We investigate the issue of model uncertainty in cross-country growth regressions using Bayesian model averaging (BMA). We find that the posterior probability is distributed among many models, suggesting the superiority of BMA over any single model. Out-of-sample predictive results support that...