This paper presents novel analysis of the relationship between temperature and income. First, using sub-national data from 12 countries in the Americas, we provide new evidence that the negative cross-country relationship between temperature and income also exists within countries and even within states. Second, we provide a theoretical framework for reconciling the substantial, negative association between temperature and income in the cross-section with the even stronger short-run effects of temperature estimated by panel models. The theoretical framework suggests that half of the negative short-term effects of temperature may be offset in the long run through adaptation.