Bayesian Evidence Synthesis to Infer Unobserved Population Dynamics: An Application to International Migration Into the United States, 2000–2019

For the United States, detailed estimates of the number of resident migrants and the rates of migrant arrival are valuable for understanding population dynamics and for determining the impact of economic and political changes that influence migration. The goal of this analysis was to derive estimates of the U.S. foreign-born population and how this population has changed in recent years, as well as estimates of recent and historical immigration volumes. Using data from large population surveys (the 2000 U.S. decennial census and 2001–2019 American Community Survey (ACS)), a Bayesian evidence synthesis was conducted to pool survey data across years while accounting for various biases and logical constraints that apply to these data. This analysis produced highly disaggregated estimates of the foreign-born population residing in the United States over the period 2000–2019, as well as estimates of immigration volume for 1950–2019. These population estimates demonstrated high in- and out-of-sample predictive performance, with substantially greater precision than that for raw survey estimates. Estimated immigration flows tracked other available time series, although with higher precision and with the potential to include undocumented immigration not represented in other immigration data. This study documents immigration from 100 countries of origin into the United States and demonstrates how the results of repeated cross-sectional population surveys can be used to infer migration dynamics that are difficult to measure directly.