WethankseminarparticipantsatIESEUniversitatPompeuFabraandtheAEDE2010meetingsfortheirthoughtfulcommentsCasasArceacknowledgessupportfromtheMinisteriodeCienciayTecnolog ID: 209430
Download Pdf The PPT/PDF document "WomenandPower:Unwilling,Ineective,orHel..." is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
WomenandPower:Unwilling,Ineective,orHeldBack?PabloCasas-ArceUniversitatPompeuFabra,BarcelonaGSE,andPublic-PrivateSectorResearchCenteratIESEBusinessSchoolAlbertSaizUniversityofPennsylvaniaMarch2011AbstractWedevelopamodelthatnestspreviousexplanationsforwomenunder-representationinpositionsofpower.Focusingondemocraticelectoraldynamics,ourframeworkdelin-eatesthethreetypesofmechanismsthatmaybeatplay:consumerdemand,candidatesupply,andinternalpartydynamicsbeyondelectoralmarkets.WeuseSpain'sEqual-ityLaw,requiringa40percentfemalequotainelectorallists,totestthealternativetheories.Thelawwasenactedbythesocial-democraticpartyafterthesurprisepar-liamentaryelectoralresultsfollowingtheMadridterroristbombings,andwasthereforecompletelyunexpectedbyregionalpoliticalmachines.Thelawonlyappliedtotownswithpopulationsabove5000,sowecanuseatreatment-control,before-and-afterdis-continuitydesigntolearnabouttheimpactoffemalepoliticiansinlocalelections.Ourevidenceismostconsistentwiththeexistenceofentrenchedmale-dominatedpoliticalmachinescapturingin uentialpowerpositionswithintheparties.JEL:J16,J71Keywords:femalepoliticalrepresentation WethankseminarparticipantsatIESE,UniversitatPompeuFabra,andtheAEDE2010meetingsfortheirthoughtfulcomments.Casas-ArceacknowledgessupportfromtheMinisteriodeCienciayTecnologÃa(SEJ2007-64340).SaizacknowledgesassistancefromtheWhartonGlobalInitiativesFund.Casas-Arce:pablo.casas@upf.edu;Saiz:saiz@wharton.upenn.edu.1 1IntroductionOvertheworld,malesarelargelyover-representedinpowerfulpositionsineitherthepublicortheprivatearenas.Only17percentofthemembersoftheUSCongressandSenate,forinstance,arewomen.Severalreasonsmayexplainwhywomendonotreachin uentialpositionsmoreoften.Womenmaybediscriminatedagainstbyinsidersoccupyingpowerfulpositions,whichhistoricallyhavebeenmen.Womenmayhaveapreferenceforalternativeoccupationsthatoerabetterbalancebetweenprofessionalandpersonallife.Additionally,inthecaseofpoliticalrepresentation,votersmayhavepreferencesformalerepresentatives.Partieswouldthereforerationallyrespondtothisbiasby eldingfewerwomencandidates.Wearguethatitispossibletotestamongthesehypothesesbylookingattheeectsofintroducingfemalecandidatequotas.Whenpartypoliticalmachinesdonotdiscriminateagainstwomentheychoosetheircandidatestomaximizeelectoralresults.Wewouldonlyobservefewerwomenthanmenifeithertherearenotenoughpotentialfemalecandidates,oriftheirvotersprefermalepoliticians.Hence,anyconstraintimposedbyfemalequotasontheparty'sselectionofcandidatesmustresultinworseelectoraloutcomes,eitherbecausenotenoughtalentedwomenareavailabletoful llthequota,orbecausevotersdislikehavingmorewomen.Incontrast,whenpartymachinesdiscriminateagainstwomen,theynominatefewerwomenthanwouldbeoptimalattheexpenseofbetterelectoralresults.Asaresult,femalequotasshouldincreasethevotingsharesofpartiesthatwere eldinglessfemalecandidates.Weexploitanexogenouschangeinelectoralrulestotesttheimplicationsofthedierenttheoriesoffemaleunder-representation.Spain'sEqualityLawoersusanunparalleledopportunitytolearnabouttheseissues.Thelawwaspassedin2007topromotetheequalitybetweenmenandwomen.Thepassageofthelaw(anindirecteectoftheMadridterroristbombings)hadbeencompletelyunanticipatedbylocalpoliticalmachines,candidates,andvoters.Itrequiredpoliticalpartiesto eldcandidatelistsforlocalelectionswithaminimum40percentoffemalecandidates.1However,thequotasonlyappliedtomunicipalitieswithmorethan5000inhabitants.Usingaregressiondiscontinuitydesign,wecancontrolforfactorsthatdieredacrossmunicipalitiesandthatchangedfromoneelectionyeartoanother,suchasvoters'preferencesorattitudestowardswomen.Wefocusonallelectorallistsandoutcomesinboththeelectionspriortothelaw(2003)andafter(2007).We ndthatthelaweectivelyincreasedthepresenceofwomeninthelistsby8percent-agepointsinthetreatedmunicipalitieswithrespecttothecounterfactual.Thisrepresented 1TheelectionoftowncouncilorsinSpainisdoneusingclosedlists,whereeachlistmustcontainanumberofcandidatesequaltothecouncilseatsthatarebeingcontended.2 anincreaseof25percentinthenumberoffemalecandidates.Moreover,thelawforcedpar-tiestomaintainthesameminimumof40percentofwomenonevery ve-positionbracketofthelist.Asaresult,thenumberofwomenonthetop5positionsalsoincreasedbyasimilaramount.Voterturnoutinthemunicipalitiesaectedbythelawwasnotreducedasaresultofthelargernumberofadditionalwomencandidates.Thiswastrueregardlessofthemagnitudeofthelocalchangeintheshareoffemalecandidates,whichdependedontheaveragedistancetothe40percentquotainthepre-treatmentelections.Thenumberofnullorblankvotes,whichcouldbethoughtofasprotestvotes,measuringdisapprovalofthepoliticalsystem,wasnotaectedeither.Theseresultsarenotconsistentwiththeexistenceofmajorvoteraversiontofemalecandidates.Moreover,therewassigni cantheterogeneityintheinitialnumberofwomenbylistandmunicipality.Apriori,thisfactcouldhavebeeninterpretedassignalingthepresenceofgenderpreferencesamongvotersthatarecorrelatedwithideology(e.g.conservativevotersinruralareasmayhavedierenttastes).However,we ndthatpartiesthatwereforcedtomakelargerincreasesinthenumberoffemalecandidatesimprovedtheirelectoralperformancerelativetotheotherparties.Thustheevidenceisnotconsistentwiththeexistenceofvoter-speci ctastesagainstfemalecandidates.Onthecontrary,atthemargin,votersseemedtobehappierwithmorebalancedlists,suggestingthatpartieswerenotoptimizingthegendercompositionoftheirlistspriortothelaw.Thisisanimportantresultinthecontextofthecon ictingexistingresearchoftheimpactofwomenonelectoraloutcomes.PapersbasedonvotersurveysintheUSclaimto ndevidenceofvoterdiscriminationbutcorrelationalevidencesuggestsaneutralorpositiveimpactoffemalepoliticiansonelectoraloutcomes(Dolan,2004,2005).Inthiscontext,weprovideevidenceofapositiveimpactofwomenpoliticiansusingacleanquasi-experimentaldesignthatexogenouslyincreasedthenumberoffemalecandidatesinatreatmentgroupofmunicipalities,butdidnotinasimilarcontrolgroup.Wefurther nd,asthepartydiscriminationtheorypredicts,thatthepositiveeectofwomenonvoteswasconcentratedinmunicipalitieswithnon-competitiveelections,whereonepartyhadenjoyedanabsolutemajorityinthe2003elections.Consistentwiththis nding,womenhadbeenmoreseverelyunderrepresentedinthesaferseatsofincumbentpartiesintheseuncompetitiveenvironmentspriortothelaw.Wealsoshowthatpartiesdidnothavemajorproblems ndingsuitablefemalecandidatesto llinthequotas.TheEqualityLawwasnotassociatedwithincreasedlistattritionordi cultiesfornewliststoarise.Partiesdidnotneedtoretainpastfemalecandidatesmoreoftenortopromotefemale llercandidatestotoppositionseither.Togetherwiththe3 resultsonelectoraloutcomes,thisevidenceisnotconsistentwiththeexistenceofmajorsupplyconstraintsofhigh-qualityfemalecandidates.Theseresultsdonotseemtobedrivenbywomen'slackofadesiretocompeteforapositioninthelist.Whilewomenmayshyawayfromcompetitiononaverage,alargenumberofwomenincompetitivelistsleveragedthequotatoattainhigherpositionsthantheminimumonesguaranteedbythelaw.Thisevidenceisnotconsistentwithgeneralizedfemalepreferencesforlowerspotsinthelists.Theevidenceseemsmostconsistentwiththeexistenceofentrenchedmale-dominatedpoliticalmachines,especiallyinlesscompetitiveenvironments.Finally,we ndthatdiscriminationagainstfemalesbypartymachineshadnotimpacteddisproportionallythesupplyofhighly-quali edfemalecandidates.NamesandsurnamesarequiteinformativeofeducationalandsocioeconomicstatusinSpain.However,thenewwomenattractedbythequotaswerenotonaverageofhighersocioeconomicbackgrounds,ascapturedbytheirnames.Thepapercontributestoanextensiveliteratureondiscriminationinlabormarkets(Becker,1957).Thereisclearevidenceofgenderdiscriminationinhiring(GoldinandRouse,2000),andproductmarkets(AyresandSiegelman,1995).Arelatedliteratureonethnicdis-criminationhasstudiedmarket-drivenpreferencesforsegregation(seeSaizandWachter,2011,forarecentexampleandreferencestherein)andinotherlessconventionalenviron-ments(Kahn,1991;Szymaski,2000;PriceandWolfers,2010).Municipalelectionsallowustostudyaveryrelevantsetting,yetonewhereaverygoodperformancemeasureisavailable:electoralresults.Moreover,becausequotaswereunexpectedandonlyimposedonsomemu-nicipalities,wehaveanessentiallyrandomselectionoftreatedanduntreatedmunicipalities,thatallowsustoeliminatetheeectsofpotentiallyconfoundingfactors.Ourpapercomplementspreviousevidenceabouttheimportanceofinternalpoliticalmachinestoaccountforfemaleunder-representation.Sanbonmatsu(2002) ndsthatwomenarelesslikelytoberepresentedinUSstatelegislatureswherepartieshavemorecontrolovernominations.BaguesandEsteve-Volart(2009)showthatpartiesplacefemalecandidatesstrategicallyinthepositionsthatarelesslikelytoaectthechancesofelectionofmalecandidates.Thepaperalsorelates,somewhatdistantly,toalargerecentliteratureonthepolicyeectsoffemalesinpower(e.g.ChattopadhyayandDu o,2004;Beamanetal.,2009).2Thisliteraturehasfocusedmainlyonthepolicyoutcomesoffemaleelected 2Theliteratureontheimpactofwomeninpoweronparliamentaryvotes,budgetlevels,budgetcomposi-tion,governmentstability,andgovernmente ciencyisnowquitelarge.Notethatwedonothavemuchtocontributetothisspeci cliterature,sincewearenotexaminingthepolicyimpactoftheadditionalelectedwomenduetothequota.,Instead,weusethenaturalexperimenttolearnaboutthecausesoffemaleunder-representationinthelists.Forthereaderinterestedonwomenpoliticiansandoutcomes,otherexamples4 leaders,eectsofquotasonthenumberofelectedfemales,orthechangeinattitudestowardswomenoncetheyareelected,andnotontheintrinsiccausesthataccountforwomenunder-representation.Finally,thepaperisrelatedtoanemergingliteraturethattriestoexplaingenderin-equality.Severalexplanationshavebeenproposed,suchastheexistenceofdierencesingenderattitudestowardscompetition(GneezyandRustichini,2004;Gneezyetal.,2003).Thesetheoriesarefocusedonexplainingwhytheaveragewomanmaybelesslikelytoseekpowerfulpositions;however,theycannotbythemselvesaccountforthelackofwomenatthetopsince,inacompetitiveenvironment,therecouldbeenoughavailablewomenattherighttailsoftheambition-abilityjointdistributiontosatisfyhypotheticalcustomerdemandformorebalancedgenderallocationsintherelativelyfewpowerfulpositions.Asweargueusinganelectoralmodel,wewillnotbeabletofullyunderstandfemaleunder-representationinpositionsofpowerunlessweconsidertheinteractionbetweenthedemand-sideinpoliticalmarkets(voterpreferences),themarginalsupplyofquali edfemaleinputsatthetailsoftheabilitydistribution,andtheindustrialorganizationofthemarket(theroleofpartiesandthedegreeofelectoralcompetition).Thepaperproceedsasfollows.Insection2wedescribetheSpanishmunicipalelectionsandtheir2007reform.Insection3wedevelopasimplemodelofelectoralcompetitionthatneststhetheoriesthatcanpotentiallyexplainlowfemalerepresentation.Themodelillustrateswhateachofthesetheoriespredictstheeectsoftheelectoralreformshouldbe.Section4describesthedatathatweusetotestthedierenttheories,andsection5explainstheresults.Somealternativeexplanationsareassessedinsection6.Weoersomeconcludingremarksinsection7.2MunicipalElectionsandElectoralReformTheelectionoftowncouncilorsinSpainisdoneusingclosedlists.Thismeansthatvotersvoteforalist,ratherthanaperson.Eachlistmustpresentanumberofcandidatesequaltothenumberofcouncilseatsatstake(11formunicipalitiesbetween2,000and5,000inhabitants;13formunicipalitiesbetween5,000and10,000).Alargenumberoflistsconcurundertheumbrellaofnationalorregionalpoliticalparties,albeittheexistenceoflocalindependentlistsisnotuncommon.Votersvoteforthelistsasawhole,andthecouncilseatsareapportionedproportionallytothenumberofvotesaccordingtothed'Hontmethod.The ofthisburgeoningliteratureinclude:Welch(1985),Swers(1998),Rehavi(2007),Clots-Figueras(2009),FunkandGathman(2010),DePaolaetal.(2010),GallarduciandPasserman(2010),FerreyraandGyourko(2010),CavalcantiandTavares(2011).Campa(2010)studiestheimpactofquotasinSpainontheprovisionofchildcare.5 law,however,establishesaminimumvotethresholdof5%inorderforalisttoqualifyfortheapportionmentofcouncilseats.Thecouncilmembersarethendrawnfromeachlistusingtheexactorderinginwhichthecandidatesappearlisted.Uponconveningforthe rsttime,thecouncilelectsamayor,typicallythe rstpersoninthemost-votedlist.Importantly,onlythe rstpersonineachlistcanbeconsideredintheinitialmayoralvote.Thecouncilalsoactsasarepresentativelegislativebodypassingandenactingalllocalbudgets,laws,regulations,andtaxcodesforaperiodoffouryears.OnMarch22,2007,theLawfortheequalityofwomenandmenwaspassedbytheParliament.3Itrequiredallcandidatelistsforthemunicipal(andother)electionstocontainatleast40percentofcandidatesofeachgender.Moreover,inordertopreventpartiesfromplacingallwomenatthebottomofthelist,thelawrequiredthisproportiontobemaintainedforeverybracketof vepositions.Nevertheless,thelawspeci edanexceptiontothequotainthosemunicipalitieswithfewerthan5,000inhabitants.Thelawappliedforthe rsttimetothemunicipalelectionsthatfollowedonMay27,2007.InthepreviouselectionofMay25,2003,nosuchlegalchangehadbeencontemplated.Indeed,thepassingofthelawwasmadepossiblebytheresultsofthe2004generalelection,whichwerelargelyunanticipated(Montalvo,2009).OnlydaysbeforetheMadridtrainbombingsofMarch11th2004theChristian-DemocraticParty(PP)waswidelyexpectedtowintheelections.Thebombingsandpost-attackmanagementfromtheincumbentpartychangedthesentimentofmanyvoterstowardstheSocial-Democraticparty(PSOE),whichwontheelectionsfourdaysaftertheterroriststrike.Itisthereforequiteunlikelythattheshareoffemalecandidatesinthemunicipalelectionsof2003re ectedananticipationofthefemalequotasthatwereimposedin2007.3FemaleRepresentationandQuotas:AComprehensiveFrameworkThereareseveralreasonswhythenumberofwomeninpowerfulpositionsmaybelow.First,politicalpartiesmaybehaveasiftheywerediscriminatingagainstwomenby eldinglesswomeninelectablepositions(Sanbonmatsu,2002,2006;BaguesandEsteve-Volart,2009).Thiscouldbedueto(traditionallymale)powerfulelitesexplicitlyfailingtopromotewomen.Alternativelyitcouldbeduetointernalorganizationaldynamicsthatimplicitlyfavormen(Reubenetal.,2010),orbythedesireofwomentoavoidcompetition(Gneezyetal.,2003,NiederleandVesterlun,2007)forpowerwithintheparties(non-adversarialbehaviorwithin 3Thislawhasabroadreach,aectingboththepublicandprivatesectors.Butwedescribehereonlytheimplicationsforthemunicipalelectionsthatwestudyinthispaper.6 organizations).Alltheseexplanationscanbeencompassedunderwhatwenamethepartydiscriminationtheory.Asecondreasonforobservinglowfemaleparticipationresidesinthetastesofvoters.Aslongasvotersprefertobegovernedbymen,politicalpartieswillrationallyrespondbyrestrictingaccesstowomen.Theywilldosoinordertoincreasethechancesofwinning,evenifthepartiesthemselvesdonothaveatastefordiscrimination.Wecallitthevoterpreferencetheory.Du oandTopalova(2004)andBeamanetal(2009) ndevidencethatmalevotersinIndiareportpreferringmalepoliticians.However,theyalso ndthatdiscrim-inatorypreferencesdissipatewithbetterknowledgeofthecandidate.Theseareimportantcontributionsinthecontextofvillagesinadevelopingcountrywithverylargesocialstrati -cation.Neverthelesslessisknownabouteectivevotingpreferencesforwomencandidatesinmoredevelopedcountries,withexistingresearchinpoliticalscienceprovidingargumentsonbothsidesbasedoncorrelationalandsurveyevidence(seeDolan,2004,2005,foraUS-baseddiscussion).Finally,lowfemaleparticipationmayalsobetheresultofself-selectionintopoliticalactivitiesleavingpartieswithaworsepooloffemalecontenders.Thisexplanationisrelatedtostatisticaldiscriminationtheories(Arrow,1973,andKnowlesetal.,2001).Potentialfemalecandidatesmaybelessablethanmalecontendersifpoliticsisinherentlymorecom-petitive,andwomenshyawayfromcompetitiveenvironments(non-adversarialbehaviorinopenmarkets).Talentedwomenmayalsoavoidacareerinpoliticsifsuchjobsoeraworsefamily-lifebalancethanotheralternatives(LawlessandFox,2005;GoldinandKatz,2008;Bertrandetal.,2010).Ingeneral,asdocumentedbyCrossonandGneezy(2008),thereareanumberofpsychologicaldierencesbetweengendersthatcouldhavehigh-abilityfemalesseethemselvesaslesssuitablepoliticians(LawlessandFox,2005).Wetermthisexplanationtheself-selectiontheory.Inthissectionwedevelopatheoreticalframeworkthatencompassesthesethreetheoriesoflowfemalerepresentation.Themodelhighlightsthedierentpredictionsthatthethreetheoriesmakeabouttheeectsofimposingfemalequotasonparties'lists.Thesepredictionswillhelpusdiscernamongtheempiricalvalidityofthedierenttheories.Themodelisalsohelpfulasanintellectualplaceholderthatconsolidatestheideasinwhatisaverydiverseliteratureonwomeninpower.3.1TheSetupConsidertwopartieslocatedattheextremesoftheinterval[0;1].Eachpartymustchooseacontinuumofcandidateswithamassof1.Thefractionofcandidatesthatareelectedfrom7 thislistisdeterminedbythefractionofvotesthateachpartygets.Thereisacontinuumofvoterslocateduniformlyon[0;1].Theirlocationonthelinerepresentstheirpolicypreferences.Voterscareforthepoliciesoftheparties,fortheaveragecompetenceoftheircandidates,andtheshareofwomenontheirlists.Werepresentthepreferencesofavoterlocatedinx2[0;1]as:u(x;p)=v�t 2jp�xj+Ap�dv(x)Wp,whereu(x;p)representstheutilityfromvotingforthepartylocatedinp2f0;1g.Theterm�t 2jp�xjcapturesthedisutilityfromvotingapartywhosepolicypdiersfromthepreferredpolicyx.Theparametertmeasuresthedegreeofcompetitionamongthetwoparties(howsubstitutabletheyareforvoters).Apistheaverageabilityofpartyp'scandidates,andWpistheshareofwomenonitslist.Theparameterdv(x)0measuresvoterx'sdistasteforvotingforfemalecandidates.Weassumethatdv(x)isamonotonicfunction,andsatis esjd0v( x)j1forallx2[0;1].4Alternatively,avotermayalsoabstainfromvoting,andobtainu(x;;)=u .Thisutilitycanbethoughtofastheopportunitycostofvoting.Weassumethatpartiescannotchoosetheirideologyandpolicies(setatthenationallevel)butcanchoosetheidentityoftheircandidates.Theydososimultaneously,álaCournot.Eachpartyhasapoolofpotentialcandidatestochoosefrom.Letfpmandfpwbethedensityfunctionsforthedistributionsofabilityofmenandwomencontenderstobecomecandidatesforpartyp.Weassumethroughoutthatthereexistsaconstantapgsuchthatwecanwritefpg(x)=fg(x�apg)forp2f0;1g,g2fm;wg:thedistributionofabilityisthesameforbothparties,exceptforameanshifter.5Wedenotethecorrespondingdistributionfunctionswithcapitalletters.Partieschoosetheircandidatelistswiththeobjectiveofmaximizingp=Vp=(V0+V1)�dpWp,whereVpisthenumberofvotesreceivedbypartyp,anddpisthepartymachine'sdistasteforhavingwomenonitslist.Therefore,bothpartieswanttoincreasetheirshareofthevotes,butmayalsoliketodecreasethenumberofwomenontheirlist.Giventheirpreferences,voterslocatedon[0; x0]voteforparty0,voterslocatedon[ x1;1]voteforparty1,andvoterslocatedon( x0; x1)abstain,where xpdenotesthemarginalvoterofpartyp.6Hence,wehavethatV0= x0andV1=1� x1.Noticethatwhennovoterabstains,themarginalvotersofbothpartiescoincide,andwewillwrite x= x0= x1. 4Aswillbecomeclearlater,tofullycomprehendtheempiricalevidence,itisimportanttohavehet-erogeneityinthevoters'preferencesfordiscrimination,andthisheterogeneitybeingcorrelatedwiththepreferencesfortheparties.Ontheotherhand,assumingthatd0v( x)isnottoolarge(inabsoluteterms)guaranteesthattheparties'objectivefunctioniswellbehaved.5Thisencompasses,forinstance,thecaseinwhichabilityisnormallydistributedwiththesamevariance2forallpartiesandgenders,whilethemeanmayvarybypartyandgender.Theassumptionalsoallowsthevariancetovarybygender,aslongasforeachgenderbothpartieshavethesamevariance.6Thisresultfollowsfromtheassumptionthatjd0v( x)j1,thatguaranteesthatthevoters'preferenceoverthelocationofthepartiesdominatesthepreferenceoverthenumberofwomen.8 Thismodelcaptureswithsimplicitythethreetheoriesdescribedabove.Thepartydis-criminationtheorystatesthatdp�0.Underthevoterdiscriminationtheory,dv(x)�0.And nally,theselectiontheorycanbecapturedbythedierencebetweenfpwandfpm.Inthefollowingsectionwedescribetheimplicationsofthesetheoriesfortheselectionofcandidates.3.2TheEconomicsofGenderSelectionForagivenshareofwomen,partiesareinterestedinmaximizingtheaverageabilityoftheircandidates.Theywillconsequentlychoosethebestmaleandfemalecandidatesamongthepoolofcontendersofeachgender.Therefore,partiesoptimallysetstandardsa pmanda pwsuchthatp'slistcontainsallmalecontenderswithaa pmandallfemalecontenderswithaa pw.Asaresult,wehaveAp=1a pmafpm(a)da+1a pwafpw(a)da;andWp=1a pwfpw(a)da=1�Fpw�a pw=Fpm�a pm;wherethelastequalityholdsbecausethenumberofmenandwomenadduptounity.Sincelistshavea xedsize,determiningthestandardforwomenalso xesthestandardformen.Indeed,da pm da pw=�fpw�a pw=fpm�a pmand,eectively,partyponlyhasoneinstrumenttomaximizeitsobjectives:a pw.Considerwhathappenswhenu issu cientlylowtoinduceallvoterstoturnouttovote.Themarginalvoterisnowimplicitlyde nedbytheequation: x=1 2+1 t[(A0�A1)�dv( x)(W0�W1)].Asaresult,partyp'sobjectiveismaxa pwVp�dpWp;whereV0= xandV1=1� x.The rst-orderconditionforthisproblemisdVp da pw�dpdWp da pw=0.Applyingtheimplicitfunctiontheoremtotheequationthatde nes x,anddierentiatingApandWpwithrespecttoa pw,wecanwritethe rst-orderconditionas:a pw�a pm=dv( x)+dp[t+d0v( x)(W0�W1)]:Thisequationde nestheoptimalselectionpolicyofpartyp,determiningthelevelsoftheabilitystandardsformenandwomen.7 7Thesecondderivativeoftheobjectivefunctionisnegativeattheoptimumwhendvisconstant.Therefore,9 Whenneitherpartiesnorvotersdiscriminatedv=dp=0.Inthiscase,a pw=a pm,sothatmenandwomenfacethesamestandard.Partiesmaximizetheaverageabilityoftheircandidatelist,regardlessofthegenderofthecandidates.This,however,doesnotmeanthatwomensharethepositionsequallywithmen:moremenarechosenontheparty'slistiftheyaremorepoliticallyablethanwomen,andfpm rst-orderstochasticallydominatesfpw.Wheneitherpartiesorvotersdiscriminate,theabilitystandardforwomenincreaseswhilethestandardformenislowered.dvanddpdeterminethentheabilitydierencethatpartyprequirestoacceptawomenonthelist.Togainsomeintuition,suppose rstthatvotersdiscriminate,butpartiesdonot,sothatdv(x)�0anddp=0.Whenoptimizingthecandidatelist,partiesareactuallymaximizingtheutilityofthemarginalvoter x.Sincevoter xhasatastefordiscriminationofdv( x),thisvoteriswillingtosubstituteawomanforamaninordertoreduceWpaslongasthedierenceinabilitybetweenthemanandthewomanissmallerthandv( x).Thereforepartiestargetingthismarginalvoterchoosealistsuchthat:a pw�a pm=dv( x).8Inthislist,theabilityofthemarginalwomanexceedstheabilityofthemarginalmanbyexactlydv( x).Ontheotherhand,supposethatvotersdonothavediscriminatorytastes,butpartiesdo,sothatdv(x)=0anddp�0.Nowtheoptimalselectionpolicysetsa pw�a pm=dpt.Partiesarewillingtosacri cevotesinordertodecreasethenumberofwomen.Thebene tsforpartypfromdroppingawomenfromthelist,dp,hastobeweightedagainstthecostinvotes,whichdependsonthedegreeofcompetitiont.Iftislarge,theelectionislesscompetitive,andtheelectoralresultsarelesssensitivetothenumberofwomenonpartyp'slist.Hencethispartycanuseaselectionpolicythatismoreresponsivetoitstastefordiscrimination.Ontheotherhand,iftislow,thecontestisverycompetitive.Votersdonothaveastrongpreferenceforaparty'spolicy.Asaresult,votersaremoresensitivetotheabilityofpartycandidates.Adecreaseinthequalityofcandidatesresultsinmorevotesbeinglosttotheotherparty.Consequently,partiesdonotrespondsostronglytotheirpreferencesfordiscrimination.Whenvotersandpartiesdiscriminatesimultaneously,thedierencesinquali cationstan-dardsformenandwomenre ectthepreferencesfordiscriminationofboth.Partiescandecreasethenumberofwomenwithoutcost(indeed,theybene t)aslongastheirmarginalvoterdemandsit.Whenthedierenceinstandardssurpassesdv( x),increasinga pw�a pmtoreducethenumberofwomencomesatthecostofalossofvotes.Asearlier,partieswill thereexistsa�0suchthatthesecond-orderconditionsaresatis edaslongasjd0v( x)jforallx2[0;1].8Noticethattheoptimalselectionofcandidatesfollowsamarginalistargument:howmuchmoreableshouldthemarginalwomanbethanthemarginalman,forapartytostayindierent.Itisthereforenotsurprisingthattheoptimalpolicydoesnotdependonthedistributionoftalentformenandwomen.Althoughthe nalnumberofwomenthatwillbeselecteddoesdependonthosedistributions.10 stillwanttodoitaslongasthatcostdoesnotexceeddp.Thelossofvotesstilldependsonhowcompetitivetheelectionis.Butnowitalsodependsonhowmuchpreferencesfordiscriminationchangeasthemarginalvotershiftsclosertotheparty'slocation.Thisaddi-tionaleectd0v( x)(W0�W1)ispositiveforthepartywithfewerwomenifitsmoreloyalvotersalsohaveastrongertastefordiscrimination:thispartywould ndthatreducingthenumberofwomenhasasmallercost.Asthemarginalvotergetsclosertoitslocation,hediscriminatesmore,andtherefore,penalizesthepartylessforthesmallernumberofwomen.Whenturnoutislowerthan100%,theresultsareanalogous.Ifu issu cientlyhigh,thevotersinintermediatelocationsabstain.Now,partyp'smarginalvoterisindierentbetweenvotingthatpartyandabstention.Therefore:Vp=2 t[v�u +Ap�dv( xp)Wp],andpartyp'sobjectiveisnow:maxa pwVp V0+V1�dpWp:The rst-orderconditionforthisproblemleadstoasimilarconditioncharacterizingtheoptimalpolicyofpartyp:a pw�a pm=dv( xp)+dp(V0+V1)2 V1�pt 2+(�1)pd0v( xp)Wp:Ifvotershavediscriminatorytastes,partiesrespondbymaximizingthepreferencesofthemarginalvoter.However,themarginalvoterisnowdierentforeachparty.Hence,partiesmayselectdierentnumberofwomensimplybecausetheyaretargetingdierentvoters.Likewise,whenpartiesdiscriminate,theresponsetotheirtastefordiscriminationdependsonthecompetitivenessoftheelectiont.However,thesizeoftheirconstituencynowbecomesimportant.Toseewhy,considerwhattranspireswhenpartyplosesonevoter.Thisvoterwillabstain,insteadofvotingfortheotherparty,asbefore.Whenpartypobtainsthemajorityofthevotesintheelection,anylostvotetranslatesnowintoasmallerdecreaseinthelist'svotingshare.9Therefore,theparty ndsitlesscostlytodiscriminate.Thiseectisanalogoustotheeectoflowcompetition.Toobtainsomeintuitionfortheeectsofintroducingfemalequotas,consider rstwhathappenswhenonepartyisunilaterallyforcedtoincreasethenumberofwomenonitslist.Ifthispartydiscriminatedagainstwomen,theaverageabilityofitscandidateswillincrease,andhencesowillvoters'preferencesforthatparty.Theparty'selectoralresultsimproveasaresult.Onthecontrary,whenvotersdiscriminate,orwhenfemalecontendersarelesstalentedthanmalecandidates,increasingthenumberofwomendecreasesthevoters'utility 9NoticethatthechangeintheshareofvotesofpartypduetoachangeinVp,whichisgivenbyV1�p=(V0+V1)2,isdecreasinginVp.11 fromvotingthatparty,whichwillobtainworseelectoralresults.Thisintuitionsuggeststestingfortheexistenceofpartydiscriminationbylookingattheeectsofaforcedincreaseinthenumberofwomen.However,inpractice,quotasareimposedonallparties.Wemustthereforetakeintoaccounttheequilibriumimplicationsofimposingquotasonbothparties.3.3EquilibriumEectsofFemaleQuotasTheeectofimposingfemalequotasonthecandidatelistsdependsinasystematicwayontheparametersfpw;fpm;dv;anddp.Nevertheless,thethreetheoriesoflowfemaleparticipationdescribedabovemakedierentpredictionsabouttheoutcomes.First,considerthehypothesisofpartydiscrimination.Toisolatetheeectsofthistheory,assumethatvotersdonothaveapreferencefordiscrimination,andtherearenodierencesinthegenderdistributionsoftalent.Underthisstrongnullhypothesis,ifparty0selectedfewerwomenonitslist,itmustbebecauseitdiscriminatedmore(d0�d1)attheexpenseofhighercandidateabilities.Asaresult,imposingbindingfemalequotasincreasestheaverageabilityofitscandidatesbymorethaninparty1.Sincevotersvaluetheabilityofcandidates,party0becomesmoreattractive,andgainsmorevotesthanparty1doeswhenquotasareimposed.Indeed,whennovoterabstains,party0canonlygainvotesattheexpenseofparty1,whosevotesarereduced.Theeectofquotasonturnoutispositive,aslongastheoutsideoptionofvotersisbinding,andsomeofthemdecidenottovote.Otherwise,turnoutremainsunaected.Thefollowingpropositionsummarizestheseresults:Proposition1(Partydiscrimination).Supposethatd0;d1�0,whilefpw=fpmforp2f0;1g,anddv=0.SupposealsothatW0W11 2whennofemalequotasareimposed.Then,introducingafemalequotaW 2W1;1 2(weakly)increasesturnout,andincreasesthenumberofvotesforparty0bymorethanforparty1.Indeed,ifturnoutis100%withoutfemalequotas,thenafemalequotadecreasesthenumberofvotesforparty1.Proof.SeeAppendixA. Notefromproposition1thattheresultsforonepartyarealwaysrelativetowhattheotherpartyisdoing.Ifthetwopartieshavethesameinitialnumberofwomen,thenquotashavenoeectsontheelectoralresultsbecausetheyaectbothpartiesequally.Onlyrelativeincreasesinthenumberofwomencausedbythequotaaectspositivelytheelectoralresults.Iffemaleunder-representationwasdrivenbyvoterdiscriminationagainstwomen,theop-positeistrue.Inthiscase,partiesrationallyrespondedtothevoters'tastefordiscriminationbydecreasingtheshareofwomenontheirlists.Quotas,therefore,decreasetheutilityof12 voters,andmayleadtoadecreaseinturnout.Moreover,ifparty0startswithalowershareofwomenitmustbebecauseitsvotershaveastrongerpreferencefordiscrimination.Theirutilitywoulddecreasebymorethanthatofparty1'svotersfortworeasons.First,havingastrongertastefordiscrimination,party0'svoterssuermorefromanygivenincreaseinthenumberofwomen.Furthermore,becauseparty0hadfewerwomen,itisalsoforcedtoincreasethenumberofwomenbymorethanparty1.Therefore,quotaswillhurtparty0more,anditwilllosemorevotesthanparty1,asaresult.10Proposition2(Voterpreferences).Supposethat,forallx,dv(x)�0andmonotonic,whilefpw=fpmforp2f0;1g,andd0=d1=0.SupposealsothatW0W11 2whennofemalequotasareimposed.Then,introducingaquotaW 2W1;1 2(weakly)decreasesturnout,anddecreasesthenumberofvotesforparty0bymorethanforparty1.Proof.SeeAppendixA. Notethattheexistenceofdierencesinthenumberofwomenbeforethequotaswereintroducedimpliesthatturnoutwaslessthan100%.Dierencesinthenumberofwomencanonlyarisewhenpartiestargetvoterswithdierentpreferences.Finally,supposethatwomenunder-representationwasnotduetopartiesorvotersdis-criminatingagainstfemalecandidates.Inthatcase,bothpartiesseta pw=a pm.Theself-selectionhypothesisthusrequiresfpmto rst-orderstochasticallydominatefpw.Asaresult,thepartywiththelowernumberoffemalecandidatesmusthavefewerhighabilityfemalecontenders,relativetothenumberofmalecontenders.Hence,forcingthepartiestohaveanequalnumberofmaleandfemalecandidatesdecreasestheaverageabilityofthisparty'slistbymore.Thistheoryalsopredictsthatquotaslowerturnout,sincetheydecreasetheutilityofvoters.Proposition3(Self-selection).Supposethatapwapmforp2f0;1g,whiledv=0,andd0=d1=0.SupposealsothatW0W11 2whennofemalequotasareimposed.Then,introducingaquotaW 2W1;1 2(weakly)decreasesturnout,anddecreasesthenumberofvotesforparty0bymorethanforparty1.Indeed,ifturnoutis100%withfemalequotas,thenafemalequotaincreasesthenumberofvotesforparty1.Proof.SeeAppendixA. Itisdi culttodistinguishthevoter-preferencesandself-selectiontheoriesinthedata,astheybothmakesimilarpredictions.Butthepartydiscriminationtheorymakesoppositeclaims.Hence,wecantestwhetherpartiesdiscriminatedagainstwomenbylookingatthe 10Themonotonicityofdv(x)ensuresthisisthecase.13 eectsoftheincreaseinfemaleparticipationcausedbyquotasontotalvoterturnout,andontheshareofvotesreceivedbyagivenparty.Totheextentthatpartiesdiscriminate,weexpecttoseeapositiveeectofanincreaseinfemaleparticipationonvoterturnout.Thetheoryalsosuggeststhatthevariablethatdeterminesthechangeintheperformanceofagivenpartyisnottheabsoluteincreaseinitsnumberofwomencausedbythequota.Instead,itistherelativeincreasewithrespecttotheotherpartythatisrelevant.Totheextentthatpartiesdiscriminated,arelativeincreaseinfemaleparticipationinagivenpartywillaectpositivelyitsperformanceintheelections.Wenowtakethesepredictionstothedata.4TheDataTheSpanishStateDepartment(MinisteriodelInterior)collectsinformationrelatedtotheelectoralprocess.Onrequest,weobtainedanon-con dentialsubsetoftheirdata.Concretelywewereprovidedwiththenameofallcandidatesbylistinallmunicipalballotsinthe2007and2003elections11,theirgenderin2007(adisclosurerequiredbytheEqualityLaw),theirlista liationtomajorparties,informationabouteachindividual'spositioninthelists,andonwhethertheyresultedelected.Wealsohaveinformationaboutthenumberofvotesforalllistspresented,fractionofnullorblankvotes,theestimatedmunicipalpopulationforlegalpurposes,andthenumberofregisteredvotersineachtown.Weimputedgenderin2003byusingthe rstnameofthecandidate.NamesinSpainhaveverystronggenderorientationandonlyaverysmallportionofcandidatesin2007hadnamesthatcouldbeascribedtoeithersex,whichmadetheimputationstraightforward.Sincethelawappliedonlytomunicipalitieswithmorethan5,000inhabitants,wecanonlyobtainmeaningfulresultsaroundthisthreshold.Populationsizesfollowanapproxi-matelyParetodistribution,sothereisaverylargeamountofverysmallmunicipalitieswithpopulationinthefewhundredsthatprobablydonotprovideuswithaverygoodcomparisongroup.Inordertoensurebothcomparabilityandlargeenoughsamplesizeswethereforerestrictoursampletothosemunicipalitieswithpopulationsbetween2,000and8,000inhab-itants.However,theresultsarenotintheleastsensitivetovariationsinthisthreshold,aswedemonstratelater.Wealsocollecteddataonunemploymentrates(averygoodlocalsocioeconomicstatusindicatorinSpain),andothereconomiccharacteristicsofthetowns.Theseadditionalcharacteristicsnevermatteredforanyresultsinthepaperandaremostlyomittedinsofarmostspeci cationsincludemunicipal xedeects.Furthermore,theevi-denceisconsistentwiththeexperimentaldesign:therewerenodierencesinobservableson 11Othercharacteristics,suchasthebirthdate,havebeensuppressedfromthedataforcon dentiality.14 bothsidesoftheEqualityLawdiscontinuity12.5EmpiricalEvidence5.1EqualityLawandtheSurgeinFemalePoliticiansThelegalreformimplementedin2007requiredallpoliticalpartiestopresentcandidatelistscontainingatleast40percentofwomen.However,itisnotobviouswhatthepercentageofwomenpoliticianswouldhavebeenintheabsenceofthequota.InTable1weexaminenonparametricallywhetherthelawhadarealsizableeectontheproportionofwomeninthecandidatelistsofthe2007municipalelections.Wedosobypresentingtransitionmatricesforrepeatlistsbynumberofwomeninboththe2003and2007elections(preandpostquota).Repeatlistsarethosethatusedthesamepartya liationinbothelections.Infact,72percentofpartycandidaturesin2007werealsopresentinthepreviouselections.13Itisveryclearthatthelawwasbindingandthatmostrepeatlistsquicklycompliedwiththenewlegislation.Thedierencesbetweenthecontrolmunicipalities(withpopulationsbetween2,000and4,999)andtreatment(between5,000and8,000)arestark.Thedistribu-tionoflistsbynumberofwomenwasquitesimilarinthe2003elections.However,almostallcandidatureswithlessthan vewomeninthelistin2003inthetreatedmunicipalitiestransitionedtowardhaving veormore.TheEqualityLawallowedforroundingthepro-portionofwomentotheclosestinteger,andthusinordertocomplyacandidatureshouldhaveatleast5women.However,mosttreatedcandidaturesdecidedtoincludeatleast6women,perhapsbecauseofuncertaintyabouttheapplicationofthenumericalthresholds.Only1.13%ofthelistsseemednottofullycomply.14InTable2weusearegressiondiscontinuityapproachtoassesstheactualaveragetreat-menteectofthelawonfemalepoliticalparticipation.Thisdesignallowsustousetheshareofwomeninthelistsofthecontrolmunicipalitiesasacounterfactual,andtoestimatetheaveragetreatmenteectofthelawonfemaleparticipationandotheroutcomes.Theregressionformatalsoallowsustoincludeallelectorallistsbymunicipalitypreandposttreatment.Incolumn1westartbyusingonlythedatafromthe2007electionsatthemunicipallevel.Themodelthatweestimatetakestheform:Wi07= DiEL+Pd( 0;Popi07)+"i07(1) 12InterestedreadershouldseeAppendixTable213Westudypatternsoflistattritionlater,andwewillshowthatlistattritionisnotaprobleminthisrepeatsampleestimation.14Itisalsopossiblethatthegenderofsomeofthecandidatesmayhavebeenrecordedwitherror.15 Wherethesubscriptidenotesamunicipality;Wi07isthedependentvariableofinterest:theshareofwomenconcurringtothemunicipalelectionsin2007ineachtown;DiEL=0ifPopi075000andDiEL=1ifPopi07&#x]TJ/;༕ ;.9;Ւ ;&#xTf 1;.34; 0 ;&#xTd [;5000isthetreatmentdummysignifyingthelocaladoptionoftheEqualityLaw,Popi07isthemunicipalitypopulationin2007,andPd( 0;Popi07)isapolynomialofdegreedwithavectorofparameters 0thatvaryonbothsidesofthepopulationdiscontinuitythatde nesthelaw(5000inhabitants).Notethatthepolynomialonpopulationisfullyinteractedbytheinequalitylawdummyanddierentparametersareallowedonbothsidesofthediscontinuity.Wechooseafunctionalformwhered=3,butresultsarenotsensitiveatalltoourspeci cationofPd( 0;Popi07)(suchashigherorderpolynomials,includingasimplelinearpopulationcontrol,oromittingtheterminthemunicipality xedeectsregressionsthatwe nallyimplementthroughout).Theresultsshowa7.2%increaseintheshareofwomenintheparties'listsforthemunicipalitiesthatwereaectedbythelaw,ascomparedtounaectedmunicipalities.Theidenti cationstrategyisbasedontheassumptionthattheshareofwomenpoliticiansinatownmaychangecontinuouslywithrespecttopopulationlevels,sothatanyeectsassociatedtomovingfromamunicipalityjustbelow5000inhabitantstooneaboveareduetothelaw.Incolumn2weshowtheresultstoberobusttoregionaldynamicsbyincluding xedeectsforeachofthe50provincesinthecountry.15Wenextcheckiftheresultscouldbedrivenbyexogenousfactorsthatchangediscon-tinuouslywhenamunicipalityreaches5000inhabitants.Firstwerunaplaceboregressionforthe2003elections,wherenoequalitylawexisted.Theresultsincolumn3yieldaverypreciselyestimatedzero,suggestingthattheeectsin2007weredrivenbythelegalchange.Tocontrolforunobservablevariablesthatdieronbothsidesofthediscontinuitymoregen-erally,wenextspecifyempiricalmodelsthatusebothelectionyears:preandpostEqualityLaw.Theequationthatweestimatebecomes:Wit= DiELD07+D07+Pd( 0;Popit)+XMi+"it(2)Inthisequationtisatimesubscript(t=03,07),D07isapost-treatmentdummy(foryear2007).Themodelissaturatedwith1,437municipal xedeects(Mi),oneforeachtownintherelevantsample,andallvariationiscomingfromchangeswithinmunicipalities.Theresultsincolumn4showconclusivelythattheincreaseinwomenpoliticianswaslargerintreatedmunicipalitiesafterthetreatment.Inquantitativeterms,treatedmunicipalitiessawanincreasein8.5percentagepointsinthefractionofwomeninlists.Sincethefractionof 15Inunreportedregressionswealsocontrolledfortheunemploymentrateandothermeasuresoflocalindustrialcompositionandamenitieswithoutanychangeintheresults.Noneofthecontrolswassigni cantinexplainingwomenpoliticalpower.16 womeninthetreatedsamplewas33percentbeforetheEqualityLawthisimpliesaverylargeshocktothenumberofwomencandidatesof25percent.16OneoftheinterestingaspectsoftheEqualityLawisthatitrequiredthepresenceofatleasttwowomenforevery vecandidatesasrankedinthelists.Thus,partiescouldnotplaceallwomenatthebottomofthelist.Thiswasquiteimportantforthenewwomentohavechancesofbeingelected.78.49%ofthelistsinthetreatedintervalofmunicipalities(anaverageof3listsbymunicipality)obtained5seatsorlessinthe2007elections.Only2.5percentoflistsobtainedmorethaneightseats.Therefore,individualsinlowerpositionsinthelistof13arewidelyperceivedastokencandidates.We nd,incolumn5,thatthelawdidhaveaverysimilarcausalimpactinthetop5positionswithrespecttothecounterfactual.However,incolumn6wealsoshowthattheincreaseinthepercentageofelectedwomenwasaboutonlyhalfofthemagnitudeoftheincreaseinwomenparticipation.Thissuggeststhatthenewwomenwereplacedinlesselectablepositionsinthoseliststhathadchancesofwinning,anaspecttowhichwecomebacklater.17Overall,thedatashowaverysigni canteectoftheequalitylaw,increasingtheshareofwomeninthemunicipallistsand,toasmallerextent,theshareofelectedwomen.Moreover,theeectsareveryconsistentthroughoutallthespeci cations,suggestingthisnaturalexper-imentprovidesuswitharobustempiricaldesign.Hereafterwetreattheincreaseinfemalepoliticiansinthetreatedmunicipalitiesasexogenousandexploitthissourceofvariationtoidentifythecausaleectoffemalepoliticalcandidacyonturnoverandelectoralresults.5.2AvailabilityofQuali edorWillingWomenItispossiblethatfemalepoliticalunder-representationbeexplainedbythelowersupplyofquali edandwillingfemalecandidates.Afterall,accordingtotheSpanishstatisticalo ce(InstitutoNacionaldeEstadÃstica)only34.08%ofpartya liatesinthelargestpartieswerewomenin2007,andanestimatefor2003isofabout33%(Verges,2006).Yetthesestatisticscannotaccountfortheevenlowerpercentageofwomenatthetopofthelocallistsin2003(only16%),ortop2(22%),orforthesmallershareofelectedwomenin2003(28%),suggestingthatwomena liatedwithpartieswereplacedinlesselectablepositions.Furthermore,thenumberofrelevantpositionsinthelistsisverysmallanditshouldbestraightforwardto ndenoughfemalecandidatesifvoterdemandjusti edincreasingthefemaleshare.Indeed,weprovideherefurtherevidencethatisinconsistentwithparties 16InAppendixTable2wevarythebandwidtharoundthetreatmentthresholdof5000inhabitantsinvari-ousways,andtheresultsareentirelyconsistentthroughout,notonlyintermsofthesignofthecoe cients,butalsointhesizeoftheeects.17Iftheorderoffemalecandidatesinthelistswererandom,wewouldobtainthesamecoe cientfortheeectofthelawontheshareoffemalecandidatesandfemaleelectedcandidates.17 havingtroubleto ndquali edfemalecandidatesfortheirlists.The rsttesttoseeiftherewasalackofquali edfemalecandidatesimpliesanalyzingthepatternsoflistsurvivalafterthelaw.Thisanalysisisalsoimportanttoassesstowhichextentpreviousresultscouldbeexplainedbylistattritioncorrelatedwiththelaw.InTable3,column1,wepresenttheresultsofalinearmodelwhereweaccountfortheprobabilityofa2003listnotrepeatingin2007.Theregression,whichincludesnationalparty xedeectsandpopulationsizepolynomials,doesnot ndanyrelationshipbetweenlistattritionandfemalequotas.Thenumberofwomenin2003wasnotgenerallyassociatedwithattritioneither,andtheonlyvariablethatpredictsattritionwastheshareofvotein2003(more-votedpartiestendedtorepeat).Theresultsincolumn2showfurtherthatitdidnotappeartobesubstantiallyhardertocomeupwithminimally-quali edfemalecandidatesforthoseliststhatwerefarfromthe40percentfemaletargetinthetreatedmunicipalities.Finally,incolumn3,weseethatthelawwasnotassociatedwithalowerpropensityfornewliststoariseeither.Hereweareusing2007listdataandthedependentvariableisadummythattakesvalueoneifthelistdidnotappearinthepreviouselections(anewlist).Iftheabsenceofquali edorwillingwomenhadbeenamajorissuewewouldhaveexpectedto ndsomeeectonlistformation,ifonlyforthemoremarginallists.Anothermarginofadjustmentifquali edwomenwerescarcewouldhavebeentoin-creasetheretentionrateofexistingwillingfemalecandidates.Sincewehavethefullnameofcandidateswecantracetheirparticipationacrossyearsandlists.Interestingly,intheuntreatedmunicipalitiesin2007theprobabilitythatawomanhadalreadybeeninthelistonthepreviouselection32percentwasmuchlowerthantheequivalentpercentageformen:43.55percent.Womencandidatesexperiencedhigherturnoverbetweenelectionsthanmen,afactthatisnotexplainedbythepositioninthelisttheyoccupied(similarresultsareobtainedwhenconditioningonlistrank).Therefore,ifquali edwomencandidatesweretrulyscarceonewouldexpectmoreeortsgearedtowardincreasingtheirretentionrate.InTable4,weestimatelinearprobabilisticmodelswiththefollowingstructure:Xcit='1DiELD07+'2D07+Pd('0;Popit)+XMi+"cit(3)whereXcitisadummyorindicatorvariabledescribinganattributeofcandidatecintowniattimet.Incolumn1thedummyvariabletakesvalueunityifthecandidateisawomen:'1hereistheimpactofthequotaontheshareofwomen.18Incolumn2thedependentvariableisnowadummythattakesvalueoneifthecandidateisawomenappearingforthe 18Theestimatesarenowslightly-dierentlyestimated.Concretely,inTable2eachmunicipalityisanobservationwhereasherewearegivingmoreweighttomunicipalitieswithmorecandidates.Notethatwearenowgenuinelyinterestedincandidate-levelestimates.18 rsttimeinthelistin2007.Thecoe cientestimatesareverysimilartothoseincolumn1:partiesdidnot llinthenewfemalepositionsrequiredbythequotabyincreasingtheretentionofwomen,butjustbroughtinnewwomentothelists.Incolumns3and4werepeattheseregressions,thistimefocusinginthetop vepositions.Whilethequotahadtoimplyanabsoluteincreaseinthenumberofwomenoverall,partiescouldhaveeasilyshiftedexistingincumbentlower-rankedcandidatesintothetoppositions.Whatwe nd,however,isthat81percentofthesetop-5positionswereactually lledwithnewwomen,andthisshareincreasesto87percentforthetop-3positions.Whenpartieswereforcedtoplacefemalesintoppositionsnewwomensteppedintotakeonthechallenge.Allinall,existingandnewpartiesintownsaectedbyfemalequotasdidnotmakeeortstoincreasetheretentionrateofpastwomencandidatesanddidnothaveanyproblems llingtheirlistswithmoreandnewfemalecandidates.Atthetop,partiesdidnotneedtopromotelower-rankedcandidates.Therewasnolistattritionorincreaseddi cultiesfornewliststoarise.Thisevidenceisnotconsistentwiththeexistenceofmajorsupplyconstrainsonquali edfemalecandidates.5.3DoVotersDislikeFemalePoliticians?5.3.1VoterTurnoutandProtestVotesAswediscussedearlier,animportanttesttogaugetheexistenceofvoterpreferencesordierencesinqualitybetweenmaleandfemalecandidatesistoexaminechangesinturnoutratesassociatedwithanexogenousincreaseinthenumberoffemalepoliticians:didsomevotersstayathomemoreoftenratherthanvotingforfemalecandidates?Intable5weanalyzetherelationshipbetweenfemale-politicianfrequencyandvoterparticipation.Voterturnoutisquitehighintheselocalelections(withanunweightedaverageof73percentofregisteredvotersacrossthetreatmentandcontrolmunicipalitiesattendingtothepolls).Itisverydi cultaprioritoknowifthepresenceofwomenpoliticiansisfrowneduponbysomevoters.Infact,inTable5column1,simpleOLSregressionsatthemunicipalityandyearlevel(standarderrorsclusteredbymunicipality)pointintheoppositedirection:municipalitieswithmorewomenintheliststendtohavehighervoterturnouts.Atownwithatenpercentagepointhighershareoffemalepoliticianscanbeexpectedtohavea1.45percentageincreaseinvoterturnout.Ofcourse,thismayonlyre ectthefactthatintownswithamoreparticipativeculture(orsocialcapital)womenarealsomorelikelytobecomepoliticians,aninterestingfactbyitselfthatwered- agforfutureresearch.However,theEqualityLawprovidesusforanexogenoussourceofvariationtoexaminethecausalimpact19 offemalepoliticiansoncitizeninterestinlocalpolitics.Todosowedeploytheempiricalmodel:VTit= 1FemShait+ 2D07+Pd(0;Popit)+XMi+"it(4)WhereVTstandsforthevoterturnoutrateintowniattimet=03;07;FemShaitistheshareoffemalecandidatesintheelectionsbymunicipalityandyear,andtherestofthenotationisconsistentwithpreviousspeci cations.InordertotakeadvantageofthelargeexogenousincreaseinwomenparticipationweinstrumenttheshareofwomencandidateswithDiELD07,whichiseectivelyanindicatorforthetreatedmunicipalitiesafterthelawwasenacted.Theresults,inTable5column2,suggestthatthepositiveassociationbetweenwomenpoliticalactivismandvoterparticipationwasnotcausal.Thecoe cientisarelativelypreciselyestimatedzero.Column3usesamoreelaborateinstrumentalvariablestrategy.Theshareofwomencandidatesbymunicipalityin2003wasnotaectedbythechangeinlaw,asitcouldnotbeanticipatedatthattimethatsuchlawwouldbepassed.Therefore,wecanconstructameasureofhowbindingthelawwasforeachmunicipalityasthedierencebetweentheshareoffemalecandidatesin2003andthe40percentrequiredin2007.WenowthereforeuseDiELD07maxf0;0:4�FemShai03gasalternativeinstrument,yieldingthesamenon-signi cantresults.Theadvantageofthisalternativeinstrumentisthatnowweareexploitingdierencesinhowbindingthelawwasacrosstreatedmunicipalities.Inmunicipalitieswherewomenwerealreadyactiveabovethe40percentthresholdthelawdidnotrepresentasigni cantchangetotheelectorallandscape.Butitdidinmunicipalitieswerewomenweremoreseverelyunderrepresented.Infact,incolumn4,wegofurtherandlimitthesampletotreatedmunicipalitiesonly(between5000and8000inhabitants)thereforedispellinganypotentialconcernsaboutthequalityofthecontrolgroup.Allofthevarianceiscomingnowfromthedosageintensityinthetreatmenteect(law-drivenchangeinthefemalepoliticianshare)onthetreatedmunicipalities;theestimatesareidentical.Finally,incolumn5werepeatasimilarspeci cationasincolumn2,thistimefocusingontheeectsoffemalepoliticiansontheshareofblankornullvotes.Thesevotesdonotaecttheresultsoftheelectionsbutvoterscouldusethemasamessageofdisapprovalwiththepoliticalprocess.Totheextentthatvoterswereunhappywithanincreasingleadingroleofwomenintheelectoralprocess,wemayseeaneectonthistypeofvotes.However,theresultsshownosigni canteectoffemaleparticipationontheshareofprotestvotes.Whilewedonotknowthegenderoftheelectorate(thisinformationiskeptcon dential)weremarkthatthesezeroresultsdonotchangeacrossmunicipalitiesbytheshareofwomeninthepopulation.Itappearstherefore,unlikelythatanincreaseinthepropensityofwomen20 tovoteosetsacorrespondingdecreaseinmalevoters.Notefurtherthat,inanycase,vote-maximizingpartiesshouldnotcareabouttheidentityoftheelectorate,butfocusonthetotalnumberofvotes.Overall,thepatternsonvoterturnovercannotoeraconclusivetesttodiscriminateamongthedierenttheoriesoflowfemaleparticipation.Howeverthedataaregenerallynotsupportiveoftheoriesofvoterdemandfordiscrimination,aswedidnotseeadecreaseinvoterturnoutoranincreaseinprotestvoteswhenandwherethenumberofwomenpoliticiansincreasedby25percentduetothequotas.5.3.2FemalePoliticiansandElectoralResultsTheprevioustestscanbeespeciallypowerfulagainstalternativesunderthenullhypothesisthatvotershadstrongpreferencesagainstfemalecandidates,andthosepreferenceswerequitehomogeneousacrosslists.However,votersofdierentideologiesmayhavedivergingpreferences.Inthatcase,partieswithamorebiasedelectoratecouldoptimallychooseto eldfewerfemalecandidates.Asdiscussedearlier,analternative,ifobservationallyequivalent,theoryhasdierentpartiesbeingdierentlyendowedwithhigh-qualityfemalecandidates.Partiesthatsystematicallyhavemoredi culty ndingquali edwomenpoliticiansshould eldlessfemalecandidates.Boththeoriessuggestthatexogenouslyforcingmorefemalecandidatesintolistsinwhichtheseareunderrepresentedshouldreducetheirelectoralsuccess.Thereforewenextexamineifthelistsinwhichthelawwasmorebindinglostelectoralgroundafterthelaw.TheintuitionforthisapproachcanbereinforcedbylookingbackatTable1.Itisclearthattreatedelectoralliststhatconcurredbothtothe2003and2007electionsandhadarelativelysmallnumberofwomenontheirlistshadtomakeverylargeadjustmentsvis-a-vislistsinthecontrolgroup.Conversely,thegrowthintheshareofwomenwasmuchsmallerornonexistentintheliststhathad elded5ormorewomenin2003withrespecttothecounterfactual.Thissuggestsusingthedistanceofthelisttothe40percentquotainthetreatedmunicipalitiesasalist-speci cinstrumentfortheexogenousgrowthinthelists'femaleshare.Concretely,themodelthatweestimatetakestheform:4VotShaki=14Femki+XMi+XPAk+"ki(5)4VotShakiisthechangeinthevoteshareforpartykinmunicipalityibetween2003and2007,and4Femkithechangeinthenumberofwomenintheparty'slist.PAkstandforparty xedeects:onedummyforeachpartythat eldedlistsinmorethan10municipalitiesandanotherdummyfortheuna liatedlists(only11percentofthetotal).Theparty xedeectscapturechangesinvotersentimenttowardthenationalorregionalparties,21 whichalsotendtobedominantinlocalelections.Notethatitisimportantheretoincludemunicipal xedeects(PMi)becauseoverallchangesinvotingshareshouldaddtozeroacrosslistsinthesamemunicipalityandtheimpactofasymmetricincreaseinthedependentvariablesacrossalllistsshouldwashout.Suchsymmetricchangeswithinmunicipalitieswillbecapturedbythemunicipal xedeects.Therefore,theresultsshouldbeinterpretedasrelationshipsbetweenthevariables'deviationsfromtheirmunicipalmeans,whichisinlinewithwhatthemodelsuggestsistheappropriatecomparisontotestthevarioustheories.Inthisspeci cation,weinstrument4FemShakiwiththevariable:DiELmaxf0;5�Femki;03g:thepre-treatmentdistancetotheminimumquotatarget(5womeninthetreatedsample)interactedbyanEqualityLawtreatmentdummy.Intuitively,morewomenwereforcedintoliststhatusedtohavefewerfemalecandidatesinordertocomplywiththequota(indeedthe rststageyieldsacoe cientofoneandtheinstrumentsareverystrong).InTable6,column1,we rstlimittheestimationtothemunicipalitiesexposedtothelaw.Allthevariancehereiscomingfromtheshocktofemalecandidatesimpliedbytheinitialnumberofwomenineachlist.Theevidenceisnotsupportiveofthehypothesisthatthevotesharedecreasedinthoselistswhereahighernumberwomenthanoptimalwasimposedbythelaw.Infact,surprisingly,wedo ndpositiveeectsofbringingmorewomentotheselists.Theseresultsarecon rmedincolumn2,whereweincludetheobservationsfromthecontrolgroup(unaectedbythelaw).DuetotheunpredictabilityoftheenactmentofthelawaftertheMadridbombingsandthesubsequentunexpectedresultsinthegeneralelectionsof2004,itisveryunlikelythatpartiesintreatedmunicipalitieshadanticipatedtheeectofthelawbystrategicallychoosingthenumberoffemalecandidatesin2003.Yetoneoftheconcernsinthesespeci cationsisthattheinitialnumberofwomencouldbesignalingpartyattributeswhosevaluationswerechangingbetween2003and2007.Whilelistswithlesswomenin2003mayhavebeendierent,notethatbecauseweareusing rstdierencesintherelevantvariablesweareeectivelyremovinganylist-speci ceectsthatdonotchangeintime.Furthermore,wedidnot ndanyevidencethatthenumberofwomenin2003waspredictiveofchangesinvotingsharesinthecontrolgrouporinanalternativesampleofalluntreatedmunicipalitieswithpopulationsbelow2000inhabitants.However,theexistenceofacontrolgroupallowsustobeabletocontrolforthenumberofwomenin2003ontheright-handside,whichwedoincolumn3.19Theresultsarenowevenlargeralbeitmuchmoreimpreciselyestimated:thelaggedfemalecandidatecontrolis,ofcourse,highlyco-linearwiththeinstrument.Clearly, 19Theresultsarerobusttoassumingnon-linearitiesintheeectofpreviousfemalecandidates.Includingasetof xedeectsforeachnumberofwomenin2003didnotchangethecoe cientsincolumn3,andthepre-treatment-women xedeectscannotbedeemedtobestatisticallydierentfromzero.22 thereisnoevidenceofanegativeimpactofwomenonvotingshares.Moreover,wecannotrejectanulldirecteectofthelaggednumberofwomencandidatesonvotingsharegrowth,whichistheassumptionthatwecarryforward.Asecondconcernwiththeseresultsisthattheymaybedrivenbyhighercandidateturnover.Indeed,oneoftheeectsofthequotawastoincreasethenumberofnewcandi-dates.Coulditbethatvotersrelativelydislikewomenbutlikenewcandidatesandthatthelattereectdominatedafterthelaw?Incolumn4wealsocontrolfortheshareofcandidateswhowerenewtothelistin2007,butresultsdonotchange.We nallycontrolforpotentialincumbenteectsincolumn5byintroducingadummyforliststhatwerethemostvotedinthepreviouselections.Whilewedo ndevidenceofwear-downforincumbentlists,thatdoesnotchangetherelevantresults.Theresultsarequitepreciselyestimated.Quantitatively,considerthehypotheticalcaseofapartythatranwith2womencandidatesin2003,whereastheremainingpartieswere elding5each.Assumingthattheotherpartiesdidnotchangethenumberofwomen,thishypotheticallistwouldberequiredto eld3morewomenbylaw,increasingitsvotingshareby2.7percentagepoints.Theseresultsareclearlyinconsistentwiththehypotheticalexistenceofvoterdiscriminationorwithlowerpoliticalabilityoffemalecandidates.Instead,theresultssuggestthepresenceofpartydiscrimination.Toexaminefurtherthenatureofthepositiveimpactofwomen,weseparatethesampleoflistsbythelevelofcompetitionintheirmunicipalities.Electoralresultsatthemunicipallevelarequitepersistent.Thecorrelationbetweenthenumberofseatsforrepeatlistsacrossthetwoelectionswas0.86.Atthetop,75percentofthewinninglistsin2003repeatedaswinnersin2007.Losinglistsin2003,conversely,onlyhadan11percentchanceofwinning.Thepercentageofrepeatwinninglistsincreasestocloseto90percentinthosemunicipalitiesthatwedeemnon-competitive.Inourde nition,non-competitivemunicipalitiesarethosewherethelargestpartyobtained7ormoreseats(outofatotalof13)in2003,anabsolutemajority.Incolumn6(Table6)wepresentseparateresultsfornon-competitivemunicipalitiesandcontrastthemtothoseinmorecompetitiveenvironments(column7).Theresultssuggestthatthepositiveimpactofwomenonelectoralreturnswasconcentratedinnon-competitivemunicipalities.Partiesthereseemedtohavedeviatedmorefromtheoptimalgendermix.InTable7weexaminethisissuefurther.Concretelywenowaskiftheinitialshareofwomenpoliticianshadbeenlowerinnoncompetitivemunicipalitiesrelativetoothers.ThesearepurelydescriptiveOLSregressionswiththeshareofwomencandidatesin2003ontheleft-hand-sideandadummyfornon-competitivemunicipalitiesinthatyearontheright-hand-side,togetherwithpopulationpolynomials.Theresultsincolumn1donotallowus23 toinferanydierences.However,whenwefocusonthemorerelevanttop5positions,thereissomeindicationthatpartiesinnon-competitivetownswere eldinglesswomen(column2).Wethenaddadummyforthewinninglist,andinteractitwiththenon-competitiveelectionindicator.Thedescriptiveevidenceincolumn3isquiteconclusive:entrenchedwinningpartiesinnon-competitiveelectionswere eldingfemalesinelectablepositionsatarate3.3percentagepointsbelowtheaverage(representinga10%decreaseinthenumberofsuchelectablefemales).Inotherwords,mentendedtobemoreprevalentinrelativelysafepositionsforcouncilseats.Thisobservation,togetherwiththefactthatitwaspreciselyinthosesamepositionsthatwomenincreasedvotes,goagainsttheviewthatmalesgenerallyperformbetteraspoliticiansandthatmoremaleswerethusrequiredbypartymachinestowintheseelections.ThisevidenceisveryconsistentwithSanbonmatsu's(2006) ndingthatwomenarelesslikelytorunforandholdstatelegislativeo ceswherepartiesarestronger;andwithBaguesandEsteve-Volart(2010)who ndthatmalesaretypicallyselectedforthesafestseatsintheSenate.Theresultsarethereforemoreconsistentwiththeviewthatlowerfemalepoliticalac-tivismatthelocallevelisnotduetovoterpreferencesorlowability,buttointernaldynamicswithintheparties.Inincumbentpartiesthatoperatedinlesscompetitiveenvironmentsthesegenderpowerdynamicsweremoreprevalent,perhapsbecauseoftheexistenceofentrenchedmalecandidatesandthelesserneedtoappealtobroaderconstituencies.Onthecontrary,incloserelectionscompetitionseemedtohavemitigatedpartyinsiders'abilitytodiscriminate,thereby eldingthemostelectablecandidatesforthelist.6AlternativestoVoters'Tastes6.1DidWomenShyAwayfromCompetitivePositions?Alloftheevidencesofarpointstotheexistenceofinternalpartydynamicstoexplainlowerfemalepoliticalpower,asopposedtovoterpreferencesoralackofwomenwhoarequali edandwilling.Politicalpartiesseemtobehaveasiftheyhavepreferencesfor eldingmalecandidates(d0;d1�0,inthemodel).Oneexplanationforthesepreferencescouldcomefromthebehaviorofwomenincompetitivesituations.Anumberofpapersdemonstratethatwomenfeellesscomfortableandarelessproduc-tiveincompetitiveenvironments(NiederleandVesterlun,2007).Thismayleadfemalesinpoliticalpartiestoavoidcompetitionbyavoidingbeinginthelistorbyavoidingconfrontingothercandidatesforbetterpositionswithinthelists.Observablyequivalenthypotheseshavefemalecandidateswantingtoavoidthestressorresponsibilityassociatedwithmorepow-24 erfulorvisiblepositions,orweightingthenegativeimpactofchild-rearingovertheabilitytosuccessfullyperformasaleader.Didwomensystematicallyshyawayfromin uentialpositions?Awaytotestastrongnullhypothesisonthoselinesistoconsiderwhattheoutcomeshouldbeifwomenfocusedsolelyonavoidingconfrontationwithmalesforhigh-responsibilityormore-visiblepositions.Inthiscase,themoreintrinsically-motivatedmalecandidateswouldbecompletelysuccessfulincapturingthehighestpositionsinthelist.BecausetheEqualityLawhadwomenoccupytwopositionsinthemorein uentialtop vespotsofeachlist,wewouldexpectwomentosystematicallyoccupypositions4and5iftheirobjectivewastominimizepoliticalresponsibility.InTable8,weexaminetheimpactofthelawontheshareofwomenatdierentlevelsinthelists.Theseareregressionsatthelistlevel(conditioningonpartiesthatappearinbothelections),usingboththepre-treatmentandpost-treatmentobservationsofboththetreatedandcontrolmunicipalitiesandlist xedeectsinordertocontrolforpartyandmunicipalunobservables.Theresultscon rmthattheaveragetreatedlistexperiencedanincreaseintheshareofwomenof8percentagepointsinthetop5positions(column1).Interestingly,however,theimpactofthelawwasofonly4.6percentagepointsinthetop3positions(column2)andof0.36percentagepointsinthetopposition(mayoralcandidate).Thissuggeststhat,indeed,womencandidatestendedtoplaceinthelowerpositions:theyweretwicemorelikelytooccupypositions4and5thanpositions1,2,and3.Ontheotherhand,itseemsthataconsiderablenumberofthenewwomeninthelistsusedthelawtogainbargainingpower,obtainingaccesstobetterpositionsthanthoseregulatedbylaw.Incolumns4through6werepeattheexercise,thistimefocusingontheliststhathadwonthe2003election.Intheselists,wherethetop3positionsconsistedofsafeseatsfortheincumbentparty,womenhadahardertimecrackingthroughtheglassceilingofposition4.Infact,thenewwomenneverdislodgedexistingmayoralcandidatesintheseincumbentlists.Ontheotherhand,thenewwomenforcedintoatleastthetop4and5positionsbythequotawerequitelikelytomakeitintothetop3positionsandasnumberoneinthenon-incumbentlists(albeitstillbelowwhatonewouldexpectfromequal-chancerandomization).Clearly,manyofthesewomentooktheopportunityprovidedbythelawandleveragedtheirincreasedbargainingpowerto ghtforthetoppositions.Thisevidenceseemsinconsistentwithastrongnullhypothesisofgeneralizedfemalelackofintrinsicmotivationforpowerorobservationallyequivalenthypotheses.Whilepreviousliteratureincontestablyestablishesthatwomenarelesskeentocompeteonaverageandchildrearingmightbeanissueformanywomen,thisneednotaectmuchequilibriumoutcomesforthefemalesatthevery25 righttailofthecompetitivenessdistributionifi)thenumberofpowerpositionsisverysmallsothattherearemanymorequali ed/ambitiouswomenthanpositions;ii)consumerspreferamorebalancedgendermix;iii)marketsarecompetitivesothatentrenchedmaleinsiderscannotdiscriminateagainstwomen.Indeed,boththefactsthatthesenewwomenatthetopincompetitivelistsneededthelawtoleverageontheirambitions,andthattheircounterpartsinincumbentpartiesdidnotmakeittothetopappearmoreconsistentwiththeexistenceofentrenchedmaleinsiderinterests,andwithperpetuatinggender-powerdynamicswithintheparties'politicalmachines.6.2EndogenousSupplyofQuali edWomen:UsingHigh-StatusNamesasMarkersAnadditionalexplanationfortheevidencesofarhasthesupplyofavailablequali edwomenendogenouslyincreasingafterthequotainthetreatedmunicipalities.Byincreasingthelikelihoodofbeingselected,quotascouldmakeitmoreappealingforhighly-quali edwomentoparticipateinthepoliticalprocess.Alternatively,thequotascouldhavehelpedbreakabadNashequilibrium:priortothequota,quali edwomendidnotshowupbecauseofaperceivedlackofopportunities,andpartieswere eldinglesswomenforlackofquali edcandidates.Inthatcase,votersmayhavereactedpositivelytothehigherqualityofthenewfemalecandidates,ratherthanshowingtheirpreferenceformoregender-balancedlists.Notethatifpartieswereinformedandforward-looking,thesestoriesimply,intheend,discriminatorybehavior:politicalmachineswereoeringlowerpositionstowomen,thereforescreeningouthighly-quali edambitiousfemalesatthecostoffuturelowervotingshares.However,thelawmadeitmorelikelyforwomentobeelectedinallparties.Thus,ageneraltendencyforthelawtoattractmorehighly-quali edwomenmighthaveresultedinawash,andneednothavebeenassociatedwithparty-speci celectoralgains.Furthermore,ahypotheticalendogenousincreaseinthesupplyofhigh-qualitywomenacrosstheboardcannotexplainwhyitispreciselythenon-competitivemunicipalitiesthatdrivetheelectoralresults.Inorderforanendogenouschangeinthedistributionofavailablewomenpoliticianstoexplaintheresults,thiscandidatequalityoughttohaveincreasedsubstantiallymoreinnon-competitiveincumbentpartiesthaninothers.Buteveninthatcase,suchatheorycouldnotexplainwhynon-competitiveincumbentpartiesusedto eldlesswomencandidatespriortothelawinthe rstplace.Moreover,thisalternativeexplanationalsoseemsinconsistentwiththefactthatincumbentpartiesweremorelikelytoplacethenewwomenpreciselyin26 positions4and5,thelowestpositionsrequiredbythelaw.Ifthenewwomeninincumbentpartieswerenowrelativelymorequali edthanthenewwomeninotherparties,wewouldhaveexpectedtheoppositetohappenabsentdiscriminatorypreferences.Yet,perhapsmoreelaboratestoriescouldbeproducedtohypothesizetheexistenceofanendogenous-qualityeect.Wecanaddressthisconcernmoredirectlybysearchingforevidenceofachangeinthequalityofthefemalecandidatepoolafterthequotainthetreatedmunicipalities.Unfortunately,wedonothaveanyinformationaboutthecandidatesthemselvesbesidestheirnameandgender.However,wecanusetheinformationaboutsocioeconomicstatusconveyedbynamestogaugewhetherhigher-statuscandidateswereattractedbythequotas.Arecentliteraturehaspointedtotheinformationalcontentofcharacteristicallyethnicnames(FryerandLevitt,2004)orthesurnamechoiceofmarriedwomenintheUS(GoldinansShim,2004).Nameassignationentailsaconsciouschoicebyparents,andisthereforein uencedbytheirculturalmilieu,revealingcluesabouttheirsocioeconomicstatus(SES).Tooperationalizethisidea,westartbycalculatingthepowerstatusofallfemale rstnamesthatappearedinthe2003municipallists.Wehave70,556femalecandidatesand1,388distinctive rstnames.Female rstnamesinthissamplefollowaPareto-likedistributionwithafewnamesaccountingforalargefractionofthepopulation.Asurprising38percentofthecandidateshaveMariaoroneofitscomposites(e.g.MariaRosa)astheir rstname.Thetop20namesaccountfor60percentoffemalecandidates.Inordertoestimateaccuratestatusmeasureswefocuson212nameswithmorethan25observations(encompassing93percentofthefemalesample).Forsucheachnameandthealternativecategoryothernamewecalculatetheaverageshareofwomenthatwereelectedinthe2003elections.20Wecreateapowerindexforeachnamebydividingtheshareofelectedwomenbearingthatnamebytheaverageshareofelectedwomenin1993(approximatelyonein vewomenresultedelected).InTable9,panelA,columns1and2wedisplaythenameswiththehighestandlowestrelativeeligibility.Low-statusnamesinparticularseemtofollowtwopatterns:theyconformtoourpriorsaboutnamingconventionsinlowSESpopulations,orcorrespondtonamesthatbecamepopularonlyrecently(e.g.LaiaorMireiainthemid80sand90s)andthereforesingleoutveryinexperiencedcandidates.Interestingly,thepervasiveMariahasnoinformationalcontent,withanelectabilityratioclosetoone.Smallernamesintheothercategoryare,onaverage,alsonon-informative.Notethatthereismoreinformationonthelefttailofthestatusdistribution:manyofourtoppowerfulnamesarealreadyclosetotheaverageintermsofelectability,butlow-statusnamesaregenuinelylesspowerful.We 20Notethatthismeasurecombinesthelikelihoodofjoiningawinninglistwiththelikelihoodofbeingclosertothetopofthelist.27 willthereforefeelmorecomfortableinferringlowpowerstatusin2007tocandidateswithnamesforwhichwecalculatearelativelylowelectabilityin2003.WecanalsoobtaininformationabouttheSESofcandidatesfromtheirsurnames.While rstnamesarisefromaconsciouschoicebyparents,surnamesarepassedonforgenerations.Therefore,theirinformationalcontentisin uencedbyinheritedattributes,suchasgenes,income,education,...(Guell,etal.,2007).Fortunatelyforourstudy,thetwopapersweareawareofstudyingtheinformationalcontentofsurnamesbothfocusonthecaseofSpain.Collado,OrtÃn,andRomeu(2008)useyellowpagesandothersourcesand ndthatindividualswithlow-frequencysurnamesinSpaintendtobemorelikelytobedoctors,lawyers,anduniversityprofessors.Individualswithraresurnameshadalsobeenmorelikelytobeliterateinthe1890historicalCensus.Withrespecttopoliticalpower,theseauthorsalso ndlow-frequencysurnamestobeover-representedamongstpoliticalcandidatesinthe2004nationalcongressandSenateelections,butthiseectissolelydrivenbytheconservativeandnationalistparties.21SimilarlyGuelletal.(2007)usecon dentialCensusdataand ndthatthemorefrequentasurnameis,thelesseducationshouldyouexpectherholdertohaveachieved.PeopleinSpainusetwosurnames: rstthefather's rstsurname,andthenthemother's rstsurname.Atraditionalwaytoachievehigh-statussurnamedierentiationinSpainhasbeentohyphenatetwosurnames(e.g.Garcia-Parra)andthenuseanotheronefromtheparentsasasecondsurname(seeSalazar-Acha,1991).Similarstrategieshavehistoricallyimpliedusingprepositionstocompoundasurnamewithanotheroneorageographiclocation(e.g.SanchezdelaEnsenada).Suchdouble-barreledsurnamestendtobecarriedbyindividualsofmeasurablyhighsocioeconomicstatus(Colladoetal.,2008).TheinsightsaboveallowustoinferhigherlikelihoodofabettereducationandSESinthefollowingways:i)wecreateadummythattakesvalueoneifthecandidateisfemaleandher rstnamepower-status(relativeelectabilityin2003)isbelow0.9,correspondingtothelowestquartileinthenamedistribution22;ii)wecreateadummythatequalsoneifthecandidateisfemaleandhasarare rstsurname;raresurnamesaresuchthatthenumberoffemalecandidatesbearingthatsurnamewaslessthanorequalto5(outof70,556)in2003(thiscorrespondstothelowestquartileofthefemaledistributionbysurnamefrequency);iii)wecreateadummyequalto1forwomenwitha rstsurnameinthelistofthetop17surnamesencompassing25percentofthetotalcandidatepopulationin2003.23Those 21Notethatmostrecentimmigrantsthatarrivedafter2002werelargelyineligibletovoteorjoinalistinthe2003,2004,and2007elections.Priortothatperiod,immigrationlevelsinthecountrywereverylow.22Resultsarenotsensitivetochangingthisthresholdoronfocusingontheright-tailofthestatusdistri-bution.23Forbothmeasuresii)andiii)we ndnoadditionalvalueintermsofpredictingelectabilityinadding28 extremelycommonsurnames(e.g.Gonzalez,Garcia)canbeseeninTable9,panelA,column3;iv)wecreateadummythattakesvalueoneifthecandidateisafemaleandhersurnameislongerthantheconventional3words.Tocalculatethenumberofwordsweseparatehyphenatedsurnamesintotwowordsandcountaswordstheusualconnectingprepositions(e.g.de,la).Remarkably,we ndthesurnamestrategiestobesuccessfulinreplicatingthepatternssuggestedbythepreviousliterature.Focusingonthelikelihoodofeachfemalecandidateresultingelectedin2003(about20percentonaverage),we ndthatwomenwithlow-frequencysurnameswere1.5percentagepointsmorelikelytobeelectedthanthemedian,andthatwomenwiththeverycommonsurnameswerelesselectableinthesameproportion.Individualswithnameslongerthantheconventional3wordswere2percentagepointsmorelikelytobeelected,aremarkable10percentincreaseinsuchprobability.Theseare,inourview,largeeects.Reassuringly,surnamerarity,notbearingoneofthecommonsurnames,andnamelengthallincreasesigni cantlythelikelihoodofbearingahigh-status rstname,whichfurthervalidatestheinformationalcontentofthesealternativemeasures.24InTable9,panelB,welookatwhethertheEquality-Lawquotaincreasedtheeducationalandsocialstatusofthewomencandidatesintreatedmunicipalitiesascapturedbythenamemeasures.Todoso,wepresentmodelsofthe2003and2007electionswhereeachcandidaterepresentsanobservation,andincludemunicipal xedeectstocontrolforlocalunobservables.Ontheleft-handside,thefemalestatusmeasurestakevaluezeroforallmales,andoneorzeroforfemalesdependingonwhethertheybearahighorlow-statusname.Themaincontrolsontheright-hand-sideare:abaselinedummyforfemalecandidates,adummyforfemalecandidatesinmunicipalitiesaectedbythelaw,adummyforfemalecandidatesin2007,andadummyforfemalecandidatesinmunicipalitiesaectedbythelawin2007.Thecoe cientofinterestisthelatterone.Notethatthefully-interactedmodelallowsustointerpretthecoe cientofthefemaledummyasthebaselineprobabilityofwomentobeinaloworhigh-statusgroup,andtobenchmarktheothercoe cientsaccordingly.Inordertogivetheendogenousqualityhypothesisabetterchancewepresentseparateregressionsforthetop5positionsandtherest.Presumably,high-qualitywomenspeci callyattractedbythepowerquotaswouldwanttogoatthetopofthelists.25TheresultsinPanelBofTable9donotdisplayanyevidenceofanincreaseinthesocioeconomicstatusoffemalesinthequotamunicipalitiesafterthelaw.26Notethatwe theinformationalcontentofthesecondsurname.24Theinterestedreadercanconsultappendixtable3fortheseresults.25Whenwepoolbothsamples,theresultsaresimilarbothquantitativelyandintermsofstatisticalsigni cance.26Wealsorepeatedthesameanalysisforthenon-competitivemunicipalities,wheretheeectofthequotas29 havealargeamountofobservationsandare,therefore,likelytopickuprelativelysmallbutsystematiceects(infact,theregressionsdocapturesmalltemporaltrendsinsomeofthevariables).Ifanything,theresultstendtoindicatethatlower-statusfemalesweremorelikelytobeincludedinthetop5positionsafterthequotas(lowerstatusnames,lessrareandmorecommonsurnames,andshorternames),buttheindividualcoe cientsarenotstatisticallysigni cant.Weconcludethatitisunlikelythatthequotasresultedinmarkedincreasesinthequalityofthefemalecandidatesinthe eld.7ConclusionRecentresearchhasemphasizedtheexistenceofdierencesintastesandpsychesbetweengenders(CrossonandGneezy,2008).LawlessandFox(2005)conductedasurveyofpotentialwomencandidatesintheUSandreportthatwomentendtoseethemselvesaslessaptpoliticalleadersonaverage.Wearguethat,whiletheseissuesarecertainlyimportanttoaccountfortheoutcomesexperiencedbytheaveragefemale,theydonotnecessarilyexplainwhywomenareunderrepresentedinpositionsofpower.Insofarthereisdemandforitinelectoralmarkets,andpartymachinestrytomaximizevotesinacompetitiveenvironmentelectorallistsshouldincludealargefractionofwomencandidates,easilydrawnfromtheveryrighttailofthefemaleabilityandambitionjointdistribution.Itisnotpossibletounderstandfemaleunder-representationintherelatively-fewtoppositionsindependentlyfromthemarketdemandfortalentedfemalepoliticiansandthebehaviorofpartymachinesonthesupply-side(Becker,1957).Toframethisdebate,wedevelopedamodelthatnestsmostoftheexplanationsofwhywomenareunderrepresentedinpowerfulpositions.Somefactorsthataccountforthisproblemmaybemediatedbymarkets.Onthedemandside,customersinthepoliticalprocess(voters)mayprefermalecandidates.Onthesupplyside,theremaybelesswillingquali edfemalecandidates.Alternatively,thefactorsthatdeterminethelowavailabilityoffemalecandidatesmaystemfrominternalprocesseswithinthepoliticalparties.Wethentookthepredictionsofthedierenthypothesestothedata.WeusedSpain'sEqualityLawasanaturalexperimentprovidinganexogenousincreaseinfemalepoliticalparticipation.ThetimingofthepassageoftheInequalityLawin2007,establishinga40percentquotaofwomeninallelectorallists,wasexogenousandquiteunpredictable.TheEqualityLawwaspassedinMarch2007,justtwomonthsbeforethemunicipalelections.AnimportantfeatureoftheEqualityLaw,wasthatitonlyappliedin2007tomunici- ismostpronounced.Wedonot ndanyevidenceofanincreaseintheSESoffemalecandidatesinthosemunicipalitieseither.30 palitieswithpopulationsabove5000inhabitants.Wethereforeimplementedadiscontinuitydesignthatcomparedtreatedandcontrolmunicipalitiesaroundthatthresholdbeforeandafterthelaw'spassage.Empirically,wefoundthatthelawincreasedthefemalecandidatesharebyeightper-centagepoints,representinganincreaseof25percentinthenumberofwomencandidates.Suchlargeshocktothenumberoflocalfemalepoliticiansdidnothaveanyadverseimpactonvoterturnoutorprotestballotsinthetreatedmunicipalities.Wealsofoundthatpartiesdidnothavemajorproblems ndingsuitablefemalecandidatesto llinthequota.Thefemalequotawasnotassociatedwithincreasedlistattritionordi cultiesfornewliststoarise.Partiesdidnotneedtoretainpastwomencandidatesmoreoftenortopromotefemale llercandidatestotoppositions.Thisevidenceisnotconsistentwiththeexistenceofmajorsupplyconstraintsoffemalecandidates.Inaddition,partiesthatwerefurtherfromthe40percentfemalethresholdandwerethereforeforcedto eldinmorefemalecandidatesexperiencedincreasesintheirvoteshares.Theevidenceisnotconsistenteitherwiththeexistenceofgeneralizedstrongpreferencesformalecandidatesamongstvoters.Wefurtherfoundthatthepositiveeectofwomenonvoteswasconcentratedinmunic-ipalitieswithnon-competitiveelections,whereonepartyhadenjoyedanabsolutemajorityinthe2003elections.Inthesemunicipalities,theincumbentlistshad eldedanumberofwomenthatwassigni cantlybelowtheaverageincompetitiveelections.Moremenhadbeen eldedinelectionsweretheincumbentpartyhadalargenumberofsafeseats,butthesepartiessubsequentlyfaredbetterwithmorewomeninthenextelections.Allofthesefactsareatoddswiththeclaimthatmoremenarerequiredtowinelections.WerepartiesandfemalepotentialcandidateslockedinabadNashequilibrium,withgoodwomancandidatesnotshowingupforlackofprospectsandpartiesoeringlowpositionstofemalesduetolackofquali cations?Weshowedthatfurtherresultsarenotconsistentwiththishypothesis.Thequotas-forcingpartiestooersomeofthetoppositionstowomen-wereunlikelytohaveattractedwomenofhighereducationalorsocioeconomicstatus.NamesandsurnamesconveySESandeducationalstatusinSpain;howevertherewasnodierentialgrowthinthestatusofnamesorsurnamesofthefemalecandidatesinthetreatedmunicipalities.Theexplanationforlowerfemalerepresentationseemstorevolvearoundtheinternaldynamicsofcandidatechoiceswithintheparties:partiesbehaveasiftheydisplayeddis-criminatorytastes.Arewomenthemselvesoptingforlesspowerfulpositions?Whilewomenmayshyfromcompetitiononaverage,alargenumberofwomenincompetitivelistslever-agedtheEqualityLawtoattainhigherpositionsthantheminimumonesguaranteedbythe31 quota.However,femalecandidateshadmoredi cultyattainingtoppositionsinrelativelysafetopspotsinincumbentlists.Theevidenceseemsoverallmoreconsistentwiththeexistenceofentrenchedmale-dominatedpoliticalmachines,especiallyinlesscompetitiveenvironments.Theresultsdemonstratethatforcingpartiestoacceptmorewomenthroughquotasonthecandidatelistscanincreasefemaleparticipationwithoutnecessarilydecreasingtheelectabil-ityofthecandidates.Butatthesametime,competitionmayalsolimitthescopefordiscriminationfrompartymachines.Hence,alternativepolicyresponsescouldimplymorerelianceonpartyprimariesatalllevels,openlists,votingforcandidatesratherthanlists,andmorecompetitiveelectoralenvironmentswithlesssafeseatsatthedispositionofpartymachines.Wehypothesizethatsuchmeasurescouldalsoinducemorefemalestoparticipateinthepoliticalprocess,withouthavingtoimposestrictquotas,atestablehypothesisforfutureresearchonthistopic.Infactwomenrepresentationismuchlargerinproportionalelectoralsystems(Norris,2006,NorrisandKrook,2011)whereeachvotecounts,asopposedtomajoritariansystemswheremanyseatsarede-factoownedandnon-competitivelyallo-catedbythemachinesofthelocallydominantparty,especiallyafterredistricting.Eectingthebehaviorofpoliticalmachinesvis-a-visgenderissuesandunderstandinghowcompe-titionchangesthesedynamicscouldthusbethekeystoimprovefemalechancesatequalparticipation.References[1]Arrow,KennethJ.1973.Thetheoryofdiscrimination,O.AshenfelterandA.Rees(eds.),DiscriminationinLaborMarkets,Princeton,NJ:PrincetonUniversityPress.[2]Ayres,Ian,andPeterSiegelman.1995.Raceandgenderdiscriminationinbargainingforanewcar,TheAmericanEconomicReview,85(3):304-321.[3]Bagues,ManuelF.,andBertaEsteve-Volart.2010.Cangenderparitybreaktheglassceiling?Evidencefromarepeatedrandomizedexperiment,TheReviewofEconomicStudies,77(4):1301-1328.[4]Bagues,ManuelF.,andBertaEsteve-Volart.2009.Arewomenpawnsinthepoliticalgame?EvidencefromelectionstotheSpanishSenate,mimeo.[5]Beaman,Lori,RaghabendraChattopadhyay,EstherDu o,RohiniPande,andPetiaTopalova.2009.Powerfulwomen:Doesexposurereducebias?TheQuarterlyJournalofEconomics,124(4):1497-1540.32 [6]Becker,Gary.1957.Theeconomicsofdiscrimination.Chicago:UniversityofChicagoPress.[7]Bertrand,Marianne,ClaudiaGoldin,andLawrenceF.Katz.2010.Dynamicsofthegendergapforyoungprofessionalsinthe nancialandcorporatesectors,AmericanEconomicJournal:AppliedEconomics,2:228-255.[8]Campa,Pamela(2010).LegislativeGenderQuotasandPublicChildcareProvision:Quasi-ExperimentalEvidence.Mimeo:StockholmUniversity.[9]Cavalcanti,TiagoV.andJoseTavares(2011),WomenPreferLargerGovernments:Growth,StructuralTransformation,andGovernmentsize.EconomicInquiry,49:155171.[10]Chattopadhyay,Raghabendra,andEstherDu o.2004.Womenaspolicymakers:Evi-dencefromarandomizedpolicyexperimentinIndia,Econometrica,72(5):1409-1443.[11]Clots-Figueras,Irma.2009.`AreFemaleLeadersGoodforeducation?EvidencefromIndia,'Mimeo,UniversidadCarlosIII.[12]Collado,M.Dolores,OrtÃn,IgnacioOrtuñoandRomeu,Andres,(2008),SurnamesandsocialstatusinSpain.InvestigacionesEconomicas,32-3,pp.259-287.[13]Crosson,Rachel,andUriGneezy,2008.GenderDierencesinPreferences.JournalofEconomicLiterature,47:2,127[14]DePaola,Maria,VincenzoScoppa,andRosettaLombardo.2010.Cangenderquotasbreakdownnegativestereotypes?Evidencefromchangesinelectoralrules,JournalofPublicEconomics,94:344-353.[15]Dolan,KathleenA.2004.VotingforWomen.HowthePublicEvaluatesWomenCan-didates.WestviewPress:Boulder,Colorado.[16]Dolan,KathleenA.2005.WomenCandidatesinAmericanPolitics:WhatWeKnow,WhatWeWanttoKnow.Presentedatthe2006meetingoftheMidwestPoliticalScienceAssociation,April20-23,2005.[17]Du o,EstherandPetiaTopalova(2004)/UnappreciatedService:Performance,per-ceptions,andWomenLeadersinIndia.MITWorkingPaper.[18]Einav,L.andL.Yariv(2006)"What'sinaSurname?TheEectsofSurnameInitialsonAcademicSuccess",JournalofEconomicPerspectives,20,175-188.33 [19]Ferreira,FernandoandJosephGyourko(2010).DoesGenderMatterforPoliticalLead-ership?TheCaseofU.S.MayorsMimeo:universityofPennsylvania.[20]Fryer,R.G.,Jr.andS.D.Levitt(2004)"TheCausesandConsequencesofDistinctivelyBlackNames",QuarterlyJournalofEconomics,119(3),767-805.[21]Funk,PatriciaandChristinaGathman,2010GenderGapsinPolicyMaking:EvidencefromDirectDemocracyinSwitzerland,Mimeo:UniversitatPompeuFabra.[22]Gallarduci,StefanoandDanielePasserman,2010.GenderInteractionswithinHierar-chies:EvidencefromthePoliticalArena.WorkingPaper:BostonUniversity.[23]Gneezy,Uri,MurielNiederle,andAldoRustichini.2003.Performanceincompetitiveenvironments:genderdierences,TheQuarterlyJournalofEconomics,118(3):1049-1074.[24]Gneezy,Uri,andAldoRustichini.2004.Genderandcompetitionatayoungage,TheAmericanEconomicReview:PapersandProceedings,94(2):377-381.[25]Goldin,Claudia,andLawrenceF.Katz.2008.Transitions:Careerandfamilylifecyclesoftheeducationalelite,TheAmericanEconomicReview:PapersandProceedings,98(2):363-369.[26]Goldin,Claudia,andCeciliaRouse.2000.Orchestratingimpartiality:Theimpactofblindauditionsonfemalemusicians,TheAmericanEconomicReview,90(4):715-741.[27]Goldin,C.andM.Shim(2004),"Makinganame:Women'sSurnamesatMarriageandBeyond",JournalofEconomicPerspectives,18(2),143:160.[28]Güell,MaiaandJoséV.RodriguezMora&ChrisTelmer,2007."IntergenerationalMobilityandtheInformativeContentofSurnames,"EconomicsWorkingPapers1042,DepartmentofEconomicsandBusiness,UniversitatPompeuFabra.[29]Montalvo,JoseG.(2009).Votingafterbombings:anaturalexperimentontheeectofterroristattacksondemocraticelectionsforthcomingintheReviewofEconomicsandStatistics[30]Kahn,LawrenceM.2000.Thesportsbusinessasalabormarketlaboratory,TheJournalofEconomicPerspectives,14(3):75-94.34 [31]Knowles,John,NicolaPersico,andPetraTodd.2001.Racialbiasinmotorvehiclesearches:Theoryandevidence,TheJournalofPoliticalEconomy,109(1):203-229.[32]Niederle,Muriel,andLiseVesterlund.2007.DoWomenShyAwayfromCompetition?DoMenCompeteTooMuch?QuarterlyJournalofEconomics,122(3):10671101.[33]Norris,Pippa(2006).TheImpactofElectoralReformonWomen'sRepresentation.ActaPolitica,Volume41,Number2.[34]Norris,PippaandMonaLenaKrook(2011).GenderEqualityinElectedO ce:ASixStepActionPlan.OSCEDraftReport.[35]Lawless,JenniferandRichardFox.2005.ItTakesACandidate:WhyWomenDon'tRunforPoliticalO ce.NewYork:CambridgeUniversityPress[36]Price,Joseph,andJustinWolfers(2010)RacialdiscriminationamongNBAreferees,TheQuarterlyJournalofEconomics,125(4):1859-1887.[37]Rehavi,M.Marit(2007)SexandPolitics:DoFemaleLegislatorsAectStateSpend-ing?Mimeo,UniversityofCalifornia,Berkeley[38]Reuben,Ernesto,PedroRey-Biel,PaolaSapienza,andLuigiZingales(2010)TheEmer-genceofMaleLeadershipinCompetitiveEnvironments.IZADiscusionPaperNo.5300.[39]Sanbonmatsu,Kira(2002)PoliticalPartiesandtheRecruitmentofWomentoStateLegislatures.JournalofPolitics64:791-809.[40]Sanbonmatsu,Kira(2006)WhereWomenRun:GenderandPartyintheAmericanStates.TheUniversityofMichiganPress:AnnArbor.[41]Saiz,AlbertandSusanWachter(2011)ImmigrationandtheNeighborhood.AmericanEconomicJournals:Policy(May,2011).[42]Salazar-Acha,J.(1991)GénesisyevoluciónhistóricadelapellidoenEspaña,RealAcademiaMatritensedeHeráldicayGenealogÃa,Madrid.[43]Szymanski,Stefan.(2000).AmarkettestfordiscriminationintheEnglishprofessionalsoccerleagues,TheJournalofPoliticalEconomy,108(3):590-603.[44]Swers,MicheleL.(1998).AreWomenMoreLikelytoVoteforWomenÃsIssueBillsThanTheirMaleColleagues?LegislativeStudiesQuarterly,23,435-448.35 [45]Verge,Tania(2006)MujeryPartidospoliticosenEspana:lasestrategiasdelospartidosysuimpactoinstitucional,1978-2004,RevistaEspanoladeInvestigacionesSociologicas,115:165-196.[46]Welch,Susan(1985)AreWomenMoreLiberalthanMenintheU.S.Congress?Leg-islativeStudiesQuarterly,10,125-13436 AppendixA:ProofsThroughoutthissection,wedenotethevariablesthatarechosenunderthequotawiththesuperscriptq.Weusetodenotethechangeinavariableaftertheintroductionofthequota.Proofofproposition1.Becausefpm=fpw,wecanwriteffw=fmandapapw=apm.LetFbethecorrespondingdistributionfunction.Eachpartychoosesthebestcandidatesforeachgender.Asaresult,thereisaone-to-onerelationshipbetweena pwandWp.Indeed,wecanwriteWp=1a pwfpw(a)da=1a pwf(a�ap)da=1a pw�apf(a)da=1�F�a pw�ap:Becauselistsareofa xedsize,thereisalsoaonetoonerelationshipbetweena pwanda pm,de nedbytheequation:1�F�a pw�ap�F�a pm�ap=0.Asaresult,ifW0W1,wemusthavea 0w�a0�a 1w�a1anda 0m�a0a 1m�a1(whenturnoutis100%,italsofollowsthatd0�d1).Moreover,underthequota,a qpwischosentosatisfyWqp=W becausethequotaisbindingforbothparties.Asbothpartieshavethesameshareofwomenafterthequotaisimposed,wehavethata q0w�a0=a q1w�a1anda q0m�a0=a q1m�a1.Finally,noticethat,becauseW11 21�W0,wehavethata 1w�a1�a 0m�a0.Weshownextthatforallvotersx2[0;1],u(x;0)increasesmorethanu(x;1)aftertheintroductionofthefemalequota,i.e.u(x;0)u(x;1).Toseethis,noticethatu(x;p)=Ap�dvWp=Ap,whereAp=a pwa qpwafpw(a)da�a qpma pmafpm(a)da:Noticethatbecausea pwa qpwa qpma pm,Ap0andthereforesoisu(x;p).Furthermore,wecanwriteA0=a 0wa q0waf(a�a0)da�a q0ma 0maf(a�a0)da=a 0w�a0a q0w�a0af(a)da�a q0m�a0a 0m�a0af(a)da=A1+a 0w�a0a 1w�a1af(a)da�a 1m�a1a 0m�a0af(a)da�A1+(a 0w�a0)[F(a 0w�a0)�F(a 1w�a1)]�(a 0m�a0)[F(a 1m�a1)�F(a 0m�a0)]A1;wherethelastinequalityfollowsfromthefactthatF(a 0w�a0)�F(a 1w�a1)=F(a 1m�a1)�37 F(a 0m�a0)0.Sincetheutilityfromvotingeitherpartyincreases,turnoutcannotdecrease.Indeed,ifturnoutislessthan100%withoutthequota,itmustincreaseafterthequotaisintroduced.WealsoneedtoseethatV0�V1.Ifthereisfullparticipationwithoutthequota,thenV0=1�V1.Themarginalvoter,locatedatx=V0isindierentbetweenthetwopartieswhennoquotaispresent.However,becauseu(x;0)�u(x;1)0,thisvotermuststrictlyprefertovoteforparty0whenthequotaisintroduced.Asaresult,thequotamovesthemarginalvoterclosertoparty1,sothatV0�0�V1.WhenthereisapartialturnoutintheabsenceofanyquotawehaveV0+V11.Supposethatforeachvoterweincreasetheutilityu(x;p)ofvotingeachpartypby.Then,aslongas x0+ x1+4 t1,thenumberofvotesofeachpartyincreasesbyexactly2=tbecausetransportationcostsarelinearandvotersareuniformlydistributedontheline.27Ifquotasdonotleadtofullparticipation,thenVp=2Ap=t.ItfollowsimmediatelythatV0�V1.Instead,ifquotasresultinfullparticipation,wecanincreasetheutilityfromvotingeachpartyintwosteps.First,increaseu(x;p)by1p=minA1;t 4(1�V0�V1) .Inthesecondstepweincreasetheutilityofvotingforpartypby2p=Ap�minA1;t 4(1�V0�V1) .Thisway,wehavethat1p+2p=Ap,whichisthefullincreasecausedbythefemalequota.Theincreaseinthe rststepisthesameforbothparties.Moreover,becauseV0+V1+41p t1,V0andV1increasebothbythesameamount.Theincreaseinutilityinthesecondstepislargerforparty0thanforparty1.Moreover,eitherthereisfullparticipationafterthe rststep(ifA1t 4(1�V0�V1))ortheutilityofvotingforparty1doesnotincreaseinthesecondstep(ifA1t 4(1�V0�V1)).Ineithercase,V0increasesbymorethanV1inthissecondstep.Therefore,addingtheeectsofthetwosteps,wegetthatV0�V1. Proofofproposition2.Noticethat1�F�a pw�ap�F�a pm�ap=1�F�a pw�ap�F�a pw�dv( xp)�ap=0,sincepartiesseta pw�a pm=dv( xp).Therefore,itiseasytoseethata pw�apisincreasingindv( xp).Asaresult,thenumberofwomenWpisadecreasingfunctionofdv( xp).Ifturnoutis100%,themarginalvoteristhesameforbothparties.Asaresult,bothpartiessetthesamedierenceforthestandardsofmenandwomen,asa pw�a pm=dv( xp)=dv( x).Becausethedistributionofabilityformenandwomenisthesame,thismeansthatbothpartiesmustbechoosingthesamenumberofwomen.Therefore,ifW0W1itmust 27Noticethatthevoterlocatedin x0isindierentbetweenvotingparty0andabstainingbeforetheincreaseinutility.Iftheutilityofeveryvoterincreasesby,thenthevoterthatisindierentbetweenvotingparty0andabstainingisnowlocactedin x0+2=t,asu( x0+2=t;0)+=u( x0;0)=u .Similarly,thevoterlocatedin x1�2=tisindierentbetweenvotingparty1andabstainingaftertheincreaseinutility.Moreover,becauseV0+V1+4 t1,thesevoters(weakly)preferstovoteforparties0and1,respectively,thanvotingfortheotherparty.Therefore,suchachangeinutilityincreasesbothpartiesvotesby2=t.38 bebecausesomevotersabstain,andthemarginalvotersofthetwopartiesdier.Moreover,itmustalsobethecasethatdv( x0)�dv( x1).Becausedvismonotonic,itmustthereforebedecreasing.Weshownextthat,forallvotersx2[0;1],u(x;0)decreasesmorethanu(x;1)aftertheintroductionofthefemalequota,i.e.u(x;0)u(x;1).Toseethis,noticethatu(x;p)=Ap�dv(x)Wp.Thechangeinutilityu(x;p)isdecreasingindv(x).Asdv(x)isdecreasinginx,itfollowsthatu(x;0)u( x0;0)forallx x0,andu(x;1)�u( x1;1)forallx� x1.De nethefunction (dv)=a wa qwafw(a)da�a qma mafm(a)da�dva qwa wf(a)da,where1�F(a w)�F(a m)=1�F(a qw)�F(a qm)=0,1�F(a qw)=W ,anda w�a m=dv.Itiseasytoseethat (dv( xp))=u( xp;p).Thismeansthat (dv( xp))measuresthechangeinutilityofpartyp'smarginalvoteraftertheintroductionofquotas.Ifwedierentiatewithrespecttodvweobtain:@ @dv=a wf(a w)@a w @dv+a mf(a m)@a m @dv�dvf(a w)@a w @dv�a qwa wf(a)da=�a qwa wf(a)da:Theequalityinthesecondlinefollowsfromthefactthata w�a m=dv,@a w @dv=1+@a m @dv,and[f(a w)+f(a m)]@a w @dv=f(a m).28Combiningalltheseelements,weobtaina wf(a w)@a w @dv+a mf(a m)@a m @dv=dvf(a w)@a w @dv.Finally,noticethat@ @dv0whenevera wa qw(asisthecaseforanydv2[dv( x1);dv( x0)]).Since isadecreasingfunctionofdv,anddv( x0)�dv( x1),itimmediatelyfollowsthatu( x0;0)u( x1;1)0.Moreover,becauseu(x;0)u( x0;0)forallx x0,andu(x;1)�u( x1;1)forallx� x1,wemusthaveV0V10. Proofofproposition3.NoticethatW0W11 2impliesthatF0w(a 0w)�F1w(a 1w)andF0m(a 0w)F1m(a 1w).BytheassumptionthatFpg(a)=Fg(a�apg),itfollowsthata 0w�a0w�a 1w�a1wanda 0m�a0ma 1m�a1m.Moreover,becausedv=d0=d1=0,partiesdonotdiscriminateagainstwomenintheabsenceofquotas,andtheybothseta pw=a pm.Italsofollowsthatu(x;p)=Ap.Themarginalmanandwomanofagivenpartyhavethesameability.Theintroductionofthequotathereforeforcespartiestoseta qpwa qpm.Moreover,sincebothpartiesendupwiththesamenumberofwomenunderthequotas,itmustbethecasethata q0g�a0g=a q1g�a1g 28Thelastequalityfollowsfromdierentiatingtheequation1�F(a w)�F(a m)=0withrespecttodv,andsubstituting@a m @dv=@a w @dv�1.39 forbothg2fm;wg.Asbefore,wehavethat:Ap=a pwa qpwafpw(a)da�a qpma pmafpm(a)da:Noticethatbecausea qpwa pw=a pma qpm,Ap0andthereforesoisu(x;p).Furthermore,wecanwriteA0=a 0wa q0wafw(a�a0w)da�a q0ma 0mafm(a�a0m)da=a 0w�a0wa q0w�a0wafw(a)da�a q0m�a0ma 0m�a0mafm(a)da=A1+a 0w�a0wa 1w�a1wafw(a)da�a 1m�a1ma 0m�a0mafm(a)da=A1+a 0wa 1w�a1w+a0wafw(a�a0w)da�a 1m�a1m+a0ma 0mafm(a�a0m)daA1+a 0w[Fw(a 0w�a0w)�Fw(a 1w�a1w)]�a 0m[Fm(a 1m�a1m)�Fm(a 0m�a0m)]=A1;wheretheinequalityfollowsfromthefactthatFw(a 0w�a0w)�Fw(a 1w�a1w)=Fm(a 1m�a1m)�Fm(a 0m�a0m).Sincetheutilityfromvotingeitherpartydecreases,turnoutcannotincrease.Indeed,ifturnoutislessthan100%withoutquotas,itmustdecreaseafterthequotasareintroduced.DenotebyVqpandVpthenumberofvotersvotingforpartypwithandwithoutquotas,respectively.WealsoneedtoseethatV0V1.Ifthereisfullparticipationwithquotas,thenthereisalsofullparticipationwithoutquotas.Themarginalvoter,locatedatx= x0isindierentbetweenthetwopartieswhennoquotaispresent.However,becauseu(x;0)u(x;1),thisvotermuststrictlyprefertovoteforparty1whenquotasareintroduced.Asaresult,V00V1.WhenthereisapartialturnoutunderquotaswehaveVq0+Vq11.Iftheabsenceofquotasdoesnotleadtofullparticipationeither,thenVp=2Ap=t,becausetransportationcostsarelinearandvotersareuniformlydistributedontheline.ItisimmediatethatV0V1.Instead,iftheabsenceofquotasresultinfullparticipation,wecandecreasetheutilityfromvotingeachpartyintwosteps.First,decreasetheutilityofvotingpartypforvoterxbyadding1p=maxfA0;u �u( x0;0)gtou(x;p).(Noticethatthemarginalvoter x0isindierentbetweenvotingeitherofthetwoparties,andtherefore,u �u( x0;0)=u �u( x0;1).)Inthesecondstepwedecreasetheutilityu(x;p)byadding2p=Ap�maxfA0;u �u( x0;0)g.NoticethatafterthetwostepstheutilityofvotingforpartypdecreasebyAp(thefulldecreasecausedbythefemalequota).Thedecreaseinthe rst40 stepiscommontobothparties.Moreover,itstillleadstofullparticipation,asallvotersstillprefertovoteforoneofthetwopartiesratherthanabstaining.Asaresult,sincetheutilityofthemarginalvoterfromvotingeachofthepartiesdropsbythesameamount,thisvoterisstillindierentbetweenthem.Hence,V0andV1remainunchanged.Thedropinutilityinthesecondstepislargerforparty0thanforparty1.Moreover,eithertheutilityofvotingforparty1doesnotdecreaseinthesecondstep(ifA1u �u( x0;0)),orthemarginalvoter x0isindierentbetweenvotingeitherpartyorabstaining(ifA1u �u( x0;0)).Ineithercase,V0decreasesbymorethanV1inthissecondstep(noticethatinthesecondcaseVpdecreasesby22p=t).Therefore,addingtheeectsofthetwosteps,wegetthatV0V1. 41 Number of of Candidatures with N Women in 2003 - Pre-Law (Rows) TotalPercentage 0 8 10 13 14 12 5 0 1 0 63 0.00% 1 3 22 60 56 45 31 1 1 1 220 2.19% 2 3 25 93 154 77 26 2 0 515 7.64% 3 1 20 80 187 225 154 50 8 2 727 17.88% 4 2 4 34 111 224 234 77 17 0 703 25.23% 5 0 4 16 49 91 203 77 1 460 24.40% 6 0 2 2 15 40 51 44 7 1 162 15.97% 7 0 0 1 4 7 8 6 3 0 29 5.62% 8 0 0 0 0 0 0 1 0 0 1 1.01% 0 0 0 0 0 1 0 0 0 1 0.03% Total1787299590779764282582,881Percentage 0.59% 3.02% 10.38% 20.48% 27.04% 26.52% 9.79% 2.01% 0.17% Number of of Candidatures with N Women in 2003 - Pre-Law (Rows) TotalPercentage 0 0 0 0 1 6 4 1 0 12 0.90% 1 0 0 0 0 2 8 25 1 0 36 2.70% 2 0 0 0 0 3 28 90 7 0 128 9.62% 3 0 0 0 1 54 180 21 1 259 19.46% 4 0 0 0 1 4 52 179 51 2 289 21.71% 5 1 0 0 0 1 40 194 63 4 303 22.76% 6 0 0 0 0 0 25 141 51 222 16.68% 7 0 0 0 0 0 2 32 23 0 57 4.28% 8 0 0 0 0 0 0 9 12 0 21 1.58% More 0 0 0 0 0 0 4 0 0 4 0.30% Total 0.00% 0.00% 0.00% 0.15% 0.98% 16.15% 64.46% 17.28% 0.90% Quota's Impact: Number of Lists by Number of Women CandidatesTABLE 1Number of Candidatures with N Women in 2007 - Post Law (Columns) Number of Candidatures with N Women in 2007 - Post-Law (Columns) Panel B: Treatment Group; Inequality Law Applies in 2007 (Population between 5,000 and 8,000) Panel A: Control Group; Inequality Law Does not Apply in 2007 (Population between 2,000 and 8,000) Share of Elected WomenTop 5Placebo (1) (2) (3) (4) (5) (6) Equality Law Binds x 20070.0720.0760.0000.0860.0830.043(0.015)***(0.015)***(0.017)(0.005)***(0.007)***(0.008)***Year 20070.0480.0470.05(0.003)***(0.004)***(0.004)***Two-sided Population Polinomials in Distance to DiscontinuityYesYesYesYesYesYesFENoneProvinceProvinceMunicipalityMunicipalityMunicipalityYears2003-20072003-20072003-2007Observations (Municipality by Year)Number of MunicipalitiesR-squared0.330.410.240.520.370.23Standard errors in parentheses* significant at 10%; ** significant at 5%; *** significant at 1%TABLE 2Impact of Equality Law on Women's Political ActivismShare of Women in Municipal Lists All Positions in List New List in 2007 (yes=1, no=0) Equality Law Dummy0.0100.021(0.027)(0.031)Max{0, Required Women in 2007 - Women in 2003 List } x Equality Law Dummy-0.006(0.007)Women in List 20030.001-0.0010.000(0.002)(0.003)(0.003)Share of Vote in 2003 (columns 1 and 2) or 2007 (column 3)-0.174-0.159-0.229(0.016)***(0.022)***(0.021)***Party Fixed EffectsyesyesyesMunicipal Fixed Effectsyes Two-sided Population Polinomials in Distance to Discontinuity yes no yes Observations (Lists in 2003 - columns 1 and 2; Lists in 2007 - column 3) 528952895678 R-squared 0.71 0.78 0.69 Standard errors in parentheses (clustered at the municipal level)* significant at 10%; ** significant at 5%; *** significant at 1%2003 List Missing in 2007 (yes=1, no=0)TABLE 3Female Quotas Do not Impede Local List Formation Share of Candidates who are Women Share of Candidates who are Women Appearing only in the 2007 List Share of Candidates who are Women Share of Candidates who are Women Appearing only in the 2007 List (1) (2) (3) (4) Equality Law x 20070.0820.0840.080.065(0.007)***(0.006)***(0.011)***(0.009)***Two-sided Population Polinomials in Distance to Discontinuityyesyesyesyes Municipal and Year Fixed Effects yes yes yes yes Observations (All Individuals in all Lists and relevant municipality, by year) Sample (Years) 2003 and 2007 2003 and 2007 2003 and 2007 2003 and 2007 Number of municipalities R-squared 0.01 0.01 0.01 0.01 Standard errors in parentheses* significant at 10%; ** significant at 5%; *** significant at 1%TABLE 4 All Candidates Only Candidates in Top 5 Positions Female Quotas: Effects on Female Candidate Retention, Promotion, or New Arrivals Share of Blank and Null Votes OLSTown Fixed Effects + IV (1) (2) (3) (4) (5) Share of Women Candidates 0.145-0.008-0.019-0.024-0.02(0.024)***(0.044)(0.031)(0.046)(0.028)Year 2007-0.035-0.019-0.018-0.0170.001(0.002)***(0.003)***(0.003)***(0.007)**(0.002)InstrumentsNone Inequality Law Binding x 2007 Inequality Law Binding x 2007 Number of Municipalities1,4371,4371,437 421 (Treated Only) 1,437Municipality FEYes Two-sided Population Polinomials in Distance to Discontinuity YearsStandard errors in parentheses* significant at 10%; ** significant at 5%; *** significant at 1% 2003-2007 Town Fixed Effects + IVYes Yes TABLE 5Impact of a Female 40% Quota on Measures of Voter ParticipationVoter Participation RateInequality Law Binding x 2007 x Distance to 40% Target in 2003 (1)(2)(3)(4)(5)(6)(7)Change in Number of Women in List (Instrumented)0.0070.010.0240.010.0090.0160.004(0.003)**(0.003)***(0.013)*(0.003)***(0.003)***(0.005)***(0.005)Number of Women in List in 20030.014(0.010)Share of Candidates Repeating from 2003 List-0.0130.054(0.012)(0.012)***List Won 2003 Election-0.059(0.004)***Observations (Lists by Municipality in 2007)Number of municipalitiesR-squaredMunicipality Fixed EffectsyesyesyesyesyesyesyesParty Fixed EffectsyesyesyesyesyesyesyesSample (Population Size)5000-8000InstrumentWeightsSample Only TreatedNon-Competitive ElectionsCompetitive ElectionsStandard errors in parentheses* significant at 10%; ** significant at 5%; *** significant at 1%All regressions include municipality and National-Party fixed effects.All Treated and ControlTABLE 6Effects of Female Quotas: Positive Impact of Women On Electoral Results2007-2003 Change in Share of the Vote for Party2000-8000 Party Vote Share in 2003 Max{ 0 , Required Women in 2007 - Women in 2003 List } x Equality Law Dummy Women's Share in 2003 (1) (2) (3) Non-competitive Town in 2003-0.062-0.0120.008(0.060)(0.007)*(0.007) Non-competitive Town x List Won the 2003 Election -0.033 (0.011)***List won the 2003 Election-0.001(0.006)R-squared0.080.010.01Sample (Year)Sample (Population Size) Two-sided Population Polinomials Weights Observations (Lists in Municipalities in 2003) Robust standard errors in parentheses (clustered at the municipality level)* significant at 10%; ** significant at 5%; *** significant at 1% 5281 Women's Share 2003: Top 5 Positions TABLE 7Electoral Competition and Share of Women: Pre-QuotaParty Vote Share in 2003 2003 2000-8000 Yes MayorMayorMayor (1) (2) (3) (4) (5) (6) (7) (8) (9) Equality Law x 20070.0790.0460.0360.0860.0350.0000.0760.0520.052(0.007)***(0.010)***(0.017)**(0.012)***(0.015)**(0.021)(0.009)***(0.013)***(0.023)**0.050.0450.0190.0550.0650.0340.0470.0350.012(0.004)***(0.005)***(0.009)**(0.006)***(0.008)***(0.010)***(0.005)***(0.007)***(0.012)List Fixed EffectsyesyesyesyesyesyesyesyesyesTwo-sided Population Polinomials in Distance to DiscontinuityyesyesyesyesyesyesyesyesyesR-squared0.690.680.70.740.730.810.670.660.66Observations (Lists Appearing in 2003 and 2007 by Municipality and year)Standard errors in parentheses* significant at 10%; ** significant at 5%; *** significant at 1% Share of Women in List by Position of Candidate: TABLE 8Women Leveraged Bargaining Power from Quota All Lists Winner List in 2003 Lists Non-Winner in 2003 High-Status First Names Electability Relative to Mean Low-Status First Names Electability Relative to Mean Common Surnames Percentage IZASKUN 2.20 TANIA 0.15 GARCIA 3.60% IDOIA 2.13 LAIA 0.22 FERNANDEZ 2.27% VIRTUDES 1.91 LETICIA 0.25 GONZALEZ 2.18% MIREN 1.81 TAMARA 0.32 RODRIGUEZ 2.11% ALFONSA 1.76 LORENA 0.34 SANCHEZ 1.98% REYES 1.70 VANESSA 0.36 MARTINEZ 1.88% MAITE 1.59 MACARENA 0.42 LOPEZ 1.87% AMAIA 1.57 NOEMI 0.42 PEREZ 1.82% AGUSTINA 1.56 ESTER 0.47 MARTIN 1.19% AINHOA 1.53 ELISABET 0.47 GOMEZ 1.11% MONTSE 1.53 GRACIA 0.47 HERNANDEZ 0.82% TRINIDAD 1.46 MARCELINA 0.49 RUIZ 0.81% CANDELARIA 1.42 CONSOLACION 0.49 JIMENEZ 0.76% ESMERALDA 1.39 ELISABETH 0.50 DIAZ 0.74% DULCE 1.37 VICTORIA 0.50 ALVAREZ 0.71% CONCEPCIO 1.35 SABINA 0.51 MORENO 0.65% BENITA 1.35 MIREIA 0.52 MUÃOZ 0.59% MARGARIDA 1.35 SANDRA 0.52 EMMA 1.34 PATROCINIO 0.54 Total 25.08% RITA 1.33 ALBA 0.55 Top 5 Below Top 5 Top 5 Below Top 5 Top 5 Below Top 5 Top 5 Below Top 5 Equality Law x 2007 x Woman 0.008 0.001 -0.010 -0.014 0.002 0.001 -0.009 0.003 (0.007) (0.006) (0.008) (0.007)** (0.008) (0.007) (0.016) (0.013) Woman Candidate 0.161 0.193 0.247 0.244 0.247 0.254 0.164 0.145 (0.003)*** (0.003)*** (0.004)*** (0.003)*** (0.004)*** (0.003)*** (0.007)*** (0.006)*** Woman x 2007 0.019 0.014 0.028 0.041 0.002 -0.014 -0.012 -0.021 (0.004)*** (0.004)*** (0.005)*** (0.004)*** (0.004) (0.004)*** (0.009) (0.008)*** Equality Law x Woman 0.001 0.000 0.004 0.003 -0.015 -0.009 -0.014 -0.017 (0.006) (0.005) (0.007) (0.006) (0.007)** (0.006) (0.013) (0.011) Municipal FE yes yes yes yes yes yes yes yes Observations (Candidates) 54,614 71,062 54,614 71,062 54,614 71,062 54,614 71,062 Standard errors in parentheses (clustered at the municipal level) * significant at 10%; ** significant at 5%; *** significant at 1% Quotas Did not Increase Presence of High-Status NamesTABLE 9 (1) (2) (3) =1 if candidate has low-status female first name=1 if candidate is female and has rare surname=1 if candidate is female and has common surname=1 if candidate is female and has long namePanel B: Impact of Quotas on High-Status FemalesPanel A: Highest and Lowest Power-Status Female Names, and Common Surnames (1)(2)(3)(4)(5)(6)(7)(8)(9)Equality Law Binds x 20070.0890.0840.0830.0860.0870.0810.0860.0880.09(0.006)***(0.005)***(0.005)***(0.005)***(0.007)***(0.006)***(0.007)***(0.009)***(0.013)***Year 20070.0430.0470.0490.0480.050.0550.0510.0490.042(0.002)***(0.002)***(0.002)***(0.003)***(0.003)***(0.004)***(0.005)***(0.006)***(0.009)***Municipalities (Population Size)0 to 100001000 to 90001500 to 85002000 to 80002500 to 75003000 to 70003500 to 65004000 to 60004500 to 5500ObservationsNumber of MunicipalitiesR-squared0.230.410.470.520.540.610.640.650.7Polinomial in Distance to DiscontinuityFEYearsAppendix TABLE 1Impact of Equality Law Female Participation: Banwidth robustness Share of Women in Municipal Lists YesMunicipality2003-2007 Unemployment %Telephone LinesRegistered CarsBanksManufacturing PlantsCommercial EstablishmentsShopping MallsRestaurants and Bars (1) (2) (3) (4) (5) (6) (7) (8) Equality Law Binds 0.243-65.40676.693-0.0957.5773.1220.009-1.978(0.266)(91.956)(342.850)(0.376)(5.962)(8.939)(0.035)(6.002)ObservationsR-squared0.450.750.250.50.310.570.050.5 Two-sided Population Polinomials in Distance to Discontinuity FE Years Standard errors in parentheses* significant at 10%; ** significant at 5%; *** significant at 1%YesProvinceProvinceAppendix TABLE 2No Other Discontinuities at Inequality Law Binding Population Level Low Status Name (1) (2) (3) Low-Frequency Surname0.0060.014-0.015(0.004)(0.004)***(0.004)***Very Common Surname-0.01-0.0150.011(0.004)**(0.004)***(0.004)***Long Name0.0270.025-0.114(0.003)***(0.003)***(0.003)***Constant0.2050.210.331 (0.003)*** (0.005)*** (0.003)*** Region FEYesYesYearObservations70,55670,55698,941 R-squared 0 0.02 0.02 Robust standard errors in parentheses* significant at 10%; ** significant at 5%; *** significant at 1% ^ Excludes Maria, which is very often used with another first name Elected Woman Candidate Appendix TABLE 3Low-Status Surname Predicts Low Power and Low-Status Name IZA Discussion Paper No. 5645 We develop a model that nests previous explanations for women under-representation in positions of power. Focusing on democratic electoral dynamics, our framework delineates the three types of mechanisms that may be at play: consumer demand, candidate supply, and internal party dynamics beyond electoral markets. We use Spains Equality Law, requiring a 40 percent female quota in electoral lists, to test the alternative theories. The law was enacted by the social-democratic party after the surprise parliamentary electoral results following the Madrid terrorist bombings, and was therefore completely unexpected by regional political machines. The law only applied to towns with populations above 5000, so we can use a treatment-control, before-and-after discontinuity design to learn about the impact of female politicians in local elections. Our evidence is most consistent with the existence of entrenched male-dominated political machines capturing influential power positions within the parties. JEL Classification: J16, J71 Keywords: female political representation Corresponding author: Albert Saiz The Wharton School University of Pennsylvania Steinberg-Dietrich Hall, Suite 1466 3620 Locust Walk Philadelphia, PA 19104-6302 USA E-mail: saiz@wharton.upenn.edu We thank seminar participants at IESE, Universitat Pompeu Fabra, and the AEDE 2010 meetings for their thoughtful comments. Casas-Arce acknowledges support from the Ministerio de Ciencia y Tecnología (SEJ2007-64340). Saiz acknowledges assistance from the Wharton Global Initiatives DISCUSSION PAPER SERIES Forschungsinstitut zur Zukunft der ArbeitInstitute for the Study of Labor Women and Power:IZA DP No. 5645Pablo Casas-Arce