高微文章job-market-signaling

MichaelSPENCEJOBMARKETSIGNALINGAndrewMickaelSPENCE(Americaneconomist)BornNovember7,1943Montclair,NewJersey, USAFieldMicroeconomicsMarketanalysisUniversityHarvardUniversity (Ph.D.)UniversityofOxford (B.A.)PrincetonUniversity (B.A.)ContributionSignalingtheoryAwardsJohnBatesClarkMedal(1981)NobelMemorialPrizeinEconomics (2001)alongwith GeorgeA.Akerlof and JosephE.Stiglitz,fortheirmarketanalysesinsituationofasymmetryofinformationContent1.Introduction2.Hiringasaninvestmentunderuncertainty3.Applicantsignaling4.Informationalfeedbackanddefinitionofequilibrium5.Propertiesofinformationalequilibrium:anexample6.Informationalimpactofindices7.Conclusion1.IntroductionWhatisthepurposeofthisessay?analyzethephenomenaofsignalingwhichaffectthemarketoutlineamodelinwhichsignalingisimplicitlydefinedDefinethepropertiesofthesignalingequilibria,theinteractionofpotentialsignalsandtheallocativeefficiencyofthemarketWhychoosingtheJobMarketexemple?«Thisessayisaboutmarketsinwhichsignalingtakesplaceandinwhichtheprimarysignalersarerelativelynumerousandinthemarketsufficientlyinfrequentlythattheyarenotexpectedto(andthereforedonot)investinacquiringsignalingreputation»=>itcorrespondexactlytotheJobMarketsituationAlargenumberofpostulantWhicharesufficientlyinfrequentlyinthemarketsothattheyarenotexpectedtoinvestinacquiringsignalingreputation2.HiringasaninvestmentunderuncertaintyLet’sdefinethesituationintheJobMarketThehiringdecisionIdentifytheproductivecapabilitiesofanindividualAssociatealevelawagetothisexpectedcapabilitiesProblemUnknownbeforehiringEvenafter,itcantaketimetoevaluateHiringaspurchasingalotterySPENCEcomparehiringtoalotterywhere« theemployerpaysthecertainmonetaryequivalentofthelotterytotheindividualaswage »Employersareriskneutral=>Wage=individual'smarginalcontributiontothehiringorganization=>Uncertainty=>aplethoraofpersonaldata=>assessmentofthelottery(expectedmarginalcontribution)SPENCEintroduceadistinctioninthekindofdataINDICESvs.SIGNALSIndicesarepersonalattributethatareimmutablyfixed,unalterableorthatdochange,butnotatthediscretionoftheindividual.RaceNationalityAgesexSignalsarealterable,subjecttomanipulationbytheindividual.Hecaninvesttimeandmoneytochangethem.Thereisapossibilityofcontrolinsignals.education3.ApplicantsignalingTheemployerwererisk-neutral.Foreachsetofsignalsandindicesthattheemployerconfronts,hewillhaveanexpectedmarginalproductforanindividualwhohastheseobservableattributes.Thisistakentobetheofferedwagetoapplicantswiththosecharacteristics.Thereisnotmuchthattheapplicantcandoaboutindices.Signals,ontheotherhand,arealterableandthereforepotentiallysubjecttomanipulationbythejobapplicant.Individualsareassumedtoselectsignalssoastomaximizethedifferencebetweenofferedwagesandsignalingcosts.ACriticalAssumption:signalingcostsarenegativelycorrelatedwithproductivityItismostappropriatelyviewedasaprerequisiteforanobservable,alterablecharacteristictobeapersistentlyinformativesignalinthemarket.Signalingcosts:psychic,monetary,time….4.InformationfeedbackandthedefinitionofequilibriumInformationalfeedbackloopSelf-confirmingThesystemwillbestationaryiftheemployerstartsoutwithconditionalprobabilisticbeliefsthatafteroneroundarenotdisconfirmedbytheincomingdatatheygeneratedWeshallrefertothebeliefsasself-confirming.ASignalingEquilibriumGivenanofferedwageschedule,onecanthinkofthemarketasgenerating,viaindividualoptimizingdecisions,anempiricaldistributionofproductivecapabilitiesgivenobservableattributesorsignals,theemployerhassubjectivelyheldconditionalprobabilisticbeliefswithrespecttoproductility,givensignals.thesubjectivedistributionandtheoneimplicitinthemarketmechanismareidentical,overtherangeofsignalsthattheemployeractuallyobservesPropertiesofinformationalequilibria:Anexample1)Aspecificnumericalexample:*Forthetimebeing,indicesplaynopartSupposethattheemployerbelievesthatthereissomelevelofeducation:ifyy*thenproductivityistwowithprobabilityone.ExpectationA:Conditionsonbehavior:SignalingEquilibrium≠WelfareEquilibriumSignalingequilibriuminwhichgroupIIisbetteroffthanno-signaling:2-y*/2>2-q1,thusy*<2q1Ifq1>0.5(groupIIisaminority),thereexistsasignalingequilibriuminwhichthemembersofgroupIIimprovetheirpositionovertheno-signalingcase.Ifq1<0.5(groupIIisamajority),groupIIwillloseduetotheexistenceofsignaling.2)GeneralizethisexampleSignalingequilibriuminwhichgroupIIisbetteroffthanno-signaling:Conditionony*:SohowsmallaminoritygroupIIhastobetohavethepossibilityofbenefitingfromsignalingdependsupontheratioofthemarginalsignalingcostsofthetwogroups3)OtherequilibriainthesystemwithquitedifferentpropertiesExpectationB:Supposethattheemployer’sexpectationsareofthefollowingform:—Ifyy*GroupIIwithprobability1Asbefore,thelevelofythatcouldconceivablybeselectedare:y=oandy=y*GroupIpreferalwaysno-signaling,andGroupIIpreferano-signaling,onlyif:2-y*/2<2-q1y*>2q1Fory=0,thewageis2-q1.So,ify*>2q1,bothgroupsrationallysety=0.Andwehaveanequilibrium.Educationconveysnoinformationinthistypeofequilibrium.Asthereexistsasignalingequilibriuminwhicheveryonesetsy=0,thereisalsoanequilibriuminwhicheveryonesetsy=y*forsomepositivey*.—Ify=y*:GroupIwithprobabilityqGroupIIwithprobability1-qAsbefore,thelevelofythatcouldconceivablybeselectedare:y=oandy=y*Fory=0,thewageis1,whilethewagefory=y*is2-q.So,ify*<1-q,bothgroupsrationallysety=y*.Again,educationlevelconveysnousefulinformation,butinthisequilibriumindividualsarerationallyinvestingineducation.ExpectationC:ResultAnalterablecharacteristiclikeeducationwhichisapotentialsignal,becomesanactualsignalifthesignalingcostsarenegativelycorrelatedwiththeindividual’sunknownproductivity.But,thenegativecorrelationisanecessarybutnotsufficientconditionforsignalingtotakeplaceEffectivesignalingdependsnotonlyuponthenegativecorrelationofcostsandproductivities,butalsoupontherebeingasufficientnumberofsignalswithintheappropriatecostrange.6.TheinformationalimpactofindicesIndicesareattributesoverwhichonehasnocontrol,likegender,race,etc.Thinkofthemasunalterableattributesofsomething,notnecessarilyaperson.Signalsarethingsonedoesthatarevisibleandthatareinpartdesignedtocommunicate.Inasensetheyarealterableattributes.Assumption:Withineachgroupthedistributionofproductivecapabilitiesandtheincidenceofsignalingcostsarethesame.DataofthemodelIfsexandproductivityareuncorrelatedinthepopulation,sexcouldnevertelltheemployeranythingaboutproductivity.Prediction:Itmustbethoughitsinteractionwiththeeducationalsignalingmechanismthatsexistohaveanyinformationalimpact.Howcouldsexhaveaninformationalimpactonthemarket?Lookbacktothesignalingfeedbackloop:Asthereareexternalitiesinthismodel,theopportunitysetsofmenandwomenofcomparableproductivityarenotnecessarilythesame.OnepossibleequilibriumIfWandy2–y*Productivity2choosey=y*:2–y*/2>1-0Therequiredequilibriumconditionsony*Wandy*Mare:11-0because0W-(y*m)/2theny*m>2q17.ConclusionSignalingisasolutionforasymmetricinformationproblem.Signalingwilltakeplace,ifthesignalingcostsarenegativelycorrelatedwiththeindividual’sunknownproductivityandthereisasufficientnumberofsignalswithintheappropriatecostrange.Therearemanytypesofsignalingequilibriumwhenthegroupshavedifferentproperties.Indices,likesex,can’taffectseparatelythejobmarket.Theinformationalimpactofindicesisthroughtheinteractionwiththesignalingmechanism.Thankyou!