Research On The Influencing Factors Of China's Urban Education Expenditure Based On The Spatial Autoregressive Model

Education expenditure refers to the total expenditure of the government on education.Since the reform and opening up,with the rapid growth of the economy,Chinese governments at all levels have been spending more on education.Educational expenditure affects the future development of education to a great extent.Based on the data of 266 large and medium-sized cities in 2016,in this paper,we make a study on the influencing factors of educational expenditure in each city.In the first part,considering the spatial correlation of education expenditure data among cities,a spatial autoregressive model is established.Firstly,we find that there are multiple collinearities among variables through multiple collinearity test.We adopt stepwise regression method to deal with the problem of multiple collinearity,and retain the more significant variables.Secondly,we use LM spatial autocorrelation-error test to find that it is more appropriate to establish a spatial autoregressive model.The study finds that the estimated value of the spatial autoregressive coefficient rho is 0.097,which indicates that the increase of education expenditure in the relevant cities will lead to the increase of education expenditure in the city;GDP,employees of educational units,the number of students in general middle schools,the number of students in general primary schools,the total collection of books in public libraries,the average wages of on-the-job workers and the proportion of educational expenditure in financial expenditure are positively correlated with educational expenditure;whether it is western region or not has a negative correlation with educational expenditure.In the second part,considering that some independent variables and dependent variables may not be linear,we improve the spatial autoregressive model and establish a partially linear single-index spatial regression model.Because the partially linear single-index spatial regression model allows more complex non-parametric functional relationship between educational expenditure and various influencing factors,and the index item reduces the multi-dimensional explanatory variables to a single index,avoiding the problem of "dimension curse".The virtual 0-1 variable about the western region is taken as the linear part and all other continuous independent variables are taken as the index items in this paper.At the same time,we test whether the function f(?)is a linear function.The test results reject the hypothesis that the function f(?)is a linear function at the significance level of 0.05,that is,the relationship between educational expenditure and indicators is non-linear,and the paper also gives the estimated values of the function f(?)and corresponding 95% confidence interval.