Corruption and efficiency

Much of the theoretical literature on corruption and economic growth suggests that corruption harms growth, yet the empirical literature is ambiguous. One highly cited paper, Mauro (1995), estimates that corruption has a moderately harmful effect on growth, yet once more recently available data on real GDP is used, the correlation disappears. This study suggests that the reason for the ambiguous evidence is multicollinearity between the regressors---initial levels of corruption and initial levels of real GDP per capita; poor countries tend to have more opportunities for growth through convergence forces, and also have more corruption which diverts investment away from rent creation and to rent appropriation. The net effect is that on average the poor countries' economies tend to grow at about the same rate as those of the rich countries. The econometric methods typically used in research inadequately overcome this problem. A novel method is introduced in this study; levels of dynamic efficiency are estimated, and then regressed on levels of corruption. Three methods are used to estimate the best-practice growth frontier and subsequent levels of efficiency: (1) Data Envelopment Analysis; (2) Stochastic Frontier Analysis; and (3) Imposed Functional Forms based on the Solow growth model. All three methods suggest the potential for convergence---poor countries have greater opportunities for growth than do rich ones. Various models are used in the regressions of efficiency, including up to three additional explanatory variables: (1) average population growth rates; (2) enrollment rates in secondary education; and (3) net royalty payments as a fraction of GDP. In all of the models, corruption is found to be a strong predictor of inefficiency, and therefore harmful to growth. In one estimate, if Nigeria were to reduce their level of corruption by 2.5 units on a scale of 1 to 10, then it would increase its average rate of growth by approximately five percentage points---from negative 3.55 percent to positive 1.50 percent.