| As an important indicator to measure a country’s macroeconomic performance,GDP can not only reflect the prosperity and decline of a country’s economy,but also provide a basis for the formulation of macroeconomic policies.Thus,it is important to accurately grasp the development trend of GDP.However,since the data frequency of quarterly GDP and forecasting indicators are often different.The traditional methods,which construct the models by change the mixed-frequency data into the same-frequency data,will result in loss or inflated information.By summarizing relevant research methods at home and abroad,it constructs MIDAS model and its extended forms which are based on the modeling theories of the mixed-frequency models to predict quarterly GDP in China.Firstly,it introduces the modeling theories of the mixed-frequency model and two error correction models,ECM-MIDAS model and CoMIDAS model are derived from the U-MIDAS model and the same-frequency error correction model.Then,the “Keqiang Index” is used as the monthly forecasting index,it respectively constructs the univariate MIDAS models and the multivariate MIDAS models which are based on the five weight functions to predict the year-on-year growth rate of China’s quarterly GDP.Furthermore,M2 is used as the monthly forecasting index,the ECM-MIDAS model and the CoMIDAS model are constructed to predict the sequential growth rate of China’s quarterly GDP,respectively.The construction of the two mixed-frequency error correction models is not only based on the five weight function forms,it also builds unconstrained models.Then,the forecast indicators and related graphs are introduced to compare the prediction accuracy of the mixed-frequency models with the same-frequency models.The results show that the mixed-frequency models can improve the prediction effect of the same-frequency models,so the mixed-frequency forecast of China’s quarterly GDP is effective.For the mixed-frequency models which are based on the “Keqiang Index”show that to some extent,the multivariate MIDAS models can improve the prediction accuracy of the univariate MIDAS models.The “Keqiang Index” is still an important indicator to forecast GDP,but the indicative role of rail freight volume on China’s GDP has weakened.The two mixed-frequency error correction models which are based on M2 provide new ideas for the analysis of the sequential growth rate of China’s quarterly GDP.Then,these two models also show that the addition of error correction can significantly improve the prediction accuracy of the mixed-frequency models,both the ECM-MIDAS model and the CoMIDAS model.In this case,the ECM-MIDAS model has a slightly better prediction effect,but the difference of prediction errors can be ignored and theselection of the models has little effects,so these two models are complementary rather than substitute.In addition,the unconstrained mixed-frequency model is more effective when the frequency difference and the lag order are small.Although the beta weight function has better performance in the constrained mixed-frequency model,there are no absolute advantages for the five weighting functions,so it must be specifically analyzed. |
