Threshold regressions for the resource curse

This paper analyzes the behavior of cross-country growth rates with respect to resource abundance and dependence. We reject the linear model that is commonly used in growth regressions in favor of a multiple-regime alternative. Using a formal sample-splitting method, we find that countries exhibit different behaviors with respect to natural resources depending on their initial level of development. In high-income countries, natural resources play only a minor role in explaining the differences in national growth rates. On the contrary, in low-income countries, abundance seems to be a blessing but dependence restricts growth.