宏观经济模型多种估计方法的EVIEWS实现_动视

宏观经济模型多种估计方法的EVIEWS实现

2025-09-29 22:27:50 责编:小OO

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、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。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: C01
Method: Two-Stage Least Squares
Date: 06/02/11 Time: 14:08
Sample (adjusted): 1979 2009
Included observations: 31 after adjustments
Instrument list: C G C01(-1)


Variable	Coefficient	Std. Error	t-Statistic	Prob.


C	1290.053	402.7353	3.203229	0.0034
Y	0.107133	0.023150	4.627739	0.0001
C01(-1)	0.785756	0.071859	10.93471	0.0000


R-squared	0.998513	Mean dependent var		34025.26
Adjusted R-squared	0.998407	S.D. dependent var		34218.49
S.E. of regression	1365.679	Sum squared resid		52222209
F-statistic	9402.761	Durbin-Watson stat		0.743434
Prob(F-statistic)	0.000000	Second-Stage SSR		53379247

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: C01
Method: Least Squares
Date: 06/02/11 Time: 14:46
Sample (adjusted): 1979 2009
Included observations: 31 after adjustments


Variable	Coefficient	Std. Error	t-Statistic	Prob.


C	1086.594	386.5534	2.810981	0.00
C01(-1)	0.954538	0.036256	26.32772	0.0000
G	0.265581	0.058021	4.577310	0.0001


R-squared	0.998480	Mean dependent var		34025.26
Adjusted R-squared	0.998372	S.D. dependent var		34218.49
S.E. of regression	1380.725	Akaike info criterion		17.39037
Sum squared resid	53379247	Schwarz criterion		17.52914
Log likelihood	-266.5507	Hannan-Quinn criter.		17.43561
F-statistic	9198.948	Durbin-Watson stat		0.743999
Prob(F-statistic)	0.000000

用普通最小二乘法估计第二个简化式：

Dependent Variable: Y
Method: Least Squares
Date: 06/02/11 Time: 14:47
Sample (adjusted): 1979 2009
Included observations: 31 after adjustments


Variable	Coefficient	Std. Error	t-Statistic	Prob.


C	-19.134	2081.958	-0.912186	0.3695
C01(-1)	1.575455	0.195273	8.067950	0.0000
G	2.4792	0.312499	7.932794	0.0000


R-squared	0.994318	Mean dependent var		84244.67
Adjusted R-squared	0.993912	S.D. dependent var		95306.59
S.E. of regression	7436.521	Akaike info criterion		20.75796
Sum squared resid	1.55E+09	Schwarz criterion		20.673
Log likelihood	-318.7484	Hannan-Quinn criter.		20.80320
F-statistic	2449.755	Durbin-Watson stat		0.686339
Prob(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

3.二阶段最小二乘法

点击objects/new object，选择system

System: UNTITLED
Estimation Method: Two-Stage Least Squares
Date: 06/02/11 Time: 15:09
Sample: 1979 2009
Included observations: 31
Total system (balanced) observations 62


	Coefficient	Std. Error	t-Statistic	Prob.


C(1)	1290.053	402.7353	3.203229	0.0022
C(2)	0.107133	0.023150	4.627739	0.0000
C(3)	0.785756	0.071859	10.93471	0.0000
C(4)	-2538.266	948.1448	-2.677087	0.0097
C(5)	0.441390	0.007534	58.58576	0.0000


Determinant residual covariance		1.63E+13



Equation: C01=C(1)+C(2)Y+C(3)C01(-1)
Instruments: G C01(-1) C
Observations: 31
R-squared	0.998513	Mean dependent var		34025.26
Adjusted R-squared	0.998407	S.D. dependent var		34218.49
S.E. of regression	1365.679	Sum squared resid		52222209
Durbin-Watson stat	0.743434

Equation: I=C(4)+C(5)*Y
Instruments: G C01(-1) C
Observations: 31
R-squared	0.991774	Mean dependent var		346.51
Adjusted R-squared	0.991491	S.D. dependent var		42513.37
S.E. of regression	3921.722	Sum squared resid		4.46E+08
Durbin-Watson stat	0.538847

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: UNTITLED
Estimation Method: Three-Stage Least Squares
Date: 06/02/11 Time: 15:20
Sample: 1979 2009
Included observations: 31
Total system (balanced) observations 62
Linear estimation after one-step weighting matrix


	Coefficient	Std. Error	t-Statistic	Prob.


C(1)	1384.346	361.6729	3.827620	0.0003
C(2)	0.116538	0.018109	6.435173	0.0000
C(3)	0.756373	0.056038	13.49746	0.0000
C(4)	-2538.266	917.0495	-2.767861	0.0076
C(5)	0.441390	0.007287	60.57228	0.0000


Determinant residual covariance		1.55E+13



Equation: C01=C(1)+C(2)Y+C(3)C01(-1)
Instruments: G C01(-1) C
Observations: 31
R-squared	0.998459	Mean dependent var		34025.26
Adjusted R-squared	0.998349	S.D. dependent var		34218.49
S.E. of regression	1390.396	Sum squared resid		54129611
Durbin-Watson stat	0.672688

Equation: I=C(4)+C(5)*Y
Instruments: G C01(-1) C
Observations: 31
R-squared	0.991774	Mean dependent var		346.51
Adjusted R-squared	0.991491	S.D. dependent var		42513.37
S.E. of regression	3921.722	Sum squared resid		4.46E+08
Durbin-Watson stat	0.538847

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: UNTITLED
Estimation Method: Generalized Method of Moments
Date: 06/02/11 Time: 15:27
Sample: 1979 2009
Included observations: 31
Total system (balanced) observations 62
Identity matrix estimation weights - 2SLS coefs with GMM standard errors
Kernel: Bartlett, Bandwidth: Fixed (3), No prewhitening


	Coefficient	Std. Error	t-Statistic	Prob.


C(1)	1290.053	616.4117	2.092844	0.0408
C(2)	0.107133	0.027722	3.8537	0.0003
C(3)	0.785756	0.093957	8.362901	0.0000
C(4)	-2538.266	1067.430	-2.377923	0.0208
C(5)	0.441390	0.013425	32.87845	0.0000


Determinant residual covariance		1.63E+13
J-statistic		1.21E+13



Equation: C01=C(1)+C(2)Y+C(3)C01(-1)
Instruments: G C01(-1) C
Observations: 31
R-squared	0.998513	Mean dependent var		34025.26
Adjusted R-squared	0.998407	S.D. dependent var		34218.49
S.E. of regression	1365.679	Sum squared resid		52222209
Durbin-Watson stat	0.743434

Equation: I=C(4)+C(5)*Y
Instruments: G C01(-1) C
Observations: 31
R-squared	0.991774	Mean dependent var		346.51
Adjusted R-squared	0.991491	S.D. dependent var		42513.37
S.E. of regression	3921.722	Sum squared resid		4.46E+08
Durbin-Watson stat	0.538847

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。下载本文

显示全文