影响银行卡交易额的因素

中国银行界开始进行银行卡业务并没有太长的时间，但是银行卡业务的发展在整个银行体系中有着十分重要的作用，鉴于此，我们决定分析一下是哪些因素在影响银行卡业务的交易额。为了找出影响银行卡交易额的因素，我们选择了工商银行，农业银行，中国银行，建设银行，招商银行六家银行的数据（中国金融年鉴1996——2000）。设模型为Y=β1+β2X1+β3X2+β4X3+β5X4+β6X5+β7X6其中，Y——银行卡业务交易额（万元）X1——发卡机构（个）X2——发卡量（张）X3——特约商户（个）X4——取现网点（个）X5——ATM机（台）X6——POS机（台）表1obsYX1X2X3X4X5X614027090029019221730694532450335653079221315000034646138735426335513102021248328890000495471000053044118201934165954205050005021087590046600217002687161655679600483488149627145697240762000104652563531268047975615000030830088814814902762054994657083042800038788390377159140522475238149927150000446809445661841128472454717810144632003052065830049774231104045302481127120059109526112204205129562191264001315465899901461382412137691800030743885083895952898869175631814475500034014986461819564273933364961015251100003301277251168182139652859268941622670500303310979934549121179456930063175395006532581861589827207089141181960014337889135.9616325526891980770000300548377479752929519828356834206731341032522125836857584225035725078221267319001631774821871871139303206386632228561800296492471124931924779571739503234595600.377669239331997027831115127522423787000136426048639918343614006首先用OLS进行模型估计表2DependentVariable:YMethod:LeastSquaresDate:12/12/03Time:23:02Sample:124Includedobservations:24VariableCoefficientStd.Errort-StatisticProb.C-3680754.6063061.-0.6070790.5518X1-15105.4243742.79-0.3453240.7341X20.5001830.6317890.7916930.4394X3588.1427357.52131.6450560.1183X4-416.5884538.2219-0.7740090.4496X51143.4925413.3740.2112350.8352X612.18085764.49840.0159330.9875R-squared0.760328Meandependentvar24739109AdjustedR-squared0.675738S.D.dependentvar24354861S.E.ofregression13868623Akaikeinfocriterion35.96665Sumsquaredresid3.27E+15Schwarzcriterion36.31025Loglikelihood-424.5998F-statistic8.988400Durbin-Watsonstat1.811537Prob(F-statistic)0.000164由上表可以看出，模型拟合还好，但是T检验都不显著，并且有些系数还出现了与经济意义相背离的现象。说明所选模型存在问题，必须进行修正。检验是否有多重共线性。表3X1X2X3X4X5X6X110.368973970.758389210.717380000.579503820.51270456X20.3689739710.662649980.536521530.924538420.81155867X30.758389210.6626499810.848732050.834731340.89361429X40.717380000.536521530.8487320510.703326820.83280296X50.579503820.924538420.834731340.7033268210.89090771X60.512704560.811558670.893614290.832802960.890907711可以看出除X1与X2的相关性较小，其余解释变量之间存在相关关系较大。选出X1，X2，X3，X4，X5，X6中对Y影响较显著的X3进行辅助回归。表4DependentVariable:X3Method:LeastSquaresDate:12/13/03Time:01:24Sample:124Includedobservations:24VariableCoefficientStd.Errort-StatisticProb.C-2125.9793965.646-0.5360990.5985X181.8070221.443953.8149240.0013X2-0.0006800.000384-1.7703030.0936X4-0.3867300.342924-1.1277420.2742X53.2263703.4869010.9252830.3671X61.5314470.3517524.3537740.0004R-squared0.935210Meandependentvar47634.75AdjustedR-squared0.917213S.D.dependentvar31777.10S.E.ofregression9143.135Akaikeinfocriterion21.29171Sumsquaredresid1.50E+09Schwarzcriterion21.58623Loglikelihood-249.5005F-statistic51.96433Durbin-Watsonstat2.338614Prob(F-statistic)0.000000因为F=(O.935210/(6-1))/((1-0.93521O)/(24-6))=51.96413，而查表F0.05（5,18）=2.77，51.96413显著大于2.77，所以可判断模型存在多重共线性。首先用Y对X3进行单独回归。表5DependentVariable:YMethod:LeastSquaresDate:12/13/03Time:01:20Sample:124Includedobservations:24VariableCoefficientStd.Errort-StatisticProb.C-4253338.5649935.-0.7528120.4595X3608.640799.308616.1287810.0000R-squared0.630637Meandependentvar24739109AdjustedR-squared0.613847S.D.dependentvar24354861S.E.ofregression15134396Akaikeinfocriterion35.98249Sumsquaredresid5.04E+15Schwarzcriterion36.08066Loglikelihood-429.7899F-statistic37.56196Durbin-Watsonstat1.971714Prob(F-statistic)0.000004可以看出X3单独对Y的影响显著。用Y对X2X3进行回归表6DependentVariable:YMethod:LeastSquaresDate:12/13/03Time:01:21Sample:124Includedobservations:24VariableCoefficientStd.Errort-StatisticProb.C-4298503.4882566.-0.8803780.3886X20.6721450.2311022.9084330.0084X3387.7936114.59053.3841690.0028R-squared0.736697Meandependentvar24739109AdjustedR-squared0.711621S.D.dependentvar24354861S.E.ofregression13078790Akaikeinfocriterion35.72735Sumsquaredresid3.59E+15Schwarzcriterion35.87461Loglikelihood-425.7282F-statistic29.37806Durbin-Watsonstat1.909223Prob(F-statistic)0.000001X2，X3联合对Y的影响比X2，X3单独对Y的影响大。加入X1进行OLS，得表7DependentVariable:YMethod:LeastSquaresDate:12/13/03Time:01:22Sample:124Includedobservations:24VariableCoefficientStd.Errort-StatisticProb.C-3159302.5160254.-0.6122380.5473X1-20483.4227193.55-0.7532450.4601X20.6221040.2427862.5623600.0186X3483.2526171.66172.8151440.0107R-squared0.743961Meandependentvar24739109AdjustedR-squared0.705555S.D.dependentvar24354861S.E.ofregression13215626Akaikeinfocriterion35.78271Sumsquaredresid3.49E+15Schwarzcriterion35.97905Loglikelihood-425.3925F-statistic19.37102Durbin-Watsonstat1.951553Prob(F-statistic)0.000004拟合并没有显著变优，同时，X1的符号与经济意义相反，所以去除X1。用同样的做法可以去除X4,X5,X6。因此，模型为Y=—4298503+0.672145X2+387.7936x3（式1）(4882566)(0.231102)(114.5905)t=(-0.880378)(2.908433)(3.384169)R^2=0.736697df=21进行异方差检验。ARCH检验表8ARCHTest:F-statistic0.162997Probability0.919791Obs*R-squared0.587160Probability0.899366TestEquation:DependentVariable:RESID^2Method:LeastSquaresDate:12/13/03Time:01:25Sample(adjusted):424Includedobservations:21afteradjustingendpointsVariableCoefficientStd.Errort-StatisticProb.C1.80E+148.59E+132.1002250.0509RESID^2(-1)-0.0083670.246741-0.0339110.9733RESID^2(-2)0.0471040.2479600.1899640.8516RESID^2(-3)-0.1721790.251874-0.6835920.5034R-squared0.027960Meandependentvar1.64E+14AdjustedR-squared-0.143576S.D.dependentvar2.66E+14S.E.ofregression2.84E+14Akaikeinfocriterion69.57035Sumsquaredresid1.38E+30Schwarzcriterion69.76931Loglikelihood-726.4887F-statistic0.162997Durbin-Watsonstat1.909327Prob(F-statistic)0.919791WHITE检验表9WhiteHeteroskedasticityTest:F-statistic1.334828Probability0.293323Obs*R-squared5.264877Probability0.261183TestEquation:DependentVariable:RESID^2Method:LeastSquaresDate:12/13/03Time:01:26Sample:124Includedobservations:24VariableCoefficientStd.Errort-StatisticProb.C5.62E+131.12E+140.5007280.6223X221313860158366241.3458590.1942X2^2-0.3692220.274141-1.3468340.1939X3-7.49E+097.07E+09-1.0597130.3026X3^290907.9071582.571.2699720.2194R-squared0.219370Meandependentvar1.50E+14AdjustedR-squared0.055027S.D.dependentvar2.51E+14S.E.ofregression2.44E+14Akaikeinfocriterion69.27764Sumsquaredresid1.13E+30Schwarzcriterion69.52307Loglikelihood-826.3317F-statistic1.334828Durbin-Watsonstat2.251078Prob(F-statistic)0.293323由以上两种检验联合判断式1可能不会存在异方差，但是由于所用数据为截面数据，所以，也有必要进行修正。进行修正：首先用GENR生成数据LY=logy，LX2=logx2，LX3=logx3，数据如下：表10LYLX2LX3117.5111416.7715511.14841216.3919315.3445810.90160317.1790115.3652010.87888416.8361816.2020610.74936513.4292612.762299.17232767.60090213.050345.866468717.8435417.2196611.30824817.2308715.9946911.17872917.1168915.9066911.032321016.4871216.8436310.815251112.5106113.906509.409519128.76405314.251566.8977051318.1582517.5970911.403051415.3747116.5226611.313941517.0387816.3628111.129941616.9365817.2526510.725271713.1984014.996689.673949189.88328515.033063.5824071918.2071217.8198911.487912018.0248716.9122611.359282117.1013716.6918011.182632217.1675817.7123610.806062315.3406115.750499.9019862416.9846515.675878.763897将以上数据用OLS得出：表11DependentVariable:LYMethod:LeastSquaresDate:12/13/03Time:01:17Sample:124Includedobservations:24VariableCoefficientStd.Errort-StatisticProb.C-7.4912002.933308-2.5538400.0185LX20.8317970.2327053.5744680.0018LX30.9738900.1628715.9795010.0000R-squared0.862135Meandependentvar15.51324AdjustedR-squared0.849005S.D.dependentvar3.037069S.E.ofregression1.180147Akaikeinfocriterion3.285624Sumsquaredresid29.24769Schwarzcriterion3.432880Loglikelihood-36.42748F-statistic65.66139Durbin-Watsonstat1.676656Prob(F-statistic)0.000000可以得出模型拟合较好，查德宾表——沃林d统计量得,dL=1.188,dU=1.546,，因为，1.1676656大于du小与2,所以，该模型不存在自相关。所以，可以得出：模型应该为LY=—7.4912+0.831797LX2+0.97389LX3。(2.933308)(0.232705)(0.162871)t=(-2.553840)(3.574468)(5.979501)R^2=0.862135df=21由所建模型来分析可以看到银行卡业务的交易额主要与发卡量总数和特约商户个数有关，至于其它因素，如发卡机构，取现网点，ATM机和POS机台数对它的影响都并不是很大。现实生活中应该也是如此，银行卡的张数直接影响着银行卡业务的交易额，因为只有拥有了银行卡，才有可能用银行卡进行消费。而一般来说，人们在进行交易时，如果是该商家是特约商户，那么会直接使用银行卡，而免去了用现金的麻烦。而现在，人们不太习惯使用银行卡直接购物；使用它进行购物时的交易额一般都会很大。而随着时间的推移，人们会越来越减少直接的现金消费，而倾向于使用银行卡直接消费。因此，银行在发展银行卡业务的时候应该着重于卡的发行以及特约商户的增加。而不应该盲目地增设ATM机，POS机，对资源作一些不必要的浪费。 报告 软件系统测试报告下载sgs报告如何下载关于路面塌陷情况报告535n,sgs报告怎么下载竣工报告下载 的不足之处在于，由于银行卡业务发展时间不是很长，并不成熟，所以使用了24组混合数据。但银行卡业务交易额与它的影响因素的关系还是有迹可循。相信如果在10年以后再来做同样的分析，会得出有更好的效果。2