The Jackie and Jill Robinson Effect Why Do Congresswom - PDF document

TheJackie(andJill)RobinsonEffect:WhyDoCongresswomenOutperformCongressmen?SarahF.AnziaStanfordUniversityChristopherR.BerryTheUniversityofChicagoIfvotersarebiasedagainstfemalecandidates,onlythemosttalented,hardestworkingfemalecandidateswillsucceedintheelectoralprocess.Furthermore,ifwomenperceivetheretobesexdiscriminationintheelectoralprocess,oriftheyunderestimatetheirqualificationsforoffice,thenonlythemo WHYDOCONGRESSWOMENOUTPERFORMCONGRESSMEN?RelatedResearchandBackgroundIntheempiricalliteratureonthedistributionoffederalspendingacrosscongressionaldistricts,littlehasbeendonetoestimatedifferencesindistributivespendingbylegislatorsex.Meanwhile,theliteratureonthesourcesoflegislativeproductivityhasfocusedalmostexclusivelyongeneralconditionswithinthelegislatureratherthanthecharacteristicsoflegislatorsthemselves.Withintheliteratureonwomeninpolitics,however,thereisagreatdealofrelevantscholarlywork.Tostart,scholarshaveamassedevidencethatmenandwomenofequalpoliticalqualificationsdonoten-tertainthepossibilityofrunningforofficewithequalfrequencies.LawlessandFox(2005)findthatpoliticallyeligiblewomenwiththesameobjectivequalificationsasmenarelesslikelytoconsiderthemselvesqualifiedtorunforpublicoffice.Moreover,womenexpressgreatercon-cernthanmenabouttheirabilitytoraisethenecessaryfinancialsupportandwinelections(Duerst-Lahti1998;FowlerandMcClure1989;FoxandLawless2004;NWPC1994).Thedifferencesinmen’sandwomen’spoliticalam-bitionmightbetheresultofdifferencesinmaleandfemalesocialization,psychology,andpersonallifecircumstances(Burrell1994).NiederleandVesterlund(2007),forexam-ple,findthatwomenshyawayfromcompetitionmoresothanmen.Inaddition,women’spoliticalambitionmightbedampenedbytheperceptionofsexbiasinpolitics:over90%ofwomenand75%ofmeninthecandidateeligibil-itypool(e.g.,attorneys,businesspeople,educators,andpoliticalactivists)believethatthereisbiasagainstwomeninelections(LawlessandFox2005).Alargeliteraturethatcomparesmen’sandwomen’sperformanceincampaignsandelectionsconcludesthatsuchconcernsarelargelyunwarranted.Oneofthemostwell-knownfindingsintheliteratureonwomeninpoliticsisthatfemalecandidateswingeneralelectionsatthesamerateasmalecandidates(Burrell1994;Fox2006;Newman1994;Seltzer,Newman,andLeighton1997).Moreover,womenraiseasmuchmoneyduringtheircampaignsasmen(Burrell1994;Fox2006;UhlanerandSchlozman1986).Onthebasisofaggregatevotetotalsandcampaignfunds,then,manyscholarshaveconcludedthatdiscrimi-nationagainstwomeninpoliticsisaphenomenonofthepast(e.g.,Fox2006;Seltzer,Newman,andLeighton1997;SmithandFox2001).Bycontrast,severalexperimentalstudiessuggestthatvotersharborbiasagainstfemalecandidates.Rosen-wasserandDean(1989)findthatvotersprefer“mascu-Forimportantcontributions,seeHowell,Adler,Cameron,andRiemann(2000)andClintonandLipinski(2006).line”traitsincandidatesforalllevelsofpublicoffice,andHuddyandTerkildsen(1993a,1993b)showthatvoters’genderstereotypesaremostharmfultofemalecandi-datesrunningfornationaloffice.FoxandSmith(1998)presentsubjectswithaseriesofhypotheticalmaleandfemaleHousecandidatesandfindthatsignificantlyfewersubjectschoosetovoteforfemalecandidates(seealsoDolan1997).Moreover,surveyevidenceshowsthatasubstantialpercentageofAmericanadultsexpressreluctancetosup-portwomeninthepoliticalarena.TheGeneralSocialSurvey,forexample,revealsthat23%ofadultsthinkthatmostmenarebettersuitedemotionallyforpoliticsthanmostwomen(Dolan2004).Inaddition,21%ofAmeri-canadultssaythatmenmakebetterleadersthanwomen,and51%saythat“manyAmericansarenotreadytoelectawomantohighoffice”(Pew2008).ArecentGallupsur-veyshowsthat11%ofbothmenandwomensaytheywouldnotvoteforafemalepresidentialcandidateevenifshewerequalifiedforthejob,andanother11%saytheywouldvoteforaqualifiedwomanonly“withreservations”(NewportandCarroll2007).Itispossiblethatthesefig-uresunderstatetheprevalenceofsexbiasamongvoters,sincepressuretoprovidesociallydesirableresponsesof-tenpreventssomerespondentsfromadmittingtheirun-derlyingprejudices,iftheyareevenawareofthem(seeFoxandSmith1998).Evenso,11%ismorethanthepercentageofrespondentswhosaidtheywouldnotvoteforablackcandidateforpresident,andyetfewar-guethatracialdiscriminationinpoliticshasdisappeared(e.g.,SearsandHenry2005).Arecentgroupofstudiessuggeststhatthepathtocongressionalofficemaypresentmorehurdlestowomenthantomen.LawlessandPearson(2008)findthatcon-gressionalprimaryelectionsinwhichatleastoneofthecandidatesisfemaletendtoattractlargernumbersofcontenders.PalmerandSimon(2006)showthatfemaleincumbentsaresignificantlylesslikelythanmaleincum-bentstofaceuncontestedprimaryandgeneralelections.Similarly,MilyoandSchosberg(2000)findthatfemalecandidatesaresignificantlymorelikelytofacehigh-qualitychallengersthanmalecandidates.Moreover,po-liticalpartyleadersbelievethatthereisgenerallymoreuncertaintyaboutawoman’selectabilitythanaman’s;hence,theyarelesslikelytorecruitwomentorunforoffice(Sanbonmatsu2006).Notably,thewomenwhoemergeascongressionalcandidatestendtohavegreaterDuetoitstiming,thesurveymightconflatevoters’opinionsaboutwomenwiththeirviewsonHillaryClinton’scandidacy,buttheaggregateresponsesdonotdifferdramaticallyfromsurveysconductedinthe1990s(FoxandSmith1998). SARAHF.ANZIAANDCHRISTOPHERR.BERRYpriorpoliticalexperiencethanmalecongressionalcandi-dates(PearsonandMcGhee2009).Furthermore,whilefemalecandidatesraisethesameamountofmoneyasmalecandidates,Jenkins(2007)findsthattheyhavetoworkhardertodoso.Anotherlargeliteratureexaminesthedifferencesbe-tweenmenandwomenoncetheyareoffice.Thereisevidencethatfemalelegislatorsdirectmoreoftheirattentiontopolicyareasthoughtofas“women’sis-sues”(e.g.,Norton1999;Thomas1991;Swers2002).Inaddition,thepresenceofwomeninlegislatureshasbeenshowntoinfluencethenatureofpolicyoutcomes(BesleyandCase2003;ChattopadhyayandDuflo2004;Rehavi2007).Forthemostpart,however,thisliteratureisnotwellintegratedwithworkthatexaminestheperfor-manceofwomeninelectoralpolitics.Inthenextsection,weproposeatheoryofpoliticalselectionthatconnectstheperformanceofwomenincampaignsandelectionswiththeirperformanceonceinoffice.ATheoryofPoliticalSelection:TheJackie(andJill)RobinsonEffectIn1947,JackieRobinsonbecamethefirstAfricanAmer-icantoplayMajorLeagueBaseball.Heiswidelyreveredasoneofthegreatestplayersinthehistoryofthegame.Thisisnocoincidence.IfRobinsoncouldhavebeenre-placedeasilybyawhiteplayer,noteamwouldhavebeenwillingtotakeachanceonhim,giventhewidespreadbigotryofthetime.Robinsontobebetterthanal-mostanywhiteplayerinordertoovercometheprejudiceofowners,players,andfans.Ofcourse,thisstoryisnotuniquetoRobinson.PascalandRapping(1972)foundthatblackMajorLeagueBaseballplayersin1967outper-formedwhiteplayersineveryposition.Noristhestoryuniquetobaseball.Thereiswidespreadevidencethatblackathleteshavehistoricallyfacedhigherperformancestandardsforentryintoprofessionalsportsthanwhiteathletes(seeKahn1991).Moregenerally,Becker(1957)pioneeredtheideathatworkerswhofacediscriminationThisviewisexplainedbyHankAaron,himselfanAfricanAmer-icanformerballplayeranderstwhileholderofthemajorleaguecareerhome-runrecord.AccordingtoAaron(1999),“JackieRobin-sonhadtobebiggerthanlife.HehadtobebiggerthantheBrooklynteammateswhogotupapetitiontokeephimofftheballclub,big-gerthanthepitcherswhothrewathimorthebaserunnerswhodugtheirspikesintohisshin,biggerthanthebenchjockeyswhoholleredforhimtocarrytheirbagsandshinetheirshoes,biggerthantheso-calledfanswhomockedhimwithmopsontheirheadsandwrotehimdeaththreats.”inthelabormarketmustperformbetterinordertoearnthesamewageasotherworkers.Wesuggestthatasimilarperformancepremiumisdemandedoffemalepoliticianswhenthereissexdis-criminationintheelectorate.Ifvotersareprejudicedagainstwomen,thenawomanmustbebetterthanthemansherunsagainstinordertowin.Moreover,ifwomenanticipatediscriminationbyvoters,orsimplyun-derestimatetheirownqualifications,thenonlythemostformidablewomenwillrunforofficetobeginwith.Inei-thercase,ourpredictionthatsex-basedselectionwillleadthewomeninofficetoperformbetter,onaverage,thanthemenflowsnaturallyfromtheliteratureonpoliticalagency,whichfocusesontwoissues:moralhazardandselection.Inmoralhazardmodelsofelections,originatingwithBarro(1973)andFerejohn(1986),thedesiretobere-electedinthefuturemotivatespoliticianstoexerteffortwhileinoffice.Citizensvoteretrospectively,reelectingtheincumbentonlyifhisperformanceisaboveathresh-oldlevelchosentomaximizetheincumbent’sincentivetoworkandhencethevoter’sexanteexpectedutility.Othercontributionsthatfocusonelectionsassanction-ingdevicesforinducingeffortfrompoliticiansincludeAusten-SmithandBanks(1989),Seabright(1996),andPersson,Roland,andTabellini(1997).Asecondbodyoftheoryconceivesofelectionsasdevicesforselectinghigh-qualitypoliticians,or“goodtypes,”intooffice(e.g.,Gordon,Huber,andLanda2007;Zaller1998).Inthisview,votersuseinformationgleanedfromcampaignsandfromincumbents’performanceinofficeassignalsaboutintrinsiccharacteristicsofcandi-dates,suchastalentorhonesty.Electionsselectgoodtypesandfilteroutbadtypes,buttheydonotalterpoliticians’behaviorinoffice.Therehavebeenattemptstoadjudicatebetweentheelectoralselectionandmoralhazardmodels(e.g.,Fearon1999)aswellasattemptstounifythem(e.g.,Forarecentsurveyoftheeconomicsofdiscrimination,seeRodgers(2006).Itmayseemthatamoreobviousanalogyiswithracialdiscrim-inationinpolitics.However,theuseofrace-consciousdistrictingconfoundstheproblem.Wereturntothisissueattheendofthearticle.Technically,somevotersmightreverse-discriminate,thatis,givepreferencetofemalecandidates.Ourpredictionstillholdsaslongastheproportionthatdiscriminatesisgreaterthanthepropor-tionthatreverse-discriminates.Itisworthnotingthatthegreaterthelevelofdiscriminationbyvoters,donors,andgatekeepers,thegreatershouldbetheobservedqualitydifferentialforwomenwhowinelections.Ofcourse,ifdiscriminationisstrongenough,itispossiblethatnoqualityadvantagewillbesufficienttoovercomeit,inwhichcaseweshouldnotobservewomenwinningelections. WHYDOCONGRESSWOMENOUTPERFORMCONGRESSMEN?Ashworth2005;BanksandSundaram1998;Besley2006).Wedonotstakeoutapositiononsuchissuesbutratheremphasizethattheimplicationsofsex-basedselectionarethesameineitherframework.Weassumethatper-formanceisafunctionofacandidate’sinnateabilityandhereffort.Inotherwords,acandidatewillperformbet-terinofficeifsheismoreable,worksharder,orboth.Therefore,ifvotersdiscounttheabilityoffemalecandi-dates(electoralselectionmodel),orifvotersdemandahigherperformancethresholdforwomen(moralhazardmodel),thenthewomenwhowinwillperformbetterinofficethanthemenwhowin,onaverage.Furthermore,iffemalecandidatesanticipatethattheywillfacediscrim-inationintheelectionorotherwiseunderestimatetheirchancesforelectoralsuccess,thewomenwhorunforofficewillbethosewhoexpecttoexceedthehigherper-formancethresholddemandedbyvoters.Wenotethatinorderforthispredictiontohold,theattributesthatmakesomeoneahigh-qualitycan-didatemustberelatedtotheattributesthatmakeherahigh-qualitylegislator.Ifthetwowereuncorrelated,thenwewouldnotexpecttoobserveadifferencebetweentheperformanceofmaleandfemalelegislatorsinof-fice.However,aslongascandidatequalityandlegislatorqualityarepositivelycorrelated,theprocessofsex-basedselectionshouldresultinalegislatureinwhichtheaver-agefemalerepresentativeoutperformstheaveragemalerepresentative.Importantly,ourtheoryofsex-basedselectioncanaccommodatetwoapparentlyconflictingstrandsofevi-dencefromtheexistingliteraturediscussedabove.Ononehand,asizableproportionofvotersappearstobebiasedagainstfemalecandidatesinpresidentialelections(e.g.,NewportandCarroll2007)andinhypotheticalHouseelectioncandidatematch-ups(FoxandSmith1998).Ontheotherhand,numerousstudiesoffemalecandidates’votetotalsandsuccessratesinHousegeneralelectionshavefoundthattheydojustaswellasmalecandi-dates(e.g.,Burrell1994).Ourtheoryshowsthatthereisnoinconsistencybetweenthesesetsoffindings.Ifonlyhigher-qualityfemalecandidateswillactuallyrunforof-fice,thenwewouldnotnecessarilyexpecttoobserveavoteorcampaignfundingdifferentialbetweenmaleandfemalecandidatesevenifthereis,infact,discriminationbyvotersanddonors.Yet,iftheaveragefemalecandi-dateisofhigherqualitythantheaveragemalecandidatebutreceivesthesameamountoffundingandwinsthesamenumberofvotes,sheisclearlynotonequalfoot-ingwiththeman.Therefore,existingstudiesthatsimplyAgain,wefindananalogoussituationinprofessionalsports.Onaverage,blackplayersintheNBAearnsalariesequaltothoseofcomparewomen’sandmen’svotesharesarenotdirectlyinformativeaboutthepresenceorabsenceofdiscrimina-tionbyvoters.Theworkingsofthecandidateselectionstageconfoundmeasurementofvoterdiscriminationattheelectoralstage.Theoryaside,theexistingevidencesuggeststhatbothfemaleself-selectiononqualityvoterdiscriminationareatwork.Iftheaveragewomanrunningforofficewereofhigherqualitythantheaveragemanandvotersdidnotdiscriminate,thenweshouldobservefemalecandidateswinningathigherratesthanmen.Buttheydonot.Ifvotersdiscriminatebutwomendonotself-selectbasedonquality—implyingthattheaveragefemalecandidateisequalinqualitytotheaveragemale—thenweshouldseewomenlosingmoreoftenthanmen.Buttheydonot.Ifthetwooccurincombination,suchthatvotersdiscrimi-nateagainstfemalecandidatesandfemalecandidatesself-screeninanticipationofthatdiscrimination,wewouldobservefewerbutmorequalifiedwomenrunningforof-ficeandpossiblyequalelectoralsuccessratesformaleandfemalecandidates.Thislastsetofcircumstancesistheonemostconsistentwithexistingempiricalevidence(PearsonandMcGhee2009).Importantly,however,ourtheoreticalpredictionholdsregardlessofwhetherdiscriminationbyvotersoccursaloneorincombinationwithself-screeningbycandidates:ineithercase,thewomenwhorunandwinwillperformbetter,onaverage,thanthemenwhorunandwin.Weemphasizethatwearearguingthatwomenhavemoreinnatepoliticaltalentthanmen,nordoweclaimthatallfemalecandidatesoutperformtheirmalecounterparts.Ourtheorysimplyidentifiesaconnectionbetweentheeconomicsofdiscriminationandmodelsofpoliticalagency:whensexdiscriminationispresentamongvoters,womenmustbebetterthantheirmalecounterpartstobeelected.Ifwomenanticipatesuchdis-crimination,oriftheyunderestimatetheirchancesforelectoralsuccess,thenonlythemostqualifiedwomenwillruninthefirstplace.Therefore,onaverage,thewomenweobserveinofficewillperformbetterthanthemen,allelseequal.whiteplayers.Someseethispayparityasevidencethatdiscrimina-tionhasbeenovercome.Otherssuggestthatblackplayersarebetteronaveragethanwhiteplayersandthatsalaryequalityisevidenceofdiscriminationratherthanitsabsence.SeeKahnandShererThereis,ofcourse,anotherscenariotoconsider.Inthecasethatthereisnodiscriminationbyvotersandpotentialfemalecandidatesdonotself-screen,wewouldexpectthatfemalecandidateswouldwinatratesequaltomalecandidatesandthattherewouldbenoperformancepremiumonthepartoffemalepoliticians. SARAHF.ANZIAANDCHRISTOPHERR.BERRYEmpiricalStrategyandDataAsourprimarymeasureofalegislator’sperformance,welooktohersuccessindeliveringfederalprogramspendingtoherhomedistrict.Theuseofspendingasanindica-torofincumbentperformancehasstrongempiricalandtheoreticalfoundations.Empirically,congressionalschol-arshavelongobservedthatafundamentalandexplicitgoalofmembersistobringhomefederaldollars,andthisobservationhasbeenacentralthemeintheclassicsonCongress(Fenno1966,1978;Ferejohn1974;Fiorina1981;Mayhew1974).Thereisevidencethatsucheffortsbolsteranincumbent’sreelectionprospects(AlvarezandSaving1997;BickersandStein1996;LevittandSnyder1997;Sellers1997;SteinandBickers1995).Moreover,membersofCongressthemselvesappeartobelievethattheymustservetheirconstituentsthroughbothcaseworkandprojectworktobuildthereputationnecessaryforfu-tureelectoralsuccess(Cain,Ferejohn,andFiorina1987).Thereisalsoastrongtheoreticalmotivationforus-ingdistrictspendingasanindicatoroflegislatorperfor-mance.Inparticular,Ashworth(2005)presentsamodelinwhichreelection-mindedincumbentsfaceafundamentaltrade-offbetweenallocatingtheirresourcestowardpro-ducingdistrict-specificbenefits,suchasfederalprogramspending,ornationalpublicgoods,suchaslegislationorbureaucraticoversight.Voterslearnabouttheabilityofincumbentsbyobservingtwosignals,whichareafunc-tionofthepolitician’seffortonthetwotasks,andreelectthosepoliticianswhomtheybelieveareofhighability.AcentralresultfromAshworth’smodelisthatpoliticianshaveanincentivetobiastheirefforttowardtasksthatvot-ersobservewithlessnoise.Thislogicfavorsthededicationofefforttosecuringdistrict-specificprojects,whicharemoreinformativesignalsoftheincumbent’sabilitythanarenationalpublicgoodsandhencereceivegreaterweightwhenvotersupdatetheirbeliefs.Inotherwords,itistheobservabilityofprogramspendingthatmakesitthemostefficientpathwayforpoliticianstosignaltheirqualitytoconstituents.Withtheseempiricalandtheoreticalmotivations,weadoptthenotunfamiliarassumptionthatlegislatorsareuniversallymotivatedtodirectprojectsandfundingtotheirdistricts(e.g.,Evans2004).Furthermore,whilesomeprogramspendingisformulaic,weassumethatarepre-sentative’stalentandeffortplayanimportantroleinthelogrolling,agendasetting,coalitionbuilding,andotherdeal-makingactivitiesthatcharacterizedistributivepoli-tics.Oftwolegislatorswhocomefromdistrictswithsim-ilarcharacteristics(orwhorepresentthesamedistrictatdifferenttimes),theonewhosucceedsindirectingmorespendingtoherdistrictcanbedeemedtohaveperformedbetterinthecontextofthisfundamentalpoliticalpursuit.Ofcourse,werecognizethatdeliveringfederalbenefitstothehomedistrictisonlyoneaspectofalegislator’sjob.Therefore,weroundoutouranalysisoflegislativeper-formancebyexamininglegislators’billsponsorshipandcosponsorshipactivity.FederalOutlaysDataTocomparefederalprogramspendingincongressionaldistrictsrepresentedbymenandwomenintheU.S.HouseofRepresentatives,weusedatafromtheFederalAssis-tanceAwardDataSystem(FAADS).FAADSisacompre-hensivesourceforfederaldomesticspendingprogramsandreportsexpendituresofabout1,000programs,in-cludingagriculturalprograms,educationgrants,researchgrants,largeentitlementprograms,andmanyothers.WeaggregatedtheFAADSrecordstoproduceadatasetthatincludes9,135federaloutlaysobservationsforcon-gressionaldistrictandfiscalyearcombinations,trackingapproximately$20.8trillioninfederalexpendituresfrom1984to2004(in2004dollars).WeattributethefederaloutlaysforeachfiscalyeartothememberofCongresswhorepresentedthecongressionaldistrictinthecalendaryearprior.Congressionaldistrictboundariesareredrawnevery10yearsduetodecennialreapportionmentandredis-tricting,andthereforewehadtotracedistrictsovertimeinconstructingourpanel.Afteradecennialredistrict-ing,somedistrictsremainessentiallyintactwhileotherschangebeyondrecognition.Weconsideradistricttobeacontinuousentityacrossredistrictingperiodsifthema-jorityofthelandareaofthepost-redistrictingdistrictismadeupofpre-redistrictingdistrictlandarea.Other-wise,whenwecouldnotmatchanewdistrictclearlytoapreexistingdistrict,thenewdistrictistreatedasanewunitfollowingthedecennialredistricting.Consequently,thepanelincludes733uniquedistrictentitiesoverthe21-yeartimeperiod.FAADSreportsawardtransactionsandrecipientcon-gressionaldistrictsaccordingtotheinitialrecipient.Thisposesaproblemforawardsmadetostatesforredistri-butionthroughoutthestate:FAADSgrosslyinflatesfed-eraloutlaystothecongressionaldistrictsthatcontainstatecapitolbuildings.WeimproveabituponLevittandSeeAppendixAforadetaileddescriptionofthedata.Themajoromissionisdefense:militaryspendinganddefenseprocurementprogramsarenotincluded.Theresultspresentedinthefollowingpagesarenothighlysen-sitivetothelandareathresholdusedformatchingdistrictsacrossyears.SeeAppendixA. WHYDOCONGRESSWOMENOUTPERFORMCONGRESSMEN?Snyder’s(1995)treatmentofthisissuebyincludingacontrolvariableequaltothefractionofthestatecapitolcountycontainedineachcongressionaldistrict,weightedbythestatepopulation.Moreimportantly,weincludedistrictfixedeffects,asexplainedbelow,whichcontrolfortime-invariantfactors,suchasbeingpartofthestatecapitolcounty.Finally,wenotethattheFAADSdataincludeagreatdealoffederalspendingbybroad-basedentitlementpro-grams,suchasSocialSecurityandMedicaid,thedistribu-tionsofwhicharedeterminedbyformula.Ithardlyseemsappropriatetoattributethiskindofspendingtothepo-liticalskillandeffortofadistrict’srepresentative.Inordertoseparatebroad-basedentitlementprogramsfrompro-gramsthatrepresentdiscretionaryspending,weadoptatacticusedbyLevittandSnyder(1995,1997).Specifi-cally,wecalculatecoefficientsofvariationindistrict-levelspendingforeachprogramcontainedintheFAADSdataandusethecoefficientstoseparateprogramsintotwocategories:low-variationprogramshavecoefficientsofvariationlessthan3/4,andhigh-variationprogramshavecoefficientsofvariationgreaterthanorequalto3/4.Thelow-variationcategoryincludes26programs,mostofwhichareprogramswithintheVeteransBenefitsAd-ministration,theCentersforMedicare&MedicaidSer-vices,andtheSocialSecurityAdministration,whichmakeup76%ofthespendinginourdata.Thehighvariationcategorycompriseshundredsofsmallerprograms.Intheanalysisthatfollows,weexaminespendingfromhighvariationprogramsonly,inexpectationthatlegislatorabilityandeffortplayalargerroleinthedistributionofthistypeofspending.Weadjustthespendingdatato2004dollars.Themeanvalueofdistrict-levelhighvariationprogramout-laysrangesfrom$398millionin1984to$753millionin2003.Themedianvalueincreasesfrom$151millionto$361million.Ofthe9,135congressionaldistrictandfiscalyearcombinations,8,307observationsrepresentannualoutlaystodistrictsledbymalelegislators,andtheremain-ing828areforwomen-leddistricts.Thereare112uniquewomeninthedataset.Thirty-eightstateshadatleastonefemalememberofCongressduringthistimeperiod.Ofthe733uniquedistrictsweobserveacrossthethreeredistrictingplans,133hadafemalerepresentativeforatleastonecongress.DetailsareprovidedinAppendixA.Wetransformthisvariabletoanaturallog.Ourresultsarenotsensitivetothechoiceofthethresholdfordefininghigh-variationprograms.SeeAppendixA.SeeAppendixBforadiscussionofresultsusingspendingfromlow-variationfederalprograms.IdentificationStrategyForourmainanalysis,weuseadifferences-in-differencesapproach,basedondistrictandyearfixedeffects,toiden-tifytheeffectofhavingafemalerepresentativeonadis-trict’sreceiptoffederalprogramoutlays.Essentially,weaskwhetheradistrictreceivesmorefederalspendingdur-ingtheyearsinwhichitsendsawomantoWashingtoncomparedtotheyearswhenitsendsaman.Becausewerelyonvariationwithindistrictsovertimeforidentifica-tion,wecaneliminateany(observableorunobservable)time-invariantattributesofadistrictthatcouldinfluenceboththelikelihoodofelectingawomanandtheflowoffederalspending.Importantly,thedistrictfixedeffectsalsosubsumetime-invariantheterogeneityacrossstates,suchasthewell-knownresultthatsmallerstatesreceivegreaterfederaloutlays,onapercapitabasis,duetomalap-portionmentintheSenate(e.g.,Lee1998).Wespecifythefollowingbasicmodel:ln(outlayswheresubscriptdenotescongressionaldistrictsanddenotestime.Thevariableofinterestis,whichisabinaryindicatorvariablecodedoneifthepersonrep-resentingdistrictattimeisfemale,zeroifmale.Weincludeyearindicators,,tocontrolforgeneralchangesinspendingovertime.Thevectordenotesotherlegislatorcharacteristicsthatmayinfluencespending.Wecontrolforparty,whichisexpectedtoaccountforthetraditionalRepublicanpref-erenceforfiscalconservatismwhichconceivablytempersthepushformorespendingtothehomedistrict.anticipationthatlegislatorsinthemajoritypartyarebet-terpositionedtosecuremoneyfortheirdistricts,wein-cludeanindicatorformajoritypartystatusduringyearsinwhichCongressandthepresidencyarecontrolledbythesameparty.Unifiedgovernmentoccursinonlyfivefiscalyearswithinourstudyperiod:in1994,1995,and2002to2004.Wealsoaddameasureofseniority:thenumberoftermsalegislatorhasservedasoftheyearoftheoutlays.Finally,weintroduceavariablethatequalsamember’stwo-partyvictorymarginintheprecedingcongressionalelection,whichcontrolsforthepossibilitythatelectorallyvulnerablemembersreceivepriorityindiscretionaryspending(Shepsle1978).Thevectorcapturesafairlyrichsetofobservableattributesofcongressionaldistricts:populationlivinginAlvarezandSaving(1997)findthatDemocratsreapgreaterelectoralbenefitfromfunnelingporktotheirdistrictsthandoRepublicans.RecallthatFAADSdataarereportedinfiscalyearsandthatlegislatorcharacteristicsarelaggedbyoneyear. SARAHF.ANZIAANDCHRISTOPHERR.BERRYurbanareas,AfricanAmericanpopulation,population65yearsofageorover,numberoffarmersandfarmmanagers,foreign-bornpopulation,medianfamilyin-come,unemployedpopulation,populationinthearmedforces,populationinpublicschool,andpopulationem-ployedinmanufacturingandconstruction.Althoughtheexistingliterature(e.g.,Burrell1994)andourownsupplementaryanalysisshowthatdistrictdemographiccharacteristicsarenotparticularlyimportantpredictorsofthepresenceofafemalelegislator(seeAppendixB),weremainconcernedthatunmeasureddistrictcharacteris-ticspredictbothlegislatorsexandtheamountoffederalspendingreceivedbyadistrict.Wethereforeincludecon-gressionaldistrictfixedeffects,,toaccountforunob-servable,time-invariantdistrictcharacteristics.Finally,,andareregressioncoefficients,andisanerrorterm.Evenwithabroadsetofcontrolvariables,theun-observable,time-variantpredictorsoffederalspendingwithinaparticulardistrictarelikelytobecorrelatedacrosstimeperiods.Furthermore,thegeographicdistributionoffederalspendinglikelyreflectstheeffectsofsenatorsaswellasthequalityandeffortofHousemembers,suggest-ingthattheremaybecorrelationacrossdistrictswithinastate.Consequently,weuserobuststandarderrorsclus-teredbystatethroughoutouranalysis.AnalysisandResultsTable1presentstheresultsofourfixedeffectsmodelsofhighvariationprogramspending.Model(1)includesdistrictcharacteristics,legislatorcharacteristics,anddis-trictandyearfixedeffects,asdescribedabove.Themainresultisclear:withindistrictsovertime,roughly9%morefederalspendingisbroughthomewhenthereisawomanrepresentingthedistrictinCongressthanwhenthesamedistrictisrepresentedbyaman.Thedistrictfixedeffectssubsumeanyattributesthatdonotchangeovertime,includingtheunchangingat-tributesofthestatesinwhichtheyarelocated.However,Demographicvariablesfor1984to2001comefromScottAdler’s(2003)“CongressionalDistrictDataFile.”Alldemographicdatafor2002to2004comefromthe2000U.S.Census.DetailsareprovidedinAppendixA.Insomeinstances,explainedbelow,weusestatefixedeffectsbecauseofdatalimitations.ThemodelsinTable1have9,067observations,ratherthan9,135,duetomissingvaluesforelectoralmarginandtermsinoffice.Wealsoexcludefiveobservationsthatrecordednegativehigh-variationoutlays.Seetheappendixfordetails.1LegislatorSexandDiscretionaryFederalDomesticSpending (1)(2)(3)(4) Female00910120PrefemaleTrend0PostfemaleTrend0Republican0790022)(0023)(0Majority000300420028)(0028)(0031)(0Terms00100020003)(0003)(0004)(0Margin00000023)(0023)(0Population03503612730)(0732)(0StateCapitol006200620Age65andOlder211)(0211)(0Black01610160Construction00150171)(0173)(0PublicSchool299)(0295)(0Farmers004900450029)(0030)(0ForeignBorn051)(0052)(0Manufacturing133)(0133)(0MedianIncome151)(0152)(0Unemployed01030111)(0111)(0ArmedForces034)(0034)(0Urban00050006)(0006)(0Constant1883718813137450Observations90679067906739R-squared0660ModelDistrict&District&State&First-Specificationyearfixedyearfixedyearfixeddifferenceseffectseffectseffects Notes:Robuststandarderrorsclusteredbystateinparentheses.ThedependentvariableinModels(1)through(3)isln(high-variationfederaloutlays).InModel(4),thedependentvariableisthedifferenceinloggedoutlaysbetweentwoconsecutiveyears.Model(4)includesonlyobservationsinwhichamixed-sexcloseelection(where“close”isdefinedasawinningvoteshareoflessthan55%)resultsinachangeinthesexofadistrict’srepresentative(seetext).Outlaysareinconstant2004dollars.Female1iflegislatorisfemale.Pre-andPost-FemaleTrendsarelineartrendsforthesixyearsbeforeandsixyearsaftertheelectionofawoman,respectively.Republican1iflegislatorisRepublican.Majority1iflegislatorisamemberoftheHousemajoritypartywhenCongressandthepresidencyarecontrolledbythesameparty.Alldemographicvariablesaretransformedasnaturallogarithms.significantat10%level;significantat5%level;significantat1%level. WHYDOCONGRESSWOMENOUTPERFORMCONGRESSMEN?alingeringconcernmaybethatthereareunmeasuredtrendswithindistrictsovertimethatmakethembothmorelikelytoelectawomanandmorelikelytoreceivefederalspending.Toexplorethispossibility,wecomparetheratesofchangeinthefederalspendingreceivedbyadistrictbeforeandafteritelectsafemalerepresentative.Specifically,inModel(2),weusepre-andpostfemalelin-eartimetrendsforthethreetermsbeforeandthethreetermsaftertheelectionofawoman.Wefindthattherateofincreaseinspendingishigherafterawomaniselectedthanitwasbefore.Usingantest,werejectthatthetwotrendsareequal(p.06).Thus,wefindnoevidencethatfemalerepresentativesmerelyinheritanalreadyfa-vorabletrendinspending;instead,thetrendchangesafterawomaniselected.Next,incolumn(3),wepresentamodelthatincludesstateratherthandistrictfixedeffects.Thisapproachal-lowsustotakeadvantageofmorevariationinthedata,as38stateshadatleastonewomaninCongressduringourstudyperiod.Adisadvantageisthatthestatefixedeffectsdonotaccountforunmeasuredwithin-state,between-districtheterogeneity.Thattheresultsofthestateanddistrictfixedeffectsmodelsaresosimilar,therefore,isreassuringandstrengthensourbeliefthatdistrict-levelattributesdonotexplaintheconnectionbetweenlegisla-torsexandfederalspending.Finally,inthespiritofaregressiondiscontinuity(RD)design(Lee2008;ThistlethwaiteandCampbell1960),weestimatechangesindistrict-levelspendingfollowingcloseelectionsinwhichamalecandidateranagainstafemalecandidateandinwhichtheelectionresultedinachangeinthesexofthedistrict’srepresentative.Basedontheclosenessoftheelections,wecaninferthateachdistrictwasroughlyequallylikelytohaveelectedawomanoraman.Whenwerestrictouranalysistomixed-sexracesinwhichthewinningcandidategarneredlessthan55%ofthevote,thereare39instancesinwhichthesexofadistrict’srepresentativechangedasaresultoftheModel(4)isafirst-differencesregressioninwhichthechangeinspendingfromtheyearbeforetotheyearaftertheelectionisregressedagainstchangesinlegislatorsexandothercovariates.Thefemaleeffectfromthecloseelectionssampleis7%.Notethatbecausewehaveonly39observationsinModel(4),wedonotattempttocontrolforthefullsetofdistrictcovariates.Thisisnotamajorconcern,sincewedonotexpectdistrictattributestochangesignificantlyintwoyears.WedocontrolforlegislatorcharacteristicsthatmaychangealongwiththeSuchracesareclosebycongressionalstandards,wheretheaverageelectionisdecidedbya40%marginandonly7%ofelectionsaredecidedwiththewinningcandidateearninglessthan55%ofthevote.sexofadistrict’srepresentative—party,majoritystatus,andseniority—andthecoefficientsforthesevariablesarecomparabletotheirfixedeffectscounterparts.Whileweadmittedlyhavefewinstancesofcloseelectionsthatproduceachangeinthesexofadistrict’srepresentative,andthereforedonotputmuchstockintheseresultstakeninisolation,thefindingsfromModel(4)comportwiththosefromthefixedeffectsmodelsandprovideausefulcomplementtothem.AmongtheremainingvariablesincludedinTable1,onlyahandfuldemonstratearobustrelationshipwithfederalspending.Democraticdistrictsappeartogarnermorefederalmoneyinthestatefixedeffectsmodel,buttheresultdissolveswhendistrictfixedeffectsareintro-duced.Inotherwords,itwouldappearthatDemocratscomefromdistrictsthatareotherwisepronetoreceivefederallargess,butwithin-districtchangesinlegislatorpartyarenotsignificantlyassociatedwithchangesinspendingtothedistrict.Membershipinthemajoritypartyappearstobeuncorrelatedwithdistrictspending.Tenureinoffice,whilepositiveineveryspecification,failstoattainstatisticalsignificance.Furthermore,thesizeofacongressperson’svictorymargindoesnotappeartoin-fluencetheallocationoffederalspending.Amongthedistrictattributes,anincreasingnumberofAfricanAmer-icansisassociatedwithincreasedfederalspendingovertime.Lastly,asexpected,districtsthatcontainmoreofthepopulationofthestatecapitolcountyreceivemorefederalspending.Insummary,theunambiguousresultisthatfemalelegislatorssucceedindirectingmorediscretionaryspend-ingtotheirhomedistrictsthanmalerepresentatives.Aspendingadvantageof9%amountstoapproximatelyanextra$88percapitaperyearfordistrictsrepresentedbywomen.Giventhattheaveragedistricthas563,732resi-dents,theaggregatespendingincreaseforthedistrictisroughly$49millionwhenitsendsawomantoCapitolHill.IsItSex-BasedSelection?Theresultsoftheprevioussectionprovidestrongev-idencethatcongressionaldistrictsreceivemorefederalfundingwhentheyarerepresentedbywomenthanwhentheyarerepresentedbymen.However,wehavenotes-tablishedthatthesourceofthespendingdifferenceisthatAppendixBprovidesadditionalRDanalysisanddiscussesthestrengthsandweaknessesofourdataforthisdesign.Whilecontrarytopopularwisdom,theseresultsareconsistentwithpriorstudies,suchasKnight(2005). SARAHF.ANZIAANDCHRISTOPHERR.BERRYcongresswomenaremoreablelegislators,northattherea-sonfortheirdifferentialsuccessinofficeiswhatwerefertoassex-basedselection.Theresultsthusfarleaveopenthepossibilitythatwomenarestrongerlegislatorssimplybecausetheyaremoreattunedtotheirconstituents,morededicatedtoprocuringfundsforso-calledwomen’sissues(e.g.,Swers2002),ormorecollaborativeandcooperativeintheirlegislativeandleadershipstyle(Carey,Niemi,andPowell1998;Kathlene1994;Rosenthal1998).Inordertodemonstratethatthemechanismrespon-sibleforthefemalespendingadvantageistheonewehaveproposed,wewouldliketobeabletomeasureeithervarianceinsex-basedselectionorvarianceincandidatequalityacrossdistrictsandtime.Withrespecttothelat-ter,weknowthatfemaleHousecandidatestendtobemorequalifiedthanmalecandidatesonthebasisofraw,formalqualificationssuchasprioroffice-holdingexpe-rience(PearsonandMcGhee2009).However,qualityismuchmorethanformalqualifications,anditisultimatelyonlymeasurablethroughperformance.Inbaseball,forexample,wedonotknowwhatqualitiescauseoneplayertohitmorehomerunsthananother,anditiscertainlysomethingmorethanjusttheplayer’sexperience,butwearecomfortableconcludingthattheplayerwhohitsmorehomerunsisabetterhitter.Candidatequalityissimilarinnature.Ofcourse,ifwecannotmeasurequalityasdistinctfromperformance,wecannothopetoexploitvariationincandidatequalitytoisolateitseffectonperformance.Alternatively,tomeasuresex-basedselection,wewouldwanttoquantifyeitherthedegreeofsexdiscrim-inationinthedistrict’selectorateortheextenttowhichhigher-qualitywomenself-selectintopoliticsrelativetomen.Wewouldexpecttofindapositiverelationshipbe-tweencongresswomen’sspendingadvantageandthelevelofsex-basedselectioninthedistrict,conditionalonawomanbeingelected.Unfortunately,weknowofnosuchmeasuresatthedistrictlevel,muchlessthedistrict-by-yearlevel.Instead,weuseaverageconstituentideologyinthedistrictasaproxy,albeitasomewhatcrudeone,fortheprevalenceofsex-basedselectioninthedistrict.Wealsoexaminefederalspendingoutcomesforwomenwho,webelieve,facedfewerbarrierstoentrytopoliticsbecauseoftheirsexthanotherfemalelegislators.First,wetakeadvantageofthefactthatattitudesaboutwomeninpoliticsarecorrelatedwiththeideologyofconstituentsinadistrict.WeuseClinton’s(2006)Ofthe11%ofGalluprespondentswhoreportedthattheywouldnotvoteforawell-qualifiedfemalecandidateforpresident,63%identifiedthemselvesaseitherveryconservativeorconservative.Only36%ofthosewhosaidtheyvoteforawomanwereconservativeorveryconservative(USAToday/GallupPoll,Febru-ary9–11andMarch2–4,2007).survey-basedmeasureofdistrict-levelconstituentide-ology,whichrangesfrom–1(mostliberal)to1(mostconservative).Thismeasuredoesnotcapturevariationinconstituentideologyovertimewithindistricts,butitdoesallowustoestimatetheextenttowhichthefemalespendingadvantagevariessystematicallywithatime-invariantmeasureofconstituentideology.Ifmoreconservativedistrictsalsotendtobethosewhereaveragesexdiscriminationlevelsarehigherorwherequalifiedwomenaremorereluctanttoenterpolitics,thenourthe-orywouldpredictthatthespendingadvantageachievedbyfemalelegislatorsinmoreconservativedistrictswillbegreaterthantheadvantagereceivedbythoseinliberaldistricts.Table2presentstheresultsofthemainmodelswithaninteractionbetweenlegislatorsexanddistrict-levelconstituentconservatism,thelatterofwhichiscenteredarounditsmean.Themaineffectofdistrictideologycannotbedirectlyestimatedsinceitisconstantacrosstimeperiodsandisthereforesubsumedwithinthedis-trictfixedeffects.Thetableistruncatedtopreservespace;allfourmodelsincludethecovariateswhosecoefficientsarepresentedinTable1aswellasdistrictandyearfixedeffects.Incolumn(1),thecoefficientonlegislatorsexrepresentsthespendingadvantagethataccruestodis-trictsofaverageideology(becausetheideologymeasurehasbeenmeandeviated)whentheyhavefemalerepresen-tatives.SincetheaveragedistrictisslightlyconservativeaccordingtoClinton’smeasure,weconcludethatfemalelegislatorselectedfromamoderatelyconservativecon-gressionaldistrictdeliverapproximately13%morefed-eralspendingtotheirconstituentsthanmalelegislators.Moreimportantly,however,thecoefficientontheinter-actiontermispositiveandstatisticallysignificantatthe1%level.Thus,moreconservativedistricts—whentheyelectwomentorepresentthem—receivealargerincreaseinspendingthandistrictsthathavemoreliberalcon-stituents.Themagnitudeofthecoefficientimpliesthataonestandarddeviationincreaseinaverageconstituentconservatismisassociatedwithanadditional10%boostinfederalspendingwhenawomaniselected.Wemightsuspectthatsinceconstituentideologyislikelytobepositivelycorrelatedwithlegislatorideology,theinteractionpresentedincolumn(1)picksuptheide-ologicalleaningsofthecongresswomenthemselves.Sinceweareconcernedherewithgeneralviewsaboutwomeninpoliticsinthedistrict,column(2)entersthelegislator’sWeloseasubstantialnumberofcasesduetothefactthatClin-ton’smeasureofconstituentideologyonlyexistsaccordingtothecongressionaldistrictboundariesofthe1990s.Wherepossible,weusedthesamevaluesofthismeasureforcorrespondingdistrictsinthe1980sand2000s.SeeAppendixAfordetails. WHYDOCONGRESSWOMENOUTPERFORMCONGRESSMEN?2EvidenceofSex-BasedSelection District&DistrictMemberIdeologyIdeologyWidows (1)(2)(3) Female0120Female5840Constituent(0IdeologyMemberIdeologyFemaleNonwidow0WidowsConstant188171876413Observations740474049067R-squared0890890FixedEffectsDistrict&District&State&yearyearyeartest:WidowsNonwidows Notes:ModelsincludeallcontrolvariablesreportedinTable1.Robuststandarderrorsclusteredbystateinparentheses.Thede-pendentvariableisln(federaloutlaysbycongressionaldistrictbyyear)fromhighvariationprograms,1984–2004.Outlaysarein2004dollars.Female1iflegislatorisfemale.ConstituentIdeol-ogyistheaverageconstituentideologyinthedistrictasmeasuredbyClinton(2006).MemberIdeologyisthelegislator’sDWNom-inatescore.Widows1iflegislatorisfemalewhosucceededherlatehusbandinoffice.FemaleNonwidows1forallotherfemalelegislators.significantat10%level;significantat5%level;significantat1%level.ideologyasaseparateregressor,measuredbyhisorherNOMINATEscore(PooleandRosenthal1997,2005).Theinclusionoftheindividualmembers’ideologychangesthecoefficientsonthefemaleindicatorandtheinteractiontermonlymodestly.Notably,thecoefficientontheinter-actiontermisstilllarge,positive,andsignificantatthe1%level.Theseresultsareconsistentwithourargumentthatthemechanismdrivingthespendingadvantageisatalentandeffortdifferentialinducedbysex-basedselection.Weseefromcolumns(1)and(2)ofTable2thatthepositiveeffectoffemalerepresentationonspendingisconsider-ablylargerindistrictswherepublicattitudesarelikelytobelessfriendlytotheideaofwomeninpolitics.Ofcourse,districtideologyisaroughmeasureforsex-basedselec-tion,sowebolstertheseresultsbycomparingtwogroupsofwomenforwhom,weassume,thepoliticalselectionprocessdiffers.OneroutebywhichwomenhavehistoricallyenteredCongressisbysucceedingtheirhusbandswhopassedawaywhileinoffice(Burrell1994).Ifwidowsbenefitfromoutpouringsofpublicsympathysurroundingthedeathsoftheirhusbands,theyareunlikelytobesub-jectedtothesamedegreeofelectoralscrutinyasotherwomen.Moreover,sincetheyhavecloselyfollowedtheirhusbands’tenureinoffice,widowsmaybelessinclinedtothinkthemselvesinsufficientlyqualifiedforpoliticalof-ficethanotherwomen.Inotherwords,wewouldexpectthatwidowsarefreeofmanyofthehurdlesotherwomenmustclearonthewaytooffice;therefore,theywouldnotneedthesameedgeinqualityoreffortinordertobecomecandidatesandgetelected.Infact,widowsmayevenbeabletowinwithaqualitydisadvantagerelativetomalecandidatesthankstopublicsympathy.Ifsex-basedse-lectionisthemechanismthatcauseswomentoperformbetterinofficethanmen,thenwidowsshouldhaveasmallerspendingadvantagethanotherwomen,andpos-siblyevenaspendingdisadvantagerelativetomen.Thus,wecomparespendingoutcomesfordistrictsrepresentedbywidowswhosucceededtheirhusbandsinofficewithdistrictsrepresentedbyotherwomenasawindowontofemaleperformanceinenvironmentswithandwithoutsubstantialsex-basedselection.Column(3)ofTable2presentsatestofthesepredic-tions.Wenoteupfrontthatourpowertoconductthistestislimitedbecausewehaveonlyeightwidowsinourdataset,accountingforacombined57yearsofpresenceinthelegislature.Nevertheless,theresultsconfirmourexpectations.Wecreateseparatebinaryindicatorvari-ablesforwidowsandfemalenonwidows;malesaretheomittedcategory.Themodelincludesthefullsetofleg-islatorcharacteristics,allthedistrict-leveldemographics,andstateandyearfixedeffects.ThesmallnumberofwidowsprohibitsusfromrunningdistrictfixedeffectsWecannotrejectthenullhypothesisofnospendingadvantageforwidows,whilethefemalenonwidowef-fectislarge,positive,andstatisticallysignificant.Infact,thewidowcoefficientisnegative,suggestingthatwidowsdeliverlessspendingthanmalelegislators,althoughthisdifferenceisnotsignificant.Antestallowsustore-jectthehypothesisthatthecoefficientsforwidowsandnonwidowfemalesareequalatp0.054.Thisresultlendssupporttothehypothesisthatsex-basedselectionWearegratefultoLindaFowlerforsuggestingthisidea. SARAHF.ANZIAANDCHRISTOPHERR.BERRYexplainsthefemalespendingadvantage.Ifwidowsareheldtoahigherstandardthanmalecandidatesbyvoterslikelytounderestimatetheirqualificationsforpolitics,thenweshouldexpectfemalelegislatorswhosucceedtheirlatehusbandstoperformbetterthanmalelegislators.Theresultspresentedincolumn(3)ofTable2showthatthisisthecase.Ofcourse,werecognizethattheremaybeotherreasonswhywidowsarelesseffectiveinoffice.Theseresults,whileconsistentwithourtheory,arenotdispositive.AlternativeExplanationsTheprecedingresultsshowthatcongresswomen’sspend-ingadvantagecannotbeexplainedbythedistrictstheyrepresentandisevenlargerindistrictswherewomenareelectedamidstchallengingconditions.Inthissection,weaddressthequestionofwhetherthereissomeothercor-relateofbeingfemale,apartfromthesex-basedselection,thatcanexplaincongresswomen’ssuccessingarneringfederalspendingfortheirdistricts.ItiswellknownthatelectorallyvulnerablemembersofCongressseekadditionalspendingfortheirdistricts(e.g.,Cain,Ferejohn,andFiorina1987).Isitpossiblethatwomenresponddisproportionatelytoelectoralvulnera-bilitybyseekingmorefederalspendingfortheirdistricts?WetestforthisinModel(1)ofTable3byestimatingadistrictfixedeffectsmodelthatincludesaninteractionbetweenthefemaleindicatorandthecandidate’selectoralmarginintheprecedingelection.(Toconservespace,onlythecoefficientsfortheprimaryindependentvariablesofinterestarereported,althoughthefullsetofcontrolvari-ablesisincludedinallthemodelsreportedinTable3.)Ifitistruethatwomenresponddisproportionatelytoelec-toralvulnerability,weshouldfindanegativecoefficientontheinteractionterm.Infact,however,thecoefficientispositiveandinsignificant.Wecanthereforedismissthepossibilitythatelectoralvulnerabilityisatthesourceofwomen’sspendingadvantage.Next,weinvestigatetheroleofpartisanshipandide-ology.ThewomeninCongressduringourstudyperiodaremorelikelytobeDemocrats(65%)thanthemenare(50%).Womenarealsomoreideologicallyliberal:theaverageNOMINATEscore(PooleandRosenthal1997,2005)forafemalememberofCongressis0.15,whiletheaverageforcongressmenis0.05.InModel(2)ofTa-ble3,weestimateaninteractionbetweenthefemaleandRepublicanindicators.WhilefemaleRepublicansdemon-TheNOMINATEscalerangesfrom1(mostliberal)to1(mostconservative).strateamodestedgeoverfemaleDemocrats,thediffer-enceisnotstatisticallysignificant.InModel(3),wees-timatetheinteractionbetweenthefemaleindicatorandtheNOMINATEscores.Again,thepointestimatesuggeststhatconservativewomengarnermorespendingthanlib-eralwomen,buttheinteractionisnotsignificant.Basedontheseanalyses,werejecttheideathatpartisanshiporideologycanexplainthefemalespendingdifferential.Asanextstep,weexploreoneavenuethroughwhichwomenmayattaintheiraddedspending:committeeas-signments.Observingthatwomenachievemoredesir-ablecommitteeassignmentswouldbeconsistentwithourtheoryofsex-basedselection.However,theobservationwouldalsobeequivalenttothealternativeexplanationthatpartiesdisplayfavoritismtowardwomeninthecom-mitteeassignmentprocess,perhapsbecausetherearefewfemalemembersandtheirpresenceontopcommitteesisvaluableforotherreasons,suchaspublicrelations.WeusetheGrosecloseandStewart(1998)Housecommitteedesirabilityscorestoplacevaluesonthecommitteeport-foliosofindividuallegislatorsineachyear.Wefindthat,controllingforseniority,womenhaveslightlylessdesir-ablecommitteeportfolios,althoughthedifferenceisnotstatisticallysignificant(notshown).Inanycase,whenwecontrolforacompletesetofcommitteeindicatorvari-ablesaswellasindicatorvariablesforcommitteechairs,committeerankingminoritymembers,andpartyleaders,whichwedoinModel(4)ofTable3,theestimatedfemalespendingadvantageisessentiallyunaffected.Womendonotattaintheirspendingadvantagemerelybysecuringbettercommitteeassignments.Inanalysisnotshown,wealsoinvestigatethenatureofthespendingthatwomenbringhometotheirdis-tricts.Womeninpoliticsscholarshavefoundthatfe-malepoliticiansaremoreactiveinareasconsideredtobe“women’sissues”(e.g.,Swers2002).Ifwomenderivetheiradvantageinspendingprimarilyfromfederalprogramsthatreflecttraditional“female”legislativepriorities,wemightbedissuadedthatitiswomen’stalentandeffortthatdrivethespendingeffect.Tothecontrary,wefindthatthefemalespendingad-vantageispresentacrossadiversesetoffederalprograms.Weestimatefixedeffectsmodelsofspendingfromeachofthefouragenciesresponsibleforthegreatestamountofhighvariationprogramspendingfrom1984to2004:theTheseresultsarepresentedinAppendixB.Thenullfindingsforcommitteechairs,rankingminoritymem-bers,andpartyleadersarecontrarytoexpectations.However,wenotethatthesepositionschangerelativelyinfrequently,andsotheestimatesareveryimpreciseinthecontextofdistrictfixedeffects.ThefollowingresultsaredetailedinAppendixB. WHYDOCONGRESSWOMENOUTPERFORMCONGRESSMEN?3AlternativeExplanations High-VariationProgramBillsBillsSpendingSponsoredCosponsored (1)(2)(3)(4)(5)(6) Female008500550093009328425057)(0041)(0Republican0100040022)(0022)(0023)(074)(7Terms00200011003)(0003)(0003)(0004)(0Margin00020003000500052025)(0023)(0022)(0023)(0FemaleMargin0FemaleRepublican0IdeologyFemaleIdeology0CommitteeChair02110047)(2RankingMinority02605351031)(203)(18HouseLeader0991224)(555)(35MajorityParty9523DistancefromMedian37410903)(25Constant18831891187518662151511)(345Observations906790679067906747524752CommitteeIndicatorsIncluded?NoNoNoYesYesYesR-squared0260 Notes:ThedependentvariableforModels(1)–(4)isln(federaloutlaysbycongressionaldistrictbyyear)fromhighvariationprograms,1984–2004.Outlaysarein2004dollars.ThedependentvariableforModel(5)isthenumberofbillssponsoredpercongress,andthedependentvariableforModel(6)isthenumberofbillscosponsoredpercongress.Models(1)–(4)includeallcontrolvariablesreportedinTable1,districtfixedeffects,andyearfixedeffects.Model(4)includesindicatorvariablesforeachcommittee.Models(5)–(6)includealldistrictdemographiccontrolsreportedincolumn(1)ofTable1,congressfixedeffects,anddummyvariablesformembershiponeachstandingcommittee.ForModels(1)–(4),standarderrorsareclusteredbystate.ForModels(5)–(6),standarderrorsareclusteredbymemberofCongress.Ideologyisthelegislator’sDWNominatescore.significantat10%level;significantat5%level;significantat1%level.DepartmentsofAgriculture,HealthandHumanServices,Transportation,andEducation.OnlyAgriculturefailstodemonstrateaspendingadvantageforwomen.Womenhaveaclearadvantageinsecuringfundsfortheirdis-trictsfromTransportation,HealthandHumanServices,andEducation.Infact,thecoefficientforTransportationislargestinmagnitude,aresultthatisparticularlysug-gestivesincetransportationisanareaidentifiedbycon-gressionalscholarsasespeciallyamenabletoporkbarrelpolitics(e.g.,Ferejohn1974).Itisthereforenotthecasethatwomenonlyhaveanadvantageinsecuringspendingforprogramsrelatedto“women’sissues.” SARAHF.ANZIAANDCHRISTOPHERR.BERRYLastly,weinvestigatewhethercongresswomen’ssuc-cessinsecuringfederalfundingfortheirconstituentscomesattheexpenseofattentiontoanotherimportantaspectoftheirjob—policymaking.Asatestofwhethercongresswomenarelesseffectivepolicymakersthancon-gressmen,weanalyzebillsponsorshipandcosponsorshippatternsformaleandfemalemembersoftheHouse.WeusedatamadeavailablebyFowler(2006),whocompiledsponsorshipinformationforeverypieceoflegislationproposedinCongresssince1973.Weexaminethenumberofbillssponsoredandcosponsoredbywomenrelativetomenfrom1984to2004,modelingeachasafunctionoflegislatorsex,party,andmajoritystatusandincludingindicatorvariablesforcommitteechairs,rankingminoritymembersofcom-mittees,andpartyleaders.Toaccountforthepossibilitythatelectorallyvulnerablemembersarespurredintoac-tion,wealsocontrolforthemembers’votemarginintheprecedingelection.Inaddition,becausesomecom-mitteesprovidemoreopportunitiesforlegislativeactivitythanothers,weincludeafullsetofcommitteemember-shipindicatorvariables.WealsoincludethedemographicvariablesfromTable1tocontrolforthepossibilityamem-bermightsponsormorebillsifsheisfromadistrictwhereconstituentsareparticularlyattentivetolegislativebehav-ior.Weexpectthatitiseasierforideologicallymoderatememberstoworkwithlargernumbersoftheircolleaguesthanmoreextrememembers,soweincludeeachlegis-lator’sideologicaldistancefromthelegislatorwiththemedianNOMINATEscoreineachcongress.Finally,weincludecongressionaltermfixedeffectstoaccountforgeneralchangesinsponsorshipandcosponsorshipovertime.Toconservespace,wedonotreportthecoefficientsforthedemographicvariables,committeevariables,orfixedeffects.Clearly,itisthecasethatwomenneglecttheirrolesaspolicymakers.Infact,Model(5)ofTable3demon-stratesthatcongresswomensponsormorelegislationthancongressmen.Onaverage,womensponsoraboutthreemorebillspercongress,whichisadifferenceofroughly17%relativetothememberaverageof18bills.Womenarealsomoreactiveinsupportingthelegislationoftheircol-leaguesthroughcosponsorship.Congresswomencospon-sorabout26morebillspercongressthancongressmen,asseeninModel(6).Inresultspresentedintheappendix,wefindthatwomenalsogarnercosponsorshipsupportfromagreaternumberoftheirpeers,whichsuggeststhatSeeAppendixAfordetails.Thereare4,752observationsinModels(5)and(6),ratherthanthefull4,785,duetomissingvaluesforTerms(24missing)andMargin(9missing).SeeAppendixAfordetails.womenhavestrongernetworksofcollaborationwiththeircolleaguesthancongressmen.Thereareobviouslimitationstocountingbillsasameasureoflegislators’attentivenesstopolicymaking.Inparticular,thedecisiontocosponsorabillisrela-tivelycostless.Inmorecomprehensiveexaminationsofcongressionalpolicymaking,however,VoldenandWise-man(2009)andVolden,Wiseman,andWittman(2010)trackeachbillintroducedinthe97to110con-gressesthroughallstagesofthelegislativeprocess—fromintroductiontosigning—andfindthatwomenscoresig-nificantlyhigherontheirmeasureof“legislativeeffec-tiveness”thanmendo.Inshort,women’sbillsmakeitfurtherinthelegislativeprocessandaremorelikelytobeconsidered“important,”asmeasuredbymediacover-age.Notonlydoesthisevidencerefutetheargumentthatwomenpaycloseattentiontodistrict-levelspendingattheexpenseofpolicymaking,butitisalsoconsistentwiththeideathatpolicymakingisyetareainwhichcongresswomenoutperformcongressmen.Ifwebelievetheevidencethattheaveragewomanunder-estimatesherqualificationsrelativetotheaverageman,thenitisreasonabletoconcludethatawomanwhoiden-tifiesherselfasacandidatefornationalofficeismorequalifiedthantheaveragemalecandidate.Ifittakesmoretalentandgreatereffortforfemalecandidatestobetakenseriouslybyvoters,campaigncontributors,andpartygatekeepers,thenthewomenwhosucceedintheelectoralprocessarelikelytobemoretalentedandhardworkingthanthemenwhodothesame.Becauseofthis,thewomenwhoareelectedtoCongressareactuallypoisedtobemoreeffectivelegislatorsthantheirmalecounterparts.Ourtheoryofsex-basedselectionmakespreciselythispoint.Itdoesnotmatterwhetherwomenareelectedtopublicofficeatlowerratesthanmenbecausetheyperceivetheirownqualificationsdifferentlyorbecausebiasagainstwomenintheelectorateproducesabarriertoentryforthem.Thecentralimplicationofsex-basedpoliticalselectionisthatthewomenweobserveinofficewill,onaverage,outperformthemen.Wetestthisimplicationusinglegislators’successindirectingfundstotheirhomedistrictsasourprimarymeasureofperformance.Thefederalspendinganalysisprovidesstrongempiricalsupportforthepredictionthatwomenoutperformmen.Allelseequal,congressionaldistrictsreceiveroughly9%morehighvariationfederalprogramspendingwhentheyarerepresentedbywomen.Thisspendingbonusamountstoapproximately$88per WHYDOCONGRESSWOMENOUTPERFORMCONGRESSMEN?capita,or$49millionintotal,fordistrictsthathaveawomaninWashingtoninagivenyear.AccordingtotheestimatescontainedinLevittandSnyder(1997),theaddi-tionof$88percapitainhighvariationprogramspendingproducesanelectoralrewardfortheincumbentofalmost2%ofthepopularvote.However,ourresultsarenotinvulnerabletocriticism.Withoutadirectwaytomeasurelegislatorabilityoreffort,wecannotdefinitivelyshowthatthesefactorsexplainfe-malesuccessinoffice.Inanearliersection,weconsideredasetofcompetingexplanationsforthespendingdiffer-entialandbroughteachonetothedata.Theresultsallowustorejectthepossibilitythatwomen’selectoralvulnera-bility,differingideologyorpartisanship,oradvantageouscommitteeassignmentscanexplaintheconnectionbe-tweenlegislatorsexandspending.Moreover,itisnotthecasethatfemaleHousemembersmanagetoexcelinse-curingfederalspendingfortheirdistrictsbyneglectingpolicymaking:theyactuallysponsorandcosponsormorebillspercongressthantheirmalecounterparts.Whilesuchevidencecannotsubstituteforadirecttestoftherelationshipbetweenlegislatorsexandabilityoreffort,itdispelsseveralreasonablecompetingexpla-nations.Forexample,onemightconjecturethatpoliti-calpartyleadersintentionallychanneldisproportionatefundingtowomen’sdistricts,eithertoprotecttheirrel-ativelysmallcadresoffemalerepresentatives,orsimplytomakeitobviousthattheydonotdiscriminateagainstthem.Alternatively,perhapsfemalelegislatorsfeeltheneedtoworkharderinordertoprovethemselvestotheircolleaguesinthemale-dominatedHouse.Whiletheseareallplausibleexplanationsforafemalespendingadvantageingeneral,theycannotaccountforwhyitdoesnotapplytowomenwhosucceedtheirlatehusbandsinofficeorwhyitisgreaterindistrictswhereconstituentsaremoreconservative.Anyalternativeexplanationforourfindingswouldhavetoaccountforallofthesepatterns,aswellasthefactthatwomensponsorandcosponsormorebillsthantheirmalecounterparts.Webelievethatourthe-oryofsex-basedselectionprovidesthemostlogicalandparsimoniousexplanationforthesefindings.Inclosing,wenotethatourtheoreticalcontribu-tiondoesnotapplyuniquelytowomenortothemea-suresofperformancethatwehavechosen.Futureresearchmightlookforotherareasinwhichfemalesexcelinoffice.Inaddition,ourtheorysuggeststhatmembersofothergroupsthatsufferfromdiscriminationbytheelectoratealsomustperformbetterinordertobeelected.Futurere-searchmightapplyasimilaranalysistoAfricanAmericansorLatinosinCongress.However,weanticipatethattheuseofrace-consciousdistricting,inparticularmajority-minoritydistricts,willseriouslyconfoundtestingofthetheory.Ifracialdistrictingmakesiteasierforminori-tiestobeelected,thenthereisnoreasontoexpectthatthoseinofficewillperformanybetterthanaverage.Ofcourse,politicalselectionisnotbasedsolelyoncandi-dates’personalattributes.Wemightexpect,forexample,aRepublicanelectedfromahistoricallyDemocraticdis-tricttodemonstrateasimilarqualityadvantage.Theseareempiricalquestionsthatwemayexploreinfutureresearch.Atthemostgenerallevel,ourresultshighlighttheimportanceofconnectingresearchonwomeninpolitics,modelsofpoliticalagency,andtheeconomicsofdiscrim-ination.WomenaresomeofthemosteffectivepoliticiansinCongress.Oneonlyhastolooktothepoliticalselectionprocesstounderstandwhy.ReferencesAaron,Henry.1999.“JackieRobinson.”InTIMEMagazine’sTIME100:The100MostInfluentialPeopleoftheCenturyJune14.Adler,E.Scott.2003.“CongressionalDistrictDataFile,98Congresses.”UniversityofColorado,Boulder,CO.Alvarez,R.Michael,andJasonL.Saving.1997.“Deficits,Democrats,andDistributiveBenefits:CongressionalElec-tionsandthePorkBarrelinthe1980s.”PoliticalResearchQuarterly50(4):809–31.Ashworth,Scott.2005.“ReputationalDynamicsandPoliticalCareers.”JournalofLaw,EconomicsandOrganizationAusten-Smith,David,andJeffreyS.Banks.1989.“ElectoralAc-countabilityandIncumbency.”InModelsofStrategicChoiceinPolitics,ed.P.Ordeshook.AnnArbor:UniversityofMichi-ganPress,121–48.Banks,JeffreyS.,andRangarajanK.Sundaram.1998.“OptimalRetentioninAgencyProblems.”JournalofEconomicTheory82(2):293–323.Barro,RobertJ.1973.“TheControlofPoliticians:AnEconomicModel.”PublicChoice14(1):19–42.Becker,GaryS.1957.TheEconomicsofDiscrimination.Chicago:UniversityofChicagoPress.Besley,Timothy.2006.PrincipledAgents?ThePoliticalEconomyofGoodGovernment.Oxford:OxfordUniversityPress.Besley,Timothy,andAnneCase.2003.“PoliticalInstitutionsandPolicyChoices:EvidencefromtheUnitedStates.”Jour-nalofEconomicLiterature41(1):7–73.Bickers,KennethN.,andRobertM.Stein.1991.FederalDo-mesticOutlays,1983–1990:ADataBook.Armonk,NY:M.E.Sharpe.Bickers,KennethN.,andRobertM.Stein1996.“TheElectoralDynamicsoftheFederalPorkBarrel.”AmericanJournalofPoliticalScience40(4):1300–1326.Burrell,BarbaraC.1994.AWoman’sPlaceIsintheHouse:CampaigningforCongressintheFeministEra.AnnArbor:UniversityofMichiganPress. SARAHF.ANZIAANDCHRISTOPHERR.BERRYCain,Bruce,JohnFerejohn,andMorrisFiorina.1987.ThePer-sonalVote.Cambridge,MA:HarvardUniversityPress.Carey,JohnM.,RichardG.Niemi,andLyndaW.Powell.1998.“AreWomenStateLegislatorsDifferent?”InWomenandElectiveOffice,ed.SueThomasandClydeWilcox.NewYork:OxfordUniversityPress,87–102.CenterforAmericanWomenandPolitics(CAWP).2009.“FactsonWomeninStateLegislatures.”NewBrunswick,NJ:Cen-terforAmericanWomenandPolitics.Chattopadhyay,Raghabendra,andEstherDuflo.2004.“WomenasPolicyMakers:EvidencefromaRandomizedPolicyExperimentinIndia.”Econometrica72(5):1409–43.Clinton,Joshua,andJonLipinski.2006.“MeasuringLegislativeAccomplishment,1877–1994.”AmericanJournalofPoliticalScience50(1):232–49.Dolan,Kathleen.1997.“GenderDifferencesinSupportforWomenCandidates:IsThereaGlassCeilinginAmericanPolitics?”WomenandPolitics17:27–41.Dolan,KathleenA.2004.VotingforWomen:HowthePublicEvaluatesWomenCandidates.Boulder,CO:WestviewPress.Duerst-Lahti,Georgia.1998.“TheBottleneck:WomenBecom-ingCandidates.”InWomenandElectiveOffice,ed.SueThomasandClydeWilcox.NewYork:OxfordUniversityPress,15–25.Evans,Diana.2004.GreasingtheWheels:UsingPorkBarrelProjectstoBuildMajorityCoalitionsinCongress.NewYork:CambridgeUniversityPress.Fearon,JamesD.1999.“ElectoralAccountabilityandtheCon-trolofPoliticians:SelectingGoodTypesversusSanctioningPoorPerformance.”InDemocracy,Accountability,andRep-resentation,ed.A.Przeworski,S.C.Stokes,andB.Manin.Cambridge:CambridgeUniversityPress,55–97.Fenno,RichardF.,Jr.1966.ThePowerofthePurse:Appropri-ationsPoliticsinCongress.Boston:Little,BrownandCom-pany.Fenno,RichardF.,Jr.1978.HomeStyle:HouseMembersinTheirDistricts.Boston:Little,BrownandCompany.Ferejohn,JohnA.1974.PorkBarrelPolitics.Stanford,CA:Stan-fordUniversityPress.Ferejohn,John.1986.“IncumbentPerformanceandElectoralControl.”PublicChoice50(1–3):5–25.Fiorina,MorrisP.1981.“SomeProblemsinStudyingtheEf-fectsofResourceAllocationinCongressionalElections.”AmericanJournalofPoliticalScience25:543–68.Fowler,LindaL.,andRobertMcClure.1989.PoliticalAmbitionNewHaven,CT:YaleUniversityPress.Fowler,JamesH.2006.“ConnectingtheCongress:AStudyofCosponsorshipNetworks.”PoliticalAnalysis14(4):456–87.Fox,RichardL.2006.“CongressionalElections:WhereAreWeontheRoadtoGenderParity?”InGenderandElections,ed.SusanJ.CarrollandRichardL.Fox.NewYork:CambridgeUniversityPress,97–116.Fox,RichardL.,andJenniferL.Lawless.2004.“EnteringtheArena?GenderandtheDecisiontoRunforOffice.”Ameri-canJournalofPoliticalScience48(2):264–80.Fox,RichardL.,andEricR.A.N.Smith.1998.“TheRoleofCan-didateSexinVoterDecision-Making.”PoliticalPsychology19(2):405–19.Gordon,Sanford,GregoryHuber,andDimitriLanda.2007.“ChallengerEntryandVoterLearning.”AmericanPoliticalScienceReview101(2):303–20.Groseclose,Timothy,andCharlesStewartIII.1998.“TheValueofCommitteeSeatsintheHouse:1947–91.”AmericanJour-nalofPoliticalScience42(2):453–74.Howell,William,ScottAdler,CharlesCameron,andCharlesRiemann.2000.“DividedGovernmentandtheLegislativeProductivityofCongress1945–94.”LegislativeStudiesQuar-terly25:285–312.Huddy,Leonie,andNaydaTerkildsen.1993a.“GenderStereo-typesandthePerceptionofMaleandFemaleCandidates.”AmericanJournalofPoliticalScience37:119–47.Huddy,Leonie,andNaydaTerkildsen.1993b.“TheConse-quencesofGenderStereotypesforWomenCandidatesatDifferentLevelsandTypesofOffice.”PoliticalResearchQuarterly46(3):503–25.Jenkins,Shannon.2007.“AWoman’sWorkIsNeverDone?Fund-RaisingPerceptionandEffortamongFemaleStateLegislativeCandidates.”PoliticalResearchQuarterlyJeydel,Alana,andAndrewJ.Taylor.2003.“AreWomenLegis-latorsLessEffective?EvidencefromtheU.S.HouseintheCongress.”PoliticalResearchQuarterlyKahn,LawrenceM.1991.“DiscriminationinProfessionalSports:ASurveyoftheLiterature.”IndustrialandLaborRelationsReview44(3):395–418.Kahn,LawrenceM.,andPeterD.Sherer.1988.“RacialDif-ferencesinProfessionalBasketballPlayers’Compensation.”JournalofLaborEconomics6(1):40–61.Kathlene,Lyn.1994.“PowerandInfluenceinStateLegisla-tivePolicymaking:TheInteractionofGenderandPositioninCommitteeHearingDebates.”AmericanPoliticalScienceReview88(3):560–76.Knight,Brian.2005.“EstimatingtheValueofProposalPower.”AmericanEconomicReview95(5):1639–52.Lawless,JenniferL.,andRichardL.Fox.2005.ItTakesaCan-didate:WhyWomenDon’tRunforOffice.NewYork:Cam-bridgeUniversityPress.Lawless,JenniferL.,andKathrynPearson.2008.“ThePrimaryReasonforWomen’sUnderrepresentation?ReevaluatingtheConventionalWisdom.”JournalofPolitics70(1):67–82.Lee,DavidS.2008.“RandomizedExperimentsfromNon-RandomSelectioninU.S.HouseElections.”JournalofEconometrics142(2):675–97.Lee,FrancisE.1998.“RepresentationandPublicPolicy:TheConsequencesofSenateApportionmentfortheGeographicDistributionofFederalFunds.”JournalofPoliticsLevitt,StevenD.,andJamesM.Snyder,Jr.1995.“PoliticalPar-tiesandtheDistributionofFederalOutlays.”AmericanJour-nalofPoliticalScience39(4):958–80.Levitt,StevenD.,andJamesM.Snyder,Jr.1997.“TheImpactofFederalSpendingonHouseElectionOutcomes.”JournalofPoliticalEconomy105(1):30–53.Mayhew,DavidR.1974.Congress:TheElectoralConnectionNewHaven,CT:YaleUniversityPress. WHYDOCONGRESSWOMENOUTPERFORMCONGRESSMEN?Milyo,Jeffrey,andSamanthaSchosberg.2000.“GenderBiasandSelectionBiasinHouseElections.”PublicChoiceNationalWomen’sPoliticalCaucus(NWPC).1994.WhyDon’tMoreWomenRun?AstudypreparedbyMellman,Lazu-rus,andLake.Washington,DC:NationalWomen’sPoliticalNewman,Jody.1994.PerceptionandReality:AStudyComparingtheSuccessofMenandWomenCandidates.Washington,DC:NationalWomen’sPoliticalCaucus.Newport,Frank,andJosephCarroll.2007.Analysis:ImpactofPersonalCharacteristicsonCandidateSupport.Princeton,NJ:GallupNewsService.Niederle,Muriel,andLiseVesterlund.2007.“DoWomenShyAwayfromCompetition?DoMenCompeteTooMuch?”QuarterlyJournalofEconomics122(3):1067–1101.Norton,NoelleH.1999.“UncoveringtheDimensionalityofGenderVotinginCongress.”LegislativeStudiesQuarterly24(1):65–86.Palmer,Barbara,andDennisSimon.2006.BreakingthePoliticalGlassCeiling:WomenandCongressionalElections.NewYork:Routledge,TaylorandFrancisGroup.Pascal,AnthonyH.,andLeonardA.Rapping.1972.“TheEconomicsofRacialDiscriminationinOrganizedBase-ball.”InRacialDiscriminationinEconomicLife,ed.An-thonyH.Pascal.Lexington,MA:LexingtonBooks,119–Pearson,Kathryn,andEricMcGhee.2009.“WhyWomenShouldWinMoreOftenThanMen:ReassessingGenderBiasinU.S.HouseElections.”Unpublishedmanuscript,Univer-sityofMinnesota.Persson,Torsten,GerardRoland,andGuidoTabellini.1997.“SeparationofPowersandPoliticalAccountability.”Quar-terlyJournalofEconomics112(4):1163–1202.PewResearchCenter.2008.“MenorWomen?Who’stheBetterLeader?”SocialandDemographicTrends,August25.Poole,KeithT.,andHowardRosenthal.1997.Congress:APolitical-EconomicHistoryofRollCallVoting.NewYork:OxfordUniversityPress.Poole,KeithT.,andHowardRosenthal.2005.DW-NOMINATEData:98Congressto108CongressRehavi,M.Marit.2007.“SexandPolitics:DoFemaleLegislatorsAffectStateSpending?”Workingpaper,UniversityofBritishColumbia.Rodgers,William,ed.2006.HandbookontheEconomicsofDis-crimination.Northampton,MA:EdwardElgar.Rosenthal,CindySimon.1998.WhenWomenLead.NewYork:OxfordUniversityPress.Rosenwasser,ShirleyM.,andNormaG.Dean.1989.“Gen-derRoleandPoliticalOffice:EffectsofPerceivedMasculin-ity/FemininityofCandidateandPoliticalOffice.”PsychologyofWomenQuarterly13:77–85.Sanbonmatsu,Kira.2006.WhereWomenRun:Gender&PartyintheAmericanStates.AnnArbor:UniversityofMichiganPress.Seabright,Paul.1996.“AccountabilityandDecentralizationinGovernment:AnIncompleteContractsModel.”EuropeanEconomicReview40(1):61–89.Sears,DavidO.,andP.J.Henry.2005.“OverThirtyYearsLater:AContemporaryLookatSymbolicRacism.”AdvancesinExperimentalSocialPsychologyVol.37,ed.M.P.Zanna.NewYork:AcademicPress,95–149.Sellers,PatrickJ.1997.“FiscalConsistencyandFederalDistrictSpendinginCongressionalElections.”AmericanJournalofPoliticalScience41(3):1024–41.Seltzer,RichardA.,JodyNewman,andM.VoorheesLeighton.SexasaPoliticalVariable.Boulder,CO:LynneReinner.Shepsle,KennethA.1978.TheGiantJigsawPuzzle:DemocraticCommitteeAssignmentsintheModernHouse.Chicago:Uni-versityofChicagoPress.Smith,EricR.A.N.,andRichardL.Fox.2001.“TheElectoralFortunesofWomenCandidatesforCongress.”PoliticalRe-searchQuarterly54(1):205–21.Stein,RobertM.,andKennethN.Bickers.1995.PerpetuatingthePorkBarrel:PolicySubsystemsandAmericanDemocracyNewYork:CambridgeUniversityPress.Swers,MicheleL.2002.TheDifferenceWomenMake.Chicago:UniversityofChicagoPress.Thistlethwaite,D.,andD.Campbell.1960.“Regression-DiscontinuityAnalysis:AnAlternativetotheExPostFactoExperiment.”JournalofEducationalPsychology51:309–17.Thomas,Sue.1991.“TheImpactofWomenonStateLegislativePolicies.”JournalofPolitics53(4):958–76.Uhlaner,CaroleJean,andKayLehmanSchlozman.1986.“Can-didateGenderandCongressionalCampaignReceipts.”Jour-nalofPolitics48:30–50.Volden,Craig,andAlanE.Wiseman.2009.“LegislativeEf-fectivenessinCongress.”Workingpaper,TheOhioStateUniversity.Volden,Craig,AlanE.Wiseman,andDanaE.Wittmer.2010.“WhyAreWomenMoreEffectiveLawmakersinCongress?”Workingpaper,TheOhioStateUniversity.Zaller,John.1998.“PoliticiansasPrizeFighters:ElectoralSe-lectionandtheIncumbencyAdvantage.”InPoliticiansandPartyPolitics,ed.JohnG.Geer.Baltimore:JohnsHopkinsUniversityPress,125–85.SupportingInformationAdditionalSupportingInformationmaybefoundintheonlineversionofthisarticle:LegislatorSexandFederalSpendingTypePredictingLegislatorSexwithDistrictDemographicsAlternativeSpecificationsforTable1ResultsRegressionDiscontinuitySourcesoftheFemaleSpendingAdvantageLegislatorSexandPolicymakingPleasenote:Wiley-Blackwellisnotresponsibleforthecontentorfunctionalityofanysupportingmaterialssup-pliedbytheauthors.Anyqueries(otherthanmissingmaterial)shouldbedirectedtothecorrespondingauthorforthearticle.