基于队列角度的高龄人口死亡率预测

Mortality Projections for the Elderly Population Based on Cohort Perspective

传统的死亡率预测模型大多以时期效应为出发点,使用人类死亡率数据库(HMD)中的年龄–时期数据。而队列预测通过捕捉不同的队列效应来提高对死亡率发展动态的捕捉,而这也被传统的年龄–时期角度出发的模型所忽视。跟经典的死亡率预测模型相比,本文提出了一种基于年龄–队列数据的C-STAD死亡率预测方法,并对日本和加拿大的成年男性队列死亡率进行分析,通过两次样本外检验来评估模型预测的准确性。结果表明,使用队列数据构建的死亡率模型的误差远小于使用时期数据的误差,具有良好的预测效果。最后文章对两个国家的死亡率从队列角度和时期角度进行预测。

Most of the traditional mortality prediction models take period effects as the starting point and use age-period data from the Human Mortality Database (HMD). Cohort prediction improves the capture of mortality dynamics by capturing different cohort effects, which are also ignored by traditional age-period models. Compared with the classical mortality prediction model, this paper proposes a C-STAD mortality prediction method based on age-cohort data, and analyzes the adult male mortality cohort in Japan and Canada, and evaluates the accuracy of the model prediction through two out-of-sample validations. The results show that the error of the mortality model based on cohort data is far less than that of the period data, and it has a good prediction effect. Finally, the mortality rates of the two countries are predicted from the perspective of cohort and period.