Cleaning the Slate Schools Choice and Educational Outcomes in Spain

Embed Size (px)

Citation preview

  • 8/12/2019 Cleaning the Slate Schools Choice and Educational Outcomes in Spain

    1/25

    Cleaning the slate? School choice and educational outcomes in SpainAuthor(s): Cecilia Albert and Carlos Garca-SerranoSource: Higher Education, Vol. 60, No. 6 (December 2010), pp. 559-582Published by: SpringerStable URL: http://www.jstor.org/stable/40930311.

    Accessed: 27/03/2014 06:03

    Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at.http://www.jstor.org/page/info/about/policies/terms.jsp

    .JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of

    content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms

    of scholarship. For more information about JSTOR, please contact [email protected].

    .

    Springeris collaborating with JSTOR to digitize, preserve and extend access toHigher Education.

    http://www.jstor.org

    This content downloaded from 103.24.98.179 on Thu, 27 Mar 2014 06:03:32 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/action/showPublisher?publisherCode=springerhttp://www.jstor.org/stable/40930311?origin=JSTOR-pdfhttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/stable/40930311?origin=JSTOR-pdfhttp://www.jstor.org/action/showPublisher?publisherCode=springer
  • 8/12/2019 Cleaning the Slate Schools Choice and Educational Outcomes in Spain

    2/25

    HighEduc 2010) 60:559-582DOI 10.1007/s0734-0 0-93 5-9

    Cleaningthe slate?Schoolchoiceand educationaloutcomesn SpainCeciliaAlbert Carlos Garca-Serrano

    Published nline: February010 Springercience+BusinessMedia B.V. 2010Abstract Using atafrom nation-wideurveynSpanishworkers,his aper xaminestwo ssues: hedeterminantsf school hoice distinguishingetweenublic ndprivateschools) ndthe nfluencef school ype nthe ndividual'sducationalttainment.orthat,ivariaterobit odelsre stimated.he nalysiss carriedut eparatelyygenderandbirthohort.We find hat ndividuals hohave ttendedprivatechool omefromhouseholds ith igherocio-culturalevel nd hat avingttendedprivatechool oesnot aise he robabilityfreachinghigher egree,ncepotentialndogeneityfschoolchoice nd ducationalutcomesre akennto ccount. owever,he atteresultependson thenstrumentalariable sed or bservedchool ypenmodels fdemand or ighereducation.Keywords School hoice Higherducation Human apitalIntroductionThe aimofthis apers to nvestigatehe pparentorrelationetweenchool hoice ndsubsequentducationalutcomesf individuals.n particular,hequestionwe try oanswers whetherttendingprivatechool eneratesrelevantositivempactn theindividuals'robabilityfobtaininghigher egree. hepotentialelationshipetweenschool hoice nd ducationalutcomes ay ot ave nexplanations simples it eemsatfirstight,ince tmighte the esult f he xistencef electionias. t spossiblehatfamilies hochoose privatechool we betterharacteristicso raise he ducationallevelof their hildren.hesefactors aybe higherncome, igherttainedducationallevel ffathernd/or other,igher referenceor ducation,nd/origherocial apitaland betterabourmarketonnections.n otherwords, chool hoicewouldnotbe anexogenous ariablewhen ne tries oestimatehe ffect fschool ype neducationalC. Albert C. Garca-SerranoEl)Departamentoe Fundamentose Economa Historia conmica,Universidad e Alcal,Plaza Victoria , 28802 Alcal de Henares,Madrid, paine-mail: [email protected]

    Springer

    This content downloaded from 103.24.98.179 on Thu, 27 Mar 2014 06:03:32 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/12/2019 Cleaning the Slate Schools Choice and Educational Outcomes in Spain

    3/25

    560 HighEduc 2010) 60:559-582achievement singa one-equationmodel.This fact should be appropriatelyaken ntoaccount nempirical esearchnorder o obtainunbiased stimates f the nfluenceftheschool choice variable.

    A numberfprevious tudieshave contributedo thedebateon the uality fprivate s.publiceducation nd to theanalysisof theeffect feducational uality n labourmarketoutcomes.Meanwhile, ther uthors ave focussed he ttentionn the nfluence fprivateschooling n measures feducational chievement,uch as completing ost-compulsorysecondary ducation, tartingheuniversityr obtaining university egree Evans andSchwab 1995; Sander and Krautmann 995; Card and Krueger1996; Neal 1997). Thesestudieshighlightheimportancef usingdata sets thatprovidedetailed nformationodiscount heeffect f observedpersonal ndfamily haracteristicsnd toadequately akeaccountof theendogenousnature f school choice and educational utcomes.The contributionf thispaper ies in two areas.First,t aims to contributeo thefieldand to extend urrent esearch y nvestigatingherelationshipetween chool choice andeducational utcomes na particularuropean ountrySpain): thegreat ulkofempiricalevidence on this ssue comes from he US but s nearlynon-existentorEuropeancoun-tries,whose institutionalnd cultural rameworksre quitedifferent.econd,thepaperperformseparate nalysesofprivate chool effects ordifferentubsamples. he samplesare divided ccording ogender ndbirthohort,which dds new nsightso the mpiricalliterature.his differentiatednalysis s potentiallyelevant,ince tmight appen hat heeffectifany)of school choice differs ormen andwomen ndforpeoplewho were bornand studied ndifferenteriods f time.Our data set allows us to make thisdistinction.The structure f the papergoes as follows. Section two providesa review of theliteratureelated o schoolchoice and educational chievementnd a brief escriptionftheSpanisheducational ontext. n sectionthree,we discuss themethodology sed toinvestigatehe effects f school typeon educational utcomes.Sectionfourpresentshedatabasewe use in the nvestigationndprovides first,impledescriptivenalysis n thebasic variables. ectionfivereportsheresults ftheempirical nalysison school choiceand its nfluencen educational utcomes.Finally, ection ix concludes.Previous literature nd institutionalontextPrevious iteratureA number f studies mainly omingfrom heUS) have contributedo thedebateon thequalityof private/Catholic/religiouschoolingand public schoolingand theirrelativeeffectiveness. owever,measuring hequalityof education s not a problem-freeask,since it is notdirectly bservable. everalstudieshave tried o examinethe educationalquality y ooking t therelationshipetweennputssuchas spending n education r theproportionetween tudentsnd teachers) ndoutputssuch as thestudents' nowledgemeasured ytest cores).Thislineof researchsee Hanushek1986)was initiatedwith heworks f Colemanandhisassociates Colemanet al. 1982;Colemanand Hoffer 987)andfollowed ubsequentlyyseveral uthorsMurnane tal. 1985;CardandKrueger 992).However, uch testshave importantrawbacks: heymaybe cultural, exualor ethnicalbiased;theymayonlymeasure he tudents'bility o answer given ort fquestions utnotother; eachersmayinstructtudents ow to answerthetest,generatingn artificialincrease fpositive esults; nd so on.

    Springer

    This content downloaded from 103.24.98.179 on Thu, 27 Mar 2014 06:03:32 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/12/2019 Cleaning the Slate Schools Choice and Educational Outcomes in Spain

    4/25

    HighEduc 2010) 60:559-582 561Due to these limitations,ther authorspropose alternative riteria o evaluate thequality f education SanderandKrautmann995;Card andKrueger 996). In particular,theyuse measures f educational utcomes such as completing igh chool or obtaining

    a university egree) thathave evident economic consequences,since there s over-whelmingvidence thatdemonstrateshe benefitsf additionalyearsofeducation.Theirresults or he US are mixed.On the one hand,some of themfind vidence thatprivateCatholicschools increase tudent chievementwithrespect o publicschools,especiallyforminoritiesEvans and Schwab 1995; Neal 1997). On the otherhand,others uthorsfind hatonly religiousprivate chools (but notnon-religious rivate chools) increaseindividual utputs elative o publicschools (Figlio and Stone 1999). Thus,this fieldofresearch eems to provethehigher ualityof private ducation Betts 1995; Evans andSchwab 1995) and thepositive ffect f educational ualityon labour market utcomessuch as wages (Sandyand Duncan 1996; Altonji nd Dunn 1997; Neal 1997). However,McEwan (2000), after eviewinghis mpirical iterature,oncludesthatprivateCatholicsecondary chools in the US have positiveeffectsn improving ighschool graduationand college attendance ut almost no effect n individual chievementn standardisedtests.InEurope, handful f studieshave nvestigatedhis ssue.Theirresults end osuggestthat ublic choolsdo notfallbehind rivate chools, n average.For nstance, ertola ndChecchi (2002) find hat talianpublic schools are associated with better erformance,followed yreligious rivate chools and ay private chools.Moreover,VanderberghendRobin 2004), usingdata of standardised estsfrom he PISA survey, ive evidence thatpublicschools can outperformrivate chools in France and Austria.And Feinstein ndSymons1999) find hat ttendance tprivatechools does not n average ffectndividualperformancenthe UK. The panorama s even moredeserted or pain,since no previousstudies xist.InstitutionalontextGivenourtaskof nvestigatinghe nfluence f schoolchoiceoneducational utcomes ndthedatawe use (referredopeoplewhowere born ntheperiod1935-1979), the distinctpolitical, ocial andeducational hanges ucceeded nSpainattract ttentions potentiallyrelevant actors hatmayhave influenced hedistributionf students etweenpublicandprivate chools and therelationshipetween thetypeof schoolingand theeducationalachievement,ince one may presume hatthe institutionalrameworknder which theindividuals akedecisionscan influencehesedecisions.In Spain,therewere and stillare) threewaysofprovision f compulsoryducation:public schools,privatebut government-dependentchools and purely private chools.Whilepublic chools arecompletely undedwithpublicmoney ndpurely rivate choolschargehigh uition eescovering ll costs whichmake them ccessibleonlytotheupperclass), privatebut government-dependentchools are oftenreligiously ffiliated nddependfinanciallyn thegovernmentsincethey repartially undedwithpublicmoney),bringingboutmoderate ees that he middleclass can afford. his three-partystem fprovision an be considered Spanishpeculiaritynd is based on therelevant istoricalweight ftheCatholic nstitutionsnthestate, ngeneral, nd intheschooling ystem,nparticularinthis ase, throughhe xtensive etofreligious choolsfunded uringheXXcenturyndmainlybased in thecities).

    Springer

    This content downloaded from 103.24.98.179 on Thu, 27 Mar 2014 06:03:32 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/12/2019 Cleaning the Slate Schools Choice and Educational Outcomes in Spain

    5/25

    562 HighEduc 2010) 60:559-582We have to note that hemajorityf the ndividualsn our samplewere n educationunder he ducational awspassed during heFrancoist eriod. The last one was passed n1970.One oftheobjectives f this aw was to tackle heproblem f schoolposts' shortage,

    thus tarting huge ncrease f the chooling upply.Thisbroughtbout the nrolmentfnearly ll children nder14 years fage. Another bjectivewas tofight gainst choolingfailure nd to modernise he educational ystem adopting he most progressive'peda-gogical deas thatwere npracticenWestern urope n the ate- 960s).Moreover, hefirstdemocraticgovernmentsafter he fall of the dictatorshipn 1975) started drive toimprove hepublic schoolingand to finance hemajor partof the private ne, whichallowed thecountryo achievethe universalizationfthecompulsoryducation.How are the tudents istributedetween rivatendpublic chools?We have used datafrom arious ssues of theAnuarioEstadstico e Espaa (Estadsticade la Enseanza)toconstruct seriesof the number fstudents egisteredn schoolsaccording o the schoolproperty. his series spans from he late- 950s to the early-2000s.The proportionfstudentsnprivatechools remainedntherange f 26-28% until 971 andthenumpedtonearly 0% inonly4 years justafter hepassing f the1970-law).Sincethen, herewas asteadydecline untilreaching nearly-stable roportionf 34% (from1995 onwards).Although hisproportionan be interpreteds a equilibriumnthecompetitionetweenbothsectors, he nspection f the flowsof studentsn the ast decade suggests hat hepublicschool is increasingts shareamongtheworking lass (as a consequenceof thestrongntryfimmigrantsn thecountryincethemid-1990s)whiletheprivate chool semphasisingtselitistnatureFernndezEnguita2003).Turning ow to thefunctioningfthehigherducation ystemnd the volution f thedemandfor ducation,t s worth oting hatnearly ll post-secondaryducation nSpainis suppliedby universities,o higher ducation s the same as universityducation. naddition, hemajority f Spanishuniversities re public institutions,undedby publicbudgets theycurrentlyccountfor90% of totaluniversitytudents,lthough hispro-portionwashighernthe1980s and 1990s); tuition ccounts or essthan ne-fourthfthecurrentncome funiversities;igherducations defined s a rightnSpain's democraticconstitution,o scholarshipsre available for ow income studentsalthough hetake-uprate ndthe fforts percentagefGDP is low as compared ointernationaltandards,eeOECD 2009); and all studentswho completehighschool may attenduniversityfterentrance xaminationsalthoughnotnecessarilyn the courseof study f their hoice,somethinghatwasespecially rue uringhe 1980sand 1990sdue to therapid ncrease fthe number f students).Some figures re in orderhere to highlighthedramaticgrowth f the demandforschoolingregisteredince the 1960s: in 1977,theproportionf individualsn the 19-24age groupwho demanded niversityducationwas 14.9% formenand 13.4%forwomen;theseproportionsmounted o27.0 and39.6%, respectively,n 2004. Thus,the ncrease nhighereducationdemandwas 12 and 26% points,respectively,n a span of nearly30 years.21 The main laws in forceduringthe Francoistperiodregarding he educational systemwere thefollowing:the 1938 law on the reform f the secondaryeducation, the 1943 law on university nd the 1945 law onprimary ducation. For a detailed analysis of the educational systemduringthatperiod, see de PuellesBenitez (1986).2 Figures come from the second quartersof the Spanish Labour Force Survey (EPA). We consider thatindividualsdemandedforhigher ducationif either heyhad a university egreeortheywerestudying t theuniversityin the4 weeks previous to the interview).i Springer

    This content downloaded from 103.24.98.179 on Thu, 27 Mar 2014 06:03:32 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/12/2019 Cleaning the Slate Schools Choice and Educational Outcomes in Spain

    6/25

    HighEduc 201 ) 60:559-582 563Giventhesecharacteristicsf theSpanishexperience,ome authors ave focussed heattentionn examiningwhich factorsin particular, amily ackgroundnd income andemploymentxpectationsccording othe evelofeducation)would nfluencehedemand

    foruniversityducation.Theydocument hat hisdemand s largern thecase ofwomen,which eflects he ncreasing roportionsfwomen ngaged npost-compulsoryducationin Spain. One reasonbehind this result ould be relatedto a discriminationrgument:womenwould need a higherevelofqualificationhanmendo inorder ocompete nthelabourmarketMora 1997; Albert 000). These studies lso stress he mportancef theeconomic and cultural ackground f thehome: higheducational ttainmentnd socio-economic tatus fparentsncreases ubstantiallyheir hildren's robabilityfattendinguniversityPeraitaand Snchez 1998; Gonzlez and Dvila 1998; Albert 000).

    Methodologyfor theanalysisof school choice and educational outcomesWhentryingo examine the ikelyeffect f school typeon educational chievement,crucial ssue one mustnecessarily ayattentiono is that fthe election ias,since tmayseriously ffect he results f theinvestigation.uch bias comes from he factthat hedecision of attendingither private choolor a publicschool is notexogenous. f it ismore ikely hat ndividuals r familieswith ertain haracteristicssuchas higher amilyincome,higher arents' ducationalevels,ormorepreferenceor ducation,ome of thembeingunobserved ortheresearcher)hoose a private chool,the estimation f a modelthatdoes not take into account thisendogeneitywould overstate he effects f privateeducation n subsequent ducational ttainment.To see howthisbias works,et us consider hefollowingingle-equationmodel.Let Ybe thevariablemeasuringhe ducational utcome itmayconsist fgraduatingrom ighschool or reaching college/universityegree),so that t takesvalue 1 if the ndividualgraduates nd value 0 otherwise. he latentmodel is thefollowing:Y*= xi+ Xi]+Si2 + vi (1)Yj* s thenetbenefithe ndividual eceives from raduating, is a vector f individualcharacteristics, is a dummy ariable or chool choice 1 ifprivatend0 ifpublic), ndv,is a normally istributedandom rrorwith ero meanand unitvariance.Thismodelmaybe estimated singa single-equationrobitsee Evans and Schwab 1995).However, et us assumethat heparents f a child areconcernedwithhisfuture,o thatthey rekeento ncur nthe ost ofsending heir hildto a privatechoolortodevotemoretime oread or to do thehomeworkwithherorthey lace a higher alue on education ndonobtaining oodmarks.nmoreformal erms,he ype fschool variable hatweplace intheestimation f theeducationaloutcomesequationwould be correlatedwiththe errorterm nd theresultswouldbe biased. This impliesthat model similar o that ketchedpreviously ould erroneouslyttribute ll the success of the individuals o theprivateschoolthey ttendedntheir hildhood.3To properlyake ccount fthis ndogeneity,t snecessary oset out a model with woequations. n addition o theprevious quationthat ndicates he benefitshe ndividualswould receive ftheygraduate,nother quation s needed thatrefers o thebenefitshe3 A similarproblemexists ifprivate chools are able to select potential tudents hrough ersonal interviewsor to expel studentson either academic or disciplinarygrounds.

    Springer

    This content downloaded from 103.24.98.179 on Thu, 27 Mar 2014 06:03:32 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/12/2019 Cleaning the Slate Schools Choice and Educational Outcomes in Spain

    7/25

    564 HighEduc 2010) 60:559-582individualsmayobtainwhether hey ttend private chool or a public school. Let usconsider hefollowingatentmodel:

    S=Z,S + i, (2)S,* is thenetbenefit fattending private chool,Z, is a vector fobserved ndividualandfamily ariablesitconsists f the haracteristicsncludednXplussome nstrumentalvariable or ttending private chool),and it s the rror erm.According o thismodel,familywill choose to send her childto a private chool ifthe net benefits re positive(Si* > 0).To allow for hepossibilityhat he unobserved eterminantsf an individual's du-cational utcome nd theunobserved eterminantsf a family's ecisionto enrol he hildina private choolarecorrelated, e assume that , nd/,re distributedivariate ormal,withE[Vj]= E[n' = 0, var[Vj] var[n] 1 and cov[vhfi' p. Because the decisionswe model aredichotomous,here refour ossiblestates f theworld, ndthe ikelihoodfunctionorrespondingo thissetof events s a bivariate robit. f thep is statisticallysignificantlyifferenthan0, the assumptions re satisfied nd estimatingwo probitssimultaneouslys a betterption; therwise,stimatingwoprobitseparately ouldbe thepreferredption.The aim of forming consistent stimator or2 becomes manageableif one canconstructnstrumentsorS,. A variableZ' would be a valid instrumentorS if it iscorrelatedwithS (cov[Z'hS' ^ 0) and uncorrelatedwith the errortermof Y (cov[Z]hv]= 0) (see Wooldridge 006; Greene2008). Checking hefirst onditions easy:wecan estimate q. 2 andverifywhetheru is correlatedwith itonce theother ariablesZ,have beencontrolled or.However, t s notpossibleto contrasthe secondcondition:wemustbe confidenthat t s true nd base this onfidence n economicand/or nstitutionalreasoning elated o theproblem nder tudy.Several previousstudies have used religiousaffiliationi.e. whether he individualdefines herself s an active member of some religiousorganization) s a potentialinstrumentorprivate/religiouschooling see Neal 1997; Sander 1999). The rationaleresides n that family emandforreligious nstructionhould nfluencehe decisionofsending heir hild to a private choolbut notnecessarily hesubsequent tudent erfor-mance. As a result, nce the observedpersonal nd family haracteristicsre controlledfor,2 should ndicatewhether aving tudied n a private chool generates ubsequentbenefits r not. Obviously,the credibility f the bivariateprobitresultsrestson theassumptionhatpeople who are religiousdo not exhibit largerpropensityo graduatefrom igh choolorfrom niversityhan therwisedentical eoplewho are notreligious.An obviousproblemwiththeproposed nstruments thatreligious ffiliationmaybe aby-productan outcome)ofprivate/religiouschooling nd notprecedetheattendancefthe individual n private/religiouschools. This happensto be if, as in our data set,religious ffiliationeferso themomentwhen he urveywas passedand,therefore,s nota characteristicrevious o the ntrynto chool.Thisreason, oupledwith hefact hat heincidence freligious ffiliations extremelyow in the urveywe use (8.2%, see Table 5of theAppendix),which seems suspicious n a countryike Spain with mplereligioustradition,as prevented s to use thisvariable s an instrument.The empiricaliteratureas considered lternativenstruments:ither heproportionfCatholicswithrespect o totalpopulationn each county/stater thenumber fCatholic(private) choolsper quaredmile see Neal 1997;FiglioandLudwig2000; Newhouse ndBeegle 2006). However, heconsistencyf estimatess based on thecritical ssumptionthatlocal privateschool proximitys uncorrelatedwith unobserveddeterminantsf Springer

    This content downloaded from 103.24.98.179 on Thu, 27 Mar 2014 06:03:32 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/12/2019 Cleaning the Slate Schools Choice and Educational Outcomes in Spain

    8/25

    HighEduc 2010) 60:559-582 565students' cademic achievement. his assumption as been questioned horoughlyn theUS contextAltonjiet al. 2005).We have tried threedifferent,lternative nstruments:he father's itizenship, heindividual'sunionmembershipnd the size of themunicipality herethe intervieweeslived whentheywere 16 yearsold. Regarding he first ariable,the incidenceof non-Spanish itizenships very ow,as Spainhas become an immigrantountrynlysince themid-1990s.As for he secondone,we have notdetected nysort frelationshipetweenbeingmemberf a trade nion t themoment f the nterviewndhaving ttended publicschool,whichhas prevented s to use it.Finally, hevariablewe have decided to use asinstruments the ize ofmunicipalityf residence t the ge of16.The mainreason s thatprivate chooling purelyprivate chools and private utgovernment-dependentchools)expanded mainly n urban areas (FernndezEnguita 2003), so a strongrelationshipbetween hatvariable ndhaving ttended publicschool is expected.As we will see in"Results",this urns ut to be thecase. Moreover, ollowing ard 1993), we examine hedirect ontributionfmunicipalityize tothedemandforhigherducation ndexplore hevalidity f an instrumentonstructed s the interaction etweenmunicipalityize andhavinggrownup in regionswithdivergingchool mix.Dataset and descriptivesThe data we use in thispiece of research ome from he"WorkingConditions urvey"(Encuestade Calidad de Vida enel Trabajo,ECVT). TheECVT is a yearly urvey hathasbeen carried utbytheSpanishMinistryfLabour and Immigrationince 1999. This is anationally epresentativeandom ample survey f all employedndividuals ver 15 yearsof age and living n households.4 mploymentmeanshavingbeen working t least 1 hduringhereference eek theprevious ne tothe nterview eek).Sampling esign akesinto account three variables: autonomouscommunityregion),municipalityize, andnumber f nhabitantsnthecensus' section. electionofemployedndividualswithin hehouseholdsone perhousehold) s purely andom.5This survey ontains nformationn a wide array f variables.These maybe groupedintoworkers' ndividual haracteristicssuch as gender, ge, marital tatus,number fchildren,nd level ofeducation) nd characteristicsf the firmswhere hey re workingandthe ob they reperformingsector,ndustryffiliation,irmize,occupation,workinghours, enure,nd type fcontract).Furthermore,he ECVT includes a set of questionsregarding heperiodwhen theindividualswere at school and the momentwhentheywere 16. Firstof all, thesurveyallows one toelicitsome informationn thetype fschoolthe ndividuals ttendedwhentheywere in school age. In particular,hequestionnaire sks the following uestion:"What ype fschool didyouattendmainly ntilyouwere16 years ld?" The answers reas follows:"Public", "Religious", "Other ypeofprivate chool", "I did notattend nyschool",and "I do notknow". naddition,he urvey rovidesnformationnsomefamilybackgroundharacteristicshen the ndividualswere at theage of 16: where hey ived,whetherhey ived at family ome ornot, nd whatwas theeducational evel,thetype fjob and theprofessionaltatus f thehouseholdhead.4 Ceuta and Melilla (the Spanish cities in North-Africa) re excluded fromthe sample frame.5 More information n the survey can be found in the Ministryof Labour and Immigrationweb site:www.mtin.es.

    Springer

    This content downloaded from 103.24.98.179 on Thu, 27 Mar 2014 06:03:32 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/12/2019 Cleaning the Slate Schools Choice and Educational Outcomes in Spain

    9/25

    566 HighEduc 2010) 60:559-582Table 1 Distributionf ndividuals etween ublic ndprivatechools ndproportionf ndividuals ithuniversityegree, y gendernd birth ohort

    Distribution % Universityegree% Public % Private Men Women

    Total 75.1 24.9 18.5 28.5GenderMan 78.6 21.4Woman 68.5 31.5Birth ohort1930s-1940s 1935-1949) 75.9 24.1 15.4 16.31950s 1950-1959) 73.6 26.4 18.6 25.51960s 1960-1969) 75.1 24.9 18.4 29.51970s 1970-1979) 76.0 24.0 21.3 36.0Spain,ECVT (1999-2004)'Private'ncludes eligiousentre,rivateentre,nd otherype f school

    In thispaper,we use a pool of thefirst ix ECVT files ontaininghe ndividual ata(this mpliesusing heeditions rom 999 to2004). To beginwith, hose ndividualswithan age rangingbetween25 and 64 years in each surveyeditionhave been selected.Therefore, e haveexcludedpeoplewhomay tillbe engaged neducation those etween16 and 24- and people who legallycannot work-thoseaged 65 or more-.This initialsample amounted o a totalof 32,462 observations ut has been purgedafterhavingselectedonlythose ndividualswithvalid observationsnall thevariables elevant or heanalysis.Thismeansthatwe have excludedobservations elonging o individuals itherwithmissingvalues or included n the "I do notknow" categoryn thecorrespondingvariables.Moreover,we have eliminatedhe ndividualswho declarethat I did not attend nyschool" and "I do notknow"in the ype f schoolvariable both ategoriesmountedo a2% of the fullsample).The categories Religious" (17.5%) and "Othertypeofprivateschool" (7.3%) have beenmergednto new one called "Private".6Once all thispurgingprocesshas been undertaken,hefinal ample comprises 6,924observations.The people included n our sample were born in the period 1935-1979. We havegrouped hese ndividualsccording o thedecade in which heywereborn 'birth ohort')as a waytoproxy hedistinctocial and nstitutionalhanges ucceeded nSpainthatmayhave influencedhe distributionf students etweenpublicand private chools.Table 1reportshisdistributionygender ndbirth ohort ccording o oursample.The shareofprivate chools remained airlytable round 5% duringhewholeperiod fanalysis theonly xceptions that fthepeopleborn n the econdhalf f the1930s,who show a lowerincidence),with womendeclaring higherproportionhanmen. In addition,Table 1provides he shareof people holding university egree n our sample.The rise in thedemandforhigherducationhas beenoutstanding,nparticularmongwomen:whilethat

    proportionasmorethan uadrupledn a periodof about40 yearsforwomen, t has lessthandoubledformen.6 Thesurvey oesnot llowonetoproperlydentifyurely rivatechools ndprivatechoolspartiallyfundedwith ublicmoneyeducacin rivada oncertada). his is thereasonwhywe will refer ere opublic chools ndtherestprivate). Springer

    This content downloaded from 103.24.98.179 on Thu, 27 Mar 2014 06:03:32 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/12/2019 Cleaning the Slate Schools Choice and Educational Outcomes in Spain

    10/25

    HighEduc 2010) 60:559-582 567Table 2 Distributionf intervieweesnd householdheads when the formerwas 16 by educationalattainment,nd ncidence fprivatechools

    Interviewees'ducational ttainmentHousehold eads' educational ttainmentDistribution % Private Distribution % Private

    Illiterate - - 3.8 8.9Less than rimary 4.5 7.8 33.6 14.7Primaryducation 18.7 9.3 35.2 23.5Compulsoryecondary 25.1 14.1 13.2 29.6Vocationalraining-I 7.9 23.6 2.0 37.4Vocational raining-II 8.8 26.8 1.9 46.3Post-compulsoryecond 12.3 36.4 3.2 51.9University-short 10.3 42.9 3.1 53.8University-long 9.7 48.5 3.3 62.4Master nd doctorate 2.0 57.7 0.6 65.1Others 0.8 38.8 0.2 38.3Total 100.0 24.9 100.0 24.9Spain,ECVT (1999-2004)

    To take step urther,able2 offers firstescriptivepproximationo thepotentialassociationetweenchool hoice nd ducationalutcomes.wo ndicatorsreprovided:the irstne s the istributionf ntervieweesndhousehold eadswhen he ormer as16years ldbyeducationalevels; he econd ndicators theproportionf ndividualswho ttendedrivatechoolswith break ownby nterviewee'sndhousehold ead'sattainedducationalevel.Thefindingsthat he roportionf ndividualsho ttendedprivatechool ncreasesteadily ith he ducationalevel attainedyeither he nter-viewees hemselvesr thehousehold eads.Thus, ncomparison ithndividuals hoattendedublic chools,he neswho tudiednprivatechools eem ocomefrommorefavouredamilyackgrounds.his esultanbe betterraspedfwe ook tTable 5 of heAppendix,here descriptionfthe ull et f variablessed nthe mpiricalnalysissprovided.he nformationromhis able hows hathe ducationalevel f he ouseholdhead shigheror he roupfpeoplewho ttendedprivatechool.As a consequence,heproportionsf publicand private choolstudents horeached universityegreeamountedo 15.6 nd41.3%,respectively.Allpreviousnformationanbe taken sprima acieevidence hatomepersonalndfamilyharacteristicsreassociatedwith heprobabilityfbeing nrolledna privateschool ndobtaininguniversityegree.nwhat ollows, eattemptofind utwhetherattendingprivatechoolwhichppearso becorrelatedithmore dvantageousamilybackground)eneratesrelevantmpactn the ducationalutcomesf ndividuals,ncethe ffectfobservedersonalharacteristicssdiscountednd he ndogenousaturefschool hoice ndeducationalutcomess taken nto ccount.ResultsThedependentariablen ur ducationalttainmentq. 1takes he alue f hendividualhasgraduatedromollege/universitynd he alue otherwise.heregressorsncludednthe stimationsepresenthe haracteristicshat ould nfluencehe hoicewe want o

    Springer

    This content downloaded from 103.24.98.179 on Thu, 27 Mar 2014 06:03:32 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/12/2019 Cleaning the Slate Schools Choice and Educational Outcomes in Spain

    11/25

    568 HighEduc 201 ) 60:559-582analyse. According o the humancapital theory,hefamilybackground,he direct ndopportunityosts and futurencomes re deemedto be themaindeterminantsf educationdemand Freeman1986; Ordovensky 995). The probabilityhat personwilldecide tocontinue ull-timeducations a functionf various ocioeconomic ariableswhich eflecttheperceivedvalue of remainings opposed to abandoning heeducationsystem.Theprivate ost ofuniversitytudys madeupof three omponents:hedirectmonetaryost ofeducation, hemonetaryndnon-monetaryostsof access to theuniversity,nd the costsrelated opostponingne's incorporationntothe workforcend sacrificinghepotentialincome hatmight ave been earned n order o invest n human apital.And somefamilycharacteristicsincome, ulturalevel)are also determinantsf demandfor ducation incetheymay nfluence hemarginal enefitsnd costs of educational pportunity.Therefore,he set of explanatory ariables ncluded n themodels refer o personalcharacteristicsgender nd age) and,above all, characteristicsf thefamily ackgroundwhen the individualswere 16 yearsold: whetherhey ivedwiththeirparents;whethertheyhad brothersrsisters;whowas thehouseholdhead;professionaltatus, ccupation,and ndustryffiliationfthehouseholdhead;region fresidence; ndfather'sitizenship.We havealso included s control ariables hemaximumducationalevel attained ytheoldestbrother/sisterf the ntervieweendbytheheadof thehousehold.A descriptionfthefull etofregressorss providednTable 5 of theAppendix.We assume that all these variablesadequatelycapturethe family ocio-economicbackground,he ncome ndemploymentxpectations,ndthe osts nd benefitsssociatedwith emainingneducation hatmodelthedemandfor ducationEq. 1)but lso influencethe choolchoice Eq. 2). Inparticular,urhypothesissthat he evel of studies eached ythehead ofthehouseholdhas a significantnfluencen enrolmentn theuniversity,incechildren fbetternformedamilies, r familieswith onger raditionfeducation,will bemore ikely ogotouniversity.he samesort fhypothesisoldsfor heprofessionaltatusandthetype f ob variables egardinghe householdheadwhenthe nterviewee as 16,since we expect hat eoplewho comefromwealthieramilieswhoseheadwas employer,professionalrsalariedworkernwhite-collarigh-skilledccupations)willexhibit igherprobabilitiesfgoingtouniversity. e also expect hat he ame characteristicseterminethehouseholds' decisionof school attendance:f private chools are considered o besuperior n average to public schools, individualsfrommore educated and wealthierhouseholds hould xhibit higher robabilityfhaving ttended private chool.

    Since single-equationunivariate robit) stimates f theprivate chooleffectwill bebiased upwards f private chool attendances correlatedwithunobserved actors hatdetermineccess tohigher ducation nduniversityuccess,we have estimated ivariateprobitmodelsof educational ttainmentsingmeasuresof thesize of themunicipalitywherethe ntervieweesivedwhentheywere 16 yearsold as an instrumentorprivate/publicschool attendance. able 6 provides he V estimates f themodels described nEqs. 1 and2 for heoverallsampleand formen andwomen eparately. he fact hat heestimatedorrelationetween he unobserved ariables feducational ttainmentndtheunobserved ariables fschool choice p) is significantlyifferenthan0 (see the bottomof Table 6) helpsus to determinehat stimatingwoprobits imultaneouslys a betteroption han stimatingwoprobits eparately.Before addressing he questionof whether rivateschools are, on average,moreeffectivet raising he educationaloutcomesof individuals,we first xaminethe rela-tionship etween heobserved haracteristicsf individuals nd thetypeof school theychoose.While thedeterminantsf schooltype reof interestntheir wnright,hey lsomay provide omeguidancetothe extent ndnature f selectionbias due to unobserved) Springer

    This content downloaded from 103.24.98.179 on Thu, 27 Mar 2014 06:03:32 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/12/2019 Cleaning the Slate Schools Choice and Educational Outcomes in Spain

    12/25

    HighEduc 201 ) 60:559-582 569individualharacteristics.he results ndicate hatwomen ndpeopleover 35 years fageexhibita higherprobability f havingattended religious/privatechool, once otherfactors re taken nto account.This probability iseswiththe educational evel of thehouseholdhead whenthe nterviewee as 16 yearsold andwith heeducational evel oftheoldestbrotherifany). t is also higherfthehead of thehouseholdworked na white-collarhigh-skilledccupation r was employerrprofessionalon the ontrary,hildren fwage and salaryworkersholdinga fixed-termontract re the ones withthe lowestprobabilityfhaving ttended religious/privatechool). It shouldbe stress hat heheadof the household's educational ttainment,rofessional tatus nd occupationwhen theintervieweewas 16 and the size of the residenceunit are theregressorshat show thelargest ffects n theprobabilityfhaving ttended private chool. These results eem toconfirm he mpressionhatprivate chools tendtobenefit rom ositive election.7Moreover, he haracteristicshat nfluence heprobabilityfhaving eacheduniversityinstructiongreewith he ones that nfluence heprobabilityfprivate choolattendance.First, he demand forhigher ducation s larger n the case ofyoungpeople and women(the atter esult eflectshe ncreasing roportionf women ngaged nhigherducationnSpain). Second,as we expected, heprobabilityfreaching universityegree s largernfamilieswhose head worked nhigh-skilled ccupations r was employer rprofessional,and thehigher he educational evel of the household head and of the oldest brother,confirmingreviousresults y other uthorsDuncan 1994; Gonzlez and Dvila 1998;Albert 000). These variables refamily ackgroundharacteristicshatmainly eflect hefamily's conomicconditions nd expectationst thetimethe ndividualwas adolescent.Whatare, then, heeffects fprivate choolattendancen the ndividual's ducationaloutcomes? able 3 presentsV estimates f the ffect f schooltype private s.public)onthe probabilityf havingreached a universityegree forthe full sample and forthesubsamples fmenand women.Columns 1) and 2) showthe oefficientsfthevariablesof nterestnunivariate robitmodels for chool choice anddemandforhigher ducation,whilecolumns 3) and (4) reporthe V bivariate robit stimates f the mpact fprivateschoolingon educational attainment. hree differentpecifications f the instrument(strategies , B andC) have been tried.Having grownup in large municipalities as a strong ositiveeffect n bothprivateschooling nddemandfor ducation.The use ofsize of themunicipalitys an exogenousdeterminantf school choice ('strategyA') yields V estimates hat ndicate hatprivateschooling rings bout significantositive ffect neducational ttainment.nparticular,themarginalffect f schooltype hows thathaving ttended private chool nstead fapublic chool ncreases heprobabilityfobtaining higher egreeby3.4% points or hefull ample,3.1% pointsfor hesubsample f men and 5.6% pointsfor hesubsample fwomen see Table 4). These figuresmply hat hepredicted robabilityorpublicschoolstudentswould have changedfrom 2.3 to 25.7% ifthe ndividualhad attended privateschool instead fa publicschool when he/shewas a child;thecorresponding iguresre21.0 and 24.1% formen and 25.3 and 30.9% forwomen see Table 4).87 Moreover,ither eing n onlychild orhaving rothersr sisters utbeingthe oldestone,havingSpanishfather,nd not ivingwiththeparents t theage of 16 are factors orrelated ith higherprobabilityfhaving ttended privatechool we mustnotethat tudyingn a privatechoolmaybeassociatedwith eing ent o a boardingchool).8 Since chool ypes a dummy,hemarginalffects the ample verage f thedifferencenthepredictedprobabilitieshen hedummy ariable s set to one or zero. The predictedrobabilityorpublic choolstudentsdummy 0) is calculatedholding ther ersonal ndfamilyharacteristicsonstantnd cor-rectinghepotentialndogeneityf the choolchoicevariable.

    Springer

    This content downloaded from 103.24.98.179 on Thu, 27 Mar 2014 06:03:32 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/12/2019 Cleaning the Slate Schools Choice and Educational Outcomes in Spain

    13/25

    cdccd>

    cd>'S3'S1

    570 HighEduc 2010) 60:559-582

    l si1 ' s ~ s L ' s ~ ssL '2 Sis d 1ot 1llui1 c-gdjL ' sill ' S ' spiisic-gd SdC 65t S

    w. 3 * r^ # t-* on * 00 # o m # o # 00 # 00S J=3 MJ -00 MO - O 00 (S- (SOO (SW --tisisS J=3 OidcdJ -00 MO - O 00 (S-di0dS0SSOO (SW --I I 1J ' sigi ' bIsI ' s||si|o -x- ^ * *

    I C8S S o3d C Sodt

    M M-gC o^dC dSd^ d^dCd^c * *5 x:3 Tf t^ - 00 ^ 00 t o - 00 o o Tfso- OfMooow-)g .2PS2 dnd^d^ dvod^dS dr d^ 0^ 6^O,Cw www www wwwlw

    1 I 1|. i|l s|s| aliisil si1 ' S^8^' ^^sislr11 ls d ?ldl 1letl(O * * * *111 ' SrK ' S|^^ ' 8 S2 S

    OQc^ow dwdw dwdC dwdCdwc * * *

    Isis E1S Edl?e ciEdic:ci * * * *

    S 'S -g 2 ^ ^N^ ?S .> ..? .> ..^S2 toi t l|S tot -lgII l|l II 1 lili l|l ilm

  • 8/12/2019 Cleaning the Slate Schools Choice and Educational Outcomes in Spain

    14/25

    HighEduc 2010) 60:559-582 571

    e1I

    coS 'S.t -S8~I as i io.B * *5 a sS 3 2 g .2?= - d ^ o P i^ I Ci, I C

    c -r * *

    I "a1 - -5-Sco- 4|0 S

    S. X"8Ci oSciS 21 1 L 2 so1 1 .S- I B 1 1S .2? - < 3g SUS i i | -a 5 "S I1 | J SSSS Sx """ -fc- *S ^ ^ """ fcO1. }l?rd>^^ 12 1- lijw. *d>^^ i > I - 7jwd cealii I^^^ilI II

    li Siili P9, 2 '-5 -v Springer

    This content downloaded from 103.24.98.179 on Thu, 27 Mar 2014 06:03:32 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/12/2019 Cleaning the Slate Schools Choice and Educational Outcomes in Spain

    15/25

    572 HighEduc 2010) 60:559-582Table 4 Predicted robabilitiesf obtaining universityegree publicschoolstudents)ndmarginaleffectsf school ype private s. public), y gendernd birth ohortIV estimates,strategy ')

    Predictedrobabilitiespublic chool) MarginalffectsTotal Men Women Total Men Women

    Total 0.223 0.210 0.253 0.034** 0.031** 0.056**1935-1949 0.169 0.209 0.163 0.028** 0.028** 0.0271950-1959 0.201 0.204 0.210 0.033** 0.036** 0.0271960-1969 0.224 0.208 0.259 0.035** 0.023** 0.082**1970-1979 0.250 0.229 0.309 0.044** 0.039** 0.046All estimationsontrol or hevariablesncludednTable 6 oftheAppendix,xcept hose elated othe1970-1979birth ohortwhich o not nclude hevariable Educationevel of olderbrother"** (*) Coefficients statisticallyignificantt 1% (5%)Inspectionf the stimatedmarginalffects or hebirthohort ubsamples eveals wointerestingssues.First, hemarginalffectsncrease n absolute erms rom hefirsto thelast cohort nder tudyfrom .8 to 4.4% points)butremain table nrelative ermstheestimated rivate ffectpremium mounts o 15-17% of thepredicted robability orpublic school students ortheoverallsampleand forall birth ohort ubsamples).Thelatterresult would suggestthat the impactof private choolingon theprobability freaching universityegreehas not lterednspite f the nstitutional,olitical nd socialchangesthat ccurredn Spain throughoutheobservation eriod.Second,thepreviouspatterns drivenbymen, incein the case of womenthe coeffi-cients nd themarginalffectsf the chooldummyrenotstatisticallyignificantor llbirth ohorts avefor he 1960-1969cohort.We canonly peculate bout ome factorshatmightause this esult. hree nterconnectedacts rein orderhere:womendeparted roma significantower evel of education s compared o men;householddecisionsweretosendwomen nhigher roportionsoprivate chools see Table 1); and theapparentwayofreducinghegendergap in the abourmarket uring he ast 40 yearshas consisted fwomenrapidly ncreasing heirparticipationn higher ducation, rrespectivef theirinitial ituation,lthoughmore ntenselyn urban reas (Mora 1997; Albert 000). Bothfactorsmight e picking p somesort frelationshipetween he ize of themunicipalitywhere he ndividualsivedwhen heywere16 years ld andtheunobserved eterminantsofreaching universityegree.This ssue eadsus towonderwhetherhat ariable s a valid nstrumentor chooltype.The estimationf theschool choice equationgivesus theresult hat, omparedwith hecategory f referenceless than10,000 nhabitants),avinggrownup ineither medium-sized or a largemunicipalityignificantlyncreases heprobabilityfhavingattendedprivate chool (see columns 1) and (3) of Table 3).9 This means thatone of thekeyidentificationonditionsf ourmodelholds.Unfortunately,e cannot ontrasthe econdcondition,.e. cov[Z]hv]= 0 (that he ize ofmunicipalityf residence t the ge of 16 isuncorrelated ithunobserved eterminantsf individuals' ducational utcomes).Althoughheexpansion f thedemandforhigher ducation hathas occurrednSpainin the ast 40 yearshas beencoupledwith heextension f thesupply hroughoutll theSpanishterritoryrespondingnmany ases to criteria f territorialompensationor essdeveloped, ural reas),wecannot eject hepossibilityhat ocation rmunicipalityize is9 Identical esults found or ll birthohorts. Springer

    This content downloaded from 103.24.98.179 on Thu, 27 Mar 2014 06:03:32 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/12/2019 Cleaning the Slate Schools Choice and Educational Outcomes in Spain

    16/25

    HighEduc 2010) 60:559-582 573a keydeterminantfuniversityttendance.nordero testt,weproposeousea moreelaboratenstrumentfollowingard1993,whousedvariationnuniversityroximityoinstrumentniversityttendance).dentificationouldbe providedythe nclusionfmunicipalityize in theeducationalttainmentquationin order ocapturets directcontributiono theoutcome)nd, n the chool hoice quation,ythe dditionf theinteractionetweenmunicipalityize anddummyariablesapturingegionalifferencesreflectingivergingolitical/religiousollective referenceshathavehistoricallyon-tributingocross-regionaleterogeneitynthe choolmix.To implementhis pecificationfthe nstrument,ehave ried wodifferentays.nthe irstne 'strategy'), theargestategoryfmunicipalityizehasbeen efts a 'largemunicipality'ndicator1 = '> 100,000nhabitants';= 'otherwise')nd he17Spanishregions ave beengroupedn twocategories1 = 'above average hareof privateschools'; = 'otherwise'),ccordingo their ivergingistributionsn the choolmixduringeveral ecadeswhich an be related o historicalifferences).nthe econd ne('strategy), the hree riginalategoriesfmunicipalityize and the aboveaverageshare fprivatechools' ndicatoravebeenused.Thekey ssumptionnthese trategiesis that he irectontributionfmunicipalityize touniversityttendancehould ot aryacross egions ith ifferentollectivereferencesor rivateducation.IV estimatesased on these dentificationtrategiesreprovidedn Table 3. Thecoefficientsn columns1) and 2) confirmhat heeffects fhavingived n a largemunicipalityrebigger or ndividualsiving n regionswith argerhares fprivateschools. hecorrespondingV estimatesfprivatechoolingnhigherducationemandaredisplayedn columns3) and 4),alongwith he irect ffectfhaving rown p n alargemunicipality.he IV estimatednfluencefhaving ttended privatechool sstatisticallynsignificantonly nderstrategy thempactormen sslightlyignificant).Moreover,hepoint stimatesf thedirect ffect fmunicipalityize on educationalattainmentresignificantlyifferentrom . Theseestimatesast some doubts n theassumptionhatmunicipalityize s anexogenouseterminantfschool hoice.To sumup, V estimatesased on the nteractionfmunicipalityize andregionaldifferencesnthe choolmix requite ifferentromV estimatesased nmunicipalitysize alone.Furthermore,stimatesf thedirect ffect f ocation n demand orhighereducationreuniformlyarge nd tatisticallyignificant.ssuminghatmunicipalityizecan be excluded romhe ducationalttainmentquation,t eems hat oth ocationndits nteractionith egionalchool-mixndicatorsannot eused s instrumentsor choolchoice.ConclusionsThe aim of thispaperhas beento investigateherelationshipetween he chool ype(public rprivate)ttendedyindividualsndtheirubsequentducationalutcomes,using atafor pain theWorkingonditionsurvey,CVT) for he irstime. ince hisdataset rovides s with large ample,thas also beenpossible ocarryut a differ-entiatednalysis ygendernd ndividual'sirthohort.The nvestigationas tried otake nto ccount hepotentialndogeneityrisingromthe ikely orrelationetween heunobservedharacteristicsssociatedwith he choolchoicemadebythehouseholdsnd theunobservedharacteristicsssociatedwith du-cational erformance.heresultsbtainednthe conometricstimationsuggesthat hisendogeneitysrelevant:fwe donot ake t nto ccount, e willprobablybtain iased

    Springer

    This content downloaded from 103.24.98.179 on Thu, 27 Mar 2014 06:03:32 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/12/2019 Cleaning the Slate Schools Choice and Educational Outcomes in Spain

    17/25

    574 HighEduc 2010) 60:559-582results.Nevertheless,lthoughwe have tried o isolate all the unobserved ariablesthatcould affect ntrancenhigherducation,we cannot eject hepossibilityhat omehiddeneffects ad not beenconsideredbecause they renoteasilymeasurable).

    One relevant indingthat greeswithprevious nes in theempirical iterature)s thatindividualswho have attended privatechool nSpaincomefrom ouseholdswithhighersocio-culturalevel. This means thatfamilybackground haracteristicsthe householdhead's educational ttainment,rofessionaltatus nd occupation),whichmainlyreflectthefamily's conomicconditions nd expectationst the time the individualwas ado-lescent,are important redictors f school choice. They are also the main variablesinfluencinghe demandforhigher ducation.The keyconclusionfrom hispaper s thatpeoplewho attended religious rprivateschooluntil he ge of 16 do not eemto exhibit higher robabilityfreaching niversityinstruction,nce either bservedpersonal nd household haracteristicshentheywere16 yearsold or unobserved actors re controlled or.When we estimate ivariate robitmodels whichuse the size ofmunicipalityf residence t theage of 16 as instrumentorprivate choolattendance,hefindings that he mpact fprivate chooling s significant:themarginalffectfhaving ttended private chool nstead f a publicschool ncreasestheprobabilityf obtaining higher egreeby 3.4% pointsfor thefullsample,whichimplies n estimated rivate ffect remiumfabout15% of thepredicted robabilityorpublicschool studentsand thisgeneralresultholdsformen and womenand for ll thebirth ohorts onsideredn theanalysis).However, hese nferencesely n therestrictivessumptionhathavinggrownup in alargemunicipalityas no effect n educational ttainmentpartfrom he effect hroughschool choice. To testthisassumptionwe have used the fact that ocationhas a largerimpact n the chooling hoicesofpeoplewho ive inregionswith arger hares fprivateschools. The use of an interactionfmunicipalityize as of theage of 16 and regionalindicatorss an instrumentalariableforobserved chooltype n modelsof demandforhigherducation ivesriseto estimatesn which he mpact fprivate chooling anishes.In sum,these results o not allow one to concludethatprivate chool attendance erseallowed individuals o reachhigherevelsof educationwhencompared o publicschoolattendancen Spainduring heobservation eriod thesecondhalfof theXX century).

    AppendixSee Tables 5 and 6.Table 5 Descriptivesf thefull ample.ECVT (1999-2004)

    Total %) Public %) Private%) ObservationsEducational utcome: niversityegree 22.00 15.61 41.26 5,924TypeofschoolPublic 75.09 100.00 0.00 20,217Private/religious 24.91 0.00 100.00 6,707GenderMan 64.96 68.02 55.75 17,491Woman 35.04 31.98 44.25 9,433

    Springer

    This content downloaded from 103.24.98.179 on Thu, 27 Mar 2014 06:03:32 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/12/2019 Cleaning the Slate Schools Choice and Educational Outcomes in Spain

    18/25

    HighEduc 2010) 60:559-582 575Table 5 continued

    Total %) Public %) Private%) ObservationsAge group25-29 14.96 15.03 14.75 4,02730-34 16.24 16.60 15.15 4,37235-39 17.44 17.34 17.74 4,69640-44 16.16 16.11 16.34 4,35245^9 12.95 12.55 14.15 3,48750-54 10.61 10.49 10.97 2,85655-59 7.58 7.70 7.22 2,04060-64 4.06 4.19 3.68 1,094Did you ive withyour arents henyouwere16?Yes 97.31 97.76 95.96 26,200No 2.69 2.24 4.04 724Who was thehousehold ead whenyouwere 16?Father 95.67 95.90 94.96 25,757Mother 3.23 3.00 3.92 870Otherman/woman 1.10 1.10 1.12 297Brother/sisters?Only hild 9.16 8.73 10.45 2,467Theoldest fbrothers/sisters 30.64 29.50 34.07 8,249Brothers/sistersutnot heoldest 60.20 61.76 55.48 16,208Educationalevel oftheoldest rotherLess than rimaryducation 3.51 4.29 1.18 946Primaryducation 10.09 11.98 4.38 2,716Compulsoryecondaryducation 10.34 11.50 6.84 2,784Vocationalrainingfirstevel) 2.14 2.21 1.94 576Vocationalrainingsecond evel) 2.51 2.42 2.76 675Post-compulsoryecondaryducation 4.23 3.74 5.71 1,139Universityegreefirstevel) 3.03 2.38 4.98 815Universityegree second evel) 4.10 2.87 7.81 1,105Doctorate/masteregree 0.54 0.30 1.27 146Without lderbrothers 59.51 58.31 63.13 16,022Job fthehousehold eadwhen he ntervieweeas 16White-collarigh-skilledob 18.55 12.99 35.29 4,994White-collarow-skilledob 10.92 10.14 13.28 2,941Blue-collar igh-skilledob 44.71 49.28 30.95 12,039Blue-collarow-skilledob 25.81 27.58 20.47 6,950Industryf thehousehold ead when he ntervieweeas 16Agriculture 24.68 29.02 11.58 6,644Building 14.84 16.51 9.83 3,996Energy, ater,ndextractives 3.25 3.28 3.13 874Metaltransformationndchemicals 7.55 6.98 9.26 2,033Equipment roduction 2.89 2.57 3.83 777Othermanufacturingndustries 10.13 9.51 12.02 2,728

    4 Springer

    This content downloaded from 103.24.98.179 on Thu, 27 Mar 2014 06:03:32 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/12/2019 Cleaning the Slate Schools Choice and Educational Outcomes in Spain

    19/25

    576 HighEduc 2010) 60:559-582Table 5 continued

    Total %) Public %) Private%) ObservationsTraditionalervices 19.80 18.42 23.96 5,331Total %) Public %) Private%) ObservationsProductionervices 4.71 3.46 8.51 1,269Social services 3.34 2.47 5.93 898Personal ervices 2.14 2.18 2.00 575Public ervices 6.68 5.60 9.94 1,799Professionaltatus fthehousehold ead when he nterviewee as 16Public ectormployee 14.71 13.05 19.71 3,961Private ector mployeewith emporaryontract 48.92 49.52 47.11 13,172Privateectormployeewith ermanentontract 4.41 5.36 1.57 1,188Employer/professionalith mployees 4.58 3.48 7.90 1,232Professional/self-employed 23.97 24.95 21.01 6,454Other 3.41 3.64 2.70 917Educationalevelofthehousehold eadwhen he ntervieweeas 16Illiterate 3.78 4.59 1.36 1,018Less than rimaryducation 33.61 38.17 19.86 9,049Primaryducation 35.23 35.90 33.20 9,484Compulsoryecondaryducation 13.16 12.34 15.66 3,544Vocationalrainingfirstevel) 1.96 1.64 2.95 529Vocationalrainingsecond evel) 1.87 1.34 3.47 503Post-compulsoryecondaryducation 3.19 2.04 6.65 859Universityegreefirstevel) 3.12 1.92 6.72 839Universityegreesecond evel)or more 3.91 1.93 9.86 1,052Othersnon-officialtudies) 0.17 0.14 0.27 47Size of themunicipalityhen he ntervieweeas 16Less than10,000 nhabitants 30.78 35.53 16.46 8,28810,000-100,000nhabitants 29.09 29.46 27.97 7,833Morethan100,000 nhabitants 40.12 35.00 55.57 10,803Father's itizenshipSpanish 95.53 95.37 96.03 25,721Non-Spanish 4.47 4.63 3.97 1,203Member f a religious rganizationYes 8.22 7.39 10.71 2,213No 91.78 92.61 89.29 24,711

    Survey ear1999 15.81 15.27 17.44 4,2582000 16.94 17.91 14.05 4,5622001 17.26 16.49 19.59 4,6472002 17.70 17.39 18.64 4,7652003 17.02 17.30 16.18 4,5822004 15.27 15.65 14.10 4,110Observations 26,925 20,218 6,707

    } Springer

    This content downloaded from 103.24.98.179 on Thu, 27 Mar 2014 06:03:32 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/12/2019 Cleaning the Slate Schools Choice and Educational Outcomes in Spain

    20/25

    HighEduc 2010) 60:559-582 577

  • 8/12/2019 Cleaning the Slate Schools Choice and Educational Outcomes in Spain

    21/25

    578 HighEduc 2010) 60:559-582

    IiSO

    I ...... ^ L L ta ......***** $ ^ c$ & r7 r73 ^n^Pn-riP ^ ~. ^ "l-t^^PJ^3 h 4 O; ^ - > en O r- o www w^-^^w.^w. S^#^ N/ ^^ '*S ^,y N*^ S_i^ s_,> s^ QQ f*N, ^ QQ '^/ '^s v^v '^/ ^^ ',^/ s,^ ^^ om - r^ooooT)vO O^qr-pr^pvpcnO^ e r^aT ^ecieecici-oo ^-jTtyo(Nin

  • 8/12/2019 Cleaning the Slate Schools Choice and Educational Outcomes in Spain

    22/25

    HighEduc 2010) 60:559-582 579

    11veIS

    O jt ar: ^v^-v #***-*#**# ^ 5k3 T ^ ^ t in m ioioon - -S2^ S g Ja s |i t, is li, fi s 1111*1 -il i I 1z 1 1 &Hil ljiili. -I g f w 1 1 1 I I 8 e .- 8- Ililil wili Mlinls,CU UJ PU S 0 N Springer

    This content downloaded from 103.24.98.179 on Thu, 27 Mar 2014 06:03:32 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/12/2019 Cleaning the Slate Schools Choice and Educational Outcomes in Spain

    23/25

    580 HighEduc 2010) 60:559-582

    11

    U1IS.as CN1 8 S 8 On STtg "g t^ On Tt> c/D m I OsI1DI1 S OcX O PO O Os 4) i2 oo o vo o I - Bui lo cI III.s? axi~i1 S II-^ -Sois #

    H c oCIcn ^co|J=11 1si'S 'S -o 7j .2s .s -'S 1 s | ^ ^a 1 I

    eu J O P *^ Springer

    This content downloaded from 103.24.98.179 on Thu, 27 Mar 2014 06:03:32 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/12/2019 Cleaning the Slate Schools Choice and Educational Outcomes in Spain

    24/25

    HighEduc 2010) 60:559-582 581ReferencesAlbert, . (2000). Higher ducation emand nSpain:The influencef labourmarketignal ndfamily

    background. igher ducation,0(2), 147-162.Altonji, .G.,& Dunn, . A. (1997).Using iblingsoestimatehe ffectf chool ualitynwages.ReviewofEconomics ndStatistics,8(4),665-671Altonji, .G., Elder,T. E., & Taber,C. R. (2005). An evaluationf instrumentalariable trategiesorestimatinghe ffects fCatholic chooling. ournalfHumanResources,0(4), 791-821.Bertola,G., & Checchi,D. (2002). Sortingndprivate ducationn Italy.Centre orEconomicPolicyResearch, iscussion aperNo. 3198.Betts, . R. (1995). Does schoolqualitymatter? vidence romhenationalongitudinalurveyfyouth.Review fEconomics ndStatistics,7(2),231-250.Card,D. (1993).Usinggeographicariationncollegeproximityoestimate he eturnoschooling. BERWorking aperSeries,Working aperNo. 4483.Card,D.,& Krueger, . B. (1992).Does school ualitymatter? eturnsoeducation nd the haracteristicsofpublic chools n theUnited tates.JournalfPoliticalEconomy, 00(1), 1-^0.Card,D.,& Krueger, . B. (1996).Schoolresourcesnd tudentutcomes: n overview fthe iteraturendnew evidence romNorth ndSouthCarolina.Journal fEconomic erspectives,0(4),31-50.Coleman, .,& Hoffer,. (1987).Public, atholic,ndprivatechools: he mportancef ommunity.ewYork:Basic Books.Coleman,J., Hoffer, ., & Kilgore, . (1982). Highschoolachievement:ublic, catholic, ndprivateschools ompared.New York:Basic Books.de PuellesBenitez,M. (1986). Educacin ideologia n la Espaa contempornea. adrid: d. Labor.Duncan, . (1994). Likefather,ikeson, ikemother,ikedaughter.ournal fHumanResources, 9(4),950-988.Evans,W. N., & Schwab,R. M. (1995). Finishing igh chool andstartingollege:Do catholic choolsmake difference?uarterlyournalf Economics. 10(4). 941-974.Feinstein, ., & Symons, . 1999). Attainmentnsecondarychools.Oxford conomic apers,51, 300-321.Fernndeznguita,M. (2003). La segunda eneracin a estaqu. Papelesde EconomaEspaola,98,238-261.Figlio,D., & Ludwig,J. 2000). Sex,drugs,ndcatholic chools:Privatechoolingndnonmarketdo-lescent ehaviour. BER Working aper7990.Figlio, .,& Stone, . 1999).Areprivatechools eally etter? esearch nLaborEconomics,3, 115-140.Freeman, . B. (1986). Demandfor ducation.nO. Ashenfelter R. Layardeds.),Handbook fLaborEconomics, orth olland:Amsterdam.Gonzlez,B., & Dvila,D. (1998). Economic ndculturalmpedimentsouniversityducation nSpain.Economics fEducationReview, 7('), 93-1 3.Greene,W. H. (2008). Econometricnalysis6th d.). Upper addle River:Prentice all.Hanushek, . A. (1986).The economics fschooling: roductionndefficiencynpublic chools.JournalofEconomic iterature,XIV,1141-1 77.McEwan,P. (2000). Comparinghe ffectivenessfpublic ndprivate chools:A review fevidencendinterpretations.ccasionalpaperno. 3. NewYork:TeachersCollege,ColumbiaUniversity.Mora,J.G. (1997). EquitynSpanishhigherducation. igher ducation, 3, 233-249.Murnane, ., Newstead, ., & Olsen,R. 1985).Comparing ublic ndprivatechools:Thepuzzling ole fselectivityias. Journal fBusiness ndEconomic tatistics,, 23-35.Neal,D. (1997).Theeffectsfcatholicecondarychoolingneducationalchievement.ournalfLaborEconomics, 5(1),98-123.Newhouse, ., & Beegle,K. (2006). The effect fschooltype n academic chievement.vidence romIndonesia. ournalfHumanResources, 1(3). 529-557.OECD (2009). Education t a glance.OECD.Ordovensky,. F. (1995). Effects f nstitutionalttributesnenrolmenthoice: mplicationsorpostsec-ondary ocationalducation. conomics fEducation eview, 4,335-350.Peraita, , & Snchez,R. (1998). The effect f family ackgroundn children's evel of schoolingattainmentnSpain.Applied conomics, 1, 1-32.Sander,W. (1999). Privatechools ndpublic choolachievement. ournal fHumanResources, 4(4),697-709.Sander,W.,& Krautmann,. (1995).Catholicchools, ropoutates,nd ducationalttainment.conomicInquiry,3,217-233.

    Springer

    This content downloaded from 103.24.98.179 on Thu, 27 Mar 2014 06:03:32 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/12/2019 Cleaning the Slate Schools Choice and Educational Outcomes in Spain

    25/25

    582 HighEduc 2010) 60:559-582Sandy,J.,& Duncan,K. (1996). Does private ducation ncrease arnings? asternEconomicJournal,22(3), 303-312.Vanderberghe,.,& Robin, . (2004).Evaluatinghe ffectivenessfprivateducationcross ountries:comparisonf methods. abourEconomics, 1,487-506.Wooldridge,.M. (2006). Introductoryconometrics:modernpproach.USA: South-western.

    i Springer