This paper demonstrates the importance of earnings-sensitive migration in response to local variation in labor demand. We use geographic variation in the depth of the housing bust to examine its effects on the migration of natives and Mexican-born individuals in the U.S. We find a strong effect of the housing bust on the location choices of Mexicans, with movement of Mexican-born population away from U.S. states facing the largest declines in construction and movement toward U.S. states facing smaller declines. This effect operated primarily through interstate migration of Mexicans previously residing in the U.S. and, to a lesser extent, through slower immigration rates from Mexico in states with larger housing declines. There is no evidence that return migration to Mexico played an important role in immigrants' migration response. We also find no impact of the housing bust on natives' location choices. We interpret these results as the causal impact of the housing bust on migration after confirming that they are robust to controls for immigrant diffusion and a pre-housing-bust false experiment. Finally, we show that cross-state variation in local labor demand had a stronger impact on overall construction employment than on construction employment among native-born workers, which suggests that mobile Mexican-born workers helped smooth geographic differences in labor market outcomes for natives.