关于房价问题的初步分析
引言：近改革开放20多年来，从来没有哪一个行业像房地产业这样盛产亿万富翁，各种富豪排行榜上，房地产富豪连年占据半壁江山；“中国十大暴利行业”中，房地产业每年都是“第一名”。是什么造就了这样的状况。房地产的问题，在开发商，政府，购房者三者来看，就是一场完完全全的博弈。而这场博弈的焦点则是房价问题。如果说开发商与政府之间的博弈是围绕“土地”这个关键词，那么整个房地产市场则在价格上开展了新一轮的对峙。先是开发商与购房者在房价涨跌上僵持不下；再有开发商与政府之间的土地成本论；最后则是关于房地产是否归为暴利行业的争执，“价格”成了市场关注的焦点。而对于房价的构成因素，至今仍然是不透明的。公布房价成本成为另政府极为头疼的一件事。房价成本是一个非常复杂的集合体，并且项目间差异性较大，同时还有软资产、品牌等组成部分，特别是现在的商品房，追求品质、功能完善以及个性化成本构成越来越难衡量。
写作目的：通过对一系列影响房价的基本因素的分析，了解对其主要因素和次要因素。并对这些因素进行统计推断和经济意义上的检验。选择拟和效果最好的最为结论。在一定层面上分析房地产如此暴利的因素。当然笔者的能力有限，并不能全面的分析这一问题。仅仅就几个因素进行分析。
写作方法：理论分析及计量分析方法，将会用到Eviews软件进行帮助分析。
关键词：房价成本拟合优度
现在我们以2003年的数据，选取30个省市的数据为例进行分析。在Eviews软件中选择建立截面数据。现在我们以2003年的数据，选取31个省市的数据为例进行分析。令Y=各地区建筑业总产值。（万元）X1=各地区房屋竣工面积。（万平方米）X2=各地区建筑业企业从业人员。（人）X3=各地区建筑业劳动生产率。（元/人）X4=各地区人均住宅面积。（平方米）X5=各地区人均可支配收入。（元）
数据如下：
YX1X3X2X4X5
126985214254.800569767.0129961.024.7714013882.62
5208402.1465.800238957.0147063.023.0957010312.91
7799313.4748.300989317.070048.0023.167107239.060
5401279.1313.300591276.089151.0022.996807005.030
2576575.1450.700265953.061074.0020.053107012.900
101707943957.100966790.082496.0020.235107240.580
3469281.1626.800303837.077486.0020.705907005.170
4401878.2181.300441518.068033.0020.492006678.900
119580343609.200505185.0153910.029.3453014867.49
2794935417730.002727006.100569.024.435309262.460
3127277916183.902429352.127430.031.0233013179.53
6227073.4017.600910691.066407.0020.754806778.030
5493441.2952.100553611.0108288.030.298709999.540
3593356.2750.900574705.070826.0022.619806901.420
148136189139.8002072530.60728.0024.480808399.910
6345217.3433.600932901.066056.0020.200906926.120
8729958.4840.8001048763.81761.0022.902807321.980
8188402.4969.7001119106.74553.0024.425807674.200
151632428105.0001492820.101932.024.9328012380.43
2818466.1721.600353700.077472.0024.173207785.040
394053.0121.500061210.0055361.0023.432007259.250
5862095.4939.600817997.069432.0025.724408093.670
122533748784.6002070534.59748.0026.358507041.870
2122907.980.3000293310.072152.0018.194306569.230
3967957.2248.700522470.069238.0024.929407643.570
293427.0121.300036593.0073205.0019.929908765.450
4404362.1580.000410311.093212.0021.750506806.350
2236860.1327.200449409.046857.0021.113806657.240
747325.0242.9000101501.061046.0019.105506745.320
1080546.578.700088225.0061459.0022.255006530.480
3196774.1450.800203375.095835.0020.781107173.540
先用Eviews软件进行White检验：
WhiteHeteroskedasticityTest:
F-statistic2.779810Probability0.049670
Obs*R-squared26.27412Probability0.156948
TestEquation:
DependentVariable:RESID^2
Method:LeastSquares
Date:12/22/05Time:21:50
Sample:131
Includedobservations:31
VariableCoefficientStd.Errort-StatisticProb.
C6.08E+122.29E+130.2655390.7960
X51.64E+083.88E+090.0423700.9670
X5^287293.54453712.30.1923980.8513
X5*X4380671243.56E+080.1068100.9171
X5*X31363.5556160.0700.2213540.8293
X5*X2-17464.3650393.75-0.3465580.7361
X5*X1-453312.21215201.-0.3730350.7169
X4-9.71E+111.83E+12-0.5314860.6067
X4^24.28E+106.46E+100.6617200.5231
X4*X3-1905048.1949296.-0.9773010.3515
X4*X2-1901040317319142-1.0976530.2981
X4*X14.23E+084.15E+081.0208010.3314
X3-1386946034509844-0.4018990.6962
X3^241.8184322.625401.8482960.0943
X3*X2517.0981231.19542.2366280.0493
X3*X1-14772.938469.467-1.7442580.1117
X21.51E+083.45E+080.4388530.6701
X2^22050.2611851.4101.1074050.2940
X2*X1-67170.5950453.24-1.3313430.2126
X17.80E+086.17E+090.1264300.9019
X1^21246362.746355.01.6699320.1259
R-squared0.847552Meandependentvar1.17E+12
AdjustedR-squared0.542656S.D.dependentvar1.78E+12
S.E.ofregression1.21E+12Akaikeinfocriterion58.69986
Sumsquaredresid1.46E+25Schwarzcriterion59.67127
Loglikelihood-888.8478F-statistic2.779810
Durbin-Watsonstat1.809921Prob(F-statistic)0.049670
结果显示为没有异方差。
DW值为1.809921，没有自相关。
做多重共线性检验：
X5X4X3X2X1
X51.0000000.6865130.2798510.8362410.418307
X40.6865131.0000000.4778860.5408810.538697
X30.2798510.4778861.0000000.1250290.960871
X20.8362410.5408810.1250291.0000000.271375
X10.4183070.5386970.9608710.2713751.000000
可以看出有多重共线性。
采取逐步回归法：
第一次回归，我们可以根据T检验值和可决系数看出：X1的效果最好：
DependentVariable:Y
Method:LeastSquares
Date:12/22/05Time:21:16
Sample:131
Includedobservations:31
Variable Coefficient Std.Error t-Statistic Prob.
X1 1651.403 87.67703 18.83508 0.0000
C 903234.0 502408.2 1.797809 0.0826
R-squared 0.924432 Meandependentvar 7446408.
AdjustedR-squared 0.921826 S.D.dependentvar 7227629.
S.E.ofregression 2020815. Akaikeinfocriterion 31.93824
Sumsquaredresid 1.18E+14 Schwarzcriterion 32.03076
Loglikelihood -493.0427 F-statistic 354.7601
Durbin-Watsonstat 1.930762 Prob(F-statistic) 0.000000
依次加入X2，X3，X4，X5：
可得，加入X2后的效果最好：
DependentVariable:Y
Method:LeastSquares
Date:12/22/05Time:21:16
Sample:131
Includedobservations:31
Variable Coefficient Std.Error t-Statistic Prob.
X2 60.57577 9.136899 6.629795 0.0000
X1 1547.354 57.83197 26.75604 0.0000
C -3711880. 765709.2 -4.847637 0.0000
R-squared 0.970594 Meandependentvar 7446408.
AdjustedR-squared 0.968493 S.D.dependentvar 7227629.
S.E.ofregression 1282914. Akaikeinfocriterion 31.05893
Sumsquaredresid 4.61E+13 Schwarzcriterion 31.19771
Loglikelihood -478.4134 F-statistic 462.0886
Durbin-Watsonstat 2.098685 Prob(F-statistic) 0.000000
再加入X3，X4，X5
加入X3，回归：
DependentVariable:Y
Method:LeastSquares
Date:12/26/05Time:10:09
Sample:131
Includedobservations:31
Variable Coefficient Std.Error t-Statistic Prob.
X1 1392.586 243.1554 5.727144 0.0000
X2 64.15614 10.72532 5.981748 0.0000
X3 0.924103 1.409311 0.655713 0.5176
C -4115494. 988624.2 -4.162850 0.0003
R-squared 0.971055 Meandependentvar 7446408.
AdjustedR-squared 0.967838 S.D.dependentvar 7227629.
S.E.ofregression 1296176. Akaikeinfocriterion 31.10765
Sumsquaredresid 4.54E+13 Schwarzcriterion 31.29268
Loglikelihood -478.1686 F-statistic 301.9308
Durbin-Watsonstat 2.037807 Prob(F-statistic) 0.000000
加入X4，回归：
DependentVariable:Y
Method:LeastSquares
Date:12/26/05Time:10:09
Sample:131
Includedobservations:31
Variable Coefficient Std.Error t-Statistic Prob.
X1 1569.186 66.74467 23.51029 0.0000
X2 64.04945 10.56258 6.063810 0.0000
X4 -69455.16 102797.7 -0.675649 0.5050
C -2476469. 1985261. -1.247428 0.2230
R-squared 0.971083 Meandependentvar 7446408.
AdjustedR-squared 0.967870 S.D.dependentvar 7227629.
S.E.ofregression 1295550. Akaikeinfocriterion 31.10668
Sumsquaredresid 4.53E+13 Schwarzcriterion 31.29171
Loglikelihood -478.1536 F-statistic 302.2316
Durbin-Watsonstat 2.298423 Prob(F-statistic) 0.000000
加入X5，回归：
DependentVariable:Y
Method:LeastSquares
Date:12/26/05Time:10:10
Sample:131
Includedobservations:31
Variable Coefficient Std.Error t-Statistic Prob.
X1 1511.624 60.28105 25.07627 0.0000
X2 39.25698 15.77525 2.488517 0.0193
X5 316.7476 193.7661 1.634691 0.1137
C -4428358. 863348.9 -5.129279 0.0000
R-squared 0.973242 Meandependentvar 7446408.
AdjustedR-squared 0.970269 S.D.dependentvar 7227629.
S.E.ofregression 1246240. Akaikeinfocriterion 31.02907
Sumsquaredresid 4.19E+13 Schwarzcriterion 31.21410
Loglikelihood -476.9506 F-statistic 327.3477
Durbin-Watsonstat 1.861895 Prob(F-statistic) 0.000000
我们发现加入X3，X4，X5的效果都不好，T检验都不充分。
于是我们只保留X1，X2再回归，得：
DependentVariable:Y
Method:LeastSquares
Date:12/22/05Time:21:16
Sample:131
Includedobservations:31
Variable Coefficient Std.Error t-Statistic Prob.
X2 60.57577 9.136899 6.629795 0.0000
X1 1547.354 57.83197 26.75604 0.0000
C -3711880. 765709.2 -4.847637 0.0000
R-squared 0.970594 Meandependentvar 7446408.
AdjustedR-squared 0.968493 S.D.dependentvar 7227629.
S.E.ofregression 1282914. Akaikeinfocriterion 31.05893
Sumsquaredresid 4.61E+13 Schwarzcriterion 31.19771
Loglikelihood -478.4134 F-statistic 462.0886
Durbin-Watsonstat 2.098685 Prob(F-statistic) 0.000000
得出回归函数为：
Y=1547.354X1+60.57577X2-3711880
结论：我们总认为房产总价值与许多成分有关，其实在最后我们看到并不是这样。但现实中房价成本具有相当大的难度。不管是资金成本很难简单地以招拍挂价格进行测算，还是融资成本比较难核算。而且房地产的利润要以综合成本衡量。种种原因构成了房价成本确定的难度。而房产行业的暴利，开发商的暴利是来源于开发商的阶层优越感和特殊占有地位，而与之相对的是老百姓的阶层卑微感和相对剥削感。房地产业的暴利如果继续维持，考验的不仅是中国经济的稳定，更是老百姓忍耐的限度。而且这种房产的暴利行为导致了从2003年10月开始的通货膨胀，并造成了中国越来越大的金融风险。我国房价的公开将会采取怎么样的方式，笔者将和大家一起拭目以待。
