第九章案例分析分布滞后模型

《第九章案例分析分布滞后模型》由会员分享，可在线阅读，更多相关《第九章案例分析分布滞后模型（9页珍藏版）》请在装配图网上搜索。

1、精品文档，仅供学习与交流，如有侵权请联系网站删除第九章 案例分析【案例7.1】 为了研究19551974年期间美国制造业库存量Y和销售额X的关系，用阿尔蒙法估计如下有限分布滞后模型：将系数(i=0,1,2,3)用二次多项式近似，即则原模型可变为其中在Eviews工作文件中输入X和Y的数据，在工作文件窗口中点击“Genr”工具栏，出现对话框，输入生成变量Z0t的公式，点击“OK”；类似，可生成Z1t、Z2t变量的数据。进入Equation Specification 对话栏，键入回归方程形式Y C Z0 Z1 Z2点击“OK”，显示回归结果（见表7.2）。表7.2表中Z0、 Z1、Z2对应的系数

2、分别为的估计值。将它们代入分布滞后系数的阿尔蒙多项式中，可计算出的估计值为：从而，分布滞后模型的最终估计式为：在实际应用中，Eviews提供了多项式分布滞后指令“PDL”用于估计分布滞后模型。下面结合本例给出操作过程：在Eviews中输入X和Y的数据，进入Equation Specification 对话栏，键入方程形式Y C PDL(X, 3, 2)其中，“PDL指令”表示进行多项式分布滞后（Polynomial Distributed Lags）模型的估计，括号中的3表示X的分布滞后长度，2表示多项式的阶数。在Estimation Settings栏中选择Least Squares(最小二

3、乘法)，点击OK，屏幕将显示回归分析结果（见表7.3）。 表7.3需要指出的是，用“PDL”估计分布滞后模型时，Eviews所采用的滞后系数多项式变换不是形如（7.4）式的阿尔蒙多项式，而是阿尔蒙多项式的派生形式。因此，输出结果中PDL01、PDL02、PDL03对应的估计系数不是阿尔蒙多项式系数的估计。但同前面分步计算的结果相比，最终的分布滞后估计系数式是相同的。【案例7.2】 货币主义学派认为，产生通货膨胀的必要条件是货币的超量供应。物价变动与货币供应量的变化有着较为密切的联系，但是二者之间的关系不是瞬时的，货币供应量的变化对物价的影响存在一定时滞。有研究表明，西方国家的通货膨胀时滞大约为

4、23个季度。在中国，大家普遍认同货币供给的变化对物价具有滞后影响，但滞后期究竟有多长，还存在不同的认识。下面采集19962005年全国广义货币供应量和物价指数的月度数据（见表7.4）对这一问题进行研究。表7.4 19962005年全国广义货币供应量及物价指数月度数据月度广义货币M2 (千亿元）广义货币增长量M2z（千亿元）居民消费价格同比指数tbzs 月度广义货币M2 (千亿元）广义货币增长量M2z（千亿元）居民消费价格同比指数tbzs Jan-9658.401Oct-00129.522-0.9518100Feb-9663.7785.377109.3Nov-00130.99411.472110

5、1.3Mar-9664.5110.733109.8Dec-00134.61033.6162101.5Apr-9665.7231.212109.7Jan-01137.54362.9333101.2May-9666.881.157108.9Feb-01136.2102-1.3334100Jun-9668.1321.252108.6Mar-01138.74452.5343100.8Jul-9669.3461.214108.3Apr-01139.94991.2054101.6Aug-9672.3092.963108.1May-01139.0158-0.9341101.7Sep-9669.643-2.6

6、66107.4Jun-01147.80978.7939101.4Oct-9673.15223.5092107Jul-01149.22871.419101.5Nov-9674.1420.9898106.9Aug-01149.94180.7131101Dec-9676.09491.9529107Sep-01151.82261.880899.9Jan-9778.6482.5531105.9Oct-01151.4973-0.3253100.2Feb-9778.9980.35105.6Nov-01154.08832.59199.7Mar-9779.8890.891104Dec-01158.30194.2

7、13699.7Apr-9780.8180.929103.2Jan-02159.63931.337499May-9781.1510.333102.8Feb-02160.93561.2963100Jun-9782.7891.638102.8Mar-02164.06463.12999.2Jul-9783.460.671102.7Apr-02164.57060.50698.7Aug-9784.7461.286101.9May-02166.0611.490498.9Sep-9785.8921.146101.8Jun-02169.60123.540299.2Oct-9786.6440.752101.5Ju

8、l-02170.85111.249999.1Nov-9787.590.946101.1Aug-02173.25092.399899.3Dec-9790.99533.4053100.4Sep-02176.98243.731599.3Jan-9892.21141.2161100.3Oct-02177.29420.311899.2Feb-9892.024-0.187499.9Nov-02179.73632.442199.3Mar-9892.015-0.009100.7Dec-02185.00735.27199.6Apr-9892.6620.64799.7Jan-03190.48835.481100.

9、4May-9893.9361.27499Feb-03190.1084-0.3799100.2Jun-9894.6580.72298.7Mar-03194.48734.3789100.9Jul-9896.3141.65698.6Apr-03196.13011.6428101Aug-9897.2990.98598.6May-03199.50523.3751100.7Sep-9899.7952.49698.5Jun-03204.93145.4262100.3Oct-98100.87521.080298.9Jul-03206.19311.2617100.5Nov-98102.2291.353898.8

10、Aug-03210.59194.3988100.9Dec-98104.49852.269599Sep-03213.56712.9752101.1Jan-99105.51.001598.8Oct-03214.46940.9023101.8Feb-99107.7782.27898.7Nov-03216.35171.8823103Mar-99108.4380.6698.2Dec-03221.22284.8711103.2Apr-99109.2180.7897.8Jan-04225.101933.87913103.2May-99110.0610.84397.8Feb-04227.050721.9487

11、9102.1Jun-99111.3631.30297.9Mar-04231.65464.60388103Jul-99111.4140.05198.6Apr-04233.627861.97326103.8Aug-99112.8271.41398.7May-04234.84241.21454104.4Sep-99115.0792.25299.2Jun-04238.427493.58509105Oct-99115.390.31199.4Jul-04234.8424-3.58509105.3Nov-99116.5591.16999.1Aug-04239.729194.88679105.3Dec-991

12、19.8983.33999Sep-04243.7574.02781105.2Jan-00121.221.32299.8Oct-04243.74-0.017104.3Feb-00121.58340.3634100.7Nov-04247.135583.39558102.8Mar-00122.58070.997399.8Dec-04253.20776.07212102.4Apr-00124.12191.541299.7Jan-05257.752834.54513101.9May-00124.0533-0.0686100.1Feb-05259.35611.60327103.9Jun-00126.605

13、32.552100.5Mar-05264.58895.2328102.7Jul-00126.3239-0.2814100.5Apr-05266.992662.40376101.8Aug-00127.791.4661100.3May-05269.22942.23674101.8Sep-00130.47382.6838100数据来源：中国经济统计数据库，为了考察货币供应量的变化对物价的影响，我们用广义货币M2的月增长量M2Z作为解释变量，以居民消费价格月度同比指数TBZS为被解释变量进行研究。首先估计如下回归模型得如下回归结果（表7.5）。表7.5Dependent Variable: TBZSM

14、ethod: Least SquaresDate: 07/03/05 Time: 17:10Sample(adjusted): 1996:02 2005:05Included observations: 112 after adjusting endpointsVariableCoefficientStd. Errort-StatisticProb. C101.43560.397419255.23580.0000M2Z0.0683710.1518720.4501900.6535R-squared0.001839 Mean dependent var101.5643Adjusted R-squa

15、red-0.007235 S.D. dependent var2.911111S.E. of regression2.921623 Akaike info criterion4.999852Sum squared resid938.9472 Schwarz criterion5.048396Log likelihood-277.9917 F-statistic0.202671Durbin-Watson stat0.047702 Prob(F-statistic)0.653460从回归结果来看，M2Z的t统计量值不显著，表明当期货币供应量的变化对当期物价水平的影响在统计意义上不明显。为了分析货币

16、供应量变化影响物价的滞后性，我们做滞后6个月的分布滞后模型的估计，在Eviews工作文档的方程设定窗口中，输入TBZS C M2Z M2Z(-1) M2Z(-2) M2Z(-3) M2Z(-4) M2Z(-5) M2Z(-6)结果见表7.6。表7.6Dependent Variable: TBZSMethod: Least SquaresDate: 07/03/05 Time: 17:09Sample(adjusted): 1996:08 2005:05Included observations: 106 after adjusting endpointsVariableCoefficient

17、Std. Errort-StatisticProb. C100.04920.584318171.22400.0000M2Z-0.0110370.140613-0.0784930.9376M2Z(-1)0.0161690.1379980.1171660.9070M2Z(-2)0.0530440.1368080.3877230.6991M2Z(-3)0.0286790.1431550.2003330.8416M2Z(-4)0.1308250.1391830.9399510.3496M2Z(-5)0.1377940.1425020.9669650.3359M2Z(-6)0.2487780.14339

18、41.7349240.0859R-squared0.055557 Mean dependent var101.1377Adjusted R-squared-0.011904 S.D. dependent var2.347946S.E. of regression2.361879 Akaike info criterion4.629264Sum squared resid546.6902 Schwarz criterion4.830278Log likelihood-237.3510 F-statistic0.823546Durbin-Watson stat0.094549 Prob(F-sta

19、tistic)0.570083从回归结果来看，M2Z各滞后期的系数逐步增加，表明当期货币供应量的变化对物价水平的影响要经过一段时间才能逐步显现。但各滞后期的系数的t统计量值不显著，因此还不能据此判断滞后期究竟有多长。为此，我们做滞后12个月的分布滞后模型的估计，结果见表7.7。表7.7Dependent Variable: TBZSMethod: Least SquaresDate: 07/03/05 Time: 17:09Sample(adjusted): 1997:02 2005:05Included observations: 100 after adjusting endpointsV

20、ariableCoefficientStd. Errort-StatisticProb. C98.356680.467897210.21020.0000M2Z-0.1676650.121743-1.3772030.1720M2Z(-1)-0.0320650.111691-0.2870840.7747M2Z(-2)-0.0009950.111464-0.0089250.9929M2Z(-3)0.0042430.1138150.0372760.9704M2Z(-4)0.1065810.1127270.9454800.3471M2Z(-5)0.0432170.1131610.3819080.7035

21、M2Z(-6)0.1175810.1184600.9925750.3237M2Z(-7)0.1404180.1155711.2149880.2277M2Z(-8)0.2208750.1143681.9312710.0567M2Z(-9)0.1408750.1153541.2212470.2253M2Z(-10)0.1804970.1158951.5574100.1230M2Z(-11)0.2469110.1255431.9667520.0524M2Z(-12)0.3923590.1300583.0167980.0034R-squared0.317136 Mean dependent var10

22、0.7830Adjusted R-squared0.213913 S.D. dependent var1.890863S.E. of regression1.676469 Akaike info criterion4.000434Sum squared resid241.7072 Schwarz criterion4.365158Log likelihood-186.0217 F-statistic3.072325Durbin-Watson stat0.265335 Prob(F-statistic)0.000906表7.7显示，从M2Z到M2Z(-11)，回归系数都不显著异于零，而M2Z（-

23、12）的回归系数t统计量值为3.016798，在5显著性水平下拒绝系数为零的原假设。这一结果表明，当期货币供应量变化对物价水平的影响在经过12个月（即一年）后明显地显现出来。为了考察货币供应量变化对物价水平影响的持续期，我们做滞后18个月的分布滞后模型的估计，结果见表7.8。表7.8Dependent Variable: TBZSMethod: Least SquaresDate: 07/03/05 Time: 17:08Sample(adjusted): 1997:08 2005:05Included observations: 94 after adjusting endpointsVar

24、iableCoefficientStd. Errort-StatisticProb. C97.414110.370000263.28150.0000M2Z-0.0836490.094529-0.8849000.3791M2Z(-1)-0.1167440.093984-1.2421610.2181M2Z(-2)-0.1199390.094428-1.2701560.2080M2Z(-3)-0.0929930.095720-0.9715090.3345M2Z(-4)-0.0329120.095823-0.3434680.7322M2Z(-5)-0.0238910.097813-0.2442560.

25、8077M2Z(-6)0.0172900.1006450.1717940.8641M2Z(-7)0.0282880.0975700.2899290.7727M2Z(-8)0.0487080.0958770.5080210.6129M2Z(-9)0.0259950.0975690.2664220.7907M2Z(-10)0.1182470.0967641.2220110.2256M2Z(-11)0.1574080.1025581.5348150.1291M2Z(-12)0.2712810.1123162.4153260.0182M2Z(-13)0.3257600.1092172.9826840.

26、0039M2Z(-14)0.3962420.1070463.7016010.0004M2Z(-15)0.3354820.1067763.1419410.0024M2Z(-16)0.2708110.1072222.5256970.0137M2Z(-17)0.2000240.1092781.8304150.0712M2Z(-18)0.1696960.1015471.6711140.0989R-squared0.610520 Mean dependent var100.6085Adjusted R-squared0.510519 S.D. dependent var1.795733S.E. of r

27、egression1.256348 Akaike info criterion3.480597Sum squared resid116.8024 Schwarz criterion4.021724Log likelihood-143.5881 F-statistic6.105105Durbin-Watson stat0.308938 Prob(F-statistic)0.000000结果表明，从滞后12个月开始t统计量值显著，一直到滞后16个月为止，从滞后第17个月开始t值变得不显著；再从回归系数来看，从滞后11个月开始，货币供应量变化对物价水平的影响明显增加，再滞后14个月时达到最大，然后逐

28、步下降。通过上述一系列分析，我们可以做出这样的判断：在我国，货币供应量变化对物价水平的影响具有明显的滞后性，滞后期大约为一年，而且滞后影响具有持续性，持续的长度大约为半年，其影响力度先递增然后递减，滞后结构为型。当然，从上述回归结果也可以看出，回归方程的不高，DW值也偏低，表明除了货币供应量外，还有其他因素影响物价变化；同时，过多的滞后变量也可能引起多重共线性问题。如果我们分析的重点是货币供应量变化对物价影响的滞后性，上述结果已能说明问题。如果要提高模型的预测精度，则可以考虑对模型进行改进。根据前面的分析可知，分布滞后模型可以用子回归模型来代替，因此我们估计如下子自回归模型：在Eviews工作

29、文档的方程设定窗口中，输入TBZS C TBZS(-1)估计结果见表7.9。表7.9Dependent Variable: TBZSMethod: Least SquaresDate: 07/10/05 Time: 23:48Sample(adjusted): 1996:03 2005:05Included observations: 111 after adjusting endpointsVariableCoefficientStd. Errort-StatisticProb. C5.3487921.9386842.7589820.0068TBZS(-1)0.9466700.0190814

30、9.613710.0000R-squared0.957596 Mean dependent var101.4946Adjusted R-squared0.957207 S.D. dependent var2.828904S.E. of regression0.585200 Akaike info criterion1.784126Sum squared resid37.32798 Schwarz criterion1.832947Log likelihood-97.01900 F-statistic2461.520Durbin-Watson stat1.779257 Prob(F-statistic)0.000000【精品文档】第 9 页