【正确答案】上题中的(1)已经给出了每个公司单个方程的OLS估计，这也就是固定效应变系数模型的OLS估计。也可以在EViews软件中一次估计。在EViews软件中，在出现的pool estimation对话框中，将“Y?”输入到“DependentVariable”栏内，将“X1?X2?”输入到“Cross section specific Coefficients”栏内，在“intercept”选项栏内选择“Fixed effects”，在“Weighting”选项栏内选择“No Weighting”，点击OK按钮后得表8-13。
   为了得到GLS估计，只需在“Weighting”选项栏内选择“Cross section Weights”即可。GLS估计结果如表8-14所示。
   

表8-13 Dependent Variable：Y？ Method：Pooled Least Squares Variable Coefficient Std.Error t-Statistic Prob. _GE--X1_GE 0.026551 0.037881 0.700903 0.4858 _GM--X1_GM 0.119210 0.019083 6.246796 0.0000 _US--X1_US 0.171430 0.052390 3.272221 0.0017 _WEST--X1_WEST 0.053055 0.104485 0.507779 0.6133 _GE--X2_GE 0.151694 0.062553 2.425046 0.0180 _GM--X2_GM 0.371525 0.027401 13.55858 0.0000 _US--X2_US 0.408709 0.102966 3.969367 0.0002 _WEST--X2_WEST 0.091694 0.373394 0.245568 0.8068 Fixed Effects _GE--C 9.956306 _GM--C -149.4667 _US--C 50.07804 _WEST--C 0.580403 R-squared 0.951157 Mean dependent var 290.9154 Adjusted R-squared 0.943256 S.D. dependent var 284.8528 S.E. of regression 67.85489 Sum squared resid 313091.5 Log likelihood -444.4040 F-statistic 189.1733 Durbin-Watson stat 0.974483 Prob(F-statistic) 0.000000

   

表8-14 Dependent Variable：Y？ Method：Pooled Least Squares Variable Coefficient Std.Error t-Statistic Prob. _GE--X1_GE 0.026551 0.015566 1.705705 0.0926 _GM--X1_GM 0.119210 0.025804 4.619724 0.0000 _US--X1_US 0.171430 0.073721 2.325409 0.0230 _WEST--X1_WEST 0.053055 0.015708 3.377633 0.0012 _GE--X2_GE 0.151694 0.025704 5.901548 0.0000 _GM--X2_GM 0.371525 0.037052 10.02704 0.0000 _US--X2_US 0.408709 0.144889 2.820836 0.0063 _WEST--X2_WEST 0.091694 0.056135 1.633465 0.1070 Fixed Effects _GE--C -9.956306 _GM--C -149.4667 _US--C -50.07804 _WEST--C -0.580403 R-squared 0.857460 Mean dependent var 318.8954 Adjusted R-squared 0.834402 S.D. dependent var 166.7454 S.E. of regression 67.85489 Sum squared resid 3130291.5 Log likelihood -401.5852 F-statistic 58.43710 Durbin-Watson stat 1.127765 Prob(F-statistic) 0.000000 Unweighted Statistics R-squared 0.951157 Mean dependent var 290.9154 Adjusted R-squared 0.943256 S.D. dependent var 284.8528 S.E. of regression 67.85489 Sum squared resid 313091.5 Durbin-Watson stat 0.974483

   可以看出，变系数固定效应模型的OLS估计与GLS估计的参数是相同的，但检验指标不相同。GLS估计显然更有效，如有着更高的t检验值，有着更高的R²等。