1、第七章分布滞后模型与自回归模型案例分析一、问题的提出和模型设定货币主义学派认为,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。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对这一问题进行研究。 广义货币M2广义货币增长量 M2z广义货币M2广义货币增长量 M2z月度(千亿元) (千亿元)居民消费价格同比指数tbzs 月度 (千亿元)(千亿元)居民消费价格同比指数tbzs
2、Jan-96 58.401 Oct-00 129.522 -0.9518 100Feb-96 63.778 5.377 109.3 Nov-00 130.9941 1.4721 101.3Mar-96 64.511 0.733 109.8 Dec-00 134.6103 3.6162 101.5Apr-96 65.723 1.212 109.7 1-Jan 137.5436 2.9333 101.2May-96 66.88 1.157 108.9 1-Feb 136.2102 -1.3334 100Jun-96 68.132 1.252 108.6 1-Mar 138.7445 2.5343
3、100.8Jul-96 69.346 1.214 108.3 1-Apr 139.9499 1.2054 101.6Aug-96 72.309 2.963 108.1 1-May 139.0158 -0.9341 101.7Sep-96 69.643 -2.666 107.4 1-Jun 147.8097 8.7939 101.4Oct-96 73.1522 3.5092 107 1-Jul 149.2287 1.419 101.5Nov-96 74.142 0.9898 106.9 1-Aug 149.9418 0.7131 101Dec-96 76.0949 1.9529 107 1-Se
4、p 151.8226 1.8808 99.9Jan-97 78.648 2.5531 105.9 1-Oct 151.4973 -0.3253 100.2Feb-97 78.998 0.35 105.6 1-Nov 154.0883 2.591 99.7Mar-97 79.889 0.891 104 1-Dec 158.3019 4.2136 99.7Apr-97 80.818 0.929 103.2 2-Jan 159.6393 1.3374 99May-97 81.151 0.333 102.8 2-Feb 160.9356 1.2963 100Jun-97 82.789 1.638 10
5、2.8 2-Mar 164.0646 3.129 99.2Jul-97 83.46 0.671 102.7 2-Apr 164.5706 0.506 98.7Aug-97 84.746 1.286 101.9 2-May 166.061 1.4904 98.9Sep-97 85.892 1.146 101.8 2-Jun 169.6012 3.5402 99.2Oct-97 86.644 0.752 101.5 2-Jul 170.8511 1.2499 99.1Nov-97 87.59 0.946 101.1 2-Aug 173.2509 2.3998 99.3Dec-97 90.9953
6、3.4053 100.4 2-Sep 176.9824 3.7315 99.3Jan-98 92.2114 1.2161 100.3 2-Oct 177.2942 0.3118 99.2Feb-98 92.024 -0.1874 99.9 2-Nov 179.7363 2.4421 99.3Mar-98 92.015 -0.009 100.7 2-Dec 185.0073 5.271 99.6Apr-98 92.662 0.647 99.7 3-Jan 190.4883 5.481 100.4May-98 93.936 1.274 99 3-Feb 190.1084 -0.3799 100.2
7、Jun-98 94.658 0.722 98.7 3-Mar 194.4873 4.3789 100.9Jul-98 96.314 1.656 98.6 3-Apr 196.1301 1.6428 101Aug-98 97.299 0.985 98.6 3-May 199.5052 3.3751 100.7Sep-98 99.795 2.496 98.5 3-Jun 204.9314 5.4262 100.3Oct-98 100.8752 1.0802 98.9 3-Jul 206.1931 1.2617 100.5Nov-98 102.229 1.3538 98.8 3-Aug 210.59
8、19 4.3988 100.9Dec-98 104.4985 2.2695 99 3-Sep 213.5671 2.9752 101.1Jan-99 105.5 1.0015 98.8 3-Oct 214.4694 0.9023 101.8Feb-99 107.778 2.278 98.7 3-Nov 216.3517 1.8823 103Mar-99 108.438 0.66 98.2 3-Dec 221.2228 4.8711 103.2Apr-99 109.218 0.78 97.8 4-Jan 225.10193 3.87913 103.2May-99 110.061 0.843 97
9、.8 4-Feb 227.05072 1.94879 102.1Jun-99 111.363 1.302 97.9 4-Mar 231.6546 4.60388 103Jul-99 111.414 0.051 98.6 4-Apr 233.62786 1.97326 103.8Aug-99 112.827 1.413 98.7 4-May 234.8424 1.21454 104.4Sep-99 115.079 2.252 99.2 4-Jun 238.42749 3.58509 105Oct-99 115.39 0.311 99.4 4-Jul 234.8424 -3.58509 105.3
10、Nov-99 116.559 1.169 99.1 4-Aug 239.72919 4.88679 105.3Dec-99 119.898 3.339 99 4-Sep 243.757 4.02781 105.2Jan-00 121.22 1.322 99.8 4-Oct 243.74 -0.017 104.3Feb-00 121.5834 0.3634 100.7 4-Nov 247.13558 3.39558 102.8Mar-00 122.5807 0.9973 99.8 4-Dec 253.2077 6.07212 102.4Apr-00 124.1219 1.5412 99.7 5-
11、Jan 257.75283 4.54513 101.9May-00 124.0533 -0.0686 100.1 5-Feb 259.3561 1.60327 103.9Jun-00 126.6053 2.552 100.5 5-Mar 264.5889 5.2328 102.7Jul-00 126.3239 -0.2814 100.5 5-Apr 266.99266 2.40376 101.8Aug-00 127.79 1.4661 100.3 5-May 269.2294 2.23674 101.8Sep-00 130.4738 2.6838 100 Eviews 上机具体操作:利用 ev
12、iews3.0 进行分析第一步:建立数据1 新建工作文档:file-new-workfile,在打开的 workfile range对话框中的 workfile frequency 中选择 monthly,start date 输入 1996-1,end date 输入 2005-5,点击 ok。2 输入数据(先是 data y x2 x3然后是将 excel 中的数据复制过来即可)并保存本题在命令窗口输入 data TBZS M2Z,并点击 name 命名为 GROUP01.然后将上面的数据录入。第二步 分析数据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 的月增长量用广义货币
13、 作为解释变量,以居民消费价格月度同比指ZM2数 TBZS 为被解释变量进行研究。首先估计如下回归模型: tZuTBZSt20t在命令窗口输入 ls TBZS C M2Z ,并点击 name 命名为 EQ01.得到如下回归Dependent Variable: TBZSMethod: Least SquaresDate: 12/20/12 Time: 10:23Sample(adjusted): 1996:02 2005:05Included observations: 112 after adjusting endpointsVariable Coefficient Std. Error t
14、-Statistic Prob. C 101.4356 0.397419 255.2358 0.0000M2Z 0.068371 0.151872 0.450190 0.6535R-squared 0.001839 Mean dependent var 101.5643Adjusted R-squared -0.007235 S.D. dependent var 2.911111S.E. of regression 2.921623 Akaike info criterion 4.999852Sum squared resid 938.9472 Schwarz criterion 5.0483
15、96Log likelihood -277.9917 F-statistic 0.202671Durbin-Watson stat 0.047702 Prob(F-statistic) 0.653460从回归结果来看, 的 t 统计量值不显著,表明当期货币供应量ZM2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.在命令窗口输入 ls TBZS C M2Z M2Z(-1) M2Z(-2) M2Z(-3) M2Z(-4) M2Z(-5) M2Z(-6),并点击 name 命名为 EQ02.Dependent Vari
16、able: TBZSMethod: Least SquaresDate: 12/20/12 Time: 10:27Sample(adjusted): 1996:08 2005:05Included observations: 106 after adjusting endpointsVariable Coefficient Std. Error t-Statistic Prob. C 100.0492 0.584318 171.2240 0.0000M2Z -0.011037 0.140613 -0.078493 0.9376M2Z(-1) 0.016169 0.137998 0.117166
17、 0.9070M2Z(-2) 0.053044 0.136808 0.387723 0.6991M2Z(-3) 0.028679 0.143155 0.200333 0.8416M2Z(-4) 0.130825 0.139183 0.939951 0.3496M2Z(-5) 0.137794 0.142502 0.966965 0.3359M2Z(-6) 0.248778 0.143394 1.734924 0.0859R-squared 0.055557 Mean dependent var 101.1377Adjusted R-squared -0.011904 S.D. dependen
18、t var 2.347946S.E. of regression 2.361879 Akaike info criterion 4.629264Sum squared resid 546.6902 Schwarz criterion 4.830278Log likelihood -237.3510 F-statistic 0.823546Durbin-Watson stat 0.094549 Prob(F-statistic) 0.570083从回归结果来看, 各滞后期的系数逐步增加,表明当期货币供ZM2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
19、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.在命令窗口输入 ls TBZS C M2Z M2Z(-1) M2Z(-2) M2Z(-3) M2Z(-4) M2Z(-5) M2Z(-6) M2Z(-7) M2Z(-8) M2Z(-9) M2Z(-10) M2Z(-11) M2Z(-12),并点击 name 命名为 EQ03.Dependent Variable: TBZSMethod: Least SquaresDate: 12/20/12 Time: 10:30Sample(adjusted): 1997:02 2005:05Included obs
20、ervations: 100 after adjusting endpointsVariable Coefficient Std. Error t-Statistic Prob. C 98.35668 0.467897 210.2102 0.0000M2Z -0.167665 0.121743 -1.377203 0.1720M2Z(-1) -0.032065 0.111691 -0.287084 0.7747M2Z(-2) -0.000995 0.111464 -0.008925 0.9929M2Z(-3) 0.004243 0.113815 0.037276 0.9704M2Z(-4) 0
21、.106581 0.112727 0.945480 0.3471M2Z(-5) 0.043217 0.113161 0.381908 0.7035M2Z(-6) 0.117581 0.118460 0.992575 0.3237M2Z(-7) 0.140418 0.115571 1.214988 0.2277M2Z(-8) 0.220875 0.114368 1.931271 0.0567M2Z(-9) 0.140875 0.115354 1.221247 0.2253M2Z(-10) 0.180497 0.115895 1.557410 0.1230M2Z(-11) 0.246911 0.1
22、25543 1.966752 0.0524M2Z(-12) 0.392359 0.130058 3.016798 0.0034R-squared 0.317136 Mean dependent var 100.7830Adjusted R-squared 0.213913 S.D. dependent var 1.890863S.E. of regression 1.676469 Akaike info criterion 4.000434Sum squared resid 241.7072 Schwarz criterion 4.365158Log likelihood -186.0217
23、F-statistic 3.072325Durbin-Watson stat 0.265335 Prob(F-statistic) 0.000906上表显示,从 M2Z 到 M2Z(-11), 回归系数都不显著异于零(P 值均大于0.05),而 M2Z(-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。在命令窗口输入 ls TBZS C M2Z M2Z(-1) M2Z(-
24、2) M2Z(-3) M2Z(-4) M2Z(-5) M2Z(-6) M2Z(-7) M2Z(-8) M2Z(-9) M2Z(-10) M2Z(-11) M2Z(-12) M2Z(-13) M2Z(-14) M2Z(-15) M2Z(-16) M2Z(-17) M2Z(-18),并点击 name 命名为 EQ04.Dependent Variable: TBZSMethod: Least SquaresDate: 12/20/12 Time: 10:33Sample(adjusted): 1997:08 2005:05Included observations: 94 after adjust
25、ing endpointsVariable Coefficient Std. Error t-Statistic Prob. C 97.41411 0.370000 263.2815 0.0000M2Z -0.083649 0.094529 -0.884900 0.3791M2Z(-1) -0.116744 0.093984 -1.242161 0.2181M2Z(-2) -0.119939 0.094428 -1.270156 0.2080M2Z(-3) -0.092993 0.095720 -0.971509 0.3345M2Z(-4) -0.032912 0.095823 -0.3434
26、68 0.7322M2Z(-5) -0.023891 0.097813 -0.244256 0.8077M2Z(-6) 0.017290 0.100645 0.171794 0.8641M2Z(-7) 0.028288 0.097570 0.289929 0.7727M2Z(-8) 0.048708 0.095877 0.508021 0.6129M2Z(-9) 0.025995 0.097569 0.266422 0.7907M2Z(-10) 0.118247 0.096764 1.222011 0.2256M2Z(-11) 0.157408 0.102558 1.534815 0.1291
27、M2Z(-12) 0.271281 0.112316 2.415326 0.0182M2Z(-13) 0.325760 0.109217 2.982684 0.0039M2Z(-14) 0.396242 0.107046 3.701601 0.0004M2Z(-15) 0.335482 0.106776 3.141941 0.0024M2Z(-16) 0.270811 0.107222 2.525697 0.0137M2Z(-17) 0.200024 0.109278 1.830415 0.0712M2Z(-18) 0.169696 0.101547 1.671114 0.0989R-squa
28、red 0.610520 Mean dependent var 100.6085Adjusted R-squared 0.510519 S.D. dependent var 1.795733S.E. of regression 1.256348 Akaike info criterion 3.480597Sum squared resid 116.8024 Schwarz criterion 4.021724Log likelihood -143.5881 F-statistic 6.105105Durbin-Watson stat 0.308938 Prob(F-statistic) 0.0
29、00000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2R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
30、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自回归模型来代替,因此我们估计如下自回归模型: t1-tt uTBZSTZS在命令窗口输入 ls TBZS C TBZS(-1)得到回归结果Dependent Variable: TBZSMethod: Least SquaresDate: 12/20/12 Time: 10:43Sample(adjusted): 1996:03 2005:05Included observations: 111 a
31、fter adjusting endpointsVariable Coefficient Std. Error t-Statistic Prob. C 5.348792 1.938684 2.758982 0.0068TBZS(-1) 0.946670 0.019081 49.61371 0.0000R-squared 0.957596 Mean dependent var 101.4946Adjusted R-squared 0.957207 S.D. dependent var 2.828904S.E. of regression 0.585200 Akaike info criterion 1.784126Sum squared resid 37.32798 Schwarz criterion 1.832947Log likelihood -97.01900 F-statistic 2461.520Durbin-Watson stat 1.779257 Prob(F-statistic) 0.000000