計(jì)量經(jīng)濟(jì)學(xué)論文我國稅收收入影響因素的實(shí)證研究

1、計(jì)量經(jīng)濟(jì)學(xué)期末論文我國稅收收入影響因素的實(shí)證研究姓名：郭瑞班級(jí)：2010國際經(jīng)濟(jì)與貿(mào)易1班學(xué)號(hào)：1002013023時(shí)間：2012年12月16日摘要：稅收是我們國財(cái)政收入的基本因素，也影響著我國經(jīng)濟(jì)的發(fā)展。本文通過查閱相關(guān)文獻(xiàn)以及搜索相關(guān)的網(wǎng)站信息對(duì)分析我國稅收收入影響因素進(jìn)行一系列的文獻(xiàn)綜述，并通過Eviews計(jì)量經(jīng)濟(jì)學(xué)軟件對(duì)稅收收入的影響因素包括選取國內(nèi)生產(chǎn)總值、財(cái)政支出、商品零售價(jià)格指數(shù)進(jìn)行分析，得出相關(guān)結(jié)論并對(duì)我國財(cái)政收入方面給出一些建議。關(guān)鍵詞：稅收收入、國內(nèi)生產(chǎn)總值、財(cái)政支出、商品零售價(jià)格指數(shù)、計(jì)量分析目錄引言3一、理論綜述4（一）文獻(xiàn)綜述41.國內(nèi)生產(chǎn)總值對(duì)稅收收入的影響42.財(cái)

2、政收入對(duì)稅收收入的影響4（二）現(xiàn)狀分析4二、實(shí)證分析5（一）變量選取5（二）數(shù)據(jù)取得5（三）模型的建立與構(gòu)造6（四）模型檢驗(yàn)81.經(jīng)濟(jì)意義檢驗(yàn)82.統(tǒng)計(jì)檢驗(yàn)83.計(jì)量檢驗(yàn)8（1）多重線性檢驗(yàn)8（2）鄒氏檢驗(yàn)13（3）異方差檢驗(yàn)14（4）自相關(guān)檢驗(yàn)20（五）模型修正22三、結(jié)論分析及政策建議22（一）結(jié)論分析22（二）政策建議23參考文獻(xiàn)24引言自1985年實(shí)行的利改稅的稅改以來，稅收占財(cái)政收入的比重逐年上升，90年代已高達(dá)96%。而1994年實(shí)施的全面稅制改革又使得稅收收入有了新的變化。稅收組織財(cái)政收入、調(diào)控經(jīng)濟(jì)運(yùn)行和監(jiān)督經(jīng)濟(jì)活動(dòng)職能的發(fā)揮，成為國家非常關(guān)心的問題。從進(jìn)入新世紀(jì)，我國的經(jīng)濟(jì)發(fā)展

3、面臨著巨大的機(jī)遇和挑戰(zhàn)。在新經(jīng)濟(jì)背景下，基于知識(shí)和信息的產(chǎn)業(yè)發(fā)展迅猛，全球經(jīng)濟(jì)發(fā)展一體化日漸深入，中國成功加入WTO。新形勢(shì)下的經(jīng)濟(jì)發(fā)展是經(jīng)濟(jì)穩(wěn)定和協(xié)調(diào)增長(zhǎng)的結(jié)果，由于稅收具有聚財(cái)與調(diào)控的功能，因而它在實(shí)現(xiàn)經(jīng)濟(jì)發(fā)展的過程中將發(fā)揮非常重要的作用，研究稅收收入的影響因素對(duì)我國有著重要的意義。一、理論綜述（一）文獻(xiàn)綜述高淑紅在我國稅收收入的影響因素分析一文中運(yùn)用多重共線性檢驗(yàn)和加權(quán)最小二乘估計(jì)法等計(jì)量經(jīng)濟(jì)學(xué)檢驗(yàn)方法對(duì)稅收收入與其影響因素做了相關(guān)計(jì)量分析，得出了以下分析結(jié)果與結(jié)論：1.國內(nèi)生產(chǎn)總值對(duì)稅收收入的影響國內(nèi)生產(chǎn)總值與稅收收入成正相關(guān)。這表明，國內(nèi)生產(chǎn)總值的增加會(huì)帶來稅收的增加。正如前面所述，

4、經(jīng)濟(jì)是稅收收入的源泉，稅收的增長(zhǎng)離不開經(jīng)濟(jì)的增長(zhǎng)，稅收收入受經(jīng)濟(jì)發(fā)展的影響，而國內(nèi)生產(chǎn)總值在很大程度上就反映我國的經(jīng)濟(jì)的發(fā)展?fàn)顩r。2.財(cái)政收入對(duì)稅收收入的影響稅收收入與財(cái)政支出顯著的正相關(guān)。這表明，隨著財(cái)政支出的增加，稅收收入也會(huì)相應(yīng)的增加，而且，其系數(shù)為0.7009，遠(yuǎn)高于國內(nèi)生產(chǎn)總值的系數(shù)。估計(jì)其原因，因?yàn)閲腋疄榱死瓌?dòng)經(jīng)濟(jì)增長(zhǎng)，常常實(shí)施加大財(cái)政支出力度，從而使經(jīng)濟(jì)得到發(fā)展，各項(xiàng)稅收相應(yīng)的都有所增加，進(jìn)而增加了稅收的總收入。（二）現(xiàn)狀分析我國的社會(huì)主義市場(chǎng)經(jīng)濟(jì)體制還不完善，各方面運(yùn)作還需要政府實(shí)施一定的宏觀職能，職能的有效實(shí)施得宜于充足的財(cái)政力量，其中稅收占很大比重。1、經(jīng)濟(jì)增長(zhǎng)仍是稅

5、收收入高增長(zhǎng)的主要決定因素, 稅收收入與經(jīng)濟(jì)增長(zhǎng)之間有著正的線性相關(guān)性。另外,我國稅收收入增長(zhǎng)具有較大的慣性。2、我國稅收收入增長(zhǎng)速度略慢于經(jīng)濟(jì)增長(zhǎng)速度,稅制改革勢(shì)在必行。另外, 稅收是我國財(cái)政收入的主要來源, 稅收收入大幅度增長(zhǎng),通過財(cái)政支出政策的運(yùn)用,有力支持了經(jīng)濟(jì)和社會(huì)各項(xiàng)事業(yè)的發(fā)展。二、實(shí)證分析（一）變量選取為了全面反映中國稅收增長(zhǎng)的全貌，選擇包括中央和地方稅收的“國家財(cái)政收入”中的“各項(xiàng)稅收”（簡(jiǎn)稱“稅收收入”）作為被解釋變量，以反映國家稅收的增長(zhǎng)；選擇“國內(nèi)生產(chǎn)總值（GDP）”作為經(jīng)濟(jì)整體增長(zhǎng)水平的代表；選擇中央和地方“財(cái)政支出”作為公共財(cái)政需求的代表；選擇“商品零售物價(jià)指數(shù)”作為

6、物價(jià)水平的代表。Y稅收收入（億元）X1國內(nèi)生產(chǎn)總值（億元）X2國家財(cái)政支出（億元）X3商品零售價(jià)格指數(shù)（以1980年為基期100）（二）數(shù)據(jù)取得以下數(shù)據(jù)來源于中國統(tǒng)計(jì)年鑒，單位均為億元。年 份國內(nèi)生產(chǎn)總值國家財(cái)政支出商品零售物價(jià)指數(shù)（上年=100)稅收收入19804545.6241228.83106571.719814891.5611138.41102.4629.8919825323.3511229.98101.9700.0219835962.6521409.52101.5775.5919847208.0521701.02102.8947.3519859016.0372004.25108.82

7、040.79198610275.182204.911062090.73198712058.622262.18107.32140.36198815042.822491.21118.52390.47198916992.322823.78117.82727.4199018667.823083.59102.12821.86199121781.53386.62102.92990483742.2105.43296.91199335333.924642.3113.24255.3199448197.865792.62121.75126.88199560793.736823.72114

8、.86038.04199671176.597937.55106.16909.82199778973.039233.56100.88234.04199884402.2810798.1897.49262.8199989677.0513187.679710682.58200099214.5515886.598.512581.512001109655.218902.5899.215301.382002120332.722053.1598.717636.452003135822.824649.9599.905920017.312004159878.328486.89102.806225718200518

9、3867.933930.28100.777430866200621087140422.73101.028237636表1. 1980-2006年我國稅收收入相關(guān)因素統(tǒng)計(jì)表（三）模型的建立與構(gòu)造在EVIEWS軟件中輸入數(shù)據(jù)，觀察Y與三個(gè)解釋變量X1、X2、X3之間的散點(diǎn)圖，如圖1、圖2、圖3所示：圖1圖2圖3由以上散點(diǎn)圖發(fā)現(xiàn)存在較強(qiáng)的線性關(guān)系，故此選擇建立線性模型。建立模型：、利用EVIEWS軟件對(duì)數(shù)據(jù)進(jìn)行普通最小二乘回歸，得到如圖4結(jié)果：Dependent Variable: YMethod: Least SquaresDate: 12/16/12 Time: 12:50Sample: 198

10、0 2006Included observations: 27VariableCoefficientStd. Errort-StatisticProb.  C-6357.3062589.143-2.4553710.0221X1-0.0111910.014037-0.7972610.4335X20.9670820.07682112.588750.0000X357.1184124.003452.3795920.0260R-squared0.994954    Mean dependent var8681.087Adjusted R-squ

11、ared0.994296    S.D. dependent var9909.343S.E. of regression748.4057    Akaike info criterion16.20972Sum squared resid12882553    Schwarz criterion16.40170Log likelihood-214.8312    F-statistic1511.718Durbin-Watson stat0

12、.691548    Prob(F-statistic)0.000000圖4Y = -6357.306 - 0.011191*X1 + 0.967082*X2 + 57.11841*X3 （2589.143） （0.014037） （0.076821） （24.00345）t =(-2.455371) (-0.797261) (12.58875) (2.379592)=0.994954 =0.994296 F=1511.718（四）模型檢驗(yàn)1.經(jīng)濟(jì)意義檢驗(yàn)我國稅收收入與財(cái)政支出及商品零售物價(jià)指數(shù)呈正相關(guān)關(guān)系，當(dāng)國內(nèi)其他因素不變時(shí)，財(cái)政支出每增加1單位，我

13、國稅收收入增加0.967082單位；當(dāng)其他因素不變時(shí)，商品零售物價(jià)指數(shù)每增加1單位，我國稅收收入增加57.11841單位，兩者與稅收收入呈正相關(guān)符合現(xiàn)實(shí)經(jīng)濟(jì)意義，但模型中國內(nèi)生產(chǎn)總值與稅收收入呈負(fù)相關(guān)，不符合現(xiàn)實(shí)經(jīng)濟(jì)意義。2.統(tǒng)計(jì)檢驗(yàn)由=0.994954 ，=0.994296與1十分接近，說明模型擬合優(yōu)度很好。F統(tǒng)計(jì)量等于1511.718大于5%顯著性水平下F（3，23）的臨界值3.03，表明模型整體的顯著性較高。除X1外，X2與X3的t檢驗(yàn)值均大于5%顯著性水平下自由度為23的臨界值1.711，通過了變量的顯著性檢驗(yàn)。故還須對(duì)模型進(jìn)行計(jì)量經(jīng)濟(jì)學(xué)檢驗(yàn)并作出修正。3.計(jì)量檢驗(yàn)（1）多重線性檢驗(yàn)對(duì)

14、各解釋變量進(jìn)行多重共線性檢驗(yàn)利用EVIEWS軟件得到各變量間相關(guān)系數(shù)矩陣表：X1X2X3X110.984833-0.407265X20.9848331-0.416781X3-0.407265-0.4167811表2. X1、X2、X3相關(guān)系數(shù)矩陣表從系數(shù)矩陣表中看出，X1與X2之間的相關(guān)系數(shù)較高，可能存在多重共線性。修正多重共線性.用EVIEWS分別對(duì)Y與各解釋變量X1、X2、X3做最小二乘回歸：Dependent Variable: YMethod: Least SquaresDate: 12/16/12 Time: 14:11Sample: 1980 2006Included observ

15、ations: 27VariableCoefficientStd. Errort-StatisticProb.  C-1143.176559.4057-2.0435540.0517X10.1610650.00658424.463690.0000R-squared0.959902    Mean dependent var8681.087Adjusted R-squared0.958298    S.D. dependent var9909.343S.E. of regression2023.59

16、2    Akaike info criterion18.13432Sum squared resid1.02E+08    Schwarz criterion18.23031Log likelihood-242.8134    F-statistic598.4724Durbin-Watson stat0.170737    Prob(F-statistic)0.000000圖5Y = -1143.176 + 0.161065 * X1

17、 （559.4057） （0.006584）=0.959902 DW=0.170737Dependent Variable: YMethod: Least SquaresDate: 12/16/12 Time: 14:13Sample: 1980 2006Included observations: 27VariableCoefficientStd. Errort-StatisticProb.  C-292.7317212.2144-1.3794150.1800X20.8925750.01434062.244310.0000R-squared0.993589 &#

18、160;  Mean dependent var8681.087Adjusted R-squared0.993332    S.D. dependent var9909.343S.E. of regression809.1614    Akaike info criterion16.30106Sum squared resid16368556    Schwarz criterion16.39705Log likelihood-218.0643 

19、   F-statistic3874.355Durbin-Watson stat0.501126    Prob(F-statistic)0.000000圖6Y = -292.7317 + 0.892575 * X2 (212.2144) (0.014340)=0.993589 DW=0.501126Dependent Variable: YMethod: Least SquaresDate: 12/16/12 Time: 14:14Sample: 1980 2006Included observations: 27Vari

20、ableCoefficientStd. Errort-StatisticProb.  C68011.8528622.302.3761840.0255X3-564.9916272.0256-2.0769790.0482R-squared0.147161    Mean dependent var8681.087Adjusted R-squared0.113047    S.D. dependent var9909.343S.E. of regression9332.439  &

21、#160; Akaike info criterion21.19157Sum squared resid2.18E+09    Schwarz criterion21.28756Log likelihood-284.0862    F-statistic4.313843Durbin-Watson stat0.179687    Prob(F-statistic)0.048232圖7Y = 68011.85 + 564.9916 * X3 (28622.30) (27

22、2.0256)=0.147161 DW=0.179687以上3個(gè)方程根據(jù)經(jīng)濟(jì)理論和統(tǒng)計(jì)檢驗(yàn)得出，財(cái)政支出X2是最重要的解釋變量（t檢驗(yàn)值=62.24431也最大），從而得出最優(yōu)簡(jiǎn)單回歸方程Y=f(X2)。.對(duì)模型進(jìn)行逐步回歸，在初始模型的基礎(chǔ)上加入解釋變量X1與X3，得到如下回歸結(jié)果：加入X1，Dependent Variable: YMethod: Least SquaresDate: 12/16/12 Time: 14:32Sample: 1980 2006Included observations: 27VariableCoefficientStd. Errort-StatisticPr

23、ob.  C-218.4640240.3033-0.9091180.3723X1-0.0105150.015337-0.6855710.4996X20.9489780.08353911.359650.0000R-squared0.993712    Mean dependent var8681.087Adjusted R-squared0.993188    S.D. dependent var9909.343S.E. of regression817.8773   

24、; Akaike info criterion16.35574Sum squared resid16054157    Schwarz criterion16.49972Log likelihood-217.8025    F-statistic1896.345Durbin-Watson stat0.526704    Prob(F-statistic)0.000000圖8Y = -218.4640 + -0.010515 *X1 + 0.948978 * X2(2

25、40.3033) (0.015337) (0.083539)=0.993712加入X3,Dependent Variable: YMethod: Least SquaresDate: 12/16/12 Time: 14:37Sample: 1980 2006Included observations: 27VariableCoefficientStd. Errort-StatisticProb.  C-6394.6562568.992-2.4891690.0201X20.9069500.01448062.636270.0000X356.7307423.815652.3820

26、780.0255R-squared0.994815    Mean dependent var8681.087Adjusted R-squared0.994383    S.D. dependent var9909.343S.E. of regression742.7027    Akaike info criterion16.16291Sum squared resid13238574    Schwarz criterion16.3

27、0689Log likelihood-215.1993    F-statistic2302.212Durbin-Watson stat0.652300    Prob(F-statistic)0.000000圖9Y = -6394.656 + 0.906950 * X2 + 56.73074 * X3 (2568.992) (0.014480) (23.81565)=0.994815由以上數(shù)據(jù)構(gòu)成表格如下：(X1)(X2)(X3)Y=f(X2)-292.7317(212.2144)0.892575(0.01434

28、0)0.993589Y=f(X1,X2)-218.4640(240.3033)-0.010515(0.015337)0.948978(0.083539)0.993712Y=f(X3,X2)-6394.656(2568.992)0.906950(0.014480)56.73074(23.81565)0.994815Y=f(X1,X2,X3)-6357.306(2589.143)-0.011191(0.014037)0.967082(0.076821)57.11841(24.00345)0.994954表3. 稅收收入模型估計(jì)結(jié)果分析：在最優(yōu)簡(jiǎn)單回歸方程Y=f(X2)中引入X1，值略有提高。雖然X

29、2與X1高度相關(guān)，在X1的引入對(duì)參數(shù)影響不大，的符號(hào)不滿意，可以是“多余變量”，暫時(shí)刪除；模型中引入X3，使值由0.993589提升到0.994815，正號(hào)也合理，進(jìn)行t檢驗(yàn)，不顯著。從經(jīng)濟(jì)理論分析，X3應(yīng)該是重要變量，雖然X2與X3高度相關(guān)，但不影響的顯著性和穩(wěn)定性，因此，可能是“有利變量”，暫時(shí)保留；最后在Y=f(X3,X2)的基礎(chǔ)上引入X1，=0.994954幾乎沒有增加，其他兩個(gè)參數(shù)系數(shù)沒有多大影響，可以確定X1是多余變量，應(yīng)從模型中刪除。得出最后回歸模型是：Y = -6394.656 + 0.906950 * X2 + 56.73074 * X3 (2568.992) (0.0144

30、80) (23.81565)=0.994815由于剔除了變量X1，故模型已不存在多重共線性，且各解釋變量前得系數(shù)均符合經(jīng)濟(jì)意義，模型擬合度上升，各變量t檢驗(yàn)值上升。在其他因素保持不變的情況下，財(cái)政支出每增加1億元，商品零售物價(jià)指數(shù)增加1%，稅收收入增加57.6377億元。（2）鄒氏檢驗(yàn)考慮到1980-2006年時(shí)間跨度較大，政府財(cái)政支出及商品零售物價(jià)指數(shù)均發(fā)生了較大的變化，有必要對(duì)模型進(jìn)行參數(shù)的穩(wěn)定性檢驗(yàn)。將數(shù)據(jù)分為1980-1992年和1993-2006年兩組分別進(jìn)行普通最小二乘回歸結(jié)果如下：1980-1992年：Dependent Variable: YMethod: Least Squa

31、resDate: 12/16/12 Time: 15:47Sample: 1980 1992Included observations: 13VariableCoefficientStd. Errort-StatisticProb.  C-3271.7351116.480-2.9304020.0150X21.0799520.07083115.246950.0000X325.7728610.765052.3941240.0377R-squared0.965039    Mean dependent var1855.634Adjusted

32、 R-squared0.958047    S.D. dependent var999.6892S.E. of regression204.7616    Akaike info criterion13.68074Sum squared resid419273.0    Schwarz criterion13.81112Log likelihood-85.92483    F-statistic138.0159Durbin-Watson

33、 stat1.601545    Prob(F-statistic)0.000000圖10記此時(shí)的殘差平方和為RSS1=4192731993-2006年：Dependent Variable: YMethod: Least SquaresDate: 12/16/12 Time: 16:10Sample: 1993 2006Included observations: 14VariableCoefficientStd. Errort-StatisticProb.  C-10058.024408.677-2.2814140.0434X20

34、.9409590.02693934.929190.0000X384.4832740.020972.1109750.0585R-squared0.992858    Mean dependent var15019.01Adjusted R-squared0.991560    S.D. dependent var10277.24S.E. of regression944.1875    Akaike info criterion16.72594Sum squared resid

35、9806391.    Schwarz criterion16.86288Log likelihood-114.0816    F-statistic764.6048Durbin-Watson stat0.739942    Prob(F-statistic)0.000000圖11記此時(shí)的殘差平方和為RSS2=9806391結(jié)合首次回歸的結(jié)果中殘差平方和RSSR=13238574，根據(jù)鄒氏參數(shù)穩(wěn)定性檢驗(yàn)的方法構(gòu)造F統(tǒng)計(jì)量： = =2.06<F（3，21）=3.07F統(tǒng)計(jì)

36、量小于了5%顯著性水平下的臨界值，接受參數(shù)穩(wěn)定的前提假設(shè)條件，因此通過了鄒氏參數(shù)結(jié)構(gòu)穩(wěn)定性檢驗(yàn)，此數(shù)據(jù)不存在結(jié)構(gòu)性差異。（3）異方差檢驗(yàn)異方差檢驗(yàn)首先利用EVIEWS做出殘差平方項(xiàng)resid2與X2、X3的散點(diǎn)圖12、圖13所示：圖12圖13由以上散點(diǎn)圖表示可能存在異方差。圖14由圖14顯示回歸方程的殘差分布有明顯的擴(kuò)大趨勢(shì)，表明方程存在異方差。再利用EVIEWS進(jìn)行懷特檢驗(yàn)，結(jié)果如下：a.有交叉項(xiàng)：White Heteroskedasticity Test:F-statistic7.109815    Probability0.000495Obs*R-

37、squared16.97331    Probability0.004551Test Equation:Dependent Variable: RESID2Method: Least SquaresDate: 12/16/12 Time: 16:34Sample: 1980 2006Included observations: 27VariableCoefficientStd. Errort-StatisticProb.  C-8759545.38461050-0.2277510.8220X21309.274610.68812.143

38、9330.0439X220.0001200.0010830.1109340.9127X2*X3-12.453015.971275-2.0854860.0494X3106713.5691508.00.1543200.8788X32-214.11253101.802-0.0690280.9456R-squared0.628641    Mean dependent var490317.6Adjusted R-squared0.540222    S.D. dependent var807591.7S.E. of reg

39、ression547602.9    Akaike info criterion29.45762Sum squared resid6.30E+12    Schwarz criterion29.74558Log likelihood-391.6778    F-statistic7.109815Durbin-Watson stat1.626934    Prob(F-statistic)0.000495圖15此時(shí)=16.9733大于5%

40、顯著性水平下自由度為5的分布臨界值11.07，因此存在異方差。b.無交差項(xiàng)White Heteroskedasticity Test:F-statistic6.769393    Probability0.001038Obs*R-squared14.89671    Probability0.004920Test Equation:Dependent Variable: RESID2Method: Least SquaresDate: 12/16/12 Time: 16:41Sample: 1980 2006Inc

41、luded observations: 27VariableCoefficientStd. Errort-StatisticProb.  C25487592373326460.6827160.5019X238.7262445.214360.8565030.4010X220.0002760.0011590.2377760.8143X3-459782.0682613.6-0.6735610.5076X322062.6313116.5070.6618410.5149R-squared0.551730    Mean dependent va

42、r490317.6Adjusted R-squared0.470226    S.D. dependent var807591.7S.E. of regression587810.0    Akaike info criterion29.57177Sum squared resid7.60E+12    Schwarz criterion29.81174Log likelihood-394.2189    F-statistic6.76

43、9393Durbin-Watson stat1.530228    Prob(F-statistic)0.001038圖16此時(shí)=14.89671大于5%顯著性水平下自由度為4的分布臨界值9.49，因此存在異方差。模型異方差的修正定義w1=1/sqr(resid2)作為權(quán)數(shù)，對(duì)模型進(jìn)行加權(quán)最小二乘回歸結(jié)果如下：Dependent Variable: YMethod: Least SquaresDate: 12/16/12 Time: 17:34Sample: 1980 2006Included observations: 27Weighting seri

44、es: W1VariableCoefficientStd. Errort-StatisticProb.  C-6305.814136.0376-46.353450.0000X20.9291040.004854191.42140.0000X355.192881.39214539.645940.0000Weighted StatisticsR-squared0.999998    Mean dependent var3565.330Adjusted R-squared0.999998    S.D.

45、 dependent var15337.09S.E. of regression22.06084    Akaike info criterion9.129924Sum squared resid11680.33    Schwarz criterion9.273906Log likelihood-120.2540    F-statistic324652.5Durbin-Watson stat1.336304    Prob(F-st

46、atistic)0.000000Unweighted StatisticsR-squared0.993942    Mean dependent var8681.087Adjusted R-squared0.993437    S.D. dependent var9909.343S.E. of regression802.7850    Sum squared resid15467129Durbin-Watson stat0.528265圖17Y = -6305.814 +

47、0.929104 * X2 + 55.19288 * X3 (136.0376) (0.004854) (1.392145)進(jìn)行加權(quán)最小二乘修正后的模型擬合度達(dá)到接近百分之百，同時(shí)各解釋變量的t檢驗(yàn)值均顯著提高，表面解釋能力增強(qiáng)，整個(gè)模型的解釋能力提高。再對(duì)修正后的模型進(jìn)行懷特檢驗(yàn)結(jié)果如下：a.有交叉項(xiàng)White Heteroskedasticity Test:F-statistic1.920492    Probability0.133585Obs*R-squared8.472079    Probability

48、0.132066Test Equation:Dependent Variable: STD_RESID2Method: Least SquaresDate: 12/16/12 Time: 17:46Sample: 1980 2006Included observations: 27VariableCoefficientStd. Errort-StatisticProb.  C66719.7345195.421.4762500.1547X2-0.4809610.717617-0.6702190.5100X222.56E-061.27E-062.0157220.0568X2*X

49、30.0034930.0070170.4977990.6238X3-1170.753812.5882-1.4407700.1644X325.1798173.6449141.4211080.1700R-squared0.313781    Mean dependent var432.6049Adjusted R-squared0.150395    S.D. dependent var698.1207S.E. of regression643.4859    Akaike in

50、fo criterion15.96481Sum squared resid8695556.    Schwarz criterion16.25277Log likelihood-209.5249    F-statistic1.920492Durbin-Watson stat2.177267    Prob(F-statistic)0.133585圖18此時(shí)=8.472079小于5%顯著性水平下自由度為5的分布臨界值11.07，因此不存在異方差。b.無交叉項(xiàng)White Het

51、eroskedasticity Test:F-statistic2.421456    Probability0.078863Obs*R-squared8.253446    Probability0.082723Test Equation:Dependent Variable: STD_RESID2Method: Least SquaresDate: 12/16/12 Time: 17:48Sample: 1980 2006Included observations: 27VariableCoefficientS

52、td. Errort-StatisticProb.  C57113.6940163.961.4220130.1690X2-0.1245820.048643-2.5611340.0178X222.52E-061.25E-062.0209930.0556X3-1011.855734.3831-1.3778300.1821X324.5412093.3528631.3544270.1893R-squared0.305683    Mean dependent var432.6049Adjusted R-squared0.179444    S.D. dependent var698.1207S.E. of regression632.3896    Akaike info criterion15.90246Sum squared resid8798165.    Schwarz criterion16.14243Log likelihood-209.6833