计量经济学论文薪资微观影响因素的计量分析打印.docx

1、计量经济学论文薪资微观影响因素的计量分析打印计量经济学课程论文某国薪资影响因素的计量分析摘要本文主要运用OLS采取数据对工人工资的微观因素分析。由此得出 影响薪资最主要的因素是工作经验，以帮助大学生在择业就业时了解 自己的优势劣势，及时增强自己的能力，增加工作经验，以求在职场 中获得更高薪资和更好的表现。AbstractThis paper mainly uses the OLS,take the analysis of data on the micro factors workers wages. Con clusi on the main in flue nee factors of s

2、alary is working experience,to help students understand their own adva ntages and disadva ntages in the employme nt,to enhance their ability,work experience,in order to get higher pay and better performa nee in the workplace关键词薪资影响因素回归分析一.弓I言我国大学扩招后，大学生就业难的问题已经是一个不争的现象， 且有可能越来越难的趋势。这个方面和国际经济形式近 3年来连

3、遭打 击，一方面和中国经济结构体制和教育改革落后有关， 更和当今大学生的就业观滞后有关。据统计，2013年全国高校毕业生将超过 700 万，这些高校学子的就业问题成为社会和学校关注的焦点。 那么我们通常关注的工作的薪水受自身的什么因素的影响呢？就此问题我搜集了关于薪水影响因素的数据，并且运用 Eviews3.0进行多元回归分二、数据搜集本文所采用数据均来自于薛薇-基于SPSS的数据分析Employee data，真实性和权威性很高。三、计量经济模型（一）模型的建立Y =內+ B2X2+ （3 3X3+ B4X4+ B5X5+ 伍X6+ P7X7+ 伍X&+U其中：Y现在薪资（美元/年），X2

4、性别X 3教育程度X 4 年龄X5 初 始工作工资X 6 工作时间 X 7工作经验 X 8行业类别U 随 机扰动项夫一性别，1代表男性，2代表女性（虚拟变量）X3教育程度，以年为单位，表示学习时间的长短X7工作经验，以月为单位，表示过去工作的时间长短X6工作时间，从被雇佣开始工作的时间X8行业类别，1表示管理者，2表示非管理者（虚拟变量）Dependent Variable: YMethod: Least SquaresDate: 11/03/13 Time: 15:36Sample: 1 471Included observations: 470Excluded observations:

5、1VariableCoeffici entStd.Errort-StatisticProb.X22384.251331.5970 -84.136131.335128151.8583-9.13708784.82123.0379550.0025X3159.68732.0765400.0384X448.88423-1.7211300.0859X50.07439317.947070.0000X632.579344.6611850.0000X75.630314-1.622830.105388X811488.07-3936.1501393.907&2416310.0000C3577.955-1.10011

6、20.2719R-squared0.83816Mean34491.339dependent varAdjusted0.83571S.D.dependent17119.69R-squared7varS.E. of regression6938.92Akaike info20.544566criterionSum squared2.22E+1Schwarz20.61524resid0criterionLog likelihood-4819.9711.88828F-statistic341.8326Durbin-WatsonProb(F-statistic)0.000000stat3由上表，模型估计

7、有以下结果Y二-3936.150+ 2384.251X2+331.5970X3-84.13613X4+1335128X5 +151 8583X6-9.137088X7+11488O *8+Use= ( 3577.955) (784.8212) (159.6873) (48.88423) (0.074393)(32.57934) (5.630314) (1393.907)t= (-1.100112) (3.037955) (2.076540! (-1.721130) (17.94707)(4.661185 (-1.622838) (&241631)R2=0.838169 Adjusted R2=

8、0.835717 F-statistic二 341.8326 ,n=471(2)参数估计的检验与修正由上表，该模型的可决系数较高，F检验值=341.8326,明显显著。 除X7所有变量的符号也和预期效果一致，说明，但 a =0.05时，t(471-7)=1.9,，只有X2和X8的系数的t检验显著，这表明很可能 存在多重共线性。尽管回归拟合的很好，但是解释变量的t统计量多 个不显著，X7工作经验的系数符号和经济意义相反，也表明模型中 解释变量确实存在多重共线性。(1)多重共线性的检验Stepl.计算各变量的相关系数。相关系数矩阵X2X3X4X5X6X7X8X210.35507640-0.0447

9、91290.456769850.073620590.169619100.3168719801331896742542705526183659X30.355076401-0.281159640.633194640.04915221-0.2512 佃540.60718718013388989462017772248873X4-0.04479129-0.281159641-0.007889480.052798320.80394398-0.0857662418967889898987472354242214236X50.456769850.63319464-0.007889481-0.02265064

10、0.046450230.782384434254946298987469899309238487X60.073620590.049152210.05279832-0.0226506410.00064724-0.0043018227055017777235469899788327574373X70.16961910-0.2512 佃540.803943980.046450230.000647241-0.081802972618224824223092378832794078X80.316871980.60718718-0.085766240.78238443-0.00430182-0.08180

11、2971365987314236848757437394078可以看出多个变量之间存在多重共线性。Step2.采用逐步回归法，来检验和解决多重共线性问题。分别做 Y对x2-x8的一元回归，结果如下元回归结果变量X2X3X4X5X6X7X8参数估计15482.93909.69-209.53821.907769133.443-15.849035885.4值34165T统计量10.956519.0798-3.14626840.136441.69609-2.1138829.293956728可决系数0.203790.437000.0207140.7745120.006090.009430.646607

12、2688修正可决 系数0.202100.435800.0186210.7740310.003970.007320.6458502764其中，加入X5的修正的可决系数最大，以 X5为基础，顺次加入 其他变量逐次回归。加入新变量的回归结果 1变量X5X5 X20.776619X5X2X30.7929X5X2X3X40.8020X5X2X3X4X60.81032X5X2X3X4X6X70.811971X5X2X3X4X6X7X80.835717经比较，新加入变量X3后，方程的修正的可决系数为改进最大， 且各参数t检验显著，所以选择保留 X3X8再加入其他新变量逐步 回归。加入新变量的回归结果 2变量

13、X5X3X8X2X4X6X7Adjusted R2X5,X3,X81.290676 736.25501751.16 -0.818020X5X3,X8X21.220554681.533912188.172292.555-0.819954X5X3,X8X2X41.290785442.227211807.882243.606-134.82290.819954X5X3,X8X2X4X61.325221369.557111700.961925.569-147.602156.7427-0.835137X5X3,X8X2X4X6X71.335128331.597011488.072384.251-84.13

14、613151.8583-0.835717经比较，新加入变量后，尽管方程的修正的可决系数都有较大改进， 参数X4, X7t检验不显著，且使原有变量的t检验值也向不显著方向 发展，所以说明X4 X7、X2引起了严重的多重共线性，应予剔除， 使模型得到改善。Step3.因此，剔除了多重共线性后的模型为丫=价+ B3X3+ B5X5+ 血X8+ 36X6 + U再次经过回归，结果为Depe ndent Variable: 丫Method: Least SquaresDate: 11/03/13 Time: 20:40Sample: 1 471In eluded observati ons: 471Va

15、riable Coeffieie Std. Error t-Statistie Prob.ntX3684.2104152.28284.4930240.0000X51.3091640.07112218.407240.0000X811702.431421.7338.2311010.0000X6150.238733.168644.5295390.0000C-11332.353224.420-3.5145380.0005R-squared0.825694Mea ndependent34492.3var0Adjusted R-squared0.824198S.D. dependent var17101.

16、4820.6038S.E. of regressi on7170.430Akaikeinfocriteri on8Sum squared resid2.40E+10Schwarz criteri on20.64798551.866Log likelihood-4847.21F-statistic36Durbi n- Watson stat1.857537Prob(F-statistic)0.000000Y = -11332.35 + 684.2104* X3+ 1.309164* X5+ 11702.43* X8+ 150.2387* X6(-3.514538) (4.493024) (18.

17、40724) (8.231101)(4.529539)R2=0.825694 DW= 1.857537 F=551.866(2)异方差的检验(White检验)Stepl.相关图形分析1015Q0 02 04 皿 Of MH从这两个图可以粗略看出，随 X3和X5的增加，Y的离散程度有 稍微逐步变大的趋势，是否存在异方差还不能判断。Step2.由于是多元的回归，所以采取含交叉项的 White检验White Heteroskedasticity Test:F-statisticObs*R-squared5.54153264.13656ProbabilityProbability0.0000000.0

18、00000Test Equati on:Depe nde nt Variable: RESIDA2Method: Least SquaresDate: 11/03/13 Time: 21:05Sample: 1 471In eluded observati ons: 471VariableCoefficie ntStd. Errort-StatisticProb.C3.17E+086.53E+080.4850590.6279X5-15447.0316063.10-0.9616470.3367X5A2-0.3373300.109760-3.0733350.0022X5*X3204.9039102

19、1.5400.2005830.8411X5*X815068.246389.6252.3582350.0188X5*X6329.9374174.88291.8866200.0598X311482926381272550.3011740.7634X3A263893.821258502.0.0507700.9595X3*X8-3399660623420222-1.4515920.1473X3*X6-154541.3336894.1-0.4587240.6467X84.04E+084.41E+080.9145190.3609X8*X6-1548520.3236506.-0.4784540.6326X6

20、-7964860.13481353-0.5908060.5549X6A234842.5878728.460.4425660.6583R-squared0.136171Mean depe ndent var50869257Adjusted R-squared0.111598S.D. dependent var1.65E+08S.E. of regressi on1.55E+08Akaike info criterion40.58625Sum squared resid1.10E+19Schwarz criteri on40.70974Log likelihood-9544.061F-statis

21、tic5.541532Durbin-Wats on stat1.847878Prob(F-statistic)0.000000由上表，Obs*R-squared概率0.05，拒绝原假设，表明模型存在异方差。Step3 .消除异方差采用加权最小二乘法（WLS对异方差进行修正。经过尝试，选用的权数为w=1/x5最为合理。用权数的回归结果为再次回归的结果为:Depe ndent Variable: YMethod: Least SquaresDate: 11/04/13 Time: 21:14Sample: 1 471In cluded observati ons: 471Weighti ng se

22、ries: W2VariableCoefficientStd. Errort-StatisticProb.X51.6872700.09509017.744000.0000X3398.335997.815434.0723220.0001X89654.8361904.6115.0691910.0000X6114.846121.593135.3186430.0000C-10179.172233.259-4.5579860.0000Weighted StatisticsR-squared0.741192Mean depe ndent var27948.34Adjusted R-squared0.738

23、970S.D.dependent var9809.63719.8875S.E. of regressi on5011.848Akaike infocriteri on6Sum squared resid1.17E+10Schwarz criteri on19.93166333.64040.000000Log likelihood-4678.519F-statisticDurbin-Wats on stat1.931204Prob(F-statistic)Un weightedStatisticsR-squared0.809798Mean depe ndent var34492.3017101.

24、482.61E+1Adjusted R-squared0.808165S.D.dependent varS.E. of regressi on7490.273Sum squared resid再进行含交叉项的White HeteroskedastWhite检验icity Test:F-statistic0.783391Probability0.678086Obs*R-squared10.26727Probability0.671946Durb in-Wats on stat1.909834Test Equati on:Depe ndent Variable: STD_RESIDA2Method

25、: Least SquaresDate: 11/04/13 Time: 21:15Sample: 1 471In eluded observati ons: 471VariableCoefficientStd. Errort-StatisticProb.C-2.12E+0Q3.67E+08-0.5774260.5639X58-5553.8619033.223-0.6148260.5390X5A20.0178380.0617250.2890010.7727X5*X3126.7006574.47180.2205510.8255X5*X82056.0253593.2610.5721890.5675X

26、5*X6-7.81360498.34690-0.0794490.9367X31712651.214411950.0798770.9364X3A2274770.0707729.90.3882410.6980X3*X8-9847646.13170567-0.7477010.4550X3*X6-83172.82189455.4-0.4390100.6609X81.36E+082.48E+080.5498260.5827X8*X6-185780.51820077.-0.1020730.9187X66262947.7581357.0.8260980.4092X6A2-28416.5044273.64-0.6418380.5213R-squared0.021799Mean depe ndent var248519708695731239.4350Adjusted R-squared-0.006027S.D.dependent var