School Entry Educational Attainment and

Kevin Langz

Description

Niveau: Supérieur, Doctorat, Bac+8
School Entry, Educational Attainment and Quarter of Birth: A Cautionary Tale of LATE Rashmi Baruayand Kevin Langz August 2011 Abstract Partly in response to increased testing and accountability, states and dis- tricts have been raising the minimum school entry age, but existing studies show mixed results regarding the e?ects of entry age. These studies may be severely biased because they violate the monotonicity assumption needed for LATE. We propose an instrument not subject to this bias and show no e?ect on the educational attainment of children born in the fourth quarter of mov- ing from a December 31 to an earlier cuto?. We then estimate a model that reconciles the di?erent IV estimates including ours. We ?nd that one standard instrument is badly biased but that the other diverges from ours because it estimates a di?erent LATE. We also ?nd that an early entry age cuto? that is applied loosely (as in the 1950s) raises educational attainment but one that is strictly enforced lowers it. We are grateful to Josh Angrist, Garry Barrett, Sandy Black, Jim Heckman, Caroline Hoxby, Claudia Olivetti, Daniele Paserman and participants in seminars at the Australian National Uni- veristy, Boston University, University of Chicago, MIT, UC Irvine, NYU, Pomona College, Singapore Management University, Tilburg University, the Tinbergen Institute, the University of Wollongong, SOLE and Econometric Society for helpful comments and suggestions.

• school entry


• studies ?nd


• educational attainment


• mum age


• ?nding no


• reduces human capital


• no child


• age

Sujets

• Tilburg University
• Barrett
• Singapore Management University
• Olivetti
• Educational attainment

Informations

• Publié par: profil-zyak-2012
• Nombre de lectures: 12
• Langue: English

Extrait

SchoolEntry,EducationalAttainmentand QuarterofBirth:ACautionaryTaleofLATE ∗ RashmiBarua † andKevinLang ‡ August2011 Abstract Partlyinresponsetoincreasedtestingandaccountability,statesanddis- trictshavebeenraisingtheminimumschoolentryage,butexistingstudies showmixedresultsregardingtheeffectsofentryage.Thesestudiesmaybe severelybiasedbecausetheyviolatethemonotonicityassumptionneededfor LATE.Weproposeaninstrumentnotsubjecttothisbiasandshownoeffect ontheeducationalattainmentofchildrenborninthefourthquarterofmov- ingfromaDecember31toanearliercutoff.Wethenestimateamodelthat reconcilesthedifferentIVestimatesincludingours.Wefindthatonestandard instrumentisbadlybiasedbutthattheotherdivergesfromoursbecauseit estimatesadifferentLATE.Wealsofindthatanearlyentryagecutoffthatis appliedloosely(asinthe1950s)raiseseducationalattainmentbutonethatis strictlyenforcedlowersit. ∗ WearegratefultoJoshAngrist,GarryBarrett,SandyBlack,JimHeckman,CarolineHoxby, ClaudiaOlivetti,DanielePasermanandparticipantsinseminarsattheAustralianNationalUni- veristy,BostonUniversity,UniversityofChicago,MIT,UCIrvine,NYU,PomonaCollege,Singapore ManagementUniversity,TilburgUniversity,theTinbergenInstitute,theUniversityofWollongong, SOLEandEconometricSocietyforhelpfulcommentsandsuggestions.Theusualcaveatapplies. BaruaacknowledgesfundingunderNSF-AERAgrantREC-0634035.Langacknowledgesfunding fromNICHDundergrantnumberR03HD056056-01andNSFundergrantnumberSEC-0339149. Theopinionsexpressedherearethoseoftheauthorsandnotnecessarilythoseofthefundingagen- .seic † [e-mail],SchoolofEconomics,SingaporeManagementUniversity ‡ [e-mail],DepartmentofEconomics,BostonUniversityandNBER 1

1Introduction Overthelastfourdecadesmanystatesandschooldistrictshaveincreasedthemini- mumageatwhichchildrenmayenterkindergarten.Inthe1960schildrenfrequently enteredkindergartenwhentheywereconsiderablylessthanfiveyearsold(orfirst gradewhentheywerelessthansixyearsold).Thiswasformallypermittedinmany stateswhereas,inotherstates,itwasrelativelyeasytogetaroundtherules.Today, thirty-eightstateshavecutoffdatesrequiringchildrenenteringkindergartentobe fiveyearsoldbeforeOctober16oftheyearinwhichtheyenterkindergarten,and someoftheremainingstateshavedistrictsthatapplyastricterstandard. Whetherdelayedentryimprovesorworsenseducationoutcomesiscontroversial. Manyrecentstudiesinbotheducationandeconomicshavebeendevotedtoobtain- ingconsistentestimatesoftheeffectsofschoolentryageonshortrunandlongrun outcomes.AngristandKrueger(1992)addressthepotentialendogeneityofentryage byusingquarterofbirthasaninstrumentforentryage.They(AngristandKrueger, 1991)showthathistoricallyindividualsborninthefirstquarterstartedschoollater thanthoseborninthefourthquarter,completedlesseducationandearnedlessthan thosebornintherestoftheyear. 1 Criticsofthisapproacharguethatquarterof birthmaybedirectlyrelatedtostudentoutcomesorparentalsocioeconomicstatus. 2 BucklesandHungerman(2008)provideevidencethatchildrenbornatdifferenttimes intheyearareconceivedbywomenwithdifferentsocioeconomiccharacteristics.To addressthisissue,severalresearchershaveexploitedthevariationinstatelawsgov- erningentryage(or,the“legalentryage”)toidentifyitseffectontestscores,wages, educationalattainmentandotheroutcomes. 3 However,sinceentryagedependson bothstatelawanddateofbirth,thepotentialendogeneityofdateofbirthremains problematicforthisapproach. Theaimofthispaperisthreefold.First,weaddresscertainunder-appreciated issuesintheinstrumentalvariableliterature.ImbensandAngrist(1994),Angristand Imbens(1995)andAngrist,ImbensandRubin(1996)showthatwithheterogeneous treatmenteffects,undercertainconditions,IVidentifiestheLocalAverageTreatment 1 SeealsoMayerandKnutson,1999andCahanandCohen,1989. 2 Bound,JaegerandBaker,1995;BoundandJaeger,2000. 3 AllenandBarnsley,1993;BedardandDhuey,2006,2008;CascioandLewis,2006;Datar,2005; DobkinandFerreira,2010;ElderandLubotsky,2009;FertigandKluve,2005;Fredrikssonand Ockert,2006;PuhaniandWeber,2007. 2

Effect(LATE).Onecondition,termed“monotonicity,”generallytreatedasanunim- portantregularitycondition,requiresthatwhiletheinstrumentmayhavenoeffecton someindividuals,allofthosewhoareaffectedshouldbeaffectedunidirectionally.We arguethatbothstandardinstruments,quarterofbirthandlegalentryage,maypro- videinconsistentestimatesofLATEbecausetheyviolatemonotonicity. 4 Therefore, weproposeaninstrumentthatsatisfiesmonotonicityandgivesconsistentestimatesof theLATEofschoolentryageoneducationalattainment.Ourtwo-sampletwo-stage leastsquares(TS2SLS)results,consistentwithAngristandKrueger,showalarge negativeeffectofschoolentryageoneducationalattainmentwhentheIVisquarter ofbirth.Usingthe“legalentryage”instrumentyieldsasmallerbutstillsubstantial adverseeffectthoughitfallsshortofstatisticalsignificanceatconventionallevels.Fi- nally,whenweusetheconsistentestimatorthatmeetsthemonotonicityrequirement, theeffectofschoolentryageoneducationalattainmentisveryclosetozero. ComparingthedifferentIVestimatesdoesnottelluswhethertheydivergebecause thetraditionalestimatorsareinconsistentorbecausetheyaremeasuringdifferent LATEs.Asecondaimofthispaperistoaddressthisissueandtoreconcilethe differentIVestimatesincludingours.Toachievethis,wedevelopasimplemodelof schoolentryandeducationalattainmentanduseindirectinferencetoestimatethe parametersofthemodel.WethensimulateourmodeltocomparetheIVestimates withtheirrespectiveLATEs.“Quarterofbirth”provestoberobusttothefailureof monotonicitybutlegalentryagedoesnot. Third,weareinterestedinthebroaderquestionoftheoptimalageatwhichtostart schooland,inparticular,optimalpolicyregardingschoolentryage.Legislation,such astheNoChildLeftBehind(NCLB)actof2001,hasputgreatpressureonschools toimprovestudentperformanceontests.Somestateshaverespondedbyraising theschoolentryage(DemingandDynarski,2008;Stipek,2006).Giventhehistorical natureofourdata,theinstrumentalvariablesapproachcapturestheeffectofdelaying schoolentryasitwaspracticedinthe1950s.Butschoolentryagelawsarenow enforcedmuchmorestrictly.Thus,weconductapolicyexperimentusingsimulated datatostudytheeffectofhavingstrictentryagerules(thecurrentpractice)on averageeducationalattainment.Weshowthattheentryagethatmaximizesachild’s eventualeducationalattainmentvariesconsiderablyfromaboutage4.5yearstowell 4 Aswediscussinsomedetaillater,dependingonthesizeofthebandaroundthediscontinuity, similarconcernsmayariseforanalysesbasedonregressiondiscontinuity. 3