We show that the standard aggregate-data methodology employed to examine the relationship between recessions and mortality may be biased when people move in response to recessions. We offer direct evidence of this “migration bias” using a historical setting that provides both large-scale longitudinal microdata and an exogenous shock to local economic conditions. This allows us to estimate the recession-mortality relationship using methodologies that differ only by whether they account for migration. Results indicate that the recession increased overall mortality when migration is taken into account. Applied to the same data, standard methodologies, which ignore the role of migration, fail to recover this effect.