The Economic Costs of Corruption: A Survey and New Evidence

by Axel Dreher♣ Thurgau Institute of Economics, Hauptstrasse 90, CH-8280 Kreuzlingen, Switzerland, and University of Konstanz, Dept. of Economics, Germany, E-mail: and Thomas Herzfeld♦  Department of Agricultural Economics, University of Kiel.

June 2005


This paper reviews the empirical literature on the economic costs of corruption. Corruption affects economic growth, the level of GDP per capita, investment activity, international trade and price stability negatively. Additionally, it biases the composition of government expenditures. The second part of the paper estimates the effect of corruption on economic growth and GDP per capita as well as on six possible transmission channels. The results of this analysis allows to calculate the total effect of corruption: An increase of corruption by about one index point reduces GDP growth by 0.13 percentage points and GDP per capita by 425 US$.

1. Introduction

The harmful effects of corruption on countries’ economic development are widely acknowledged in the economics literature. Using formal as well as empirical approaches several authors show that corruption detracts investors, reduces the productivity of public expenditures, distorts the allocation of resources and thus lowers economic growth. These findings are reflected in the strategies of multinational organizations like the World Bank, the International Monetary Fund, and the OECD, among others. The most well-known examples are the 1997 World Bank Anti-corruption Strategy, the OECD Convention on Bribery of Foreign Public Officials in International Business (1997) and the recent (2003) United Nations Anti-corruption Treaty.
[footnote 1: More information about international treaties and strategies against corruption are available at the World Bank (, OECD (,2686,en_ 2649 _37447_1_1_1_1_37447,00.html), and Transparency International (]

While the consequences of corruption on certain aspects of the economy have frequently been investigated, attempts to quantify the overall costs of corruption on the economy have only recently been made (DREHER, KOTSOGIANNIS and MCCORRISTON 2004a, 2004b, 2005). What are the quantitative costs of corruption? This is the question our paper attempts to answer. We present cross-section regressions estimating the impact of corruption on the most important variables identified in the previous literature. Combining the direct and the most important indirect effects allows us to derive estimates of the overall costs of being more corrupt than the average sample country on economic growth.

The article proceeds as follows.
The next section reviews the empirical literature on the consequences of corruption. In section 3 we present our econometric results. Finally, section 4 concludes.

2. Previous Literature

Since corruption is an old phenomenon of human history almost all centuries experienced discussion of this topic. Regarding the 20th century two popular examples of this discussion are HUNTINGTON (1968) and MYRDAL (1971), contemporaneously representing, however, opposing points of view. According to the first, corruption is a phenomenon of modernization that is not likely to affect economic development, whereas the latter clearly expects corruption to impose strong obstacles to a country’s development. This discussion and the related literature are reviewed in BARDHAN (1997), JAIN (2001) and AIDT (2003). More formally, SHLEIFER and VISHNY (1993) show how corruption might affect economic welfare. They distinguish between corruption with theft and those without, the latter being characterised by an additional bribe besides the regular price for obtaining a certain service or good. The bureaucrat acts as monopolistic supplier, thus corruption reduces equilibrium demand and therefore public revenues decline and welfare losses arise. Alternatively, the bureaucrat does not charge the regular price but still demands a bribe. While the impact on supply is not straightforward to quantify, welfare losses arise at least in the form of reduced public revenues. They might be complemented by welfare losses due to lower supply, depending on the specific situation. Empirical studies estimating welfare effects usually focus on the impact of corruption on aggregated indicators instead of actual prices of single services or goods.

[footnote 2: One exception is DI TELLA and SCHARGRODSKY (2003). They find a significant cut of prices for homogeneous hospital inputs over the period of an anti-corruption campaign in Buenos Aires.]

Until recently, the availability of appropriate measures of corruption posed the main obstacle to empirical research. This changed substantially over the second half of the 90s, however. Basically, three different groups of corruption indices emerged: First, indices like those of the International Country Risk Guide (ICRG) and Business International (BI) base on the assessment of country experts. The second group is derived from surveys among foreign or native businesspeople or the broad public. Examples are the indicators reported in the 1997 World Development Report (WDR), by the World Economic Forum (WEF) and by the Institute for Management Development (IMD).  The third and last group consists of so called ‘polls of polls’ and includes the Corruption Perception Index (CPI) of Transparency International and the Graft-index developed by the World Bank. Both indices are constructed using several corruption indicators with the aim of enlarging country coverage and reducing measurement error.

Starting with the seminal work of MAURO (1995) – employing the BI index – a broad empirical research based on those indices emerged. As another example, KNACK and KEEFER (1995) introduced the ICRG index to the literature, JOHNSON ET AL. (1998) and LAMBSDORFF (1998) were among the first to use Transparency International’s index.

The availability of proxies for the degree of corruption inspired new research and resulted in more than 50 papers on the economic consequences of corruption. [footnote 3: The literature analyzing the effects of institutional quality on economic growth is much broader and is reviewed by ARON (2000) and HALL and JONES (1999).]

Figure 1 in the article shows the number of published articles, working papers and book chapters over the period 1995-2005.

The bulk of recent research focuses on determinants of economic welfare like, e.g., the level of per capita GDP and its growth rate, the quality of the public infrastructure, public expenditure allocation, total investment and foreign direct investment. In the following, we discuss the impact of corruption on these variables. However, before proceeding to the main findings of the empirical studies, some general remarks should be raised.

First, existing empirical analyses mainly base on cross-sectional approaches. Results from these studies may overestimate the impact of corruption on the dependent variable because they do not control for unobserved country specific characteristics that do not vary over time.

Second, most empirical studies do not carefully disentangle the high correlation between explanatory variables like indices of institutional quality, investment and corruption – to some extent challenging their results.

Third, while it is obvious that richer countries display a lower level of corruption, causality between corruption and economic wealth is still debated. Some studies apply two-stage least squares approaches and instrument corruption with an index of ethnolinguistic fractionalization or origins of the legal system to take endogeneity into account. However, the instruments employed clearly affect the results.

Finally, the different indices of corruption available differ regarding the period of time and number of countries  they cover.

[footnote 4: Almost half of the studies surveyed here use either the Transparency International (TI) or the International Country Risk Guide (ICRG) indices, which cover a large sample of countries and years. Surprisingly, the Graft-index of KAUFMANN, KRAAY and ZOIDO-LOBATON (1999a; 2002) and KAUFMANN, KRAAY and MASTRUZZI (2003) which shows by far the broadest country coverage is used in only one tenth of these studies. Figure A1 in Appendix A presents the different corruption indices and their frequency of use in empirical research.]

Although correlation among the different indices of corruption is usually rather high, results of empirical studies might in some cases be driven by the underlying choice of corruption index.

GDP per capita and GDP growth
Among the first empirical cross-country analyses of the consequences of corruption on economic development, MAURO (1995) focuses on GDP per capita growth. To measure corruption, he employs the corruption index provided by Business International. Among the 68 countries in the sample, and independent of the method of estimation, more corrupt countries experience both statistically significant lower GDP growth and investment rates. Various studies confirm these results. Among others, GDP per capita growth is used as dependent variable in MAURO (1996), BRUNETTI (1997), POIRSON (1998), LI, XU and ZOU (2000), MO (2001), ABED and DAVOODI (2002), LEITE and WEIDMANN (2002) and GYIMAH-BREMPONG (2002). All of these studies find a statistically significant negative impact of corruption on economic growth in at least some of the estimated specifications. In a more recent study MÉON and SEKKAT (2005) analyse how the interaction of corruption and indices of good governance affect economic growth. Besides the significantly negative impact of corruption on GDP per capita growth, the interaction of corruption and the rule of law as well as corruption and government effectiveness affect growth rates significantly negative. This leads MÉON and SEKKAT to conclude that corruption will be even more detrimental to growth in environments of weak rule of law and low government effectiveness.

However, the negative impact of corruption on growth is not always confirmed. According to PELLEGRINI and GERLAGH (2004), there is no statistically significant direct relationship once other relevant factors are controlled for. There are, however, indirect effects of corruption on economic growth, as corruption negatively affects investment, schooling, trade policies and political stability. Even more surprising, BARRETO (2001) finds a significantly positive direct relationship between GDP per capita growth and corruption (employing the same indicator of corruption as MAURO, 1995).

ROCK and BONNETT (2004), finally, analyze the newly industrializing East Asian countries which are frequently cited as an exception to the general rule of the negative impact corruption has on growth. Despite their comparably high levels of perceived corruption, these countries experienced high growth rates over an extended period of time. Accordingly, ROCK and BONNETT find a significantly positive impact of corruption on growth in large East Asian countries, whereas the impact on other developing countries’ growth rates remains negative and statistically significant. [footnote 5: The group of large East Asian countries includes China, Indonesia, South Korea, Thailand and Japan. The result applies only in the largest sample using the Graft-index compiled by KAUFMANN, KRAAY and ZOIDO-LOBATON (1999a).]

One explanation might be that a strong centralized government can limit the negative effects of bribery compared to a decentralized corrupt bureaucracy (SHLEIFER and VISHNY, 1993). REJA and TALVITIE (2000) provide another explanation. They argue that corruption in Asia is part of the fixed costs of doing business, whereas it is a variable cost component in Africa.

The impact of corruption on the level of per capita GDP has also been frequently investigated. Not surprisingly, most of these papers find a negative impact of corruption on the level of economic development (EHRLICH and LUI, 1999; KAUFMANN, KRAAY and ZOIDO-LOBATON, 1999b; NEEMAN, PASERMAN and SIMHON, 2004 and WELSCH, 2004).

The results of these cross-section approaches have, however, recently been challenged. According to ISLAM (2004), the unobserved fixed country effects and high multicollinearity between explanatory variables are likely to bias the estimation of the impact of corruption on per capita GDP. Whereas he finds a significantly negative relationship between corruption and GDP per capita in a cross-section model, estimating the same model in first differences – thus eliminating the unobserved fixed effects and reducing the correlation between exogenous variables – the impact of corruption is no longer significant. The results of ISLAM, unfortunately, suffer from very limited country coverage, and the inclusion of only two explanatory variables – corruption and total investment.


SEE for the continuation of this paper the original text, go to: