用带趋势的ARIMA模型分年龄预测死亡率  

Forecasting Mortality Using ARIMA Model with the Trend

作　　者：范勇[1] 朱文革[1] 汪丽萍[1] 

机构地区：[1]上海财经大学金融学院,上海200433

出　　处：《保险研究》2015年第5期23-30,共8页Insurance Studies

基　　金：上海财经大学研究生创新基金项目"最优退休年龄;基本养老保障与混合年金制度研究"(CXJJ-2014-333)资助

摘　　要：针对中国1994～2011年中国人口死亡率数据，在Reer修匀、平滑、去趋势后用ARIMA方法分年龄进行拟合、预测。结果表明带趋势的ARIMA模型在平均误差比（MAPE）、光滑度和BIC等指标上优于经典Lee-Catrer模型，其预测结果与随机波动趋势法相比，期望大致相同而方差更小，且原始的抽样调查数据在青少年阶段死亡率整体偏低。Using Chinese mortality data from1994 to 2011, this paper fit the data and forecasted future mortality rate with ARIMA model after applying Reer＇sgraduation, smoothing and detrending to the data. It showed that the ARIMA model with trend was superior to the classical Lee-Carter model for the model criterion indices of MAPE, smooth degree and BIC. Comparing with the Stochastic Volatility and Trend Method, our forecasted result was the same for the estimation of expectation but smaller for variance. Mortality rate of teenagers in the earlier years was relatively lower according to the original sample survey data.

关 键 词：ARIMA模型 Lee-Carter模型 Beer修匀法 随机波动趋势法 

分 类 号：F840.62[经济管理—保险]