ebook img

Climate treaties and approaching catastrophes PDF

16 Pages·2013·1 MB·English
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview Climate treaties and approaching catastrophes

JournalofEnvironmentalEconomicsandManagement](]]]])]]]–]]] ContentslistsavailableatSciVerseScienceDirect Journal of Environmental Economics and Management journal homepage: www.elsevier.com/locate/jeem $ Climate treaties and approaching catastrophes Scott Barretta,b,n aSchoolofInternationalandPublicAffairs&EarthInstitute,ColumbiaUniversity,NewYork,NY10027,UnitedStates bPrincetonInstituteforInternationalandRegionalStudies,PrincetonUniversity,Princeton,NJ08544,UnitedStates a r t i c l e i n f o a b s t r a c t Articlehistory: If the threshold that triggers climate catastrophe is known with certainty, and the Received7November2011 benefits of avoiding catastrophe are high relative to the costs, treaties can easily coordinatecountries’ behavior so asto avoid thethreshold.Where thenetbenefits of Keywords: avoidingcatastrophearelower,treatiestypicallyfailtohelpcountriescooperatetoavoid Climatechange catastrophe,sustainingonlymodestcutsinemissions.Theseresultsareunaffectedby Internationalenvironmentalagreements uncertainty about the impact of catastrophe. By contrast, uncertainty about the Catastrophe catastrophicthresholdnormallycausescoordinationtocollapse.Whethertheprobability Cooperation densityfunctionhas‘‘thin’’or‘‘fat’’tailsmakeslittledifference. Coordination &2012ElsevierInc.Allrightsreserved. Uncertainty Enforcement 1. Introduction Thereisuniversalagreement,codifiedintheFrameworkConventiononClimateChange,thatatmosphericconcentra- tions of greenhouse gases should be stabilized ‘‘at a level that would prevent dangerous [my emphasis] anthropogenic interference with the climate system.’’ The Kyoto Protocol tried to get countries to reduce their emissions without identifyingathresholdfordangerousinterference,butKyotofailed,mainlyforlackofanenforcementmechanism.1The recentlynegotiatedCopenhagenandCancunagreementsalsolackenforcement,andarelegallynon-bindingbesides,but unlikeKyototheseagreementsidentifyadangerousthreshold—a21Celsiusincreaseinaverageglobaltemperature.Inthis paperIask,Canthefearofcrossingacatastrophicthresholdovercometheenforcementchallenge?Canitmakeclimate treatiesmoreeffective? The theory of international environmental agreements has generally offered a gloomy prognosis for cooperation, particularly on this issue.2 To this point, however, the literature has considered only continuous abatement benefit functions.InthispaperItakethebenefitfunctiontobediscontinuousatathreshold.Thissimplechangegivesrisetoanew result.Ifthethresholdisknownwithcertainty,andthelossfromcatastrophevastlyexceedsthecostsofavoidingit,then the collective action problem changes fundamentally. Rather than cooperate to limit emissions, countries need only $TheideaforthispapercametomeasIwasthinkingofhowtorespondtoaquestionposedbyWilliamNordhausataseminarIwasgivingatYalea fewyearsago.IamgratefultoApurvaSanghi,ArthurCampbell,AstridDannenberg,ClaudeHenry,ErinMansur,GeirAsheim,KennethArrow,MarkCane, MatthewKotchen,MartinWeitzman,RobertKeohane,RobertMendelsohn,RobertSocolowandThomasSchellingforcommentsonapreviousdraft. Iamalsogratefultotheeditor,twoanonymousreferees,andparticipantsatseminarsgivenatColumbia,Harvard,Princeton,Stanford,theWorldBank, andYale(alaterseminar)fortheircomments.AnearlierversionofthispaperwasgivenattheinvitationofMichaelFinusasakeynotelectureatthe EnvironmentalProtectionandSustainabilityForum,UniversityofExeter,April2011. nCorrespondenceaddress:SchoolofInternationalandPublicAffairs&EarthInstitute,ColumbiaUniversity,NewYork,NY10027,UnitedStates. Fax:þ12128544782. E-mailaddress:[email protected] 1SeeBarrett[7]. 2Forsurveysoftheliterature,seeMichaelFinus[13],UlrichJ.Wagner[26],andBarrett[6,7]. 0095-0696/$-seefrontmatter&2012ElsevierInc.Allrightsreserved. http://dx.doi.org/10.1016/j.jeem.2012.12.004 Pleasecitethisarticleas:S.Barrett,Climatetreatiesandapproachingcatastrophes,JournalofEnvironmentalEconomics andManagement(2013),http://dx.doi.org/10.1016/j.jeem.2012.12.004i 2 S.Barrett/JournalofEnvironmentalEconomicsandManagement](]]]])]]]–]]] coordinatetoavoidcatastrophe.Underthesecircumstances,climatetreatiescansustaintheefficientoutcome.Essentially, natureherselfenforcesanagreementtoavoidcatastrophe. Here is why: when the net benefit of steering clear of the threshold is very high, avoiding catastrophe is a Nash equilibrium.(Imainlyrestrictattentioninthispapertosymmetricequilibriainpurestrategies.)Sinceeachcountryistoo smalltoavertcatastropheonitsown,however,therealsoexistsaParetoinferiorNashequilibriuminwhichthethreshold is crossed. Under these circumstances, the challenge is for countries to coordinate on the mutually preferred equilibrium—a task for which treaties are exquisitely suited. When these circumstances do not apply, there exists a unique Nash equilibrium that sits on the wrong side of the threshold. Since there also exist other feasible, mutually preferredoutcomes,countrieswillwanttocooperatesoastoimproveonthisinefficientequilibrium(asinaprisoners’ dilemmagame).Atreatycanhelphere,too,butcredibleenforcementmechanismsforsupplyingthiskindofglobalpublic goodaregenerallyweak.3RelativetotheinefficientNashequilibrium,whichiscommontobothsituations,coordination, whenpossible,almostalwayssucceedsbetterthancooperation. Weitzman [27] has drawn our attention to ‘‘uncertain catastrophes with tiny but highly unknown probabilities,’’ showingthat,iftheprobabilityofeverlargercatastrophesdoesnotfallfasterthanwelfarelossesincreaseasweventure deeperintothetailsoftheprobabilitydensityfunction,thentheexpectedgainfromapolicytoreduceemissionswillbe infinite.Underthesecircumstances,policyshoulddoeverythingpossibletoreduceemissionsasquicklyaspossible,even thoughdoingsocannotguaranteethatcatastrophewillbeavoided.4Hecallsthisresult,appropriatelyenough,‘‘thedismal theorem.’’ Nordhaus[18]:21argues,tothecontrary,thatanyconceivablecatastrophicoutcomecanbeavoidedbypolicy,butthat doingsorequires‘‘solvingtheglobalpublicgoodsproblembygatheringmostnationstogethertotakecollectiveaction.’’5 Weitzman does not address the collective action aspects of this challenge, but his theorem should carry through in a decentralizedsetting.Ifitpaystheworldasawholealwaystodevotemoreresourcestoreducingathreatthatcannotbe eliminated,nomatterthecost,thenitshouldpayindividualcountries(havingthesamepreferences,andfacingthesame uncertainty as assumed by Weitzman) always to devote more resources to reducing the same existential threat, even when the benefits of doing so are widely dispersed. That is, under Weitzman’s assumptions, coordination should be unnecessary. Every country should want to put its economy on a ‘‘climate war’’ footing, irrespective of whether other countriesjointheminthiseffort. Underpinningthedismaltheoremisthestrongassumptionthatutilityisunbounded.6Ifthisassumptionisrelaxed,so that preferences are not very risk averse as consumption approaches zero, a different outcome emerges, one that is consistentwithWeitzman’spolicyrecommendationof‘‘arelativelymorecautiousapproach[emphasisadded]’’toreducing greenhousegasemissions([27]:13).Inthispaper,forsimplicity,Itakeutilitytobelinear.Whilethisassumptionrulesout Weitzman’s dismal theorem, I shall show that it nonetheless implies that countries may want to adopt a much more cautiousapproachascomparedtooneinwhichthethreatofcatastrophecanbeignored.Unfortunately,underthesesame circumstances, I also find that the uncertain prospect of approaching catastrophe has little if any effect on the non- cooperative outcome, or the ability of an international agreement to sustain collective action. The reason is that uncertainty(aboutthethresholdforcatastrophe)makestheexpecteddamagefunctioncontinuous,renderingcoordina- tion ineffective, and restoring the main results reported previously in the literature—that cooperation is needed, but difficulttosustain.Inthismodel,whethertheprobabilitydensityfunction(pdf)hasthinorfattailsmakeslittledifference. Uncertaintyaboutthemagnitudeofcatastrophicdamagesisalsorelativelyunimportant.Itisuncertaintyinthethreshold thatmatters. Acertainthresholdcanbeinterpretedasadiscreteuniformdistributionhavingasinglevalue.Aninterestingextension ofthismodel,whichtiesmyearlieranalysisofcertaintytogetherwithmylateranalysisofuncertainty,isthecontinuous uniformdistribution.Thisdistributionexhibitsacriticaldiscontinuity(inthedistribution’srightsupport),beingfat-tailed throughoutitsrangeandzero-tailedelsewhere.Ishowthat,forthisspecialpdf,theresultIobtainedpreviouslyforthe modelofcertaintyextendstouncertainty—countriesmaybeabletocoordinateontheright-sidediscontinuitysoasto guaranteeavoidingcatastrophe.However,ascomparedwiththecertaintycase,Ifindthatopportunitiesforcoordination areextremelylimited.Evenunderthefavorableconditionsofthisspecialpdf,cooperationwillalmostcertainlybeneeded and,asusual,difficulttosustain. Myoverallconclusionisthatfearofapproachingcatastropheswithuncertainthresholdsstrengthenstheimperativeto cooperatewithoutimprovingtheprospectsforcollectiveaction. 3Thisistrueeveninarepeatedgamesolongastreatiesmustberenegotiation-proof;seeBarrett[5,6,7].Theone-shotgamedevelopedinthispaper obeysananalogousproperty. 4Iftheconditionsthatgiverisetothisresultarerelaxedevenslightly,thenthethreatofcatastropheraisesnumerouspolicyissues;seeKouskyetal.[15]. 5ToNordhaus[18],threeconditionsmustholdforcatastrophetobecauseforconcern:policyfailure,asjustnoted,hightemperaturesensitivityto concentrations,andextremelyconvexdamages.Inthispaper,becausetemperatureissubsumedintheanalysis(damagesarerelatedtoaggregate emissions),thelattertwoconditionswillbesatisfiedwhentheabatementbenefitfunctionisdiscontinuous(or,intheuncertaintycase,whenthe varianceisverysmall)andthelossduetocrossingthethreshold(parameterXinthispaper)isverylarge. 6Arrow[2]notesthat,forindividuals,theassumptionimpliesthat‘‘thevalueofstatisticallifeisinfinite,aconclusionclearlycontrarytoallempirical evidenceandtoeverydayobservation.’’ Pleasecitethisarticleas:S.Barrett,Climatetreatiesandapproachingcatastrophes,JournalofEnvironmentalEconomics andManagement(2013),http://dx.doi.org/10.1016/j.jeem.2012.12.004i S.Barrett/JournalofEnvironmentalEconomicsandManagement](]]]])]]]–]]] 3 Althoughthefocusofthispaperisonclimatechange,themodelcanbeapplied,withsuitablemodifications,toother situations.Considertheproblemofspacedebris.Ifthevolumeofdebrisinlowearthorbitexceedsacriticallevel,onemore collision dramatically increases the probability of more collisions, creating a cascading effect that could render orbital spaceunusable(KesslerandCour-Palais[14]).Here,catastrophecanbeavoidedbylimitingadditionstothestockofdebris, orbydesigningsatellitesandrocketboosterssothat theycanbemaneuveredto a‘‘graveyard’’ordecayingorbitwhen they become obsolete. Another example is antibiotic resistance. Resistant strains of a pathogen sometimes have high evolutionaryfitnessinthepresenceofadrugtreatment,butsufferafitnesscostrelativetodrug-sensitivestrainswhenthe drugisremoved(Smith[23]).Ifuseofthedrugislow,thesensitivestrainswillwinout.Ifuseishigh,theresistantstrains willthrive.Inbetweenthereexistsacriticalthresholdfordruguse,withthepriorityforpublichealthbeingtostayonthe goodsideofthethreshold.Consider,finally,exploitationofabiologicalresourcesuchasaspeciesoffishforwhichthere existsastrictlypositiveminimumviablepopulationlevel.Economicoverexploitationoftheresourcemaybeaproblem even at higher stock levels, but exploitation below the critical level would be disastrous; crossing it would cause the speciestogoextinct.Thismaybeaparticularlyimportantissueforfishpopulationssuchastunathataggregate,ifschool sizeisindependentofstockabundance,andharvesteconomicsdependonschooling(ClarkandMangel[11]). Notwithstandingtheimportanceoftheseandotherapplications,mymainconcerninthispaperiswithclimatechange. I shall introduce my model in Section 3, but it will help to begin by putting this model in the context of a previously publishedexperiment,whichwasalsoconcernedwithclimatechange. 2. Anexperimentincatastropheavoidance TheexperimentbyMilinskietal.[17]simulatesthe‘‘collective-risksocialdilemma’’inpreventing‘‘dangerousclimate change.’’Intheirexperimenttherearesixplayers.Eachisgivenh40.Thegameisplayedovertenperiods.Ineachperiod, everyplayermustchoosetocontributeh0,h2,orh4withoutcommunicating.If,attheendofthegame,atleasth120has beencontributed,dangerousclimatechangeisavertedwithcertainty,andeachplayergetsapayoffequaltotheamountof money he has left (there are no refunds in this game). If less than h120 has been contributed, each player loses all the money she has left with probability 0.9. In their experiment, Milinski et al. [17] played the game with ten groups of students,onlyhalfofwhichsucceededinavoidingthethreshold. Therearetwosymmetricpurestrategyequilibria.Inone,everyplayercontributesh0everyperiod,givingeachplayer anexpectedpayoffofh4.Intheother,everyplayercontributesh2everyperiod,givingeachacertainpayoffofh20.(Of course,therealsoexistmanyasymmetricpurestrategyequilibriainwhichdifferentplayerscontributedifferentamounts, possiblyindifferentperiods).Thelatterequilibriumisefficient;theformerisnot. In contrast to the conventional representation of the climate change game, the game with catastrophic damages is pleadingforcoordination.Why,then,didhalfthegroupsintheMilinskietal.[17]experimentfailtocoordinate?Themain reasonisalmostsurelythat,byconstruction,theplayerswerenotallowedtocommunicate,letaloneformulateatreatyto coordinatetheircontributions.7 Theusualwayofmodelinganinternationalenvironmentalagreementisinthreestages.8Instageone,countrieschoose independentlywhethertobeapartyornon-partytotheagreement.Instagetwo,partieschoosetheiractions(inthiscase, contributions) so as to maximize their collective payoff. Finally, in stage three, non-parties choose their actions with the aim of maximizing their individual payoffs. A treaty is self-enforcing if, given the treaty and participation level, non-participantsdonotwanttochangetheirbehavior;if,giventheparticipationlevel,partiestothetreatydonotwantto changetheobligationsexpressedinthetreaty;andif,giventheparticipationdecisionsofothercountries,eachcountry doesnotwanttochangeitsdecisionofwhethertobeapartyornon-partytothetreaty. Applyingthisnotionofaself-enforcingtreatytotheMilinskietal.[17]game,itwillhelptoassumethatcontributions aremadeinasingleperiodandthateachplayercancontributeanyamountuptohisorherendowment.Ifparticipation werefull,thetreatywouldthentelleachcountrytocontributeh20,nettingeachcountryapayoffofh20.Wereacountry todropoutofthisagreement,theremainingparties,choosingcollectively,wouldchangetheircontributions.Theywould reason that, if they contributed an amount Y in total, then, taking this contribution as given, the non-party would contributeanamountZ¼h120–Yforh120ZYZh84andZ¼h0forYoh84.Knowingthis,thefiveremainingsignatories could do no better than to contribute Y¼h84 collectively (h16.80 each). This would net each of the five parties h23.20, whereasthesolenon-signatorywouldgetjusth4.Recallthat,werethiscountrynottowithdraw,itwouldgetapayoffof h20.Obviously,withthetreatywritteninthisway,nocountryhasanincentivetowithdraw,startingfromasituationin 7AsimilarexperimentbyTavonietal.[25]confirmstheimportanceoftheassumptionaboutcommunication. 8SeeBarrett[7].Thisthree-stageformulationimbuespartiestotheagreementwiththespecialabilitytocommittoplayingacollectiveabatement level—theso-called‘‘Stackelbergleadership’’assumption.Ifnon-signatorieshavedominantstrategies,thisassumptionisindistinguishablefromthe Nashassumption.IntheMilinskietal.model,however,thepayofftoavoidingcatastrophejumpsatthethreshold,implyingthatcountriesmaynothave dominantstrategies.Themainjustificationfortheleadershipassumptionisthecustomarylawprinciple,pactasuntservanda:treatiesarebinding.Parties toanagreementareexpectedtofulfilltheirobligations,withcustombeingenforcedoutsidethetreaty(thatis,outsidethemodel).Notethatthisisa minimalcommitment,sinceinternationallawalsosaysthatcountriesarefreetoparticipateintreatiesornotastheyplease—anassumptioncapturedin Stage1oftheabovemodel.Inthispaper,theleadershipassumptionisalsoimportantforequilibriumselection—itensuresthat,inacoordinationgame, themoreefficientequilibriumwillbesupportedbyaself-enforcingagreement. Pleasecitethisarticleas:S.Barrett,Climatetreatiesandapproachingcatastrophes,JournalofEnvironmentalEconomics andManagement(2013),http://dx.doi.org/10.1016/j.jeem.2012.12.004i 4 S.Barrett/JournalofEnvironmentalEconomicsandManagement](]]]])]]]–]]] which participation is full. Hence, a treaty comprising six signatories, each of which contributes h20, is self-enforcing. Moreover,whilethereexisttwoNashequilibriainpurestrategiesintheunderlyinggame,theself-enforcingagreement justnotedisuniqueandsupportstheParetoefficientequilibrium. Tosumup,whileMilinskietal.[17]claimthatcountriesmayfailtoavertcatastrophewhendoingsoisfeasibleand efficient,simpletheorysuggestthat,solongascountriesarepermittedtonegotiateatreaty,catastropheshouldberather easy to avoid. In the remainder of this paper I inquire into whether this result could be expected to hold in richer (analytical)environments. 3. Aclimatechangecatastrophegame Assumethatcountrieshavesymmetricpayofffunctions.Denotingcountryi’sabatementbyq andaggregateabatement i byQ,whereQ¼PN q andN isthenumberofcountries,countryi’spayoffisassumedtobegivenby i¼1 i p ¼8<bQ(cid:2)cq22i ifQZQ : ð1Þ i :bQ(cid:2)X(cid:2)cq2i ifQoQ 2 In (1), parameter b stands for each country’s marginal benefit of abating ‘‘gradual’’ climate change, cq represents i countryi’smarginalabatementcosts,andXdenotesthedamageseachcountrysuffersfrom‘‘catastrophic’’climatechange. ThesedamagesareexperiencedifandonlyifglobalabatementfallsshortofQ,thethreshold.9IfweletX¼0,(1)collapses tothestandardmodelintheliteratureoninternationalenvironmentalagreements.10 It is worth noting the important differences between (1) and Milinski et al. [17] model. First, in (1) contributions (abatementlevels)arebeneficialevenifthethresholdisexceeded;abatementreduces‘‘gradual’’aswellas‘‘catastrophic’’ climatechange.Second,Milinskietal.varyX(morespecifically,theexpectedvalueofX),butnotthethreshold,and(1) allowsustovaryboth.11Third,in(1),marginalabatementcostsareincreasing,whereasMilinksietal.implicitlyassume that they are constant. Fourth, Milinski and coauthors take abatement (contributions) to be discrete, and (1) allows abatementtobevariedcontinuously.Finally,while(1)assumesthatXiscertain,andMilinksietal.letXbeuncertain,later inthispaperIshallletbothXandQ beuncertain. IncommonwithMilinskietal.[17]andWeitzman[27],model(1)abstractsawayfromthedynamicsofclimatechange. It is, however, able to capture the key difference between ‘‘gradual’’ and ‘‘abrupt and catastrophic’’ climate change. It assumesthatgradualclimatechangeissoslowthatmarginalbenefitsareconstant,andthatabruptclimatechangeisso fastthattransitionisinstantaneous.12Themodelcanbeinterpretedascompressingperhapsacenturyofdecision-making intoasingleperiod.Sincetheimpactsofabruptchangewillunfoldmoreslowlythanthis(someoftheimpactscouldtake amillenniumormoretoplayout),Xshouldbeinterpretedascapturingthefullconsequences,includingintothedistant future,ofdecisionstakenthiscenturytoavoidorexceedathreshold.Theparameterbshouldbeinterpretedsimilarlyfor gradualclimatechange. Inthefullcooperativeoutcome,theaggregatepayoffwillbe 8>><bQN(cid:2)Pc2q2i ifQZQ PFC¼ i ð2Þ >>:bQN(cid:2)XN(cid:2)Pcq22i ifQoQ: i IfX¼0,maximizationof(2)yieldsQFC¼bN2=c.AssumeQ4bN2=c(avoidingthethresholdrequiresabatingmorethan isoptimalforaddressingonly‘‘gradual’’climatechange).13Thentherearetwopossibilities.Eitheritwillpayallcountries collectivelytomeetthethresholdðQFC¼QÞ,just,oritwillnotpaytomeetthethresholdðQFC¼bN2=cÞ.Uponsubstituting 9Damagesarenormallyrelatedtotemperature,temperaturetoconcentrations,andconcentrationstoanemissionsprofile.However,thereisstrong evidencethattemperaturecanberelateddirectlytocumulativeemissions(Allenetal.[1];Zickfeldetal.[28]).Inthispaper,Itakebusinessasusual emissionsasgiven.Myfocusisonreductionsfromthislevel:thelevelofabatement.Thegreateristhelevelofabatement,thesmallerwillbecumulative (total)emissions,andthelowerwillbetemperatureand,therefore,damages.Emissionswillneedtobereducedtoacriticallevel,denotedherebyQ,ifa dangerousthreshold,suchasthe21Cchangeintemperature,istobeavoided. 10SeeespeciallyBarrett[7]. 11Milinskietal.[17]varytheexpectedvalueofXbyvaryingtheprobabilitythatdamageswillbe‘‘catastrophic’’giventhatcontributionsfallshortof thethreshold. 12Inamodelwithincreasingmarginalbenefits,internationalcooperationstillachievesverylittle[3],soitisnottheconstancyofmarginalbenefits thatmatterssomuchascontinuityinthebenefitfunction. 13Thisassumptionimpliesthatafocusonlyon‘‘gradual’’climatechangecouldcausetheworldtocrossadangerousthreshold.InthewordsofStern [24],whowarnsofpossibletippingpoints,‘‘Itisimportanttobeclearthatthe‘‘climatepolicyramp’’advocatedbysomeeconomistsinvolvesareal possibilityofdevastatingclimaticchanges.’’Ofcourse,itisconceivablethatthefullcooperativeabatementlevelcouldbedeterminedsolelyby‘‘gradual’’ climate change ðQobN2=cÞ. Suppose, then, that a treaty addressing only ‘‘gradual’’ climate change sustained the abatement level, Qn. Then, if QoQnrbN2=c,considerationof‘‘catastrophic’’climatechangewouldhavenoeffectontheusualanalysis;abatementtoavoid‘‘gradual’’climatechange wouldavoidcatastropheinthebargain.If,tothecontrary,QnoQobN2=c,considerationgivento‘‘catastrophic’’climatechangemightimprovethe prospectsforcollectiveaction,whilestillfallingshortofsustainingthefullcooperativeabatementlevel.Inthiscase,theanalysiswouldproceedinthe samewayasinthispaper. Pleasecitethisarticleas:S.Barrett,Climatetreatiesandapproachingcatastrophes,JournalofEnvironmentalEconomics andManagement(2013),http://dx.doi.org/10.1016/j.jeem.2012.12.004i S.Barrett/JournalofEnvironmentalEconomicsandManagement](]]]])]]]–]]] 5 thesevaluesinto(2),itiseasytoshowthatitwillpaytomeetthethresholdifandonlyif b2N2 cQ2! XZ (cid:2) bQ(cid:2) : ð3Þ 2c 2N2 Fig.1illustratestherelationship(whichisnon-linearduetomarginalcostsincreasing).Iftheimpactofcatastrophic climatechange,X,is‘‘small,’’catastropheisworthavoidingonlyifQ,thethreshold,is‘‘small’’(sothatthecostsofavoiding catastrophearelow).IfQ is‘‘large,’’catastropheisworthavoidingonlyifXis‘‘large.’’ In the non-cooperative outcome, each country i will maximize (1), taking as given the abatement choices of other countries.ThereareatmosttwosymmetricNashequilibriainpurestrategies.Inone,everycountryiplaysq ¼b=c,and i the threshold is exceeded. In theother, every country plays q ¼Q=N, and thethreshold is avoided, just. Suppose every i countryjaiplaysq ¼Q=N.Theniwillplayeitherq ¼Q=Norq ¼b=c.Uponsubstitutingthesevaluesin(1),itiseasyto j i i show that country i will prefer to play the former abatement level (in which case, avoiding catastrophe is a Nash equilibrium)ratherthanthelatterifandonlyif b2 bQ cQ2! XZ (cid:2) (cid:2) : ð4Þ 2c N 2N2 The top part of the Fig. 2 shows the space in which (3) and (4) both hold—that is, the space in which catastrophe avoidanceisboth(i)collectivelyoptimaland(ii)asymmetricNashequilibrium.Thebottompart,unchangedfromFig.1, Fig.1. Optimalcatastropheavoidance. Fig.2. Whencoordinationsufficesandcooperationisneeded. Pleasecitethisarticleas:S.Barrett,Climatetreatiesandapproachingcatastrophes,JournalofEnvironmentalEconomics andManagement(2013),http://dx.doi.org/10.1016/j.jeem.2012.12.004i 6 S.Barrett/JournalofEnvironmentalEconomicsandManagement](]]]])]]]–]]] showsthecombinationsofXandQ forwhichavoidingcatastropheisinefficient.Finally,sandwichedbetweenthesetwo spacesisamiddleareainwhich(3)holdsbut(4)doesnothold—thatis,catastropheavoidanceiscollectivelyoptimalbut cannotbesupportedasasymmetricNashequilibrium.(Notethatthismiddleareaonlyexistsbecauseabatementreduces ‘‘gradual’’aswellas‘‘catastrophic’’climatechange.)Thefigureshowsthatcatastropheavoidancecannotbesustainedby coordinationifXisextremelysmall,evenifQ isclosetothefullcooperativeabatementlevelforgradualclimatechange. OnlyifXis‘‘bigenough’’willitbepossibleforcountriestocoordinatetoavoida‘‘catastrophe.’’ In comparing (3) and (4), we see that coordination can always be relied upon to avoid catastrophe when doing so is optimal provided either N¼1 (no externality) or b¼0 (no benefit to abatement apart from avoiding catastrophe). The former result is to be expected. The latter result is surprising.14 Intuitively, the incentives to avoid ‘‘catastrophic’’ climate change should be helped when the actions needed to do this also reduce ‘‘gradual’’ climate change. Imagine, however,thateverycountryiplaysq ¼Q=N.Then,takingtheabatementlevelsofothercountriesasgiven,acountrythat i deviatesunilaterallywillsufferasmallerlosswhentheabatementbytheothercountriesreducesgradualclimatechange. Ironically,theeffectofabatementinreducinggradualclimatechangeincreasestheincentiveforacountrytodeviatefrom anagreementseekingtoavertcatastrophicclimatechange. Tosummarize: Proposition1. Wheninequality(3)doesnothold,sothattheimpactof‘‘catastrophic’’climatechangeislowrelativetothecost ofavoidingcatastrophe,thecollectiveactionproblemisaprisoners’dilemmagameforreducing‘‘gradual’’climatechange.When (4)holds,sothattheimpactofcatastrophicclimatechangeishighrelativetothecostofpreventingcatastrophe,thecollective actionproblemisacoordinationgameforavoidingcatastrophe.Inbetweenthesesituations,(3)holdsbut(4)doesnothold,and thecollectiveactionproblemisaprisoners’dilemmagameforavoidingcatastrophe.15 4. Climatetreatiesforavoidingcatastrophes Whatroleisthereforclimatetreatiesinavertinganapproachingcatastrophe?Startfromasituationinwhichevery country participates in a treaty requiring that each play q ¼Q=N. What conditions must hold for such a treaty to be i self-enforcing? Considerfirsttheincentivesforacountrytowithdrawfromthetreaty.Uponwithdrawing,thedeviatingcountrywill takeasgiventhebehavioroftheremainingN(cid:2)1cooperatingcountriesasthisisspecifiedintheself-enforcingagreement. Supposethat,onceiwithdraws,theagreementinstructsthesecountriestoplayQ .Then,uponwithdrawing,iwilleither (cid:2)i wanttoplayq ¼Q(cid:2)Q ,toensurethatthethresholdismet,oritwillletthethresholdslipandplayq ¼b=c.Giventhat i (cid:2)i i theothercountriesabateQ ,iwillprefertoplaytheformerlevelratherthanthelatterif (cid:2)i c(cid:2) (cid:3)2 (cid:4) b(cid:5) c(cid:4)b(cid:5)2 bQ(cid:2) Q(cid:2)Q Zb Q þ (cid:2)X(cid:2) ð5Þ 2 (cid:2)i (cid:2)i c 2 c Solvingthequadratic,theminimumvalueofQ whichensuresthat(5)ismetisgivenbyQ^ ¼Q(cid:2)b=c(cid:2)pffi2ffiffiXffiffiffi=ffiffifficffiffi.Hence, (cid:2)i (cid:2)i offtheequilibriumoffullparticipation,iftheN(cid:2)1cooperatingcountriesplayQ^ ,theniwillplayq ¼b=cþpffi2ffiffiXffiffiffi=ffiffifficffiffiand (cid:2)i i catastrophewillbeavoided.If,tothecontrary,theyplayQ(cid:2)ioQ^(cid:2)i,theniwillplayqi¼b=c,andthecatastrophicthreshold willbecrossed. How, then, will the N(cid:2)1 remaining signatories want to play? If they play Q(cid:2)ioQ^(cid:2)i, then they cannot do better collectively than to play Q ¼bðN(cid:2)1Þ2=c. Knowing how the deviant country will behave, they will prefer to abate (cid:2)i Q ¼Q^ ratherthanQ ¼bðN(cid:2)1Þ2=cifandonlyif (cid:2)i (cid:2)i (cid:2)i c Q(cid:2)b=c(cid:2)pffi2ffiffiXffiffiffi=ffiffifficffiffi!2 bðN(cid:2)1Þ2 b! c(cid:4)bðN(cid:2)1Þ(cid:5)2 bQ(cid:2) Zb þ (cid:2)X(cid:2) : ð6Þ 2 N(cid:2)1 c c 2 c If(6)holds,andifishouldwithdrawfromtheagreementwithfullparticipation,thetreatywillinstructtheotherN(cid:2)1 countries to play (not Q ¼QðN(cid:2)1Þ=N, the Nash assumption, but) Q ¼Q^ . Under these conditions, country i will not (cid:2)i (cid:2)i (cid:2)i wanttowithdrawif !2 rffiffiffiffiffiffi!2 c Q c b 2X bQ(cid:2) ZbQ(cid:2) þ : ð7Þ 2 N 2 c c Rewriting, we see that inequality (7) requires b=cþpffi2ffiffiXffiffiffi=ffiffifficffiffiZQ=N; a country that withdraws is punished by being movedtoabatemoreasanon-signatorythanasasignatory(theoppositeoffreeriding!). 14ForMilinskietal.’s[17]model,thereisnomiddlewedge;efficientcatastropheavoidanceisalwaysaNashequilibrium. 15Obviously,theseresultsapplyspecificallytomodel(1),whichislinear-quadratic(linearbenefitsandquadraticcosts).However,itiseasytosee thattheseresultsalsoapplytothemodeldiscussedinSection2,whichislinear–linearwithb¼0,exceptthat,inthismodel,thereisnocollectiveaction problemforlimitinggradualclimatechange.ThepaperbyBarrettandDannenberg[10],writtenafterthispaperwassubmitted,showsthatProposition1 alsoholdsforamodelwithlinearbenefits(withb40)andstepwise-linearmarginalcosts(theirmodelwasdevelopedforthepurposeofconductingan experimentaltestofthetheorypresentedinthispaper);compareFig.1intheirpaperwithFig.2inthisone. Pleasecitethisarticleas:S.Barrett,Climatetreatiesandapproachingcatastrophes,JournalofEnvironmentalEconomics andManagement(2013),http://dx.doi.org/10.1016/j.jeem.2012.12.004i S.Barrett/JournalofEnvironmentalEconomicsandManagement](]]]])]]]–]]] 7 If(6)and(7)hold,nocountrycangainbywithdrawingunilaterallyfromtheagreementthatavoidscatastrophe.Itis easytoshow(seetheAppendix)that,if(7)holds,thensowill(6)hold.Moreover,inequality(7)reducesto(4),whichisthe conditionforwhencatastropheavoidancecanbesustainedbycoordination.Hence,wehave: Proposition 2. When (4) holds, so that the impact of catastrophic climate change is high relative to the cost of preventing catastrophe, a self-enforcing international environmental agreement can sustain the full cooperative outcome, ensuring that catastropheisavoided.Undertheseconditions,thetreatyservesasacoordinatingdevice. Thisanalysisassumesfullparticipation.Itiswellknownthat,inmostcases,treatiescanonlysupportalowerlevelof participation.So,mightcollectiveactiontoavoidcatastrophebesustainedevenif(7)failstohold? Even if (7) were violated, inequality (6) could still hold. That is, the N(cid:2)1 signatories could have an incentive to play pffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffi Q(cid:2)b=c(cid:2) 2X=c, knowing that the sole non-signatory would then abate b=cþ 2X=coQ=N (in which case the non- signatory would be a true free rider). Moreover, this agreement comprising N(cid:2)1 countries would be self-enforcing, provideditcouldbesustainedbycoordination.However,itiseasytoshowthattheconditionsthatenableN(cid:2)1countriesto coordinate to avoid catastrophe also enable N countries to coordinate to sustain the full cooperative outcome (see the Appendix).Moreover,byextendingthislogictootherparticipationlevels,itisclearthat,if(7)failstohold,itwillnotbe possibletosustaincoordinationtoavoidcatastrophebyanynumberofcountries(rN). Ofcourse,when(7)holds,itcouldbepossibleforfewerthanNcountriestocoordinatesoastoavoidcatastrophe(thisisjust anotherwayofsayingthattheremayexistalargenumberofasymmetricNashequilibriaforavoidingcatastrophe).However, coordinationbyfewerthanNcountries(withtheothersactingasfreeriders)offersnoadvantageovercoordinationbythe grandcoalition—onlyadisadvantage,sinceitismorecostlyforN(cid:2)1countriestosustainthethresholdthanforNcountriesto doso(withmarginalcostsincreasing,cost-effectiveabatementrequiresthateverycountryabate thesamequantity).Since everycountry’sexpectedpayoff(withtheexpectationtakenbeforeanagreementisadopted,whennocountryknowsifitwill beasignatoryoranon-signatory)isgreaterforanagreementcomprisingNcountriesthanonecomprisingN(cid:2)1 (orfewer) countries,providedbothagreementsareself-enforcing,inthenegotiationstageallcountrieswillfavorauniversalagreement. Moreover,oncesuchanagreementhasenteredintoforce,if(7)holds,nocountrywillhaveanincentivetowithdraw. Together,theseresultsimply: Proposition3. Ifthereexistsaself-enforcinginternationalenvironmentalagreementthatavoidscatastrophebycoordinating countries’abatement,thenthisagreementwillbeuniqueandsustainfullparticipation. Whathappensifavoidingcatastropheisefficientandyetcoordinationfails?Thiswouldputusinthemiddleterritoryof Fig.2.Countriescouldstilltrytocooperatetoavoidcatastropheinthisregion,byusingstrategiestoovercometheprisoners’ dilemma,asintheusualmodelofaninternationalenvironmentalagreement.Inthiscase,signatorieswouldreducetheir abatementinthebeliefthat,bydoingso,thethresholdwouldbeblown.However,thecooperationproblemisnowdifferent from the usual one, which assumes only ‘‘gradual’’ climate change. For model (1), a deviation triggers a harsher punishment—the‘‘catastrophic’’lossduetocrossingthethreshold.Itmay,therefore,sustaingreatercooperation.16 ForittobecollectivelyrationalfortheN(cid:2)1signatoriestoplayQ ¼bðN(cid:2)1Þ2=cafterihaswithdrawn,inequality(6) (cid:2)i mustnothold.Knowingthis,countryiwillnotwanttowithdrawif b2½2ðN(cid:2)1Þ2þ1(cid:3) cQ2! XZ (cid:2) bQ(cid:2) : ð8Þ 2c 2N2 Condition(8)isthesameas(3)exceptfortheintercept.Itisstraightforwardtoshowthattheintercepttermin(8)isno greaterthantheintercepttermin(3)forNr3.Thatis,whencoordinationfails,atreatycanstillbereliedupontosustain the full cooperative outcome—meaning, avoid catastrophe—if Nr3. It is well known that, with the functional forms assumedin(1),aself-enforcinginternationalenvironmentalagreementcansustainthefullcooperativeoutcomeevenin theabsenceofcatastropheforNr3.Considerationofcatastrophethusdoesnothingtochangethisresult. ForN43,however,matterscanbedifferent.Cooperationmayfailentirelytoavoidcatastrophe,oritmaysucceedin avoidingcatastrophebutonlywithinaportionofthemiddlewedgeshowninFig.2.17Inthisregion,inequality(8)holds but(6)doesnothold. Inequality (6) is awkward to work with, but an example can illustrate where and why cooperation can be effective within this middlewedge. Suppose N¼20, b¼c¼1, and Q¼600.18 Then the vertical distance of thecooperation wedge shown in Fig. 2, as defined by inequalities (3) and (4), requires 420:54XZ50 (coordination succeeds for XZ420:5; 16‘‘Punishment’’appearsintheone-shotmodelofaninternationalenvironmentalagreementwhenthetreatyspecifiesthatcooperatingcountries willlowertheirper-countryabatementasthenumberofparticipatingcountriesfalls.Wenormallythinkof‘‘punishment’’inthecontextofarepeated game.Butthestagegamemodelofaninternationalenvironmentalagreementisparticularlysuitedtoexploringtheenforcementofparticipation, whereas the repeated game model is better suited to exploring the enforcement of compliance. I have shown elsewhere that there is a natural correspondencebetweenthesetwoapproaches[5–7]. 17Usingotherfunctionalforms,self-enforcingagreementscansustainfullcooperationforN43withoutconsideringthepossibilityofcatastrophe. Theseincludethequadratic-quadratic[3]andlinear-linear[6]models.However,inbothofthesecases,aself-enforcingagreementcansustainthefull cooperativeoutcomeonlywhenthisoutcomeisnearlyidenticaltothenon-cooperativeoutcome. 18YoucanthinkofthisassumptionassuggestingthatcollectiveactionbytheG-20groupofcountriescouldpreventcatastrophicclimatechange. Pleasecitethisarticleas:S.Barrett,Climatetreatiesandapproachingcatastrophes,JournalofEnvironmentalEconomics andManagement(2013),http://dx.doi.org/10.1016/j.jeem.2012.12.004i 8 S.Barrett/JournalofEnvironmentalEconomicsandManagement](]]]])]]]–]]] catastropheavoidanceisinefficientforXo50).Condition(8)requiresXZ211:5;theimpactofcatastrophemustbelarge enoughtodeterwithdrawal.Ofcourse,XmustalsobesmallenoughthattheN–1countriesremainingintheagreement are willing to cross the threshold should i decline to cooperate. Upon substituting we find that (6) will not hold if p pffi7ffiffi2ffiffiffi2ffiffiffiðffiXffiffiffiffiþffiffiffiffi4ffiffiffi1ffiffiffi8ffiffiffi:ffi5ffiffiffiÞffio599(cid:2) ffi2ffiffiXffiffiffi,anditcanbeshownthatthereisnovalueforXwithintherelevantrangeð420:54XZ211:5Þ for which reversal of inequality (6) is satisfied. For these parameter values, cooperation to avoid catastrophe fails completely. Cooperation will have a better chance of avoiding catastrophe if Q is bigger. Raising the threshold helps because it makes credible the threat by the N(cid:2)1 remaining signatories to lower their abatement dramatically should country i withdraw. For example, if the threshold is increased to 4000, the vertical range of the cooperation wedge becomes19,800:54XZ16,200(coordinationsucceedsforXZ19,800.5;avoidingcatastropheisinefficientforXo16,200), and cooperation can succeed in avoiding catastrophe—but only over a small portion of this range ð16,371:5ZXZ 16,361:5Þ. ItcanalsobeshownthatifNcountriescannotsustaincooperationtoavoidcatastrophe,thenneithercanfewerthanN countries do so.19 Of course, in the event that cooperation to avoid catastrophe fails, a self-enforcing international environmentalagreementcomprisingthreecountries(forNZ3)couldstillsucceedinreducing‘‘gradual’’climatechange. However, such an agreement would reduce emissions just a little relative to the non-cooperative outcome; as is well known,eachofthethreesignatorieswouldabateq ¼3b=candeachnon-signatoryjustb=c.Cooperationtolimit‘‘gradual’’ s climatechangewouldhaveonlyamodestimpactonemissions. Tosumup: Proposition4. When(3)holdsbut(4)doesnothold,cooperationmaysucceedinavertingcatastrophe,butonlyunderspecial circumstances. Moreover, even when these special circumstances apply, cooperation improves welfare relatively little.20 Coordination,bycontrast,succeedsspectacularlywhenthenetbenefitsofavoidingcatastropheareparticularlylarge. Itshouldnotsurpriseusthatcoordinationmakesallthedifference.TheMontrealProtocolsucceededpartlybecausethe strategicapplicationoftraderestrictionstransformedprotectionoftheozonelayer,ordinarilyaprisoners’dilemma,intoa coordination game [4,6]. Network externalities can play the same role in a treaty that strategically targets technology standardsratherthanemissionlimits[6,8].Inbothofthesecases,aself-enforcingtreatysustainscoordinationwithfull participation.21Thenoveltyinthepresentpaperisthatheretheopportunityforcoordinationisnotstrategicbutrathera giftfromMotherNature. Having established the preeminent role of coordination in averting catastrophe under conditions of certainty, in the remainderofthispaperIshalllimitmyattentiontotheconditionsunderwhichcoordinationcanreduce,ifnoteliminate, theprobabilityofcrossingacatastrophicthresholdunderconditionsofuncertainty. 5. Impactandthresholduncertainty [27]emphasizesfat-taileduncertaintyinclimatesensitivity,compoundedbyuncertaintiesintranslatingtemperature changes into welfare changes. Here I distinguish between uncertainty in the impact of catastrophe (X) and uncertainty (cid:2) (cid:3) aboutthethresholdthattriggerscatastrophe Q . Thetwouncertaintiesarebothverysubstantial,butalsoverydifferent.Onewaytounderstandthenatureofthreshold uncertaintyistoconsidertherecentattention-gettingpaperbyRockstro¨metal.[19].Theyidentifya‘‘planetaryboundary’’ intermsofatmosphericCO concentrationsof350ppmbyvolume(ppmv)soas‘‘toensurethecontinuedexistenceofthe 2 largepolaricesheets,’’whilenotingthatthepaleoclimaticrecordimplies‘‘thatthereisacriticalthresholdbetween350 and 550ppmv.’’ Uncertainty about this threshold is purely scientific (and much of it is substantially irreducible). Uncertaintiesabouttheimpactsassociatedwiththelossofpolarice,bycontrast,arebothscientificandeconomic.Lossof theGreenlandIceSheet,forexample,wouldlikelycausesealeveltorisebetween2and7moveraperiodof300toover 1000years(Lenton[16]).ThevalueofXwillobviouslydependontheextentofsealevelriseanditsrateofchange,butit willalsodependonthevaluesattachedtothesechanges,includingdiscounting.Muchhasbeenmadeintheeconomics literatureoftheimportanceofthesevaluesandtheiruncertainties,butIshallnowshowthat,asregardstheprospectsfor collectiveaction,uncertaintyabouttheimpactofcrossingathresholdisrelativelyunimportant.Itisscientificuncertainty aboutthethresholdthatreallymatters. Considerfirstuncertaintyabouttheimpactofcatastrophe,X.Itisstraightforwardtodemonstratethatalltheresults shownthusfarcarrythroughifwesubstitutetheexpectedvalueofX,E(X),forX(this,again,isassumingthatcountriesare 19Forexample,itwouldonlypayN(cid:2)1countriestocooperatesoastoensurethatthethresholdwasavoidedwhentheothercountry,thefreerider, reducedemissionsbyb=c,ifbQ(cid:2)ðc=2ÞððQ(cid:2)b=cÞ=ðN(cid:2)1ÞÞ2ZbðbðN(cid:2)1Þ2=cþb=cÞ(cid:2)X(cid:2)ðc=2ÞðbðN(cid:2)1Þ=cÞ2:Butifthisconditionapplies,then(6)surelyholds, whichmeansweareinacoordinationsituation. 20RecallthatthewedgeforcooperationinFig.2borderstheregioninwhichcatastropheavoidanceisinefficient. 21InBarrett[4,6],traderestrictionssustainafirstbest.InBarrett[6,8],technologystandardsmaysustainonlyasecondbest.Inthepresentpaper, theprospectofcatastrophemaysustainafirstbest;itwouldsustainasecondbestonlyundertheconditionsdescribedtowardstheendoffootnote13. Withthresholduncertainty,asexplainedinthenextsection,acoordinatingtreatycanonlysustainasecondbest,exceptforaspecialcasediscussedin Section6. Pleasecitethisarticleas:S.Barrett,Climatetreatiesandapproachingcatastrophes,JournalofEnvironmentalEconomics andManagement(2013),http://dx.doi.org/10.1016/j.jeem.2012.12.004i S.Barrett/JournalofEnvironmentalEconomicsandManagement](]]]])]]]–]]] 9 Fig.3. (a)Prospectofcatastrophehaslittleeffectonthefullcooperativeoutcome.(b)Prospectofcatastrophehasprofoundeffectonthefullcooperative outcome.(c)Prospectofcatastrophewith‘‘thin’’and‘‘fat’’tails. riskneutral).22Uncertaintyaboutthemagnitudeofcatastrophicdamagesdoesnotfundamentallyalterthenatureofthe cooperation challenge. So long as the catastrophic threshold is certain, and E(X) is large, countries will be able to coordinatesoastoavertcatastrophe.23 Consider now uncertainty about the threshold, Q. In particular, let Q be a random variable with a continuous (cid:2) (cid:3) cumulativeprobabilitydistributionFðQÞ¼Pr QrQ .Werecountriestocooperatefully,theywouldnowmaximize E(cid:7)Pc(cid:8)¼bQN(cid:2)Xcq2i (cid:2)XN½1(cid:2)FðQÞ(cid:3), ð9Þ 2 i whichrequires bN(cid:2)cqþXNfðQÞ¼0, ð10Þ i wherefðQÞistheprobabilitydensityfunction(pdf).Notethat,ingeneral,Eq.(10)willnotbesufficientforamaximum. Theeffectofuncertaincatastrophedependsonthefunctionf.Ifthepdfhasinfinitesupports,fwillalwaysbepositive, andtheprospectofcatastrophewillcommendgreaterabatement,comparedtoasituationinwhichcatastropheisignored. Recallthatwhencatastropheiscertain,abatementinthefullcooperativeoutcomemayormaynotbeaffected. Fig. 3a illustrates Eq. (10) for a plausible pdf (fðQÞ).24 In the figure, the bell curve placed on top of the aggregate marginal benefit of avoiding ‘‘gradual’’ climate change represents XNfðQÞ, the expected aggregate marginal benefit of catastrophe avoidance. For the pdf shown in this figure, concern about catastrophic climate change increases the full cooperativeabatementlevelveryslightlyabovebN2=c,theoptimalabatementlevelforgradualclimatechangeignoring catastrophic climate change. If abatement is increased beyond QFC, net benefits fall. They later increase before falling again,butsolongasareaAexceedsareaB,asitdoesinFig.3a,thenetbenefitsofincreasingabatementbeyondQFCwillbe negative,anditwillbeoptimaltolimitaggregateabatementtoQFC. Fig.3blookssimilarto3a,butinthiscaseareaBexceedsareaA,andtheoptimalaggregateabatementlevelincreases dramatically,relativetobN2=c,duetotheuncertainprospectofcatastrophe.Obviously,whatmattersinthesefiguresis thesizeofareaArelativetoB.Theserelativevaluesdependinturnonthevaluesofb,X,andcaswellasthepdfforthe threshold. Does the possibility of ‘‘fat tails’’ matter? As shown in Fig. 3c, a thickening of the tails of the pdf changes the full cooperative outcome very little, with the full cooperative abatement level increasing only slightly, from QFC to QFC.25 1 2 Weitzman,ofcourse,getsaverydifferentresult,butasnotedintheintroduction,heassumesthatsocialpreferencesare stronglyriskaverse. Thisbriefdiscussionsummarizesjustafewofmanypossibleconstructions.Myaimhereisnottobecomprehensive, but to illustrate how sensitive or insensitive policy recommendations can be to small changes in the constituent parts ofEq.(10). 22IamalsoassumingthatE(X)exists.Forsomefat-taileddistributions,expectedvaluedoesnotexist. 23ThisresultshouldhavebeenanticipatedfrommypreviousdiscussionoftheMilinskietal.[17]experiment.Recallthat,inthisexperiment,the valueofX(andnotQ)wasuncertain.AnexampleofasituationinwhichthethresholdwascertainbuttheimpactswereuncertainistheMillenniumBug orY2Kproblem.Ofcourse,thisturnedoutnottobethecatastropheexpected,thoughweshallneverknowwhetherthisisbecausetheactiontakento avertacrisissucceededorearlierfearsofcalamitywereoverblown.Thetheorypresentedhereisconsistentwithbothinterpretations. 24Allthepdfsdrawninthispaperaresymmetric,butmydiscussionandanalysisdoesnotdependonthisassumption.Therearereasonstobelieve thepdfforclimatesensitivityexhibitspositiveskewness(RoeandBaker[20]). 25Thediagramcanbethoughtofascomparingthenormalandt-distributions,theformerbeingthin-andthelatterthick-tailed. Pleasecitethisarticleas:S.Barrett,Climatetreatiesandapproachingcatastrophes,JournalofEnvironmentalEconomics andManagement(2013),http://dx.doi.org/10.1016/j.jeem.2012.12.004i 10 S.Barrett/JournalofEnvironmentalEconomicsandManagement](]]]])]]]–]]] Tosumup,wehave: Proposition5. Uncertaintyabouttheimpactof‘‘catastrophic’’climatechangehasnoeffectonthefullcooperativeabatement level.Bycontrast,uncertaintyaboutthethresholdmaycauseabatementinthefullcooperativeoutcometoincreasealotrelative tothefullcooperativeoutcomeforlimiting‘‘gradual’’climatechange. Ifcountriesfailtocooperate,eachcountryiwillchooseq tomaximize i cq2 EðpÞ¼bQ(cid:2) i (cid:2)X½1(cid:2)FðQÞ(cid:3): ð11Þ i 2 Thesolutionrequires b(cid:2)cqþXfðQÞ¼08i: ð12Þ i Whatdeterminesthegap betweenthefullcooperative and non-cooperativeoutcomes?Lookfirst atFig. 4a. The top portion of the figure shows the full cooperative outcome. Though positioned somewhat differently, this portion of the figure is analogous to the situations depicted in Fig. 3a–c. The bottom portion of the figure shows the non-cooperative outcome.Asdrawn,Fig.4ashowsthattheprospectofuncertaincatastrophehasasubstantialeffectonthefullcooperative outcome(raisingitfrombN2=c toQFC), butvirtuallynoeffectonthenon-cooperativeabatementlevel(which,ignoring catastrophe,isbN=c,andwhichincreasesjustalittletoQNC,assumingthatthepdfhasaninfinitesupportontheleftside). Uncertaintyaboutthethresholdrobsusofanychanceforcoordinationonthefullcooperativelevelofabatement.Indeed, comparedtothecertaintycase(assumingthatthecertainthresholdisequaltoitsexpectedvalueinFig.4a),uncertainty aboutthethresholdmakescountrieswanttoabatemore,andyetcausesthemtoabateless. Fig. 4b illustrates a situation in which the prospect of uncertain catastrophe increases abatement in both the full cooperativeandnon-cooperativeoutcomes(again,relativetoasituationinwhichtherewasonlygradualclimatechange), while at the same time narrowing the gap between the non-cooperative and full cooperative outcomes (ignoring catastrophe, this gap is bN2=c(cid:2)bN=c; taking catastrophe into account, it is QFC(cid:2)QNC). In this case, even though the thresholdisuncertain,theprospectofcatastrophecreatesanopportunityforcoordinationonthemutuallypreferred(but stillinefficient)Nashequilibrium.Itisobviousfromthefigure,however,thatthecircumstancesthatsupportthismore cheerfulsituationarenotnecessarilytobeexpected.Theyincludenotonlyahighimpact,X,butalsoaverylowvariancein thepdfandasmallN—circumstancesthatarenotfavoredbyclimatechange. Again,myaimherehasnotbeentobecomprehensive,buttoindicatewhencollectiveactionislikelytosucceedorfail. Themainresultsmaybestatedsimply: Proposition6. Comparedtothecertaintycase,uncertaintyabouttheimpactofcatastrophehasnoeffectoncollectiveaction. By contrast, uncertainty about the threshold is critical, being very likely to transform the collective action problem from a coordinationgameunderthresholdcertaintytoaprisoners’dilemmaunderthresholduncertainty. Fig.4. (a)Uncertainprospectofcatastrophewidensthegapbetweenthenon-cooperativeandfullcooperativeoutcomes.(b)Uncertainprospectof catastrophenarrowsthegapbetweenthenon-cooperativeandfullcooperativeoutcomes. Pleasecitethisarticleas:S.Barrett,Climatetreatiesandapproachingcatastrophes,JournalofEnvironmentalEconomics andManagement(2013),http://dx.doi.org/10.1016/j.jeem.2012.12.004i

Description:
I am grateful to Apurva Sanghi, Arthur Campbell, Astrid Dannenberg, Claude .. Knowing how the deviant country will behave, they will prefer to abate.
See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.