Uncovering the determinants of corruption

Identifying the real causes of corruption has proven difficult because of limited data availability, lack of a unifying theoretical framework, and endogeneity concerns. Combining data for a comprehensive list of 36 potential determinants of corruption across 123 countries (including 87 percent of the world population), we use Instrumental Variable Bayesian Model Averaging to account for model uncertainty and endogeneity. In addition to income levels, the extent of primary schooling emerges as a powerful predictor of lower corruption levels. Rule of law and FDI matter particularly in developing countries. Finally, we find some evidence for trade freedom, political rights, government size, and religious fractionalization to contain corruption. These findings offer feasible avenues for policymakers, as corruption does not seem to be driven by deeply rooted cultural attributes.