Expansion Of Mortality Model And Prediction Of Chinese Mortality

As the economy developing, the social environment improving, the level of medical technology enhancing and people’s awareness of health increasing, the overall trend in death rates is clearly downward, human life expectancy is prolonging and the impact of longevity risk for national pension systems, enterprise annuity, the insurance company pension business is increasingly prominent. Since the longevity risk stems from the unexpected changes of mortality, more accurate predictions of the mortality will become an important aspect on the research of longevity risk. In view of this, this paper will have further research on mortality model and forecast the Chinese mortality.Our research is on the basis of fuzzy Lee-Carter model. Firstly, the Weighted Least Square method for the fuzzy model is proposed which the weight is deaths corresponding to the specific year and age. And four approaches including two methods for Lee-Carter model and two methods for fuzzy Lee-Carter model are compared:Singular Value Decomposition, Weighted Least Square for Lee-Carter model and Ordinary Least Square, Weighted Least Square for the fuzzy model. By computing the error and MSE, we find that the WLS method for fuzzy model has the best fitting effect. Secondly, by using a definition about degree of fuzzy sets’nearness, we propose a new index for evaluating the performance of different fuzzy models. Finally, under the fuzzy model, the extreme value theory (EVT) is applied to have prediction of mortality factor.Due to the lack of Chinese mortality data, the EVT method which makes well forecast for Japanese mortality is impossible to apply into prediction of Chinese mortality. So we use the co-integration theory to construct error correction model between Chinese and Japanese male mortality. This model provides a new idea for Chinese mortality prediction.