Should I stay or should I go? Return migration from the United States

Return migration is important, but how many migrants leave and who is poorly understood. This paper proposes a new method for estimating return migration rates using aggregated repeated cross-sectional data, treating the number of migrants in a group who arrived in a particular year as an unobserved fixed effect, and the observed number (including, importantly, observed zeroes) in the arrival or subsequent years as observations from a Poisson distribution. Compared to existing methods, this allows us to estimate return rates for many more migrant groups, allowing more in-depth analysis of the factors that influence return migration rates. We apply this method to US data and find a decreasing hazard, with most returns occurring by eight years after arrival, when about 13% of migrants have left. The return rate is significantly lower for women, those who arrive at a young age, and those from poorer; it is higher for those on non-immigrant visas for work or study. We also provide suggestive evidence that, conditional on their country of origin, those with lower education are more likely to return.