ContentslistsavailableatScienceDirect
Journal
of
Economic
Behavior
&
Organization
jo u r n al ho me p ag e :ww w . e l s e v i e r . c o m / l o c a t e / j e b o
Time-varying
fairness
concerns,
delay,
and
disagreement
in
bargaining
Emin
Karagözo˘glu
a,b,∗,
Kerim
Keskin
caBilkentUniversity,Turkey bCESifo-Munich,Germany cKadirHasUniversity,Turkey
a
r
t
i
c
l
e
i
n
f
o
Articlehistory:
Received18September2017 Receivedinrevisedform 27November2017 Accepted3January2018 Availableonline9January2018 JELclassification: C72 C78 D03 Keywords: Bargaining Deadlineeffect Delay Disagreement Fairnessconcerns Justicesensitivity
a
b
s
t
r
a
c
t
Westudyanalternating-offers,bilateralbargaininggamewhereplayersmayderive disu-tilityfromacceptingsharesbelowwhattheydeemasfair.Moreover,weassumethatthe valuestheyattachtofairness(i.e.,theirsensitivitytoviolationsoftheirfairnessjudgments) decreaseovertime,asthedeadlineapproaches.Ourresultsofferanewexplanationto delaysanddisagreementsindynamicnegotiations.Weshowthatevenmutually compati-blefairnessjudgmentsdonotguaranteeanimmediateagreement.Wepartiallycharacterize conditionsfordelayanddisagreement,andstudythechangesinthelengthofdelayin responsetochangesinthemodelparameters.
©2018ElsevierB.V.Allrightsreserved.
1. Introduction
Delays in reaching an agreement and stalemates are ubiquitous features of real life bargaining encounters. Labora-tory experiments on bargaining games report similar observations (see Roth et al., 1988; Babcock et al., 1995; Babcock and Loewenstein, 1997; Gächter and Riedl, 2005; Karagözo˘glu and Riedl, 2015; Karagözo˘glu and Kocher, 2016; among others). Roth et al. (1988), who observed subjects’tendency to reach agreements towards the deadline, labeled this phe-nomenon as the deadline effect.1 Babcock et al. (1995), Babcock and Loewenstein (1997), Gächter and Riedl (2005), and
Karagözo˘gluandRiedl(2015)observedthatdelaysinreachingagreementsanddisagreementsarepositivelycorrelatedwiththe incompatibilitybetweenthebargainers’fairnessjudgments.Basedontheseobservations,theyinformallyarguedthat fairmind-edness, combined withbiasedandincompatiblefairness judgments,couldbeone ofthereasons behind delays,last-minute agreements,andevendisagreements.Standardbargainingmodelswithcommonknowledgeofrationality,complete/perfect
infor-∗ Correspondingauthorat:BilkentUniversity,DepartmentofEconomics,06800Bilkent,Ankara,Turkey. E-mailaddresses:karagozoglu@bilkent.edu.tr(E.Karagözo˘glu),kerim.keskin@khas.edu.tr(K.Keskin).
1Thisisalsofrequentlyobservedoutsidethelab.Forinstance,laborunionsandemployersusuallyreachagreementsjustbeforeastrikestarts,a
phenomenonknownasthe“eleventhhourdeal”.Settlementsinpre-trialnegotiationsareusuallyreachedjustbeforethecourtdate.Similarly,in2011and 2013,DemocratsandRepublicansreachedanagreementinthenegotiationsabouttheU.S.debtceiling,justbeforethedeadline.
https://doi.org/10.1016/j.jebo.2018.01.002 0167-2681/©2018ElsevierB.V.Allrightsreserved.
mation,anddiscountingfailtocapturethesephenomena.Inthosemodels(seeRubinstein,1982),theplayerwhomakesthefirst offerusesalltherelevantinformationandmakesanofferthattheotherplayercannotreject.Thisresultsinanimmediateagreement. Delaysanddisagreementsinnegotiationscanhaveimportanteconomicimplications.Forinstance,strikesorlockoutsdueto disagreementsbetweenlaborandmanagement(e.g.,theNBAlockoutinthe2011–2012season,theUSAirwaysunionworkers’ strikein2014),continuingconflictsbetweencountries(e.g.,theCyprusconflict,theIsraeli-Palestinianconflict),postponeddecisions duetostalematesinpoliticalnegotiations(e.g.,delaysingovernmentformationduetolengthycoalitionnegotiationsinBelgiumin 2011andintheNetherlandsin2017),mergerandacquisitionnegotiations(e.g.,theVerizon–Yahoomerger),orlitigationnegotiations (e.g.,thePennzoilvs.Texacocase)implygreatwelfarecostsnotonlyforthenegotiatingpartiesinvolvedbutalsoforthethirdparties whohaveinterestsconnectedtothenegotiation(e.g.,fans,tvbroadcastingcompanies,sportsbrandsintheNBAlockoutexample). Therefore,understandingthefactorsthatcausebargainingdelaysanddisagreementsisofutmostimportance.Overthelastthree decades,bargainingscholarsofferedvarioustheoreticalexplanationsfordelayanddisagreements.2Asymmetricinformationorlack
ofcommonpriorsonimportantbargainingcharacteristics(e.g.,discountfactors,outsideoptions),stochasticallyevolvingmodel parameters(e.g.,stakesize),oruncertaintyregardingplayertypesandreputationalconcernsarethemainreasonsforequilibrium delayanddisagreementinthosepapers.
Inthecurrentpaper,westudyanalternating-offersbargaininggamewithaknowndeadlinebetweentwoplayerswhereplayers mayderivedisutilityfromacceptingsharesbelowwhattheydeemasfair.Moreprecisely,eachplayerhassomesubjectivejudgment regardingthefairdivisionofthepie,whichiscommonknowledge;andifshereceivesasharebelowwhatshedeemsasfair,then shederivesdisutilityfromthat.Anaturalexampleaddressedinvariousbargainingexperimentsisasfollows(seeGantneretal., 2001;BirkelandandTungodden,2014;Karagözo˘gluandRiedl,2015;BoltonandKaragözo˘glu,2016):Supposethattwoplayers,Alan andBetty,exertedeffortstojointlyproducethepietheyarebargainingover.Alan’seffortproduced70%ofthepie,whereasBetty’s effortproduced30%ofthepie.Now,Alanbelievesthattheappropriatejusticenormisequityandhencethepieshouldbedivided inproportiontotheircontributions(i.e.,70–30division);whereasBettybelievesthattheappropriatejusticenormisequalityand hencethepieshouldbedividedequally(i.e.,50–50division).Eachplayerwouldexperiencedisutilityfromreceivingsharesbelow theirfairshares(70vs.50,respectively).
Importantly,weallowtheweightstheplayersattachtofairnessconcerns(i.e.,theirsensitivitytoviolationsoftheirfairness judgments)todiminishovertime.Accordingly,asthedeadlineapproaches,theplayerscarelessandlessaboutfairnessandbecome moreandmorematerial-gain-oriented.Thismodelingchoiceisinspiredbytherecentexperimentalfindings,whichhighlightthe primacyofeconomicconcernsoverfairnessconcerns—especiallyundertime/cognitivepressure(seeMooreandLoewenstein,2004; Knochetal.,2006;KnochandFehr,2007;Halalietal.,2013;Hochmanetal.,2015;amongothers)andthetemporalinstability ofjusticesensitivity(seeFortinetal.,2016foranexcellentreview).Furthermore,itisalsoinlinewiththehabituationandcostly self-controlargumentsinthetheoreticalliterature,whichwouldimplythatanagentwhorepetitivelyreceivesastimulithatis cognitivelydisturbingorcostlytohandlestartstobecomelessresponsive(seeKaragözo˘glu,2014forhabituation;andFudenberg andLevine,2006;Dreberetal.,2016forcostlyself-control).
We,first,analyticallyshowthatdelayand/ordisagreementmayexistintheequilibriumofthisbargaininggame.Weprovide necessaryandsufficientconditionsfortheexistenceofdelayanddisagreement.Ourresultsshowthatwhetherplayerswillbeableto reachanagreementandifsowhendependonmultiplefactorssuchasplayers’fairnessjudgments,thepaceofthedecreaseintheir weightsforfairness,theidentitiesofplayers(i.e.,proposerorresponder)indifferentperiods,andthehorizonlength.Furthermore, oursimulationresultsshowthatourmodelcanproduceempiricallyrelevantoutcomepatternssuchasU-shapedorJ-shaped distributionofagreementtimes,whichareusuallyobservedindynamicbargainingexperiments(seeRothetal.,1988;Gächterand Riedl,2005;Karagözo˘gluandRiedl,2015;Sullivan,2016).Ourcomparativestaticanalysesrevealnon-trivialinteractionsbetween themodelparametersandbargainingdelay.Forinstance,moredemandingfairnessjudgmentsorstrongerconcernsforfairness mayleadtofasteragreements.Finallyandsomewhatsurprisingly,weshowthatmutuallycompatiblefairnessjudgmentsdonot guaranteeanimmediateagreement.
Itisworthwhileemphasizingthatalltherelevantinformationiscommonknowledgeinourmodel.Thereisnoincompleteor imperfectinformation.Playersareperfectlyforward-looking.3Ourmodelisthefirsttoexplainbothdelayanddisagreementina
finite-horizon,alternatingoffersbargainingframeworkwithoutresortingtousualincompleteorimperfectinformation assump-tions.Moreover,itisthefirstmodeltoformallystudytime-varyingfairnessconcerns.Tothebestofourknowledge,Birkelandand Tungodden(2014)presentthecloseststudytooursinthattheyalsoutilizefairnessconcerns.TheseauthorsworkonastaticNash bargainingmodel,withoutanytime-varyingcomponentinfairnessjudgments.Naturally,theirmodelcanexplaindisagreements, butnotdelay.
Theorganizationofthepaperisasfollows.InSection2,weintroduceourmodel.InSection3,wefirstpresentouranalytical resultsontheexistenceofequilibriuminvolvingdelayanddisagreement.Later,wepresentsimulationresultsontherelationships betweenthemodelparametersandthelengthofdelay.Section4concludes.
2. Themodel
Weinvestigateatwo-playerfinite-horizonalternatingoffersbargaininggamealaRubinstein(seeRubinstein,1982).LetN={1, 2}bethesetofplayerswhobargainoveradivisiblepie(withanormalizedsizeof1)forafinitenumberofperiodsT.Inperiodt=1, player1proposesadivisionofthepiechoosingastrategy(anoffer)from[0,1].Observingplayer1’soffer,player2chooseswhether toacceptorreject:{a,r}.Ifsheacceptstheoffer,thentheproposeddivisionisimplementedandthegameends.Incasesherejects
2 SeeRubinstein(1985),AdmatiandPerry(1987),GulandSonnenschein(1988),Cho(1990),Cramton(1991),FershtmanandSeidmann(1993),Maand
Manove(1993),JehielandMoldovanu(1995),MerloandWilson(1998),AbreuandGul(2000),Bac(2000),Yildiz(2004),FeinbergandSkrzypacz(2005), Ali(2006),SimsekandYildiz(2014),Romm(2016),Fanning(2016)amongothers.
theoffer,thegameproceedstothenextperiod.Inperiodt=2,itisplayer2’sturntomakeanoffer,choosingastrategyfrom[0, 1],afterwhichplayer1istodecidewhethertoaccept(a)orreject(r)theoffer.Ifplayer1accepts,thentheproposeddivisionis implementedandthegameends;ifsherejects,thenthegameproceedstoperiod3inwhicheverythingfollowsasinthefirstperiod. ThisbargainingprocedurecontinuesuntiloneoftheplayersacceptsanofferoruntilperiodTends.Incasetheformerhappens,each playergetssomeutilitybasedontheagreeddivision,whereasifthelatterhappens,thenthegameendswithadisagreementwhich yieldsautilityof0tobothplayers.
Players have fairness concerns. More precisely, each player i∈N believes that it is fair for her to receive a share ϕi∈[0,1]of thepie.If playerigetsashare xi<ϕi,thensheexperiences adecreaseintheutilityderivedfromthematerial
gain of having xi∈[0, 1].4 This type of preferences is commonly represented with an additively separable utility
func-tion in the literature (see Bolton, 1991; Fehr and Schmidt, 1999; Bolton and Ockenfels, 2000). The novel feature of our model isthe introduction of time-varying fairnessconcerns. Accordingly,each utilityfunction has different weightsfor the material gain and thedisutility comingfrom an unfair division, and theweight (relative importance) for fairness concerns weakly decreases through time. We normalize the weight of the material gain inboth utility functions to 1, and we let ˛i:{1,...,T}→[0,∞)beanon-increasingfunctionofplayeri’srelativeweightsforherfairnessconcernsateachperiod.Ifplayers
agreeonanallocation(x,1−x)atsomeperiodt∈{1,...,T},theirutilitiesfromsuchanagreementaregivenby: u1((x,1−x),t)=x−˛1(t)·max{ϕ1−x,0}
u2((x,1−x),t)=1−x−˛2(t)·max{ϕ2+x−1,0}
Playersdiscountfuturepayoffswithıi∈(0,1).Thismeansthattheutilitiesgivenabovearediscountedbyarateofıit−1when
evaluatedinthefirstperiod.Alltherelevantinformation(e.g.,fairnessconcerns,theinitialvaluesandtheevolutionoftheweights attachedtofairness,thediscountfactors,functionalforms,etc.)iscommonknowledge.WedenotethisgamebyT.
3. Theresults
Inwhatfollows,wefirstpresenttheanalyticalresultsontheexistenceofequilibrium,delay,anddisagreements.Later,wepresent simulationresultsonthelengthofdelay.
3.1. Analyticalresults
Throughoutthepaper,anequilibriummeans(purestrategy)subgameperfectNashequilibrium.Wedenotetheoffermadeby player1atsomeperiodtbyxtandtheoffermadebyplayer2atsomeperiodtbyyt.Asaconvention,nomatterwhotheproposeris, anofferalwaysindicatestheshareofplayer1.Toputitdifferently,player1proposesthesharehewantstoreceive,whereasplayer 2proposestheshareshewantstogive.Ourfirstpropositionconcernstheexistenceofequilibrium.
Proposition1. LetTbeaT-period,bilateralbargaininggamewithtime-varyingfairnessconcerns.AnequilibriumofTexists. Proof. Tisafinite-horizon,completeinformation,extensiveformgamewithcontinuousutilityfunctionsandcompactstrategy
setsateachdecisionnode.Theexistenceresultaccordinglyfollows.
Itisworthmentioningthattheremayexistmultipleequilibriaandthatallequilibriaarepayoffequivalent:utilizingbackward induction,onecanseethatifaplayerpreferstomakeanofferthatcannotberejected,thenthereexistsauniqueoptimaloffer;but ifaplayerpreferstomakeanofferthatwillberejected,thenthereexistsinfinitelymanyoptimaloffers.Thelatterwouldleadtoa multiplicityofequilibria,whereastheformerguaranteesthatallequilibriaarepayoffequivalent.Notethattheplayersmaynotbe abletosimplyagreeontheequilibriumpayoffsinthefirstperiodsincethosepayoffsmaynotbefeasibleinthefirstperioddueto higherweightstheyattachtofairnessatthebeginningofthegame.
Oursecondpropositionconcernstheexistenceofdisagreementsanddelayedagreementsontheequilibriumpath.Forthat purpose,itisenoughtoanalyzeatwo-periodbargaininggame.
Proposition2. Let2beatwo-period,bilateralbargaininggamewithtime-varyingfairnessconcerns.Then,(i)thereexistsanequilibrium
of2withadisagreementand(ii)thereexistsanequilibriumof2withadelayedagreement(i.e.,anagreementinthesecondperiod). Proof. SeeAppendixA.
ThecorollarybelowshowsthatthesameresultsfollowforanyT-periodbargaininggame.
Corollary1. LetTbeaT-period,bilateralbargaininggamewithtime-varyingfairnessconcerns.Then,(i)thereexistsanequilibriumof
Twithadisagreementand(ii)thereexistsanequilibriumofTwithadelayedagreement(i.e.,anagreementsometimeafterthefirst
period).
Proof. TheproofsimplyfollowsfromProposition2sinceonecanaddarbitrarynumberofperiodstothetwo-periodgameanalyzed inProposition2.
4Notethatplayers’fairnessjudgmentsareself-servinginthattheydonotexperiencedisutilityiftheyreceiveasharelargerthantheirfairshare.
Introducingthecounterpartforxi>ϕiisneitheressentialforourmodelnorchangesourresultsqualitatively.Hence,weoptforsimplicity.Satisfaction
withadvantageousinequityincognitivelydemandingsituationsissupportedbyexperimentalresearch,aswell(seeFalkandFischbacher,2006;Vanden Bosetal.,2006).
TheproofofProposition2maygivetheimpressionthattheincompatibilitybetweenplayers’fairnessjudgmentsnecessitatesa delayedagreement,ifnotadisagreement.Bycontrast,Proposition3showsthatevenwhenthereisnosuchconflictbetweenfairness judgments,itispossibletohavedelayedagreements.Thepropositionfurtherindicatesthatdisagreementisimpossibleinthatcase. Proposition3. Let2beatwo-period,bilateralbargaininggamewithtime-varyingfairnessconcerns.Letϕ
1+ϕ2<1,i.e.,thereexists
anallocationwhichsatisfiesbothplayers’fairnessconcerns.Then,(i)theremaystillexistanequilibriumwithadelayedagreementbut(ii) thereexistsnoequilibriumwithdisagreement.
Proof. SeeAppendixA.
Noticethattheformerstatementisaboutapossibility.Whenthereisnoconflictbetweenplayers’fairnessjudgments,an immedi-ateagreementisgenerallyreached;however,theoppositemightalsobetrueinsomecases.Forinstance,incasethebestacceptable offerinthefirstperiodisunfairtooneoftheplayers5andtheweightsattachedtofairnessdecreasedramaticallyinthenextperiod,
thenthatplayermayprefertoproceedtothenextperiodinwhichshemayfindsuchanunfairoffermoreacceptableduetohaving alowerweightforfairnessconcerns.Thelatterstatementinthepropositionisbasedonthefactthatifplayers’fairnessjudgments arecompatible,eithertheyimmediatelyagreeoriftheydonot,theremustbeabetteralternativeinthefuture,whichcannotbea stalemate.
Proposition4presentsanecessaryandsufficientconditionfortheexistenceofadisagreementinequilibrium.Infact,itboils downtogeneralizingthecondition(*)givenintheproofofProposition2.
Proposition4. LetTbeaT-period,bilateralbargaininggamewithtime-varyingfairnessconcerns.AnequilibriumofT involves
disagreementifandonlyif 1+˛2(T)·(1−ϕ2)
1+˛2(T) <
˛1(T)·ϕ1
1+˛1(T). Proof. SeeAppendixA.
Theleft-hand-sideoftheinequalityinthepropositionistheshareofthepiewhichyieldsautilityof0toplayer2inperiodT, therebymakingplayer2indifferentbetweenagreementanddisagreement,whereastheright-hand-sideistheshareofthepiewhich yieldsautilityof0toplayer1inthesameperiod.Accordingly,theleft-hand-siderepresentsthehighestofferplayer2iswillingto acceptinperiodTandtheright-hand-siderepresentsthelowestofferplayer1iswillingtomakeinthesameperiod.Incasethe formerislessthanthelatter,thereexistsnoofferthatisacceptabletobothplayers.Thisimpliesadisagreement.Notethatifsucha conditionissatisfiedinperiodT,similarconditionswouldbesatisfiedinanyperiodt<T.
3.1.1. TheLengthofDelay:APartialCharacterization
Anaturalquestionatthispointis:Canweidentifythelengthofdelay(orequivalently,theagreementperiod)?Itturnsoutthat writingthelengthofdelayasafunctionofthemodelparametersisfarfromtrivial;andhence,isleftforfutureresearch.Herewe takeanintermediatestep.Inordertopartiallycharacterizetheagreementperiod,wemakeuseofthefollowinginterpretation: Considerafinitehorizon,alternatingoffersbargaininggamewithnofairnessconcerns(i.e.,withstandardutilityfunctions).Inthe firstperiod,playersbargainoverasetwithalinearfrontierpassingfromtheutilitypairs(0,1)and(1,0).6Asamatteroffact,a
similarinterpretationfollowsforeachperiodwiththedifferencethatthebargainingsetinperiodtisalwayslargerthanthatin periodt+1foreveryt∈Nduetothediscountfactorıi∈(0,1)(seeFig.1).Thisisthereasonwhythereisanimmediateagreement
inthestandardmodel:supposethatplayersagreeinsomeperiodt>1andgetdiscountedutilitiesof ¯u1and ¯u2,wecaneasilysee
thatplayer1couldoffer ¯u2toplayer2inthefirstperiodandreceiveahigherutilitythan ¯u1;acontradiction.
Inourmodelwithtime-varyingfairnessconcerns,thereisanumberofpossibilitiesregardinghowtheabove-describedbargaining setschangeacrossperiods.Forinstance,ifthereisaconflictbetweenplayers’fairnessconcernsandif˛i(t)issufficientlyhighforeach
player,thenitmaybethatthebargainingsetinperiodtdoesnotincludeautilitypairgreaterthanorequalto(0,0),invectorterms (seeS1inFig.2).Insuchaperiod,weknowforsurethatthereexistsnoagreement.Ontheotherhand,when˛
i(t)issufficientlylow
foreachplayer,thenthereexistallocationsyieldingpositiveutilitiestobothplayers(seeS2inFig.2).Notethatifthisisthecasein
someperiodt,thenforeveryperiodt>t,therealwaysexistallocationsyieldingpositiveutilitiestobothplayers.Furthermore,note thatduetotime-varyingfairnessconcerns,wecannotsimplysaythatthebargainingsetinperiodtisalwayssmallerorlargerthan thebargainingsetinperiodtforsomeperiodst,t>t.Thereasonisthatalthoughthediscountfactorshrinksthebargainingsetin
thenextperiod,non-increasingfairnessconcernsexpandthebargainingset.Andtheseparametersmaybesuchthatthebargaining setsinperiodtandtdonotcontainoneanother(seeS3,S4,andS5inFig.2).
Becauseofthesedifferentformseachbargainingsetmaytake,itisaverydifficulttasktoidentifytheagreementperiodinclosed form.Thisisthereasonwhyweturntoapartialcharacterization.Westartwiththefollowingremark.
Remark1. LetStdenotethebargainingsetinperiodt.IfS1doesnotincludeautilitypairgreaterthanorequalto(0,0),i.e.,iffor
everyx∈[0,1],
x−˛1(1)·(ϕ1−x)<0 and 1−x−˛2(1)·(ϕ2+x−1)<0,
5 Althoughthereexistallocationswhichbothpartiesdeemasfair,an‘optimal’offermightbeunfairtooneofthepartiesduetotheasymmetrycreated
bytheidentityoftheproposerinthefinalperiod.
6 Althoughthemodelrestrictsplayerstochooseautilitypaironthefrontier,wecanalsoassumethatplayersareabletochoosefromthewholebargaining
Fig.1.Apossiblecollectionofbargainingsetsinthestandardmodel.
thenthereexistsdelayordisagreement.
Fromnowon,weavoidtrivialcasesandonlyconsidertheonesinwhichS1includesatleastoneutilitypairgreaterthan(0,0),
invectorterms.Inthefollowing,wedefinetwoconceptsaboutthecorrespondingbargainingsetsineachperiodandusethem toprovideadditionalconditionsforhavingadelayedagreement.First,wedefinedominationinbargainingsetsanddescribethe conditionsforSt
⊂St.
Definition1. LetTbeaT-period,bilateralbargaininggamewithtime-varyingfairnessconcerns.Aperiodtdominatesperiodt /=t
inbargainingsetsifSt ⊂St.
Thefollowingobservationsummarizestheconditionsforthebargainingsetinperiodttodominatethebargainingsetinperiod t /=t.
Observation1. LetTbeaT-period,bilateralbargaininggamewithtime-varyingfairnessconcerns.Considertwoperiodstandtsatisfying
u1
1−˛2(t)·ϕ2 1+˛2(t), ˛2(t)·ϕ2 1+˛2(t) ,t >ıt−t 1 ·u1 1−˛2(t)·ϕ2 1+˛2(t), ˛2(t)·ϕ2 1+˛2(t) ,t u2 ˛ 1(t)·ϕ1 1+˛1(t), 1−˛1(t)·ϕ1 1+˛1(t) ,t >ıt−t 2 ·u2 ˛ 1(t)·ϕ1 1+˛1(t), 1−˛1(t)·ϕ1 1+˛1(t) ,t. If 1−˛2(t)·ϕ2 1+˛2(t)<ϕ1 and 1−˛1(t)·ϕ1 1+˛1(t)>ϕ2,thenperiodtdominatesperiodtinbargainingsets;andif 1−˛2(t)·ϕ2
1+˛2(t)>ϕ1
or1−˛1(t)·ϕ1
1+˛1(t) >ϕ2,
thenperiodtdominatesperiodtinbargainingsetsonlywhent<t.
Thisobservationisbasedonthefactthatdominationinbargainingsetscanbeidentifiedusingtheendpointsofthecorresponding bargainingfrontiers.Thelatterpartoftheobservationindicatesthatwhenthereexistsnostrongconflict(i.e.,whenthereexistsa utilitypairwhichsatisfiesthefairnessconcernofoneplayerandyieldsapositiveutilityfortheotherplayer),onlyanearlierperiod candominateaparticularperiodinbargainingsets.7Utilizingthisobservation,wepresentthefollowingnecessaryconditionfora
delayedagreement.
Remark2. Ifforeveryi∈{1,2}andeveryt∈{1,...,T}, ˛i(t)
˛i(1)>ı t−1 i ,
7RemembertheresultinProposition3:evenwhenthereisnoconflictinplayers’fairnessconcerns,t=1doesnotnecessarilydominatethefollowing
Fig.2. Apossiblecollectionofbargainingsetsinourmodel.
thent=1dominatesalltheotherperiodsinbargainingsets.Accordingly,thereisanimmediateagreementinanyequilibrium ofT.Inordertohaveadelayedagreementinequilibrium,theinequalityabovemustbeviolatedforsomei∈{1,2}andsome
t∈{1,...,T}.
Inotherwords,iftheinequalityinRemark2issatisfiedforbothplayersandforanyperiod,thenthediscountfactorsaresufficiently lowsuchthatdiscountingshrinksthebargainingsetsinthefollowingperiodsmorethannon-increasingfairnessconcernsexpands thebargainingsetintherespectiveperiod.Thisimpliesthatthefirstperioddominatesalltheotherperiodsinbargainingsets,so thatplayerswouldbebetteroffagreeinginthefirstperiod.
Below,weprovidethedefinitionofdominationthroughtheunionofbargainingsets,whichwillbeusefulin(partially)identifying theagreementperiod.
Definition2. LetTbeaT-period,bilateralbargaininggamewithtime-varyingfairnessconcerns.Acollectionofperiodst1,...,tk
dominatesanotherperiodtthroughtheunionofbargainingsetsif St⊂
i∈{1,...,k}S ti
.
Furthermore,aperiodtisundominated,ifnocollectionofperiodsdominatesperiodtthroughtheunionofbargainingsets. Now,wearereadytodescribeapropertyoftheagreementperiodandanothersufficientconditionfordelay.
Remark3. InaT-periodbargaininggame,theagreementperiodt*isundominated.8Thus,ifthefirstperiodisdominatedbysome
collectionofperiods,thenthereexistsdelayordisagreement;andifthereexistsauniqueperiod ˆtwhichisundominated,then ˆtis theagreementperiod.
Asmentionedabove,theremaybemultipleundominatedperiodsinthepresenceoftime-varyingfairnessconcerns.Inthatcase, inwhichperiodplayerswillagreedependsonmultiplefactorssuchastheidentityoftheproposersintheundominatedperiods, theshapeofbargainingsetsintheundominatedperiods,theendpointsofbargainingfrontiersintheundominatedperiods,etc.For suchsituations,wearguethatsolvingfortheequilibriumistheonlywaytodeterminetheagreementperiod.
3.1.2. TheLengthofDelay:AnIllustrativeExample
Wenowprovideathree-periodexampleinwhichwecharacterizethesubgameperfectNashequilibrium.Thisistounderstand howtheequilibriumconditionsinourmodelappearandhowthevaluesofmodelparametersdeterminewhethertherewillbe anagreementandifsowhen.Notethat(i)anyT-periodgamecanbeanalyzedinasimilarmannerforanyoddT ∈Nand(ii)the equilibriumanalysisbelowcanbeextendedtothecaseswhereTiseven.
Example.Considerthebargaininggame3andassumethatthedisagreementconditionpresentedinProposition4isnotsatisfied.
AlthoughT=3,westicktogeneralnotationinthefollowinganalysis.
ConsiderperiodTinwhichplayer1istheproposer.Theanalysisforheroptimalofferhasalreadybeenmadeintheproofof Proposition4.Giventheassumptionof“nodisagreement”specifiedabove,player1offers
x∗ T=
1+˛2(T)·(1−ϕ2)
1+˛2(T)
whichwillbeacceptedbyplayer2.
ConsiderperiodT−1inwhichplayer2istheproposer.Sheanticipatesthatifherofferisrejected,thenshewillreceiveautility of0inperiodT,whereasplayer1willreceiveautilityofı1·u1((x∗T,1−x∗T),T).Accordingly,player2makesanoffersoastomake
player1indifferentbetweenaandr,ifsuchanofferyieldsplayer2autilitynotlessthan0.Wedivideourequilibriumanalysisinto sixcases:
Forthesecases,wedefine mT−1=ı1x ∗ T+˛1(T−1)·ϕ1 1+˛1(T−1) and nT−1=ı1· [x∗ T−˛1(T)·(ϕ1−x∗T)]+˛1(T−1)·ϕ1 1+˛1(T−1) . CaseA[ϕ1≤ı1x∗T≤x∗Tandu2((ı1xT∗,1−ı1x∗T),T−1)≥0]:
Inthiscase,player1’sofferint=Tisfairforher.Suchanofferyieldsherautilityofx∗
T.Player2makesthefollowingoffer
y∗
T−1=ı1x∗T.
CaseB[ϕ1≤ı1x∗T≤x∗Tandu2((ı1x∗T,1−ı1x∗T),T−1)<0]:
Itturnsoutthatanofferthatwillbeacceptedbyplayer1yieldsplayer2anegativeutility.Hence,player2canmakeanyofferless thanı1x∗T,andanysuchofferwillberejectedbyplayer1.
CaseC[ı1x∗T≤ϕ1≤x∗Tandu2((mT−1,1−mT−1),T−1)≥0]:
Inthiscase,player1’sofferint=Tisfairforher.Suchanofferyieldsherautilityofx∗
T.Player2makesthefollowingoffer
y∗
T−1−˛1(T−1)·(ϕ1−y∗T−1)=ı1x∗T
y∗
T−1=mT−1.
CaseD[ı1x∗T≤ϕ1≤x∗Tandu2((mT−1,1−mT−1),T−1)<0]:
Itturnsoutthatanofferthatwillbeacceptedbyplayer1yieldsplayer2anegativeutility.Hence,player2canmakeanyofferless thanmT−1,andanysuchofferwillberejectedbyplayer1.
CaseE[ϕ1≥x∗Tandu2((nT−1,1−nT−1),T−1)≥0]:
Inthiscase,player1’sofferint=Tisunfairforher.Suchanofferyieldsherautilityofx∗
T−˛1(T)·(ϕ1−x∗T).Player2makesthe
followingoffer y∗ T−1−˛1(T−1)·(ϕ1−y∗T−1)=ı1·[x∗T−˛1(T)·(ϕ1−x∗T)] y∗ T−1=nT−1. CaseF[ϕ1≥x∗Tandu2((nT−1,1−nT−1),T−1)<0]:
Itturnsoutthatanofferthatwillbeacceptedbyplayer1yieldsplayer2anegativeutility.Hence,player2canmakeanyofferless thannT−1,andanysuchofferwillberejectedbyplayer1.
Tosumup,player2’sbestofferisdescribedbythefollowingpiecewiseset-valuedfunction:9
y∗ T−1=
⎧
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎨
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎩
ı1x∗T ifCaseA mT−1 ifCaseC nT−1 ifCaseE [0,ı1x∗T) ifCaseB [0,mT−1) ifCaseD [0,nT−1) ifCaseFConsiderperiodT−2inwhichplayer1istheproposer.IncasesA,C,andEabove:player1anticipatesthatifherofferisrejected, thenshewillreceiveautilityof
UT−1≡ı1·u1((y∗T−1,1−y∗T−1),T−1),
whereasplayer2willreceiveautilityof ı2·u2((y∗T−1,1−y∗T−1),T−1).
Accordingly,player1makesanoffersoastomakeplayer2indifferentbetweenaandr,ifsuchanofferyieldsplayer1autilitynot lessthanUT−1.Wedivideourequilibriumanalysisintosixcases:
Forthesecases,wedefine kT−2=ı2·(1−y∗T−1), mT−2= 1+˛2(T−2)·(1−ϕ2)−ı2·(1−y∗T−1) 1+˛2(T−2) and nT−2= 1+˛2(T−2)·(1−ϕ2)−ı2·[1−y∗T−1−˛2(T−1)·(ϕ2+y∗T−1−1)] 1+˛2(T−2) . Case1A[ϕ2≤kT−2≤1−y∗T−1andu1((kT−2,1−kT−2),T−2)≥UT−1]:
Inthiscase,player2’sofferint=T−1isfairforher.Suchanofferyieldsherautilityof1−y∗
T−1.Player1makesthefollowingoffer
x∗
T−2=kT−2.
Case1B[ϕ2≤kT−2≤1−y∗T−1andu1((kT−2,1−kT−2),T−2)<UT−1]:
Itturnsoutthatanofferthatwillbeacceptedbyplayer2yieldsplayer1anundesirableutility.Hence,player1canmakeanyoffer higherthankT−2,andanysuchofferwillberejectedbyplayer2.
Case1C[kT−2≤ϕ2≤1−y∗T−1andu1((mT−2,1−mT−2),T−2)≥UT−1]:
Inthiscase,player2’sofferint=T−1isfairforher.Suchanofferyieldsherautilityof1−y∗
T−1.Player1makesthefollowingoffer
1−x∗
T−2−˛2(T−2)·(ϕ2+x∗T−2−1)=ı2·(1−yT−1∗ )
x∗
T−2=mT−2.
Case1D[kT−2≤ϕ2≤1−y∗T−1andu1((mT−2,1−mT−2),T−2)<UT−1]:
Itturnsoutthatanofferthatwillbeacceptedbyplayer2yieldsplayer1anundesirableutility.Hence,player1canmakeanyoffer higherthanmT−2,andanysuchofferwillberejectedbyplayer2.
Case1E[ϕ2≥1−y∗T−1andu1((nT−2,1−nT−2),T−2)≥UT−1]:
Inthiscase,player2’sofferint=T−1isunfairforher.Suchanofferyieldsherautilityof1−y∗
T−1−˛2(T−1)·(ϕ2+y∗T−1−1).
Player1makesthefollowingoffer 1−x∗
T−2−˛2(T−2)·(ϕ2+x∗T−2−1)=ı2·[1−y∗T−1−˛2(T−1)·(ϕ2+y∗T−1−1)]
x∗
T−2=nT−2.
Case1F[ϕ2≥1−y∗T−1andu1((nT−2,1−nT−2),T−2)<UT−1]:
Itturnsoutthatanofferthatwillbeacceptedbyplayer2yieldsplayer1anundesirableutility.Hence,player1canmakeanyoffer higherthannT−2,andanysuchofferwillberejectedbyplayer2.
IncasesB,D,andFabove:player1anticipatesthatifherofferisrejected,thenshewillreceiveautilityof VT−1≡ı21·u1((x∗T,1−x∗T),T),
whereasplayer2willreceiveautilityof0.Accordingly,player1makesanoffersoastomakeplayer2indifferentbetweenaandr,if suchanofferyieldsplayer1autilitynotlessthanVT−1.Wedivideourequilibriumanalysisintotwocases:
Forthesecases,wedefine pT−2=1+˛12(+T˛−2)·(1−ϕ2)
2(T−2) . Case2A[u1((pT−2,1−pT−2),T−2)≥VT−1]:
Inthiscase,player2’sofferint=T−1isirrelevantandplayer1’sofferinperiodTyieldsplayer2autilityof0.Player1makesthe followingoffer
x∗
T−2=pT−2.
Case2B[u1((pT−2,1−pT−2),T−2)<VT−1]:
Itturnsoutthatanofferthatwillbeacceptedbyplayer2yieldsplayer1anundesirableutility.Hence,player1canmakeanyoffer higherthanpT−2,andanysuchofferwillberejectedbyplayer2.
Tosumup,player1’sbestofferisdescribedbythefollowingpiecewiseset-valuedfunction: x∗ T−2=
⎧
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎨
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎩
kT−2 ifCase1A mT−2 ifCase1C nT−2 ifCase1E pT−2 ifCase2A (kT−2,1] ifCase1B (mT−2,1] ifCase1D (nT−2,1] ifCase1F (pT−2,1] ifCase2BThiscompletesouranalysisofequilibriuminclosedform.Now,weconsiderdifferentsetsofparametervaluesinordertoillustrate thatagreementinanyperiodispossible.Wefirstfix
ϕ1=ϕ2=0.8 and ı1=ı2=0.8
If˛1(T)=˛2(T)=2,thenitfollowsbyProposition4thatwehaveadisagreement.Instead,assumingthat
˛1(T) =˛2(T)=1
˛1(T−1) =˛2(T−1)=2
wehaveadelayedagreementinperiodT=3.Assumingthat ˛1(T) =˛2(T)=2/3
˛1(T−1) =˛2(T−1)=3/4
˛1(T−2) =˛2(T−2)=2
wehaveadelayedagreementinperiodT−1=2.Finally,consideringstandardpreferenceswithnofairnessconcerns,wewouldhave animmediateagreementinperiodT−2=1.♦
3.2. Simulations
Inthissubsection,wereportsimulationresultsidentifyingtheagreementperiodandanalyzehowitrespondstochangesinthe modelparameters.10Forthesakeoftractability,weconsiderthefollowingfunctionalformfor˛
i(·):
˛i(t)=iati
foreveryi∈{1,2},andusethefollowingvaluesforthemodelparameters T ∈ {20,21}
ıi ∈ {0.95}
ϕi ∈ {0.1,0.2,...,0.9,1}
i ∈ {6,24}
ai ∈ {0.1,0.2,...,0.8,0.9}
foreveryi∈{1,2},whichcorrespondstoatotalof64,800numericalresults.11
ThenumberofperiodsTischoseninordertocapturesufficientamountofvariabilityintheagreementperiods.The func-tionalformfor˛i andthecorrespondingparametervaluescoverafairlywiderangeof interestingsituations.For instance,if
ai=0.1and i=6,thenplayer i alwaysvalues materialgain morethanshe valuesfairness; ifai=0.9and i=24,thenplayer
i cares aboutfairnessat the beginningof thegame so much thather materialgain never becomes relativelymore impor-tantthroughoutthegame;andinalloftheremainingcases,fairnessisrelativelymorevaluableforplayeriatthebeginning, butitbecomesrelativelylessvaluablesomewhereduringthegameasthedeadlineapproaches.Theswitchingperiod signif-icantly depends ontheactualvaluesof these parameters:ai governs thespeed of decreaseinplayer i’sweight for fairness
concerns and i determines whethersuch a speed is enough to make thematerial gain relativelymore important before
theend.
10ThesimulationsareconductedinVisualStudiousingC++astheprogramminglanguage.
11Duetospacelimitations,wereportasmallsubsetoftheseresultsinFigs.3and5.Allequilibriumagreementperiodscanbefoundintheonlineappendix
Fig.3.TheagreementperiodswhenT=20.(Boldfacenumbersinrowsandcolumnsarethevaluesforϕ1andϕ2,respectively.)
Toprovideaclearunderstanding,herewerestrictthesetofparametervaluesandreporttheresultsinFig.3.12Asitcanbeseen
inthisfigure,weobserveawiderangeofagreementperiods;withanimmediateagreementwhenplayers’beliefabouttheirfair sharesaresufficientlylow,andwiththeagreementperiodgenerallyincreasingasϕi,ai,oriincreasereachingamaximumvalueof
19whenϕ1=ϕ2=0.8,a1=a2=0.8,and1=1=24.
Below,wepresenttwoobservationsfromoursimulationresults.Thefirstobservationisaboutacommonlyreportedfindingin theexperimental/empiricalliteratureonbargaining:thedistributionofagreementperiodsisU-orJ-shaped,meaningthatplayers mostlyagreeeitherclosetothebeginningorclosetothedeadline,withlatterbeingobservedmorefrequently,buttheyrarelyagree inthemiddleofthegame(seeRothetal.,1988;Kessler,1996;GächterandRiedl,2005;Karagözo˘gluandRiedl,2015;Sullivan,2016; VassermanandYildiz,2016,amongothers).Thisobservationcanbesupportedbyourtheoreticalmodelforsomevaluesofmodel parameters.Forexample,considera1=a2=0.9and1=2=6whenT=20andı1=ı2=0.95.Settingperiods1–7asthebeginningpart
andperiods15–20astheendpartofthegame,andassumingthateachplayer’sbeliefaboutherfairshareisuniformlydistributed over{0.2,...,0.9},13wecanarguethatthedistributionofagreementperiodsisU-shaped(seethechartontheleftinFig.4).Asfor
anotherexample,considera1=a2=0.8,1=2=24,andϕ1,ϕ2∈{0.3,...,0.9}keepingtheotherparametervaluesunchanged.Now,
thedistributionofagreementperiodsbecomesJ-shaped(seethechartontherightinFig.4).
Thesecondobservationisthattheagreementperiodisnotnecessarilymonotonicinaplayer’sbeliefaboutherfairshareϕior
inherfairnessweightparameterai.Thatis,asoneoftheseparametersincreasesforaplayer,itmayturnoutthatplayersreachan
agreementmorequickly.Forexample,fixinga1=a2=0.8and1=2=6,playersagreeinthe15thperiodwhenϕ1=0.8andϕ2=0.5;
whereastheyagreeinthe13thperiodwhenϕ1remainstobe0.8andϕ2increasesto0.6.Althoughsuchcasesarequiteuncommon
inthedomaincoveredbytheparametervaluesweused,theydoexist,preventingustoreachageneralmonotonicityresult. Theintuitionbehindthissomewhatcounterintuitiveresultisasfollows.Considerthenumericalexampleabove.Wecanseevia backwardinductionthatplayerswouldagreeinthe17thperiodyieldinganexpectedutilityof0.3399toplayer2inbothcases.From thatagreementplayer1’sexpectedutilityisequalto0.2727incase(i)ϕ2=0.5andto0.2667incase(ii)ϕ2=0.6.Simply,inthelatter
case,player1hastoofferahighersharetoplayer2inordertomakeherindifferentbetweenacceptingandrejectingtheoffer,so thatplayer1’sownexpectedutilityturnsouttobelower.Now,incase(i),player1hasachancetoofferanamountyielding0.3399 toplayer2inthe15thperiodincreasingherownexpectedutilityto0.2910.Thus,playersagreeinthe15thperiod.Ontheother hand,incase(ii),player1doesnotwishtoofferanexpectedutilityof0.3399toplayer2inthe15thperiod.Thereasonisthatplayer 2’sfairamountisnowhighersothatanunfairdivisionwouldgiveherahigherdisutility(incomparisontotheformercase),sothat makingplayer2indifferentbetweenacceptingandrejectingturnsoutnottobebeneficialforplayer1.Thus,theydonotagreeinthe 15thperiod.Finally,inthe13thperiodofbothcases(i)and(ii),player1istoofferanexpectedutilityof0.3399toplayer2inorderto makeherindifferentbetweenacceptingandrejectingtheoffer.However,incase(i),player1wouldmaketheofferonlyifityields anexpectedutilitynolessthan0.2910toherself;whereasincase(ii),shewouldmaketheofferonlyifityieldsanexpectedutility
12 ForT=21,forwhichthelastproposerisplayer1,theresultsarerelegatedtotheAppendixA. 13 Arguably,casesinwhichϕ
Fig.4.U-orJ-shapedagreements.
nolessthan0.2667toherself.Giventhattheofferthatmakesplayer2indifferentyieldsanexpectedutilityof0.2854toplayer1, theofferismadeincase(ii)leadingtoanagreementinthe13thperiod,butnotmadeincase(i).Asimilarresultaswellasasimilar interpretationarevalidforai.14
4. Conclusion
Westudyafinitehorizonbargaininggamebetweentwoplayerswhovaluenotonlytheabsoluteamounttheyreceivebutalsothe relativeshare.Thenoveltyofourmodelisthattheweightsplayersattachtofairnessdecreaseovertimeasthedeadlineapproaches. Thus,asthepushcomestoshove(i.e.,asthedeadlineapproachesandlosingthemoneyonthetablebecomesmoresalient),players becomemorefocusedonmaterialgainsandlessfocusedonfairness.Thismodelingassumptionisinspiredbytherecentexperimental studiesonhotvs.coldpsychologicalstates,therelativeprimacyofeconomicvs.socialconcerns,timepressure,andthetemporal instabilityofjusticesensitivity.
Ourresultsshedlightontheinfluenceoffairnessjudgmentsondelays,last-minuteagreements,anddisagreementsindynamic bargainingsituations.We,first,analyticallyshowthatdelayordisagreementmayexistintheequilibriumpath.Then,weprovide necessaryandsufficientconditionsfordisagreementanddelay.Wealsofindthatmutuallycompatiblefairnessjudgmentsdonot guaranteeanimmediateagreement.Ouranalyticalresultsshowthatwhetherplayerswillbeabletoreachanagreementandifso whendependsonmultiplefactorssuchasplayers’fairnessjudgments,thepaceofthedecreaseintheirweightsforfairness,the iden-titiesofplayers(i.e.,proposerorresponder)indifferentperiods,andthehorizonlength.Ourmodelproducesempiricallyrelevant outcomepatternssuchasU-shapedorJ-shapeddistributionofagreementtimes,whicharefrequentlyobservedindynamic bargain-ingexperiments.Finally,oursimulationresultsrevealsomeinterestingcomparativestaticresults.Forinstance,moredemanding fairnessjudgmentsorstrongerconcernsforfairnessmayleadtofasteragreements.
Herewepresentedanempiricallyinspiredmodelofbargaining,whichsuccessfullymatchesmultipleregularitiesinthe experi-mental/empiricaldata.Tothebestofourknowledge,thisisthefirstfinite-horizonalternating-offersbargainingmodelthatproduces delayanddisagreementinequilibriumwithoutresortingtousualincompleteorimperfectinformationassumptions.15Obviously,
tobetterunderstandtherelationshipbetweentime-varyingfairnessconcernsanddelay/disagreementinbargaining,experiments particularlydesignedtotestthepredictionsofourmodelshouldbeconducted.
Thisisafirststepinincorporatingtime-varyingfairnessjudgmentsintobargainingmodels.Afurtherstepwouldbetomodelthe micro-foundationsoftime-varyingfairnessjudgments.Toachievethisgoal,webelievethatexperimentsinvestigatingthechangesin brainactivityintheregionsthatreacttounfairness—asaproxyforchangesinthevaluepeopleattachtofairness—duringnegotiations withfMRIstudiesoralternativemethodssuchasskinconductancewouldbenaturalcomplements.
Acknowledgments
Wewouldliketothanktheeditor,twoanonymousreviewers,andGaryCharnessforusefulcommentsandsuggestions.Weare gratefultoHamideTuranforherhelpinwritingthecodeusedforthesimulationresults.Wealsothankaudiencesatthe“Bargaining: TheoryandExperiments”workshopatBilkentUniversityandthe2ndTurkishExperimentalandBehavioralEconomicsWorkshop at ˙IstanbulBilgiUniversity.EminKaragözo˘gluthanksTÜB˙ITAK(TheScientificandTechnologicalResearchCouncilofTurkey)forthe post-doctoralresearchfellowship,andMassachusettsInstituteofTechnology,DepartmentofEconomicsfortheirhospitality.The usualdisclaimersapply.
AppendixA.
ProofofProposition2. Considert=2inwhichplayer2istheproposer.Shemakesanoffer,whichmakesplayer1indifferent betweenplayingaandr.Notingthatplayingrwouldyieldplayer1autilityof0,wefindplayer2’soptimalofferasfollows:
14Asamatteroffact,onecanfindnumericalexamplesforthenon-monotonicityina
iintherestrictedsetofexampleswehavereportedinFigs.3and5. 15Incorporatinguncertaintyintotheevolutionofplayers’fairnesspreferencesovertimemaybeareasonableextension.Nevertheless,itwould
y∗ 2−˛1(2)·(ϕ1−y∗2)=0 y∗ 2= ˛1(2)·ϕ1 1+˛1(2)
Noticethatplayer2wouldnotprefertomakethisoffer,iftheremainingpieamountyieldsheranegativeutility;i.e.,if16
1−y∗
2−˛2(2)·(ϕ2+y∗2−1)<0. (*)
Ifthisisthecase,player2makessomeofferlessthany∗
2,whichwillthenberejectedbyplayer1.So,therewillnotbeanyagreement
int=2.Ifthisisnotthecase,therewillbeanagreementontheallocation(y∗ 2,1−y2∗).
Now,tocompletetheproof,wemustshowthatthebargainingproceedstot=2.Considert=1.Notethatplayer1istheproposer andthatsheanticipatestheoutcomeoft=2.Shemakesanofferthatmakesplayer2indifferentbetweenplayingaandr.Notingthat playingrwouldyieldplayer2autilityof
u2((y∗2,1−y∗2),2)
int=2,wedividetheproblemintotwocases: (i) ϕ2≤1−y∗2 and (ii) ϕ2>1−y∗2.
Incase(i),player2’sutilityis1−y∗
2.Dividethiscaseintotwosubcases:
(i.a) ϕ2≤ı2·(1−y∗2) and (i.b) ϕ2>ı2·(1−y∗2).
Incase(i.a),player1’soptimalofferis: x∗
1=1−ı2·(1−y∗2)
Incase(i.b),player1’soptimaloffercanbefoundbythefollowing: 1−x∗ 1−˛2(1)·(ϕ2+x1∗−1)=ı2·(1−y∗2) x∗ 1= 1+˛2(1)·(1−ϕ2)−ı2·(1−y∗2) 1+˛2(1)
Incase(ii),player2’sutilityis1−y∗
2−˛2(2)·(ϕ2+y∗2−1).Player1’soptimaloffer—whichshouldbeunfairforplayer2—canbe
foundbythefollowing: 1−x∗ 1−˛2(1)·(ϕ2+x∗1−1)=ı2·[1−y∗2−˛2(2)·(ϕ2+y∗2−1)] x∗ 1= 1+˛2(1)·(1−ϕ2)−ı2·[1−y∗2−˛2(2)·(ϕ2+y∗2−1)] 1+˛2(1) Tosumup,ifx∗
1yieldsanegativeutilitytoplayer1int=1foroneoftheabovecases,thenforthatparticularcase,thereis
noagreementinthefirstperiod.Notethatthisispossible,if˛1(1)issufficientlyhigh.Ifthishappensincase(i),thenthereisan
agreementint=2.Ifthishappensincase(ii),thenthereisanagreementint=2iftheequation(*)isnotsatisfied,andthereis disagreementiftheequation(*)issatisfied.
ProofofProposition3. Fortheformer,weprovideanexampleinwhichthereexistsdelay.Consideratwo-periodmodelwith ı1 =ı2=0.9,
ϕ1 =ϕ2=0.4,
˛1(2) =˛2(2)=0.25,and
˛1(1) =˛2(1)=1.
Considert=2inwhichplayer2istheproposer.Shemakesanoffersoastomakeplayer1indifferentbetweenaandr.Thisoffer isy∗
2=0.08,whichyieldsautilityof0toplayer1andadiscountedutilityof0.828toplayer2.Now,considert=1.Anticipatingthe
outcomeoft=2,player1’soffercouldbetotake0.172forherself,andtherebymakingplayer2indifferentbetweenplayingaandr. However,suchanofferwouldyieldautilityof0.172−0.228=−0.056<0.Then,player1wouldprefermakinganofferwhichplayer 2rejects.Hence,inequilibrium,agreementisreachedwithdelay.
Forthelatter,supposethatthereexistsdisagreement.ThisalsomeansthatplayersdidnotagreeuntilperiodT(i.e.,theyreached thefinalperiod).ConsiderperiodT.Withoutlossofgenerality,assumethatplayer1istheproposer.Inthisperiodthereexistsanoffer forplayer1whichyieldsapositiveutilitytoherselfandautilityof0toplayer2,sinceϕ1+ϕ2<1byassumption.Sincedisagreement
16 Thisisnotnecessarilyplayer2’sutility,becauseif1−y∗
2isfairforher,thenherutilitywouldbe1−y∗2.Insuchacase,player2getsapositiveutilityfor
yieldsautilityof0tobothplayers,player1woulddeviatetosuchanofferandplayer2wouldaccept.Thisisacontradiction.Hence, thereisalwaysanagreement.
ProofofProposition4. AssumethatTisodd(forevenT,similarargumentsfollow).ConsiderperiodTinwhichplayer1isthe proposer.Shemakesanofferthatmakesplayer2indifferentbetweenplayingaandr.Notingthatplayingrwouldyieldplayer2a utilityof0,wefindplayer1’soptimalofferasfollows:
1−x∗ T−˛2(T)·(ϕ2+xT∗−1)=0 x∗ T= 1+˛2(T)·(1−ϕ2) 1+˛2(T)
Now,ifsuchanofferisnotbeneficialforplayer1,i.e.,if x∗ T−˛1(T)·(ϕ1−x∗T)<0 x∗ T<˛1 (T)·ϕ1 1+˛1(T)
thenplayer1wouldoffersomethinghigherthanx∗
Tandplayer2wouldreject.
Tosumup,if 1+˛2(T)·(1−ϕ2)
1+˛2(T) <
˛1(T)·ϕ1
1+˛1(T),
thereisnoagreementinthefinalperiod.Noticethatifthisisthecase,thentherewouldnotbeanagreementinthepreviousperiods either.Tobemoreprecise,ifthereisnoallocationthatyieldsautilitypairgreaterthan(0,0)inperiodT,thenthereisnoallocation thatyieldssuchautilitypairinthepreviousperiods.Thisisbecause˛i(t)≥˛i(T)foreveryi∈{1,2}andt<T.Thiscompletesthe
proof.
SimulationresultswhenT=21:
Fig.5. TheagreementperiodswhenT=21.(Boldfacenumbersinrowsandcolumnsarethevaluesforϕ1andϕ2,respectively.)
References
Abreu,D.,Gul,F.,2000.Bargainingandreputation.Econometrica68(1),85–87.
Admati,A.D.,Perry,M.,1987.Strategicdelayinbargaining.Rev.Econ.Stud.54(3),345–364.
Babcock,L.,Loewenstein,G.,1997.Explainingbargainingimpasse:theroleofself-servingbiases.J.Econ.Perspect.11(1),109–126. Babcock,L.,Loewenstein,G.,Issacharoff,S.,Camerer,C.,1995.Biasedjudgmentsoffairnessinbargaining.Am.Econ.Rev.85(5),1337–1343. Bac,M.,2000.Signalingbargainingpower:strategicdelayversusrestrictedoffers.Econ.Theory16(1),227–237.
Birkeland,S.,Tungodden,B.,2014.Fairnessmotivationinbargaining:amatterofprinciple.TheoryDecis.77(1),125–151.
Bolton,G.E.,Karagözo˘glu,E.,2016.Ontheinfluenceofhardleverageinasoftleveragebargaininggame:theimportanceofcredibleclaims.GamesEcon. Behav.99,164–179.
Bolton,G.E.,Ockenfels,A.,2000.ERC:atheoryofequity,reciprocity,andcompetition.Am.Econ.Rev.90(1),166–193. Bolton,G.E.,1991.Acomparativemodelofbargaining:theoryandevidence.Am.Econ.Rev.81(5),1096–1136. Cho,I.K.,1990.Uncertaintyanddelayinbargaining.Rev.Econ.Stud.57(4),575–595.
Cramton,P.,1991.Dynamicbargainingwithtransactioncosts.Manage.Sci.37(10),1221–1233.
Dreber,A.,Fudenberg,D.,Levine,D.K.,Rand,D.,2016.Self-control,socialpreferencesandtheeffectofdelayedPayments.WorkingPaper.MIT. Falk,A.,Fischbacher,U.,2006.Atheoryofreciprocity.GamesEcon.Behav.54,293–315.
Fanning,J.,2016.Reputationalbargaininganddeadlines.Econometrica84(3),1131–1179.
Fehr,E.,Schmidt,K.M.,1999.Atheoryoffairness,competition,andcooperation.Q.J.Econ.114(3),817–868. Feinberg,Y.,Skrzypacz,A.,2005.Uncertaintyaboutuncertaintyanddelayinbargaining.Econometrica73(1),63–91.
Fershtman,C.,Seidmann,D.,1993.Deadlineeffectsandinefficientdelayinbargainingwithendogenouscommitment.J.Econ.Theory60(2),306–321. Fortin,M.,Cojuharenco,I.,Patient,D.,German,H.,2016.Itistimeforjustice:howtimechangeswhatweknowaboutjusticejudgmentsandjustice
effects.J.Organ.Behav.37,S30–S56.
Fudenberg,D.,Levine,D.K.,2006.Adualselfmodelofimpulsecontrol.Am.Econ.Rev.96(5),1449–1476. Gächter,S.,Riedl,A.,2005.Moralpropertyrightsinbargainingwithinfeasibleclaims.Manage.Sci.51(2),249–263.
Gantner,A.,Güth,W.,Königstein,M.,2001.Equitablechoicesinbargaininggameswithjointproduction.J.Econ.Behav.Organ.46(2),209–225. Gul,F.,Sonnenschein,H.,1988.Ondelayinbargainingwithone-sideduncertainty.Econometrica56(3),601–611.
Halali,E.,Bereby-Meyer,Y.,Ockenfels,A.,2013.Isitallabouttheself?Theeffectofself-controldepletiononultimatumgameproposers.Front.Hum. Neurosci.7,240.
Hochman,G.,Ayal,S.,Ariely,D.,2015.Fairnessrequiresdeliberation:theprimacyofeconomicoversocialconsiderations.Front.Psychol.6,747. Jehiel,P.,Moldovanu,B.,1995.Negativeexternalitiesmaycausedelayinnegotiation.Econometrica63(6),1321–1335.
Karagözo˘glu,E.,Kocher,M.,2016.Bargainingundertimepressure,CESifoWorkingPaper#5685.
Karagözo˘glu,E.,Riedl,A.,2015.Performanceinformation,productionuncertainty,andsubjectiveentitlementsinbargaining.Manage.Sci.61(11), 2611–2626.
Karagözo˘glu,E.,2014.Adynamicmodelofdissonancereductioninamodularmind.J.Neurosci.Psychol.Econ.7(1),15–32. Kessler,D.,1996.Institutionalcausesofdelayinthesettlementoflegaldisputes.J.LawEcon.Organ.12(2),432–460.
Knoch,D.,Fehr,E.,2007.Resistingthepoweroftemptations:therightprefrontalcortexandself-control.Ann.N.Y.Acad.Sci.1104,123–134. Knoch,D.,Pasqual-Leone,A.,Meyer,K.,Treyer,V.,Fehr,E.,2006.Diminishingreciprocalfairnessbydisruptingtherightprefrontalcortex.Science314,
829–832.
Ma,C.A.,Manove,M.,1993.Bargainingwithdeadlinesandimperfectplayercontrol.Econometrica61(6),1313–1339. Merlo,A.,Wilson,R.,1998.Efficientdelaysinastochasticmodelofbargaining.Econ.Theory11(1),39–55.
Moore,D.,Loewenstein,G.,2004.Self-interest,automaticity,andthepsychologyofconflictofinterest.Soc.JusticeRes.17,189–202. Romm,A.,2016.Buildingreputationattheedgeofacliff.WorkingPaper.HebrewUniversityofJerusalem.
Roth,A.E.,Murnighan,J.K.,Schoumaker,F.,1988.Thedeadlineeffectinbargaining:someexperimentalevidence.Am.Econ.Rev.78(4),806–823. Rubinstein,A.,1982.Perfectequilibriuminabargainingmodel.Econometrica50(1),97–110.
Rubinstein,A.,1985.ABargainingmodelwithincompleteinformationabouttimepreferences.Econometrica53(5),1151–1172. Simsek,A.,Yildiz,M.,2014.Durability,Deadline,andelectioneffectsinbargaining.WorkingPaper.MIT.
Sullivan,S.P.,2016.Whywaittosettle?Anexperimentaltestoftheasymmetric-informationhypothesis.J.LawEcon.59(3),497–525.
VandenBos,K.,Peters,S.L.,Bobocel,D.R.,Ybema,J.F.,2006.Onpreferencesanddoingtherightthing:Satisfactionwithadvantageousinequitywhen cognitiveprocessingislimited.J.Exp.Soc.Psychol.42(3),273–289.
Vasserman,S.,Yildiz,M.,2016.Pretrialnegotiationsunderoptimism.WorkingPaper.MIT. Yildiz,M.,2004.Waitingtopersuade.Q.J.Econ.119(1),223–248.