1. Top of page
  2. Abstract
  3. Introduction
  4. 2 Topics of corruption research
  5. 3 Methodological challenges
  6. 4 Empirical corruption analyses on Southeast Asia
  7. 5 Conclusion
  8. References

This paper surveys the empirical literature on corruption in Southeast Asia with a focus on the methodological approach that the contributions take to identify the extent, determinants, and consequences of corruption and the remedies against it. We present the major topics that empirical corruption research has focused on and point out the methodological challenges that this line of research has to address. We discuss the empirical corruption studies on Southeast Asia and describe the empirical approach that they have taken.


  1. Top of page
  2. Abstract
  3. Introduction
  4. 2 Topics of corruption research
  5. 3 Methodological challenges
  6. 4 Empirical corruption analyses on Southeast Asia
  7. 5 Conclusion
  8. References

Corruption involves breaking the rules by public officials for private gain (Banerjee et al. 2012b), which distorts the resource allocation and induces economic inefficiencies (Bardhan 1997). In corrupt environments, resources are diverted from productive activities towards various forms of rent seeking (Krueger 1974), and talents are misallocated to occupations that are individually rewarding but socially not the most productive (Acemoglu and Verdier 1998). The secrecy involved in corrupt transactions, and the uncertainty created place an extra burden on the economy (Shleifer and Vishny 1993). Uncertainty arises not only because corrupt contracts, being illegal, cannot be enforced, but also because politicians and bureaucrats deliberately create situations of regulatory uncertainty to make bribery a necessary remedy to the problem they created (Tanzi 1998). Firms or individuals have to bribe the official in order to reduce their uncertainty or excessive delays in their dealings with the government. Misallocation of resources, increase in the costs of doing business, and an endogenous system of red tape reduce the productivity of capital (Lambsdorff 2003a), reduce investment (Knack and Keefer 1995; Mauro 1997), and lower FDI (Wei 2000; Lambsdorff 2003b). As a result of all this, corruption diminishes economic growth (Mauro 1995, 1997) and compromises development (Bardhan 1997). In addition, corruption often creates a hugely inequitable distribution as it favours those individuals and firms with political connections or the best ‘corruption technology’ rather than those with the highest skills or the most efficient production technology.

Due to its potentially huge economic costs and its equity and ethical implications, economic analysis has been trying to assess the extent and consequences of corruption in various contexts and to search for potential remedies. The quantitative analysis on corruption serves not only descriptive purposes, but it is essential for understanding the mechanisms of corruption and for developing successful anti-corruption strategies.

Yet, corruption, clandestine by its very nature, is very challenging to measure. As it is typically not directly observable,1 approaches to its measurement have to be derived that are feasible and reasonably accurate and create unbiased, representative estimates. Earlier empirical approaches relied predominantly on assessments by expert committees as well as on survey evidence on corruption perceptions or on experiences by the business community or the population at large. Although readily available, such measures suffer from much larger potential perception biases than other, more direct ways of measuring corruption.

In the past few years, a new generation of empirical analyses has emerged which go beyond perception measurements and are considerably more reliable. Direct observation of corruption requires experimental settings or the tracking of public expenditures through specialised surveys and audits. The extent of corruption can often also be estimated indirectly through publicly available information such as market outcomes, firms' balance sheets, or fiscal accounts. While more circumferential, these alternative approaches offer evidence that is much less dependent on context-specific perceptions and easier to quantify and analyse.

Empirical analyses of corruption are even more complicated as corruption is very context specific. It depends on the culture (Fisman and Gatti 2002b; Barr and Serra 2010), the institutional setting (including the legal system), and the state of development, among other things (Treisman 2000; Paldam 2002; Baksi et al. 2009, etc.). Therefore, it is not a priori clear to what extent results derived in one context apply to another context. In other words, while economic analysis has no context-specific analytical tools, it needs to take into consideration the context-specific determinants of corruption, some of which are not easily quantifiable.

This paper critically surveys the empirical literature that has analysed corruption in the context of Southeast Asia. While we report on the findings of these contributions, our focus is on the methodological approaches that the studies have taken in order to overcome the methodological challenges described above (and below).2

Corruption is a severe problem for many Southeast Asian countries. In the most recent Corruption Perceptions Index (CPI) of Transparency International (2012), several countries in the region figure in the most corrupt third. Table 1 gives an excerpt of the online version of the 2012 CPI; the original table lists 176 countries. Southeast Asian countries are highlighted.3

Table 1. Transparency International's corruption perceptions index (CPI)
  1. The score ranges from 1.0 (most corrupt) to 100 (not corrupt), 176 countries covered.

Source: Transparency International (2012), available at, accessed 20 March 2013.
1New Zealand90
 14Hong Kong77
 19United States of America73
 79Sri Lanka40
113Timor Leste33
174North Korea8

The paper proceeds as follows. The next section discusses the research questions that empirical analyses of corruption have sought to answer. The third section elaborates on the methodological difficulties of empirical corruption research. Section 4 presents empirical corruption analyses for Southeast Asia, sorted by the econometric approach that they use. Section 5 concludes.

2 Topics of corruption research

  1. Top of page
  2. Abstract
  3. Introduction
  4. 2 Topics of corruption research
  5. 3 Methodological challenges
  6. 4 Empirical corruption analyses on Southeast Asia
  7. 5 Conclusion
  8. References

Empirical research on corruption seeks to answer the following questions: (1) What are the magnitude and incidence of corruption; (2) What are the determinants of corruption; (3) What are the consequences of corruption; and (4) How do we fight corruption most effectively? In this section, we highlight the issues around which the empirical research on corruption has evolved, without claiming to be exhaustive. We intend to set the stage for the more detailed description of corruption studies in the context of Southeast Asia, which we therefore leave out in this section.

2.1 Magnitude and incidence

Corruption is an elusive concept. The most common definition of corruption is misuse of public office for private gain. Corruption thus includes a wide array of activities, for instance accepting bribes in exchange for favourable treatment, such as the award of public procurement contracts or neglected prosecution in case of criminal offences, but it also comprises nepotism or the absenteeism of public officials, who steal the contracted time for different purposes, and theft of government transfers. Private gains may be monetary or in kind and may be directed to the public official or to his friends and family.

The magnitude and incidence depends on the measurement concept (cf. section 4). To establish an overall incidence/intensity pattern across regions or countries, a measure is called for that is broad enough to capture different forms of corruption and is not specific to the situation in which corruption may arise. Such a measure may be based on perceptions about corruption (cf. section 4.1), on corruption experiences (cf. section 4.2) reported by individuals or firms, on corruption convictions (Glaeser and Saks 2006), or differences in behaviour in experimental settings (cf. section 4.3). Alternatively, corruption could be inferred from measurable consequences of corruption such as the market valuation of firms (cf. section 4.4) or the leakage in public spending (cf. section 4.5). The resulting corruption measures differ in terms of context specificity and may not be representative for overall corruption levels.4 While the incidence analysis is descriptive, it lends itself to being extended into a causal analysis that asks for the root causes of corruption.

2.2 Determinants

The most fundamental question of empirical corruption research is what the circumstances are that breed corruption. Due to reasons of data availability, most of research on cross-country differences in corruption actually explains variations in corruption perceptions (cf. section 4.1) and not necessarily the actual incidence of corruption.

Among the factors that have been in the focus of attention are the stage of development,5 the economic system, the distribution of political power, and the institutional framework (democracy vs. autocracy, origin of the legal system,6 centralised vs. decentralised structure, and security of property rights), and of course the incentives that bureaucrats and politicians face (including the effectiveness of law enforcement). There is widespread evidence that politicians abuse their power to increase their personal wealth (inter alia Menes 2006 for the US; Baland and Robinson 2006 for Chile; Goldstein and Udry 2008 for Ghana).7 The power structure is thus also an important determinant of corruption.

The evidence on the relationship between democracy and corruption is mixed. Ades and Di Tella (1999) and Fisman and Gatti (2002a) find no correlation between democracy or political rights and corruption perceptions. Treisman (2000) shows that a long exposure to democracy (but not its actual level) reduces perceived levels of corruption. Chowdhury (2004) finds press freedom and democracy to have a significant negative effect on corruption. Brunetti and Weder (2003) show that freedom of press significantly reduces corruption. Lederman et al. (2005) demonstrate that democracies, parliamentary systems, a free press, and a stable political system are all associated with lower corruption levels. Rock (2009) suggests that an inverted U-shape relationship exists between democracy and corruption, with the turning point to exit around ten to 12 years into democracy.8 Persson et al. (2003) study the role of electoral rules for corruption in democracies and find that open list electoral systems are less prone to corruption than those with closed (party) lists as in the latter system corrupt individuals cannot be held accountable. Corruption is larger in systems with smaller electoral districts as barriers to entry are larger.

One major institutional choice is the extent to which authority is centralised. Fisman and Gatti (2002a) and Arikan (2004) show that decentralised states are strongly and significantly associated with lower levels of corruption perceptions. Treisman (2000) finds the opposite—more strongly decentralised countries are perceived as more corrupt. He measures decentralisation with a dummy that portrays a decentralised constitution, whereas Fisman and Gatti (2002a) measure decentralisation by the share of expenditures of subnational tiers of government in total government expenditures. Both measures are less than perfect: A federalism dummy does not capture the actual distribution of competence between the centre and the subnational jurisdictions, which may be very different across states; the expenditure share of subnational jurisdictions does not measure the extent of actual authority over spending as expenditures of subnational governments can be mandated by central laws. Thus, in interpreting the results, it is important to recognise how exactly decentralisation is measured; results need not be stable with respect to the measure used. Fan et al. (2009) suggest that corruption increases with the number of tiers of the government, pointing towards uncoordinated corruption (‘overgrazing’). They argue that the higher the revenue share of local governments, the lower corruption.9 Focusing on local governments only and taking into account the degree of independence that they enjoy, Ivanyna and Shah (2010) find that decentralisation strongly reduces corruption. Their decentralisation index is based on de facto expenditure autonomy and tax autonomy at the local level.10

Income inequality has been shown to be positively correlated with corruption (Gyimah-Brempong 2002; Glaeser and Saks 2006), education to be negatively correlated (Glaeser and Saks 2006). Better education may empower individuals to resist extortion attempts as they may know the official regulations and possible complaint mechanisms better.

Salary levels of public bureaucrats have been surmised to affect corruption as higher paid officials have more to lose in case of detection. Van Rijckeghem and Weder (2001) show that perceived corruption levels are significantly negatively related to the relative pay of bureaucrats. Di Tella and Schargrodsky (2003) find that after a corruption crack down in Buenos Aires, when audit levels had come back to intermediate intensity, corruption was negatively associated with wage levels. Schulze et al. (2013) show that registered corruption in Russia declines with relative salaries, but at a diminishing rate, and increases after a turning point. This U-shape relationship turns out to be very robust with respect to the corruption measure used (registered incidents, convictions, corruption experiences) and the measures of relative salary levels.

The lack of competition in markets has been found to be associated with higher levels of corruption (Ades and Di Tella 1999). Countries pursuing active industrial policies experience higher levels of corruption (Ades and Di Tella 1997; Emerson 2006). Openness has been shown to reduce corruption (Krueger 1974; Ades and Di Tella 1999; Treisman 2000). This is interpreted as a sign that increased foreign competition reduces the rents in an economy and thus the incentives of bureaucrats to get a share. Moreover, more restrictive regulations increase the dealings with the government and thus the opportunities for government officials to extract bribe payments. Tavares (2007) argues that economic liberalisation is most successful in reducing perceived levels of corruption when it is accompanied by a concurrent democratisation process, while it might even increase corruption otherwise.

These determinants—institutions, the level of development and democracy, trade policies, and power structures—are endogenous themselves, and thus causality might run both ways, making the identification of causality a daunting task.11 The following determinants are (more) exogenous to corruption.

Ethnic divisions may lead to higher corruption and stronger patronage politics. Easterly and Levine (1997) show in a cross-country framework that the reduction in growth performance by ethnically divided societies is brought about largely by an increase in corruption. Olken (2006) shows that corruption is larger in ethnically divided villages (see also section 4.5).12

Gender has been frequently studied as a determinant of corruption. Dollar et al. (1999) show that a higher political representation of women is associated with lower corruption. Swamy et al. (2001) find that a higher participation of women in market work and in government is associated with lower corruption levels.13 Goetz (2007) and Sung (2003), however, argue that this relationship is not causal in the sense that women react differently to corruption, but rather that political systems that are more open for women are less prone to corruption.14 Vijayalakshmi (2008) does not find gender to have explanatory power for levels of corruption in the context of local governments in India.15

Resource rents have been found to increase corruption in countries that have poor institutions but not in economies with high quality institutions (Bhattacharyya and Hodler 2010). Vicente (2010) shows how the announced oil discoveries in Sao Tome and Principe (1997–99) led to a sharp increase in corruption, while in Cape Verde, an otherwise similar country, corruption continued to decline. Other papers on the ‘natural resource curse’ include van der Ploeg (2011) and Arezki and van der Ploeg (2011).

Social norms are an important determinant of corruption but, due to their elusive nature, hard to capture. A notable exception is Fisman and Miguel (2007), who show that the pattern of unpaid parking tickets by diplomats accredited at the UN in New York follows the pattern of corruption in their home countries: diplomats from more corrupt countries violated parking regulations in New York more often. However, social learning over time makes longer serving diplomats to become more similar in their behaviour—once unpunished, diplomats from low corruption countries also increase their violations. A similar result was found for undergraduate students in a corruption experiment by Barr and Serra (2010). By contrast, the relationship between corruption culture and student behaviour turned out to less clear-cut in the experiment of Cameron et al. (2009) (see also section 4.3).

While most of the analyses described above either use cross-country perception data or experimental data, a line of research has looked at firm characteristics. Svensson (2003) shows in a sample of Ugandan firms that firms that had extensive dealings with the bureaucracy had to pay more bribes (because they sold to the government or engaged in trade) as well as those with higher current or expected future profits. Firms with better outside options had to pay fewer bribes. Clarke and Xu (2004) show that in the utility sector of 21 transition economies, more profitable firms, newer firms, and those with overdue payments pay larger bribes.16

2.3 Consequences

The empirical evidence of the effects of corruption on economic activity is very clearly a bleak one. Corruption reduces growth significantly on the macro level (Mauro 1995, 1997; Gyimah-Brempong 2002, Johnson et al. 2011, see also above)17 but also on the firm level. Fisman and Svensson (2007) show for a sample of Ugandan firms that corruption is harmful to firm growth, much more than taxation is. Dal Bó and Rossi (2007) find for a sample of 80 electricity distribution firms from 13 Latin American countries for the years 1994 to 2001 that firms exposed to higher corruption levels are less efficient. Corruption of bank officials constrains firm growth (Beck et al. 2005). Khwaja and Mian (2005) show that politically connected firms in Pakistan received preferential loans from government banks, borrowed more as a consequence and displayed a 50 per cent higher default rate. The ensuing misallocation resulted in a 0.3 to 1.9 per cent GDP loss every year.

Corruption does not grease any wheels either: Hunt and Laszlo (2006) show for Peru and Uganda that services do not improve with a bribe, and that individuals refusing to pay a bribe are severely punished. A special inefficiency is reported by Bertrand et al. (2007) who demonstrate how bribes can secure drivers' licenses to people that do not know how to drive. Corruption also affects the composition of government spending: As corruption is easier in some expenditure categories, especially for non-standard high-tech products for which reference prices do not exist (Shleifer and Vishny 1993), corrupt governments' expenditures will shift accordingly. Education expenditures decline (Mauro 1998), military expenditures increase (Gupta et al. 2001). The capturing of public funds by officials for private gain reduces undoubtedly the quality of public services; it results in worse educational infrastructure and leads to worse educational outcomes (Reinikka and Svensson 2005; Ferraz et al. 2012). Moreover, higher corruption is associated with lower confidence in public institutions (Clausen et al. 2011).

2.4 Remedies

Proposals to fight corruption are of two—related—kinds. The first type of remedy is a change of institutions in such a way that bureaucrats have fewer entry points to extract rents through corruption. Rose-Ackerman (1978) suggested that competition between bureaucrats could drive down corruption payments. Likewise, simpler and fewer regulations reduce interactions with bureaucrats and their discretionary scope. For instance, simple tariffs are better than quotas and import licenses allocated by some bureaucrats who enjoy leeway in their decisions. No protection at all is of course first best. The same holds true for the tax system: simple tax systems that do not allow for special tax credits and exemptions are easier to administer, more predictable, and provide fewer entry points for extortion and corruption by tax officials. To strengthen democratic institutions and the rule of law in general, for instance through reforms of the juridical system and the police as well as through stricter financial auditing, is a promising approach.

The second type of remedy is to strengthen various accountability mechanisms so that the costs of corruption increase. Among those are better information transmission and better transparency through enhanced press coverage, increased participation, more stringent and more frequent audit mechanisms, better education of the populace, higher salaries for civil servants, and increased yardstick competition. While the first remedy type reduces the scope to extort bribe payments, the second reduces the incentives to do so.

One obvious way to reduce corruption is to increase competition in the goods market (Ades and di Tella 1999). One direct way to do so is to open up the country to foreign competition by reducing trade barriers. Another way is to increase transparency through publicly available information (cf. Lindstedt and Naurin 2010).

Reinikka and Svensson (2004) report on a large capitation grant to finance non-wage expenditures for schools in Uganda in 1991–95. They find that only 13 per cent of the funds actually reached the schools, with schools that had lower bargaining power, that is, those in poorer neighbourhoods, receiving an even lower share, and the lion share disappearing in the pockets of local officials. In reaction, the government published the monthly amount disbursed to the local officials in local newspapers. In 2001, 80 per cent of the funds reached their destination, with those schools receiving a higher share, which were closer to the nearest newspaper outlet and therefore had potentially better informed head teachers (Reinikka and Svensson 2005).18

Another large-scale experiment involving centralised audits and the increasing of transparency has been taking place in Brazil since 2003, where the federal government randomly audited municipalities and made the reports publicly available. By comparing electoral outcomes of municipalities audited before the 2004 elections with those that were audited after the elections and had the same level of corruption, Ferraz and Finan (2008) show that publishing the audit reports reduced the re-election chances of corrupt mayors. The effect was stronger in municipalities with local radio stations that disseminated the information. Moreover, electoral accountability mechanisms also affected the behaviour of mayors: those who did not yet reach their term limits and stood for re-election misappropriated considerably less resources than mayors in their last term, who were not facing re-election incentives (Ferraz and Finan 2011).

Björkman and Svensson (2009) show how an intervention of local non-government organisations (NGOs) in Uganda, training villagers to get involved in the healthcare provision, has led to better monitoring of health services by the communities, to larger utilisation of health facilities and to better health outcomes. However, a number of further studies also show that monitoring by service beneficiaries is not a panacea: it requires intensive training and presupposes both local empowerment and the overcoming of information and collective action problems in order to be successful. Banerjee et al. (2010) show for instance that a program that taught the local population to monitor teaching quality and outcomes in India did not improve teacher effort nor student outcomes as it failed to increase monitoring by the community. Even the more successful central audits can have unintended side effects: Cisneros et al. (2013) show that in municipalities of the Brazilian Amazon, deforestation, which was not directly observed by the auditors, increased considerably in the aftermath of randomised fiscal audits. Moreover, the adverse effects were concentrated in municipalities that were found especially corrupt by the auditors while their mayor was standing for re-election.

In the framework of this survey, we can hardly do justice to the many contributions that report on failed or successful attempts to curb corruption. Further recent contributions include, among others, Banerjee et al. (2012a), Duflo et al. (2012), Zitzewitz (2012), or Reinikka and Svensson (2011).

After addressing the main issues analysed by corruption research, we now shift the focus to look at the methodological challenges (section 3) and to give a broad overview over the contributions on corruption in Southeast Asia, organised by the method that they use (section 4).

3 Methodological challenges

  1. Top of page
  2. Abstract
  3. Introduction
  4. 2 Topics of corruption research
  5. 3 Methodological challenges
  6. 4 Empirical corruption analyses on Southeast Asia
  7. 5 Conclusion
  8. References

Quantitative empirical research has two major strengths. First, it allows for testing the real-life relevance of theoretical hypotheses. While it is relatively easy to formulate hypotheses on, say, particular drivers of corruption that are deemed important and to find some anecdote to back up that claim, without a methodologically sound empirical analysis, one cannot be sure at all whether this claim is justified. The anecdote may be a correct narrative for the specific circumstance it describes, but it may be specific to that particular circumstance and a generalisation of the importance of a particular driver for corruption may thus be misleading or outright wrong. A systematic quantitative approach, which carefully takes into account relevant characteristics of different situations that it analyses, allows for the falsification of such a claim and thus provides a far better foundation than just casual observation or even in-depth case studies.

Second, quantitative analyses, carefully designed, allow quantifying causal relationships. This is imperative for policy formation and implementation. For instance, if empirical observations suggest that corruption is fostered by low wages of civil servants, a possible course of action may be to raise salaries. By how much should they be raised? The answer to this question necessitates knowing the quantitative reaction of the actors to all such policy changes.19 In particular, an optimal policy response requires that the marginal costs of raising salaries are equated to the marginal gains in terms of reduced corruption. While the marginal costs of raising salaries are relatively easy to determine, the reduction in corruption levels is much harder to calculate. The reduction in corruption incidents then needs to be transformed into an economic gain. How much do we gain (in monetary units) if corruption incidents are reduced by X per cent? Again, that requires quantifying these relationships. Moreover, we need to check whether raising the salaries of the civil servants is the best instrument; it may well be that increased auditing is much more cost-effective in reducing corruption. But then, by how much should we increase auditing? This in turn requires knowing the marginal cost function for increased auditing as well as the marginal economic gain from reducing corruption through increased auditing. In other words, for a rational policy formation it is absolutely inevitable to identify and quantify the relevant relationships through sound empirical analyses.20

Yet, what is required of a methodologically sound empirical analysis?21

3.1 Representativeness

The sample used for the empirical analysis should be representative for the research question. The data should be randomly sampled from the relevant population in order to ensure representativeness. If the sample is biased, and that sample bias is unknown and uncorrected for, estimates will be biased. For instance, if the research question is the incidence and magnitude of corruption in manufacturing firms in a certain district and the interviewer selects larger firms more often than their share in the population of firms because they are easier to locate and it is easier to find an interviewee, the estimates will most likely be biased as larger firms typically have a lower burden of bribes relative to their revenues.22 Therefore, a correct sampling procedure is of pivotal importance for the validity of the results.23

Sometimes, specific groups are deliberately oversampled if their number would be too small to derive meaningful results if sampled randomly. For instance, if the research question is to identify and quantify the determinants of corruption for manufacturing firms such as size, location, ownership, sector, etc. in a given district, and there are only very few large firms in this district, the researcher may want to sample all of them to get reasonably precise estimates (see below). If not corrected for, analyses of such a biased sample will result in biased estimates. Therefore, such a choice based sampling needs to account for the biased sample through the use of sample weights. This requires knowing the population share of the oversampled subcategory.

A second source of bias arises from systematic measurement errors, which occur—also in randomly selected samples—if the data compiled are systematically biased. Such a bias may occur in interviews if the respondents have a systematic answering bias because they want to appear in a more favourable light or because they want to express their concern as expression of their public identity or for reasons of political correctness.24 For instance, they might be reluctant to admit to having bribed as this is illegal or deemed immoral. They might also deem it to be socially inacceptable to openly criticise government officials in front of foreign researchers (cf. Svensson 2005). These biases may be very context specific (influenced by culture, socioeconomic position and gender of the respondent, etc.) and may also depend on the person of the interviewer (interviewer bias). To reduce such answering biases, the questions must be framed accordingly. For instance, many questionnaires on corruption ask for corruption payments of a typical firm of the same size and in the same line of business as the respondent but not for the bribes made by the respondent. Moreover, government officials are unsuitable interviewers; locals working for NGOs, business associations, or universities may be more appropriate.

Similar biases may occur in national or regional accounts as they get manipulated. If, for example, corruption in illegal logging is analysed, the quantity logged may simply be misreported. As a consequence, the corruption tax for log exports may be overestimated as the quantity exported is understated. If there is corruption in the customs department to save duties or to evade capital controls, trade figures (volumes and prices) may be misreported (Schulze 2000).25 The public expenditure tracking analyses (see section 4.5) take into account the incentives to misreport budgetary figures.

The fact that people make mistakes in interviews or in public records etc. does not invalidate the results—it only makes them less precise.26 Only if people make systematic misrepresentations of the true data, the results will be biased and thus not valid.

Lastly, all important control variables need to be included in the regression. This requires to know what the relevant parameters of the situation are and to obtain the values of the variables. If important variables are not included in the regression equation, and if these variables are correlated with the error term, an omitted variable bias will be the consequence.27 For instance, assume that the determinants of corruption incidents at the firm level are analysed and that information on the ethnicity of the owner is not available. If minority ethnicity owners are targeted more often by corrupt officials, and the minority ethnicity is overrepresented in certain sectors, a regression model that leaves out the ethnicity of the owner will falsely capture the effect of this discrimination in the sector dummy. The conclusion that follows is that the specifics of that sector (rather than the ethnicity of the owner) lead to larger corruption payments, which is wrong and misleads the policy response. Since budgets are limited, researchers have to carefully weigh the sample size against the number of questions asked per unit of observation.28 Some of the relevant variables may be unobservable and thus appropriate econometric techniques, notably fixed effects panel regressions, have to be applied in order to address this issue.

3.2 Precision

A second requirement is that empirical results are sufficiently precise. For that, the sample size has to be sufficiently large. Statistical inference relies on the basic concepts of the law of large numbers and the central limit theorem. The law of large numbers states that the sample average of a randomly distributed variable converges in probability to its expected value as the number of draws from the population approaches infinity. The central limit theorem states that, as the number of draws (n) of this identically and independently distributed random variable with finite variance σ2 increases, the difference between sample mean and the expected value, blown up by the factor inline image, converges in distribution to a normal distribution with zero mean and variance σ2.29 In other words, large n implies precision and allows hypothesis testing based on asymptotic tests that are based on the normal distribution. That calls for a large sample rather than for a small number of observations.

Data precision is a recurring issue in many contexts. One context is again interview data: If questions are posed in a very general way such as ‘how do you rate the level of corruption in the mayor's office on a scale of 1 (none) to 5 (very high)’, precision will be quite low as the answer will depend on individuals' ideas what a high, medium, or low level of corruption would be. That idea may differ substantially across individuals, and it may differ systematically between different groups of the society as a common yardstick is missing (see Kaiser et al. 2006 and section 4.1). This may result in biased estimates. For instance, if the poor are more likely to regard higher levels of corruption as normal than others, they will state lower levels of corruption for the same factual corruption intensity, and the data—taken at face value—will show that (perceived) corruption is lower for the poor, while in fact it is not.

More precise questions such as ‘how much did you have to pay for an identity card?’ will give substantially more precise estimates of the corruption level, if reported prices are compared to official prices. The disadvantage is that these questions refer to one particular transaction only and may give a biased picture of the overall situation.30 Moreover, these questions mostly refer to petty corruption as standard transactions are experienced by a large number of people and thus are easily measurable for a large sample. They may not be the most important ones.

Other sources of imprecision are lacking knowledge on part of the respondents and recollection biases. This calls for timely interviews and questions that refer to contexts the respondents are familiar with.

3.3 Relevance and realism

The research design should address real situations; thus it should be realistic. Laboratory experiments have the merit that they take place in a controlled environment and thus can focus on one particular mechanism. As all relevant variables are under the control of the researcher, omitted variable biases cannot occur in a carefully designed lab experiment (see above). The downside of this approach is that lab experiments portray a hypothetical situation, and we cannot be sure that behaviour in laboratories carries over to real-world situations, especially when investigating behaviour with moral implications.31

Particularly useful are empirical analyses that investigate situations in which corruption potentially affects many people in significant ways, such as corruption in education, in health service provision, in public procurement, etc. Controlled field experiments are especially relevant; other very useful data can be gathered by analysing public records or private firms' accounts (see below).

3.4 Identification

Empirical analysis seeks to establish causal relationships. Yet, the fact that two variables are correlated does not establish causality. The explanatory variable under consideration could be endogenous. This would be the case, if confounding factors caused the correlation of these two variables without any causal relationship between them. Alternatively, causality could run both ways, thus making it difficult to establish to what extent a variation in X caused a variation in Y and converse.

In our context for example, we might seek to answer whether, and if so to what extent, a free press reduces corruption levels. District level data might show a negative correlation between the shares of people having access to newspapers and perceived corruption. Though suggestive, this does not establish causality. The share of newspaper readers might be endogenous; a possible confounding factor may be education levels. Education levels may influence readership size; at the same time, they may affect corruption levels as people have more means to resist extortion attempts, for instance because they are able to read regulations and prices for public services. Even in the (implausible) absence of any causal influence of newspaper circulation on corruption, there would be a correlation between these variables. Stated differently, the estimated correlation between corruption and newspaper circulation would not inform us to what extent it is causal as part of it could be driven by a confounding third factor, education. This hypothetical problem could be addressed by adding average education levels in the population as an additional control to the regression of corruption on media access.32 However, if the average years of education proxy only imperfectly for the quality of human capital in the population, some omitted variable bias will still persist.

Reverse causality may occur in our example if newspapers circulation reduces actual corruption levels by exposing them; but as perceived corruption is measured, the exposure of corruption may lead to an increase in perception about corruption, other things being equal. As a consequence, the effect of newspaper circulation on corruption may be underestimated.

This has led to a number of econometric techniques that can be used to establish causality.33 Instrumental variables help to extract the exogenous part of the endogenous explanatory variable by regressing it first on variables that are uncorrelated to the error term and affect the endogenous variable only through the variable they instrument for. Here, the difficulty lies in finding appropriate instruments that are econometrically valid and strong and make theoretically sense (cf. Angrist and Krueger 2001). For instance, if media access is partly explained by some topological factors that influence the ease with which a radio or television signal can travel across space (hilliness or forest cover), these factors can be used as instruments for radio or television ownership. Such a strategy would rely on only the part of the variation in television or radio access that can be explained by the topological conditions (and other exogenous factors), and hence would be unaffected by issues of reverse causality or omitted variable bias. Instrument validity is a major issue here: if topological conditions not only affect media access but also institutional quality, for instance by reducing the accessibility of the region and hence the strength of central control, they will also have a direct effect on corruption. If that were the case, topological factors would be invalid instruments that once again led to biased (overestimated) results: media access instrumented by topology would not only capture the effects of media control but also the strength of the central control.

A second approach is to find units of observation that are very similar except for the variable of interest and to compare the value of the endogenous variable. These matching approaches are an established procedure in this line of research. These approaches rely on constructing a control group that resembles the group in all relevant characteristics except for the variable to be studied. For instance, if ethnicity were suspected to influence the corruption in teachers' absenteeism, a specific form of corruption, one could find for each individual from ethnicity A a set of individuals from ethnicity B who had very similar characteristics except for the ethnicity. The problem here is to identify all relevant characteristics (so that individuals differ in their ethnicity and nothing else) and to only focus on individuals for who close matches exist.

Comparable observations can also be identified based on regression discontinuity approaches, which have become very popular in the economic literature. For instance, one could compare corruption outcomes in villages and municipalities that are just receiving transmission signals of television or radio stations with those that are just outside the reach of these signals. This is a very data hungry method: it would require observations from a relatively large number of geographically close municipalities that lie just within/outside the transmission range. The identifying assumption is then that the villages are structurally similar except for the fact that one group receives the television/radio signal and the other does not (implying that no other village characteristic changes discontinuously at the transmission threshold).

A third valuable, and potentially the most rigorous approach, is to set up controlled field experiments in which treatment and control groups are randomly selected. For example, the effect of newspaper circulation on corruption levels could be investigated by giving out free newspapers to one group of villages but not to the other. The identifying assumption is that both sets of villages are initially structurally very similar. Such an exogenous variation in newspaper access could then identify the effect of increased newspaper circulation on corruption. The difficulty with this approach is that field experiments are typically resource intensive and can be carried out only by researchers affiliated with international organisations, governments, large NGOs, etc. In addition, it is often not entirely clear to what extent the effect of the specific intervention is context specific and whether it is representative for a range of similar or related interventions. While the internal validity of a well-executed experiment is by definition very high (and hence it yields a truly causal effect for the given sample and application), the external validity of such results, that is, their applicability to other contexts can never be simply assumed but has to be carefully considered on a case-by-case basis. For instance, the success of a small-scale anti-corruption beneficiary training implemented by a well-organised NGO might not be reproducible by a large government organisation, which might lack credibility, the political backing of the affected unions, skills, or motivated personnel to implement the same strategy.34

Related to the third approach are natural experiments that use an exogenous variation in variables that are hypothesised to affect corruption levels. For instance, if the existence of natural resource rents is thought to increase corruption, a natural experiment would be to compare the development of corruption levels of countries before and after they found natural resources with those countries that are comparable but did not find natural resources. An example for such a difference in difference strategy is Vicente (2010).

4 Empirical corruption analyses on Southeast Asia

  1. Top of page
  2. Abstract
  3. Introduction
  4. 2 Topics of corruption research
  5. 3 Methodological challenges
  6. 4 Empirical corruption analyses on Southeast Asia
  7. 5 Conclusion
  8. References

In what follows, we will report on the empirical analyses of corruption in Southeast Asia, sorted by the methods they use.35 We will comment on the relative strengths and weaknesses of these approaches and make reference to the most important contributions in order to relate the papers surveyed adequately to the literature. Before we turn to the empirical approaches that use regression techniques or experiments, however, we note a number of very insightful contributions which, while not carrying out formal empirical analyses themselves, are informed by empirical evidence and provide an insightful picture of corruption in Southeast Asia.

Hill (2012) provides an overview over the corruption landscape in Indonesia. He argues that corruption is due to poor civil service remuneration, complex regulatory environment, low risk of detection despite the establishment of the corruption eradication commission, and permissive social norms towards corruption. He argues that growth and corruption could co-exist under Soeharto's rule because of an otherwise relatively prudent macroeconomic policy, that is, significant investment in education and infrastructure, non-excessive taxation as the state was partly financed by oil and gas revenues and a relatively open economy, and a stable—albeit repressive—political environment. (Of course, growth could have been much higher in absence of corruption.) He argues that after democratisation and decentralisation, corruption has become more decentralised and less organised, but no less intense.

McLeod (2008) shows how the Soeharto regime forced the bureaucrats to become part of a corrupt system in Indonesia—‘the Soeharto franchise'—by setting their salaries strongly below market levels and thereby forcing them to supplement their income through illegal means. At the same time, ‘wet’ positions that allowed extraction of illegal payments were scarce, and lower levels of bureaucracy were overstaffed. Competition for those ‘wet’ positions was resolved by selling them to the highest bidder, which in turn put pressure on successful candidates to recoup their expenses after being promoted. The Judiciary was brought in line by denying it to review laws for their constitutionality, by receiving salaries below market levels, and by assigning the personnel management to the executive. Judges, ruling against the franchise (the first family, cronies, state-owned enterprises, large foreign firms) were repositioned to unattractive posts, not promoted, and overruled by higher courts. That led to corruptibility of many judges who supplemented their comparatively low income illegally. As a consequence, corruption is deeply engrained in the bureaucracy, and since performance was not the dominant promotion criterion, civil service lacks competence and skills. Thus, civil sector reform is indispensable for eradicating corruption and promoting prudent economic policy formulation and implementation (McLeod 2005, 2006).

Thailand, even though relatively poor on econometric studies, has been subject of a host of qualitative analyses on governance and corruption. The seminal book on corruption in Thailand is by Phongpaichit and Piriyarangsan (1996); a more recent study on corruption and patronage is by Phongpaichit and Baker (2004). Mutebi (2008) argues that corruption in Thailand could thrive despite an existing comprehensive anticorruption mechanism because a ‘powerful business-politics nexus’ could capture key regulatory processes and because contestability was low as Premier Thaksin held a comfortable majority in legislature and controlled the executive.

Quah (2001) describes the Singapore success story: Corruption was virtually eradicated through a substantial increase in civil service salaries and other accompanying measures.

There is a host of narratives on corruption in Southeast Asia, which is impossible to do justice to. Guggenheim (2012) reports on corruption in research in Indonesia with large kickbacks to the contract-rewarding agencies, Butt (2011) documents the various attempts to make the corruption eradication commission less effective, and Crouch (2010, ch. 6) describes the struggle to make the judiciary less corrupt (job buying in the judiciary and buying of court rulings).

We now turn to the empirical analyses that use econometric techniques to address one of the issues laid out in section 2. We begin with the most obvious and oldest approach, the measurement of perceived corruption.

4.1 Perception-based analyses

The most classical approach to corruption measurement relies on expert opinions or various national surveys of corruption perceptions and produces aggregate, mostly country-level corruption indices. These are primarily used to inform foreign businesses and policymakers on the overall extent of corruption in any given country. Two currently most widely used indices are the CPI of Transparency International and the Control of Corruption index from the World Bank Governance Indicators database.36 Both are constructed as composite indices, based on a relatively large number of perception-based surveys. The Indonesian branch of Transparency International has put out a regional corruption perception index for the largest 50 cities.37 These indices are used for the ranking of various countries or regions in terms of their perceived levels of corruption. Their main advantages lie in their large country coverage, the relatively constant methodology over time, and their easy accessibility. Due to their merits, they have been widely used in cross-country studies for the economic analysis of the causes and consequences of corruption.38 Corruption perceptions have been used in subnational studies as well. These studies have the distinct advantage that the institutional setup across the units of observation is very similar within a country and thus a lot of—potentially unobserved—factors that create heterogeneity can be excluded as a source of omitted variable bias (such as the legal system, general attitudes towards corruption, etc.).

Azfar and Gurgur (2008) measure the effect of perceived corruption on a battery of health outcomes in 80 municipalities in the Philippines. They use household level data for corruption perceptions as well as government officials' perceptions of corruption and relate these perceptions to health outcomes either on the municipal level (for immunisation rates as dependent variable) or on the household level (for household satisfaction with health services, average waiting time for immunisation of children, access to public health clinics, etc.). They find a significant negative relation between perceived corruption and various measures of health outcomes. The study uses cross-section data as the survey was carried out only once in 2000, which makes it hard to establish causality. Moreover, a number of other control variables (and some of the dependent variables) are perception based as well so that possible endogeneity of regressors make causal interpretation difficult.39

Hofman et al. (2009) analyse the pattern of corruption in decentralised Indonesia using the Governance and Decentralization Survey (GDS) 2003, which was fielded in 150 randomly selected districts (out of 348 at that time). They show that decentralisation has not reduced corruption and that the level of perceived corruption differs strongly between institutions, law enforcement being the worst. They show in a regression analysis that districts with a higher ethnic fragmentation and those with a higher political fragmentation exhibit higher levels of perceived corruption; districts with higher revenues from local taxes have less corruption. Districts in Java have less corruption, those in Papua more.

Suryadarma (2012) uses the district-level CPI of Transparency International in Indonesia to study whether spending for education has a positive effect on school outcomes. In a pooled ordinary least squares (OLS) regression framework at the district level, he regresses separately junior and secondary school enrolment rates and test scores on spending for education, the CPI, and an interaction between private spending and the CPI. He finds that additional spending has a positive effect on enrolment only for regions with low corruption40 and that spending has no effect on test scores. Due to limited availability of data, his data set contains only a total of 61 observations in two rounds. His data do not allow him to address issues of reverse causality, and the corruption perception does not refer to schools in particular but to perceptions of businessmen of the general corruption climate.

The major drawback of this perception-based approach lies in the subjectivity of corruption perceptions, be they formed by experts or by the surveyed population. Perceptions about what types of behaviour can be considered corrupt do vary considerably across countries but also within countries, albeit to a lesser extent, and strongly reflect cultural norms and individual attitudes (Bertrand and Mullainathan 2001). Reported corruption perceptions might be for instance affected by surveyor biases or by the strength of the norm of not speaking ill about one's own government (see above).41 They might also reflect different levels of information.42 Perceptions also strongly depend on the overall quality of institutions. For instance, a free press is more likely to expose and report extensively on corruption cases, increasing, everything else equal, the perceived extent of corruption. Moreover, while the public is well aware of corruption encountered in everyday life (like bribe payments to public officials or doctors), it might be less able to assess the true extent of graft, misuse of political office, or outright theft from public accounts.43

Studies comparing corruption perceptions of the public with the real extent of corruption do indeed find substantial discrepancies between perceptions and reality. Mocan (2008) documents a clearly non-linear (concave) relationship between the CPI index and average individual corruption experiences in 20 countries (cf. section 4.2). Donchev and Ujhelyi (2011) also compare the correlates of corruption perception indices with the correlates of average survey-based corruption experiences (measured by individual crime victim and enterprise surveys) in a cross-country setting. They find that average corruption perceptions are only weakly correlated with average corruption experience. Moreover, for the same level of average corruption experience, corruption perceptions are more favourable in more democratic, wealthier and traditionally Protestant countries. These factors thus bias corruption perceptions downwards from the actual level of corruption. Fisman and Miguel (2007) compare actual behaviour of public officials (diplomats originating from a given country) with average corruption perceptions in an ingenuous way: they relate the frequency of unpaid parking tickets of UN diplomats in New York to their host country's perceived corruption levels. They find a clear positive correlation between perceptions and behaviour, indicating that the overall corruption culture in the home country does affect (although of course does not explain entirely) differences in the behaviour of diplomats abroad.

Micro-level comparisons of perceptions with actual measures of corruption can shed additional light on the mechanisms causing corruption perceptions and reality to differ. Olken (2007) evaluates the relative effectiveness of different ways of monitoring public road building projects (by central audits vs. by the project beneficiaries) in a controlled field experiment in Indonesian villages. In the experiment, external auditors measured theft of building materials, unpaid wages, the extent of overpricing, and false invoicing. Comparing the fairly precise assessments of theft by the auditors with survey-based evidence on the villagers' perceptions about theft and corruption in the village projects, Olken (2009) shows that, once again, corruption perceptions and reality are only weakly correlated. Although villagers were more likely to suspect corruption in the road project in those villages where auditors also identified more theft, for a 1 per cent increase in missing expenditures, the likelihood of reporting corruption increased only by 0.08 per cent. This is far below the unitary elasticity one would expect if villagers were perfectly able to assess the real extent of corruption. More detailed analysis of the correlations between perceptions and missing expenditures also showed that villagers were substantially more likely to detect corruption in terms of prices (over-invoicing or unpaid wages) than in terms of quantities (material theft). This corroborates the argument that perceptions can better reflect reality in cases when individuals are better qualified to make judgments and observe the outcomes more directly: when wages are not paid, this will be easily detected by the villagers, whereas it requires much more expertise to detect when officials are cheating by reducing road quality. Moreover, Olken (2009) shows that corruption perceptions are systematically related to individual characteristics. Exploiting only within-village variation, and hence holding the corruption environment constant, he shows that individual biases are very substantial in explaining corruption perceptions: More educated, wealthier and male villagers reported considerably more corruption, while individuals more closely involved with the project reported less corruption. The statistical analysis also suggests that these differences cannot be explained by misreporting only, but that they also reflect true differences in perceptions.

The subjectivity and context specificity of corruption perceptions renders them less appropriate for causal or comparative analysis. Their inherent limitations cannot be easily addressed and hence, newer strands of the literature started to search for other ways to measure corruption.

4.2 Survey-based evidence of corruption experiences

Survey data on the actual experience with corrupt officials or corruption payments offers a somewhat more objective way to assess the extent of corruption. Instead of measuring corruption perceptions in general, it focuses more strongly on individual or firm experiences with bribes and other illegal payments. The World Bank Enterprise Surveys (WBES) routinely ask firms whether they had to make informal gifts/payments when applying for permits, licenses, water, or telephone connections.44 The standardised International Crime Victims Surveys (ICVS) ask individual respondents whether they had to pay bribes to any government official (customs/tax/police officers, municipal officials, teachers, doctors, etc.) during the last year.45 The ICVS data have been used for instance by Mocan (2008) to relate direct experience with corruption of 55,000 individuals in 30 countries to individual and country characteristics. He finds that young people, men, people living in large cities and those with higher socioeconomic status report more frequently having paid bribes.

By referring to specific experiences instead of to general views and opinions, such questionnaires are likely to suffer less from perception biases. However, they are still prone to the usual errors of self-reporting as the surveyed might prefer not to reveal the truth. Respondents in firm-level surveys might be reticent to answer sensitive information. For instance Clausen et al. (2010) find that about 13 per cent of respondents in a World Bank firm-level survey in Nigeria are reticent and much less likely to admit to sensitive acts if questions are worded in a way that implicates personal wrongdoing. Jensen et al. (2010) find that non-response is higher in countries with less press freedom. Azfar and Murrell (2009) suggest a method of identifying reticent respondents and improving accuracy of survey data. They find for a sample of Romanian company officials that 10 per cent of the respondents are reticent.

While fairly general questions about whether bribes or informal gifts have been paid in a specific situation are usually considered less sensitive and are more likely to be answered truthfully, it is hard to ensure the quality of more sensitive information, for instance on the actual amount of bribes. Due to these limitations, the ICVS or the Global Corruption Barometer by Transparency International only ask whether a person has paid a bribe to a government official during the last year but do not ask for the actual amount. Thus, they only measure the overall prevalence of bribe taking in the country, but neither its magnitude nor the prevalence of other types of corrupt activities. Yet, if these questions can be asked in a situation free of mutual suspicions, they reveal interesting information about the sheer extent of corruption in the economy. The World Bank Enterprise Surveys try to get even closer estimates of the magnitude of bribes by additionally asking what share of a contract value a ‘typical establishment’ in the industry would have to pay to government officials, or what an establishment ‘like this one’ would have to pay in form of informal gifts. By not asking for specific experiences here but referring to the average firm, these questionnaires try to enhance truth revealing behaviour.46

Kuncoro (2004) uses a Special Survey on Governance conducted by the Institute of Economic and Social Research at the University of Indonesia (LPEM-FEUI) that sampled 64 out of around 300 districts at that time, districts with stronger presence of manufacturing firms being oversampled. This survey asks for the bribes paid in connection with licensing, taxation, etc., time spent with bureaucrats, and the efficacy of bribing, that is, whether bribes help to achieve the promised services. As many firms report not to have paid bribes at all,47 he uses OLS, Tobit, and Heckman selection models and finds that bribes as a share of production costs are larger for high tax payers, for firms in oil-rich regions, and for younger firms. Smaller firms pay more bribes as a share of production costs than larger ones, and more bribes are paid by firms that perceive a higher regulatory burden and spend more time with bureaucrats. Since a number of regressors are endogenous (tax payments, time spent with officials, regulatory burden, bribe efficacy), Kuncoro's analysis reveals interesting associations, but it is not a causal analysis in the strict sense of the term.

McCulloch et al. (2010) use data from the Rural Investment Climate Survey that surveyed 2500 non-farm small and micro-enterprises in rural Indonesia. As licensing is not enforced, they analyse the determinants of formalisation and show that older and larger firms and those paying more taxes and ‘other levies’ are more likely to be formalised (that is, hold business licenses). After controlling for the endogeneity of the formal status in an instrumental variable approach, however, they show that formalisation reduces corruption and tax payments—formality thus curbs extortion by government officials. They also find that larger firms (as measured by fixed assets, sales, and employment) pay more bribes and that the small and micro-enterprises that have older or female managers/owners and those with managers/owners residing in the village where the firm is located are less subject to extortion.

Henderson and Kuncoro (2011) provide a nice example for context-specific interview techniques. For their analysis on the effects of political decentralisation on firms' corruption payments in Indonesia, they instructed surveyors to engage in the Indonesian ‘conversation among friends’, giving many examples of various ‘gifts’ that firms might be paying to government officials in various situations. It turned out that while firms were unlikely to reveal exact bribe amounts, they were quite comfortable with estimating what fraction of their total costs was devoted to ‘smoothing business operations’. Henderson and Kuncoro use a pooled sample of 1862 firms in 37 districts in Java created by two surveys in 2001 and 2004 (with little overlap of firms) and regress the bribes as a share of total costs on firm characteristics (size and export) and the vote share of the main secular parties (Golkar and PDIP) in the district council elections of 1999. They show that in districts with a high share of secular parties, the corruption level was initially lower (although not significantly so) but decreased less than in districts with a stronger Islamic party representation. Based on a cross-sectional sample of 2,474 firms from 87 districts of Java (collected in 2004), they show that districts with a larger vote share of Islamic parties in 1999 experienced lower levels of corruption. They use Tobit specifications in both setups as a substantial share of firms (more than a quarter in the first setup and more than a third in the second) report not to have paid bribes at all; they also use instrumental variables approaches to account for a possible endogeneity of the 1999 vote share.

Rand and Tarp (2012) analyse the effects of corruption and formalisation on firm growth in Vietnam using two rounds of a survey of small and medium-sized enterprises (from 2005 and 2007) with 1,661 SMEs in a balanced panel. They find that larger, more formal, and more profitable firms as well as those with a higher capital-labour ratio and with more interactions with the government (inspections, government assistance, etc.) are more likely to pay bribes. Bribes harm firm growth, formality increases bribes, but the growth-reducing effect of these additional bribes is outweighed by a positive growth effect of formality as such.

Overall, corruption experience surveys seem to be more effective in revealing the frequency of corruption payments and potentially less effective in accurately determining their size. They potentially suffer from response biases but provide considerably more objective information than purely perception-based approaches. By their nature, they only cover those forms of corruption that affect firms and citizens on a regular basis, most notably bribe payments, while being less useful for quantifying theft or the economic value of political connections.

4.3 Experimental evidence

The economic analysis of corruption has only recently started to incorporate experimental insights on individual or group behaviour. The majority of this evidence comes from computer laboratory settings and investigates determinants of attitudes towards corruption by examining individual behaviour in public goods or bribing games.48 The analysed settings are usually framed to invoke the experience of actual corruption events. Even though played under artificial circumstances, the games provide considerable economic incentives to engage in corruption as the participants win actual money by opting for giving or taking bribes, while players involved in monitoring incur real costs. Although behaviour in laboratory experiments does not necessarily correspond to real-life behaviour, differences in behaviour across subjects can shed light on their different attitudes or beliefs about the social acceptability of corruption.

Cameron et al. (2009) set up a one-shot bribe game between a corrupt official, a firm, and a citizen who can punish corrupt behaviour. They play the same game with university students from India, Indonesia, Singapore, and Australia to analyse the role of cultural factors in explaining the propensity to engage in corrupt behaviour as well as the propensity to punish it. Surprisingly, they find Indonesian students to be significantly less likely to engage in bribe payments and more likely to punish bribing behaviour than their Singaporean counterparts, although Singapore is overall ranked as one of the least corrupt countries, while corruption is a common phenomenon in Indonesia. They attribute this difference to the saliency of the issue of corruption in current Indonesian society. This is also supported by the post-game survey: when asked to explain their behaviour, Indonesian participants were significantly more likely to cite moral arguments. Alatas et al. (2009) repeat the experiment with Indonesian public servants and compare the results with the students' behaviour. Once again surprisingly, they find that public officials are considerably less likely than students to engage in corrupt behaviour in a game. Indonesian public servants seem to be even more sensitised towards the illegality of corrupt behaviour than Indonesian students are. These results convey interesting information about overall—expressed—attitudes, yet they are derived in a hypothetical laboratory experiment (even if they involve real money). They do not imply that the same public officials would not accept corruption payments in a real-life setting.49

As opposed to the somewhat artificial nature of computer lab experiments, experimental or quasi-experimental evidence from real-life settings quantifies corrupt behaviour by directly observing it. Existing evidence is restricted to a handful of ingenious field experiments, all of which strive for nearly random sampling or rely on observations from a full population, and hence are well suited for quantitative inference. Olken and Barron (2009) investigate the amount of bribes that truck drivers have to pay if their truck is found to be overweight at a weighting point located along the two main transport routes in Aceh. They employ undercover surveyors, who accompanied the drivers on their routes and recorded detailed information about the circumstances of the encounter with the public officials as well as the amount of the bribe payments.50 They observe that drivers change their bribe offers from checkpoint to checkpoint and that prices vary considerably within the same trip. The amount of bribe payments increases with the bargaining power of the officials (number of officials at the checkpoint and the visibility of guns), even after controlling for trip and checkpoint-direction-month fixed effects. Bribe payments also increase with the value of the cargo and the age of the trucks, which shows that officials are successful at price discrimination. Olken and Barron (2009) also show that an exogenous decrease in the number of checkpoints in Aceh (due to the withdrawal of the military from the province) has lead to an increase in the amount of bribes paid in the neighbouring province of North Sumatra, which is in line with what the economic theory of endogenous price determination would predict.51

Both computer lab and field experiments offer very specific evidence by their nature. The evidence of computer lab experiments is restricted to the experimental subjects, although variations in the subject pool can yield interesting insights. Moreover, even though capturing overall attitudes to the social acceptability of corruption fairly precisely, these studies cannot claim to describe true behaviour outside of the laboratory setting. The external validity of the evidence from observational field experiments is much higher, but it relies crucially on their ‘undercover’ nature. The ‘invisibility’ of the observers is very important here as it allows them to observe real-life corruption events. Despite the benefit of the much richer description of the mechanisms of corruption, observational field experiments are always specific to the economic context and are better suited to assess everyday and smaller scale forms of corruption.

4.4 Inference from market information

Another promising approach to quantify the extent of corruption in an economy is to measure the value of political connections of firms using market information. The underlying assumption is that if firms receive economic benefits from their political connections in the form of licenses, public contracts, better access to credit, or market protection, efficient financial markets should price in these benefits, which thus should be reflected in those firms' market valuation. A comparison of the market valuation of firms with and without political ties should thus shed light on the value of political connections. Yet, only event studies can establish causality in a rigorous manner. From observing a higher market valuation of politically connected firms, it is yet unclear whether politicians merely prefer to sit on the board or hold shares of the best performing firms, or whether firms also directly benefit from such political ties.

In her pioneering cross-country study, Faccio (2006) analyses the stock market performance of more than 20 thousand publicly traded firms from 47 countries. She defines firms as politically connected if their large shareholders or members of the management also hold a major political office or are related to a top politician. She finds that globally about 8 per cent of the market capitalisation is represented by politically connected firms. In her event study, she finds considerable improvements in a firm's market valuation if large shareholders or company officers enter politics, while politicians entering the board do not yield significantly positive returns. Cheung et al. (2012) investigate a global sample of 166 large bribery cases that were discovered later on, and quantify the effect of bribe payments on the firm value at the time when the preferential contracts were awarded to the firms in question. They find sizeable economic rents to bribe payments (in a magnitude of 10 dollars of additional benefit for 1 dollar of bribes paid) and that bribe amounts vary with the rank of the official who is taking the bribe and the overall performance of the firm, while the size of benefits is unrelated to characteristics of firms or officials. The major drawback of this approach is that focusing on revealed cases only is subject to selectivity biases as the likelihood of getting discovered depends on officials' and firms' characteristics.52

The seminal event study on the value of political connections is situated in Southeast Asia. Fisman (2001) identifies the extent of economic benefits of political connections in Soeharto's Indonesia by investigating the effects of unexpected news about the dictator's deteriorating health on the relative market valuation of firms connected to the Soeharto family. The study measures the extent of political connections of the various firms listed on the Jakarta Stock Exchange based on expert opinions from Indonesian business consultancies on how strongly each firm's profitability depends on its political ties. Using this approach, Fisman finds very sizable reactions to the news about Soeharto's bad health among the firms with the strongest political connections. Building on Fisman's approach, Leuz and Oberholzer-Gee (2006) show that politically connected Indonesian firms were less likely to seek financing in the global capital markets than their less connected counterparts. Moreover, firms with ties to the Soeharto family suffered relative losses during the 1997/98 crisis and under the Wahid government, but not under the more Soeharto-friendly Habibie regime.

Analysing the stock market performance of Malaysian firms, Johnson and Mitton (2003) show that politically connected firms experienced larger losses during the 1997/98 crisis but also experienced larger gains once the government introduced capital controls in 1998. They show that 32 per cent of the gain in valuations of firms connected to PM Mahathir after the imposition of capital controls can be attributed to an increase in the value of political connections. These connections contributed 17 per cent to the overall market valuation of these firms in September 1998.

As all the above examples show, market information combined with a thorough mapping of the major political and economic players can be very successfully used to measure the extent of cronyism in an economy. The quantitative results do rely on extensive knowledge of the political economic structure of the analysed countries, but the applied methods are easily transferable between different contexts.

4.5 Expenditure tracking

Administrative corruption in the form of theft or leakage of funds can be identified directly by comparing budget accounts with the realisations of specific projects. Olken and Pande (2011) label this approach ‘estimation by subtraction’ since it subtracts missing funds or the estimated cost of missing project outputs from the claimed financial expenses and thus quantifies the amount of funds not accounted for.

While expenditure claims can be relatively easily inferred from administrative data, the actual financial outcomes are generally assessed either based on official audit information or on specialised surveys. Since central audit information is rarely available,53 representative public expenditure tracking surveys (PETS) or quantitative service delivery surveys (QSDS) can be used to estimate the realised extent of public service provision. Public expenditure tracking surveys are designed to uncover how much of the centrally designated funds actually reach the different administrative tiers and especially their final destinations like schools or hospitals (Reinikka and Svensson 2006). They thus measure leakage taking into account the incentives to misreport. The quantitative service delivery surveys focus less on monetary transactions and scrutinise more closely the actual results of service delivery in terms of the quality of public facilities, the number of public personnel, and user rates. They can also be used to assess absenteeism through unannounced facility visits. When comparing with officially planned expenditures, both survey types can help to infer the amount of leakage in the system of public finances. Such surveys yield better measures of corruption since they are less prone to perception biases (Reinikka and Svensson 2006).

The use of such expenditure tracking methods has been first demonstrated in the seminal studies of Reinikka and Svensson (2004, 2005) on misappropriated education expenditures in Uganda in the mid-1990s, which found that 87 per cent of capitation grants to primary schools never reached the schools themselves.

Olken (2006) provides a nice example of this approach by estimating the loss of public funds in a rice subsidy program in Indonesia. He analyses the ‘special market operation’ program (Operasi Pasar Khusus, OPK), which distributed rice to poor Indonesian households in the years following the 1997/98 crisis at a highly subsidised rate. He compares figures on the yearly amount of rice allotted for distribution to village households from district warehouses with survey evidence on the amount of subsidised rice consumed in the same villages, and finds that at least 18 per cent of the allotted rice did not reach the households at all. In this study, information on whether households received subsidised rice comes from well-established household survey instruments.54 Although it is conceivable that households under-report their consumption of subsidised rice in such surveys, there is no reason to expect under-reporting to be systematically concentrated in few villages only, which is exactly what Olken observes in reality. It is also unlikely that the low reported consumption of subsidised rice resulted simply from interviewer effects (for instance due to laxer interview procedures in some villages) as the data do not seem to suffer from misreporting with respect to other variables (for example, educational attainment) as compared to the census. The study shows that especially villages with high-ethnic heterogeneity tend to experience more corruption, which is in accord with the findings of the early cross-country literature on corruption perceptions (Mauro 1995, La Porta et al. 1999).

Mehta and Jha (2012) study theft in a program for subsidised rice in the Philippines in 2006 run by the Philippines National Food Authority. They compare the expenditures for subsidised rice from a Family Income Expenditure Survey with the allocations of subsidised rice for 13 regions and find that almost half of the allocations got stolen; however those regions that were allocated more rice had a higher share of stolen rise pointing towards a very high marginal pilferage rate. The analysis is hampered by the very few observations in a one-time cross-section analysis.

Olken (2007) reveals theft in road building projects in Indonesia by combining expenditure tracking and direct measurement of corruption within a field experiment. He sets up a fully randomised field experiment in order to compare the effectiveness of central vs. grass-root monitoring in building local village roads within the Kecamatan Development Project village development program. He quantifies material theft by digging samples of every newly built road, estimating the value of materials used for building it and comparing his estimates with the value of materials stated on the village accounts. He combines his estimates on missing materials with survey information on local prices (in order to detect overpricing) and on local wages and participation in local public works (to detect overstating wages). Based on the collected local price and wage information and the estimates of material usage, Olken produces an independent assessment of the project expenditures. He finds that nearly 25 per cent of the total expenditures on village accounts are missing in reality, and can thus be attributed to corruption. When addressing the effects of external monitoring on corruption, he finds that village officials who expected the visit of external auditors reduced material theft (by about one third) but did not adjust reported prices. This is in line with the expectation that corruption should occur in less easily detectable forms, that is, in form of material theft rather than in form of providing false information on market prices, which are more readily observable. Moreover, he finds that invitations to community meetings increased the local citizens' involvement in the discussion process about a public road building project but did not succeed in reducing material theft or overpricing. They only reduced over-billing of labour expenses; this is not surprising as the local population is both better informed about their own labour inputs and most willing to monitor proper payments for labour.55

4.6 Law enforcement data

Corruption can be assessed by using law enforcement data. Most approaches use corruption convictions as endogenous variable. As legal systems and zeal of law enforcement may differ substantially across jurisdictions such approach is only sensible for subnational analyses for countries with centralised and sufficiently unbiased police force and judiciary. Studies include inter alia Glaeser and Saks (2006) and Alt and Lassen (2013) for the USA, and Schulze, Sjahrir and Zakharov (2013) for Russia. There is no study on Southeast Asia yet.

4.7 Administrative overspending

A phenomenon closely related to corruption is administrative overspending, which is often used to provide additional ‘perks’ to bureaucrats such as luxurious offices, official cars, or expensive travels. The concept does not only refer to clearly illegal activities but to all forms of public spending that are excessively directed to the own needs of bureaucrats/politicians and do not benefit the public. Administrative overspending can be assessed by analysing public accounts but identifying it is far from easy. The general challenge is to identify which part of the administrative spending is necessary and warranted and what can be considered as overspending. While making this distinction would be clearly very difficult in a national context, analysing the regional variation of the spending patterns of local governments can help to trace common characteristics of big spenders.

Sjahrir et al. (2013a) address this issue by analysing the budgets of Indonesian local governments since the administrative decentralisation of 2001. They focus on the costs of the administrative overhead. While on average, districts spend about 30 per cent of their budget for administrative purposes, in some districts up to 60 per cent of total expenditures is used for administration.56 The quantitative analysis regresses real per capita district expenditures on fiscal size, local GDP per capita, measures of geographic location, and remoteness as well as further controls of the socioeconomic environment in any district.

The results show that more educated and better informed local populations tend to put some constraints on administrative spending. Administrative expenditures are ceteris paribus lower in districts that have a better educated population (as measured by literacy rates) as well as a better informed population (proxied by local access to media outlets). The process of pemekaran, the splitting up of districts after decentralisation, has led to considerable increases in the share of administrative expenditures, especially in the first few years after the splitting up. This increase is natural to some extent as newly established district administrations have to invest more heavily into a new physical infrastructure (building new offices, buying cars, etc.).

The results also document the presence of an election cycle in administrative expenditures: since local elections happen at different times, and only after the incumbent district head served a full term, this exogenous variation in timing can be used to identify election cycle effects. The results show that in the election year, there is a shift from capital and goods expenditures to expenditures on ‘other’, not clearly specified spending categories. These unspecified ‘other’ expenditures tend to increase in the election year, but only after 2004, when district populations started to directly elect the heads of the local governments (who were previously elected by the local parliaments instead), which increased their legitimacy and discretionary power. More interestingly, this increase is concentrated in districts where the incumbent stands for re-election. These pieces of evidence point towards incumbents' reallocating district funds and using accounting tricks, potentially in order to finance their re-election campaigns (Sjahrir et al. 2013b).

This approach offers an indirect view on local corruption. However, it addresses an aspect of corruption that has been disregarded so far: administrative overspending. Not illegal as such, excessive administrative overspending is a wasteful activity that misuses the political office for personal gain. Such an exploration into administrative overspending relies crucially on the large variation in the characteristics of local governments and constituencies within the same legal environment and can only be applied in larger federal systems.

5 Conclusion

  1. Top of page
  2. Abstract
  3. Introduction
  4. 2 Topics of corruption research
  5. 3 Methodological challenges
  6. 4 Empirical corruption analyses on Southeast Asia
  7. 5 Conclusion
  8. References

Corruption by its very nature is hard to grasp conceptually and very difficult to measure empirically. Corruption is not very sharply defined as ‘misuse of public office for private gain’ and as such comprises a host of activities such as absenteeism of public servants, embezzlement of government funds and resources, providing preferential access to public services or public contracts, lowering of tax payments and dropping or reducing criminal charges, manipulating court rulings, all in exchange for bribes or kickbacks. As corruption is so multidimensional, corruption measurement is either limited to one specific activity, such as corruption payments of manufacturing firms or corruptibility of traffic police, etc. or it relates to the perception of the overall climate of corruption.

We have argued that corruption perceptions are not very reliable as indicators of the intensity of actual corruption since they are prone to perception biases because a clear yardstick is missing and illicit activity is only imperfectly observable, and to answering biases because respondents might be reluctant to reveal illegal behaviour—even of others—to the interviewer. Better suited are surveys that relate to corruption experiences of specific groups, such as manufacturing surveys that ask for typical corruption payments in their line of business. These data potentially still suffer from the same biases, albeit to a much lesser extent.

Because of these substantial limitations, the economic analysis of corruption has recently turned to quantitative measurement approaches that are narrower in focus as they are linked to specific activities. These measures rely on direct observation of corrupt actions, frequently in experimental settings (lab, field, and natural experiments), public expenditure tracking, law enforcement data, or on indirect inference of corrupt behaviour based on observed market outcomes or outcomes of public and private finance. All of these newer approaches are relatively data intensive and presuppose an adequate knowledge of the economic, business, and regulatory environment. A possible drawback of these approaches is that it is not a priori clear to what extent the estimates derived are indicative for the general corruption environment. At the same time, they result in information that can be readily analysed by quantitative methods and lends itself to causal inference. We have surveyed these approaches and pointed out their main strengths and possible weaknesses.

Evidently, none of the presented methods is fully immune to measurement errors. Perceptions and self-reported information depend on the cultural context and can be strongly biased. Experiments might also induce some observational biases due to experimenter effects, although certainly less than the self-reported measures. Markets can misprice available information, and public accounts can be manipulated to some extent. What is decisive, however, is whether these measurement errors are systematically related to specific factors that would undermine causal interpretation of the results. For instance, if the low levels of perceived and reported corruption are a result of high social trust or the cultural norm of loyalty towards one's government, a finding on the negative economic effects of corruption cannot be clearly attributed to corruption only. The lack of adequate controls for social trust or cultural norms will bias the corruption coefficients and mask an underlying relationship. A similar argument could be made with respect to media freedom, which might increase reporting on corruption and hence lead to higher perceived corruption, while at the same time reducing the actual extent of corruption.

Measurement issues notwithstanding, carefully constructed observational measures of corruption can serve as powerful tools for informing economic policies. Even if markets made errors in pricing in the economic value of political connections, changes in market valuations would still provide considerable relevant though imprecise information. Behavioural responses in laboratory experiments do clearly reflect the role of attitudes towards corruption, albeit they are derived in an artificial setting. Well-designed social experiments yield causal results once the potential sources of bias are excluded or addressed. Finally, misreporting in public accounts might lead to underestimating the extent of corruption, but such misreporting can often be traced by a careful study design and hence it does not necessarily change the behavioural relationships. Measurement errors arising from direct or indirect observation are considerably less culture specific, and there exist well-defined procedures that quantitative research can (and should) follow to limit their extent. What is clear by now is that, if implemented carefully, observational measures of corruption are by far less specific to the cultural context and yield themselves more to common methods of quantitative analysis than earlier subjective corruption measures.

Even though the measurement of corruption is challenging, not the least because corrupt activities are illegal and thus clandestine, empirical economics has made significant progress in assessing magnitude, determinants, and consequences of corruption as well as remedies for it. We have argued that quantitative analyses of corruption are absolutely essential for a rational formulation of anti-corruption policies as they provide necessary information on the extent and determinants of corruption and the best avenue of approach to fight corruption, which creates inefficiencies and injustice and severely compromises development. We have presented the econometric studies on corruption in Southeast Asia and pointed out their achievements and their limits. Given the importance of corruption as an obstacle for development in Southeast Asia, it is remarkable that there are relatively few empirical analyses for the region. Both from a methodological perspective and from the perspective of the application studied, there is ample scope—and need—for further empirical research on corruption in Southeast Asia. As corruption is such a complex and multidimensional issue, and no empirical study, even if carefully designed, can ever capture the complexity of the problem entirely, a consistent and comprehensive picture emerges only through the synopsis of a number of methodologically sound analyses.

  1. 1

    For an exception cf. Olken and Barron (2009).

  2. 2

    See Svensson (2005), Pande (2007), and Olken and Pande (2011) for recent surveys on the economics of corruption, all of which have a less specific focus on methodology and none on Southeast Asia. We do not include theoretical contributions in our survey.

  3. 3

    See section 4.1 for a discussion of the limitations of such indices.

  4. 4

    This trade-off between precision and generality of the results is also found in investment climate surveys with some surveys aiming at measuring the costs of very specific transactions, such as the World Bank's Cost of doing business survey (, while others being more general such as World Bank's Enterprise surveys (

  5. 5

    Virtually all cross-country studies find a negative association between GDP per capita levels and corruption perceptions. That does not, however, establish a causal relationship as corruption hampers growth and growth affects GDP per capita, but then the levels GDP per capita also affect corruption. The strength of institutions is highly negatively associated with corruption in almost all studies on the subject, which is not surprising at all—it borders a tautology.

  6. 6

    Treisman (2000) finds that countries with protestant traditions and British legal origin have lower perceived corruption, other things being equal.

  7. 7

    Baland and Robinson (2006) show how the introduction of secret ballot affected electoral corruption and thus land rents. This introduction made vote buying by landowners impossible, which was tied to the rural employment relationship, and thus reduced the value of land (that had previously incorporated the value of political rents). Menes (2006) shows how corrupt politicians in American cities 1880–1930 were constrained in the level of corruption as people and firms could move freely between cities.

  8. 8

    For similar evidence cf. Sung (2004).

  9. 9

    Bardhan (1997) shows that for developing countries the success of decentralisation might be hampered by local capture (cf. also Bardhan and Mookherjee 2000). Other empirical works on decentralisation and corruption include Fisman and Gatti (2002b) and Hofman et al. (2009).

  10. 10

    Other studies that find a negative relationship between decentralisation and corruption include Crook and Manor (2000) and Huther and Shah (1998).

  11. 11

    For instance, a trade policy that uses import quotas to be allocated by bureaucrats on a discretionary basis provides manifold entry points for bureaucratic corruption; yet, in a corrupt environment, regulations may be stipulated deliberately such that corruption is facilitated. Thus, in this example, bad institutions may cause corruption, but corruption may also cause bad institutional design.

  12. 12

    Cf. Collier (2000) and Alesina et al. (2003) for the influence of ethnic diversity on growth, and Franck and Rainer (2009) for the influence of ethnic favouritism on public service delivery in health and education in Sub-Saharan Africa.

  13. 13

    Swamy et al. (2001) show gender differences in attitudes towards corruption. Using a firm survey for Georgia, they show that male-owned or managed firms are more likely to give bribes than firms owned or managed by women.

  14. 14

    In an experimental setting, Frank and Schulze (2000) and Schulze and Frank (2003) show that women are less prone to corruption, and that this effect is due only to the larger aversion towards the risk of detection and punishment. Lambsdorff and Frank (2011) show that men are more inclined to reciprocal corruption than women.

  15. 15

    If women are considered to be less susceptible to corruption—still a debated point—the increase in women's political representation may reduce corruption. There is ample related evidence that enhanced women's representation changes policy outcomes: Chattopadhyay and Duflo (2004) show for India that in village councils that were headed by women as a consequence of a 30 per cent quorum for village council heads, infrastructure investment was more in line with women's needs. Beaman et al. (2009) show for Indian village councils that after ten years with a women quota, women are more likely to stand for election and to be elected due to a change in voters' attitudes. See also Clots-Figueras (2011).

  16. 16

    For other studies on the influence of firm characteristics on corruption see Safavian et al. (2001), Rand and Tarp (2012) and McCulloch et al. (2010).

  17. 17

    Johnson et al. (2011) find that corruption reduces growth performance of US states.

  18. 18

    For cross-country evidence on the corruption-reducing effect of a free press cf. Brunetti and Weder (2003). There is a growing literature on the positive effects of a free press on governance quality in general. Contributions to this literature include Besley and Burgess (2002), Snyder and Strömberg (2010), and Reinikka and Svensson (2011).

  19. 19

    It requires knowing the functional form and the relevant parameters; if the relevant relationship between relative salaries of civil servants and corruption is U shaped, raising the salaries too much may not only be wasteful, it may be counterproductive, cf. Schulze et al. (2013).

  20. 20

    One frequent critique of this argument is that reliable estimates are hard to come by and that data are of poor quality. To be sure, sound empirical analyses are often challenging, especially for clandestine activities such as corruption. Yet, policy makers necessarily make these calculations, explicitly or implicitly, when making their decisions and so does everybody else who voices an opinion on what should be done (and has thought about the problem). People may not be aware of the relevant trade-offs, but they implicitly weigh these options when deciding on a course of action. We argue that these trade-offs should be made explicit and that decisions/views should be based on sound empirical research rather than on anecdotal evidence, hearsay, or even no evidence at all. Thus, this critique is beside the point; it simply neglects the necessity to make a decision, which should be made on the best available basis.

  21. 21

    This section briefly highlights some of the challenges; for a systematic, formal, and comprehensive treatment of these issues consult the econometric literature.

  22. 22

    Sample biases often occur also if interviews are made by telephone. People with low opportunity costs are more likely to answer the telephone and to be willing to do the interview. Thus, retirees and unemployed people may be oversampled. If they tend to answer systematically different from the rest of the population, the estimates will be biased.

  23. 23

    Sample selection problems occur also if the appearance of an observation in the population depends on factors that affect the dependent variable at the same time. Consider, for instance, that the determinants for the regional incidence of corruption in manufacturing firms are investigated and that one factor analysed is the proximity to the district capital. A sample selection bias may occur because more remote locations may be less likely to have manufacturing enterprises at all, which is the prerequisite for observing corruption. Therefore a simple OLS regression may lead to biased estimates; a Heckman selection model may correct for such a bias.

  24. 24

    For an analysis of expressive behaviour, see Hillman (2010) and (28.10.2012).

  25. 25

    The systematic mismatch between different types of customs accounts (in exporting vs. importing country) can be used for quantifying tax evasion and related corruption (cf. Fisman and Wei, 2004 and 2009). This is an old idea, which goes back to Morgenstern (1950, ch. 9) and Bhagwati (1964 1974). For further references cf. Schulze (2000, 108–14).

  26. 26

    This is often misunderstood by scientists who do not use statistical methods. Random noise in the data increases standard errors, but does not create biased estimates. Thus it reduces significance levels of estimated parameters but does not bias them.

  27. 27

    If they are not correlated with the error term, estimates will be unbiased, but the explanatory power of the analysis will be reduced.

  28. 28

    Obviously, one extreme choice for this trade-off is to rely on case studies, in which the sample size is extremely low, often one, and the number of questions very large. Results are obviously not representative and cannot inform other situations (unless complemented by additional large n studies), but the chance that relevant variables are omitted is low.

  29. 29

    More formally, for a sequence {xi}i≡ℕ of i.i.d. random variables with E(xi) = μ and Var(xi) = σ2 < ∞ and n approaching infinity inline image

  30. 30

    A potential additional disadvantage may be that such questions may get relatively low response rates and may suffer from recollection biases; this does, however, depend on the circumstances and is not a ubiquitous phenomenon.

  31. 31

    There is evidence, however, that in many circumstances lab experiments portray real-world behaviour quite well. Armantier and Boly (2012) show remarkably similar results in a corruption experiment implemented in the lab (in Canada and Burkina Faso) and in the field (in Burkina Faso). See also Harrison and List (2004) and Levitt and List (2007) on the relationship between laboratory and field experiments.

  32. 32

    Education and newspaper circulation would be highly correlated. Such a multi-collinearity problem would not bias estimates but would make point estimates less reliable by increasing the standard errors.

  33. 33

    The reader is referred to the econometric literature.

  34. 34

    The advantage of this approach—a clear identification of causality—has made it extremely popular to the extent that some journals are focusing on papers following this approach. If this focus is exclusive it reduces the scope of questions analysed to those on which experiments can be constructed with the issue of external validity remaining. Such an exclusive focus thus seems inadvisable. For a related discussion cf. Ravallion (2012).

  35. 35

    Although we cover all of Southeast Asia, the lion share of the quantitative contributions discussed concern Indonesia. That is a reflection of the literature that has focused on Indonesia, not a deliberate selection on our part.

  36. 36

    For their methodology see and (accessed 20 March 2013). The corruption index by the International Country Risk Guide (ICRG), which has also been used extensively in economic analyses, has the major drawback of mixing expert views on the extent of corruption with the political risks from corrupt practices. As a consequence highly corrupt but stable (and often undemocratic) countries are given a relatively lower corruption score (cf. Lambsdorff 2007:238).

  37. 37

    This survey has been carried out in 2004, 2005, and 2008, cf. Accessed 7 February 2013.

  38. 38

    See for classical studies using the ICRG measures Mauro (1998), Knack and Keefer (1995), or La Porta et al. (1999), for an early use of CPI Treisman (2000), and Fisman and Gatti (2002a) for a comparison of indices.

  39. 39

    An underlying dissatisfaction with government performance could affect the perception of health services, corruption, and capacity of the health sector—another explanatory variable—thereby creating a correlation without necessarily establishing a causal relationship.

  40. 40

    The CPI ranges from zero (very corrupt) to ten (not corrupt). The direct effect of the CPI is negative, the interaction effect positive.

  41. 41

    Hofman et al. (2009) show that politicians and bureaucrats perceive corruption as less severe than members of the civil society or the business community.

  42. 42

    Kaiser et al. (2006) show for the Governance and Decentralization Survey (GDS) 1+, fielded in 2004 and much smaller than GDS 1, that respondents connected to the elite perceived corruption to be much higher than unconnected individuals. They also show that 49 per cent of the respondents were overcharged for the issuing of an ID card, but only 36 per cent stated that they had paid an informal fee in addition to the formal fee. In other words, a significant share of individuals did not realise that they had to pay a bribe when applying for the ID card.

  43. 43

    Cf also Knack and Keefer (1995), Golden and Picci (2005), and Seligson (2006) for weaknesses of perception-based corruption measures.

  44. 44

    See for a list of countries and questions.

  45. 45
  46. 46

    Even such very general wordings result in non-negligible non-response rates. Out of 66,171 firms surveyed around the world between 2006 and 2011 using standardised WBES questionnaires, about 16 per cent did not answer the questions related to the value of informal payments of a firm ‘like this one’. See,

  47. 47

    Out of 1,808 firms, 1,333 reported to have paid bribes; for those which had, bribes amounted to 11 per cent of the total production costs on average.

  48. 48

    The first controlled laboratory experiment on corruption is Frank and Schulze (2000). For a survey of laboratory experiments on corruption, see Abbink (2006) and Lambsdorff (2012).

  49. 49

    In another lab experiment, Barr and Serra (2010) examine corruption behaviour of immigrant students in Oxford. They establish a positive correlation between undergraduate students' likelihood to engage in corruption and their home country's CPI levels, but fail to find the same relationship for graduate students. They show that the difference between graduates and undergraduates in their sample arises due to two channels. Social learning seems to play a role as corrupt behaviour decreases with the time spent in the UK, but this cannot explain the statistical difference between the two groups of students. Moreover, graduate students also appear to be subject to a different selectivity as their general attitudes to the overall institutional environment and corruption in their home countries are likely to affect their decision to study abroad.

  50. 50

    Interestingly, they find that in interviews based on the recollection of past events, truck drivers substantially overstate the amount of bribes they had to pay as they are likely to use this information to elicit higher compensation payments from their employers.

  51. 51

    Similarly, direct observations on corruption were also used by Bertrand et al. (2007) in analysing the mechanisms and extent of corruption involved in obtaining a driver's license in New Delhi, India.

  52. 52

    Jayachandran (2006) investigates the value of political connections in the USA: When Senator Jeffords left the Republican Party in May 2001, tipping the control of the US Senate in favour of the Democratic Party, firms donating to the Republican (Democratic) Party lost (gained) significantly in market valuation. Ferguson and Voth (2008) show that the firms supporting Hitler experienced extra-normal stock returns between January and March 1933 when Hitler took power (so-called ‘Machtergreifung’). Beyond the effect of political connections on market valuation or the access to the global capital markets, empirical studies have also investigated the effect of political connections on (preferential) access to public bank loans (Khwaja and Mian 2005, Li et al. 2008), the size of tax payments or the likelihood of corporate bailouts (Faccio et al. 2006).

  53. 53

    The public audit lottery program in Brazil is an almost unique exception: it publishes centralised fiscal audit reports on randomly selected municipalities, which can be used to assess the value of misappropriated funds and the extent of administrative irregularities (Ferraz and Finan 2008).

  54. 54

    He uses information on households consuming subsidised rice from Susenas, the large-scale national household survey, and SSD (Survey Seratus Desa) or the ‘100 villages survey’.

  55. 55

    Interestingly, the evaluation of the experiment would have yielded different results if Olken had relied on perception-based measures of corruption instead (Olken 2009). For instance, he finds that ethnic heterogeneity tends to increase corruption perceptions in the village while at the same time decreasing the amount of actually missing funds. The mechanism at work might be less trust and social cohesion leading to more monitoring and hence less corruption. This supports the arguments against the use of perceived corruption measures.

  56. 56

    These budget figures are already corrected for evident misclassifications of specific expenditure items from other sectors to administrative spending.


  1. Top of page
  2. Abstract
  3. Introduction
  4. 2 Topics of corruption research
  5. 3 Methodological challenges
  6. 4 Empirical corruption analyses on Southeast Asia
  7. 5 Conclusion
  8. References
  • Abbink, K., 2006. ‘Laboratory experiments on corruption’, in S. Rose-Ackerman (ed.), International Handbook on the Economics of Corruption, Edward Elgar Publishing Inc., Northampton:418437.
  • Acemoglu, D. and Verdier, T., 1998. ‘Property rights, corruption and the allocation of talent: a general equilibrium approach’, Economic Journal, 108(450):13811403.
  • Ades, A. and Di Tella, R., 1997. ‘National champions and corruption: some unpleasant interventionist arithmetic’, Economic Journal, 107(443):10231042.
  • Ades, A. and Di Tella, R., 1999. ‘Rents, competition and corruption’, American Economic Review, 89(4):892993.
  • Alatas, V., Cameron, L., Chaudhuri, A., Erkal, N. and Gangadharan, L., 2009. ‘Subject pool effects in a corruption experiment: A comparison of Indonesian public servants and Indonesian students’, Experimental Economics, 12(1):113132.
  • Alesina, A., Devleeschauwer, A., Easterly, W., Kurlat, S. and Wacziarg, R., 2003. ‘Fractionalization’, Journal of Economic Growth, 8(2):155194.
  • Alt, J.E. and Lassen, D.D., 2013. ‘Enforcement and Public Corruption: evidence from the American States.’ Journal of Law, Economics, and Organization, forthcoming, doi:10.1093/jleo/ews036.
  • Angrist, J.D. and Krueger, A.B., 2001. ‘Instrumental variables and the search for identification: from supply and demand to natural experiments’, Journal of Economic Perspective, 15(4):6985.
  • Arikan, G.G, 2004. ‘Fiscal decentralization: a remedy for corruption’, International Tax and Public Finance, 11(2):175195.
  • Arezki, R. and van der Ploeg, F., 2011. ‘Do natural resources depress income per capita?’, Review of Development Economics, 15(3):504521.
  • Armantier, O. and Boly, A., 2012. Comparing corruption in the lab and in the field in Burkina Faso and in Canada.
  • Azfar, O. and Gurgur, T., 2008. ‘Does corruption affect health outcomes in the Philippines?’, Economics of Governance, 9(3):197244.
  • Azfar, O. and Murrell, P., 2009. ‘Identifying reticent respondents: assessing the quality of survey data’, Economic Development and Cultural Change, 57(2):387411.
  • Baksi, S., Bose, P. and Pandey, M., 2009. ‘The impact of liberalization on bureaucratic corruption’, Journal of Economic Behaviour and Organization, 72(1):214224.
  • Baland, J.M. and Robinson, J.A., 2006. ‘Land and power: theory and evidence from Chile’, NBER Working Paper No. 12517.
  • Banerjee, A.V., Banerji, R., Duflo, E., Glennerster, R. and Khemani, S., 2010. ‘Pitfalls of participatory programs: evidence from a randomized evaluation in education in India’, American Economic Journal: Economic Policy, 2(1):130.
  • Banerjee, A.V., Chattopadhyay, R., Keniston, D. and Singh, N., 2012a. ‘Can institutions be reformed from within? Evidence from a randomized experiment with the Rajasthan police’, NBER Working Paper No. 17912.
  • Banerjee, A.V., Mullainathan, S. and Hanna, R., 2012b. ‘Corruption’, NBER Working Paper No. 17968.
  • Bardhan, P., 1997. ‘Corruption and development: a review of issues’, Journal of Economic Literature, 35(3):13201346.
  • Bardhan, P. and Mookherjee, D., 2000. ‘Capture and governance at local and national levels’, American Economic Review, 90(2):135139.
  • Barr, A. and Serra, D., 2010. ‘Corruption and culture: an experimental analysis’, Journal of Public Economics, 94(1112):862869.
  • Beaman, L., Chattopadhyay, R., Duflo, E., Pande, R. and Topalova, P., 2009. ‘Powerful women: does exposure reduce bias?’, Quarterly Journal of Economics, 124(4):14971540.
  • Beck, T., Demirgürc-Kunt, A. and Maksimovic, V., 2005. ‘Financial and legal constraints to growth: does firm size matter?’, Journal of Finance, 60(1):137177.
  • Bertrand, M., Djankov, S., Hanna, R. and Mullainathan, S., 2007. ‘Obtaining a driver's license in India: an experimental approach to studying corruption’, Quarterly Journal of Economics, 122(4):16391676.
  • Bertrand, M. and Mullainathan, S., 2001. ‘Do people mean what they say? Implications for subjective survey data’, American Economic Review, 91(2):6772.
  • Besley, T. and Burgess, R., 2002. ‘The political economy of government responsiveness: theory and evidence from India’, Quarterly Journal of Economics, 117(4):14151451.
  • Bhagwati, J., 1964. ‘On the Underinvoicing of Imports’, Bulletin of the Oxford University Institute of Statistics, 26(4):389397, reprinted in Bhagwati, Jagdish, 1974. Illegal Transactions in International Trade: Theory and Measurement, North Holland, Amsterdam.
  • Bhattacharyya, S. and Hodler, R., 2010. ‘Natural resources, democracy and corruption’, European Economic Review, 54(4):608621.
  • Björkman, M. and Svensson, J., 2009. ‘Power to the people: evidence from a randomized field experiment on community-based monitoring in Uganda’, Quarterly Journal of Economics, 124(2):735769.
  • Brunetti, A. and Weder, B., 2003. ‘A free press is bad news for corruption’, Journal of Public Economics, 87(78):18011824.
  • Butt, S., 2011. ‘Anti-corruption reform in Indonesia: an obituary?’, Bulletin of Indonesian Economic Studies, 47(3):381394.
  • Cameron, L., Chaudhuri, A., Erkal, N. and Gangadharan, L., 2009. ‘Propensities to engage in and punish corrupt behaviour: experimental evidence from Australia, India, Indonesia and Singapore’, Journal of Public Economics, 93(7–8):843851.
  • Chattopadhyay, R. and Duflo, E., 2004. ‘Women as policy makers: evidence from a randomized policy experiment in India’, Econometrica, 72(5):14091443.
  • Cheung, Y.-L., Rau, R. and Stouraitis, A., 2012. ‘How much do firms pay as bribes and what benefits do they get? Evidence from corruption cases worldwide’, NBER Working Paper No. 17981.
  • Chowdhury, S.K., 2004. ‘The effect of democracy and press freedom on corruption: an empirical test’, Economics Letters, 85(1):93101.
  • Cisneros, E., Hargrave, J. and Kis-Katos, K., 2013. ‘Unintended consequences of anti-corruption strategies: public fiscal audits and deforestation in the Brazilian Amazon’, University of Freiburg IEP Discussion Paper No. 24, forthcoming.
  • Clarke, G.R.G. and Xu, L.C., 2004. ‘Privatization, competition, and corruption: how characteristics of bribe takers and payers affect bribes to utilities?’, Journal of Public Economics, 88(9–10):20672097.
  • Clausen, B., Kraay, A. and Murrell, P., 2010. Does respondent reticence affect the results of corruption surveys? Evidence from the World Bank Enterprise Survey for Nigeria, World Bank Policy Research Working Paper No. 5415.
  • Clausen, B., Kraay, A. and Nyiri, Z., 2011. ‘Corruption and Confidence in Public Institutions: Evidence from a Global Survey’, World Bank Economic Review, 25(2):212249.
  • Clots-Figueras, I., 2011. ‘Women in politics: evidence from the Indian states’, Journal of Public Economics, 95(78):664690.
  • Collier, P., 2000. ‘Ethnicity, politics, and economic performance’, Economics and Politics, 12(3):225245.
    Direct Link:
  • Crook, R. and Manor, J., 2000. ‘Democratic decentralization’, World Bank OED Working Paper No 11.
  • Crouch, H., 2010. Political Reform in Indonesia after Soeharto, Institute of Southeast Asian Studies, Singapore.
  • Dal Bó, E. and Rossi, M., 2007. ‘Corruption and inefficiency: theory and evidence from electric utilities’, Journal of Public Economics, 91(2007):939962.
  • Di Tella, R. and Schargrodsky, E., 2003. ‘The role of wages and auditing during a crackdown on corruption in the city of Buenos Aires’, Journal of Law and Economics, 46(1):269292
  • Dollar, D., Fisman, R. and Gatti, R., 1999. ‘Are women really the fairer sex? Corruption and women in government’, Journal of Economic Behaviour and Organization, 46(4):423429.
  • Donchev, D. and Ujhelyi, G., 2011. ‘What do corruption indices measure?’, Working paper, University of Houston. Accessed February 7, 2013.
  • Duflo, E., Hanna, R. and Ryan, S.P., 2012. ‘Incentives work: getting teachers to come to school’, American Economic Review, 102(4):12411278.
  • Easterly, W. and Levine, R., 1997. ‘Africa's growth tragedy: policies and ethnic divisions’, Quarterly Journal of Economics, 112(4):12031250.
  • Emerson, P.M., 2006. ‘Corruption, competition and democracy’, Journal of Development Economics, 81(1):191212.
  • Faccio, M., 2006. ‘Politically connected firms’, American Economic Review, 96(1):369386.
  • Faccio, M., Masulis, R.W. and McConnell, J.J., 2006. ‘Political connections and corporate bailouts’, Journal of Finance, 61(6):25972635.
  • Fan, S., Lin, C. and Treisman, D., 2009. ‘Political decentralization and corruption: evidence from around the world’, Journal of Public Economics, 93(12):1434.
  • Ferguson, T. and Voth, H.-J., 2008. ‘Betting on Hitler—The value of political connections in Nazi Germany’, Quarterly Journal of Economics, 123(1):101137.
  • Ferraz, C. and Finan, F., 2008. ‘Exposing corrupt politicians: the effects of Brazil's publicly released audits on electoral outcomes’, Quarterly Journal of Economics, 123(2):703745.
  • Ferraz, C. and Finan, F., 2011. ‘Electoral accountability and corruption: evidence from the audits of local governments’, American Economic Review, 101(4):12741311.
  • Ferraz, C., Finan, F. and Moreira, D.B., 2012. Corrupting learning: evidence from missing federal education funds in Brazil, NBER Working Paper No. 18150.
  • Fisman, R., 2001. ‘Estimating the value of political connections’, American Economic Review, 91(4):10951102.
  • Fisman, R. and Gatti, R., 2002a. ‘Decentralization and corruption: evidence across countries’, Journal of Public Economics, 83(3):325345.
  • Fisman, R. and Gatti, R., 2002b. ‘Decentralization and corruption: evidence from a US federal transfer program’, Public Choice, 113(12):2535.
  • Fisman, R. and Miguel, E., 2007. ‘Corruption, norms, and legal enforcement: evidence from diplomatic parking tickets’, Journal of Political Economy, 115(6):10201048.
  • Fisman, R. and Svensson, J., 2007. ‘Are corruption and taxation really harmful to growth? Firm level evidence’, Journal of Development Economics, 83(1):6375.
  • Fisman, R. and Wei, S.-J., 2004. ‘Tax rates and tax evasion: evidence from “missing imports” in China’, Journal of Political Economy, 112(2):471496.
  • Fisman, R. and Wei, S.-J., 2009. ‘The smuggling of art, and the art of smuggling: uncovering the illicit trade in cultural property and antiques’, American Economic Journal: Applied Economics, 1(3):8296.
  • Franck, R. and Rainer, I., 2009. ‘Does the leader's ethnicity matter? Ethnic favouritism, education and health in Sub-Saharan Africa’, American Political Science Review, 106(2):294325.
  • Frank, B. and Schulze, G.G., 2000. ‘Does economics make citizens corrupt?’, Journal of Economic Behaviour and Organization, 43(1):101113.
  • Glaeser, E. and Saks, R., 2006. ‘Corruption in America’, Journal of Public Economics, 90(6–7):10531072.
  • Goetz, A.M., 2007. ‘Political cleaners: women as the new anti-corruption force?’, Development and Change, 38(1):87105.
  • Golden, M., Picci, L., 2005. Proposal for a new measure of corruption: illustrated with Italian Data, Economics and Politics 17:3775.
  • Goldstein, M. and Udry, C., 2008. ‘The profits of power: land rights and agricultural investment in Ghana’, Journal of Political Economy, 116(6):9811022.
  • Guggenheim, S., 2012. ‘Indonesia's quiet springtime: knowledge, policy and reform’, in A. Reid (ed.), Indonesia Rising: the repositioning of Asia's third giant, Institute of Southeast Asian Studies, Singapore:141169.
  • Gupta, S., de Mello, L. and Sharan, R., 2001. ‘Corruption and military spending’, European Journal of Political Economy, 17(4):749777.
  • Gyimah-Brempong, K., 2002. ‘Corruption, economic growth, and income inequality in Africa’, Economics of Governance, 3:183209.
  • Harrison, G. and List, J., 2004. ‘Field Experiments’, Journal of Economic Literature, 42(4):10131059.
  • Henderson, J.V. and Kuncoro, A., 2011. ‘Corruption and local democratization in Indonesia: the role of Islamic parties’, Journal of Development Economics, 94(2):164180.
  • Hill, H., 2012. ‘Corruption and development: the Indonesian experience’, in S. Khoman (ed.), A Scholar for All: essays in honour of Medhi Krongkaew, Thammasat University Press, Bankok:150164.
  • Hillman, A., 2010. ‘Expressive behaviour in economics and politics’, European Journal of Political Economy, 26(4):404419.
  • Hofman, B., Kaiser, K. and Schulze, G.G., 2009. ‘Corruption and decentralization’, in C.J.G. Holzappel and M. Ramstedt (eds.), Decentralization and Regional Autonomy in Indonesia, Institute for Southeast Asian Studies, Singapore:99113.
  • Hunt, J. and Laszlo, S., 2006. ‘Who loses from bribery and who are its beneficiaries?’, McGill University, mimeo,
  • Huther, J. and Shah, A., 1998. ‘Applying a simple measure of good governance to the debate on fiscal decentralization’, Washington, World Bank Policy Research paper No.1894.
  • Ivanyna, M. and Shah, A., 2010. ‘Decentralization (localization) and corruption: new cross-country evidence’, in Peerasit Kamnuansilpa and B.P. Brereton (eds.), Proceedings of The First International Conference on Local Government, November 18–19, 2010, Khon Kaen, Thailand, College of Local Administration, Khon Kaen University, Khon Kaen:87124.
  • Jayachandran, S., 2006. ‘The Jeffords effect’, Journal of Law & Economics, 49(2):397425.
  • Jensen, N.M., Li, Q. and Rahman, A., 2010. ‘Understanding corruption and firm responses in cross-national firm-level surveys’, Journal of International Business Studies, 41:14811504.
  • Johnson, S. and Mitton, T., 2003. ‘Cronyism and capital controls: evidence from Malaysia’, Journal of Financial Economics, 67(2):351382.
  • Johnson, N., LaFountain, C. and Yamarik, S., 2011. ‘Corruption is bad for growth (even in the United States)’, Public Choice, 147(3):377393.
  • Kaiser, K., Pattinasarany, D. and Schulze, G.G., 2006. ‘Decentralization, governance and public services in Indonesia’, in G.E. Peterson , P. Smoke and E.J. Gómez (eds), Decentralization in Latin America and Asia: An Inter-Disciplinary Perspective, Edward Elgar Press, Cheltenham:164207.
  • Khwaja, A.I. and Mian, A., 2005. ‘Do lenders favour politically connected firms? Rent provision in an emerging financial market’, Quarterly Journal of Economics, 120(4):13711411.
  • Knack, S. and Keefer, P., 1995. ‘Institutions and economic performance: cross-country tests using alternative institutional measures’, Economics and Politics, 7(3):207227.
  • Krueger, A.O., 1974. ‘The political economy of the rent-seeking society’, American Economic Review, 64(3):291303.
  • Kuncoro, A., 2004. ‘Bribery in Indonesia: some evidence from micro-level data’, Bulletin of Indonesian Economic Studies, 40(3):329354.
  • Lambsdorff, J.G., 2003a. ‘How corruption affects productivity’, Kyklos, 56(4):459476.
  • Lambsdorff, J.G., 2003b. ‘How corruption affects persistent capital flows’, Economics of Governance, 4(3):229243.
  • Lambsdorff, J.G., 2007. The Institutional Economics of Corruption and Reform: Theory, Evidence, and Policy, Cambridge University Press, Cambridge.
  • Lambsdorff, J.G., 2012. ‘Behavioural and experimental economics as a guidance to anti-corruption’, in D. Serra and L. Wantchekon (eds.), New Advances in Experimental Research on Corruption (Research in Experimental Economics, Volume 15), Emerald Group Publishing Limited, Bingly:279300.
  • Lambsdorff, J.G and Frank, B., 2011. ‘Corrupt reciprocite: experimental evidence on a men's game’, International Review of Law and Economics, 31(2):116125.
  • La Porta, R., Lopez-de-Silanes, F., Shleifer, A. and Vishny, R., 1999. ‘The quality of government’, Journal of Law, Economics and Organization, 15(1):222279.
  • Lederman, D., Loayza, N. and Reis Soares, R., 2005. ‘Accountability and corruption: political institutions matter’, Economics and Politics, 17(1):135.
  • Leuz, C. and Oberholzer-Gee, F., 2006. ‘Political relationships, global financing, and corporate transparency: evidence from Indonesia’, Journal of Financial Economics, 81(2):411439.
  • Levitt, S.D. and List, J.A., 2007. ‘What do laboratory experiments measuring social preferences reveal about the real world?’, Journal of Economic Perspectives, 21(2):153174.
  • Li, H., Meng, L., Wang, Q. and Zhou, L.-A., 2008. ‘Political connections, financing and firm performance: evidence from Chinese private firms’, Journal of Development Economics, 87(2):283299.
  • Lindstedt, C. and Naurin, D., 2010. ‘Transparency is not enough: making transparency effective in reducing corruption’, International Political Science Review, 31(3):301322.
  • Mauro, P., 1995. ‘Corruption and growth’, Quarterly Journal of Economics, 110(3):681712.
  • Mauro, P., 1997. ‘The effects of corruption on growth, investment, and government expenditure: across–country analysis’, in K.A. Elliott (ed.), Corruption and the Global Economy, Institute for International Economics, Washington DC:83107.
  • Mauro, P., 1998. ‘Corruption and the composition of government expenditure’, Journal of Public Economics, 69(2):263279.
  • McCulloch, N., Schulze, G.G. and Voss, J., 2010. ‘What determines firms' decisions to formalize? Evidence from rural Indonesia’, University of Freiburg IEP Discussion Paper No. 13.
  • McLeod, R., 2005. ‘The struggle to regain effective government under democracy in Indonesia’, Bulletin of Indonesian Economic Studies, 41(3):367386.
  • McLeod, R., 2006. ‘Private sector lessons for public sector reform in Indonesia’, Agenda, 13(3):275288.
  • McLeod, R., 2008. ‘Inadequate budgets and salaries as instruments for institutionalizing public sector corruption in Indonesia’, South East Asia Research, 16(2):199223.
  • Mehta, A. and Jha, S., 2012. ‘Corruption, food subsidies, and opacity: evidence from the Philippines’, Economics Letters, 117(3):708711.
  • Menes, R., 2006. ‘Limiting the reach of the grabbing hand graft and growth in American cities, 1880 to 1930’, in E. Glaeser and C. Goldin (eds.), Corruption and Reform: Lessons from America's Economic History, University of Chicago Press, Chicago: 6393.
  • Mocan, N., 2008. ‘What determines corruption? International evidence from microdata’, Economic Inquiry, 46(4):493510.
  • Morgenstern, O., 1950. On the Accuracy of Economic Observations, Princeton University Press, Princeton.
  • Mutebi, A.M., 2008. ‘Explaining the failure of Thailand's anti-corruption regime’, Development and Change, 39(1):147171.
  • Olken, B.A., 2006, ‘Corruption and the costs of redistribution: micro evidence from Indonesia’, Journal of Public Economics, 90(4–5):853870.
  • Olken, B.A., 2007. ‘Monitoring corruption: evidence from a field experiment in Indonesia’, Journal of Political Economy, 115(2):200249.
  • Olken, B.A., 2009. ‘Corruption perceptions vs. corruption reality’, Journal of Public Economics, 93(7–8):950964.
  • Olken, B.A. and Barron, P., 2009. ‘The simple economics of extortion: evidence from trucking in Aceh’, Journal of Political Economy, 117(3):417452.
  • Olken, B.A. and Pande, R., 2011. ‘Corruption in developing countries’, NBER Working Paper No. 17398.
  • Paldam, M., 2002. ‘The cross-country pattern of corruption: economics, culture and the seesaw dynamics’, European Journal of Political Economy, 18(2):215240.
  • Pande, R., 2007. ‘Understanding political corruption in low income countries’, in T.P. Schultz and J. Strauss (eds.), 2008, Handbook of Development Economics Vol. 4, Elsevier, Amsterdam:31553184.
  • Persson, T., Tabellini, G. and Trebbi, F., 2003. ‘Electoral rules and corruption’, Journal of the European Economic Association, 1(4):958989.
  • Phongpaichit, P. and Piriyarangsan, S., 1996. Corruption and Democracy in Thailand, Silkworm Books, Chiang Mai.
  • Phongpaichit, P. and Baker, C., 2004. Thaksin: The Business of Politics in Thailand, Silkworm Books, Chiang Mai.
  • Quah, J.S.T., 2001. ‘Combating corruption in Singapore: what can be learned?’, Journal of Contingencies and Crisis Management, 9(1):2935.
    Direct Link:
  • Rand, J. and Tarp, F., 2012. ‘Firm-level corruption in Vietnam’, Economic Development and Cultural Change, 60(3):57195.
  • Ravallion, M., 2012. Fighting poverty one experiment at a time. A review essay on Abhijit Banerjee and Esther Duflo, poor economics, Journal of Economic Literature, 50(1):10314.
  • Reinikka, R. and Svensson, J., 2004. ‘Local capture: evidence from a central government transfer program in Uganda’, Quarterly Journal of Economics, 119(2):679705.
  • Reinikka, R. and Svensson, J., 2005. ‘Fighting corruption to improve schooling: evidence from a newspaper campaign in Uganda’, Journal of the European Economic Association, 3(2–3):259267.
  • Reinikka, R. and Svensson, J., 2006. ‘Using micro-surveys to measure and explain corruption’, World Development, 34(2):359370.
  • Reinikka, R. and Svensson, J., 2011. ‘The power of information in public services: evidence from education in Uganda’, Journal of Public Economics, 95(7–8):956966.
  • Rock, M., 2009. ‘Corruption and democracy’, Journal of Development Studies, 45(1):5575.
  • Rose-Ackerman, S., 1978. Corruption: A Study in Political Economy, Academic Press, London/New York.
  • Safavian, M.S., Graham, D.H. and Gonzalez-Vega, C., 2001. ‘Corruption and microenterprises in Russia’, World Development, 29(7):12151224.
  • Schulze, G.G., 2000. The Political Economy of Capital Controls, Cambridge University Press, Cambridge.
  • Schulze, G.G. and Frank, B., 2003. ‘Deterrence versus intrinsic motivation: experimental evidence on the determinants of corruptibility’, Economics of Governance, 4(2):143160.
  • Schulze, G.G., Sjahrir, B.S. and Zakharov, N., 2013. ‘Corruption in Russia, University of Freiburg IEP Discussion Paper No. 22, forthcoming.
  • Seligson, M., 2006. The measurement and impact of corruption victimization: survey evidence from Latin America. World Development, 34:381404.
  • Shleifer, A. and Vishny, R.W., 1993. ‘Corruption’, Quarterly Journal of Economics, 108(3):599617.
  • Sjahrir, B.S., Kis-Katos, K. and Schulze, G.G., 2013a. ‘Administrative spending in Indonesian districts: the role of local politics, University of Freiburg IEP Discussion Paper No. 23, forthcoming.
  • Sjahrir, B.S., Kis-Katos, K. and Schulze, G.G., 2013b. ‘Political Business Cycles in Indonesia's Local Elections’, University of Freiburg, Department of International Economic Policy, mimeo.
  • Snyder, J.M. and Strömberg, D., 2010. ‘Press coverage and political accountability’, Journal of Political Economy, 118(2):355408.
  • Sung, H.-E., 2003. ‘Fairer sex or fairer system? Gender and corruption revisited’, Social Forces, 82(2):703723.
  • Sung, H.-E., 2004. ‘Democracy and political corruption: a cross-national comparison’, Crime, Law and Social Change, 41(2):179194.
  • Suryadarma, D., 2012. ‘How corruption diminishes the effectiveness of public spending on education in Indonesia’, Bulletin of Indonesian Economic Studies, 48(1):85100.
  • Swamy, A.V., Knack, S., Lee, Y. and Azfar, O., 2001. ‘Gender and corruption’, Journal of Development Economics, 64(1):2555.
  • Svensson, J., 2003. ‘Who must pay bribes and how much? Evidence from a cross Section of firms’, Quarterly Journal of Economics, 118(1):207230.
  • Svensson, J., 2005. ‘Eight questions about corruption’, Journal of Economic Perspectives, 19(3):1942.
  • Tanzi, V., 1998. ‘Corruption around the world: causes, consequences, scope and cures’, IMF Staff Papers, 45(4):559594.
  • Tavares, S.C., 2007. ‘Do rapid political and trade liberalizations increase corruption?’, European Journal of Political Economy, 23(4):10531076.
  • Transparency International, 2012. ‘2012 Corruption Perceptions Index’, Available at <>, accessed 20 March 2013.
  • Treisman, D., 2000, ‘The causes of corruption: a cross-national study’, Journal of Public Economics 76(3):399458.
  • Van der Ploeg, F., 2011. ‘Natural resources: curse or blessing?’, Journal of Economic Literature, 49(2):366420.
  • Van Rijckeghem, C. and Weder, B., 2001. ‘Bureaucratic corruption and the rate of temptation: do wages in the civil service affect corruption, and by how much?’, Journal of Development Economics, 65(2):307331.
  • Vicente, P.C., 2010. ‘Does oil corrupt? Evidence from a natural experiment in West Africa’, Journal of Development Economics, 92(1):2838.
  • Vijayalakshmi, V., 2008. ‘Rent-seeking and gender in local governance’, Journal of Development Studies, 44(9):12621288.
  • Wei, S.-J., 2000. ‘How taxing is corruption on international investors?’, Review of Economics and Statistics, 82(1):111.
  • Zitzewitz, E., 2012. ‘Does transparency reduce favouritism and corruption? Evidence from the reform of figure skating judging’, NBER Working Paper No. 17732.