Bayesian Hierarchical Model for Estimating Age and Sex Patterns of International Migration Using Socioeconomic Covariates

Aysha Basheer , Max Planck Institute for Demographic Research
Arkadiusz Wisniowski, University of Manchester
Maciej Jan Danko, Max Planck Institute for Demographic Research
Emilio Zagheni, Max Planck Institute for Demographic Research

Migration is a key driver of demographic change, yet the lack of reliable and consistent data constrains detailed analysis. This study implements a hierarchical Bayesian model to estimate age- and sex-specific bilateral migration flows among 31 European countries (2009–2022). The model incorporates socioeconomic covariates such as life expectancy, unemployment rate, and educational attainment to examine how contextual factors shape migration patterns. A Poisson model with log-normal random effects is used to capture overdispersion and unobserved heterogeneity. The framework integrates demographic effects and a measurement-error layer to reconcile discrepancies between sending- and receiving-country reports, producing coherent estimates with quantified uncertainty across origin, destination, age, sex, and time. Implemented in JAGS using MCMC sampling, it applies partial pooling across countries and demographic groups to improve estimation in sparse or inconsistent data settings. Results show that including socioeconomic covariates enhances model performance and reveals systematic variations in migration intensity linked to broader social and economic conditions. The estimated age–sex migration profiles align with theoretical expectations, while demonstrating that higher unemployment is associated with increased out-migration and higher life expectancy with reduced mobility. These findings confirm that embedding socioeconomic dimensions within a Bayesian framework captures heterogeneities in migration dynamics that remain hidden in purely demographic models. The proposed approach advances migration modelling by offering a probabilistically coherent framework that improves interpretability, accounts for uncertainty, and enables more robust estimation of disaggregated migration flows across demographic and contextual dimensions.

See extended abstract

 Presented in Session 2. Bayesian Demographic Modeling