Accounting for geographic and social heterogeneity in models of human migration

Human migration significantly influences demography and economics in our globalized world. The impacts on origin and destination countries can vary widely: positively by alleviating labor shortages and contributing to cultural diversity, and negatively by causing brain drain or social integration challenges. Migration studies aim to unravel these complex dynamics by analyzing movement patterns and identifying underlying drivers. This cumulative thesis explores heterogeneities in human mobility and suggests modeling approaches to improve performance under specific scenarios. Departing from the common idea of universal laws for human migration, I am trying to explain a large range of applications and behaviors with a single modeling framework. We begin by examining spatial heterogeneities associated with internal migration, focusing on traveled distance and direction. Our investigation employs two models: the Radiation Model, which predicts migration flows based on population sizes and distribution without fitting parameters, and the Gravity Model, which uses population size and traveled distance, allowing for multiple fitting parameters. In the first part, I address heterogeneities connected to traveled distance. Using data on internal migration from the USA and Mexico, we demonstrate that the Radiation Model consistently underpredicts long-distance moves, whereas the Gravity Model performs well for large distances. We propose an extension of the Radiation Model by introducing a single parameter. This parameter can be interpreted as an increasing loss of information about potential destinations with increasing traveled distance. Our extended model fits the data across the full spectrum of traveled distances while remaining analytically tractable. The second part investigates the moving direction for internal migration using the Radiation Model. We identify a conceptual discontinuity in the original model related to how populations are counted. Additionally, neglecting migration direction, such as towards or away from large metropolitan areas, can significantly affect migrant number estimations. Our proposed modification incorporates the angle between alternative potential destinations, considering that migrants may have directional preferences. This modification addresses the conceptual inconsistency and enhances the model's fit to observational migration data without introducing fitting parameters. Next, we shift focus to demographic heterogeneities in international migration projections. We extend a global Gravity-type Model of bilateral migration to include age, sex, and education dimensions while accounting for classical economic and social drivers. This enhanced model allows coupling with a widely used cohort-component population model, enabling future migration projections until 2050 under various socioeconomic scenarios. Furthermore, the enhanced migration model fits the data better than the original model without demographic dimensions and is more sensitive to socioeconomic scenario choices. When integrated with the population projection model, it aligns stylized migration assumptions with quantitative estimates of key migration drivers, capturing interactions between migration and other demographic processes. We conclude that complex movement patterns can only be explained by accounting for heterogeneities in individual and local factors driving human migration. While minimalist approaches can provide insights into general movement, they are insufficient to explain the complex migration streams found in data and should be extended for more accurate results. We demonstrated this for two spatial heterogeneities (traveled distance and direction) and demographic heterogeneities (age, sex, and education). Providing some fundamental insights into these topics, future research should extend the identification and model implementation of heterogeneities to further improve our understanding and prediction of human migration patterns.