*** -4.57 -0.068*** -5.30
F-Test of excluded instruments 2799*** 1364*** 1214*** 784***
Anderson canon. corr. likelihood 3481*** 3388*** 2991*** 2105***
ratio stat.
Anderson-Rubin test 47.60*** 47.05*** 45.96*** 5.44*
Hansen J statistic 1.80 1.59 0.36
Number of observations 15755 15248 13331 14281
Prob > F 0.00 0.00 0.00 0.00
Notes: The reference group consists of Age<30, Man, Single/Living Together, Other Employment Status. ***, ** and * denote
significance at 1%, 5% and 10%, respectively. Robust standard errors.

To check for robustness, we are going to use a variety of instruments in our 2SLS
regressions reported in Table 5-3. In specification (11), we use an index of perceived honesty

   
115
 

as an instrument for SAB71. In the second one, we also use Generalized Trust as an instrument.
A seen previously, generalized trust did not affect the justifiability of corruption. However, as
Table 5-3 shows generalized trust is a good instrument for conditional corruption. Next, we
also consider a second index of perceived honesty72. The results indicate that the variable
SAB remains statistically significant in all the 2SLS. Table 5-3 also reports the results of the
Anderson canonical correlation likelihood-ratio test to test whether the equation is identified
as a measure of instrument relevance. The test shows that the null hypothesis can be rejected,
indicating that the model is identified and the instruments are relevant in all cases. Table 5-3
further shows that the F-tests for the instrument exclusion set in the first-stage regression are
statistically significant in all cases. In addition, we test for the validity of the instruments
using a Hansen test of overidentifying restrictions. Table 5-3 indicates that the null
hypothesis that the excluded instruments are not correlated with the error term cannot be
rejected. Thus, the results confirm the validity of the instruments.

Table 5-4 Causality discussion (Filtering)

Depend. V.: Justifiability of Corruption Coeff. z-Stat. Marg. Effects
(Highest Value = Never Justified) Weighted Probit
INDEPENDENT V. (see specifications)
Specification (15)
Filtered SAB using specification (4) -0.090*** -5.21 -0.030

Specification (16)
Filtered SAB using specification (5) -0.086*** -4.89 -0.029

Specification (17)
Filtered SAB using specification (6) -0.10*** -5.33 -0.034

Specification (18)
Filtered SAB using specification (7) -0.10*** -5.25 -0.034
Control of Corruption 0.21*** 32.32 0.071
Notes: Summary of four regressions. The reference group consists of Age<30, Man,
Single/Living Together, Other Employment Status. ***, ** and * denote significance at 1%,
5% and 10%, respectively. Robust standard errors.

                                                            
71
 Index covering the average value of the following questions: According to you, how many of your
compatriots do the following: Claiming state benefits to which they are not entitled. Cheating on tax if they have
the chance. Paying cash for services to avoid taxes. Speeding over the limit in built-up areas. Taking the drug
marijuana or hash. Driving under the influence of alcohol (scale from 1 to 4).
72
Index covering the average value of the following questions: Speeding over the limit in built-up areas. Taking
the drug marijuana or hash. Driving under the influence of alcohol (scale from 1 to 4).   

   
116
 

In addition, to deal with a potential causality problem, we filter out a possible bias in the
conditional cooperative effect. Again, the causality problem may arise because an
individual’s justifiability of bribing might lead to the expectation that others behave in the
same way. Thus, individuals with a higher social norm of compliance have a lower
perception that others are bribing. To deal with this possibility, we calculate first the average
perceived corruption for each country. In the next step, we calculate the average perceived
corruption in each country for individuals having the lowest justifiability of corruption. In a
further step, we build the difference between both average values. This variable may measure
a particular bias in perceived corruption due to the level of social norms. In a last step, we
add this bias to the individual values of the group with the lowest justifiability of corruption
values. As a consequence, each of the individuals with the highest social norm of compliance
now has higher perceived corruption values. Hence, the values between the group with the
higher and lower justifiability of corruption values are brought closer together, depending on
the perceived corruption situation in each country. This procedure may help to better isolate
the existence of a conditional corruption. Table 5-4 presents the results for the filtered
perceived corruption variable using specifications in line with Table 5-1 and 5-2. The
coefficient remains highly statistically significant and, although the marginal effects have (in
general) decreased from previous estimates, they still are very high.
5.4.2 Micro Level using the WVS
In a next step we are going to use an alternative data source to check whether the previously
obtained results remain robust. As discussed, we are using a slightly different proxy for
conditional corruption. The WVS provides the possibility to explore a large set of countries
and further regions. This also provides the opportunity to explore the relevance of conditional
corruption at the macro level. We work with average values within each country using for our
dependent variable the 4 point scale (0 to 3). Figure 5-1 shows a relatively strong negative
correlation (Pearson r=-0.42), significant at the 0.01 level. Looking at the linear relationship
in a simple regression shows that conditional corruption can explain 18 percent of the total
variance of the justifiability of corruption.
In general, empirical support for a theoretical foundation depends not only on the validity
of the theory but also on the quality of the data. It is not possible to ascertain with survey data
whether respondents are truthful in their answers as truth is not observable by the
interviewers (Kanazawa 2005). To validate statements one could explore the correlation
between respondents’ statements and the Control of Corruption variable at the macro level

   
117
 

using country averages. Figure 5-3 shows an expected positive correlation (Pearson r=0.38)
statistically significant at the 0.05
Working with the WVS we are also able to control for the similar independent
variables 73 . Table 5-5 presents the first results. We explore regressions with regional or
country fixed effects. Moreover, we provide evidence with and without the income variable.
In addition, we also include sequentially the macro corruption variable Control of Corruption.
In all the specifications the variable Perceived Level of Corruption is statistically significant
with marginal effects between 0.6 and 3.5 percentage points. In addition, the macro variable
Control of Corruption is also statistically significant with marginal effects close to 6
percentage points. To deal with the social desirability problem we also change the cut-off
point (see last specification in Table 5- 1). The values 1 and 2 in the original scale have been
coded as 1 and all other values as 0. The coefficient is highly statistically significant,
reporting even larger marginal effects than comparable results in specification (20). Thus, we
can conclude that conditional corruption is also observable when using alternative data
sources. The control variables show similar tendencies. A higher level of political interest is
correlated with a lower justifiability of corruption. Risk averse and married people are also
less inclined to justify corruption. On the other hand, self-employed individuals are more
likely to justify corruption. Similarly, we also observe an age and gender effect. However, the
effects of religiosity, urbanization and income are less strong.
                                                            
73
 See definition of the variables in previous footnotes.  

   
118
  

Table 5-5 Conditional corruption using WVS

Dependent Variable: Justifiability Coeff z-Stat. Marg. Coeff. z-Stat. Marg. Coeff. z-Stat. Marg. Coeff. z-Stat. Marg. Coeff. z-Stat. Marg. Coeff. z-Stat. Marg.
Of Corruption (19) (20) (21) (22) (23) (24)
Weighted Probit Change of cut-off point
Conditional Corruption
Perceived Level Of Corruption -0.112*** -11.90 -0.035 -0.020* -1.81 -0.006 -0.034*** -2.81 -0.010 -0.035** -2.50 -0.011 -0.043*** -2.95 -0.013 -0.041*** -3.35 -0.009
(Plc)
Control Of Corruption 0.19*** 15.18 0.059 0.19*** 14.09 0.058
Formal And Informal Education
Politicial Interest 0.025*** 3.02 0.008 0.031*** 3.55 0.010 0.021** 2.17 0.006 0.036*** 3.15 0.011 0.030** 2.52 0.009 0.051*** 5.18 0.011
Formal -0.002 -0.64 -0.001 0.003 0.77 0.001 0.006 1.25 0.002 0.005 0.95 0.002 0.010* 1.75 0.003 0.006 1.34 0.001
Demographic Factors
Age 30-49 0.20*** 9.91 0.064 0.20*** 9.20 0.059 0.19*** 8.51 0.059 0.18*** 6.52 0.055 0.18*** 6.43 0.057 0.19*** 8.36 0.043
Age 50-64 0.41*** 15.37 0.12 0.40*** 14.46 0.11 0.40*** 13.50 0.11 0.38*** 10.83 0.11 0.38*** 10.23 0.11 0.39*** 12.87 0.077
Age 65+ 0.57*** 16.11 0.15 0.53*** 14.45 0.14 0.53*** 13.04 0.14 0.55*** 11.50 0.14 0.53*** 10.71 0.14 0.53*** 12.55 0.094
Female 0.14*** 8.89 0.045 0.14*** 8.67 0.044 0.13*** 7.40 0.041 0.16*** 7.58 0.051 0.15*** 6.83 0.048 0.16*** 8.61 0.035
Marital Status
Married 0.10*** 4.99 0.033 0.12*** 5.67 0.038 0.14*** 5.81 0.042 0.13*** 4.64 0.041 0.13*** 4.53 0.042 0.13*** 5.35 0.029
Widowed 0.086** 2.18 0.026 0.084** 2.01 0.025 0.092** 2.04 0.028 0.095* 1.87 0.029 0.072 1.36 0.022 0.098** 2.08 0.021
Divorced 0.020 0.48 0.006 -0.001 -0.02 0.000 0.029 0.64 0.009 0.026 0.49 0.008 0.024 0.43 0.007 -0.0001 0.00 0.000
Separated 0.069 1.22 0.021 0.015 0.26 0.005 0.008 0.14 0.003 0.066 0.87 0.020 0.054 0.69 0.017 -0.0002 0.00 0.000
Employment Status
Selfemployed -0.078*** -2.64 -0.025 -0.088*** -2.87 -0.028 -0.10*** -3.09 -0.032 -0.098*** -2.65 -0.032 -0.11*** -2.70 -0.034 -0.087** -2.59 -0.020
Risk Attitudes
Risk Averse 0.084*** 4.95 0.026 0.071*** 4.03 0.022 0.068*** 3.49 0.021 0.080*** 3.43 0.025 0.075*** 3.07 0.023 0.073*** 3.70 0.016
Urbanization
Urbanization 0.001 0.40 0.000 -0.008** -2.30 -0.002 -0.007* -1.80 -0.002 -0.010** -2.40 -0.003 -0.008* -1.77 -0.002 -0.005 -1.23 -0.001
Religiosity
Church Attendance 0.0002 0.04 0.0001 0.018*** 3.86 0.006 0.017*** 3.40 0.005 -0.003 -0.58 -0.001 -0.005 -0.86 -0.002 0.021*** 4.02 0.005
Economic Situation
Income -0.006 -1.42 -0.002 -0.011** -2.48 -0.004
Regional Fixed Effects YES NO NO YES YES NO
Country Fixed Effects NO YES YES NO NO YES
Pseudo R2 0.038 0.082 0.11 0.059 0.058 0.099
Number of observations 37759 37759 32096 20793 18914 37759
Prob > chi2 0.00 0.00 0.00 0.00 0.00 0.00
Notes: The reference group consists of Age<30, Man, Single/Living Together, Other Employment Status. ***, ** and * denote significance at 1%, 5% and 10%, respectively. Robust standard errors.

119 
 
 

Table 5-6 2SLS results (WVS)

Dependent Variable: Justifiability of Corruption Coeff t-Stat. Coeff. t-Stat.
(25) (26)
Conditional Corruption
Perceived Corruption (PLC) -0.044** -2.43 -0.061*** -3.04
Formal And Informal Education
Political Interest 0.007** 2.35 0.003 0.84
Formal -0.001 -0.94 0.000 -0.29
Demographic Factors
Age 30-49 0.07*** 9.32 0.066*** 8.53
Age 50-64 0.13*** 14.96 0.13*** 14.19
Age 65+ 0.17*** 16.23 0.17*** 14.87
Female 0.045*** 8.85 0.042*** 7.66
Marital Status
Married 0.039*** 5.27 0.042*** 5.28
Widowed 0.032*** 2.83 0.040*** 3.35
Divorced 0.012 0.89 0.027* 1.94
Separated 0.031* 1.67 0.033* 1.68
Employment Status
Selfemployed -0.024** -2.43 -0.031*** -2.94
Risk Attitudes
Risk Averse 0.027*** 5.14 0.022*** 3.79
Urbanization
Urbanization 0.001 0.81 0.001 0.91
Religiosity
Church Attendance -0.0002 -0.15 -0.002 -1.16
Economic Situation
Income 0.0004 0.34
Country Fixed Effects YES YES
First stage regressions:
Generalized Trust -0.30*** -28.14 -0.30*** -25.68
F-Test of excluded instruments 791.73*** 659.46***
Anderson canon. corr. likelihood ratio stat. 956.50*** 810.50***
Anderson-Rubin test 5.88** 9.20***
Number of observations 36296 30968
Prob > F 0.00 0.00
Notes: The reference group consists of Age<30, Man, Single/Living Together, Other Employment Status. ***,
** and * denote significance at 1%, 5% and 10%, respectively. Robust standard errors.

Table 5-6 presents 2SLS estimations using generalized trust as an instrument for perceived
corruption (in line with Table 5-3). Looking at the first stage regressions and the diagnostic
tests we can conclude that generalized trust is a good instrument74. The results also show that
Perceived Corruption (PLC) remains statistically significant, providing therefore further
support for previous findings. We report additional findings in Table 5-7 obtained with a
filtered PLC variable using previous specifications. Also here we observe that the PLC
coefficient is always statistically significant with a negative sign. Thus, even after filtering we
can conclude that conditional corruption matters.
                                                            
74
 The WVS does not provide the possibility to consider an index of perceived honesty.  

120 
 
 

Table 5-7 Causality discussion filtering with WVS Data

Depend. V.: Justifiability of Corruption Coeff. z-Stat. Marg. Effects
(Highest Value = Never Justified) Weighted Probit
INDEPENDENT V. (see specifications)
Specification (27)
Filtered PLC using specification (19) -0.13*** -12.87 -0.041

Specification (28)
Filtered PLC using specification (20) -0.020* -1.75 -0.006

Specification (29)
Filtered PLC using specification (21) -0.026** -2.08 -0.008

Specification (30)
Filtered PLC using specification (22) -0.035** -2.50 -0.011
Control of Corruption 0.19*** 15.18 0.059

Specification (31)
Filtered PLC using specification (23) -0.043*** -2.95 -0.013
Control of Corruption 0.19*** 14.09 0.058
Notes: Summary of four regressions. The reference group consists of Age<30, Man,
Single/Living Together, Other Employment Status. ***, ** and * denote significance at 1%, 5%
and 10%, respectively. Robust standard errors.

We conduct a further robustness test to deal with a potential “social desirability” bias using
the EVS and WVS. We run a two-stage approach where the previous estimations were just the
first stage. First, respondents decide whether or not to answer that corruption is never justified
(“socially correct response”). In a second stage, given the decision to answer something other
than the socially correct response, individuals report a value from the remaining scale (1 to 9).
The results are not reported in a table but indicate that our conditional corruption variable is
always statistically significant. Despite trying to check the causality relationship one should
note that providing a clear causality relationship is quite problematic working with such micro
survey data. To some extent we see these results as more precisely estimated partial
correlations and not fully precise estimates of a causal relationship (see also Guiso et al. 2003).
In the next stage we are going to explore the importance of conditional corruption at the
macro level over time in order to explore hypothesis 2.
5.4.3 Macro Level Using a Large Panel Data Set
In the previous analysis we were only able to explore conditional corruption in a cross-sectional
setting. In this next step, it is highly relevant to bring in the time dimension to see the potential
dynamics of conditional corruption. This requires the use of a panel data set. Therefore, we are
going to work with a large international panel data set that covers 18 years (1986 till 2003). As

121
 
 
 

discussed previously, we use the ICRG data to get a measurement of (perceived) corruption.
The panel analysis will help us to see whether corruption is contagious. Sah (2005, p. 6), e.g.,
stresses “If past experiences have convinced some citizens that corruption is more pervasive in
the economy, then they are more likely to cheat. Likewise, if their past experiences have
convinced some bureaucrats that cheating is more pervasive in the economy, then they are more
likely to choose to be corrupt…Through these dynamic relationships, future levels of cheating
and corruption in the economy become explicitly linked to past levels of cheating and
corruption in the economy…”. A contagion effect can increase the demand for corruption as
individuals perceive additional opportunities for bribing (Goel and Nelson 2007). It can also
affect the supply of corruption as potentially corrupt bureaucrats are aware of the high
probability that one can be corrupt without being caught and penalized. Moreover, bureaucrats
could also try to introduce lax enforcement and punishment strategies for corruption (Goel and
Nelson 2007). Similarly, the criminal literature has stressed that the prevalence of a given type
of criminal behaviour may change the propensity of others to engage in that same behaviour. It
affects the perceptions about the net return of such a behaviour (information function) and also
the probability of arrests or constraints (Ludwig and Kling 2006, Cook and Goss, 1996; Becker
and Murphy, 2000; Manski, 1993, 2000). A contagion effect in crimes has been observed in the
areas of assassinations, hijackings, kidnappings, and serial murders (Bikhchandi, Hirshleifer
and Welch 1998). However, only a limited number of studies have provided empirical evidence
focusing on corruption. Evidence is mainly available from the US. Goel and Nelson (2007), use
state-level U.S. data in a cross-sectional analysis data over the period 1995–2004, and find that
the effect of neighbouring corruption is positive and statistically significant - showing therefore
that corruption does appear to be contagious. A 10% increase in corruption in neighbouring
states appears to increase corruption in a state somewhere in the range of 4–11%. Our study
adds to the literature by using panel data and exploring a contagion effect at the international
level. In line with studies on contagion in general, we are going to use lagged values to explore
its importance75. To isolate such an effect we are going to control for further factors such as law
and order 76 , democratic accountability 77 , economic performance, or the level of openness.
Discretion in the application of rules enhances corruption. On the other hand, a strong legal
system that penalizes deviance reduces the incentives to act illegally (high Law and Order
                                                            
75
 Similarly, Becker (1996) stresses in a general framework that individual’s consumption (Ci) depends on that of
other individuals in the past (Cjt-1) and on individual I consumption in a previous period (Cit-1).  
76
 The ‘law’ sub-component measures the strength and impartiality of the legal system, while the ‘order’ sub-
component is an assessment of popular observance of the law.
77
Measures how responsive the government is with its people.  

122
 
 
 

value). In addition, a more encompassing and legitimate state increases the willingness to
contribute. If citizens perceive that their interests (preferences) are properly represented in
political institutions and they receive an adequate voice, their identification with the state
increases, their willingness to be corrupt decreases (high Democratic Accountability value).
Moreover democratic accountability helps to limit the abuse of political power by selfish
politicians, when citizens cannot completely foresee incumbents’ preferences, elements of
direct democracy also empower them with an instrument for controlling the government. Levi
(1988) points out that a possible method of creating or maintaining compliance is t provide
reassurance by the government. A government that precommits itself with direct democratic
rules imposes restraints on its own power and thus sends a signal that taxpayers are seen as
responsible persons. Voting possibilities also provide utility in themselves. Citizens value the
right to participate, because it produces a kind of procedural utility as the opportunity set
increases which fosters the moral costs of behaving illegally and enhances rule obedience
(Torgler and Schneider 2009). Moreover, political involvement and political attention is
correlated with income as political attention may be a luxury good and therefore people pay
more attention to corrupt activities and are better able to take actions against these officials
(Glaeser and Saks 2006). Thus we would observe a negative correlation between GDP per
capita and corruption. Moreover, economic rents will decrease with a higher level of economic
competition. Ades and Di Tella (1999) find that corruption is higher in countries where
domestic firms are protected from foreign competition. We use data provided by Dreher (2006)
that measure three main dimensions of openness: economic, social and political globalization.
The overall index of globalization covers not less than 23 variables.
Thus, to test our second hypothesis, we propose the following baseline equations:

CRit =  + 1 CTRLit +2 CRi(t-1) +3 LOit+ 4 DAit + 4 GLit + TDt +REGIONi + it

where i indexes the countries in the sample and t denotes the time period. CRit denotes the level
of corruption (higher values=lower level corruption) and CRi(t-1) is the one year lag of
corruption. LOit is our law and order variable, DAit the proxy for democratic accountability and
GLit the proxy for globalization. The regressions also contain several control variables, CTRLit,
including GDP per capita and the population size. We control for time as well as regional
invariant factors including fixed time, TDt, and fixed regional effects, REGIONi78. it denotes
the error term.
                                                            
78
 We differentiate between Europe, Latin America, North America, North Africa, Sub Saharan Africa, Pacific,
Asia, Caribbean and Australia. 

123
 
 
 

Table 5-8 presents the results. We report beta or standardized regression coefficients to
compare magnitude, which reveals the relative importance of the variables used. To obtain
robust standard errors in these estimations, we use the Huber/White/Sandwich estimators of
standard errors. First we only include our lagged corruption variable. The coefficient is highly
statistically significant with a high beta coefficient. We observe that the lagged corruption value
together with time and regional fixed effects already explain more than 60 percent of the total
variance of the variable corruption. In the next regression, we add our control variables together
with the globalization index as independent variables. The results show that the coefficient for
Corruption(t-1) (CRi(t-1)) is still statistically significant at the 1% level reporting the highest beta
coefficients among the used independent variables. Table 5-8 also shows that economic
development and globalization have a negative influence on corruption. However, it should be
noted that this effect disappears once you control for governance/institutional factors in the
third specification reported in Table 5-8. An increase in Globalization and GDP per capita leads
to a decrease in corruption. Moreover, a faster growing population has a positive effect on
corruption. The third specification introduces governance and institutional factors. We find that
both factors, Law and Order (LOit) and Democratic Accountability (DAit), are statistically
significant. Also here we find that the past level of corruption has the strongest relationship
with our dependent variable, followed by institutional/governance variables and globalization.
Thus, the macro results show that corruption is not independent of the past experiences.
However, one should note that such macro analysis has the disadvantage that we may not only
measure a potential contagion effect but also path-dependency that may not be directly related
to conditional corruption.
In sum, the micro and macro evidence generated in this chapter suggests that social forces
and past experiences matter. Conditional corruption is a key factor in understanding corruption.

124
 
 
 

Table 5-8 Evidence at the macro level

Dependent Variable: Corruption Beta t-stat. Beta t-stat. Beta t-stat.
(higher values = lower level of corruption) (32) (33) (34)
Corruption (t-1) 0.60*** 19.80 0.55*** 15.17 0.46*** 12.78

Rule of Law 0.20*** 7.85

Democratic Accountability 0.18*** 7.52

Log (GDP per capita) 0.11*** 3.39 -0.048 -1.46

Log(population) -0.042*** -2.68 -0.056*** -3.98

Globalization Index 0.22*** 3.13 0.19*** 3.08

Region Fixed Effects Yes Yes Yes

Time Fixed Effects Yes Yes Yes
Prob > F 0.00 0.00 0.00
R2 0.67 0.71 0.75
Number of observations 1439 1059 1059
Notes: Estimations with robust standard errors. OLS coefficients = standardized/beta coefficients. *, ** and ***
denote significance at the 10%, 5% and 1% level, respectively.

5.5 Conclusion

Traditional economics assumes that preferences are independent of the behavior of everyone
else and also independent of past and future consumption. Therefore choices affect only the
agents directly involved. However, in the last few decades economists have paid more attention
to the structure of preferences. For example, social interactions, an aspect that has long been
discussed by important figures such as Adam Smith (1759/1976), Karl Marx (1849), Thorstein
Veblen (1899) or James Duesenberry (1949), have gained importance in economics. In this
chapter we explore whether and to what extent group dynamics or social forces and past
experiences affect corruption. In other words, we explore theoretically and empirically whether
conditional cooperation matters (hypothesis 1) and whether corruption is contagious
(hypothesis 2). We use the notion of “conditional corruption” for these effects. The
experimental economics literature has explored (pro-)social preferences through designs that
implement own and others’ material payoffs. We observe models of reciprocity, inequity
aversion, or altruism in the literature (see Rabin 1993, Charness and Rabin 2002, Fehr and
Schmidt 1999, Bolton and Ockenfels 2000, Andreoni and Miller 2002). We have presented a
theoretical framework that allows derivation of these two hypotheses. The theoretical part is
supplemented with empirical evidence on conditional corruption and contagion. Interestingly,
only a limited number of studies on corruption have explored this question. Similar discussions

125
 
 
 

on social interactions can be found in the crime literature or more specifically in the literature
on information cascades, network externalities, fads, herd behaviour or bandwagon effects (see,
e.g., Banerjee 1992, Bikchandani, Hirshleifer, and Welch 1992, 1998, Katz and Shapiro 1985).
However, as a novelty we present a large amount of empirical evidence that explores this
question in the area of corruption. First we use two data sets at the micro level followed by a
large international panel data set at macro level covering almost 20 years. The results clearly
indicate that the willingness to be corrupt is influenced by the perceived activities of peers and
other individuals. Moreover, the panel data set at the macro level also indicates that the past
level of corruption has a strong influence on the current corruption level which indicates that
contagion matters. The results clearly show that conditional corruption matters. The findings
therefore underscore the relevance of social interactions. The results are of particular
importance in politics as genuine information is weak and incentives to collect information are
limited due to the possibility of free-riding (Wintrobe 2006). When developing policy strategies
it is recommended to take into account that individuals are not acting in isolation. Social
interactions and group dynamics are highly relevant in the understanding of corruption. A
critical mass of cooperative individuals is required to induce a positive dynamic process of
conditional cooperation. On the other hand, a society which has many non-compliant
individuals will inherit a weak social norm which leads to a shift to a non-cooperative situation
similar to a “corruption trap”. Thus, policies should take into account that we may observe a
path-dependent process within a society.

126
 
 
  

Chapter Six Social interaction and Corruption: Within-country Evidence79

6.1 Introduction

Corruption is a widespread phenomenon affecting human societies throughout time and space.
Contemporaneous corruption scandals not only occur in developing countries such as Nigeria,
India, and China but also in developed economies such as France, Germany and United States.
Even in Scandinavian countries, like Sweden and Norway, supposedly free-from-corruption,
managers of state owned companies have been found to be taking bribes (for an overview see
Rose-Ackerman, 1999).
Corruption in the public sector is recognised to be the greatest obstacle to development
(Kaufmann, 1997). A higher level corruption is associated with lower investment and economic
growth (Mauro 1995; World Bank, 1997). Corruption weakens the effect of industrial policies
and induces private sectors to violate tax and regulatory laws. Foreign direct investment is also
depressed by a high level of corruption (Wei, 2000). Anticorruption policies are therefore very
important since corruption can induce great harm to countries. Some stress that bribery may
increase the overall efficiency of an economic system (e.g., Lui, 1985). However, Rose-
Ackerman (1999) argues that issues such as tax evasion, violation of environmental rules,
certification of unqualified people for public benefits, and grants of immunity to organized
crime do not have such an effect. In addition, bureaucrats have an incentive to delay
transactions in order to extract higher payments (see Rose-Ackerman, 1997)
Reducing corruption requires a thorough understanding of its causes. Sizable literature has
emerged to investigate the determinants of corruption. Current research associates corruption
with cultural tradition, economic development, political institutions and government policies.
For example, in his comprehensive cross-country study, Treisman (2000) finds that Protestant
traditions, history of British rule, long exposure to democracy, higher average income and high
levels of imports depress corruption, while decentralization encourages it. Brunetti and Weder
(2003) present evidence that press freedom can control corruption. Using a within-country data
set, Glaeser and Saks (2006) document that economic development and education decrease
corruption while income inequality and racial fractionalization may increase corruption in
America. However, few have explored the impact of social interaction on corruption. A notable
exception is Goel and Nelson (2007) who, with state-level U.S. data between 1995 and 2004,
                                                            
79
 This chapter has been submitted to the Journal of Policy Modelling. 

127
 
 
 

show that the effect of neighbouring corruption on local corruption is significantly positive. In
other words, corruption is contagious. Contagion effects have been observed in other illegal
activities such as assassinations, hijackings, kidnappings, and serial murders as referred to by
Bikhchandi, Hirshleifer and Welch (1998). The relevance of social interaction and crime is
explored by Glaeser, Sacerdote and Scheinkman (1996) who focus on the United States (across
cities and across precincts in New York). The results indicate that social interaction models
provide a framework for understanding variances of cross-city crime rates. Individuals are more
likely to commit crimes when those around them do. Focusing on corruption, Dong, Dulleck
and Torgler (2008) find using cross-sectional micro data that conditional cooperation matters.
The willingness to engage in corruption is influenced by the perceived activities of other
individuals.
In this chapter we explore the effect of social interaction on the incidence of corruption both
theoretically and empirically in the context of China. China is an interesting country to analyse,
not only because it is the largest transitional and developing country, but also because
corruption has become more rampant in China since the economic reform was launched in 1978.
Even the Chinese government has admitted that corruption “is now worse than during any other
period since New China was founded in 1949. It has spread into the Party, into government
administration and into every part of society, including politics, economy, ideology and culture”
(Liang, 1994, p. 122). Such widespread corruption has caused severe consequences in China,
including economic losses estimated to have been between 13.2 and 16.8% of China’s GDP in
the late 1990s (Hu 2001). Not surprisingly, such rampant corruption has generated much
literature, especially in sociology and political science (e.g., White, 1996, and Gong, 2006).
From an economic perspective, Yao (2002) argues that corruption in China is generated by the
Chinese political system, which grants and protects privileges. With a unique corruption
measure, Cai, Fang and Xu (2009) find that corruption has a substantially negative effect on the
productivity of Chinese firms. Nevertheless, there is a lack of studies that comprehensively
analyse the economic underpinnings of corruption in China. We therefore explicitly study the
impact of social interaction on the incidence of corruption, and find a statistically significant
relationship between social interaction and corruption. This suggests that like other crimes, the
incidence of corruption is significantly affected by social interaction. The rest of this chapter is
structured as follows: Section 6.2 presents a theoretical model. Section 6.3 describes our
empirical analysis and results. Section 6.4 concludes the chapter.

128
 
 
  

6.2 Theoretical Model

In this section we investigate theoretically social interaction in the context of corruption.

Following Aidt’s survey (2003), we identify three related theoretical articles, which are
summarized in Table 6-1 below. These articles, though providing stylized facts related to social
interaction, cannot explain thoroughly the effect of social interaction on the incidence of
corruption because they do not introduce social interaction explicitly into their models. In terms
of social interaction theory, Sah (1988, 2007) and Avdvig and Moene (1990) only study the
effect of local interaction, while Lui (1986) simply investigates the effect of global interaction80.

Table 6-1 Literature summary

Crucial point Approach Stylized fact
It is harder to audit corrupt officials in The different levels of
Lui (1986) societies where corruption is more The overlapping-
corruption across regions under
prevalent. generations model
the same deterrence scheme
An individual’s perception of the
corruption level is stochastically
influenced by the real level that he faced
Sah (1988, 2007) in the past, and this perception affects his The overlapping- The different levels of
current and future corrupt act, which in generations model corruption across regions
turn exert stochastic influences on the
current and future real corruption level.
The probability of corruption is related to Simple dynamic model The different levels of
Avdvig and Moene (1990)
its established frequency. corruption across regions

Nevertheless, a growing body of research considering the role of social interaction in

economic outcomes has emerged during the last two decades. According to Zanelia (2004, p. 4),
social interaction is the “direct interdependences, not mediated by markets and enforceable
contracts, between individual decisions and the decisions and characteristics of others within a
common sociological group”. Economic models that have embedded social interactions “seem
particularly adept to solve a pervasive problem in the social science, namely the observation of
large differences in outcomes in the absence of commensurate differences in fundamentals”
(Scheinkman, 2008, p. 2). Sah (1991) states that an individual’s environment influences her
propensity for crime and thereby explains the obvious difference between the crime
participation rates of societal groups with similar economic fundamentals. With two models of
social interaction, Glaeser, Sacerdote and Scheinkman (1996) provide a framework to interpret
the cross-city variation in crime rates.
In line with social interaction research we employ the interactions-based approach (Blume
and Durlauf, 2004) to explore bureaucratic corruption. Specifically, we use the binary choice

                                                            
80
 We will discuss these terms later. 

129
 
 
 

model with social interactions developed by Brock and Durlauf (1995) to model corrupt
behaviour.
We consider a population of I homogenous bureaucrats. In the presence of social
interactions, each bureaucrat chooses one of two actions: corruption or non-corruption, which is
coded by 1, 1 . The space of all possible sets of actions by the population is denoted
by , , . Thus , , , , , represents the choices of all
bureaucrats other than i.
The utility of the bureaucrat i is assumed to be
, 6 1
Here U ωi i ωi is a private component of the utility. U ωi is the deterministic private
utility decided by the bureaucrat i’s choice, which is expressed below, as:
, 1
6 2
, 1
Here represents the bureaucrat’s wage, is the bribe a corrupt bureaucrat accepts and the
probability that his corrupt act is not detected. A corrupt bureaucrat will lose his job and hence

all his income if his corrupt act is detected. Let and , we can

easily rewrite into the form

6 3
is the random private utility independently and identically distributed across bureaucrats.
In our model it represents the moral shock (moral cost) of taking one of the actions. Following
Brock and Durlauf (2001) and Glaeser and Scheinkman (2002), we further assume that is
extreme-value distributed. Thus the difference between 1 and 1 is logistically
distributed,
1 1 1 ; 0 6 4
1
, in (1), however, is the social component of the utility, namely social utility
associated with a bureaucrat’s choice. We assume that it captures a pure conformity effect,
hence,
,
,
2

, 1

, 1 6 5

130
 
 
 

Here , >0 are measures of the disutility of nonconformity. denotes the bureaucrat i’s
subjective expectation of the bureaucrat j’s choice. With the above assumptions we have,

1 1 1

2 2 ,

1
6 6
1 exp 2 ∑ ,

Since the bureaucrats are homogenous, we can assume that , 0 , and

. Thus,
1
1 6 7
1 exp 2
1· 1 1 · 1

6 8

The joint set of choices obeys (because is independently distributed),

1
6 9
1 exp 2

It is obvious that the corrupt decision of a bureaucrat depends on his expectation of others’
decisions. However, there are two different ways in which each bureaucrat interacts with others,
namely local interaction and global interaction. According to Brock and Durlauf (2001), local
interaction means that each bureaucrat interacts directly only with his neighbourhood in the
population, while global interaction implies that each bureaucrat interacts directly with every
other bureaucrat of the population. Actually, people often interact with each other in both ways
though they assign different weights to these interactions. To reflect this fact, we assume,
1 , 0 1 6 10
where the expectation formed from the local interaction can be further expressed as,
1
, 1, … , 6 11 

is the number of bureaucrat i’s neighbours. And the expectation formed from the global
interaction, on the other hand, can be expressed as,
1
, 1, … , 6 12
1

We eventually have the following equations

131
 
 
 

1
1 6 13
1 exp 2 1

1 6 14

From above we can deduce,

1
0; 0 6 15

1
0; 0 6 16

We therefore conclude that the social interaction, including local interaction and global
interaction, does matter in a corrupt act. As can be seen in 6-15 and 6-16 the incidence of
corruption is positively related to both local interaction and global interaction.

6.3 Empirical Work

The model above generates two testable implications on the relationship between the incidence
of corruption and the social interaction, presented in equations 6-15 and 6-16. We plan to test
these implications using within-country panel data from China.
6.3.1 Data and Methodology
Among related studies, Goel and Nelson (2007) use the cross-sectional within-country data of
America, while Attila (2008) and Dong, Dulleck and Torgler (2008) employ cross-country data
sets. We prefer within-country panel data in such a context. Studies on corruption could have
attributed the different levels of corruption to the cultural and institutional difference across
regions rather than the social interactions. In addition, one can stress that social interactions are
triggered by the institutional condition within a country. Thus, it is difficult to estimate the
importance of social interaction in explaining the different corruption levels across countries. If
we use cross-country data set, cultural and institutional variations across countries are hard to
be proxied and fully controlled. Using within-country data, especially those of a country
homogenous in culture and institutions like China, however, can mitigate this kind of problem.
Moreover, we can further control for regional heterogeneity when using within-country panel
data since it allows us to control for the state- and time- invariant variables in the econometric
analysis (Hisao, 2003).
Goel and Nelson (1998), Fisman and Gatti (2002) and Glaeser and Saks (2006) use the
corruption convictions of states to measure state-level corruption in America. We use a similar
measure, namely the registered cases on corruption in the procurator’s offices of provinces, to

132
 
 
 

proxy the provincial corruption level in China. Using conviction data 81 has the strength of
dealing with a less subjective measure of corruption offering also the opportunity to work with
longer time spans. In addition, they are not subjected to the problems of sampling error and
survey non-response (Glaeser and Saks, 2006). On the other hand, there is the disadvantage that
the conviction rate is driven by the quality of the detection process. This weakness, however,
will not trouble us in our current study since the quality of local judicial systems in China is
basically homogeneous. In addition, we will control local anti-corruption efforts in our
regressions.
Following the definition of global interaction, we use the average of the corruption levels in
the neighbour provinces to measure the global interaction between the bureaucrats, which
therefore can also be called the neighbouring effect. According to the definition of local
interaction, we need to find the average corruption level of closely interacting bureaucrats at the
beginning of a period when a bureaucrat makes a corrupt decision. We assume that closely
interacting bureaucrats are bureaucrats within the same province. We therefore choose the
corruption level of this province in the last period to proxy the local interaction between
bureaucrats in the province, which hence can be also referred to as the historical effect. Sah
(2005, p. 6), e.g., stresses “(…) if their past experiences have convinced some bureaucrats that
cheating is more pervasive in the economy, then they are more likely to choose to be
corrupt…Through these dynamic relationships, future levels of cheating and corruption in the
economy become explicitly linked to past levels of cheating and corruption in the economy…”.
Besides the key variables discussed above we also employ a set of control variables which
are commonly used in corruption regressions to minimize omitted variable bias. Treisman
(2000) suggested that corruption is associated with historical and cultural traditions, levels of
economic development, political institutions and government policies. Since there are no
substantial differences in history, culture and institutions between Chinese provinces, we only
focus here on the economic and policy controls. Similar to Goel and Nelson (1998), we use
provincial per capita expenditures for police, procuratorate, court and judiciary to proxy
anticorruption efforts of each province. This is insofar also important as we are focusing on
                                                            
81
  Theoretically conviction rates and the number of registered cases of corruption are different. However in China
they are actually the same. In most cases in China suspect officials are first investigated by the discipline
inspection commission of the Chinese Communist Party and its local branches. Only after they have obtained
enough evidence, the discipline inspection commissions will refer corrupt cases to the procuratorates, and
procuratorates then register the cases. Furthermore the courts and the procuratorates are both controlled by the
Chinese government. Therefore in few circumstances the courts will reject public prosecutions against corrupt
cases. 
 

133
 
 
 

registered cases on corruption (influenced by regional anticorruption efforts). According to the

two comprehensive studies on the causes of corruption, Treisman (2000) and Glaeser and Saks
(2005), more educated and richer areas have less corruption. These studies also suggest
government regulation and the relative wage of the public sector as potential determinants of
corruption. Ades and Di Tella (1999) show the tendency that an increase in economic rents, due
to the resource abundance or a decrease in competition, leads to an increase in corruption.
Furthermore, Fisman and Gatti (2002) find that, contrary to Treisman (2000), fiscal
decentralization depresses corruption in America. As mentioned in the introduction, Brunetti,
and Weder (2003) also show that the media substantially controls corruption. Finally, Swamy,
Knack, Lee and Azfar (2001) find that countries with more parliamentary seats held by women
tend to have less corruption. We therefore control for such potential determinants of corruption.
The detailed description of all the explanatory variables is presented in Table 6-2. We measure
the female representation in politics in Chinese provinces with the female representation in the
National People’s Congress, the only legislative house in China. In line with Zhang and Zou
(1998) we use the ratio of per capita provincial government expenditure to per capital central
government expenditure to proxy fiscal decentralization among provinces.
Because the definition and thus statistical calibre of the crime of corruption and bribery was
changed with a 1997 amendment to China’s criminal law, we ensure comparability by
collecting data only for 1998 to 2007. Looking at the summary of corruption levels by region,
there is a fairly wide degree of regional variation that ranges from 1.77 in Tibet to 5.01 in
Tianjin (see Table 6-4 in the Appendix).

Table 6-2 Variables description, 1998—2007

Variable Description Mean Std. Dev. Source
Cases Provincial registered cases on corruption in the procurator’s 3.14 0.96 China Procuratorial
office per 100,000 population Yearbook
Border Unweighted average of Cases in neighbouring provinces 3.04 0.64
Anticorruption Per capita expenditure for police, procuratorate, court and 112.43 103.41
judiciary
Income Logarithm of the per capita gross provincial product 9.15 0.63
Education Fraction of the population over 6 with college completed 5.44 4.31
Wage Ratio of the government employee’ wage to the average 1.13 0.13 China Statistical
wage Yearbook
Openness Ratio of export to gross provincial product 14.45 22.62
Decentralization Ratio of per capita provincial consolidated spending to per 38.20 19.52
capita central consolidated spending
Resource The fraction of employment in the mining sector 4.93 3.75
Regulation Relationship between the market and the government 6.72 2.04 Fan, Wang, and Zhu
(2010)
Media Annual newspapers circulation per capita 41.38 88.07 China Statistical
Female Female representation in the National People’s Congress 0.22 0.041 Yearbook

134
 
 
 

Our basic specification is as follows:

′
, =α Casesi,t-1 + , + , β , 6 17
where i and t denote provinces and years respectively and j is the lag value. however
indicates unobserved province fixed effects. The vector , includes all control variables
discussed above. We choose one-year lagged values for explanatory variables because there
must be intensive investigations before the corruption cases are registered in the procurator’s
offices.
We first perform pooled OLS to obtain primary results. However, to identify the causal
effect of social interaction on corruption, we need to address the endogeneity problem in our
estimation. We first include province fixed effects in our panel regressions to control for the
unobserved provincial characteristics influencing both corruption and its determinants
especially social interaction to deal with potential endogeneity biases. Mo (2001, p. 70)
describes a corruption problem as “an institutional problem that lasts for a long period”. Thus,
since the major source of potential bias in our regressions may be time-invariant historical
factors, we choose fixed-effect regressions as the most suitable tool for investigating the
relationship between corruption and social interactions.
However, fixed effect regressions do not necessarily estimate the causal effect of social
interaction on corruption. First, fixed effects regressions cannot remove endogeneity biases
generated by time-varying omitted factors affecting both corruption and its determinants
(especially social interaction). Second, the lagged independent variable Casesi,t-1 is indeed
correlated with , for s , which according to Wooldridge (2002), biases our fixed effects
OLS estimation. The standard strategy to deal with such potential biases is the instrumental
variables method. Anderson and Hsiao (1981) suggest to first-difference the equation like 6-17
to remove individual effects:
′
∆ , =α ∆Casesi,t-1 + ∆ , +∆ , β ∆ , 6 18
Casesi,t-2 is then used as the instrument for ∆Casesi,t-1 to obtain more consistent estimates since
it is uncorrelated with ∆ , as long as , are not serially correlated. However, the instrumental
variable estimator suggested by Anderson and Hsiao (1981) is not efficient because all further
lags of , can also be used as additional instruments since they are uncorrelated with ∆ , .
Arellano and Bond (1991) therefore derive a GMM estimator with all these instruments to
estimate the model more efficiently than the Anderson and Hsiao (1981) estimator.
Furthermore, based on Arellano and Bover (1995), Blundell and Bond (1998) develop a system
GMM estimator since the above lagged-level instruments in the Arellano and Bond (1991)

135
 
 
 

estimator becomes weak when the autoregressive process is too persistent in the dynamic
model. In their system estimator lagged differences are used as instruments for the level
equation such as 6-17 while the lagged levels are used as instruments for an equation such as 6-
18. We therefore estimate our model with this Arellano-Bover/Blundel-Bond system estimator.
In our case the lags of , are used as its instruments since there might be a reverse
causality between , and , . Using the same method we instrument some of other
corruption determinants, namely income, education, openness and regulation which might
potentially be endogenous.
The correlation matrix (Table 6-5) in the Appendix indicates potential multicollinearity
issues. To minimize the consequence of multicollinearity, we first adopt a parsimonious
specification including only measures of social interactions and anticorruption efforts. Then
some control variables which are not highly correlated with each other are added into the
specification. Finally we run regressions with all the discussed control variables. The process
allows us also to better check the robustness of the results.
6.3.2 Results
The findings are presented in Table 6-3. We start with OLS estimation to obtain primary results
(see specification (1), (4), and (7)). Then fixed effects regressions are performed to deal with a
potential endogeneity bias (see specification (2), (5) and (8)). Finally, the Arellano-
Bover/Blundel-Bond system estimator is used to get the results (see (3), (6), and (9)). In the
first three result columns of Table 6-3 we run regressions with a parsimonious specification
where corruption mainly depends on social interaction when anticorruption efforts are
controlled. In the second three columns we only include control variables which are not highly
correlated with other explanatory variables into our specification to minimize multicollinearity.
In the third three columns of Table 6-3, results of the full specification are presented.
Overall the results presented in Table 6-3 indicate that there is a positive and highly
statistically significant relationship between social interactions and corruption. Both, global and
local interactions matter and the findings are quite robust through all the specifications.
Furthermore, the effect of social interaction on corruption is sizable. Other things being equal,
one standard deviation increase in local interaction ( , ) raises provincial registered
cases on corruption per 100,000 people between 41% and 78% of a standard deviation, while
an increment of global interaction , by one standard deviation is associated with an
extra 11% to23% increase of a standard deviation in provincial registered cases on corruption

136
 
 
 

per 100,000 people. Thus, it looks as if social interaction is a key element in understanding
corruption.
Besides the social interaction, other explanatory variables are also observed to generally
have the expected effects on corruption in our econometric analysis 82 . The results in the
regressions indicate that anticorruption efforts and fiscal decentralization significantly decrease
corruption, while resource abundance is observed to substantially increase corruption.
According to the Arellano-Bover/Blundell-Bond estimation in Column (6) and (9), though
insignificant, deregulation, the relative wage of the public sector and female representation in
the National People’s Congress are negatively correlated with corruption.. In Column (9),
education reduces corruption, while higher income is weakly correlated with the higher
incidence of corruption (not statistically significant), which seems to contradict most previous
studies. Such a result might be driven by the transitional nature of Chinese society. Actually
countries making the transition to a market economy often experience unprecedented corruption
(Levin and Satarov 2000; Paldam and Svendsen 2000). China specifically began its transitional
process when economic reform loosened up its economy; however, political reform has lagged
behind. Therefore, in the absence of institutional and legal constraints, the government
continues to play an extensive role in China’s economic environment. One unavoidable
consequence of such an involvement is corruption, a type of corruption that becomes more
pervasive when government power is widened through increased economic activity. As a result,
regions with higher income levels may be more corrupt. Trade openness and media variables in
our regression (9) had an unexpected sign, which might be due to multicollinearity. In effect, in
Column (10) and (11) when trade openness and media are included without other highly
correlated variables, they both have expected signs and are even statistically significant.

                                                            
82
 Although some are not statistically significant.  

137
 
 
 

Table 6-3 Corruption and social interaction

Annual Cases (1998—2007)
Pooled Fixed effects Arellano- Pooled Fixed effects Arellano- Pooled Fixed effects Arellano- Arellano- Arellano-
OLS OLS Bond GMM OLS OLS Bond GMM OLS OLS Bond GMM Bond GMM Bond GMM
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11)
Casest-1 0.81*** 0.43*** 0.64*** 0.78*** 0.41*** 0.58*** 0.74*** 0.41*** 0.52*** 0.64*** 0.64***
(0.040) (0.064) (0.11) (0.043) (0.063) (0.11) (0.046) (0.064) (0.078) (0.12) (0.11)
Bordert 0.16*** 0.35*** 0.31*** 0.17*** 0.23** 0.34*** 0.12** 0.22* 0.24** 0.42*** 0.34**
(0.044) (0.11) (0.11) (0.046) (0.12) (0.092) (0.049) (0.12) (0.10) (0.15) (0.14)
Anticorruptiont-1 -0.0014*** -0.0021*** -0.0018*** -0.00089*** -0.0016* -0.00027 -0.0029*** -0.0015 -0.0017
(0.00029) (0.00071) (0.00043) (0.00033) (0.00087) (0.00077) (0.00094) (0.0017) (0.0020)
Resourcet-1 0.011 0.072** 0.075** 0.015 0.072** 0.090***
(0.0098) (0.031) (0.036) (0.0100) (0.032) (0.029)
Decentralizationt-1 -0.014** -0.025*** -0.020*** -0.018*** -0.025*** -0.020*** -0.020** -0.021***
(0.0060) (0.0035) (0.0062) (0.0041) (0.0039) (0.0070) (0.0085) (0.0074)
Regulationt-1 -0.015 -0.063* -0.043 -0.052** -0.035 -0.071
(0.012) (0.035) (0.046) (0.022) (0.040) (0.048)
Femalet-1 -0.52 0.57 -0.48 -0.79 0.43 -2.79 -0.75 -0.071
(0.95) (1.15) (2.11) (1.02) (1.21) (1.99) (2.53) (1.92)
Waget-1 0.021 0.43 -0.23 -0.077 0.37 -0.32 -0.88 -0.74
(0.20) (0.37) (0.59) (0.23) (0.38) (0.45) (0.64) (0.59)
Incomet-1 0.17 -0.065 0.073
(0.11) (0.22) (0.26)
Educationt-1 0.0100 -0.027 -0.052*
(0.018) (0.031) (0.031)
Opennesst-1 0.0062** 0.0037 0.020** -0.0046*
(0.0027) (0.0051) (0.0085) (0.0025)
Mediat-1 -0.000026 -0.0037 0.000044 -0.0024*
(0.00060) (0.0030) (0.0012) (0.0015)
Constant 0.23 0.91** 0.35 0.40 0.82 0.62 -0.36 1.62 1.33 0.81 1.05
(0.15) (0.37) (0.28) (0.30) (0.69) (1.03) (0.77) (2.06) (2.19) (0.92) (0.80)
AR(2)Test [0.21] [0.18] [0.20] [0.17] [0.16]
R-squared 0.77 0.83 0.78 0.84 0.78 0.85
Observations 279 279 279 277 277 277 276 276 276 278
Notes: Robust standard errors in parentheses; p-values in brackets; ***, **, and * denote significance at 1%, 5%, and 10%, respectively.

138 
 
 

6.4 Conclusion

In this chapter, we explore both theoretically and empirically whether social interaction
influences the incidence of corruption in China. We first present an interaction-based model on
corruption that leads to the theoretical prediction that corruption is positively associated with
social interaction. We differentiate in the chapter between local interaction (proxied as the
lagged corruption values as closely interacting bureaucrats are bureaucrats within the same
region), and global interaction (average of corruption levels in neighbour provinces). Then we
test the theoretical prediction applying an empirical analysis using province-level data in China
over the period 1998 to2007. Empirical evidence clearly indicates that social interactions, both
local and global interaction, have a significantly positive effect on the corruption rate in China.
Our findings therefore underscore the relevance of social interaction, an aspect that has long
been discussed in economics (see, e.g., Smith, 1759/1976, Veblen, 1899 and Duesenberry,
1949). Interestingly, many traditional models have treated cooperation or compliance with rules
as an isolated case. However, individuals do not normally act as isolated individuals playing a
game against nature.. The behaviour of others (individuals or regions) is important to
understand compliance. Hence, theories of pro-social behaviour, which take the impact of
behaviour or the preferences of others into account, are promising. The concept of pro-social
behaviour can be widely applied in daily life. For example, the broken windows theory suggests
that “signs of inappropriate behaviour like graffiti or broken windows lead to other
inappropriate behaviour (e.g. litter or stealing)” (Keizer et al. 2008, p.1685). The theory has
strongly influenced law enforcement strategies in several US cities such as New York, Chicago,
Baltimore, Boston and Los Angeles aiming at maintaining order by dealing more aggressively
with minor offenses (Harcourt and Ludwig 2006).
There are important policy implications based on our findings. Regional corruption is
affected by neighbourhood corruption. Successful anti-corruption activities in one area have
positive spillover effects on reducing corruption in other (contiguous) areas. To efficiently
control corruption neighbouring areas should either coordinate their individual anti-corruption
efforts with regional agreements or policy makers should take spillover effects into account
when allocating resources. This is particularly relevant when corruption is widespread. On the
other hand, a critical mass of cooperative behaviour (low level of corruption) can induce a
positive dynamic process of conditional cooperation. Moreover, previous corruption levels have
a significant effect on the current corruption level. Evolution of corruption is a path-dependent
process. Policies should take such path-dependent processes within a society into account. The

139 
 
 

closer a region is to the threshold or tipping point, the easier it is to influence the dynamic
conditional cooperative processes. However, identifying such a tipping point is not without
problems. A possibility is to change underlying institutional conditions at the local level. In
general, rigorous anti-corruption measures need to be carried out over a long period to control
corruption in areas where corruption is pandemic. As suggested by Aidt (2003), a “big push”
like the one that took place in Hong Kong in the 1970s, might be needed to address the
corruption levels in areas where previous corruption rates have been high.

140
 
 
 

Appendix
Table 6-4 Average annual registered cases on corruption per capita across regions in China (1998-2007)
Average annual registered Average annual registered Average annual registered
Region Region Region
cases per 100,000 Pop. cases per 100,000 Pop. cases per 100,000 Pop.
Tianjin 5.01 Shaanxi 3.15 Yunnan 2.61
Heilongjiang 4.77 Qinghai 3.08 Hunan 2.59
Jilin 4.50 Ningxia 3.08 Hainan 2.59
Liaoning 4.12 Hubei 3.05 Beijing 2.59
Shanxi 3.83 Guizhou 2.95 Chongqing 2.49
Hebei 3.67 Zhejiang 2.91 Anhui 2.36
Inner
Shandong 3.62 2.77 Sichuan 2.35
Mongolia
Xinjiang 3.41 Shanghai 2.77 Gansu 2.05
Fujian 3.40 Jiangsu 2.71 Guangdong 2.05
Henan 3.35 Guangxi 2.64 Tibet 1.77
Jiangxi 3.29

Table 6-5 Pairwise correlation coefficients between variables

Corruption Border Anticorruption Income Education Wage Openness Regulation Media Resource Female Decentralization
Corruption 1.00
Border 0.32 1.00
Anticorruption -0.20 0.25 1.00
Income 0.04 0.34 0.78 1.00
Education 0.06 0.44 0.79 0.75 1.00
Wage -0.05 0.01 0.23 0.38 0.14 1.00
Openness -0.08 0.42 0.76 0.75 0.81 0.27 1.00
Regulation -0.15 -0.03 0.25 0.54 0.33 0.31 0.39 1.00
Media -0.06 0.40 0.58 0.51 0.78 0.09 0.75 0.15 1.00
Resource 0.39 0.04 -0.37 -0.30 -0.22 -0.33 -0.43 -0.27 -0.26 1.00
Female -0.24 0.23 0.19 0.14 0.24 0.20 0.29 0.08 0.49 -0.15 1.00
Decentralization -0.08 0.28 0.52 0.39 0.46 0.01 0.46 -0.08 0.41 -0.18 0.17 1.00

141 
 
 

Chapter Seven Consequences of Corruption: Chinese Evidence83

7.1 Introduction

States, whether they are benevolent or repressive are in a position of having control over the
distribution of benefits and costs and therefore hold a certain level of discretionary power
which can in turn lead to corruption (Rose-Ackerman, 1999). Thus, corruption is a widely
observed phenomenon in all manner of different societies. In this chapter we are going to focus
on China. It has been stressed that corruption has become more rampant in China since
economic reform was launched in 1978. The Chinese Government has admitted that corruption
“is now worse than during any other period since New China was founded in 1949. It has
spread into the Party, into government administration and into every part of society, including
politics, economy, ideology and culture” (Guoqing Liang, 1994, p. 122). Widespread
corruption has caused severe consequences in China. According to Hu (2001), the economic
loss due to corruption in China was estimated in the late 1990s to be between 13.2 and 16.8%
of Chinese’s GDP. Although many articles have emerged about corruption in China (e.g.,
Yao’s 2002 theoretical paper and Cai, Fang and Xu’s 2009 empirical paper), there is still a lack
of systematic analysis on the consequences of corruption in China. Such a shortcoming reduces
the possibility for policy makers to assess the exact magnitude of the harmfulness of corruption
in China and therefore to derive anti-corruption strategies that are suitable for China.
In general, there has been an increasing number of economic studies on the consequences of
corruption since the 1990s, most of which focus on the effect of corruption on economic
development. The transformation of the socialist economies was one of the main reasons for
this surge in interest since institutional weaknesses and corruption surfaced as major obstacles
to market reforms and economic development (Abed and Gupta, 2002). Studies have emerged
which explored the relationship between economic growth and corruption (Mauro, 1995). Most
of these studies work with cross-sectional data using common corruption proxies such as the TI,
the ICRG and the World Bank Quality of Governance ratings (control of corruption). Such
indices reflect an indirect way of measuring corruption focusing mainly as Tanzi (2002, p. 39)
stresses on “perceptions and not objective and quantitative measures of corruption”. Naturally,
one can therefore criticize that such data is subject to many biases. Treisman (2007), for
example, pointed out that corruption perception data actually reflects impressions of the
                                                            
83
 This chapter has been submitted to the China Journal. 

142 
 
 

intensity of corruption instead of the corruption phenomenon itself. They measure beliefs that
may be driven by other social and economic conditions (see also Knack, 2006). In other words,
if the meaning of corruption is subjective, the values can vary among countries consequently
reducing the possibility of comparing the level of corruption among countries (Glaeser and
Saks, 2006). This, therefore, also reduces the effectiveness of statistical analysis. Cross-
country estimations may also be affected by omitted variable biases. Enormous unobservable or
unmeasurable differences in institutions and cultures between countries may also induce
estimation biases. Institutional and cultural frameworks that typify specific countries might
influence the size of corruption. Such features cannot always be controlled in a satisfactory
manner. In other words homogeneity reduces omitted variable biases. It may be useful to
complement such studies with within-country data. We therefore present within country
evidence focusing on China. Interestingly, not many studies have used within country data. It
certainly requires there to exist a significant level of within country variation as well as a
regional institutional structure. The U.S. with its 50 states provides, for example, an interesting
case study. Glaeser and Saks (2006) therefore explore corruption in America using information
on the amount of corruption in each of the states in the U.S. to explore state characteristics that
are associated with corruption and how corruption affects the economic development at the
state level. Once one focuses on within country data, alternative proxies of corruption can be
evaluated. The strength is the ability to focus on more concrete measures of corruption. Glaeser
and Saks (2006) use the number of government officials convicted for corruption practices
through the Federal justice department. The obvious shortcoming of such a variable is that the
proxy is driven by the quality and efficiency of the judicial system itself. If the judicial system
is inefficient or even corrupt, a large share of corrupt activities remains unobserved. Regional
differences in the efficiency of the judicial system may also bias within country comparisons.
One way to deal with this problem is to focus, as done by Glaeser and Saks (2006), on federal
convictions where one can assume that the federal judicial system is relatively isolated from
local corruption, therefore treating similar people across regions or in their case across states.
Nevertheless, it is still unclear whether and to what extent local information and efforts have
helped to reveal corruption that was treated at the federal judicial system.
Studying China may provide obvious advantages. On the one hand, China is a centralized
country with unified legal and administrative systems, which is dominated by the Han
nationality with Confucian values in most of its regions (for the detailed evidence, see
http://english.gov.cn/about.htm). Specifically, in China the heads of local governments are
actually determined by the central government and their promotions depend mainly on whether
143
 
 
 

they can faithfully carry out the regulations and policies of the central government. Furthermore,
the Supreme People's Court supervises local people's courts and actually exerts crucial
influence on the appointment and removal of the presidents and vice presidents of local people's
courts. In addition to this the Supreme People's Procuratorate has a similar but an even stronger
impact on local people's procuratorates. All of these factors ensure the homogeneity of the
Chinese political and judicial systems across the various regions. It therefore also minimizes the
potentially omitted-variable bias in the econometric analysis. On the other hand, there are great
economic differences between the rich Eastern provinces and the poor Western provinces. The
American data, which focuses on states, reduces the possibility of generalizing the results to
global differences in corruption and on the economic development across countries. Glaeser
and Saks (2006, p. 1054), for example, state: “No state today is as poor or as corrupt as many
countries in the developing world, and so relying on variation across the states in the US limits
research to a small part of the distribution of both independent and dependent variables”.
Compared to the U.S., we observe in China a stronger variance in the economic conditions
which may help to increase the generalizability of the results. Table 7-1, for example, shows
that the GDP per capita of Shanghai, which is close to that of Hungary, is nearly nine times as
high as the GDP per capita of Guizhou province, which approximates that of Cameroon.

Table 7-1 GDP (PPP) per capita of Chinese regions in 2008 (Intl. $)
Beijing 16577 Anhui 3810 Chongqing 4741
Tianjin 14590 Fujian 7922 Sichuan 4044
Hebei 6112 Jiangxi 3887 Guizhou 2321
Shanxi 5365 Shandong 8701 Yunnan 3310
Inner Mongolia 8472 Henan 5153 Tibet 3646
Liaoning 8221 Hubei 5223 Shaanxi 4799
Jilin 6184 Hunan 4608 Gansu 3185
Heilongjiang 5714 Guangdong 9886 Qinghai 4573
Shanghai 19232 Guangxi 3936 Ningxia 4706
Jiangsu 10421 Hainan 4517 Xinjiang 5232
Zhejiang 11102

The key innovative aspect of this chapter is hence to provide within country rather than
cross-country evidence, focusing on the impact of corruption on the development in China
using both provincial and city-level data. The diversity of China allows us to explore corruption
in a within-country environment. This allows, compared to cross-country studies, a better
control for unobserved culture or institutional differences, since it holds them constant. In
addition, we are going to explore how corruption affects a large set of factors such as economic
development, income inequality, public expenditures, foreign direct investment, and pollution.
144
 
 
 

Glaeser and Saks (2006) found a weak negative association between corruption and
economic growth in the US. Fisman and Svensson (2007), using the survey data from Ugandan
firms, found that bribery is negatively correlated with firm growth. With a survey of Chinese
firms (World Bank, 2006), Cai, Fang and Xu (2009) employed the entertainment and travel cost
of Chinese firms as a proxy for corruption measure and found that corruption substantially
decreases firm performance in China. We will conduct a similar empirical approach as Glaeser
and Saks (2006) to increase the comparability of the results. The chapter is structured as
follows: Section 7.2 reviews related literature. Section 7.3 presents empirical analysis. Section
7.4 concludes.

7.2 Literature Review

Corruption has significant influences on many aspects of societies (Lambsdorff, 2005). We will
focus here on its impact on economic development, which is a major concern in China since it
is the largest developing country in the world. According to Deardorff (2006), and Myint and
Krueger (2009)84, economic development is the increase in the economic, political, and social
well-being of people in a country with sustained growth from a simple, poor country into a
modern, prosperous country. Its scope includes economic growth, income distribution, public
goods (public expenditures) and environmental quality. Here we will provide a short literature
review on the influence of corruption in these subareas of economic development.
Economic growth is a fundamental part of economic development. Economic growth
always improves the living standard of the public by increasing both private income and social
services. Poor countries may experience economic growth without development in some cases.
No country, however, can sustain economic development without growth. It is therefore
important to investigate the impact of corruption on economic growth when studying the
relationship between corruption and economic development. There is indeed an existing
theoretical debate on the effect of corruption on economic growth. Some authors emphasize
that corruption can promote economic growth (“grease the wheels”). Leff (1964) and
Huntington (1968) argue that bribes can be used as an incentive instrument to influence public
officials, inducing an improvement in the quality of civil services. Lui (1985) also shows in his
model that bribes can efficiently accelerate the bureaucratic process. However, one can criticize
that bureaucrats have an incentive to delay transactions in order to extract higher payments (see
                                                            
84
    Deardorff, A., 2006. Economic development. Deardorff’s Glossary of International Economics.
(http://www-personal.umich.edu/~alandear/glossary/e.html)
Myint, H., Krueger, A.O., 2009. Economic development. Encyclopædia Britannica.
(http://www.britannica.com/EBchecked/topic/178361/economic-development) 

145
 
 
 

Rose-Ackerman, 1997). Other researchers stress that corruption decreases economic growth
(“sand the wheels”). For example, Murphy, Shleifer and Vishny (1991) point out that most
talented people are allocated into rent-seeking activities instead of productive ones in corrupt
societies. It therefore lowers economic growth since the unproductive rent seeking activities
only bring positive returns to the rent seekers instead of to the whole society (Krueger, 1974).
From a different angle, Shleifer and Vishny (1993) stated that corrupt officials may distort
investment projects to those offering better opportunities for corruption. In other words, corrupt
bureaucracy will not award the services to the most efficient producers, but instead to the
producer who offers the largest bribes. In general, most of the empirical literature supports a
negative association between corruption and economic growth. Mauro (1995) finds empirically
that corruption lowers productive investment significantly thus also reducing economic growth.
Mo (2001) reports further that through the channels of political instability, the level of human
capital and the share of private investment, corruption significantly hinders economic growth.
Pellegrini and Gerlagh (2004) also provide evidence that corruption reduces economic growth
via its effect on investment and trade policy. Recently, a number of studies stress that the
correlation between corruption and growth is conditional on the institutional quality. Meon and
Sekkat (2005) observe that corruption depresses economic growth especially in countries with a
low quality of governance. However, Mendez and Sepulveda (2006) find a quadratic
relationship between corruption and growth in free countries instead of not-free ones, which
implies the existence of the growth maximizing level of corruption. Aidt, Dutta and Sena (2008)
using a threshold model report that corruption exerts a significant negative effect on economic
growth in regimes with good governance, while having no effect on growth in regimes with
poor governance. Moreover, Meon and Weill (2008) provide empirical evidence using a panel
of 54 countries that corruption is beneficial (or at least less harmful) in countries with weak
institutions. These papers provide indirect evidence that supports the “grease the wheels”
hypothesis insofar as corruption is only beneficial in weak institutions while being harmful
elsewhere (Aidt, 2009).
As an engine of economic growth, foreign direct investment is suggested to be negatively
correlated with corruption in previous literature. Field, Sosa and Wu (2006) employed a Nash
bargaining game to find that the corruption in a host country influences the competitiveness of
foreign firms and thus also affects their expected profits in the host market. Consequently the
decisions of foreign firms to invest in a country are affected by the corruption level of the host.
Empirically, using two-year bilateral flows between 14 source and 45 host countries, Wei
(2000a) found a significantly negative association between perceived corruption in hosts and
146
 
 
 

inbound foreign direct investment. Corruption acts like a tax on FDI reducing the attractiveness
of FDI. Smarzynska and Wei (2000) also provided firm-level evidence that corruption impedes
inbound FDI and moves the ownership structure to joint ventures. Habib and Zurawicki (2002)
utilized a sample of 89 countries to find the negative effect of corruption on FDI, while Akcay
(2001) failed to identify a statistically significant link between corruption and FDI.
Furthermore, Egger and Winner (2005) analysed the impact of corruption on FDI with a sample
of 73 countries and detected a clear positive relationship between corruption and inward FDI.
However, noting that FDI is an indicator of openness, Larrain and Tavares (2004, 2007)
presented cross-country evidence that there exists reverse causation between corruption and
FDI. Due to the regional data and regional differences in China researchers may be able to
explore the relationship between corruption and FDI within a country rather than focusing on
cross-country evidence. Cole, Elliott and Zhang (2009) indeed have found that FDI is attracted
to the provinces with greater anti-corruption efforts in their within China analysis.
Being an important indicator of economic development, income distribution also influences
economic growth (Barro, 2000). Corruption, however, is observed to significantly affect the
income distribution. As discussed, better connected people in society have increased
opportunity and incentives to bribe and belong mainly to the high-income groups within a
country (Tanzi, 1995). This could lead to a reduction in the level of social services available to
the poor (Rose-Ackerman, 1999). For example, in China under the Mandarins, as in medieval
Europe, wealthy individuals in society choose their principal career in government services
where it was possible to generate bribes and tax revenues for private benefits (Baumol, 1990).
Dabla-Norris and Wade (2002) show that in the absence of credit markets only wealthy agents
have the chance to overcome the nonconvexity in income-earning possibilities (“born into rent-
seeking”, p. 454). Li, Xu and Zou (2000) used a variant of the rent-seeking model developed by
Murphy, Shleifer and Vishny (1993) to conclude that corruption influences income inequality
in a reversed U-shape way. Furthermore, they presented empirical evidence to support this
although the quadratic terms of corruption indicators were not significant in most of their
specifications (Begovic, 2006). In general, however, we still observe a lack of empirical
evidence on how corruption affects income inequality. Based on the discussion about the
channels through which corruption influences income inequality, Gupta, Davoodi and Alonso-
Terme (2002) provided robust cross-country evidence that corruption monotonically increases
income inequality. We add to the literature an analysis that uses within country data instead of
cross-country data to better isolate the unobserved institutional and cultural factors.
 

147
 
 
 

Public expenditures actually have a twofold effect on economic development. On the one
hand, Public expenditures such as those on infrastructure, education and science stimulate
economic growth. On the other hand, public expenditures such as those on health and other
social services improve the social wellbeing of the public. Corruption can substantially
influence public expenditures. It can “adversely affect the provision of publicly provided social
services” (Gupta, Davoodi, and Tiongson, 2002, p. 245). Mauro (1998) stresses that measuring
the effects of corruption on the composition of government expenditure may help to quantify
the severity of the principal-agent problem between citizens and politicians. He argued that
corrupt politicians may increase the government expenditure that it is easier to collect bribes
from and decrease the expenditure which provides fewer bribery opportunities. For example,
corrupt officials will choose goods whose exact value is difficult to monitor to maintain secrecy.
Moreover, classical rents such as the allocation of transfer and welfare payments enjoy
substantial discretionary power. Mauro (1998) presented evidence that corruption significantly
reduces government expenditure on education due to the fact that education does not provide as
many lucrative opportunities for corrupt officials compared to other spending components.
Corruption also reduces spending on operations and maintenance and increases large
government capital spending (Tanzi and Davoodi, 1997). On the other hand, Gupta, de Mello
and Sharan (2001) using a panel of 120 countries during 1985–1998, found that corruption is
positively correlated with government spending on the military.
Now we turn to the relationship between corruption and the environment quality, an
important qualitative indicator of economic development. Lopez and Mitra (2000) studied
cooperative and non-cooperative interactions between government and private firms and
conclude that introducing corruption moves the Kuznets environmental curve up: the pollution
level corresponding to every income level is constantly above the socially optimal level due to
corruption. Considering both the direct effect of corruption on pollution and the indirect effect
of corruption which influences pollution by reducing per capita income, Welsch (2004) found
with simultaneous equations that the direct effect of corruption on pollution was positive while
the indirect effect of corruption is either positive or negative and was also numerically smaller.
Therefore corruption aggravates pollution overall, especially in developing countries. However,
with the similar empirical strategy but controlling the endogeneity problems in regressions,
Cole (2007) provides cross-country evidence that the positive direct impact of corruption on air
pollution emissions is dominated by the negative indirect impact of corruption. The total effect
of corruption on air pollution emissions is hence negative in the countries that are not the

148
 
 
 

richest. This contradicting evidence demands an investigation into the actual mechanism for the
relationship between corruption and pollution.
It is quite possible that bureaucratic corruption affects pollution mainly through
environment policy making since special interest groups often exert undue influence on policy
makers by lobbying and offering bribes. This is another symptom of corruption. Several studies
about the influence of corruption on the formation of environment policy have therefore
emerged. Pellegrini and Gerlagh (2006a, b) found empirically with cross sectional samples that
corruption has a substantially negative effect which is more significant and larger in magnitude
than that of income and democracy on the stringency of the environment policy. Considering
the interactions between corruption and political stability, trade liberalization and FDI
respectively, Fredriksson and Svensson (2003), Damania, Fredriksson and List (2003) and Cole,
Elliott and Fredriksson (2006) provided both theoretical evidence from the lobbying models
and empirical evidence from the cross-country analyses that corruption not only significantly
reduces the stringency of environmental policy but also modifies the effects of other
determinants of environment policy such as political stability, trade liberalization and FDI.
Furthermore Fredriksson, Vollebergh and Dijkgraaf (2004) adopted a similar approach to find
theoretically and empirically that corruption reduces the stringency of energy policy thus
lowering energy efficiency. Indeed, the existing literature on the linkage between corruption
and the stringency of environmental policy concludes that institutional quality influences the
way policy makers respond to environmental concerns. Since the formation of environmental
policy is likely to be ‘representative of many other forms of government decision making’
(Fredriksson and Svensson, 2003), results here might be illuminative in the research of the
relationship between corruption and other public policies.

7.3 Empirical Analysis

7.3.1 Data and Methodology

China is administratively divided into 22 provinces, 5 autonomous regions and 4 municipalities,
all of which are placed directly under the Central Government. A province or an autonomous
region is subdivided into (autonomous) prefectures and/or prefecture-level cities. In this chapter
we will use two different regional data sets to explore the causes of corruption in China. The
first one is a province-level data set which consists of all 31 provincial areas in the mainland of
China. To ensure the comparability of the data, we collected data only from 1998 to 2007 as the
definition and hence the statistical calibre of the crime of corruption and bribery was changed in
1997 due to an amendment to the Criminal Law of China. It should also be noted that we have

149
 
 
 

not included data from Hongkong, Macao and Taiwan due to the obvious differences in the
political and legal systems between these areas and other parts of China.
In this dataset, corruption data are derived from the China Procuratorial Yearbooks. We
collected the number of annual registered cases on corruption in procurator’s office by region.
Glaeser and Saks (2006, p. 1058) state: “Because the conviction data are less subjective, cover
a longer time span, and are not subject to the problems of sampling error and survey non-
response, we believe that using these data has distinct advantages over the survey-based
evidence”. We then divided these registered cases by the regional population in order to obtain
the regional registered corruption cases rate per 100,000 people. An overview of ranking
corruption levels by region is presented in Table 7-2. We can observe there is a fairly wide
degree of variation across regions ranging from 1.77 in Tibet to 5.01 in Tianjin85.

Table 7-2 Average annual registered cases on corruption per capita across regions in China (1998-2007)
Average annual registered Average annual registered Average annual registered
Region Region Region
cases per 100,000 Pop. cases per 100,000 Pop. cases per 100,000 Pop.
Tianjin 5.01 Shaanxi 3.15 Yunnan 2.61
Heilongjiang 4.77 Qinghai 3.08 Hunan 2.59
Jilin 4.50 Ningxia 3.08 Hainan 2.59
Liaoning 4.12 Hubei 3.05 Beijing 2.59
Shanxi 3.83 Guizhou 2.95 Chongqing 2.49
Hebei 3.67 Zhejiang 2.91 Anhui 2.36
Inner
Shandong 3.62 2.77 Sichuan 2.35
Mongolia
Xinjiang 3.41 Shanghai 2.77 Gansu 2.05
Fujian 3.40 Jiangsu 2.71 Guangdong 2.05
Henan 3.35 Guangxi 2.64 Tibet 1.77
Jiangxi 3.29

Glaeser and Saks (2006) have shown that the conviction rates they used are positively
correlated with the survey of state house reporters’ perception of public corruption. In our case,
we check the robustness by using an alternative proxy for corruption. The second data set we
use is a data set of 120 prefecture-level cities in China which comes mainly from the survey on
the investment climate of Chinese prefecture-level cities conducted by World Bank and the

                                                            
85
We here do not use the provincial number of officials investigated in registered cases on corruption per 100,000
population to measure the regional corruption levels in China, which might be closer to the approach of Glaeser
and Saks (2006) since this corruption measure is only available in the period from 2003 to 2007. However, the
extremely high correlation between this measure and the provincial number of registered cases on corruption per
100,000 population ensures the qualification of the provincial number of registered cases on corruption per
100,000 population as a measure of regional corruption levels in China. 

150
 
 
 

Enterprise Survey Organization of China in 2005 (World Bank, 2006). The survey sampled 100
industrial firms in each city (except in four municipalities where 200 industrial firms were
sampled) to evaluate the investment climate of 120 cities covering almost all of the Chinese
provinces. In their paper, Cai, Fang and Xu (2009) used the entertainment and travel costs
relative to the sales of firms as an indirect measure of corruption in Chinese firms since
“Chinese managers commonly use the entertainment and travel costs accounting category to
reimburse expenditures used to bribe government officials, to entertain clients and suppliers, or
to accommodate managerial excess” (Cai, Fang and Xu, 2009, p.1). Similar to them, we will, in
our second data set, utilize the average value of this measure of firms investigated in the above
survey in a city as a proxy for the corruption level of the city. We refer to World Bank (2006)
for the detailed description of this measure. Other variables in our two datasets are described in
Table 7-10 in the Appendix.
To identify the casual effect of corruption, we need to address the endogeneity problem in
our econometric analysis. Our first strategy to deal with a potential endogeneity bias is to
control for unobservable province-specific factors influencing both corruption and our
dependent variables by including province fixed effects and year fixed effects in our first data
set. Mo (2001, p.70) concluded: “Corruption is commonly considered an institutional problem
that lasts for a long period.” Fixed effect regressions are therefore suitable for the investigation
of the relationship between corruption and our explained variables since the major source of
potential bias in our regressions is mainly province-specific, historical factors which are time-
invariant.
However, fixed effects regressions do not necessarily identify the causality between
corruption and our dependent variables when there are time-variant omitted factors influencing
both corruption and our dependant variables. Our second strategy hence is to adopt the
instrumental variable approach to estimate the causal effect of corruption in both data sets we
have. Commonly use instruments for corruption in the literature are, e.g., ethnolinguistic
fragmentation (for example, Mauro, 1995), origins of the country's legal system or colonial
history (for example, Wei, 2000b), and even regional dummies (for example, Mo, 2001).
However, Andrei Shleifer and William Dickens pointed out, when commenting on Wei (2000b),
that these instruments are not sufficient even if they are exogenous because they are actually
associated with all aspects of institutional quality other than corruption (p.351, Wei, 2000b).
We therefore follow the spirit of Fisman and Svensson (2007) and Cai, Fang and Xu (2009) and
use averages of corruption levels in neighbouring regions as the instrument for local corruption
levels as we are working with within-country data. This kind of approach has been widely used
151
 
 
 

in various economic areas such as empirical industry organization (for example, Hausman,
Leonard and Zona, 1994, Nevo, 2001). As mentioned before, political and legal institutions in
Chinese regions are determined by the central government and hence homogeneously.
Although China is a large country, economic development and social conventions are similar in
neighbouring regions. For example, Confucian culture dominates in most Chinese regions. We
hence are able to follow previous studies to decompose a regional corruption level C into two
terms: one area-specific: C (the area here includes the region and its neighbours), and the other
particular to the region: u
C C u
  
where i indicates areas, and j represents regions in one area. The first term C is a function of
the common economic, institutional, cultural and social factors in the area and can be expressed
by the average corruption values in neighbouring regions. The second term u however is the
idiosyncratic component determined by the regional characteristics. As main regional attributes
such as income, education, population, urbanization and openness are controlled in our
regressions we can plausibly assume that C is uncorrelated withu . We therefore use the
average of corruption levels in neighbouring regions C as the instrument for the regional
corruption level C . However, there is still a possibility in our provincial analysis that C is
correlated with u since Dong and Torgler (2010a) provide empirical evidence that corruption
is contagious among Chinese provinces, which implies that C also affects C in our case. We
therefore use a lagged variable: average corruption value of neighbouring regions in 1998 to
instrument the regional corruption during 1998-2007 in our provincial analysis to exclude the
possibility since provincial corruption levels in the following years cannot affect corruption
levels of their neighbouring provinces in 1998.
Because our instrument is time-invariant in the investigation period, we cannot control for
province fixed effects in our IV regressions. We instead follow Dong and Torgler (2010b) to
divide Chinese into eight areas (Northeastern China, Northern Coastal Areas of Seaboard of
China, Eastern Coastal Areas of China, Southern Coastal Areas of China, Middle Reaches of
the Yellow River, Middle Reaches of the Yangtze River, Southwestern China and Northwestern
China) and then include both area and year fixed effects in our IV regressions. Since Chinese
provinces in the same area are to a large extent homogeneous, the area fixed effects therefore
can capture most provincial characteristics.

152
 
 
 

The other important identification issue that we need to address is the influence of the data
noise especially in our first provincial data set. Glaeser and Saks (2006) found “the annual
fluctuations in convictions to be too noisy to identify any interesting relationships” (p. 1062).
To mitigate the impact of data noise, we have included time fixed effects in our panel
regressions. More importantly, since the data noise, largely measurement error, is idiosyncratic
to the region (see Angrist and Krueger, 2001), using grouped averages as instruments can
provide a more consistent estimation even in the presence of measurement error.
7.3.2 Corruption, Economic Growth and Income Distribution
We first test the effect of corruption upon economic growth. Levine and Renelt (1992)
identified in their cross-country analysis three robust variables in determining growth: the
initial level of real GDP per capita related to the conditional convergence hypothesis, the
average share of investment in GDP, and education attainment as a proxy for human capital.
Demurger (2001), however, observed that trade openness and infrastructure endowment
significantly influences economic performance in Chinese provinces. Since China is now a
newly industrialized country, the difference in industrialization among the Chinese regions may
be an important reason for the difference in economic growth across them. We therefore
include all those factors as control variables in the specification when investigating the impact
of corruption upon economic growth in China. Specifically, we measure regional trade
openness using the ratio of import and export to the Gross Regional Product (GRP henceforth)
and regional education attainment as the regional share of the population how completed
college. The infrastructure endowment of Chinese regions is proxied by the average regional
road mileage per 10000 people.
We perform the provincial analysis at first. We start with a commonly used specification
that does not consider corruption to test the appropriateness of our baseline specification for the
explanation of economic growth. Then, we focus on the relationship between corruption and
economic growth. We first run pooled-OLS regressions to obtain the primary results, followed
by fixed effects regressions with province and year dummies to deal with the potential
endogeneity problem. We also adopt the IV approach is to deal with a potential endogeneity
bias using as discussed the average level of corruption in neighbouring provinces in 1998 as an
instrument for provincial corruption.
Results in Table 7-3 are in line with the consensus in growth regressions (Levine and Renelt,
1992), which justifies our specification. The negative parameter on initial income level
indicates income convergence among provinces in China. Investment and industrialization

153
 
 
 

strongly encourage economic development, while education and openness insignificantly

promote economic growth. The provincial infrastructure endowment seems negatively
correlated with economic growth because high endowment of infrastructure might lead to low
infrastructure investment (a part of total investment) which is positively associated with
economic growth. More importantly, provincial corruption is observed to have a negative effect
on economic growth in China. Previous literature such as Mo (2001) and Glaeser and Saks
(2006) also provided evidence that corruption retards economic growth.

Table 7-3 Effect of corruption on economic growth: cross-province evidence

Annual Growth Rate of GRP per capita (1998—2007)
Pooled Fixed effects Pooled Fixed effects Fixed effects
OLS OLS OLS OLS 2SLS
(1) (2) (3) (4) (5)
Corruption -0.00021 0.0044 -0.024***
(0.0021) (0.0029) (0.0090)
Income -0.0022 -0.18*** -0.0022 -0.18*** -0.018**
(0.0059) (0.029) (0.0059) (0.029) (0.0088)
Education 0.0083* 0.0058 0.0084* 0.0059 0.0060
(0.0043) (0.0077) (0.0044) (0.0077) (0.0050)
Investment 0.13*** 0.13*** 0.13*** 0.13*** 0.13***
(0.014) (0.023) (0.015) (0.023) (0.023)
Openness -0.0027 0.011 -0.0029 0.014 0.0040
(0.0077) (0.016) (0.0081) (0.016) (0.013)
Industrialisation 0.080*** 0.29*** 0.081*** 0.29*** 0.13***
(0.022) (0.080) (0.023) (0.081) (0.050)
Infrastructure -0.031 -0.056** -0.031 -0.058** -0.035
(0.020) (0.024) (0.020) (0.024) (0.026)
Constant 0.042 1.45*** 0.043 1.46*** 0.19**
(0.044) (0.24) (0.044) (0.24) (0.077)
R-squared 0.28 0.67 0.28 0.68
First stage
F test of excluded IV 23.98[0.00]
Anderson canon. Corr. LM statistic 22.07[0.00]
Observations 310 310 310 310 310
Notes: Robust standard errors in parentheses, p-values in brackets, ***, **and * denote significance at
1%, 5% and 10% respectively. The education attainment is henceforth expressed in logarithm. All
explanatory variables are hereafter lagged by one year in our panel estimations except that income
variable in Columns (5) & (6) of the table is lagged by two years.

Next we check the result with the city-level dataset. We use as an alternative corruption
measure, namely the average entertainment and travel costs relative to the sales of sample firms
in Chinese cities (ETC hereafter) to re-examine the relationship between corruption and
economic growth in China. Cai, Fang and Xu (2009) found that “ETC is a mix that includes
‘grease money’ to get better government services, ‘protection money’ to lower tax rates” (p.2),
as well as other expenditures. They observed that some components of ETC promote firm
performance though the overall effect of ETC is negative. They, however, recognize that their
firm-level findings do not necessarily mean that ETC expenditures are socially “grease the

154
 
 
 

wheels” or “sands the wheels”. We here attempt to locate the social influence of ETC as an
indirect corruption measure by exploring its effect on regional economic growth.
We directly investigate the effect of corruption on economic growth in Chinese cities with a
specification similar to the above. Since the corruption measure here comes from industrial
firms, we therefore focus on its effect on the growth rate of the industrial output per capita in
cities, which is highly correlated with the growth rate of GRP per capita. Following Fisman and
Svensson (2007), we add the average tax burden of firms in cities, measured by the average
taxes and fees relative to firm sales, as a control in our previous specification. Due to the lack
of education data, we however employ local public library collections per 100 people in 2003 to
proxy for general education levels in cities. Similar to the provincial analysis, we start with the
OLS regression and run the 2SLS regression afterwards. The instrumental variable we use in
the city-level analysis is the average corruption levels of other cities within the same province
(neighbouring cities)86.
Results in first three columns of Table 7-4 generally support those in Table 7-3. The
coefficient of initial industrial output per capita is significantly negative. Investment and
industrialization substantially promote economic growth, while education and trade openness
insignificantly encourage economic development. Consistent with Fisman and Svensson (2007),
the average tax burden here, though not robust, appears to lowers economic growth, while the
effect of regional infrastructure stock on economic growth seems ambiguous. More importantly,
similar to our finding in Table 7-3, corruption measured by ETC also has a negative effect on
economic growth though it is statistically insignificant. It is worth noting that our instrument
seems weak in the IV regression in Column (3). However, Imbens (2008) point out that even
with very weak instruments, 2SLS in the just-identified case is not substantively misleading
though its confidence intervals tend to be wide: “In this case better estimators are generally not
available”, while “improved methods for confidence intervals based on inverting test statistics
have been developed” (p.1). We therefore use a conditional likelihood ratio (CLR) test
proposed by Moreira (2003) which is robust to weak instruments to construct confidence
intervals. According to CLR test, the p-value of the corruption variable in Regression (3) is
0.22, which support our original finding.
Till now we only have weak evidence about the effect of corruption on economic growth.
We are therefore looking for further evidence in the next columns of Table 7-4. We utilize the
shares of firms which believe in the need for informal payment to obtain bank loans in cities
                                                            
86
 For the four municipalities in our city-level data set, we use averages of corruption levels of their neighbour cities to
instrument their corruption levels respectively. 

155
 
 
 

(Loan pay) as another (indirect) corruption measure though it is constrained to the credit area.
However, most banks in China are controlled by the government. Bank managers hence act as
if they were government officials. Interestingly, this corruption measure is observed to have a
significantly positive effect on economic growth (“greases the wheels”). As discussed
previously, the informal payment to obtain bank loans is one of the “grease money”
components of the ETC measure. Its positive effect is reasonable because the larger share of
firms which believe in the need for informal payment to obtain bank loans in Chinese cities
reflects a larger opportunity for firms there to circumvent the appraisals of government banks
with bribery to obtain loans. Compared to those in cities where (government) banks are
incorrupt but bureaucratic, the firms in cities with corrupt banks make more informal payment
to bank officials and hence have easier access to bank loans though at some cost. They
consequently grow faster.
Furthermore, we also found that government red tape proxied by the average days per year
that enterprise staff must spend interacting with government bureaucracies in cities (Red tape)
significantly impedes economic growth. It is natural that firms in cities with more bureaucratic
red tape have to spend more time and money to go through or circumvent the red tape. The
money that firms have spent on red tape is of course one of the “protection money” parts of
ETC. This kind of “protection money” has a socially negative effect on economic performance
(“sands the wheels”) by a wasting productive resource although an individual firm might
benefit from it. It hence can also be labelled as the “sand money”.
Though ETC includes both “grease money” and “sand money” (“protection money”) as
discussed before, it is difficult to thoroughly isolate the two components of ETC since we do
not know the detailed composition of ETC in each firm. We nonetheless utilize a two-stage
estimation procedure to explore ETC greases and sands the wheels simultaneously in Chinese
cities. As discussed above the Loan pay variable and the Red tape variable are highly correlated
with the “grease” component and the “sand” component of ETC respectively. We hence first
run two regressions where ETC is a dependent variable and the Loan pay (grease aspect) and
the Red tape (sand component) are the independent factors respectively (see Column (4) and (5)
in Table 7-4). Then we put the fitted values of ETC in the two regressions as proxies for the
“grease” and “sand” component of ETC respectively into the previous growth regressions to
test their effects on economic growth. Results in Column (8) to (9) confirm previous
conjunctures. The proxy for the “grease” component of the ETC does have a significant
positive effect on the economic growth of Chinese cities, while the proxy for the “sand” part of
ETC indeed has a substantially negative impact on economic growth. Including the two
156
 
 
 

components simultaneously into Regression (10), we find a positive and negative effect on
economic growth respectively, and the positive effect seems slightly stronger when looking at
the standardized (beta) coefficients. Such a result is consistent with our previous finding that
the corruption measured by ETC has an overall (insignificantly) negative effect on growth.
It is worth noting that the corruption measure in our first province-level data set (regional
registered cases on corruption in procurator’s office) may also be a comprehensive corruption
measure indeed since officials involved in corruption cases probably accept “grease money”
and/or “sand money”. The insignificant effect of this measure on growth supports our inference.
We, however, cannot decompose this measure as what we have done to the ETC due to the lack
of detailed information.
Above results indicate that corruption greases and sands the wheels simultaneously. The
overall effect of corruption on growth depends on which effect dominates. It appears that the
relationship between growth and corruption is not robust or even not significant in some
scenarios in China. This has also been observed by Mauro (1995) and Mo (2001), among others
in cross-country analysis. Recently, Mendez and Sepulveda (2006) and Adit, Dutta and Sena
(2008) documented in their empirical studies that the effect of corruption on economic growth
depends on the quality of institutions. We therefore may be able to plausible conclude that
corruption can have both, positive and negative effects, and the overall effect may depend on
the balance between components, which is probably determined by the institutional quality. In
sum, previous studies fail to explore single elements of corruption. An analysis as the one done
here through decomposing potential elements can provide new insights in the literature.

157
 
 
 
Table 7-4 Effect of corruption on economic growth: cross-city evidence
Average growth rate of industrial output per capita (2004-2007) ETC Average growth rate of industrial output per capita (2004-2007)
OLS OLS 2SLS OLS OLS OLS OLS OLS OLS OLS
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
ETC -0.019 -0.30
(0.040) (0.31)
Loan pay 0.0051* 0.12***
(0.0031) (0.0085)
Red tape -0.0011* 0.017***
(0.00065) (0.00087)
Grease component 0.043* 0.052** {0.160}
(0.026) (0.026)
Sand component -0.067* -0.082** {-0.161}
(0.039) (0.041)
industrial output per capita -0.13*** -0.13*** -0.10*** -0.13*** -0.12*** -0.13*** -0.12*** -0.12***
(0.031) (0.031) (0.038) (0.033) (0.030) (0.033) (0.030) (0.031)
Tax -0.040*** -0.037** 0.012 -0.041*** -0.035** -0.041*** -0.035** -0.035**
(0.014) (0.016) (0.057) (0.014) (0.015) (0.014) (0.015) (0.015)
Education 0.00031 0.00030 0.00013 0.00033 0.00034 0.00033 0.00034 0.00037
(0.00023) (0.00023) (0.00030) (0.00022) (0.00025) (0.00022) (0.00025) (0.00025)
Investment 0.37*** 0.38*** 0.47*** 0.41*** 0.39*** 0.41*** 0.39*** 0.44***
(0.12) (0.12) (0.17) (0.12) (0.12) (0.12) (0.12) (0.13)
Openness 0.015 0.015 0.0060 0.019 0.013 0.019 0.013 0.017
(0.032) (0.031) (0.033) (0.031) (0.031) (0.031) (0.031) (0.031)
Industrialisation 0.64*** 0.60** 0.017 0.67*** 0.60*** 0.67*** 0.60*** 0.62***
(0.22) (0.24) (0.62) (0.21) (0.23) (0.21) (0.23) (0.22)
Infrastructure 0.0046 0.0035 -0.013 0.0098 -0.00041 0.0098 -0.00041 0.0048
(0.039) (0.039) (0.041) (0.039) (0.039) (0.039) (0.039) (0.038)
Constant 1.65*** 1.66*** 1.78*** 1.55*** 1.64*** 1.55*** 1.64*** 1.52***
(0.25) (0.25) (0.29) (0.27) (0.25) (0.27) (0.25) (0.27)
R-squared 0.32 0.32 0.33 0.33 0.71 0.79 0.33 0.33 0.35
First Stage
F test of excluded IV 3.17[0.078]
Anderson canon. corr. LM statistic 3.33[0.068]
Observations 118 118 118 118 118 120 120 118 118 118
Notes: Robust standard errors in parentheses, p-values in brackets, beta coefficients in braces. ***, **and * denote significance at 1%, 5% and 10% respectively. ETC, Loan pay, Red tape, and Tax are
derived from World Bank (2006). Investment, Openness, and Industrialization are measured by their averages in the period 2004-2007 respectively. Education and Infrastructure are henceforth
represented by their values in 2003 and 2004 respectively. And we use the 2004 value of Industrial output per capita in regressions.

158 
 
 

An important issue that closely relates to economic growth is income distribution. This is
also a key aspect of economic development. Since we only have the Gini coefficients of
Chinese provinces before 2001, we have to find an alternative measure of income inequality in
our analysis. Kanbur and Zhang (1999) and Sicular, Yue, Gustafsson and Li (2007) concluded
that the urban-rural income gap is a main source of the overall inequality in China. We
therefore use the ratio of the per capita income of urban households to that of rural households
to proxy for income inequality in Chinese regions.
We adopt a specification similar to Li, Xu and Zou (2000) to examine the relationship
between corruption and income inequality. To address the reverse causality from income
inequality to corruption we adopt the IV approach in both province-level and city-level
analyses using the same instrument for corruption as in the previous regressions. It should be
noted that the results in Table 7-5 are consistent with Li, Xu and Zou (2000) and Gupta,
Davoodi and Alonso-Terme (2002) in that corruption substantially increases income inequality.
Furthermore, income, though weakly, has a negative impact on income equality, which is also
in line with previous findings such as Lundberg and Squire (2003).

Table 7-5 Effects of corruption on income inequality

Urban-rural income ratio
Province level City level
Pooled Fixed effects Fixed effects OLS 2SLS
OLS OLS 2SLS
(1) (2) (3) (4) (5)
Corruption -0.19*** 0.0019 0.22** 0.97* 5.82*
(0.025) (0.022) (0.096) (0.58) (3.31)
Income -0.55*** -0.59** -0.090 -0.55 -0.54
(0.090) (0.24) (0.11) (0.42) (0.70)
Education 0.10 -0.16 -0.25*** 0.0062* 0.0062
(0.10) (0.11) (0.095) (0.0033) (0.0052)
Expenditure 4.09*** -1.72 1.75*** -0.0022 -5.26
(0.41) (1.20) (0.56) (7.54) (13.38)
Industrialisation 1.55*** 3.94*** -0.36 8.16 15.57
(0.34) (0.89) (0.49) (7.88) (10.21)
Urbanization 0.19 -0.0035 0.72*** 0.19 -1.01
(0.18) (0.097) (0.18) (1.08) (1.98)
Constant 7.19*** 6.79*** 4.00*** 2.32 -5.87
(0.67) (2.00) (0.93) (3.58) (8.29)
R-squared 0.63 0.94 0.10
First stage
F test of excluded IV 41.07[0.00] 9.88[0.00]
Anderson canon. corr. LM statistic 32.71[0.00] 10.05[0.00]
Observations 284 284 284 113 113
Notes: Robust standard errors in parentheses, p-values in brackets. ***, **and * denote significance at
1%, 5% and 10% respectively. In city-level analysis, Urbanization and Expenditure are hereafter
represented by their values in 2003, and Income is represented by its value in 2004 henceforth.

159 
 
  
 

7.3.3 Corruption and Foreign Direct Investment

Mauro (1995), Mo (2001), and Pellegrini and Gerlagh (2004) have documented the fact that
investment is the main channel through which corruption influences economic growth:
“Corruption is found to lower investment, thereby lowering economic growth” (Mauro, 1995, p.
681). We will focus here on the effect of corruption upon a special kind of investment, namely
foreign direct investment. Does corruption deter inward foreign direct investment in China as
some prior literature suggested? According to Wei (2000a, b), China seems to be a puzzle in
terms of its relationship between corruption and FDI inflows. China has been the largest
developing host of FDI for 16 consecutive years while simultaneously being reported in
international surveys as being severely corrupt. Wei (2000b) performs a cross-country analysis
including a Chinese dummy and concludes that “corruption is just as damaging to FDI into
China as it is elsewhere” (p. 321) basing this statement on the fact that coefficients on the
corruption variable and the Chinese dummy in regressions are significantly negative. This
insightful finding, however, is not fully convincing. The negative coefficient on corruption in
their cross-country analysis does not necessarily mean that corruption significantly deters FDI
inflows in China since China-related data only cover 2% of the sample used. Moreover the
negative coefficient on the China dummy might be due to some unobservable factors rather
than on corruption since country corruption levels have been controlled simultaneously in the
regressions. To try to generate a solid finding of the linkage between corruption and FDI
inflows in China, we perform a within-country analysis here, controlling for the endogeneity
problem with both the fixed-effects and instrumental variable approaches.
Similar to Harms and Ursprung (2002), we use the regional average annual FDI inflow per
capita as the dependent variable. Besides corruption, we, following (Wei, 2000a, b) and Egger
and Winner (2005), introduce several common controls such as income level and education
attainment in our regressions. We first perform the provincial analysis. We start with pooled-
OLS regressions. To address the omitted variable bias and the reverse causation between
corruption and FDI that Larraín and Tavares (2003, 2007) pointed out, we then adopt the fixed
effects and the IV approaches. To check the robustness of our provincial results, we also
perform the city-level analysis. Similarly, we run the OLS regression and then the IV regression
in the city-level analysis.

160 
 
  
 
Table 7-6 Effect of corruption rate on inbound FDI
FDI per capita
Province level City level
Pooled Fixed effects Fixed effects OLS 2SLS
OLS OLS 2SLS
(1) (2) (3) (4) (5)
Corruption -0.66*** -0.84** -2.45** -8.74** -12.55
(0.23) (0.37) (0.96) (3.45) (8.08)
Income 7.11*** 7.97*** 3.12*** 14.1*** 14.01***
(0.93) (3.03) (0.91) (3.91) (3.71)
Education -0.66 1.05* 1.67*** 0.032 0.033
(0.45) (0.55) (0.49) (0.050) (0.047)
Population -1.03*** 8.56* -1.74** -7.95* -8.40*
(0.37) (4.63) (0.72) (4.43) (4.32)
Industrialisation -0.0019 -0.19*** 0.21*** -16.54 -24.76
(0.029) (0.064) (0.068) (17.45) (25.24)
Expenditure -0.052 -0.075 0.042 0.80* 0.80*
(0.038) (0.073) (0.047) (0.46) (0.45)
Wage -3.55*** 9.00*** 2.78 10.3* 10.4*
(1.13) (2.98) (2.07) (5.96) (5.62)
Infrastructure 0.26*** 0.14*** 0.23*** -9.42** -9.32**
(0.029) (0.031) (0.033) (4.41) (4.17)
Constant -17.4*** -207*** -50.5*** -113* -101
(5.71) (50.2) (17.7) (63.3) (69.6)
R-squared 0.82 0.93 0.60
First stage
F test of excluded IV 15.37[0.00] 12.34[0.00]
Anderson canon. corr. LM statistic 13.43[0.00] 11.83[0.00]
Observations 303 303 303 116 116
Notes: Robust standard errors in parentheses, p-values in brackets. ***, **and * denote significance
at 1%, 5% and 10% respectively. The Wage variable is expressed in logarithm. In our city-level
analysis, Population are henceforth measured by their values in 2004.

Table 7-6 shows results consistent with previous findings. In general corruption hinders
inward foreign direct investment significantly in China87. Hence the nexus of corruption and
FDI in China, which has seemed to be perplexing in our cross-country comparison, does indeed
not contradict prior theoretical and empirical findings. In this sense, China is a normal country.
Additionally, we observe that FDI is substantially attracted to the provinces with higher income
level and education attainment, which is also in line with previous literature. Average wage has
a weak positive effect on FDI inflows in Chinese regions, which may reflect the fact that
foreign investors in China prefer developed regions with better investment climate to
underdeveloped regions with lower wage levels.
We notice that Cole, Elliott and Zhang (2009) found in Chinese provinces a positive
relationship between FDI and annual registered corruption cases per 100,000 persons which
they interpret as a measure for provincial anti-corruption efforts. Their identification, however,
                                                            
87
 In regression (5) in Table 7-6 corruption is significant at 12% level. 

161 
 
  
 

seems problematic since they did not effectively address both the endogeneity bias and the
potential bias introduced by the presence of the lagged dependent variable. Furthermore, it is
not appropriate for them to interpret the provincial registered corruption cases per 100,000
persons as regional anticorruption efforts. There are actually not substantial differences among
provincial anticorruption efforts in China because principal anti-corruption laws and policies in
China are actually made by the Central Government and the provincial leaders including the
local anticorruption leaders are also all appointed by the Central. A local anti-corruption leader
will be replaced if he is obviously incompetent to control local corruption (Wu, 2008).
Following the spirit of Fisman and Gatti (2002), and Glaeser and Saks (2006), we stress that
this kind of measure mainly indicates the level of local corruption rather than provincial
anticorruption effort, or it cannot vary as much as they showed across provinces.
7.3.4 Corruption and Public Expenditures
Bureaucratic corruption also affects the supply of public goods. Following Tanzi and Davoodi
(1997) and Mauro (1998), we examine the effects of corruption on public expenditures in
Chinese regions. We explore whether regional corruption influences the composition of
government expenditure with a specification richer than those in prior studies. Similar to Mauro
(1998), we use ratios of public expenditures to GDP as dependent variables in our regressions.
Results are reported in Table 7-7. Consistent with those in Tanzi and Davoodi (1997) and
Mauro (1998), corruption significantly decreases government expenditures on education and
science, while substantially increases public expenditure on social security. According to
Mauro (1998)’s theoretical analysis, our findings are reasonable. For example, “it seems easier
to hand out a disability pension to a healthy person than to give a teaching job to an unqualified
person.” (Mauro, 1998, p. 264)
We then retest provincial findings with the city-level data. We add a dummy to indicate
municipalities in city level regressions since province-level cities have more autonomy in
public expenditures than others cities in our sample which are prefecture-level. We conduct
first OLS and then IV regressions to investigate the effect of ETC upon government
expenditures. From Table 7-7 we can find that our city-level results generally support the cross-
province findings.

162 
 
  
 
Table 7-7 Effects of corruption on public expenditures
Expenditure on social security/GRP Expenditure on science /GRP Expenditure on education /GRP
Province level (1998-2007) City level (2005) Province level (1998-2007) City level (2005) Province level (1998-2007) City level (2005)
Pooled FE FE OLS 2SLS Pooled FE FE OLS 2SLS Pooled FE FE OLS 2SLS
OLS OLS 2SLS OLS OLS 2SLS OLS OLS 2SLS
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15)
Corruption 0.077 -0.19*** 0.32*** 0.32*** 1.20*** -0.0012 0.0086 -0.10*** -0.0046 -0.023* -0.17*** -0.020 -0.35*** -0.017 -0.47
(0.047) (0.055) (0.12) (0.11) (0.46) (0.0054) (0.0073) (0.035) (0.0038) (0.014) (0.034) (0.031) (0.11) (0.11) (0.41)
Income -0.75*** -1.14** -0.50*** -0.55*** -0.34** 0.0068 -0.10* 0.034 -0.0012 -0.0024 -0.39*** -1.24*** -0.47** -0.70*** -0.80***
(0.11) (0.44) (0.13) (0.10) (0.16) (0.015) (0.055) (0.030) (0.0034) (0.0042) (0.15) (0.36) (0.22) (0.098) (0.13)
Education 0.37*** 0.036 0.012 -0.00029 -0.0012 -0.015 -0.011 -0.014 0.00019*** 0.00018*** -0.32** 0.21 -0.58***
(0.10) (0.11) (0.089) (0.0011) (0.0014) (0.012) (0.014) (0.014) (0.000067) (0.000069) (0.14) (0.13) (0.11)
Population -0.47*** -2.09** -0.36*** -0.11 -0.094 -0.071*** 0.20 -0.10*** -0.0023 -0.0037 -1.00*** 0.39 -0.74*** -0.19* -0.21**
(0.060) (0.86) (0.053) (0.12) (0.14) (0.0067) (0.13) (0.023) (0.0037) (0.0045) (0.14) (0.84) (0.13) (0.099) (0.099)
Urbanization -0.40 -0.44 0.65*** 0.84 0.42 -0.028 0.028 0.11*** 0.016 0.020 0.0068 0.57*** 0.60*** -0.031 0.23
(0.30) (0.38) (0.23) (0.54) (0.51) (0.030) (0.041) (0.039) (0.019) (0.020) (0.22) (0.21) (0.21) (0.26) (0.39)
Industrialisation 0.28*** -0.075 0.94*** -0.020 -0.053*
(0.082) (0.15) (0.25) (0.022) (0.032)
Researchers 0.0045*** -0.0050** -0.00049 0.0017 0.0057
(0.00093) (0.0025) (0.0021) (0.017) (0.015)
Students 0.0015*** 0.00026 0.00023 -0.011 -0.034
(0.00046) (0.00060) (0.00044) (0.036) (0.047)
Municipality 0.77** 0.89** 0.068** 0.066** 1.05*** 1.03***
(0.30) (0.39) (0.027) (0.027) (0.27) (0.31)
Constant 11.4*** 29.9*** 8.86*** 6.06*** 3.20 0.62*** -0.47 0.60 0.057 0.11* 14.4*** 9.29 14.8*** 9.51*** 11.1***
(1.22) (7.48) (1.54) (1.49) (2.41) (0.13) (1.04) (0.37) (0.042) (0.069) (1.61) (7.21) (2.41) (1.53) (2.05)
R-squared 0.35 0.87 0.29 0.37 0.86 0.62 0.95 0.47
First stage
F test of excluded IV 37.58[0.00] 12.63[0.00] 8.37[0.00] 8.60[0.00] 36.26[0.00] 10.12[0.00]
Anderson canon. corr. LM statistic 31.74[0.00] 12.00[0.00] 9.90[0.00] 9.24[0.00] 33.61[0.00] 10.18[0.00]
Observations 228 228 228 120 120 228 228 228 228 228 228 120 120
Notes: Robust standard errors in parentheses, p-values in brackets. ***, **and * denote significance at 1%, 5% and 10% respectively. In city-level analysis, we use 2005 values of
Population, Students, Researchers and Industrialization.

163 
 
 

7.3.5 Corruption and the Environment

In this section we will turn to the impact of corruption on the environment. We first
investigate the effect of corruption upon pollution. Welsch (2004) and Cole (2007) apply
simultaneous equations to estimate the effect of corruption on the pollution as they assume
that besides its direct effect on pollution, corruption also has an indirect impact on pollution
through lowering income level. However, according to Glaeser and Saks (2006), Gundlach
and Paldam (2009) and Dong and Torgler (2010b), the causality between income and
corruption is mainly from income to corruption. We therefore are able to estimate the effect
of corruption on pollution with a single equation. We use SO2 emission per capita and soot
emission per capita as alternative measures for pollution emissions. Due to the existence of
the environmental Kuznets curve (Dasgupta, Wang and Wheeler, 2002), we include income
per capita, its quadratic term and even its cubic term in our specification. Following
Fredriksson, Vollebergh and Dijkgraaf (2004), we also investigate the influence of corruption
on energy efficiency, an important determinant of the environment in China. We employ the
energy intensity index (energy consumption per unit of gross regional product) to measure
regional energy efficiency. We set up a specification similar to the previous study and use the
same estimation procedure and instrumental variable as above.
Provincial results in Table 7-8 show that corruption exerts a positive effect on per capita
emissions. It is in line with the findings of Welsch (2004) and Cole (2007) on the direct effect
of corruption on pollution. Furthermore, similar to Fredriksson, Vollebergh and Dijkgraaf
(2004), we also observe that corruption significantly enhances energy intensity and thus
reduces energy efficiency in China.
We then check the provincial findings with the city-level data. Here we employ SO2
emission by industry per capita and soot emission by industry per capita as alternative
measures for pollution emissions as corruption measure here (ETC) is drawn from the survey
of industrial firms in Chinese cities. We therefore control for industrialization of Chinese
cities in the city-level regressions. We run OLS regressions and 2SLS regressions with the
same instrument as that in the previous city-level analysis. Our city-level findings clearly
support the positive influence of corruption on pollution emissions in China.

164 
 
  

Table 7-8 Effect of corruption and the environment

SO2 emission per capita Soot emission per capita Energy efficiency
Province level City level (2005) Province level City level (2005) Province level
Pooled FE FE OLS 2SLS Pooled FE FE OLS 2SLS Pooled FE FE
OLS OLS 2SLS OLS OLS 2SLS OLS OLS 2SLS
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13)
Corruption 0.0093 0.0040 0.072*** 0.0017 0.054 0.022*** -0.0032 0.043*** 0.019 0.090* 0.0022** 0.00018 0.0051**
(0.0062) (0.0047) (0.022) (0.034) (0.096) (0.0037) (0.0022) (0.010) (0.016) (0.047) (0.00092) (0.00019) (0.0023)
Income -17.49*** -10.2*** -13.56*** 22.88** 15.79* -4.01** -3.29*** -13.51*** 2.92 6.43 -0.0067*** -0.0038** -0.0051*
(4.77) (1.76) (4.43) (11.10) (8.99) (2.04) (0.92) (3.08) (3.22) (4.07) (0.0016) (0.0015) (0.0029)
Income2 1.87*** 1.16*** 1.50*** -2.59** -1.79* 0.43* 0.36*** 1.45*** -0.32 -0.72
(0.52) (0.19) (0.47) (1.26) (1.02) (0.22) (0.10) (0.33) (0.36) (0.46)
Income3 -0.066*** -0.043*** -0.055*** 0.097** 0.067* -0.016* -0.013*** -0.052*** 0.011 0.027
(0.019) (0.0070) (0.016) (0.048) (0.039) (0.0080) (0.0037) (0.011) (0.014) (0.017)
Education 0.0053 0.0086*** 0.010*** -0.00013 -0.00021 0.0019 0.00059 0.0040** -0.00010 -0.00020 -0.00017 0.000084 0.000085
(0.0046) (0.0024) (0.0039) (0.00031) (0.00033) (0.0020) (0.0012) (0.0016) (0.00014) (0.00015) (0.00018) (0.000051) (0.00024)
Population -0.0086 0.21** -0.0019 -0.12*** -0.11*** -0.0066 0.016 -0.0047 -0.036** -0.031** -0.0048*** 0.0015 -0.00082
(0.0100) (0.090) (0.012) (0.039) (0.040) (0.0043) (0.036) (0.0058) (0.014) (0.015) (0.0012) (0.0017) (0.0012)
Urbanization 0.035 0.027 -0.053 0.059 0.084 0.046 0.023 -0.036 0.058 0.033 0.0021 0.0015** -0.0019
(0.048) (0.031) (0.055) (0.14) (0.12) (0.031) (0.022) (0.031) (0.065) (0.058) (0.0035) (0.00065) (0.0044)
Openness -0.050 0.0025 0.093** -0.0013 -0.016 -0.043*** -0.012 0.0082 -0.028 -0.033*
(0.032) (0.028) (0.037) (0.039) (0.039) (0.015) (0.012) (0.017) (0.018) (0.019)
Constant 54.58*** 27.47*** 40.75*** -66.20** -45.53* 12.40** 10.00*** 41.68*** -8.65 -18.97 0.11*** 0.038** 0.069***
(14.62) (5.15) (13.98) (32.35) (26.24) (6.23) (2.70) (9.65) (9.53) (12.10) (0.016) (0.018) (0.026)
R-squared 0.084 0.94 0.38 0.20 0.95 0.18 0.49 0.99
First stage
F test of excluded IV 74.32[0.00] 13.23[0.00] 74.32[0.00] 13.23[0.00] 23.87[0.00]
Anderson canon. corr. 53.53[0.00] 12.89[0.00] 53.53[0.00] 12.89[0.00] 15.78[0.00]
LM statistic
Observations 284 284 284 120 120 284 284 284 120 120 108 108 108
Notes: Robust standard errors are in parentheses. The significant levels of 1%, 5% and 10% are noted by ***, **and *. Industrialization here is measured with their values in 2005 in the city-
level analysis. Openness is measured by its value in 2004.

165 
 
 

We proceeded to investigate the mechanism through which corruption affects pollution or

environmental quality. Fredriksson and Millimet (2004) suggested that political institutions
influence the formation of economic policy including environmental policy. It is therefore
plausible to conjecture that corruption affects environmental quality mainly through the
channel of environmental policy formation. We measure the stringency of environment
policy with the ratio of industrial waste water meeting discharge standards in Chinese regions.
Then we examine our conjecture using routine estimation procedures. Regional corruption is
observed to substantially loosen environmental regulations in China (see first five columns in
Table 7-9). Moreover, trade openness also appears to decrease environmental policy
stringency. In addition, Damania, Fredriksson and List (2003) argued that corruption not only
directly influences environmental regulation but also modifies the effects of other
determinants of environmental policy such as trade openness. Detailedly, Damania,
Fredriksson and List (2003) developed a lobbying model with the endogenous formation of
environmental regulation and concluded that when government corruption is high, trade
openness tends to raise local environmental policy stringency if trade policy is protective,
which was confirmed by their cross-country empirical analysis. With the specification similar
to Damania, Fredriksson and List (2003), we look here for the within-country evidence that
corruption indirectly influences the environmental policy stringency through the channel of
trade openness. We in turn perform Pooled OLS, fixed effects and fixed effects-IV
regressions. The results in the last five columns of Table 7-9 show that regional corruption
has an overall negative effect on the stringency of environmental policy (at the mean level of
regional trade openness), which is consistent with our previous findings.

Figure 7-1 Marginal effects of trade openness on environmental stringency conditional on corruption
Marginal effect of openness

0
3.53 Corruption

-1.06

Note: Marginal effect is calculated with the 2SLS results in Column (8) of Table 7-9

166 
 
 

Similar to Damania, Fredriksson and List (2003), the effect of trade openness on the
provincial environmental policy stringency are substantially modified by provincial
corruption levels. However, we do not obtain significant city-level evidence probably due to
data noise. As shown in Figure 7-1, the marginal effects of trade openness on the
environmental policy stringency, which are initially negative, increase with local corruption
level and become positive at within-sample levels of corruption. However, the mechanism
that Damania, Fredriksson and List (2003) suggested is not applicable in China since
lobbying is not allowed in China. A plausible interpretation of our results may be consistent
with Li and Zhou (2005) who observed that the probability of promotion or termination of
local leaders depends mainly on their economic performance in China, where the central
government essentially controls the mobility of local government leaders. We therefore
assume that local leaders in China may consider both, promotion benefits and corrupt
incomes, when they maximize their utilities. Moreover, the regional corruption level
determines the relative importance of bribery vs. promotion to the leaders. The local
environmental policy to some extent becomes a tool of the provincial leaders in China. When
the local level of corruption is low, implying the difficulty in engaging in corrupt activities,
the regional leader’s weight on promotion relative to bribery is added. Since exports are an
important source of economic growth in China, the local leader is likely to loosen the local
environmental policy under the pressure of the export industries which play an important role
in the local economy in order to encourage regional economic growth to earn a promotion for
himself. In a highly corrupt region, however, the relative importance of bribery versus
promotion to the local leader rises. He may have an incentive to extort bribes from local firms
including export-oriented firms by the reinforcement of local government regulations
including environmental ones, provided that there are enough local firms especially those
export-oriented so that he can extort lots of bribes without destroying the local economy
which may cause his termination. According to this mechanism, trade openness increases the
stringency of environmental policy if the local corruption level is high, while it decreases the
stringency of environmental policy if the local corruption level is low. Our interpretation
based on the unique Chinese political system with centralized personnel control, one of the
“Chinese characteristics”, can therefore explain our empirical results. Finally, we conclude
that corruption affects environmental quality mainly by influencing regional environmental
policy stringency directly and indirectly in China.

167 
 
 
Table 7-9 Effect of corruption on environment policy
Ratio of industrial waste water meeting discharge standards
Province level City level (2005) Province level City level (2005)
Pooled Fixed effects Fixed effects OLS 2SLS Pooled Fixed effects Fixed effects OLS 2SLS
OLS OLS 2SLS OLS OLS 2SLS
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
Corruption 0.023*** 0.0092 -0.14** 0.18 -6.96* 0.0063 0.018* -0.16*** -0.38 -7.05*
(0.0077) (0.0079) (0.056) (1.68) (4.22) (0.0098) (0.010) (0.052) (1.82) (4.22)
Openness -0.053** -0.18*** -0.34*** 1.99** 0.97 -0.16*** -0.12* -1.06*** 0.87 2.74
(0.026) (0.063) (0.060) (0.89) (1.19) (0.041) (0.070) (0.27) (1.43) (3.53)
Corruption* Openness 0.045*** -0.028** 0.30*** 100.5 -321.1
(0.016) (0.013) (0.11) (126.3) (461.0)
Income -5.72 -1.24 4.40 -254 -573 -1.68 -2.28 31.6*** -89.3 -706
(4.45) (3.95) (5.32) (335) (423) (4.56) (4.08) (12.0) (330) (579)
Income2 0.74 0.21 -0.38 33.4 69.6 0.30 0.32 -3.39** 14.5 84.6
(0.48) (0.44) (0.57) (36.9) (46.9) (0.49) (0.45) (1.32) (36.5) (64.9)
Income3 -0.030* -0.0099 0.010 -1.41 -2.77 -0.015 -0.014 0.12** -0.69 -3.33
(0.017) (0.016) (0.020) (1.35) (1.73) (0.018) (0.016) (0.048) (1.34) (2.42)
Education 0.075** 0.042* 0.075*** -0.014 -0.0084 0.079** 0.040 0.10*** -0.012 -0.0015
(0.031) (0.024) (0.021) (0.011) (0.013) (0.031) (0.024) (0.022) (0.011) (0.015)
Urbanization -0.041 -0.093** 0.017 -3.18 -2.18 -0.035 -0.089** 0.10* -2.57 1.63
(0.035) (0.046) (0.048) (3.06) (3.32) (0.035) (0.045) (0.055) (2.99) (4.44)
Industrialisation -0.39*** -0.40 1.28*** 14.28 1.23 -0.37*** -0.41 0.66 15.58 4.30
(0.12) (0.49) (0.38) (9.60) (13.21) (0.12) (0.49) (0.45) (9.54) (12.99)
Population 0.100*** -0.032 -0.023 2.81* 1.92 0.11*** -0.063 0.084 1.91 0.62
(0.011) (0.22) (0.033) (1.50) (1.78) (0.012) (0.22) (0.053) (1.17) (1.62)
Infrastructure 0.0062*** 0.0034*** 0.0073*** -2.81* -3.32** 0.0053*** 0.0034*** 0.000056 -1.89 -2.05
(0.00098) (0.0012) (0.0013) (1.66) (1.60) (0.0011) (0.0012) (0.0033) (1.41) (1.40)
Municipality 1.87 1.73 2.31 4.22
(2.95) (3.38) (2.81) (3.65)
Constant 13.9 2.52 -16.0 673 1629 1.51 5.98 -98.1*** 206 2033
(13.8) (11.6) (16.4) (1004) (1263) (14.1) (12.2) (36.3) (987) (1714)
R-squared 0.61 0.85 0.36 0.62 0.85 0.33
First stage
F test of excluded IVs
Corruption 8.06[0.00] 8.51[0.00] 4.07[0.02] 4.81[0.01]
Corruption* Openness 3.92[0.02] 1.02[0.36]
Anderson canon. corr. LM statistic 9.17[0.00] 9.45[0.00] 5.61[0.02] 10.72[0.00]
Observations 277 277 277 120 120 277 277 277 118 118
Notes: Robust standard errors in parentheses, p-values in brackets. ***, **and * denote significance at 1%, 5% and 10% respectively.

168 
 
 

7.4 Conclusion

Cross-country analyses in corruption with subjective survey data are suffering from a number
of biases. A comprehensive case study of a representative country may provide a helpful
supplement to these studies. Limited information is available outside the US. In this chapter,
we investigate the consequences of corruption using two objective data sets and alternative
corruption measures across Chinese regions. Glaeser and Saks (2006) pointed out that the
noise of corruption data, the small sample size and the narrow variation in cross regions make
it difficult for researchers to identify relationships between corruption and other variables in a
within-country analysis. The relatively great regional disparity in China mitigates the
problem of narrow variation across sub-nations in within-country analysis. We also employ
both the fixed effects approach and the instrumental variables approach to address potential
endogeneity problem which may bias our estimation. The fact that our results are basically
consistent with prior findings somehow validates our analysis. Furthermore, two
complementary data sets and alternative corruption measures in our analysis contribute to the
robustness of our findings.
Our research identifies adverse influences of corruption on economic development which
have been observed in cross-country studies. Like prior studies, we first observe that
corruption appears to lower economic growth. Unlike previous research, we subsequently
obtain solid evidence that corruption has simultaneously positive and negative effects on
economic growth. The impact of corruption detected in literature, either negative or positive,
might be the balance of the two simultaneous effects in a specific institutional environment.
Corruption also affects the income distribution in China which is an important aspect of
economic development. Similar to cross-country analyses, we find that corruption
considerably increases income inequality in China. We also find that regional corruption
significantly reduces inbound foreign direct investment, a main source of economic growth in
Chinese regions. This finding sheds new light on the “China puzzle” (Wei, 2000b): the
seemingly positive relationship between corruption and FDI inflows in China in previous
cross-country comparisons. As to the impact of corruption on public expenditures, we
observe that corruption significantly decreases government spending on education and
science, while increases public expenditure on social security. Turning to the nexus between
corruption and the environment, we observe that corruption substantially aggravates pollution
mainly through loosening environment regulations. Furthermore, corruption in China is also

169 
 
 

observed to modify the effect of trade openness on the environment policy stringency in the
similar way to that suggested in previous studies.
In summary, our study casts new light in a broadly manner on the consequences of
corruption especially in developing countries and hence is a constructive complement to
previous research about the consequences of corruption.

170 
 
  

Appendix
Table 7-10 Data description
Variable Data Description Source Mean S.D.
Corruption Regional registered cases on corruption in procurator’s China Procuratorial Yearbooks 3.09 0.94
office per 100,000 population
Average entertainment and travel costs relative to sales World Bank (2006) 1.13 0.45
of investigated firms in cities
Income Logarithm of per capita real gross regional product China Statistical Yearbooks 9.21 0.65
China City Statistical Yearbooks 9.47 0.66
Education Provincial fraction of the population over 6 with China Statistical Yearbooks 5.44 4.31
college completed
Public library collections per 100 people in a city China City Statistical Yearbooks 46.57 56.61
Openness Regional ratio of import and export to gross regional China Statistical Yearbooks 0.30 0.37
product China City Statistical Yearbooks 0.39 0.72
Industrialization Regional industrial contribution to gross regional China Statistical Yearbooks 0.37 0.10
product China City Statistical Yearbooks 0.48 0.091
Expenditure Ratio of regional government expenditure to gross China Statistical Yearbooks 0.16 0.11
regional product China City Statistical Yearbooks 0.086 0.030
FDI Regional inward foreign direct investment per capita China Statistical Yearbooks 4.65 6.68
China City Statistical Yearbooks 12.34 20.06
Urbanization Regional share of urban population China Statistical Yearbooks 0.32 0.16
China City Statistical Yearbooks 0.38 0.19
Investment Regional investment in the fixed assets/ gross regional China Statistical Yearbooks 0.41 0.14
product
Infrastructure Regional road mileage per 10000 people China Statistical Yearbooks 9.49 14.22
World Bank (2006) 2.41 0.52
Loan pay Average investigated firms’ expectation of informal World Bank (2006) 7.22 4.72
payments for loans in a city
Tax Average taxes and fees relative to firms’ sales of firms World Bank (2006) 4.94 1.40
investigated in a city
Red tape Average days per year that enterprise staff must spend World Bank (2006) 60.54 21.39
interacting with four major government bureaucracies
(tax administration, public security, environmental
protection, and labour and social security)
Population Logarithm of regional population (ten millions) China Statistical Yearbooks 8.03 0.89
(million (city)) China City Statistical Yearbooks 8.48 0.56
Municipality Dummy which equals 1if a city is among 4
municipalities, 0 otherwise
Others Regional number of middle school students China Statistical Yearbooks
China City Statistical Yearbooks
Regional number of research workers China Statistical Yearbooks
China City Statistical Yearbooks
Industrial output and employment China City Statistical Yearbooks
regional public expenditures China Statistical Yearbooks
China City Statistical Yearbooks
All environmental data China Statistical Yearbooks
China City Statistical Yearbooks

171 
 
 

Chapter Eight Conclusion

8.1 Summary of Findings

This thesis comprehensively studies the causes and consequences of corruption in both cross-
country and within-country contexts, mainly focusing on China. The thesis commences by
extensively investigating the causes of corruption in chapters 2, 3, 4, 5 and 6.
Chapter 2 adopts the standard economic approach to explore the causes of corruption in
China using two different sets of data at regional levels (provinces and cities). It is observed
that regions with more anti-corruption efforts, higher education attainment, Anglo-American
historic influence, higher openness, more access to media, higher relative wages of
government employees, and a greater representation of women in legislature are markedly
less corrupt; while the social heterogeneity, deregulation and abundance of resources,
substantially breed regional corruption. Moreover, fiscal decentralization is discovered to
depress corruption significantly. This study also finds that there is currently a positive
relationship between corruption and the economic development in China that is mainly driven
by the transition to a market economy. The findings above are fairly conceivable since this
study has addressed the potential endogeneity bias using the IV approach, and checked their
robustness with different specifications and data sets.
In Chapter 3 this thesis investigates the relationship between one of the informal aspects
of political institution. Specifically, this study uses the micro-level data from the World
Values Survey to explore the impact of political interest on corruption. To obtain robust
results, this study utilizes alternative measures for corruption: perceived corruption and the
justifiability of corruption, and three different proxies for political interest in the empirical
analysis. With the weighted ordered probit estimation (including 2SLS), this study presents
both cross-country and within-country evidence that a high level of political interest helps to
reduce the level of corruption within a society.
This thesis then continues by exploring the influence of democracy on corruption in
Chapter 4. To make the results more solid, this study establishes a political economy model
demonstrating that the effect of democracy on corruption is conditional on income
distribution and property rights protection. Then a cross-national panel data is constructed.
Employing fixed-effect and instrumental variable approaches, this study provides empirical
evidence that the effect of democracy upon corruption depends on the level of property rights

172 
 
 

protection and income inequality, which suggests that in order to effectively control
corruption we may highlight protection property rights and the mitigation of income
inequality rather than just democratization.
Chapter 5, using the social economic approach, performs a cross-country analysis to test
whether social interaction or social norms plays an important role in determining corruption
in a cross-country context. The study first builds a simple behavioural model to explain why
engaging in corruption results in a disutility of guilt. Guilt itself depends on the average guilt
levels of others within a country. This model also explores whether - and to what extent -
group dynamics or socialization and past experiences affect corruption. The empirical section
presents evidence using two data sets at a micro level and a large macro level panel data set
covering almost 20 years. The results indicate that the willingness to engage in corruption is
influenced by the perceived activities of peers and other individuals. Moreover, the panel data
set at the macro level indicates that past levels of corruption have a strong impact on the
current corruption level.
The thesis, however, turns to examine with the objective data whether the above findings
hold within the Chinese context in Chapter 6. Using Chinese province-level panel data, this
study clearly indicates that social interaction has a significantly positive effect on the
corruption rate in China. Therefore both cross-country and within-country findings
underscore the relevance of social interaction in corruption.
Chapter 7， however， comprehensively evaluates the consequences of corruption with
complementary Chinese data sets and alternative corruption measures. Adopting a novel
approach it provides solid evidence that corruption can simultaneously have both positive and
negative effects on economic development. The overall impact of corruption might be the
balance of the two effects, which may depend on the specific institutional environment.
Corruption is also observed to considerably increase the income inequality in China.
Furthermore this study also finds that corruption in China significantly decreases government
spending on education, science and public health, while it significantly increases public
expenditure on social security. And local corruption is observed to significantly reduce FDI
in Chinese regions. This finding sheds new light on the “China puzzle” that China is the
largest developing host of FDI while it is appears to be very corrupt. Finally the study
documents that corruption substantially aggravates pollution probably through a loosening of
the environmental regulation, and that it modifies the effects of trade openness and FDI on

173 
 
 

the stringency of environmental policy in a manner similar to that observed in literature to

date.
Overall, this thesis adds to the current literature by a number of novel findings concerning
both the causes and the consequences of corruption.

8.2 Policy Implications

The policy implications are discussed in the context of China. The wide-ranging adverse
effects of corruption on Chinese society disclosed in the thesis call for effective policies to
curb corruption. Based on the findings regarding the causes of corruption in this study, this
study provides some suggestions to design coherent anti-corruption measures in China.
Economic development and marketization appear to increase corruption during the
transitional process of China. However, considering the fact that economic development and
marketization substantially determine Chinese transition, our policy suggestion would be to
promote economic growth and marketization in order to accelerate this transitional process.
Corruption may be depressed by the economic development and marketization in China after
the accomplishment of the transition, which destroys the foundation of business-government
collusion.
The results here also highlight the view that an equal society, namely one with income
equality, gender equality and racial equality, has less corruption. A reduction in corruption is
an additional benefit of a society with more socio-economic equality. Furthermore, one of the
beneficial aspects of a more educated population is to decrease corruption.
The negative effects of trade openness and fiscal decentralization on corruption suggest
that competition, either between enterprises or governments, plays an important role in
controlling corruption. Policies inducing more competition in either commercial or political
markets are encouraged while any “marketization” policies leading to a monopoly due to the
government-business collusion are undesirable. Moreover, breeding corruption is one of the
detrimental aspects of the resource curse. Much attention should be paid to the role of
anticorruption in the resourceful provinces.
The media in China, though controlled by the government, acts as a watchdog over
corruption. More press freedom is however expected since according to Brunetti and Weder
(2003), freer press controls corruption more effectively. Additionally, although high salary
significantly prevents officials from corruption in China, as Rose-Ackerman (1999) argued, a
high salary of the public sector can be justified only if its productivity increases while its size

174 
 
 

decreases. Otherwise high pay in the public sector itself is a kind of corruption though
individual corruption acts may decrease in the public sector with high salary.
Social interactions however are found to be highly relevant in the incidence of corruption.
Regional corruption is affected by neighbourhood corruption. Anti-corruption in an area has a
positive spillover effect in reducing corruption in contiguous areas. This requires either
neighbouring areas to coordinate their individual anti-corruption efforts with regional
agreements or policy makers should take spillover effects into account when allocating
resources to corruption controls. Moreover, previous corruption levels have a significant
effect on the current corruption level. The evolution of corruption is a path-dependent process.
Rigorous anti-corruption measures need to be carried out for a long period to control
corruption in areas where corruption is pandemic. As suggested by Aidt (2003), a “big push”
like the one that took place in Hong Kong in the 1970s, might be needed to address the
corruption levels in areas where previous corruption rates have been high.
China is now on the track for democratisation. Some scholars insist that democratisation
can always curb corruption. However, this study argues that democracy is not a panacea for
corruption. The effect of democracy actually depends on the level of property rights and on
income inequality. To control corruption, protection of property rights, mitigation of income
inequality and economic development might be more essential, especially in poor countries.

8.3 Further Research

The validity of the findings on corruption to a large extent depends on the accuracy of
corruption measures that the researchers have. As discussed before, both the subjective cross-
country data and the objective within-country data that are currently utilized in the research
of corruption have their own inherent constrains. “If we cannot correctly measure corruption,
our ability to conduct empirical tests is severely impaired” (Banerjee et al., 2009, p.1).
However, measuring corruption accurately is a large challenge for economists as corruption
is illicit and hence secretive. One of the tasks of the next wave of research is therefore the
enhancement of our ability to measure corruption precisely.
A major innovative method developed recently in measuring corruption is to use quasi-
experimental and experimental approaches to measure corruption. This can facilitate us in our
attempts to better understand the causes and consequences of corruption at the micro level.
For example, Bertrand et al. (2007) carried out a field experiment to investigate the process
of obtaining the driving license in India. In their experiment, some applicants are observed to
make extralegal payments and to obtain licenses without attending the driving test. The

175 
 
 

results of their experiment show that corruption does distort resource allocation. Barron and
Olken (2007) also designed a field study in which surveyors accompanied truck drivers on
304 trips in two Indonesian provinces. The surveyors observed more than 6,000 illegal
payments to the traffic police, soldiers, and weigh station attendants during these trips. With
the data, they empirically show that as Shleifer and Vishny (1993) suggest, market structure
substantially influences the level of bribes. These studies allow us to measure corruption
more accurately and are also very helpful in testing the micro-theories of corruption.
However, no field experiment on corruption has been performed in a Chinese context. To
further understand the causes and consequences of corruption in China, a series of field
experiments measuring corruption needs to be carried out in order to test corruption theories
within the unique context of China. Actually there are many interesting candidates for field
experiments in China. For example, to obtain an entrance into key schools for their children,
parents in Chinese cities often have to make extralegal payments to these schools and to their
principals. Furthermore patients in China usually have to send their doctors red envelopes if
they hoped to be better treated. These are some potential field experiments that the author
plans to explore in future work.
 

176 
 
 

References

Abed, G.T., and Gupta, S., 2002. Governance, Corruption & Economic Performance. International
Monetary Fund.
Acemoglu, D., 1995. Reward Structures and the Allocation of Talent. European Economic Review 39,
17-33.
Acemoglou, D., Johnson, S., Robinson, J.A., 2001. The Colonial Origins of Comparative
Development: An Empirical Investigation. American Economic Review 91, 1369–1401.
Adam, C., 1993. Testing for Regime Shifts in Short-Sample African Macroeconomic Data: A Survey
of Some Monte Carlo Evidence. Working Paper 93.1, The Study of African Economies, University
of Oxford.
Adams, C. 1993. For Good and Evil. The Impact of Taxes on the Course of Civilization. Madison
Books, London.
Adams, J.S., 1965. Inequity in Social Exchange. In Advances in Experimental Social Psychology, eds.
Berkowitz, Leonard. New York: Academic Press, 167-299.
Ades, A., Di Tella, R., 1996. The Causes and Consequences of Corruption. IDS Bulletin 27 (2), 6–10.
Ades, A., Di Tella, R., 1997. The New Economics of Corruption: A Survey and Some New Results.
Political Studies 45, 496-515.
Ades, A., Di Tella, R., 1999. Rents, Competition and Corruption. American Economic Review 89,
982-993.
Adsera, A., Boix, C. and Payne, M., 2003. Are You Being Served? Political Accountability and
Quality of Government. Journal of Law, Economics and Organization 19, 445-490.
Aidt, T.S., 2003. Economic Analysis of Corruption: A Survey. The Economic Journal 113, 632-652.
Aidt, T.S., 2009. Corruption, Institutions, and Economic Development. Oxford Review of Economic
Policy 25, 271-291.
Aidt, T.S., Dutta, J. and Sena, V., 2008. Governance Regimes, Corruption and Growth: Theory and
Evidence. Journal of Comparative Economics 36, 195-220.
Akcay, S., 2001. Is Corruption an Obstacle for Foreign Investors in Developing Countries? a Cross-
Country Evidence. Yapi Kredi Economic Review 12, 27-34.
Akerlof, G.A., 1980. A Theory of Social Custom, of Which Unemployment May be One
Consequence. The Quarterly Journal of Economics 94, 749-75.
Alesina, A., Angeletos, G.M., 2005. Corruption, Inequality, and Fairness. Journal of Monetary
Economics 52, 1227-1244.
Alesina, A., Devleeschauwer, A., Easterly, W., Kurlat, S., Wazciarg, R., 2003. Fractionalization.
Journal of Economic Growth 8, 155–194.
Alm, J., 1996. Explaining Tax Compliance. In Exploring the Underground Economy, Susan Pozo, ed.
Kalamazoo: W. E. Upjohn Institute for Employment Research, 103-128.

177 
 
 

Alm, J., Martinez-Vazquez, J. and Torgler, B. 2006. Russian Attitudes Toward Paying Taxes- Before,
During, and After the Transition. International Journal of Social Economics 33, 832-857.
Alm, J., McClelland, G. H., and Schulze, W. D., 1999. Changing the Social Norm of Tax Compliance
by Voting, KYKLOS 52, 141-171.
Anderson, T., and Hsiao, C., 1981. Estimation of Dynamic Models with Error Components. Journal of
the American Statistical Association 76, 598-606.
Andreoni, J. and Miller, J., 2002. Giving According to GARP: An Experimental Test of the
Consistency of Preferences for Altruism. Econometrica 70, 737-753.
Andreoni, J. and Vesterlund, L., 2001. Which Is the Fair Sex? Gender Differences in Altruism,
Quarterly Journal of Economics 116, 293-312.
Andvig, J.C. and Moene, K.O., 1990. How Corruption May Corrupt. Journal of Economic Behaviours
& Organization 13, 63-76.
Angrist, J.D., Krueger, A.B., 2001. Instrumental Variables and the Search for Identification: From
Supply and Demand to Natural Experiments. Journal of Economic Perspectives 15, 69-85.
Arellano, M., and Bond, S., 1991. Some Tests of Specification for Panel Data: Monte Carlo Evidence
and an Application to Employment Equations. Review of Economic Studies 58, 277-297.
Arellano, M.., and Bover, O., 1995. Another Look at the Instrumental-Variable Estimation of Error-
Components Models. Journal of Econometrics 68, 29-51.
Attila, G., 2008. Is Corruption Contagious? An Econometric Analysis. Norwegian Institute of
International Affairs Working Paper, No. 742.
Banerjee, A.V., 1992. A Simple Model of Herd Behavior. Quarterly Journal of Economics 107, 797-
817.
Banerjee, A., Hanna, R., and Mullainathan, S., 2009. Corruption. Prepared for the Handbook of
Organizational Economics.
Banfield, E.C., 1958. The Moral Basis of a Backward Society. New York: Free Press.
Bardhan, P. and Mookherjee, D., 2006. Decentralization, Corruption and Government Accountability.
In: Rose-Ackerman, S. (Eds.), International Handbook on the Economics of Corruption.
Cheltenham: Edward Elgar. pp. 161-188.
Barro, R., 2000. Inequality and Growth in a Panel of Countries. Journal of Economic Growth 5, 5-32.
Barron, P. and Olken, B.A., 2009. The Simple Economics of Extortion: Evidence from Trucking in
Aceh. Journal of Political Economy 117, 417-252.
Basu, S. and Li, D., 2000. A Theory of the Reform of Bureaucratic Institutions. Econometric Society
World Congress 2000 Contributed Papers no. 1271.
Baumol, W. J., 1990, “Entrepreneurship: Productive, Unproductive, and Destructive. Journal of
Political Economy. 98, 893-921.
Becker, G.S., 1968. Crime and Punishment: An Economic Approach. Journal of Political Economy 76,
169-217.

178 
 
 

Becker, G.S. 1976. The Economic Approach to Human Behavior. Chicago: The University of
Chicago Press.
Becker, G.S. 1996. Accounting for Tastes. Cambridge, MA: Harvard University Press.
Becker, G.S. and Murphy, K.M., 2000. Social Economics: Market Behaviour in a Social Environment.
Cambridge, MA: Belknap / Harvard Press.
Begovic, B., 2006. Economic Inequality and Corruption. Working paper, Center for Liberal-
Democratic Studies,
Belgrade, Serbia.
Berninghaus, S.K. and Güth, W., 2002. Buying Versus Hiring - An Indirect Evolutionary Approach.
Discussion Papers on Strategic Interaction 2002-39, Strategic Interaction Group, Max Planck
Institute of Economics.
Besley, T., Ghatak, M., 2009. Improvement and Extension of Property Rights. In: Rodrik, D. and
Rosenzweig, M.R. (Eds.), Handbook of Development Economics 5, Elsevier, Burlington.
Bertrand, M., Djankov, S., Hanna, R., and Mullainathan, S., 2007. Obtaining a Driver's License in
India: An Experimental Approach to Studying Corruption. The Quarterly Journal of Economics
122, 1639-1676.
Bertrand, M., and Mullainathan, S., 2001. Do People Mean What They Say? Implications for
Subjective Survey Data. American Economic Review 91, 67-72.
Bikhchandani, S., Hirshleifer, D. and Welch, I. 1992. A Theory of Fads, Fashion, Custom, and
Cultural Change as Informational Cascades. Journal of Political Economy 100, 992-1026.
Bikhchandani, S., Hirshleifer, D. and Welch, I., 1998. Learning from the Behavior of Others:
Conformity, Fads, and Informational Cascades.” Journal of Economic Perspectives 12, 151-170.
Billger, S.M., Goel, R.K., 2009. Do Existing Corruption Levels Matter in Controlling Corruption?
Cross-Country Quantile Regression Estimates. Journal of Development Economics 90, 299-305.
Bird, R.M., Martinez-Vazquez, J. and Torgler, B., 2006. Societal Institutions and Tax Effort in
Developing Countries. In The Challenges of Tax Reform in the Global Economy, eds. Alm, James,
Jorge Martinez-Vazquez and Mark Rider. New York: Springer, 283-338.
Bird, R.M., Martinez-Vazquez, J., and Torgler, B., 2008. Tax Effort in Developing Countries and
High Income Countries: The Impact of Corruption, Voice and Accountability. Economic Analysis
& Policy 38. 55-71.
Blume, L. E. and Durlauf, S.N., 2004. The Interactions-Based Approach to Socioeconomic Behaviour.
In Social Dynamics, Steven N. Durlauf, H. Peyton Young (eds.), The MIT Press.
Blundell, R.W., and Bond, S.R., 1998. Initial Conditions and Moment Restrictions in Dynamic Panel
Data Models. Journal of Econometrics 87, 115-143.
Boettke, P., Chamlee-Wright, E., Gordon, P., Ikeda, S., Leeson, P.T., and Sobel, R., 2007. The
political, economic, and social aspects of Katrina. Southern Economic Journal 74, 363–376.

179 
 
 

Bohara, A.K., Mitchell, N.J., Mittendorff, C.F., 2004. Compound Democracy and the Control of
Corruption: A Cross-Country Investigation. Policy Studies Journal 32, 481-499.
Bolton, G.E. and Ockenfels, A., 2000. ERC: Theory of Equity, Reciprocity, and Competition.
American Economic Review 90, 166-193.
Boyce, C. J., 2010. Understanding Fixed Effects in Human Well-Being. Journal of Economic
Psychology 31, 1–16.
Bradhan, P., 1997. Corruption and Development: A Review of Issues. Journal of Economic Literature
35, 1320-1346.
Brambor, T., Clark, W.R., Golder, M., 2006. Understanding Interaction Models: Improving Empirical
Analyses. Political Analysis 14, 63-82.
Brock, W. A. and Durlauf, S.N., 1995. Discrete Choice with Social Interactions I: Theory. NBER
Working Papers No. 5291.
Brock, W. A. and Durlauf, S.N., 2001. Discrete Choice with Social Interactions. The Review of
Economic Studies 68, 235-260.
Brunetti, A. and Weder, B., 2003. A Free Press is Bad News for Corruption. Journal of Public
Economics 87, 1801-1824.
Cadot, O., 1987. Corruption as a Gamble. Journal of Public Economics 33, 223-244.
Cai, H., Fang, H. and Xu, C.L., 2009. Eat, Drink, Firms and Government: An Investigation of
Corruption from Entertainment and Travel Costs of Chinese Firms. Journal of Law and Economics,
forthcoming.
Carpini, D. and Keeter, S., 1996. What Americans Know about Politics and Why It Matters. New
Haven, CT: Yale University Press.
Charness, G. and Rabin, M., 2002. Understanding Social Preferences with Simple Tests. The
Quarterly Journal of Economics 117, 817-869.
Cervellati, M., Fortunato, P., Sunde, U., 2008. Hobbes to Rousseau: Inequality, Institutions And
Development. Economic Journal 118, 1354-1384.
China Continuation Committee (2007), Christian Occupation of China. Beijing, China: Chinese
Social Science Press.
Chowdhury, S., 2004. The Effect of Democracy and Press Freedom on Corruption: An Empirical Test.
Economics Letters 85, 93-101.
Cole. M.A., 2007. Corruption, Income and the Environment: an Empirical Analysis. Ecological
Economics 62, 637-647.
Cole, M.A., Elliott, R.J.R., Fredriksson, P., 2006. Endogenous Pollution Havens: Does FDI Influence
Environmental Regulations? Scandinavian Journal of Economics 108, 157-178.
Cole, M.A., Elliott, R.J.R., Zhang, J., 2006. Corruption, Governance and FDI Location in China: A
Province-Level Analysis. Journal of Development Studies 45, 1494-1512.
Coleman, J.S., 1990. Foundations of Social Theory. Cambridge, MA: Harvard University Press.

180 
 
 

Cook, P.J., and Goss, K.A., 1996. A Selective Review of the Social-Contagion Literature. Working
paper, Sanford Institute of Policy Studies, Duke University.
Dabla-Norris, E. and Wade, P., 2002. Production, Rent Seeking, and Wealth Distribution. in
Governance, Corruption, and Economic Performance, ed. by G. T. Abed and S. Gupta.
International Monetary Fund, Washington.
Dahl, R. A., 1974. Who Governs? Yale University Press, New Haven.
Damania, R., Fredriksson, P.G., List, J.A., 2003. Trade Liberalization, Corruption, and Environmental
Policy Formation: Theory and Evidence. Journal of Environmental Economics and Management
46, 490–512.
Dasgupta, S., Wang, H., Wheeler, D., 2002. Confronting the environmental Kuznets curve. Journal of
Economic Perspectives 16, 147–168.
Davoodi, H.R. and Tanzi, V., 1997. Corruption, Public Investment, and Growth. IMF Working Paper,
WP/97/139.
Deininger, K., Squire, L., 1996. A New Data Set Measuring Income Inequality. World Bank
Economic Review 10, 565-591.
Demurger, S., 2001. Infrastructure Development and Economic Growth: An Explanation for Regional
Disparities in China? Journal of Comparative Economics 29, 95-117.
De Soto, H., 1989. The Other Path: The Invisible Revolution in the Third World. New York: Harper&
Row.
Dollar, D., Fisman, R. and Gatti, R., 2001. Are Women Really the "Fairer" Sex? Corruption and
Women in Government. Journal of Economic Behaviours and Organization 46, 423-429.
Dong, B. and Torgler, B., 2009. Corruption and Political Interest: Empirical Evidence at the Micro
Level. Journal of Interdisciplinary Economics 21, 295-325.
Dong, B., Dulleck, U. and Torgler, B., 2008. Conditional Corruption. CREMA Working Paper Series
2008-29.
Dong, B. and Torgler, B., 2010a. Social Interaction and Corruption: Evidence from China, draft,
Queensland University of Technology.
Dong, B. and Torgler, B., 2010b. The Causes of Corruption: Evidence from China. Working Paper
series No. 257, School of Economics and Finance, Queensland University of Technology.
Dreher, A., 2006. Does Globalization Affect Growth? Evidence from a New Index of Globalization.
Applied Economics 38, 1091-110.
Duesenberry, J.S., 1949. Income, Saving, and the Theory of Consumer Behavior. Cambridge, MA:
Harvard University Press.
Eagly, A. H. and Crowley, M., 1986. Gender and Helping Behavior: A Meta-Analytic Review of the
Social Psychological Literature. Psychological Bulletin 100, 283-308.
Eberhard, W., 1965. Chinese Regional Stereotypes. Asian Survey 5, 596-608.

181 
 
 

Eckel, C. C. and Grossman, P. J., 2001. Chivalry and Solidarity in Ultimatum Games. Economic
Inquiry 39, 171-188.
Egger, P. and Winner, H., 2005. Evidence on Corruption as an Incentive for Foreign Direct
Investment European Journal of Political Economy 21, 932– 952.
Elster, J. 1989. The Cement of Society: A Study of Social Order. Cambridge: Cambridge University
Press.
Elster, J. 1989. Social Norms and Economic Theory. Journal of Economic Perspectives 3, 99-117.
Emerson, P.M., 2006. Corruption, Competition and Democracy. Journal of Development Economics
81, 193-212.
European Values Questionnaire. 1999. Tilburg University, The Netherlands.
Falk, A., and Fehr, E., 2002. Psychological Foundations of Incentives. European Economic Review
46, 687-724.
Falk, A, Fischbacher, U. and Gächter, S., 2003. Living in Two Neighborhoods – Social Interaction in
the Lab. CESifo Working Paper No. 954, Munich.
Fan, G., Wang, X. and Zhu, H., 2010. The Marketilization Index of China: 2009 Report on the
Relative Progress of Marketization in Different Regions. Beijing: Economic Science Publishing
House.
Fehr, E., Fischbacher, U., von Rosenbladt, B., Schupp, J, and Wagner, G.G. 2003. A Nation-Wide
Laboratory Examining Trust and Trustworthiness by Integrating Behavioral Experiments into
Representative Surveys. CESifo Working Paper No. 866, Munich.
Fehr, E., and Gächter, S. 1997. Reciprocity and Economics: The Economic Implications of Homo
Reciprocans. Working Paper, University of Zurich.
Fehr, E., and Schmidt, K.M.., 1999. A Theory of Fairness, Competition, and Cooperation. The
Quarterly Journal of Economics 114, 817–868.
Field, A., Sosa, L., and Wu, X., 2006, Impacts of Endogenous Bribes on Foreign Direct Investment.
Manuscript, Department of Economics, University of North Carolina, Chapel Hill.
Fischbacher, U., and Gächter, S. 2006. “Heterogeneous Social Preferences and the Dynamics of Free
Riding in Public Goods.” CeDEx Discussion Paper No. 2006-01, Centre for Decision Research
and Experimental Economics, Nottingham.
Fischbacher, U., Gächter, S., and Fehr, E., 2001. Are People Conditionally Cooperative? Evidence
from a Public Goods Experiment. Economics Letters 71, 397–404.
Fisman, R. and Gatti, R., 2002a. Decentralization and Corruption: Evidence across Countries. Journal
of Public Economics 83, 325-345.
Fisman, R. and Gatti, R., 2002b. Decentralization and Corruption: Evidence from U.S. Federal
Transfer Programs. Public Choice 113, 25–35.
Fisman, R. and Svensson, J., 2007. Are Corruption and Taxation Really Harmful to Growth? Firm
Level Evidence. Journal of Development Economics 83, 63-75.

182 
 
 

Ford, R. C. and Woodrow, D. R., 1994. Ethical Decision Making: A Review of the Empirical
Literature. Journal of Business Ethics 13, 205–221.
Fredriksson, G. and Svensson, J., 2003. Political Instability, Corruption and Policy Formation: The
Case of Environmental Policy. Journal of Public Economics 87, 1383-1405.
Fredriksson, Per G., List, J. and Millimet, D.L., 2003. Bureaucratic Corruption, Environmental Policy
and Inbound US FDI: Theory and Evidence. Journal of Public Economics 87, 1407-1430.
Fredriksson, Per G., and Millimet, D.L., 2004. Comparative Politics and Environmental Taxation.
Journal of Environmental Economics and Management 48, 705-722.
Fredriksson, P., Vollebergh, H.R.J. and Dijkgraaf, E., 2004. Corruption and Energy Efficiency in
OECD Countries: Theory and Evidence. Journal of Environmental Economics and Management
47, 207–231.
Freedom House, 2009. Freedom on the Net: A Global Assessment of Internet and Digital Media.
Freille, S., Haque, M.E. and Kneller, R., 2007. A Contribution to the Empirics of Press Freedom and
Corruption. European Journal of Political Economy 23, 838-862.
Frey, B. S. and Feld, L. P., 2002. Deterrence and Morale in Taxation: An Empirical Analysis, CESifo
Working Paper No. 760.
Frey, B.S., and Meier, S., 2004a. Pro-social Behavior in a Natural Setting. Journal of Economic
Behavior and Organization 54, 65-88.
Frey, B.S., and Meier, S., 2004b. Social Comparisons and Pro-social Behavior: Testing ‘Conditional
Cooperation’ in a Field Experiment. American Economic Review 94, 1717-1722.
Frey, B. S., and Stutzer, A., 2000. Happiness, Economy and Institutions. Economic Journal 110, 918-
938.
Frey, B. S., and Stutzer, A., 2002. Happiness and Economics. Princeton University Press, Princeton.
Frey, B.S., and Torgler, B., 2007. Tax Morale and Conditional Cooperation. Journal of Comparative
Economics 35, 136–159.
Gächter, S., 2006. Conditional Cooperation: Behavioral Regularities from the Lab and the Field and
Their Policy Implications. unpublished manuscript, University of Nottingham.
Gerring, J. and Thacker, S.C., 2005. Do Neoliberal Policies Deter Political Corruption? International
Organization 59, 233-254.
Gërxhani, K., 2002. Tax Evasion in Albania: An Institutional Vacuum?” paper presented at the
Annual Meeting of the European Public Choice Society, Belgirate.
Glaeser, E.L., LaPorta, R., Lopez-De-Silanes, F., Shleifer, A., 2004. Do Institutions Cause Growth?
Journal of Economic Growth 9, 271-303.
Glaeser, E.L., Sacerdote, B. and Scheinkman, J.A., 1996. Crime and Social Interactions. The
Quarterly Journal of Economics 111, 507-548.
Glaeser, E.L. and Saks, R.E., 2006. Corruption in America. Journal of Public Economics 90, 1053-
1072.

183 
 
 

Glaeser, E.L. and Scheinkman, J.A., 2002. Non-Market Interactions. in Advances in Economics and
Econometrics: Theory and Applications, Eight World Congress, M. Dewatripont, L.P. Hansen, and
S. Turnovsky (eds.), Cambridge University Press.
Glaeser, E.L., Scheinkman, J., Shleifer, A., 2003. The Injustice of Inequality. Journal of Monetary
Economics 50, 199-222.
Glover, S. H., Bumpus, M. A., Logan, J. E. and Ciesla, J. R., 1997. Reexamining the Influence of
Individual Values on Ethical Decisionmaking. Journal of Business Ethics 16, 1319-1329.
Goel, R.K. and Nelson, M.A., 1998. Corruption and Government Size: A Disaggregated Analysis.
Public Choice 97, 107-120.
Goel, R.K., Nelson, M.A., 2005. Economic Freedom versus Political Freedom: Cross-Country
Influences on Corruption. Australian Economic Papers 44, 121-133.
Goel, R.K., and Nelson, M.A., 2007. Are Corrupt Acts Contagious? Evidence from the United States.
Journal of Policy Modeling 29, 839-850.
Goldsmith, A.A., 1999. Slapping the Grasping Hand: Correlates of Political Corruption in Emerging
Markets. American Journal of Economics and Sociology 58, 865-883.
Gong, T., 2006. Corruption and Local Governance: the Double Identity of Chinese Local
Governments in Market Reform. The Pacific Review 19, 85-102.
Gordon, J.P.P., 1989. Individual Morality and Reputation Costs as Deterrents to Tax Evasion.
European Economic Review 33, 797-805.
Gordon, S.C., 2009. Assessing Partisan Bias in Federal Public Corruption Prosecutions. American
Political Science Review 103, 534-554.
Gottfredson, M. R. and Hirschi, T., 1990. A General Theory of Crime. Stanford: Stanford University
Press.
Gradstein, M., 2007. Inequality, Democracy and the Protection of Property Rights. The Economic
Journal 117, 252-269.
Guiso, L., Sapienza, P. and Zingales, L., 2004. The Role of Social Capital in Financial Development.
American Economic Review 94, 526-556.
Gundlach, E. and Paldam, M., 2009. The Transition of Corruption: From Poverty to Honesty.
Economics Letters 103, 146-148.
Gupta, S., Davoodi, H.R., and Alonso-Termé, R., 2002. Does Corruption Affect Income Inequality
and Poverty? Economics of Governance 3, 23-45
Gupta, S., Davoodi, H.R., and Tiongson,E.R., 2002. Corruption and the Provision of Health Care and
Education Services, in Governance, Corruption, and Economic Performance, ed. by G. T. Abed
and S. Gupta, International Monetary Fund, Washington.
Gupta, S., de Mello, L. and Sharan, R., 2001. Corruption and Military Spending. European Journal of
Political Economy 17, 749–777.

184 
 
 

Güth, W, and Yaari, M.E., 1992. Explaining Reciprocal Behavior in Simple Strategic Games: An
Evolutionary Approach. in Explaining Process and Change: Approaches to Evolutionary
Economics, eds, Witt Ulrich, Michigan University Press, 23-34.
Habib, M. and Zurawicki, L., 2002. Corruption and Foreign Direct Investment. Journal of
International Business Studies 33, 291-307.
Hall, R.E., and Jones, C.I., 1999. Why do Some Countries Produce so Much More Output per Worker
than Others? Quarterly Journal of Economics 114, 83-116.
Harms, P. and Ursprung, H.W., 2002. Do Civil and Political Repression Really Boost Foreign Direct
Investments? Economic Inquiry 40, 651-663.
Harcourt, B.E., and Ludwig, J., 2006. Broken Windows: New Evidence from New York City and a
Five-City Social Experiment. University of Chicago Law Review 73, 271-320.
Hartog, J., Ferrer-i-Carbonell, A. and Jonker, N., 2002. Linking Measured Risk Aversion to
Individual Characteristics. Kyklos 55, 3-26.
Hayek, F.A., 1960. The Constitution of Liberty. University of Chicago Press, Chicago.
Heldt, T., 2005. Conditional Cooperation in the Field: Cross-Country Skiers? Behavior in Sweden.
unpublished manuscript, Uppsala University.
Henrich, J.. 2004. Cultural Group Selection, Coevolutionary Processes and Large-Scale Cooperation.
Journal of Economic Behavior and Organization 53, 3-35.
Hirschman, A. O. 1970. Exit, Voice, and Loyalty. Cambridge (Mass.): Harvard University Press.
Hooper, J., 1995. The New Spaniards. London: Penguin Books.
Hofstede, G. and Hofstede, G.J., 2005. Cultures and Organizations: Software of the Mind. New York:
McGraw Hill.
Hsiao, C., 2003. Analysis of Panel Data. Cambridge University Press, New York.
Hu, A., 2001. Corruption and Social Inequality. Jiangsu Social Sciences 2001(3), 51-53.
Huntington, S.P., 1968. Political Order in Changing Societies. New Haven, Yale University Press.
Huntington, S.P., 1991. The Third Wave: Democratization in the Late Twentieth Century. University
of Oklahoma Press, Norman.
Inglehart, R. et al., 2000. Codebook for World Values Survey. Ann Arbor: Institute for Social
Research.
Jain, A.K., 2001. Corruption: A Review. Journal of Economic Surveys 15, 71-121.
Jensen, P.S., Paldam, M., 2007. The Pattern of Democracy in the 20th Century: A Study of the Polity
Index. In: Casas, J. and Schwartz, P. (Eds.), Public Choice and the Challenges of Democracy.
Edward Elgar, Cheltenham.
Junger, M., 1994. Accidents. in: T. Hirschi and M. R. Gottfredson (eds.), The Generality of Deviance.
New Brunswick: Transaction Publishers: 81-112.
Kahan, D.M., 1998. Social Meaning and the Economic Analysis of Crime. The Journal of Legal
Studies 27, 609-622.

185 
 
 

Kahneman, D. and Tversky, A., 1979. Prospect Theory: An Analysis of Decision under Risk,
Econometrica 47, 263-291.
Kanazawa, S., 2005. Who Lies on Surveys, and What Can We Do About It? Journal of Social,
Political and Economic Studies 30, 361-370.
Kanbur, R and Zhang, X., 1999. Which Regional Inequality? The Evolution of Rural–Urban and
Inland–Coastal Inequality in China from 1983 to 1995. Journal of Comparative Economics 27,
686–701.
Kasper, W., and Streit, M. 1999. Institutional Economics: Social Order and Public Policy.
Cheltenham, UK: Edward Elgar.
Katz, M. and Shapiro, C., 1985. Network Externalities, Competition, and Compatibility. American
Economic Review 75, 424-440.
Kaufmann, D., 1997. Corruption: The Facts. Foreign Policy 107, 114-131.
Kaufmann, D., and Kraay, A., 2002. Growth without Governance. Policy Research Working Paper
Series 2928, World Bank.
Kaufmann, D., Kraay, A., and Mastruzzi, M., 2003. Governance Matters III: Governance Indicators
for 1996-2002. World Bank Economic Review 18, 253-287.
Keser, C., and van Winden, F., 2000. Conditional Cooperation and Voluntary Contributions to Public
Goods. Scandinavian Journal of Economics 102, 23-39.
Keizer, K., Lindenberg, S., and Steg, L., 2008. The Spreading of Disorder. Science 322, 1681–1685.
Kim, C.K., 1994. The Effects of Public Transfers and Tax Rate Changes on Reported Income:
Experimental Evidence. Ph.D. dissertation, University of Pittsburgh.
Klitgaard, R., 1988. Controlling Corruption. Berkeley, CA: University of California Press.
Knack, S. 1999. Aid Dependence and the Quality of Governance. IRIS Center, Working
Paper,University of Maryland.
Knack, S., 2006. Measuring Corruption in Eastern Europe and Central Asia: A Critique of the Cross-
country Indicators. Policy Research Working Paper Series 3968, World Bank.
Kornai, J., 1990. The Affinity between Ownership Forms and Coordination Mechanism: The
Common Experience of Reform in Socialistic Countries. Journal of Economic Perspective 4, 131-
147.
Krueger, A. O., 1974. The Political Economy of the Rent-Seeking Society, American Economic
Review, 64, 291-303.
Kuenzi, M., 2006. Non-Formal Education and Community Development in Senegal, Community
Development Journal 41, 210-222.
Kunicova, J., and Rose-Ackerman, S. 2005. Electoral Rules and Constitutional Structures as
Constraints on Corruption. British Journal of Political Science 35, 573-606.

186 
 
 

Kurzban, R., McCabe, K., Smith, V.L. and Wilson, B.J., 2001. Incremental Commitment and
Reciprocity in A Real-Time Public Goods Game. Personality and Social Psychology Bulletin 27,
1662-1673.
Lambsdorff , J.G., 2004. Background paper to the 2004 Corruption Perceptions Index.
http://www.icgg.org/downloads/CPI_Methodology.pdf
Lambsdorff, J.G., 2005. Consequences and Causes of Corruption – What do We Know from a Cross-
Section of Countries? Discussion Papers no. V-34-05. Department of Economics, University of
Passau.
Landman, T. and Larizza, M., 2009. Inequality and Human Rights: Who Controls What, When, and
How. International Studies Quarterly 53, 715–736.
LaPalombara, J., 1994. Structural and Institutional Aspects of Corruption. Social Research 61, 325-50.
La Porta, R., Lopez-De-Silanes, F., Shleifer, A. and Vishny, R.W., 1999. The Quality of Government.
The Journal of Law, Economics and Organization 15, 222-279.
Larraín, B.F. and Tavares, J., 2004. Does Foreign Direct Investment Decrease Corruption? Latin
American Journal of Economics 41: 217-230.
Larraín, B.F. and Tavares, J., 2007. Can Openness Deter Corruption? The Role of Foreign Direct
Investment. CEPR Discussion Papers no. 6488.
Leff, N.H., 1964. Economic Development through Bureaucratic Corruption. American Behavioural
Scientist 8, 8-14.
Leeson, P.T. and Sobel, R.S., 2008. Weathering Corruption. Journal of Law and Economics 51, 667-
681.
Leite, C. and Weidmann, J., 1999. Does Mother Nature Corrupt? Natural Resources, Corruption and
Economic Growth. IMF Working Paper no. 99/85.
Lenin, V.I., 1916. The Socialist Revolution and the Right of Nations to Self-Determination.
Vorbote 2.
Lewis, A. 1982. The Psychology of Taxation. Oxford: Martin Robertson.
Levi, M., 1988. Rules and Revenue. Berkeley/Los Angeles/London: University of California Press.
Levine, R. and Renelt, D., 1992. Sensitivity Analysis of Cross-Country Growth Regressions.
American Economic Review 82, 942-963.
Levin, M. and Satarov, G., 2000. Corruption and institutions in Russia. European Journal of Political
Economy 16, 113-132.
Li, H., and Zhou, L.-A., 2005. Political Turnover and Economic Performance: the Incentive role of
Personnel Control in China. Journal of Public Economics 89, 1743-1762.
Li, S.-M., 2001. Regional Development in China: States, Globalization and Inequality. China Review
International 8, 572-578.
Li, H., Xu, L.C., and Zou, H., 2000. Corruption, Income Distribution, and Growth. Economics and
Politics 12, 155-182.

187 
 
 

Li, S. and Hou, Y., 2003. Analysis on Socio-economic Characteristics of Chinese Regions. Research
Report, Department of Development Strategy and Regional Economy, Development Research
Centre of the State Council.
Liang, G., 1994. The Practical Encyclopaedia of Anti-Corruption in China and Foreign Countries.
Beijing: Xinhua Press.
Liu, A. P. L., 1983. The Politics of Corruption in the People’s Republic of China. American Political
Science Review 77, 602-623.
Lopez, R. and Mitra, S., 2000. Corruption, Pollution, and the Kuznets Environment Curve. Journal of
Environmental Economics and Management 40, 137-150.
Lu, Y., Png, I.P.L., and Tao, Z., 2010. Do Institutions not Matter in China? Evidence from
Manufacturing Enterprises. http://www.comp.nus.edu.sg/~ipng/research/instn.pdf.
Ludwig, J., and Kling, J.R., 2006. Is Crime Contagious? Working Papers Series 85, Center for
Economic Policy Studies, Department of Economics, Princeton University.
Lui, F.T. 1986. A Dynamic Model of Corruption Deterrence. Journal of Public Economics 31, 215-
236.
Lui, F.T., 1985. An Equilibrium Queuing Model of Bribery. Journal of Political Economy 93, 760-
781.
Lundberg, M., and Squire, L., 2003. The Simultaneous Evolution of Growth and Inequality.
Economic Journal, 113, 326–344.
Manion, M., 1996. Corruption by Design: Bribery in Chinese Enterprise Licensing. Journal of Law,
Economics, & Organization 12, 167-195.
Mauro, P., 1995. Corruption and growth. Quarterly Journal of Economics 110, 681-712.
Mauro, P., 1998. Corruption and the Composition of Government Expenditure. Journal of Public
Economics 69, 263–279.
Mauro, P., 2004. The Persistence of Corruption and Slow Economic Growth. IMF Staff Papers 51, 1.
Manski, C.F., 1993. Identification of Endogenous Social Effects: The Reflection Problem. Review of
Economic Studies 60, 531-542.
Manski, C.F., 2000. Economic Analysis of Social Interactions. Journal of Economic Perspectives 14,
115-136.
Martin, R., and Randal, J., 2005. Voluntary Contribution to a Public Good: A Natural Field
Experiment. unpublished manuscript, Victoria University of Wellington, Wellington.
Martinez-Vazquez, J., and McNab, R.M., 2000. The Tax Reform Experiment in Transition Countries.
National Tax Journal 53, 273-298.
Marx, K. 1849. Wage Labor and Capital. in K. Marx and F. Engels, Selected Works, Vol.1,
Moscow: Progress Publishers.
Mason, R. and Calvin, L. D., 1984. Public Confidence and Admitted Tax Evasion. National Tax
Journal 37, 490-496.

188 
 
 

Mears, D. P., Ploeger, M. and Warr, M., 2000. Explaining the Gender Gap in Delinquency: Peer
Influence and Moral Evaluations of Behavior, in: R. D. Crutchfield, G. S. Bridges, J. G. Weis and
C. Kubrin, Crime Readings. Thousand Oaks: Pine Forge Press: 143-148.
Mendez, F., and Sepulveda, F., 2006. Corruption, Growth and Political Regimes: Cross Country
Evidence. European Journal of Political Economy 22, 82-98.
Meng, X., Gregory R. and Wang Y., 2005. Poverty, inequality, and growth in urban China, 1986–
2000. Journal of Comparative Economics 33, 710-729.
Méon, P.-G. and Sekkat, K., 2005. Does Corruption Grease or Sand the Wheels of Growth? Public
Choice 122, 69-97.
Méon, P.-G. and Weill, L., 2008. Is Corruption an Efficient Grease? Bank of Finland, Institute for
Economies in Transition, BOFIT Discussion Papers 20/2008.
Mishra, A., 2006. Persistence of Corruption: Some Theoretical Perspectives. World Development 34,
349–358.
Mo, P.H., 2001. Corruption and Economic Growth. Journal of Comparative Economics 29, 66-79.
Mobarak, A.M., 2005. Democracy, Volatility and Development. The Review of Economics and
Statistics 87, 348-361.
Mocan, N., 2008. What Determines Corruption? International Evidence from Microdata. Economic
Inquiry 46, 493-510.
Mohtadi, H., Roe, T.L., 2003. Democracy, Rent Seeking, Public Spending and Growth. Journal of
Public Economics 87, 445-466.
Montinola, G.R., Jackman, R.W., 2001. Sources of Corruption: A Cross-Country Study. British
Journal of Political Science 32, 147-170.
Montinola, G. Qian, Y. and Weingast, B.R., 1995. Federalism, Chinese Style: The Political Basis for
Economic Success in China. World Politics 48, 50-81.
Murphy, K.M., Shleifer, A., and Vishny, R.W., 1991. The Allocation of Talent: Implications for
Growth. Quarterly Journal of Economics 106, 503-530.
Murphy, K.M., Shleifer, A., and Vishny, R.W., 1993. Why is Rent-Seeking so Costly to Growth?
American Economic Review 83, 409-414.
Myles, G.D. and Naylor, R.A., 1996. A Model of Tax Evasion with Group Conformity and Social
Custom. European Journal of Political Economy 12, 49-66.
Naylor, R.A., 1989. Strikes, Free Riders and Social Customs. Quarterly Journal of Economics 104,
771-805.
Nevo, A., 2001. Measuring Market Power in the Ready-to-Eat Cereal Industry. Econometrica 69,
307-442.
Nowell, C. and Tinkler, S., 1994. The Influence of Gender on the Provision of a Public Good. Journal
of Economic Behavior and Organization 25, 25-36.
Oates, W. E. 1999. An Essay on Fiscal Federalism. Journal of Economic Literature 37, 1120-1149.

189 
 
 

Ostrom, E., 2000. Collective Action and the Evolution of Social Norms. Journal of Economic
Perspectives 14, 137-158.
Paldam, M., 2001. Corruption and Religion: Adding to the Economic Model. Kyklos 54, 383-413.
Paldam, M. and Svendsen, G.T., 2000. Missing Social Capital and the Transition in Eastern Europe.
Working Papers 00-5, Aarhus School of Business, Department of Economics.
Pappaport, J. and Sachs, J.D., 2003. The United States as a Coastal Nation. Journal of Economic
Growth 8, 5-46.
Paternoster, R., Simpson, S., 1996. Sanction Threats and Appeals to Morality: Testing a Rational
Choice Model of Corporate Crime. Law and Society Review 30, 549-583.
Pellegrini, L. and Gerlagh, R., 2004. Corruption's Effect on Growth and its Transmission Channels.
Kyklos 57, 429-456.
Pelligrini, L. and Gerlagh, R., 2006a. Corruption, Democracy and Environmental Policy: An
Empirical Contribution to the Debate. Journal of Environment and Development 15, 332-354.
Pelligrini, L. and Gerlagh, R., 2006b. Corruption and Environmental Policies: What are the
Implications for the Enlarged EU? European Environment 16, 139-154.
Pellegrini, L. and Gerlagh, R., 2008. Causes of Corruption: A Survey of Cross-Country Analyses and
Extended Results. Economics of Governance 9, 245–263.
Persson, T., Tabellini, G., 1994. Is Inequality Harmful for Growth? American Economic Review 84,
600-621.
Plümper, T. and Troeger, V., 2007. Efficient Estimation of Time-Invariant and Rarely Changing
Variables in Finite Sample Panel Analyses with Unit Fixed Effects. Political Analysis 15, 124–139.
Polinsky, M.A. and Shavell, S., 1999. Corruption and Optimal Law Enforcement. Working Paper 171,
John M. Olin Program in Law and Economics, Stanford Law School.
Polinsky, M.A. and Shavell, S. 2000. The Economic Theory of Public Enforcement of Law. Journal
of Economic Literature 38, 45-76.
Posner, R.A., 1997. Social Norms and the Law: An Economic Approach. American Economic
Review 87, 365-369.
Rabin, M., 1993. Incorporating Fairness into Game Theory and Economics. American Economic
Review 83, 1281-1302.
Rabin, M., 1998. Psychology and Economics. Journal of Economic Literature 36, 11-46.
Rappaport, J. and Sachs, J., 2003. The United States as a Coastal Nation. Journal of Economic Growth
8, 5 –46.
Rege, M., 2004. Social Norms and Private Provision of Public Goods. Journal of Public Economic
Theory 6, 65-77.
Reiss, M. C. and Mitra, K., 1998. The Effects of Individual Difference Factors o the Acceptability of
Ethical and Unethical Workplace Behaviors. Journal of Business Ethics 17, 1581-1593.
Rock, M.T., 2009. Corruption and Democracy. Journal of Development Studies 45, 55 – 75.

190 
 
 

Rose-Ackerman, S., 1978. Corruption: A Study in Political Economy. New York: Academic Press.
Rose-Ackerman, S. 1997. The Political Economy of Corruption, in: K. A. Elliott (ed.). Corruption and
the Global Economy. Washington D.C.: Institute for International Economics: 31-60.
Rose-Ackerman, S., 1999. Corruption and Government: Causes, Consequences, and Reform.
Cambridge: Cambridge University Press.
Sachs, J.D. and Warner, A.M., 2001. The Curse of Natural Resources. European Economic Review 45,
827-838.
Sah, R.K., 1988. Persistence and Pervasiveness of Corruption: New Perspectives. Working Paper
Series, 560, Economic Growth Center, Yale University.
Sah, R., 1991. Social Osmosis and Pattern of Crime. Journal of Political Economy 99, 1272-1295.
Sah, R.. (2005). Large and Persistent Differences in Corruption across Countries and Regions.
Working Paper (http://www.economics.smu.edu.sg/events/Paper/Sah2.pdf)
Sah, R., 2007. Corruption across Countries and Regions: Some Consequences of Local Osmosis.
Journal of Economic Dynamics & Control 31, 2573–2598.
Saha, S., Gounder, R., Su, J.-J., 2009. The Interaction Effect of Economic Freedom and Democracy
on Corruption: A Panel Cross-Country Analysis. Economics Letters 105, 173-176.
Shang, J., and Croson, R., 2005. Field Experiment in Charitable Contribution: The Impact of Social
Influence on the Voluntary Provision of Public Goods. unpublished manuscript, University of
Pennsylvania, Philadelphia.
Scheinkman, J. A., 2008. Social Interactions. in The New Palgrave Dictionary of Economics (2nd ed.),
S. Durlauf and L. Blume (eds.), Palgrave Macmillan.
Shleifer, A., Vishny, R.W., 1993. Corruption. Quarterly Journal of Economics 108, 599–617.
Sicular, T., Yue, X., Gustafsson, B. and Li, S., 2007. The Urban–Rural Income Gap and Inequality in
China. Review of Income and Wealth 53, 93 – 126.
Smarzynska, B.K. and Wei, S.-J., 2000, Corruption and the Composition of Foreign Direct Investment:
Firm-Level Evidence. World Bank Discussion Paper no. 2360.
Smith, A. 1759/1976. The Theory of Moral Sentiments. Oxford: Clarendon Press.
Sonin, K., 2003. Why the Rich May Prefer Poor Protection Of Property Rights. Journal of
Comparative Economics 31, 715-31.
Spichtig, M., and Traxler, C., 2007. Social Norms and the Evolution of Conditional Cooperation.
Discussion Papers in Economics 1398, University of Munich, Department of Economics.
Stutzer, A., 1999. Demokratieindizes für die Kantone der Schweiz. Working Paper No. 23. Institute
for Empirical Research in Economics, University of Zurich.
Sung, H., 2003. Fairer Sex or Fairer System? Gender and Corruption Revisited. Social Forces 82,
703-723.
Sung, H.-E., 2004. Democracy and Political Corruption: A Cross-National Comparison. Crime, Law
& Social Change 41, 179–194.

191 
 
 

Svensson, J., 2003. Who Must Pay Bribes and How Much? Evidence from a Cross Section of Firms.
Quarterly Journal of Economics 118, 207-230.
Swamy, A., Knack, S., Lee, Y. and Azfar, O., 2001. Gender and Corruption. Journal of Development
Economics 64, 25-55.
Tanzi, V., 1995. Corruption: Arm’s-Length Relationship and Markets, in: G. Fiorentini and S.
Peltzman (ed.), The Economics of Organized Crime. Cambridge: Cambridge University Press.
Tanzi, V., 2002. Corruption Around the World: Causes, Consequences, Scope, and Cures, in: G. T.
Abed and S. Gupta (eds.), Governance, Corruption and Economic Performance. Washington:
International Monetary Fund: 19-58.
Tanzi, V. and Davoodi, H., 1997., Corruption, Public Spending and Growth. IMF Working Paper,
WP/97/139.
Taylor, P.D., 1979. Evolutionarily Stable Strategies with Two Types of Player. Journal of Applied
Probability 16, 76-83.
Tirole, J., 1996. A Theory of Collective Reputation (with applications to the persistence of corruption
and to firm quality). Review of Economic Studies, 1-22.
Tittle, C., 1980. Sanctions and Social Deviance: The Question of Deterrence. New York: Praeger.
Tocqueville, A.de, 1835. Democracy in America.
http://ebooks.adelaide.edu.au/t/tocqueville/alexis/democracy/.
Torgler, B., 2005. Tax Morale and Direct Democracy. European Journal of Political Economy 21,
525-531.
Torgler, B., 2006. “The Importance of Faith: Tax Morale and Religiosity.” Journal of Economic
Behavior and Organization 61: 81-109.
Torgler, B., 2007. Tax Compliance and Tax Morale: A Theoretical and Empirical Analysis.
Cheltenham, UK: Edward Elgar.
Torgler, B., and Schneider, F., 2009. The Impact of Tax Morale and Institutional Quality on the
Shadow Economy. Journal of Economic Psychology 30, 228-245.
Torgler, B., and Valev, N.T., 2006. Corruption and Age. Journal of Bioeconomics 8, 133-145.
Torgler, B. and Valev, N. T. 2007. Public Attitudes Toward Corruption and Tax Evasion:
Investigating the Role of Gender Over Time. School of Economics and Finance Discussion Papers
and Working Papers Series 214, Queensland University of Technology
Traxler, C., 2006. Social Norms and Conditional Cooperative Taxpayers. Discussion Papers in
Economics 1202, Department of Economics, University of Munich.
Treisman, D., 2000. The Causes of Corruption: A Cross-National Study. Journal of Public Economics
76, 399-457.
Treisman, D., 2007. What Have We Learned About the Causes of Corruption From Ten Years of
Cross-National Empirical Research? Annual Review of Political Science 10, 211–244.

192 
 
 

van Rijckeghem, C. and Weder, B., 2001. Bureaucratic Corruption and the Rate of Temptation: Do
Wages in the Civil Service affect Corruption, and by How Much? Journal of Development
Economics 65, 307-331.
Veblen, T. 1899. The Theory of the Leisure Class. New York: MacMillan.
Verba, S., Schlozman, K. H. and Brady, H., 1995. Voice and Equality: Civic Voluntarism in
American Politics. Cambridge: Harvad University Press.
Wedeman, A., 2004. The Intensification of Corruption in China. The China Quarterly 180, 895-921.
Wei, S.-J., 2000a. How Taxing is Corruption on International Investors. Review of Economics and
Statistics 82, 1-11.
Wei, S.-J., 2000b. Local Corruption and Global Capital Flows. Brookings Papers on Economic
Activity, 2000(2), 303-354.
Weingast, B., 1995. The Economic Role of Political Institutions: Market-Preserving Federalism and
Economic Development. Journal of Law, Economics, and Organization 11, 1-31.
Welsch, H., 2004. Corruption, Growth, and the Environment: a Cross-Country Analysis.
Environment and Development Economics 9, 663–693.
White, G., 1996. Corruption and the Transition from Socialism in China. Journal of Law and society
23, 149-169.
Wintrobe, R., 2006. Rational Extremism: The Political Economy of Radicalism. Cambridge
University Press: Cambridge.
Wooldridge, J.M. (2002). Econometric Analysis of Cross Section and Panel Data. MIT Press:
Cambridge.
World Bank, 1997. Helping Countries Combat Corruption: The Role of the World Bank. Poverty
Reduction and Economic Management, World Bank.
World Bank, 2000. Anticorruption in Transition: A Contribution to the Policy Debate. World Bank
Publication. Washington D.C.: World Bank.
World Bank, 2006. Governance, Investment Climate and Harmonious Society. Washington D.C.:
World Bank.
Wu, Y., 2008. Fiscal Decentralization, Corruption and Governance. China Economic Quarterly 7,
1045-1060.
Xu, K.-N. and Wang, J., 2006. An Empirical Study of a Linkage between Natural Resource
Abundance and Economic Development. Economic Research Journal 2006(1), 78-89.
Yang, Z. and Ye, F., 1993. Studies on Semi-colonization of Qing Dynasty. Beijing: Higher Education
Press.
Yao, S., 2002. Privilege and Corruption: The Problems of China's Socialist Market Economy.
American Journal of Economics and Sociology 61, 279-299.
You, J.-S. and Khagram, S., 2005. Inequality and Corruption. American Sociological Review 70, 136-
157.

193 
 
 

Zager, M. A., 1994. Gender and Crime, in: T. Hirschi and M. R. Gottfredson (eds.), The Generality of
Deviance. New Brunswick: Transaction Publishers: 71-80.
Zanella, G., 2004. Social Interactions and Economic Behaviour. Working paper No. 411, Department
of Economics, University of Siena.
Zhang, T. and Zou, H.-F., 1998. Fiscal Decentralization, Public Spending, and Economic Growth in
China. Journal of Public Economics 67, 221-240.

194