在计量经济学的Endogeneity：有助易变的估计t .ppt

《在计量经济学的Endogeneity：有助易变的估计t .ppt》由会员上传分享，免费在线阅读，更多相关内容在PPT专区-天天文库。

1、EndogeneityinEconometrics:InstrumentalVariableEstimationMingLUEndogeneityOmittingvariablebiasSimultaneityMeasurementerrorCanweignoretheomittedvariablesbias?Itcanbesatisfactoryiftheestimatesarecoupledwiththedirectionofthebiasesforthekeyparameters.Canweuseprox

2、ytoeliminateomittedvariablebias?–Sometimes.CanFEestimationsolveomittingvariableproblem?Firstdifferencingorfixedeffectsestimationeliminatestime-constantvariables.Inaddition,thepaneldatamethodsdonotsolvetheproblemoftime-varyingomittedvariablesIdeaofIVEstimation

3、Exogenousvariable.IndirecteffectsofIV.ExampleWhatcanserveasIVforedu?Mother’seducation?Numberofsiblings?Thereportofothers?Adummyvariablethatisequalto1ifamanisborninthefirstquarteroftheyear.AngristandKrueger(1991).(Problematic.)InChina,theyearsofprimaryedu?IVfo

4、rskippedclass?Thedistancefromhometoschool.OtherexamplesofIVIVforinstitution:Language?History?Mauro(1995)使用人口的种族和语言构成作为腐败的工具变量，HallandJones(1999)用距离赤道的距离和以西欧语言为第一语言的程度作为制度质量的工具变量，LaPortaetal.(1997,1998,1999)把法律的起源作为各种法律结构的工具变量。Acemoglu,Johnson,andRobinson(2001

5、,2002)使用殖民地时代(1500年前后)的死亡率和人口密度作为制度的工具变量IVforschoolchoice:Numberofsteams?IdentificationReferto(15.9)and(15.10)The(asymptotic)standarderrorofSSTisthetotalsumofsquaresofthexiSelf-selectionAngrist(1990)studiedtheeffectthatbeingaveteranintheVietnamwarhadonlifetim

6、eearnings.DraftlotterynumberisagoodIVcandidateforveteran.SomeadditionalwordsaboutnaturalexperimentandDIDPropertiesofIVwithaPoorInstrumentalVariablePoorIVcancauseseriousbias.R2MostregressionpackagescomputeanR-squaredafterIVestimation,usingthestandardformula:R2

7、=1-SSR/SST,whereSSRisthesumofsquaredIVresiduals,andSSTisthetotalsumofsquaresofy.R2canbenegativeinthiscase.IVESTIMATIONOFTHEMULTIPLEREGRESSIONMODELstructuralequationEstimationEfficientIVEquation(15.26)isanexampleofareducedformequation,whichmeansthatwehavewrit

8、tenanendogenousvariableintermsofexogenousvariables.TWOSTAGELEASTSQUARES2SLSinwordsThefirststageistoruntheregressionin(15.36),whereweobtainthefittedvaluesyˆ2.ThesecondstageistheOLSregressi