81285 Resource Windfalls and Emerging Market Sovereign Bond Spreads: The Role of Political Institutions ¨ ckner* Rabah Arezki and Markus Bru We examine the effect that revenue windfalls from international commodity price booms have on sovereign bond spreads using panel data for 38 emerging market econ- omies during the period 1997-2007. Our main finding is that commodity price booms lead to a significant reduction in the sovereign bond spread in democracies, but to a significant increase in the spread in autocracies. To explain our finding we show that, consistent with the political economy literature on the resource curse, revenue wind- falls from international commodity price booms significantly increased real per capita GDP growth in democracies, while in autocracies GDP per capita growth decreased. JEL codes: C33, D73, D74, D72, H21. I. INTRODUCTION Some researchers have argued that international commodity price booms may spawn an over-accumulation of external debt in commodity exporting countries that increases the risk of external debt default (e.g. Krueger, 1987; Berg and Sachs, 1988).1 We examine this hypothesis empirically by analyzing how the spread on sovereign bonds reacted in these countries to the booms and slumps of the export-relevant commodity prices. Changes in the spread on sovereign bonds reflect changes in investors’ beliefs of the risk that a country * International Monetary Fund (Arezki, corresponding author) and University of Adelaide (Bruckner). Contact e-mails: rarezki@imf.org; markus.bruckner@adelaide.edu.au. We thank three anonymous referees, the editor Elisabeth Sadoulet, and members of the editorial board for helpful comments and suggestions. We are grateful to Amine Mati for providing us with his dataset on sovereign bond spreads and to Daniel Lederman for providing us with his dataset on export diversification. The views in this paper are those of the authors alone and do not necessarily represent those of the IMF or IMF policy. All remaining errors are our own. Bru ¨ ckner gratefully acknowledges the financial support of the Spanish Ministry of Science and Technology provided by CICYTECO2008- 04997. 1. The recent concern that Dubai may default on its external debt is an example par excellence that higher commodity prices may be associated with a higher risk of external debt default. Further examples are, among others, Russia and Nigeria. THE WORLD BANK ECONOMIC REVIEW, VOL. 26, NO. 1, pp. 78– 99 doi:10.1093/wber/lhr015 Advance Access Publication May 18, 2011 # The Author 2011. Published by Oxford University Press on behalf of the International Bank for Reconstruction and Development / THE WORLD BANK. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com 78 ¨ ckner Arezki and Bru 79 defaults on its external debt. An increase in the spread on sovereign bonds is in turn a cost for the bond issuing country that may trigger in a self-fulfilling way the default on its external debt. Both for investors and policy makers, it is therefore important to have knowledge about how international commodity price shocks, which induce large upturns and downturns in foreign currency revenues in emerging market economies, affect the spread on sovereign bonds. We find that increases in international commodity prices for exported com- modity goods are associated with a significant reduction in sovereign bond spreads on average. However, the reduction in the spread on sovereign bonds is particularly large in countries with sound democratic institutions and strong political checks and balances. In autocratic regimes and countries where the political rule is characterized by weak checks and balances, windfalls from international commodity prices lead to a significant increase in the spread on sovereign bonds. The heterogeneous response of sovereign bond spreads to international com- modity price shocks sheds new light on the resource curse literature, that has argued for the importance of political institutions in determining whether windfalls from natural resources are a curse or a blessing for the economic development of resource exporting countries (e.g. Mehlum et al., 2006; Robinson et al., 2006).2 We provide further evidence in this direction by showing that, consistent with the political economy model developed in Mehlum et al. (2006), international commodity price booms significantly increased real per capita GDP growth in countries with sound democratic insti- tutions. In countries with autocratic institutions, revenue windfalls from inter- national commodity price booms led to a significant decrease in output growth. Hence, while our empirical results are consistent with general equili- brium models that predict a countercyclical relationship between sovereign bond spreads and the business cycle in emerging market economies (e.g. Arellano, 2008), our results highlight the importance of political economy factors in shaping the relationship between commodity price shocks and sover- eign bond spreads in these countries. The remainder of our paper is organized as follows. Section II describes the data. Section III discusses the estimation strategy. Section IV presents the main results. Section V concludes. I I . D ATA COMMODITY REVENUE WINDFALLS. We construct a country-specific international commodity export price index that captures revenue windfalls from 2. See also Van der Ploeg (2010) for a review and overview of the resource curse literature. 80 THE WORLD BANK ECONOMIC REVIEW international commodity prices as: Y ComPIi;t ¼ ComPricec;tui;c c [C where ComPricec,t is the international price of commodity c in year t, and ui,c is the average (time-invariant) value of exports of commodity c in the GDP of country i.3 We obtain data on annual international commodity prices from UNCTAD Commodity Statistics and our data on the value of commodity exports are from the NBER-United Nations Trade Database. The commodities included in our index are aluminum, beef, coffee, cocoa, copper, cotton, gold, iron, maize, oil, rice, rubber, sugar, tea, tobacco, wheat, and wood. In case there were multiple prices listed for the same commodity we used a simple average of all the relevant prices. We note that even though some of the countries in our sample are net resource importers (in sum, across all commodities) our commodity export price index captures that there may still be some commodities for which the country is an exporter. For example, according to Lederman and Maloney (2008) Egypt is a net natural resource importer. However, Egypt also exports a significant amount of crude oil. When the international price of oil increases Egypt experi- ences a positive revenue windfall, and this is captured by our export price index. On the other hand, when the international prices of other commodities increase Egypt experiences a negative terms of trade shock but not necessarily a negative revenue shock (which depends among other things on the structure of ad valorem import duties). We therefore follow the resource curse literature (e.g. Sachs and Warner, 1995, 2001) and focus on a gross export price index as our measure for resource windfalls. As a robustness check we will present estimates that are restricted to the sample of countries that are net natural resource exporters. SOVEREIGN BOND SPREADS. Our data on the spread on sovereign bonds are from the Emerging Markets Bond Index Global (EMBI Global). The bond spreads are measured against a comparable US government bond and are period averages for the whole year. POLITICAL INSTITUTIONS. Our two main measures of political institutions are the average (time-invariant) Polity2 score from the Polity IV database (Marshall and Jaggers, 2009) and the average (time-invariant) checks and balance score from the Database of Political Institutions (Beck et al., 2001). The Polity2 score is based on the constraints placed on the chief executive, the competitiveness of political participation, and the openness and competitiveness of executive recruitment. The Polity2 score ranges from 2 10 to þ 10, with higher values 3. This functional form of the commodity export price index follows common practice in the literature. See for example Collier and Goderis (2007) and the references cited therein. ¨ ckner Arezki and Bru 81 indicating stronger democratic institutions. The checks and balance score is based on the number of veto players in the political system, their respective party affiliations, and the electoral rules. The checks and balance score ranges between 1 to 6, with higher values indicating stronger checks and balances. Following Persson and Tabellini (2003, 2006) and the Polity IV project we also construct an autocracy indicator variable that takes on the value of unity in countries with negative (average) Polity2 scores. The main purpose of this auto- cracy indicator variable is to facilitate the interpretation of the results from the regression analysis. Note that we use countries’ average polity and checks and balance scores because we want to capture long-run and thus more fundamental differences in countries’ political institutions. Countries’ political institutions are also highly persistent as about three-fourths of the countries in our sample did not experience changes in their political institutions score. OTHER CONTROL VARIABLES. Data on real per capita GDP are from the Penn World Table, version 6.3 (Heston et al., 2009). Data on corruption are from Political Risk Service (2010). Data on ethnic fractionalization are from Alesina et al. (2003). Data on the Herfindahl index of export diversification are from Lederman and Xu (2010). Data on the Gini coefficient are from the World Development Indicators (2010). Data on British colonial origin, French colonial origin, and historical settler mortality are from Acemoglu et al. (2001). Descriptive statistics of these variables are provided in Data Appendix Table 1. A list of countries included in the sample is provided in Data Appendix Table 2. I I I . E S T I M AT I O N S T R AT E GY To examine the effects that revenue windfalls from international commodity price booms have on sovereign bond spreads, we estimate the following econo- metric model: DlogðSpreadi ; tÞ ¼ ai þ bt þ hDlogðComPIi;t Þ þ ui;t where ai are country fixed effects and bt are year fixed effects. ui,t is an error term that is clustered at the country level. As a baseline regression, we estimate the average marginal effect h that commodity price booms have on sovereign bond spreads. We then examine how this marginal effect varies as a function of countries’ political institutions by estimating: DlogðSpreadi;t Þ ¼ ai þ bt þ cDlogðComPIi;t Þ þ dDlogðComPIi;t Þ Ã Poli þ ei;t where Poli is a measure of cross-country differences in political institutions. In order for the estimate on the parameter c to reflect the average marginal effect we compute Poli for the Polity2 score as the Polity2 score of country i minus the Polity2 sample average. Formally: Poli ¼ Polity2i - Avg.(Polity2). We do the same for the checks and balance score. This rescaling does not affect the 82 THE WORLD BANK ECONOMIC REVIEW parameter estimate d but it is useful for interpretation purposes as it ensures that the parameter estimate c reflects the average marginal effect (i.e. the effect for the “average” country). Note that our measures of political institutions Poli are time-invariant and therefore we do not need to control for them in the fixed effects regression (the reason is that the direct effect of these variables on the sovereign bond spread is already accounted for by the country fixed effects ai). We estimate both static and dynamic panel data models. For the dynamic panel data model we report system-GMM estimates (Blundell and Bond, 1998) as the presence of country fixed effects leads the fixed effects estimator to produce inconsistent estimates.4 We address the important issue of political institutions being correlated with other cross-sectional variables that could possibly affect the relationship between commodity price booms and sovereign bond spreads by including additional interaction terms in the regression. In particular, we include in all regressions an additional interaction term between DComPI and cross-country differences in GDP per capita. In addition, we use instrumental variables tech- niques to further address endogeneity biases. In particular, we build on the seminal work of Acemoglu et al. (2001) and instrument the political insti- tutions interaction term Pol*DComPI with the interaction between DComPI and indicator variables for colonial origin and historical settler mortality. We test the validity of these instrumental variables using the Hansen test. I V. M A I N R E S U L T S Table 1, column (1) presents our estimates of the average marginal effect that resource windfalls from international commodity price booms have on sover- eign bond spreads in the largest possible sample of 38 emerging market econ- omies during the period 1997-2007. The main finding is that commodity windfalls lead on average to a significant reduction in sovereign bond spreads. Panel A presents panel data estimates that control for country fixed effects and Panel B presents panel data estimates that control in addition to the country fixed effects for year fixed effects. The panel data estimates reported in column (1) imply that an increase in the commodity export price index of size 1 stan- dard deviation would significantly reduce the spread on sovereign bonds on average by over 0.1 standard deviations. Column (2) of Table 1 shows that the marginal effect of international com- modity price booms on the spread on sovereign bonds significantly varies across countries as a function of cross-country differences in political 4. In the system-GMM estimation we use the first and second lags as instruments for the lagged dependent variable to reduce the concern that too many moment conditions are used (for further discussion on this issue see e.g. Roodman, 2009). We note that the dynamic panel data bias associated with the fixed effects estimator is bounded of order T 21, where T is the time-series dimension of the panel (see Nickell, 1981). For comparison purposes we also report estimates from the fixed effects estimator. ¨ ckner Arezki and Bru 83 T A B L E 1 . Commodity Windfalls, Political Institutions, and the Spread on Sovereign Bonds (Static Panel Regression) DSpread Panel A: Controlling for Country Fixed Effects (1) (2) (3) (4) LS LS LS LS DComPI 2 10.950*** 2 7.417** 2 29.694*** 2 8.072*** ( 2 3.03) ( 2 2.26) ( 2 4.27) ( 2 2.83) DComPI* 2 2.610*** Avg. Polity2 Score ( 2 2.81) DComPI* 55.815*** Autocracy Indicator (4.15) DComPI* 2 16.939*** Avg. Checks & Balance Score ( 2 3.35) DComPI* 0.001** 0.004*** 0.002*** Avg. GDP Per Capita (1.98) (4.21) ( 2 2.63) Country Fixed Effects Yes Yes Yes Yes Year Fixed Effects No No No No Observations 291 291 291 291 Panel B: Controlling for Country and Year Fixed Effects (1) (2) (3) (4) LS LS LS LS DComPI 2 6.127* 2 1.644 2 20.727*** 2 3.108 ( 2 1.72) ( 2 0.37) ( 2 3.46) ( 2 0.74) DComPI* 2 2.121** Avg. Polity2 Score ( 2 2.33) DComPI* 45.676*** Autocracy Indicator (3.57) DComPI* 2 11.420** Avg. Checks & Balance Score ( 2 2.17) DComPI* 0.002** 0.004*** 0.002** Avg. GDP Per Capita (2.13) (3.88) (2.09) Country Fixed Effects Yes Yes Yes Yes Year Fixed Effects Yes Yes Yes Yes Observations 291 291 291 291 Note: The method of estimation is least squares. t-values (in brackets) are based on Huber robust standard errors that are clustered at the country level. The dependent variable is the log- change in the spread on sovereign bonds. DComPI stands for the log-change in the international commodity export price index (see page 3 in the paper for a detailed explanation of how this vari- able is constructed). The cross-section (average time-series) dimension of the panel is 38 (7.7). *Significantly different from zero at the 10 percent significance level, ** 5 percent significance level, *** 1 percent significance level. institutions. The estimated interaction effect between revenue windfalls from international commodity price booms and the Polity2 score is negative and statistically significant at the 5% level. The point estimate on the interaction term implies that at the sample maximum Polity2 score (democracies), an increase in the commodity export price index of size 1 standard deviation 84 THE WORLD BANK ECONOMIC REVIEW F I G U R E 1. Commodity Windfalls, Political Institutions, and the Spread on Sovereign Bonds Note: The left-hand side figure shows the relationship between changes in countries’ commodity export price index and the spread on their sovereign bonds for countries that had on average a strictly positive Polity2 score. The right-hand side figure shows the relationship between changes in countries’ commodity export price index and the spread on their sovereign bonds for countries that had on average a negative Polity2 score. would significantly reduce the spread on sovereign bonds by over 0.3 standard deviations. On the other hand, at the sample minimum Polity2 score (autocra- cies), a shock of similar magnitude would be associated with a significant increase in the spread on sovereign bonds by 0.2 standard deviations. Column (3) of Table 1 shows that we obtain similar heterogeneity in the marginal effect of international commodity price booms on sovereign bond spreads when we discretize the Polity2 score into an autocracy indicator vari- able that is unity for negative Polity2 scores and zero otherwise. The significant positive coefficient on the autocracy interaction term implies that in autocracies revenue windfalls from commodity price booms significantly increased the spread on sovereign bonds, while in democracies sovereign bond spreads sig- nificantly decreased. Figure 1 illustrates this nonlinear relationship graphically. We show in column (4) of Table 1 as a robustness check on our measure of political institutions, that windfalls from international commodity price booms significantly decreased sovereign bond spreads in countries with strong checks and balances, while in countries with weak checks and balances the sovereign bond spreads significantly increased.5 Table 2 shows that our results are robust to controlling for lagged changes in the sovereign bond spread. Columns (1) to (3) present the least squares 5. We document in Appendix Table 1 that the results in Table 1 are robust to outliers. In particular, we report in columns (1)-(3) of Appendix Table 1 median (quantile) estimates, and in columns (4)-(6) least-squares estimates that exclude observations which fall in the top/bottom 1 percentile of the distribution of the change in the commodity export price index. T A B L E 2 . Commodity Windfalls, Political Institutions, and the Spread on Sovereign Bonds (Dynamic Panel Regression) DSpread (1) (2) (3) (4) (5) (6) LS LS LS GMM GMM GMM DComPI 4.123 2 16.369*** 1.984 2 0.032 2 14.994*** 2 2.224 (0.85) ( 2 2.76) (0.41) ( 2 0.01) ( 2 2.78) ( 2 0.56) DComPI* 2 2.305** 2 1.685** Avg. Polity2 Score ( 2 2.34) ( 2 2.43) DComPI* 49.324*** 33.101*** Autocracy Indicator (3.37) (2.68) DComPI* 2 10.407** 2 8.086** Avg. Checks & Balance Score ( 2 2.02) ( 2 2.17) DComPI* 0.003*** 0.005*** 0.002** 0.002*** 0.003*** 0.001* Avg. GDP Per Capita (2.87) (4.00) (2.56) (2.89) (1.91) (2.53) L.DSpread 0.183*** 0.182*** 0.180*** 0.241*** 0.231*** 0.232*** (3.73) (3.65) (3.58) (5.06) (5.22) (4.84) Hansen J, p-value . . . 0.232 0.220 0.259 AR(2) test, p-value . . . 0.125 0.151 0.134 Country Fixed Effects Yes Yes Yes Yes Yes Yes Year Fixed Effects Yes Yes Yes Yes Yes Yes Observations 253 253 253 253 253 253 Note: The method of estimation in columns (1)-(3) is least squares; columns (4)-(6) system-GMM (Blundell and Bond, 1998) with two-step Windmeijer (2005) small sample correction. t-values (in brackets) are based on Huber robust standard errors that are clustered at the country level. The Arezki and Bru dependent variable is the log-change in the spread on sovereign bonds. DComPI stands for the log-change in the international commodity export price index (see page 3 in the paper for a detailed explanation of how this variable is constructed). The cross-section (average time-series) dimension of the ¨ ckner panel is 37 (6.8). *Significantly different from zero at the 10 percent significance level, ** 5 percent significance level, *** 1 percent significance level. 85 86 THE WORLD BANK ECONOMIC REVIEW estimates and columns (4) to (6) present the system-GMM estimates. The dynamic panel data estimates reveal a significant positive autocorrelation in the log-change of the sovereign bond spreads. Importantly, they show that the interaction between changes in the commodity export price index and political institutions remains statistically significant at the 5% level when we allow for dynamics in the dependent variable. So far we only controlled in our regressions for an interaction term between changes in the commodity export price index and cross-country differences in GDP per capita. The GDP per capita interaction control is important because there exists a large literature that has argued for a positive effect of cross- country per capita income differences on political institutions (see for example Barro, 1999, or Przeworski et al., 2000). To demonstrate that the interaction between political institutions and commodity price windfalls is robust to additional interaction controls we report in Table 3 estimates when controlling for an interaction between changes in the commodity export price index and ethnic fractionalization, an interaction between changes in the commodity export price index and the Gini coefficient, an interaction between changes in the commodity export price index and a Herfindahl index of export diversifica- tion, and an interaction between changes in the commodity export price index and an indicator variable that is unity if the country is a net natural resource importer. Some of these additional interaction controls are indeed statistically significant. But nevertheless, the inclusion of these additional interaction con- trols on the right-hand side of the estimating equation continues to produce a significant interaction effect between commodity price booms and political institutions. Table 4 shows that we obtain similar results to our baseline estimates if we restrict the sample to the natural resource net-exporting countries. The natural resource net-exporting countries are strongly affected by the booms and slumps in the international commodity prices. It is thus reassuring from the standpoint of identification that in this restricted sample our results continue to hold. We can go even further and examine the relationship between commodity price windfalls, political institutions and sovereign bond spreads using instru- mental variables techniques that correct for possible endogeneity bias of the estimated interaction effect. Building on the seminal work by Acemoglu et al. (2001), we use historical settler mortality data and indicator variables of countries’ colonial origin as instrumental variables for political institutions. Table 5 reports our two-stage least squares estimates where the political insti- tutions interaction term is instrumented by the interaction between changes in the commodity export price index and the Acemoglu et al. instruments for institutions. The main result is that the political institutions interaction con- tinues to be significant in the instrumental variables regression. Also, with the exception of the autocracy interaction term the Hausman test does not indicate a significant difference between the least squares and instrumental variables estimates. We also note that the quality of the instrumental variables is good as T A B L E 3 . Commodity Windfalls, Political Institutions, and the Spread on Sovereign Bonds (Robustness to Additional Interaction Control Variables) DSpread (1) (2) (3) (4) (5) (6) LS LS LS GMM GMM GMM DComPI 2 16.203** 2 24.292*** 2 19.010** 2 17.839* 2 24.920** 2 20.853** ( 2 2.06) ( 2 2.70) ( 2 2.00) ( 2 1.92) ( 2 2.24) ( 2 2.16) DComPI* 2 2.572*** 2 2.305** Avg. Polity2 Score ( 2 3.74) ( 2 2.56) DComPI* 31.557*** 27.595** Autocracy Indicator (2.97) (2.35) DComPI* 2 8.884* 2 9.517** Avg. Checks & Balance Score ( 2 1.87) ( 2 2.10) DComPI* 0.007*** 0.006*** 0.006*** 0.006*** 0.006*** 0.005*** Avg. GDP Per Capita (5.95) (5.81) (4.67) (5.11) (4.82) (4.48) DComPI* 12.794 9.712 22.785 2 10.773 7.279 20.443 Ethnic Fractionalization (0.86) (0.59) (1.25) (0.63) (0.43) (1.02) DComPI* 2 1.755*** 2 1.479** 2 2.264*** 2 1.497*** 2 1.280** 2 1.885*** Avg. Gini Coefficient ( 2 3.40) ( 2 2.22) ( 2 4.02) ( 2 2.93) ( 2 2.08) ( 2 4.41) DComPI* 56.488*** 35.402** 42.894*** 50.247*** 32.609** 38.955*** Avg. Export Diversification (3.28) (2.24) (2.65) (3.60) (2.24) (2.87) DComPI* 0.631 2.646 1.482 1.079 2.362 0.286 Nat. Res. Importer Indicator (0.02) (0.08) (0.04) (0.05) (0.11) (0.01) L.DSpread 0.194*** 0.193*** 0.195*** 0.246*** 0.245*** 0.246*** (3.49) (3.56) (3.58) (3.73) (3.79) (3.75) Hansen J, p-value . . . 0.376 0.367 0.377 AR(2) test, p-value . . . 0.192 0.190 0.197 Country Fixed Effects Yes Yes Yes Yes Yes Yes Year Fixed Effects Yes Yes Yes Yes Yes Yes Observations 247 247 247 247 247 247 Arezki and Bru Note: The method of estimation in columns (1)-(3) is least squares; columns (4)-(6) system-GMM (Blundell and Bond, 1998) with two-step Windmeijer (2005) small sample correction. t-values (in brackets) are based on Huber robust standard errors that are clustered at the country level. The ¨ ckner dependent variable is the log-change in the spread on sovereign bonds. DComPI stands for the log-change in the international commodity export price index (see page 3 in the paper for a detailed explanation of how this variable is constructed). The cross-section (average time-series) dimension of the panel is 35 (7.1). *Significantly different from zero at the 10 percent significance level, ** 5 percent significance level, *** 1 percent significance level. 87 88 T A B L E 4 . Commodity Windfalls, Political Institutions, and the Spread on Sovereign Bonds (Robustness to Restricting the Sample to Natural Resource Exporting Countries) DSpread (1) (2) (3) (4) (5) (6) LS LS LS GMM GMM GMM DComPI 10.729* 2 10.365 7.876 5.774 2 13.272** 3.766*** (1.64) ( 2 1.42) (1.23) (1.04) ( 2 2.32) (3.69) DComPI* 2 2.274*** 2 2.039** Avg. Polity2 Score ( 2 3.31) ( 2 2.15) DComPI* 45.230*** 41.569*** Autocracy Indicator (3.02) (3.00) DComPI* 2 9.329 2 10.762** THE WORLD BANK ECONOMIC REVIEW Avg. Checks & Balance Score ( 2 2.07) ( 2 2.23) DComPI* 0.003*** 0.005*** 0.002*** 0.002** 0.004*** 0.002** Avg. GDP Per Capita (3.11) (3.78) (2.83) (2.38) (3.26) (2.11) L.DSpread 0.205*** 0.217*** 0.202*** 0.198*** 0.206*** 0.189*** (3.00) (3.38) (2.90) (3.24) (3.58) (3.00) Hansen J, p-value . . . 0.281 0.359 0.301 AR(2) test, p-value . . . 1.000 1.000 0.999 Country Fixed Effects Yes Yes Yes Yes Yes Yes Year Fixed Effects Yes Yes Yes Yes Yes Yes Observations 125 125 125 125 125 125 Note: The method of estimation in columns (1)-(3) is least squares; columns (4)-(6) system-GMM (Blundell and Bond, 1998) with two-step Windmeijer (2005) small sample correction. t-values (in brackets) are based on Huber robust standard errors that are clustered at the country level. The dependent variable is the log-change in the spread on sovereign bonds. DComPI stands for the log-change in the international commodity export price index (see page 3 in the paper for a detailed explanation of how this variable is constructed). The cross-section (average time-series) dimension of the panel is 17 (7.4). *Significantly different from zero at the 10 percent significance level, ** 5 percent significance level, *** 1 percent significance level. T A B L E 5 . Commodity Windfalls, Political Institutions, and the Spread on Sovereign Bonds (Robustness to Instrumental Variables Estimation) (1) (2) (3) (4) (5) (6) 2SLS 2SLS 2SLS 2SLS 2SLS 2SLS Panel A: Second Stage (Dependent Variable is DSpread) DComPI 2 8.497 2 23.636*** 2 16.827* 2 10.780* 2 26.276*** 2 20.367** ( 2 1.01) ( 2 2.92) ( 2 1.66) ( 2 1.64) ( 2 3.54) ( 2 2.43) DComPI 2 3.019*** 2 3.141*** * Avg. Polity2 Score ( 2 5.54) ( 2 6.60) DComPI* 46.731*** 48.514*** Autocracy Indicator (4.67) (5.38) DComPI* 2 14.079*** 2 15.197*** Avg. Checks & Balance Score ( 2 3.50) ( 2 4.19) DComPI* 0.006*** 0.006*** 0.005*** 0.007*** 0.008*** 0.007*** Avg. GDP Per Capita (5.80) (6.55) (4.42) (8.71) (9.22) (7.36) DComPI* 2.511 4.035 24.284 25.202* 26.442* 49.703** Ethnic Fractionalization (0.21) (0.32) (1.44) (1.66) (1.68) (2.52) DComPI* 2 0.877 2 0.162 2 1.321 2 1.626** 2 0.877 2 2.049*** Avg. Gini Coefficient ( 2 1.19) ( 2 0.18) ( 2 1.87) ( 2 2.10) ( 2 0.98) ( 2 2.73) DComPI* 47.443** 18.220 30.884 43.143** 12.781 25.692 Avg. Export Diversification (2.36) (0.80) (1.50) (2.24) (0.58) (1.31) L.DSpread 0.233*** 0.232*** 0.233*** (3.41) (3.41) (3.40) Hansen J, p-value 0.336 0.467 0.319 0.221 0.399 0.218 Hausman test, p-value 0.776 0.028 0.967 0.724 0.083 0.645 Panel B: First Stage (Dependent Variable is DComPI*Polity Variable) DComPI* 2 4.184*** 0.407** 2 0.870*** 2 4.139*** 0.411** 2 0.859*** Log Settler Mortality ( 2 4.94) (2.11) ( 2 8.58) ( 2 4.76) (2.11) ( 2 8.15) Arezki and Bru DComPI* 2 4.081*** 0.078 2 0.570*** 2 4.169*** 0.070 2 0.590*** British Colony ( 2 3.56) (0.30) ( 2 4.30) ( 2 3.52) (0.26) ( 2 4.29) ¨ ckner (Continued ) 89 90 TABLE 5. Continued (Robustness to Instrumental Variables Estimation) DComPI* 2 7.171*** 0.335* 2 1.812*** 2 7.275*** 0.362* 2 1.814*** French Colony ( 2 8.34) (1.77) ( 2 18.25) ( 2 8.63) (1.88) ( 2 18.82) Country Fixed Effects Yes Yes Yes Yes Yes Yes THE WORLD BANK ECONOMIC REVIEW Year Fixed Effects Yes Yes Yes Yes Yes Yes Observations 148 148 148 128 128 128 Note: The method of estimation is two-stage least squares. t-values (in brackets) are based on Huber robust standard errors that are clustered at the country level. Panel A shows the second-stage estimates and Panel B shows the first-stage estimates. The dependent variable in Panel A is the log-change in the spread on sovereign bonds. The dependent variable in Panel B, columns (1) and (4) is the interaction between DComPI and countries’ average Polity2 score; in columns (2) and (5) of Panel B the dependent variable is the interaction between DComPI and countries’ autocracy indicator; in columns (3) and (6) of Panel B the dependent variable is the interaction between DComPI and countries’ average checks and balance score. DComPI stands for the log-change in the international commodity export price index (see page 3 in the paper for a detailed explanation of how this variable is constructed). The cross-section (average time-series) dimension in columns (1)-(3) of the panel is 19 (7.8); columns (4)-(6) 18 (7.1). *Significantly different from zero at the 10 percent significance level, ** 5 percent significance level, *** 1 percent significance level. ¨ ckner Arezki and Bru 91 the first-stage F-statistic easily exceeds the Stock and Yogo (2005) critical values for instruments to be declared weak and the Hansen test does not reject that the instruments are uncorrelated with the second-stage error term. As an intermediate step to explain the heterogeneity in the marginal effect that international commodity price booms have on sovereign bond spreads, we report in Table 6 the effect that international commodity price booms have on countries’ real per capita GDP growth. We find that higher international prices for exported commodity goods are associated with a significant increase in real per capita GDP growth in democracies. But in countries with deeply autocratic regimes, windfalls from international commodity prices are associated with a significant decrease in real per capita GDP growth. Taking for example the esti- mates in column (5) of Table 6, a one standard deviation increase in the export price index growth rate was associated with a significant increase in real per capita GDP growth in the democracy sample by about 0.29 standard deviations while in the autocracy sample it was associated with a significant reduction in GDP per capita growth by about 0.16 standard deviations. Similarly, columns (4) and (6) show that the marginal effect of commodity price booms on GDP per capita growth is significantly increasing in countries’ Polity2 and checks and balances scores. So much so, that at sample maximum Polity2 and checks and balances scores a commodity windfall was associated with a significant increase in GDP per capita growth while at sample minimum Polity2 and checks and balances scores a commodity windfall was associated with a signifi- cant decrease in GDP per capita growth. The estimates in Table 6 therefore show that while in countries with strong political institutions a plausibly exogenous windfall from international commodity price booms was associated with a significant increase in GDP per capita growth, in countries with weak political institutions it was associated with a significant decrease. The political economy model developed in Mehlum et al. (2006) can provide an explanation for this heterogeneous response in real per capita GDP growth: in countries with grabber friendly political institutions, revenue wind- falls from international commodity price booms increase rent-seeking activity and lead to a crowding out of production activity. Democratic institutions, in particular, stronger checks and balances constrain politicians in their policy space. Relative to an autocratic regime, politicians are also held more accounta- ble to the public. Hence, in a more democratic regime the expected returns to rent-seeking activities are lower. This in turn means that production activity will remain strong in the democratic regime despite the high rents that are rea- lized in the commodity exporting sector when international commodity prices are booming. In the autocratic regime, on the other hand, where there are rela- tively high gains from specializing in grabbing activities, production activity will be crowded out in the presence of a revenue windfall. Thus, revenue wind- falls from international commodity prices may be associated with lower per capita GDP growth in more autocratic regimes. 92 T A B L E 6 . Commodity Windfalls, Political Institutions, and Economic Growth DGDP (1) (2) (3) (4) (5) (6) LS LS LS GMM GMM GMM DComPI 0.164 2.890*** 0.610 0.470 2.100*** 0.732 (0.24) (3.61) (0.83) (0.96) (5.66) (1.40) DComPI* 0.375*** 0.219*** Avg. Polity2 Score (2.65) (2.85) DComPI* 2 5.623*** 2 3.328*** Autocracy Indicator ( 2 3.94) ( 2 4.40) DComPI* 1.417** 0.948*** THE WORLD BANK ECONOMIC REVIEW Avg. Checks & Balance Score (2.25) (2.71) DComPI* 2 0.001* 2 0.001*** 2 0.001 2 0.001** 2 0.001*** 2 0.001* Avg. GDP Per Capita ( 2 1.69) ( 2 3.28) ( 2 1.28) ( 2 2.24) ( 2 3.36) ( 2 1.76) L.DGDP 0.020 0.014 0.017 0.172 0.172 0.170 (0.30) (0.21) (0.24) (1.60) (1.60) (1.16) Hansen J, p-value . . . 0.815 0.833 0.822 AR(2) test, p-value . . . 0.887 0.877 0.968 Country Fixed Effects Yes Yes Yes Yes Yes Yes Year Fixed Effects Yes Yes Yes Yes Yes Yes Observations 253 253 253 253 253 253 Note: The method of estimation in columns (1)-(3) is least squares; columns (4)-(6) system-GMM (Blundell and Bond, 1998) with two-step Windmeijer (2005) small sample correction. t-values (in brackets) are based on Huber robust standard errors that are clustered at the country level. The dependent variable is the log-change in real GDP per capita. DComPI stands for the log-change in the international commodity export price index (see page 3 in the paper for a detailed explanation of how this variable is constructed). The cross-section (average time-series) dimension of the panel is 37 (6.8). *Significantly different from zero at the 10 percent significance level, ** 5 percent significance level, *** 1 percent significance level. ¨ ckner Arezki and Bru 93 Table 7 provides further evidence on this political economy channel by doc- umenting that political institutions played a key role in shaping the relationship between commodity windfalls and corruption. The significant positive auto- cracy interaction term in the corruption equation implies that in autocracies commodity windfalls are associated with a significant increase in corruption. On the other hand, in democracies and countries with strong checks and bal- ances commodity windfalls did not lead to a significant increase in corruption. This result is consistent with the political economy literature that has high- lighted the importance of political institutions in shaping political leaders’ incentive constraints and thus economic outcomes (e.g. North, 1990; Acemoglu et al., 2001). The growth results in Table 6 are in line with the political economy model developed in Mehlum et al. (2006). However, an open and conceptually inter- esting question is whether beyond their effect on GDP per capita growth com- modity price booms exhibit significant effects on sovereign bond spreads. The business-cycle literature on the link between GDP per capita growth and sover- eign bond spreads has argued for a countercyclical average relationship between economic growth and sovereign bond spreads (see e.g. Neumeyer and Perri, 2005; Aguiar and Gopinath, 2006; or Arellano, 2008). Given this litera- ture which does not emphasize the role of political institutions but instead argues for a countercyclical relationship between economic growth and sover- eign bond spreads in an environment where financial markets are incomplete, it is interesting to explore whether beyond their effects on economic growth the interaction between commodity price booms and political institutions still matters for sovereign bond spreads. To explore the above issue Table 8 reports estimates of the effects that com- modity price booms have on sovereign bond spreads when GDP per capita growth is included as a right-hand- side regressor in the sovereign bond spreads estimating equation. Because we condition in this regression on GDP per capita growth the estimates should be interpreted as capturing the effects that com- modity price booms (and the interaction between commodity price booms and political institutions) have on sovereign bond spreads beyond the effects that these variables have on GDP per capita growth. We report in Table 8 both least squares and system-GMM estimation. To address possible reverse effects of changes in the sovereign bond spreads on GDP per capita growth we instru- ment GDP per capita growth with the lagged first differences. The main result in Table 8 is that, conditional on GDP per capita growth, the interaction effect between commodity price booms and political institutions are quantitatively smaller, but still statistically significant for the majority of the specifications. Hence, while the effect on aggregate output is clearly of first-order importance, we find that commodity price booms and political institutions exhibit additional effects that go beyond aggregate output. This result highlights the importance of political institutions in shaping the relationship between resource windfalls and the spreads on sovereign bonds; it is also consistent with our 94 T A B L E 7 . Commodity Windfalls, Political Institutions, and Corruption Corruption (1) (2) (3) (4) (5) (6) LS LS LS GMM GMM GMM DComPI 2 15.920 2 32.482 2 17.700 2 7.077 2 20.706 2 5.011 ( 2 0.93) ( 2 1.57) ( 2 1.02) ( 2 0.38) ( 2 0.87) ( 2 0.26) DComPI* 2 3.593** 2 2.671* Avg. Polity2 Score ( 2 2.35) ( 2 1.88) DComPI* 59.695*** 45.972** Autocracy Indicator (2.94) (2.07) DComPI* 2 16.864** 2 9.179 Avg. Checks & Balance Score ( 2 2.19) ( 2 1.07) THE WORLD BANK ECONOMIC REVIEW DComPI* 0.005** 0.008*** 0.005*** 0.003 0.005* 0.002 Avg. GDP Per Capita (2.32) (2.97) (2.64) (1.55) (1.92) (1.15) L.Corruption 0.439*** 0.437*** 0.441*** 0.515*** 0.512*** 0.518*** (6.04) (6.14) (6.06) (4.35) (4.30) (4.39) Hansen J, p-value . . . 0.833 0.789 0.837 AR(2) test, p-value . . . 0.366 0.440 0.331 Country Fixed Effects Yes Yes Yes Yes Yes Yes Year Fixed Effects Yes Yes Yes Yes Yes Yes Observations 242 242 242 242 242 242 Note: The method of estimation in columns (1)-(3) is least squares; columns (4)-(6) system-GMM (Blundell and Bond, 1998) with two-step Windmeijer (2005) small sample correction. t-values (in brackets) are based on Huber robust standard errors that are clustered at the country level. The dependent variable is the corruption score from Political Risk Service. The corruption score is rescaled so that higher values indicate more political cor- ruption. DComPI stands for the log-change in the international commodity export price index (see page 3 in the paper for a detailed explanation of how this variable is constructed). The cross-section (average time-series) dimension of the panel is 35 (6.9). *Significantly different from zero at the 10 percent significance level, ** 5 percent significance level, *** 1 percent significance level. T A B L E 8 . Commodity Windfalls, Political Institutions, and Sovereign Spread (Effect Beyond GDP Per Capita Growth) DSpread (1) (2) (3) (4) (5) (6) LS LS LS GMM GMM GMM DComPI 5.789* 2 5.272 5.390 3.450 2 8.975 2.184 (1.74) ( 2 0.70) (1.61) (1.10) ( 2 1.51) (0.64) DComPI* 2 0.777 2 1.327* Avg. Polity2 Score ( 2 0.96) ( 2 1.67) DComPI* 27.461* 28.059** Autocracy Indicator (1.80) (2.22) DComPI* 2 4.673 2 7.751* Avg. Checks & Balance Score ( 2 1.02) ( 2 1.72) DComPI* 0.002*** 0.004*** 0.002*** 0.001*** 0.003*** 0.002** Avg. GDP Per Capita (3.66) (2.95) (3.26) (2.99) (2.94) (2.48) DGDP 2 4.073*** 2 3.892*** 2 4.089*** 2 3.471*** 2 3.430*** 2 3.509*** ( 2 3.91) ( 2 3.80) ( 2 3.98) ( 2 3.68) ( 2 3.71) ( 2 3.69) L.DSpread 0.148*** 0.148*** 0.146*** 0.209*** 0.205*** 0.201*** (2.82) (2.89) (2.76) (4.56) (4.59) (4.34) Hansen J, p-value (DGDP) . . . 0.199 0.184 0.194 Hansen J, p-value (L.DSpread) . . . 0.197 0.199 0.227 Country Fixed Effects Yes Yes Yes Yes Yes Yes Year Fixed Effects Yes Yes Yes Yes Yes Yes Observations 253 253 253 253 253 253 Note: The method of estimation in columns (1)-(3) is least squares; columns (4)-(6) system-GMM (Blundell and Bond, 1998) with two-step Windmeijer (2005) small sample correction. t-values (in brackets) are based on Huber robust standard errors that are clustered at the country level. The Arezki and Bru dependent variable is the log-change in the spread on sovereign bonds. DComPI stands for the log-change in the international commodity export price index (see page 3 in the paper for a detailed explanation of how this variable is constructed). The cross-section (average time-series) dimension of the ¨ ckner panel is 37 (6.8). *Significantly different from zero at the 10 percent significance level, ** 5 percent significance level, *** 1 percent significance level. 95 96 THE WORLD BANK ECONOMIC REVIEW finding that political institutions significantly affect the relationship between resource windfalls and corruption. V. C O N C L U S I O N We investigated in this paper the effects that international commodity price booms have on sovereign bond spreads using panel data for 38 emerging market economies during the period 1997-2007. Our main finding is that revenue windfalls from international commodity price booms lead to a signifi- cant reduction in sovereign bond spreads in emerging market economies with sound democratic institutions. In countries with more autocratic institutions revenue windfalls lead on the other hand to a significant increase in the sover- eign bond spreads. To explain this heterogeneity in the marginal effect that international com- modity price booms have on sovereign bond spreads, we showed that revenue windfalls from international commodity price booms lead to a significant increase in real per capita GDP growth in countries with sound democratic institutions. In countries with deeply autocratic regimes, revenue windfalls lead to a decrease in real per capita GDP growth. Our empirical results are consist- ent therefore with general equilibrium models that predict a countercyclical relationship between sovereign bond spreads and the business cycle in debtor countries (e.g. Arellano, 2008). However, our empirical results also highlight the importance of political economy factors in shaping the relationship between commodity price booms and sovereign bond spreads. Further research, in particular, theoretical contributions along the lines of Cuadra and Saprinza (2008) may therefore be of interest in advancing our understanding of the relationship between revenue windfalls from international commodity price booms, economic growth, and the spread on sovereign bonds in emerging market economies. We conclude on a cautious note that our empirical analysis is based on a relatively short time period. Ideally, an empirical analysis of the effects of com- modity price booms on sovereign bond spreads should include also the 70s and 80s. Manzano and Rigobon (2007) argued that the commodity boom of the 70s led many of the developing (in particular, Latin American countries) to overborrow. When commodity prices collapsed in the 80s, these countries had large debt to GDP ratios and were unable to service their debt, leading to a debt crisis. There exist, unfortunately, no panel data on sovereign bond spreads for the 70s and 80s. This means that we are unable to cover in our analysis the 70s and 80s. We thus end on a note that interestingly, and in line with our results, many of the developing countries were much less democratic in the 70s and 80s than they are today. ¨ ckner Arezki and Bru 97 REFERENCES Acemoglu, D., S. Johnson, and J. Robinson (2001). “The Colonial Origins of Comparative Development: An Empirical Investigation.” American Economic Review 91: 1369– 1401. Aguiar, M., and G. Gopinath (2006). “Defautable Debt, Interest Rates, and the Current Account.” Journal of International Economics 69: 64– 83. Alesina, A., A. Devleeschauwer, W. Easterly, S. Kurlat, and R. Wacziarg (2003). “Fractionalization.” Journal of Economic Growth 8: 155–194. Arellano, C. (2008). “Default Risk and Income Fluctuations in Emerging Markets.” American Economic Review 98: 690– 712. Barro, R. (1999). “Determinants of Democracy.” Journal of Political Economy 107: 158– 183. Beck, T., G. Clarke, A. Groff, P. Keefer, and P. Walsh (2001). “New Tools in Comparative Political Economy: The Database of Political Institutions.” World Bank Economic Review 15: 165 –176. Berg, A., and J. Sachs (1988). “The Debt Crisis: Structural Explanations of Country Performance.” Journal of Development Economics 29: 271– 306. Blundell, R., and S. Bond (1998). “Initial Conditions and Moment Restrictions in Dynamic Panel Data Models.” Journal of Econometrics 87: 115– 43. Collier, P., and B. Goderis (2007). “Commodity Prices, Growth, and the Natural Resource Curse: Reconciling a Conundrum.” CSAE Working Paper 2007–15. Cuadra, G., and H. Sapriza (2008). “Sovereign Default, Interest Rates, and Political Uncertainty in Emerging Markets.” Journal of International Economics 76: 78 –88. Heston, A., R. Summers, and B. Aten (2009). “Penn World Table Version 6.3”, Center for International Comparisons of Production, Income and Prices at the University of Pennsylvania, August 2009. Krueger, A. (1987). “Origins of the Developing Countries’ Debt Crisis, 1970 to 1982.” Journal of Development Economics 27: 141–163. Lederman, D., and W. Maloney (2008). “In Search of the Missing Research Curse.” World Bank Policy Research Paper 4766. Lederman, D., and L. Xu (2010). “Commodities and Structural Volatility.” World Bank, mimeo. Manzano, O., and R. Rigobon (2007). “Resource Curse or Debt Overhang?” in Natural Resources and Development: Neither Curse nor Destiny. Editors: D. Lederman, and W. Maloney, Stanford Press. Marshall, M., and K. Jaggers (2009). Polity IV Project: Dataset Users’ Manual. Center for Global Policy, George Mason University (www.cidcm.umd.edu/polity). [Polity IV Data Computer File, Version p4v2004. College Park, MD: Center for International Development and Conflict Management, University of Maryland.] Mehlum, H., K. Moene, and R. Torvik (2006). “Institutions and the Resource Curse.” Economic Journal 116: 1 – 20. Nickell, S. (1981). “Biases in Dynamic Models with Fixed Effects.” Econometrica 49: 1417– 1426. Neumeyer, P., and F. Perri (2005). “Business Cycles in Emerging Economies: The Role of Interest Rates.” Journal of Monetary Economics (2): 345–380. North, D. (1990). Institutions, Institutional Change, and Economic Performance. New York: Cambridge University Press. Persson, T., and G. Tabellini (2003). The Economic Effects of Constitutions. MIT Press, Cambridge. ——— (2006). “Democracy and Development. The Devil in Detail.” American Economic Review Papers and Proceedings 96 (2): 319– 324. Political Risk Service (2010). International Country Risk Guide. Online Database. Przeworski, A., M. Alvarez, J. Cheibub, and F. Limongi (2000). Democracy and Development: Political Institutions and the Well-Being of the World, 1950-1990. Cambridge, UK: Cambridge University Press. 98 THE WORLD BANK ECONOMIC REVIEW Robinson, J., R. Torvik, and T. Verdier (2006). “Political Foundations of the Resource Curse.” Journal of Development Economics 2: 447–468. Roodman, D. (2009). “A Note on the Theme of Too Many Instruments.” Oxford Bulletin of Economics and Statistics 71: 135– 158. Sachs, J., and A. Warner (1995). “Natural Resource Abundance and Economic Growth.” NBER Working Paper 5398. ——— (2001). “The Curse of Natural Resources.” European Economic Review 45: 827–838. Stock, J., and M. Yogo (2005). “Testing for Weak Instruments in Linear IV Regression.” In Identification and Inference for Econometric Models: Essays in Honor of Thomas Rothenberg, ed. D. Andrews, and J. Stock, 80–108. Cambridge: Cambridge University Press. Van der Ploeg, F. (2010). “Natural resources: curse or blessing?” Journal of Economic Literature, forthcoming. WDI (2010). World Development Indicators. Online Database. Windmeijer, F. (2005). “A Finite Sample Correction for the Variance of Linear Efficient Two-Step GMM.” Journal of Econometrics 126: 25– 51. A p p e n d i x Ta b l e 1 . Robustness to Outliers DSpread (1) (2) (3) (4) (5) (6) Median Median Median Excluding Excluding Excluding Regression Regression Regression Max/Min Max/Min 1% Max/Min 1% 1% DComPI 2 0.478 2 16.023*** 2 0.310 2 5.858 2 30.213*** 2 7.860 ( 2 0.12) ( 2 2.91) ( 2 0.07) ( 2 0.77) ( 2 4.28) ( 2 1.19) DComPI* 2 1.879** 2 3.371** Avg. Polity2 ( 2 2.39) ( 2 2.42) Score DComPI* 29.437** 55.051*** Autocracy (2.42) (4.26) Indicator DComPI* 2 7.734 2 16.651** Avg. Checks ( 2 1.43) ( 2 2.54) & Balance Score DComPI* 0.002** 0.003** 0.002* 0.001 0.004*** 0.001 Avg. GDP Per (2.08) (2.40) (1.75) (0.12) (3.13) (1.18) Capita Country Fixed Yes Yes Yes Yes Yes Yes Effects Year Fixed Yes Yes Yes Yes Yes Yes Effects Observations 291 291 291 284 284 284 Note: The method of estimation in columns (1)-(3) is maximum likelihood; columns (4)-(6) least-squares. The least-squares regressions in columns (4)-(6) exclude observations where the change in the commodity export price index is in the top/bottom 1 percentile. The dependent variable is the log-change in the spread on sovereign bonds. DComPI stands for the log-change in the international commodity export price index (see page 3 in the paper for a detailed explanation of how this variable is constructed). The cross-section (average time-series) dimension of the panel is 38 (7.7). *Significantly different from zero at the 10 percent significance level, ** 5 percent significance level, *** 1 percent significance level. ¨ ckner Arezki and Bru 99 Data Appendix Table 1. Descriptive Statistics Mean Std. Dev. Min Max Obs. DLog Sovereign Bond Spread (DSpread) 2 0.11 0.39 2 2.02 1.32 291 DLog Export Price Index (DComPI) 0.002 0.006 2 0.02 0.04 291 Polity2 Score 4.98 5.35 27 10 291 Checks and Balance Score 3.20 1.43 1 6 291 GDP Per Capita 9189 5085 1236 21331 291 Ethnic Fractionalization 0.42 0.23 0.002 0.85 289 Export Concentration 0.11 0.19 0.006 0.98 282 Gini 43.16 9.16 27 60.4 291 Corruption 2.41 0.96 1 5 278 Settler Mortality 206.5 486.6 17.7 2004 148 Data Appendix Table 2. List of Countries Country Observations Spread Polity2 GDP GINI Ethnic Frac Algeria 4 748.88 23 5432 0.35 0.34 Argentina 10 2135.98 7.9 12956 0.5 0.26 Brazil 10 684.82 8 8666 0.58 0.54 Bulgaria 10 420.05 8.7 7303 0.3 0.4 Chile 8 132.52 9.2 15765 0.55 0.19 China 10 102.81 27 5209 0.42 0.15 Colombia 10 446.4 7 6919 0.58 0.6 Croatia 9 2288.7 0.7 11209 0.29 0.82 Cuba 7 305.35 27 7706 0.27 0.37 Dominican Republic 6 539.44 8 8194 0.51 0.43 Ecuador 10 1271.83 6.6 5351 0.56 0.66 Egypt 6 195.49 2 4.5 5102 0.32 0.18 El Salvador 5 259.21 7 5325 0.51 0.2 Greece 2 89.99 10 19117 0.34 0.16 Hungary 8 69.66 10 14881 0.27 0.15 Indonesia 3 249.39 8 4944 0.39 0.74 Korea, Republic of 7 255.87 8 18806 0.32 0 Lebanon 9 400.77 7 7679 0.6 0.13 Malaysia 10 197.84 3 14952 0.43 0.59 Mexico 10 315.8 7.6 10226 0.49 0.54 Morocco 9 379.89 26 4855 0.4 0.48 Nigeria 10 908.19 3.5 1664 0.45 0.85 Pakistan 6 492.48 2 3.8 3112 0.31 0.71 Panama 10 346.38 9 7464 0.55 0.55 Peru 10 434.95 7 5339 0.51 0.66 Philippines 10 414.73 8 3918 0.45 0.24 Poland 10 155.88 9.6 11568 0.33 0.12 Russia 10 972.75 5.2 9718 0.39 0.25 South Africa 10 234.17 9 9223 0.35 0.75 Thailand 9 170.31 7.4 7713 0.43 0.63 Tunisia 5 148.6 24 9034 0.41 0.04 Turkey 10 488.33 7 6569 0.42 0.32 Ukraine 7 677.71 6.2 7696 0.3 0.47 Uruguay 6 508.43 10 10962 0.45 0.25 Venezuela 10 715.42 6.2 10689 0.48 0.5 Vietnam 2 158.72 27 3492 0.38 0.24