紐約聯(lián)儲(chǔ)-美國貨幣政策對外國公司的影響（英）-2022.11-52正式版

NO.1039NOVEMBER2022

TheImpactofU.S.MonetaryPolicyonForeignFirmsJuliandiGiovanni|JohnRogersTheImpactofU.S.MonetaryPolicyonForeignFirmsJuliandiGiovanniandJohnRogersFederalReserveBankofNewYorkStaffReports,no.1039November2022JELclassification:E52,F40AbstractThispaperusescross-countryfirm-leveldatatoexploretheimpactofU.S.monetarypolicyshocksonfirms’sales,investment,andemployment.WeestimateasizableimpactofU.S.monetarypolicyontheaverageforeignfirm,whilecontrollingforothermacroeconomicandfinancialvariablesliketheVIXandexchangeratefluctuationsthataccompanyU.S.monetarypolicychanges.Wethenquantifytheroleofinternationaltradeexposureandfinancialconstraintsintransmittingmonetarypolicyshockstofirms,allowingforabetteridentificationoftheimportanceofexternaldemandeffectsandtheinterestratechannel.Wefirstexploitcross-countrysector-leveldataonintermediateandfinalgoodstradetoshowthatgreaterglobalproductionlinkagesamplifytheimpactofU.S.monetarypolicyatthefirmlevel.Wethenshowthattheimpactvariesalongthefirm-leveldistributionofproxiesforfirms’financialconstraints(forexample,sizeandnetworth),withtheimpactbeingsignificantlyattenuatedforlessconstrainedfirms.Keywords:U.S.monetarypolicyspillovers,foreignfirms,internationalproductionlinkages,financialconstraints_________________Giovanni:FederalReserveBankofNewYork(email:julian.digiovanni@).Rogers:FudanUniversity(email:johnrogers@).TheauthorsthankWenchuanDong,NeelLahiri,andYijingRenforprovidingexcellentresearchassistance.Thispaperpresentspreliminaryfindingsandisbeingdistributedtoeconomistsandotherinterestedreaderssolelytostimulatediscussionandelicitcomments.Theviewsexpressedinthispaperarethoseoftheauthor(s)anddonotnecessarilyreflectthepositionoftheFederalReserveBankofNewYorkortheFederalReserveSystem.Anyerrorsoromissionsaretheresponsibilityoftheauthor(s).Toviewtheauthors’disclosurestatements,visit/research/staff_reports/sr1039.html.IntroductionTheimpactofU.S.monetarypolicyontherealeconomyisalong-studiedtopic,andonethatisofprimaryimportancetounderstandtodayastheFedandothercentralbankshaveenteredaglobaltighteningcycle.Thesepolicyactionsarenottakeninavacuum,andsomeeconomists,suchasObstfeld(2022)andWei(2022),havearguedthatthereisriskofcentralbanksdampeningaggregatedemandexcessively.Indeed,spilloversofU.S.monetarypolicymayimpactforeigneconomiesviaseveralchannelsindependentlyofdomesticpolicyactions.Thispapermergesfirm,sectoralandmacroeconomicdataforalargecross-sectionofcountriestoquantifyhowinternationaltradeexposureandthefinancialchannelofinterestratechangesaffecttransmissionofU.S.monetarypolicyshockstoforeignfirmactivity.Westudythesetwochannelsgiventhattherecentconfluenceofescalatingprotectionism,Covid-19,disruptedsupplychains,Brexit,OFACsanctions,corporatedelistings,andgeopoliticaltensionshasraisedquestionsaboutwhetherthedecades-longtrendtowardglobalizationintradeandfinancialmarkets,aswellastheriseof“megafirms”(Autor,Dorn,Katz,PattersonandvanReenen,2020),isreversing.Such“end-of-globalization”considerationsareimportantforglobalwelfaredependingonthedegreetowhich–andchannelsthroughwhich–shockssuchasmonetarypolicytighteningarefundamentallytransmitted.Focusingonthefirmlevelisparticularlysalientgiventheroleof“granular”firmsindrivingaggregatefluctuations(Gabaix,2011).Firmheterogeneityfurtherinteractswithexposuretotheworldeconomy,particularlyviainternationaltrade,toplayalargeroleinaggregateinter-nationalbusinesscycleco-movement(diGiovanni,LevchenkoandMejean,2014,2018,2022;WeiandXie,2020).WebeginbyestimatingtheeffectofU.S.monetarypolicyshocksonthechangeintheaverageforeignfirm’sinvestment-to-capitalshare,sales-to-capitalshare,andemploymentgrowthinagivencountry.Ourmethodologyutilizesapanelregressionmodel,whichallowsustocontrolfortime-varyingfirm-levelandmacroeconomicvariables,aswellasarichsetofnon-timevaryingfixedeffects.ThemainresultsimplythatthetighteningofU.S.monetarypolicyhasastatisticallysignificantcontractionaryeffectonthechangeofafirm’sinvestmentandsalesratios,whileemploymentgrowthalsofalls,butnotsufficientlytodetectastatisticallysignificanteffect.Resultsarealsoeconomicallymeaningful.Forexample,aonepercentagepointcontractioninU.S.monetarypolicytranslatestoafallintheinvestmentratioequivalenttosixty-fivepercentofthemedianchangeintheinvestmentratioacrossoverthesampleperiod.Wethenaskhowthisspillovereffectvariesalongmultiplecountrydimensions.Forexample,wedocumentsignificantdifferencesbetweenemergingmarketeconomies(EMEs)andadvancedeconomies(Kalemli-Ozcan,2019).WealsoexaminehowfinancialaccountandtradeopennessattheaggregatelevelsaffectthemagnitudeofU.S.monetarypolicytransmissiontoforeignfirms.1Wenextturntoamorein-depthanalysisoftheimpactofafirm’stradeexposure,bothtotheworldeconomyaswellaswithrespecttotheUnitedStatesalone.Todoso,weconstructfourexport-orientedmeasuresoftradeusingcross-countrysector-leveldataonintermediateandfinalgoodstradeaswellassectoraloutputsourcedfromtheWorldInput-OutputDatabase(WIOD)fromTimmer,Dietzenbacher,Los,StehreranddeVries(2015).Specifically,weconstructacountry-sector’s(i)totalexports-to-outputratio,(ii)finalgoodsexports-to-outputratio,(iii)intermediateexports-to-outputratio,and(iv)export-basedweightedoutdegree.Thelattermeasurecaptureshowimportantasectorisassupplierofintermediatesintheproductionofoneunitofitscountry-sectorexportpartners’output.Weinteractthesevariableswiththemonetarypolicyshockinthenextsetofregressions,focusingontheimpactonfirm-levelinvestment.Theapproachallowsustoidentifyhowthevariationintradeexposureimpactsshockspillovertotheaveragefirmwithinacountry-sector.Giventhatweexploitvariationatthecountry×sector×yearlevel,weareabletocontrolfortime-varyingfixedeffectsatthecountryand/orsectorlevel.WedocumentthattotalexportexposureplaysasignificantroleinthetransmissionofU.S.monetarypolicyshockstofirminvestment.Movementsalongthedistributionofcountry-sectorexportopennessfromlow(bottomdecile)tohigh(topdecile)amplifiestheimpactoftheshockbyfortypercentrelativetotheimpactontheaveragefirm.Interestingly,decomposingthetotalexport-to-outputratio,wefindthatitisintermediategoodsandservicestradethatdrivestheoverallexportexposurefindings,bothfortradewiththewholeworldandbilaterallywiththeUnitedStates.Finally,theestimatedcoefficientontheweightedoutdegreemeasureisalsoeconomicallyandstatisticallysignificant,indicatingthatit’snotjusttheimportanceofoverallintermediateexportsindrivingthetransmissionofU.S.monetarypolicyshockstoforeignfirms,butalsotheamplificationofdemandshocksviaglobalproductionlinkages.Theresultsofthisexternaldemandchannelviainternationaltradeandproductionlinkagesisinlinewithrecentfindingsintheliteratureusingmoreaggregateddata,suchasBr¨auningandSheremirov(2021)anddiGiovanniandHale(2022).Toprovideevidenceontheroleofdifferentialfinancialconstraintsontheinterestratechannel,werunpanelregressionsinteractingproxiesoffinancialconstraints(sizeornetworth)withtheU.S.monetarypolicyshock.Thisallowsustoexploittime-varyingfirm-levelvariationintheinteractionvariabletoidentifythistransmissionmechanism,andtherebyincludeanexhaustivesetoftime-varyingfixedeffectsatthecountry×sector×yearlevelalongwithnon-timevaryingfirmfixedeffects.Resultsshowthatforeignfirmswithlowerfinancialconstraintsareabletoattenuatetheimpactofmonetarypolicyshocksontheirinvestment,consistentwithrecentmicrostudiesofdomesticfirmsbyCloyne,Ferreira,FroemelandSurico(2020)andOttonelloandWinberry(2020).Themagnitudeofthiseffectislarge.Forexample,movingovertheinterquartilerangeofthefirm-networthdistributionimpliesthatlessfinanciallyconstrainedfirmsareabletoattenuatetheimpactofU.S.monetarypolicyshocksbyroughlyone-quarteroftheimpactonthemeanfirm.2Thefinalsetofheterogeneityregressionscombinesthetradeexposuremeasureswiththefinan-cialconstraintproxiesinordertoestimatethejointimpactofthesechannels.Theseregressionsyieldsomeinterestingresults.First,themagnitudeandsignificanceofthetradeandfinancialinter-actioncoefficientsdonotchangedramaticallywhenincludedtogether.Second,ourquantificationexercisesimplythatthedampeningeffectoflooserfinancialconstraintsoflargerfirmsdominatestheamplificationeffectofgreatertradeexposure.Whilethetrademeasuresareatthecountry-sectorlevel,largefirmstendtodominatetheexportmarket(Melitz,2003;FreundandPierola,2015),soouroverallresultspointtothese“granular”foreignfirmsbeingimpactedlessonnetbyU.S.monetarypolicyshocksgiventhechannelsidentifiedinourregressions.RelatedLiteratureTheempiricalliteratureoncross-borderspilloversofmonetarypolicyshocksisvoluminous.Mostofthisresearch,includingearlypapersontheGlobalFinancialCycle,reliedonaggregatedata.PioneeringresearchontheGFCincludesRey(2013),Rey(2016),Kalemli-Ozcan(2019),HanandWei(2018),andMiranda-AgrippinoandRey(2020).EarlyworkonspilloversfromU.S.monetarypolicyshocksincludesEichenbaumandEvans(1995),Rogers(1999),KimandRoubini(2000),FaustandRogers(2003),andFaust,Rogers,SwansonandWright(2003),whofocusedonforeigninterestratesandexchangeratesinVARs.Rogers,ScottiandWright(2014)examinetheeffectsofunconventionalmonetarypolicybytheFed,BOE,ECB,andBOJoncross-borderbondyieldsandstockprices,aswellasexchangerates.1Br¨auningandSheremirov(2021)documentthattradeplaysakeyroleinexplainingcross-countryheterogeneityintheeffectsofU.S.monetaryshocksonaggregateoutput,interestrates,andtradeflowsinalargepanelofcountries.Degasperi,HongandRicco(2021)findthataU.S.monetarypolicytighteninghaslargecontractionaryeffectsonbothadvancedandemergingeconomies,withfinancialchannelsdominatingoverdemandandexchangeratechannelsinthetransmissiontorealvariables.Onthemicroside,Br¨auningandIvashina(2020)examinetheroleofU.S.monetarypolicyinaffectingcreditconditionsofEMEfirms.Theyshowthatthespilloverisstrongerinhigher-yieldingandmorefinanciallyopenmarketsandforfirmswithahigherrelianceonforeignbankcredit.Morais,Peydr′o,Rold′an-Pe?naandRuiz-Ortega(2019)analyzetheuniverseofcorporateloansinMexico,matchedwithfirmandbankbalance-sheetdata,toidentifythespillovereffectsofadvancedeconomymonetarypolicyshocks.TheyfindthatatighteningofforeignmonetarypolicyincreasesSeealsoGeorgiadis(2016),whofindsthatthemagnitudeofU.S.monetarypolicyspilloversdependsonahostofreceivingcountrycharacteristics,includingtradeandfinancialintegration,exchangerateregime,andparticipationinglobalvaluechains;Dedola,RivoltaandStracca(2017),whofindthatasurpriseU.S.monetarypolicytighteningleadstoadollarappreciation,declineinforeignindustrialproduction,realGDP,andinflation,andriseinunemploymentinapanelofadvancedandemergingeconomies;andKearns,SchrimpfandXia(2019),whomeasuremonetarypolicyshocksforsevenadvancedeconomycentralbanksandspilloversto47advancedandemergingmarketeconomies.Theyfindnoevidencethatspilloversarerelatedtoreallinkagessuchastradeflows,butsomeimportanceforexchangerateregimes,withthekeycountrycharacteristicbeingfinancialopenness.3thesupplyofcreditofforeignbankstoMexicanfirmsandthatthisoccursviatheirrespective(country’s)banks.DiGiovanniandHale(2022)examinespilloversofU.S.monetarypolicyshockstosectoralstockreturns.Theyderiveamodelinwhichfirmsinallcountriesareaffectedbyamonetaryshock,byanamountthatisproportionaltoafirm’sglobalproductionlinkages,andfindthattheglobalproductionnetworkplaysthekeyroleintransmittingU.S.monetarypolicyshockstocross-borderstockreturns,evenconditioningonfinancialchannelvariables.InadditiontoourpaperbeingrelatedtothelargeliteratureoninternationalspilloversofU.S.monetarypolicy,itiscloselyrelatedtoClaessens,TongandWei(2012)andDao,MinoiuandOstry(2021).AlthoughneitherofthesepapersexaminesU.S.monetarypolicy,Claessensetal.(2012)examinehowtheglobalfinancialcrisisaffectedfirms’profits,sales,andinvestment,thefocusofourpaper.Theyfindthatthecrisishadabiggernegativeeffectonfirmswithgreatersensitivitytobusinesscycleandtradedevelopments,particularlyincountriesmoreopentotrade.Daoetal.(2021)examinetherelationshipbetweenrealexchangeratefluctuationsandfirm-levelinvestmentandgrowth.Theyshowthatrealdepreciationboostprofits,investment,andassetgrowthoftradablesectorfirmsthathavehigherlaborsharesandarerelativelymorefinanciallyconstrained,interpretingthisfindingasevidenceforan“internalfinancingchannel.”2Ourpaperisalsorelatedtoworkontradeandtransportcostsininternationaltradeandmacro(ObstfeldandRogoff,2000;AndersonandvanWincoop,2004).Theseauthorspresentevidencethattotaltradecostsinrichcountriesarelarge,withanestimatedadvaloremtaxequivalentofabout170percent,whilepoorcountriesfaceevenhighertradecosts.Morerecentestimatesindicatethatglobaltradecostshavedeclinedby15percentbetween2000and2018(/).Clearly,thereisalotofvariationacrosstime,countries,andgoods,featuresofthedatathatweexploit.OurforensicinvestigationofthelinkagesfromU.S.monetarypolicyshockstocross-borderfirms’investment,sales,andemploymentuncoversakeyrolefortradenetworks,consistentwithBr¨auningandSheremirov(2021)anddiGiovanniandHale(2022).Finally,ourpaperisalsorelatedtotheliteratureontheinvestmentchannelofmonetarypolicytransmissioninclosedeconomies.Thisliteratureemphasizestheimportanceoffirmheterogeneityforthetransmissionofmonetarypolicy,withmuchattentionpaidto“balancesheeteffects.”Thebalancesheetchannelbroadlyreferstofeedbackeffectsbetweenthehealthofborrowers’balancesheets,asmeasuredforexamplebynetworth,andinvestment(e.g.oroutput,assetprices,etc.).Inthisframework,financiallyconstrainedfirmsborrowinordertoundertakeproductivelong-termprojects.Firmsfinanceprojectsbyissuingclaimstolenders.Thecashflowsassociatedwithfirms’projectsareexposedtoanaggregateshockthatmaygeneratefluctuationsinnetworth,WhileputtingthefinishingtouchesontheARCversionofthispaper,webecameawareofcontemporaneousworkinprogressbyArbatli-Saxegaard,Firat,FurceriandVerrier(2022).Theseauthorsalsoexaminethecross-bordereffectsofU.S.monetarypolicyshocksinalargepaneloffirm-leveldata.AlthoughwedomuchmoreanalysisoftradenetworkchannelsandusedifferentmeasuresofFedmonetarypolicyshocksandinvestment,thetwopapershaveasimilarfocus.4whichcouldinturntriggerliquidationsofcapitalandaffectinvestment,thepriceofcapital,andaggregateoutput.Monetarypolicyshocks,forexample,wouldgiverisetosucheffects.SeminalpapersincludeBernankeandGertler(1989)andKiyotakiandMoore(1997)andmorerecentworkbyCloyneetal.(2020),OttonelloandWinberry(2020),Caglio,DarstandKalemli-Ozcan(2021).Therestofthepaperproceedsasfollows. Section2describesourempiricalmethodology.Section3describesthedataandpresentssummarystatistics.Section4providesregressionresultsthatfocusontheroleofcountry-levelcharacteristics,whileouranalysisoffirm-leveltradeandfinancialconstraintheterogeneityisinSection5.Section6concludes.MethodologyWefirstestimatetheunconditionalimpactofU.S.monetarypolicyshocksonforeignfirms’annualchangeininvestment,sales,andemployment.Ourregressionanalysisfollowstheapproachesusedinaclosed-economysettingbyrunningpanelregressions,whereweallowforthepossibilityoftracingoutthedynamicimpulseofendogeneousvariablesusinglocalprojections(Jord`a,2005).Thebaselineregressionthatestimatestheaverageeffectofmonetarypolicyshocksonallfirmsis:Yfsc,t+h?Yfsc,t?1=α+βMPtUS?1+Zfsc,t?1γ′+Xc,t?1δ′+εfsc,t+h,(1)wherefdenotesafirm,sthesectorandcthecountry.Yfsc,t+histhefirm-leveloutcomemeasuredinyeart+h,h=0,1,2,...,T.Thefirm-leveloutcomesareeither(i)theinvestment-to-laggedfixedcapitalratio(It/Kt?1),(ii)thesales-to-laggedfixedcapitalratio(St/Kt?1),or(iii)logemployment(lnEt).Giventheuseofannualdata,ourbaselineistoestimatethemodelforh=0only.Inthiscase,theleft-handsideof(1)measureseithertheannualchangeintheinvestmentorsalesshares,orannualemploymentgrowth.MPtUS?1istheU.S.monetarypolicyshockvariablefromBu,RogersandWu(2021)(BRW)att?1,thusaccountingforthelaggedimpactofmonetarypolicyontherealeconomy.Asdescribedbelow,theBRWshockisameasureofmonetarysurprisescenteredoneachoftheeightFOMCmeetingsperyear.Tomatchourannualfirm-levelrealvariables,weaggregatetheeightshockobservationsthroughoutthecalendaryear,whichiscustomaryintheliterature.Thistimingissuefurthermotivatestheuseofalaggedmonetarypolicyshockvariableasopposedtoacontemporaneousone.3Ifamonetarypolicytightening(loosening),MPUS>0(MPUS<0),depresses(stimulates)firms’activity,wewouldexpectthatβ<0.Wefurthercontrolforotherstandardfirm-levelcontrols,Z,whichwelagoneperiod.Thesevariablesincludefirmsize(measuredasthelogoftotalassets),networth,andchangeinthecashflow-to-assetratio.4Wealsoincludethelagofmacroeconomiccontrols,X,whichmayvaryat3Weexperimentedwithadditionallags,butthisdidnotyieldanyadditionalinsights.4Wealsoexperimentbyincludingfirmage,Tobin’sQ,andothermeasuresoffirms’financialhealthsuchaschangesinitsleverageratio.Includingtheseregressorsdidnotimpactourresultsbutcutthesamplesizeinseveralcases5thecountryorgloballevels.TheseincludedomesticrealGDPgrowth,changeinthelognominalexchangerateagainsttheU.S.dollar,thechangeinshort-termdomesticinterestrates,andlogVIX.Giventhepanelsetup,wearealsoabletoincludeasetofnon-time-varyingfixedeffects,α,(e.g.,atthecountry,sector,orfirm-level).Finally,εistheerrorterm.Giventhatthemonetarypolicyshockisrepeatedacrossallfirmsinagivenyear,weclusterstandarderrorsattheannuallevelandfurtherclusteratthefirmleveltocontrolforpotentialautocorrelationintheerrors.Equation(1)isausefulbaselinespecificationtoestimatetheimpactofU.S.monetarypolicyontheaveragefirminagivencountry.Wecanthen“unpack”thepotentialheterogeneousimpactsofmonetarypolicybyallowingforβtovaryacrossmultipledimensions.Tobegin,weexaminehowtheimpactofU.S.monetarypolicyonforeignfirmsvariesacrosscountriesviasimplesamplesplitsandinteractionswithcountrycharacteristics.Forexample,weexaminewhetherβdiffersbe-tweenemergingmarketeconomies(EMEs)andindustrialcountries.Wealsoexaminehowfinancialaccountandtradeopennessattheaggregatelevelsimpacttheestimatesofβ.RoleofTradeLinkagesChangesinU.S.monetarypolicymayimpactforeignfirms’activitydirectlygiventheresultingcontraction/expansionofU.S.demand.Thischannelmightbeexpectedtohaveanout-sizedimpactonfirmsorsectorsdependingonhowinvolvedtheyareininternationaltrade.Further,giventheexpansionofglobalproductionnetworksovertime,firmsthataremoreintegratedinglobalvaluechainsmaybeevenmoreimpactedgivenspilloversacrosscountriesarisingfromthechangeinU.S.monetarypolicy.Givendatalimitations,weareforcedtoexploittradeheterogeneityatthecountry-sectorlevelratherthanfirmlevelinourestimation.5Thus,ourfirstextendedregressionspecificationexploitsheterogeneityacrosscountry-sectorswithinayear:Yfsc,t+h?Yfsc,t?1=α+β1MPtUS?1+β2(Tradecs,t?1×MPtUS?1)(2)θTradecs,t?1+Zfsc,t?1γ′+Xc,t?1δ′+εfsc,t+h,whereTradecsisameasureofacountry-sector’stradeexposure.Weconstructseveralmeasuresofacountry-sector’sexposuretodemandshocksbyexploitingheterogeneityinacountry-sectors’linkstoeithertheworldorU.S.onlyviaexportsalongfourpossibledimensions:acountry-sector’s(i)totalexports-to-outputratio,(ii)finalgoodsexports-to-outputratio,(iii)intermediateexports-to-outputratio,and(iv)export-basedweightedoutdegree.AswedescribeinSection3,theweightedoutdegreemeasurecaptureshowimportantacountry-sector’soutputisforallothercountry-sectors’(forexample,Italianfirmsdonotreporttheagevariableinourdataset).Therefore,inordertomaximizesamplesizeweconstraintheinclusionoffirm-levelcontrolsinthefinalanalysis.SeediGiovannietal.(2022)forevidencethatsalesgrowthoffirmsmoreexposedtotradearemoresensitivetochangesinworldGDP.6productionandthuscapturestheimportanceofacountry-sectorintheglobalvaluechain.6Notethatwealsoexploredrelatedimport-basedmeasures,consistentwiththeideathatthegeneralequilibriumimpactofU.S.monetarypolicyshocksmayalsofeedthroughtofirms/sectors’costsviaimports,butresultsbasedonimportmeasureswereneversignificantsoweomitforbrevity.ConditionalonU.S.monetarypolicyhavingagreaterimpactonfirmsinsectorsthathavelargertradeexposuremeasures,wewouldexpectthatβ2<0.Anotabledifferencebetweenthisandourbaselineestimationisthatthesetoffixedeffects(α)maynowvaryovertime,allowingustocontrolforunobservedtime-varyingcountry-and/orsector-levelcharacteristics(e.g.,howacountry’stradeopennessvariesovertime).7Regression(2)issimilartotheregressionwithfirmheterogeneitythatwedescribebelow,howeverherethetime-varyingfixedeffectscannotbeasgranularbecausethetradevariablesonlyvaryatthecountry×sectorlevelwithinayear.However,byexploitingdifferencesintradepatternsaswellasthetypeoftrade(intermediatevs.finalgoods),weareabletoestimatemicro-levelresponsesthatvaryforafirmgivenitscountry-sectors’exposuretodifferenttradechannels,whilealsocontrollingfortime-varyingfirm-levelvariables.RoleofFirmFinancialConstraintsWealsoestimatetheroleoffirmfinancialconstraintsinaffectingthetransmissionofU.S.mon-etarypolicyshocks.Weextendthebaselineregression(1)toallowforheterogeneouseffectsatthefirmlevel,conditioningonstandardfirm-levelmeasuresoffinancialconstraints:Yfsc,t+h?Yfsc,t?1=α+β1MPtUS?1+β2(Zfcs,t?1×MPtUS?1)+Zfsc,t?1γ′+Xc,t?1δ′+εfsc,t+h,(3)wherewenowallowfortheimpactofmonetarypolicytovarybyfirmcharacteristicZthroughβ2.Giventhatfirmcharacteristicsmayvarybothwithinandacrosscountries(e.g.,thelargestfirminGermanymaybelargerthanthelargestfirminThailand),wenormalizeallfirm-levelinteractionvariableswithinacountry-yearaswedescribebelowinthedatasection.Followingtheliterature,twofirmcharacteristicsthatweusetoproxyforfirmfinancialconstraintsaresizeandnetworth.Weusethelogoftotalassetsforourmeasureofsize.Whilethismaygenerateanaturalcorrelationwiththetwoendogenousvariablesthataredeflatedbythelagoffixedassets(investmentandsalesratios),wehavealsoestimatedallregressionswithlogemploymentasameasureofsizeinsteadandobtainedsimilarresults.8ThenetworthvariableisdefinedasthelogofthedifferenceintotalassetsSeeCarvalho(2014)foramoredetaileddiscussionoftheweightedoutdegreeandotherpossibleproductionnetworksufficientstatistics.7Notethatdoingsoeliminatesthepossibilityofestimatingtheaverageimpactofmonetarypolicy,β1.Therefore,wefirstrunregressionswithouttimeeffectsinordertoestimatetheimportanceoffirmheterogeneityrelativetotheaverageeffectofmonetarypolicyonallfirmsinaneconomy.8Asnotedinthefirmdynamicsliterature(Cloyneetal.,2020,andothers),agealsoplaysaroleindependentlyofsizeornetworth.Wehavealsorunregressionswithageandresultswerequalitativelysimilartousingsizeornetworth,butreducedsamplesizequitesubstantially,soweomittheseresults.7andtotalliabilitiesandisonemeasurethatproxiesforthedifferencesinfirms’collateral/abilityto¨borrow(seeGopinath,Kalemli-Ozcan,KarabarbounisandVillegas-Sanchez(2017);diGiovanni,¨Kalemli-Ozcan,UluandBaskaya(2021);Caglioetal.(2021)).IflargerorhighnetworthfirmsarelessimpactedbyU.S.monetarypolicyshocksbecausetheirfinancialconstraintsarelessbinding,wewouldexpectthatβ2>0.Themoststringentsetoffixedeffectsmaynowvaryatmoregranularlevelsalongthetimedimensions–specificallyatthecountry×sector×yearlevel–sinceidentificationoftheinteractiontermsisexploitingvariationatthefirm×yearlevel.Therefore,weareabletoidentifydifferentialimpactsoffinancialconstraintswithinayearalongthefirmdistributionwhilecontrollingfortime-varyingcountry×sectorcharacteristicsorshocks.Giventheliteraturethatstudiesthebalancesheeteffectofexternalshocks(e.g.,exchangeratechanges),weextendtheestimationof(3)alongseveraldimensions.Forexample,weinteractothermacrovariables,suchaschangesinthe