We criticize existing empirical results on the detrimental effects of natural resource dependence on the rate of economic growth after controlling for institutional quality, openness, and initial income. These results do not survive once we use instrumental variables techniques to correct for the endogenous nature of the explanatory variables. Furthermore, they suffer from omitted variables bias as they overestimate the effect of initial income per capita and thus underestimate the speed of conditional convergence. Instead, we provide new evidence for the impact of natural resource dependence on income per capita in a systematic empirical cross-country framework. In addition to a significant negative direct impact of natural resources on income per capita, we find a significant indirect effect of natural resources on institutions. We allow for interaction effects and provide evidence that the natural resource curse is particularly severe for economic performance in countries with a low degree of trade openness. Adopting policies directed toward more trade openness may thus soften the impact of a resource curse. We also check the robustness of our results by using a variety of instruments and also employing the ratio of natural capital rather than natural resource exports to national income as an explanatory variable. We find evidence that resource abundance, measured by the stock of natural capital, also induces a resource curse, but less severely for countries that are relatively open.