1Western financial investment in education and economic growth Empirical Analysis ofAbstract: Since the Schultz human capital theory began, more and more investment in education, economists began to affect a countrys economic development as the endogenous variables through the use of panel unit root tests and panel cointegration analysis of this more effective method of data in an integrated way to examine the western financial investment in education funding and long-term relationship between economic growth based on Granger causality test found that the western financial investment in education and economic growth exists between the two long-term stable equilibrium relationship, financial investment in education causes of economic growth, increasing lag, the conclusion is very robust. Keywords: financial investment in education; economic growth; panel data CLC number: F810.4 literature flag code: A Article ID :1673-291X (2010) 01-0014-02 I. Introduction 2Analysis method is: the western financial investment in education and the relationship between economic growth cointegration test, on the basis of further causality test, both from the statistic point of view whether a causal relationship, the final analysis and discussion of financial investment in education and the economy growth in the long-term equilibrium relationship between changes. The introduction of the 12 western provinces and autonomous regions, the panel data from 1996-2008, which due to the representativeness of the data and taking into account availability, was chosen western provinces of the state education budget (GAE) as representatives of investment in education variable, while the western provinces selected per capita (GDP) as a measure of economic growth indicators, the data from the , and its treatment on the number. Second, the model and measurement methods (A) model and data sources Comprehensive results of previous studies, combined with Chinas specific conditions, this paper also uses Apergis et al (2007) model to test the western 3financial investment in education and economic growth cointegration relationship between: yit = 0i + 1iFit + it (1) Which, yit that the logarithm of per capita GDP growth, Fit the western financial investment in education, Xit is a set of control variables. (B) the measurement methods 1 panel unit root test Before starting the cointegration test, we first need to examine the model variables are stationary or non-stationary, that is, each sequence contains a unit root. Im, Pesaran and Shin (1999,2002) proposed a heterogeneous panel data (Heterogenous panel data) unit root test, referred to as the IPS test. IPS unit root test and compared to other panel unit root tests there is less restrictive and more effective advantages. IPS-type test test is: yi, t = iyi, t-1 + ij yi, t-j + zi, t + i, t (2) Which, yi, t represents the model of each sequence; zi, t is a fixed effects or time trends include the decision variables. IPS test to relax the cross-sectional time-series regression coefficients of the first order from this constraint must be the same conditions to test the null hypothesis is H0: i = 0 (i = 1,2, ., N) is assumed to be prepared: i0 (i = 1,2, . N1), i = 0 (i = N1 +1, N1 +2, ., N). IPS statistic is the average of individual 4ADF test statistics constructed on the basis of the amount. (2) panel cointegration test We use the Pedroni (1999) panel cointegration test proposed method, because Gutierrez (2003) pointed out that when T becomes large, Pedroni test than Kao (1999) and Larsson et al (2001) test is more effective. Pedroni tests do not exist Cointegration test of the null hypothesis is mainly assumed cointegration regression by calculating the return of more than. Pedroni constructs seven to regression residuals statistics, four of which are within the group with combined dimensions (within dimension) described, denoted Panel v, Panel rho, Panel PP, Panel ADF, four different statistics is assumed cross-sectional regression coefficients have the same self, the other three dimensions with a group (between dimension) described, denoted Group rho, Group PP, Group ADF, three different statistics assuming different from cross-sectional regression coefficient taking into account the nature of this small sample of data, the Panel ADF test in Pedroni and Group ADF statistic than the other statistics have better small sample properties, so the model in small samples the 5main reference Panel ADF and Group ADF statistic to determine whether there is cointegration. 3 panel cointegration estimation Given variable is the case of the use of cointegration Stock and Waston (1993) dynamic least squares method to estimate long-term relationship, mainly due to co-integration method of least squares regression equation estimated as endogenous variables and serial correlation is biased In the DOLS estimation equation, the e