The Statistical Analysis For The Effect Of Accounting Profit On Predicting Consumer Price Index

The consumer price index(CPI)is an important macroeconomic indicator that reflects the level of national price changes and inflation within a certain period of time.The prediction of CPI can help to formulate relevant macroeconomic policies and stabilize prices.Accounting profit is an important indicator to measure the output of an enterprise in an accounting period.The enterprise directly participates in pricing as a producer,so it is significant to apply accounting profit in predicting consumer price index.Through reading domestic and foreign literature,the author found that most of the CPI forecasts at this stage fail to take accounting profits into consideration.This article focuses on the forecasting effect of accounting profits on CPI,and studies the impact of corporate behavior on macroeconomic policies and macroeconomic indicators.This article selects China's CPI and accounting profit quarterly data from the fourth quarter of 2002 to the second quarter of 2020 for empirical analysis.Firstly,the impulse response plots from VAR model find that after a positive shock is applied to accounting profits,CPI shows an equally positive impulse response with a few lags and turns negative in subsequent periods,and similar results are obtained for the sub-sector data,and accounting profits are the Granger cause of CPI.Secondly,different models are built for total profit and sub-sector profit respectively.Among them,ARIMAX(3,0,0),ARIMAX(2,1,3)and LSTM neural network models are built for total profit respectively.The models have good prediction results with a lag of five to ten periods.The KNN regression model and the LSTM neural network model were developed for the sub-sector profit data.Among them,the KNN regression model can better predict the fluctuation of CPI,and the LSTM neural network model has the least prediction error.However,the single LSTM neural network model fails to predict the fluctuation of CPI.Existing literature either uses time series models to forecast consumer price index or uses machine learning to analyze the influencing factors of CPI.This paper tries to use two hybrid models,KNN Reg-LSTM and VAR-LSTM,and proposes a new fusion approach to improve the forecasting effect of LSTM neural network model.And through the empirical analysis,it is found that not only the prediction accuracy of the total profit model is improved,but also the fluctuation trend of the CPI is predicted.Through reading the literature and empirical analysis of Chinese CPI data,this paper explains the direction and mechanism of the influence of micro-firm economic behavior on the counteraction of macroeconomic variables and economic policies,and provides new ideas for studying the counteraction of micro-firm behavior on macroeconomic indicators.