Resource Rents, Institutions, Corruption & Conflicts in Sub-Saharan Africa, Study notes of Business Administration

Download Resource Rents, Institutions, Corruption & Conflicts in Sub-Saharan Africa and more Study notes Business Administration in PDF only on Docsity! 7 can be pursued in the political arena. This indicator ranges from 1 to 10, with greater values denoting more competition. Table 2. Descriptive Statistics Tables 1 and 2 describe all other variables used in our empirical analysis as well as some summary statistics. Appendix table 1 provides a list of countries included in our empirical analysis. 3. Estimation Strategy We now explain our estimation strategy that allows us to estimate the effect of resource rents on changes in corruption (and other outcome variables of interest). Specifically, we estimate the model: ΔCorruptionit = µi+αCorruptionit-1 + γt +βResource Rentsit + λInteraction it + uit where µi are country fixed effects that capture unobservable time-invariant country characteristics, and γt are year fixed effects that capture shocks common to all countries. The parameter estimate β Mean Std. Dev. Min Max Observations ΔCorruption -0.01 0.10 -0.92 0.69 662 ΔInternal Conflicts 0.01 0.17 -0.92 0.92 662 ΔGovernment expenditure 0.01 0.18 -2.14 1.72 1719 Resource Rents 0.98 1.86 -4.61 5.37 5289 Polity2 Score 2.08 0.45 0.69 3.08 1596 Executive Constraints Score 1.02 0.46 0.00 1.95 1596 Political Competition Score 1.18 0.50 0.00 2.27 1596 Ethnic Fractionalization 0.63 0.24 0.00 0.93 1710 Note: All variables are taken in log. Polity2 score has been rescaled adding 12 to the original score. Docsity.com 8 reflects, therefore, the (short-run) marginal effect of resource rents on changes in corruption.5 The parameter estimate λ captures the marginal effect of the interaction between resource rents and country-specific Polity2 score made time invariant by taking the average over the sample period. We also controlled for lagged corruption (Lagged Corruptionit-1), which captures convergence effects in the level of corruption as corruption scores are bounded. We present estimates using least squares estimation but also system-GMM estimation (Blundell and Bond, 1998) to deal with possible biases arising from dynamic panel data estimates in the presence of fixed effects. The error term uit is clustered at the country level and may hence be arbitrarily serially correlated within countries. 4. Main Results Resource Rents and Corruption. Table 3 summarizes our estimation results for the link between within-country variation in resource rents and within-country variation in corruption. Column (1) shows the least squares estimates where control variables are country fixed effects as well as year fixed effects (both jointly significant at the 1% level). Using the column (1) estimate, the impact of a shock to corruption at time t would take about 3.9 years to dissipate by one-half.6 The obtained point estimate on our resource rents measure is -0.03, which is, however, not statistically significant at conventional levels. Because a higher corruption score taken from Political Risk Services (2009) indicates less corruption, the point estimate in column (1) implies that a one standard deviation increase in the resource rents increases corruption by about 0.6 standard deviations.7 In column (2) we show that this adverse link between oil rents and corruption remains statistically insignificant when omitting to control for time effects. However, column (3) which is our preferred Docsity.com 11 assessment of political violence from Political Risk Services (2009), a higher score indicating less internal political violence. In column (1) we estimate the individual effect of resource rents on internal conflicts and find that resource rents decrease in a statistically significant manner the assessment of internal conflicts. In columns (2) to (4) we also include an interaction term between resource rents and the Polity2 score. As can be seen from columns (2) and (3), the estimates of the interaction are statistically significant and quantitatively large. So much so that for a country with the same level of democracy as in Mauritius, an increase by one standard deviation in resource rents barely improves the assessment of internal stability (0.1 standard deviation). However, for a country with a lower level of democracy comparable to the one in Senegal, an increase by one standard deviation in resource rents leads to an almost 1.2 standard deviation increase in the assessment of internal stability. We do find, however, that neither the estimate of the individual effect of resource rents nor the estimate of interaction between resource rents and the Polity2 score are significant when using GMM estimator. Table 4. Resource Rents, Democracy and Internal Conflicts Oil Rents and Government Expenditure. What explains the moderating role of political institutions in the relationship between resource rents and internal conflicts in Sub-Saharan Africa? There could clearly be a number of possible reasons but a useful way in answering this question is to focus on aspects of the distribution of resource rents between the political elite and the masses. (1) (2) (3) (4) OLS OLS OLS GMM Lagged Internal Conflicts -0.360*** -0.367*** -0.288*** -0.146* [0.036] [0.039] [0.034] [0.078] Resource Rents 0.045*** 0.144*** 0.167*** 0.008 [0.013] [0.049] [0.047] [0.019] Interaction with Polity2 -0.045** -0.053** -0.003 [0.022] [0.020] [0.012] Country Effects Yes Yes Yes Yes Time Effects Yes Yes No Yes R-squared 0.301 0.303 0.177 Observations 623 602 602 599 Countries 29 28 28 28 Note: The dependent variable is Δinternal conflict. Robust standard errors are reported in brackets. *Significantly different from zero at 90 percent confidence, ** 95 percent confidence, *** 99 percent confidence. Internal Conflict Docsity.com 12 Extending transfers to the population may be an effective way to quell the masses following a resource boom. It may, however, prove harder for democracies to effectively quell the masses through redistribution to the public. This can be the case because of the scrutiny of government actions resulting from the existence of constraints on the executive power and checks and balances. Those constraints and checks are likely to hamper accrued redistribution disguised as additional government spending. In contrast, in less democratic countries where no such constraints and checks and balance exist, the content and the volume of government spending may increase drastically following a resource bonanza so as to quell the masses. By doing so, the political elite in less democratic countries may significantly reduce the risk of conflict and thus preserve its rent income from resource revenues. Table 5. Resource Rents, Democracy and Government Spending In Table 5 we provide evidence supporting our argument by documenting that higher resource rents lead to more (less) government spending in less (more) democratic countries. In columns (1) we estimate the individual effect of resource rents on government spending and find that resource rents increase in a statistically and economically significant manner government spending. A one standard deviation increase in resource rents leads to an increase by 0.16 standard deviation in government spending. In columns (2) to (4) we also include an interaction term between resource rents and the Polity2 score. As can be seen from columns (2) and (5), the estimates of the (1) (2) (3) (4) OLS OLS OLS GMM Lagged Expenditure -0.196** -0.130* -0.120* -0.105 [0.078] [0.066] [0.066] [0.073] Resource Rents 0.016*** 0.046*** 0.045*** 0.017* [0.006] [0.017] [0.015] [0.009] Interaction with Polity2 -0.016** -0.016** -0.008** [0.006] [0.006] [0.003] Country Fixed Effects Yes Yes Yes Yes Time Effects Yes Yes No Yes R-squared 0.065 0.051 0.017 Observations 1438 1312 1312 1265 Countries 43 39 39 39 Note: The dependent variable is Δgovernment spending. Robust standard errors are reported in brackets. *Significantly different from zero at 90 percent confidence, ** 95 percent confidence, *** 99 percent confidence. Government Spending Docsity.com 13 interaction are statistically significant and quantitatively large.9 So much so that for a country with the same level of democracy as in Mauritius, an increase by one standard deviation in resource rents decreases government spending by 0.08 standard deviations whereas for a country with a level of democracy as in Senegal government spending would increase by 0.1 standard deviation. We attribute this dichotomous effect to the ability of political elites in autocracies to effectively redistribute to the public in periods of resource bonanza rendered possible by the lack of scrutiny they face. Our findings suggest that the mechanisms through which resource rents affect corruption cannot be separated from political systems. Table 6. Robustness Checks Robustness Checks. To ascertain that our main results are not driven by other factors than political systems we further control for the level of ethnic fractionalization. Indeed, Sub-Saharan Africa is the most ethnically fractionalized continent (see Alesina et al., 2003). Furthermore, ethnic fractionalization has been used either as an instrument for corruption (see Mauro, 1995) or as a key determinant of the likelihood of conflicts (see Collier and Hoeffler, 2004). Table 6 documents the results of our previous regressions adding to the specification an interaction between resource rents and the country specific level of ethnic fractionalization. Albeit some estimates have become less (1) (2) (3) (4) (5) (6) OLS GMM OLS GMM OLS GMM Lagged Corruption -0.164*** -0.067 [0.030] [0.046] Lagged Internal Conflicts -0.369*** -0.151** [0.038] [0.076] Lagged Government Expenditure -0.131* -0.099 [0.066] [0.074] Resource Rents -0.007 -0.032** 0.155* 0.015 0.046** 0.018* [0.033] [0.017] [0.056] [0.021] [0.017] [0.011] Interaction with Polity2 -0.002 0.017** -0.053* -0.012 -0.017** -0.009* [0.019] [0.007] [0.031] [0.012] [0.007] [0.006] Interaction with Ethnic Fractionalization -0.048 -0.003 -0.021 -0.049** -0.006 -0.006 [0.033] [0.014] [0.061] [0.022] [0.018] [0.010] R-squared 0.133 0.304 0.054 Observations 602 599 602 599 1284 1276 Countries 28 28 28 28 38 38 Corruption Internal Conflicts Government Expenditure Note: The dependent variables are respectively Δcorruption, Δinternal conflicts and Δgovernment expenditure. Robust standard errors are reported in brackets. *Significantly different from zero at 90 percent confidence, ** 95 percent confidence, *** 99 percent confidence. All regressions included fixed and time effects but estimates are not shown. Docsity.com