Deep learning four decades of human migration

Clinical Snapshot

Key Findings

• P Value: Not applicable — this is a modelling/validation study; performance assessed via held-out predictive accuracy rather than hypothesis testing

• Effect Size: Model outperforms existing five-year flow estimates on held-out validation data; specific performance metrics (e.g., RMSE, MAE, correlation coefficients) not reported in abstract

• Primary Outcome: Annual origin-destination migration flow estimates for 230 countries and regions from 1990 to present, with propagated uncertainty bounds

• Nnt Or Sensitivity: Not applicable in traditional clinical sense; analogous metric is predictive accuracy on held-out data versus existing quinquennial estimates — reported as superior but without specific sensitivity/specificity or RMSE values in the abstract

• Confidence Interval: Confidence bounds generated for all estimates via uncertainty propagation; specific numerical intervals not reported in abstract — available in full paper and released dataset

Clinical Application

High feasibility for research and policy applications: all data, code and trained models are publicly released, enabling direct integration into demographic, epidemiological and health system planning workflows. Requires statistical expertise to appropriately interpret uncertainty bounds and understand model limitations. Not a point-of-care clinical tool. Directly relevant to Australian practice and policy. Australia's Department of Home Affairs, the Australian Bureau of Statistics and the Australian Institute of Health and Welfare all require high-resolution bilateral migration data for workforce planning, Medicare and PBS demand modelling, and refugee health service planning. The dataset's annual resolution is particularly valuable for modelling post-COVID migration recovery, the impact of the Pacific Australia Labour Mobility (PALM) scheme, and net overseas migration projections used in Commonwealth and state budget forecasting. RACGP workforce planning — particularly for general practice in regional and rural areas heavily dependent on overseas-trained doctors — would benefit from finer-grained migration flow data. TGA and PBS demand modelling for migrant-prevalent conditions (e.g., tuberculosis screening, hepatitis B, haemoglobinopathies) could be refined using these estimates. The dataset's coverage of Asia-Pacific bilateral flows is of particular strategic relevance given Australia's regional demographic ties. Researchers, public health planners, health economists and policymakers requiring high-resolution migration flow data for workforce modelling, infectious disease epidemiology, refugee health planning, social determinants of health analysis and demographic projections

Abstract

Human migration is a fundamental driver of global demographic change, shaping population structure, labour markets and social policy across countries1-3. Although long-term migration patterns are often linked to economic development4, they can shift rapidly in response to shocks such as conflict, environmental crises and political change5. Despite its importance, migration remains difficult to measure consistently: existing data are sparse, concentrated in high-income settings and are fragmented across incompatible definitions, temporal resolutions and data types6-8. Past efforts have relied on partial datasets, including flow records, stock estimates and model-based reconstructions with limited coverage9-14. A central challenge is therefore to construct a globally consistent, high-resolution account of migration flows over time. Here we present a new dataset of annual origin-destination migration across 230 countries and regions from 1990 to the present, integrating diverse data sources into a unified modelling framework. By combining official statistics, census-based stocks, net migration estimates and past flow reconstructions, our approach produces temporally detailed and spatially comprehensive estimates that substantially extend existing resources. Using an ensemble of deep recurrent neural networks informed by geographic, economic, cultural and political covariates, we capture both persistent trends and short-term responses to changing conditions-all while propagating uncertainty to generate confidence bounds. Our results outperform existing five-year flow estimates on held-out data and provide finer temporal resolution, revealing previously obscured dynamics in global migration patterns. This framework highlights regions in which uncertainty remains high and data collection is most urgently needed. By releasing all data, code and trained models, we provide a transparent and reproducible foundation for future work. These advances enable a more timely and detailed understanding of human mobility, with implications for research and policy in an increasingly dynamic global system.

References

1. 1.Gaskin, T., & Abel, G. J. (2026). Deep learning four decades of human migration. Migration Studies. https://doi.org/10.1093/migration/mnae014
2. 2.Abel, G. J., & Sander, N. (2014). Quantifying global international migration flows. Science, 343(6178), 1520–1522. https://doi.org/10.1126/science.1248676
3. 3.Azose, J. J., & Raftery, A. E. (2019). Estimation of emigration, return migration, and transit migration between all pairs of countries. Proceedings of the National Academy of Sciences, 116(1), 116–122. https://doi.org/10.1073/pnas.1820871116