我国城镇居民消费影响因素分析.docx

1、我国城镇居民的消费影响因素分析姓名：毕雅颖班级：统计0901学号：2009111039指导教师：郑华章内容摘要 随着经济水平的不断发展，人们的消费水平也在不断提高，从社会生产的角度看，消费是其最终环节，消费能否顺利实现是决定企业能否顺利完成生产的重要原因，也是决定经济增长的主要因素，因此消费对于经济增长有着重要的作用。研究影响消费的因素对于我国的经济发展有很重要的经济意义。本文根据2009年我国城镇居民的消费数据和影响其变量的数据，试图运用数学方法和计量经济的理论和方法的有关知识，通过建立经济模型研究影响消费水平的各个因素以及这些因素对消费水平的影响程度大小。然后根据模型的最终结果制定一些有利

2、于消费水平提高的政策，从而促进经济的增长，提高人们的生活水平。 关键词：城镇居民、消费水平、经济增长、计量经济模型目 录一、 前言二、 理论背景三、 模型的选择与建立四、 数据来源分析（一） 原始数据表（二） 变量的基本统计信息（三） 数据的散点图（四） 模型的设定形式五、 模型的估计六、 模型的应用七、 经验解释八、 总结和建议我国城镇居民消费影响因素分析一、 前言 消费是人类生产的目的，而生产是社会的核心活动，在社会在生产中，生产必须围绕消费需求来进行，消费需求同时对生产有重要的导向作用。生产的消费是相辅相成的，生产者生产的产品的规模，质量，档次直接决定了消费的规模，质量和档次，消费能否顺

3、利实现，决定了生产能否顺利完成。消费行为是指消费者受需求动机的影响而做出购买决定、修改购买方案、完成购买过程的行为。消费者行为过程既是消费者的思维、心理过程，也是不断采取行动、产生方案、解决问题的过程。影响消费者行为的因素是多方面的，由社会的、历史的、经济的等多方面极其复杂的因素。从扩大需求的角度来看，消费是刺激经济增长的主要方式。所以研究消费的影响因素对经济增长有重要的经济意义。 二、理论背景从经济学的角度看，消费支出和经济发展成共同的变化趋势，经济增长越快，消费支持增长也越快；反之，经济增长越慢，消费支出增长也越慢。为了全面的反映影响消费支出的因素，本文在分析了影响居民消费支出的影响因素的

4、基础上，研究有关参考文献，主要利用计量经济学方法理论，通过对数据的准确分析，建立模型并检验，研究居民消费支出的函数关系。从而得出具体的影响消费支出的因素与消费支出间的定量关系。从而得出具体的消费支出模型。三、 模型的选择与建立本模型研究2009年我国城镇居民的消费支出与各主要影响因素之间的定量关系其中，模型中的被解释变量为Y 各城市居民的人均消费支出消费性支出(元)城乡居民人民币存款年底余额（亿元）总计-居民消费价格指数(上年=100)国内生产总值/增加值（当年价，亿元）城镇居民消费水平（元）平均每人可支配收入（元）城镇人口-人口数(万人)YX1X2X3X4X5X6北京17,893.30146

5、72.198.4612,153.0324,044.0026738.481,492.00天津14,801.354885.999.037,521.8517,475.0021402.01958.00河北省9,678.7513551.199.3217,235.4812,195.0014718.253,025.00山西省9,355.108099.499.587,358.3110,617.0013996.551,576.00内蒙古12,369.873914.099.739,740.2514,784.0015849.191,293.00辽宁省12,324.5812030.9100.0515,212.4914

6、,774.0015761.382,607.00吉林省10,914.444614.4100.077,278.7512,061.0014006.271,461.00黑龙江省9,629.606430.1100.198,587.0010,592.0012565.982,123.00上海20,992.3513707.399.5915,046.4531,608.0028837.781,702.00江苏省13,153.0020080.699.5734,457.3015,965.0020551.724,295.00浙江省16,683.4817833.498.4722,990.3521,251.0024610.

7、812,999.00安徽省10,233.986619.599.0910,062.8211,301.0014085.742,581.00福建省13,450.577078.898.1912,236.5315,739.0019576.831,864.00江西省9,739.995092.799.347,655.1810,033.0014021.541,914.00山东省12,012.7317082.8100.0033,896.6516,027.0017811.044,576.00河南省9,566.9911207.499.4219,480.4611,884.0014371.563,577.00湖北省10

8、,294.078163.599.5912,961.1012,080.0014367.482,631.00湖南省10,828.237809.899.6413,059.6913,000.0015084.312,767.00广东省16,857.5031411.497.6539,482.5621,098.0021574.726,110.00广西10,352.384686.297.867,759.1612,585.0015451.481,904.00海南省10,086.651282.999.321,654.2110,215.0013750.85424.00重庆市12,144.064908.798.366

9、,530.0113,314.0015748.671,475.00四川省10,860.2011575.2100.8014,151.2811,701.0013839.403,168.00贵州省9,048.292676.198.723,912.6811,223.0012862.531,135.00云南省10,201.814668.6100.406,169.7511,661.0014423.931,554.00西藏9,034.31226.4101.41441.369,563.0013544.4169.00陕西省10,705.676743.8100.528,169.8012,223.0014128.76

10、1,641.00甘肃省8,890.793026.9101.253,387.5610,765.0011929.78860.00青海省8,786.52711.3102.651,081.2710,845.0012691.85233.00宁夏10,280.00967.7100.751,353.3113,151.0014024.70288.00新疆9,327.553050.8100.744,277.0510,546.0012257.52860.00模型中初步确定的解释变量为X1 各城市居民人民币储蓄存款年底余额X2 各城市居民消费价格指数总计X3 各城市生产总值X4 各城市居民消费水平X5 各城市居民人

11、均可支配收入X6 各城市人口数量初步确定的模型形式为Y=F(X1 ,X2,X3,X4,X5,X6)四、 数据来源与分析（一）原始数据表以上数据来源于中经网统计数据库，中国统计年鉴2009（二）变量的基本统计信息 为了模型参数估计、检验、各种系数计算和预测的方便，各变量的基本统计分析如表2、3、4、所示表2 各变量的基本统计分析Y X1 X2X3X4X5X6Mean11628.978348.70099.6696811783.914010.30216276.952037.484Median10352.386619.50099.590008585.0012195.0014371.561702.000

12、Maximum20992.3531411.40102.6500039483.531608.0028837.786110.000Minimum8786.520226.400097.65000441.369565.00011929.7869.00000Std. Dev.2978.7916799.9971.0984709730.44801.4124326.4661354.169Observations31313131313131表3 简单相关系数yX1X2X3X4X5X6y10.622032-0.4708310.4926820.9757550.9703040.312512X10.6220321-0.

13、4567620.9412150.6022420.6187730.892691X2-0.479831-0.4567621-0.40544-0.40352-0.49614-0.425885X30.4926810.941215-0.4054410.4787670.5094470.943031X40.9757550.602242-0.403520.47876710.9588580.290134X50.9703040.618773-0.496140.5094470.95885810.307145X60.3125120.892691-0.4258850.9430310.2901340.3071451表4

14、变量的协方差YX1X2X3X4X5X6Y858696512193254-1517.9901381962513505458121015461219940X11219325444748039-3301.7616026796019028601176169897955016X2-1517.990-3301.7611.167713-4193.775-2059.641-2281.881-613.0737X31381962560267960-4193.77591626510216462912075498912025114X41350545819028601-2059.64121646291223098951

15、92759601825579X51210154617616989-2281.8812075498919275960181144941741440X612199407955016-613.073712025114182557917414401774620(三)、样本数据散点图由以上几幅图可以看出，Y与X4 X5有较强的线性关系。与X1 X2 X3 X6呈现一定的指数关系，初步考虑选择模型的形式为对数模型形式。（四）根据散点图，模型的初步形式设定为LnY=0+1lnX1+2lnX2+3lnX3+4lnX4+5lnX5+6X6+五模型的估计对六元对数模型进行回归估计，得到结果表5表5Dependen

16、t Variable: LYMethod: Least SquaresDate: 12/14/11 Time: 15:57Sample: 1 31Included observations: 31VariableCoefficientStd. Errort-StatisticProb.C4.1638827.5964470.5481360.5887LX10.0071490.0530150.1348490.8939LX2-0.6687141.470204-0.4548440.6533LX3-0.0315410.071650-0.4402160.6637LX40.4685330.1242923.76

17、96200.0009LX50.3977390.1945312.0446020.0520LX60.0233630.0885040.2639800.7940R-squared0.957720Mean dependent var9.334675Adjusted R-squared0.947150S.D. dependent var0.224760S.E. of regression0.051670Akaike info criterion-2.892192Sum squared resid0.064075Schwarz criterion-2.568388Log likelihood51.82898

18、Hannan-Quinn criter.-2.786640F-statistic90.60823Durbin-Watson stat1.803176Prob(F-statistic)0.000000 由表看出，解释变量的t检验不显著，模型拟合效果不好，且X3的经济意义不合理，所以采用对数模型的效果不理想，故将模型的设定形式改成：Y=0+1X1+2X2+3X3+4X4+5X5+6X6+2、对六元线性模型进行最小 二乘回归估计。回归估计结果为Dependent Variable: YMethod: Least SquaresDate: 12/14/11 Time: 00:21Sample: 1 3

19、1Included observations: 31VariableCoefficientStd. Errort-StatisticProb.C13348.1614660.660.9104750.3716X10.0515360.0603690.8536850.4017X2-112.3481139.7690-0.8038130.4294X3-0.0489940.051808-0.9456890.3537X40.3540550.0832974.2504940.0003X50.2717540.1116302.4344030.0227X60.1187050.3922940.3025920.7648R-

20、squared0.970888Mean dependent var11628.97Adjusted R-squared0.963610S.D. dependent var2978.791S.E. of regression568.2411Akaike info criterion15.71865Sum squared resid7749551.Schwarz criterion16.04245Log likelihood-236.6390Hannan-Quinn criter.15.82420F-statistic133.3994Durbin-Watson stat1.973291Prob(F

21、-statistic)0.000000由回归结果可以看到解释变量的t检验有了明显的改善，所以选择线性函数模型。但是X6的P值较大，对去除X6的模型进行回归估计，结果为表6 表 6 Dependent Variable: YMethod: Least SquaresDate: 12/14/11 Time: 11:32Sample: 1 31Included observations: 31VariableCoefficientStd. Errort-StatisticProb.C15895.0911783.561.3489210.1895X10.0601880.0521931.1531950.2

22、597X2-135.2491115.3507-1.1725030.2520X3-0.0370620.032985-1.1235940.2719X40.3584920.0804934.4536840.0002X50.2534730.0921492.7506880.0109R-squared0.970777Mean dependent var11628.97Adjusted R-squared0.964932S.D. dependent var2978.791S.E. of regression557.8213Akaike info criterion15.65794Sum squared res

23、id7779116.Schwarz criterion15.93549Log likelihood-236.6981Hannan-Quinn criter.15.74841F-statistic166.0965Durbin-Watson stat1.978266Prob(F-statistic)0.000000各个变量的t检验有显著改善，模型的拟合程度提高，但是X3的经济意义不合理。可能是由多重共线性引起的。3对五元回归模型进行异方差检验残差平方和各个变量之间的散点图如下：因为散点图没有递增或递减的趋势，因此采用怀特检验，结果如表 7表7Heteroskedasticity Test: Whi

24、teF-statistic1.654992Prob. F(20,10)0.2078Obs*R-squared23.80740Prob. Chi-Square(20)0.2509Scaled explained SS8.858907Prob. Chi-Square(20)0.9845Test Equation:Dependent Variable: RESID2Method: Least SquaresDate: 12/14/11 Time: 12:03Sample: 1 31Included observations: 31VariableCoefficientStd. Errort-Stat

25、isticProb.C1.05E+095.56E+081.8935790.0875X1995.335111539.570.0862540.9330X12-0.0138770.029287-0.4738130.6458X1*X2-11.50459114.4724-0.1005010.9219X1*X30.0288010.0363830.7916010.4470X1*X4-0.1280400.060923-2.1016770.0619X1*X50.1154200.0742741.5539770.1512X2-2084773110701581-1.9480980.0800X22103361.6519

26、50.391.9896210.0747X2*X331.6930092.295850.3433850.7384X2*X4-103.823285.07361-1.2203930.2503X2*X579.5949382.062520.9699300.3550X3-2903.1659240.424-0.3141810.7598X32-0.0091450.010231-0.8938730.3924X3*X40.0681770.0386781.7626580.1084X3*X5-0.0764520.051453-1.4858670.1681X410839.288729.0411.2417500.2427X

27、420.0762750.0983940.7751990.4562X4*X5-0.1435570.211632-0.6783360.5129X5-8545.8408462.986-1.0097900.3364X520.0759940.1099660.6910670.5052R-squared0.767981Mean dependent var250939.2Adjusted R-squared0.303942S.D. dependent var272872.4S.E. of regression227657.7Akaike info criterion27.73251Sum squared re

28、sid5.18E+11Schwarz criterion28.70392Log likelihood-408.8539Hannan-Quinn criter.28.04917F-statistic1.654992Durbin-Watson stat2.331770Prob(F-statistic)0.207750由于F的伴随概率较大，不存在异方差。4检验模型的自相关由于DW=1.978266,dL=1.09，dU=1.83, 即dUDW4-dU,模型没有自相关。该模型虽然 F检验均合格，但是t检验不是都合格，且存在多重共线性，X3的回归系数-0.0370620,根据经济理论和经验知，经济意义不

29、符合，故排除该五元模型。 (二)采用逐步回归法选择模型首先进行一元回归，由回归结果得到各解释变量的R2值分别是X1：R2=0.386923,X2:R2=0.229806,X3:R2=0.242735,X4:R2=0.952097X5:R2=0.941490,可以看出，X6:R2=0.097663.X4对Y的影响最大，故先引入X4，再引入X5，得到二元回归模型结果如下： 表8,Dependent Variable: YMethod: Least SquaresDate: 12/14/11 Time: 17:13Sample: 1 31Included observations: 31Variab

30、leCoefficientStd. Errort-StatisticProb.C1911.187468.61214.0783990.0003X40.3492690.0749874.6577260.0001X50.2963960.0832193.5616430.0013R-squared0.967033Mean dependent var11628.97Adjusted R-squared0.964678S.D. dependent var2978.791S.E. of regression559.8365Akaike info criterion15.58493Sum squared resi

31、d8775672.Schwarz criterion15.72371Log likelihood-238.5664Hannan-Quinn criter.15.63017F-statistic410.6675Durbin-Watson stat1.721668Prob(F-statistic)0.000000R2=0.967033，模型的拟合程度提高，t检验显著，继续引入X2,回归结果如下： 表9Dependent Variable: YMethod: Least SquaresDate: 12/14/11 Time: 17:15Sample: 1 31Included observation

32、s: 31VariableCoefficientStd. Errort-StatisticProb.C17471.2811307.541.5450990.1340X40.3804160.0772014.9275920.0000X50.2441090.0902872.7036890.0117X2-151.9561110.3348-1.3772270.1798R-squared0.969197Mean dependent var11628.97Adjusted R-squared0.965774S.D. dependent var2978.791S.E. of regression551.0813

33、Akaike info criterion15.58156Sum squared resid8199646.Schwarz criterion15.76659Log likelihood-237.5141Hannan-Quinn criter.15.64187F-statistic283.1789Durbin-Watson stat1.969738Prob(F-statistic)0.000000R2=0.969197，模型拟合拟合程度进一步提高，去掉x2,继续引入X3,得到回归结果如下： 表10Dependent Variable: YMethod: Least SquaresDate: 1

34、2/14/11 Time: 17:15Sample: 1 31Included observations: 31VariableCoefficientStd. Errort-StatisticProb.C1923.374486.52283.9533080.0005X40.3496570.0764004.5766340.0001X50.2941630.0865063.4005020.0021X30.0015880.0124370.1276750.8994R-squared0.967053Mean dependent var11628.97Adjusted R-squared0.963392S.D

35、. dependent var2978.791S.E. of regression569.9375Akaike info criterion15.64884Sum squared resid8770377.Schwarz criterion15.83388Log likelihood-238.5571Hannan-Quinn criter.15.70916F-statistic264.1653Durbin-Watson stat1.710884Prob(F-statistic)0.000000R2=O.967053,模型的拟合程度有所提高，但是t检验不通过，继续引入X1,得到如下回归结果： 表

36、11Dependent Variable: YMethod: Least SquaresDate: 12/14/11 Time: 18:20Sample: 1 31Included observations: 31VariableCoefficientStd. Errort-StatisticProb.C1993.395492.76714.0453090.0004X40.3474160.0759044.5770780.0001X50.2868830.0856023.3513370.0024X10.0118080.0193680.6096780.5472R-squared0.967481Mean

37、 dependent var11628.97Adjusted R-squared0.963867S.D. dependent var2978.791S.E. of regression566.2253Akaike info criterion15.63578Sum squared resid8656499.Schwarz criterion15.82081Log likelihood-238.3545Hannan-Quinn criter.15.69609F-statistic267.7589Durbin-Watson stat1.677146Prob(F-statistic)0.000000