A Statistical Study On Tax Revenue Forecast

Since the reform and opening up,with the rapid development of China’s economy,the tax work has also made continuous progress.The tax collection and management work has undergone several changes,and the leverage of adjusting the economy has become increasingly prominent.Tax revenue is an important part of the financial revenue of a country or region.It is an important means to help the government play a regulatory role.It is an efficient method to maintain social equity and improve efficiency.The level of regional economic development,industrial structure,tax system structure and tax policy are important factors that affect the tax revenue of a region.The statistical study of tax revenue forecast is theoretically conducive to the exploration of methods;In practice,it provides a basis for tax decision-making and the formulation of tax revenue and expenditure plans.In this regard,taking Henan Province as an example,this paper studies the problem of tax revenue forecasting,and explores the application of statistical forecasting in the field of taxation.This paper analyzes the connotation,types and methods of economic forecasting.With the support of economic theory,it considers economic growth,national income determination,macro tax burden,tax elasticity and functional budget,and extends it to the field of tax revenue forecasting.This paper reviews the change process of tax revenue in Henan Province,analyzes the fluctuation trend and characteristics,and finally selects three main models: short-term multi factor VAR model,medium-term mixed forecasting model and long-term grey forecasting model from different time dimensions of short-term,medium-term and longterm.The results are compared with ARIMA model and exponential smoothing model.The forecasting methods include univariate forecasting model,multivariable forecasting model and combination forecasting model.The vector autoregressive model selects the variables as the year-on-year growth rate of tax revenue,industrial added value above Designated Size,fixed asset investment in the province,total retail sales of social consumer goods and total import and export in the province.The mixed forecasting variables are tax revenue data and PMI,and the grey forecasting variables are the cumulative deformation of tax revenue data;At the same time,in order to explore better prediction accuracy,the combined prediction analysis is carried out.The combined prediction weight is based on the proportion of the reciprocal of root mean square error in its total,which can give higher weight to the model with high prediction accuracy and make the combined prediction effect more scientific and reliable.The results of predicted values and real values of various models are visually displayed in the form of charts.The specific research results of this paper are as follows: according to the information criteria,the lag order of the short-term var prediction model is determined to be 2,and the overall prediction accuracy is3.79%,which is better than the results of the two time series models,showing good adaptability to abnormal fluctuations of data.ARIMA model and exponential smoothing model have poor response to abnormal fluctuations of data.The results of combined prediction analysis from July to December show that the prediction accuracy of the short-term model is improved;The model selected for medium-term mixing prediction is Midas(3,3,2)-ar(4)(m=3,k=3,h=2,p=4)model under the exponential Almon weight function.The final prediction accuracy is 3.82%,which is better than 6.5% and 5.4% of the two models,and the combined prediction accuracy is 3.22%,which can be improved to some extent,but it is not obvious enough;The sample length of the long-term grey prediction model is 8,and the overall prediction accuracy is 5.21%,which is slightly lower than the previous short-term and medium-term prediction accuracy.It may be due to the long-term prediction,but it is still referential,which is close to the prediction accuracy of the exponential smoothing model of 5.42%.The prediction accuracy of ARIMA model is slightly lower,which is 6.34%,and the combined prediction result is 5.64%,which can also be improved.Finally,the overall results of the article are summarized.This paper holds that the short-term var forecast,medium-term mixing forecast and long-term grey forecast data have good prediction effects,and can be used to predict the tax revenue of Henan Province.The combined forecast can optimize the prediction accuracy to a certain extent,but depending on the prediction results and weight selection of the previous models,the prediction effect of the time series model is slightly worse,which is considered to be the reason for the long prediction period;In addition,three problems should be paid attention to in the process of Tax Forecasting: first,the forecasting process needs to combine quantitative analysis with qualitative analysis;Second,combine the medium and long-term trend analysis with the impact analysis of emergencies;Thirdly,the selection of forecasting methods should be flexible,which should consider the actual situation and model structure,and be flexibly applied based on the comprehensive consideration of experience,reality and theory.The innovation of this paper lies in the application and method innovation.The introduction of VAR model for short-term tax revenue forecast,the improved mixing data model for medium-term tax revenue forecast,and the grey model for long-term tax revenue forecast belong to application innovation;Analyzing the data of different dimensions of tax revenue in Henan province belongs to the level innovation of forecasting methods;Secondly,taking Henan Province as an example,this paper studies the prediction of provincial tax data,which can provide theoretical and tool reference for the prediction of tax revenue in other regions.The deficiency lies in that the discussion focuses on the macro level data,and the micro investigation is insufficient.