Kumulative Dissertation, Otto-Friedrich-Universität Bamberg, 2025

Demographic forecasting is essential for planning infrastructure, healthcare, and social services, yet traditional population projections—often deterministic and scenario-based—fail to capture uncertainty or provide probabilistic insights. This thesis advances probabilistic population forecasting by developing Bayesian methods for mortality, fertility, and net-migration, enabling accurate age-specific population predictions at national and regional scales.
The first chapter introduces spatial extensions to the Lee-Carter and Age-Period-Cohort models, improving regional mortality forecasts by accounting for spatial correlation and using Bayesian stacking for robustness. The second chapter addresses pandemic-induced mortality shocks, proposing a model for correlated jumps that better reflects the gradual decline of excess deaths observed during events like COVID-19. Finally, the thesis integrates these methods into a probabilistic cohort-component framework, applying it to subnational regions in Upper Franconia, Bavaria. By combining regionalized forecasts of mortality, fertility, and migration—including Dirichlet regression for migration patterns—this work enhances the precision and reliability of small-area population projections.