Working PaperPDF Available

Insights from a multilevel analysis of corruption determinants in developing and transition countries

Authors:

Abstract and Figures

Because of shared norms of ethics, trust, and coordination prevailing in a given society or a social group, corrupt micro-level decisions are related to each other. In micro empirical analysis, this interdependence of corruption decisions can be addressed through the hierarchical modelling of corruption data. Exploiting a baseline sample of 34,358 bribe reports of firms from 71 developing and transition countries, I use a three-level estimation framework to re-examine the contribution of five major determinants of corruption emphasized by the literature: the economic and human development levels, the size of governments, trade openness and democracy. Multilevel estimations stress that the negative effect of income per capita on bribery is found to be mostly driven by improvement in human capital, more particularly by the decline in fertility rates. They also allow reconciling some contrasting findings of the literature on other corruption determinants. First, higher school attendance may increase corruption incidence by inducing larger public spending, but may also reduce the size of bribes by improving the scrutiny over public agents. Second, public intervention has a positive effect on bribery through public spending, but a negative effect through taxation. Third, trade openness has a positive effect on bribery, explained by state interventions on the one hand and by the country’s geographical distance from main world markets on the other hand. Fourth, results reveal a negative effect of political rights and press freedom on corruption, but also show that young democracies may experience higher corruption levels because of larger scope for private transactions. All in all, the significance of random coefficients across estimations, and the sensitivity of some coefficient estimates to their inclusion, point that the contribution of key corruption determinants emphasized in this study is nevertheless highly context-dependent.
Content may be subject to copyright.
Insights from a multi-level analysis of corruption
determinants in developing and transition
countries
Joël Cariolle, Research Officer at the Foundation for International Development Study and Research, Clermont-
Ferrand (Ferdi), France. Email: joel.cariolle@ferdi.fr
Tuesday, March 06, 2018
Abstract
Because of shared norms of ethics, trust, and coordination prevailing in a given society or a social group,
corrupt micro-level decisions are related to each other. In micro empirical analysis, this
interdependence of corruption decisions can be addressed through the hierarchical modelling of
corruption data. Exploiting a baseline sample of 34,358 bribe reports of firms from 71 developing and
transition countries, I use a three-level estimation framework to re-examine the contribution of five
major determinants of corruption emphasized by the literature: the economic and human development
levels, the size of governments, trade openness and democracy. Multilevel estimations stress that the
negative effect of income per capita on bribery is found to be mostly driven by improvement in human
capital, more particularly by the decline in fertility rates. They also allow reconciling some contrasting
findings of the literature on other corruption determinants. First, higher school attendance may
increase corruption incidence by inducing larger public spending, but may also reduce the size of bribes
by improving the scrutiny over public agents. Second, public intervention has a positive effect on bribery
through public spending, but a negative effect through taxation. Third, trade openness has a positive
effect on bribery, explained by state interventions on the one hand and by the country’s geographical
distance from main world markets on the other hand. Fourth, results reveal a negative effect of political
rights and press freedom on corruption, but also show that young democracies may experience higher
corruption levels because of larger scope for private transactions. All in all, the significance of random
coefficients across estimations, and the sensitivity of some coefficient estimates to their inclusion, point
that the contribution of key corruption determinants emphasized in this study is nevertheless highly
context-dependent.
* I am grateful to Olivier Cadot, Jean-Louis Combes, Vianney Dequiedt, Bernard Gauthier, Michaël
Goujon, Patrick Guillaumont, Frédéric Lesné, Sébastien Desbureaux, and Alexander Sarris for useful
discussions and comments. I also thank participants of the 2016 EPCS meeting in Freibourg (Germany),
and of the 8th Annual Joint Workshop in socio-Economics at Paris 1 University.
2
1. Introduction
The economic literature generally focuses on the demand side of corrupt transactions, arising from
public agents, and depicts it as the results of a tension between public agents’ own interest and the
general interest (Banfield, 1975). According to various studies, corruption is an individually-driven
phenomenon, resulting from a cost-benefit analysis made by public agents. By contrast, the literature
on the supply side of corruption depicts corrupt transactions as an expense or an investment made by
private agents to obtain undue preferential treatments from the public sector, thereby yielding
important private payoffs (Martin et al., 2007; Lambsdorff, 2002; Bhagwati, 1982). In this setting,
corruption is compatible with corporate objectives, and hence results from a tension between an
organization’s pecuniary objectives and the legal and social norms of ethics and integrity prevailing in a
society.
Moreover, social sciences analyse corruption as an informal institutional arrangement, contributing to
the stability and predictability of exchanges between economic agents, when formal arrangements are
dysfunctional (Andvig, 2006; Graeff, 2005; Blundo et al., 2006; Williamson, 2009). Therefore, corrupt
transactions are also determined by holistic factors that go beyond individual and organizational
motives. As a result, because of shared norms of ethics, trust, and coordination prevailing in a given
society or a social group, corrupt individual decisions may be related to each other.
In micro empirical analysis, this interdependence of corruption decisions may result into estimation
biases. This so-called “intra-class correlation problem can however be addressed through multi-level
modelling of corruption data. I therefore apply this empirical framework to a baseline sample of 34,358
bribe reports of firms from 71 developing countries in order to re-examine the contribution of key
corruption determinants: the economic and human development processes, the alleged oversize of the
State, trade openness and democracy.
In line with previous empirical researches, multi-level estimates confirm that economic development,
measured by income per capita, significantly and negatively contributes to bribe prevalence. However,
this negative effect of income per capita is found to be mostly driven by improvement in human capital,
more particularly by the decline in fertility rates. In fact, raising the fertility rate by one child per women
is found to almost double bribe prevalence in the baseline sample, which is equivalent to the effect of a
10% decline in GDP per capita on bribe payments. Regarding education, higher school attendance is
associated with smaller bribe payments but higher corruption incidence. This contrasting evidence on
the relationship between human capital and corruption is found to be partly mediated by the public
spending channel. One explanation is that more educated people require more public spending, and
hence larger rents in the economy, but induce a better monitoring of public action (Eicher et al, 2009).
In this regard, public intervention is found to have a positive effect on bribery through public spending,
but a negative through taxation. This result suggests that a larger government size may be both
associated with larger rents and better institutions. Larger state intervention is also found to mediate
part of the positive effect of trade openness on bribery, but this positive effect also lies in countries
inclination to natural openness. Last, results reveal a negative effect of political rights and press
freedom on corruption, but also show that young democracies may experience higher corruption levels
because of larger scope for private transactions. All in all, the significance of random coefficients across
estimations, and the sensitivity of coefficient estimates to their inclusion, point that the contributions of
key corruption determinants emphasized in this study are nevertheless highly context-dependent.
3
The next section reviews business and socio-economic studies highlighting the contextual origin of
corrupt transactions. The third section sets a multi-level model of corruption prevalence that takes into
account the contexts of corrupt transaction occurrence. In the fourth section, a three-level empirical
analysis of five key corruption determinants is undertaken and confronted to the literatures findings.
The fifth section concludes.
2. The contextual origin of corruption prevalence
The economic literature generally views corruption from the demand-side of corrupt transactions
(Martin et al, 2007). Defined by the World Bank as the abuse of public office for personal gains,
corruption is often confined to a consequence of State interventions aimed at correcting market failures
(Shleifer and Vishny, 1993; Banerjee, 1997; Acemoglu and Verdier, 2000). Standard economic
approaches therefore depict corruption as an individually-driven phenomenon, resulting from public
agents’ discretionary power over public resources allocation and the lack of accountability in the public
sector (Klitgaard, 1988). In this setting, corrupt transactions result from a cost-benefit trade-off, and
arise from the tension between public agent’s private interests and the general interest (Banfield,
1975).
On the other hand, business studies view corruption from the supply side of corrupt transactions. They
point that firms decisions to engage in corruption may fully match their objective of profit maximization
(Banfield, 1975), so that corrupt transactions result from the prevalence of profitability objectives over
organizational and societal value of ethics. This point is also highlighted by social science studies who
stress how holistic norms of honesty, integrity and trust in a society strongly frame individual corrupt
decisions (Blundo et al., 2006).
Therefore, by confining corrupt transactions to a demand-sided and individually-driven decision,
mainstream economic approach eludes the importance of the context of corrupt transactions, that is,
how constraints at the group level (the community, the industry, the country, and so on) affect
individual corruption incentives.
1
In this sub-section, I present a short review of studies stressing how
the context matters for the understanding of corruption prevalence.
2.1. Business corruption: a tension between firm profitability and societal values of ethics
The increasing role of private companies and other non-state actors in the delivery of essential goods
and services raises the concern for the complex issue of the supply side of corruption (Transparency
International, 2009a; Rose-Ackerman, 2007; Pope, 2000). In fact, addressing the supply side of
corruption is a difficult task because it is far less understood and documented than the demand side of
corruption, emerging mostly from the public sector (Kaufmann, 2005).
Studies in business ethics give some elements of understanding of the mechanisms underlying corrupt
behaviours in a private organisation. These studies consider private sector corruption as the result of a
tension between firm profitability and societal values of ethics (Martin et al., 2007; Nguyen and Cragg,
2012; Rose-Ackerman, 2007; Banfield, 1975). Results of experimental research on corrupt business
1
In general, these studies confine the question of ethics, social trust, or reciprocity to a “moral cost function”, without properly
addressing the complex relationships between them.
4
practices suggest that profitability may in certain circumstances prevail over morality, depending on
cultural orientations of societies (Husted, 1999). In fact, it has been suggested that hiring staff with high
ethical values does not necessarily lead to creating robust enough conditions to prevent business sector
corruption (Nguyen and Cragg, 2012; Rose-Ackerman, 2007; Vitell et al. 2000, Banfield, 1975). In
particular, corruption within an organisation can arise when norms of friendship and solidarity among
staff are antagonists to corporate objectives (Rose-Ackerman, 2007).
Thus, the study of business corruption highlights the existing tension between firms’ pecuniary
objectives and norms of ethics and honesty which prevail within a given society or group of economic
agents. The next sub-section addresses how social norms of corruption or honesty frame corrupt deals.
2.2. Corruption as result of social norms, reciprocity and trust
In socio-economic studies, a particular attention has been paid to the effects of social capital and its
manifestations such social norms of honesty, ethics, and trust on corrupt exchanges (Lambsdorff
and Frank, 2011; Graeff, 2005). In this regard, the importance of reciprocity as a key principle guiding
corrupt deals has been investigated (Lambsdorff and Frank, 2011; Graeff, 2005). According to Graeff
(2005), some social norms foster corruption by systematizing the reciprocity in corrupt deals. Those
‘corruption norms’ are defined by the author as
“the expectation that one can usually offer or accept a corrupt deal in a certain situation”.
2
By specifying how corrupt agents should behave in particular situations, corruption norms structure
corrupt exchanges, especially when agents and clients do not know each other, and so, may not trust
each other.
In some contexts, honest behaviours may even be socially discredited and considered as morally
reprehensible, when for instance a public agent does not take advantage of his position to enrich and
redistribute corruption proceeds among his kin or colleagues (Blundo et al., 2006). In certain countries
people are even recruited on the condition that they actively seek bribes and share the proceeds with
the individuals who recruited them (Kodi, 2008).
When such social norms of corruption do not fully operate, the level of trust between individuals can
ensure the reciprocity in corrupt exchanges (Graeff, 2005). A few economic analyses addressed the
structuring role of interpersonal trust, by emphasizing that frequent and prolonged interactions
between corrupt actors allow them to reinforce reciprocity in corrupt deals (Andvig and Moene, 1990).
Kingston (2007) stresses the role of informal shared activities between bribe-demanders and bribe-
payers, collateral to corrupt exchanges, to enforce corrupt arrangements. Corruption may therefore be
encouraged through various types of network membership such as kinship or friendship, age, ethnic
group, gender, social/religious cast or religion so that corruption may be persistent even in societies
with broad civic and ethical norms.
However, other studies consider that trust and social norms of honesty may be self-reinforcing features
of a society. Uslaner (2005) points out that trust when defined as a value expressing the belief that
2
Graeff, P. “Why should one trust in corruption ?” in eds. Lambsdorff, G, Taube, M. and M. Schramm, The New Institutional
Economics of Corruption, Chapter 3, Routledge, 2005, p.44.
5
others are part of your moral community”
3
discourages corrupt behaviours by enhancing the
coexistence of citizens and contributing to the respect of laws, values of ethics and morality. Letki
(2006) nuances slightly this assertion. She distinguishes interpersonal trust from citizen’s trust in public
institutions, and finds that the mere trust in institutional actors such as the police, the justice, the
parliament, or the army contributes to people’s adhesion to moral and civic values. In line with this
argument, Fisman and Miguel (2007, 2008)’s analysis of UN diplomats’ inclination to break the law
suggests that corruption prevalence in a given country depends on people’s “sentiments towards their
own country’s laws”
4
. Thus, according to these studies, the frontier between trust and adherence to
“corrupt” or “ethical” social norms is porous.
Socio-economic studies stress that this frontier is strongly shaped by the organisation of exchanges
within societies. Following Max Weber’s theory of modernization, Andvig (2006) depicts corrupt
societies as dynamic hybrid systems where emerging and ancient coordination modes confront each
other. From this overlap of coordination modes may emerge persistent inefficient states, characterized
by state capture by private interest and/or patrimonial forms of corruption. In his framework, systemic
corruption results from the confrontation between older illegal but legitimate and newer legal but
illegitimate norms of coordination:
- patrimonial corruption stems from the persistence of family/friendship transactions while
political, bureaucratic or commercial transactions should be the norm;
- commercial corruption stems from the persistence of family/friendship transactions or
political/hierarchical transactions while market transactions should be the norm;
- and the phenomenon of state capture arises from the illegitimate intrusion of market-based or
kinship/friendship transactions in the area of political transactions.
Therefore, while economic studies insist on various cross-country features that give public agents’
incentives to engage in corruption, business and socio-economic studies reframe corrupt transactions
within their context of occurrence. In the next sub-section, I explain why a multi-level empirical
framework is particularly suited for the contextual analysis of cross-country determinants of corruption
prevalence.
3. A multi-level empirical framework for the analysis of corruption prevalence
3.1. A multilevel estimation framework
Multi-level or hierarchical models depict a hierarchical system in which units of observations nested
within groups, and groups nested within higher-level groups (Hox, 2010). These models, widely applied
to human capital analysis (Hitt et al., 2007), exploit the hierarchical structure of micro datasets to relax
the hypothesis of independence of observations within different levels of the data. The relevance of
these models for the analysis of corrupt deals is that, because of shared social norms of ethics, trust,
and coordination modes prevailing in a given society or a social group, corrupt individual decisions may
3
Uslaner, E.M. “Trust and corruption” in eds. Lambsdorff, G, Taube, M. and M. Schramm, The New Institutional Economics of
Corruption, Chapter 5, Routledge, 2005, p.76.
4
Fisman, R., and E. Miguel “Nature or Nurture? Understanding the Culture of Corruption” in eds. Fisman, R., and E. Miguel,
Economic Gangsters, Chapter 4, Princeton University Press, 2008, p.100.
6
be correlated with each other. Applying a multi-level framework for the analysis of bribery determinants
seems therefore particularly suitable, by allowing a contextualization of economic agents’ decisions
within the different groups to which they belong.
A multilevel corruption model is useful when there is one dependent corruption variable measured at
the lowest level, e.g. the firm or individual experience or perception of corruption offences, explained
by various corruption determinants measured at all levels of the data. The three-level “country-sector-
firm” model adopted in this paper is a hierarchical system of equations where the dependent variable,
the firm k’s experience of bribery (Yi,j,k), is explained by firms’ characteristics (Hi,j,k) (e.g. large or small
firm), the sector j’s characteristics (Zi,j) (e.g. sector’s tax pressure) and the country i’s characteristics (Xi)
(e.g. government’s public spending). In this multi-level estimation framework, we would regress:
            (1)
With  a country-sector intercept, ,  and are country-sector regression coefficients, and
 a usual zero-mean and constant-variance error term. Therefore, contrary to usual single-level
models, a three-level model assumes that each sector, in each country, is characterized by a different
intercept and different slope coefficients. In other words, while the single-level model assumes that
observations are independent across levels of the data, multilevel models relax this hypothesis.
By setting random coefficients in the model, this three-level model posits that, in our example, the
relationship between the firm’s size and bribery ( ), the relationship between the sector-level tax
pressure and firm bribery (, or the relationship between public spending () and firm bribery
could change according to sector or country’s characteristics. In regards to the random intercept
interpretation, a higher value means that specific characteristics in sector j and country i are associated
with higher corruption prevalence. To sum up, in this three-level model of corruption prevalence, the
intercept and slope coefficients are allowed to vary randomly across countries and sectors, and by this
way, allows the various contexts corruption prevalence to be accounted for.
The intercept and regression coefficients can therefore be expressed as a function of lower-level
random coefficients. For the sake of simplicity, and because this paper reviews the contribution of
traditional country-level determinants of corruption through a multi-level approach, we apply a random
slope model for Xi only, and exclude the possibility of cross-level interactions between random
coefficient and the factor variable Xi
5
, so that equation (1) can be re-written as:
            (2)
In this model, the random intercept  can be expressed as follows:
    with  a zero-mean constant-variance country-level error term (2.1)
    with  a zero-mean constant-variance sector-level error term (2.1.1)
Which gives:
5
They could also be a function of relevant explanatory variables, but the resulting cross-level interactions are
complex to interpret, and this complexification of the model can lead to problems of convergence during
estimations.
7
       (2.1.2)
And the random-slope can be expressed as:
    with  a zero-mean constant-variance country-level error term. (2.2)
    with  a zero-mean constant-variance sector-level error term. (2.2.1)
Which gives:
       (2.2.2)
By subtitling equations (2.1.2) and (2.2.2) into equation (2), we get:
                  (3.1)
With      a zero-mean constant-variance country-level error term.
The segment          in equation contains the fixed coefficient, while the
segment         contains the random coefficients. Equation (3.1) can be rearranged
as follows:
                    (3.2)
With =     the random intercept, and =     is the random slope.
As results, this three-level model relaxes the hypothesis of independence between observations at the
country and sector levels, and by this way, controls for intra-class correlation between reported-
corruption offences that bias estimations.
3.2. The data
In this review of macro-level determinants of firm-level bribery, multi-level estimations are conducted
exploiting self-assessments of firms’ experience of bribery in conducting business drawn from the World
Bank Enterprise Surveys (WBES). Compared to indicators based on experts’ assessment of corruption,
this survey data has the advantage of being conceptually more precise and less-exposed to biases than
the former (Razafindrakoto and Roubaud, 2010; Hallward-Driemeier and Pritchett, 2015).
Two dependent variables of the WBES reflecting corruption prevalence among firms are used. The first
dependent variable is the size of informal payments reported by firms, expressed as a share of their
total sales. This dependent variable is bi-dimensional since an increase in this variable can be both
induced by an increase in the incidence and/or an increase in the size of bribes. However, it has been
contended that for different reasons respondents may under-report or over-report bribe amounts
(Clarke, 2011). One way to circumvent this problem of under/over-reporting is to focus on bribe
incidence by computing a second dependent variable equal to one if it has reported an informal
payment and zero if it has reported no informal payment. I therefore used these two complementary
dependent variables the bribe-payment variable and the bribery-incidence variable in this empirical
analysis.
8
Moreover, the WBES dataset allows exploiting information on firms’ sector of activity and controlling for
a range of firms’ characteristics that are expected to affect their inclination to engage in corruption.
Building on studies on the determinants of firm-level corruption (Svensson, 2003; Hellman et al., 2003;
Fisman and Svensson, 2007; Dabla-Norris and Gradstein, 2008; Diaby and Sylwester, 2015), I control for
the logarithm of firms’ total annual sales, for their share of direct and indirect exports in total sales, for
their size (using dummy variables for medium-size and large-size firms), for the share of public
ownership, for their share of working capital funded by internal funds, their share of working capital
funded by public and private commercial banks, and their sector of activity (using sector dummies).
Macroeconomic determinants of bribe prevalence are therefore tested, controlling for this set of firms’
characteristics. These determinants consist of the economic development process, the human
development process, the size of the State, trade and democracy.
6
Each corruption determinant, its
corresponding proxy, and data sources are presented in table 1 below. Descriptive statistics are
presented in Appendix A.
Table 1. Corruption determinants and data sources.
Corruption determinant
Proxy
Data sources
Economic development
GDP per capita
World Development Indicators (WDI)
Human Development
1. Fertility rate
2. Primary enrolment ratio
UNESCO7
State size
1. Public spending
2. Public expenditure on education
3. Tax revenue
1. International Monetary Fund (IMF)
2. WDI
3. IMF
Trade
1. Trade intensity (% of exports +
imports in GDP)
2. Natural openness determinants:
a. Remoteness from world
markets
b. Log of population
1. WDI
2. a. FERDI
b. WDI
Democracy
1. a. Freedom of the press index
b. Civil liberty index
c. Political rights index
2. Durability of the polity
1. Freedom House
2. Polity IV
3.3. Empirical specification
Pooled three-level estimations of the following baseline econometric model are conducted:
      
    (4)
6
Mechanisms underpinning the effect of these determinants are discussed in section 4.
7
Drawn from Teorell et al. (2015)’s database, University of Gothenburg.
9
Subscripts i j, k refer to countries, sector and firms respectively. Bribei,j,k is the variable of bribe
prevalence, Xi the vector of country-level corruption determinants, Yi,j,k the vector of micro-level
controls, dj the sector dummies and εi,j,k a zero-mean constant-variance error term.
Three-level country-sector-firm maximum-likelihood estimations of equation (4) are conducted. Only
model calibrations with significant random components are kept. Moreover, when the binary variable of
corruption incidence is used as dependent variable, I perform a probabilistic linear multi-level modelling
of equation (2), in order to avoid convergence problems (Caudill, 1988).
8
If the sign and significance of
the resulting estimated coefficients is informative, I do not dare interpreting their strength.
3.4. Endogeneity issue: the problem of intra-class correlation
There are various reasons to expect that multi-level estimates of country-level determinants of
corruption reflect their causal effects on firm bribery. Recent analyses on the effect of macro-level
explanatory variables on micro-level decisions argue that a transaction undertaken by a single firm
should have no macro-level effects (Héricourt and Poncet, 2015; Paunov and Rollo, 2015; Farla, 2014).
In such a framework, the issue of reverse causality bias from corruption towards macro-level corruption
determinants should therefore be mitigated by the use of firm-level corruption data.
However, some studies have shown that micro-level interdependencies between firms are such that a
micro phenomenon may have aggregate consequences (Gabaix, 2011; Acemoglu et al., 2012; Chaney,
2014). This concern is heightened by the fact that corrupt transactions may be contagious within a
group of firms (Andvig and Moene, 1990). As a result, one corrupt transaction could have macro-level
consequences, and this issue of intra-class correlation could be a source of reverse causality.
The advantage of the three-level estimation framework set earlier is that it controls for intra-class
correlation (Hox, 2010) in so far as intra-class correlation is correctly modeled, i.e. this contagion effect
plays at the sector and/or the country-levels. Moreover, the relative small size of bribe payments in the
sample amounting to 1.36% of total sales in average (see Appendix A) limits the magnitude of this
contagion effect. Last, because secrecy over corrupt deals may prevail at different layers of the society,
multi-level analysis theoretically takes into account sector-level and country-level unobserved factors
influencing firms inclination to report bribe payments and to under or over-report bribe payments, and
therefore should also tackle problems of misreporting (Clark, 2011).
9
For all these reasons, I assume
estimates of country-level corruption determinants, hereafter emphasized, to be unaffected by reverse
causality and measurement error.
4. Three-level empirical analysis
This empirical analysis starts by questioning the negative contribution of the economic development
process to corruption prevalence. While income per capita is usually found to explain a significant part
of cross-country differences in corruption prevalence, this relationship is not clear-cut and hides
8
Due to the presence of dummy variables in our model inducing convergence problem, when applying a nonlinear mixed effect
model (logit or probit).
9
Note that problems of misreporting are also addressed by the use of a binary dependent variable of bribery incidence,
hereafter explained.
10
complex and sometimes conflicting mechanisms. Then, four additional mechanisms that could underlie
the development-corruption nexus are discussed and tested within a three-level empirical framework:
human capital, state interventions, openness, and democracy.
4.1. Is economic development detrimental to corruption?
The development process is considered as a major determinant of corruption prevalence in the
empirical literature. It is commonly argued that wealthier countries undergo lesser corruption, because
corruption decreases with improved life standards, on the one hand; and because of the many
institutional, sociological, and demographic changes which usually accompany the development process
on the other hand (Treisman, 2000).
However, putting aside the endogenous nature of the relationship, the development-corruption nexus
is a catch-all phenomenon reflecting various and sometimes antagonist mechanisms. In fact, higher
income leads to a range of improvements in human capital, public resources managements, the rule of
law, and so on, which are expected to drag down corruption levels, but these socio-economic
transformations may also create new grounds for corrupt transactions (Andvig, 2006). From these two
competing arguments, it is possible to derive the following opposite hypotheses:
H1: Corruption will be lower in more economically developed countries, when populations are less needy,
more educated, when the management of public resources is efficient and when institutions are better.
H1’: Corruption will grow with economic development, when new opportunities to enrich are not framed
within the rule of law, when public funds are allocated with discretion, and when citizens cannot properly
monitor public decision-making.
In order to have a proper assessment of the effect of GDP per capita on corruption prevalence, three-
level estimations using WBES data on firms’ bribe reports are run and compared to single-level
estimations. Results of valid multi-level models
10
are presented in table 2. When the bribe payment (BP)
dependent variable (continuous) is used, three-level estimations are compared to OLS estimations.
When the bribery incidence (BI) variable (binary) is used, three-level estimations (linear probability
model) are compared to Logit estimations. In a first step, corruption dependent variables are regressed
over firms’ characteristics, the GDP per capita variable being included in a second step.
First, random intercepts are all found to significantly vary across countries and sectors, even though
intra-class correlation coefficients
11
- which inform on the proportion of variance explained at the
country and sector levels indicate that most of the variance is explained at the firm-level. However,
intra-class correlation is substantial when the bribery incidence variable is used, as around 15%-20% of
the variance is explained by within-country variations. Second, estimations stress the negative
contribution of firms’ total sales and the positive contribution of their indirect exports to bribery. Access
10
Models with significant random components.
11
The country-level and sector level intra-class correlation coefficients are calculated according to Siddiqui et al. (1996)’s
method:
 




, and  



With 
 

the variance estimates of the country-level random intercept, the sector-level random intercept, and the
residual, respectively. Another interpretation of these coefficient is that they reflect the correlation between two randomly
chosen firms in a given country or in given sector.
11
to external finance and internal funding are both found to deter firms’ bribery, while the share of public
ownership is found to reduce corruption incidence.
12
Third, the last two sets of regressions highlight the significant negative effect of income per capita on
bribery. Single and multi-level estimates show that increased GDP per capita reduces both the average
amount and the incidence of bribery in a sample of 71 developing countries. According to the third set
of estimations, a 10% increase in the average GDP per capita results into a 0.67 percentage point
decrease in the size informal payments (see figure 3), which is substantial as the country average
informal payments lies around 1.3% of firms’ total sales (see Appendix A).
To conclude this sub-section, economic development is negatively and strongly related to corruption
prevalence, but this evidence does not tell much on its transmission channels. In what follow, I test
whether the human capital is an important channel of the relationship between income and corruption,
by emphasizing the role of fertility and education in curbing corruption.
Figure 3. GDP per capita and bribe payments
12
This latter finding is consistent with the findings of Hellman et al. (2003), who show that public firms resort to influence
rather than bribery to obtain political favours.
13
4.2. Human capital and corruption
The New Economic Growth Theory considers long-run growth as being endogenously determined by
demographic factors (Brezis and Young, 2014). In parallel, it has been stressed that demographic
change is closely related to human capital evolution (Li et al., 2013; Becker and Tomes, 1994), and that
a healthy and educated population is necessary for the well-functioning of institutions (Acemoglu et
al., 2001; Glaeser et al., 2004; Svensson, 2005). To check whether the wealth-corruption nexus
previously evidenced relies on the human development process, I run a second series of regressions
including demographic and education variables
13
together with the GDP per capita.
In a first step, the total fertility rate is used as a general proxy for human development, as changes in
fertility are narrowly associated with the development process, more specifically with socio-economic
changes that accompany the demographic transition. High fertility is indeed correlated with low
access to health services and infrastructures, low educational attainment, a large and youth
population, low wages, low productivity and low saving rates (Varvarigos and Arsenis, 2015; Kuznets,
1960; Becker, 1960). As a result, high fertility rates may be a key determinant of corruption
prevalence, by i) reducing quantities of public goods and services per capita, which may tempt citizens
to bribe public agents to “jump the queue” (Banerjee, 1997; Fisman and Gatti, 2002); and by ii)
reducing the empowerment of citizens, and thereby, their ability to monitor public decision-makers.
Therefore, the following hypothesis can be derived from these arguments.
H2: corruption will be higher in countries with large and low-skilled population, and will therefore
increase with fertility rates.
A fertility rate variable is therefore introduced in the corruption equation (3) together with the GDP
per capita. Three-level estimations using the bribe payment (BP) and the bribe incidence (BI) variables
are reported in the first two sets of regressions in table 3. In each set, I estimate the model with i)
country/sector-level random intercepts only (the random intercept model), and ii) with
country/sector-level random intercepts and random slope(s) (the random slope model). But only
models with significant random parameters are reported.
The first two sets of regressions evidence a strong and significant positive effect of fertility rates on
bribe payments and bribery incidence. Estimates with the BP variable reported in the first set of
regressions show that the GDP per capita has no longer significant effect once the fertility variable is
introduced; thereby suggesting that human capital is a structural determinant of corruption and a
critical channel of the GDP per capita-corruption nexus. Moreover, adding a random slope to the
fertility parameter makes sense at the country-level with the BP and BI dependent variables, and at
the sector-level with the BP dependent variable.
In a second stage, to check whether the effect of fertility passes through citizens’ empowerment, the
specification is refined by adding a variable of basic educational attainment: the gross enrolment rate
in primary. A more educated population should indeed allow a better monitoring of policies and
rulers, but this effect may be ambiguous as more educated people may be correlated with increased
rents in an economy (Eicher et al, 2009). The hypothesis testing is detailed below.
13
drawn from the UNESCO Institute for Statistics. See descriptive statistics in Appendix A.
14
H3: Corruption will be lower in countries with higher educational attainment, because a more educated
population allows a better monitoring of public decision-making.
H3’: Corruption will be higher in countries with higher educational attainment, because a more
educated population leads to the creation of new rents in the economy.
The last two regressions in table 3 test the education channel alone. Fertility and education variables
are then introduced together with the GDP per capita. Estimates are reported in table 4 (columns (1),
(2), (4), (5)). Once controlling for schooling and including a country-level random slope, higher fertility
rate is found to increase bribe payments but to reduce bribery incidence. Therefore, including a
random component in the fertility coefficient slope reverses the sign of its estimated effect on BI,
suggesting that the direct effect of fertility on BI is strongly affected by country-level unobserved
heterogeneity. Moreover, estimates in column (5) of table 4 supports that the positive effect of
fertility on BI evidenced in table 3 (column (6)) is mainly channelled through educational attainment.
Regarding educational attainment, this primary enrolment rate is found to reduce bribe payments but
to increase bribery incidence (tables 3 and 4). Perhaps this variable probably provides a quantitative
rather than qualitative assessment of educational attainment, and the positive effect of schooling on
bribe incidence may result from the increase in rents induced by large pupil inflows in public schools.
To examine this possibility, I introduce in the corruption equation a policy variable measuring the
share of public expenditures on education in GDP.
14
Estimates of valid models are presented in
columns (3) and (6) of table 4. Once public expenditures on education are taken into account, the
positive effect of primary enrolment and the negative effect of fertility on bribery incidence disappear
(column (6)). An interesting fact is that the contrasting effect of primary enrolment on bribery
evidenced in columns (2) and (5) is now reflected in the contrasting effect of public expenditures on
education. This finding therefore highlights the ambiguous effect of schooling on corruption
prevalence: while an educated population may improve the monitoring of public officials and thereby
reduce the amounts of the latter may extract form their rents (H3), it may increase the number of
corrupt transactions by increasing the size of public spending (H’3).
To sum up, the relationship between the development process and corruption is not as
straightforward as surmised. Contrary to the conventional wisdom, the growth process may create
new opportunities for corrupt transactions at early stages of development, when citizens are not
sufficiently empowered to scrutinize public actions, and when governments may be overwhelmed by
the increasing demand for public goods and services. In this regard, previous results pointed out the
role of increased public spending in channelling the positive effect of human development on
corruption levels. The next section therefore further addresses the relationship between state
interventions and corruption.
14
Thereby eluding the question of rent creation in the private education sector.
15
Table 3. Human Capital and bribery (1)
Dep. Var.
Bribe payments (BP)
Bribe incidence (BI)
BP
BI
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
GDP per capita
-0.00003
(0.00007)
-0.00001
(0.00004)
-0.00003
(0.00006)
0.00001
(0.00004)
-0.00002**
(0.00001)
-0.00001
(0.00001)
-0.0003***
(0.0001)
-0.00003***
(0.0000)
Fertility rate
0.673***
(0.132)
0.697***
(0.137)
0.652***
(0.131)
0.681***
(0.138)
0.057***
(0.021)
0.067***
(0.028)
1ary enrollment
ratio
-0.013
(0.012)
0.005***
(0.002)
Dummies
Firms sizes & sectors
Country-level random effect parameters
Intercept
0.786***
0.000
0.744***
0.000
0.028***
0.002***
1.086***
0.044***
Slope fertility
0.062***
0.061***
0.020**
Sector-level random effect parameters
Intercept
0.000
0.000
0.000
0.000
0.001***
0.001***
0.008***
0.001***
Slope fertility
0.007***
0.007*
R2 / Wald Stat
166.5***
154.7***
157.8***
146.9***
143.2***
130.6***
137.4***
143.7***
LR Chi2
342.4***
354.9***
348.8***
360.7***
2935.8***
2946.2***
434.0***
2601.9***
#Countries (#obs)
40(18.052)
Firm controls not reported. Standard errors in parenthesis. *significant at 10%; **significant at 5%; ***significant at 1%.
Table 4. Human Capital and bribery (2)
Dep. Var.
Bribe payments
Bribe incidence
(1)
(2)
(3)
(4)
(5)
(6)
GDP per capita
-0.00005
(0.0001)
-0.0000
(0.0001)
-0.0000
(0.0001)
-0.00001
(0.0001)
-0.000
(0.0001)
-0.000
(0.0001)
Fertility rate
0.668***
(0.126)
0.681***
(0.130)
0.675***
(0.126)
0.005***
(0.002)
-0.758***
(0.328)
-0.071
(0.160)
Primary enrolment ratio
-0.015
(0.011)
-0.017*
(0.010)
-0.013
(0.010)
0.059***
(0.024)
0.044***
(0.015)
0.003
(0.011)
Public exp. education
-0.246***
(0.075)
0.212**
(0.93)
Dummies
Firms sizes & sectors
Country-level random effects
Intercept
0.691***
0.000
0.000
0.038***
8.503***
0.000
Slope fertility
0.053***
0.049***
1.126***
0.159***
Slope pub. exp. edu.
0.065***
Sector-level random effects
Intercept
0.000
0.000
0.000
0.001***
0.001***
0.001***
Slope fertility
0.007**
0.006**
R2 / Wald Stat
163.9***
152.7***
165.2***
149.3***
116.7***
109.5***
LR Chi2
269.4***
301.4***
311.5***
2449.4***
2761.7***
2841.4***
#Countries (#Firms)
40(18.052)
Firm controls not reported. Standard errors in parenthesis. *significant at 10%; **significant at 5%; ***significant at 1%.
16
4.3. Are larger states more corrupt?
Over the last decades, state interventions in the economy and the oversize of the public sector have
been pinpointed as being a major source of corruption. By contrast, the expansion of market-based
transactions through increased domestic and foreign competition i.e. deregulation, and the
privatization of state-owned enterprises, and trade openness has been viewed as a lever for
dragging firms’ profits down and therefore discouraging bribe payments (Shleifer and Vishny, 1993;
Lambsdorff, 2005; Sandholtz and Koetzle, 2000; Ades and Di Tella, 1999).
In fact, large public sector size is often depicted as a burden inciting private and public agents to
exploit it or to get rid of it through malpractices. Notably, red tape
15
may foster the bribery of
bureaucrats to ‘get things done’ or ‘to make things going faster’ (Guriev, 2004; Aidt, 2003; Lambsdorff,
2002; Tanzi, 1998). Moreover, increased public expenditure may enlarge the scope of public resources
under the discretion of public agents charged with their allocation; while higher tax rates may raise
the amount of bribes asked or offered for tax exemption and evasion (Gauthier and Goyette, 2014;
Gauthier and Reinikka, 2006; La Porta et al., 1999; Tanzi, 1998). For all these reasons, an increased
scope for state interventions may increase the size of the “corruption pie”.
However, several arguments in support to a deterrent effect of public action on corruption can be
invoked. Indeed, it has been contended that an increased state intervention often tracks the long run
growth process (Peacock and Scott, 2000), generally accompanies the openness of economies (Rodrik,
1998), and sometimes results from improved economic and democratic institutions (Rodrik, 2000).
Moreover, red tape does not systematically create opportunities for corruption since it may also be
associated with better screening and higher internal administrative controls (Wilson, 1989). The effect
of taxation has also to be nuanced since higher tax revenue may result from higher tax rates, but also
from a larger tax base or a better firm tax compliance (Hibbs and Piculesco, 2010). Last but not least, it
is argued that the growth of the private sector, illustrated by the wave of privatizations in the 90’s has
increased the supply of corrupt transactions (Transparency International, 2009a; Rose-Ackerman,
2007; Pope, 2000). It is therefore possible to derive two competing hypotheses over the state size-
corruption nexus.
H4: Corruption will be higher in countries with greater state interventions, if these interventions result
into stronger monopoly and discretionary powers over public rents.
H4’: Corruption will be lower in countries with greater state interventions, if these interventions result
into efficient public goods and service delivery and effective regulation of market-based transactions.
To test H4 against H’4, the share of public expenditure in GDP is used as a first proxy for public sector
size, and is introduced alongside the GDP per capita in the corruption equation. Estimates of valid
models are reported in the first two sets of estimates in table 5 (columns (1) to (7)). Results show that
there is no evidence of a significant effect of public expenditure on bribe payments. A positive
significant effect of public spending on bribe incidence (second set of regressions) is found with the
random intercept (column (5)) and the sector random-slope models (column (7)), but is no longer
significant with the country random-slope model.
15
red tape refers to excessive and/or poorly-designed bureaucratic rules that imply non-pecuniary costs for agents dealing
with bureaucracy (Banerjee, 1997).
17
Moreover, because i) public spending may affect corruption through increased tax burden, and ii)
public spending and taxation may have distinct effects on corruption prevalence, I use as a second
proxy for public sector size: the share of revenue collected from goods and services (G&S) taxation in
GDP.
16
This variable is tested alone with GDP per capita (columns (8) to (11)), and then tested jointly
with the public expenditure variable (columns (12) and (13)). Random intercept models (columns (8)
and (10)) support a negative effect of G&S taxation on corruption, suggesting that higher tax revenue
results in lower bribe prevalence (Hibbs and Piculescu, 2010). However, including a country-level
random slop component neutralizes this negative effect of tax on bribe payments (column (9)), and
even reverses it into a significant and positive effect on bribery incidence variable (column (10)). This
uncertainty over the coefficient sign is finally lift up when public expenditure and tax revenue variables
are introduced together in the random slop models (columns (12) and (13)). Each variable is found to
have a separate effect on corruption outcomes: while a 10% increase in tax revenue results into a 3%
decrease in the average bribe payment, a 10% increase in public expenditure is found to raise the
average bribe payment by 1%.
In other words, the positive effect of increased tax revenue on bribe incidence evidenced in column
(11) seems channelled through the effect of increased public spending on rent-seeking and
corruption, highlighted in the literature. Once this latter effect is taken into account, the net effect of
tax revenue on bribery is negative and may reflect the positive effect of the quality of tax policies and
tax administrations on firm integrity or public spirit.
Despite this contrasting evidence on the effect of public sector “oversize” on corruption prevalence,
principles of competition that guided international institutions’ agenda towards a lower scope for
public interventions were the same that motivated policies supporting increased openness of
domestic markets. The effect of trade openness on bribery is therefore discussed and analysed in the
next sub-section.
4.4. Are opened States less corrupt?
In the same way as excessive taxation has been depicted as a source of corruption by increasing the
amount of bribes required for tax exemption (La Porta et al., 1999; Tanzi, 1998), higher trade barriers
may create opportunities for politicians and custom officers to extort money or to sell favourable
treatments to domestic and international private companies (Dutt and Traca, 2010; Dutt, 2009; Gatti,
2004; Hellman, et al., 2003). Therefore, trade openness is expected to reduce the number and the size
of corrupt transactions through lowered trade barriers and increased foreign competition. Moreover,
Wei (2000) stressed that “natural openness”, i.e. trade openness determined by structural factors
such as country size and geography, is associated stronger institutional safeguards against corruption.
The author argues that countries’ natural inclination to trade incites governments to invest in
institutions protecting foreign investors and traders from corrupt practices. Therefore both structural
and policy-induced trade openness is likely to be detrimental to corruption.
However, Knack and Asfar (2003) have shown that the negative effect of trade on corruption is due to
a selection bias in the country-coverage of corruption perception indices, which tend to under-
represent poorly-governed small nations. Moreover, over the last decades, the worldwide
16
Drawn from the IMF database.
18
privatization of public services along with the removal of barriers to trade and financial flows created a
fertile ground for the internationalization of private corrupt practices, especially towards countries
with weak and/or non-democratic institutions (TI, 2009; Nellis, 2009; Hellman et al., 2003). Bribery is
indeed a common and widespread mean for international companies to win contracts abroad, to
avoid regulations, or to unduly influence policy-making, especially in the developing and transition
world (Hellman et al., 2003). Therefore, it is possible to derive from these arguments two competing
hypotheses on the trade openness-corruption relationship.
H5: Corruption will be lower in opened economies, since lower trade barriers, foreign competition, and
larger natural openness are detrimental to corruption.
H5’: Corruption will higher in opened economies, since trade openness exposes countries to imported
foreign corrupt practices.
Table 6 reports three-level estimates of the effect of trade openness and bribe payments and
incidence. First, a variable of trade intensity the ratio of export plus imports on GDP is included
together with the GDP per capita in the corruption equation. A significant and positive effect on
bribery incidence is evidenced, but no significant effect on the size of bribe payments. Second, public
expenditures and tax revenue variables are included as control to test whether the effect of trade
openness on bribery depends on the size of governments. As pointed out by Rodrik (1992), the
positive effect of trade openness on economic performance is uncertain, and its deterrent effects on
corruption prevalence may depend on the extent of state interventions (Rodrik, 1998), so that
previous estimations may suffer from omitted variable bias. Results in columns (3) and (6) put in
evidence a direct positive effect of trade intensity the size of bribe payments and an indirect effect
passing through the size of public interventions. Last, to check whether the effect of natural or
structural openness (Wei, 2000) could interfere with the sign and significance of estimated
relationships, an index of remoteness from world markets
17
and a proxy for country size (the logarithm
of the population) are added into the corruption equation. Results are reported in columns (7) and (8).
Including these variables in the regression highlights the positive and significant contribution of
remoteness from world markets to bribe prevalence, and turns the effect of the trade intensity on
bribe payments insignificant. Therefore, these estimations suggest that i) trade openness has a direct
effect on bribery relying on the countrys remoteness from world markets and independent from state
interventions, and that ii) trade openness has an indirect effect on bribery, mediated by the size of
state interventions.
These results are consistent with Rodrik (1998)’s findings that successful experiences of integration
into international trade are often accompanied by larger state interventions. They also support the
findings of Wei (2000) according to whom remote countries face structural handicaps that preclude
them from building good institutions. Moreover, these authors pointed out the importance of
democratic institutions to make state interventions efficient and to cushion the economic turmoil
resulting from international trading. The next section examines the link between democracy and
corruption prevalence.
17
Remoteness is an index between 0 and 100. It is the trade-weighted average distance from the nearest countries to reach
50% of the world market, and adjusted for landlockness. This index is used by the UN-DESA for the calculation of the
Economic Vulnerability Index and its methodology is presented here: http://byind.ferdi.fr/en/indicator/evi/build
21
4.5. Are democratic institutions detrimental to corruption?
The lack of accountability and transparency in public decision-making is a feature shared by corrupt
and non-democratic countries. By supporting effective economic regulations and administrative rules,
transparent procedures, law-enforcement institutions and strong watchdog and oversight bodies
18
,
democracy represents a strong corruption deterrent. Democratic institutions indeed reduce
opportunities for corrupt transactions, by allowing an improved scrutiny of voters upon political
decisions, fostering political competition and supporting the freedom of media (Lambsdorff, 2002;
Treisman, 2000, 2007; Sandholtz and Koetzle, 2000; Bhattacharyya and Hodler, 2010, 2015).
According to Sandholtz and Koetzle (2000), this virtuous effect of democracy on governance depends
on how well institutionally established norms of democracy are. This idea is also supported by
Treisman (2000, 2007), who shows that only democracy older than 40 consecutive years are
significantly associated with lower corruption levels.
The relationship between democracy and corruption may however be unstable at intermediary levels
of democracy, as pointed out by Treisman (2007, p. 228): “Perceived corruption always decreases as
democracy increases from 3 to 1 on the FH scale or as authoritarianism softens from 7 to 6, but the
effects of movements between 6 and 3 are more erratic.” Therefore, based on this literature, the
following conflicting hypotheses on the effect of democracy on corruption can therefore be made:
H6: Corruption will be lower in well-entrenched democratic countries.
H6’: Corruption will be higher in countries experiencing a transition towards democracy.
As a first evidence on the effect of democracy on corruption, I separate the sample between
democratic and less democratic countries, and test H4 against H4’, and H5 against H5’ in each sub-
sample. Democratic and less democratic countries are identified according to Freedom House (FH)’s
democracy’s status
19
, based on three indices reflecting three dimensions of modern democracies: the
extent of civil liberties (CL), of political rights (PR), and the freedom of the press (FotP). Using the
combined CL and PR country status on the one hand, and the FotP status on the other hand, countries
are split between free countries and less free countries (i.e. countries with partly-free and non-free
status). The effects of openness and state interventions are then re-estimated using these separate
sub-samples. Results are reported in table 7.
First, estimations show that, in PR-CL free countries (columns (1) and (2)), the effect of public
spending and taxation is no longer significant, trade intensity significantly increases bribe payments,
and remoteness fosters bribery incidence. By contrast, in PR-CL less-free countries (columns (3) and
(4)), the effects of state interventions and remoteness are much stronger and more significant, while
the effect of trade intensity is no longer significant. Therefore, the effects of natural openness and
state interventions evidenced in table 6 are found to mostly hold in less-free countries.
Second, when countries are split according to their FotP status (columns (5) to (8)), estimations
highlight the virtuous effect of greater press freedom on trade-related variables. In fact, the effect of
18
the parliament, civil society organizations, the media, or supreme audit institutions.
19
Description of indices is given at https://freedomhouse.org/. The Press Freedom Index ranges from 0 (the most free) to
100 (the least free). The Civil Liberties and Political Rights Indices range from 1 (the most free) to 7 (the least free).
22
both trade intensity and remoteness on bribe payments and incidence turns negative and 1%-
significant in countries with free press.
20
By contrast, the effect of remoteness from world markets on
bribery is positive and significant in countries with less-free press. Regarding domestic variables, the
effects of state interventions on the economy are significant in both sub-samples, but stronger in the
sample of FotP less-free countries than in the sample of FotP free countries.
In short, civil liberties and political rights on the one hand, and press freedom on the other hand
appear as complementary corruption deterrent: while the former are found to neutralize the effect of
internal determinants of corruption, i.e. related to state interventions, the latter is found to mitigate
the effect of external trade-related causes of corruption. Moreover, estimations also show that the
negative effect of GDP per capita holds in free countries, not in less-free countries.
Table 7. Trade, state interventions and bribery in free and less-free countries
Political rights (PR) and civil liberties (CL) status
Freedom of the Press status
Free
Not free and partly free
Free
Not free and partly free
Dep. var.:
BP
BI
BP
BI
BP OLS (b)
BI
BP
BI
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
GDP per
capita
-0.0001**
(0.00005)
0.000
(0.000)
0.0002
(0.0006)
0.000
(0.000)
-0.0001***
(0.00002)
-0.00001***
(0.0000)
-0.0002
(0.0003)
-0.00004
(0.00003)
Trade
0.019***
(0.007)
0.001
(0.001)
0.053
(0.039)
0.001
(0.003)
-0.011***
(0.003)
-0.002***
(0.0008)
0.023
(0.023)
0.002
(0.002)
Remoteness
0.016
(0.013)
0.003**
(0.0017)
0.221***
(0.052)
0.013***
(0.004)
-0.008**
(0.004)
-0.002**
(0.0008)
0.116***
(0.043)
0.009*
(0.005)
Log pop.
0.042
(0.069)
0.007
(0.010)
-0.019
(0.265)
-0.018
(0.019)
0.065**
(0.028)
0.009*
(0.005)
-0.039
(0.231)
0.002
(0.024)
Pub. spend
-0.016
(0.019)
-0.003
(0.003)
0.125*
(0.074)
0.014***
(0.006)
0.013***
(0.004)
0.003***
(0.001)
0.135**
(0.070)
0.014**
(0.007)
Tax rev.(a)
0.013
(0.076)
0.006
(0.009)
-1.479***
(0.405)
-0.134***
(0.029)
-0.126***
(0.032)
-0.013***
(0.006)
-0.830***
(0.334)
-0.071**
(0.033)
Dummies
Firms sizes & sectors
Country-level random effect parameters
Intercept
0.000
0.000
0.000
0.000
.
0.003
0.000
0.034***
Slope Pub.
spend.
0.001
0.0001*
0.064***
0.0004***
.
0.087***
0.0008***
Slope tax
rev.
0.019***
0.0003***
1.066***
0.004***
.
0.836***
0.006***
Sector-level random effect parameters
Intercept
0.000
0.001***
0.000
0.001***
.
0.001***
0.000
0.001***
Slope Trade
0.00002***
0.00005***
.
0.00005***
R2 / Wald
Stat
65.6***
59.0***
105.8***
148.2***
R2=0.03
139.5***
115.8***
142.7***
LR Chi2
381.9***
1233.9***
360.2***
2302.1***
26.1***
649.7***
3695.4***
#Countries (obs)
23(10,936)
25(12,180)
13(5364)
34(17,752)
Micro controls not reported. Standard errors in parenthesis. *significant at 10%; **significant at 5%; ***significant at 1%. (a) General goods
and services tax revenue. (b) OLS regression with standard errors robust to heteroscedasticity and clustered by countries.
20
Since the LR test did not reject the superiority of the linear model over the multi-level model, so I run instead OLS
estimations and report estimates in column (5). The resulting OLS estimates are consistent with multi-level estimates with
the bribe incidence variable, reported in column (6).
24
To test the overall effect of democracy on bribe prevalence, CL, PR, and FotP indices are regressed
together with the GDP per capita, within a random-slope three-level framework. Estimates
21
are
reported in table 8 (columns (1) to (8)) and are consistent across different valid calibrations of the
random-slope model. They display contrasting evidence on the effect of these three dimensions of
democracy on bribe prevalence. In fact, while increased political rights and media independence have,
as expected, a significant deterrent effect on corrupt transactions, greater civil liberties are found to
foster bribery. This last evidence suggests that the greater civil liberties may result in a larger scope for
private initiatives, a larger private sector size, and hence an increased supply of bribes.
To explore this possibility, I enter in the corruption equation the four sub-indices from Freedom House
underlying the CL index: the rule of law index, the freedom of expression index, the associational
rights index, and the personal autonomy index. If the positive effect of civil liberties on bribe
prevalence passes through increased private initiatives, the latter index should be positively related to
bribery. Estimates of the random slope model support a positive 5%-significant effect of personal
autonomy on bribe payments (column (9)).
This contrasting evidence on the effect of democratic institutions on corruption questions the
importance of the maturity of political systems for the study of the democracy-corruption nexus. In
fact, this positive effect of personal autonomy on corruption prevalence may be explained by a larger
a scope for private corrupt transaction combined with the relative ineffectiveness of anti-corruption
safeguards prevalent in young democracies (Treisman, 2000). To check whether the positive effect of
personal autonomy holds when controlling for the longevity of political regimes, I add to the previous
model a variable of polity durability drawn from the Polity IV database.
22
Three-level estimates
reported in columns (11) and (12) of table 8 support a negative and significant effect of political
regime’s durability on bribe payments, but not on bribery incidence. More importantly, controlling for
the durability of the polity neutralizes the positive effect of personal autonomy on bribe payments, as
well as the negative effect of the rule of law on bribery incidence. This result therefore suggests that i)
the stability of political institutions, whether democratic or not, is also a significant corruption
deterrent and that ii) greater civil liberties in young democracies may lead to higher corruption levels
because of an ineffective rule of law and larger scope for private transactions.
5. Conclusion
Corruption results from individual choices, but these choices are also influenced by norms of ethics,
trust, and coordination prevailing in a given society or a social group. Corrupt micro-level decisions
may therefore be related to each other. In micro empirical analysis, this interdependence of
corruption decisions can be addressed through the multi-level modelling of corruption data. In a first
step, a literature review sheds light the contextual nature of corrupt transactions and motivates the
multi-level analysis of corruption prevalence. In a second step, an empirical study of the contribution
21
Their sign has been reversed so that their interpretation be not misleading, i.e. that a positive (negative) coefficient reflects
a positive (negative) effect on bribery.
22
The polity durability variable is the number of years since the most recent regime change (defined by a three point change
in the democracy variable score over a period of three years or less) or the end of transition period defined by the lack of
stable political institutions.
25
of key corruption determinants emphasized by the economic literature is conducted within a three-
level empirical framework.
In line with previous empirical researches, multi-level estimates support that economic development,
measured by income per capita, significantly and negatively contributes to bribery prevalence.
However, this straightforward relationship hides complex and sometimes conflicting mechanisms.
First, estimations highlight that this negative effect of income appears to be mostly driven by change
in human capital-related factors, which contrasting effect on corruption is partly mediated by the size
of public spending (Eicher et al, 2009). Second, estimations show that state interventions have strong
but again contrasting effects on corruption: while larger public expenditures increase corruption
prevalence, higher tax revenues are negatively associated with corruption. This evidence suggests that
an increased scope for state intervention may stimulate rent-seeking behaviours by inducing
redistribution (Tanzi, 1998; Tornell and Lane, 1999), but on the other hand, increased state
interventions often track the institutional development process (Peacock and Scott, 2000; Rodrik,
2000). Third, results stress that trade openness is partly mediated by state interventions but also by
structural factors such as countries geographical distance from world markets. This results therefore
finds echo in the conclusions of Rodrik (1992, 1998), who stresses the role of government size for
trade policy performances, and those of Wei (2000), who highlights the contribution of natural
openness to the quality of institutions. Last, estimations emphasize the direct and indirect effects of
democratic institutions on bribe prevalence, and also stress contrasting evidence: while increased
political rights and media independence have, as expected, a significant deterrent effect on corrupt
transactions, greater civil liberties are found to foster bribery. This result corroborates the findings of
Treisman (2007) and stresses the importance of the maturity of political systems for the study of the
democracy-corruption nexus, suggesting that the transition towards democracy may temporarily
widen a scope for private corrupt transactions.
26
REFERENCES
Acemoglu, D. and T. Verdier (2000) “The Choice between Market Failures and Corruption”, American
Economic Review, 90:194-211.
Acemoglu, D., Johnson, S., and Robinson, J. A. (2001) The Colonial Origins of Comparative
Development: An Empirical InvestigationAmerican Economic Review, 91(5): 1369-1401.
Acemoglu, D., Carvalho, V. M., Ozdaglar, A., and Tahbaz‐Salehi, A. (2012) “The network origins of
aggregate fluctuations” Econometrica, 80(5): 1977-2016.
Ades, A., and Rafael Di Tella (1999) “Rents, Competition, and Corruption” American Economic Review,
89(4):982-993.
Aidt, T.S. (2003) “Economic analysis of corruption: a survey” The Economic Journal, 113(491):632-652.
Alesina, A., and R. Wacziarg (1998) “Openness, country size and government” Journal of Public
Economics, 69(3):305-321.
Andvig, J.C. (2006) “Corruption and Fast Change”, World Development, 34(2):328-340.
Andvig, J.C., and K.O. Moene (1990) “How corruption may corrupt”, Journal of Economic Behaviour &
Organization, 13:63-76.
Banerjee, A.V. (1997) “A Theory of Misogovernance” The Quarterly Journal of Economics, 112(4):1289-
1332.
Banerjee, A.V., and R. Pande (2007) “Parochial Politics: Ethnic Preferences and Politician Corruption”,
National Bureau of Economic Research.
Banfield, E.C. (1975) “Corruption as a feature of governmental organization”, The Journal of Law and
Economics, 18(3):587-605.
Becker, Gary S. "An economic analysis of fertility." In Demographic and economic change in developed
countries. Columbia University Press, 1960. 209-240.
Becker, G. S., and N. Tomes (1994). “Human capital and the rise and fall of families.” In Human Capital:
A Theoretical and Empirical Analysis with Special Reference to Education (3rd Edition) (pp. 257-298).
The University of Chicago Press.
Bird, R.M., and P-P. Gendron, (2007). The VAT in Developing and Transitional Countries, Cambridge:
Cambridge University Press.
Boadway, R., and M. Sato (2009). Optimal Tax Design and Enforcement with an Informal Sector,
American Economic Journal: Economic Policy, 1(1): 1-27.
Blundo, G., J-P.O. de Sardan, and S. Cox, Everyday Corruption and the State: Citizens and Public
Officials in Africa. London: Zed Books, 2006.
Brezis, E. S., and Young, W. (2014) “Population and economic growth: Ancient and modern”, The
European Journal of the History of Economic Thought, 23(2):1-27.
Cadot, O. (1987) “Corruption as a gamble”, Journal of Public Economics, 33(2):223-244.
Caudill, S. B. (1988). “Practitioners corner: An advantage of the linear probability model over probit or
logit”, Oxford Bulletin of Economics and Statistics, 50(4):425-427.
27
Chand, S. K., and Moene, K.O. (1999) “Controlling fiscal corruption” World Development, 27(7):1129-
1140.
Chaney, T. (2014) "The Network Structure of International Trade." American Economic
Review, 104(11): 3600-3634.
Clarke, G.R.G. (2011). “How Petty is Petty Corruption? Evidence from Firm Surveys in Africa”. World
Development, 39(7):1122-1132.
Dabla-Norris E., Gradstein, M. and G. Inchauste (2008). “What causes firms to hide output? The
determinants of informality”, Journal of Development Economics, 85:1-27.
Diaby, A., and K. Sylwester (2015). “Corruption and Market Competition: Evidence from Post-
Communist Countries”. World Development, 66:487499.
Dreher, Axel, and F. Schneider (2010) "Corruption and the shadow economy: an empirical
analysis." Public Choice, 144(1-2): 215-238.
Dutt, P. (2009) “Trade protection and bureaucratic corruption: an empirical investigation”, Canadian
Journal of Economics, 42:155-183.
Dutt, P., and D. Traca (2010) “Corruption and Bilateral Trade Flows: Extortion or Evasion?”, The Review
of Economics and Statistics, 92(4):843-860
Eicher, T., García-Peñalosa, C., and T. van Ypersele, (2009) “Education, corruption, and the distribution
of income” Journal of Economic Growth, 14(3):205-231.
Farla, K. (2014) “Determinants of firms’ investment behaviour: a multilevel approach”, Applied
Economics, 46(34):4231-4241.
Fenner, G. “Financial institutions and the fight against corruption”, in ed. Transparency International,
Corruption in the Private Sector, Global Corruption Report, 2009.
Fisman, R., and E. Miguel “Nature or Nurture? Understanding the Culture of Corruption” in eds.
Fisman, R., and E. Miguel, Economic Gangsters, Chapter 4, Princeton University Press, 2008.
Fisman, R., and E. Miguel (2007) “Corruption, Norms, and Legal Enforcement: Evidence from
Diplomatic Parking Tickets”, Journal of Political Economy, 115(6):1020-1048.
Fisman, R., and R. Gatti (2002) “Decentralization and corruption: evidence across countries”, Journal of
Public Economics, 83:325-345.
Fisman, R., and Svensson, J. (2007) “Are corruption and taxation really harmful to growth? Firm level
evidence” Journal of Development Economics, 83(1): 63-75.
Gabaix, X. (2011). The granular origins of aggregate fluctuationsEconometrica, 79(3): 733-772.
Gatti, R. (2004) “Explaining corruption: Are open countries less corrupt?”, Journal of International
Development, 16:851861.
Gauthier, B., and J. Goyette (2014) “Taxation and corruption: theory and firm-level evidence from
Uganda”, Applied Economics, 46(23): 2755-2765.
Gauthier, B., and R. Reinikka (2006) “Shifting Tax Burdens through Exemptions and Evasion: an
Empirical Investigation of Uganda”, Journal of African Economies, 15(3), 373-398.
28
Glaeser, E. L., La Porta, R., Lopez-de-Silanes, F., and Shleifer, A. (2004) “Do institutions cause
growth?” Journal of economic Growth, 9(3): 271-303.
Guriev, S. (2004) “Red Tape”, Journal of Development Economics, 73:489-504.
Hellman, J.S., Jones, G., and D. Kaufmann (2003) “Seize the state, seize the day: state capture and
influence in transition economies”, Journal of Comparative Economics, 31(4):751-773.
Héricourt, J., and S. Poncet (2015). Exchange rate volatility, financial constraints, and trade: empirical
evidence from Chinese firms. The World Bank Economic Review, 29(3): 550-578.
Hallward-Driemeier, M., and L. Pritchett (2015). “How Business is Done in the Developing World: Deals
versus Rules”, The Journal of Economic Perspectives, 29(3):121-140.
Hellman, J. S., Jones, G., and D. Kaufmann (2003). “Seize the state, seize the day: state capture and
influence in transition economies”, Journal of Comparative Economics, 31(4):751-773.
Hibbs, D.A., and V. Piculescu (2010) “Tax Toleration and Tax Compliance: How Government Affects the
Propensity of Firms to Enter the Unofficial Economy”, American Journal of Political Science, 54:18-33.
Hindriks, J., Muthoo, A., and M. Keen (1999). Corruption, extortion and evasion. Journal of Public
Economics, 74: 395430.
Hox, J.J. Multi-level Analysis, Techniques and Applications, Quantitative Methodology Series, Rouledge,
2010.
Human development report, Sustainability and equity: a better future for all, United Nations
Development Programme, 2011.
Husted, B. W. (1999) “Wealth, culture, and corruptionJournal of International Business Studies,
30(2):339-359.
Kaufmann, D. (2005) “Myths and Realities of Governance and Corruption” in eds. Porter, M.E., and K.
Schwab, The Global Competitiveness Report 2005-2006, The World Economic Forum, pp.81-98.
Kingston, C. (2008) “Social structure and cultures of corruption”, Journal of Economic Behaviour &
Organization, 67:90-102.
Kingston, C. (2007) “Parochial corruption”, Journal of Economic Behaviour & Organization, 63:73-87.
Klitgaard, R., Controlling corruption, University of California Press, 1988.
Knack, S., and O. Azfar (2003) “Trade intensity, country size and corruption”, Economics and
Governance, 4:18.
Knack, S. and P. Keefer (1997) “Does social capital have an economic payoff? A cross-country
investigation”, The Quarterly Journal of Economics, 112(4):1251-1288.
Kodi, M., Corruption and Governance in the RDC during the transition period (2003-2006), monograph
series No.148, Institute for Security Studies, 2008.
Kuznets, Simon. "Population change and aggregate output." In Demographic and economic change in
developed countries. Columbia University Press, 1960. 324-351.
29
La Porta, R., Lopez-de-Silanes, F., Shleifer, A. and and R Vishny (1999) “The quality of the government”,
The Journal of Law, Economics, and Organization, 15:222-279.
Lambsdorff, J.G. (2005) “Consequences and Causes of Corruption What do We Know from a Cross-
Section of Countries?”, Passau/Germany: Universität Passau.
Lambsdorff, J.G. (2002) “Corruption and Rent-seeking”, Public Choice, 113(1-2)97-125.
Lambsdorff, J.G., and B. Frank (2011) “Corrupt reciprocity – Experimental evidence on a men’s game”,
International Review of Law and Economics, 31(2):116-125.
Lesné, F. "Trois façons d’évaluer la corruption : comment comparer les indicateurs ?" Ferdi, Document
de travail I16, Novembre 2013.
Letki, N. (2006) “Investigating the Roots of Civic Morality: Trust, Social Capital, and Institutional
Performance”, Political Behaviour, 28(4):305-325.
Li, H., Liang, Y., Liu, Z., and X. Wang (2013) “Human Capital in China, 1985-2008,” Review of Income
and Wealth , 59(2):212-234.
Mauro, P. (2004) “The persistence of corruption and slow economic growth”. IMF Staff Paper, 51:1-18.
Nguyen, A., and W. Cragg (2012) “Inter-organizational Favour Exchange and the Relationship, Between
Doing Well and Doing Good”, Journal of Business Ethics, 105:53-68.
Olson, M. (1993) “Dictatorship, Democracy and Development”, The American Political Science Review,
87(3):567-576.
Paunov, C., and V. Rollo (2015), “Overcoming Obstacles: The Internet’s Contribution to Firm
Development”, The World Bank Economic Review, 29(Supplement), 192-204.
Peacock, A., and A. Scott (2000) “The curious attraction of Wagner’s law”, Public Choice, 107(1-2):1-
17.
Persson, T, and G. Tabellini (2004) “Constitutions and Economic Policy”, Journal of Economic
Perspective, 18:75-98.
Pope, J. (2000), “The Private Corporate Sector” in Ed. Pope, J., TI Source Book 2000 - Confronting
Corruption: the Elements of a National Integrity System, Berlin: Transparency International.
Razafindrakoto, M., and F. Roubaud (2010) “Are international databases on corruption reliable? A
comparison of expert opinion surveys and household surveys in sub-Saharan Africa”, World
development, 38(8):10571069.
Rodrik, D. (2000) “Participatory Politics, Social Cooperation, and Economic Stability”, American
Economic Review, 90(2):140-144.
Rodrik, D. (1998) “Why Do More Open Economies have Bigger Governments?”, Journal of Political
Economy, 106(5):997-1032.
Rodrik, D. (1992). The limits of trade policy reform in developing countries. The Journal of Economic
Perspectives, 6(1): 87-105.
Rose-Ackerman, S. (2007) “Measuring Private Corruption”, U4 Brief No.5, U4 Anti-Corruption Resource
Center.
30
Rose-Ackerman, S. (1996) “The Political Economy of Corruption – Causes and Consequences”, Public
Policy for the Private Sector, Note no.74, The World Bank.
Sandholtz, W., and W. Koetzle (2000) “Accounting for corruption: Economic structure, democracy, and
trade”, International Studies Quarterly, 44:31-50.
Shleifer, A., and R.W. Vishny, (1993) "Corruption", The Quarterly Journal of Economics, 108(3):599-
617.
Siddiqui, O., Hedeker, D., Flay, B.R., and F.B. Hu (1996) “Intraclass correlation estimates in a school-
based smoking prevention study: Outcomes and mediating variables, by gender and ethnicity.”
American Journal of Epidemiology, 144:425-433.
Sovacool, B. K. (2016) “Countering a corrupt oil boom: Energy justice, Natural Resource Funds, and Sao
Tome e Principe's Oil Revenue Management Law”. Environmental Science & Policy, 55:196-207.
Svensson, J. (2005) “Eight questions about corruption”, The Journal of Economic
Perspectives, 19(3):19-42.
Svensson, J. (2003) “Who Must Pay Bribes and How Much? Evidence from a Cross Section of Firms”,
The Quarterly Journal of Economics, 118(1):207-230.
Tanzi, V. (1998) “Corruption around the World: Causes, Consequences, Scope, and Cures”, IMF Staff
Paper, 45(4):559-594.
Teorell, J., Samanni, M., Holmberg, S., and B. Rothstein, The Quality of Government Dataset, version
6Apr11. University of Gothenburg: The Quality of Government Institute, 2011.
Tornell, A. & Lane, P. R. (1999) The voracity effect. The American Economic Review, 89 (1), 2246
Transparency International, Corruption in the Private Sector, Global Corruption Report, 2009.
Transparency International, “Using the OECD Guidelines to Tackle Corporate Corruption”, Working
Paper, No.3, 2008.
Treisman, D. (2007) “What Have We Learned About the Causes of Corruption from Ten Years of Cross-
National Empirical Research?”, Annual Review of Political Sciences, 10:211-244.
Treisman, D. (2000) “The Causes of Corruption: a Cross-National Study”, Journal of Public Economics,
76(3):399-457.
Uslaner, E.M., “Trust and corruption” in eds. Lambsdorff, G, Taube, M. and M. Schramm, The New
Institutional Economics of Corruption, Chapter 5, Routledge, 2005.
Varvarigos, D., and P. Arsenis (2015). Corruption, fertility, and human capital”, Journal of Economic
Behavior & Organization, 109:145-162.
Vitell, S.J., Dickerson, E.B., and T.A. Festervand (2000) “Ethical problems, conflicts and beliefs of small
business professionals”, Journal of Business Ethics, 28:15-24.
Wei, S-J. (2000) “Natural Openness and Good Government”, Working Paper No.7765, National Bureau
of Economic Research, Cambridge.
Wilson, J.Q., Bureaucracy: What Government Agencies Do and Why They Do It. Basic Books, NY, 1989.
31
32
APPENDICES
A.SUMMARY STATISTICS
A.1. Sample summary statistics
Variable
Obs
Mean
Std. Dev.
Min
Max
Sources
Bribe payments (% of sales)
34358
1.356437
4.953842
0
11.4911
WBES
Bribery incidence
34358
0.196228
0.3971488
0
1
WBES
GDP per capita
34358
3369.933
3070.527
146.3977
21013.9
WDI
Fertility rate
29660
3.487835
1.51194
1.4
6.559
UNESCO - QoG database
Primary enrolment ratio
30099
107.2765
12.1074
43.6712
137.6601
UNESCO - QoG database
Public spending
32702
6.008457
10.05859
-27.83
64.002
IMF
Tax revenue
27936
5.121478
2.542863
0.1362075
13.065
IMF
Trade intensity (% of trade in
GDP)
33968
69.02034
26.63353
22.1183
155.6252
WDI
Remoteness index
30812
64.14002
17.97551
17.517
100
Ferdi
Log population
34358
15.74181
2.702494
9.920542
21.51861
WDI
PR_scores
34358
3.534344
1.885392
1
7
Freedom House
CL_scores
34358
3.491763
1.469576
1
7
Freedom House
fotp_score
34358
53.00675
18.66132
15
94
Freedom House
Durability
33337
17.48121
16.03764
0
63
Polity IV
Log total sales
34358
16.97554
2.974218
0
35.53203
WBES
% firms public ownership
34358
0.5515455
6.253454
0
100
WBES
% indirect exports
34358
2.476369
12.4893
0
100
WBES
% of direct exports
34358
5.955108
19.54836
0
100
WBES
Internal funding
34358
68.46624
34.14051
0
100
WBES
Bank funding
34358
12.06221
22.96764
0
100
WBES
Dummy large size
34358
0.1759416
0.380776
0
1
WBES
Dummy medium size
34358
0.3252518
0.4684757
0
1
WBES
A.2. Baseline sample composition, by region
#observations
Region
2006
2007
2008
2009
2010
2011
2012
Total
Sub-Saharan Africa
3,129
5,759
0
1,184
564
840
0
11,476
East-Asia and Pacific
0
0
0
2,311
0
0
1,830
4,141
Eastern Europe and Central Asia
0
1,204
0
0
0
0
2,342
3,546
Latin America and Caribe
5,650
0
0
123
6,878
0
0
12,651
Middle-East and North Africa
0
0
0
0
0
519
0
519
33
South Asia Region
0
517
358
544
0
461
145
2,025
Total
8,779
7,480
358
4,162
7,442
1,820
4,317
34,358
A.3. Baseline sample composition, by country
#Observations
Bribe
payments (%
of sales)
Bribe
incidence
71 Countries
2006
2007
2008
2009
2010
2011
2012
Total
Afghanistan
0
0
358
0
0
0
0
358
2.4944134
0.38547486
Angola
359
0
0
0
233
0
0
592
3.3440878
0.38513514
Antigua & Barbuda
0
0
0
0
125
0
0
125
0
0
Argentina
510
0
0
0
0
0
0
510
1.3784314
0.18823529
Bahamas, The
0
0
0
0
107
0
0
107
0.17757009
0.07476636
Bangladesh
0
517
0
0
0
0
0
517
2.4880542
0.77369439
Belize
0
0
0
0
135
0
0
135
0.05925926
0.01481481
Bhutan
0
0
0
239
0
0
0
239
0.31799163
0.05020921
Bolivia
343
0
0
0
177
0
0
520
2.2096154
0.25
Botswana
257
0
0
0
204
0
0
461
0.97635575
0.13232104
Brazil
0
0
0
123
0
0
0
123
6.2439024
1
Bulgaria
0
725
0
0
0
0
0
725
0.69241379
0.10482759
Burkina Faso
0
0
0
217
0
0
0
217
1.0737327
0.06451613
Burundi
258
0
0
0
0
0
0
258
4.6015504
0.5503876
Cabo Verde
0
0
0
106
0
0
0
106
0.61320755
0.01886792
Cameroon
0
0
0
286
0
0
0
286
2.8776224
0.42657343
Central Afr. Rep.
0
0
0
0
0
127
0
127
3.5590551
0.37795276
Chad
0
0
0
127
0
0
0
127
2.5748031
0.34645669
Chile
754
0
0
0
894
0
0
1,648
0.25788835
0.0315534
China
0
0
0
0
0
0
1,830
1,830
0.18743169
0.03934426
Colombia
675
0
0
0
794
0
0
1,469
1.0633084
0.09121852
Cote d'Ivoire
0
0
0
112
0
0
0
112
11.491071
1
Croatia
0
479
0
0
0
0
0
479
0.51356994
0.07306889
Dominica
0
0
0
0
138
0
0
138
0
0
Dominican Rep.
0
0
0
0
299
0
0
299
0.37458194
0.05685619
Ecuador
462
0
0
0
306
0
0
768
0.8125
0.10286458
El Salvador
328
0
0
0
268
0
0
596
1.1073826
0.12583893
Eritrea
0
0
0
127
0
0
0
127
0
0
Gambia, The
135
0
0
0
0
0
0
135
4.6814815
0.5037037
Ghana
0
471
0
0
0
0
0
471
2.0006369
0.29723992
Grenada
0
0
0
0
136
0
0
136
0.16911765
0.07352941
Guatemala
309
0
0
0
413
0
0
722
1.2451524
0.08033241
Guinea
189
0
0
0
0
0
0
189
5.7015873
0.82010582
Guinea-Bissau
102
0
0
0
0
0
0
102
3.5745098
0.53921569
Guyana
0
0
0
0
126
0
0
126
0.75396825
0.15079365
Honduras
204
0
0
0
251
0
0
455
1.4549451
0.12747253
Indonesia
0
0
0
1,024
0
0
0
1,024
0.43164063
0.12402344
Iraq
0
0
0
0
0
519
0
519
1.9788054
0.23121387
Kenya
0
646
0
0
0
0
0
646
2.6866715
0.70897833
34
Lao PDR
0
0
0
323
0
0
145
468
0.7542735
0.16666667
Malawi
0
0
0
108
0
0
0
108
0.4537037
0.10185185
Mali
0
444
0
0
127
0
0
571
1.357268
0.18739054
Mauritania
191
0
0
0
0
0
0
191
4.6068063
0.80104712
Mexico
829
0
0
0
83
0
0
912
1.2653509
0.17653509
Mozambique
0
463
0
0
0
0
0
463
1.6274298
0.13390929
Namibia
277
0
0
0
0
0
0
277
0.83104693
0.11552347
Nepal
0
0
0
305
0
0
0
305
0.46229508
0.06557377
Nicaragua
195
0
0
0
283
0
0
478
1.2887029
0.12133891
Nigeria
0
1,891
0
0
0
0
0
1,891
1.9649637
0.4235854
Panama
387
0
0
0
23
0
0
410
3.197561
0.24390244
Paraguay
145
0
0
0
225
0
0
370
4.2351351
0.32702703
Peru
431
0
0
0
764
0
0
1,195
0.58493724
0.09539749
Philippines
0
0
0
964
0
0
0
964
1.0829876
0.16804979
Russian Fed.
0
0
0
0
0
0
2,342
2,342
0.95260461
0.11016225
Rwanda
208
0
0
0
0
153
0
361
1.6803324
0.13296399
Senegal
0
494
0
0
0
0
0
494
1.5665992
0.22469636
South Africa
0
878
0
0
0
0
0
878
0.61264237
0.09339408
Sri Lanka
0
0
0
0
0
461
0
461
0.31670282
0.07158351
St. Kitts and Nevis
0
0
0
0
118
0
0
118
0.20338983
0.04237288
St. Lucia
0
0
0
0
136
0
0
136
0
0
St. Vincent
0
0
0
0
126
0
0
126
0.02380952
0.01587302
Suriname
0
0
0
0
152
0
0
152
0.38157895
0.07236842
Swaziland
280
0
0
0
0
0
0
280
1.2628571
0.40357143
Tanzania
387
0
0
0
0
0
0
387
3.7372093
0.47286822
Togo
0
0
0
101
0
0
0
101
0.91089109
0.12871287
Trinidad & Tobago
0
0
0
0
265
0
0
265
0.17358491
0.07169811
Uganda
486
0
0
0
0
0
0
486
3.7965021
0.47736626
Uruguay
78
0
0
0
417
0
0
495
0.16767677
0.04040404
Venezuela, RB
0
0
0
0
117
0
0
117
3.0598291
0.36752137
Zambia
0
472
0
0
0
0
0
472
1.1213983
0.16101695
Zimbabwe
0
0
0
0
0
560
0
560
0.18214286
0.04107143
Total
8,779
7,480
358
4,162
7,442
1,820
4,317
34,358
1.3564369
0.19622795
35
A.3. Baseline sample composition, by sector
Sector
# observations.
Bribe payments (%
of sales)
Bribe incidence
Textiles
1,859
0.76747837
0.13824637
Leather
303
1.6729472
0.28382838
Garments
2,795
1.2458604
0.20930233
Food
3,960
1.2328808
0.19318182
Metals and machinery
2,203
0.96466992
0.14298684
Electronics
508
0.63799213
0.11220472
Chemicals and pharmaceuticals
1,771
1.1147657
0.17617165
Wood and furniture
609
1.3449884
0.27422003
Non-metallic and plastic materials
1,807
0.8990213
0.13779745
Auto and auto components
126
0.3015873
0.1031746
Other manufacturing
4,655
1.7614892
0.22534909
Retail and wholesale trade
7,371
1.3556819
0.18925519
Hotels and restaurants
1,377
1.3424409
0.24473493
Other services
3,232
1.7796684
0.21194307
Other: Construction, Transportation, et
1,782
2.0249227
0.2633352
Total
34,358
1.3564369
0.19622795
... Cariolle [6] has suggested multilevel framework for esti mation of corruption that considers the economic and human development levels, the size of governments, trade openness, and democracy. ...
Article
Purpose.Development of a scientific and methodological approach to the identification of the most impactful determinants on corruption using multivariate adaptive regression splines. Methodology. Methodological tools of the research methods are comparison, grouping, bibliometric analysis, and multivariate adaptive regression splines in the form of piecewise linear functions. Findings. Systematization of the literary sources and approaches for factors influencing corruption indicates that most empirical studies are based on using panel data. Panel data allows you to insert general patterns, but does not consider the patterns of the national economy. For the study on corruption in Ukraine, 15 influencing factors were selected, characterizing the institutional, economic and social environment. Based on the constructed MAR Spline models, three regression equations were obtained that describe the linear dependence of the level of corruption in governance on the selected factors. The paper found that the relevant factors influencing corruption in Ukraine are: tax burden, general government final consumption expenditure, average monthly wage in Governance and rule of law. Originality.The proposed approach makes it possible to determine the dynamics of the degree of factor influence on the level of corruption in the country. The paper defines the threshold values of statistically significant indicators at which the maximum degree of correlation with the corruption perception index is achieved. Practical value.The regularities between the level of corruption and economic, institutional and social factors revealed by the research results can be used in the development of tools to fight corruption in Ukraine. The formation of an effective anti-corruption system will strengthen financial stability in the country and increase the level of public trust in society.
Working Paper
Full-text available
The evidence of a “voracity effect” of revenue windfalls reducing growth by fostering rent-seeking and corruption is widely documented by the literature. However, the reverse hypothesis of a “scarcity effect” of revenue declines, stimulating corruption by creating shortages, has theoretical foundations but little empirical support. This paper fills this gap by providing an empirical analysis of the voracity and scarcity effects of aggregate export windfalls and downfalls on firm-level bribery. Exploiting 24,920 bribery reports from firms located in 49 developing and transition countries, multilevel estimations of these effects are conducted. Results support a positive effect of both export surges and slumps on bribe payments and incidence, when democratic and financial institutions are weak; and a negative effect when institutions are stronger. Estimated relationships are more pronounced for manufactures than service firms, and stay robust when the sample is restricted to small-medium non-exporting firms and to domestic firms, which lowers the eventuality of reverse causality from firm corrupt transactions to aggregate export shocks. Therefore, consistent with the literature, this paper provides additional evidence on the importance of institutional safeguards against corrupt practices in times of abundance. But more importantly, it provides new insights into their importance in times of shortage.
Article
Full-text available
This paper analyzes the conflicts of interest arising from the " revolving door ". The revolving door is a common phenomenon, and it is unlikely that most of it can be explained by 'regulatory capture', a practice that is unlawful. Therefore, there is a need for a new framework. This paper proposes a framework wherein conflicts of interest arising from the revolving door are not unlawful, as is in the case of regulatory capture, but still lead to economic distortions. The paper introduces a market for bureaucratic capital, which explains why in equilibrium, the government allows this unethical, yet not unlawful, conflict of interest to persist. Our first result is that the political elite finds it optimal to allow the existence of the revolving door, as well as the creation of bureaucratic capital. The second result is that in equilibrium, the revolving door leads to an excessive level of bureaucratic capital. As a consequence, the interconnection of elites and the existence of the revolving door actually lead to lower economic growth.
Article
Organizations * Armies, Prisons, Schools * Organization Matters Operators * Circumstances * Beliefs * Interests * Culture Managers * Constraints * People * Compliance Executives * Turf * Strategies * Innovation Context * Congress * Presidents * Courts * National Differences Change * Problems * Rules * Markets * Bureaucracy and the Public Interest
Article
Because government intervention transfers resources from one party to another, it creates room for corruption. As corruption often undermines the purpose of the intervention, governments will try to prevent it. They may create rents for bureaucrats, induce a misallocation of resources, and increase the size of the bureaucracy. Since preventing all corruption is excessively costly, second-best intervention may involve a certain fraction of bureaucrats accepting bribes. When corruption is harder to prevent, there may be both more bureaucrats and higher public-sector wages. Also, the optimal degree of government intervention may be nonmonotonic in the level of income
Article
Discussion of how to combat corruption have focused more sharply on the recipients of bribes than on those who pay them. A more balanced approach, which is emerging, promises to make anticorruption efforts more effective. It is early days, however. Changing corporate habits takes time and is difficult. Many corporations have been paying bribes around the world for decades. But, through legislative and regulatory actions; new official interventions; and the work of civil society; the media, and public prosecutors, the heat has been turned up on the bribe givers. Corporations, as a result, are starting to respond. The agenda of actions designed to combat corruption by influencing the supply side - penalizing the payment of bribes in international business transactions - is getting longer and more substantive. Such efforts have not received sufficient publicity, but as the relevant facts and trends become better known, they should further encourage leaders in the public and private sectors, who, meanwhile, are showing courage and skill in influencing the demand side of the corruption equation by penalizing the receipt of bribes.
Conference Paper
Elections serve two functions in representative democracies. First, they select political actors who enact public policies in the light of constituents' preferences. Second, they permit citizens to hold their representatives accountable and to punish them if they enrich themselves in corrupt or self-serving ways. In other words, elections provide both incentives for politicians to enact certain kinds of policies and constraints on politicians' malfeasance. In this chapter, we focus on the second of these two functions and investigate how different electoral systems constrain corrupt rent-seeking, holding constant other political, economic and social factors.
Chapter
This book shows that in calling economics the ‘dismal science’, Thomas Carlyle was profoundly wrong. The influence of economic thinking on other social sciences is bringing about a theoretical integration of all the social sciences under one overarching paradigm. The ten chapters of this book illustrate the intellectual advances that account for this unified view of economics and societies. The key theme that emerges is the interaction between political, economic, legal, and social forces. Examples of this include the political influence of corruption and special interest groups, the organizational structure of a government, the effect of commercial law, and the differences between communities with high and low social fragmentation. All these affect and are affected by economic conditions.