Allis Albèr: Estimation of mortality rates with multi-population models

Accurate mortality prediction is crucial in the insurance industry, directly impacting risk assessment and financial planning. The Lee-Carter model has long been a pioneer in mortality modeling, and it is widely used due to its simplicity and effectiveness. However, as the need for more precise predictions across multiple populations has grown, several extensions of the Lee-Carter model have been developed to address this challenge. This thesis investigates the performance of various multi-population mortality models, all of which are extensions of the Lee-Carter framework. A key focus is on their out-of-sample predictive ability, which was evaluated using relative Poisson deviance. Additionally, the DUS model, a Swedish approach, is included in the analysis. While no single model consistently outperformed others across all populations, the additive model showed a slight overall advantage. These findings highlight the complexity of selecting an optimal model and the necessity for continued refinement in mortality modeling.

Keywords: Lee-Carter, mortality rates, multi-population models, Poisson Lee-Carter, DUS, addi-
tive model