Research on the Application of Mixed Frequency Model in Macroeconomic Forecasting in China
In recent years,there have been significant shortcomings in China's macroeconomic forecasting models,such as a significant decline in stability and a significant decrease in prediction accuracy.It is urgent to improve macroeco-nomic forecasting models.By integrating dynamic factor model and mixing data sampling model,a factor mixing data sam-pling model FA-MIDAS was constructed,which achieved deep mining and information extraction of over 120 macroeco-nomic indicators,and the quarterly GDP growth rate of China is predicted both in-and out-of-sample.The results show that,the fitting of FA-MIDAS model within the sample is better than that of the model without mixing data and the model that converts monthly data into quarterly data by simple averaging.Compared with the same frequency model,the FA-MI-DAS model can improve prediction accuracy.When the data is in a relatively stable stage,adding the autoregressive terms in the model will indeed improve the prediction accuracy.However,when the data fluctuates greatly,adding autoregressive terms will have an inertial trend which will reduce the prediction accuracy of the model.With the increase of data released in the current quarter,the accuracy of nowcasting and forward multi-step prediction of GDP will be improved accordingly.
mixed frequency modelFA-MIDASdynamic factor modeleconomic forecastingnowcasting
NETL
NSTL
万方数据
