07EnvironmentalandSocialScoresMethodologyandFieldInformationABloombergProfessionalServicesOfferingBloombergTerminalESGBloombergESGEnvironmentalandSocialScores1ContentsIntroduction..................................................................................................................................................................2ApproachtoBloombergEnvironmentalandSocialScores.............................................................................................3IssuePrioritiesandScoreTemplateInputs.....................................................................................................................5ScoreFrameworkandIssuePriorities........................................................................................................................8ScoreFrameworkandStructure.....................................................................................................................................8IssuePriorities................................................................................................................................................................9ScoringMethodology.................................................................................................................................................11PrinciplesofQuantitativeESScoresMethodology.......................................................................................................11ParametricApproachtoScoring...................................................................................................................................11FieldAttributes..............................................................................................................................................................13FieldTransformation.....................................................................................................................................................14FieldScoring................................................................................................................................................................14ScoreAggregation......................................................................................................................................................25Sub-IssueScores.........................................................................................................................................................25IssueScoresandDisclosureFactors............................................................................................................................25TechnicalDescription...................................................................................................................................................26PillarScores.................................................................................................................................................................30PillarDisclosure............................................................................................................................................................31EnhancementsandLimitations.................................................................................................................................32Endnote........................................................................................................................................................................33VersionReleaseDateDescriptionDecember2020MethodologyforEnvironmentalandSocialScoresBloombergESGEnvironmentalandSocialScores2IntroductionInrecentyears,environmental,socialandgovernance(ESG)issueshaveincreasinglyinfluencedandimpactedbusinessdecisions.Businessesneedtomanagetheeffectsofclimatechangeandshiftingsocietalnorms,amongotherkeyissues,makingtransparency,disclosure,andmeasurabilityofESGissuesparamount.AdvancesincorporatereportingonESGissuesshouldhelpinvestorsandotherstakeholdersbetterunderstandhowbusinessescanandwillrespondtothisdynamiclandscape–butonlyiftheinformationcanbeevaluatedeasilyandtransparently.Inthisdocument,BloombergoutlinesthemethodologyforproprietaryEnvironmentalandSocialScores(ESScores).Thisinitiative,throughwhichwewillreleasescoresforallsectors,isgroundedinBloomberg’sdecade-longefforttochampionuseful,comparable,andconsistentsustainabilitydisclosuresandtheiruseinfinancialdecision-making.Theincreasedavailabilityofaccurateandtimelyinformationhelpstodrivegrowthinallmarkets.Sustainablefinanceandinvestingwilllikewisebenefitfromstandardized,transparentcompanydata.Asastewardofthisinformation,BloombergplaysakeyroleinfacilitatingcorporateESGreporting.Bloombergalsoservesasacentralpointofcontactinmaintainingaplatformtosupporttheanalytics,includingdevelopingmeasuresforensuringtheintegrityandcomparabilityofthedata.Data-drivenscoresareapowerfulvehicleformakingESGdataaccessibleandintelligibletoinvestorsandotherinterestedparties.Byintroducingtransparent,data-drivenESGScores,Bloombergishighlightingthevalueofsustainabilityinformationandpromotingimproveddisclosurefromawiderangeofbusinessesaroundtheworld.InadditiontoESScores,BoardCompositionScoreshavebeenreleasedasafirststeptowardcomprehensivegovernance(G)scores.ThiseffortwillintegrateseamlesslywithBloomberg’sdata,analyticsandresearchsolutions,Bloombergnewsandmediaplatforms,andthecompany’ssupportformarketinfrastructuretoimproveboththesupplyanddemandsideofsustainablefinance.Indevelopingproprietaryscores,Bloomberg’sworkhasbeeninformedbyinvestorswhoflaggedthecriticalchallengeofdeployingcompany-reportedESGdataintheinvestmentdecision-makingprocess.BloombergESGscoressummarizecorporatesustainabilitydataandsimplifytheintegrationofESGanalysisintobusinessandinvestmentanalysis.Bloomberg’sESScoreshavebeendevelopedbyagroupofspecializedcross-businesscontributorsinresearchandconsultationwithexternalexpertsandthroughactiveengagementwithclientstolearnfromtheirobservationsandexperiences.KeyContributorstoBloomberg’sEnvironmentalandSocial(ES)ScoresSustainableFinanceSolutionsGlobalDataQuantitativeResearchIndexResearchBloombergIntelligenceBloombergNEFBloombergLawandGovernmentBloombergESGEnvironmentalandSocialScores3ApproachtoBloombergEnvironmentalandSocialScoresBloomberg’srationalefordevelopingESScoresisdrivenbythegrowingdemandforESGinformation,theemergenceofglobalsustainabilityframeworks,andthespecificnatureofreportedESGdataandperformancescores.Sustainabilityframeworks,suchasthoseoftheSustainabilityAccountingStandardsBoard(SASB)andtheTaskForceonClimate-relatedFinancialDisclosures(TCFD),1areemblematicoftheshifttowardviewingbusinessrelevanceandfinancialmaterialityasthekeyconsiderationsforselectingindustry-specificESGthemesanddataforuseindecision-makingandreporting.Bloomberg’sgoalindevelopingtheESScoresistoevaluateandpresentESGdatasothatitcanbe:1)integratedintoanalysismoreeasily,2)mademorecomparablewithinspecificindustries,3)usedtoidentifyleadersandlaggards,and4)deployedtocomplementorvalidateresearch-drivenperformancescores.ThekeychallengethatBloomberg’susersfaceinthiseffortistheoverwhelmingnumberofcompany-andthird-party-reportedESGfields,includingboththesourcesandthedatatypes.Lookingatenergyandclimatechangealone,forexample,Bloombergoffersalmost200differentcompany-reporteddatafields.Usersofsustainabilityinformationarealsofacedwithnon-comparabledisclosuresandsubstantialdisclosuregapswithinandacrossfirms.Finally,whileframeworksandreportingrecommendationsareuseful,interpretingandderivinginsightsfromthisguidancerequiressubstantialresearch,informedjudgment,andanalyticalresources.Research-drivenscoreshavethusfarstoodasthemainsummaryindicatorsofperformanceforinvestmentmanagers,assetowners,corporateleaders,andothers.However,suchscoresoftendivergeforreasonsthatarenotnecessarilycleartousersofthisinformation.Inordertoaddressthesechallenges,BloomberghasbroughtarangeofresourcesandassetstogetherindevelopingtheESScores.Thegoalsandprocessesincludethefollowing:•ImprovingESGdatainexistingdatafieldswithmoreprecision,coverage,andhistoryandthroughnewindustry-specificdatasets•ProvidinganorganizingstructureandtaxonomytoclarifyhierarchiesandESissuesandtotranslatefromreportingframeworks•Offeringguidanceonmaterialitytohighlightandrankkeyissuesbyindustry•Leveragingfundamentalresearchtoprovidecontextandpracticalinsights•Usingquantitativeanalysistoincorporatestandingguidelinesandprinciplesandtoaggregatedataforanalyticaluse.Bloomberg’sESGScoresbringtogethervariousdatasourcesofferedontheBloombergTerminal®service,principallycompany-reportedsustainabilityinformationandfinancialfundamentalsdata,withproprietaryresearchassetsandanalytics,suchasBloombergIntelligenceESGresearchandtheBloombergIndustryClassificationStandard(BICS).Bloomberg’sESScoresproductissupportedbyevidence,research,consultation,andanalyticalrigor.BloombergIntelligence’sdedicatedESGteamconductstop-downmaterialityassessmentsbyindustry;thesearedetailedinthisdocument.Abroader,bottom-upconsultationassessesexistingandnewdatatodeterminetheirsuitabilityfordescribingandquantifyingmaterialsustainabilityissues.RigorousquantitativetechniquesarethenappliedtoensurethatmeaningfulsignalsarereflectedintheScores.2BloombergESGEnvironmentalandSocialScores4Bloomberg’sProcessforESScoresProductDevelopmentExternalFrameworksExternalframeworksusedtoprioritizesustainabilityperformancedriversandtogroupissuesKeyIssueResearchResearchonkeyissuesthroughbothproprietaryanalysisandexternalconsultationtospecifythedatafieldsbestalignedwithprioritythemesESGDataAnalysisAnalysisofexistingESGdata,especiallydisclosureandsuitability,todescribekeysustainabilityissuesandactivitymetricsfornormalizationNewESGFieldsDevelopmentandcollectionofnewESGdatafieldstofillgapsinindustry-specificsustainabilityinformationTaxonomy/GroupingGroupingofsustainabilitythemesandcorrespondingdatafieldsIssuePrioritiesResearch,evaluation,andassignmentofissueprioritiesbyindustryDataAttributesAssignmentoffieldqualifiersforimportanceofdisclosure,polarity(e.g.,positive=betterorworse),fit,consistency,andqualityofdatafieldsQuantitativeDataSurveyQuantitativeevaluationofqualityandcoverageofESGdataFieldScoreDeploymentoffield-level,industry-specificscoringmethodologiesAggregationandWeightingIncorporationofweightingschematoaggregatefromBloombergFieldScoresuptoSub-Issues,Issues,andPillarsReviewandFeedbackInternalscorereviewandfeedbacktocalibrateandrefineindustryscoringmodelsBloombergESGEnvironmentalandSocialScores5IssuePrioritiesandScoreTemplateInputsBloomberg’sESScoresdrawonmajorsustainabilityreportingframeworksusedbypubliccompaniesaroundtheworldtohighlightthemostmaterialsustainabilityissues.Bloombergidentifiesdisclosedcorporateinformationthatalignswiththeseissues,particularlywithregardtocorporatestrategy,operations,andpriorities,transformingthisinformationintoausefultoolforinvestmentdecision-makingandothertypesofcompetitiveanalysis.Byembracingmaterialityasthecentralconcept,Bloomberg’sapproachfocusesonthedriversofoperatingperformanceandtheimpactsofsustainableoperatingstrategiesontheenvironmentandsociety.WhiledevelopinginputsfortheESScoresmethodology,BloombergIntelligenceconductedanindustry-specificassessmentofsustainabilityissues,prioritizingandrankingthesetsofIssuesusingproprietaryandexternalsources.TheanalysisofwhatisrelevantandmaterialforeachindustryprovidescontextforthecategorizationofSub-Issues,aswellasdatafieldselection,normalization,andtransformationdecisions.DetailsonBloomberg’ssectorspecificanalysisandIssueprioritizationcanbefoundinourESScoresMethodologyIndustryGuide(IndustryGuide)ontheterminalatBESG<GO>underBloombergESGScores.TheIndustryGuidewillalsoprovidedetailsonbusinessactivitiesforcompanieswithineachsub-industry,explanations,andexamplesforhoweachESIssuescoredimpactsthesub-industry,companyspecificexamplesofESrisksandopportunities,aswellasdetailedmetricsandfieldsusedtoscoreeachsub-industry.ThisanalysisandtheinputsofBIindustryanalystsalsodriveswhichlevelofBICSisselectedtodeterminepeergroupsandscoringtemplates.Forexample,theChemicalsindustriesarescoredinthefollowingBICSLevel4groups:Basic&DiversifiedChemicals,SpecialtyChemicals,andAgriculturalChemicalstodifferentiatethesustainabilityrisksandopportunitiesofthesub-industriesfromoneanother.ESGIntegrationandGuidingFrameworksAsanexampleofthisconsultativeprocess,BloombergreferredtothefollowingESGreportingframeworksaswellasnumerousindustryassociationsduringthedevelopmentofitsESScores.ForacompletelistofframeworksandindustryassociationsbysectorpleaserefertotheIndustryGuide.NameofReportingFrameworkTypeSustainabilityAccountingStandardsBoardNGOTaskForceonClimate-relatedFinancialDisclosures(TCFD)NGOGRINGOCDPNGOTheIndustryGuidewillalsoprovidedetailsonbusinessactivitiesforcompanieswithineachsub-industry,aswellasexplanationsandexamplesforhoweachESIssuescoredimpactsthesub-industry,transparency,anddisclosure.DisclosureasaDimensionofPerformanceCorporate-reportedsustainabilityinformationisattheheartoftheESScoresandBloomberghasinvestedinimprovingthedatasetsthatsupportscoring,withworkundertakentofillgapsandalignthescoringprocesswithemergingsustainabilityframeworks.Giventhedualobjectivesofevaluatingperformanceandincentivizingdisclosure,theESScoresapproachfocusesonimprovingthetransparency,sophistication,andcomparabilityofsustainability-relatedinformationprovidedbyindividualcompanies.TheESGdatausedforESScoresconsistsofvoluntarydisclosurescapturedonlyfromdirect(primary)sourcesinordertoensureaccuracyandconsistencywithoriginalcorporateinformation.Thesesourcesincludesustainabilityreports,annualfilings,proxystatements,corporategovernancereports,supplementalreleases,andcompanywebsites.CertainfieldsarederivedbyBloombergusingcompany-reportedunderlyingdatatoincreasecomparabilityandstandardizationsuchastheEmployeeFatalityRate(F1442),whichusesEmployeeFatalities(ES053)andTotalBloombergESGEnvironmentalandSocialScores6Employees(ES043)asinputsintothecalculation.OtherfieldsusedintheESScoresareobtainedfromfundamentalequitiesdatasets,suchasdataonrevenues.Bloombergrunssophisticated,multi-layerqualitycontrolsystemstoensurethatthedataconformstothehigheststandards.Inaddition,onlycomparableandcomprehensivedataareincludedintheproduct.EnvironmentalandsocialdatafeedingtheESScoresareupdatedannuallyandarealignedwiththefiscalyearend.Itisimportanttonotethatsustainabilityreportingmayhavelagsandreportingdatescandependonthecompany,industry,andregion.Newcompany-reporteddatafields,includingbothquantitativeandqualitativefields,havebeenaddedforthedevelopmentofESScores,oftentofillgapsinindustry-specificsustainabilityinformation.ThefieldIDsforthenewlyaddeddatabeginwithSAfollowedbya3-digitidentifier(e.g.,SA001).Someofthesenewlyaddedfieldsareuniversalforallindustriesandprovideamorecompletelookatacompany’ssustainabilityperformance.Thesefieldshavebeenbackfilledforallavailablehistory.Otherfieldsarespecifictoanindustryandhavedatastartingfromfiscalyear2015.ThesefieldscanbeaccessedviaESGD<GO>orbycreatingcustomtemplatesonFAESG<GO>tovisualizeontheBloombergTerminal®service.Whentherecommendationsofsustainabilityreportingframeworksreachbeyondcurrentcorporatedisclosurepatterns,morewidelyreportedproxyfieldshavebeenintroducedtoenhanceexistingandnewdata.Insomecases,Bloomberghasmodifiedthekeyperformanceindicatorsrecommendedbyreportingguidelinesinordertoenhancestandardizationacrossregions,jurisdictions,andindustries.Futureversionsofthesescoresmayintegratethird-partyinformation.Fieldsthatwouldideallybescoredbutarenotbeingdisclosedatthepresenttimeareaddedtoawatchlist.Allfieldsusedbyeachsector’sscoringframeworkandthosetobetrackedforfutureinclusionintheESScoresappearontheBloombergTerminal®serviceintheESGMaterialityFramework,whichcanbefoundonBESG<GO>underBloombergESGScoresandbyrunningDOCS2093704<GO>.Whilethedatasetincludesbothquantitativeandbinaryqualitative“policy”fields,theformerareweightedmoreheavilytoemphasizethevalueofthequantitativedisclosures.Thequantitativemethodologyhasintegratedadisclosurefactortoincentivizecompaniestoreportonmaterialsustainabilityissuesandtoinstalltransparencyasadimensionofcorporateperformance.Corporatesustainabilitydisclosuresvarywidelybyindustryanddimension.Environmentaldatafieldsbenefitfrombetterquantitativedisclosures.Socialfieldsoftenhavelimiteddisclosure,althoughthisislikelytoevolveovertime.AnyIssuebeingscoredthatreliesheavilyonqualitativedatafieldswillenjoyhigheraveragedisclosuremarksbecausethesefieldsindicatetheexistenceofcorporatedisclosures.However,thefieldsdonotassessthequalityofthosedisclosures.Incaseswherescoringreliesheavilyonqualitativefields,IssueScoresmayreflectalowerevaluationofcompanyperformance,notnecessarilyduetopoormarksondisclosedinformation,butratherasaresultoflackofdisclosure,whichisanimportantdistinction.TheAppendicesintheIndustryGuidesfeaturenotesontransparencyinthescoringmodels,includingtheestimationmethodsusedandprovisionofscoringtemplatesbyindustry.BloombergESGEnvironmentalandSocialScores7IssuePriorityandtheCurrentStateofDisclosureSincetheframeworksusedtodevelopBloomberg’sESScoresdonotassignweightingstotheIssuesthattheyidentifyasimportant,Bloomberghasdevelopedathree-partassessmenttodetermineIssuepriorities:•Probability:EachIssuewasassignedarankingofhigh,medium,orlowtorepresenttheprobabilityoftheIssue(cost/opportunity)materializing.•Magnitude:EachIssuewasassignedarankingofhigh,medium,orlowtorepresentthemagnitudeorpotentialseverityofthefinancialcostoropportunity.•Timing:EachIssuewasassignedaclassificationofshort-term,medium-term,orlong-term.Short–termsuggeststhatthefinancialimpactcanoccurwithin2years.Medium-termindicatesthatthefinancialimpactismorelikelytooccurin2-5years,andlong-termin5–10years.Thefinancialimpactofmedium-andlong-termIssuesmaybemoredependentonphysicalandregulatorychanges.Forexample,environmentalfines,whichwouldbecapturedunderEcologicalImpact,haveamediumprobabilityandahighmagnitude.In2019,ValereceivedUSD125.5millioninfinesfornon-compliancewithenvironmentallawsandregulationsasaresultoftheirtailingdamruptureattheCórregodoFeijãoMine.3AsthiseventresultedinfatalitiesandinjuriesofemployeesitwouldalsobecapturedinfatalityandsafetyincidentfieldsunderOccupationalHealth&SafetyManagement.InMetals&MiningandSteelindustries,EnergyManagementalsohasahighmagnitudeandasthecostsanduseofenergyarecontinuous,theprobabilityisalsohigh.ThisdistinctioninprobabilityisonereasonwhytheEnergyManagementIssuehasapriorityof1forallsub-industries,whileEcologicalImpacthasalowerpriority.Activitymetrics,detailedforeachindustryinAppendix1oftheIndustryGuides,arealsoselectedtonormalizedatafieldswhereappropriate.Whereanoptimalactivitywasnotavailable(e.g.,industryproductionmetrics),themethodologyusesmoreuniversalmetricssuchassalesrevenueornumberofemployees.SummarytablesineachIndustryGuidedescribekeyIssuesandshowtransparencyintoIssuePriorities.Detailsarealsoprovidedontheavailabilityofassociatedquantitativedata,aswellasadditionaltransparencytoexplaintheassignmentofIssuePrioritiesforeachindustry.SummaryinformationaboutdisclosureisalsoincludedineachIssuesection.Notethatfieldgroupswithonlyqualitativedatareceivelowerweightings(asdiscussedinthisMethodology).Fieldlistsassociatedwitheachindustry,includingtheactivitymetrics,areincludedintheIndustryGuideAppendices,asareboxplotsthatillustratedisclosuredispersion.BloombergESGEnvironmentalandSocialScores8ScoreFrameworkandIssuePrioritiesScoreFrameworkandStructureBloomberg’sESScoresarestructuredintothefollowinghierarchy:Pillars,Issues,Sub-IssuesandFields.EachIssuecontainsatleastoneSub-Issue,whichaggregatesassociatedESdatafields.Pillars,IssuesandSub-IssuesToseealistofthespecificIssuesthatarescoredforeachsectorandindustrypleaserefertotheIndustryGuides.ENVIRONMENTALAirQualityGHGEmissionsManagementAirEmissionsGHGEmissionsAirEmissionsPoliciesGHGEmissionsPoliciesGHGRegulationClimateExposureGHGTargetTransitionRiskSustainableProductEcologicalImpactGreenProductEcosystemProtectionEnvironmentalFinesWasteManagementEnvironmentalIncidentsHazardousWasteGenerationHazardousWasteRecyclingEnergyManagementWasteGenerationEnergyConsumptionWasteRecyclingRenewableEnergyUseWaterManagementEnvironmentalSupplyChainManagementWastewaterSustainableSourcingWaterUseWaterUsePoliciesBloombergESGEnvironmentalandSocialScores9SOCIALCommunityRights&RelationsOccupationalHealth&SafetyManagementCommunity&HumanRightsFatalitiesCommunityRelationsHealth&SafetyFinesHealth&SafetyPoliciesEthics&ComplianceSafetyIncidentsBusinessEthicsCompetitiveBehaviorOperationalRiskManagementLegal&RegulatoryManagementOperationalIncidentsOperationalPreparednessLabor&EmploymentPracticesLaborActionsProductQualityManagementOrganizedLaborProductQuality&SafetyTrainingSocialSupplyChainManagementSupplierSocialComplianceIssuePrioritiesBloombergIntelligencehasprioritizedandrankedindustry-specificrisksandopportunitiesassociatedwithmaterialsustainabilitythemesandembodiedinBloombergIssues.Theserankingsareprovidedbyindustry,withrationalesforeachIssue’sprioritylevel,whichreflectsthefollowinginputs:•InternaldiscussionsandinterviewswithBloombergIntelligenceindustryanalysts.•AnalysisandnewsbyBloombergIntelligenceandBloombergLawandGovernmentthathighlightfinancialimpactsrelatedtokeyenvironmentalandsocialrisksintheindustry(e.g.,litigation,fines,reputational/brandrisk,employeeturnover).•Bloombergproprietaryresearchondatarelatedtoindustryactivitiesandoperationsandtheirimpacts.•Academic/scientificstudiesthatpointtoindustryexposuretothehighlightedfactors.•Regulatoryactionsinrelevantjurisdictionstolimit,track,andcontrolnegativeimpactsassociatedwithindustryactivitiesandoperations.ThelevelofBICSthatisusedforeachscoredgroupofcompaniescanvarydependingonthealignmentofthebusinessactivitiesandtheirrelativesustainabilityexposures.Industryandsub-industryspecificpriorityrankingscanbefoundintheIndustryGuides;asanillustrativeexample,theOil&Gaspriorityrankingsareprovidedinthetablesbelow.BloombergESGEnvironmentalandSocialScores10HeatMapofIssuePrioritiesfortheOil&GasIndustry–EnvironmentalIssuesandPrioritiesExploration&ProductionIntegratedOilsMidstreamRefining&MarketingDrilling&SupportOilfieldServices&EquipmentAirQualityClimateExposureEcologicalImpactEnergyManagementGHGEmissionsManagementWaterManagementHeatMapofIssuePrioritiesfortheOil&GasIndustry–SocialIssuesandPrioritiesExploration&ProductionIntegratedOilsMidstreamRefining&MarketingDrilling&SupportOilfieldServices&EquipmentCommunityRights&RelationsEthics&ComplianceLabor&EmploymentPracticesOccupationalHealth&SafetyManagementOperationalRiskManagementDarkgreenrepresentsthehighestprioritiesandgrayrepresentsthelowestpriorities.BloombergESGEnvironmentalandSocialScores11ScoringMethodologyBloomberg’sapproachtoscoringESGperformanceischaracterizedbyabottom-up,model-drivenmethoddrivenprimarilybyself-reported,publiclyavailableinformationthatresultsinafullytransparent,parametric,rules-basedscoringframework.Itfeatures:•Qualitativeinputfromresearchanalystsandindustryexpertsforidentifyingappropriatefieldsandmetrics,aswellastheirrelevancetospecificissuesandindustries.•Statisticalanddatasciencetechniquestoassistinidentifyingpeergroups.•Factoranalysistoaidinidentifyinguniqueenvironmentalandsocialissues.•Incentivesforimprovedtransparencyanddisclosure,sothatthebestscoresreflectbothgoodsustainabilityperformanceandgooddisclosure.ThefollowingsectionsdetailtechnicalprotocolsforgeneratingEandSscores.Attributesassignedtodatafieldsandtheapproachtofieldtransformationsaredescribedbelow.Field-levelscoringapproachesareillustratedinthenextsection,followedbyweightingandaggregationdecisions.Additionaldetails,suchasscoringtemplatesbyindustryanddatafieldswithnodisclosuretobemonitoredinfutureversions,areavailableintheAppendices.PrinciplesofQuantitativeESScoresMethodologyOnekeygoalofBloomberg’squantitativeapproachtoscoringEandScompanyperformanceistobuildanextendible“toolkit”toscorecorporateESdatabasedonthetypeofdata.InanalyzingtheEandSdatatodevelopthesescores,certainprinciplesandbenefitsoftheapproachadoptedforBloomberg’sproprietaryESScoreshaveemerged:•Guidelinesorprinciples,whenavailable,mustbetranslatedintoquantitativethresholds.•Normalizingdataisnecessarytocomparemetrics.•Modelchoicehelpsaddresssizebias.•Factoranalysisaidsinidentifyingindividualissuesfrommultiplerelatedfields.•Datasciencetechniquesassistinidentifyingpeergroups.•Scoringfromfieldsuptohigherlevelsrequiresanaggregationapproachthatrewardsconsistentperformanceandpenalizesunevenperformance.•Disclosuremustbeascoredimensionduetolimitedreportingandwidereffortstoimprovetransparencyandstandardizationofreporting.ParametricApproachtoScoringQuantitativeanalysisofcorporateESGperformanceinthecurrentmarketplaceforsustainablefinanceandinvestingdataandanalyticsreliesheavilyonquantitativeandqualitativeanalysesbyresearchersandonawiderangeofinformationsources,includingcompanyreporting,NGOmonitoringanddata,governmentinformation,andvariousnewssources.Theseinformationsourcesandtheresultingsummaryanalytics(i.e.,scores)havevaryinglevelsoffrequencyandtransparency.Asaresult,thedriversofESGScorescanbedifficulttodiscern.Muchlikeotherinvestmentresearch,theresultsmaycapturesubjective,potentiallyinvisibleweightingsandheuristicsthatleadtobiasesnoteasilyunderstoodbyauser.Bloomberg’sapproachtotheseESScores,aswellasouraforementionedGscores,aimstodeliveratransparent,parametric,rules-basedscoringframeworkdrivenbyself-reportedcompanyinformation.Theuseofaparametricapproachthatcloselyapproximatestheempiricaldistributionisintendedtoaddressthechallengesthatariseincommonlyusedapproachestorankingsustainabilityperformancebypercentiles:BloombergESGEnvironmentalandSocialScores12•Parameterscanbeestimatedinarobustmannertolimitsensitivitytooutliers.•Parameterscanbefixedorslowlyadjustedovertime,forinstance,usingdatafromthethreepreviousyearstoestimateparametersforagivenyear’sscoring(asinthecurrentmethodology).•Companiescanbescoredassoonasdatainthecurrentscoringyearisavailablewithouthavingtowaitforallcompaniestoreportbecauseparametersareestimatedfromhistoricaldata.ParametricAnalysisPercentilesBucketingScorescompaniesastheyreportRequiresmost/alldatatobereportedforapeergrouptorankPreservestrendswithslower-movingparametersSuffersfromsmallpeergroupsorlow-disclosureindustriesBetterIdentifiesperformancedifferencesinpeergroupsCanobscuretrendsindataovertimeIsmorerobusttooutliersCanbesensitivetooutliersThetableabovehighlightsthedifferencesbetweenusingparametricapproachestoscoringvs.bucketingintopercentiles.Whilepercentilesbucketingisasimpleapproachtosortingcompanies(orcompanieswithinadefinedpeergroup)bythevalueofafieldandtoassigningscoresthatcorrespondtosomequantile(i.e.,percentiles),thedrawbackisthatdataisnotalwaysavailableatthesametime.Asaresult,quantilescannotbecomputeduntilalldataforareportingyearisavailable.RoleofPeerGroups:Akeyconsiderationintheestimationprocessisestablishingsuitablepeergroupsfromwhichdatacanbesampledtodetermineappropriateparameters.Thegeneralapproachtopeergroupsistocomparedataandevaluateforstatisticaldifferences.Ifstatisticaltestssuggestthatthedifferencesinthedataacrossindustriesarenotsignificant,thenpoolingthedataacross2ormoreindustriesisconsideredasameansofincreasingthenumberofdatapointsavailablefordeterminingscoringparameters.Often,however,thereisinadequatedatatodeterminethatthedatasetsarestatisticallydifferent.Inthatcase,theESScoresapproachdeferstotheopinionoffundamentalindustryandESGanalysts’judgement.Incaseswhereestablishedexternalbenchmarksorwidespreadconsensusofidealstandardsexist,FieldScorescanbecomputedforallcompaniesagainstoneuniversalstandard,i.e.,anestimateofasinglesetofparametersforallcompanies.Asabenefit,eachcompany’sscorewouldbedirectlycomparable,inanabsoluteway,toeveryothercompany’sscore,regardlessofsectorsandindustries.Asecondadvantageofuniversalscoringisthatpoolingallavailabledatatoestimateoneuniversalrelationshipacrossallcompanieslendsgreaterstatisticalconfidencetotheparameterestimates.InBloomberg’sESScoresmethodology,fatalitiesprovideanexampleofsuch“absolutescoring.”Inrecognizingtheuniversalvalueofahumanlife,wescorefatalitiesforallcompanies,regardlessofindustryorsize,toacommonstandard.Allbinaryfieldsarealsoabsolute,bytheirnature,becausetheyrepresentonlytwostates—acompany’sdisclosureofagivenpolicyorthelackofone.However,theevaluationofseveralfieldsismeaningfulonlywithinsmallerpeergroups.Scope1GreenhouseGasCarbonDioxideEmissionsisonesuchfield.Theexpectedquantityofacompany’semissionsisafunctionofthenatureandtheamountoftheactivitythatproducesthem.Emissionsproducedbyupstreamexplorationandproductionactivitiesarenotcomparabletothoseproducedbydownstreamdistributionactivitiesbecausethenatureofthoseactivitiesisdifferent.Asaresult,scoresevaluatedrelativetosmaller,industry-specificpeergroupsare“relative,”i.e.,rankedonlywithinthatpeergroup(industry,inthecaseabove).Inotherwords,acompanythatscoresa7onaparticularfieldisbetterthanapeerthatscoresa5,butnodirectwayisavailabletocompareittoacompanyfromadifferentpeergroupthatalsoscoresa7inthatfield.ThepeergroupusedforeachfieldislistedintheappendicesoftheIndustryGuides.BloombergESGEnvironmentalandSocialScores13Finally,itisworthnotingthatscoringbasedonempiricaldistributions,regardlessofparametric-orpercentile-drivenapproach,canleadtohighscoresthatmaynotcorrespondtoanESGgoal.Forexample,ifallparticipantsinagivenindustryhavehighpollutionlevels,thebestrelativeperformerwillstillreceiveahighscore.Wherequantitativeguidanceisavailable,thecurrentversionoftheESScoresattemptstointegratemeaningfulthresholds,suchasa1%ofrevenuesmaterialitytest.However,giventherarityofsuchbenchmarks,Bloomberg’scurrentapproachcorrespondsmostcloselyto“bestinclass”approachestointegratingESGintoportfolioconstruction.FutureiterationswillbeabletoincludethresholdssuchasthosesetbytheEuropeanUnion’sTaxonomyforclimatechangemitigationandadaptation.FieldAttributesDatafieldsareaggregatedintoSub-Issues,Issues,andPillars.TheinputofBIanalystsdeterminestherelativeimportanceofvariousscoresintheaggregationprocess.Furthermore,keyattributesarecapturedforuseasscoreinputs.IndustryspecificmodelinputsareprovidedinAppendix2intheIndustryGuide.BIIssuePriorityisdeterminedbyBloombergIntelligenceresearchanalyststoreflectthevaryingdegreesofrelevanceandimportanceofvariousenvironmentalorsocialIssuestoaparticularindustry.ThisvaluedeterminestheweightassignedtoeachIssueinaggregatedEandSscores.FieldTypeindicateswhetherfieldvaluesarequantitativeorbinary.Quantitativefieldtypeshavevaluesthatarenumerical.Binaryfieldtypeshaveeither“Yes”or“No”valuesandrepresentBloomberg’sEandSpolicyfields—indicatingwhetherornotacompanyhasdisclosedinformationonaparticulartopic.Assuch,binaryfieldsdonotassessthequalityofdisclosure.Fit/QualityvaluesarealsodeterminedbyBIanalysts.ThesevaluescanbeHigh,Medium,orLowandareusedtoweightindividualFieldScoresintoaggregatedSub-IssueScores.•High(H):ThemetricisagoodmeasureofwhatiscalledforinvariousESGreportingframeworks,andthedataiscomparable.•Medium(M):Themetriciseitheragoodmeasure(asabove),orthedataiscomparable,butnotboth.•Low(L):Themetricisnotagoodmeasureandthedataisnotcomparable,orthefieldisaqualitativePolicyfield.4Polarity(positiveornegative)isusedtoreflectactivitiesthatdecreaseorincreaseE,S,financial,operational,orreputationalrisks.Inotherwords,positivepolarityisassignedwhereahigherfieldvaluemeanslowerEorSrisk,orhigherEorSopportunityand,therefore,ahigherscore.DisclosureFactor(DF)determinationsaremadetoguidethetreatmentofmissingdatawithinthescoringframework.EachfieldisassignedaDFratingofA,B,orC.•DFRatingA:FieldsusedinthescoringmodelthatarecalledforbyamultitudeofESGdisclosureframeworksandinvestorsreceiveaDFofA.•DFRatingB:FieldsusedinthescoringmodelthatareaproxyforfieldscalledforbyamultitudeofESGdisclosureframeworksorarecalledforbyonlyalimitednumberofESGdisclosureframeworksandinvestorsreceiveaDFofB.•DFRatingC:Fieldsusedinscoringthatarenotdisclosedbycompanies,butinsteadarederivedbasedonnon-ESGinformation(forexample,valueofcarboninpubliclydisclosedoil&gasreserves)receiveaDFofC.Activitymetricsareusedtonormalizesustainabilityperformancerelativetooperatingorfinancialmetrics.Forinstance,GHGemissionsscalemostcloselywiththeeconomicsofproduction,whereasspendingonworkertrainingscalesmostcloselywithnumberofemployees.Incaseswhereprecisescalingquantitiesarenotavailable,moreuniversalmetrics,suchasRevenuesareused.BloombergESGEnvironmentalandSocialScores14Activitymetricsthatquantifytheamountofactivitymaydifferdependingonindustryfundamentalsandtheavailabilityofwidelyreporteddata.Formanyindustries,revenuesareusedasthecommonactivitymetricduetothewiderangeofproductsandthelackofproduct-specificfinancialdata.Fieldsthatarealreadynormalized—suchasper-unitmeasures,percentages,andstandardizedcalculationslikeTotalRecordableIncidentRate(TRIR)—arenotassignedcorrespondingactivitymetrics.Fieldsthatarecategoricalorbinary,includingpolicyfields,arenotassignedcorrespondingactivitymetrics.Similarly,fieldsthatarescoredregardlessofcompanysizeoractivitylevels—suchasnumberofsitesinenvironmentallysensitiveareas—donothaveactivitymetrics.TaxonomyInclusionStartYear(four-digityear)isusedtodeterminethefirstfiscalyearforwhichafieldwillbeincludedinthescoringframework.Ifthevalueisblankthefieldwillbeincludedinthescoringframeworkfromfiscalyear2015.Forinstance,theGlasgowFinancialAllianceforNetZero(GFANZ)waslaunchedin2021.Thus,fieldSA904,whichindicateswhetheracompanyisaGFANZsignatory,has2021asitsTaxonomyInclusionStartYear.TaxonomyInclusionEndYear(four-digityear)indicatesthatafieldisnolongerconsideredmaterialorrelevanttothescoringframeworkandwillbeexcludedfromthescoringframeworkafterthespecifiedfiscalyear.Inotherwords,thespecifiedfiscalyearwillbethelastyearforwhichsuchafieldwillaffectscoring.Ifthisvalueisblank,itmeansthatthefieldremainsmaterialandwillbeincludedinthescoringframework.FieldTransformationBeforescoring,itissometimesusefultoapplyafieldtransformationtocalculate“clean”valuesforadesiredmetricortoreducetheimpactofmissingdatabyusingaproxycalculation.ThegoaloffieldtransformationsistomaximizethesuitabilityofthedataspecifiedforevaluationinthescoringmodeltothesustainabilityIssuebeingscored.Asanexample,fieldSA023:PercentageofHazardousWasteRecycledistransformedtorepresenttheamountofhazardouswastenotrecycledasapercentoftotalwaste.Thus,thepolarityforscoringchangesfrompositivetonegativebecauseofthisfieldtransformation.Thistransformationisdesignedtoensurethatacompanywithnohazardouswasteisnotpenalized.Thetransformationformulais((100-SA023)/100)SA016.Thiscanbeinterpretedas:•SA023isthepercentageofhazardouswasterecycled•SA016isthepercentageofallwastethatishazardous.•100-SA023:isthepercentageofhazardouswastenotrecycled.•(100-SA023)/100SA016isthehazardouswastenotrecycledasapercentageoftotalwaste.If,forexample,60%ofhazardouswastewasnotrecycledand50%ofallwastewashazardous,then30%ofallwastewasunrecycledhazardouswaste.Mostfieldtransformationswillapplytoallsectors;however,therearesomeformulasthatareusedtoincreasetheamountofavailabledatathatarenotneededforindustrieswithadequatereporting.FieldScoringThisprocessusesvariousfield-levelscoringmethodsthatdependonFieldType,Unit,Polarity,andpresenceofanActivityMetric.ResultingFieldScoreswillrangefrom0to10.Aparametricapproachisusedtoscorefields.Forallfieldtypes,parametersareestimatedempiricallyforpeergroups,withtheexceptionofbinaryfields,categoricalscoringfields,andFatalities.Scoresarecomputedforthecurrentyear’sdatausingparametersthathavebeenestimatedfromdatathatcorrespondstothethreeyearspriortothecurrentyear.BloombergESGEnvironmentalandSocialScores15Dependingonthetypeoffieldandthenatureandavailabilityofaccompanyingdata,parametersareestimated,andfieldsarescoredusingdifferentstatisticaltechniques.Eachmethodisdescribedwithsomeillustrativeexamplesinthissection.Appendix1intheIndustryGuidecontainsatablethatliststhespecifictechniqueandpeergroupusedforeachfield.Finally,inafewcases,somefieldsarenotrelevanttoallcompaniesinasectorand,so,maynothavevaluesdisclosed.AnexamplewouldbeSA141-NuclearWaste.Insuchcases,asetofconditionsisusedtodetermineifthefieldisrelevanttoacompany.Ifitisfoundtoberelevant,thefieldisscorednormallyfollowingtheprocessdetailedabove.However,ifthefieldisdeterminedtonotberelevant—inthisexample,becauseacompanydoesnothaveanynuclearoperations—thenthefieldisawardedapre-determinedscoreandfulldisclosurefactorpoints.BloombergESGEnvironmentalandSocialScores16IntensityFieldsAlargenumberoffieldtypesrepresentvariousquantities(e.g.,numberofspills,amountofemissions)thatneedtobescaledbyanactivitymetricbeforetheycanbescored.Regressiontechniquesareemployedhere.Parametersthatdescribetherelationshipbetweentheactivitymetricandfieldvaluesforparticularpeergroupsareestimated.Theassumeddistributionineachcasedependsonthetypeoffieldvalue.Thus,separateprocessesarerequiredtoestimateregularIntensityfieldsandCountIntensityfields.RegularIntensityFieldsESGanalystscommonlyuseintensityratiostodetermineacompany’senvironmentalorsocialimpactasitpertainstoaparticularESGfield,especiallythoserelatedtotheeconomicsofproductionsuchasgreenhousegasemissions.TheintensityratioiscalculatedasFieldValue/ActivityLevel.Accordingly,Intensityfieldsarepairedwithactivitymetricstoallowthescoringofthefieldvaluerelativetoacompany’sactivitylevel.Thisisdonetoaddresstheextenttowhichthesizeofacompanyortheextentofitsphysicalprocessesthatdriveemissionsmayinfluencetheassociatedsustainabilityexposures.Anexamplewouldbetoanalyzeandrankcompanies’GHGScope1emissionsrelativetoproduction.Bloomberg’sempiricalanalysissuggeststhatthiscanbeenhancedintwoways:•Therelationshipsareclosertolinearonlogarithmicscales,ratherthannominal.•PlottingActivityversusFieldValueshowsdiminishingmarginalintensityforsomefieldsandindustries.Inotherwords,incertainindustries,ascompanyactivitygrows,theincrementalamountofimpact(emissionsproduced,forexample)perunitofactivityfalls.Oneconsequenceofthisisthatnominalratioscanbebiasedagainstsmallercompaniesforactivitiesthatdisplayapatternofdiminishingmarginalintensity.Themethodologyproposedhereisflexibleandappearstobetterfittheeconomicsofproductionandscalepatternsthatareevidentinthedata.Infurtherdetail,thecurrentestimationprocessforscoringthesefieldsisasfollows.Acompany’sintensityratioistypicallyexpressedas:𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼=𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴Thiscanberewrittenas:𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼=𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼×𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝛾𝛾Weintroducetheexponentγtocapturepossiblenonlinearitybetweenproductionactivitiesandtheirimpacts.Ifγ=1,thentherelationshipislinear;γ<1wouldimplymarginalimpactthatdecreaseswithincreasinglevelsofactivity.Rewritingthisequationinlogarithmictermsandintroducinganinnovationtermε,wehave:𝑙𝑙𝑙𝑙𝑙𝑙𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼=𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒+𝛾𝛾𝛾𝛾𝛾𝛾𝛾𝛾𝛾𝛾𝛾𝛾𝛾𝛾𝛾𝛾𝛾𝛾𝛾𝛾𝛾𝛾𝛾𝛾+𝜀𝜀Thisallowsustoutilizeregressiontechniquestoestimatethenecessaryparameters.Theintensitytermistheintercept,theelasticityterm(γ)istheslopeco-efficientandtheinnovation(ε)isthecompany-specificterm.εisassumedtobenormallydistributed,withameanofzeroandastandarddeviationofσ.BloombergESGEnvironmentalandSocialScores17Wefirstestimatetheintensity,elasticity,andstandarddeviationoftheinnovations(σ)forallcompaniesinagivenpeergroup(e.g.,industry)foraspecificyear.Wethenaveragetheestimatedintensities,elasticities,andstandarddeviationoftheinnovationsoverthethreemostrecentyears.Thisprovidesuswithamodeltopredictacompany’sImpactgivenitslevelofActivity.TheFieldScoreforCompanyiiscalculatedintermsofthenormalizedinnovationas:𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑖𝑖=10×�1−𝐶𝐶𝐶𝐶𝐶𝐶�𝜀𝜀𝑖𝑖𝜎𝜎��Thatis,εiisthedifferencebetweenacompany’sactualandexpectedImpact(inlogterms)andtheScoreisafunctionofhowmuchlowerorhigherthecompany’sImpactisrelativetowhatispredictedbythepropertiesestimatedfromitspeergroup.CDFisthecumulativedistributionfunctionforthestandardnormaldistribution(withameanof0andastandarddeviationof1).Accordingtothismodel,acompanythathasalowerimpactthanwhatispredictedfromthepropertiesofitspeergroup(i.e.,anegativeinnovation)willreceiveahighscoreandviceversa.Finally,itisworthpointingoutthatinallcaseswheretheelasticity(γ)is1—whichisthetraditionalassumption—therankingonintensities(Impact/Activity)isequivalenttotherankingoninnovations(ε).Figure1showsanexampleforscoringScope1GreenhouseGas(GHG)emissions.Thedashedlinerepresentsthelinethatcanbedrawnfromtheestimatedintercept(𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙)andslopeco-efficient(γ).Anydatapointsonthislineareconsidered“average”andwouldreceiveascoreof5.Pointsabovetheline,coloredinshadesofred,indicatecompaniesthathavehigher–than-averageGHGemissionsthantheirpeers,adjustedfortheirsize(i.e.,theactivitymetric—Revenueinthiscase),andreceivelowerscores.Similarly,pointsbelowtheline,coloredinshadesofgreen,representcompaniesthathaveperformedbetterrelativetotheirpeersandthusreceivehigherscores.Asanexample,in2018,ChemicalscompanyLGChem(051910KS)reported5.42millionmetrictonsofScope1GHGemissionsanditsRevenuewasapproximately$25.6billion.Thispointliesbelowtheestimatedlineandresultedinascoreof7.8.AirLiquide(AIFP)reportedsimilarRevenue,approximately$24.8billion,buthad15.39millionmetrictonsofGHGemissionswhichresultedinascoreof2.9.Figure1:ExampleofScoringScope1GHGCO2Emissions(F0947)BloombergESGEnvironmentalandSocialScores18CountIntensityFields—NegativePolarityManycountfieldssuchasthosethatenumerateincidentsandspills,havenegativepolarity.Weestimatetheparametersforthesefieldsusingregressiontechniquesaswell,butratherthanimplementingthe(log-log)linearregressionillustratedpreviously,weassumeanegativebinomialdistributionforcountfields.Thiscapturesthenatureofthevalues(beingintegers)andshowsthatmanyofthevaluesarezero.Figure2showsthescoringcurvefortheNumberofSignificantSpills.ThedashedlinerepresentsthecombinationoftheNumberofSignificantSpillsandRevenuethatwouldbeconsidered“average”andreceiveascoreof5.Valuesthatarereportedasexactly0areshownasstars.In2018,ChemicalscompanyRongshengPetroChemicalCompany(002493CH)hadRevenueofapproximately$13.8billion.ItreportednoSignificantSpillsandreceivedascoreof10asaresult.Ontheotherhand,EastmanChemicalCompany(EMNUS)hadapproximately$10.2billioninRevenuebutreported54SignificantSpills.ThisissignificantlyhigherthantheNumberofSpillsconsideredaverageforthatlevelofRevenueand,accordingly,resultedinascoreof0.8.Figure2:2018ScoringCurveforNumberofSignificantSpills(SA215)Zero-inflatedIntensityFieldsSomefieldshavevaluesthatcanbeassumedtobecontinuouslydistributed,butwithapositivemassatzero;anexamplewouldbethevolumeornumberofhydrocarbonspills.5Insuchcases,weuseaTweediedistribution(specifically,acompoundPoisson–gammadistribution)toaccountforthenatureofthedistribution.Fieldvaluesequalto0areassignedascoreof10,whileotherfieldvaluesarescoredbasedontheactivitymetricandtherelevantparametersestimatedforthefield.Figure3showsthatOil&GascompanyMedcoEnergy(MEDC)reported356m3ofhydrocarbonsspilledin2018.Ithad$5.8billionofTICandreceivedascoreof0.38.KosmosEnergy(KOS)reported0m3ofhydrocarbonsspilledandreceivedascoreof10.ItsTICof$5.1billionwasnotarelevantinputintothescorecalculationsinceallcompaniesreporting0m3ofspillsareawardedaperfectscore.BloombergESGEnvironmentalandSocialScores19Figure3:ExampleofScoringHydrocarbonSpills(ES249)PercentageFieldsPercentagefieldslieonarangefrom0to100,whichmeansvaluesofeither0or100shouldgetfullcredit(i.e.,ascoreof10),dependingonthepolarity.Forsuchfields,thecumulativedistributionfunction(CDF)forbetadistributionisusedtodefineascoringcurve.Betadistributionsareafamilyofcontinuousprobabilitydistributionsdefinedonanintervalfrom0to1parametrizedbytwopositive-shapeparameters,denotedbyαandβ.Inthefollowingexample,thescoreappearsontheY-axis,whilethefieldvalueappearsontheX-axis.Figure4providesanexample;Chemicalscompany,NittoDenkoCorporation(6988JP)—whichreported20.6%HazardousWasteoutofTotalWastediscardedin2018—receivedascoreof6.0;PPGIndustries(PPGUS)reported49.7%inthesameyearandreceivedascoreof1.8.Figure4:2018ScoringCurveforPercentageofHazardousWaste(SP016)BloombergESGEnvironmentalandSocialScores20RateFieldsFieldsforwhichtheunitisaratearescoredslightlydifferentlydependingonpropertiesoftheratefieldsuchasitspolarity,theexpectedmodeofthedistribution,andwhetherzeroratesarecommon.Forexample,incidentratesandfatalitiesbothhavenegativepolaritiesandcommonlyreportedzerovalues.Ontheotherhand,somefieldsthatwouldtypicallybescoredasregularintensityfieldsarereportedasrates.Inthecaseofaregularintensityfield,aswehavedescribedpreviously,wetakelogsofthefield’svalueanditscorrespondingactivitymetricandscoreitusingstandardregressiontechniquesandanassumednormaldistribution.Consequently,whentherelationshipisreportedasarate—e.g.,TrainingspendingperemployeeorEnergyperunitofproduction—alognormaldistributionisthemostappropriateanditisequivalenttoscoringthefieldasaregularintensityfieldwhentheelasticityoftheintensityisbelievedtobe1.RateTypeHasZeros?ModePolarityDistribution#ofParametersExamplesNotesEpisodicRateY0-Exponential1TotalrecordableincidentrateExponentialdistributionfitsdatacharacteristics(0isacommonincidentrate)andoneparameterleadstostableestimatesFatalityRateY0orhigher-Gamma2FatalityrateAnextraparameterisneededtofitthetailtobetterdifferentiatelowperformanceIntensityRateNStrictlypositive+or-Lognormal2Trainingspendingperemployee,EnergyperunitofproductionDataandmodearestrictlypositiveEpisodicRateFields—NegativePolarityFieldsforwhichtheunitisincidentsperunittimeandhaveanegativepolarityassociatedwiththemarescoredusingcurvesspecifiedbytheexponentialdistribution.Thisguaranteesascoreof10forfieldswithavalueof0.Figure5showsanexampleofsuchafield.BharatPetroleum(BPCL)reportedatotalrecordableincidentrateforcontractorsof0.05(per200,000hoursworkedorper100contractors)in2018andreceivedascoreof9.02,whileEQTCorporation(EQT)reportedarateof0.79andreceivedascoreof1.73.BloombergESGEnvironmentalandSocialScores21Figure5:2018ScoringCurveforTotalRecordableIncidentRate–Contractors(ES261)FatalityRateFields—NegativePolarityForfatalityrates,thescoringcurveisspecifiedbyagammadistributionandtheparametersarenotestimatedfromthedata,butaredefinedtoresultinascoringcurvewiththefollowingproperties:•Afatalityrateof0willresultinascoreof10.•Anyfatalityrateabove0willhaveitsscoreadjustedbyafactorof0.7.Thus,themaximumpossiblescoreforanon-zerofatalityrateis7.•Afatalityrateof1(in1,000)willgetascorenear1,whileratesgreaterthan2.5willreceivescoresnear0.AsseeninFigure6,JohnWoodGroup(WG/)hadafatalityrateof0.02in2018anditsscorewas5.13.BaytexEnergy(BTE)hadafatalityrateof0.57andsoitsscorewas1.36.Figure6:TotalWorkforceFatalityRateScoringCurve(RX389)BloombergESGEnvironmentalandSocialScores22RateIntensityFieldsSometimes,afieldthatwouldtypicallybescoredasaregularintensityfieldisreportedasarate.Insuchcases,weestimatetheparametersforarateintensityfieldusingalognormaldistribution.Themeanandstandarddeviationareestimatedtocorrespondtothecentraltendencyandwidthofthedistribution.Sincethereisnopreviouslydeterminedidealrateforthesepositivepolarityratefields,asisthecasewithnegativepolarityratefields,thefieldvaluethatreceivesascoreof10isalsodeterminedempiricallyforthatfield.Figure7showsthescoringcurveforEnergyPerUnitofProduction,whichhasanegativepolarity.In2018,NucorCorporation(NUE)reported1.36MegaWatthours(MWh)ofenergyconsumedperunitofproduction,resultinginascoreof9.8.Severstal(CHMF),ontheotherhand,reported6.68MWhofenergyconsumedperunitofproductionandreceivedascoreof2.3.Figure7:2018ScoringCurveforEnergyPerUnitofProduction(ES494)Figure8showsanexampleofapositivepolarityrateintensityfield:TrainingSpendingperEmployee.Saipem(SPM)receivedascoreof0.98for$187spentperemployeein2018,whileKSInnovation(096770)scored9.43for$3,118spentperemployee.BloombergESGEnvironmentalandSocialScores23Figure8:2018ScoringCurveforTrainingSpendingPerEmployee(RX321)CategoricalScoringFieldsIncertaincases,wheredataisinsufficientforestimatingparametersreliably,thescoresaredeterminedbygroupingrangesofoutcomesintocategoriesandassigningthesamescoreforallvalueswithinarange(completelistofcategoricalscoringfieldsandassociatedvaluesinAppendix3).Suchfieldsmayormaynothaveassociatedactivitymetrics.Twoexamplesofthistypewouldbe:1)thenumberofenvironmentalfinesand2)theamountofenvironmentalfines.Thefirstexampleillustratestheconceptthatcompaniesfinedasmallnumberoftimeswillreceiveapenaltyforonetoninefines,whereascompaniesfinedasubstantialnumberoftimeswillreceiveharsherscorepenalties.Inthesecondexample,fortheamountofenvironmentalfines,1%ofrevenuesisusedasacommonlyacceptedthresholdforfinancialmateriality,withthedeepestpenaltyat1%ormoreofrevenue.Amountsgreaterthan0andupto1%willreceiveasmallpenalty.ValueScore0101-9610-993100+0Figure9:CategoricalScoringforNumberofEnvironmentalFines(ES032)ValueScore0%10Greaterthan0%andupto1%7Greaterthan1%andupto2%2Greaterthan2%andupto5%1Greaterthan5%0Figure10:CategoricalScoringforAmountofEnvironmentalFines(ES033)withActivityMetricdefinedasSalesRevenueTurnover(IS010)BloombergESGEnvironmentalandSocialScores24BinaryFieldsBinaryfieldsrepresentBloombergESGpolicyfields,whicharegivenavalueofYor1,ifthecompanydisclosesontheselectedtopic,andavalueofNor0,ifthecompanydoesnotdiscloseonthattopic.Inotherwords,thesefieldsareindicativeofdisclosure,notthecompany’sactualperformanceonthetopicthatisdescribedinthedisclosure.IfthefieldisBinary,thecompanygetseitherafullcredit(ascoreof10),orzerocredit(ascoreof0)basedonthefieldvalue(Y/N),aswellastakingpolarity(positivevs.negative)intoaccount.Asanexample,SA161-GHGEmissionsReductionPolicyindicateswhetherornotthecompanyhasdisclosedapolicyorastrategytoreduceGHGemissionsspecifically.Ifthecompanyhasdisclosed,itgetsascoreof10;otherwise,itgetsascoreof0.Binary+FieldsBinary+fieldsaresimilartoBinaryfields.However,inthiscasethereisaquantitativeelementunderpinningthepolicythatisdisclosed.AnexampleofsuchafieldisSA559-NetZeroEmissionsTargetwhichindicateswhetheracompanyhassetEmissionsTargetstobecomecarbonneutralbyatargetdateinthefuture.SimilartoaBinaryfield,acompanywillgetafullcredit(ascoreof10)ornocredit(ascoreof0)basedonthefieldvalue(YorN,respectively).Inadditiontothis,thecompanywillalsoearndisclosurecredit(relativetoitsdisclosurerating)dependingonthefieldvalue(Y/N).IfthefieldvalueisYitsdisclosurepointswillcontributetothepointsearned,andifthefieldvalueisN,itwillnotreceivethosedisclosurepoints.BloombergESGEnvironmentalandSocialScores25ScoreAggregationGeneratingacompositethatdescribesperformanceacrossbroadersustainabilityissuescanbecomplexinlightofthedisclosureissuesdiscussedpreviously.Asaconsequence,Bloomberg’sproprietaryapproachtowardaggregationattemptstorewardconsistentperformanceandpenalizeunevenperformance.However,italsoworkstotemperthepenaltiesbymakinguseofotherattributes,suchasresearch-drivenBIIssuePriorities.FieldScores,asdescribedhere,rolluptoSub-IssueScores,IssueScores,andPillarScores.DisclosureasadimensionofperformanceistakenintoaccountattheIssuelevel,whereaDisclosureFactorisintroducedtosummarizetheavailabilityofquantitativefieldsforscoring.Bloomberg’sapproachtoaggregationattheIssuelevelemphasizesquantitativedisclosure.Becausepolicyperformancecanbeperfect,evenwhilequantitativedisclosureispoororevenzero,theaggregationapproachaimstominimizethepotentialtoscorewellthroughdisclosureofqualitativeinformationalone.Sub-IssueScoresSub-IssueScoresareaggregatedfromFieldScoresbyaweightedaverage,dependingontheFit/Qualityattribute.TheFit/Qualitylevelisusedtodeterminetheweight,whereHigh=9,Medium=4,andLow=1.Ifacompanydoesnotdiscloseonagivenfieldinagivenyear,thatfieldisignored,resultinginaredistributionoftheweightsattachedtoeachfield.ThelevelsforeachfieldaredeterminedbyBIanalysts,asdescribedinprevioussections.AslistedinFigure1,forexample,F0947(Scope1GreenhouseGas/CarbonDioxideEmissions)isassignedaHighlevelintheChemicalssector,whileSA119(CommunityEngagementPolicy)isassignedaMediumlevel.IssueScoresandDisclosureFactorsSub-IssueScoresareaggregatedintoIssueScores.InadditiontocapturingsustainabilityperformanceIssuesprioritizedbymaterialityrank,IssueScoreshighlightdisclosureperformance.6ByincorporatingthelevelofdisclosureattheIssueScorelevel,Sub-IssueandFieldScoresreflectonlydisclosedperformance(orareblankduetolackofavailabledata).OnlyquantitativefieldsandBinary+fieldsareconsideredinthecalculationoftheDisclosureFactor.First,binarypolicyfieldsreflectdisclosureinthemselves;ifBloombergdoesnotfindevidenceofacertainpolicy,thefieldissetto“No”andisnotblank.Second,companiesmaydisclosequalitativeinformation,butnotsupplydataforthequantitativefields.Forexample,intheWaterManagementIssue,manycompaniesreportonWaterPolicies(ES247),butfewercompaniesdisclosetheamountoffreshwatertheywithdraw(SA020).Itishardertoevaluatequalitativedisclosuresgiventheirwide-rangingcontentandcomplexity.Asaresult,Bloomberg’sapproachtoaggregationattheIssuelevelemphasizesquantitativedisclosuretoavoidaperfectpolicyperformancescore,withpoororzeroquantitativedisclosures.Essentially,theaggregationapproachaimstominimizethepotentialtoscorewellbydisclosingonlyqualitativeinformation.Inordertoaccomplishthis,thedisclosurelevelsarebuiltintoaseparateDisclosureFactor,whichisthenaggregatedintoperformancescoringattheIssueScorelevel.Bloomberg’sdualgoalofmeasuringperformanceandincentivizingdisclosuredictatesthatitwouldbeanincompletemeasureofacompany’sperformancetoaggregatetoahigher-levelscorebyaveraginglower-levelscoresonlyfordatathathasbeendisclosed.BloombergESGEnvironmentalandSocialScores26Insum,thisapproachattheIssueScorelevelismotivatedbythefollowingprinciples:•Nouseofimputationforundisclosedvalues:Bloomberg’sproprietaryscorescurrentlyreflectself-reported,publiclyavailableinformation,orthelackthereof.Thisadvancesthegoalofincentivizingtransparency.Usingimputationwouldhavetheoppositeeffect.7•Desirabilityofbothgooddisclosureandgoodperformance:Thebestscoresshouldcomefromtransparencyanddecision-usefulsustainabilitydisclosures,aswellasfromgoodsustainabilityperformance.Ifonlyoneaspectisgood,thescoresarecapped.•Incentivesfordisclosure:Bloomberg’scorevalueistransparency.BloombergESGScoresofferanincentivetocompanydisclosure,evenifadisclosurereflectspoorperformance.Exceptfortheveryworstcases,evenapoorperformanceinagivencategoryshouldresultinabetterscorethanundisclosedperformance.Insummary,IssueScoresworkasfollows:1.DetermineanaverageperformancescoreofSub-IssueScores;thisisthe“PerformanceScore.”2.MeasuredisclosureofquantitativeandBinary+fields,withtheresultbeingaweightedpercentage,calleda“DisclosureFactor.”3.TheDisclosureFactordeterminesaperformancerange.4.ScaleandshiftthePerformanceScoreintothedisclosure-drivenrange,asnotedinFigure11.DisclosureFactorIssueScoreRange00-310-10Figure11:TargetIssueScoreRangesforVaryingLevelsofDisclosure5.ZerodisclosureresultsinPerformanceScoresbeingadjustedtoa0-3range.PerfectdisclosureresultsinPerformanceScoresbeingadjustedtoa0-10range.6.Additionally,thereisadisclosure-incentiveIssueScoreboostforallbutthelowestPerformanceScores(everythingabove1.5).TechnicalDescriptionIssueScoresareafunctionoftheweightedgeneralizedmean(p-mean)oftheSub-IssueScores(i.e.,thePerformanceScore)andaDisclosureFactor(DF).TheSub-IssueScoreweightsaregivenby:•Onequarter(¼)iftheSub-Issueonlycontainsbinarypolicyfields•One(1)otherwiseP-meansareusedtorewardexcellenceacrosstheboardandtopenalizelessconsistentperformancebetweenthevariousSub-Issuesbeingaggregated.AswithallBloombergESG(sub)scores,weuseweightedshiftedp-meanswiththepowerp=0.5andshifts=1.8𝑀𝑀(𝑥𝑥,𝑤𝑤,𝑝𝑝,𝑠𝑠)=��𝑤𝑤𝑗𝑗�𝑥𝑥𝑗𝑗+𝑠𝑠�𝑝𝑝𝑛𝑛𝑗𝑗=1�1/𝑝𝑝−𝑠𝑠BloombergESGEnvironmentalandSocialScores27Asnotedearlier,DisclosureFactors(DF)areassignedatthefieldleveltodeterminetreatmentofmissingdatainthescoresmodel,withthefollowingratingsofA,B,andC.Eachconnotesexpectationsaboutthenatureofdisclosure.PolicyfieldsarebinaryandalwayshaveavalueofYes(thespecifiedpolicyispubliclydisclosed)orNo(thepolicyisnotpubliclydisclosed),andthusdonotrequiremissingdatatreatmentanddonothaveaDisclosureFactor.Acompany’sDisclosureFactorscoreforanIssueiscomputedasaweightedpercentageoverallfieldswithinanIssuetopic.Pointsareearnedifthefieldisdisclosedandtheactivitymetricisdisclosure(whererelevantonly),meaningthatthepointsareearnedwhereweareabletogenerateafieldscore;thepointvaluedependsonafield’sDisclosureFactorasshowninFigure12.CertainfieldsarelessdetailedortransparentthantheirAorBcounterparts.ThesefieldsaredesignatedA-orB-andareassigned25%ofthepointvalueofthefullletterrank.Forexample,theoptimalreportingprocedurefortheTotalRecordableIncidentRateistoreportforemployees(ES121)andcontractors(ES261)separately.Accordingly,bothofthosefieldshaveaDisclosureRankofA.However,ifacompanyonlydisclosesanaggregatednumberfortheentireworkforce(SA201),thatfieldhasaDFofA-andreceivessomepoints,butnotthefullpointvalueforthedisclosure.DisclosureFactorPointValueA5A-1.25B2B-0.5C0Figure12:PointValuesforDisclosureFactorCalculationTheDisclosureFactoristhencomputedasDF=∑PointValueiffieldcanbescoredFields∑PointValueFieldsThisresultsinanumberbetween0and1.TheDisclosureFactorisusedtodetermineUpperandLowerTargets(UTandLT)forIssueScores.Bothtargetsincreasewithincreasinglevelsofdisclosure.ThisisillustratedinFigure12.TheUpperTargetis3iftheDFis0,anditis10iftheDFis1.Thus,acompanythatdoesnotdiscloseanyquantitativeinformationcannotreceiveanIssueScoregreaterthan3.Furthermore,disclosinginformationthatonlypertainstoquantitativefieldsinwhichperformanceisgooddoesnotnecessarilyresultinahighIssueScore,astheDFwillaccountforthelackoffulldisclosurebycappingthescoreattheUT.TheUTisacurvedlinethatincreasesmoredramaticallyatlowerlevelsofdisclosure,reflectingthehighermarginalvalueofnewdisclosureatlowlevelsofdisclosure.UT=3+�√DF×(10−3)�TheLT,forperformancescores1.5orgreater,9is0.45iftheDFis0anditis4iftheDFis10.TheLTismeanttoprovideascoreincentivetoincreasedisclosure,evenforfieldswhereperformanceisnotexemplary,asallbutthelowestPerformanceScores(thosebelow1.5)willseeIssueScoresproportionallyflooredattheLTthatcorrespondstotheDF.BloombergESGEnvironmentalandSocialScores28GiventheSub-Issuep-meanM(i.e.,thePerformanceScore)andUpperandLowerTargets,theformulafortheIssueScoreis𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆(𝑀𝑀,𝐷𝐷𝐷𝐷)=�𝐿𝐿𝐿𝐿1.5×𝑀𝑀𝑀𝑀<1.5𝐿𝐿𝐿𝐿+��𝑈𝑈𝑈𝑈−𝐿𝐿𝐿𝐿8.5�×(𝑀𝑀−1.5)�1.5≤𝑀𝑀≤10Visually,Figures14-16showhowtheUpperTarget(UT)andLowerTarget(LT)defineatargetrangeofpossibleIssueScoresalongtheY-axis.Thistargetrangeisafunctionofthelevelofdisclosure.Onlyverypoorperformance(i.e.,aPerformanceScoreoflessthan1.5)resultsinascorebelowtarget.0123456789100.00.10.20.30.40.50.60.70.80.91.0IssueScoreDisclosureFactorUTLTFigure13:UpperandLowerTargets(UTandLT)asaFunctionofDisclosureFactorFigure13:UpperandLowerTargets(UTandLT)asaFunctionofDisclosureFactorBloombergESGEnvironmentalandSocialScores29Figure14:IllustrationofIssueScoreWhenDisclosureFactor=0Figure15:IllustrationofIssueScoreWhenDisclosureFactor=0.5012345678910012345678910IssueScorePerformanceScoreLTUTIssueScore012345678910012345678910IssueScorePerformanceScoreLTUTIssueScoreBloombergESGEnvironmentalandSocialScores30Figure16:IllustrationofIssueScoreWhenDisclosureFactor=1PillarScoresThePillarScoreisaweightedgeneralizedmean(p-mean)ofIssueScores,wheretheweightsaredeterminedbytheIssuePriorityranking.Thevaluesofthepowerandshift(p=0.5ands=1)arethesameastheyareinthecalculationofIssuescores.Step1:InitialWeightsBIIssuePriorityrankingsareassignedweights(w̃)thataredeterminedasafunctionofaranking(k)oftheirrelativeimportance.𝒘𝒘̃=𝟏𝟏+𝒆𝒆𝟎𝟎.𝟓𝟓×(𝟑𝟑−𝒌𝒌)Notethattheweightsdonotdecreaseinalinearfashion.ThisreflectstherelativelyhighimportanceofthetoprankingsandimpliesthatlowerpriorityIssueshavealessereffectonPillarScores,butstillserveanessentialdisclosurerole.Step2:WeightingAdjustmentIssueScorescontainingonlybinaryfieldshavetheirweightreducedby80%toreflectthatquantitativefieldshavebetterscoringpowerthanbinaryfields.Forexample,theOperationalRiskManagementIssueforOil&GasMidstreamProducersonlyreflectswhetherornotacompanyhasanEmergencyResponseandPreparednessPolicy(SA086,whichisapartoftheOperationalPreparednessSub-Issue).012345678910012345678910IssueScorePerformanceScoreLTUTIssueScoreBloombergESGEnvironmentalandSocialScores31PillarDisclosureThePillarDisclosuremeasuresthelevelofdisclosureacompanyoffersforthefieldsundereachofitspillars.UnliketheDisclosurefactorthatisusedtocomputeIssueScores,thePillarDisclosurerequiresonlytherelevantQuantitativeorBinary+fieldtobedisclosedanddoesnotrequiretheactivitymetrictobedisclosedtoearntheassociatedPoints.ThePillarDisclosurevalueiscomputedasfollows:PD=∑PointValueifdisclosedFields∑PointValueFieldsThisresultsinanumberbetween0and1.ThePillarDisclosuredoesnotaffectthePillarScoreandshouldbeusedforinformationalpurposesasameasureofdisclosureonly.BloombergESGEnvironmentalandSocialScores32EnhancementsandLimitationsBloomberg’sESScoresareintendedtointroducetransparentscoringforcompaniesbasedoncompany-discloseddata,proprietaryfundamentalindustryresearchandproprietaryquantitativetransformationsandanalyses.Currentlimitations,primarilydrivenbytheunevenandrapidlyevolvingnatureofcompanyreporting,aswellasformalframeworksusedbycompaniestoguidetheirreportingdecisions,shouldultimatelybeaddressed,andovercomeandthusenabletheenhancementofBloomberg’sESScores.Enhancementswillfocusonimprovingtheeffectivenessofthescoresinassessingbothsustainabilityperformanceandthequalityandcomprehensivenessofdisclosure.Theseenhancementsmayresultinchangestothescoringtemplatesandparametersthatcurrentlydrivethequantitativemodel.Assessmentsandpotentialchangesentailthefollowingactivities:•Sectorreview:Eachsectorframeworkwillbereviewedonceperyeartoevaluatechangesinguidanceembodiedinframeworks,corporatedisclosure,availabilityofcompany-reportedandthird-partydata,andotherkeydrivers.Notably,fieldsidentifiedasonwatchfordisclosurewillbere-examinedforinclusion.Datafieldsmayalsoberemoved.•Newdataandimputation:Dataenhancementsmayincludeimputationofkeydatafields(e.g.,greenhousegasemissions),inclusionofadditionalproprietarydata(e.g.,ESnewssentimentdata,geolocateddata,orasset-levelinformation)orintegrationofthird-partydata(e.g.,governmentorNGOdata).•Regulatoryactionorcollectiveconsensus:Newadditionstothemethodologycouldbeintroducedinresponsetoregulatorybehavior,suchasthatintroducedbytheEuropeanUnionTaxonomyoradditionalimpactmeasurement,accordingtotheSustainableDevelopmentGoals.•Restatements:Restatementsmaytakeplaceonoccasionduetocorrectionsordatabackfills,forexample.Anyrestatementswillbemanagedwithconsultationandexpertjudgmentandchangeswillbecommunicatedclearly.Giventhechallengesofincompletereportingfromindustry,Bloomberg’sESScoresaimtosummarizerisksthatcanfeasiblybeassessedthroughacombinationofdatacollection,research,andanalysis.However,thedatafieldsandanalysisusedinthesescoresarenotexhaustive,butratherrepresentativeofsustainabilityrisksandopportunities.ESScoreusersareexpectedtocombinethesescoreswiththeirownanalysisandjudgmenttodeterminetheirsuitabilityfortheintendedgoal.Bloombergexpectstocommunicatechangestothemethodologytousersonanannualbasis,atleast.Bloombergalsointendstoestablishaprocesstofacilitatestakeholdercommunicationandeffectiveconsiderationoffeedbackabouttrendsindisclosure,marketparticipants’useofscores,andthequantitativemodels,amongotheritems.Bloombergintendstotakescoringmethodologydecisionswithinternalandexternalconsultationviaappropriatescoresgovernanceandnormaluserfeedbackchannels.InadditiontotheESScores,Bloombergdeterminesand/ormakesavailableanumberofothercalculatedvalues,suchasindices,fixings,financialrates,andothervalues.AswithallsuchBloomberg-calculatedvalues,usagerestrictionsapply.Specifically,unlessexpresslyagreedinwritingbyBloomberg,neithertheESScoresnoranyinformationordataprovidedinconnectionwiththeESScoresmaybeusedforanyofthefollowingpurposes:(i)valuationoraccountingpurposes;(ii)todetermineanyinterestorotheramountspayableunderorinrespectofafinancialinstrumentorafinancialcontract;(iii)todeterminethepriceatwhichafinancialinstrumentmaybeboughtorsoldortradedorredeemed;(iv)todeterminethevalueofafinancialinstrument;or(v)tomeasuretheperformanceofaninvestmentfund,includingwithoutlimitation,forthepurposeoftrackingreturnordefiningtheassetallocationofaportfolioorofcomputingperformancefees.BloombergESGEnvironmentalandSocialScores33Endnote1BoththeSustainabilityAccountingStandardsBoard(SASB)andtheTaskForceonClimate-relatedFinancialDisclosures(TCFD)receivedfoundingsupportfromBloomberg.2ThisisdescribedintheMethodologywithsupportinginformationinAppendices.32019SustainabilityReport,Vale,http://www.vale.com/EN/investors/information-market/annual-reports/sustainability-reports/Sustainability%20Reports/Relatorio_sustentabilidade_vale_2019_alta_en.pdf4Policyfieldscanbe“M”iftheyaretheonlyfieldsinanIssuecategory.5Acontinuousrandomvariableisoneforwhichthesetofpossiblevalues(itsrange)isinfinite,ascontrastedtoadiscreterandomvariablethathasacountablesetofpossiblevalues(e.g.,therollofadie).Sincethecontinuousrandomvariablehasaninfinitenumberofpossiblevalues,theprobabilityofobservinganysinglespecificvalueiszero(onlyrangesofitsvaluescanhaveanon-zeroprobability).However,sincemanycompaniescanreportzerohydrocarbonspills,theprobabilityofobservingavalueofexactlyzeroisnotzero.Hence,weutilizeadistributioninwhichtheprobabilityofobservingazerovalueisinflated.6IssueScoresarenotjustaboutincorporatingdisclosure.Theyareinsomesensethemostvaluablesub-scores—theGoldilocks“justright”pointwhereindividualESGissuesarescoredwithoutthedetailof“toomany”FieldScores.7Goingforward,dataimputationmaysupportmoretacticalfieldinclusionwithadjustmentstoavoiddisclosurebias.Anyimputationswouldseektoavoiddisclosurebiasasaconcernsothat,whenafieldisincludedthatfewcompaniesdisclose,companiesarenotgivenanincentivetoavoiddisclosingsubparresults.8Thep-valueof0.5ischosenasthemidpointofthevaluesthatrepresentanarithmeticmean(p=1)andageometricmean(p=0).Theshiftvalues=1ischosentoavoidlargepenaltiesforscoresnear0.9Moreprecisely,whentheweightedp-meanofpolicyanddisclosedquantitativeFieldScoresis1.5orgreater.
VIP
