《安徽省财政收入与经济增长的回归模型分析.doc》由会员分享，可在线阅读，更多相关《安徽省财政收入与经济增长的回归模型分析.doc（12页珍藏版）》请在三一办公上搜索。
1、安徽省财政收入与经济增长的回归模型分析摘要：财政收入与经济增长之间存在着高度的相关性，本文在相关经济学理论的基础上，对安徽省财政收入与经济增长间关系做了实证分析，并得出结论，要保持一地区或一个国家经济的可持续增长，财政收入与经济增长之间应形成相互依存的长期稳定关系.关键词：财政收入，经济增长，回归分析 财政收入与经济增长之间存在着相互依存、相互制约的关系，正确认识二者之间的关系，对促进我省经济增长有重要作用。一 理论分析财政收入是政府部门的公共收入，表现为政府部门在一定时期内所取得的货币收入。在西方经济学教科书中，国内生产总值（GDP）是指经济社会（即一国或一地区）在一定时期内运用生产要素所生
2、产的全部最终产品（物品和劳务）的市场价值，是国民经济活动最终成果的总量指标。研究过财政收入与经济增长之间关系的学者很多。最先比较明确提出国家财政税收原则的是威廉配第，他在代表作赋税论中，比较深刻地分析了税收与国民财富、税收与国家经济实力之间的关系。亚当斯密在其著作国富论一书中，综合了自由主义学说的观点，主张对经济实行自由放任的政策，认为政府应当减少干预或者不干预，政府只应作为“守夜人”存在。斯密之后，许多经济学家从不同角度提出了不同的财政税收观点，比如瓦格纳在其代表著作财政学中提出了社会政策的财政理论，认为财政收入增长能够随着经济增长自动增加。哈勃格计算了税收的超额负担，进而发现课税扭曲了消费
3、者对课税商品与其他商品的选择。我国学者高培勇认为，应当根据实际情况合理科学地确定财政收入和财政支出，不能简单的量入为出。一个地区或一个国家要保持经济的可持续增长，财政收入与经济增长之间应形成相依相存的长期稳定关系，并且，只有合理的财政收入水平才能对GDP的增长产生积极的影响，这一命题可以根据拉弗曲线得以证明。在经济学界，美国供给学派经济学家拉弗知名度颇高，以其“拉弗曲线”而著称于世。拉弗曲线表明：在税率增长的初期， GDP迅速增长；当税率增长超过某一点，尽管其增长率不变，但GDP的增长率迅速下降，甚至出现负增长，图中表示为EB线段。当税负大于A点时，过高的税率反而导致政府税收收入的减少，长期来
4、看会抑制居民消费、储蓄和投资的积极性，从而抑制经济的可持续发展，因此ABC区域被称之为“禁区”，政府部门需要在OAC区域征税。 二实证分析安徽省财政收入与生产总值表年份财政收入（亿元）生产总值（亿元）199052.89658.02199148.18663.60199255.14801.16199373.211069.841994108.761488.471995147.002003.581996193.142339.251997230.812669.951998262.072805.451999280.852908.592000290.422902.092001309.553246.71200
5、2346.653519.722003412.293923.112004520.714759.32005656.555350.172006816.516112.520071034.737360.9220081326.058851.6620091551.2610062.821相关说明经济增长可以用GDP来表示，建立计量经济模型，解释财政收入与经济增长之间的关系。本文财政收入和GDP数据均来源于安徽省统计年鉴2010 2一元回归模型的建立 Y=+X+其中：Y为各年的财政收入，X为各年的GDP，为常数项，为回归系数，为随机变量。模型估计在Eviews软件包, 进行OLS估计。3.检验结果Depende
6、nt Variable: YMethod: Least SquaresDate: 06/08/11 Time: 20:52Sample: 1990 2009Included observations: 20VariableCoefficientStd. Errort-StatisticProb.C-151.050623.83424-6.3375430.0000X0.1597040.00528330.227060.0000R-squared0.980680Mean dependent var435.8385Adjusted R-squared0.979607S.D. dependent var4
7、32.9014S.E. of regression61.82064Akaike info criterion11.18099Sum squared resid68792.24Schwarz criterion11.28056Log likelihood-109.8099F-statistic913.6752Durbin-Watson stat0.198163Prob(F-statistic)0.000000得出回归方程为: Y= 151.05 + 0.16X (6.34) (30.23)R-squared=0.981 DW=0.198 S.E=61.821 F=913.675 T=20(括号中
8、的数字表示参数估计值对应的t统计量) 下面对模型进行平稳性检验，自相关检验和异方差检验。（1） 平稳性检验1.1首先使用图示法X和Y均呈现递增，很可能部平稳。1.2用ADF法对x进行平稳性检验，得Null Hypothesis: X has a unit rootExogenous: Constant, Linear TrendLag Length: 1 (Automatic based on AIC, MAXLAG=4)t-StatisticProb.*Augmented Dickey-Fuller test statistic-0.0055940.9924Test critical val
9、ues:1% level-4.5715595% level-3.69081410% level-3.286909*MacKinnon (1996) one-sided p-values.Warning: Probabilities and critical values calculated for 20 observationsand may not be accurate for a sample size of 18Augmented Dickey-Fuller Test EquationDependent Variable: D(X)Method: Least SquaresDate:
10、 06/27/11 Time: 22:42Sample (adjusted): 1992 2009Included observations: 18 after adjustmentsVariableCoefficientStd. Errort-StatisticProb.X(-1)-0.0007820.139792-0.0055940.9956D(X(-1)0.6940830.3654451.8992830.0783C-19.40139137.6275-0.1409700.8899TREND(1990)21.7485441.639710.5223030.6096R-squared0.7652
11、65Mean dependent var522.1789Adjusted R-squared0.714965S.D. dependent var427.5773S.E. of regression228.2778Akaike info criterion13.89213Sum squared resid729550.8Schwarz criterion14.08999Log likelihood-121.0292Hannan-Quinn criter.13.91942F-statistic15.21393Durbin-Watson stat1.823266Prob(F-statistic)0.
12、000110结果显示在=5%的水平下，不能拒绝原假设，即x是非平稳的。同理对Y做ADF检验，也没有通过检验。2.1偏相关系数检验Date: 06/22/11 Time: 17:08Sample: 1990 2009Included observations: 20AutocorrelationPartial CorrelationACPACQ-StatProb. |*|. |*|10.7640.76413.5160.000. |* |.*| . |20.482-0.24419.2000.000. |*. |. *| . |30.222-0.12820.4750.000. | . |.*| .
13、|4-0.037-0.22020.5130.000. *| . |. | . |5-0.1800.05021.4640.001. *| . |. |* . |6-0.1740.14922.4100.001. *| . |.*| . |7-0.197-0.22223.7290.001. *| . |. | . |8-0.198-0.04925.1630.001.*| . |. *| . |9-0.225-0.18927.1850.001.*| . |.*| . |10-0.326-0.20731.8660.000*| . |. *| . |11-0.417-0.12940.3700.000*|
14、. |. | . |12-0.4100.00449.6080.000由偏相关（PAC）也可推断出，y和x之间存在着一阶自相关2.2布罗斯-戈弗雷（b-g）检验或者说是LM检验Test Equation:Dependent Variable: RESIDMethod: Least SquaresDate: 06/22/11 Time: 17:52Sample: 1990 2009Included observations: 20Presample missing value lagged residuals set to zero.VariableCoefficientStd. Errort-S
15、tatisticProb.C-12.5633613.16756-0.9541150.3534X0.0046360.0029731.5595070.1373RESID(-1)0.9395690.1428806.5759260.0000R-squared0.717809Mean dependent var7.11E-14Adjusted R-squared0.684610S.D. dependent var60.17179S.E. of regression33.79225Akaike info criterion10.01582Sum squared resid19412.58Schwarz c
16、riterion10.16518Log likelihood-97.15821Hannan-Quinn criter.10.04498F-statistic21.62140Durbin-Watson stat1.124328Prob(F-statistic)0.000021Obs*R-squared项对应的伴随概率p=0.000151,小于0.05的显著水平，说明存在一阶自相关。2.3异方差检验（不带交叉项的White异方差检验）White Heteroskedasticity Test:F-statistic4.250745Probability0.031843Obs*R-squared6.
17、667446Probability0.035660Test Equation:Dependent Variable: RESID2Method: Least SquaresDate: 06/08/11 Time: 21:36Sample: 1990 2009Included observations: 20VariableCoefficientStd. Errort-StatisticProb.C7357.0201735.9144.2381250.0006X-2.4315580.873201-2.7846500.0127X20.0002478.46E-052.9147670.0097R-squ
18、ared0.333372Mean dependent var3439.612Adjusted R-squared0.254946S.D. dependent var3217.214S.E. of regression2776.987Akaike info criterion18.83360Sum squared resid1.31E+08Schwarz criterion18.98296Log likelihood-185.3360F-statistic4.250745Durbin-Watson stat0.726673Prob(F-statistic)0.031843Obs*R-square
19、d项的伴随概率p=0.037,小于0.05的显著水平，说明存在异方差。3.1现在的问题是这个模型中x和Y不平稳，既存在自相关，又存在异方差，该如何处理呢？我们试着设立模型LnY=+LnX+ 。检验结果如下:首先进行平稳性检验。对lnx和lny进行ADF检验Null Hypothesis: LNX has a unit rootExogenous: Constant, Linear TrendLag Length: 3 (Automatic based on AIC, MAXLAG=4)t-StatisticProb.*Augmented Dickey-Fuller test statistic
20、-4.0790380.0278Test critical values:1% level-4.6678835% level-3.73320010% level-3.310349*MacKinnon (1996) one-sided p-values.Warning: Probabilities and critical values calculated for 20 observationsand may not be accurate for a sample size of 16在5%的显著水平下，通过了检验。同理lnyNull Hypothesis: LNY has a unit ro
21、otExogenous: Constant, Linear TrendLag Length: 4 (Automatic based on AIC, MAXLAG=4)t-StatisticProb.*Augmented Dickey-Fuller test statistic-3.9744190.0351Test critical values:1% level-4.7283635% level-3.75974310% level-3.324976*MacKinnon (1996) one-sided p-values.Warning: Probabilities and critical v
22、alues calculated for 20 observationsand may not be accurate for a sample size of 15也通过了检验。在命令窗口输入 命令：ls lny c lnx 得Dependent Variable: LNYMethod: Least SquaresDate: 06/08/11 Time: 22:07Sample: 1990 2009Included observations: 20VariableCoefficientStd. Errort-StatisticProb.C-4.6658110.229532-20.327470
23、.0000LNX1.2941930.02880144.935250.0000R-squared0.991164Mean dependent var5.596989Adjusted R-squared0.990673S.D. dependent var1.058707S.E. of regression0.102244Akaike info criterion-1.628269Sum squared resid0.188169Schwarz criterion-1.528696Log likelihood18.28269F-statistic2019.177Durbin-Watson stat0
24、.354984Prob(F-statistic)0.000000得出回归方程为: lnY= -4.67 + 1.29 lnX (-20.33) (44.94)R-squared=0.99 DW=0.35 S.E=0.10 F=2019.18 T=20(括号中的数字表示参数估计值对应的t统计量)再次使用怀特检验进行检验，得到Heteroskedasticity Test: WhiteF-statistic3.775495Prob. F(2,17)0.0440Obs*R-squared6.151271Prob. Chi-Square(2)0.0462Scaled explained SS4.779
25、216Prob. Chi-Square(2)0.0917Test Equation:Dependent Variable: RESID2Method: Least SquaresDate: 06/14/11 Time: 23:26Sample: 1990 2009Included observations: 20VariableCoefficientStd. Errort-StatisticProb.C0.4386410.2351611.8652840.0795LNX-0.1034650.060696-1.7046540.1065LNX20.0061600.0038881.5844650.13
26、15R-squared0.307564Mean dependent var0.009408Adjusted R-squared0.226100S.D. dependent var0.013370S.E. of regression0.011762Akaike info criterion-5.910468Sum squared resid0.002352Schwarz criterion-5.761108Log likelihood62.10468Hannan-Quinn criter.-5.881311F-statistic3.775495Durbin-Watson stat1.310581
27、Prob(F-statistic)0.043977Obs*R-squared的P值为0.046仍小于显著性水平0.05，说明还有一定的异方差再次使用LM检验，（滞后向为1）Breusch-Godfrey Serial Correlation LM Test:F-statistic14.03623Prob. F(1,17)0.0016Obs*R-squared9.045061Prob. Chi-Square(1)0.0026Test Equation:Dependent Variable: RESIDMethod: Least SquaresDate: 06/22/11 Time: 18:59S
28、ample: 1990 2009Included observations: 20Presample missing value lagged residuals set to zero.VariableCoefficientStd. Errort-StatisticProb.C-0.0600930.175536-0.3423430.7363LNX0.0079460.0220360.3605810.7229RESID(-1)0.6895340.1840483.7464960.0016R-squared0.452253Mean dependent var8.01E-16Adjusted R-sq
29、uared0.387812S.D. dependent var0.099517S.E. of regression0.077865Akaike info criterion-2.130211Sum squared resid0.103069Schwarz criterion-1.980851Log likelihood24.30211Hannan-Quinn criter.-2.101054F-statistic7.018115Durbin-Watson stat1.248890Prob(F-statistic)0.005997Obs*R-squared项对应的伴随概率p=0.00260.05
30、,已经消除了序列自相关再次使用怀特检验，Heteroskedasticity Test: WhiteF-statistic0.530792Prob. F(2,16)0.5981Obs*R-squared1.182194Prob. Chi-Square(2)0.5537Scaled explained SS0.811413Prob. Chi-Square(2)0.6665Test Equation:Dependent Variable: RESID2Method: Least SquaresDate: 06/22/11 Time: 19:16Sample: 1991 2009Included o
31、bservations: 19Collinear test regressors dropped from specificationVariableCoefficientStd. Errort-StatisticProb.C0.0071080.0051801.3720480.1890GRADF_022-0.0005560.000540-1.0295210.3185GRADF_032-0.0137970.018361-0.7514450.4633R-squared0.062221Mean dependent var0.001807Adjusted R-squared-0.055002S.D. dependent var0.002584
