Corruption and the institutional environment for growth.

Heckelman, Jac C. ; Powell, Benjamin

INTRODUCTION

The development policy community widely believes that reducing corruption would improve growth rates in less developed countries. Since 1996, the World Bank has supported more than 600 anticorruption programs and governance initiatives developed by its member countries and publicly sanctioned 338 firms and individuals for corrupt practices. The World Bank also maintains an Institutional Integrity Department that investigates corrupt practices with a staff of more than 50 employees and consultants, and expenditures of more than $10 million annually (World Bank, 2007). (1) According to Institutional Integrity Department director Suzanne Rich Folsom, 'Corruption has a devastating impact on the capacity of governments to function properly; on the private sector to grow and create employment; on the talents and energies of people to add value in productive ways; and ultimately on societies to lift themselves out of poverty' (World Bank, 2007). (2)

Other development agencies express similar sentiments. For example, according to USAID:

   Corruption ... undermines economic development. In the private
   sector, corruption increases the cost of business through the price
   of bribes themselves, the management cost of negotiating with
   officials, and the risk of breached agreements or detection.
   Although some claim corruption reduces costs by cutting red tape,
   an emerging consensus holds that the availability of bribes induces
   officials to contrive new rules and delays. (USAID) (3)

Despite these widely held beliefs, some economists, going back to at least Left (1964) and Huntington (1968), believe that corruption can enhance growth by allowing individuals to pay bribes in order to circumvent inefficient rules and bureaucratic delays. Simply put, in much of the third world corruption is needed to get things done. If corruption is reduced without corresponding changes to eliminate inefficient rules, business activity and economic growth may slow down. If a first best solution of 'good rules' is unavailable then corruption that avoids some of the restrictions created by bad rules becomes a second best solution and an alternative path to growth.

To investigate this hypothesis we examine the empirical relationship between corruption and growth when we interact political and economic institutions with corruption. Previous studies (Aidt et al., 2008; Mendez and Sepulveda, 2006; Meon and Sekkat, 2005) have examined, with mixed results, how political institutions impact the relationship between corruption and growth. We directly build on this literature by first including proxies for political institutions into our analysis and then interacting them with a measure of corruption. We find the results differ dramatically depending on the type of institutions considered. In particular, corruption is found to be more beneficial to growth for greater levels of democracy, a seemingly perverse result, but one that is consistent with Mendez and Sepulveda (2006). The reverse, however, is found when considering economic institutions; corruption is more beneficial when economic freedom is low, and the benefits diminish as economic freedom improves. We also find that among the different types of economic freedom, this result is driven primarily by the size of government and extent of regulatory burdens.

Our paper proceeds as follows. In the next section we discuss other studies that have examined the relationship between corruption and growth and in particular we focus on the recent studies that have incorporated political institutions into their empirical analysis and describe how our study differs. A description of our data follows in the subsequent section. Our empirical methodology and results are presented in the penultimate section. The final section highlights specific findings and offers suggestions for further research in this area.

LITERATURE REVIEW

Corruption and growth

Numerous academic articles give credence to the development policy community's views about corruption. On the theoretic level Shleifer and Vishny (1993) argue that, for example, when it is necessary to get permission from many individuals for a project, and each has veto power over approval, the cost of corruption will rise and slow economic growth. Myrdal (1968) argues that corrupt officials may use their arbitrary power to create delays and barriers that would not otherwise exist in order to collect more bribes. Krueger (1974) represents a classic study of socially inefficient rent-seeking through corrupt trade restriction enforcement. In cases of corruption such as these, the de facto institutional environment would restrict economic activity more than the de jure legal restrictions on the official books.

However, there is also reason to believe that corruption could be good for economic growth. Lui (198S) shows that corruption can shorten the amount of time waiting in queues. (4) In the face of bureaucratic delays that slow business formation or restrictions that prevent businesses and consumers from exploiting potential gains from trade, corrupt officials who circumvent inefficient rules could actually enhance growth. Some positive level of corruption may even be growth-maximizing in countries with relatively efficient rules because as corruption decreases it becomes increasingly costly to eliminate it entirely, much like crime in general (Klitgaard, 1988). Colombatto (2003) also analyzes corruption theoretically in a variety of different institutional environments and finds that in some cases corruption can be efficient in developed countries as well as in totalitarian ones.

The empirical literature using cross-country data to estimate how corruption affects growth is mixed, reflecting the various offsetting theoretic effects corruption may have. Mauro (1995) produced the seminal study for empirically investigating corruption's impact on growth for a wide cross-section of countries. He found that higher levels of corruption significantly decrease both investment and economic growth, but his findings were sensitive to the choice of specification. Poirson (1998) and Leite and Weidmann (1999) found that corruption has a negative effect on growth. Mo (2001) found that corruption decreases growth after controlling for investment, but that the effect of corruption becomes insignificant once education is controlled. Gyimah-Brempong (2002) found that corruption decreased growth rates and increased income inequality among a sample of exclusively African nations. However, Brunetti et al. (1997) found inconclusive results and Wedeman (1997) found that many corrupt countries have rapid growth rates. In Svensson's (2005) survey article on corruption, he updated Mauro's calculations and although he found corruption did have a negative coefficient it was not statistically significant. Svensson concluded that 'to the extent we can measure corruption in a cross-country setting, it does not affect growth' (p. 39). (5)

Our hypothesis is that the empirical literature is unable to sort out whether corruption is beneficial or harmful to growth in a cross-country setting because most earlier studies have not controlled for institutional quality. In countries where the institutional environment is relatively good, we expect that corruption will mainly take the form of rent-seeking activities that slow growth. In contrast, when the institutional environment presents a low level of economic freedom, we expect that entrepreneurs will use bribes to circumvent cumbersome regulations and thus corruption will promote growth) Next we review the few existing studies that have examined how corruption effects growth while including measures of institutional quality and explain how we add to that literature.

Controlling for institutions

Recent empirical studies have begun to examine corruption's impact on economic growth contingent on a country's institutional environment. Typically political, rather than economic, institutions have been the focus. Mendez and Sepulveda (2006) use the Freedom House democracy index, which measures civil liberties and political rights. After splitting countries into groups classified as 'free' or 'not-free', they find no relationship between corruption and growth in 'not-free' countries but a small, positive, growth-maximizing level of corruption in 'free' countries.

Aidt et al. (2008) control for political institutions using the voice and accountability index, one of five indicators of governance constructed by Kaufmann et al. (1999). This index attempts to measure the degree to which citizens participate in the selection of their government and have the ability to hold government officials responsible for policy outcomes. Aidt et al. also find a nonlinear relationship between corruption and growth once political institutions are controlled, but the pattern is somewhat different from the findings of Mendez and Sepulveda (2006). Aidt et al. conclude that when political institutions are of 'low quality', corruption has little impact on growth. However, unlike Mendez and Sepulveda, they find that 'high quality' political institutions result in corruption being harmful to growth.

Meon and Sekkat (2005) examine whether corruption 'greases the wheels' or 'sands the wheels' of economic growth when institutional quality and corruption interact. Their measure of institutional quality combines both political and some economic institutions. They use all five of Kaufmann et al.'s (1999) indicators of governance, namely: (A) 'voice and accountability', (B) 'lack of political violence', (C) 'government effectiveness', (D) 'regulatory burden', and (E) 'rule of law'. Meon and Sekkat find that the 'regulatory burden' and 'voice and accountability' measures are not significant in any of their specifications when they interact them with corruption. They find, however, that the 'rule of law' and 'government effectiveness' measures are consistently statistically significant when interacted with corruption, and that as institutional quality decreases corruption becomes more harmful to growth. They conclude that, on net, net corruption 'sands the wheels' of economic growth, supporting the conventional view, rather than 'greasing the wheels' by allowing individuals to circumvent bad governance.

Each of these three studies furthers our knowledge of how political institutional quality impacts the relationship between corruption and growth. Mendez and Sepulveda (2006) and Aidt et al. (2008) both find no relationship between corruption and growth in countries with low quality political institutions but they reach conflicting conclusions in countries with high quality political institutions. Meon and Sekkat (2005) find that corruption is harmful for growth overall and that it is even more harmful in countries with low quality political institutions. None of these studies support the view that corruption can increase growth in countries with low quality institutions.

However, none of these studies directly control for the role of economic institutions while investigating the interplay between corruption and democracy. Furthermore, only Meon and Sekkat (2005) examine the connection between economic institutions and the effect of corruption on growth, but their measure of economic institutions is quite limited.

We build on the current corruption-institutions growth literature by examining the relationship between corruption and growth while controlling for, and interacting, both the quality of political institutions and economic institutions. Existing studies have advanced our understanding by pointing to a nonlinear relationship between corruption and growth contingent on institutional quality. Our main interest is in testing specific areas in which corruption may be able to grease the wheels to circumvent inefficient regulations and improve growth. We expect that if corruption is to grease the wheels of growth, this would occur in places in which the economic freedoms beneficial to growth are restricted and barriers prevent voluntary exchange from exploiting the gains from trade. (7) When economic, rather than political, freedom is low, corruption is most likely to improve growth. To test this, we interact both general levels of economic freedom and specific sub-areas of economic freedom with corruption, while controlling for the overall level of democracy. Our results are the reverse of those found for the interrelationship between democracy and corruption growth. We find that corruption 'greases the wheels' of growth when economic freedom is low but the benefit of corruption diminishes when economic institutions improve. Among the different areas of economic freedom (described below), this effect is driven primarily by the separate categories of government size and regulation.

DATA

Our empirical focus is on the differential impact of corruption on growth, dependent upon the level of economic freedom. We first describe our measures of corruption, democratization, and economic freedom, and compare them against other measures in the literature. We then briefly describe our other variables.

Our measure of corruption comes from Transparency International's (1995-2000) Corruption Perceptions Index (CPI), which has been utilized in many studies. The CPI is an 'index of indexes' that averages scores from 16 different surveys of the perceived level of corruption in a country. A nation must have a score for at least two of the surveys to be included in the CPI. The index is scaled from 0 (most corrupt) to 10 (most clean). In our empirical analysis we have inverted the index so that greater values represent more, rather than less, corruption. The recent studies that examine corruption and growth while controlling for political institutions (Meon and Sekkat, 2005; Mendez and Sepulveda, 2006; Aidt et al., 2008) use a variety of different measures of corruption, but the CPI is the only measure used in all of them. Thus, our choice of the CPI as a measure of corruption better enables comparison of our results with these studies.

The CPI has been calculated on an annual basis since 1995. The underlying survey scores on which the CPI is based are not available, however, and the number of surveys used to calculate the index and the number of countries covered varies from year to year. The original 1995 index covered only 41 countries, but by 2000 it included 90 countries. The CPI was chosen as our measure of corruption because of data availability, country coverage, and ability to compare our results to prior studies. Although the index is not consistently measured over time, due to the varying number of surveys included, the fact that the CPI is based on multiple surveys with different methodologies helps to reduce measurement error associated with a single survey. We add a robustness check replicating our main regression with only 2000 index values to account for the changes in the index.

Alternative indexes of corruption are available from the Institute for Management Development (IMD) and the International Country Risk Guide (ICRG). The IMD index covers only 50 countries, however. The ICRG measures the risk involved in corruption rather than the perceived level of corruption. These two proxies for corruption can conceivably differ from each other because public attitudes toward corruption vary between countries (Svensson, 2005, p. 22). However, Mendez and Sepulveda (2006) have shown that for the countries in which the three indexes overlap, the CPI is highly correlated with both the ICRG index (0.91) and the IMD index (0.96), thereby suggesting that the choice of corruption index is unlikely to greatly influence the estimation results. Indeed, in their comparative study, Mendez and Sepulveda (2006) showed their fundamental results were robust across all three measures.

There may also be some concern that there could be problems with relying on the CPI and related indexes because they are based on subjective perceptions. Unfortunately, as noted by Svensson (2005), 'hard evidence' on corruption is available for only a limited set of countries. The strong correlations noted above among the standard indexes at least suggest that our results from using the CPI will not be driven by a particular form of subjectivity.

Our measure of political institutions is the traditional Polity IV index. The Polity IV index ranks a country's political institutions by giving each country a score from -10 to 10, ranging from pure autocracy to consolidated democracy. Polity IV scores are based on: the presence of institutions and procedures through which citizens can express their preferences; presence of institutionalized constraints on the executive; civil liberties for citizens in political participation and their daily lives; the extent of suppression of competitive political participation; and whether chief executives are chosen in a regularized process within the political elite. For ease of comparison to our economic freedom variable described below, and for interpreting interaction coefficients involving this democracy proxy, we have rescaled the Polity IV data to range from 0 to 10.

We chose to measure democracy using Polity IV data rather than the index of political rights and civil liberties from Freedom House used by Mendez and Sepulveda (2006). (8) The political rights index incorporates a direct measure of corruption, which would hinder estimation of an independent relationship between corruption and democracy. Similarly, the civil liberties index is based in part on respect for the rule of law, which many economists treat as a proxy for economic institutions in its own right to measure property rights (Knack and Keefer, 1995; Barro, 1997; Aron, 2000), and is part of the index we use to measure the quality of economic institutions. Given that we want to specifically differentiate between democratic and economic institutions with separate proxy variables, relying on the civil liberties index would be problematic. (9) Still, for our sample, the Polity values are strongly correlated with both the political rights index (0.91) and civil liberties index (0.83). (10)

Our measure of economic institutions comes from Gwartney and Lawson's (2006) Economic Freedom of the World Annual Report. Their economic freedom of the world (EFW) index currently uses 37 criteria to measure freedom levels in five broad areas: size of government; legal structure and property rights; access to sound money; freedom to exchange with foreigners; and regulation of credit, labor, and business. Each area score is based on the average value of the different components in that area (see Appendix A). Each component is assigned a value from 0 (least freedom) to 10 (most freedom). The overall index value is the simple average of the five area scores. (11)

The EFW index provides a more direct measure of restrictive policies for which the 'grease the wheels' form of corruption would be necessary to circumvent. Meon and Sekkat's (2005) measures of government effectiveness, regulatory burden, and rule of law have come the closest thus far to measuring the inefficient economic institutions that corruption might circumvent. Indeed, proxies for the latter two are included as part of the EFW index. However, Meon and Sekkat's government effectiveness measure also includes variables that could themselves be proxies for measures of corruption, such as the independence of the civil service from political pressures and the credibility of the government's commitment to policies. Their regulatory burden measure seems more directly related to the form of corruption that could potentially grease the wheels of development, but it is never statistically significant in their regressions. Area 5 of the EFW provides a broader measure of regulation that represents 14 specific components of regulation across the areas of credit, labor, and business. One of the five components of business regulation represents 'irregular payments' (C.v.), which could be considered a measure of corruption. As such, this component is dropped from the EFW index we use. (12) Area 2 of the EFW index includes five measures of the legal structure and protection of property rights similar to Meon and Sekkat's rule of law measure.

The EFW index also has the advantage over the Kaufmann et al. (1999) index in its coverage of the size of government (Area 1), which includes measures of government spending, transfers, ownership of enterprises and investment, and tax rates. Clearly, there are opportunities for corruption in these areas that could either grease or sand the wheels of development. The inclusion of measures for access to sound money (Area 3) and freedom to trade with foreigners (Area 4) in the EFW index provide additional areas of institutional variation in which corruption could potentially grease the wheels. Meon and Sekkat (2005) do include a measure of trade openness but do not interact it with corruption. Also, their measure includes exports and imports as a percent of GDP, which reflects non-regulatory factors, such as geography, in which corruption could not lead to greasing the wheels of development. (13) The EFW index focuses directly on trade barriers in which corruption could potentially increase efficiency and growth. It uses various measures to assess tariff and regulatory barriers to trade, and exchange rate and capital controls. In all of these areas, corruption has the potential to actually improve economic growth. For example, if the de jure rule is a 500% tariff rate, and customs officials can be bribed to allow goods in for less than that, international trade, and potentially growth, could increase. Thus, we focus on a measure of institutions that, although still imperfect, more directly measures the types of economic policies for which corruption could potentially minimize the harm done by restrictions and thus promote growth. Because separate area scores are available, we can further investigate which, if any, of the economic institutions affect the relationship between corruption and growth.

Although most of the components comprising the EFW index are based on objective data, a variety of subjective measures are used to determine Area 2 for legal structure and protection of property rights in which objective measures are less readily available. These measures are collected from outside sources published by ICRG and Global Competitiveness Report. Thus, all the component scores that comprise the EFW are at least easily verifiable, which is not true of several other indexes (eg, Heritage Foundation Index of Economic Freedom) that have been used to proxy for the extent of economic freedom. It is important to note that many of the economic environment data used by Meon and Sekkat (2005) are also based on subjective evaluations.

Our base regression also includes the starting level of GDP per capita and investment to GDP ratio, both taken from the World Bank's World Development Indicators. Another standard determinant of growth is the extent of human capital. Country coverage for such variables is often incomplete. As such, our primary regressions do not include any controls for human capital, but we do consider the effect of including an education variable from Barro and Lee (2000) as a robustness check. Finally, we also include a set of regional dummies to control for any remaining unobserved heterogeneity that may differ systematically by geographic location. (14)

ANALYSIS

Descriptive statistics

Our period of analysis is determined by the availability of CPI data, which begins in 1995, therefore, like Aidt et al. (2008), we are limited to explaining short-run growth rates. To avoid the potential for single-year anomalies, we use average values for the explanatory variables (except starting year GDP to capture the 'catch-up' effect). Averages over several years are also important for our main variables of focus because corruption and economic freedom are likely long-term phenomena that only change slowly and also because our measures of corruption are surveys of peoples' opinions, which are likely influenced by many prior years' experience with corruption. (15) To ensure a sufficient number of years for both averaging the independent variables and still avoiding as much endogeneity as possible by not overlapping with the growth period used as the dependent variable, we use 1995-2000 values to explain per capita growth over 2000-2005. Thus, like the prior studies we compare our results to, we are limited to cross-sectional analysis rather than panel estimations due to data availability.

Some concern remains about the possible simultaneity of corruption and growth. We make a modest attempt to minimize the problem by using independent variables from 1995-2000 to explain 2000-2005 growth. However, the possibility remains that there is a longer-term omitted variable that impacts both corruption in the late 1990s and growth in the early 2000s. At this point, data limitations do not allow us to do more to address the issue, so due caution is in order when interpreting our results.

The corruption variable is the average of all the years from 1995-2000 in which a country was rated. The economic freedom variable is the average value of 1995 and 2000 (annual data are not available until after 2000). The investment and democracy variables are the average for all years from 1995-2000. Initial GDP is measured in constant 2000 international dollars PPP. Our preliminary, complete data set includes 82 countries (see Appendix B). Finally, for the education variable we use the average for the number of years of schooling for 1995 and 2000 from the Barro-Lee data set. (16) Including education reduces the sample to 72 countries (see Appendix B).

Descriptive statistics for the full set of 82 countries appear in Table 1. Among the economic freedom categories, freedom is highest on average for Area 3 (sound money) and Area 4 (international trade) and lowest for Area 5 (regulatory burden) and Area 1 (size of government). The other four columns stratify the sample by mean levels of democracy and economic freedom. Countries are evenly divided on their levels of economic freedom, with the mean approximately equal to the median, but the distribution is much more skewed in terms of the level of democracy, with the median democratic nation well above the average level.

The average levels in both the higher democracy (III) and higher economic freedom (V) samples are higher for initial wealth and investment, but lower for growth and corruption, compared to the respective lower level samples (II and IV). In the higher democracy (III) compared to lower democracy (II) samples, economic freedom is higher overall and in each separate category, except for Area 1 (government size) in which it is basically the same. The average level of democracy is also higher for countries with above average level of economic freedom (IV versus V). Every area of economic freedom is higher on average when the overall level of economic freedom is above average.

In our regression analysis, we will keep the five areas of economic freedom distinct, as studies have shown that their marginal impacts on growth differ (Carlsson and Lundstrom, 2002; Heckelman and Knack, 2008). Table 2 presents the correlation matrix among corruption, democracy, and the areas of economic freedom. The areas of economic freedom are not highly correlated with each other. The highest correlation is only 0.61, between Area 2 (legal structure and property rights) and Area 4 (international trade). Area 1 (government size) is the least correlated with other areas of economic freedom, with only very weak positive correlations of 0.23 to Area 5 (regulation) and 0.12 to Area 3 (sound money), and inverse correlations of 0.13 to Area 4 (international trade) and 0.43 to Area 2 (legal structure and property rights). Others have shown that even the various components within each area are often not highly correlated with each other and may be more highly correlated with freedom components grouped into other areas (Caudill et al., 2000; Heckelman and Stroup, 2000). For the current version of EFW, 38 distinct pieces of data comprise the index components and subcomponents, therefore it is impractical to include each component variable separately in the regression analysis. As a compromise, we separate out the five areas of economic freedom in the regressions, keeping in mind that misspecification is still possible if the underlying components of each area have different marginal effects on growth. (17)

Democracy is correlated with the overall economic freedom index at the moderate level of 0.49. The highest correlation is with Area 4 (international trade) at only 0.57, and there is basically no correlation with Area 1 (government size), registering only at 0.02.

Consistent with the large differences in average corruption level presented in Table l, Table 2 shows that corruption is highly inversely correlated with the overall EFW index, and with four of the five individual economic freedom areas. In particular, the correlation between corruption and freedom Area 2 (legal structure and property rights) is -0.93. Corruption is positively correlated only with Area 1 (government size), but the correlation is modest at 0.30. The inverse correlation between corruption and democracy (-0.53) is also much weaker than between corruption and overall economic freedom.

Regression analysis for corruption, democracy, economic freedom, and growth

As described above, the independent variables include log initial GDP, investment, democracy, corruption, and a set of regional dummies to control for other inter-regional heterogeneity. A White test rejected the null of no heteroskedasticity. (18) A general White robustness correction did not affect the standard errors very much. A similar issue confronted Clarke (1995) and Folster and Henrekson (1999) in their growth regressions, so they adopted Weighted Least Squares (WLS) as their preferred estimator. Because they used panel data, their weights for each country observation were based on the standard deviation of the country residuals. In a pure cross-section format such as ours, we are unable to follow suit.

Instead, we note that Folster and Henrekson (1999) suggested growth will tend to vary less among larger countries because growth is measured as the average of growth in subregions, and subregions in larger countries tend to be more economically integrated, which, due to regional policy, factor mobilization, and other reasons, will tend to smooth growth, compared to smaller countries. On the basis of this rationale, we use country population as our proportional weighting variable. (19) This estimation technique was also used by Heckelman and Knack (2008) for their growth regressions.

Table 3 presents the WLS regressions. (20) In column I, each estimated variable has the expected sign and is statistically significant at the 5% level. The negative coefficient on initial GDP supports evidence for conditional convergence, and higher levels of investment, democracy, and corruption are all shown to promote growth.

Column II includes measures of economic freedom. These estimates show that not all areas of economic freedom have the same impact. In particular, only freedom in Areas 2 and 5 are significantly beneficial to growth, while an increase in freedom in Area 1 is marginally harmful to growth. On the basis of the relative magnitudes of the area coefficients, the net impact from an equal across-the-board increase in every area of economic freedom is positive for growth overall.

A comparison of estimates from regressions I and II also shows the importance of controlling for economic freedom. The signs and significance for initial GDP, investment, and corruption remain the same, but in regression II the impact of initial GDP and corruption are now enhanced, while the effect for investment is reduced (but still statistically significant). Because of the high degree of inverse correlation between corruption and economic freedom, failure to control for economic freedom can lead to a bias in the estimated impact of corruption. In additional regressions not presented in the table, we also find that if only one area of economic freedom is included at a time, the estimated coefficient for corruption remains positive and significant for Areas 1, 2, and 5 at 5%, and Area 4 at 10%. The estimated coefficient for corruption is also positive but not quite significant at 10% when controlling only for Area 3 freedom. Thus, no matter which area of economic freedom was included, corruption was never found to reduce growth, and typically benefits growth. The increased impact of corruption shown in column II appears to be driven primarily by controlling for Area 2 (legal structure and property rights), as the coefficient on corruption jumps to 6.3 (t-statistic = 5.3).

The other major change between estimates from columns I and II is the effect of democracy is now negative, and weakly significant at the 10% level. Thus, controlling for economic freedom reverses the impact of democracy on growth. Even when controlling for just one area of economic freedom at a time (additional regressions not reported in the table) the impact of democracy is always reduced, and falls below the 5% level of significance for Area 3, and below 10% for Area 4. Again, Area 2 appears to be the critical area needed to be controlled to generate the negative and significant coefficient on democracy (t-statistic = -2.1). Using the Freedom House index of democratic freedoms and the overall EFW Index, Gwartney et al. (1999) find that economic freedom significantly benefits growth when controlling for democracy, but the reverse does not hold. Depending on the particular area of economic freedom controlled, and the level of statistical significance accepted, our results can be interpreted as broadly supportive of their findings.

Differing effects of corruption on growth

In separate samples, Mendez and Sepulveda (2006) found the effect of low levels of corruption to be small but beneficial in countries rated democratically 'free' by Freedom House, and that it had no impact in 'not free' countries, suggesting a relationship on the interplay of democracy and corruption, which they hypothesize might follow Klitgaard's (1998) theory. We can assess the findings of Mendez and Sepulveda in our framework by interacting our Polity democracy variable with corruption. The coefficient on corruption by itself captures the effect of corruption in a purely autocratic nation (democracy = 0), and the coefficient on the interaction term captures changing effects of corruption as democracy begins to improve.

Our estimates, presented in column III, are roughly in accord with Mendez and Sepulveda (2006), but stronger overall. We find that at the lowest levels of democracy, corruption is harmful to growth but becomes less harmful and eventually beneficial as the level of democracy increases. The point at which the net impact of corruption becomes positive occurs at a democracy level of only 6.5, which is below the mean level of democracy in our sample (see Table 1). Note that investment is no longer statistically significant using this specification.

The effect of corruption found so far appears to be somewhat counterintuitive. Our analysis in the opening sections suggests that the positive impact from corruption would be greatest when economic freedom is limited. To test this hypothesis, we interact our measures of corruption and economic freedom in column IV.

As expected, the estimated coefficient for corruption is positive and statistically significant, indicating that corruption benefits growth when other institutions and policies repress economic freedom to the maximum level possible (EFW = 0). In contrast, when economic freedom improves, the interaction term suggests corruption becomes less beneficial. Eventually, corruption does become harmful, but only at a high EFW Index value of 9.6, which, while theoretically possible, is not matched by any country in our sample. (21) At the mean level of EFW, if an average level of corruption country such as Greece (inverted corruption average = 5.15) were to reduce its corruption down to United States levels (2.36), then, ceteris paribus, its growth rate would be expected to fall by roughly two standard deviations over the 5-year period. If, on the other hand, a country with a top EFW score of 10 were to do the same, its growth would be expected to increase an additional 2.6 percentage points above its norm.

We also find that replacing democracy by economic freedom in the interaction term returns the significant impact of investment, perhaps further indicating the democracy interaction was a misspecification. Finally, note also that the estimated coefficients and t-statistics on the separate economic freedom areas all become more positive than in either columns II of III, with Areas 3 and 4 now also considered statistically significant. Only Area 1 has a negative coefficient, but contributes a negligible effect. These findings are consistent with the view that when a country has poor economic institutions, corruption can allow individuals to avoid inefficient rules that would otherwise slow growth even more. (22) Our analysis is in direct contrast to Mendez and Sepulveda (2006), who found low levels of corruption had no impact in 'not free' countries but a small amount of corruption was beneficial in 'free' countries, when they investigated only political rather than economic freedom.

Thus far we have treated the index values as cardinal, rather than ordinal. If the index values are viewed as strictly ordinal in nature then the interaction terms created and marginal impacts we estimate do not have much meaning. An alternate method to interactions is to split the sample based on the EFW values and determine if corruption has a different estimated coefficient across the samples. We do not have enough observations to run quantile regressions, so instead we split the sample into halves. A simple dummy for the democracy index is used to indicate if the polity value is greater than its median. To allow for greater differentiation on our primary variable of interest, a series of dummy variables are used to indicate if the corruption index is in any particular quartile. Results for this regression are presented in Table 4. (23)

Among the low EFW sample countries, greater values of corruption result in higher growth rates, with the fourth quartile of corruption yielding a statistically significant coefficient. The least corrupt countries have the lowest expected growth, with each successive quartile increasing growth monotonically. Moving from the lowest to highest quartile of corruption would increase growth by roughly a third of the standard deviation in growth. (24)

For the high EFW sample countries, every corruption quartile generates a negative coefficient suggesting corruption is always harmful on average. However, none of the coefficients are statistically significant.

Thus, these results generally support the findings from column IV of Table 3, which treated the indexes as continuous variables. More corruption aids growth only when economic freedom is low.

Robustness

We also tested the sensitivity of the corruption effect on growth found previously to different specifications. To conserve space, Table 5 reports only the coefficient and t-statistics on the corruption and interaction term for each modification, but the control variables otherwise remain the same as the final specification in column IV of Table 3.

Our main comparison has been to differentiate between freedom for a nation based on its level of democracy versus its type of economic institutions and policies. We therefore drop democracy from the specification to allow the corruption-economic freedom interplay to be independent of the degree of democratization. Dropping democracy also allows for the inclusion of two additional country observations for Iceland and Luxembourg, which were missing Polity data. As shown in row 1 of Table 5, the results remain substantially the same to this alteration.

We also note that Mo (2001) found the effect of corruption to differ depending on whether investment or education levels were controlled. As shown in row 2, dropping investment has no appreciable effect on either the corruption or interaction term.

Next we added a control for education in row 3. Using the Barro-Lee data dropped 10 observations (Appendix B denotes the lost country observations by an asterisk), but results remained substantively the same. We present results using the average number of years of schooling as the human capital proxy, but alternative variables from Barro and Lee (2000) were also utilized with similar results.

The degree of corruption may differ systematically in different parts of the world (Clague, 2003; Aidt et al., 2003), in which case the regional dummies may be obfuscating the direct effects from corruption. As a check, in row 4 the regional dummies are removed. Dropping the set of regional dummies suggests the impact of corruption will turn negative at a slightly lower level of overall economic freedom, but still above the highest rated nation in our sample.

These results show that the effects of corruption interacted with economic freedom estimated in column IV of Table 3 are robust to the previously described specification changes. The effects of corruption, interacted with democratization reported in column III of Table 3, are not. In unreported regressions, we repeated each of the specification changes described in rows 2-4 of Table 5 when the interaction term involved democratization, as in column III of Table 3. Doing so always yielded the same estimated signs for corruption (negative) and the interaction term (positive) as presented in column III of Table 3, but significance levels varied dramatically. Dropping investment does not alter significance of either term, but neither is significant when adding the education variable, and the corruption term by itself is not significant when dropping the regional dummies, although the interaction term remains significant. Thus, not only does the interaction of corruption with economic freedom alter previous findings of the effect of corruption which utilized democracy as the proxy for institutional quality, but the results are more robust as well.

The remaining rows in Table 5 consider if the effect of corruption is dependent on the particular type of economic freedom. We expect corruption to be more beneficial when the economic institutions necessary for growth are lacking. Judging from the results in Table 3, this appears to be most evident when freedom in Areas 2 and 5 are low. Thus we expect the effect of corruption to be positive when these areas in particular are low but eventually to become negative when freedom reaches a high enough level.

Conversely, the positive impact on growth from Areas 3 and 4 was less robust. Thus, our expectations on how corruption effects growth dependent on these types of freedom are much more tentative. Similarly, because larger values for Area 1 reduce growth at the margin (although generally not to a significant degree), we might speculate that the effect of corruption relative to the level of freedom in Area 1 would be the reverse of Areas 3 and 4. However, given that the negative impact of freedom in Area 1 is basically absent in the preferred specification (IV), this reverse effect is likely to be absent as well.

The estimated effects of corruption dependent upon the specific category of economic freedom are presented in the last five rows of Table 5. We find that freedom in Areas 2-4 does not affect corruption, and that corruption is only found to have a significant impact at all when the interaction is with Area 4. Although greater freedom in Area 4 would reduce the benefits of corruption, the estimated t-statistic on the interaction term suggests this reduction is not statistically significant. For Areas 1 (size of government) and 5 (regulation), however, corruption is again shown to be beneficial in the absence of freedom in these areas, but the benefit falls rapidly as economic freedom improves and eventually becomes harmful to growth. The estimated turning point of corruption occurs when freedom in Area 1 is close to its mean and median values. For Area 5, this occurs at a somewhat higher level; although the estimated value of 6.7 would appear to be feasible for a great number of countries, even the top-rated countries in this area just barely manage to miss this threshold. (25)

Our results suggest that when freedom from big government and regulations are low, corruption appears to be a beneficial way to circumvent growth-retarding government presence and regulations that would otherwise hinder productivity. When government size is small or freedom from regulation is already high, corruption becomes harmful to growth. The interpretation of our result for Area 5 (freedom from regulation) is straightforward. When there are pervasive regulations that limit potential gains from trade, corruption allows entrepreneurs to bypass official regulations and further capitalize on growth opportunities. As economies become freer from regulation, corruption serves this beneficial purpose less often. The interpretation of our result for Area 1 (government size) requires greater elaboration. Area 1 measures government consumption spending as a percent of total consumption spending, transfers, and subsidies as a percent of GDP, government enterprises and investment as a percent of total investment, and marginal tax rates. One might theorize that corruption in this area of government would divert government spending away from the optimal provision of public goods and toward private interests while reducing growth in the process. Our results do not support such a view. Alternatively, if the political process predominantly serves private interests anyway, perhaps the introduction of explicit corruption actually enhances the process of allocating government funds by directing funds to those most willing to pay for the transfer rather than the most politically connected or largest voting bloc. If the highest bidder is best able to make efficient use of the resource, then corruption in Area 1 might actually move resources to their highest valued use and thus promote growth. Our results are consistent with this perspective but this conjecture remains more speculative than our conclusion regarding freedom from regulation. (26)

CONCLUSIONS

In theory, corruption can be either harmful or beneficial to growth, depending on the quality of the institutional environment. Several cross-country studies have found that corruption slows growth, but these findings are not universally robust. Recent attempts to control for the quality of institutions when examining the impact of corruption on growth have pointed to a potential nonlinear relationship among them that depends on the quality of the institutions. However, most of the institutional measures previously employed control for the quality of political rather than economic institutions. By directly examining how controlling for economic freedom impacts corruption's effect on growth, we have more directly tested some areas in which corruption may allow entrepreneurs to circumvent bad economic policies that would otherwise reduce growth.

Our initial findings are that corruption can have a positive effect on growth, most likely by allowing people to circumvent inefficient public policies. We further find that the benefits of corruption fall as the economic institutional environment improves. By breaking down the economic freedom index into each of its five areas, we find that corruption is only significantly growth enhancing when countries have low levels of freedom in the areas of government size, freedom to trade internationally, and regulation of credit, labor, and business. When economic freedom in the areas of government size and regulations improve, the benefits of corruption for growth significantly decline, and eventually turn negative.

These findings suggest that policy efforts to lower corruption across the board may not always improve economic growth rates. Instead, the particular form of corruption and the institutional quality of the country need to be addressed. In some cases, eliminating corruption may improve growth. But in other instances, such as in which inefficient rules limit entrepreneurial opportunities, ending corruption by solving principle agent problems might not improve growth. In these cases corruption with inefficient institutions is a second best result. Efforts at reform should instead focus on improving economic freedom to reduce the need for corruption rather than ending the discretion of decision-makers. Only after strong economic institutions are in place would reducing corruption be likely to improve growth prospects. On the basis of our regression analysis, reducing the size of government and decreasing the regulation of credit, labor, and business would do the most to alleviate the need for corruption to enhance growth. Furthermore, improvements in the legal structure and property rights, sound money policies, and freer trade, would also have direct benefits to growth independent of corruption.

Much research remains to be done. Current measures of corruption limit our ability to measure the specific forms of corruption occurring in different countries. More detailed measures of corruption could prove quite illuminating. Given that changes in institutions and policies often affect growth through an investment channel, further studies of the effect of the direct effect of corruption on investment when controlling for, and interacting with, economic freedom could also prove useful. Most importantly, additional research is necessary to deal with simultaneity issues. We attempt to minimize simultaneity issues by using 1995-2000 corruption to explain 2000-2005 growth, but long-term underlying factors, such as culture and other informal institutions, could influence both corruption and growth. Future research might make use of instrumental variables to address this issue. Finally, additional research could also examine in what instances the opportunity for corruption leads public officials to create inefficient institutions in the first place.

APPENDIX A

Components of economic freedom of the world

Area 1: Size of government: Expenditures, taxes, and enterprises

(a) General government consumption spending as a percentage of total consumption

(b) Transfers and subsidies as a percentage of GDP

(c) Government enterprises and investment as a percentage of total investment

(d) Top marginal tax rate and threshold at which it applies

(i) top marginal income tax rate (and threshold)

(ii) top marginal income and payroll tax rate (and threshold)

Area 2: Legal structure and property rights

(a) Judicial independence

(b) Impartial courts

(c) Protection of intellectual property

(d) Military interference in rule of law and the political process (e) Integrity of the legal system

Area 3: Access to sound money

(a) Average annual growth of the money supply in the last 5 years minus average annual growth of real GDP in the last 10 years

(b) Standard inflation variability during the last 5 years

(c) Recent inflation rate

(d) Freedom to own foreign currency bank accounts domestically and abroad

Area 4: Freedom to trade internationally

(a) Taxes on international trade

(i) Revenue from taxes on international trade as a percentage of exports plus imports

(ii) Mean tariff rate

(iii) Standard deviation in tariff rate

(b) Regulatory barriers to trade

(i) Non-tariff barriers

(ii) Compliance cost of importing and exporting

(c) Actual size of trade sector compared to expected size

(d) Difference between official exchange rate and black-market rate

(e) International capital market controls

(i) Foreign ownership/investment restrictions

(ii) Restrictions on the freedom of citizens to engage in capital market exchange with foreigners

Area 5: Regulation of credit, labor, and business

(a) Credit market restrictions

(i) Ownership of banks--percentage of deposits held in privately owned banks

(ii) Competition--domestic banks face competition from foreign banks

(iii) Extension of credit--percentage of credit extended to private sector

(iv) Avoidance of interest rate controls and regulations that lead to negative real interest rates

(v) Interest rate controls

(b) Labor market regulations

(i) Impact of minimum wage

(ii) Hiring and firing practices (determined by private contract)

(iii) Share of labor force whose wages are set by centralized collective bargaining

(iv) Unemployment benefits

(v) Use of conscripts to obtain military personnel

(c) Business regulations

(i) Price controls

(ii) Burden of regulation

(iii) Time with government bureaucracy

(iv) Ease of starting a new business

(v) Irregular payments (omitted from this study because it is a measure of corruption)

APPENDIX B

See Table B1.

Table B1: Sample of countries

Albania (a)         Latvia (a)
Argentina           Lithuania (a)
Australia           Malawi
Austria             Malaysia
Bangladesh          Mauritius
Belgium             Mexico
Bolivia             Morocco (a)
Botswana            Namibia (a)
Brazil              The Netherlands
Bulgaria            New Zealand
Cameroon            Nicaragua
Canada              Nigeria (a)
Chile               Norway
China               Pakistan
Colombia            Paraguay
Costa Rica          Peru
Cote d'Ivoire (a)   Philippines
Croatia (a)         Poland
Czech Republic      Portugal
Denmark             Romania
Ecuador             Russia (a)
Egypt               Senegal
El Salvador         Slovak Republic
Estonia (a)         Slovenia
Finland             South Africa
France              South Korea
Germany             Spain
Ghana               Sweden
Greece              Switzerland
Guatemala           Tanzania
Honduras            Thailand
Hungary             Tunisia
India               Turkey
Indonesia           Uganda
Ireland             Ukraine (a)
Israel              United Kingdom
Italy               United States
Jamaica             Uruguay
Japan               Venezuela
Jordan              Zambia
Kenya               Zimbabwe

(a) Missing education data from Barro-Lee.

Acknowledgements

We thank participants at the 2007 Southern Economics Association, and the 2007 Association of Private Enterprise conferences, the editor, and two anonymous referees, for valuable comments, Robert Lawson for providing a recalculated index of economic freedom with its corruption component removed, Lisa Verdon for providing data, and Kyle Jackson for excellent research assistance. The usual caveat applies.

REFERENCES

Aidt, T. (2003): Economic analysis of corruption: A survey. Economic Journal 113: 632-652.

Aidt, T, Dutta, J and Sen& V. (2008): Governance regimes, corruption and growth: Theory and evidence. Journal of Comparative Economics 36: 195-220.

Aron, J. (2000): Growth and institutions: A review of the evidence. The World Bank Research Observer 15: 99-135.

Ayal, EB and Karras, G. (1999): Components of economic freedom and growth: An empirical study. Journal of Developing Areas 32: 327-338.

Bardhan, P. (1997): Corruption and development: A review of issues. Journal of Economic Literature 35: 1320-1346.

Barro, R. (1997): Determinants of economic growth. MIT Press: Cambridge, MA.

Barro, R and Lee, J-W. (2000): International data on educational attainment: Updates and implications, http://www.cid.harvard.edu/ciddata/ciddata.html, accessed March 2007.

Brunetti, A, Kisunko, G and Weder, B. (1997): Credibility of rules and economic growth--Evidence from a world wide private sector survey. Background paper for the World Development Report 1997, The World Bank: Washington, D.C.

Carlsson, F and Lundstrom, S. (2002): Economic freedom and growth: Decomposing the effects. Public Choice 112: 335-344.

Caudill, SB, Zanella, FC and Mixon Jr, FG. (2000): Is economic freedom one dimensional? A factor analysis of some common measures of economic freedom. Journal of Economic Development 25: 17-40.

Clague, C. (2003): The international campaign against corruption: An institutionalist approach. In: Heckelman, JC and Coates, D (eds). Collective Choice: Essays in Honor of Mancur Olson. Springer-Verlag: Berlin.

Clarke, GRG. (1995): More evidence on income distribution and growth. Journal of Development Economics 47: 403-427.

Colombatto, E. (2003): Why is corruption tolerated? Review of Austrian Economics 164: 363-379.

Folster, S and Henrekson, M. (1999): Growth and the public sector: A critique of the critics. European Journal of Political Economy 15:337-358.

Goel, R and Nelson, M. (2005): Economic freedom versus political freedom: Cross-country influences on corruption. Australian Economic Papers 44: 121-133.

Graeff, P and Mehlkop, G. (2003): The impact of economic freedom on corruption: Different patterns for rich and poor countries. European Journal of Political Economy 19: 605-620.

Gwartney, J and Lawson, R. (2006): Economic freedom of the world: 2006 annual report. The Fraser Institute: Vancouver, Canada.

Gwartney, J, Lawson, R and Holcombe, R. (1999): Economic freedom and the environment for economic growth. Journal of Institutional and Theoretical Economics 155: 643-663.

Gyimah-Brempong, K. (2002): Corruption, economic growth, and income inequality in Africa. Economics of Governance 3: 183-209.

Heckelman, JC and Knack, S. (2008): Foreign aid and market liberalizing reform. Economica 75: 524-548.

Heckelman, JC and Stroup, MD. (2000): Which economic freedoms contribute to growth? Kyklos 53: 527-544.

Huntington, S. (1968): Political order in changing societies. Yale University Press: New Haven, CT.

Jain, AK. (2001): Corruption: A review. Journal of Economic Surveys 15: 71-121.

Kaufmann, DA, Kraay, A and Zoido-Lobaton, P. (1999): Governance matters. Policy Research Working Paper 2195. The World Bank: Washington D.C.

Klitgaard, R. (1988): Controlling corruption. University of California Press: Berkeley, CA.

Knack, S and Keefer, P. (1995): Institutions and economic performance: Cross-country tests using alternative institutional measures. Economics and Politics 7: 207-227.

Krueger, A. (1974): The political economy of the rent seeking society. American Economic Review 64: 291-303.

Left, N. (1964): Economic development through bureaucratic corruption. American Behavioral Scientist 82: 337-341.

Leite, C and Weidmann, J. (1999): Does mother nature corrupt? Natural resources, corruption, and economic growth. International Monetary Fund Working Paper, 99/85.

Lui, FT. (1985): An equilibrium queuing model of bribery. Journal of Political Economy 93: 760-781.

Mauro, P. (1995): Corruption and growth. Quarterly Journal of Economics 110: 681-712.

Meon, P-G and Sekkat, K. (2005): Does corruption grease or sand the wheels of growth? Public Choice 122: 69-97.

Mendez, F and Sepulveda, F. (2006): Corruption, growth and political regimes: Cross country evidence. European Journal of Political Economy 22: 82-98.

Mo, PH. (2001): Corruption and economic growth. Journal of Comparative Economics 29: 66-79.

Munck, G and Verkuilen, J. (2002): Conceptualizing and measuring democracy: Evaluating alternative indices. Comparative Political Studies 35: 3-34.

Myrdal, G. (1968): Asian drama: An inquiry into the poverty of nations Vol. 2. The Twentieth Century Fund: New York.

Paldam, M. (2002): The big pattern of corruption, economics, culture and the seesaw dynamics. European Journal of Political Economy 18: 215-240.

Poirson, H. (1998): Economic security, private investment, and growth in developing countries. International Monetary Fund Working Paper, 98/4.

Rose-Ackerman, S. (1999): Corruption and government, causes, consequences and reform. Cambridge University Press: Cambridge, UK.

Shleifer, A and Vishny, RW. (1993): Corruption. Quarterly Journal of Economics 108: 599-617.

Svensson, J. (2005): Eight questions about corruption. Journal of Economic Perspectives 19: 19-42.

Transparency International. (1995-2000): Corruption Perceptions Index, www.transparency.org/ policy_research/surveys_indices/cpi, accessed February 2007.

Wedeman, A. (1997): Looters, rent-scrapers, and dividend-collectors: Corruption and growth in Zaire, South Korea, and the Philippines. Journal of Developing Areas 31: 457-478.

World Bank. (2007): World development indicators online www.worldbank.org, accessed February 2007.

(1) Full website address: http://webworldbankorg/WBSITE/EXTERNAL/TOPICS/EXTPUBLIC- SECTORANDGOVERNANCE/EXTANTICORRUPTION/0,contentMDK:21205078~menuPK:384461~ pagePK:64020865~piPK:149114~theSitePK:384455,00.html, accessed May 2007.

(2) Full website address given in note 1 above.

(3) Full website address: www.usaid.gov/our_work/democracy_and_governance/technical_areas/ anti corruption/, accessed May 2007.

(4) In the quote above, USAID acknowledged that corruption may be able to reduce costs associated with bureaucratic red tape.

(5) For earlier surveys of the literature on corruption see Bardhan (1997), Rose-Ackerman (1999), Jain (2001), and Aidt (2003).

(6) We recognize that inefficient institutional environments that do not allow much economic freedom may in some cases be intentionally created to allow officials to extract bribes. Furthermore, the underlying cultural capital could jointly cause both corruption and low levels of economic freedom.

(7) Paldam (2002), Graeff and Mehlkop (2003), and Goel and Nelson (2005) have used the indexes of economic freedom to examine how economic freedom impacts corruption. They have generally found that the more economic freedom a country has the lower the level of corruption present; however, their studies did not examine how this relationship affects growth. As we show below, the inverse relationship between corruption and economic freedom is important for predicting the impact corruption will have on growth.

(8) For a general comparison between alternative measures of democracy, see Munck and Verkuilen (2002).

(9) Scores for each of the separate components of the political rights and civil liberties indexes are not publicly available, therefore the Freedom House democracy index cannot be purged of these particular elements.

(10) The Freedom House scores are based on rank, therefore lower values represent more freedom. The indexes were inverted in order to match the Polity method of higher values representing greater democracy, and thus result in positive correlations.

(11) Because each area contains a different number of components, this aggregation method (which weights each area equally) does not weight all the individual components equally.

(12) We thank Robert Lawson for rescaling the Area 5 scores and the overall EFW index without component C.v. when he sent us the EFW data.

(13) Area 4 (freedom to exchange with foreigners) component C of the EFW index includes a measure of the actual size of the international trade sector compared to the expected size. This measure may suffer from similar problems, but it does take account of structural and geographic characteristics of the country when calculating the expected size. Furthermore, it only accounts for 4% of the overall EFW score (20% of a country's Area 4 score).

(14) The regions include: Latin America, Asia, Europe, Africa, and Middle East. North America (comprised of Canada and United States; Mexico is included as part of Latin America) is the default region not included.

(15) The same is true for some of the components of economic freedom, described above.

(16) Alternative education measures from Barro and Lee (2000), such as primary and secondary completion rates, were also utilized but did not appreciably affect our results.

(17) See Ayal and Karras (1999) and Heckelman and Stroup (2000), who show differing marginal impacts of the economic freedom components on growth using earlier versions of the EFW that contained fewer total components.

(18) The test statistic of 2.36 is distributed as F(13, 68) with p value of 0.0l. An alternative test statistic of N x R2- 25.52 is distributed as [chi square] (13) with p-value of 0.02.

(19) Our results do not substantially differ if we applied a two-step WLS, using the residuals from OLS as the weighting proportion.

(20) WLS contains no constant term, as the intercept is now the inverse of population (relative to the mean). As such, we do not report R2 measures because they do not retain their normal interpretation without a constant term. We also do not report specific estimates for the various regional dummies (Latin America, Asia, Europe, Africa, Middle East).

(21) The five highest-rated countries for EFW over the period 1995-2000 are United States (8.28), New Zealand (8.27), United Kingdom (8.06), Ireland (8.05), and Switzerland (8.00).

(22) As noted earlier, the CPI is not consistently measured over time due to the changing inclusion of varying surveys. As a check on our results, we ran another regression matching specification IV, but using only the 2000 values for all independent variables. This reduced our sample to 71 nations but our conclusions are similar. The coefficient on corruption is positive and significant, and the coefficient for the interaction with the EFW index is negative and significant. The estimated turning point for when additional corruption becomes harmful occurs at a slightly lower, but still quite high. rating for EFW at 8.26. roughly in accordance with US levels. The democracy index remains statistically insignificant.

(23) Neither country from the default region (Canada and US) is in the low EFW sample. To avoid perfect multicollinearity, one of the regional dummies used in previous regressions (see note 14) was dropped for that regression.

(24) The weighted standard deviation of growth in the low EFW sample is 96.9 with a mean of 25.5.

(25) The highest-rated countries for Area 5 are New Zealand (6.63), United States (6.60), Namibia (6.59), and United Kingdom (6.57).

(26) Another reason we are less certain about Area 1 than Area 5 is that the finding of significance of Area 5 was more robust across specifications in Table 3.

JAC C HECKELMAN [1] & BENJAMIN POWELL [2]

[1] Department of Economics, Wake Forest University, Carswell Hall, Winston-Salem, NC 27109, USA. E-mail: heckeljc@wfu.edu

[2] Department of Economics, Suffolk University, 8 Ashburton PL, Boston, MA 02108, USA. E-mail: benjaminwpowell@gmail.com

Table 1: Descriptive statistics: Means with standard deviation in
parentheses

Sample              Full sample      Democracy      Democracy
                                      < mean          > mean
                        (I)            (II)           (III)

Growth             11.87 (11.73)   12.21 (16.19)   11.72 (9.47)
Log initial GOP     8.92 (1.09)     8.07 (0.99)    9.27 (0.93)
Investment/GDP     22.03 (4.72)    21.05 (5.39)    22.44 (4.41)
Corruption index    5.32 (2.27)     6.92 (1.03)    4.66 (2.32)
Democracy           7.94 (2.55)     4.40 (1.83)    9.40 (0.68)
EFW index           6.28 (1.00)     5.59 (0.72)    6.56 (0.97)
  Area 1 index      5.55 (1.61)     5.59 (1.33)    5.54 (1.72)
  Area 2 index      6.20 (1.75)     5.02 (0.91)    6.69 (1.79)
  Area 3 index      7.17 (2.19)     6.10 (1.80)    7.61 (2.19)
  Area 4 index      7.05 (1.00)     6.29 (0.84)    7.36 (0.90)
  Area 5 index      5.48 (0.57)     5.11 (0.52)    5.64 (0.52)
N                       82              24              58

Sample               EFW index      EFW index
                      < mean          > mean
                       (IV)            (V)

Growth             13.89 (14.81)   9.74 (6.78)
Log initial GOP     8.58 (1.04)    9.27 (1.04)
Investment/GDP     21.58 (5.10)    22.51 (4.31)
Corruption index    6.70 (1.19)    3.88 (2.25)
Democracy           6.65 (2.79)    9.29 (1.31)
EFW index           5.47 (0.59)    7.11 (0.55)
  Area 1 index      5.32 (1.50)    5.80 (1.71)
  Area 2 index      5.20 (1.04)    7.25 (1.75)
  Area 3 index      5.59 (1.81)    8.84 (0.97)
  Area 4 index      6.48 (0.95)    7.64 (0.65)
  Area 5 index      5.12 (0.45)    5.87 (0.41)
N                       42              40

Table 2: Correlations between corruption, democracy, and economic
freedom

             Corruption   Democracy    EFW    Area 1   Area 2   Area 3

Democracy      -0.529
EFW index      -0.760       0.493
  Area 1        0.306      -0.024     0.206
  Area 2       -0.927       0.500     0.706   -0.430
  Area 3       -0.604       0.345     0.898    0.117   0.581
  Area 4       -0.670       0.574     0.699   -0.128   0.611    0.537
  Area 5       -0.632       0.439     0.781    0.228   0.537    0.590

             Area 4

Democracy
EFW index
  Area 1
  Area 2
  Area 3
  Area 4
  Area 5     0.527

Table 3: Growth regressions

Specification                              I

Intercept                          -16.47 (-0.76)
Log initial GDP per capita          -3.59 ** (-2.88)
Investment                           2.18 ** (8.62)
Democracy                            1.12 ** (2.69)
Corruption                           3.50 ** (2.93)
Corruption x Democracy
Corruption x EFW index

Economic freedom
Area 1: Size of government
Area 2: Legal structure
Area 3: Sound money
Area 4: Freedom to trade
Area 5: Regulation

Corruption effect turns positive
Corruption effect turns negative
Regression F-statistic             142.47 **

Specification                             II

Intercept                          -43.22 (-1.40)
Log initial GDP per capita          -5.79 ** (-4.71)
Investment                           0.97 ** (2.57)
Democracy                           -1.21 * (-1.90)
Corruption                           6.09 ** (3.73)
Corruption x Democracy
Corruption x EFW index

Economic freedom
Area 1: Size of government          -2.31 * (-1.88)
Area 2: Legal structure              5.01 ** (2.30)
Area 3: Sound money                  0.05 (0.05)
Area 4: Freedom to trade            -0.17 (-0.11)
Area 5: Regulation                   9.63 ** (2.26)

Corruption effect turns positive
Corruption effect turns negative
Regression F-statistic             124.86 **

Specification                             III

Intercept                           94.03 * (1.85)
Log initial GDP per capita          -5.06 **(-4.32)
Investment                           0.50 (1.30)
Democracy                          -13.98 ** (-3.55)
Corruption                         -11.13 ** (-2.03)
Corruption x Democracy               1.72 ** (3.28)
Corruption x EFW index

Economic freedom
Area 1: Size of government          -1.38 (-1.17)
Area 2: Legal structure              4.93 ** (2.42)
Area 3: Sound money                 -0.69 (-0.66)
Area 4: Freedom to trade             0.89 (0.61)
Area 5: Regulation                   7.36 * (1.82)

Corruption effect turns positive   Democracy > 6.48
Corruption effect turns negative
Regression F-statistic             134.18 **

Specification                              IV

Intercept                          -190.01 ** (-3.90)
Log initial GDP per capita           -5.87 ** (-5.21)
Investment                            0.75 ** (2.13)
Democracy                             0.07 (0.11)
Corruption                           24.57 ** (4.72)
Corruption x Democracy
Corruption x EFW index               -2.55 ** (-3.70)

Economic freedom
Area 1: Size of government           -0.08 (-0.06)
Area 2: Legal structure               7.94 ** (3.69)
Area 3: Sound money                   4.15 ** (2.79)
Area 4: Freedom to trade              6.92 ** (2.92)
Area 5: Regulation                   10.62 ** (2.72)

Corruption effect turns positive
Corruption effect turns negative   EFW > 9.64
Regression F-statistic              139.59 **

* significant at 10%; ** significant at 5%.

Notes: Regressions estimated by Weighted Least Squares 310.00
population. t-statistics in parentheses. Regressions also include
regional dummies not reported. N=82.

Table 4: Growth regressions split sample by EFW

                                                  EFW [greater than
Sample                         EFW < median      or equal to] median

Log initial GDP per capita    -2.89 * (-2.85)      2.11 (1.06)
Investment                     1.60 ** (14.92)     0.68 * (1.89)
Democracy > median             2.92 * (1.84)      -4.98 * (-1.53)
Corruption 1st quartile      -15.82 (-0.31)      -21.34 (-0.95)
Corruption 2nd quartile       -7.19 (-0.57)      -23.07 (-1.03)
Corruption 3rd quartile        3.67 (0.36)        -8.44 (-0.42)
Corruption 4th quartile       19.00 * (1.81)     -12.73 (-0.65)
N                                   41                   41
Regression F-statistic       142.47 **           124.86 **

* significant at 10%; ** significant at 5%.

Notes: Regressions estimated by Weighted Least Squares using initial
population. t-statistics in  parentheses. Regressions also include
regional dummies not reported.

Table 5: Sensitivity results for the effect of corruption on growth

Specification change       N      Corruption      Corruption x EF

1. Drop democracy          84   24.31 ** (5.32)   -2.51 ** (-4.34)
2. Drop investment         82   26.46 ** (5.02)   -2.81 ** (-4.03)
3. Include education       72   25.12 ** (4.10)   -2.57 ** (-3.04)
4. Drop regional dummies   82   27.26 **( 5.67)   -3.05 ** (-5.09)
5. EF Area 1 interaction   82   13.52 ** (7.01)   -2.26 ** (-5.45)
6. EF Area 2 interaction   82    3.39 (0.61)       0.30 (0.51)
7. EF Area 3 interaction   82    6.63 (1.54)      -0.06 (-0.14)
8. EF Area 4 interaction   82   11.63 ** (1.97)   -0.76 (-0.98)
9. EF Area 5 interaction   82   39.29 ** (5.86)   -5.89 ** (-5.06)

Specification change       Corruption effect
                            turns negative

1. Drop democracy             EFW > 9.69
2. Drop investment            EFW > 9.42
3. Include education          EFW > 9.77
4. Drop regional dummies      EFW > 8.94
5. EF Area 1 interaction    EF areal > 5.97
6. EF Area 2 interaction          --
7. EF Area 3 interaction          --
8. EF Area 4 interaction
9. EF Area 5 interaction   EF area 5 > 6.67

* significant at 10%; ** significant at 5%.

Notes: Regressions estimated by Weighted Least Squares using initial
population. t-statistics in parentheses. Each regression, except where
otherwise noted, also includes: log initial GDP per capita,
investment, measure of democracy, economic freedom for Areas 1-5, and
a set of regional dummies. The interaction for Corruption x EF uses
the overall Economic Freedom of the World Index except in the last
five rows in which EF is limited to a specific area score.