DownloadedfromDownloadedfrom/rof/LowCarbonMutualFunds*MarcoCeccarelli1,StefanoRamelli2,andAlexanderF.Wagner3AbstractClimatechangeposesnewchallengesforportfoliomanagement.Inournot-yet-lowcarbonworld,investorsfaceatrade-offbetweenminimizingtheirexposuretoclimaterisksandmaximizingthebene?tsofportfoliodiversi?cation.Thisarticleinvestigateshowinvestorsand?nancialintermediariesnavigatethistrade-off.AfterthereleaseofMorningstar’snovelcarbonriskmetricsinApril2018,mutualfundslabeledas“l(fā)owcarbon”experiencedasigni?cantincreaseininvestordemand,es-peciallythosewithhighrisk-adjustedreturns.Fundmanagersactivelyreducedtheirexposureto?rmswithhighcarbonriskscores,especiallystockswithreturnsthatcorrelatedmorewiththefunds’portfoliosandwerethuslessusefulfordiversi?ca-tion.These?ndingsshedlightonwhetherandhowclimate-relatedinformationcanre-orientcapital?owsinalowcarbondirection.V46M.Ceccarellietal.DownloadedfromDownloadedfrom/rof/Howshouldinvestorsbehaveinthefaceofclimate-relatedrisksandtheenergytransitiontoalowcarbonworld?Toanswerthisquestion,itisimportanttorecognizethataccount-ingforclimaterisksininvestmentdecisionsbringsinvestorsbothbenefitsandcosts.Ontheonehand,shunningcarbon-intensive,“brown”assetscanreduceaninvestor’sexposuretoclimaterisks.Theseriskshaveyettofullymaterialize,bothintermsofphysicalconsequencesandsocietalreactions,andmanyobserversbelievethattheyarecurrentlyunderestimatedinassetprices(StroebelandWurgler,2021).Ontheotherhand,inournot-yet-lowcarboneconomy,excluding“brown”assetsandinvestingonlyinthoseconsidered“green”requireinvestorstoforegoopportunitiestodiversify.Thistrade-offisparticularlysalientinassetmanagement,whereportfoliodiversification,notonlythefeaturesofindi-vidualsecurities,playsacrucialroleinreducingoverallinvestmentrisk(Markowitz,1952).Inthisarticle,westudyhowinvestorsandassetmanagersnavigatethistrade-off.Wefocusonthemutualfundindustry,whichrepresentsanimportantshareofglobalfinancialmarkets,1andexploitaquasi-naturalexperimentinvolvingasuddenincreaseinboththeavailabilityandsalienceofinformationoncarbonrisk(climatetransitionrisk),thatis,theclassofriskderivingfromthetransitiontoalowercarboneconomy.AswedescribeinmoredetailinSection2,onApril30,2018,Morningstar,themostimportantdataproviderinthemutualfundindustry,releasedanewPortfolioCarbonRiskScorederivedfromfirm-leveldataprovidedbySustainalytics,whichMorningstarhascontrolledsince2017.ThenoveltyofMorningstar’sPortfolioCarbonRiskScoreishighlightedbythefactthatitcor-relatesonlymildlywithotherportfoliometrics,basedonpreviouslyavailableenvironmen-talscoresfromSustainalytics,Refinitiv,andMSCIKLD.Basedonitsnewcarbonriskscore,combinedwithrelativelystandardinformationonfirms’fossilfuelinvolvement(FFI),Morningstaralsoissuedaneco-labelformutualfunds—thelowcarbondesignation(LCD).WeusealargesampleofactiveEuropeanandUSmutualfundstostudyinvestors’andfundmanagers’reactionstotheseinformationshocksproducedbythepublicationofMorningstar’sPortfolioCarbonRiskScoreanditsassociatedLCDeco-label.WedeveloptheconceptualframeworkguidingourempiricalanalysesinSection3.Wefirstconfirmthat,inlinewithextantliterature(e.g.,Engleetal.,2020;BoltonandKacperczyk,2021a),individuallowcarbonsecuritiesarelessriskythanotherfirms,bothintermsofexposuretonegativeclimatechangenewsandrealizedreturnvolatility.Wethenshiftourfocustotheportfoliolevel.Onemaynaivelythinkthattheriskpropertiesoflowcarbonfundsshouldmirrorthoseoftheirlowcarbonholdings.Such,wefind,isnotthecase.Theinvestmentriskofaportfoliodependsnotonlyonthevarianceofitsindivid-ualholdings’returns,butalsoonthecovarianceofthesereturns(Markowitz,1952).Empirically,whilelowcarbonfundshavelowerexposuretoclimaterisks,theirvolatilityisnotlowerthanthatofmoreconventionalfunds.Infact,wefindthatthemutualfundswiththelowestcarbonriskscoreshavehighervolatilitythanthosewithmedianscores.Thesourceofthisresultisthehighdegreeofindustryconcentration(Kacperczyk,Sialm,andZheng,2005)oflowcarbonfunds.ThesefundsoverweightIT,retail,andhealthcarefirms,LowCarbonMutualFunds47Downloadedfrom/rof/whiletheyunderweightenergy,materials,andutilityfirms.Beyondtheindustryconcentra-tion,thefactthatlowcarbonfundsholdfewerstocksdoesnotsignificantlyfurtherexplainDownloadedfrom/rof/InSection4,westudythereactionsofmutualfundinvestorstotheApril2018informa-tionshock.Fundsreceivingthe“LowCarbonDesignation”enjoyedasubstantialincreaseintheirmonthlyflowsrelativetootherfunds.TheeconomicimpactoftheLCDlabelcorre-spondstoanaverageincreaseinflowsofapproximately36basispointseachmonththroughtheendof2018;thisincreaseisequaltoabouttwo-thirdsoftheeffectonflowscausedbyaone-standard-deviationstrongermonthlyfinancialperformance.Beforethenewdatabecameavailable,investorslikelyusedMorningstar’ssustainabilityGlobesasanimperfectproxyforexposuretocarbonrisk.Intuitively,ifafundwithfewGlobesreceivedtheLCD,itwouldcomeasalargersurprisetoinvestors.Consistentwiththislogic,wefindlargereffectsonflowsinsuchsituations.Inaddition,LCD-labeledfundswithstrongrisk-adjustedperformanceexperiencedamorepronouncedflowpremium.Moreover,afterthepublicationoftheLCDlist—butnotbefore—qualifyingforthelowcarboneco-labelresultedinparticularlylargeextraflowsinmonthsofgreaterattentiontoclimatechange,asmeasuredbyGooglesearchintensity.Alltheseresultsareconsistentwithinvestorstakingboththebenefitsandthecostsintoaccountwheninvestinginlowcar-bonfunds.InSection5,weemployadatasetofmonthlyportfolioholdingstostudythereactionsoffundmanagerstothereleaseofMorningstar’sportfolioandfirm-levelcarbonriskinfor-mation.Weshowthat,afterApril2018,fundmanagersactivelyrebalancedtheirportfoliostoreducetheircarbonrisk.Onaverage,relativetotheperiodbeforethepublicationofMorningstar’scarbonriskmetrics,mutualfundsreducedtheirpositionintheaveragehighcarbonriskfirmbyabout0.17basispointsoftheirassetsundermanagement(AUM)permonth.Thiseffectiseconomicallymeaningful,consideringthatthemedianmonthlypos-itionchangeiszeroforthewholesampleand2.8basispointsfornon-zeropositionchanges.Managersreactedtocarbonrisknotonlywithaone-shotrebalancingoftheirport-folios,butalsobyintegratingthenewinformationintotheirflow-driveninvestmentdeci-sionsaftertheinitialshock.Inparticular,weobservethatfundsexperiencinglargenegativenetflowssoldhighcarbonriskassetsmoreaggressivelythandidotherfunds,whilefundsexperiencinghighinflowsincreasedtheirstakesinlowcarbonriskassets.Furthercross-sectionalevidenceindicatesthat,asweexpected,fundswithhigherexanteindustryconcentrationreactedmorestronglytothereleaseofthenewcarbonriskin-formation.Forthesefunds,shiftingtolowercarbonriskassetsislesslikelytodecrease(andmayevenincrease)theirdiversification.Theyarealsolikelytoserveclientswhoarelessinterestedinbroaddiversificationinthefirstplace.Importantly,wefindthatwhenmanagersreducedtheirpositionsinstockswithascoreofmediumorhighcarbonrisk,theydidsomoreaggressivelyforthosewithahigherreturncovariancewiththeremainderoftheportfolio,consistentwithanattempttopreservediversification.Thisarticlecontributes,first,byprovidinginsightsintothebenefitsandcostsofgreeninvestmentproducts.Existingresearchsuggeststhatfirmswithbetterenviron-mentalperformancehavelowerexposuretoclimate-relatedrisks,andarepricedac-cordingly(e.g.,Engleetal.,2020;BoltonandKacperczyk,2021a,2021b;HuynhandXia,2021;Ilhan,Sautner,andVilkov,2021;Ramellietal.,2021b;Hsu,Li,andTsou,48M.Ceccarellietal.Downloadedfrom/rof/2022).However,howtheriskpropertiesofindividualgreensecuritiestranslatetotheportfoliolevelisstilllargelyunexploredand,asweshow,notobvious.Thetrade-offattheportfoliolevelthatwehighlightinthisDownloadedfrom/rof/Second,wecomplementtheliteratureonwhetherandwhyinvestorsprefersociallyre-sponsibleinvestmentproducts(e.g.,Bollen,2007;Renneboog,terHorst,andZhang,2011;RiedlandSmeets,2017;Bassenetal.,2019;HartzmarkandSussman,2019;Barber,Morse,andYasuda,2021;Bauer,Ruof,andSmeets,2021;Geczy,Stambaugh,andLevin,2021;AndersonandRobinson,2022).Theresponsestothequasi-naturalexperimentthatweanalyzehighlightboththecostsandbenefitsofsociallyresponsibleinvestmentproducts,crucialforunderstandingthecomplexityofinvestorbehavioronsustainabilityissues.Intermsofcosts,lowcarboninvestingasksinvestorstopayapriceintermsoflowersectoraldiversification,atleastintheshortterm.Genericsustainableratings/products,incontrast,areusuallybasedon“bestinclass”approachespreciselytoallowinvestorstonotgiveupanysectoraldiversification.Intermsofbenefits,theeventweanalyzeallowsafocusoninvestors’specificclimate-relatedpreferences.AsdocumentedbyHartzmarkandSussman(2019),theinvestorswestudyhadalreadyself-selectedintofundsbasedontheirgenericsustainabilitypreferences.Ourresultsindicatethatboththecostandbenefitsidesoflowcarboninvestingshapeinvestorresponses.Third,wecomplementtheliteratureonprofessionalmoneymanagerbehavior.Severalstudiesconsiderfundmanagerbehaviorasafunctionoftraditionalfinancialperformancemetrics,butinrecentyears,ESGfactors,andclimate-relatedconsiderationsinparticular,havegainedimportanceintheindustry.Forinstance,Krueger,Sautner,andStarks(2020)andIlhanetal.(2023)providesurveyevidenceontheimportanceofclimaterisksforinsti-tutionalinvestors.BoltonandKacperczyk(2021a)showthatinstitutionalinvestorsapplycarbon-relatedscreensandChoi,Gao,andJiang(2023)documentadecreaseininstitution-alinvestors’exposuretocarbon-intensivedomesticfirmsafter2015.Fundmanagerschangetheirholdingsaftershiftsinclimateriskperceptionduetonaturaldisasters(Alok,Kumar,andWermers,2020)orextremeheatevents(Alekseevetal.,2021).Gantchev,Giannetti,andLi(2022)studyfundmanagers’tradingbehaviorwithrespecttofirms’sus-tainability,focusingonthepricepressureimplicationsonindividualstocks.Ourarticlecontributestothisliteraturebystudyinghowfundmanagersactivelychangedtheirport-folioholdingsfollowingincreasedtransparencyonclimaterisksinthemutualfundindustry.LowCarbonMutualFunds49Downloadedfrom/rof/Downloadedfrom/rof/2.1EmpiricalSettingOnApril30,2018,MorningstarlaunchedonitsplatformthePortfolioCarbonRiskScore,ameasuredesignedtohelpitsclientsbetterassessaportfolio’sexposuretocarbonrisk(alsoknownasclimatetransitionrisk),thatis,theriskduetothetransitionfromafossilfuelrelianteconomytoalowercarboneconomy.3Onthesameday,MorningstarassigneditsLCDlabeltofundswithlowcarbonriskscoresandlowlevelsoffossilfuelexposure;thisheuristicisaimedathelpingclientseasilyidentifymutualfundswhoseportfoliosalignwiththetransitiontoalowcarboneconomy.4Figure1showstheportfoliocarbonriskscoreandtheLCDlabel,asseenonMorningstar’sfundreport.DetailsonthemethodologyunderlyingthesemetricsareinMorningstar(2018a,2018b).Theportfoliocarbonmetricsarebasedonfirm-levelcarbonriskscoresfromtheESGdataproviderSustainalytics;thesescoreswerealsodisclosedforthefirsttimeattheendofApril2018.5Thesimultaneousreleaseoffirm-levelandfund-levelcarbonriskscoreswaspossiblebecauseMorningstarhascontrolledSustainalyticssince2017(initiallywitha40%stake,whichincreasedto100%in2020).Accordingtothetwodataproviders,thefirm-levelcarbonriskscorequantifiesacompany’sexposureto,andmanagementof,materialclimatetransitionrisk.Itattemptstocapturethedegreetowhichafirm’seconomicvalueisatriskinthetransitiontoalowcarboneconomy(Morningstar,2018b).TableA1intheSupplementaryAppendixprovidesthesummarystatisticsoffirm-levelcarbonriskscoresineachGlobalIndustryClassificationStandard(GICS)sector.Firmsinhigh-emittingsectors(e.g.,energy,materials,andutilities)havethehighestmeancarbonriskscores,butthereissubstantialvariabilityinthismeasurewithinallsectors.ToreceivetheLCDlabel,afundhastocomplywithtwocriteria:(i)a12-monthaveragePortfolioCarbonRiskScorebelow10(outof100)and(ii)a12-monthaverageFFIratingbelow7%.AsofApril2018,havingaPortfolioCarbonRiskScorebelow10impliesbeingamongthe29%best-performingfundsonthisdimension.A12-monthportfolioFFIratingbelow7%representsa33%under-weightingoffossilfuel-relatedcompanies,relativetotheglobalequityuniverse.6ThereleaseofMorningstar’scarbonmetricsthusrepresentedadoubleshocktoinvest-ors:ashocktotheavailabilityofcarbon-relatedinformationthroughthefirm-levelandfund-levelcarbonriskscoresandashocktoitssaliencethroughtheLCDlabel.Thearrival50M.Ceccarellietal.DownloadedfromDownloadedfrom/rof/Figure1.Morningstardirectsnapshot.ofthesenewdataispotentiallyrelevantbothtofundmanagersandtotheirclients.7Morningstarrepresentativeshaveconfirmedtousthattheydidnotcommunicatethere-leaseofthesemetricstoeitherfundmanagersorclientsinadvanceoftheirpublicationonApril30,2018.Asseenfurtherbelow,ouranalysesofpre-publicationtrendsofinvestorandfundmanagerbehaviorareindeedconsistentwiththereleaseofthenewdatanotbeinganticipated.Webaseouranalysesontwomaindatasets,coveringtheperiodfromApril2017(1yearbeforeourmaineventofinterest)toSeptember2019:Fund-levelmonth-endinformation(fromMorningstarDirect)andindividualhistoricalportfolioholdings(fromMorningstarOnDemand).Wecomplementthesetwodatasetswithfirm-levelcharacteristicsfromCompustatCapitalIQandSustainalytics.Inwhatfollows,webrieflydescribeourdata.2.2.a.Fund-levelcharacteristicsFromMorningstarDirect,weobtainsurvivorship-bias-freedata(allinUSD)forallactiveopen-endmutualfundsdomiciledinEuropeandtheUSA.Toworkwitharelativelyhomo-geneoussample,wedropfundsclassifiedbyMorningstaraspurefixedincome,sector-specific,orinvestingexclusivelyoutsidetheUSAandEurope.Weareleftwithtwentycategoriesofequityandbalancedfunds.8LowCarbonMutualFunds51Downloadedfrom/rof/Whilemutualfundsissueseveralshareclassestotarget-specificinvestorgroupsorgeog-raphies,theunderlyingportfolioisthesameregardlessofclass.Consequently,weconductourmainanalysesatthefundlevel.Inaggregatingdatafromtheshareclasstothefundlevel,wecomputefunds’returnsandvolatilityasvalue-weightedaveragevaluesacrossdif-ferentshareclasses.Fundassets(inUSD)arethesumofafund’sAUMinallitsshareclasses.Werequirefundstohaveatleast1millionUSDDownloadedfrom/rof/FollowingSirriandTufano(1998),wecomputeflowsasthemonthlygrowthofAUM,netofreinvestedreturns.Wewinsorizeflowsatthe1stand99thpercentiles.FollowingHartzmarkandSussman(2019),wealsocomputeameasureofnormalizedflows:First,wesplitthesampleintodecilesoffundsize;second,werankfundsaccordingtonetflowswith-ineachsizedecileandcomputepercentilesofthenetflowrankings.Thesepercentilescor-respondtothenormalizedflowvariable.Returnisthetotalmonthlyreturn(inpercentagepoints),asreportedbyMorningstar.Weestimatethereturnvolatilityasthestandarddeviationofreturnsoverthepast12months.Wealsocollectotherinformationabouteachfund,includingitsage,itsMorningstarcategory,itsfinancialperformancerating(theMorningstarStars,ona1–5scale,with5indicatingatopfinancialperformer),anditsgenericsustainabilityrating(theMorningstarGlobes,ona1–5scale,with5indicatingatopsustainabilityperformer).Toaccountfortheimpactonflowsofchangesinafund’sfinancialperformancerating(DelGuercioandTkac,2008),wedefinethevariableΔStarstoindicateanupgrade(1)oradowngrade(–1)inthefund’sStarsratingfromthepreviousmonth.Similarly,toaccountfortheimpactonflowsofchangesinafund’sgenericsustainabilityrating(Ammannetal.,2018;HartzmarkandSussman,2019),wedefinethevariableΔGlobestoindicateanup-grade(1)oradowngrade(–1)inthefund’sGlobesratingfromthepreviousmonth.WeclassifyobservationswithmissingStarsorGlobesasnochange.PanelAofTableIshowssummarystatisticsforfund-monthobservations,fromApril2017toSeptember2019,forwhichinformationonflowsisavailable.PanelBprovidesasnapshotofthestatisticsasoftheendofApril2018.Thesamplecoverssome13,600funds,ofwhich17–18%obtainedMorningstar’sLCDeco-label.PanelAinTableA2intheSupplementaryAppendixshowsthegeographicaldistribu-tionofoursampleasofApril2018.Around9,000fundsaredomiciledinEuropeand4,000intheUSA,ofwhich18%receivedtheinitialLCD.PanelsBandCinthesametableshowtheshareoflowcarbonfundsfordifferentvaluesofMorningstar’sgenericsustain-abilityratings(Globes)andoverallfinancialperformanceratings(Stars).HighglobesandhighstarsfundsaremorelikelytoreceivetheLCD.However,evenamongfundswithoneortwoglobes,oroneortwoStars,asignificantfractionobtainedthelowcarboneco-label.TableA3intheSupplementaryAppendixexploresthecorrelationsofthenewdatawithpreviouslyavailablefirm-levelenvironmentalscores.ItshowsthatthePortfolioCarbonRiskScoreonlymildlycorrelateswithmetricsinvestorsmayhaveself-computed,basedonexistinginformation(wecalculatedthesemeasuresbasedonportfolioholdingsasofApril2018).Inparticular,thePortfolioCarbonRiskScorehasacorrelationof–0.27withaport-folio’sSustainalytics’environmentalscore,–0.08withaportfolio’sRefinitiv’senvironmen-talscore,and–0.19withaportfolio’sMSCI–KLD’senvironmentalscore.Overall,thelowcorrelationofthePortfolioCarbonRiskScorewithpriorenvironmentalmetricsconfirmstherelevanceoftheApril2018informationshocks.52M.Ceccarellietal.Downloadedfrom/rof/Downloadedfrom/rof/DescriptivestatisticsofactivemutualfundsdomiciledinEuropeandtheUSAforwhichinfor-mationonMorningstar’sLCDand?owsisavailable.PanelAcoversallfund-monthobserva-tionsfromApril2017toSeptember2019,whilePanelBisasnapshotfromtheendofApril2018.PanelCcoversallfund-?rm-monthobservationsfromApril2017toSeptember2019.LCDisanindicatorequalto1forfundsthatobtainedtheLCDlabelattheendofApril2018.CRandFFIareMorningstar’sportfolioCRandFFIscores.Flows(inpercentagepoints)isthemonthlygrowthofassets,netofreinvestedreturns.Normalized?owsiscomputedfollowingHartzmarkandSussman(2019).Returnisthemonthlynetreturn.LogassetsisthelogofAUM,inUSD.Volatilityisthestandarddeviationofreturnsintheprevious12months.Ageisthenumberofyearssincetheinceptionoftheoldestshareclass.GlobesistheMorningstarsustainabilityrat-ing,ona1–5scale.StarsistheMorningstaroverall?nancialperformancerating,ona1–5scale.ΔGlobesandΔStarsindicateifafundreceivedadowngrade(–1)oranupgrade(1)intheMorningstarGlobesratingorStarsrating,respectively.Positionchange(inbasispoints)isthechangeinthenumberofsharesheldbyfundfinstockifrommontht–1tomontht,valuedatthepriceofmontht–1,dividedbyAUMinmontht–1.LowCR(?rm),MediumCR(?rm),andHighCR(?rm)areindicatorsequalto1for?rmswithCRscoresbetween0and9.99(low),be-tween10and29.99(medium),orabove29.99(high),and0otherwise.FFI(?rm)isanindicatorequalto1for?rmsderivingasigni?cantshareoftheirrevenuesfromfossilfuel-relatedactiv-ities.Churnrateisameasureofhowfrequentlyfundmanagersrotatetheirpositionsonallthestocksinaportfolio.PositionweightisthepercentageofAUMinvestedina?rm.PanelA:Fund-levelvariables,fromApril2017toSeptember2019NMinp25Meanp50p75MaxSDLCD379,0860.000.000.180.000.001.000.39CR237,3030.238.3910.1510.0611.4645.603.44FFI334,9010.003.067.016.209.5584.225.85Flows379,086-19.53-1.60-0.03-0.291.2132.824.74Normalized?ows379,0861.0027.0049.3849.0072.00100.0027.24Return379,086-90.60-1.090.410.612.2428.493.31Logassets379,08613.8216.8218.4018.3519.8626.022.06Volatility379,0760.041.742.782.513.5726.531.46Age379,0861.006.2614.0112.6518.89119.3210.12Globes275,7781.002.003.053.004.005.001.13Stars237,3151.002.003.153.004.005.001.06ΔGlobes379,086-1.000.000.000.000.001.000.32ΔStars379,086-1.000.00-0.000.000.001.000.30PanelB:Fund-levelvariables,snapshotattheendofApril2018Nminp25Meanp50p75MaxSDLCD13,0560.000.000.180.000.001.000.39CR8,9970.239.0310.7010.6211.9445.583.47FFI13,0130.002.956.705.929.0870.995.53Flows13,056-19.53-2.24-0.88-1.60-0.0432.824.88Normalized?ows13,0561.0026.0048.8047.0071.00100.0027.44Return13,056-9.790.472.041.823.4513.912.11Logassets13,05613.8616.8418.4218.3619.8925.932.05(continued)LowCarbonMutualFunds53Downloadedfrom/rof/Downloadedfrom/rof/PanelB:Fund-levelvariables,snapshotattheendofApril2018Nminp25Meanp50p75MaxSDVolatility13,0560.121.732.242.302.728.650.80Age13,0561.005.8013.6312.2518.52118.2410.14Globes9,3581.002.003.023.004.005.001.14Stars9,8871.002.003.163.004.005.001.05PanelC:PortfolioholdingsNminp25Meanp50p75MaxSDPositionchange12,786,149-82.510.00-0.070.000.0083.7213.23Positionweight12,398,4360.000.060.780.331.1146.201.10CR(?rm)12,786,149-0.001.3511.059.0615.6481.0911.37HighCR(?rm)12,786,1490.000.000.060.000.001.000.24MediumCR(?rm)12,786,1490.000.000.400.001.001.000.49LowCR(?rm)12,786,1490.000.000.541.001.001.000.50FFI(?rm)12,786,1490.000.000.100.000.001.000.30Return(?rm)12,500,884-0.37-0.040.010.010.051.000.08Volatility(?rm)9,737,9992.65