The importance of the reference population for coherent mortality forecasting models

Søren Kjærgaard, Max-Planck Odense Center on the Biodemography of Aging
Vladimir Canudas-Romo, Max-Planck Odense Center on the Biodemography of Aging
James W. Vaupel, Max-Planck Institute for Demographic Research and Max-Planck Odense Center on the Biodemography of Aging

Coherent forecasting models that take into consideration mortality changes observed in different countries are today among the essential tools for demographers, actuaries and other researchers interested in forecasts. Medium and long term life expectancy forecasts are compared for two multi-population mortality models aiming to find the optimal of the set of countries to use as reference population and analyse the importance of the selection of countries. The two multi-population mortality models used are the Li-Lee model and the Double-Gap life expectancy forecasting model. The reference populations is calculated taking into account all the possible combinations of a set of 20 industrialized countries. The different reference populations possibilities are compared by their forecast performance. The results show that the selection of countries for multi-population mortality models has a significant effect on the model's life expectancy forecasts. A small reference population tends to perform better compared to a large group of countries. Even when countries share similar regional history and mortality development, this does not imply that those countries are the optimal reference population among them.