1、 Research contents and methods
（1） Research content
There are many factors that affect China's fiscal revenue. Due to different research methods and perspectives, the analysis results will also be different. Through data collection from various parties and combined with the actual situation of China's economy, this paper selects six main influencing factors: agricultural added value (x1), industrial added value (x2), construction added value (x3), total population (x4), final consumption (x5), and disaster affected area (x6). The data of the paper is from the 2015 National Statistical Yearbook.
（2） Research methods
There are many factors that affect the national fiscal revenue. This paper mainly makes a quantitative study on the factors that affect the national fiscal revenue, and uses the stepwise regression analysis method to study the factors that affect the national fiscal revenue. The basic idea of stepwise regression is to introduce variables into the model step by step. For each explanatory variable, F test and t test should be carried out. When a new explanatory variable is introduced and the original explanatory variable becomes no longer significant, the newly introduced explanatory variable will be eliminated. To ensure that the regression equation contains only significant variables before each new variable is introduced. Regression analysis is to establish a regression equation based on the optimal combination of explanatory variables to predict the development trend of the explanatory variables, which needs a lot of data as support. This paper uses statistical software eview for auxiliary analysis.
2、 Research process and analysis
（1） Establishment of econometric model
According to the selection of influencing factors of fiscal revenue, we can establish the following regression analysis model:
Using data and Eviews software analysis, using the least square method, the model results are as follows:
（2.493493） （-2.849060） （7.742058） （0.969910） （-2.045158） （4.135405）
（-0.692149）
R2=0.999291 F=5403.738
According to the above results, r2=0.999291 It shows that the fitting degree of the model is very high. At a given significance level α= 0.05. The critical value of t-test is 2.064. According to the model results, the T values of X3, X4 and X6 are all less than the critical value and cannot pass the t-test. It indicates that there may be multicollinearity between explanatory variables.
According to the correlation coefficient matrix, the correlation coefficient between each explanatory variable is high, especially the correlation coefficient between X1 and X2 is as high as 0.994919. It shows that there is multicollinearity between the explanatory variables.
（2） Stepwise regression analysis
1. Y carry out univariate regression for each explanatory variable. Use the least square method to calculate the univariate regression equation of Y for each explanatory variable, as follows:
Through the analysis of the above univariate regression equation, it is known that the R2 of the univariate regression equation with X3 as the explanatory variable is the largest, so X3 is selected as the first explanatory variable of the regression model to form the univariate regression model.
2. Gradual regression. According to the univariate regression model, the remaining five explanatory variables are added to the regression model by using the stepwise regression analysis method to form the following binary regression model, as shown below:
By analyzing the above binary regression equation, it can be seen that R2 in the binary regression model formed by adding the new variable X4 is the largest. It shows that the fitting degree of this model is very high. Retain the explanatory variable x4, and establish a binary regression model with X3 and X4 as explanatory variables. Then, add the remaining explanatory variables to the binary regression models of X3 and X4 respectively to obtain the following four ternary regression models, as shown below:
By analyzing the above ternary regression model, it can be concluded that R2 in the ternary regression model formed by adding the new variable X2 is the largest. It shows that the fitting degree of this model is very high, so the explanatory variable X2 is retained, and a ternary regression model with X3, x4, X2 as explanatory variables is established. Add the remaining explanatory variables to the ternary regression models of X3, X4 and X2 respectively to obtain the following three quaternary regression models, as shown below:
Similarly, R2 in the quaternion regression model formed by adding a new variable X5 is the largest. It shows that the fitting degree of this model is very high, which is in line with the actual fiscal revenue of our country. Retain the explanatory variable X5, and establish a quaternion regression model with X3, x4, X2, X5 as explanatory variables. Add the remaining explanatory variables to the quaternion regression models of X3, x4, X2 and X5 respectively to obtain the following two five variable regression models, as shown below:
Similarly, R2 in the five variable regression model formed by adding X1 is the largest. It shows that the fitting degree of this model is very high, which is in line with the actual fiscal revenue of our country. Retain the explanatory variable x1, and establish a five variable regression model with X3, x4, X2, X5, X1 as explanatory variables. Add the remaining explanatory variables X6 to the five variable regression model of X3, x4, X2, X5, X1 to obtain the six variable regression model, that is, the most needed regression model, as follows:
3. Establishment of stepwise regression model. Through the above stepwise regression model analysis, the factors affecting China's fiscal revenue are added to the model in turn, and finally the equation between China's fiscal revenue and the six influencing factors through stepwise regression analysis is obtained as follows:
Y=32887.09+0.5925254x3-0.273504x4+0.321920x2+0.344622x5 -0.809875x1-0.035997x6
According to the stepwise regression, the final regression equation is shown in the above figure. The actual value is highly close to the fitted value, as shown in the following figure, and the maximum error is only 3.3%.
3、 Conclusion
The above result analysis adopts eviews7.0 software. The regression model established through stepwise regression analysis has a very good fitting effect, and the maximum error is only 3.3%. The model shows that agricultural added value, construction added value, final consumption and industrial added value have a particularly prominent impact on China's fiscal revenue. Among them, the impact factor of agricultural added value is -0.809875, indicating that the impact of agricultural added value on fiscal revenue is particularly significant, and the relationship between the two changes in the opposite direction, that is, the increase of total agricultural output value has a negative impact on the increase of fiscal revenue. According to China's national conditions, China's agricultural tax has been abolished as early as 2006, while the total agricultural expenditure in fiscal expenditure is increasing year by year. Therefore, the agricultural added value analyzed above has a negative impact on the increase of fiscal revenue. From the final multiple regression model, we can see that the added value of the construction industry has a greater impact on fiscal revenue. For every 100 million yuan increase in the added value of the construction industry, the fiscal revenue will increase by 59.3 million yuan. It shows that China should strengthen infrastructure construction and vigorously develop the construction industry to promote economic growth. Increase national fiscal revenue.