试探交通运输发展与国民经济的关系.doc

试探交通运输发展与国民经济的关系小组成员: 金融学院2001级袁新熠40104017 李泓良 40104031 王阳 40104028童运超40104005 罗卫平 40104029 陈东 40104003指导教师 周游日期 2004年5月摘要：本文主要通过对我国1991年到2002年交通运输业的发展状况与国民经济的发展之间的关系进行多因素实证分析。建立以国民经济指标为应变量，交通运输业的经济指标为自变量的多元线性回归模型，试图探索交通运输发展与国民经济的关系。首先，我们收集了相关的数据，利用EVIEWS软件对计量模型进行了参数估计，并建立了理论模型。然后，进行检验并对模型加以修正。最后，我们结合相关的理论对所得的分析结果作了经济意义的分析。一.问题的提出与猜测： “要想富，先修路”是我们大家都耳熟能详的一句话，改革开放以来，我国的交通运输业有了很大的发展，表现在运输线路长度上和客货运送量上都大幅度增长，与此同时，我国的经济发展也快速发展。二者的同步发展是否存在着某种联系？在此，我们猜测两者之间存在着一定的联系，根据“要想富，先修路”这一经验，我们猜测交通运输业对国民经济的发展具有先行作用，也即交通运输业对国民经济的发展具有促进作用。以下，我们将根据这一设想，收集相关数据，并估计和检验，希望能够找出二者之间是否存在关系，如若有，它们是什么样的关系？以及它们在多大程度上相关？二.数据的搜集： 在进行实证分析的过程中，所需要的数据，应是能够很好代表两者水平的指标。就国民经济而言，GDP应该是最合适的指标，因为我们探究的是经济总量的问题，我们选取了各年我国的GDP总量（虽然人均GDP也同样有用，但明显不及总量GDP）；对于我国交通运输发展状况的水平指标，可选择的余地较大，但我们发现它们之间存在着明显的相关性，为了尽量避免多重共线形和使模型更加简洁精确，我们选取了四个最具有代表性的指标，它们分别是全国全年客运总量（用X1表示），全国全年货运总量（X2），截止当年全国铁路总里程数（X3），截止当年全国公路总里程数（X4）。 本文中数据的起止时间是1991年到2002年，一共12年的数据。数据来源于中经专网和国家统计局网站。三.对模型的猜测： 我们假设以上四个变量和GDP之间存在以下的关系，待估计方程为： Y=m+aX1+bX2+cX3+dX4+u YGDP X1全国全年客运总量 X2 全国全年货运总量X3 全国铁路总里程数 X4 全国公路总里程 接下来我将利用样本数据对参数进行估计。四.数据:（单位：1万人，2万吨，3公里，4公里）年份指标X1X2X3X4Y199187090793628857800104110021662.51992891960100486058100105670026651.91993979430106995558600108350034560.51994110092411330465900011178004667019951241770118587159700115700057494.919961271387124636964900118580066850.519971329770122457266000122640073142.719981370410121245466400127850076967.219991388321126874667400135170080579.420001471849133555468700140270088254.920011526602137765070100169800095727.9200215995551458555720001732000103553.6五.模型的参数估计：利用EVIEWS软件，用OLS方法估计得：Dependent Variable: YMethod: Least SquaresDate: 05/07/04 Time: 14:30Sample: 1991 2002Included observations: 12VariableCoefficientStd. Errort-StatisticProb. X10.0799470.00703511.364680.0000X20.0148970.0102531.4530240.1895X4-0.0010450.003393-0.3081220.7670X31.1290280.2166425.2114970.0012C-124807.79976.147-12.510620.0000R-squared0.999179 Mean dependent var64343.00Adjusted R-squared0.998710 S.D. dependent var27118.34S.E. of regression973.9812 Akaike info criterion16.89500Sum squared resid6640476. Schwarz criterion17.09704Log likelihood-96.36999 F-statistic2130.105Durbin-Watson stat2.285535 Prob(F-statistic)0.000000可得模型：Y=0.07994666486*X1+0.01489726364*X2+1.129028112*X3-0.001045403313*X4- 124807.7471 在上面的ols的结果中我们可以看出，变量x2与x4的p值未获得通过，我们在接下来的过程中进行检验和修正。六.计量经济学检验及其修正 1.多重共线性检验 用EVIEWS软件，得相关系数矩阵表：X1X2X4X3X110.9776489310990.8836835472450.952768261281X20.97764893109910.9039279280610.934040427024X40.8836835472450.90392792806110.908697841494X30.9527682612810.9340404270240.9086978414941由上可以看出，整体上线形回归拟合较好，但x2，x4变量的参数的t检验的p值大于0.05，所以t值并不显著，而且x4的系数符号与经济意义不符。两种方法结合一起来看，解释变量确实存在多重共线性。 下面我们利用逐步回归法（变量剔除法）进行修正： (1)运用ols方法逐一求y对各个解释变量的回归.结合经济意义和统计检验选出拟合效果最好的一元线形回归方程. (a)对x1与y回归:Dependent Variable: YMethod: Least SquaresDate: 05/07/04 Time: 14:35Sample: 1991 2002Included observations: 12VariableCoefficientStd. Errort-StatisticProb. X10.1107930.00252243.934120.0000C-74544.213215.745-23.181010.0000R-squared0.994846 Mean dependent var64343.00Adjusted R-squared0.994331 S.D. dependent var27118.34S.E. of regression2041.904 Akaike info criterion18.23217Sum squared resid41693735 Schwarz criterion18.31298Log likelihood-107.3930 F-statistic1930.207Durbin-Watson stat1.906367 Prob(F-statistic)0.000000(b)对x2与y回归:Dependent Variable: YMethod: Least SquaresDate: 05/07/04 Time: 14:35Sample: 1991 2002Included observations: 12VariableCoefficientStd. Errort-StatisticProb. X20.1749220.01151815.187110.0000C-146349.013973.68-10.473190.0000R-squared0.958446 Mean dependent var64343.00Adjusted R-squared0.954290 S.D. dependent var27118.34S.E. of regression5797.865 Akaike info criterion20.31938Sum squared resid3.36E+08 Schwarz criterion20.40020Log likelihood-119.9163 F-statistic230.6482Durbin-Watson stat0.888799 Prob(F-statistic)0.000000(c)对x3与y回归:Dependent Variable: YMethod: Least SquaresDate: 05/07/04 Time: 14:36Sample: 1991 2002Included observations: 12VariableCoefficientStd. Errort-StatisticProb. X35.1100790.40722112.548650.0000C-263000.126162.96-10.052380.0000R-squared0.940287 Mean dependent var64343.00Adjusted R-squared0.934316 S.D. dependent var27118.34S.E. of regression6950.126 Akaike info criterion20.68192Sum squared resid4.83E+08 Schwarz criterion20.76274Log likelihood-122.0915 F-statistic157.4687Durbin-Watson stat1.038107 Prob(F-statistic)0.000000(d)对x4与y回归:Dependent Variable: YMethod: Least SquaresDate: 05/07/04 Time: 14:36Sample: 1991 2002Included observations: 12VariableCoefficientStd. Errort-StatisticProb. X40.1043460.0162546.4196870.0001C-68969.2521080.72-3.2716750.0084R-squared0.804735 Mean dependent var64343.00Adjusted R-squared0.785208 S.D. dependent var27118.34S.E. of regression12568.17 Akaike info criterion21.86673Sum squared resid1.58E+09 Schwarz criterion21.94755Log likelihood-129.2004 F-statistic41.21238Durbin-Watson stat0.498566 Prob(F-statistic)0.000076由以上可以得知拟合程度最好的方程是: Y = 0.110793012*X1 - 74544.21148 (43.93412) (-23.18101) R-squared=0.994846 S.E=2041.909 F=1930.207(2)逐步回归,将其余的解释变量逐一代入上式中,得如下几个模型:(a) 将x2代入:Dependent Variable: YMethod: Least SquaresDate: 05/07/04 Time: 14:42Sample: 1991 2002Included observations: 12VariableCoefficientStd. Errort-StatisticProb. X10.1012700.0122198.2876300.0000X20.0156680.0196550.7971340.4459C-81478.419295.350-8.7655020.0000R-squared0.995186 Mean dependent var64343.00Adjusted R-squared0.994116 S.D. dependent var27118.34S.E. of regression2080.176 Akaike info criterion18.33061Sum squared resid38944181 Schwarz criterion18.45184Log likelihood-106.9837 F-statistic930.2357Durbin-Watson stat1.615796 Prob(F-statistic)0.000000(b)将x3代入:Dependent Variable: YMethod: Least SquaresDate: 05/07/04 Time: 14:43Sample: 1991 2002Included observations: 12VariableCoefficientStd. Errort-StatisticProb. X10.0885620.00401522.059660.0000X31.1069820.1904635.8120640.0003C-117587.07567.214-15.539010.0000R-squared0.998916 Mean dependent var64343.00Adjusted R-squared0.998675 S.D. dependent var27118.34S.E. of regression987.2214 Akaike info criterion16.83998Sum squared resid8771456. Schwarz criterion16.96121Log likelihood-98.03990 F-statistic4145.611Durbin-Watson stat2.228882 Prob(F-statistic)0.000000(c)将x4代入:Dependent Variable: YMethod: Least SquaresDate: 05/07/04 Time: 14:45Sample: 1991 2002Included observations: 12VariableCoefficientStd. Errort-StatisticProb. X10.1037740.00502420.655740.0000X40.0083170.0052611.5809420.1483C-76371.763213.872-23.763160.0000R-squared0.995966 Mean dependent var64343.00Adjusted R-squared0.995070 S.D. dependent var27118.34S.E. of regression1904.137 Akaike info criterion18.15376Sum squared resid32631643 Schwarz criterion18.27499Log likelihood-105.9226 F-statistic1111.059Durbin-Watson stat2.072357 Prob(F-statistic)0.000000由以上可以得知拟合程度最好的方程是:Y = 0.08856170882*X1 + 1.106981758*X3 - 117587.0399 (22.05966) (5.812064) (-15.53901) R-squared=0.998916 S.E=987.2214 F=4145.611(3)再将x2,x4代入上式:(a)将x2代入:Dependent Variable: YMethod: Least SquaresDate: 05/07/04 Time: 14:53Sample: 1991 2002Included observations: 12VariableCoefficientStd. Errort-StatisticProb. X10.0805730.00634212.705150.0000X31.0958900.1771036.1878530.0003X20.0135100.0086741.5575160.1580C-123134.87881.530-15.623200.0000R-squared0.999168 Mean dependent var64343.00Adjusted R-squared0.998856 S.D. dependent var27118.34S.E. of regression917.2336 Akaike info criterion16.74180Sum squared resid6730539. Schwarz criterion16.90344Log likelihood-96.45082 F-statistic3202.405Durbin-Watson stat2.187638 Prob(F-statistic)0.000000(b)将x4代入:Dependent Variable: YMethod: Least SquaresDate: 05/07/04 Time: 14:54Sample: 1991 2002Included observations: 12VariableCoefficientStd. Errort-StatisticProb. X10.0883540.00427020.693000.0000X31.0702520.2271334.7119940.0015X40.0011200.0032530.3443850.7394C-116405.08675.275-13.418020.0000R-squared0.998932 Mean dependent var64343.00Adjusted R-squared0.998531 S.D. dependent var27118.34S.E. of regression1039.430 Akaike info criterion16.99193Sum squared resid8643317. Schwarz criterion17.15357Log likelihood-97.95160 F-statistic2493.118Durbin-Watson stat2.219465 Prob(F-statistic)0.000000由以上可知x2,x4对y的影响并不显著,故将其删去,得如下模型:Y = 0.08856170882*X1 + 1.106981758*X3 - 117587.0399 (22.05966) (5.812064) (-15.53901) R-squared=0.998916 S.E=987.2214 F=4145.6112.异方差检验(1) ARCH检验我们首先对模型进行ARCH检验,得结果如下:首先对模型滞后三期:ARCH Test:F-statistic0.376599 Probability0.774398Obs*R-squared1.658810 Probability0.646130Test Equation:Dependent Variable: RESID2Method: Least SquaresDate: 06/09/04 Time: 08:57Sample(adjusted): 1994 2002Included observations: 9 after adjusting endpointsVariableCoefficientStd. Errort-StatisticProb. C2.23E+081.53E+081.4626810.2034RESID2(-1)0.2334990.3637610.6419020.5492RESID2(-2)-0.2335370.297622-0.7846750.4682RESID2(-3)-0.0031150.259411-0.0120070.9909R-squared0.184312 Mean dependent var2.04E+08Adjusted R-squared-0.305100 S.D. dependent var2.72E+08S.E. of regression3.11E+08 Akaike info criterion42.24730Sum squared resid4.82E+17 Schwarz criterion42.33495Log likelihood-186.1128 F-statistic0.376599Durbin-Watson stat0.623651 Prob(F-statistic)0.774398再对模型滞后一期:ARCH Test:F-statistic1.133281 Probability0.314803Obs*R-squared1.230212 Probability0.267366Test Equation:Dependent Variable: RESID2Method: Least SquaresDate: 06/09/04 Time: 09:01Sample(adjusted): 1992 2002Included observations: 11 after adjusting endpointsVariableCoefficientStd. Errort-StatisticProb. C2.93E+081.88E+081.5589780.1534RESID2(-1)0.3355980.3152471.0645570.3148R-squared0.111837 Mean dependent var4.08E+08Adjusted R-squared0.013153 S.D. dependent var5.16E+08S.E. of regression5.12E+08 Akaike info criterion43.10906Sum squared resid2.36E+18 Schwarz criterion43.18140Log likelihood-235.0998 F-statistic1.133281Durbin-Watson stat1.004647 Prob(F-statistic)0.314803由以上可知,由于F和obs*R-squared的p值都大于0.05,所以其不显著,不能拒绝原假设,所以模型不存在异方差.(2)white检验：我们运用white检验对模型进行异方差的检验,得到如下结果:White Heteroskedasticity Test:F-statistic1.419397 Probability0.337822Obs*R-squared6.502552 Probability0.260340Test Equation:Dependent Variable: RESID2Method: Least SquaresDate: 05/07/04 Time: 15:01Sample: 1991 2002Included observations: 12VariableCoefficientStd. Errort-StatisticProb. C-677835114.18E+08-0.1620610.8766X1-67.63222320.5373-0.2109960.8399X121.03E-055.57E-050.1841590.8600X1*X30.0007550.0070950.1064720.9187X33719.79919588.760.1898950.8557X32-0.0395290.225308-0.1754440.8665R-s