Modelling and predicting forced migration

Area of Science:

• Computational Social Science
• Demography
• Econometrics

Background:

• Migration models, particularly flow Fixed-Effects (FE) gravity models, are widely used for understanding and forecasting human mobility.
• Existing FE models struggle to accurately capture the temporal dynamics of human movement, especially in complex scenarios like forced migration.
• The drivers of migration are multifaceted, including economic, geopolitical, and environmental factors, necessitating more robust modeling approaches.

Purpose of the Study:

• To derive and evaluate a novel Flow-Specific Temporal Gravity (FTG) model for migration analysis.
• To compare the predictive performance of the FTG model against traditional FE gravity models.
• To assess the impact of training data length on the accuracy of both FTG and FE models.

Main Methods:

• Development of the Flow-Specific Temporal Gravity (FTG) model, grounded in random utility theory but empirically less restrictive than FE models.
• Training and comparison of both FTG and FE models using EUROSTAT migration data.
• Inclusion of climate, economic, and conflict indicators as covariates in the models.

Main Results:

• The predictive accuracy of migration models is significantly influenced by the length of the available time-series training data.
• The FTG model demonstrates increasing predictive accuracy as the duration of migration time-series data lengthens.
• Conversely, the predictive performance of the FE model degrades with longer time-series data.

Conclusions:

• The FTG model offers superior predictive capabilities for migration dynamics compared to conventional FE models, particularly with extended historical data.
• The FTG model's enhanced performance suggests its utility for more reliable migration projections and forecasts.
• Future research should leverage the FTG model for analyzing complex migration patterns and drivers.