Prediction Of U.S. Inflation Based On Factor Model With Targeted Predictors

Inflation rate forecasting research has become a popular theme among academics and economic researchers.Forecasting the inflation rate not only helps the central bank and other government agencies to formulate monetary policies in real time to stabilize prices,thereby preventing market shocks caused by inflation,but also helps financial institutions and investors make better investment decisions.But at the same time,the inflation rate forecast is also a challenging forecast series.The quality of the forecast results has a lot to do with the forecasting model methods used.In the current research on forecasting inflation rate,the model methods commonly used by scholars at home and abroad include Phillips curve model,time series models,etc.When predicting the inflation rate,many potentially relevant economic variables are usually involved,and there is often a strong correlation between these economic variables,which leads to the unavailability of many forecasting models or poor forecasting effects.Therefore,this article proposes a new predictive model method is introduced—a factor model with targeted predictors and control variables,and the model method is used to predict the inflation rate in the United States.The core idea of the factor model with targeted predictors is to eliminate redundant variables,and only use variables that have predictive capabilities for predictive variables to perform factor estimation and establish a predictive model,thereby reducing prediction errors.Based on the factor model with targeted predictors,this paper proposes a factor model with targeted predictors and control variables.The innovation of this model lies in the estimation of target predictors and the selection of control variables: First of all,this article uses the "hard" threshold method to screen out variables that have predictive power on the U.S.inflation rate from many potentially relevant economic variables,and estimate target predictors from them,and then establish a target factor model;Secondly,canonical correlation analysis is used to screen out the groups of variables with the strongest correlation with the target predictor from the remaining variables in this article,and the estimated factors are introduced into the target factor model as the control variables,and then the factor model with targeted predictors and control variables is obtained.After constructing the factor model with targeted predictors and control variables,this paper confirms the predictive performance of the model method through simulation experiments,and then conducts empirical research on the basis of simulation experiments.This paper is based on three different models,namely,factor model,the target factor model and the target factor model introducing control variables respectively predict the inflation rate in the United States.The selected sample data contains 118 economic variables.These monthly data come from the Federal Reserve Economic Database from January 1959 to mid-July 2020.According to the results of the empirical research,this paper finds that using the target factor model that introduces control variables to predict the US inflation rate has the smallest forecast error.In the short-term and forecasting and long-term forecasting processes,the forecasting effect of the model is significantly improved.This proves that the target factor model method introduced in this paper with control variables is effective in predicting the inflation rate.