Identifying the robust determinants of corruption among cultural, economic, institutional, and geographical factors has proven difficult. From a policy perspective, it is important to know whether inherent, largely unchangeable attributes are responsible or if institutional and economic attributes are at work. Accounting for model uncertainty, we use Bayesian Model Averaging (BMA) to analyze a comprehensive list of 36 potential corruption determinants across 123 countries (covering 87 percent of the world population). The BMA methodology sorts through all 68,719,476,736 possible model combinations (236) in order to carve out the robust correlates. We then take a step toward alleviating endogeneity concerns in an Instrumental Variable BMA framework. Our results indicate that cultural factors are largely irrelevant, whereas particular economic and institutional characteristics matter. The rule of law emerges as the most persistent predictor with a posterior inclusion probability (PIP) in the true model of 1.00, whereas we find strong evidence for government effectiveness (PIP of 0.88), urbanization (0.85), and the share of women in parliament (0.80) as meaningful determinants of lower corruption levels. In developing countries, the extent of primary schooling enters as a powerful factor with a PIP of 1.00.
- Bayesian model averaging
- Instrumental variable Bayesian model averaging
- Political institutions
ASJC Scopus subject areas
- Geography, Planning and Development
- Sociology and Political Science
- Economics and Econometrics