Are recessions good for health? A large literature following Ruhm (2000) addresses this question by applying a fixed-effects approach that implicitly assumes either that recessions do not generate a substantial migratory response, or that such responses are accurately reflected in intercensal population estimates. These assumptions may pose a serious methodological concern in settings, such as developing countries, that are characterized by weak social safety nets, mobile populations, and poor intercensal data. We illustrate this point by drawing on a natural experiment—the recession in Britain’s cotton textile-producing regions caused by the U.S. Civil War (1861-1865)—to provide evidence that migration-induced bias can substantially affect, and even reverse, estimates of the recession-mortality relationship. To deal with this bias, we propose a strategy based on accounting for mortality spillovers in migrant-receiving locations. Applying this methodology to our historical setting, we find evidence that, if anything, the recession we study increased mortality. In contrast, we show that existing approaches, which do not account for migration bias, would lead us to exactly the opposite conclusion. After adjusting for migration, we do find evidence that infant mortality fell, but this was offset by large increases in mortality among the elderly.