Common Factor Mortality Model And Its Comparative Study

In recent decades,the mortality rate of the population in various countries has continued to improve,the life expectancy of the population has increased year by year,and the proportion of the elderly population has been increasing.This change affects the development of all sectors of society.In order to effectively deal with this problem,it is necessary to accurately predict the future mortality rate.Due to the lack of mortality data broken down by age in China,the use of the original single-population mortality model in long-term prediction may cause unreasonable crossover or deviation of mortality prediction results,which will affect the accuracy of mortality prediction.The common factor mortality model based on multi-populations can absorb these population deviations by simulating multiple population mortality data at the same time to make up for the shortcomings of the single-population mortality model.In response to the above situation,this thesis applies the multi-population common factor mortality model to the Chinese male and female mortality data on the basis of the single-population model to study its application in China.The specific work is as follows:(1)Further popularize the single-population APC model,develop a new mortality model(EAPC model),and perform numerical simulation on it based on the mortality data of Chinese males and females,and obtain the estimated values of model parameters and the predicted values of mortality;(2)According to the change ideas of the augmented common factor model and its two age reduction models,this thesis transforms the additional time factor of the augmented common factor model into a common time factor,and proposes two innovative time reduction models(ACF3 model and ACF4 model),And classify it as a common factor mortality model,and fit a common factor mortality model based on a multi-population group composed of Chinese male and female mortality data;(3)Use standard residuals and AFE values to compare the goodness of fit of the six mortality models,and then compare their prediction effects based on the predicted residuals and sMAPE values,and perform stability tests.The results show that the ACF3 model extended from the augmented common factor model performs better in fitting and prediction,and the model prediction results are relatively stable,which provides a more scientific basis for multi-population common factor mortality modeling.It is helpful for us to predict the future population mortality rate more accurately.