Indpro Prediction Based In Hierarchical Dynamic Factor Model

In the information age,economic indicators are becoming more and more abundant,and factor models can reduce the dimensionality of high-dimensional data,which brings convenience to research.However,the factors extracted by the factor model have no practical economic significance,which greatly limits the application of the factor model in economic analysis.Based on this,Moench(2013)proposed a hierarchical dynamic factor model to capture the structural information between indicators at different levels.This paper establishes a hierarchical dynamic factor model for 128 US macroeconomic indicators,breaking through the idea of stratifying based on regions in previous studies,and divides the indicators into six categories according to economic significance,namely income and output,labor,consumption and investment,price and currency credit,and interest rate and exchange rates.Two methods of buttom-up and top-down are used to extract common common factors and block-level factors,the number of each block factor is determined according to the information criterion and the explanatory power of the factors,and the economic significance of each factor is analyzed.Finally,a prediction model for the public factors and block-level factors of the industrial production index is established and compared with the prediction results of the time series model.In addition,this article makes a comparative analysis of the consumer price index,the total number of employees,the federal funds rate to establish a hierarchical dynamic factor model and its prediction model,and the conclusion of the industrial production index.The main conclusions of this paper are as follows: 1.This paper considers that after extracting common factors,different types of economic indicators have nonlinear structural information.If the indicators are not stratified,this type of information will be directly trapped in the residuals or lost,so this article classifies economic indicators according to their actual economic significance,and extracts the structural information between different types of indicators relatively completely.2.The factors extracted by the hierarchical dynamic factor model are explanatory,which broadens the application of the factor model in various fields.In this paper,the data set is divided into six blocks,and the extracted factors correspond to the corresponding blocks.The factors are highly correlated with the data of the corresponding blocks,which can describe the changing trend of the factors within the block.3.The RMSE of the forecast model based on the hierarchical dynamic factor model is better than the time series model that uses a single variable for forecasting.The main reason is that,on the one hand,the amount of information used by the model is much greater than that used by the time series model.On the other hand,the extracted factors take into account the structural information of the data set,which makes the prediction more accurate.At the same time,in order to improve the prediction accuracy of the model,this paper tries to use different numbers of block-level factors and common factor predictors,and finally determines the optimal number of factors.The common factor selects the first factor that contributes the most,six categories of block-level factors.The numbers are 1,2,1,1,1,3 respectively.