Zhigang Wang, Xiaojun Wang, Xuebin Zhang

【Abstract】

Forecasting the survival’ s number of isolated population consisting of various groups, is the base of the related economic, social issues research. However, when studying the related problems, a general way of obtaining prediction is to estimate expectation of each group’ s survival’ s number, which is categorized according to some characteristics, and then the whole survivors ’ s number can be obtained by summing them. Since this method ignores the correlation of between the different characteristics of the population, the volatility of survival’ s number of people is usually underestimated. In this paper, the Lee? Carter model will be extended by capturing mortality correlations between groups, and we can obtain the whole survival’ s number in one step, using the extended model. Then the existing methods and new methods are compared by theoretical analysis and numerical simulation methods. The conclusion is, only when the expectation can be accurately estimated by the two methods, the old methods can’ t work well both in predicting variability and distribution, so the new method is recommended.

【Keywords】

number of survivors, Lee-Carter model, forecasting