EUROPEANCENTRALBANK

EUR○SYSTEM

ValentinaDeCicco,IsabellaGschossmann,

ChristofferKok

WorkingPaperSeries

Banklendingimplicationsofclimatestresstests

No3088

Disclaimer:ThispapershouldnotbereportedasrepresentingtheviewsoftheEuropeanCentralBank(ECB).TheviewsexpressedarethoseoftheauthorsanddonotnecessarilyreflectthoseoftheECB.

ECBWorkingPaperSeriesNo30881

Abstract

Doclimatestresstestsaffectbankcreditsupplytobrownfirms?Usingadifference-in-differencesapproachanddetaileddataonindividualbankloansintheeuroarea,thispaperprovidesnovelevidenceontheeffectsoftheECB’s2022climateriskstresstest.Despitenocapitalimplicationsorpublicdisclosures,participatingbankssignificantlyreducedcredittogreenhousegas-intensiveindustriesrelativetonon-participants.Amongaffectedfirms,smallerborrowersweremorenegativelyim-pacted.Notably,onlythebest-performingbanksintheclimatestresstestsignif-icantlyreducetheirbrowncreditafterparticipation.Thisisevidencethatbankswhicharemoreadvancedinclimateriskmanagementmoreproactivelyconsidertransitionrisksintheirlending.Incontrast,bankslessadvancedinmanagingcli-materiskdonottothesameextentdiscriminateagainstpollutingfirms.

Keywords:ClimateRisk,ClimateStressTest,BankingSupervision

JELCodes:E51,G21,G28

ECBWorkingPaperSeriesNo30882

Non-technicalsummary

Inrecentyears,centralbanksandbankingregulatorshavebecomeincreasinglycon-cernedwiththefinancialrisksposedbyclimatechange.Whileitisnotinthemandateofcentralbanksorsupervisorstoimplementclimatepolicies,theyrecognizethatclimatechangeandrelatedpoliciescanaffectboththeeconomyandfinancialstability.Toad-dressthis,bankingregulatorshavestartedencouragingbankstomanagetheirexposuretoclimate-relatedriskscarefully.Onekeytoolareclimatestresstests,whichevaluatehowwellbankscanhandleclimaterisks.Unliketraditionalstresstests,theseclimatestresstestsdonotresultinpenaltiesorpublicdisclosureofindividualresults,raisingthequestionofwhetherthisapproachisstrongenoughtopromptmeaningfulchangesinbanks’behavior.

Thisstudyinvestigateswhetherclimatestresstestsinfluencebanks’lendingpractices,particularlytowardcarbon-intensiveindustry(i.e.,“brown”)firms.UsingdatafromtheEuropeanCentralBank’s2022climatestresstest,weexamineifparticipatingbanksreducedlendingtobrownfirmscomparedtothosethatdidnotparticipate.Wefindthatbanksinvolvedinthestresstestdidindeedreducetheirloanstobrownfirms.Inparticular,smallborrowersaremoreadverselyaffectedbyparticipatingbanks’creditallocationpoliciesfollowingthestresstest.However,inherentdifferencesbetweenpar-ticipatingandnon-participatingbanksaswellasotherfactorsprevalentinoursampleperiod(e.g.,thebroaderadversemacroeconomicenvironment)havelikelyplayedaroleinthelendingdynamicsweobservebeforeandaftertheexercise.

Todigdeeper,wefocusononlythebanksthatparticipatedinthestresstestandexaminehowtheirperformanceinthetestinfluencedlendingbehavior.Ourfindingssuggestthatonlythehighest-performingbankssubstantiallyreducedlendingtobrownfirms,hintingthatstrongerclimateriskmanagementpracticesmightgivebanksmoreabilityormotivationtocutbackonhigh-emissionlending.Importantly,thisresultholdsbeyondtheconsiderationofexternaleconomicconditionsorinherentbank-leveldiffer-

ECBWorkingPaperSeriesNo30883

encesthatcouldbiasourresults.

Thefindingssuggesttwomainpolicytakeaways:First,onlybyenhancingtheirca-pacitytoproperlymeasureandstresstestclimaterisksbankswillbeabletoproperlymanagetherisks.Climatestresstestsandothersupervisoryactivitiesaimingatfoster-ingbanks’climateriskmanagementapproachesarethereforeinstrumentalinsupportingthegreentransition.Second,theresultsshowthatadditionalmeasuresmaybeneeded,especiallyforbanksthatscoredlowerinclimateriskassessments,toencourageindustry-wideclimateriskmanagement.Bothontheregulatoryandsupervisorysidearangeofmeasuresarebeingimplementedinordertoinducethebankingindustrytomoreproactivelymeasureandmanageclimate-relatedandenvironmentalrisks.Ourresultsthereforeunderscoretheimportanceofastrongandenforceablesupervisoryframework,includingclearlydefinedexpectationsaboutclimatestresstestingandriskmanagementmorebroadly,toensurethatbanksmoreeffectivelyintegrateclimaterisksintotheircreditpolicies.

ECBWorkingPaperSeriesNo30884

1Introduction

Centralbanksandbankingsupervisorshavepaidincreasingattentiontoclimatechange-relatedimplicationsinrecentyears.Whileitisnotinthemandateofcentralbanksorsupervisorstoconductclimatepolicies,theyneedtobemindfuloftheriskstotherealeconomyandthefinancialsectorthatclimatechangeandpoliciesaimedtomitigateitmayentail.

Againstthisbackground,inrecentyearsbankingsupervisorshaveprovidedexten-siveguidanceandsupervisoryexpectationstobankshighlightingtheimportancethattheyproperlyandprudentlymanagetheirclimateriskexposures.Oneexampleforsuchguidanceareclimatestresstests,inwhichregulatorsassessinstitutions’levelofpreparednessforproperlymanagingclimateriskwithoutimposingdiscipliningactionex-post.Climatestressteststhusdifferfromtraditionalsupervisorystresstestsintwokeyaspects:Participatingbanksdonothavetofearcapitalpunishmentscontingentontheirperformanceandindividualbankresultsarenotpublished.

Thislackofex-postregulatoryactionraisesthequestionwhethersuchpassivecentralbankclimatepolicyactionsaresufficienttoinducebank-levelresponseswhichalleviatetheirclimaterisks.Giventhenovelnatureofclimatestresstest,theexistingliteraturelacksempiricalworkconvincinglyidentifyingtheireffects.Thecurrentliteraturethere-forecannotfullyanswer(1)whetherclimatestresstestsinducebank-levelchangesincredit,(2)whetherthesechangeshavefirm-levelimplications,and(3)ifso,whythatisthecase.Findinganswerstothesequestionsisimportantsincecentralbankswillneedtoclearlycommunicatehowanyproposedpoliciesarejustifiedwithintheirvariousmandates.ThisisespeciallyrelevantconsideringrecentthreatsbytheEuropeanCentralBank(ECB)tofinebanksoverclimateriskshortcomings.1

Doclimatestresstestsaffectbankcreditsupplytobrown(i.e.,greenhousegasintensivesector)firms?Usingdetailedinformationonindividualbankloansintheeuroarea,

1

SomeeurozonebanksmaybefinedaftermissingECBclimategoal

,Reuters,5June2024.

ECBWorkingPaperSeriesNo30885

weuseadifference-in-differencesdesigntostudywhether,followingaclimatestresstest,participatingbanksadverselychangetheircreditsupplytobrownfirms,relativetonon-participatingbanks.TheECB’s2022climateriskstresstest,towhichaselectionofbankswasexogenouslyexposed,providesasuitablesetting.Ourresultsshowthatparticipatingbanksreducetheirlendingtobrownfirms.Infurtherdisentanglingcreditsupplyanddemand,wefindthatsmallbrownborrowersaremoreadverselyaffectedthantheirlargecounterparts.However,fixed-effectsregressionsandrobustnesstestsshowthatunobservabletime-varyingbankcharacteristicsandselectioneffectswithintreatedbanksdrivetheoverallresults.Indeed,participatinginstitutionsaresignifi-cantlydifferentfromnon-participatinginstitutionsacrossseveraldimensions.Moreover,theyendogenouslylendlessthannon-participatingbanksinthemonthsfollowingtheclimatestresstest.

ToovercomethislimitationandmorefullystudytheeffectivenessoftheECB’scli-matestresstest,werestrictoursampletoparticipatingbanksonlyandanalyzetheeffectofbank-levelperformanceintheclimatestresstest.Wefindthatonlythebest-performingbankssignificantlyreducetheirbrowncreditafterparticipation,pointingtoaselectioneffect:Best-performingbanksarelikelyinabetterspottoreducetheirbrowncredit,ormoreinclinedtodoso.Scoringbankswithoutanyex-postsupervisoryactiondoesthereforenotsufficeasasupervisorypressureinstrument.

Therestofthepaperisstructuredasfollows.Section2providesabriefreviewoftherelatedliterature.Section3explainstheinstitutionaldetailsoftheECB’s2022climateriskstresstest.Section4describestheempiricalstrategyanddata.Section5presentsthepaper’sresults,followedbyrobustnesstestspresentedinsection6.Section7presentsbankperformanceanalysesandsection8concludes.

ECBWorkingPaperSeriesNo30886

2Relatedliterature

Theliteratureonbankcreditandclimatetransitionriskshasgenerallydocumentedthatbankspriceclimatepolicyriskexposureintheirlendingbehavior.

Ivanovetal.

(

2022

)findthathigh-emissionfirmsfaceshorterloanmaturities,loweraccesstopermanentformsofbankfinancing,higherinterestrates,andhigherparticipationofshadowbanksintheirlendingsyndicates.

Delisetal.

(

2023

)and

Chava

(

2014

)alsoshowthatbankschargehigherloanratestoprice-inenvironmentalriskfacedbyhighly-exposedfirms.

KacperczykandPeydr′o

(

2022

)findthatfirmswithhighercarbonfootprintspreviouslyborrowingfrombankswhichcommittedtodecarbonizationsubsequentlyreceivelesscredit.

Reghezzaetal.

(

2022

)studytheParisAgreementin2015asaggregateclimateregulatoryeventandfindthatEuropeanbanksreallocatedcreditawayfrompollutingfirms.

Astophysicalrisks,

OuazadandKahn

(

2022

)showthatintheaftermathofnaturaldisasters,lendersaremorelikelytoapprovemortgagesthatcanbesecuritized,therebytransferringclimaterisk.

Nguyenetal.

(

2022

)findthatlenderschargehigherinterestratesformortgagesonpropertiesexposedtoagreaterriskofsealevelrisk.Conversely,

MurfinandSpiegel

(

2020

)findnoevidenceofsignificantvaluationeffectsbetweenoth-erwisesimilarhomesbutforwhichthetimetoinundationwilldifferdependingonthepaceofthesealevelrise.Overall,theliteratureonbankcreditandclimaterisksuggeststhatbank-levelresponsestosuchrisksarealreadypresentwithoutdirectsupervisoryintervention.

Anotherrelevantstrandofliteraturestudiesbank-levelandrealeffectsoftraditionalstresstests.Testingdifferenthypothesessurroundingtheeffectofbankstresstestsoncreditsupply,

Acharyaetal.

(

2018

)findsupportoftheRiskManagementHypothesis:Stress-testedbanksreducecreditsupplyparticularlytorelativelyriskyborrowerstode-creasetheircreditrisk.Theauthorsidentifyvariouschannelsatworkwhich,however,allrelyonex-postsupervisoryrequirementsforbankstoholdmorecapitalrelativeto

ECBWorkingPaperSeriesNo30887

theircreditriskexposure.Relatedly,

Cort′esetal.

(

2020

)showthatbanksmostaffectedbystresstestsreallocatecreditawayfromriskiermarketsandtowardsaferones.Theyalsoraiseinterestratesonsmallloans.Quantitiesfallmostinhigh-riskmarketswherestress-testedbanksownnobranches,andpricesrisemainlywheretheydo.Astorealeffects,

Groppetal.

(

2019

)findthatbankswithhighercapitalpunishmentsreducelend-ingtocorporateandretailcustomers,resultinginlowerasset,investment,andsalesgrowthforfirmsobtainingalargershareoftheirbankcreditfromthesuchbanks.Fo-cusingonpublicdisclosureeffects,

Flanneryetal.

(

2017

)findthatstresstestdisclosuresareassociatedwithsignificantlyhigherabsoluteabnormalreturnsandhigherabnormaltradingvolume,suggestingthattheygeneratesignificant,newinformationaboutstress-testedbanks.

Schuermann

(

2014

)furtherfindsthatthedisclosureofresultsisacriticalcomponentofstresstestsasitallowstoreestablishtrust.Theliteratureontraditionalstressteststhereforelargelyestablishesthatregulatoryactionintheformofcapitalpunishmentsorresultdisclosuresarekeytoinducebank-levelresponses.

Anothersetofpapershasconsideredtheimpactofbankingsupervision.Relyingonnewmetricsforsupervisoryscrutinyduringthe2016EU-widestresstest,

Koketal.

(

2023

)findthatthediscipliningeffectoncreditriskisstrongerforbankssubjecttomoreintrusivesupervisoryscrutinyduringtheexercise.

Hirtleetal.

(

2020

)showthatbanksthatreceivemoresupervisoryattentionholdlessriskyloanportfolios,arelessvolatile,andarelesssensitivetoindustrydownturns,butdonothavelowergrowthorprofitability.Similarly,

CorellandPapoutsi

(

2024

)showthateuroareabankspassonthecostofcomplyingwithalarge-exposureframeworktoborrowersabovetheexposurethresholdviainterestratepremiaandreductionincredit.

Abbassietal.

(

2023

)analyzetheresponseofbanks’totheassetqualityreviewconductedbytheECBandstudytheassociatedrealeffects.Theyfindthatafteritsannouncement,reviewedbanksreduceriskiersecurityholdingsandcreditsupply,withnegativespilloversonassetpricesandfirm-levelcreditavailability.Thisliteraturethereforesuggeststhatsupervisoryattentionbeyonddirectregulatoryinterventionscanalreadyinducebank-andfirm-leveleffects.

ECBWorkingPaperSeriesNo30888

Arecentsetofpapershasstudiedclimatestresstestsinparticular.Mostofthesepapershavebeennormativeinevaluatingthedesignofclimatestresstests(

Acharya

etal.

,

2023

;

Battistonetal.

,

2017

;

Jungetal.

,

2023

).

OehmkeandOpp

(

2022

)study“green”capitalrequirementsasapotentialregulatoryconsequenceofclimatestresstests.Theyfindthathighercapitalrequirementsfordirtyloanscanreducecleanlend-ing,whiledecreasesincapitalrequirementsforcleanloanscanincreasedirtylending.Thisisbecausechangesincapitalrequirementsaffectcreditallocationviathemarginalloan,whichcanbecleanordirty.

Giventhenoveltyofclimatestresstestingexercises,theliteraturelacksaclearem-piricalidentificationoftheireffects.

Fuchsetal.

(

2023

)arethefirsttoempiricallystudythebank-levelandrealeffectsofclimatestresstests.Theauthorsfindthatstress-testedbanksintheFrenchbankclimatepilotexerciseincreaseloanvolumesbutchargehigherinterestratesforhigh-emittingborrowers.Affectedfirmsincreasetheirclimateriskmanagementeffortsbutdonotexhibitchangesintheiremissions.Wecontributetotheauthors’resultsbystudyingtheECBexerciseasarelativelymoreexogenousidentifica-tionsettingandbyusingmorecomprehensivecreditdata.Notably,

Fuchsetal.

(

2023

)focusesontheFrenchclimatestresstestexercise,whichinvolvedvoluntaryparticipa-tion.Additionally,theiranalysisislimitedtosyndicatedloans,whichmayinfluencethetypesofbanksincludedinthestudy.

Thispaper’sfirstcontributionliesinitsattempttoreconcilethefirsttwostrandsofliteraturehighlightedabove.Wethereforewanttoshinelightonthefollowingques-tions:Doclimateriskspecificstresstestingandbankingsupervisionaffectbanks’creditriskmanagementinthecontextofclimaterisk?Andtowhichdegreearesupervisoryfollow-upsorregulatorypunishmentnecessarytomeaningfullyimpactbanks’manage-mentofclimaterisksintheircreditbusiness?

Astotheliteratureonclimatestresstests,onecontributionistoprovidemoreempir-icalevidencealongside

Fuchsetal.

(

2023

).Goingbeyond,thepaper’smaincontributionliesinitsuniquepositiontomoreconvincinglyidentify(1)bank-levelresponsestocli-

ECBWorkingPaperSeriesNo30889

matestresstests,(2)theirfirm-levelimplications,and(3)theunderlyingmechanismsdrivingbanks’climateriskmanagementintheircreditbusiness.

3TheECB’s2022climateriskstresstest

Climatestresstestsareanoveltooltoassessbanks’resiliencetoclimaterisks,withsev-eralregulatorshavingcompletedsuchexercisesintherecentpast(e.g.,BankofEnglandin2021,theFrenchPrudentialSupervisionandResolutionAuthority(ACPR)in2020,andtheFederalReservein2023).Whiletheseclimatestresstestingexercisessharecommonfeatures(e.g.,creditriskprojectionsunderdifferentclimateshockscenarios),theECB’s2022climateriskstresstestisuniquealongtwokeydimensions.First,itistheonlyexerciseinwhichthenamesofparticipatingbanksarenotpublicknowledge.Second,itisoneofthefewexercisesthatwasnotvoluntary,insteadtargeting104sig-nificantinstitutionsunderECBsupervisionatthetime.Furtherinstitutionaldetailsareexplainedbelow.

3.1Timelineandscope

Figure

1

describesthetimelineoftheexercise.Whiletheactualexercisewasimple-mentedinQ1andQ2of2022,theECBfirstannounceditatthestartofQ42021.Importantly,thisistheofficialdateatwhichparticipatingbankswerefirstnotifiedabouttheirinvolvementintheexercise.2TherewasnoEU-widestresstestbytheEuro-peanBankingAuthority(EBA)andnooverlapwithothercentralbanks’climatestresstestsin2022.3

2See

Lettertobanks

.NotethatthegeneralannouncementthattheECBwouldrunaclimatestresstestin2022wasmadeearlier,attheendof2020/beginningof2021.Untiltheofficialnotificationofindividualbanks,theSingleSupervisoryMechanism(SSM)conductedanindustryconsultationonthedraftmethodologyofthetest.Giventheinnovativenatureoftheexerciseandtheuncertaintyaboutthefinalmethodologyandapproach,weassumethatex-anteadjustmentsandanticipationeffectsbeforetheannouncementarenotmaterial.

3TheECB/SSMhoweverinparallelranadesktop-basedsolvencyexercise,the2022SSMVulnera-bilityAnalysis,inresponsetotheRussianinvasionofUkraineandtheenergycrisisthatfollowed.Itdidnotinvolvethebanksdirectlyandalsodidnothaveanydirectcapitalimplications,soshouldarguablynotbeinfluencingthebanks’lendingdecisions.

ECBWorkingPaperSeriesNo308810

Howdoesthe2022climateriskstresstestfitintotherestoftheECB’ssupervisoryclimateriskexercises?Theexerciseisjustoneofseveralsupervisoryactivitiestoad-dresssuchrisksintheeuroarea.Figures

A1

and

A2

placetheexerciseanditstimelineintheECB’sgeneralclimaterisksupervisoryframework.Bothfigureshighlightthethematicreviewonclimate-relatedandenvironmentalrisksasapossiblethreattoiden-tification:Consideringitscontentandtimelineoverlaps,wecouldpotentiallypickupbank-levelreactionstothethematicreview,insteadoftheclimateriskstresstest.Whilethiscannotbefullyruledout,therelativelyqualitativenatureoftheformeralleviatesthisconcern:Thethematicreviewfocusedonbanks’broaderclimateriskmanagementapproaches,includinggovernanceissues.Theclimatestresstestentailedextensivedatacollectionsandquantitativeanalyses,includingstressprojections,providingbankswithmoretangibleoutcomesintermsoftheirabilitytomeasureclimateriskinthefirstplaceandtheidentificationoftheirvulnerabilitiestosuchrisks.

3.2Structureandmodules

Theclimateriskstresstestexerciseconsistedofthreedistinctmodules:Module1com-prisedanoverarchingquestionnairetoassesshowbanksarebuildingtheirclimatestresstestcapabilitiesforuseasariskmanagementtool.4Module2wasadatacollectionandpeerbenchmarkanalysiswiththeobjectivetoassesswhetherbankswereabletoprovidegoodqualityclimatedataandallowforcomparingbanksacrossacommonsetofclimateriskmetrics.Thefirstmetriccapturedhowmuchbanksrelyonincomefromcarbon-intensiveindustries,basedontheEUTechnicalExpertGrouponSustainableFinancetaxonomyofNACE-classifiedindustries.5Thismetricreferredtotheperiodfrom1January2021to31December2021.Thesecondmetriccapturedthevolumeofgreenhousegasemissionsbanksfinance,basedonscope1toscope3emissionintensitymetricsandasat31December2020.Finally,Module3mimicsabottom-upstresstesttargetingtransitionandphysicalrisks,askingbankstoprovideprojectionsoffinancial

4See

Climateriskstresstest-SSMstresstest2022

,AnnexA.1foradetailedoverviewofthequestions.

5Seefigure

A3

forthefulllistofindustries.

ECBWorkingPaperSeriesNo308811

losses(mainlycreditriskandtoalesserextentmarketrisk)underavarietyofshort-andlong-termclimatescenarios.Figure

A4

providesanoverviewofitsstructure.

Giventhepassivestock-takingnatureofModule1andtheuncertaintysurroundingprojectionsmadeinModule3,weexpectanyobservablebank-levelreactiontooriginateinitscompletionofModule2.Whilewerefrainfromhypothesizinganymechanismsaheadofestablishingasignificantreactioninthefirstplace,thequalitativenatureofModule1andthelong-termscopeofModule3speakagainstimmediateadjustmentsincreditsupply,bothuponannouncementofandparticipationintheexercise.Module2,ontheotherhand,ledtomaterialinformationproductiononthesideofbanks,nexttosignallingclearmetricsthattheECB,asbankingsupervisor,caresabout.Bank-levelreactionstoModule2arealsolikelytodominateanyeffectsfromthethematicreviewasapassivestock-takingexercise.6

3.3Aggregateresults

Thekeyfindingsoftherespectivemodulescanbesummarizedasfollows.7Module1showedthatbankshadmadeconsiderableprogressintheirclimatestress-testingcapa-bilities.However,theexercisealsorevealedmanyshortcomings,datagapsandinconsis-tenciesacrossinstitutions:Importantly,just20%consideredclimateriskasavariableintheirloandecisions;althoughaswewillargueinthispapertheclimatestresstestitselfseemtohaveactedasacatalystforselectedbankstomoreproactivelyaccountfortheserisksintheirlendingdecisionex-post.Module2revealedthatparticipatingsignificantinstitutionsgeneratenon-negligibleincomefromactivitiesrelatedtoGHG-intensiveindustries:Theshareofinterestincomerelatedtothe22mostGHG-emittingindustriesamountedtomorethan60%oftotalnon-financialcorporateinterestincomeonaverage.Moreover,bankslackedactualdataregardingGHGemissions,withca.70%ofreportedemissionintensitiesrelyingonproxies.Finally,Module3showedthatpar-

6Alsoconsiderthatthethematicreviewcovered107significantinstitutionsand79lesssignificantinstitutions.Therearethereforeoverlapsbothinthetreatmentandcontrolbanksusedintheempiricalstrategy,alleviatingconcernsthatthethematicreviewaffectsthetwogroupsdifferentially.Seesection4.1.1fordetails.

7See

2022climateriskstresstest

fordetailsonaggregatefindings.

ECBWorkingPaperSeriesNo308812

ticipatingbanksare,tovaryingdegrees,exposedtothematerialisationofacutephysicalrisksinEurope(e.g.,droughtandheatevents,floodrisk).Therisksbanksarefacinginthisregardarecloselylinkedtothegeographicallocationoftheircreditbusinessesandcouldinsomecasesleadtonon-negligiblelosses.Banksfurtherdidnothaverobustlong-termtransitionplansandshowedlittledifferentiationbetweenpossiblelong-termscenarios.

Aggregateresultsthereforeshowedthattherearestillmanychallengesbanksarefacingwithregardtoclimateriskstresstesting.Accordingly,theexercisewasdeemedausefullearningexerciseforbanksandsupervisors,withahopetoactasacatalysttostrengthenbanks’effortstodevelopclimateriskstress-testingframeworks.8Next,weexplainhowweempiricallydeterminewhethersuchadjustmentsarereflectedinbanks’creditbusiness.

4Empiricalanalysis

Weemployasetofstandardtripledifference-in-differencespecificationstoassessthequestionempirically,wherebyweprogressivelysaturatethespecificationwithfixed-effectsacrossmultipledimensions.Thisallowsto(1)establishthatanyobservedeffectisrobustacrossdifferentspecifications,and(2)determinewhichunobservabletime-invariantortime-varyingcharacteristicsarepossiblydrivingthebaselineresults.Theempiricalstrategy,possiblethreatstoidentification,andthedataandsampleusedarediscussedinthefollowingsubsections.

8See

ECBreportongoodpracticesforclimatestress