Probabilistic forecasts of international bilateral migration flows

Area of Science:

• Demography
• Computational Social Science
• Statistical Modeling

Background:

• Accurate forecasting of global human migration is crucial for policy and planning.
• Existing models for international migration flows have limitations in accuracy and uncertainty assessment.

Purpose of the Study:

• To develop and validate a novel Bayesian hierarchical model for forecasting bilateral human migration flows.
• To assess the model's performance against existing methods and simple alternatives.
• To integrate the forecasting model into a probabilistic population projection framework.

Main Methods:

• Utilized a Bayesian hierarchical model to project 39,800 bilateral migration flows among the 200 most populous countries.
• Generated out-of-sample forecasts for 2015-2020 using data from 1990-2015.
• Integrated the migration forecasting model with a probabilistic population projection model.

Main Results:

• The proposed model produced well-calibrated out-of-sample forecasts for bilateral, country-level inflows, outflows, and net flows.
• Achieved a 61% decrease in mean absolute error compared to a leading international migration model.
• Outperformed or matched simple deterministic forecasting methods while accurately quantifying uncertainty.

Conclusions:

• The Bayesian hierarchical model offers a robust and accurate approach to forecasting global human migration flows.
• The method provides reliable probabilistic projections, including age and sex-specific flows, for future demographic scenarios.
• This enhanced forecasting capability supports better-informed policy decisions related to migration and population dynamics.