Immigration, local crowd-out and undercoverage bias

Using decadal census data since 1960, I cannot reject the hypothesis that new immigrants crowd out existing residents from US commuting zones and states one-for-one. My estimate accounts explicitly for dynamic local adjustment, it is statistically precise and robust to numerous specifications, and I show how it can be reconciled with apparently conflicting results in the literature. Exploiting my model's structure, I attribute 30% of the observed effect to mismeasurement, specifically undercoverage of new immigrants in the census. Based on a remarkably simple decomposition (and after adjusting for undercoverage), I show that population mobility accounts for 90% of local labor market adjustment (following an immigration shock), and labor demand the remainder. These results have important methodological implications for the estimation of immigration effects.