How effective is foreign direct investment in supporting economic performance in low-income countries? This paper assesses this question using meta-regression-analysis techniques on a set of 550 and 554 estimates of the impact of FDI on economic performance from 103 micro and 72 macro-studies, respectively. The results suggest that (a) the estimated effects tend to be larger in the macro/country than in the micro/firm studies, (b) the effect is significantly greater in low- than in middle-income countries, and (c) econometric method and specification choice seem central to understand the observed variation in the estimates. The paradox this study raises is how to reconcile the main lesson from the literature (that the effect emerges only for countries that have reached certain thresholds, mainly with respect to human capital and financial development) with the finding that the effects are larger for counties that are typically far from reaching such critical thresholds. We argue that considerations of the gap between private and social returns, albeit missing in most of the current academic and policy discussions, may provide the key.