Backtesting stochastic mortality models by prediction interval-based metrics

Abstract Human lifespan increments represent one of the main current risks for governments and pension health benefits providers. Longevity societies imply financial sustainability challenges to guarantee adequate socioeconomic conditions all individuals a longer period. Consequently, modelling population dynamics projecting future longevity scenarios are vital tasks policymakers. As an answer, demographic actuarial literature have been introduced compared several stochastic mortality models, although few studies thoroughly tested uncertainty concerning projections. Forecasting levels central role since they reveal potential, unexpected rise related economic impact. Therefore, present study poses methodological framework backtest in projections by exploiting metrics not yet adopted literature. Using data from Mortality Database male female populations five countries, we some numerical applications illustrate how proposed criterion works. The results show that there is no model overperforming others cases, best choice depends on considered.