Modeling and Forecasting Bilateral Migration Flows for All Countries

Nathan Welch, University of Washington
Adrian Raftery, University of Washington, Seattle

We propose a method for forecasting global human migration flows. A Bayesian hierarchical model is used to make joint probabilistic projections of 39,800 bilateral migration flows among 200 countries. The model is fit to estimates of quinquennial bilateral migration flows from 1990 through 2020. We find that the model produces well-calibrated out-of-sample forecasts of bilateral flows, total country-level inflows, total country-level outflows, and country-level net flows. We generate out-of-sample forecasts and compare them to projections from a leading model fit to the same data. Mean absolute error decreases by 60% using our model compared to a leading model of international migration. Our approach yields a well-calibrated probabilistic model of gross migration flows. We demonstrate a model-based approach to disaggregating gross flow forecasts into age and sex specific flow forecasts. These disaggregated forecasts are then used to generate global probabilistic population projections from 2020 to 2045.

Keywords: Bayesian methods / estimation, International migration, Multi-level modeling, Population projections, forecasts, and estimations

See extended abstract.

  Presented in Session 34. International Migration Projections, Determinants and Crossing Strategies