Dynamic Mortality Portfolio Forecasting Method And Its Application In Commercial Endowment Insurance

With the development of society and improvement of medical and health conditions,China’s population mortality rate continues to decrease and life expectancy per capita continues to increase,which brings mortality risk to China’s life insurance industry,and it is particularly important to accurately predict the future mortality rate of China’s population.In this context,five dynamic mortality models were firstly selected for the study.The paper adopted the national age-and sex-specific population mortality data from 1997 to 2020,which was divided into age groups of 0-90 years,and applied the iterative method to fit the five dynamic mortality models.Secondly,the paper assessed the effect of single dynamic mortality based on tests of residual plots,as well as comparisons of statistical values of AIC and BIC information measures and relative error comparisons.Furthermore,the combined prediction model was constructed using simple weighting,ordinary least squares weighting,and L1 parametric regularization weighting.After comparing the merits of the individual and combined forecasting models using mean squared error values,future population mortality projections are made based on the constructed dynamic mortality combined forecasting model.In the end,the impact of mortality reduction on our commercial pension insurance products is analyzed based on the predicted mortality values.After theoretical and empirical research,this paper mainly draws the following conclusions:(1)After analyzing the raw mortality data,it is found that the mortality rate of our population has improved significantly,and with different years as the observation perspective,it is found that the mortality rate of both men and women shows a decreasing trend as the year increases.With different ages as the observation perspective,the mortality data showed a clear trend of decreasing and then increasing,and the linear trend was more obvious in the middle and late ages.(2)In the process of fitting single dynamic mortality models,this paper finds that all single prediction models capture the time effect relatively well,but most of them do not capture the birth-year effect in the high age range as well as the birth effect in the low age range well enough.Moreover,this paper also finds that the order of merit of the single models is: RH model,Lee-Carter model,APC model,M7 model,and CBD model.(3)In the comparison of the three combined approaches with the single model,ordinary least squares weighting gave the best results and regularization the second best.Both combination models fit better than the single model,which reflects the advantages of the dynamic mortality combination prediction method in this paper.(4)In this paper,a10-year projection was performed based on a dynamic regularized combined projection method.The results of the projections show a general trend of decreasing mortality over time for the same age projections as well as higher mortality at higher ages than at lower ages for the same year projections.The paper then applies the mortality projections to the calculation of the single pure premium for a term survivor annuity and finds that the mortality risk borne by the insurance company is increasingly severe due to the decreasing mortality rate.