Sichuan Provincial Tax Revenue Forecast Model Discussion And Empirical Analysis

From the birth of the revenue, it has been highly important for the state. So the rulers of the past dynasties attached great importance to the collection of the revenue and focused on the plan and forecasting of it. Under the market economy, revenue stems from the whole economy. So the outstanding revenue plans must connect the revenue with the economic development .And forecasting the revenue with scientific approach is vital to effective revenue plan. The thesis adopts econometrics and time series tools to analyze the revenue forecasting model of Sichuan province. It has certain reference value to standardize government budget, to assign revenue rationally, to arrange expenditures and to make the revenue and economy growth inter dynamically grow.First, the author reviews some papers about the revenue forecasting model, such as Cointegration Analysis, Error Correction Model, Granger causality test, VAR model, stepwise regression and combination forecasting. This thesis analyses the impact of the macroeconomic variables on the revenue in Sichuan Province from different aspects systematically, then set up three revenue forecasting model, by the way of quantitative analysis as the basic and the qualitative as the supplement. At last, the author tests the models by using practical data and forecast the revenue of Sichuan Province in 2004 by using combination forecasting.The whole thesis is divided into 5 chapters: Chapter 1 is the preface. The first part discusses the background and the purposes of the whole thesis. The second part studies some papers about the research of revenue forecasting model, Cointegration Analysis, Error Correction Model, Granger causality test, VAR model and so on. The third part introduces the methods and the structure of the thesis.Chapter 2 carries on the analysis of the VAR revenue forecasting model of Sichuan Province. Part 1 introduces the theory of choosing GDP as the main variable, and where the data from and how to adjust it. Part 2 is the main body of this chapter, deeply analyzes the cointegration and Granger causality relationship between revenue and GDP. The result indicates that there has no long-term cointegration relationship between the two variables, and the revenue is the Granger causality of GDP, but GDP is not the Granger causality of revenue. The establishment of the VAR model is not based on the cointegration relationship and economic theory. Parameters tell us that the model is properly fitted. Part 3 uses the historical data to test the precision of the VAR model, which shows that the model gives good forecasts. Part 4 uses Unit root and Cointegration test approaches to analyze the long-term behavior among the revenue, the first industry GDP, the second industry GDP and the third industry GDP. The result indicates that there has no long-term cointegration relationship among them.Chapter 3 carries out the ECM revenue forecasting model of Sichuan Province. The first part introduces the economic theory of the relationship between the logarithm of revenue and the logarithm of GDP, the coefficient of liner regression is the revenue elasticity coefficient. It also introduces the source and adjustment of data. In the second part, Cointegration test proves that there has long-term cointegration relationship between the logarithm of revenue and the logarithm of GDP, and the revenue elasticity coefficient is 0.6114.Then we set up the ECM model based on the cointegration relationship. In part 3, we use the historical data to do quantitative test. Result indicates that the model has good forecast except the year when revenue policy was modified. In another word, the model is sensitive to the policy and not steady enough. Chapter 4 analyzes the stepwise regression revenue forecasting model of Sichuan Province. Part 1 introduces the economic theory of choosing the macroeconomic variable, and the source and adjustment of data. Part 2 adopts the method of stepwise regression through important variables flow into the model one by one. The normality assumption test an...