Geography matters: Reconsidering the effect of geography on development

We decisively reject the hypothesis that geographical factors influence long-run only indirectly, through the quality of institutions. The direct influence of geography on per capita incomes is robust to the inclusion of a sub-Saharan Africa dummy and other tests. We obtain our results by replacing the usual instrument (settlers' mortality) by stronger instruments for institutional quality (latitude, the share of the country in the temperate climatic zone). We also show that settlers' mortality suffers from endogeneity with respect to institutional quality for early colonies, because of its dependence on nineteenth-century mortality data.