JournalofPublicEconomics73(1999)3115 - PDF document

JournalofPublicEconomics73(1999)3115±20%ofGNPinsomecountries.Inseveralcountries,liketheNordicones,educationandhealthcarearealmostexclusivelypubliclyprovided,whereasinothercountries,liketheU.S.,thereisamixofprivateandpublicprovision.PublicprovisionofdaycareforpreschoolchildrenisalsoanimportantpartofpublicexpenditureintheNordiccountries.ThereisanongoingpublicdebatebothinNorthAmericaandEuropetowhatextentthepublicsectorshouldprovideprivategoods.ParticularlyintheU.S.thereisalsoadebateonwhetherindividualsshouldbeallowedtosupplementand/orupgradepubliclyprovidedgoodswithprivatepurchases.Especiallywithrespecttoeducationvouchersystemshavebeendiscussed(EppleandRomano(1998)).Standardtextbooksinpubliceconomicsprovidenoexplanationforthistypeofpublicexpenditure.However,thereisanemerging,fastgrowing,literatureattemptingtoexplaintheroleofpublicprovisionofprivategoods.Thisliteratureconsistsoftwoquitedisparatepartsprovidingverydifferentviewsonwhatthefunctionofpubliclyprovidedgoodsisintheeconomy.Onestrandofliteraturewithnormativetheoriesshowshowpublicprovisionofprivategoodsundercertainconditionscanbebene®cialandimproveef®ciency.Thesetheoriesdistinguishbetweentwobasicfunctionsofpublicprovisionofaprivategood.Oneistomitigateaninformationalconstraintthatrestrictstheredistributionthatcanbeaccomplishedbyanoptimalnonlinearincometax.Anotherfunctionistomitigateatimeinconsistencyproblem.Inanotherstrandofliteratureconsistingofvotingmodelsthereisnoef®ciencyenhancingroleforpubliclyprovidedgoods.Inthevotingmodelspubliclyprovidedprivategoodsisameansforsomegroupsintheeconomytoredistributeresourcestothemselves.Oneconclusionofanin¯uentialarticle,EppleandRomano(1996a),isthatpublicprovisionofhealthcareleadstoinef®ciency.Thetwostrandsofliteraturemodeltheeconomyinquitedifferentways.Thevotingmodelsusuallyimposehighlyrestrictiveassumptionsontheavailabletaxandtransferinstruments.Normativetheoriesontheotherhandfrequentlyusemodelswhereanonlinearincometaxisavailableforredistribution.Ourpurposeistomakeasynthesisofthetwostrandsofliterature.Twostepsareessentialinourapproach.First,weconstructapoliticaleconomyframeworkwhichestablishesacaseforachoiceofef®cientpolicy.Second,weconstructapositivetheoryofpublicprovisionofprivategoodswhereweallowthetaxinstrumentstobeas¯exibleaspossible,andwhereasanessentialpartofthemodelweincludeoneofthemechanismsforimprovedef®ciencythataredescribedinthenormativeliterature.Itisshownthattherewillbeavotingequilibriumwithpubliclyprovidedprivategoodsinmodelswherepubliclyprovidedprivategoodsdonotserveasasubstitutefortaxinstruments,andtheequilibriumwillbeParetoEppleandRomanodenotethepubliclyprovidedprivategoodashealthcare.However,thegooddoesnotpossessanyinherentpropertiessuggestingthatitisinthecapacityofhealthcarethatitshouldbepubliclyprovided. JournalofPublicEconomics73(1999)31MeltzerandRichard(1985)isanearlycontributiontotheliteratureonpubliclyprovidedprivategoods.Onecanseethispaperasaforerunnerbothforlaternormativemodelsandpositivevotingmodels.Theirarticlestudiesavotingprocess.However,incontrasttolaterarticlesonvotingmodels,publicprovisionofaprivategoodisintheirmodelnotdrivenbysevererestrictionsonthetaxinstrumentsbutbyamechanismwherebythepubliclyprovidedgoodaffectsthesizeofthetaxbase.MorerecentarticleslikeEppleandRomano(1996a),GradsteinandJustman(1996)andGouveia(1997)areexamplesofvotingmodelswheretheresultthatthereispublicprovisionofaprivategoodisdrivenbysevererestrictionsonthetaxinstruments.Forexample,inEppleandRomano(1996a),wherethepubliclyprovidedprivategoodis®nancedbyaproportionalincometax,raisondetreforpublicprovisionisthatitservesasasubstituteforademogrant.Ifthemodelallowedalineartax,therewouldnotexistavoting4,5equilibriumwithpublicprovision.Thenormativemodelsfocusontheef®ciencyenhancingroleofpubliclyprovidedprivategoods.NicholsandZeckhauser(1982)isafascinatingpaperpointingtotheroleofin-kindtransfersinimprovingtheef®ciencyofredistribu-tionschemes.Publicprovisionschemesimplythatsomeindividualswillfacequantityconstraints.GuesnerieandRoberts(1984)showthatiftheeconomyinitiallyisatasecond-bestpositionitisingeneralpossibletoachieveaParetoimprovementbyintroducingasmallquantityconstraint.BesleyandCoate(1991)maketheobservationthatdifferentgroupsofpeoplecanvaluepubliclyprovidedgoodsdifferently.Usingamodelwithaheadtaxto®nancethepubliclyprovidedgoodtheyshowhowpublicprovisioncaninduceself-selectionandachieveredistributionwithloweref®ciencycoststhanifcashtransferswereused.BoadwayandMarchand(1995)andBlomquistandChristiansen(1995)showthatevenifanoptimalnonlinearincometaxisavailableforredistributionthereisaroleforpubliclyprovidedprivategoods.TheproblemtheystudyishowtoBergstromandBlomquist(1996)studyasimilarmechanism.Intheirmodelthelaborforceparticipationofmothersincreasesifdaycareispubliclyprovided.Undercertainconditionsthiscanleadtosuchalargeincreaseinthetaxbasethattheincometaxcanbeloweredasthelevelofpubliclyprovideddaycareincreases.Thismeansthatallindividualswouldvoteforsomepublicprovisionofdaycare.LundholmandOhlsson(1998)alsostudyataxbaseeffectmechanism.However,intheirmodelwagesareendogenousandpartofthebene®tsofpubliclyprovideddaycareisduetowageHencetheschemeisbasicallythesameasthatofRoberts(1977)whichprovidesapoliticaleconomyexplanationforthechoiceofincometaxrateandalump-sumtransferincash.Thisisalsopointedoutinfootnote15inEppleandRomano(1996a).Gouveia(1996)andEppleandRomano(1996b)areotherarticleswithvotingmodels.ThefocusofBesleyandCoate(1997)isnotonpublicprovisionofprivategoods.However,oneoftheirexamplesprovidesapoliticaleconomymodelofpublicprovisionofprivategoods.BlackorbyandDonaldson(1988)provideasimilarargument.BruceandWaldman(1991),Coate(1995),CremerandGahvari(1997)andBlomquistandChristiansen(1998a)areotherexamplesofnormativetheoriesofpubliclyprovidedprivategoods. JournalofPublicEconomics73(1999)31achieveredistributionfromhigh-skilledtolow-skilledindividualswhentheidentityofthetwotypesofindividualsisnotobservedbythegovernment.Becauseofthisasymmetricinformationtheredistributionishamperedbyaself-selectionconstraint.Thepotentialroleofpublicprovisionistoalleviatetheself-selectionconstraintbygivingpartofthetransfertothelow-skilledpersonsintheformofthepubliclyprovidedgood.Thisisthemechanismwewillbuildintoourvotingmodels.InSection2weelaborateonthemechanismbywhichpubliclyprovidedprivategoodscanimprovetheef®ciencyoftheeconomy.Eventhoughthepublicprovisionofprivategoodsprovidesthemotivationforthepresentstudyourpoliticaleconomymodelswillalsoyieldresultsthatareofinterestbeyondthisparticularissue.Thecentralpoliticaleconomyconsiderationsarevalidmoregenerallyandinmodelswithotherpolicyinstrumentsaffectingthedistributionofwelfare.InSection3weintroduceamedianvotermodelandcharacterizethevotingequilibriumingeneralandintheparticularcaseinwhichweembedthepublicprovisionschemeofSection2.SincepolicymakingrarelyisdecidedthroughsimplemajorityvotingwestudyinSection4themorerealisticcaseofrepresentativedemocracywithtwopoliticalpartiescompetingforvotes.Section52.PubliclyprovidedprivategoodsasaskillscreeningdeviceAcentralpurposeofourstudyistodiscussfromapoliticaleconomyperspectivethecaseforpublicprovisionofprivategoodsthathasbeenestablishedinthenormativetheory.Asabackgroundfortherestofthepaperandforsubsequentreferenceitishelpfultosummarizethemainaspectsofthemodelwhichhasbeenthevehicleofmuchofthenormativeanalysis.Themotivationforfocusingonthisparticularmodelisnotonlythatithasbeencentralintheliterature,butmoreimportantlyithasanumberoffeaturesthatenablesustoestablishamorefundamentalcaseforpublicprovisionthandoesanumberofalternativemodels.Inparticularthecaserestsonbasicinformationasymmetriesratherthanonde®cienciesofthetaxdesignoradhocrestrictionsontheavailabletaxinstruments.Inthepoliticaleconomymodelswestudybelowpublicprovisionofprivategoodsservesasaskillscreeningdevice,allowingredistributiontotakeplacewithlessdistortionsthanifonlytaxesandcashtransferswereused.Thismechanismhasbeenstudiedby,forexample,BlomquistandChristiansen(1995,1998a),BoadwayandMarchand(1995)andCremerandGahvari(1997).Themodelconsidersaneconomywithtwotypesofindividuals,onegroupoflowproductivityandonegroupofhighproductivityindividuals.Skilllevelsarere¯ectedinwagerates.Informationaboutindividualskills(wagerates)isprivateinformationnotavailabletothegovernment.Letbehoursofwork, JournalofPublicEconomics73(1999)31before-taxlaborincomeandafter-taxincome.Letdenoteaquantityofagoodthatisonlyavailableonthemarketandbethequantityofagoodthatcanbeboughtonthemarket,butalsocanbepubliclyprovided.Forconveniencewenormalizetheproducerpricesoftoone.Individualpreferencesarerepresentedbyastrictlyquasiconcaveutilityfunction).Wede®netheconditionaldemandfunction)byMax)w.r.tWewillassumethepreferencesaresuchthat0and0,i.e.isanormalgoodandthedemandforincreasesinhoursofworkfora®xedincome.BlomquistandChristiansen(1998a,p.405)showthat0isequivalentto0,i.e.thatthemarginalvaluationoftermsofincreasesinhoursofwork.BlomquistandChristiansen(1998a)alsoshowthat,providedthat0,thebestpublicprovisionschemeistheonethatallowstheconsumerstotopupthepublicquantityattheirownexpense.Theinstrumentsavailabletothegovernmentareincometaxesandpublicprovisionof.Wedenotethepubliclyprovidedquantityby.Norestrictionsareimposeddirectlyontheshapeoftaxschedulethatcanbeused.However,theasymmetricinformationimpliesthatthepolicymustbedesignedsubjecttoaninformationconstraint.Theincome/consumptionbundleassignedtoapersonmustbechosensubjecttobeingunattractivetotheotherperson.Lettheindirectutilityfunctionconditionalonthelaboursupplyandthepublicprovision.Weuseasuperscripttoindicatetypeofindividual.thenumberofindividualsoftypetypeh1,2j.TheproblemofdesigningaParetooptimalincometaxandpublicprovisionstructureisthengivenby:111)(1)11222222222112)(3)111221)(4)222111120(5)Theconstraint(2)assignsaminimumutilityleveltoperson2.Theconstraint(3)Theconditionaldemandfunction)hasbeenusedquitemuchinthetheoreticalstudyofoptimalcommoditytaxes.However,ithasnotbeenempiricallystudied.Forsomegoodswecanmakeaninformedguessaboutthesignof.Forexample,ifisdaycareitseemslikelythatthedemandforincreasesinhoursofwork.AsdiscussedbyCremerandGahvari(1997)theavailabilityofcommoditytaxesmakeslittledifferenceforthecaseforpublicprovision.However,asdiscussedinBlomquistandChristiansen(1998b)commoditytaxesmayberedundantprovidedthatanincometaxandpublicprovisionareboth JournalofPublicEconomics73(1999)31Fig.1.TheParetofrontierwithandwithoutpublicprovision.istheself-selectionconstraintimposingthatthetaxesmustbesetinsuchawaythatperson2doesnotgainbymimickingperson1.Eq.(4)isthecorrespondingconstraintthatperson1doesnotgainbymimickingperson2.Onecanshowthatatmostoneself-selectionconstraintisbinding.Themoststudiedcase(oftencalledthenormalcase)istheonewhereredistributionisfromthehighskillgrouptothelowskillgrouptotheextentthatconstraint(3)isbinding.Thisconstraintcausestheoptimaltaxationtobesuchthatadistortionisimposedonthelowskillperson.Heisinducedtosubstituteleisureformarketconsumptionsothatthebeforetax/aftertaxincomeistoolowascomparedwiththe®rstbest.Inequality(5)isthegovernment'sbudgetconstraint.isrestrictedtozero,theproblemabovereducestothestandardoptimalincometaxproblemasformulatedbyStiglitz(1982)andStern(1982).ThesolidcurveinFig.1depictsthecorrespondingParetofrontier.Forlevelsofsuf®cientlyclosetothelaissezfairelevel,indicatedbypointA,redistributioncantakeplacewithouttheself-selectionconstraintbeingbinding.However,formoreambitiousredistributiontheself-selectionconstraintwillbind.AsshownindetailinBlomquistandChristiansen(1995,1998a),ifleisureisnonseparablefromgoods,publicprovisionofasuitableprivategoodcanalleviatetheself-selectionconstraintandgenerateaParetoimprovement.Fig.1illustrateshowpublicGiventheassumptionthat0acommoditysubsidyoncanalsobeusedtoalleviatetheself-selectionconstraint(seeEdwardsetal.(1994)andBlomquistandChristiansen(1998b)). JournalofPublicEconomics73(1999)31provisioncanshiftthefrontieroutwards.TotheleftofpointandtotherightofpublicprovisionispartoftheParetoef®cientpolicy.Onecanshowthatthepuretaxoptimizationhasauniquesolution.ThisimpliesthattoeachpointontheParetofrontierthereexistsauniquepolicypoint.For0theremightexistmultiplesolutionstotheoptimizationproblem,i.e.,theremightbepointsontheParetofrontierthatcouldbeimplementedbyalternativepolicies.Thisdoesnotcreateanyproblemsforthefollowinganalysisexceptthatitcomplicatesnotation.Fornotationalsimplicitywethereforeinthefollowingwriteasifthesolutiontotheproblemde®nedbyEqs.(1)±(5)alwaysisunique.Wewillbrie¯ycommentonthisinSection3.Togetfurtherintuitionfortheroleofpublicprovision,considerthecaseinwhichthereisredistributionfromhigh-skilltolow-skillindividualsandconstraint(3)isstrictlybinding.Notethatifthehigh-skilltypemimicsthelow-skilltype,theformerhasmoreleisureaslessworkeffortisrequiredtoobtainagivenincomewhenthewagerateishigh.Withtheassumptionthathasbeenmadeaboutthedemandfor,themimickerhasalowerdemandthanperson1.Thismeansthatitispossibletogivethelow-skillpersonatransferintermsofthatisatthesametimewithintheconsumptionleveldesiredbyperson1andbeyondtheleveldesiredbythemimicker.Thenforthelow-skillpersonthetransferisequivalenttoatransferincash,whilefromtheperspectiveofthemimickeritisinferior.Theadvantageisthatthetransferisachievedwithlessinducementforperson2tomimic.Theself-selectionconstraintissoftenedanditispossibletoachieveamorefavorableoptimum.BlomquistandChristiansen(1995)discusshowpublicprovisionofeducation,daycareandhealthcarecanbe®ttedintothisframework.Tobeconcretewecouldexemplifywithdaycare.Iflowskillwomenworkingfulltimerequire,say,1800hofdaycareandbuythisinthemarket,thentheirsituationwouldbeunchangedifoneintroducedthesameamountofpubliclyprovideddaycareandatthesametimeaddedthecostofprovisiontothetaxbill.Thesituationforhighskillwomenwhoworkfulltime,requiring1800hofdaycare,wouldalsobeunchanged.However,highskillwomenthinkingofworkinghalftimeinsteadoffulltimewillnow®ndtheformeroptionlessattractive.Thisisbecausetheaftertaxincomeearnedbyworkinghalftimehasdecreasedwithanamountequaltothecostof1800hofdaycarewhereasthedaycaretheyneedisonly900h.BlomquistandChristiansen(1995)characterizegoodssuitableforpublic1.Thegoodshouldbesuchthatitisimpossible(oratleastveryhard)toresellthepubliclyprovidedquantity.2.Thereshouldnotbeanyclosemarketsubstitutestothegoodinquestion.3.Thedemandforthepubliclyprovidedgoodshouldvarymuchasleisurevaries.4.Leisureshouldnotbeweaklyseparablefromgoods.Theideabehindthesecharacteristicsisthatitshouldbepossibletoimposea JournalofPublicEconomics73(1999)31quantityconstraintonperson2whenmimickingandthatthisquantityconstraintshouldhurt.Clearly,ifitwerepossibletoresellthepubliclyprovidedquantitythemimickerwouldnotbehurtbypublicprovision.Likewise,ifgoodsclosemarketsubstitutesitwouldnothurtthemimickermuchtohavethecompositionbetweenchangedfromthepointselectedifhecouldchoosefreelygivenhisaftertaxincome.Thattheverysamegoodastheoneprovidedcanbepurchasedinthemarketwhentoppingupispossible,posesnoproblemsasthemimickerwillnotwanttousethisoption.Characteristicensuresthatitispossibleto®ndapublicprovisionlevelthathurtsthemimickerbutisfairlyharmlessforthelowskillperson.Characteristicisanecessaryconditionfortheschemetowork.Ifleisureisweaklyseparablefromgoodsthelowskillpersonandthemimickerwouldhavethesamedemandforandpublicprovisionwouldhurtthelowskillpersonasmuchasthemimicker.Thenormativetheoryoftaxationandpublicprovisiondescribedabovederivesanumberofresults.Con®ningourattentiontotheso-callednormalcase,wemaydrawattentiontothefollowingresults:Result1.thennoParetoimprovementcanbeobtainedbypublicprovisionofxSeeBlomquistandChristiansen(1995),(1),BoadwayandMarchand(1995),(1),(BlomquistandChristiansen(1998),(andCremerandGahvari(1997),(2).Result2.SupposethatandindividualsareallowedtosupplementthepublicprovisiononthemarketthenapolicywithoptimaltaxationandoptimalpublicprovisionstrictlyParetodominatesthepuretaxoptimumSeeBlomquistandChristiansen(19982).3.AmedianvotermodelWeshallstudyavotingequilibriumofamodelwherethereareonlytwoskillclassesasinSection2.However,weshalltakeamoregeneralapproachtopolicymakingwhilestillusingthemodelabovetoinferimplicationsofthevotingequilibriumforpublicprovisionofprivategoodsinparticular.Asdiscussedintheintroductiontheliteratureonthisissuehasprovidedamajormotivationforthepresentstudy.Acentralassumptioninourvotingmodelwillbethatindividualsdifferwithrespecttohowmuchredistributiontheywant.Toestablishthebuildingblocksofourmodelweshalldiscusseachoftheseelementsinturn.Therecouldverywellexistanonmarketperfectsubstituteforthepubliclyprovidedprivategood.Fordaycarethisisclearlythecase.Careofchildrenathomeisaverygoodsubstitutefordaycareboughtinthemarket.Thisistheveryreasonwhyitisreasonablethatthemarketdemandfordaycareincreaseswithhoursofwork(decreaseswithamountofleisureavailable.) JournalofPublicEconomics73(1999)31PolicyinstrumentsAtthegenerallevelweletavectorofpolicyinstrumentsbedenotedbyisthepolicysetsuchthatthatZ.Byde®nitionisthesetofavailableandfeasiblepoliciesthatcapturesallrelevantrestrictions.Weshallinterpretthepolicyassomekindofeconomicpolicywithdistributionalimplications,butotherwisethisisaverygeneralpolicyformulation.Thetaxandpublicprovisionmodelaboveisaspecialcaseinwhichapolicyconsistsofaconstellationofbeforeand1122aftertaxincomesintendedforthehighandlowskillgroupsandavalueofthepublicprovision.Inthisspecialcaseisthesetofcombinations1122thatachievebudgetbalance,andsatisfytheself-selectionconstraintsasstatedbyEqs.(4)and(5)above.PoliticalpreferencesIngeneraleachindividualhaspreferencesoverconsumptionbundlesincludingleisure(orlabour)thatcanberepresentedbya(direct)utilityfunction.Weimposethestandardassumptionfromoptimumtaxtheorythatthebasicutilityfunctionisthesameforeverybody.Whendecidingonhisconsumption/labourchoicetheindividualissupposedtomaximisehisutility.Maximisingbehaviourinresponsetoagivenpolicythenimpliesautilitylevelindicatedbyanindirectutilityiii),whichinthespeci®cmodelisgivenbyWhenvotingtheindividualwillexpresshispreferenceforchoiceofpolicy.Thisisadecisionwhichisofadifferentnaturethanthechoicesmadeinthemarketplace.Tovoteistoexpressanopinionalsoaboutdistributionalpolicy.Thevotercanthenadoptthreedifferentattitudes.(i)Hecantakeapurelysel®shandmyopicattitudevotingforthepolicyfromwhichhisownwell-beingwillbene®tmoregivenhispresentposition.(ii)Thevotercanbebasicallysel®sh,butmayfeeluncertainabouthisownfutureandperhapsthatofhischildren.Hemaythenhaveaninsurancemotiveforsupportingtransfersthatwillmakehimanetcontributorinhispresentposition,butfromwhichhewillbene®tifinthefuturestateheisdeprivedofhispresentskill.Theperceptionofdistributionalpolicyasasocialinsuranceschemeisfairlycommonintheliterature,seee.g.Barr(1992),Sinn(1995,1996).(iii)Thevotermaybegenuinelyaltruisticandcareaboutthewelfareofotherpeople.Inpracticeitishardtodistinguishinsuranceandgenuinelyaltruisticmotivation,butinthepresentcontextidentifyingthetruemotivationisoflittleinterest.Whatweshallassumeisthatnoteverybodyismyopicallysel®sheventhoughitisperfectlypossiblethatthereissuchacategoryofpeopleinthemodel.ThereisasubgroupofindividualswhocareaboutthewelfareofpeopleintheotherskillThenotionthatpeoplearenotentirelysel®shisnotanovelidea.LindbeckandWeibull(1993)assumethatavoterismotivatedbothbyapartyidenti®cationand JournalofPublicEconomics73(1999)31theutilityshederivesfromaparticularpolicy.WithtwopartiesAandBavoterwhoisa`B-partisan'mayvoteforpartyBeventhoughshewouldderiveasomewhathigherutilityfromA'spolicy.SeealsoFiorini(1981)foradiscussion.DixitandLondregan(1998)assumethatinadditiontocaringabouttheimplicationsofincometransfersfortheirownwellbeing,votersgenuinelycareaboutsocialissues:``Thatvoters'equityconcernsarenotmererhetoricalmasksforself-interestisclear;therearerichliberalswhowillvoteforaleftistpartythatpromisesahightaxrate,andpoorlibertarianswhowillsupportarightistpartyeventhoughtheywillpersonallybene®tlittlefromitstaxandtransferpolicies''(op.cit.p.499).Ourassumptionsimplythatwhenvotinganindividualattachesaweighttotheutilityoftheowngroup,butingeneralalsotothatoftheothergroup.Itisstillpossiblethatanumberofvotersassignazeroweighttotheutilityoftheotherskillgroup,andthatvotersfromoneclasstypicallyattachagreaterweighttotheutilityofhis/herownclass.Letbetheweightgiventotheutilityofthelow-skillclassand1theweightgiventothehigh-skillclasswhere01.Thevalueofcharacterisesthepoliticalpreferencesofeachvoter,andhasacontinuousdistributionwithaclassspeci®cdensity1,2.Letdenotetheutilityofanindividualinskillclassthatisobtainedatthepolicy,andrecallthatthenumberofindividualsinthatclass.Anindividual'spoliticalpreferencesarethenrepresentedbyapolicypreferencefunction1122Inthetaxandpublicprovisionmodelthepolicypreferencefunctionwilltakethe11112222)(7)ppppThepersonwillsupportthepolicythatyieldsthehigherforthevalueofcorrespondingtohispreferences.Itisofinteresttonotethat(6)isgeneralenoughtoaccommodatetheclassof11122`mixed'preferencefunctionsforapersonwhoisamemberofclass1(andlikewiseforclass2).Wecaninterpretasa`sel®sh'partwiththeparametermeasuringthedegreeofsel®shness,whilsttheremainingpartsconstitutea`socialpart'.Weassumethat0,01,andatleastoneoftheparametersisstrictlypositive.Thesepreferencesareequally11122wellrepresentedbyafunction)whereisanarbitrarypositiveparameter.Letbede®nedbythefollowingequations:,andweseethatwearebacktoapreferencefunctionoftype(6).SolvingwithrespecttowegetDixitandLondregan(1998)alsointroduce`mixed'preferencesmadeupofasel®shpartandasocialpart,buttheyuseotherspeci®cationsoftherespectiveparts. JournalofPublicEconomics73(1999)31Wenotethat0and01asshouldbethecase.Onemightarguethattrulysocialpreferencesshouldnotbebiasedinfavourofanyclasssothat0.5.Thenweimmediatelyseethatanydeviationoffrom0.5re¯ectssel®shness.Havingassumedthateverybodyhasthesamebasicpreferencesforownconsumptionbundle,wedonotcapturethepossibilitythatpreferenceforequality(orsel®shness)andpreferenceforleisureareinterrelated.Onemightforinstancesuspectthatmorehard-workingpeoplewouldfeelmoreentitledtokeepthefruitsoftheirworkthantherichandlazy,andwouldbemoreopposedtogivetransferstoothers.Otherwiseitishardtoseeconvincingargumentsingeneralwhysocialattitudesshouldvarysystematicallywithworkpreference.votingbehaviourHowanindividualranksdifferentpoliciesaccordingto(6)onlydependsonhowthepoliciesaffect.Whenstudyinghowanindividualranksdifferentpolicieswecanthereforestudyhowheranksdifferentcombinationsof.SinceallindividualswithinaskillclassaretreatedequallywecandrawtheParetofrontierasinFig.1.Wedenotethemedianandde®nethemedianvoterasthevotercharacterizedbyThevotingequilibriumThevotingequilibriumcanbecharacterisedbythefollowingproposition.Proposition1.ThemedianspreferredpolicywillbeaCondorcetwinnerThispolicyisParetoef®cientLetusconsidertwopolicies,wheretheformeristhepolicypreferredbythemedianvoterandthelatterisafeasible,butotherwisearbitraryalternativepolicy.Thepolicyimpliestheutilitylevelsforthemmmrespectiveskilltypes.Forsimplicity,andwithoutlossofgenerality,wenowassumethereisonepersonofeachtypeSincethemedianvoterassignspositiveweightstotheutilitiesofbothskilltypeshispreferredpolicyisobviouslyParetoef®cient.Theutilitylevelscorrespondingto.Letusde®neaaaACondorcetwinningpolicyisonethatwillwininmajorityotingagainstanyotherfeasiblepolicythatmaybeproposed. JournalofPublicEconomics73(1999)31asthedifferencebetweenthewelfarelevelresultingfromandthewelfarelevelobtainedfromasperceivedbyavoteroftypetypeaV1(12a)V]2amm1211222211aV1(12a)V5a[(V2V)1(V2V)]1V2VandD9(a)5[(V2aamaammam122)].Therearethreecasestoconsider.(i)Assume®rstthataam1220,i.e.thealternativepolicyismoregeneroustotypeoneandaamlessgeneroustotypetwo.Then0.Sinceisthepreferredpolicyofthemedianvoter,obviously0.Itfollowsthatallvoterswith(andsomevoterswithgreaterthanbutcloseto)willprefer.(ii)Assumemma1122nextthat(0,(0,i.e.thealternativepolicyismoremaamgeneroustotypetwoandlessgeneroustotypeone.Then0.Itfollowsthatallvoterswith(andsomevoterswithlessthanbutcloseto)will.WemaynotethatinthecasesconsideredsofarmayormaynotmaabeaParetoef®cientpolicy.TheonlyassumptionisthatthispolicyisnotParetodominatedby,whichisthe®nalcasetobeconsidered.(iii)Assumethat11220,(0.SincethealternativepolicyisParetodominatedbymaam,everybodywillprefer.QEDmmaReadersfamiliarwiththemedianvotertheoremknowthatthetheoremisonlyapplicabletovotingoveronedimensionalissues.Sowhydoweobtainamedianvoterresulthere?ThereasonisthatthevotingisdefactooverpointsontheParetofrontier,whichisone-dimensionalsincethereareonlytwogroups.TheintuitionforProposition1isstrong.Whenvotingeachindividual'smostpreferredpointistheonethatmaximizesthepolicypreferencefunction,whichwerecognizeasaBergson-Samuelsonsocialwelfarefunction.ButweknowfromwelfareeconomicsthatamaximumofsuchasocialwelfarefunctionisnecessarilyaParetooptimum.Whenallindividualshaveapreferredpolicywhichisef®cient,itwouldindeedbesurprisingifthevotingequilibriumwerenotalsoef®cient.ReturningtothetaxandpublicprovisionmodelofSection2,wecanstate:Corollary1.AtthevotingequilibriumofthetaxandpublicprovisionmodelarevalidAccordingtoProposition1themedianvoter'spreferredpolicywillbeaCondorcetwinner,thatis,itwillwininmajorityvotingagainstanyotherproposedpolicy.Sincethemedianvoter'spreferredpolicyisParetoef®cient,publicprovisionwillbepartofthepolicyifthemedianvoterspreferredpolicyisinacertainrangeoftheParetofrontier.Wenotethatinthepresentsettingoptimalnonlinearincometaxesareavailable.Still,thevotingequilibriumwillbesuchthatpublicprovisionispartofthepolicy.Thereasonisthatbyusingpublicprovisionofaprivategoodtheef®ciencyoftheincomeredistributionisincreased.ItalsoisParetoef®cient,thereisofcoursenocasewhereitisParetodominatedbythealternativepolicy. JournalofPublicEconomics73(1999)31meansthatthecharacterizationofprivategoodssuitableforpublicprovisionthatisgiveninBlomquistandChristiansen(1995)andreproducedinSection2abovewouldbevalidinthepresentcontext.InSection2wenotedthatthesolutiontoEqs.(1)±(5)mayfailtobeunique.However,thisdoesnotcreateanyseriousdif®culties.Thereasonforthisisthatwhatdetermineshowindividualsvoteisthesolutioninutilityspace.Suppose,forexample,thattheoptimizationproblemsolvedbythemedianvoterhastwo1212thatbothyieldtheutilityallocation.Toobtainuniquenessinpolicyspacewemustaddamechanismwherebythemedianvoterchoosesoneofthetwopolicies.Themedianvoter,asallothervoters,isindifferentbetweenthetwopolicies.Henceaddingasimplerandommechanismwoulddo.Inthemodelabovetheelectoratehavebeenendowedwithsocialpreferences.Inacompanionpaper(BlomquistandChristiansen(1998c))wehavealsoanalysedwithinasimplemajorityvotingmodelthesituationinwhichallvotersareindeedentirelysel®sh.Assumingthatthelow-skillclassisthelargeritwillexploititsmajoritypositiontomaximisetheutilityoftheclassmemberssubjecttotherelevantconstraints.Thepreferredpolicyofthelow-skillindividualswillthenbeaCondorcetwinner,anditistrivialthatthispolicyisParetoef®cient.ObviouslyResults1and2arevalidinthetaxandpublicprovisionversionofthe4.RepresentativedemocracywithtwopoliticalpartiesWeshallconsideravotingmodeloftwopoliticalparties.Themodelisestablishedby®rstmakingassumptionsaboutthepoliticalpartiesandthenaboutthevoters.Thecompetitionbetweenthepartiesisperceivedasanon-cooperativegamewhichisshowntohaveaNashequilibrium.Thepropertiesofthisequilibriumareexaminedinsomedetail.ThepartiesOurmodelisinspiredbyHanssonandStuart(1984),LindbeckandWeibull(1993)andDixitandLondregan(1998).Therearetwopartiesdenotedby(leftistparty)and(rightistparty).Thenatureofthepolicysetisthesameasbefore.PriortotheelectioneachpartychoosesapolicypolicyZ;p5L,R.pWeassumethatthepolicythatisactuallyimplementedisthepolicyofthepartythatwinstheelection.TheunderlyingpresumptionisthateachpartyisabletoTheutilitylevelofthehigh-skillclassmaybeboundedfrombelowbyasubsistencelevelorareservationutilityinthecaseofmobilityoritmaybeloweredtothepointwherefurthertaxationofthehigh-skilledfailstogenerateadditionaltransferstothelow-skilledbecauseofdisincentives. JournalofPublicEconomics73(1999)31crediblycommittoacertainpolicy.Thusweruleoutthepossibilitythatapartydoesnotkeepitspromisesfromtheelectioncampaign.Itiscommonintheliteraturetomakesomekindofassumptiontotheeffectthatthepartiesimplementtheirpromisedpoliciesifwinning.Sometimesitisjustimposedasanassumptionthatthepartiesimplementtheirpromisedpolicies(e.g.DixitandLondregan(1998)),sometimestheassumptionisimplicit(e.g.LindbeckandWeibull(1987)orHanssonandStuart(1984)),andsometimesthereisanappealtounderlyingassumptions.LindbeckandWeibull(1993)assumethatatleastallvotersbelievethatthewinningpartywillimplementitsannouncedpolicy.Foratheoryofwhathappensattheelectionthismaybeasuf®cientassumption.Foratheoryofactualpolicyoneobviouslyneedstoknowalsohowthepartiesactuallybehave.LindbeckandWeibullalsoappealtotheexistenceoffutureelections(otherwisenotappearinginthemodel)tosuggestthatcheatingmaybeprohibitivelycostlyintermsoffuturelossofcredibility.BesleyandCoate(1997)pointoutthatitisnaturaltoassumethatthecitizenwhowinstheelectionimplementshispreferredpolicy(i.e.thepolicyhewouldchooseasadictator),whilecommitmenttootherpoliciesismoreproblematicduetolackofcredibility.Atthepre-electionstagetheoutcomeoftheelectionisperceivedasrandom.Oncetheoutcomeoftheelectionisknown,eachpartyevaluatesthewinningpolicyaccordingtoitssocialwelfarefunction.Weassumethereisnobene®tfromwinningtheelectionassuch.Thereisnoso-called`ego-rent'.Itfollowsthatpriortotheelectiontheobjectivefunctionofapartyistheexpectedwelfareaccordingtothepreferencesoftheparty.Forpartytheexpectedwelfareisgivenby:)(8)LRLLRRistheprobabilitythatpartywillwintheelection.Thisprobabilitydependsonthepolicychoicesofthetwoparties.Belowwedescribehowthisprobabilityisdetermined.Thesocialwelfarefunction)isgivenby1122.(9)pLpLpTheexpectedwelfarefromtheperspectiveofpartyisde®nedinasymmetricway.Weassumetheleftistpartyassignsahigherweighttothewelfareofthelowskillgroup,.Partymaximizestheexpectedwelfarew.r.t.foragivenLRLandviceversa.Acomplicationinpracticemaybethatatthetimethepolicyisdesignedthestateoftheworldthatwillmaterialiseintheperiodofof®ceisnotknownwithcertainty.Sincethereislimitedscopeforstatecontingentelectionmanifestos,itmaybeamatterofinterpretationwhethertheactualpolicyisaccordingtothepre-electionplatformwhenallowingfortheneedtoadjusttothecircumstancesthatarise.Weabstractfromuncertaintyofthiskind.SeeAlesina(1988). JournalofPublicEconomics73(1999)31votingbehaviorThevotersareassumedtohaveexactlythesortofpreferencesasinthemedianvotermodelabove.Theymaybemoreorlessmyopicallysel®sh,altruisticorconcernedwithsocialinsurance.Anindividual'spoliticalpreferencesarerepre-sentedbyapolicypreferencefunction1122(10)whichisidenticalto(6)above.Inourmodelitisoptimaltovotesincerely.Tounderstandwhythisissoitisworthwhiletoconsideranumberoftheassumptionsthathavebeenmade.First,avotercannotin¯uenceaparty'schoiceofprogrammebythreateningtovotefortheotherpartyundercertaincontingencies.Thisisbecausethereisnowayavotercancrediblycommittodoso.Everybodyknowsthatwhenelectiondaycomesalongitisintheinterestofacitizentocasthisvoteinfavourofthepolicywhichismoreinlinewithhispreferences.Doingotherwisewillonlyincreasetheprobabilitythatapolicyhelikeslessisgoingtowin.Secondly,itisassumedthatthepolicyofthewinningpartyisfullyimplemented.Iftheactualpolicyweretobedeterminedassomesortofweightedcompromisebetweenthepoliciesofthetwopartieswiththeweightsaffectedbyvoteshares,strategicvotingbehaviourmightcreepinasavotermightthenhopetodrawtheactualpolicyabitclosertohispreferredpolicybyvotingforapolicyhewouldnotliketoseefullyimplemented.(Buteventhensincerevotingisnotnecessarilyruledout;seeDixitandLondregan,1998).Thirdly,thereisonlyoneelectionperiodandthusnoincentivetotryandin¯uenceaparty'sfuturechoiceofplatform.Ourassumptionsimplythatanindividualvotesforparty11221122(11)LLRRWerewritethisas222(12)111222LRRLbedistributedas)and),01,forthelowandhighskillclass,respectively.Thenumberofindividualspreferringparty'spolicyisthengiven1122(13)Ifthevotingbehaviourisentirelydeterministiceachpartycancalculateforanycon®gurationofpolicieswhetheritisgoingtowintheelection.Thepolicieswillthenconvergeasfromtheperspectiveofanypartyawinningpolicyisalwaysbetterthanalosingpolicyaslongasthewinningpolicyisconsideredatleast JournalofPublicEconomics73(1999)31slightlypreferabletothepolicyoftheopponent.Thisscenarioisnotveryrealistic.Inpracticetherearenumerouselementswhicharebeyondthecontrolofthepartiesandwhichtheywilltreatasrandom.Intheliteratureseveralapproacheshavebeenadoptedtomodelrandomvotingbehaviour.Butitseemsthattheexactwaytomodelrandomvotingbehaviourisoflessimportance.Whatiscentralisthatthepartiescanonlyaffecttheprobabilitiesofwinningandtheexpectedoutcomeintermsofpolicyandwelfareaccordingtoaparty'spreferencescale.Acceptingthispremisewemayaswellchooseasimplerandomelement.Weknowthatinmostelectionsnoteveryonewillvoteandthatanimportantaspectofwinninganelectionistomotivatetheindividualssupportingthepartytoactuallyvote.Wedonotintendtogointothisinanydepth,butuseaquitestylizedmodeltocapturethefactthatmanypeopledonotvote.Withoutlackofgeneralityweassumethatindividualssupportingpartywillalwaysvote.However,onlyafractionoftheindividualssupportingpartywillactuallyvote.Thisfractionisarandomvariableassumedtofollowauniformdistributionwithsupport01.Let.Iftheprobabilitythatwillwinis1.Iftheprobabilitythatpartywillwiniszero.Fortheprobabilitythatpartwillwinisgivenby:(partywins)LRLRLRu.(N2m(z,z))/LRLRRewritingthisprobabilityweobtain:(partywins)AnimportantfeatureofthisexpressionisthatSometimesitisjustpostulatedthatprobabilitiesdependonpolicychoices.Sometimesmorespeci®cassumptionsaremade.ForinstanceDixitandLondregan(1998)assumethattherearegroupsofextremistvoterssupportingtheirrespectivepartiesirrespectiveofelectionmanifestos,butthesizeofeachgroupisunknowntoandistreatedasrandombytheparties.Theinclinationtogoandvotemaydependonweatherconditions,whetherthereisanin¯uenzaepidemic,theencouragementofferedbythemassmedia,etc.Onemayarguethatitisabitspecialtolinkvotingpropensitytopartyadherenceassuch.Theimportantassumptionisthattheprobabilityofvotingdiffersbetweengroupsotherwisethestochasticelementwouldjustbesimilartohavinganelectorateofrandomsizebutwithdeterministicsharesofpartysupport.Wehaveoptedforthesimplestwaytomodelasystematicdifferencebetweenclasses.Ourapproachmayevenhaveaclaimforrealismasitisknownfromseveralcountriesthatconditionsinducingahighturn-outtendtofavourcertain JournalofPublicEconomics73(1999)31NashequilibriumTostudytheNashequilibriaforthismodelweneedtostudythereaction)and).Thisiscomplicatedsinceismultidimensional.LRRLPHowever,wecanreducethedimensionalityoftheproblemandinsteadstudythe2222reactionfunctions)and).Thereasonwhyitissuf®cienttostudyLRRLthesefunctionsisthateachpartywillalwaysrespondwithaParetoef®cientpolicy.Thismeansthatwecansummarizethepolicyofpartybyasingle.SinceonlyParetoef®cientpointsarerelevantisthenuniquelydetermined.LetusestablishthatallchosenpoliciesareParetoef®cient.Proposition2.TheoptimalstrategyforapartyistoproposeaParetoef®cientRecallthattheexpectedwelfarefromthepointofviewofapartyppppW(z)2W(z)]1W(z),whereistheparty'sownpolicy,istheppopoopopolicyoftheopponentandisthewelfarelevelaccordingtothepreferencesoftheparty.Alsorecallthatthewelfarefunctionofanypartyandvoterisoftheform1122.Weassumethatforbothparties01,andthereisacontinuousdistributionofvotersfromthosewith0tothosewith1.AssumethatthepartyisgoingtomakeachoicebetweenapolicywhichisnotParetoef®cient,andaParetosuperiorpolicy,whiletakingthepolicyproposaloftheopponentas®xed.Thechoicewillhavetwoeffects.Itwillaffectthewelfarethatthepartyobtainsifitwins,anditwillaffecttheprobabilityofwinning.SincethepartyadherestotheParetoprinciple(i.e.,italwaysconsidersaParetoimprovementtobedesirable),thewelfareitobtainsifwinningmustincreaseifitpickstheParetosuperiorpolicy.SincethevotersalsoadheretotheParetoprinciple,allvoterswill®ndtheparty(atleastweakly)moreattractiveifitselectstheParetosuperiorpolicy.Becomingamoreattractivepartytovoteforthepartycannotlosesupport.Infact,sincethereisacontinuumofvoters,somewhowouldotherwisebemarginalsupportersoftheopponentwillnowvoteforthepartyunderconsideration.ChoosingaParetosuperiorpolicyisthewaytoincreaseInSection2wenotedthattheremightexistmultiplesolutionstoEqs.(1)±(5).However,wehavepursuedtheanalysisasifthesolutionswereunique.Thisisnotrestrictive.Sincepartiesandindividualsevaluatepoliciesintermsoftheutilityimplicationstheanalysiswillremainunaffectedifacertainutilityallocationcanbeimplementedbymorethanonepolicy.Ifonepartyproposesautilityallocationthatcanbeimplementedbytwoormorepoliciesonecouldaddamechanismwherebythepartychoosesauniquepolicy.Sinceitisofnoconsequencewhichpolicythatisactuallychosenasimplerandommechanismwoulddo.TheonlycaveatisthatasaspecialcasetheParetosuperiorpolicymayonlybene®tonegroup,andtheremaybevoterswhoassignazeroweighttothatgroup.Thesevoterswillthenbeindifferent,whiletheothersstrictlyprefertheParetosuperiorpolicy. JournalofPublicEconomics73(1999)31welfareifthepartywinsandthewaytoincreasetheprobabilityofwinning.QEDRecallingfromSection2thepropertiesofaParetooptimuminthetaxandpublicprovisionmodelitistrivialtostate:Corollary2.AtthepoliticalequilibriumofthetaxandpublicprovisionmodelarevalidWhendiscussingfurtherthepoliticalequilibriumitisusefultoconsiderasbenchmarksthe(hypothetical)casesinwhicheachpartyisnotexposedtopoliticalIfthepartiescouldactasdictatorstheywouldchoosepoliciesthatmaximizetheirrespectivewelfarefunctions,i.e.,objectivefunctionsoftype(9).Wewilldenotethesepoliciesasthedictatorpolicies.Wedenotethecorresponding222valuesof.TheimpliedpointsontheParetofrontierareindicatedinFig.2.DependingontheirvalueofdifferentindividualspreferdifferentpointsontheParetofrontier.Foranindividualwith0thepreferredpointwillbetheoneachievesitsmaximum.Wedenotethis.AsincreasesthepreferredpointwillmovedowntowardstherightalongtheParetofrontier.IfFig.2.TheParetofrontieranddictatorpolicies. JournalofPublicEconomics73(1999)311thepointwhereismaximizedwillbethepreferredone.Wedenotethis222Toguaranteethatthevotingproblemisnon-degenerate,i.e.thatnopartywinswithcertainty,wehavetomakeassumptionsonthedistributionof.Weassumethedistributionofissuchthatmorethan50%oftheindividualspreferavaluethatislessthan.Otherwisethepolicycorrespondingtowouldwinwithcertainty.Wealsoassumethatmorethanafraction(10.5/)oftheindividualspreferavalueofthatislargerthan.Otherwisethepolicycorrespondingtowouldwinwithcertainty.Proposition3.ThereexistsaNashequilibrium2222Wewanttostudycrucialpropertiesofthereactionfunctions)andRLLRandshowthattheresponsecurvescrossatleastonce.First,supposethatoneparty2222proposestoset.Thenthereexistsavalueofwhichthemaxmaxopponentprefersto,andwhichtheopponentcanadoptandbesureofwinningtheelection.Toseethisitsuf®cestonotethatbypickingclosetotheshareoftheelectoratefavouringthecanbemademaxmaxin®nitesimallysmall.Itobviouslyfollowsthatthebestresponseoftheopponentis22222alsosomevalueof.Wecanconcludethat,andmaxmax222222.Byanalogousreasoningwecanshowthatmaxmaxminmin222.Thesefeaturesoftheresponsefunctionsarere¯ectedinFig.3.minmin)willstartbelowthe45lineandendupabovethe45lineonthehorizontal222axisindicating)willstartabovethe45lineandendupbelowthe45lineontheverticallineindicating.HencetheresponsecurveswillcrossandtherewillbeaNashequilibrium.QEDANashequilibriumisParetoef®cientandtheResultsThecorollaryfollowsfromProposition2andtheResults1and2.QEDThecorollaryimpliesthattherewillbepublicprovisionofaprivategoodatapoliticalequilibriumprovidedthatconsumerpreferencessatisfytheassumptioninResult2andtheredistributionalambitionsaresuf®cientlylarge.WedenoteaNashequilibriumasHavingestablishedtheexistenceofaNashequilibriumwenextproceedtocharacterizefurthersuchanequilibrium.InthelemmaandpropositionbelowwenarrowdownthesetofpossibleNashequilibria.TheParetofrontieristhepartofthecurveinFig.2wherethecurveisdownwardsloping. JournalofPublicEconomics73(1999)31Fig.3.TheresponsecurvesandtheNashequilibrium.ApartypwillalwaysproposeapolicywhichitaccordingtoitswelfarefunctionoftypeconsiderstobestrictlybetterthanthatofitsopponentothesensethatWApartywillneverchooseapolicywhichitaccordingtoitswelfarefunctionoftype(9)perceivesasstrictlyworsethanthepolicyoftheopponent.Iftheopponent'spolicyisbetter,itisrationalforthepartytoadoptthatpolicybecausethenabetterpolicywillwinwithcertainty.Supposenextthatthepartiesconsiderselectingthesamepolicy.Thatpolicymustdifferfromthedictatorpolicyofatleastonepartysincethepartieshavedistinctlydifferentpreferences.Thenapartywithadifferentdictatorpolicywillgainbyswitchingtoapolicywhichitpreferstothecommonpolicy.Thenthereisalowerprobabilitythattheinitialpolicywillwinandapositiveprobabilitythatabetterpolicywillwin,sotheoutcomecanonlygetbetterfromthatparty'sperspective.QEDÅÃÃÅProposition4.ANashequilibriumwillbesuchthatVLLRRThelemmaimpliesthat.TheassumptionthatimpliesthatLRLRÅÅÅ.Partywillneverchoosea-levelbelowandpartywillneverLRL JournalofPublicEconomics73(1999)31222choosea-levelabove.Bypickingavalueofthatislessthanwouldswitchtoaninferiorpolicywithoutgettinganymorevotes.ThereasonisthatthosewhopreferavalueofthatislessthanwillalreadyvoteforpartyBychoosinga-levelslightlyabovemaygainvotesfromthosewhowouldotherwisebeindifferentbetweentheparties'policies.However,ifthedeparturefromissuf®cientlysmallthepartywillonlysufferanegligiblelossofwelfare,wereittowininanycase.Thuspartywillchooseapolicysuchthatasmalldecreaseinwouldbepreferableassuch,butwouldscareawaysomevoters.Similarly,partyRwillchooseapolicysuchthatafurtherdecreaseinwouldbeconsideredwelfaredecreasingaccordingtoitswelfareobjectivewhileattractingmorevoters.QEDThepropositionimplies:(i).Bothpartiesdeviatefromthedictatorpolicies.(ii).Eachpartywilltradeoffthequalityofthepolicyfromitsownperspectiveandthechanceofwinningtheelection.(iii).ThepoliciesconvergetosomeextentwithoutThischaracterisationhassomeimportantfeaturesincommonwithDixitandLondregan(1998),HanssonandStuart(1984)andLindbeckandWeibull(1993),whileitdiffersfromsomeotheranalysessuchasforinstanceLindbeckandWeibull(1987)wherethereiscompleteconvergencetoasinglepolicy.Itcanalsobeofinteresttoconsidertheanalyticsofthediscussionabovebyinvestigatingthe®rstorderconditions.Theobjectivefunctionsoftherespectivepartiesaregivenby:222)forparty(14)LRLLRR222)forparty(15)LRLLRR)denotesparty:sevaluationoftheotherparty'spolicyproposalexpressedintermsoftheimpliedutilitylevelfortypetwoindividuals.The®rstorderconditionsformaximizingtheseobjectivefunctionsaregivenby:0(16)0(17)Eachconditionconsistsoftwomainterms.Theformertermof(16)expressestheexpectedwelfaregainpartycanobtainbygivingaslightlyhigherutilitytoclasstwo,andthusincreasingtheprobabilityofwinningtheelectionbygainingthesupportofsomevoterswhowouldotherwisevote.Thelattertermre¯ectsthat JournalofPublicEconomics73(1999)31thereisacostbecauseinordertoobtaintheadditionalvotesthepartyhastoadoptapolicywhichislesssatisfactoryaccordingtoitsownpreferences.Thepartywilltradeofftheprobabilityofwinninganditssatisfactionwiththepolicyitproposes.Uptoapointideologicalpurityissacri®cedinordertoreducethelikelihoodoflosingtoapolicywhichisideologicallyevenlessacceptable.Thesimilartrade-offforpartyRisexpressedin(17).5.ConcludingremarksPublicprovisionofprivategoodsisaphenomenonwhichhastoalargeextentbeenneglectedbytextbooksinnormativepubliceconomics.Theconventionalviewhasbeenthatitispublicgoods,andpossiblythecloselyrelatedprivategoodsprovidedunderdecreasingaveragecosts,thatshouldbepubliclyprovided.Itshouldbethetaskofprivatemarketstoprovideprivategoods.Thisobservationisnoteasilyreconciledwiththefactthatinmanycountriesprivategoodsprovisionconstitutesalargeshareofpublicexpenditure.Twoexplanationsseempossible.Oneisthatnormativepubliceconomicsdoesnotcapturewhatisactuallygoingononthearenaofeconomicpolicy.Policymakingismotivatednotbytheef®ciencycriteriaofwelfaretheory,butratherbythedesiretodesignapolicywhichcanobtainamajorityinthevotingprocess.Theotherpossibilityisthattheconventionalnormativetheoryhasmissedoutconsiderationsthatmaybeimportantforunderstandingthepotentialroleofpublicprovisioninpromotingeconomicef®ciency.Inrecentyearsbothapproacheshaveappearedintheliterature.Politicaleconomymodelshaveexplainedpublicprovisionofprivategoodsasapoliticalphenomenon.Modernpubliceconomicsmodelshaveexplainedhowtheremaybeanef®ciencycaseforpublicprovisionofprivategoodsunderplausiblein-formationconstraints.However,thetwoapproacheshavebeenquitedisparate,andhavebeenbasedonquitedifferentassumptions.Thispaperhasintegratedthewelfareapproachandthepoliticaleconomyapproachbyinvestigatingwhetherthemechanismsappearinginthewelfaremodelsmaybesustainedbythepoliticalprocessesstudiedinpoliticaleconomy.Wehavepresentedtwopoliticaleconomymodelsinwhichthisisthecase.Wehavestudiedbothamedianvotermodelandatwo-partymodelofrepresentativedemocracy.Thecommoncentralfeatureofthemodelsisthatitisdesirablefromtheperspectiveofthedecisionmakerstoachievea(secondbest)Paretooptimalallocation.TheadherencetotheParetoprinciplebyvotersandparties,andtheabilitytocommittotheannouncedpoliciesaccountfortheef®ciencyofthepoliciesthatareselected.Asweknowfromthenormativemodels,theroleofpublicprovisionistosoftentheself-selectionconstraintandthustorelaxamajorrestrictionontheattainableParetoef®ciency.SinceParetooptimalityisdesirable,itisnotsurprisingthattheagentsofthepoliticaleconomywouldliketotakeadvantageofthesamemechanismtopromoteef®ciency.Itis JournalofPublicEconomics73(1999)31thecaseforef®ciencyalongwiththebasicinformationconstraintthatdrivethepublicprovisionresults.Thisisdifferentfromtheconventionalpoliticaleconomycaseforpublicprovisionthatleanonseverelyrestrictedpolicyoptions.Implicitinourmodelsistheargumentthatef®ciencyaccordingtothenormativetheoriesisnotnecessarilyincon¯ictwiththepoliticalobjectivesthatareattheheartofpoliticaleconomymodels.Howrobustthiscoexistenceofeconomicef®ciencyandpoliticalobjectivesis,canonlybeansweredbyconsideringawiderrangeofpoliticaleconomymodelswithdifferentassumptions.Thiswillhavetobethetopicoffutureresearch.ThemaindistinctionbetweenthenormativeandthepoliticalmodelsthatwehaveconsideredisthatthelatteralsoexplainwhichParetooptimumthatisgoingtobechosenbythepoliticalprocess(evenifnotnecessarilyinadeterministicway).Itisthecentralresultofthepresentstudythatunderplausibleinformationconstraintseconomicallyef®cientpublicprovisionofprivategoodswillbepartofpoliticalrationaldecisionswithinthecontextofmajorityvoting.Thereisnoneedtoappealtopoliticalconcernsthatareatoddswitheconomicef®ciencytoexplainthesizablepublicprovisionofprivategoods.WethankparticipantsinseminarsattheUniversityofHelsinki,theUniversityofLeicesterandUppsalaUniversityforhelpfulcommentsandsuggestions.WearealsoindebtedtoHenryOhlssonandtherefereesandtheeditorJimPoterbafortheirconstructivecomments.VidarChristiansenwouldalsoliketoacknowledgethehospitalityofthePSERCandtheEconomicsDepartmentoftheUniversityofLeicesterduringhisvisitinthesummerof1998.Alesina,A.,1988.Credibilityandpolicyconvergenceinatwo-partysystemwithrationalvoters.AmericanEconomicReview78(4),796±806.Barr,N.,1992.Economictheoryandthewelfarestate:Asurveyandinterpretation.JournalofEconomicLiterature30(2),741±803.Bergstrom,T.,Blomquist,S.,1996.Thepoliticaleconomyofsubsidizeddaycare.EuropeanJournalofPoliticalEconomy12(3),443±458.Besley,T.,Coate,S.,1991.Publicprovisionofprivategoodsandtheredistributionofincome.AmericanEconomicReview81(4),979±984.Besley,T.,Coate,S.,1997.Aneconomicmodelofrepresentativedemocracy.QuarterlyJournalofEconomics112(1),85±114.Blackorby,C.,Donaldson,D.,1988.Cashversuskind,self-selectionandef®cienttransfers.AmericanEconomicReview78(4),691±700.Blomquist,S.,Christiansen,V.,1995.Publicprovisionofprivategoodsasaredistributivedeviceinanoptimumincometaxmodel.ScandinavianJournalofEconomics97(4),547±567. JournalofPublicEconomics73(1999)31Blomquist,S.,Christiansen,V.,1998a.Toppinguporoptingout?Theoptimaldesignofpublicprovisionschemes.InternationalEconomicReview39(2),399±411.Blomquist,S.,Christiansen,V.,1998b.Pricesubsidiesversuspublicprovision.InternationalTaxandPublicFinance5(3),283±306.Blomquist,S.,Christiansen,V.,1998c.ThePoliticalEconomyofPubliclyProvidedPrivateGoods,DepartmentofEconomics,UppsalaUniversity,Workingpaper1998/14.Boadway,R.,Marchand,M.,1995.Theuseofpublicexpendituresforredistributivepurposes.OxfordEconomicPapers47(1),45±59.Bruce,N.,Waldman,M.,1991.Transfersinkind:Whytheycanbeef®cientandnonpaternalistic.AmericanEconomicReview81(5),1345±1351.Coate,S.,1995.Altruism,thesamaritan'sdilemma,andgovernmenttransferpolicy.AmericanEconomicReview85(1),46±57.Cremer,H.,Gahvari,F.,1997.In-kindtransfers,self-selectionandoptimaltaxpolicy.EuropeanEconomicReview41(1),97±114.Dixit,A.,Londregan,J.,1998.Ideology,tactics,andef®ciencyinredistributivepolitics.QuarterlyJournalofEconomics113(2),497±529.Edwards,J.,Keen,M.,Tuomala,M.,1994.Incometax,commoditytaxesandpublicgoodprovision:Abriefguide.FinanzArchiv51(4),472±487.Epple,D.,Romano,R.E.,1996a.Publicprovisionofprivategoods.JournalofPoliticalEconomy104(1),57±84.Epple,D.,Romano,R.E.,1996b.Endsagainstthemiddle:Determiningpublicserviceprovisionwhenthereareprivatealternatives.JournalofPublicEconomics62(3),297±326.Epple,D.,Romano,R.E.,1998.Competitionbetweenprivateandpublicschools,vouchers,andpeer-groupeffects.AmericanEconomicReview88(1),33±62.Fiorini,M.P.,1981.RetrospectiveVotinginAmericanElections,YaleUniversity,NewHavenCT.Gouveia,M.,1996.Thepublicsectorandhealthcare.InternationalTaxandPublicFinance3(3),329±349.Gouveia,M.,1997.Majorityruleandthepublicprovisionofaprivategood.PublicChoice93(3±4),221±244.Gradstein,M.,Justman,M.,1996.Thepoliticaleconomyofmixedpublicandprivateschooling:Adynamicanalysis.InternationalTaxandPublicFinance3(3),297±310.Guesnerie,R.,Roberts,K.,1984.Effectivepolicytoolsandquantitycontrols.Econometrica52(1),59±86.Hansson,I.,Stuart,C.,1984.Votingcompetitionswithinterestedpoliticians:Platformsdonotconvergetothepreferencesofthemedianvoter.PublicChoice44(3),431±441.Lindbeck,A.,Weibull,J.W.,1987.Balanced-budgetredistributionastheoutcomeofpoliticalcompetition.PublicChoice52,273±297.Lindbeck,A.,Weibull,J.W.,1993.Amodelofpoliticalequilibriuminarepresentativedemocracy.JournalofPublicEconomics51(2),195±209.Lundholm,M.,Ohlsson,H.,1998.Wages,taxesandpubliclyprovideddaycare.JournalofPopulationEconomics11(2),185±204.Nichols,A.L.,Zeckhauser,R.J.,1982.Targetingtransfersthroughrestrictionsonrecipients.AmericanEconomicReviewPapersandProceedings72(1),372±377.Meltzer,A.H.,Richard,S.F.,1985.Apositivetheoryofin-kindtransfersandthenegativeincometax.PublicChoice47(1),231±265.Roberts,K.W.S.,1977.Votingoverincometaxschedules.JournalofPublicEconomics8(3),329±340.Sinn,H.W.,1995.Atheoryofthewelfarestate.ScandinavianJournalofEconomics97(4),495±526.Sinn,H.W.,1996.Socialinsurance,incentivesandrisktaking.InternationalTaxandPublicFinance3(3),259±280.Stern,N.H.,1982.Optimumincometaxationwitherrorsinadministration.JournalofPublicEconomics17(2),181±211.