房地產(chǎn)價(jià)格需求影響因素的實(shí)證分析

#X4X1X5X2X31.000000-0.4998820.997441-0.576337-0.144329X4-0.4998821.000000-0.4960860.9772730.648775X10.997441-0.4960861.000000-0.564538-0.098604X5-0.5763370.977273-0.5645381.0000000.706070X2-0.1443290.648775-0.0986040.7060701.000000X3從表中可以看出，XI和X2存在嚴(yán)重的正相關(guān)性，對(duì)這個(gè)的解釋是人口密集的地方也是經(jīng)濟(jì)發(fā)展迅速的地方，可能就因?yàn)槿绱硕哿舜罅康娜丝?。這個(gè)對(duì)模型的影響，起碼在經(jīng)濟(jì)上的影響是不大的。但是X4和X5的共線性會(huì)對(duì)模型的估計(jì)產(chǎn)生較大的影響。在用逐步剔除變量法修正模型后，再用最小二乘法估計(jì)：DependentVariable:YMethod:LeastSquaresDate:06/24/07Time:17:37Sample:19912005Includedobservations:15VariableCoefficientStd.Errort-StatisticProb.C-680.56384425.953-0.1537670.8806X10.2032430.0677133.0015390.0120X20.0083700.0398340.2101260.8374X30.5326350.4665131.1417380.2778R-squared0.963002Meandependentvar1917.540AdjustedR-squared0.952911S.D.dependentvar629.7125S.E.ofregression136.6475Akaikeinfocriterion12.89586Sumsquaredresid205397.8Schwarzcriterion13.08468Loglikelihood-92.71898F-statistic95.43668Durbin-Watsonstat1.099332Prob(F-statistic)0.0000000=680.5638+0.2032X+0.0084X+0.5326Xt1t2t3tSe=(4425.953)(0.0677)(0.0398)(0.4665)T=(-0.1538)(3.0015)(0.2101)(1.1417)RA2=0.963F=95.4367這個(gè)結(jié)果從變量的系數(shù)來看基本符合經(jīng)濟(jì)意義，但是t檢驗(yàn)不是很好。所以可能還存在其它的問題，諸如異方差，自相關(guān)等。異方差檢驗(yàn)在異方差的檢驗(yàn)中，無論用那種檢驗(yàn)方法得到的結(jié)論都是不存在異方差。因?yàn)槟Ｐ蛿?shù)據(jù)為時(shí)間序列數(shù)據(jù)，所以選用ARCH檢驗(yàn)來判斷時(shí)間序列是否存在I—I、.、"八異方差A(yù)RCH檢驗(yàn)H:a，a中至少有一個(gè)不為零提出假設(shè)H°：a°=ai=0做輔助回歸et"o+HH:a，a中至少有一個(gè)不為零ARCHTest:F-statistic0.145258Probability0.709770Obs*R-squared0.167441Probability0.682396TestEquation:DependentVariable:RESIDEMethod:LeastSquaresDate:06Z24/07Time:16:06Sample(adjusted):19922005Includedobservations:14afteradjustingendpointsVariableCoefficientStd.Errort-StatisticProb.C14894.315486.9142.7145140.0188RESIDA2(-1)-0.1197930.314313-0.3811280.7098R-squared0.011960Meandependentvar13429.06AdjustedR-squared-0.070377S.D.dependentvar14158.32S.E.ofregression14648.05Akaikeinfocriterion22.15357Sumsquaredresid2.57E-KJ9Schwarzcriterion22.24486Loglikelihood-153.0750F-statistic0.145258Durbin-Watsonstat1.766831Prob(F-statistic)0.709770在Q=0.05水平下，查表得，X爲(wèi)（1）=3.8414且（n-P）R2=0.1674得（n-p）R2<X2（1）得0.05所以，不拒絕原假設(shè)，表明模型中不存在異方差在進(jìn)行自相關(guān)的檢驗(yàn)自相關(guān)檢驗(yàn)一．圖示檢驗(yàn)法1．繪制et和et-1的散點(diǎn)圖RESID如圖所示，圖像雖然比較散亂，但還是有一個(gè)大致的規(guī)律性，粗略的判斷隨機(jī)誤差項(xiàng)ut存在自相關(guān)。2．按照時(shí)間序列繪制回歸殘差項(xiàng)。從上圖可以看出，et隨著t的變化逐次有規(guī)律的變化，呈大致的鋸齒形的變化,粗略判斷存在自相關(guān)。3．DW檢驗(yàn)法DependentVariable:YMethod:LeastSquaresDate:06/24/07Time:16:23Sample:19912005Includedobservations:15VariableCoefficientStd.Errort-StatisticProb.C-680.56334425.953-0.1537670.8806X10.2032430.0677133.0015390.0120X20.0083700.0398340.2101260.8374X30.5326350.4665131.1417380.2778R-squared0.963002Meandependentvar1917.540AdjustedR-squared0.952911S.D.dependentvar629.7125S.E.ofregression136.6475Akaikeinfocriterion12.89586Sumsquaredresid205397.8Schwarzcriterion13.08468Loglikelihood-92.71893F-statistic95.43668Durbin-Watsonstat1.099332Prob(F-statistic)□.□□□OOO使用普通最小二乘法估計(jì)該模型，得Yt=-680.5638+0.2032x1t+0.0084x2t+0.5326x3tSe=(4425.953)(0.0677)(0.0398)(0.4665)T=(-0.1538)(3.0015)(0.2101)(1.1417)RA2=0.963F=95.4367從上圖可以看出DW=1.099332,樣本量為15,解釋變量個(gè)數(shù)為3個(gè)，在5%顯著性水平下，查表可知，dl=0.814,du=1.75,落在了不確定區(qū)域。采用修正的DW檢驗(yàn)法進(jìn)行檢驗(yàn)，擴(kuò)大拒絕區(qū)域，拒絕H0:P=0,而不輕易接受無自相關(guān)，認(rèn)為模型存在自相關(guān)。二．自相關(guān)模型的處理1．運(yùn)用廣義的差分法估計(jì)模型。使用et進(jìn)行滯后一期的自回歸，可得回歸方程Et=0.389647et-1對(duì)原模型進(jìn)行廣義差分，得廣義差分方程Y—0.389647Y=卩(1—0.389647)+卩(x—0.389647X)+卩(x—0.389647X)+卩(x—0.389647X)+卩tt—112111(t—1)3212(t—1)4313(t—1)t

1．由于該模型樣本量較小，用普萊斯—溫斯滕變換補(bǔ)充第一個(gè)觀測值2．做最小二乘法估計(jì)該模型DependentVariable:DYMethod:LeastSquaresDate:06/24/07Time:11:04Sample:19912005Includedobservations:15VariableCoefficientStd.Errort-StatisticProb.C656.4977323.52622.0291950.0673DX10.2275990.01815912.53389DX2-0.0014720.003447-0.4270690.6776DX3-0.2760680.448259-0.6158670.5505R-squared0.940637Meandependentvar1248.515AdjustedR-squared0.924448S.D.dependentvar378.7928S.E.ofregression104.1180Akaikeinfocriterion12.35211Sumsquaredresid119246.1Schwarzcriterion12.54092Loglikelihood-88.64079F-statistic58.10065Durbin-Watsonstat0.998024Prob(F-statistic)0.000000從該圖可以看出，用廣義差分法后，t統(tǒng)計(jì)變小，可決系數(shù)和F檢驗(yàn)也變小，且DW值仍落在無法判斷區(qū)域，可見效果并不是十分理想。2.Cochrane-Orcutt法估計(jì)模型DependentVariable:DYMethod:LeastSquaresDate:06/24/07Time:17:21Sample(adjusted):19922005Includedobservations:14afteradjustingendpointsConvergenceachievedafter18iterationsVariableCoefficientStd.Errort-StatisticProb.C3513.0903508.0301.0014450.3428DX10.2845990.0811703.5062310.0067DX2-0.0409090.049592-0.8249000.4307DX3-0.3839820.540337-0.7106340.4953AR(1)-0.0676310.129470-0.5223640.6140R-squared0.933253AdjustedR-squared0.903587S.E.ofregression112.7464Sumsquaredresid114405.8Loglikelihood-82.92428Durbin-Watsonstat0.841791Meandependentvar1285.980S.D.dependentvar363.1079Akaikeinfocriterion12.56061Schwarzcriterion12.7888531.45927Prob(F-statistic)0.000027從上圖看出，效果不是十分理想。根據(jù)以上所有分析和比較，最后得出最優(yōu)模型為：Yt=-680.5638+0.2032x1t+Se=(4425.953)(0.0677)T=(-0.1538)(3.0015)RA2=0.963F=95.43670.0084x2t+0.5326x3t(0.0398)(0.4665)(0.2101)(1.1417)將X1可支配收入滯后一期，得到估計(jì)結(jié)果DependentVariable:YMethod:LeastSquaresDate:06/24/07Time:18:19Sample(adjusted):19922005Includedobservations:14afteradjustingendpointsVariableCoefficientStd.Errort-StatisticProb.C9971.2895330.3521.8706630.0942X1(-1)0.4803220.2027432.3691170.0420X1-0.1002860.146499-0.6845510.5109X2-0.0837700.047252-1.7728410.1100X30.6635120.4245171.5629810.1525R-squared0.974117Meandependentvar1998.351AdjustedR-squared0.962613S.D.dependentvar567.0545S.E.ofregression109.6443Akaikeinfocriterion12.50481Sumsquaredresid108196.8Schwarzcriterion12.73305Loglikelihood-82.53369F-statistic84.67823Durbin-Watsonstat1.743606.Prob(F-statistic)0.000000模型的估計(jì)結(jié)果也不是很好，可能的原因是，人們的可支配收入對(duì)價(jià)格的當(dāng)期影響比較大，而其他期的收入水平對(duì)當(dāng)期沒有太大的影響。估計(jì)結(jié)果：目前這個(gè)估計(jì)結(jié)果是最優(yōu)的。Yt=-680.5638+0.2032x1t+0.0084x2t+0.5326x3tSe=(4425.953)(0.0677)(0.0398)(0.4665)T=(-0.1538)(3.0015)(0.2101)(1.1417)RA2=0.963F=95.43674、結(jié)論根據(jù)最終的模型得出的估計(jì)結(jié)果可以看出，在剔除了零售商品物價(jià)和CPI以后，模型中個(gè)變量的系數(shù)正負(fù)號(hào)基本符合經(jīng)濟(jì)意義。雖然，人口數(shù)量和匯率對(duì)房地產(chǎn)的價(jià)格沒有顯著性的影響，原因可能在于：先從人口來講，中國的人口基數(shù)很大，而且最近幾年增長速度也在減慢。所以說可能他對(duì)房地產(chǎn)價(jià)格的影響不是很顯著，這一點(diǎn)從它的系數(shù)也可以看出，但是這并不代表可以剔除這個(gè)因素，因?yàn)樗从车氖侨藗儗?duì)房屋的居住需求。再從匯率來看，九十年代的匯率制度屬于政府管制下的固定匯率制度，雖然現(xiàn)在的很多文獻(xiàn)也提到人民幣的升值將引發(fā)新一輪的房地產(chǎn)的泡沫經(jīng)濟(jì)，但是從91-05年的數(shù)據(jù)來看，影響不太明顯。但是我們有理由相信人民幣升值的預(yù)期，會(huì)推動(dòng)房地產(chǎn)價(jià)格的急速上升，刺激人們投機(jī)性的需求。最