Forecasting mortality rates with the penalized exponential smoothing state space model

It is well known that accurate forecasts of mortality rates are essential to various demographic research like population projection, and the pricing insurance products such as pensions annuities. Recent studies suggest multivariate ages usually not leading indicators in forecasting. Therefore, stochastic models including classic Lee–Carter may necessarily lead more forecasts, compared with sophisticated univariate counterparties exponential smoothing state space (ETS) model. Despite its improved forecasting accuracy, original ETS model cannot ensure age-coherence forecast rates. By introducing an effective penalty scheme, we propose a penalized significantly overcome this problem, discussions on related technical issues reduction parameter dimensionality selection tuning parameter. Empirical results based Australian males females proposed consistently outperforms models. Robust conclusions drawn when scenarios considered. Long-term analyses up 2050 comparing three further performed. To illustrate usefulness practice, application price fixed-term annuities demonstrated.