混合频率FA-MIDAS类模型构建及其应用  

Construction and Application of Mixed-frequency FA-MIDAS Class Model

作　　者：于扬 房奕彤 王维国[2] Yu Yang;Fang Yitong;Wang Weiguo(School of Statistics and Mathematics,Inner Mongolia University of Finance and Economics,Hohhot 010070,China;School of Econoics,Dongbei University of Finance and Economics,Dalian Liaoning 116025,China)

机构地区：[1]内蒙古财经大学统计与数学学院,呼和浩特010070 [2]东北财经大学经济学院,辽宁大连116025

出　　处：《统计与决策》2022年第21期36-40,共5页Statistics & Decision

基　　金：内蒙古自治区自然科学基金资助项目(2018MS07020);内蒙古财经大学教育改革专项(JXYB1603)。

摘　　要：文章构建了一类能够直接将大量不同频率指标放入同一模型的混频数据因子(FA-MIDAS)模型,深入挖掘了FA-MIDAS类模型的内部结构及驱动机制,推导出了FA-MIDAS类模型非线性最小二乘估计方法。在此基础上,引入多个高频宏观经济影响因素对我国经济增长进行了预测和监测研究。结果表明:非线性最小二乘估计方法能迅速找到FA-MIDAS类模型的收敛解;FA-MIDAS-AR模型在对经济增长的短期预测上具有领先优势,组合模型FA-MIDAS-AR-BIC对新时期经济增长的预测具有较高的时效性及精确性;按照各因子对经济增长的动力程度由高到低依次为高频消费因子、高频投资因子。This paper constructs a kind of mixed-frequency data factor(FA-MIDAS) model which can directly put a large number of different frequency indicators into the same model,deeply explores the internal structure and driving mechanism of FA-MIDAS class model,and derives the nonlinear least squares method of FA-MIDAS class models.On this basis,a number of high-frequency macroeconomic factors are introduced to forecast and monitor China’s economic growth.The results show that the nonlinear least squares estimation method can quickly find the convergent solution of FA-MIDAS class model,that FA-MIDAS-AR model has a leading advantage in forecasting the short-term economic growth,the combined model FA-MIDAS-AR-BIC having high timeliness and accuracy in forecasting economic growth in the new period,and that the driving degree of each factor to economic growth from high to low is high frequency consumption factor and high frequency investment factor.

关 键 词：混合频率 FA-MIDAS类模型 非线性最小二乘法 

分 类 号：F064.1[经济管理—政治经济学]