Validation and Mortality Forecasting: Establishing Method Validation Prior to Forecasting
Ricarda Duerst1, Christina Bohk-Ewald2
1Max Planck Institute for Demographic Research, 2University of Helsinki
Accurate mortality forecasts are essential for decision makers to plan for changing needs of pension and other social security systems. Researchers have developed a variety of methods with increasing methodological complexity to forecast mortality developments. Our study's main goal is to establish validation as a test prior to mortality forecasting to confirm that a selected method is suitable. We develop a detailed and objective validation design to assess accuracy and bias of mortality forecasts. For our analysis, we apply our validation design to short-term Lee-Carter forecasts. We use age-specific mortality rates for 24 countries and all available calendar years from the Human Mortality Database. We quantify Lee-Carter's forecast performance with the percentage error for life expectancy and lifespan disparity at birth in three analytical settings. We show that validation serves as a meaningful test to decide whether a method is appropriate to forecast mortality in a country of interest.