【正确答案】由于lnGDP与lnCONS都是一阶单整的，做lnCONS关于lnGDP的OLS回归，得表9-19所示的结果。
   

表9-19 Dependent Variable：LOG(CONS) Included observations：23 Variable Coefficient Std.Error t-Statistic Prob. C -0.364518 0.065934 -5.528577 0.0000 MA(5) 0.964861 0.006697 144.0714 0.0000 R-squared 0.998989 Mean dependent var 9.077676 Adjusted R-squared 0.998941 S.D. dependent var 1.062404 S.E. of regression 0.034570 Akaike info criterion -3.808700 Sum squared resid 0.025097 Schwarz criterion -3.709961 Log likelihood 45.80004 F-statistic 20756.58 Durbin-Watson stat 0.529594 Prob(F-statistic) 0.000000

   记该回归的残差项为e_t，对它进行ADF检验，经尝试，一个不包括截距项、趋势项与差分滞后项的检验模型如表9-20所示。
   

表9-20 ADF Test Statistic -2.371247 1/% Critical Value* -2.6756 5/% Critical Value -1.9574 10/% Critical Value -1.6238 *MacKinnon critical values for rejection of hypothesis of a unit root. Augmented Dickey-Fuller Test Equation Dependent Variable：D(E) Included observations：22 after adjusting endpoints Variable Coefficient Std.Error t-Statistic Prob. E(-1) -0.342555 0.144462 -2.371247 0.0274 R-squared 0.183479 Mean dependent var 0.004529 Adjusted R-squared 0.183479 S.D. dependent var 0.024727 S.E. of regression 0.022344 Akaike info criterion -4.720143 Sum squared resid 0.010484 Schwarz criterion -4.670550 Log likelihood 52.92157 Durbin-Watson stat 1.464892

   可见，在5/%的显著性水平下，拒绝存在单位根的假设，表明残差序列e_t是平稳的。由此知，lnGDP与lnCONS存在(1，1)阶协整关系。

【正确答案】在(1)的基础上，将残差序列e_t作为误差修正项，可建立如表9-21所示的误差修正模型。
   

表9-21 Dependent Variable：D(LOG(CONS)) Included observations：22 after adjusting endpoints Variable Coefficient Std.Error t-Statistic Prob. D(LOG(GDP)) 0.964585 0.030821 31.29639 0.0000 E(-1) -0.342796 0.150448 -2.278505 0.0338 R-squared 0.888026 Mean dependent var 0.145198 Adjusted R-squared 0.882428 S.D. dependent var 0.066773 S.E. of regression 0.022896 Akaike info criterion -4.629238 Sum squared resid 0.010484 Schwarz criterion -4.530052 Log likelihood 52.92162 Durbin-Watson stat 1.464548

   尽管D.W.值为1.46，但拉格朗日乘数检验结果如表9-22所示。
   

表9-22 F-statistic 2.142235 Probability 0.159641 Obs*R-squared 1.529664 Probability 0.216163

   可见模型已不存在序列相关性。因此，最终的误差修正模型为
   △lnCONS_t=0.9645△lnGDP_t-0.3428(lnCONS_t-1+0.3645-0.9649lnGDP_t-1)