中国季度GDP预测——断点是否改善预测精度?

Quarterly GDP Forecasting in China——Does the Breakpoint Improve Forecasting Accuracy?

结构时间序列模型使用卡尔曼滤波算法具有较高的预测精度。文章引入多点多类结构断点,构建了一组具有趋势、季节、周期、不规则等成分的基准模型,以及三组扩展模型,应用BP断点检验探测断点位置;进一步在断点位置考察不同断点类型,通过干预效应的t检验进行验证。比较四组模型的信息指数和预测方差,结果均表明中国季度GDP趋势中2008年第四季度的次贷危机为水平型断点和2020年第一季度的新冠疫情为脉冲断点拟合最佳,这也验证中国经济已成功由高速增长转向高质量增长。

Structural time series model using Kalman filter algorithm has high prediction accuracy.This paper first introduces multi-point and multi-class structural breakpoints,and constructs a set of benchmark models with factors of trend,season,period and irregularity and three groups of extended models,and also uses BP breakpoint test to detect the location of breakpoints.Then,the paper examines different types of breakpoints at breakpoint locations,and uses the intervention effect t-test to do the verification.Finally,the paper makes comparison of the information index and forecast variance of the four models.The results show that the horizontal breakpoint of the sub-prime crisis in the fourth quarter of 2008 and the pulse breakpoint of the COVID-19 outbreak in the first quarter of 2020 are the best fitting of China’s quarterly GDP trend,which also proves that China’s economy has successfully transformed from high-speed growth to high-quality growth.