预测我国人口总量的具有外生变量的半参数自回归模型

Semiparametric Autoregression Model with Exogenous Variables for Forecasting Chinese Population

由于现有人口总量线性预测模型不能理想地描述人口这一非线性问题,根据1952-2005年我国人口总量数据和GDP总量数据自身特点及建模的需要,首先对数据进行对数变换和差分运算得到平稳时间序列,然后将GDP总量这一外生变量的滞后量和人口总量的滞后量作为被解释变量建立了含有外生变量的半参数模型。利用RMSE法(即最小化拟合残差)来选择显著变量及最佳带宽,最后检验所建模型的随机误差项服从独立正态分布。结果表明所建模型能较理想地描述非线性问题,其预测精度与其它线性模型相比有了一定的提高。

In order to overcome the problems of forecasting Chinese population in the existing literature such as lacking the forecast function and the bigger error,semi-parametric model with exogenous variable using the data in 1952-2005 is constructed.The population total quantity and its lag are served as the explainary variables.The significant variable and the best bandwidth are chosen by the RMSE criterion.The result showed that the model is reasonable, and has better prediction than linear model.