Parameter heterogeneity in growth regressions

This paper explores the appropriateness of pooling many diverse countries together in growth regressions. It is shown that estimates are sensitive to small changes in the sample, and separating OECD and non-OECD countries can result in drastic differences.