1、-. z. . . . . 資料. . .計(jì)量經(jīng)濟(jì)學(xué)計(jì)量經(jīng)濟(jì)學(xué)期末論文市城鎮(zhèn)居民消費(fèi)支出相關(guān)因素的實(shí)證分析目錄一、引言1二、實(shí)證分析1一變量選取1二數(shù)據(jù)取得2三模型的建立與構(gòu)造3四模型檢驗(yàn)51.模型經(jīng)濟(jì)意義檢驗(yàn)52.統(tǒng)計(jì)檢驗(yàn)53.計(jì)量檢驗(yàn)53.1. 多重共線性檢驗(yàn) 53.2.鄒氏檢驗(yàn)83.3.異方差檢驗(yàn)113.4. 自相關(guān)檢驗(yàn)五模型修正16三、實(shí)證分析結(jié)論18四、政策建議19市城鎮(zhèn)居民最終消費(fèi)支出總額相關(guān)因素的實(shí)證分析【摘要】 本文旨在對(duì)1980-2010年市城鎮(zhèn)居民人均可支配收入、市商品零售價(jià)格指數(shù)以及市城鎮(zhèn)居民常住人口數(shù)對(duì)城鎮(zhèn)居民最終消費(fèi)支出總額變動(dòng)的影響進(jìn)展實(shí)證分析。首先利用EVIEWS

2、軟件建立了理論模型，進(jìn)而利用其對(duì)計(jì)量模型進(jìn)展了參數(shù)估計(jì)和檢驗(yàn)，并且對(duì)模型進(jìn)展了修正。最后，對(duì)所得的分析結(jié)果作出了經(jīng)濟(jì)意義的分析，得出結(jié)論，并提出一些政策建議?！娟P(guān)鍵詞】最終消費(fèi)支出總額 相關(guān)因素 模型 計(jì)量經(jīng)濟(jì)學(xué) 參數(shù)估計(jì) 檢驗(yàn)一、引言針對(duì)當(dāng)下國(guó)所存在的宏觀經(jīng)濟(jì)問題來看，要解決中國(guó)經(jīng)濟(jì)的又好又快以及可持續(xù)開展，首當(dāng)其沖的就是需要拉動(dòng)需，我國(guó)進(jìn)一步重視擴(kuò)大消費(fèi)的作用，把增加居民消費(fèi)作為擴(kuò)大消費(fèi)需求的重點(diǎn)，不斷拓寬消費(fèi)領(lǐng)域和改善消費(fèi)環(huán)境。改革開放以來，人們的收入水平尤其人均可支配收入在不斷增加，同時(shí)消費(fèi)品的種類和層次也在不斷更新提升。對(duì)于始終走在開展前沿的更是如此，這個(gè)作為未來世界金融中心、航運(yùn)中

3、心以及貿(mào)易中心的國(guó)際都會(huì)，它的居民尤其是在占絕大局部比重的城鎮(zhèn)居民，他們的最終消費(fèi)支出總額在這些年來發(fā)生著什么樣的變化，引起這些變化的相關(guān)因素又是什么，研究好這些問題，對(duì)于我國(guó)接下來的開展導(dǎo)向的制定和改變是有著積極的作用和影響的。居民最終消費(fèi)支出是指常住居民在一定時(shí)期的全部消費(fèi)性貨物和效勞支出，居民指的是從事消費(fèi)活動(dòng)的住戶和個(gè)人，不包括從事生產(chǎn)活動(dòng)的企業(yè)、事業(yè)、行政等各種類型單位。它是研究居民生活水平、消費(fèi)購(gòu)置力等的重要經(jīng)濟(jì)指標(biāo)。為了把它的增長(zhǎng)變化原因弄清楚，我們引入它的相關(guān)因素變量，從多方面逐一進(jìn)展剖析，再加以判斷。二、實(shí)證分析一變量選取1市城鎮(zhèn)居民人均可支配收入。由于城市的開展，居民的收入

4、在逐年遞增，消費(fèi)構(gòu)造以及消費(fèi)觀念也在發(fā)生著改變。從早期購(gòu)置耐用品到如今各類款式商品以及局部高檔奢侈品。人均可支配收入與消費(fèi)支出總額必然存在關(guān)系，且收入越高，相應(yīng)的消費(fèi)支出也會(huì)增加，預(yù)計(jì)兩者呈現(xiàn)正相關(guān)的關(guān)系。2市商品零售價(jià)格指數(shù)。通過此變量來說明價(jià)格的變動(dòng)對(duì)于消費(fèi)的影響，價(jià)格水平越高，相應(yīng)的消費(fèi)支出就會(huì)減少，預(yù)計(jì)兩者應(yīng)呈現(xiàn)負(fù)相關(guān)的關(guān)系。由于指數(shù)是一個(gè)相對(duì)量的經(jīng)濟(jì)指標(biāo)，這里均以1978年基期100。3市城鎮(zhèn)常住人口數(shù)。針對(duì)此文研究的目標(biāo)是最終消費(fèi)支出總額的相關(guān)影響因素，則由于我國(guó)是一個(gè)人口大國(guó)，每年的人口都是逐年遞增，故人口與消費(fèi)支出總額必然存在關(guān)系。人口越多，消費(fèi)支出也越多，預(yù)計(jì)兩者應(yīng)呈現(xiàn)正相關(guān)

5、的關(guān)系。Y市城鎮(zhèn)居民最終消費(fèi)支出總額億元*1市城鎮(zhèn)居民人均可支配收入元*2市商品零售價(jià)格指數(shù)以1978年為基期100*3市城鎮(zhèn)居民常住人口數(shù)萬人二數(shù)據(jù)取得1980-2010年市城鎮(zhèn)居民最終消費(fèi)支出總額相關(guān)因素統(tǒng)計(jì)表年份最終消費(fèi)支出總額Y人均可支配收入*1商品零售價(jià)格指數(shù)*2城鎮(zhèn)居民常住人口數(shù)*3198046.88637107.6702.43198150.44637109.2715.08198248.61659109.5731.31198354.73686109.6745.86198463.46834112760.75198586.431075130.4776.371986103.2412931

6、39.1802.561987113.861437151.4822.311988155.071723183.6838.931989183.451976214.3855.841990218.642183224.6864.461991238.922486245.9869.881992281.233009269.8875.551993445.154277317893.461994602.635868372.4910.491995773.647172420.9921.71996912.668159441.9932.1419971085.628439436.6943.0319981200.89877341

7、5.2953.6519991352.2210932404969.6320001579.4811718389.5986.1620011729.2212883384999.0720021949.21132503791018.8120032249.6614867375.41041.3920042682.4816683378.81097.620053138.418645376.67831148.9420063634.5620668377.35711173.320074363.7723622.73386.53031196.9420085014.4926674.9407.0691216.562009547

8、9.3128838404.8281236.1620106942.7731838.08411.73111254.95表1 以上數(shù)據(jù)來自2011年統(tǒng)計(jì)年鑒三模型的建立與構(gòu)造在EVIEWS軟件中輸入數(shù)據(jù)，觀察Y與三個(gè)解釋變量*1、*2、*3之間的散點(diǎn)圖，如圖1、圖2、圖3所示： 圖1 y與*1的散點(diǎn)圖圖2 y與*2的散點(diǎn)圖圖3 y與*3的散點(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)展普通最小二乘回歸，得到如下結(jié)果：Dependent Variable: YMethod: Least SquaresDate: 12/08/11 Time: 22:37Sam

9、ple: 1980 2010Included observations: 31VariableCoefficientStd. Errort-StatisticProb.C888.5900865.54901.0266200.3137*10.2366460.01785113.256820.0000*2-2.5437330.543711-4.6784660.0001*3-0.9053291.181137-0.7664900.4500R-squared0.989195Mean dependent var1509.068Adjusted R-squared0.987995S.D. dependent v

10、ar1854.239S.E. of regression203.1677Akaike info criterion13.58585Sum squared resid1114482.Schwarz criterion13.77089Log likelihood-206.5807F-statistic823.9566Durbin-Watson stat0.755904Prob(F-statistic)0.000000Y = 888.5900229 + 0.2366456885*1 - 2.543733397*2 - 0.9053293021*3四模型檢驗(yàn)1.模型的經(jīng)濟(jì)意義檢驗(yàn)除*3外，*1與*2的

11、估計(jì)系數(shù)符號(hào)均符合預(yù)期以及經(jīng)濟(jì)意義。2.統(tǒng)計(jì)檢驗(yàn)?zāi)Ｐ偷目蓻Q系數(shù)為0.989195，說明模型的擬合度較好，被解釋變量對(duì)解釋變量的解釋能力較強(qiáng)。F統(tǒng)計(jì)量等于823.9566大于5%顯著性水平下F3，31-3-1的臨界值3.35，說明模型整體的顯著性較高。除*3外，*1與*2的t檢驗(yàn)值均大于5%顯著性水平下自由度為31-3-1=27的臨界值2.052，通過了變量的顯著性檢驗(yàn)。故還須對(duì)模型進(jìn)展計(jì)量經(jīng)濟(jì)學(xué)檢驗(yàn)并作出修正。3.計(jì)量檢驗(yàn)3.1. 多重共線性檢驗(yàn)1對(duì)各解釋變量進(jìn)展多重共線性檢驗(yàn)利用EVIEWS軟件得到各變量間相關(guān)系數(shù)矩陣表：*1*2*3*11.0000000.7252340.970428*20

12、.7252341.0000000.803181*30.9704280.8031811.000000從系數(shù)矩陣表中看出，*3與*1之間的相關(guān)系數(shù)較高，可能存在多重共線性。2修正多重共線性利用EVIEWS分別對(duì)Y與各解釋變量*1、*2、*3做最小二乘回歸，回歸結(jié)果如下：Dependent Variable: YMethod: Least SquaresDate: 12/08/11 Time: 22:48Sample: 1980 2010Included observations: 31VariableCoefficientStd. Errort-StatisticProb.C-343.905580

13、.61173-4.2661970.0002*10.1967580.00614232.036440.0000R-squared0.972521Mean dependent var1509.068Adjusted R-squared0.971573S.D. dependent var1854.239S.E. of regression312.6306Akaike info criterion14.39026Sum squared resid2834399.Schwarz criterion14.48278Log likelihood-221.0491F-statistic1026.334Durbi

14、n-Watson stat0.346919Prob(F-statistic)0.000000Y = -343.9054935 + 0.1967584335*1Dependent Variable: YMethod: Least SquaresDate: 12/08/11 Time: 22:51Sample: 1980 2010Included observations: 31VariableCoefficientStd. Errort-StatisticProb.C-1277.062694.9926-1.8375190.0764*29.4024642.1697634.3334070.0002R

15、-squared0.393031Mean dependent var1509.068Adjusted R-squared0.372101S.D. dependent var1854.239S.E. of regression1469.300Akaike info criterion17.48530Sum squared resid62606447Schwarz criterion17.57782Log likelihood-269.0222F-statistic18.77842Durbin-Watson stat0.072711Prob(F-statistic)0.000161Y = -127

16、7.061938 + 9.402464436*2Dependent Variable: YMethod: Least SquaresDate: 12/08/11 Time: 22:52Sample: 1980 2010Included observations: 31VariableCoefficientStd. Errort-StatisticProb.C-8575.827717.1970-11.957420.0000*310.686330.74931714.261430.0000R-squared0.875209Mean dependent var1509.068Adjusted R-sq

17、uared0.870906S.D. dependent var1854.239S.E. of regression666.2224Akaike info criterion15.90347Sum squared resid12871716Schwarz criterion15.99598Log likelihood-244.5037F-statistic203.3884Durbin-Watson stat0.190966Prob(F-statistic)0.000000Y = -8575.827025 + 10.68632524*3可見，最終消費(fèi)支出總額與人均可支配收入的影響最大，與經(jīng)歷相符合

18、，因此選擇*1與Y的模型作為初始的回歸模型。對(duì)模型進(jìn)展逐步回歸，在初始模型的根底上參加解釋變量*2與*3，得到如下回歸結(jié)果參加*2：Dependent Variable: YMethod: Least SquaresDate: 12/08/11 Time: 22:54Sample: 1980 2010Included observations: 31VariableCoefficientStd. Errort-StatisticProb.C229.9373102.96502.2331600.0337*10.2237050.00575438.876910.0000*2-2.7929690.432

19、541-6.4571140.0000R-squared0.988960Mean dependent var1509.068Adjusted R-squared0.988171S.D. dependent var1854.239S.E. of regression201.6656Akaike info criterion13.54286Sum squared resid1138732.Schwarz criterion13.68164Log likelihood-206.9144F-statistic1254.117Durbin-Watson stat0.757159Prob(F-statist

20、ic)0.000000Y = 229.9373152 + 0.2237047768*1 - 2.792968875*2參加*3：Dependent Variable: YMethod: Least SquaresDate: 12/08/11 Time: 22:56Sample: 1980 2010Included observations: 31VariableCoefficientStd. Errort-StatisticProb.C2957.184983.21933.0076540.0055*10.2681200.02184812.271940.0000*3-4.2100841.25084

21、9-3.3657810.0022R-squared0.980436Mean dependent var1509.068Adjusted R-squared0.979039S.D. dependent var1854.239S.E. of regression268.4583Akaike info criterion14.11503Sum squared resid2017957.Schwarz criterion14.25381Log likelihood-215.7830F-statistic701.5978Durbin-Watson stat0.462125Prob(F-statistic

22、)0.000000Y = 2957.183726 + 0.2681202171*1 - 4.210084196*3 初始模型參加*2后可決系數(shù)上升，且各變量的t檢驗(yàn)值上升，說明變量的顯著性提高；參加*3后可決系數(shù)雖仍上升，但是各變量的t檢驗(yàn)值下降，說明變量的顯著性下降。這說明*3對(duì)模型的解釋能力不強(qiáng)，因此決定剔除*3，保存*1和*2。 修正后的模型為：Y = 229.9373152 + 0.2237047768*1 - 2.792968875*2由于剔除了變量*3，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，人均

23、可支配收入每增加1元，價(jià)格指數(shù)每上升1%，則最終消費(fèi)支出總額會(huì)增加0.2237億元，減少2.793億元。3.2.鄒氏檢驗(yàn)1對(duì)參數(shù)進(jìn)展鄒氏檢驗(yàn)考慮到1980-2010年時(shí)間跨度較大，居民的消費(fèi)觀念以及商品種類、價(jià)格均發(fā)生了較大的改變，因此有必要對(duì)模型進(jìn)展參數(shù)的穩(wěn)定性檢驗(yàn)。將數(shù)據(jù)分為1980-1994年和1996-2010年兩組分別進(jìn)展普通最小二乘回歸結(jié)果如下：1980-1994年：Dependent Variable: YMethod: Least SquaresDate: 12/14/11 Time: 23:23Sample: 1980 1994Included observations: 1

24、5VariableCoefficientStd. Errort-StatisticProb.C-23.5714311.68572-2.0171130.0666*10.1088800.00769514.149270.0000*2-0.0312000.136626-0.2283600.8232R-squared0.997322Mean dependent var179.5160Adjusted R-squared0.996876S.D. dependent var161.4286S.E. of regression9.022875Akaike info criterion7.414260Sum s

25、quared resid976.9473Schwarz criterion7.555870Log likelihood-52.60695F-statistic2234.625Durbin-Watson stat1.784714Prob(F-statistic)0.000000記此時(shí)的殘差平方和為RSS1=976.94731996-2010年：Dependent Variable: YMethod: Least SquaresDate: 12/08/11 Time: 23:05Sample: 1996 2010Included observations: 15VariableCoefficien

26、tStd. Errort-StatisticProb.C-3614.980932.6004-3.8762370.0022*10.2399630.00637937.614960.0000*26.0501752.2691122.6663180.0206R-squared0.991746Mean dependent var2887.649Adjusted R-squared0.990371S.D. dependent var1836.485S.E. of regression180.2137Akaike info criterion13.40302Sum squared resid389723.5S

27、chwarz criterion13.54463Log likelihood-97.52265F-statistic720.9383Durbin-Watson stat1.872683Prob(F-statistic)0.000000記此時(shí)的殘差平方和為RSS2=389723.5結(jié)合首次回歸的結(jié)果中殘差平方和RSSR=1138732，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：F統(tǒng)計(jì)量超出了5%顯著性水平下的臨界值，拒絕參數(shù)穩(wěn)定的前提假設(shè)條件，因此未通過鄒氏參數(shù)構(gòu)造穩(wěn)定性檢驗(yàn)，此數(shù)據(jù)存在構(gòu)造性差異。2對(duì)參數(shù)存在構(gòu)造性變化進(jìn)展修正由于未通過鄒氏檢驗(yàn)，參數(shù)存在構(gòu)造性差異，故此引入虛擬變量D1，在截距

28、項(xiàng)和斜率項(xiàng)分別影響模型。說明在1980-1994年與1996-2010兩個(gè)時(shí)間段，居民的消費(fèi)觀念與構(gòu)造發(fā)生了改變，因此以1995年作為臨界年份，修改后的模型為：其中D1=01995年以前 D1=11995年以后對(duì)上述模型作普通最小二乘回歸結(jié)果如下：Dependent Variable: YMethod: Least SquaresDate: 12/08/11 Time: 23:14Sample: 1980 2010Included observations: 31VariableCoefficientStd. Errort-StatisticProb.C-23.57143164.3544-0.

29、1434180.8871D1-3674.672670.0097-5.4845060.0000*10.1088800.1082281.0060240.3240*2-0.0312001.921588-0.0162370.9872D1*10.1300720.1083151.2008740.2411D1*26.3504882.4794562.5612420.0168R-squared0.996097Mean dependent var1509.068Adjusted R-squared0.995316S.D. dependent var1854.239S.E. of regression126.902

30、6Akaike info criterion12.69670Sum squared resid402607.0Schwarz criterion12.97425Log likelihood-190.7989F-statistic1275.978Durbin-Watson stat1.979337Prob(F-statistic)0.000000Y = -23.57142636 - 3674.672072*D1 + 0.1088797451*1 - 0.*2 + 0.1300721433*D1*1 + 6.350487647*D1*2模型的擬合度上升，且其中D1與D1*2兩個(gè)解釋變量的t檢驗(yàn)值大

31、于5%顯著性水平下自由度為25的臨界值，說明這兩個(gè)變量具有較強(qiáng)的解釋能力，因此保存D1與D1*2，剔除D1*1。則模型修正為：再次對(duì)上述模型作普通最小二乘回歸結(jié)果如下：Dependent Variable: YMethod: Least SquaresDate: 12/09/11 Time: 11:35Sample: 1980 2010Included observations: 31VariableCoefficientStd. Errort-StatisticProb.C147.684382.392161.7924560.0847D1-3834.322662.2472-5.7898650.

32、0000*10.2387430.00436954.648400.0000*2-2.2863550.410741-5.5664150.0000*2*D18.5851631.6524245.1954980.0000R-squared0.995872Mean dependent var1509.068Adjusted R-squared0.995236S.D. dependent var1854.239S.E. of regression127.9770Akaike info criterion12.68827Sum squared resid425830.9Schwarz criterion12.

33、91956Log likelihood-191.6682F-statistic1567.950Durbin-Watson stat1.966144Prob(F-statistic)0.000000Y = 147.6842822 - 3834.321875*D1 + 0.2387434908*1 - 2.286354563*2 + 8.585162965*2*D1此時(shí)模型的擬合度再次提高，同時(shí)各變量的t檢驗(yàn)值均通過了顯著性檢驗(yàn)，模型F值上升，說明整個(gè)模型的解釋能力增強(qiáng)，顯著性增強(qiáng)。并且通過引入虛擬變量D1修正了參數(shù)的構(gòu)造性差異。3.3.異方差檢驗(yàn)1異方差檢驗(yàn)首先利用EVIEWS做出殘差平方項(xiàng)與*

34、1、*2的散點(diǎn)圖4、圖5所示： 圖4 與*1的散點(diǎn)圖圖5 與*2的散點(diǎn)圖再利用EVIEWS進(jìn)展懷特檢驗(yàn)，結(jié)果如下：有穿插項(xiàng)：White Heteroskedasticity Test:F-statistic36.26392Probability0.000000Obs*R-squared29.37967Probability0.001081Test Equation:Dependent Variable: RESID2Method: Least SquaresDate: 12/11/11 Time: 12:58Sample: 1980 2010Included observations: 31V

35、ariableCoefficientStd. Errort-StatisticProb.C-24014.8467752.52-0.3544490.7267D15145418.1598058.3.2197940.0043D1*1-68.9418853.66990-1.2845540.2136D1*2-22261.227645.556-2.9116550.0086*1-16.8244551.44397-0.3270440.7470*120.0007690.0001395.5330490.0000*1*20.0503280.1902340.2645570.7941*1*(*2*D1)0.098888

36、0.1944560.5085340.6166*2416.10381162.1450.3580480.7241*22-1.1573623.689461-0.3136940.7570*2*(*2*D1)24.994639.6804942.5819580.0178R-squared0.947731Mean dependent var13736.48Adjusted R-squared0.921597S.D. dependent var36417.07S.E. of regression10196.96Akaike info criterion21.56899Sum squared resid2.08

37、E+09Schwarz criterion22.07782Log likelihood-323.3193F-statistic36.26392Durbin-Watson stat2.099187Prob(F-statistic)0.000000此時(shí)大于5%顯著性水平下自由度為10的分布臨界值18.31，因此存在異方差。無穿插項(xiàng)：White Heteroskedasticity Test:F-statistic33.54974Probability0.000000Obs*R-squared28.23481Probability0.000199Test Equation:Dependent Var

38、iable: RESID2Method: Least SquaresDate: 12/11/11 Time: 13:04Sample: 1980 2010Included observations: 31VariableCoefficientStd. Errort-StatisticProb.C3470.30622393.110.1549720.8782D12056518.1653884.1.2434470.2262*1-33.619795.474862-6.1407560.0000*120.0009860.0001426.9611260.0000*2100.2725229.75840.436

39、4260.6666*220.9224330.5326941.7316380.0967*2*D1-8034.5157952.597-1.0103010.3229(*2*D1)27.5995829.5839570.7929480.4359R-squared0.910800Mean dependent var13736.48Adjusted R-squared0.883652S.D. dependent var36417.07S.E. of regression12421.77Akaike info criterion21.90993Sum squared resid3.55E+09Schwarz

40、criterion22.27999Log likelihood-331.6038F-statistic33.54974Durbin-Watson stat2.314129Prob(F-statistic)0.000000此時(shí)大于5%顯著性水平下自由度為7的分布臨界值6.35，因此存在異方差。2模型異方差的修正令z2等于e2，定義w1=1/sqr(z2)作為權(quán)數(shù)，對(duì)模型進(jìn)展加權(quán)最小二乘回歸結(jié)果如下：Dependent Variable: YMethod: Least SquaresDate: 12/09/11 Time: 11:47Sample: 1980 2010Included observ

41、ations: 31Weighting series: W1VariableCoefficientStd. Errort-StatisticProb.C148.00731.134763130.43010.0000D1-3904.35874.21755-52.606950.0000*10.2391540.000626382.32880.0000*2-2.2891500.012391-184.74420.0000D1*28.7544220.18199048.103970.0000Weighted StatisticsR-squared1.000000Mean dependent var1721.7

42、21Adjusted R-squared1.000000S.D. dependent var7450.192S.E. of regression2.669503Akaike info criterion4.948352Sum squared resid185.2824Schwarz criterion5.179640Log likelihood-71.69945F-statistic12118677Durbin-Watson stat1.680378Prob(F-statistic)0.000000Unweighted StatisticsR-squared0.995867Mean depen

43、dent var1509.068Adjusted R-squared0.995231S.D. dependent var1854.239S.E. of regression128.0463Sum squared resid426292.3Durbin-Watson stat1.968072Y = 148.0072628 - 3904.358476*D1 + 0.2391537329*1 - 2.289149579*2 + 8.754421628*D1*2進(jìn)展加權(quán)最小二乘修正后的模型擬合度到達(dá)百分之百，同時(shí)各解釋變量的t檢驗(yàn)值均顯著提高，外表解釋能力增強(qiáng)，整個(gè)模型的解釋能力再次提高。再對(duì)修正后的

44、模型進(jìn)展懷特檢驗(yàn)結(jié)果如下：有穿插項(xiàng)：White Heteroskedasticity Test:F-statistic0.419673Probability0.920428Obs*R-squared5.376703Probability0.864636Test Equation:Dependent Variable: STD_RESID2Method: Least SquaresDate: 12/11/11 Time: 13:16Sample: 1980 2010Included observations: 31VariableCoefficientStd. Errort-StatisticP

45、rob.C-14.5099119.46344-0.7454960.4646D1-353.7725459.0784-0.7706150.4499D1*10.0112560.0154180.7300480.4738D1*21.5303252.1963590.6967560.4940*1-0.0117780.014778-0.7969420.4348*129.52E-093.99E-080.2384390.8140*1*25.65E-055.46E-051.0336430.3136*1*(D1*2)-5.61E-055.59E-05-1.0042130.3273*20.3548070.3338521

46、.0627650.3006*22-0.0012000.001060-1.1322810.2709*2*(D1*2)-0.0011500.002781-0.4134400.6837R-squared0.173442Mean dependent var5.976852Adjusted R-squared-0.239837S.D. dependent var2.630769S.E. of regression2.929308Akaike info criterion5.258832Sum squared resid171.6169Schwarz criterion5.767666Log likeli

47、hood-70.51190F-statistic0.419673Durbin-Watson stat2.282676Prob(F-statistic)0.920428此時(shí)小于5%顯著性水平下自由度為10的分布臨界值18.31，因此不存在異方差。無穿插項(xiàng)：White Heteroskedasticity Test:F-statistic0.445943Probability0.862685Obs*R-squared3.704580Probability0.813106Test Equation:Dependent Variable: STD_RESID2Method: Least Squares

48、Date: 12/11/11 Time: 13:18Sample: 1980 2010Included observations: 31VariableCoefficientStd. Errort-StatisticProb.C5.9012035.0824641.1610910.2575D1-395.6681375.3748-1.0540620.3028*1-0.0002970.001243-0.2387500.8134*127.81E-093.21E-080.2429150.8102*20.0094340.0521470.1809020.8580*22-2.45E-050.000121-0.

49、2024310.8414D1*21.9704751.8049651.0916970.2863(D1*2)2-0.0024310.002175-1.1177720.2752R-squared0.119503Mean dependent var5.976852Adjusted R-squared-0.148475S.D. dependent var2.630769S.E. of regression2.819315Akaike info criterion5.128501Sum squared resid182.8163Schwarz criterion5.498562Log likelihood

50、-71.49176F-statistic0.445943Durbin-Watson stat2.328660Prob(F-statistic)0.862685此時(shí)小于5%顯著性水平下自由度為7的分布臨界值6.35，因此不存在異方差。通過對(duì)模型進(jìn)展加權(quán)最小二乘回歸，修正了異方差，使模型通過了懷特檢驗(yàn)。并且再次提高了擬合優(yōu)度以及各解釋變量的t檢驗(yàn)值。使整個(gè)模型的解釋能力明顯提高。3.4. 自相關(guān)檢驗(yàn) 首先利用EVIEWS軟件作出殘差序列與時(shí)間以及滯后一期的殘差散點(diǎn)圖，如圖6和圖7所示：圖6 殘差序列與時(shí)間殘差散點(diǎn)圖圖7 殘差序列與滯后一期的殘差散點(diǎn)圖其次進(jìn)展D.W檢驗(yàn)和LM檢驗(yàn)D.W檢驗(yàn)：模型D.

51、W.值等于1.68，臨界值上下限分別為1.30和1.57，因此不存在自相關(guān)。LM檢驗(yàn)：利用EVIEWS軟件對(duì)模型進(jìn)展LM檢驗(yàn)，得到結(jié)果如下：Breusch-Godfrey Serial Correlation LM Test:Obs*R-squared3.843634Probability0.049935Test Equation:Dependent Variable: RESIDMethod: Least SquaresDate: 12/11/11 Time: 13:30Presample missing value lagged residuals set to zero.VariableCoefficientStd. Errort-StatisticProb.C-26.2519279.87834-0.3286490.7452D1614.6457695.85030.8833020.3855*1-0.0055510.004992-1.1120390.2767*20.213