宏观经济模型多种估计方法的EVIEWS实现.docx

1、宏观经济模型多种估计方法的EVIEWS实现08统计 学号：0807294 吴扬一、 问题综述建立中国宏观经济模型。宏观经济模型，是指以整个国民经济系统为研究对象，从总量水平和经济结构方面来研究国民经济各变量之间的相互作用。它可用来评价宏观经济政策、分析宏观经济结构和国民经济的发展趋势。宏观经济模型的表达可以用单一方程进行表达，也可以用联立方程组表达。本作业建立如下宏观经济模型，完备的结构式模型为其中，包含3个内生变量，即国内生产总值Y，居民消费总额C和投资总额I；3个先决变量，即政府消费G，前期居民消费总额Ct-1和常数项。可以判断，消费方程是恰好识别的方程，投资方程是过度识别的，模型可以识别

2、。数据来自题目提供。导入EVIEWS二、 各种方法的EVIEWS实现1. 狭义的工具变量法估计消费方程选取消费方程中未包含的先决变量G作为内生解释变量Y的工具变量；在工作文件主窗口点击quick/estimate equation，选择估计方法TSLS，在equation specification对话框输入消费方程，在instrument list对话框输入工具变量.点击确定，得到：Dependent Variable: C01Method: Two-Stage Least SquaresDate: 06/02/11 Time: 14:08Sample (adjusted): 1979 20

3、09Included observations: 31 after adjustmentsInstrument list: C G C01(-1)VariableCoefficientStd. Errort-StatisticProb.C1290.053402.73533.2032290.0034Y0.1071330.0231504.6277390.0001C01(-1)0.7857560.07185910.934710.0000R-squared0.998513Mean dependent var34025.26Adjusted R-squared0.998407S.D. dependent

4、 var34218.49S.E. of regression1365.679Sum squared resid52222209F-statistic9402.761Durbin-Watson stat0.743434Prob(F-statistic)0.000000Second-Stage SSR53379247得到结构参数的工具变量法估计量：2. 间接最小二乘法估计消费方程消费方程中包含的内生变量的简化方程为参数关系体系为用普通最小二乘法估计第一个简化式：Dependent Variable: C01Method: Least SquaresDate: 06/02/11 Time: 14:4

5、6Sample (adjusted): 1979 2009Included observations: 31 after adjustmentsVariableCoefficientStd. Errort-StatisticProb.C1086.594386.55342.8109810.0089C01(-1)0.9545380.03625626.327720.0000G0.2655810.0580214.5773100.0001R-squared0.998480Mean dependent var34025.26Adjusted R-squared0.998372S.D. dependent

6、var34218.49S.E. of regression1380.725Akaike info criterion17.39037Sum squared resid53379247Schwarz criterion17.52914Log likelihood-266.5507Hannan-Quinn criter.17.43561F-statistic9198.948Durbin-Watson stat0.743999Prob(F-statistic)0.000000用普通最小二乘法估计第二个简化式：Dependent Variable: YMethod: Least SquaresDate

7、: 06/02/11 Time: 14:47Sample (adjusted): 1979 2009Included observations: 31 after adjustmentsVariableCoefficientStd. Errort-StatisticProb.C-1899.1342081.958-0.9121860.3695C01(-1)1.5754550.1952738.0679500.0000G2.4789920.3124997.9327940.0000R-squared0.994318Mean dependent var84244.67Adjusted R-squared

8、0.993912S.D. dependent var95306.59S.E. of regression7436.521Akaike info criterion20.75796Sum squared resid1.55E+09Schwarz criterion20.89673Log likelihood-318.7484Hannan-Quinn criter.20.80320F-statistic2449.755Durbin-Watson stat0.686339Prob(F-statistic)0.000000得到简化式参数估计量为：由参数体系计算得到结构参数间接最小二乘估计值为3. 二阶

9、段最小二乘法点击objects/new object，选择systemSystem: UNTITLEDEstimation Method: Two-Stage Least SquaresDate: 06/02/11 Time: 15:09Sample: 1979 2009Included observations: 31Total system (balanced) observations 62CoefficientStd. Errort-StatisticProb.C(1)1290.053402.73533.2032290.0022C(2)0.1071330.0231504.6277390

10、.0000C(3)0.7857560.07185910.934710.0000C(4)-2538.266948.1448-2.6770870.0097C(5)0.4413900.00753458.585760.0000Determinant residual covariance1.63E+13Equation: C01=C(1)+C(2)*Y+C(3)*C01(-1)Instruments: G C01(-1) CObservations: 31R-squared0.998513Mean dependent var34025.26Adjusted R-squared0.998407S.D.

11、dependent var34218.49S.E. of regression1365.679Sum squared resid52222209Durbin-Watson stat0.743434Equation: I=C(4)+C(5)*YInstruments: G C01(-1) CObservations: 31R-squared0.991774Mean dependent var34646.51Adjusted R-squared0.991491S.D. dependent var42513.37S.E. of regression3921.722Sum squared resid4

12、.46E+08Durbin-Watson stat0.538847消费方程的参数估计量为投资方程的参数估计量为4. 三阶段最小二乘法System: UNTITLEDEstimation Method: Three-Stage Least SquaresDate: 06/02/11 Time: 15:20Sample: 1979 2009Included observations: 31Total system (balanced) observations 62Linear estimation after one-step weighting matrixCoefficientStd. Er

13、rort-StatisticProb.C(1)1384.346361.67293.8276200.0003C(2)0.1165380.0181096.4351730.0000C(3)0.7563730.05603813.497460.0000C(4)-2538.266917.0495-2.7678610.0076C(5)0.4413900.00728760.572280.0000Determinant residual covariance1.55E+13Equation: C01=C(1)+C(2)*Y+C(3)*C01(-1)Instruments: G C01(-1) CObservat

14、ions: 31R-squared0.998459Mean dependent var34025.26Adjusted R-squared0.998349S.D. dependent var34218.49S.E. of regression1390.396Sum squared resid54129611Durbin-Watson stat0.672688Equation: I=C(4)+C(5)*YInstruments: G C01(-1) CObservations: 31R-squared0.991774Mean dependent var34646.51Adjusted R-squ

15、ared0.991491S.D. dependent var42513.37S.E. of regression3921.722Sum squared resid4.46E+08Durbin-Watson stat0.538847 消费方程的参数估计量为投资方程的参数估计量为5. GMM（广义矩估计）System: UNTITLEDEstimation Method: Generalized Method of MomentsDate: 06/02/11 Time: 15:27Sample: 1979 2009Included observations: 31Total system (bal

16、anced) observations 62Identity matrix estimation weights - 2SLS coefs with GMM standard errorsKernel: Bartlett, Bandwidth: Fixed (3), No prewhiteningCoefficientStd. Errort-StatisticProb.C(1)1290.053616.41172.0928440.0408C(2)0.1071330.0277223.8645370.0003C(3)0.7857560.0939578.3629010.0000C(4)-2538.26

17、61067.430-2.3779230.0208C(5)0.4413900.01342532.878450.0000Determinant residual covariance1.63E+13J-statistic1.21E+13Equation: C01=C(1)+C(2)*Y+C(3)*C01(-1)Instruments: G C01(-1) CObservations: 31R-squared0.998513Mean dependent var34025.26Adjusted R-squared0.998407S.D. dependent var34218.49S.E. of reg

18、ression1365.679Sum squared resid52222209Durbin-Watson stat0.743434Equation: I=C(4)+C(5)*YInstruments: G C01(-1) CObservations: 31R-squared0.991774Mean dependent var34646.51Adjusted R-squared0.991491S.D. dependent var42513.37S.E. of regression3921.722Sum squared resid4.46E+08Durbin-Watson stat0.53884

19、7消费方程的参数估计量为投资方程的参数估计量为三、 几种方法的分析比较由上述各种结果可以看出，狭义的工具变量法（IV）、间接最小二乘法(ILS)、二阶段最小二乘法(2SLS)与广义矩阵法(GMM)，都得到了相同的参数估计量。前三种方法都是适用于恰好识别的结构方程，只是使用不同的工具变量估计得到的。三阶段最小二乘法(3SLS)是一种系统估计方法，是二阶段最小二乘法(2SLS)的推广和发展，并且都是在各个阶段采用了普通最小二乘法(OLS)，非常类似。发现3SLS的估计标准误差小于2SLS的估计标准误差，体现了3SLS估计更为有效。四、 总结对我国1978-2009年部分宏观经济数据宏观经济模型，运用EVIEWS分别运用狭义的工具变量法、间接最小二乘法、二阶段最小二乘法、三阶段最小二乘法和广义矩阵法对模型进行了估计，取得了较好的结果，并略微对各个方法进行了比较。