This paper uses Bayesian model averaging to uncover the true determinants of export diversification among 36 potential factors, and thus 2 36 potential models. Using data from 2001 to 2010, our results reveal two strong predictors: Primary school enrollment (99.7% posterior inclusion probability in the true model) raises export diversification, whereas the share of natural resources in gross domestic product (98.6%) lowers diversification levels. The importance of basic education coverage offers policymakers an opening toward diversifying exports, at least in the long run. This result is robust to accounting for the endogeneity of income levels by applying an instrumental variable BMA method . ( JEL C11, F1, F34, O1, O2, O11)