Research On Macroeconomic Variables Forecasting Based On Three-Pass Regression Filter(3PRF) Model

Accurately predicting macroeconomic trends is of great significance to the highquality and sustainable development of my country's economy.Traditional models for macroeconomic variable prediction in China usually include vector autoregressive models,principal component analysis models and extended models derived therefrom,but these traditional models exist in the use of large-scale variables,screening suitable variables,and prediction accuracy.Certain limitations.In order to improve the above problems,this paper attempts to apply a three-pass regression filtering(3PRF)model based on large-scale explanatory variables for macroeconomic forecasting,which can consider the noise variables with indirect effects and some unobservable latent variables on the target variables.,And has higher prediction accuracy than traditional models.The model is mainly used for variable prediction of stock market abroad,and its application in domestic macroeconomic research is still blank.From the perspective of the prediction performance of the 3PRF model,this study first summarizes the theories and characteristics of the four traditional prediction models and compares them with the 3PRF model.Research shows that the 3PRF model is constructed in three steps,and each step is calculated by OLS,so that the relationship between the observed value and the regression value can be found indirectly,which is more logical than the traditional model.Then,through the literature screening method,find out 70 indicators commonly used by experts and scholars for GDP growth rate prediction,construct an indicator system according to its economic significance,and build a 3PRF macroeconomic index on this basis,so that it can be more clearly demonstrate the actual economic meaning of variables after dimensionality reduction.Statistical analysis found that the 3PRF macroeconomic index based on the 3PRF model is more advantageous in fitting GDP,and the 3PRF macroeconomic index is more stable than the index obtained by the principal component analysis.The relatively unchanging data will lay a good foundation for subsequent research and analysis.Finally,taking the forecast of GDP growth rate as an example,an empirical comparative analysis of the model's prediction accuracy is carried out.The empirical results of the simulation training of the model in the training set show that the 3PRF macroeconomic index model has the best overall model fitting effect,followed by the SW method proposed by Stock and Watson.The root mean square result of the model's prediction error in the test set shows that the 3PRF macroeconomic index model has the highest prediction accuracy,followed by the SW method proposed by Stock and Watson.The results of the DM statistical test show that the 3PRF macroeconomic index model is statistically significantly superior to the SWTF model.