计量经济学eviews分析

1、数据处理与统计分析期末考核关于税收收入影响因素实证分析学院：商学院 专业：技术经济及管理 学号： 姓名： 时间： 数据处理与统计分析期末考核摘要:一国的经济增长是以其财政收入的增长为前提的，而财政收入的增长又离不开税收。本文采用我国自1978年至2012年的税收收入的主要因素的相关统计数据进行的实证分析。选取的自变量有国内生产总值、财政支出和商品零售价格指数。然后，在收集了相关数据之后，通过建立多元线性回归模型，利用EVIEWS软件对模型进行了参数估计和检验，并加以修正。最后得出国内生产总值对财政支出和商品零售价格指数对税收收入的影响模型。关键字：税收收入 影响因素 实证分析- 26 -Abs

2、tract：A countrys economic growth is the premise of its fiscal revenue growth, and revenue growth without tax. This article USES the our country tax revenue from 1978 to 1978 of the main factors related to the empirical analysis of statistical data. Selection of the independent variable has a gross d

3、omestic product (GDP), fiscal expenditure and commodity retail price index. And then, after collecting the related data, through the establishment of multiple linear regression model, use EVIEWS software to parameter estimation and model test, and modified. Finally it is concluded that gross domesti

4、c product (GDP) of fiscal expenditure and the commodity retail price index effect on the tax revenue model.Keyboard：Tax revenue Factors affecting empirical analysis目录摘要:2Abstract3引言5一、模型设定5(一)变量设定5（二）数据取得6（三）模型的建立与构造7二、经济意义检验9三、统计意义检验9（一）拟全优度检验9（二）t检验9（三）F检验9（四）结构稳定性检验10四、多重共线性11（一）多重共线性诊断11（二）多重共线性

5、补救13五、自相关20（一）自相关诊断20（二）自相关补救21六、异方差22（一）异方差诊断22七、模型应用分析与政策建议24（一）模型应用分析24（二）政策建议24参考文献26引言税收是我国财政收入的基本因素，也影响着我国经济的发展。经济是税收的源泉，经济决定税收，而税收又反作用于经济，这是税收与经济的一般原理。这几年来，中国税收收入的快速增长甚至“超速增长”引起了人们的广泛关注。科学地对税收增长进行因素分析和预测分析非常重要，对研究我国税收增长规律，制定经济政策有着重要意义。影响税收收入的因素有很多，但据分析主要的因素可能有：从宏观经济看，经济整体增长是税收增长的基本源泉，而国内生产总值是

6、反映经济增长的一个重要指标。公共财政的需求，税收收入是财政收入的主体，社会经济的发展和社会保障的完善等都对公共财政提出要求，因此对预算支出所表现的公共财政的需求对当年的税收收入可能会有一定影响。物价水平。我国的税制结构以流转税为主，以现行价格计算的GDP等指标和经营者的收入水平都与物价水平有关。因此，可以从以上几个方面，分析各种因素对中国税收增长的具体影响。为了全面反映中国税收增长的全貌，我们选用“国家财政收入“中的各项税收”（即税收收入）作为被解释变量，反映税收的增长；选择“国内生产总值”（即GDP）作为经济整体增长水平的代表；选择“财政支出”作为公共财政需求的代表；选择“商品零售价格指数”

7、作为物价水平的代表。一、模型设定(一)变量设定为了具体分析各要素对提高我国税收收入的影响大小，选择能反映我们税收变动情况的“各项税收收入”为被解释变量，选择能影响税收收入的“国内生产总值”、“财政支出”和“ 商品零售价格指数”为解释变量。Y税收收入（亿元）X1国内生产总值（亿元）X2国家财政支出（亿元）X3商品零售价格指数（上年=100）（二）数据取得以下数据来源于中国统计年鉴，单位均为亿元表1.原始数据年份YX1X2X31978519.823645.221122.09100.71979537.824062.581281.791021980571.74545.621228.8310619816

8、29.894891.561138.41102.41982700.025323.351229.98101.91983775.595962.651409.52101.51984947.357208.051701.02102.819852040.799016.042004.25108.819862090.7310275.182204.9110619872140.3612058.622262.18107.319882390.4715042.822491.21118.519892727.416992.322823.78117.819902821.8618667.823083.59102.11991299

9、0.1721781.53386.62102.919923296.9126923.483742.2105.419934255.335333.924642.6113.219945126.8848197.865792.62121.719956038.0460793.736823.72114.819966909.8271176.597937.55106.119978234.0478973.039233.56100.819989262.884402.2810798.1897.4199910682.5889677.0513187.6797200012581.5199214.5515886.598.5200

10、115301.38109655.1718902.5899.2200217636.45120332.6922053.1598.7200320017.31135822.7624649.9599.9200424165.68159878.3428486.89102.8200528778.54184937.3733930.28100.8200634804.35216314.4340422.73101200745621.97265810.3149781.35103.8200854223.79314045.4362592.66105.9200959521.59340902.8176299.9398.8201

11、073210.79401512.889874.16103.1201189738.39473104.05109247.79104.92012100614.28518942.11125952.97102（三）模型的建立与构造在EVIEWS软件中输入数据，观察Y与三个解释变量X1、X2、X3之间的散点图，如图1、图2、图3所示：图1.Y，散点图图2.Y，散点图图1.Y，散点图由以上散点图可以看出，虽然Y与X1 X2分布类似线性分布，但Y与X3散点图分布无规律，故选择取对数模型进行分析：LnY=+Ln+Ln+Ln+e用Eviews做最小二乘回归得下表：表2.回归LnY,c,LnX1，LnX2，LnX3

12、Dependent Variable: LOG(Y)Method: Least SquaresDate: 12/11/14 Time: 12:50Sample: 1978 2012Included observations: 35VariableCoefficientStd. Errort-StatisticProb.C-7.0012372.333631-3.0001470.0053LOG(X1)0.4522940.0993424.5529150.0001LOG(X2)0.6148520.1049625.8578740.0000LOG(X3)1.1583910.5012192.3111490.

13、0276R-squared0.992412Mean dependent var8.750346Adjusted R-squared0.991678S.D. dependent var1.616175S.E. of regression0.147434Akaike info criterion-0.883661Sum squared resid0.673841Schwarz criterion-0.705907Log likelihood19.46407F-statistic1351.546Durbin-Watson stat0.605241Prob(F-statistic)0.000000故原

14、始模型为：LnY=-7.0012 + 0.4523 Ln+ 0.6149 Ln+ 1.1584 Ln （2.3336） （0.0993） （0.1050） （0.5012）t (-3.0001) (4.5529) (5.8579) (2.3111)=0.9924 =0.9917 F=1351.54 D.W.=0.6052 =0.6738二、经济意义检验从上表可以看出，所作的参数估计=0.4523，=0.6149，=1.1584，且、都均为正，符合变量参数的确定范围。但模型可能会存在其他统计缺陷，需要做进一步检验才能判定。三、统计意义检验（一）拟全优度检验可以系数=0.9924，=0.9917，

15、这说明所建模型整体上对样本数据拟合很好，即解释变量“国内生产总值（）”、“财政支出（）”和“零售价格指数（）”被解释变量“税收收入（Y）”的绝大部分差异作出了解释。（二）t检验分别针对：0（j0，1，2，3），给定显著性水平0.05，查t分布表的自由度为nk31的临界值=2.042由表2中的数据可得，与、对应的t统计量分别为4.5529, 5.8579, 2.3111,其绝对值均大于2.042，这说明在显著水平0.05下，、都能拒绝：0（j0，1，2，3），也就是说，当在其他解释变量不变的情况下，各个解释变量“国内生产总值（）”、“财政支出()”和“ 商品零售价格指数（）”分别对被解释变量“各

16、项税收收入（Y）”有显著影响，但的值与2.042较接近，显著性不够，说明模型可能多重线性、自相关等问题。（三）F检验针对H0：=0，给定显著性水平=0.05，在F分布表中查出自由度为k13和nk31的临界值（3,31）2.91，由表2中得到F1351.54F（3，31）2.91，应拒绝原假设H0：=0，说明回归方程显著，即列入模型的解释变量“国内生产总值（）”、“财政支出()”和“ 商品零售价格指数（）”联合起来确实对被解释变量“各项税收收入（Y）”有显著影响（四）结构稳定性检验考虑到1978-2012年时间跨度较大，政府财政支出及商品零售物价指数均发生了较大的变化，有必要对模型进行参数的稳定

17、性检验。将数据分为1978-1994年和1996-2012年两组分别进行普通最小二乘回归结果如下：1978-1994年：表3.回归LnY,c,LnX1，LnX2，LnX3（1978-1994）Dependent Variable: LOG(Y)Method: Least SquaresDate: 12/11/14 Time: 13:27Sample: 1978 1994Included observations: 17VariableCoefficientStd. Errort-StatisticProb.C-5.5096355.072654-1.0861440.2971LOG(X1)0.46

18、97860.5715020.8220200.4259LOG(X2)0.7195740.8737360.8235600.4250LOG(X3)0.6367011.1322930.5623110.5835R-squared0.944817Mean dependent var7.350950Adjusted R-squared0.932082S.D. dependent var0.792959S.E. of regression0.206653Akaike info criterion-0.113223Sum squared resid0.555173Schwarz criterion0.08282

19、7Log likelihood4.962398F-statistic74.19298Durbin-Watson stat0.674013Prob(F-statistic)0.000000记此时的残差为=0.55521996-2012年：表4. 回归LnY,c,LnX1，LnX2，LnX3（1996-2012）Dependent Variable: LOG(Y)Method: Least SquaresDate: 12/11/14 Time: 13:33Sample: 1996 2012Included observations: 17VariableCoefficientStd. Errort

20、-StatisticProb.C-3.9719421.491948-2.6622530.0196LOG(X1)0.2985180.1483682.0120140.0654LOG(X2)0.7553850.1088796.9378650.0000LOG(X3)0.5879870.4018271.4632820.1671R-squared0.998859Mean dependent var10.15236Adjusted R-squared0.998595S.D. dependent var0.867433S.E. of regression0.032513Akaike info criterio

21、n-3.812037Sum squared resid0.013742Schwarz criterion-3.615987Log likelihood36.40231F-statistic3791.957Durbin-Watson stat0.917267Prob(F-statistic)0.000000记残差为：=0.0137则非限定条件下的残差平方和：=+=0.5689假设检验：=F=F(K -2K)=0.7377在=0.05水平下，查表得（4 16）=3.010.7377,所以检验值不在拒绝域中，不拒绝原假设，原假设成立，模型通过稳定性检验。四、多重共线性（一）多重共线性诊断1.检验相关

22、系数利用EVIEWS软件得到各变量间相关系数矩阵表：表5.相关系数logx1,logx2,logx3LOG(X1)LOG(X2)LOG(X3)LOG(X1)10.9847-0.2138LOG(X2)0.98471-0.27616LOG(X3)-0.2138-0.27611由表中数据发现Ln与Ln之间相关系数较大，可能存在多重共线性。2.辅助回归诊断建立最小二乘模型：Ln=+ LnY=+Ln+Ln+e假设检验：=0若拒绝原假设则说明存在多重共线性，反之不拒绝则不存在。 对作辅助回归表6.回归logx1 ,c,logx2,logx3Dependent Variable: LOG(X1)Method

23、: Least SquaresDate: 12/11/14 Time: 13:54Sample: 1978 2012Included observations: 35VariableCoefficientStd. Errort-StatisticProb.C-6.8476613.972304-1.7238510.0944LOG(X2)1.0417840.03114133.454080.0000LOG(X3)1.7571040.8360752.1016100.0435R-squared0.973473Mean dependent var10.69187Adjusted R-squared0.97

24、1815S.D. dependent var1.562728S.E. of regression0.262356Akaike info criterion0.243589Sum squared resid2.202584Schwarz criterion0.376904Log likelihood-1.262800F-statistic587.1612Durbin-Watson stat0.129082Prob(F-statistic)0.000000由上表得出F=587.16(3 32)=2.90,所以拒绝原假设：=0存在多重共线性。对作辅助回归表7.回归logx2 c logx1,logx

25、3Dependent Variable: LOG(X2)Method: Least SquaresDate: 12/12/14 Time: 13:37Sample: 1978 2012Included observations: 35VariableCoefficientStd. Errort-StatisticProb.C7.5982763.6936602.0571130.0479LOG(X1)0.9332090.02789533.454080.0000LOG(X3)-1.8458540.778538-2.3709240.0239R-squared0.974322Mean dependent

26、 var8.997691Adjusted R-squared0.972718S.D. dependent var1.503316S.E. of regression0.248309Akaike info criterion0.133528Sum squared resid1.973030Schwarz criterion0.266844Log likelihood0.663260F-statistic607.1113Durbin-Watson stat0.148451Prob(F-statistic)0.000000由上表得出F=607.11(3 32)=2.90,所以拒绝原假设：=0存在多重

27、共线性。由以上诊断得出，之间存在多重共线性，需要对模型进行补救。（二）多重共线性补救1.逐步回归先用LnY对Ln、Ln、Ln分步进行最小二乘回归，找出最优最简模型。 对Ln作最小二乘回归表8.回归logY,c,logx1Dependent Variable: LOG(Y)Method: Least SquaresDate: 12/12/14 Time: 14:00Sample: 1978 2012Included observations: 35VariableCoefficientStd. Errort-StatisticProb.C-2.2184530.245893-9.0220300.0

28、000LOG(X1)1.0259000.02276345.068530.0000R-squared0.984013Mean dependent var8.750346Adjusted R-squared0.983529S.D. dependent var1.616175S.E. of regression0.207422Akaike info criterion-0.252678Sum squared resid1.419788Schwarz criterion-0.163801Log likelihood6.421862F-statistic2031.172Durbin-Watson sta

29、t0.305719Prob(F-statistic)0.000000LnY=-2.2185+1.0259 Ln(0.2459) (0.0228)t (-9.0220) (45.0685)=0.9840 =0.9835 F = 2031.17 对Ln作最小二乘回归表9.回归logx1,c,logx2Dependent Variable: LOG(Y)Method: Least SquaresDate: 12/12/14 Time: 14:01Sample: 1978 2012Included observations: 35VariableCoefficientStd. Errort-Stati

30、sticProb.C-0.8407900.221792-3.7908900.0006LOG(X2)1.0659550.02432243.826180.0000R-squared0.983109Mean dependent var8.750346Adjusted R-squared0.982597S.D. dependent var1.616175S.E. of regression0.213204Akaike info criterion-0.197691Sum squared resid1.500044Schwarz criterion-0.108814Log likelihood5.459

31、589F-statistic1920.734Durbin-Watson stat0.305459Prob(F-statistic)0.000000LnY=-0.8408+1.0660Ln(0.2218) (0.0243)T (-3.7909) (43.8262)=0.9831 =0.9826 F = 1920.73 对Ln作最小二乘回归表10.回归logx1,c,logx3Dependent Variable: LOG(Y)Method: Least SquaresDate: 12/12/14 Time: 14:03Sample: 1978 2012Included observations:

32、 35VariableCoefficientStd. Errort-StatisticProb.C37.0968222.825171.6252590.1136LOG(X3)-6.0994864.911087-1.2419830.2230R-squared0.044656Mean dependent var8.750346Adjusted R-squared0.015706S.D. dependent var1.616175S.E. of regression1.603433Akaike info criterion3.837616Sum squared resid84.84287Schwarz

33、 criterion3.926493Log likelihood-65.15827F-statistic1.542521Durbin-Watson stat0.057628Prob(F-statistic)0.222998LnY=37.0968+-6.0995Ln(22.8252) (4.9111)T (1.6253) (-1.2420)=0.0447 =0.0157 F = 1.5425由以上数据可以看出，Ln与LnY拟合的最好，所以选LnY=-2.2185+1.0259 Ln为最优最简模型。以此模型为基础，逐步添加变量，作最小二乘回归。 添加表11.回归logy,c,logx1,logx2

34、Dependent Variable: LOG(Y)Method: Least SquaresDate: 12/12/14 Time: 15:05Sample: 1978 2012Included observations: 35VariableCoefficientStd. Errort-StatisticProb.C-1.6289740.219798-7.4112280.0000LOG(X1)0.5322520.0992395.3633440.0000LOG(X2)0.5210830.1031615.0511730.0000R-squared0.991105Mean dependent v

35、ar8.750346Adjusted R-squared0.990549S.D. dependent var1.616175S.E. of regression0.157117Akaike info criterion-0.781834Sum squared resid0.789946Schwarz criterion-0.648518Log likelihood16.68209F-statistic1782.781Durbin-Watson stat0.499038Prob(F-statistic)0.000000LnY=-1.6289+0.5322 Ln+0.5210Ln（0.2197）（

36、0.0992） （0.1031） t (-7.4112) ( 5.3633) ( 5.0511)=0.9911 =0.9905 F = 1782.78添加表12.logy,c,logx1,logx3Dependent Variable: LOG(Y)Method: Least SquaresDate: 12/12/14 Time: 15:06Sample: 1978 2012Included observations: 35VariableCoefficientStd. Errort-StatisticProb.C-2.3294183.133237-0.7434540.4626LOG(X1)1

37、.0260800.02366343.362580.0000LOG(X3)0.0234630.6604140.0355280.9719R-squared0.984014Mean dependent var8.750346Adjusted R-squared0.983014S.D. dependent var1.616175S.E. of regression0.210634Akaike info criterion-0.195574Sum squared resid1.419732Schwarz criterion-0.062259Log likelihood6.422552F-statisti

38、c984.8503Durbin-Watson stat0.305246Prob(F-statistic)0.000000LnY=-2.3294+1.0260 Ln+0.0234Ln（3.1332）（0.0236）（0.6604） t (-0.7434)( 43.3625)( 0.0355)=0.9840 =0.9830 F = 984.85添加、LnY=-7.0012+0.4523 Ln+0.6149 Ln+1.1584 Ln （2.3336）（0.0993） （0.1050） （0.5012）t (-3.0001) (4.5529) (5.8579) (2.3111)=0.9924 =0.9

39、917 F=1351.54以上模型进行比较，列表如下：表13.逐步模型比较模型LnY=F（）0.98400.9835LnY=F（、）0.99110.9905LnY=F（、）0.98400.9830LnY=F（、）0.99240.9917由上表可以看出：最简模型LnY=F（）添加了后，模型的、值明显增加，说明为重要变量，不能删去。最简模型LnY=F（）添加了后，模型、值变量不明显，不能判断是否为重要变量。原始模型LnY=F（、）与之前模型比较，其、值明显增加，说明原始模型较之其他模型是拟合最好的模型，所以不能删去变量，模型保持为原始模型LnY=F（、）。2.变量变换对作变量变换表14.变量变换l

40、ogy/logx1Dependent Variable: LOG(Y)/LOG(X1)Method: Least SquaresDate: 12/12/14 Time: 13:46Sample: 1978 2012Included observations: 35VariableCoefficientStd. Errort-StatisticProb.C0.5008120.1128934.4361770.00011/LOG(X1)-7.8407512.625585-2.9862870.0055LOG(X2)/LOG(X1)0.5702050.1235834.6139450.0001LOG(X3

41、)/LOG(X1)1.3139980.5659412.3217930.0270R-squared0.826984Mean dependent var0.814026Adjusted R-squared0.810241S.D. dependent var0.036045S.E. of regression0.015702Akaike info criterion-5.362897Sum squared resid0.007643Schwarz criterion-5.185143Log likelihood97.85069F-statistic49.39148Durbin-Watson stat

42、0.642049Prob(F-statistic)0.000000LnY/Ln=0.5008-7.8407*1/Ln+0.5702*Ln/Ln+1.3139*Ln/Ln(0.1128) (2.6255) (0.1235) (0.5659)t (4.4361) (-2.9862) (4.6139) (2.3217)=0.8269 =0.8102 F=49.3914 D.W.= 0.6420对作变量变换表15.变量变换logy/logx2Dependent Variable: LOG(Y)/LOG(X2)Method: Least SquaresDate: 12/12/14 Time: 13:50

43、Sample: 1978 2012Included observations: 35VariableCoefficientStd. Errort-StatisticProb.C0.5805140.1253834.6299190.00011/LOG(X2)-7.6970262.606769-2.9527080.0060LOG(X1)/LOG(X2)0.4898430.1139384.2992210.0002LOG(X3)/LOG(X2)1.2882940.5616032.2939570.0287R-squared0.715762Mean dependent var0.969640Adjusted R-squared0.688256S.D. dependent var0.033930S.E. of regression0.018945Akaike info criterion-4.987377Sum squared resid0.011126Schwarz criterion-4.809623Log