The Performance Evaluation on the General Procedure for Forecasting Mortality

This study investigates the forecasting ability of the general procedure(GP) using mortality data for South Korean males during 1983-2010. The GP was recently introduced to construct a stochastic mortality model by including every significant demographic feature in historical mortality data. We assess the GP via a comparison with seven existing stochastic mortality models, testing in-sample fit and out-of-sample prediction for three age groups: 1-79, 11-79, and 60-79. The results suggest that the GP consistently outperforms other models with regard to the Bayesian Information Criterion(BIC) and Mean Absolute Percentage Error (MAPE). This shows that the GP extracts optimal risk factors for the projections of age-specific mortality rates from mortality data. Furthermore, we examine predicted levels of uncertainty in forecasts at different ages and show how the risk can be hedged using q-forwards. This information is useful for pension providers or insurers to hedge future unexpected liabilities.