计量经济学实验报告(共8页)

1、精选优质文档-倾情为你奉上计量经济学实验报告班级 金融 学号 1 姓名 一、 实验名称：城乡居民消费、政府消费对经济增长的的影响的实证分析二、 实验准备数据城乡居民消费、政府消费对经济增长的影响的实证分析年份人均国内生产总值城镇居民消费支出总计农村居民消费支出总计人均中央财政支出人均地方财政支出199223111671.7659.2199.89161219.48776199329982110.8769.65110.70648280.9926199440442851.31016.81146.38548336.937199550463537.571310.36164.74352398.636919

2、9658463919.51572.08175.77315472.77778199764204185.61617.15204.85173542.04294199867964331.61590.3250.52701614.98225199971594615.91577.4330.11066718.310462000785849981670.13435.51518817.92683200186225309.011741.1451.943551029.1365200293986029.921834.3527.173361189.6532003105426510.941943.3574.19115133

3、3.30112004123367182.12184.7607.292981584.20852005140537942.882555.4671.171491923.75952006161658696.552829.02760.102852315.08512007195249997.473223.85865.976432901.65620082269811242.93660.71007.08653693.6729三、 实验步骤基本数据处理由于要对城乡居民消费、政府消费对经济增长的的影响，所以对数据取对数设立回归模型如下：LnYt =C+1Ln X1t +2Ln X2t +3Ln X3t +4Ln

4、X4tt+ Ut其中Y表示人均国内生产总值；X1表示城镇居民消费支出总计；X2 表示农村居民消费支出总计；X3表示人均中央财政支出；X4表示人均地方财政支出；t=1992-2008。 从散点图中可以看出lnx1、lnx2、lnx4与lny线性关系较明显，lnx3与lny线性关系不明显多重共线性检验运用OLS方法估计模型的参数，利用计量经济计算机软件Eviews计算可得如下结果Dependent Variable: LNYMethod: Least SquaresDate: 12/14/10 Time: 22:38Sample: 1992 2008Included observations: 1

5、7VariableCoefficientStd. Errort-StatisticProb. C0.0.1.0.1863LNX10.0.4.0.0010LNX20.0.3.0.0059LNX3-0.0.-0.0.3650LNX40.0.7.0.0000R-squared0. Mean dependent var8.Adjusted R-squared0. S.D. dependent var0.S.E. of regression0. Akaike info criterion-5.Sum squared resid0. Schwarz criterion-4.Log likelihood49

6、.24635 F-statistic6288.743Durbin-Watson stat1. Prob(F-statistic)0.由于LnX3的参数估计未能通过t检验，而且符号的经济意义也不合理，故认为解释变量间存在多重共线性在Eviews中计算解释变量之间的简单相关系数，得如下结果从表中可以发现LnX1与LnX2存在高度相关性。用Eviews找出最简单的回归形式Dependent Variable: LNYMethod: Least SquaresDate: 12/16/10 Time: 02:25Sample: 1992 2008Included observations: 17Vari

7、ableCoefficientStd. Errort-StatisticProb. LNX11.0.59.425630.0000C-1.0.-7.0.0000R-squared0. Mean dependent var8.Adjusted R-squared0. S.D. dependent var0.S.E. of regression0. Akaike info criterion-3.Sum squared resid0. Schwarz criterion-3.Log likelihood30.69305 F-statistic3531.405Durbin-Watson stat0.

8、Prob(F-statistic)0.Dependent Variable: LNYMethod: Least SquaresDate: 12/16/10 Time: 02:30Sample: 1992 2008Included observations: 17VariableCoefficientStd. Errort-StatisticProb. LNX21.0.28.232840.0000C-1.0.-3.0.0072R-squared0. Mean dependent var8.Adjusted R-squared0. S.D. dependent var0.S.E. of regre

9、ssion0. Akaike info criterion-1.Sum squared resid0. Schwarz criterion-1.Log likelihood18.16341 F-statistic797.0931Durbin-Watson stat0. Prob(F-statistic)0. Dependent Variable: LNYMethod: Least SquaresDate: 12/16/10 Time: 02:31Sample: 1992 2008Included observations: 17VariableCoefficientStd. Errort-St

10、atisticProb. LNX30.0.15.792120.0000C4.0.13.681490.0000R-squared0. Mean dependent var8.Adjusted R-squared0. S.D. dependent var0.S.E. of regression0. Akaike info criterion-0.Sum squared resid0. Schwarz criterion-0.Log likelihood8. F-statistic249.3911Durbin-Watson stat0. Prob(F-statistic)0. Dependent V

11、ariable: LNYMethod: Least SquaresDate: 12/16/10 Time: 02:31Sample: 1992 2008Included observations: 17VariableCoefficientStd. Errort-StatisticProb. LNX40.0.25.673970.0000C3.0.19.988720.0000R-squared0. Mean dependent var8.Adjusted R-squared0. S.D. dependent var0.S.E. of regression0. Akaike info criter

12、ion-1.Sum squared resid0. Schwarz criterion-1.Log likelihood16.58107 F-statistic659.1528Durbin-Watson stat0. Prob(F-statistic)0.从以上表可以看出LnY受LnX1影响最大，其次LnX2，再次LnX3，受LnX4影响性对较小，因此选LnY=-1.+ 1.Ln X1为初始回归模型进行逐步回归。首先，在初始模型中引入LnX2得到如下表Dependent Variable: LNYMethod: Least SquaresDate: 12/16/10 Time: 02:40Sa

13、mple: 1992 2008Included observations: 17VariableCoefficientStd. Errort-StatisticProb. LNX11.0.7.0.0000LNX20.0.0.0.3437C-1.0.-7.0.0000R-squared0. Mean dependent var8.Adjusted R-squared0. S.D. dependent var0.S.E. of regression0. Akaike info criterion-3.Sum squared resid0. Schwarz criterion-3.Log likel

14、ihood31.25695 F-statistic1761.506Durbin-Watson stat0. Prob(F-statistic)0.尽管拟合优度提高，但是LnX2的参数未能通过t检验，所以去掉LnX2.第二，将LnX4引入初始模型，可得到下表Dependent Variable: LNYMethod: Least SquaresDate: 12/16/10 Time: 13:10Sample: 1992 2008Included observations: 17VariableCoefficientStd. Errort-StatisticProb. LNX10.0.12.136

15、930.0000LNX40.0.4.0.0011C0.0.0.0.9571R-squared0. Mean dependent var8.Adjusted R-squared0. S.D. dependent var0.S.E. of regression0. Akaike info criterion-4.Sum squared resid0. Schwarz criterion-3.Log likelihood37.35711 F-statistic3617.809Durbin-Watson stat0. Prob(F-statistic)0. 可以看出引入LnX4,模型拟合优度提高，且参

16、数符号合理，变量也通过t检验，所以LnX4保留。再引入LnX3可以得到下表VariableCoefficientStd. Errort-StatisticProb. LNX10.0.17.305760.0000LNX3-0.0.-3.0.0023LNX40.0.6.0.0000C-0.0.-0.0.6440R-squared0. Mean dependent var8.Adjusted R-squared0. S.D. dependent var0.S.E. of regression0. Akaike info criterion-4.Sum squared resid0. Schwarz

17、criterion-4.Log likelihood43.66927 F-statistic4711.074Durbin-Watson stat1. Prob(F-statistic)0.可以看出引入LnX3后拟合优度再次提高且参数都通过t检验，所以LnX3保留。因此，最终的经济增长函数应以Y=f(Lnx1,LnX4,LnX3)为最优，拟合结果如下：Lny= -0.+ 0.LnX1 -0.LnX3+ 0.LnX4 序列相关性检验 经过调整后的模型变成：Lny= -0.+ 0.LNX1 -0.LNX3+ 0.LNX4 所以，只针对此模型进行序列相关性的分析。进行拉格朗日乘数检验,首先是二阶的结果

18、如下:Breusch-Godfrey Serial Correlation LM Test:F-statistic1. Probability0.Obs*R-squared4. Probability0.其次进行三阶的检验。其结果如下:Breusch-Godfrey Serial Correlation LM Test:F-statistic1. Probability0.Obs*R-squared4. Probability0.通过两种检验的结果看，此模型不存在序列相关性问题，无需调整。异方差检验首先进行没有交叉项的怀特检验，结果如下：White Heteroskedasticity Test:F-statistic1. Probability0.Obs*R-squared7. Probability0.再进行有交叉项的怀特检验，结果如下：White Heteroskedasticity Test:F-statistic0.y0.Obs*R-squared7.y0.通过检验。我们可以知道其Obs