基于分位自回归的中国人口死亡率动态预测

Dynamic Forecasting the Mortality of China's Population on Quantile Autoregression

采用1994-2012年国家统计局公布的死亡率数据,针对Lee-Carter模型的预测方法提出改进,将分位自回归方法内置到Lee-Carter模型框架中,构建中国人口死亡率的分位自回归预测模型,对未来人口死亡率的变动趋势进行预测。通过与传统的均值回归方法比较,得出结论为:人口死亡率的分位自回归预测方法能够获取更加全面的信息,能够有效应对人类预期寿命被低估的可能,从而对未来死亡率的预测更加合理、可信。

This paper improves the forecasting method of Lee-Carter model based on the data of the year 1 994-2012 from the National Bureau of Statistics.We build a Chinese mortality model with quantile autoregression method to forecast the mortality trend in the future.Comparing the predicted results with the mean regression,we can get more information to overcome the mortality underestimate and make the future mortality prediction more reasonable and reliable.