Arbeitsbereich
Migration und Mobilität
Auf einen Blick
Projekte
Publikationen
Team
Projekt
Bayesian Approaches for Estimating and Understanding Migration Patterns by Age, Sex, and Country of Birth (Dissertation)
Aysha Basheer, Arkadiusz Wiśniowski (The University of Manchester, Vereinigtes Königreich von Großbritannien und Nordirland), Maciej Danko, Emilio Zagheni
Ausführliche Beschreibung
This project estimates age- and sex-specific European migration patterns using a Bayesian hierarchical model with socioeconomic covariates and extends it to include country of birth, producing six-dimensional estimates by origin, destination, age, sex, time, and country of birth.
Description:
Reliable age- and sex-disaggregated migration data are essential for understanding population dynamics, planning policies, and making projections. However, migration statistics are often incomplete, inconsistent, or reported differently by sending and receiving countries. This project addresses these challenges by developing a comprehensive, integrated modeling framework for 31 EU/EFTA countries from 2009 to 2024, disaggregated by age, sex, country of birth, origin, destination, and time. The estimation of bilateral migration flows from limited or indirect data has long been central to demographic research. Early stock-based approaches derived age–sex profiles from census data, while recent Bayesian hierarchical models have harmonized data sources to generate coherent migration estimates. However, few studies have jointly modeled age- and sex-specific migration alongside socioeconomic drivers within a unified framework. Building on this foundation, the present study develops Bayesian hierarchical models capable of reconciling dual reported data and producing disaggregated migration estimates, thus allowing us to capture key heterogeneities that are hidden in purely demographic models.
The study involves two models:
- The first is a Bayesian Poisson log-normal model used to estimate the bilateral age- and sex-specific migration proportions over time, and to assess the influence of various socioeconomic factors, such as unemployment and educational attainment. The model treats the observed counts as noisy reports and uses a hierarchical structure to borrow information across dimensions. Moreover, the model incorporates socioeconomic covariates such as unemployment rates and educational attainment levels, as well as life expectancy as a proxy for quality of life and health care infrastructure. This approach captures how different age and sex groups respond to economic and social conditions in their origin and destination countries, including, for example, how labor market opportunities impact working-age men and women or retirees. By analyzing posterior distributions, we quantify marginal effects and explore counterfactual scenarios. For instance, our preliminary findings indicate that high unemployment in the country of origin acts as a push factor for working-age populations in particular, while higher life expectancy in the country of destination, which is likely a proxy for high quality of life, attracts migrants across multiple age groups, with the effects varying by sex.
- The second model extends the Rogers-Castro migration model into a multidimensional Bayesian framework. While the classic Rogers-Castro model captures age-specific peaks, our extension allows these patterns to vary across origin-destination pairs, sex, and time. The model successfully captures labor force peaks around ages 20-35 and retirement age patterns, with full uncertainty quantification through posterior distributions. This approach provides a more detailed picture of migration dynamics, while also serving as a useful comparison to the estimates from the first model.
Another component of this project is the integration of country of birth (CoB) as an additional dimension in the estimation framework using iterative proportional fitting (IPF). This reconstruction harmonizes immigration flows by destination and CoB with emigration flows by origin and CoB, producing consistent six-dimensional estimates. The CoB dimension is essential to distinguish between native-born and foreign-born migrants, which is important for capturing secondary migration (foreign-born individuals moving from one host country to another) and return migration (individuals moving back to their CoB). Proportional balancing ensures that total flows remain consistent between the origin and destination perspectives.
The final output will include comprehensive migration flow estimates for European countries by origin, destination, age, sex, time, and country of birth, along with uncertainty quantification. This work advances our understanding of structural drivers, such as of how unemployment differentially shapes mobility across demographic groups. Overall, these results will improve the quality of European migration data and advance our understanding of the demographic and structural drivers of international mobility. The estimates will also be suitable for use in further country-specific analyses (e.g., analyses focusing on the detailed migration patterns of single countries), and will thus help to uncover specific migration narratives across different national contexts. In the future, we plan to extend the approach to other regions of the world.