MONEY AND INFLATION N CHINA △ogM1=b+2-1+∑81△logM-+∑82△logP s=1 +∑δ,△IP,-+n2 In these equations v's are the disturbance terms and the error-correction term ss are the residuals obtained from the cointegrating regressions reported in Table 2. Since AP and AM appear to be cointegrated, the Granger Representation Theorem implies that there must be granger causality between them at least in rection (Engle and Granger, 1987). For example, the null hypothesis of noncausality from money stock to prices is rejected if either the group of coefficients on the money stock variable in the inflation equation, 25=182s,are jointly significant or the error-correction coefficient, ((1), is significant. The optimal error-correction model is specified using Hendry's (1984) general-to-specific modeling strategy. That is, a more general model was spec ified in which growth rates of prices and money stock were regressed on the full vector of exogenous and endogenous variables with a two-period lag. Through stepwise regression, insignificant variables were eliminated to yield a data congruent parsimonious specification. Once the variables and optimal lag lengths in each equation are identified, the price and money equations are pooled together to form a system of simultaneous equations. The system is estimated with the aid of Zellner's Seemingly-Unrelated Regression(SUR) method to obtain more efficient estimates, recognizing the fact that the errors across the equations exhibit significant correlation. The results are reported in Table 4. It is evident that the error-correction terms in both the price and money equations are negative and statistically significant In the next step, a series of Wald's x tests were performed to ascertain the joint significance of each of the explanatory variables in the two-equation error correction model of Table 4. The results of Granger causality tests are reported in Table 5. The results suggest that lag on the money stock variable in the price equation is statistically significant. The x value of 15.28 corresponding to the null hypothesis Ho: 82 =0, together with a highly significant error-correction coefficient, decisively rejects the null hypothesis of noncausality from money stock to prices. Judged by the values of asymptotic x statistics, the price Given the relatively small sample available, we retrieved the residual series for correction term in the EC model from the cointegration regression proposed by HansenDlog Mt 5 u0 1 w2jt21 1 O s51 n1 d1sDlog Mt2s 1 O s51 n2 d2sDlog Pt2s 1 O s51 n3 d3sDlog Wt2s 1 O s51 n4 d4sDgt2s 1 O s51 n5 d5sAPt2s 1 O s51 n6 d6sDIPt2s 1 n2t. (6) In these equations n’s are the disturbance terms and the error-correction term j’s are the residuals obtained from the cointegrating regressions reported in Table 2.10 Since DP and DM appear to be cointegrated, the Granger Representation Theorem implies that there must be Granger causality between them at least in one direction (Engle and Granger, 1987). For example, the null hypothesis of noncausality from money stock to prices is rejected if either the group of coefficients on the money stock variable in the inflation equation, ¥s51 n2 d2s, are jointly significant or the error-correction coefficient, (w1), is significant. The optimal error-correction model is specified using Hendry’s (1984) general-to-specific modeling strategy. That is, a more general model was spec￾ified in which growth rates of prices and money stock were regressed on the full vector of exogenous and endogenous variables with a two-period lag. Through stepwise regression, insignificant variables were eliminated to yield a data congruent parsimonious specification. Once the variables and optimal lag lengths in each equation are identified, the price and money equations are pooled together to form a system of simultaneous equations. The system is estimated with the aid of Zellner’s Seemingly-Unrelated Regression (SUR) method to obtain more efficient estimates, recognizing the fact that the errors across the equations exhibit significant correlation. The results are reported in Table 4. It is evident that the error-correction terms in both the price and money equations are negative and statistically significant. In the next step, a series of Wald’s x2 tests were performed to ascertain the joint significance of each of the explanatory variables in the two-equation error correction model of Table 4. The results of Granger causality tests are reported in Table 5. The results suggest that lag on the money stock variable in the price equation is statistically significant. The x2 value of 15.28 corresponding to the null hypothesis H0: d2s 5 0, together with a highly significant error-correction coefficient, decisively rejects the null hypothesis of noncausality from money stock to prices. Judged by the values of asymptotic x2 statistics, the price 10 Given the relatively small sample available, we retrieved the residual series for the error correction term in the VEC model from the cointegration regression proposed by Hansen (1990). MONEY AND INFLATION IN CHINA 679