This paper argues that the empirical trade-growth relationship should be modelled using a dynamic panel data approach and that it is best estimated with Blundell and Bond’s (1999) system-GMM estimator. This procedure remedies some econometric problems such as regressor endogeneity, measurement error and weak instruments, and allows to control for time-invariant country-specific effects such as institutions or geography. The findings are largely plausible and satisfy intuition better than previous results. They confirm the existence of a strong causal effect of trade on growth but fail to find evidence for trade as an independent factor of divergence. Hence, one cannot blame trade as such for the disappointing performance of initially poor countries.
growth empirics trade convergence generalized methods of moments