摘要
生活方式的改变、医学的进步和遗传学的新发现都会使人的预期寿命变得不确定。本文针对中国人口死亡率历史数据(0~89岁男性数据),利用贝叶斯信息准则和嵌套模型的似然比检验等方法,比较了8种目前流行的随机死亡率模型的拟合效果;同时,检验了这8种随机死亡率模型预测结果的生物合理性和稳定性,并比较了它们的预测效果。结果表明,由Lee-Carter模型拓展而来的Age-Period-Cohort模型最适合于拟合和预测中国的人口死亡率,这为我国寿险企业和养老金机构的死亡率风险管理提供了科学依据。
Life expectancy will become uncertain due to changes in people' s lifestyle, medical progress and new genetics discoveries. In this article, fitting effects of eight popular stochastic mortality models were compared on the base of Chinese historical mortality data (0 -9 years old males). Test methods such as Bayesian Information Criteri- on and likelihood ratio testing of nested model were used in the analysis. At the same time, the biological plausibility and the stability of the predicted results of these eight models were examined and their predictive effects were com- pared. The results showed that, Age-Period-Cohort model, as an extension of the Lee-Carter model, was the most suitable model for fitting and forecasting Chinese mortality rates. This provides a scientific basis for managing mor- tality risks for life insurance companies and pension institutions.
出处
《保险研究》
CSSCI
北大核心
2017年第9期15-31,共17页
Insurance Studies
基金
湖南省社科基金项目(11YBA343)
(14YBA093)
省情与决策咨询项目(2012BZZ29)
湖南省教育厅优秀青年项目(17B286)
中南林业科技大学青年基金重点项目(2012ZD05)的阶段性研究成果