Does trade openness cause higher GDP per capita? Since the seminal instrumental variables (IV) estimates of Frankel and Romer [F&R](1999) important doubts have surfaced. Is the correlation spurious and driven by omitted geographical and institutional variables? In this paper, we generalize F&R's geography-based empirical strategy to a panel setting. We observe that natural disasters affect bilateral trade, and that this effect is conditioned by geographical variables such as distance to financial centers or area. This allows us to use interactions between geography and the incidence of disasters at the bilateral level to construct an instrument for multilateral openness that varies across countries and time. The instrument can be used in panel setups where it is possible to fully control for geographical and historical determinants of countries' performances as well as for the direct effect of disasters. We find that the elasticity of income with respect to openness is about 0.69, but that substantial heterogeneity exists across country samples.
per capita income openness natural disasters instrumental variable estimation panel econometrics