DecomposingClimateRisksinStockMarketsYuanchenYang,ChengyuHuang,YuchenZhangWP/23/141IMFWorkingPapersdescriberesearchinprogressbytheauthor(s)andarepublishedtoelicitcommentsandtoencouragedebate.TheviewsexpressedinIMFWorkingPapersarethoseoftheauthor(s)anddonotnecessarilyrepresenttheviewsoftheIMF,itsExecutiveBoard,orIMFmanagement.2023JUN“TheauthorswouldliketothankKoshyMathai,PierpaoloGrippa,TannousKass-Hanna,LiamMasterson,LucyLiu,AugustusPanton,DivyaKirti,HenkJanReinders,andparticipantsattheCanada2022ArticleIVAnalyticalSeminarforcomments,andJerryChavesforresearchinputs.©2023InternationalMonetaryFundWP/23/141IMFWorkingPaperWesternHemisphereDepartmentDecomposingClimateRisksinStockMarketsPreparedbyYuanchenYang,ChengyuHuang,YuchenZhangAuthorizedfordistributionbyKoshyMathaiJune2023IMFWorkingPapersdescriberesearchinprogressbytheauthor(s)andarepublishedtoelicitcommentsandtoencouragedebate.TheviewsexpressedinIMFWorkingPapersarethoseoftheauthor(s)anddonotnecessarilyrepresenttheviewsoftheIMF,itsExecutiveBoard,orIMFmanagement.ABSTRACT:Climatechangeposesanunprecedentedchallengetotheworldeconomyandtheglobalfinancialsystem.Thispapersetsouttounderstandandquantifytheimpactofclimatemitigation,withafocusonclimate-relatednews,whichrepresentsanimportantinformationsourcethatinvestorsusetorevisetheirsubjectiveassessmentsofclimaterisks.Usingfull-textdatafromFinancialTimesfromJanuary2005toMarch2022,wedevelopmachinelearning-basedindicatorstomeasurerisksfromclimatemitigation,andthedirectionoftheriskisidentifiedthroughmanuallabels.ThedocumentedriskpremiumindicatesthatclimatemitigationnewshasbeenpartiallypricedintheCanadianstockmarket.Morespecifically,stockpricesreactpositivelytomarket-wideclimate-favorablenewsbuttheydonotreactnegativelytoclimate-unfavorablenews.Theresultsarerobusttodifferentmodelspecificationsandacrossequitymarkets.RECOMMENDEDCITATION:Yang,Yuanchen,ChengyuHuang,andYuchenZhang,2023.DecomposingClimateRisksinStockMarkets.IMFWorkingPaperNo.23/141.JELClassificationNumbers:G11,G12,Q54Keywords:ClimateMitigation;MachineLearning;AssetPricingAuthors’E-MailAddress:yyang6@imf.org;chuang@imf.org;yzhang5@imf.orgWORKINGPAPERSDecomposingClimateRisksinStockMarketsPreparedbyYuanchenYang,ChengyuHuang,YuchenZhangIMFWORKINGPAPERSDecomposingClimateRisksinStockMarketsINTERNATIONALMONETARYFUND2Contents1.Introduction.....................................................................................................................................................32.LiteratureReview............................................................................................................................................63.MethodologyandData..................................................................................................................................104.ResultsandDiscussion................................................................................................................................165.RobustnessChecks......................................................................................................................................216.Conclusion.....................................................................................................................................................26References.........................................................................................................................................................27Appendix............................................................................................................................................................31FIGURESFigure1.WordCloudofClimatePolicyKeywords.............................................................................................11Figure2.ResearchMethodologyFlowchart.......................................................................................................15Figure3.IntensityofClimateNewsCoverage....................................................................................................16Figure4.IntensityofClimateFavorableandUnfavorableNews........................................................................17TABLESTable1.PricingofClimateNewsFactorinCanadianStockMarket...................................................................18Table2.PortfolioSortingAnalysisofCanadianStockMarket............................................................................19Table3.PricingofClimateNewsFactorintheUSStockMarket.......................................................................20Table4.PricingofClimateNewsFactorintheEUStockMarket.......................................................................20Table5.PricingofClimateNewsFactorUsingAlternativeAssetPricingModels..............................................21Table6.PricingofClimateNewsFactorUsingAlternativeClimateRiskMeasures...........................................22Table7.PricingofClimateNewsFactorUsingDifferentSortingStrategies......................................................22Table8.PricingofClimateNewsFactor(Canada-specific)inCanadianStockMarket.....................................23Table9.PricingofClimateNewsFactor(WithoutESG)inCanadianStockMarket..........................................24Table10.PricingofAlternativeClimateNewsFactorinCanadianStockMarket...............................................25IMFWORKINGPAPERSDecomposingClimateRisksinStockMarketsINTERNATIONALMONETARYFUND31.IntroductionTheworld’sclimateischanging.Thereisgeneralscientificconsensusthatincreasedgreenhousegasconcentrationsareattributabletohumanactivities(IPCC,2021).Theaverageglobalsurfacetemperaturecouldriseby3-6degreesCelsiusby2100withoutimmediateactiontoslowthepaceofglobalwarming(OECD,2021).Withincreasingglobalsurfacetemperature,thelikelihoodandintensityofnaturaldisasterswillalsoincrease.Climatechangeisanexistentialthreat.Whilecountriesbroadlyagreeonreachingnet-zeroemissionsbymid-centurytoavoidthemostadverseclimatechangescenarios,uncertaintiessurroundthetransitiontowardalow-carboneconomy.Ifgovernments,firms,andhouseholdsfailtochecktemperatureriseandmitigateclimatechange,disorderlyadjustmentsinassetpriceswouldoccur,withpossibledisruptiontotheproperfunctioningofthefinancialsystemandpotentialspilloverstoothersectorsoftheeconomy.Thispapersetsouttoexaminehowthemarketviewsandpricesclimatemitigationpolicies,withafocusontheCanadianeconomy.Itiswidelyrecognizedthatrisksofclimatechangescanbedividedintotwotypes:physicalrisksandtransitionrisks(IMF,2022).Theformerreferstotherisksstemmingfromsevereclimateevents,suchasfloods,droughts,wildfires,etc.,whereasthelatterresultsfrompolicychanges,technologicaladvances,andmarketshiftsintheprocessofadjustingtoalow-carboneconomy.Ofdifferenttypesoftransitionrisks,policychangeistheonethathasattractedthemostattention.Withthenet-zeropledgeconstantlycallingformoremitigationactions,itwouldbeimportanttounderstandtheeffectivenessofexistingmitigationpolicies,andinparticular,whethercurrentpoliciesprovideadequateincentivestosteeragents’behaviortowardslowercarbonemissions.Oneofthechannelsforpoliciestoshapeagents’behavioristhroughpriceadjustmentsinfinancialmarkets.AssessingtheimpactofmitigationpoliciesisparticularlyrelevanttoCanada,whichisamongtheworld’stopcarbonemittersandfacesamajortransformationastheworldmovesawayfromfossilfuels.(GovernmentofCanada,2020;EnvironmentandClimateChangeCanada,2022)Thegovernmentispushingforstrongerclimatepoliciesatboththefederalandprovinciallevels,withprofoundimplicationsforitsoilandgassector.PrimeMinisterTrudeauannouncedduringtheUnitedNationsClimateChangeConferenceinGlasgowthatCanadawillbethefirstmajoroil-producingeconomytocapandreducepollutionfromtheoilandgassectortonetzeroby2050.Facingahardpathtodecarbonize,thesectorwillneedtocontinuetoadapt,inordertospurinnovationandpreservejobs.However,itisnotyetwellunderstoodhowthesectorwillrespondtomarketandgovernmentalregulatorypressureforgreentransition.Thekeychallengetoanalyzingmarketresponsestoclimatemitigationistodevelopatimeseriesthatadequatelycapturesrisksfromimplementingmitigationpolicies.Oneapproachistoleveragethefactthatpolicyeventsthatpotentiallysignifyshiftsinclimatepoliciesareoftencoveredbynewspapers.Infact,newsmayevenserveastheprimarysourceofinformationforinvestorstoformulatetheirsubjectiveprobabilitiesofclimaterisks.Studyingtheresponsivenessofsharepricestonewscoverageofclimatepoliciesoffersimportantadvantagesoverperformingeventstudiesaroundmajorclimate-policyevents.(Bhattacharyaetal.,2009;Cahanetal.,2009)Inparticular,anevent-studymethodologywouldnotallowtheunexpectedcomponentsofpolicyshockstobeeasilydisentangled,norwoulditallowmeasurementoftheintensityof,orsentimentIMFWORKINGPAPERSDecomposingClimateRisksinStockMarketsINTERNATIONALMONETARYFUND4towards,theseshocks.Onthecontrary,newstext,whichhasincreasinglybeenusedinfinancialeconomicsresearch,capturesunexpectednewinformationthatisjustacquiredbyinvestorsandforetellschangesinfutureinvestmentopportunities.Newsarticlesencompassdiversenarrativesandaretimelyandfocusedontherisksmostpertinenttothemarket,andinthespiritofcanonicalassetpricing,newscoveragemayserveasthebestofallpubliclyavailablesignals.(Huangetal.,2019;Barkemaetal.,2021)Weproposeanovel,machinelearning-basedmethodtoinvestigatehowmarket-wideclimatemitigationrisksarepricedinCanadianstocks.Weusestate-of-the-artNaturalLanguageProcessingmodeltoperformtextualanalysisofnewsonclimatemitigationpoliciesthatappearedinFinancialTimesover2005-2022.Risksfrommitigationareidentifiedthroughacombinationofdetailednarrativeanalysisandsupervisedlearningalgorithms.AsetofassetpricingmodelsareappliedtoCanadianoilandgascompanieslistedintheS&PTSXcompositeindextoexaminewhethermodel-generatedclimateriskfactorsareassociatedwithpositive/negativeriskpremia.Finally,weexplorewhetherCanadianfirmsaremoresensitivetonewsonclimatemitigationbycomparingthemwithrepresentativeUSandEUfirms.Ourresultsprovideevidencethatstockpricesofoilandgascompaniesincorporateinformationaboutclimatemitigationpolicies.Toavoidneutralizingpositiveandnegativenewslabels,wegeneratetwoclimatenewsindices,withonesignalingpositivenewsforclimatemitigationandthuslowertransitionrisk,andtheothersymbolizingnegativenewsforclimatemitigationandthushighertransitionrisk.Wehypothesizethatanincreaseinthepositive(negative)climatefactorsignalsstricter(lighter)mitigationpolicies,andthusshouldbebad(good)newsforoilandgascompanies.Inresponsetosuchanegative(positive)shock,investorswouldsell(buy)oilandgasstocks,thusdecreasing(increasing)theirpricesandincreasing(decreasing)theirreturns.Consequently,weshouldobserveastatisticallysignificantcoefficientontheclimateriskfactor.Theresultsconfirmourhypothesis,indicatingthatfirmshavestartedtoincorporateclimatefactorsintheirportfolioconstruction.Theresponsesofstocksintheoilandgassectortopositiveandnegativenewsaboutmitigationpoliciesareasymmetrical.Whilethereisasignificantlypositivecoefficientassociatedwithclimate-unfavorablenewsamongthegroupofoilandgascompanies,wefindthatthesensitivitytiedtoclimate-favorableisnegativebutstatisticallyinsignificant.Inotherwords,aneaseinclimatemitigationconstitutesafavorableshocktooilandgascompaniesbuttightermitigationpoliciesdonotnecessarilysignalanegativeshock.Theresultsarerobusttovariousassetpricingmodels,includingmarketmodel,Fama-Frenchthreefactormodel,Fama-Frenchfivefactormodel,Carhartfourfactormodel.CanadianfirmsareslightlymoresensitivetobothclimatefavorableandunfavorablenewsthanUSandEUfirms.Theefficiencyandextentofassetpricingcouldvaryacrossmarkets.Weconductacross-countrycomparisonbyperformingthesamesetofexercisesonUSandEUcompanies.Inthebaselinemodel,wefindthatthecoefficientsontheclimateriskfactorofUSandEUcompaniesareslightlysmallerinmagnitude,relativetothatofCanadiancompanies.Overall,USandEUoilandgascompaniesalsorespondtoclimatepolicynews,butwithdifferentdegreesofsensitivityfromthosebasedinCanada.Ourstudymakesanoriginalcontributiontotheliterature.First,weexploitnewsdatatoestablishanovelindicatorofmarket-wideclimaterisks.Previously,Engleetal.(2020)reliedonnewsdatatomeasureclimaterisksanddesignhedgingstrategies.Buildingontheirwork,Faccinietal.(2021)furtherdisentangleddifferentdimensionsofclimaterisksusinganarrativemethod.Ourpapercombinesthemeritsofbothstudies.WeIMFWORKINGPAPERSDecomposingClimateRisksinStockMarketsINTERNATIONALMONETARYFUND5presentevidenceonwhattypesofclimaterisksarepriced,enabledbyanovelmethodthatcombinesmanuallabelingandmachinelearningalgorithms.Second,wedocumenttowhatextentoilandgassectorstocksareexposedtoclimaterisks,demonstratinghowtraditionalassetpricingapproachescaninformclimatefinance.Inaddition,weperformcross-countrycomparisononclimateriskpricing,whichoffersimportantinsightsintowhichmarketsaremoreexposedtoclimatepolicyrisks.Takentogether,usingmachinelearning-basednewsmeasures,weshowthatnewsonclimatemitigationhasbeenpartiallypricedinthestocksofCanadianoilandgascompanies.Aneaseinmitigationpoliciesrepresentsapositiveshocktotheirstockpricesbutstrongermitigationpoliciesdoesnotnecessarilyleadtonegativepriceeffects.Thesefindingssuggestthatclimatemitigationriskscapturedbynewscoveragehavelimited,asymmetricimpactonstockvaluations.Therestofthepaperisorganizedasfollows.Section2reviewstheliteratureonclimateriskandassetpricing.Section3introducesourmethodologyanddata.Section4presentsanddiscussesourresults.Section5providesrobustnesschecks,andSection6concludes.IMFWORKINGPAPERSDecomposingClimateRisksinStockMarketsINTERNATIONALMONETARYFUND62.LiteratureReview2.1ClassificationofClimateRisksItiscommontoclassifyclimaterisksintophysicalriskandtransitionrisk.Theformerisdefinedasrisksarisingfromphysicalimpactofclimatechange,whereasthelatterreferstorisksresultingfrompolicy,technology,legal,andmarketchangesthatoccurduringthemovetoalow-carboneconomy.(IMF,2022)Theimpactsofclimateriskdrivers,whetherphysicalriskortransitionrisk,oneconomiesandfinancialmarketsmayvarywidelydependingongeolocationsandbecomeincreasinglyhardtopredict.Physicalriskdriverscanbefurtherclassifiedintoacuterisksassociatedwithextremeweatherevents(Network,2019),andchronicrisksrelatedtogradualclimateshifts(McKinseyGlobalInstitute,2020).Ithasbeenestimatedthatnaturaldisasters,mostofwhichclimatological,ledtoroughly$5.2trillionlossesfrom1980sto2018(MunichRe,2020).Transitionriskdriversareglobal,althoughthespecificnatureofeachriskdrivermayvarybyeconomy.Variousstresstestresultsindicatethatthelossesforfinancialinstitutionsintheeventofadisorderlyenergytransitioncouldbesizeable.Whilebanksandthenon-bankfinancialsectorhavebeenaffectedby,andhavethereforecloselymonitored,theseformsofchangeslinkedtoclimatetransition,thesynchronousnatureoftransition-relatedchangeshavethepotentialtoresultinascaleofimpactthatismuchgreaterthanpreviouslyanticipated.(NetherlandsBank,2018)Inthisstudy,wefocusonclimatetransitionrisk,particularlytheriskfrommitigationpolicies,hopingtoshedlightontheeffectofchangingpolicies,legislation,andregulationassocietyandindustryworktoreducetheirrelianceoncarbonandimpactontheclimate.2.2MeasurementofClimateRisksUnderstandingclimaterisksformsthebasisfordevelopingappropriateclimatestrategies.Whileconventionalriskassessmenttoolsmayserveasastartingpointforclimate-relatedriskmeasurement,climateriskdrivershaveuniquefeaturesthatchallengesthedirectincorporationoftheserisksintoexistingframework.Mostnotably,themostsevereconsequencesofclimaterisksareprojectedtomaterializein30to80years.Theexceedinglylongsimulationhorizonsandtheresultinguncertaintiesassociatedwithclimatemodelingnecessitategranularandsometimesinnovativemeasurementmethodologies.Theassessmentofclimate-relatedrisksbybanksandsupervisorshasfocusedonmappingshort-termtransitionriskdriversintoportfolios,bycapturingthecarbonintensityofportfolioexposures,creatinginternalclimateriskscoresorratings,etc.Whilecarbonintensityconstitutesastraightforwardmeasurementofanentity’scarbonfootprint,itproveschallengingtotrackindirectemissionsthatoccurinafirm’svaluechainandinaproduct’slifecycle.Since2014,interestinclimate-relatedfinancialriskshasbeenboostedbythedevelopmentofESGratings(HalbritterandDorfleitner,2015).ESG—shortforEnvironmental,Social,andGovernance—hasrisentobecomeanincreasinglyimportantpartofinvestmentdecisions.However,controversiesarealsorisingrelatedtoESGratingstandards,datasources,andpossiblegreenwashing.ESGratingstandardsarevastlydifferentIMFWORKINGPAPERSDecomposingClimateRisksinStockMarketsINTERNATIONALMONETARYFUND7acrossagencies,andcanthereforeresultinaninconsistentmeasureofperformance.Companies,ontheotherhand,areselectivelytransparentaboutwhattheyreport,whichmasksrelevantenvironmentalrisksandcallsintoquestionthereliabilityofsuchratingstoguideinvestmentdecisions.Moreover,ESGinvestingisincreasinglyconsideredasamerebrandingexercisedesignedtobenefitpollutingbusinesseswhichclaimtobeingreentransitionbutdevoteinsignificantandopaquespendingtocleantechnologies.Anotherpossibleapproachtoclimateriskassessmentisthrougheventanalysis.Speakingoftransitionrisk,especiallywhichstemsfrompolicychanges,onemightnaturallythinkofmajorclimate-relatedevents.Forexample,variousinternationalsummits,COP25,COP26.Onewouldimaginethatsurroundingtheseevents,therearelikelychangesintheprobabilityofclimateriskscenarios.However,usingeventstudyalonedoesnotallowustoidentifythedirectionofchange.Whileitislikelythatwiththeagreementstrickenatinternationalsummits,orthepassageofcarbonemissionsbill,weshouldseeafallintransitionrisk,buttooaggressivepolicychangesorfailuretodeliverthemitigationcommitmentscouldalsoincreasetransitionrisks.Takentogether,gapsremaininthemeasurementanduseofclimate-relatedrisks.MostexistingmeasuresrelyonperformanceindicatorsderivedfromESGscores,whichhavebeenmarkedbyincreasingcontroversiesduetothelackoftransparencyaroundtheirmethodologiesanddatasources,themisuseofESGforgreenwashing,potentialconflictofinterests,etc.(Mishra,2022)Incomparison,measuresbasedoncarbonfootprintshouldhaveanadvantage,butsuchinformationisnotwidelyavailableatthefirmlevel.Thefactthatinvestorsdemandfuturecarbontrajectoriesratherthanjustcurrentcarbonemissionsaddedtothedifficultyofusingcarbonintensitytoquantifyclimaterisks(Yang,2021).Takentogether,climateriskassessmentmethodologieshavenotyetreachedmaturity.2.3ClimateRisksandAssetPricesThereisarapidlygrowingliteratureexaminingtheeffectsofclimaterisksonfinancialmarketsandassetprices.Tobringclimateriskintoassetpricingmodels,onemustestimatethefundamentalvalueofassetsundervariousclimatechangescenarios.However,thisassessmentiscomplicatedbytheuncertaintyaboutthefutureevolutionoftheclimate,thefuturepathoftheeconomy,andthepotentialinteractionbetweentheclimateandtheeconomy.Giventhegeographicaldiversityofriskexposures,thenonlinearityofriskdistribution,andthefactthathistoricalexperiencecouldprovidelittleguidanceforfutureriskcalibration,eachsourceofuncertaintycanmakeahugedifferenceeffectonequilibriumpricesandriskpremia.Recentevidencehasshownthatstockmarkethasstartedtopriceclimaterisks(Engleetal.,2020;Faccinietal.,2021).Whiletheliteratureonclimatechangeandassetpricinghavetakendifferentdirections,studieshaveinvariablyshownthatclimatechangecouldleadtoarelativediscountofclimate-vulnerableassets.Insomecases,themovementinpricesconsistentwithwhatonemightexpect,whereasinothercases,marketseemstounder-orover-reactbeforereturningtorationallevels.Forourpurposes,wefocusontheequitymarketreactionstoclimate-relatedrisks.Hong,Li,andXu(2016)demonstratethataportfoliothatshortsfood-sectorstocksindrought-strickencountriesandlongsfood-sectorstocksindrought-freecountrieswouldhaveproduceda9.2percentannualizedreturnfrom1985to2015.Sincethepremiumislargerfordrought-affectedcountrieswheretherewasnotmuchhistoryofdroughtspriorto1980,theauthorsconcludethatclimatesurprisesaredrivingthispremium.IMFWORKINGPAPERSDecomposingClimateRisksinStockMarketsINTERNATIONALMONETARYFUND8Bansal,Kiku,andOchoa(2016)developatemperature-augmentedlong-runriskmodeltoexaminehowinfinitelylivedrationalagentscapitalizelong-termclimateriskintoforward-lookingassetprices(Delletal.,2012;Severenetal.,2018).BuiltonEpstein-Zinpreferences,theirmodelsimultaneouslymatchestheprojectedtemperaturepath,theobservedconsumptiongrowthtrajectory,thediscountrateimpliedbytherisk-freerate,andthenegativeelasticityofstockpricestotemperaturechanges.Theresultssuggestasignificantsocialcostofcarbonandmotivateearlyactiontomitigateglobalwarming.Itisworthnotingthattheyassumecertainsectoralportfolios,includingmining,oilandgasextraction,construction,transportation,andutilities,tohaveahighsensitivitytoclimate-relatedrisk,whereasthosewithalowsensitivityaremanufacturing,wholesale,retailtrade,services,andcommunications.KarpandRezai(2017)adoptanoverlappinggenerations(OLG)modelingapproach,tryingtodisentanglepeople’sincentivestotackleclimatechange.Peoplemightreducecarbonemissionstoprotectthemselves,theirwealth,orfuturegenerationsfromclimatedamage.Basedonanoverlappinggenerationsclimatemodelwithendogenousassetpriceandinvestmentlevels,theirfindingssuggestthatassetmarketscapitalizethefutureeffectsofclimatepolicy,regardlessofpeople’sconcernforfuturegenerations.Climatepolicycanexertsubtledistributionaleffectsacrossthecurrentlylivinggenerations,andmarketscanleadself-interestedagentstoundertakesignificantabatementefforts.Asmallpolicychangethatraisesthepriceofcapitalincreasesoldagents’welfareandinthemeantime,increaseswelfareofyoungagentswithahighintertemporalelasticityofsubstitution.Pankrantz,Bauer,andDerwall(2019)demonstratethatextremelyhightemperatureeventsnegativelyaffectbothrevenuesandoperatingincomeinwaysthatmarketanalystsdidnotanticipate.Theycreatefirm-specificmeasuresofheatexposureandlinkittofirmfinancials,analystforecasts,andearningsannouncementreturnsofacross-countrysampleof4,400listedfirmsfrom1995to2017.Byestimatingtheimpactofrandomlydistributedvariationinthenumberofdaysonwhichfirmswereexposedtoextremelyhightemperatures,theauthorsfindthatincreasingexposuretoheatreducesrevenuesandoperatingincome.Moreover,thedeviationofanalystestimatesfromactualperformanceindicatorsaswellastheabnormalreturnsaroundearningsannouncementsbecomesmorenegativewiththeincreaseofheatexposureatthefirms’locations.Thestudylendsevidencethatinvestorsdonotanticipatetheeconomicimplicationsofheatasafirst-orderclimatephysicalrisk.Kumar,Xin,andZhang(2019)addtotheevidencethatstockreturnsaresensitivetoabnormaltemperaturechanges.Exploitinganewmeasureoffirm-leveltemperaturesensitivity,theauthorsfindsignificantoverpricingonstockswithhightemperaturesensitivity.Morespecifically,atradingstrategythatexploitsthisinformationgeneratesanannualizedrisk-adjustedreturnof4.22%duringthe1968-2020sampleperiod.Meanwhile,thesefirmshavelowerfutureprofitability,riskiercorporatepolicies,andloweraveragereturns.Theauthorsnotethatthemispricingcouldbedrivenbyinstitutionalinvestorswhohavelowerportfolioweightsinfirmswithhightemperaturesensitivitiesandbysell-sideequityanalystswhoissueforecastthatarelessaccurateforthesefirms.Inotherwords,financialmarketsunder-reacttofirm-specificinformationaboutclimatechange,thusleadingtopredictablepatternsinstockreturns.Otherstudieshavelookedatequity-basedfinancialmetricsinsteadofpurepricemeasures.Chava(2014)concludesthatstocksassociatedwithsubstantialemissionsandclimatechangeconcernshaveahighercostofequityanddebtcapital.Theauthorusestheimpliedcostofcapitalimputedfromanalysts’earningstoformulateestimates.ElGhouletal.(2016)usecross-countrydatatoshowthatmanufacturingfirmsacross30countriesthatinvestincorporateenvironmentalresponsibilityhavealowercostofequitycapital.GinglingerIMFWORKINGPAPERSDecomposingClimateRisksinStockMarketsINTERNATIONALMONETARYFUND9andMoreau(2021)examinetheimpactofclimateriskoncapitalstructure.Relyingonnoveldatathatmeasureforward-lookingphysicalclimateriskatthefirmlevel,thestudyfindsthatfirmsfacedwithgreaterclimateriskhavelowerleverageduetobothdemandeffect(areductionintheiroptimalleverage)andsupplyeffect(areducedwillingnessonthepartoflenderstofundthem).However,onecompellingcritiquecomesfromthegovernoroftheBankofEngland,MarkCarney,whopointsoutthatwecannotrelyoncurrentactors,whosedecisionhorizonsarelikelytobelessaffectedbyclimatechange,toactcommensuratewiththeinterestsoffuturegenerationsonclimateissues(Carney,2015).Comparedwithphysicalrisk,transitionriskislessstudiedintheliterature,partlybecausethemeasurementofclimatetransitionriskprovesharderandlessstraightforward,partlyduetothefactthatclimatetransitionriskvariessubstantiallyacrosstimeandbycountry,byindustry,andevenbyfirm.Morerecently,researchershavetakenupthechallengetoexaminetransitionrisksinassetprices.Severalstudieshaveemerged,supportingthehypothesisthatinvestorspriceinatleastsometransitionrisk.AccordingtotheBlackrockInvestmentInstitute(2016),climatetransitionrisk,amongotherrisks,hasyettobefullypricedintoassetvalues.Engleetal.(2020)usestandardtoolofassetpricingtobuildportfoliosthatarehedgedagainstinnovationsinclimatechangenews.DespitegrowingcontroversiessurroundingtheirmethodologyanduseofESGscores,theydemonstratehowassetpricingapproachescaninformorevendevelopnewtopicsinclimatefinance.Faccinietal.(2021)buildsonEngle’swork.Theyfurtherdisentangledclimaterisksintoriskfromnaturaldisasters,fromglobalwarning,frominternationalsummits,andfromUSpolicychange.Theyshowedthatonlytheriskofgovernmentinterventionispriced.Ourpapersetsouttofillthegap.Wedevelopnews-basedmeasurestomeasuretransitionrisk.Morespecifically,wefocusontransitionriskfromimplementingclimatemitigationpolicies,whichisdefinedasthemonthlycountofnewsarticlesrelatedtoclimatemitigationpolicies,standardizedbythetotalnumberofmonthlypublishednewsarticles.Therationaleforthisassumption,commonintheliterature,isthatinvestors’attentionislimited,theywilleasilybeinfluencedbywhattheyread,especiallybynewswithintensemediaexposure.orestimatingtheprospectofmorestringentclimateregulation.Usingnewsnotonlyallowsustoquantifytheintensityofshock;thefactthatnewscomesatamuchhigherfrequencyalsoenablesustocaptureanyeventorsentimentchangethatmightmovethemarket.(Dougaletal.,2012;Wanetal.,2021)IMFWORKINGPAPERSDecomposingClimateRisksinStockMarketsINTERNATIONALMONETARYFUND103.MethodologyandData3.1Pre-ProcessingNewsTextDataTobeginwith,weretrievefull-textnewsdatafromFinancialTimesnewsAPI.ThedatabaseisupdatedinrealtimesinceJanuary2005andoursampleperiodstartsfromJanuary2005andendsinMarch2022.Atotalof923,862uniqueEnglishnewsarticlesarecollectedwhichcomesinjsonformatwithanaverageofabout4500newsmonthly.Then,weconductedstandardtextdatapreprocessingstepsincludinglowercaseandlemmatizationwithPartofSpeechTagging,1whichreducewordstotheirbaseformbasedontheirgrammaticalpositionsinsentences.AlltheseproceduresareperformedusingPython’sNaturalLanguageToolkit(NLTK),whichcontainstextprocessinglibrariesfortokenization,parsing,classification,lemmatization,tagging,andsemanticreasoning.Cleaneddatafilestoresnewsinparagraphformatwithnewspublishdateinformation,title,andprocessedfulltext.3.2ClimatePolicyIdentificationInthisstudy,weareinterestedinnewstextsonclimatetransitionrisk,andinparticularriskthatarisesfrompolicychangesthatfolloweconomicandsocietalshiftstowardalow-carbonandmoreclimate-friendlyfuture.Tofocusonnewstextsonclimatechangepolicies,wecreateaglossaryofclimatepolicytermsthatcanbeusedtoexcludeirrelevanttextsfromthewholesamplethroughkeywordsearch.Toconstructourlistoffilteringterms,wefirstcollectclimatechangewhitepapersfromofficialsourcessuchastheIntergovernmentalPanelonClimateChange(IPCC),theEnvironmentalProtectionAgency(EPA),andNationalOceanicandAtmosphericAssociation(NOAA)whichispartoftheUnitedStatesDepartmentofCommerce.Giventheuniquefeaturesofnewsarticles,wecomplementtheseauthoritativetextswithclimateglossariesfromnewsmedia,mostnotablyBBC,WallStreetJournal,andBloomberg.Basedonthiscorpusofauthoritativetexts,wecompileourownlistofclimatechangekeywords.Itisworthnotingthat,toavoidintroducingnoiseintothedataset,weexcludegeneraltermssuchas“climate”and“environment”whichcanalsoappearinirrelevantwordcombinationssuchas“businessclimate”and“businessenvironment”.Instead,weusemorespecificwordcombinationssuchas“climatechange”,“climatetransition”,etc.Tofurtherlimitthesampletotextsonclimatechangepolicies,weextractkeywordsonclimatepolicyinstrumentsfromIPCC,OECD,IMF,WorldBank,InternationalOrganizationforStandardization(ISO),aswellasacademicandpolicyliterature(e.g.,Guptaetal.,2007;Gorlach,2013;WorldBank,2017,2021;Michaelowa,etal.,2018;Penascoetal.,2021;USGS,2022).Thepolicyinstrumentstotackleclimatechangearecategorizedintotwotypes—market-basedinstrumentsandnon-market-basedinstruments—atthemostaggregatelevel.Morespecifically,market-basedinstrumentsrefertopoliciesthatworkthroughpriceadjustmentssothattheexternalcostsofproductionorconsumptionareincorporated.Examplesofmarket-basedinstrumentsincludecarbontaxes,emissionstrading,internationalcarbonpricefloor,etc.Onthecontrary,non-market-basedinstrumentsworkbyencouragingordiscouragingcertainbehaviorthroughnon-1Theparagraphsarelemmatizedwithpostagwhichreturnthewordstobaseformbyconsideringtheirgrammaticalpositionsinsentences.IMFWORKINGPAPERSDecomposingClimateRisksinStockMarketsINTERNATIONALMONETARYFUND11monetaryincentives,orbyimposingobligations.Typicalnon-market-basedinstrumentsaremandatory/voluntaryemissionsdisclosure,cleantechnologysupport,andcommand-and-controlregulations.InFigure1,wesummarizethelistofclimatechangeandpolicykeywordsintheformofawordcloud.ThesizesofwordsorwordcombinationsareproportionaltotheirTF-IDFscores2inthecorpus.Figure1.WordCloudofClimatePolicyKeywordsSources:IPCC,EPA,NOAA,OECD,IMF,WorldBank,ISO,andauthorcalculations.Clearly,“climatechange”isthemostsalientkeyword,showingthethemeofourstudy.Sinceweaimatmeasuringclimatetransitionriskthatstemsfrompolicychanges,termssuchas“government”,“regulation”,and“tax”alsoappearwithhighfrequency.Ourfilteringstandardistoincludenewsparagraphsthatcontainatleastonewordorwordcombinationfromtheclimatechangekeywordlist,andatleastonefromthelistofpolicyinstruments.Weendupwithatotalof3.7millionnewsparagraphsandanaverageof17.6thousandparagraphspermonthfromJanuary2005toMarch2022.3.3GeneratingTrainingDataforClimateTransitionRiskIdentificationHavingobtainedthesampleofnewsparagraphsonclimatechangepolicies,ournextgoalistoidentifywhethereachparagraphsignalsanincreaseordecreaseinclimatetransitionrisk,ataskthatisimpossibletobeaccomplishedbymanpowerduetothepurenumberofparagraphsawaitinganalysis.Weneedmachinelearningalgorithmstoperformthetask.Ideally,machinelearningmodelscouldhelpusclassifyallnewsparagraphsintopredefinedgroups—thegroupofnewsparagraphswhosecontentflagsanincreaseinclimatetransitionriskandthegroupofparagraphsthatannounceadecreaseinclimatetransitionrisk.However,toreliablytellwhetherapieceofnewssignifiesariseordropintransitionriskhasalwaysbeenachallenge,becauseassessingtheriskisnotonlysubjective,asthe2TF-IDFscoreisanumericalstatisticthatisintendedtoreflecthowimportantawordistoadocumentinacollectionorcorpus.Formoredetails,pleaserefertoTF-IDFWiki.IMFWORKINGPAPERSDecomposingClimateRisksinStockMarketsINTERNATIONALMONETARYFUND12sameclimatepolicymaybegoodnewsforsomeindustrieswhereasbadnewsforothers,butalsorequiresconsiderableexpertise.Therefore,theprocessoftextclassificationcannotbeentirelyautomated.Tosolvethisproblem,wefollowthesupervisedlearningapproachwithcustomizeddatalabels.Tolabelthenewsdatawithrelativelyhighconfidence,weconductadetailednarrativeanalysistodissectthenewscontent.Asthefirststep,werandomlyselectroughly800representativeparagraphsthatspreadacrossthesampleperiod,whichconstitutearestrictedsample.Agroupoffourresearchersthenreadthrough50paragraphsandagreeonthelabelingcriteria.Thegeneralruleisthatwelabeltheparagraphsbasedonwhetherthecontentimpliesstrictermitigationpoliciesorlightermitigationpolicies,andlabelitasclimatefavorableorunfavorable.3Examplesofmanuallylabellednewsparagraphsarepresentedintheappendixtohelpillustrateourlabelingcriteria.Therestoftherestrictedsampleisthensplitinto𝐶𝐶42,orsixportions.Eachresearcherlabelsthreeportionsindependently,sothateachportionisbeinglabelledbytworesearchers.Ifthetworesearchersgivethesamelabel,weacceptthelabelasitis.Iftheydisagree,wewouldintroduceathirdresearcherandtakethemajorityvote.Inthisway,weobtainabout800labelednewsparagraphs,whichwillbeusedbymachinelearningmodelstounderstandthecontextandmakepredictions.,3.4ApplyingMachineLearningAlgorithmsWeusethe800labelednewsparagraphstoteachthemachinelearningmodeltolabeltheremainingsampleforus.These800paragraphsaredividedintotrainingset(500),validationset(150),andtestset(150)threesplitsthatarecommonlyusedindifferentstagesofthecreationofamachinelearningmodel.Thetestdatasetusedtoprovideanunbiasedevaluationofafinalmodelfitonthetrainingdataset.Asitisneverusedintrainingprocess,itisalsocalledaholdoutdataset.Inthecaseoftextclassification,parametersrefertoconfigurationvariablesthatcapturefeaturesofthetextsuchasthelengthofaparagraph,therelationshipbetweenthefirstandlastword.Theyaretailoredtoaspecifictaskandadjustedasthemodellearns.Themodeltrainingprocessisachievedbysupervisedlearningthroughoptimizationmethodssuchasstochasticgradientdescent.Themodelthenproducesaresult,foreachparagraph—favorableorunfavorable—andcomparestheresultwithourlabel,whichistheanswerkey.Basedontheresultofthecomparison,themodelparametersareadjustedandoptimizeduntiltheaccuracyratestabilizes.3Ifthecontentofthetitleandthecontentofthetextpointtodifferentdirectionsofchangeinclimatetransitionrisk,thedirectionofthetitledominates.IMFWORKINGPAPERSDecomposingClimateRisksinStockMarketsINTERNATIONALMONETARYFUND13Figure2.SentimentClassificationStrategiesCompared4Morespecifically,wefollowedthemorerecenttransferlearningstrategyforourtaskasshowninFigure2.Weadoptedatransformerbaseddeeplearningmodelarchitecture(Vaswanietal.2017)whichdemonstratedstate-of-the-artperformancefortextclassificationtasks.WeinitiateourmodelweightsfromSIEBERT(“SentimentinEnglish”).Itisanopen-sourcemodelcheckpointinitializedfromRoBERTa-large,thenfine-tunedandevaluatedon15textclassificationdatasets.Itisapopularpretrainedmodelforgeneralpurposetextclassificationtasks(Hartmannetal.2022).Wefurtherfinetunedthemodelonourcustomizedtrainingdata.Weconductedasmallscalehyperparametertuningonourtrainingandvalidationsplit.Asourtrainingdataisrelativelysmall,wedidnotgoforthesetofhyperparametersthatmaximizetheaccuracyinvalidationsplit.Instead,wechoseasetofhyperparametersthatyieldsimilaraccuracyinbothtrainingandvalidationsplit.Finally,weevaluatedourmodelinthetestsplitandachieved70%accuracy.Thisisadecentperformancegiventhattheaverageaccuracyrateyieldedbyhumanratersisalittleover80%.Asthefinalstep,weapplythemodeltotherestofthenewsdatathroughmodelinference,andobtainthelabelsofallclimatepolicy-relatedparagraphs.Inthisway,wegetthenumberofclimatefavorablenewsparagraphsandthenumberofunfavorablenewsparagraphs,which,afterbeingscaledbythetotalnumberofnewsparagraphs,becomeourclimatenewsfactors.4FiguremodifiedbasedonHartmannetal.2022(Fig.2).LexiconsTraditionalMachineLearningTransferLearningReseasrchquestionOff-the-shelfdictionaryCustomizeddictionaryFinalmodel(baedon+/-wordfrequency)ReseasrchquestionFeatureengineering(preprocess/veterizationetc)ClassificationmodeltrainingLabeledtext(task-specific)FinalmodelReseasrchquestionText(generalbooks/wikietc)Self-supervisedpre-trainingLanguagemodel(general)Pre-trainedclassificationmodel(e.gSIEBERT)SupervisedfinetuningonsimilarclassificaiontasksModeltraining(initatedwithpre-trainedclassificationmodel)Labeledtext(task-specific)FinalmodelIMFWORKINGPAPERSDecomposingClimateRisksinStockMarketsINTERNATIONALMONETARYFUND143.5CollectingStockPriceDataThesubjectsofstudyareallCanadianoilandgascompaniesintheStandardandPoor’s(S&P)TSXindex,aswellasrepresentativeU.S.andEUcompanies.StockpricedataisgatheredfromtheBloombergTerminal.Wecollectcompanyname,indexmembership,industryclassification,monthlystockprice,absoluteCO2emissions,andtotalGHGemissionsdataforalltheconstituentsofthefollowingstockindices:S&PTSX,S&PEurope,S&PSIOP,S&PSIOS,S&PUtilities,S&PSteel,S&PMetals,andCement,whicharebenchmarkCanadianindex,Europeanindex,oilandgasexplorationandproductionselectindustryindex,oilandgasequipmentselectindustryindex,utilities,steel,metals,andcementindustries’indices,respectively.VariablesonCO2andGHGemissionsaredroppedduetothelargenumberofmissingvalues.Thefinaldatasetspansover19yearsfromJanuary2004toMarch2022,covering1751companies.OurfactorconstructionmethodcloselyfollowsthatofFamaandFrench(1992,1993).Informationonmarketpremium,sizepremium,andvaluepremium,whicharetraditionallyconsideredtobeabletocapturecross-sectionalreturnvariations,alongwiththerisk-freerate,canbefoundintheofficialwebsiteofUniversityofChicagoprofessorKennethFrench.Sinceweareinterestedincross-marketcomparisons,wegather3assetpricingfactorsfortheNorthAmericanmarketandtheEuropeanmarket.Wealsoretrieve2additionalfactors,cashearningtopriceratioanddividendyieldratio,alongsidethemomentumfactorforuseinrobustnesschecks.3.6CorrelatingClimateFactorwithStockPriceRiskpremiumiscalculatedusingasetofassetpricingmodels.Inthebaselinemodel,weestimatethefollowingequation𝑟𝑟𝑖𝑖,𝑡𝑡−𝑟𝑟𝑓𝑓,𝑡𝑡=𝛼𝛼𝑖𝑖+𝛽𝛽𝑖𝑖𝐶𝐶𝑡𝑡+𝛾𝛾𝑖𝑖𝑋𝑋𝑡𝑡+𝜀𝜀𝑖𝑖,𝑡𝑡(1)where𝑟𝑟𝑖𝑖,𝑡𝑡isthemonthlyreturnonstock𝑖𝑖,𝑟𝑟𝑓𝑓,𝑡𝑡isthemonthlyrisk−freereturn,𝐶𝐶𝑡𝑡isthenews−basedclimatenewsfactor,𝑋𝑋𝑡𝑡isavectoroffactorsthathavebeenfoundtoexplaintheperformanceofstockreturns,and𝜀𝜀𝑖𝑖,𝑡𝑡istheerrorterm.TheoverlappingmonthsbetweenthenewsdataandthestockdataarefromJanuary2005toMarch2022,leadingtoatimeseriesof206monthlyreturns.Thesignificanceandmagnitudeof𝛽𝛽𝑖𝑖areofinteresttoourstudy.Ifweobserveastatisticallysignificant𝛽𝛽𝑖𝑖,itmeansthatournovelclimateriskfactorcanexplainthevariationsinstockreturns.Wealsoinvestigatewhethertheclimatenewsfactorisrelevantforinvestorsbycreatinglong-shortportfolios.Asopposedtoregression-basedassetpricingtests,portfoliosortingrepresentsanon-parametricapproachtotestingfortherelevanceofassetpricingfactors.Webeginbysortingstocksbasedonthesensitivityofeachstock’sreturnstothemodel-generatedclimatenewsfactor.Wethencreatealong-shortportfoliothatmimicsthetradingbehaviorofbuyingstockswiththelargestclimatesensitivities,orbetas,andsellingthestockswiththelowestclimatesensitivities.Theabnormalreturnofthemimicportfolio,ifany,isestimatedusingequation(1).Thesignificanceandmagnitudeof𝛼𝛼𝑖𝑖areofIMFWORKINGPAPERSDecomposingClimateRisksinStockMarketsINTERNATIONALMONETARYFUND15interesttoourstudy.Ifthelong-shortportfolioyieldsstatisticallysignificantreturnswhenotherriskfactorsarecontrolled,itindicatesthattheclimatenewsfactorispricedbytheinvestors.Thehelicopterviewofourmethodologyispresentedintheflowchartbelow.Figure3.ResearchMethodologyFlowchartGivenourfocusontheoilandgassector,inanalternativeapproach,wesortallcompaniesbasedontheirindustryclassificationontheassumptionthatoilandgascompaniesaremoresensitivetoclimatetransitionrisks.Thecompaniesinthesamplearethusdividedintotwogroups—thegroupofoilandgascompaniesandtherest.Ourlong-shortportfolioisthenconstructedbylongingthestocksofoilandgascompaniesandshortinginthestocksofnon-oilandgascompanies.Ifweobservesignificantabnormalreturnsontheclimatenewsfactoraftercontrollingforotherriskfactors,thissuggeststhatclimateriskisrelevantforinvestorsinthestockmarket.Sources:FinancialTimesandIMFstaffcalculationsSources:BloombergandIMFstaffcalculationsSources:KennethR.FrenchDataLibraryandIMFstaffcalculationsFinancialTimesNewsdatacollectionandpreprocessClimatechangepolicyidentifactioninnewsdataIdentifyclimatetransitionrisksbymanuallableing800samplesTrainadeeplearningmodeltopredictclimatetransitionrisksfromnewsStockpricedatacollectionConstructlong-shotportfoliobasedonindustryandmonthlyreturnsMergewithcross-sectionalriskfactorsforCanadian,USandEUstocksPlugclimateriskfactorintoaccetpricingmodelandestimatealphasClimatetransitionrisksindicatorApplytrainedmodelonnewsdataandaggregateIMFWORKINGPAPERSDecomposingClimateRisksinStockMarketsINTERNATIONALMONETARYFUND164.ResultsandDiscussion4.1ClimateNewsFactorInFigure3,weplotthetimeseriesofclimatepolicy-relatednewscoverageoverthesamplehorizon,annotatedwithclimate-relevantevents.Figure4.IntensityofClimateNewsCoverageSources:FinancialTimesandauthorcalculationsTheindexregularlyshowsincreasesaroundsalientclimateevents,suchasworldclimatesummits,theannouncementofinternationallybindingclimateaccords,andthereleaseofimportantclimatereports.Wefurtherdisaggregatethenewsbasedonitsclimateimpactandgeneratetwosub-indices—climatefavorablenewsandclimateunfavorablenews.Thetimeseriesofthetwosub-indicesareshowninFigure4.IMFWORKINGPAPERSDecomposingClimateRisksinStockMarketsINTERNATIONALMONETARYFUND17Figure5.IntensityofClimateFavorableandUnfavorableNewsSources:FinancialTimesandauthorcalculationsTheindicesaretheratiosofclimatefavorable/unfavorablenewsoverthetotalnumberofnewsparagraphs.Asexpected,internationalsummitswherethepoliticalleadersofmanycountriesmeettodiscussissuesrelatedtoclimatechangeinanattempttoreduceglobalcarbonemissions,areassociatedwithtightermitigationpolicies,andareclimatefavorablenewsbyourdefinition.5Itisworthemphasizingthatthedefinitionof“news”inourstudyisnotlimitedtothearrivalofnewinformation.Rather,itisextendedtoincludetheincrementalchangeininvestors’perceptionofclimaterisks,whichcouldalsomovethemarket.Theindicatortracksreal-timeclimatenews,andthusreflectscurrentattitudestowardsclimateevents.Itmayevenreflectattitudesthatexistpriortotheeventsthemselves,giventhatnewscoverageoftenprecedesthem.However,ithaslimitationswhenitcomestocapturingtheretrospectivesignificanceofanevent.Forexample,whiletheParisConferenceresultedinalandmarkagreementwithlastingimplicationsforfutureclimateactions,itsimportancemaynotnecessarilyhavebeenappreciatedincontemporarynewscoverage,whereasthesignificanceofthePoznanCOPconferencemayprovetobelessconsequentialwhenviewedinretrospect.Wefurthervalidatetheindexbyconductingmanualchecksonarandomlyselectedsample,andthemovementofourindicatorsurroundingthiseventisalsosimilartothatofotherindicesintheliterature.(e.g.Engleetal.,2020;Faccinietal.,2021).Additionalteststovalidatetheindicatorcanbefoundintherobustnesschecksection.5Notethatthespikeismuchmoremoderateforcontroversialevents,asourindexisconstructedasthepercentageofnegativenewsratherthantherawcounts.Themovementofourindexsurroundingthiseventissimilartothatofotherindicesintheliterature.(e.g.Engleetal.,2020;Faccinietal.,2021)IMFWORKINGPAPERSDecomposingClimateRisksinStockMarketsINTERNATIONALMONETARYFUND184.2TheEffectofClimateRiskonOilandGasCompaniesWethenaddtheclimatenewsfactortotraditionalFama-Frenchthreefactormodel,whichincludesamarketfactor,asizefactor,andavaluefactor,andexaminewhethertheclimateriskelicitedbythenewsfactorisusefulforexplainingthereturnsofCanadianoilandgassectorstocks.Wehypothesizethatnewsontightermitigationpolicies,whichisdefinedasclimate-favorablenewsinthisstudy,representsabadsignaltotheoilandgassector.Inresponsetosuchanegativeshock,investorswouldshortsellstocksintheoilandgassector,thusreducingtheirprices.Asaresult,weshouldobserveanegativecoefficientontheriskfactor.Conversely,newsonlightermitigationpolicies,whichislabeledasclimateunfavorablehere,constitutesarelativelygoodsignaltotheoilandgassector.Investorswouldreactbybuyingstocksintheoilandgassector,thusboostingtheirprices.Asaresult,weshouldfindapositivecoefficientonthatriskfactor.Theresults,reportedinTable1,areconsistentwithourhypothesis.However,onlythecoefficientontheclimateunfavorablefactorisstatisticallysignificant(Model2).Intermsofeconomicmagnitude,onepercentageincreaseinthefactorvalueisassociatedwith14percentageincreaseinstockreturns.Theimpactoftheclimateriskfactorismuchlargerthanmarket,size,andvaluefactors.Table1.PricingofClimateNewsFactorinCanadianStockMarket(1)(2)Climate-favorable-2.945(4.490)Climate-unfavorable14.259(2.174)Mkt-RF1.9742.459(0.247)(0.267)SMB1.8274.395(0.340)(0.492)HML2.0044.168(0.187)(0.430)Constant5.289-1.973(7.476)(3.096)Observations13,86113,861R-squared0.4680.468Sources:Bloomberg,Frenchdatalibrary,andauthorcalculationsInvestorswithlongpositionsinoilandgasassets,onemayargue,mayperceiveanycurrentrelaxationofclimatemitigationeffortsasapostponementofstrictermeasuresinthefuture.Consequently,itremainsuncertainwhethertheexpectedreturnsonfossilfuelassetsshouldriseandincentivizefurtherinvestment,IMFWORKINGPAPERSDecomposingClimateRisksinStockMarketsINTERNATIONALMONETARYFUND19giventhatexpectedreturnsaretypicallyassessedbasedonthepresentvalueofdiscountedfuturereturns.Admittedly,investorscouldberationalandabletopriceinallpolicychangesinfutureperiods.However,ithasalsobeenwidelydocumentedintheliteraturethatinvestorsdemonstrateconsiderableshort-termism,pursuinguninformativeshort-termspeculationandneglectinglong-runfundamentals(seee.g.Bushee,1998;Frootetal.,1992;Stein,1989).Ourresultslendsupporttothelattertheory.4.3AsymmetricalResponsestoClimateFavorableandUnfavorableNewsTheresultsinTable1pointtoasymmetricalresponsesofCanadianoilandgassectorstockstoclimatefavorableandunfavorablenews.Ourfindingsfromtheportfoliosortingapproachlendfurtherevidence.Sinceanincreaseintheclimatefavorablenewsfactorsignalsstrictermitigationpolicies,andthusshouldbebadnewsforoilandgascompanies,inresponsetosuchanegativeshock,investorswouldselloilandgasstocks,thusdecreasingtheirpricesandincreasingtheirreturns.Asaresult,thelong-shortportfoliowouldyieldapositivealpha.Inasimilarvein,anincreaseinclimateunfavorablenewsshouldbeassociatedwithanegativealpha.TheresultsinTable2confirmourhypothesis,indicatingthatfirmshaveincentivestomanagetheirclimateriskexposure.Table2.PortfolioSortingAnalysisofCanadianStockMarketPanelA:Fama-FrenchThree-FactorModelClimate-favorableClimate-unfavorableTotal0.105-0.769-0.875(0.157)(0.107)(0.177)PanelB:MarketModelClimate-favorableClimate-unfavorableTotal0.045-0.745-0.882(0.157)(0.109)(0.179)Sources:Bloomberg,Frenchdatalibrary,andauthorcalculationsWhilethereisasignificantlynegativeriskpremiumassociatedwithclimateunfavorablenewsamongthegroupofoilandgascompanies,wefindthattheriskpremiumtiedtoclimatefavorablenewsispositivebutstatisticallyinsignificant.Inotherwords,aneaseinclimatemitigationconstitutesapositiveshocktooilandgascompaniesbuttightermitigationpoliciesdonotnecessarilytranslateintobadnews.4.3Cross-CountryEvidenceWhilemarket-wideclimateriskisexpectedtobeaffectcompaniesacrosscountries,theextentofpricingcouldvary,dependingonthelevelofexposuretoclimatechange,degreeofmarketefficiency,etc.CanadianfirmscouldbemoresensitivetoclimateriskfactorsthanUSandEUfirms,duetoitshigherrelianceontheoilandgassector.Weconductacross-countrycomparisonbyperformingthesamesetofexercisesonUSandEUcompanies.IMFWORKINGPAPERSDecomposingClimateRisksinStockMarketsINTERNATIONALMONETARYFUND20Table3.PricingofClimateNewsFactorintheUSStockMarketPanelA:Fama-FrenchThree-FactorModelClimate-favorableClimate-unfavorableTotal0.087-0.692-0.783(0.153)(0.107)(0.176)PanelB:MarketModelClimate-favorableClimate-unfavorableTotal0.149-0.713-0.774(0.151)(0.105)(0.173)Sources:Bloomberg,Frenchdatalibrary,andauthorcalculationsTable3displaystheaveragereturnsoflong-shortportfoliosmadeupofcompaniestradedintheUSstockmarket.InthebaselineFama-FrenchThree-Factormodel,wefindthattheaveragemonthlyabnormalreturnofUScompanies,whichstandsat0.69percentagepoints,isslightlysmallerinmagnitudecomparedwiththatofCanadiancompanies,whichisestimatedtobe0.77percentagepoints.OilandgassectorcompaniesinCanadademonstrateaslightlyhighersensitivitytoclimate-relatedtransitionriskthanthosebasedintheUnitedStates.Table4.PricingofClimateNewsFactorintheEUStockMarketPanelA:Fama-FrenchThree-FactorModelClimate-favorableClimate-unfavorableTotal0.131-0.715-0.770(0.138)(0.096)(0.160)PanelB:MarketModelClimate-favorableClimate-unfavorableTotal0.120-0.741-0.805(0.139)(0.097)(0.161)Sources:Bloomberg,Frenchdatalibrary,andauthorcalculationsWhenweturntotheEUstocks,wefindthedirectionofchangeisthesameasthatoftheCanadianandUSmarket,thoughofvaryingmagnitude.EUoilandgascompaniesreactsignificantlytoclimateunfavorablenews,ornewsonlighterclimatemitigation,whereastheirreactiontoclimatefavorablenews,ornewsonstricterclimatemitigation,israthermuted.Thefactthatcompaniestreatclimatetransitionrisksimilarlyregardlessoftheirlistinglocation,couldperhapsreflecttheglobalnatureofthesemultinationalsandtheirinvestorbase.IMFWORKINGPAPERSDecomposingClimateRisksinStockMarketsINTERNATIONALMONETARYFUND215.RobustnessChecks5.1UsingAlternativeAssetPricingModelsWehaveshownfromTable2toTable4thatourresultsarerobusttovariousassetpricingmodelspecifications,includingthemarketmodelandtheFama-Frenchthreefactormodel.Inadditionalassetpricingtests,weintroducetheFama-Frenchfivefactor(FamaandFrench,2014)andtheCarhartfourfactor(Carhart,1997).Table5.PricingofClimateNewsFactorUsingAlternativeAssetPricingModelsPanelA:Fama-FrenchFive-FactorModelClimate-favorableClimate-unfavorableTotal0.113-0.707-0.781(0.210)(0.152)(0.248)PanelB:CarhartFourFactorModelClimate-favorableClimate-unfavorableTotal0.124-0.723-0.806(0.210)(0.149)(0.243)Sources:Bloomberg,Frenchdatalibrary,andauthorcalculationsTheFama-Frenchfivefactormodelisdirectedatcapturingthesize,value,profitability,andinvestmentpatternsinaveragestockreturns.Withtheadditionofprofitabilityandinvestmentfactors,theresultsremainstable(Table5,PanelA).TheCarhartfourfactormodel,proposedbyMarkCarhartin1997,addsanextramomentumfactortotheFama-Frenchthreefactormodel.Momentumreferstothespeedorvelocityofpricechangesinastock,which,reflectedinthemodel,isthepremiumonwinnersminuslosers.Ourconclusionsremainunchangedafterintroducingthemomentumfactor,asshowninPanelBofTable5.5.2UsingAlternativeClimateRiskMeasuresOurresultsarealsorobusttodifferentdefinitionsofclimaterisk.InFaccinietal.(2021),theauthorsbelievethatanincreaseininvestorattentioncanbereflectedeitherbyanincreaseinthenumberofnewsarticlespublishedonclimatechange,and/oranincreaseintheproportionofarticlesdevotedtothetopicofclimatechange,withinagivennumberofarticlespublished.Therefore,thereisnoneedfortheclimatenewsfactortobestandardizedbydividingthenumberofclimatepolicy-relatedparagraphsbythetotalnumberofmonthlypublishednewsparagraphs.Followingthisdefinitionyieldssimilarresults(Table6).Anincreaseintheintensityofclimateunfavorablenewscoverageisassociatedwithsignificantlynegativeabnormalreturns,i.e.,apositiveshock.However,itisworthnotingthatanincreaseinthecoverageofclimatefavorablenewsalsoleadstosignificantabnormalreturns,whichhighlightstheimportanceofclimatepolicycommunicationtothepublicviamediaoutlets,asmoreinformationonclimatemitigation,whethertighterorlooserpolicymeasures,hasbeenpositivelyreceivedbythemarket.IMFWORKINGPAPERSDecomposingClimateRisksinStockMarketsINTERNATIONALMONETARYFUND22Table6.PricingofClimateNewsFactorUsingAlternativeClimateRiskMeasuresPanelA:Fama-FrenchThree-FactorModelClimate-favorableClimate-unfavorableTotal-0.352-0.390-0.384(0.054)(0.051)(0.053)PanelB:MarketModelClimate-favorableClimate-unfavorableTotal-0.331-0.357-0.355(0.057)(0.055)(0.057)Sources:Bloomberg,Frenchdatalibrary,andauthorcalculations5.3AdoptingDifferentPortfolioSortingStrategiesInanalternativeapproachtotestingthevalidityoftheclimatenewsfactor,wesortstocksindescendingorderindecileportfolios,basedonthemagnitudeoftheirestimatedclimatebetaswithrespecttothefactor.Arollingwindowofdailyobservationsoverthepastthreeyearsisusedtoestimateclimatebetas,andthewindowisrolledforwardbyone-monthateachestimationstep.Thepost-rankingvalue-weightedportfoliomonthlyreturnoverthenextmonthisthencomputed,andtheresultingspreadportfolioreturniscalculatedasthedifferencebetweenthereturnofhighest-rankingportfoliowithhighestclimatebetasminusthereturnofthelowest-rankingportfoliowithlowestclimatebetas.Table7.PricingofClimateNewsFactorUsingDifferentSortingStrategiesPanelA:Fama-FrenchThree-FactorModelClimate-favorableClimate-unfavorableTotal0.134-0.752-0.858(0.052)(0.273)(0.507)PanelB:MarketModelClimate-favorableClimate-unfavorableTotal0.133-0.550-1.324(0.046)(0.307)(0.435)Sources:Bloomberg,Frenchdatalibrary,andauthorcalculationsThebaselinemodelincludesFama-Frenchthreefactors(Table7,PanelA).Inaseriesofadditionaltests,weincludeonlythemarketfactor(PanelB)anddividestocksintoquintiles/quartiles/tercilesportfolios,theresultsstillhold.5.4FocusingonCanadianSpecificNewsWhileclimatechangeiswidelyrecognizedasamarket-widerisk(WorldEconomicForum,2022)andCanadahasbeenaleadingadvocateforglobalclimateaction(Bloomberg,2022;GovernmentofCanada,2022),itwouldstillbeworthwhiletoexplorethepriceeffectofclimate-relatednewsthathasaspecialreferencetoCanada.Outofthewholesampleofclimate-relatedtweets,wepickedthoseonCanadausinganarrayofcountryidentifiers.(Schrodt,2015)TheresultsofthisexercisearepresentedinTable8.WhilethecoefficientsareIMFWORKINGPAPERSDecomposingClimateRisksinStockMarketsINTERNATIONALMONETARYFUND23consistentwithourhypothesis,themagnitudesaresignificantlylarger,indicatingstrongereffectofCanada-specificnewsonthedomesticmarket.Table8.PricingofClimateNewsFactor(Canada-specific)inCanadianStockMarket(1)(2)Climate-favorable-5.958(9.083)Climate-unfavorable42.281(6.446)Mkt-RF-1.5143.681(0.556)(0.381)SMB6.034-0.502(0.695)(0.515)HML1.5101.184(0.177)(0.187)Constant5.564-32.278(7.883)(6.039)Observations13,86113,861R-squared0.4680.468Sources:Bloomberg,Frenchdatalibrary,andauthorcalculations5.5DroppingESGTagESG,shortforEnvironmental,Social,andGovernance,hasbecomeacrucialforceininvestingduetoclimatechange.(OECD,2020)DespitethefactthatincorporatingrisksassociatedwithclimatechangeandstrandedassetsstemmingfromtheclimatetransitionisincreasinglyrecognizedasacentralelementtotheESGenvironmentalpillar,severalinherentflaws—lackofstandardization,subjectivity,dataqualityconcerns,conflictingpriorities,greenwashingandimpactwashing—haverenderedtheindexaless-than-optimalframeworkforevaluatingclimatechangeimpacts.(Elmaltetal.,2021;HarvardBusinessReview,2022)Therefore,inthissubsection,wereruntheclimatekeywordsearchwithoutincludingtheESGtag,andthefindingsdonotqualitativelydifferfromourmainresults.IMFWORKINGPAPERSDecomposingClimateRisksinStockMarketsINTERNATIONALMONETARYFUND24Table9.PricingofClimateNewsFactor(WithoutESG)inCanadianStockMarket(1)(2)Climate-favorable-2.994(4.564)Climate-unfavorable10.035(1.530)Mkt-RF2.0392.543(0.248)(0.272)SMB1.7403.617(0.341)(0.414)HML1.9713.437(0.185)(0.332)Constant5.435-1.329(7.691)(0.212)Observations13,86113,861R-squared0.4680.468Sources:Bloomberg,Frenchdatalibrary,andauthorcalculations5.6Including“Climate-Benign”LabelThenewspassagesarecategorizedintotwogroups:"climate-favorable"and"climate-unfavorable".Nevertheless,thisbinaryclassificationmaybetoolimited,anditcouldbebeneficialtoincorporateathirdcategory,labeledas"climate-benign",toencompassinstanceswherethenewsdoesnothaveapositivenornegativeconnotationregardingclimatepolicy.Includingthisadditionalclassificationcouldenhancetheaccuracyandinformativenessoftheothertwocategoriesandpotentiallyleadtomorerobustresults.Toobtainthe“climate-benign”category,weremapthesentimentmetrictoclimatelabels.Aratioinsentimentanalysisisascorethatgaugestheconfidenceintervalofnegativeandpositivesentimentsconveyedincomments.Typically,thisisdepictedonascalerangingfrom0to1,withthelowendofthescaleindicatingpredominantlynegativeresponsesandthehighendofthescalesignifyingmainlypositiveresponses.Inpreviousexercises,weadoptabinarylabel,whichmeansallsentimentscoreshigherthanorequalto0.5arecategorizedaspositive,andthoselowerthan0.5areclassifiedasnegative.Inthenewmapping,weintroduceathirdcategory,labelingscoreshigherthan0.8aspositive,lowerthan0.2asnegative,andtherestneutral.Theneutrallabelsarealsocalled“climate-benign”labels,suggestingthatthenewscontentisneitherfavorablenorunfavorabletoclimate.Takingintoconsiderationthenew“climate-benign”labeldoesnotfundamentallychangeourresults,asshowninTable10.IMFWORKINGPAPERSDecomposingClimateRisksinStockMarketsINTERNATIONALMONETARYFUND25Table10.PricingofAlternativeClimateNewsFactorinCanadianStockMarket(1)(2)Climate-favorable-2.892(4.409)Climate-unfavorable11.028(1.681)Mkt-RF2.0572.391(0.249)(0.263)SMB1.9463.747(0.338)(0.426)HML1.9923.343(0.186)(0.320)Constant4.683-1.235(6.585)(0.197)Observations13,86113,861R-squared0.4680.468Sources:Bloomberg,Frenchdatalibrary,andauthorcalculationsIMFWORKINGPAPERSDecomposingClimateRisksinStockMarketsINTERNATIONALMONETARYFUND266.ConclusionWhileclimatechangehaslaggedbehindsocialissuesduringtheCovid-19crisis,theCOP26andthegreenracehaverecentlychangedtheequation,puttingclimateriskbackontheagendaofinvestorsandevengoesbeyondESGinvesting.Usingstate-of-the-artmachinelearningtechniques,weproposeanews-basedindextomeasuretheintensityofclimatemitigationpolicies.Theindexprovestobeeffectiveintrackingmajorclimatepolicyeventswitharelativelyhighfrequency.Weconverttheindexintoaclimateriskfactorandapplyittoasetofassetpricesmodels,inordertoestimatemarketresponsetodifferenttypesofmitigationpolicies.TheresultsconsistentlyshowthatriskfromclimatemitigationhasbeenreflectedinthestocksofCanadianoilandgascompanies.Aneaseinmitigationpolicies,whichshouldgenerallybeviewedasgoodnewsfortheoilandgassector,isassociatedwithanupwardmovementintheirstockprices.However,strongermitigationpoliciesdonotnecessarilyrepresentanegativeshock.Theoilandgassectorexperiencesadownwardcorrectioninstockvaluations,butthepricefallisnotstatisticallysignificant.WeextendtheanalysistoUSandEUstockmarketsandfindasimilarasymmetricresponsetoclimate-favorableand-unfavorablenews.Thesensitivities,however,aresmallerthanintheCanadianmarket.Ourfindingsindicatethatclimatemitigationpoliciesarerelevantforinvestorsandcurrentpoliciesproviderightincentives,asreflectedinthemovementofstockpricesintheexpecteddirection.However,mitigationpoliciescapturedbymediacoveragehavelimited,asymmetricimpactonassetvaluations.Theimpactofclimatechangeisfar-reaching.(UTZ,2022)Beyondtheoilandgassector,howtherestoftheeconomyrespondstoclimatechangepoliciescouldbeimportantavenuesforfutureresearch.IMFWORKINGPAPERSDecomposingClimateRisksinStockMarketsINTERNATIONALMONETARYFUND27ReferencesAhn,Natalie,2017.ComparingNLPMethodsforIdentifyingPolicyDecisionsinGovernmentDocuments,UniversityofCalifornia,BerkeleyUnpublishedPaper.Ashish,Vaswani.,Noam,Shazeer.,Niki,Parmar.,Jakob,Uszkoreit.,Llion,Jones.,Aidan,N.,Gomez.,Lukasz,Kaiser.,Illia,Polosukhin.(2017).AttentionIsAllYouNeed.arXiv:ComputationandLanguage,Bansal,Ravi,DanaKiku,andMarceloOchoa,2016.PriceofLong-runTemperatureShiftsinCapitalMarkets,NBERWorkingPaperSeries22529.Barkema,Jelle,BorislavaMircheva,MicoMrkaic,YuanchenYang,2021.LicensetoSpill:HowDoWeDiscussSpilloversinAIVStaffReports,IMFWorkingPaperNo.2021/134.Bhattacharya,U.,N.Galpin,R.Ray,andX.Yu.2009.TheroleofthemediaintheinternetIpobubble.JournalofFinancialandQuantitativeAnalysis44(03):657–682.Blackrock,2016.AdaptingPortfoliostoClimateChangeImplicationsandStrategiesforAllInvestors,BlackrockInvestmentInstitute.Bloomberg,October19,2022.JustinTrudeauDefendsCanada'sMinusculeClimateProgress.Bushee,B.1998.TheinfluenceofinstitutionalinvestorsonmyopicR&Dinvestmentbehavior.TheAccountingReview73(3):305–333.Cahan,R.,J.Jussa,andY.Luo.2009.Breakingnews:Howtousenewssentimenttopickstocks.MacquarieUSEquityResearch.Carhart,Mark,1997.OnPersistenceinMutualFundPerformance.TheJournalofFinance,52(1):57–82.Chava,Sudheer,2014.EnvironmentalExternalitiesandCostofCapital,ManagementScience,60:2223-2047.CIA,2022.TheWorldFactbook,availableathttps://www.cia.gov/the-world-factbook/deMarneffe,Marie-Catherine,BillMacCartney,andChristopherD.Manning,2006.GeneratingTypedDependencyParsesfromPhraseStructureParses,ProceedingsoftheInternationalConferenceonLanguageResourcesandEvaluation(LREC).Dell,Melissa,BenjaminJones,andBenjaminOlken,2012.TemperatureShocksandEconomicGrowth:EvidencefromtheLastHalfCentury,AmericanEconomicJournal:Macroeconomics,4(3):66-95.Dougal,C.,J.Engelberg,D.Garcia,andC.A.Parsons.2012.Journalistsandthestockmarket.ReviewofFinancialStudies25(3):639-679.ElGhoul,Sadok,OmraneGuedhami,HakkonKim,andKwangwooPark,2018.CorporateEnvironmentalResponsibilityandtheCostofCapital:InternationalEvidence,JournalofBusinessEthics,149:335-361.Elmalt,Dalya,DenizIgan,DivyaKirti.2021.LimitstoPrivateClimateMitigation.IMFWorkingPaperNo.2021/112.Engle,Robert,StefanoGiglio,BryanKelly,HeebumLee,JohannesStroebel,2020.HedgingClimateChangeNews,TheReviewofFinancialStudies,33(3):1184–1216.EnvironmentandClimateChangeCanada,2022.CanadianEnvironmentalSustainabilityIndicators:GlobalGreenhouseGasEmissions.IMFWORKINGPAPERSDecomposingClimateRisksinStockMarketsINTERNATIONALMONETARYFUND28Faccini,Renato,RastinMatin,GeorgeSkiadopoulos,2021,DissectingClimateRisks:AreTheyReflectedinStockPrices.Fama,EugeneF,andKennethRFrench,1992.TheCross-SectionofExpectedStockReturns.TheJournalofFinance,47(2):427–65.Fama,EugeneF,andKennethRFrench.1993.CommonRiskFactorsintheReturnsonStocksandBonds.JournalofFinancialEconomics33(1):3–56.Fama,EugeneF.andKennethRFrench,2014.AFive-FactorAssetPricingModel.Fama-MillerWorkingPaper.Fellbaum,Christiane,2005.WordNetandWordnets.InAlexBarber(ed.),EncyclopediaofLanguageandLinguistics.Elsevier,2-665.Froot,K.,D.Scharfstein,andJ.Stein.1992.HerdontheStreet:Informationalinefficienciesinamarketwithshort-termspeculation.JournalofFinance47(4):1461–1484.Ginglinger,Edith,andQuentinMoreau,2019.ClimateRiskandCapitalStructure,UniversiteParis-DauphineResearchPaper3327185,EuropeanCorporateGovernanceInstitute-FinanceWorkingPaper737/2021,http://dx.doi.org/10.2139/ssrn.3327185.Görlach,Benjamin.2013.WhatConstitutesAnOptimalClimatePolicyMix?DefiningtheConceptofOptimality,IncludingPoliticalandLegalFrameworkConditions.CECILIA2050WP1Deliverable1.1.Berlin:EcologicInstitute.GovernmentofCanada,2020.CrudeOilIndustryOverview,availableatCrudeOilIndustryOverview(nrcan.gc.ca).GovernmentofCanada,2022.MinisterGuilbeaultleadstheCanadiandelegationtotheUNClimateChangeConference(COP27)topushforglobalclimateaction.Gupta,S.,D.A.Tirpak,N.Burger,J.Gupta,N.Höhne,A.I.Boncheva,G.M.Kanoan,C.Kolstad,J.A.Kruger,A.Michaelowa,S.Murase,J.Pershing,T.Saijo,A.Sari,2007.Policies,InstrumentsandCo-operativeArrangements.InClimateChange2007:Mitigation.ContributionofWorkingGroupIIItotheFourthAssessmentReportoftheIntergovernmentalPanelonClimateChange[B.Metz,O.R.Davidson,P.R.Bosch,R.Dave,L.A.Meyer(eds)],CambridgeUniversityPress,Cambridge,UnitedKingdomandNewYork,NY,USA.HarvardBusinessReview,2022.AnInconvenientTruthAboutESGInvesting.Hong,Harrison,FrankWeikaiLi,andJiangminXu,2019.ClimateRisksandMarketEfficiency,JournalofEconometrics,208:265-281.Huang,Chengyu,SeanSimpson,DariaUlybina,AgustinRoitman,2019.News-BasedSentimentIndicators,IMFWorkingPaperNo.2019/273.Karp,Larry,andArmonRezai,2018.AssetPricesandClimatePolicy,2018MeetingPapers595,SocietyforEconomicDynamics.Klein,DanandChristopherD.Manning,2003.AccurateUnlexicalizedParsing,Proceedingsofthe41stMeetingoftheAssociationforComputationalLinguistics(ACL).Kumar,AlokandWeiXinandChendiZhang,2019.ClimateSensitivityandPredictableReturns,unpublishedpaper.IMFWORKINGPAPERSDecomposingClimateRisksinStockMarketsINTERNATIONALMONETARYFUND29Jochen,Hartmann.,Mark,Heitmann.,Christian,Siebert.,Christina,Schamp.(2022).MorethanaFeeling:AccuracyandApplicationofSentimentAnalysis.InternationalJournalofResearchinMarketing,availableathttps://doi.org/10.1016/j.ijresmar.2022.05.005McKinseyGlobalInstitute,2020.ClimateRiskandResponse:PhysicalHazardsandSocioeconomicImpacts,availableatwww.mckinsey.com/mwg-internal/de5fs23hu73ds/progress?id=aoCkcm-eu6iy2EkYtKoPdpaOKwcmtryTgHZFzqUNzJE,&dl.Michaelowa,Axel,MylesAllen,andFuSha,2018.PolicyInstrumentsforLimitingGlobalTemperatureRiseto1.5°C–CanHumanityRisetotheChallenge?,ClimatePolicy18(3):275-286,DOI:10.1080/14693062.2018.1426977Millischer,Laurent,TatianaEvdokimova,andOscarFernandez,2022.TheCarrotandtheStock:InSearchofStock-MarketIncentivesforDecarbonization.IMFWorkingPaperWP/22/231.MunichRe,2020.RisksPosedbyNaturalDisasters,availableatwww.munichre.com/en/risks/natural-disasters-losses-are-trendingupwards.html#-1624621007.NetherlandsBank,2018.AnEnergyTransitionRiskStressTestfortheFinancialSystemoftheNetherlands,OccasionalStudies,vol16-7,availableatwww.dnb.nl/media/naiiupcg/persberos_transition-risk-stress-test-versie_web_tcm46-379397.pdf.NetworkforGreeningtheFinancialSystem.2019.ACallforAction:ClimateChangeasaSourceofFinancialRisk,availableatwww.ngfs.net/sites/default/files/medias/documents/ngfs_first_comprehensive_report_-_17042019_0.pdf.IPCC,2021.ClimateChangeWidespread,Rapid,andIntensifying,availableathttps://www.ipcc.ch/2021/08/09/ar6-wg1-20210809-pr/OECD,2020.OECDBusinessandFinanceOutlook.OECD,2021.COP21:ClimateChangeinFigures,availableathttps://www.oecd.org/environment/cop21-climate-change-in-figures.htmPankrantz,Nora,RobBauer,andJeroenDerwall,2019.ClimateChange,FirmPerformance,andInvestorSurprises,unpublishedpaper.Penasco,Christina,LauraDiazAnadon,andElenaVerdolini,2021.SystematicReviewoftheOutcomesandTrade-OffsofTenTypesofDecarbonizationPolicyInstruments,NatureClimateChange11:257-272.Schrodt,Philip,2015,CountryInfo.txt:CountryNames,Codes,PlacesandLeaders,availableathttps://doi.org/10.7910/DVN/NBPRDW,HarvardDataverse,V1Severen,Christopher,ChristopherCostello,andOlivierDeschenes,2018.AForward-LookingRicardianApproach:DoLandMarketsCapitalizeClimateChangeForecasts?JournalofEnvironmentalEconomicsandManagement,89:235-254.Stein,J.C.1989.Efficientcapitalmarkets,inefficientfirms:amodelofmyopiccorporatebehavior.QuarterlyJournalofEconomics104:655–669.USGS,2022.ClimateResearchandDevelopmentProgram.UTZ,2022.ClimatePolicyRelevantSectors.WanX.,J.Yang,S.Marinov,J.Calliess,S.Zohren,andX.Dong,2021.Sentimentcorrelationinfinancialnewsnetworksandassociatedmarketmovements.ScientificReports11:3062.IMFWORKINGPAPERSDecomposingClimateRisksinStockMarketsINTERNATIONALMONETARYFUND30WorldBank,2021.WhatYouNeedtoKnowAboutIFC’sGreenBonds.WorldBank,2017.AGuidetoGreenhouseGasBenchmarkingforClimatePolicyInstruments,TechnicalNote14.WorldEconomicForum,2022.GlobalRisksReport.Yang,Yuanchen,2021.ArePassiveInstitutionalInvestorsEngagedMonitorsorRisk-AverseOwners?Both!IMFWorkingPaperNo.2021/158.IMFWORKINGPAPERSDecomposingClimateRisksinStockMarketsINTERNATIONALMONETARYFUND31AppendixTable1.SampleLabelsofNewsDataPublicationDateTitleTextLabel2008-08moresuppliersraiseenergybillsanothergovernmentoption,oneconsideredmorelikely,istoforcecompaniestopayformorecarbonemissionspermitstocovertheirpowerstationoutput.currentlymostoftheiremissionspermitsarefree.Favorable2015-10oilmajorstocollaborateoncleanerenergyeightbigoilandgasgroupsaretomeetinparisnextweekwheretheywillannounceresearchanddevelopmentco-operationoncleanerenergy,aheadofaunglobalclimatechangeconferenceindecember,accordingtothechiefexecutiveoftotal.Neutral2018-08usautosectorshouldhavebeenmorecarefulwhatitwishedforassoonashewaselected,carmakersstartedurgingmrtrumptorethinkpresidentbarackobama’splannedvehiclefueleconomyandemissionsstandardsfor2022-25.lastweek,thetrumpadministrationsetoutitsproposalsforfulfillingthatrequest...UnfavorableDecomposingClimateRisksinStockMarketsWorkingPaperNo.WP/2023/141
VIP
