A1096
Title: Bayesian multi-population Lee-Carter model applied to Japanese mortality data
Authors: Hao Chen - Hiroshima University (Japan) [presenting]
Haruhisa Nishino - Hiroshima University (Japan)
Abstract: The Lee-Carter model has been widely used for analyzing mortality. Bayesian approaches enable us to extend the model and achieve more flexible modelling. On the other hand, mortality research has recently focused on modelling data collected by categories such as regions. For analyzing these data, Bayesian multi-population mortality models are used, including a spatial conditional autoregressive (CAR) model, to prefecture-specific data from the Japanese Mortality Database (JMD). By comparing the models using the widely applicable information criterion (WAIC), the results show that the model containing the CAR outperformed other models without it. It suggests that the spatial effect should be included in analyzing prefecture-specific mortality data in Japan. The impact of great earthquakes on Japanese mortality rates is also examined, including the Great East Japan Earthquake and the Great Hanshin-Awaji Earthquake. The findings highlight the importance of considering the spatial effect in mortality research and demonstrate the benefits of using the multi-population mortality model with CAR parameters. The implications could help improve mortality forecasting and advance the understanding of mortality trends in Japan.