BISWorkingPapersNo1260

Supplychaintransmissionofclimate-relatedphysicalrisks

byDouglasK.G.Araujo,FernandoLinardiandLuisVissotto

MonetaryandEconomicDepartment

April2025

JELclassification:E32,L14,Q54,R15

Keywords:climate-relatedphysicalrisks,precipitationanomalies,supplychains,GDPgrowth

BISWorkingPapersarewrittenbymembersoftheMonetaryandEconomicDepartmentoftheBankforInternationalSettlements,andfromtimetotimebyothereconomists,andarepublishedbytheBank.Thepapersareonsubjectsoftopicalinterestandaretechnicalincharacter.TheviewsexpressedinthemarethoseoftheirauthorsandnotnecessarilytheviewsoftheBIS.

ThispublicationisavailableontheBISwebsite

()

.

?BankforInternationalSettlements2025.Allrightsreserved.Briefexcerptsmaybereproducedortranslatedprovidedthesourceisstated.

ISSN1020-0959(print)

ISSN1682-7678(online)

1

Supplychaintransmissionofclimate-relatedphysicalrisks?

DouglasK.G.Araujo1,FernandoLinardi2,LuisVissotto3

1BankforInternationalSettlements,douglas.araujo@

2BancoCentraldoBrasil,fernando.linardi@.br

3BancoCentraldoBrasil,luis.vissotto@.br

Abstract

HowdoclimateanomaliesaffectGDPgrowth,andhowdotradeconnectionshelpunder-standthisimpact?WeaddressthesequestionsexploringlocalfluctuationsintemperatureandprecipitationcoupledwithdataonsupplychainlinkagesbetweenmunicipalitiesinBrazil.GDPgrowthfallswithlocalanomalousdryspellsandtoalowerextent,alsowetspells.Muchofthiseffectisattributabletomoderatelevelsofclimateanomaly.Thisimpactissu?icientlymaterialtotransmitacrosssupplychainconnectionstoothermunicipalities.Focusingonpairsofdistantmunicipalitiestoavoidcommonclimateshocks,municipalitieswhosecustomerfirmssufferdryspellshavebetween1and2percentagepoints(p.p.)lowerGDPgrowth.Thissupplychainshockalsoleadstolowerimportgrowthandweakerlabourmarketmetrics,suggestinganoveralllowerlevelofeconomicactivity.Wealsoexaminethemajoreconomicsectorsseparatelyandfindthatagriculturalactivityismoresensitivetosupplychaintransmissionofphysicalshocks,includingmoderateones,thanmanufacturing(whichrespondsmainlytointensesup-plychainshocks)orservices.Thissuggeststhatthelocaleconomicmixcanbeapotentiallyimportantdriverofeffectheterogeneity.Usingacounterfactualanalysis,weestimatealsothatsupplychainspilloversfromclimatechangevariessubstantiallyovertheyearsbutcanleadto1p.p.lowergrowthonaverage.Keywords:Climate-relatedphysicalrisks.Precipitation

anomalies.Supplychains.GDPgrowth.JELCodes:E32,L14,Q54,R15.

1Introduction

Themacroeconomicimpactsofphysicalshocksandtheiramplificationmechanismscontinuetobeanimportantknowledgegap(Battiston,Dafermos,andMonasterolo(2021)).Thisworkdoc-umentstheimportanceofsupplychainstopaintamorecompletepictureofphysicalrisksfromclimatechange.OuranalysesconcentrateonthetransmissionofphysicalshockacrossBrazilian

*ThisworkrepresentstheviewsoftheauthorsanddoesnotnecessarilyrepresentthoseoftheBankforInter-nationalSettlementsortheBancoCentraldoBrasil.TheauthorsthankRodrigoBarradasandAlejandroParadafortheresearchassistance,TammaCarletonforthemanyinsightfulcommentsandsuggestions,ThorstenBeck(dis-cussant),MarshallBurke,JuliánCaballero,BenCohen,JonFrost,DenizIgan,EnisseKharroubi,GabrielaNodari,LuizPereira,José-LuisPeydró,KevinTracolandGoetzvanPeter,andparticipantsinthe14thBISCCAResearchNetworkon‘Macro-financialimplicationsofclimatechanceandenvironmentaldegradation’,theIIInternationalConferenceontheClimateMacro-FinanceInterface:‘NewEnvironmentalChallengesforFiscal,Monetary,andMacroprudentialPolicy’(2CMFI)andatseminarsintheBancoCentraldoBrasilandtheBancaD’Italiaformanyhelpfulcommentsthatmadethepaperbetter.

2

municipalitiesdirectlyconnectedtoeachotherbytrade.Usingauniqueadministrativedataset,weexploretheeconomiceffectsoflocalclimateanomaliesofdifferentintensitiesinaggregateandacrosssectors.Next,westudyhowtheseeconomicimpactsspreadthroughsupplychainsbeyondanylocaleffect.Importantly,ourestimatesbenefitfromBrazil’scontinentalsizetodisentangletheeffectsofsimultaneousshocksoccurringlocallyandinmunicipalitieswheresupplierorcustomerfirmsarelocated.

Ourdataandmethodologyaffordmultipledimensionsofanalyses.Forexample,theclimateanomaliescanrelatetoeitherwetordryspells,andhavedifferentlevelsofintensity.Intermsoflocation,theanomaliesaremeasuredforeachmunicipality,butinouranalysestheycanalsoenterregressionsassuppliersand/orcustomerstoeachother.Further,eachmunicipalitypairisidentifiedasbeinga(geographically)“distant”linkornot,whereitisassumedthatthedistantonesdonotsharethesamephysicalanomaliesbecauseofthedistance.Finally,theeconomicoutcomeofinterestisGDPgrowthateachmunicipality,asawholeordividedintoagriculture,manufacturingorservicessectors.

Usingsimilarmeasurementsofclimateanomaliesasinthescientificliterature,weshowthatadverseclimateanomaliesofbothmoderateandintensemagnitudeinfluencelocaleconomicout-comes.Duetotheirsheerfrequencyandthemagnitudeofthecoe?icients,moderateshocksareresponsibleforthebulkoftheeffects.Theselocalshocksinfactaresorelevantthattheyspilloverviasupplychainlinkagestoimpactotherregions’economicgrowth.Whencustomerfirmsareinmunicipalitiessufferingdrought,thissupplychainlinkleadstoadepressedlocallabourmarketandforeigntradeimportactivity,consistentwithanincreaseinslack.Shocksinbothcustomersandsuppliersalsoleadmunicipalitiestodiversifymorethelocationoftheirsupplychainconnections.Butultimatelytheeffectoflocalandsupplychain-transmittedclimateshocksisheterogeneous,dependingonthedifferentsectoralmixofeconomicactivityofeachmunicipality:inparticular,agricultureisthemostsensitiveeconomicsectortobothlocalandmoderatesupplychain-transmittedshocks;manufacturingontheotherhandismoresensitivetointensesuppliershocksandservicesGDPgrowthis,onaverage,insulatedfromtheseshocks.Thismightbere-latedtohowclimatephysicalshocksnegativelyimpactagriculturalyield,whichinturnalsoservesasinputtoothercropsandtoanimalhusbandry-bothmacroeconomicallyrelevantactivitiesinBrazil.

1

Thisworkexploresinmoredetaildifferenttypesandintensitylevelsofprecipitationanomaliesacrosssectorsandspace.ByfocusingonBraziliandatainsteadofcross-countrydata,ourworkabstractsfromimportantvariationsineconomicstructureandpolicyandthusalargershareofoutcomevariationcomesfromexposuretodifferentlocalorsupplychainclimateshocks.Anotheradvantageisthatitallowsthepossibilitytouseawealthofconsistentdatasetsatthemunicipalitylevel.Brazilisagoodlaboratoryforstudyingtheeconomicimpactofclimateshocks.Itsconti-nentalarea,geographicaldiversity,heterogeneousweatherpatternsaswellasitsvastbiodiversity,resultinvariationsintheexposuretoclimate-relatedevents(P?rtneretal.(2022)).Atthesametime,Brazilisamiddle-incomecountrywithmunicipalitiesindifferentlevelsofsocio-economicdevelopment,allofthemsubjectedtothesamefederallegalandmonetaryframeworkthatallowsforcomparabilitybetweenitsregions.Importantly,asdocumentedinFigure

1

,mostcatastrophes

1Forexample,Ahvoetal.(2023)showhowinterconnectedagriculturalsupplychainscantransmitshocksfromlowercropyieldswithinthesector.

3

inBrazilarerelatedtoclimate,ratherthanfromgeologicaldisasters.

Ourfindingscontributethreemaininsightstotheliterature.First,wedocumenteconomicallyrelevantimpactsofphysicalshocksalsoforclimateanomaliesthatarenotextremeordisastrous.Thiscanraiseawarenessabouttheimportanceofthoseshockstoclimate-relatedphysicalrisks,asopposedtothefocusonextremeeventsonly.Inaddition,moderateshockscanincreasetheex-ternallyvalidityoffindingscomparedtothoseobtainedfromlargedisasters,sincethelattershedlightonimportanteconomicquestionsthatinsomecasesmaybeapplicableonlyinsimilarlyex-tremesituations.Thesecondcontributionisthestrongevidenceonhowsupplychainconnectionstransmitclimateeconomicshocks.Sincetheseeffectsaremeasuredaftercontrollingfortimeandmunicipality,theycanbeinterpretedasoccurringaboveandbeyondeffectsofclimateanomaliesonpricesorotherincentivesthatarenotrelatedtothetradelinkages.Ourthirdcontributionrelatestouncoveringimportantaspectsthatdriveheterogeneityinresponsestolocalandremoteclimateshocks,inthiscasethroughthemixoflocaleconomicactivitybetweentheagricultural,manufacturingandservicessectors.

Supplychainarebutoneformoftransmissionoflocalshockstootherlocalities.Otherlong-recognisedchannelsincludechangesinprices(Hayek(1945),Flori,Pammolli,andSpelta(2021)),commonbankinglinks(PeekandRosengren(1997),FenderandMcGuire(2010),CortésandStrahan(2017),Ivanov,Macchiavelli,andSantos(2022)),tourismflows(AnastasiaArabadzhyanandLeón(2021)),andothers.Andsupplychainsofcoursedonottransmitonlyclimate-relatedshocks:alongliteraturedocumentshowtradelinksspreadothernaturaldisastersandevenadhoctradeevents(Lafrogne-Joussier,Martin,andMejean(2023)).Still,ourfindingsthatsupplychainconnectionstransmiteconomicshocksfromdifferenttypesofclimateanomaliesandinaheterogenousway,evenwhenshocksaremoderate,improvesourunderstandingofthecomplexeconomicconsequencesofclimaterisks.

Therestofthepaperisorganizedasfollows.Section

2

describesthedataandofferssomeback-groundinformationonchangesinaveragetemperatureandprecipitationexperiencedbyBrazilianmunicipalitiesinrecentdecadesandpresentsthenetworkoflinkagesamongfirmslocatedindifferentmunicipalities.Section

3

analysethelocalimpactofclimateanomalies,andSection

4

elaboratesonthoseresultstodocumentthetransmissionofshocksthroughtradelinkages.Sec-tion

5

breaksdowntheeconomicimpactsintodifferentbroadsectors.AcounterfactualexerciseinSection

6

demonstratestherelevanceofthesefindings.ThenSection

7

connectsthedotsandconcludes.

1.1Literature

Alongliteratureexplorestheeconomiceffectsofphysicalrisks;withsomeinsightslearnedalsofromthebroadersetofnaturaldisastersthatincludesgeologicalevents.Somestudiesfocusontheeffectoftemperaturelevelsordeviationsfromlong-termaveragesoneconomicgrowth(Dell,Jones,andOlken(2012);Burke,Hsiang,andMiguel(2015);HenselerandSchumacher(2019);MaximilianKotzetal.(2021);KalkuhlandWenz(2020)).OtherstudiesexaminetheeffectofvariabilityoftemperatureandprecipitationoneconomicgrowthorGDPpercapita(Felixetal.(2018);Damania,Desbureaux,andZaveri(2020);LettaandTol(2019);Kahnetal.(2021);M.Kotz,Levermann,andWenz(2022)).S.Acevedoetal.(2020)documentsarelationshipbetween

4

temperatureshocksandreducedinvestment,depressedlabourproductivity,poorerhumanhealth,andloweragriculturalandindustrialoutput.Deryugina(2017)documentsthefiscaleffectsofhurricanes,andtheinsuranceprovidedbysocialsafetynets.Hornbeck(2012)showstheshort-andlong-runadjustmentsincountiesaffectedbythe1930sAmericanDustBowlenvironmentalcatastrophe.

Inastudyofthe1950sdroughtintheUS,RajanandRamcharan(2023)showthatexantecreditavailabilityhadasizeableimpactonthelong-runeffectsofthedroughtthroughchangingtheabilityofaffectedtownstoadapttothecircumstancesthroughinvestmentandinnovation.Kim,Matthes,andPhan(2022)useasmoothtransitionvectorautoregression(pioneeredbyAuerbachandGorodnichenko(2012))appliedtoUSdata,findingsubstantialeffectsofclimate-relatedphys-icalrisksonindustrialproduction,consumption,unemploymentandinflation.Parker(2018)alsofindsthatclimatedisastersimpactinflation,withamorepronouncedeffectindevelopingcountries.TheworkbyGonzalez,Ornelas,andSilva(2023)illustratethemulti-sectoralimpactofphysicalrisks.Theyshowthatthe2015Marianaenvironmentaldisastercausedaffectedfarmstoreceivebroadlyhalflesspayments(aproxyforrevenue)fromnon-affectedcustomers,andcreditcardandconsumerfinancebalancestofallby8%.

Anotherstrandintheliteratureexamineshoweconomiclossespropagatebeyondlocalshocks.ThisisconsistentwiththeargumentmadebyAcemogluetal.(2012)thatsectoralshocksandtheirsecond-ordereffectcanexplainaggregateoutcomes,andbyElliott,Golub,andLeduc(2022)aboutthepossibilitythatsupplychainscantransmitevenrelativesmallshocks.Acemoglu,Akcigit,andKerr(2016)showhowthepropagationofmacroeconomicshocksthroughinput-outputandgeographicnetworkscanbeapowerfuldriverofmacroeconomicfluctuations.WenzandWillner(2022)overviewapproachestoassessextremeweathereventsalongglobalsupplychains.GiroudandMueller(2019)documenthowlocalshockspropagateacrossUSregionsthroughfirms’internalnetworksofestablishments,whileCravinoandLevchenko(2017)investigatehowmultinationalfirmscontributetothetransmissionofshocksacrosscountries.Additionally,studieshaveexaminedthepropagationofnaturaldisasterssuchasthe2011Japanearthquake(Carvalhoetal.(2020);Boehm,Flaaen,andPandalai-Nayar(2019))andHurricaneSandyintheUS(Kashiwagi,Todo,andMatous(2021)).

Usingfirm-leveldata,thepapersshowhowshockspropagatethroughsupplychainstoareasnotdirectlyhitbydisasters.BarrotandSauvagnat(2016)usenaturaldisastersintheUStodocumentlargeeffectsoncustomerfirmswhentheirsuppliersaredisrupted,inawaythatisconsistentwithspecificityoftheirinputtotheproductionchain.Dasetal.(2022)showthatsupplyanddemandshockspropagateupstreamanddownstreamintheproductionanddistributionnetwork,bothdomesticallyandabroad.Feng,Li,andWang(2023)showthatinternationaltradeconnectionsexplaincross-borderspilloversofclimateshocks.Zappalà(2023)useaglobalsectoralproductiondatatoinvestigatethepropagationofweathershockstoagricultureinamulti-region,multi-sectorproductionnetworkmodel.Hispaperexploreslinkagesacrosssectorsandspace,showingthattheselinkagescontributesignificantlytolossestimatescomparedtoestimatesfromaggregateprojectionsofGDPonclimateshocks.

5

2Data

Wecombinemultiplepublicly-availableclimateandeconomicdataatthemunicipallevelwithconfidentialBancoCentraldoBrasil(BCB)paymentsdata.Theprimaryfeaturesofthesedataaretherichnessofdataatandisaggregatedgeographicallevelinformation,andtheuseofmultiplecomplementarymeasuresofweatheranomaliesthatdonotdependonindividuallyidentifiedlargescaledisasters.Theresearchperiodspansfrom2000(localshocks)or2012(withsupplychaindata)toend-2019,asu?icientlylongperiodthatencompassesdifferenteconomicpoliciesandbusinesscycledynamicsbutavoidstheturbulentCovid-19pandemicperiod.

2.1Climatedata

Similartothescientificliteratureonclimateanomalies,ourmeasureofclimateanomaliescomesfromdivergencesinprecipitationinagivenlocationfromitshistoricaldistribution.Inparticular,weusemunicipality-levelprecipitationandtemperaturereadingssince1961tocalculatetheStan-dardizedPrecipitationEvapotranspirationIndex(SPEI)(Vicente-Serrano,Beguer??a,andLópez-Moreno2010).Thisindicatormeasuresthedivergencebetweenactualandexpectedtemperature-adjustedprecipitationinagivenlocationforagivenwindowoftime.Forexample,theone-yearadjustedprecipitationvaluesofamunicipalityarecomparedtoitshistoricalaverages.TheSPEIvaluesaremeasuredinstandarddeviations(s.d.)ofthehistoricalvaluesforeachlocation,andarethuscomparableacrosslocationsandtime.Positivereadingsindicatewetterclimatesthanthehistoricalaverageforalocation,andconverselynegativereadingspointtooccurrencesofdroughts.Thisindicatoristhebasisforthephysicalshocksinthisstudyduetotheprevalenceofthistypeofprecipitation-relatedclimateanomalyacrossBrazil(Figure

1

).

2

TheSPEImeasuredateachtimewindowreflectsdifferentphysicalimplicationsofclimateanoma-lies.Forexample,one-yearSPEIismoreinformativeabouttheeffectofdroughtonsoilhumidityandrivervolumes,whilefive-yearSPEIlevelsindicatemorestructuralimplications,suchasonundergroundwaterreservelevels.Ourpreferredmeasureofphysicalshockistheone-yearSPEI.Ayear-longdeviationcanbesu?icientlymaterialtooverwhelmshort-termresiliencemeasures(suchasinputstocksinmanufacturingfirms)whilenotlong-termenoughtoreshapehowsupplychainsarestructured.Forexample,abnormallyhighone-yearSPEIscouldindicatecasesofex-cessiverainthatleadstourbanflooding,causingbothlossoflifeandwealth(oftenalsoaffectingpoorerhouseholds’goods)aswellaslogisticobstaclestocommerce.

3

Inaddition,aone-yearshockhorizonmapswellwiththeannualfrequencyofthemunicipalGDPdata.

4

2Otherwidely-usedindicatorsofprecipitationanomalyincludetheStandardizedPrecipitationIndex(SPI),closelyrelatedtotheSPEI,andthePalmerDroughtSeverityIndex(PDSI).TheSPIfollowsasimilarcalculationastheSPEIbutwithoutanyadjustmentfortemperature,whichbiastheanomalyestimationsespeciallyformorearidregions.ThePSDI,ontheotherhand,isafairlycomplexindicatorthatadjustsfortemperaturebutalsosoilcharacteristics.Forthisreason,itismoredirectlyassociatedwiththesoil-leveldamageofdroughts.TheSPEIcanbeinterpretedasabalancebetweenthesimplicityofSPIwiththeusefulnessofthenuancesprovidedbythePDSI(Liuetal.(2024)).

3OneofvariousexamplesfromlocalnewsisaboutthedamagestotruckshaulinggoodstoandfromaS?oPaulowarehouse,at(inPortuguese):

/agronegocios/noticia/2020/02/10/chuva-em-sp-leva

-ceagesp-a-paralisar-atividades.ghtml.

4Throughoutthepaper,themainresultsareestimatedwithSPEI.ButquantitativeanalyseswiththerelatedindicatorStandardizedPrecipitationIndex(SPI)(McKeeetal.(1993),EdwardsandMcKee(1997),Guttman(1999))

6

Figure1:DeclaredenvironmentaldisastersinBrazil

Numberofdeclareddisasters

600

400

200

0

Jan2005Jan2010Jan2015Jan2020

Typeofdisaster

rain

cyclones

exhaustiongeologic

wildfiresdrought

extremetemperaturestorm

ThehistoricalweatherdatafortheSPEIaretakenfromthewidely-usedClimaticResearchUnitgriddedTimeSeries(CRUTS)monthlydataset,version4(Harrisetal.(2020)).Thisisaclimatedatasetonagridwitharesolutionof0.5°x0.5°inlatitudeandlongitudethatconsistsofweathervariablessuchasprecipitation,temperature,diurnaltemperaturerange,cloudcover,windspeed,andothers.WeusemonthlyprecipitationandaveragedailymeantemperaturetocalculatetheSPEIatthemunicipalitylevelusingtheSPEIRpackage(Beguer??aetal.2014),withthegammadistributioncalibratedwitharectangularkernel,ieallpastobservationsintheone-yearhorizonhavethesameweight.

AninterestingempiricalfeaturethatBraziloffersforthestudyofphysicalriskscomesfromthewidefluctuationsintemperatureandprecipitationconditionsobservedduringtheresearchperiod.ThisisbestseeninFigure

2

,whichportraysthe12-monthrollingaveragesofthenationalunweighedmeanofmunicipaltemperatureandprecipitationvalues.Temperaturelevelshaveremainedbroadlybelow24°Cupuntil2015,buthavebeenconsistentlyabovethatlevelsince.Andtheprecipitationlevelhasswungwidelytoapeakin2010,thengraduallyfallingtoreachlevelsthatwerenotseeninthetwodecadesbefore.

Ourdefinitionof“physicalshock”isstatistical:thepositiveandnegativeSPEIvaluesofmore

yieldsimilarresults,soarenotreported.Conceptually,boththeSPEIandSPIaredifferentwaysofcalculatingthedeviationsofprecipitationfromthelong-termaverageforthatsamelocationandtimeperiodintheyear,withthesedifferencesstandardisedaccordingtothesamedistribution.ThemaindifferencebetweentheSPEIandSPIisthattheformerrelatesonlytoprecipitation,whilethelatteralsocorrectsforchangesinthetemperatureofagivenlocationtobetterreflectthephenomenonofevapotranspiration,whichisimportantformanyphysicalphenomenasuchaswaterabsorptionbythesoiloritsusebyplants.Thesesetsofresultsarebroadlythesame.

7

Figure2:Average12-monthtemperatureandprecipitationlevelsinBrazil

Temperature(°C)

140

130

120

110

100

25.0

24.5

24.0

23.5

23.0

Jan2000Jan2005Jan2010Jan2015Jan2020

Precipitation(mm/month)

temperature

precipitation

thanones.d.(lessthanminusone)identifythe“anomalous”episodes.PositiveSPEIanomaliesarealsoreferredtoas“wetspells”throughout;conversely,negativeSPEIanomaliesare“dryspells”.Thisdefinitionbenefitsfromanintuitiveunderstandingthatanomaliesareeventstending towardsthetailsofdistributions,andallowsanaturalquantitativecomparisonoftheshocks.Toexplorevariationinintensityoftheanomalies,weconsiderallanomaliesofmorethantwos.d.(or lessthannegativetwos.d.)tobe“intense”,whiletheonesbetweenoneandtwos.d.arecalled“moderate”;thisnotationispresentedmoreformallyintheempiricalsectionsbelow.

Figure

3

showstheevolutionoftheclimateanomaliesidentifiedthisway.Notethevirtualabsenceofintensepositiveshocks.Forthisreason,theempiricalspecificationsbelowdonotincludethisparticulartypeofanomalies.

2.2Supplychaindata

ThesupplychainnetworkisbasedonconfidentialBCBdataonelectronicfundstransfersbetweenfirms’bankaccountsatdifferentbanks.

5

Thispaymentmodalityhasnoupperceilingandsettlesonthesameday,reasonsforwhichitiswidelyusedforinterfirmpayment.Correspondingly,thesepaymentsrepresent42%ofalltransactionvaluesaccordingtoBCBdata,

6

thelargestvalueshareinthepaymentssystemwithinourperiodofanalysis.Notably,thisshareremainedstableevenwiththeadventofthepopularinstantpaymentssystemPix(Duarteetal.2022).

7

This

5Transfersbetweenbankaccountsinthesamebankaresettledinternallyinthebank’ssystems(“booktransfer”)andthereforedonotgothroughtheBCBpaymentssystem.

6Availableat

.br/estatisticas/spbadendos/

.

7Availableat

.br/estabilidadefinanceira/estatisticaspix/

.

8

Figure3:Evolutionofclimateanomalies

Percentofmunicipalities

60

40

20

0

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

2014

2015

2016

2017

2018

2019

DryspellsWetspells

Ofwhich,intensedryspells

Ofwhich,intensewetspells

underscoresitsrelevanceasapaymentmethodforfirms.

Importantlyinourcase,theseelectronicfundtransfershelpdifferentiatetrade-relatedpayments:otherbusiness-to-businesspaymentssuchasexpenditureswithutilitiescompaniesareoftensettledviapaymentstubs,whichuseadifferentpaymentsrailandarenotcapturedinthisdata.Thesamefirm-to-firmpaymentsdatahavebeenusedinotherstudiesasaproxyforthesupplychainnetwork(egMartins,Schiozer,andLinardi(2023),Gonzalez,Ornelas,andSilva(2023))andalsobytheBCBinitsoversightactivities(BancoCentraldoBrasil(2015)).

Otherworksintheliteratureidentifytradelinkageswithinvoices(eg,P.Acevedoetal.(2023))orregulatoryfilingsbypublicly-tradedcompanies(eg,Qiu,Shin,andZhang(2023)).Usingpaymentsdatatoproxyforsupplychains,asdoneinthispaper,hasadvantagesanddisadvantages.Paymentsidentifyexactlyfirmsthatarecustomersandsupplierstoeachother,evenwhenthesetraderelationshipsareinformalandthusarenotdocumentedininvoices.Paymentsalsoexpandtheuniverseoffirmsforwhichthereissupplychaindatabeyondthetypicallylargefirmsthatareinsomecasesrequiredtodisclosepublicinformation.Anotheradvantageisthattheactualamountsare,bydefinition,observed.

8

However,thereareimportantlimitationsinhowmuchthepaymentsdataweusecanidentifysup-plychainrelationships.First,wecannotmeasurethewholesupplychainnetwork:asmentionedabove,thesepaymentsonlyconsiderelectronictransfersbetweenaccountsindifferentbanks,since

8OtherpapersthatusethesamedatasetincludeSilva,Amancio,andTabak(2022),Gonzalez,Ornelas,andSilva(2023),amongothers.Silva,Amancio,andTabak(2022)discussesthedatasetinmoredetail,includingtheevolutionofnetworkcentralitymeasuresoverasimilartimeperiodtothatofouranalysis.

9

intrabanktransfersaresettledinternallybythebankviabooktransfer.Cashorcreditcardpay-mentsarealsonotincluded,althoughtypicallythosewouldrelatetoverysmallfirmsormundanelow-ticketpurchasesinsteadofactualproductioninputs.TransactionssettledoutsideofBrazil,suchasforeigntrade,arealsonotconsidered.Thismeansthatsupplychainidentificationmightbesuboptimal,especiallyinsmallermunicipalitieswhereonlyoneortwobankshaveactivepres-ence(generallythelarge,gov