The Study Of China's Real GDP Prediction Based On Mixed Frequency Data Model

GDP(gross domestic product)is the core indicator of national economic accounting and an important indicator for measuring the overall economic situation of a country.So,the attention to GDP has often become a key topic in the development of the global economy.The prediction of GDP growth rate has also become a hot spot for people from all walks of life at home and abroad.Therefore,correct and reasonable judgment of GDP growth rate is not only conducive to our judgment on the future development of the macro economy,but also helps the state and the government to propose policy recommendations in line with the basic national conditions for the macroeconomic development trend.As China's economy changes from a high-speed growth stage to a high-quality development stage,the prediction of real GDP growth rate is more economic and practical.In the past,the traditional real GDP growth rate measurement model only predicted the data outside the sample,but could not predict the data in the sample in real time,which caused the defect that the prediction effect was not timely enough.At the same time,the time-frequency model of the same frequency cannot realize the modeling research using data of different frequencies.The low-frequency data can only be processed into high-frequency data by interpolation,or the high-frequency data can beconverted into low-frequency data by adding and replacing methods.However,the summation method and the substitution method usually ignore the effective value information contained in the high-frequency data,eliminating the volatility characteristic of the high-frequency data;although the interpolation method can process the low-frequency data into high-frequency data,this method will reduce the prediction accuracy of the data.With the maturity of scientific computing technology,the emergence of the mixed frequency data model has broken the constraints of traditional co-frequency data modeling.Therefore,this paper will use the mixed frequency data model to predict the real GDP growth rate in China,and conduct real-time forecasting and short-term forecasting of real GDP growth rate.According to the factors affecting the real GDP growth rate,this paper selects the growth rate of total retail sales of social consumer goods,the growth rate of tax revenue,the growth rate of fixed assets investment completion,the growth rate of total import and export,the growth rate of money supply and the national housing climate index and.The monthly data is used as a high frequency explanatory variable of the mixed data model.Secondly,establishment of the single variable mixed frequency data model and the multivariate mixed frequency data model respectively,and then compared with the mean square error of the traditional measurement benchmark models(OLS ? AR ? PDL and ADL),and draw some conclusions.The empirical results show that,firstly,whether it is the single variable mixed frequency data model or the multivariate mixed frequency data model,it has a comparative advantage when the high-frequency prediction reference period is short(except for the single variable MIDAS(m,K,h)model),but as the reference period becomes larger,the comparison advantages will be weaker,which indicates that the mixed frequency data model may be immediacy and short-term.Secondly,the impact of each high-frequency monthly explanatory variable on the real-time GDP growth rate in the low-frequency quarter is different.In the case of mixed-frequency data model,the total retail sales of consumer goods,tax revenue and fixed assets investment have strong explanatory power.In the case of full-sample prediction,the optimal lag order of the national housing boom index is smaller,which indicates that the national housing boom index contains the expected information of the real GDP growth rate in the future to a certain extent.Thirdly,through the prediction research of the multivariate mixed frequency data model,and the real-time forecast and short-term forecast of China's real GDP growth rate,the final research results show that the growth rate of China's real GDP growth rate is relatively stable in the next four quarters.Finally,the mixed frequency data model can deeply mine the effective information contained in the high-frequency explanatory variables and use it for the estimation and prediction of the model.The model estimates and predictsthe real GDP growth rate better than the traditional co-frequency measurement model,which further proves that China on the basis of the prediction research of the mixed frequency data model,the real GDP growth rate can better reflect the applicability and effectiveness of real-time forecasting and short-term forecasting,and it has very important practical significance for judging China's macroeconomic development.