货币供应量对我国股票交易金额的影响分析一、摘要为了研究货币供应量对我国股票交易金额的影响程度,建立经济模型进行估计检验,用二元回归分析的方法,通过OLS法和广义差分法进行模型修正,从而得出狭义货币供应量和广义货币供应量对股票交易额的影响变化情况。二、关键字:股票交易额狭义货币供应量广义货币供应量三、模型建立与检验由经济理论知,货币流动性与证券市场密切相关,它是货币政策调整的重要工具,它影响着股票价格和交易额,货币供应量越多,实际利率下降,增加了持有货币的机会成本,货币会由货币市场流入资本市场。反之,流通中的货币供应量越少,货币会由资本市场流向货币市场。对我国股票交易额建立股票交易金额与货币供应量的函数关系,Y表示股票市场交易额,X1表示狭义的货币供给量M1,M1=流通中现金+企业活期存款+机关团体部队存款+农村存款+个人持有的信用卡类存款;X2表示广义货币供应量,M2=M1+居民储蓄存款+企业定期存款1.建立模型如下表数据是1995-2008年的时间序列数据,即观测值是连续不同年份中的数据。表一我国股票交易金额与货币供应量资料单位:亿元年份股票交易额狭义货币供应量广义货币供应量1995812323987.160750.51996403628514.876094.919973072234826.390995.319982354438953.7104498.519993132045837.3119897.920006082753147.2134610.320013830559871.6158301.920022799070881.8185007.020033211584118.6221222.820044233495969.7254107.0200531665107278.7298755.7200690469126035.13345603.62007460556152560.1403442.22008267113166217.13475166.6资料来源:《中国统计年鉴》(2002,2009)。对时间序列数据,建立计量经济模型,并进行回归分析。我们假设先建立如下二元回归模型:Y=C+β1X1+β2X2+uiY——股票交易额X1——狭义货币供应量X2——广义货币供应量Ui——随机干扰项根据表一中的数据,利用EVIEWS软件,可得如表二所示结果:DependentVariable:YMethod:LeastSquaresDate:12/06/10Time:15:32Sample:19952008Includedobservations:14VariableCoefficientStd.Errort-StatisticProb.C-115989.052017.46-2.2298090.0475X112.889798.3142611.5503230.1493X2-3.8428802.956650-1.2997410.2203R-squared0.638700Meandependentvar82079.93AdjustedR-squared0.573010S.D.dependentvar126996.2S.E.ofregression82985.04Akaikeinfocriterion25.67812Sumsquaredresid7.58E+10Schwarzcriterion25.81506Loglikelihood-176.7468F-statistic9.722821Durbin-Watsonstat1.005954Prob(F-statistic)0.003701初步方程为:Y=-115989.0+12.88979x1-3.842880x2(-2.23)(1.55)(-1.30)R2=0.6387F=9.7228DW=1.0060模型检验:(一)经济意义检验:X2的符号不符合经济理论的假设,因此经济意义检验不能通过。(二)统计检验:1.拟合优度检验:拟合优度R2=0.6387,修正后的R2=0.5730,拟合效果不是很好,说明还有其他解释变量对被解释变量产生影响。2.T检验:在5%的显著水平下,临界值t0.025(11)=2.201,x1、x2都不能通过t检验,说明在其他解释变量不变的情况下,广义的货币供应量和狭义的货币供应量对股票交易额没有显著影响。3.F检验:在5%的显著水平下,F0.05(2,11)=3.98,F大于临界值,应拒绝原假设,说明回归方程显著。两个解释变量联合起来对被解释变量的影响是显著的。(三)计量检验A.多重共线性检验T检验和F检验综合判断法,F检验通过,但T检验不通过,说明模型很可能存在着多重共线性。相关系数判断法得到相关系数矩阵如下:X1X2X110.998197X20.9981971可以看出:X1、X2之间存在严重的正相关。多重共线性的修正:首先对y和x1进行回归分析的计算结果如下,DependentVariable:YMethod:LeastSquaresDate:10/29/10Time:22:35Sample:19952008Includedobservations:14VariableCoefficientStd.Errort-StatisticProb.C-81374.5045948.36-1.7709990.1019X12.1028890.5131794.0977720.0015R-squared0.583214Meandependentvar82079.93AdjustedR-squared0.548481S.D.dependentvar126996.2S.E.ofregression85335.26Akaikeinfocriterion25.67813Sumsquaredresid8.74E+10Schwarzcriterion25.76942Loglikelihood-177.7469F-statistic16.79173Durbin-Watsonstat1.760149Prob(F-statistic)0.001478写出如下回归分析结果:Y=-81374.50+2.102889X1(-1.77)(4.10)R2=0.5832F=16.79DW=1.76对y和x2进行回归分析结果如下:DependentVariable:YMethod:LeastSquaresDate:12/06/10Time:16:40Sample:19952008Includedobservations:14VariableCoefficientStd.Errort-StatisticProb.C-71165.8845701.29-1.5571960.1454X20.7326190.1875583.9061020.0021R-squared0.559756Meandependentvar82079.93AdjustedR-squared0.523069S.D.dependentvar126996.2S.E.ofregression87703.79Akaikeinfocriterion25.73288Sumsquaredresid9.23E+10Schwarzcriterion25.82418Loglikelihood-178.1302F-statistic15.25763Durbin-Watsonstat1.892712Prob(F-statistic)0.002087写出如下回归分析结果:Y=-71165.88+0.732619X2(-1.56)(3.91)R2=0.5598F=15.26DW=1.89显然,y与X1的R2、修正后的R2以及t值都比y与x2的好,说明一元回归最佳模型应选取x1为自变量能更好的说明问题。但是它仍然没有二元回归的模拟拟合度高。B.异方差检验1、图形检验法0.E+002.E+104.E+106.E+10050000100000150000200000X1E20.E+002.E+104.E+106.E+100100000200000300000400000500000X2E2从图可以看到,随着X1、X2的增加,e2有增加的趋势2、利用怀特检验法,可以得到如下结果:WhiteHeteroskedasticityTest:F-statistic35.44328Probability0.000030Obs*R-squared13.39530Probability0.019943TestEquation:DependentVariable:RESID^2Method:LeastSquaresDate:12/06/10Time:17:51Sample:19952008Includedobservations:14VariableCoefficientStd.Errort-StatisticProb.C4.27E+092.24E+091.9021780.0937X1-2988560.1338014.-2.2335790.0560X1^2164.9588128.25981.2861300.2344X1*X2-108.377090.89867-1.1922830.2673X21078397.500190.42.1559730.0632X2^217.6572816.042071.1006850.3030R-squared0.956807Meandependentvar5.41E+09AdjustedR-squared0.929812S.D.dependentvar6.73E+09S.E.ofregression1.78E+09Akaikeinfocriterion45.73809Sumsquaredresid2.54E+19Schwarzcriterion46.01197Loglikelihood-314.1666F-statistic35.44328Durbin-Watsonstat1.159629Prob(F-statistic)0.000030可以看到,f值和卡方检验的p值都小于0.05,拒绝原假设的,原假设是同方差,所以结果表示存在着异方差。异方差的修正:(加权最小二乘法)选用权数w=1/e2DependentVariable:YMethod:LeastSquaresDate:12/06/10Time:18:34Sample:19952008Includedobservations:14Weightingseries:WVariableCoefficientStd.Errort-StatisticProb.X1-8.8981401.577200-5.6417340.0002X23.7067830.5308036.9833440.0000C-15367.9610313.69-1.4900550.1643WeightedStatisticsR-squared0.999242Meandependentvar83661.60AdjustedR-squared0.999104S.D.dependentvar217324.2S.E.ofregression6506.171Akaikeinfocriterion20.58630Sumsquaredresid4.66E+08Schwarzcriterion20.72324Loglikelihood-141.1041F-statistic7246.848Durbin-Watsonstat1.605515Prob(F-statistic)0.000000UnweightedStatisticsR-squared0.365105Meandependentvar82079.93AdjustedR-squared0.249670S.D.dependentvar126996.2S.E.ofregression110006.1Sumsquaredresid1.33E+11Durbin-Watsonstat2.119833可以看到,R2=0.9992,修正后的R