We construct a gravity model of worldwide foreign direct investment stock (FDI) in order to study the effect of religion on FDI allocation. We establish empirically that both bilateral religious similarity and bilateral religious diversity foster FDI at the country pair level. These apparently contradicting results confirm an empirical puzzle that has already emerged in the literature, particularly in the case of trade in goods. We investigate whether the answer to this puzzle could lie on the fact that the effect of these two variables play for different types of countries, depending on the level of efficiency of their institutions.