P aper No. 907 Banks, shadow banks, and business c

1Introduction

Whatdrivesbusiness-cyclefluctuations?Theviewthatfinancialfactorsareamajor causehasbecomewidespread,supportedbyever-growingevidence.Theliterature highlightstwomainchannelsthroughwhichfinancialstressimpactstherealeconomy. Oneinvolvestheroleofhousepricesandcreditonhouseholdspending(MianandSufi 2011,2014,Ramcharan,Verani,andHeuvel2016,JensenandJohannesen2017),the otherconcernsthedisruptionofcreditonfirminvestmentandhiring(Chodorow-Reich 2014,GiroudandMueller2017,Huber2018).

Onthetheoreticalfront,researchershavemodeledthetwochannelsextensively. Financialfrictionsinmacroeconomicmodelsnowrangefromconsumersborrowing topurchasehousesandgoods(GuerrieriandLorenzoni2017,Justiniano,Primiceri, andTambalotti2018,Kaplan,Mitman,andViolante2020)tofirmsusingdebtto financeinvestmentprojects(JermannandQuadrini2012,Arellano,Bai,andKehoe 2018,KiyotakiandMoore2019)tobankscollectingdepositstofundthecorporate sector(BrunnermeierandSannikov2014,Boissay,Collard,andSmets2016,Gertler, Kiyotaki,andPrestipino2020).Whiletherelativeimportanceofthetwochannelsis debated(GertlerandGilchrist2019,Kehoeetal.2020),thislargebodyofresearch hasgreatlyimprovedourunderstandingoftheinterplaybetweenfinancialcrisesand recessions.

Yetforalltherecentprogress,existingtheoriesfailtoexplainoneempiricalregu- larity:creditspreadsonhouseholdandbusinessloansmovehandinhand.Thisfact, showninFigure1fortheeuroarea,suggeststhetwochannelsarealwaysoperating together.1^,2Whenbankstightencredit,theytightenforhouseholdsandfirmsatonce.

Figure1alsodisplaysan(inverted)indexofconsumerconfidence,aleadingindicator believedtobeacentralforcebehindaggregatespending.Consumerpessimismspikes attheeveofeachrecession,afewmonthsbeforespreadsthemselvesshootup.Thus, itappearsthatasconfidenceamongconsumersandinvestorssinks,financialmarkets respondinawaythatbanksenduprestrictingcreditacrosstheboard.Interestingly,

YvanBecard:DepartmentofEconomics,PUC-Rio,RuaMarquêsdeSãoVicente,225,RiodeJaneiro, RJ22451-900,Brazil;[e-mail]:BankofEngland,Threadneedle St,LondonEC2R8AH,UK;[e-mail]biie, AmbrogioCesa-Bianchi,CristianoCantore,SimonGilchrist,TimoHiller,MichelJuillard,Ricardo Masolo,BenjaminMoll,andRicardoReisforhelpfuldiscussions.

1Thecorrelationis0.96intheeuroareaoverthe2003-2020period.Figure6intheOnlineAppendix showsthatthisfactholdsacrossindividualcountriesoftheeuroareaaswellasintheUnitedStates.

2Tobesure,afewpapersdomodelthetwochannelstogether.Buttheyeitherabstractfromspreads (Iacoviello2005,LombardoandMcAdam2012)orresorttocorrelatedhousehold-andfirm-specific shocksinordertomatchthejointmovementsofspreadsinthedata(Geralietal.2010,Ferrante2019).

1

Figure1:ConsumerPessimismandCreditSpreadsintheEuroArea

Notes:Consumerpessimismcorrespondstotheconsumerconfidenceindicator,invertedandrescaled.Creditspreadsarethe differencebetweeninterestratesonbankloans(tohouseholdsandfirms)andtheshort-termeurointerbankrate.Shadedbars indicateCEPR-datedrecessions.

bankdepositstendtoincreaseinbadtimes,asinsureddepositorsflytosafety,implying thatthistraditionalsourceoffundingdoesnotposeathreattobanks.Bycontrast,itis welldocumentedthatnonbankfinancialinstitutions,alsoknownasshadowbanks,suf- ferrunsinperiodsofstress.3Theseinstitutionshandlevastamountsofdebtandserve asanadditionalsourceoffundingforcommercialbanks.Buttheytypicallyexhibita volatileandprocyclicalbehavior.

Inthispaper,weproposeatheoryastowhybanksadjusttheirlendingstandards simultaneouslyonalltheirborrowers.Thekeyelementisafinancialsectorinwhich traditionalbanksinteractwithshadowbanks.Traditionalbanks(henceforthbanks) transformdepositsfrominvestorsintoloansfortherealsectoroftheeconomy.Shadow banksdonotsupplyloansdirectlybutabsorbalargechunkofriskwhichthebanks areunwillingtobear.Thisisbasedonthereal-worldobservationthatshadowbanks providemainstreambankswithanarrayofservicesincludingsecuritization,insurance, andliquidationofnon-performingloans.Suchservicesrepresentatransferofriskfrom originallenderstooutsideinvestorsandthepriceatwhichtheyaretradeddependson howeasilyshadowbanksraisefundsonthefinancialmarkets.

Wefirstlayoutourtheoryinasimpletwo-periodbankingmodel(Section2).Banks holdaportfolioofriskyassetsbutarenotallowedtodefaultontheirdepositliabilities. Tomitigatethiscreditrisk,theymayeitherrequireasufficientlylargeamountofphysical assettobepostedascollateralortheymayshiftpartoftheirportfoliotoshadowbanks.

3Shadowbanksincludemoneymarketfunds,privateequityfunds,hedgefunds,insurancecompanies, securitieslenders,andstructuredinvestmentvehicles.GortonandMetrick(2012),Pozsaretal.(2013) andCovitz,Liang,andSuarez(2013),amongothers,describehowinvestorsranontheseinstitutions duringthelastfinancialcrisis.

2

Shadowbankspooltogethermanyloans,therebyeliminatingidiosyncraticrisk,and transformthemintotradableasset-backedsecurities,whichtheyselltoinvestors.We showthatsecuritizationenablesbankstoexpandtheirbalancesheet:foragivenamount ofcollateralintheeconomy,lendingishigher.

Whathappenswhenconfidenceamonginvestorsdips,saywhentheyrealizethat securitiesareriskierthandeposits?Werefertothissituationasamarketsentimentshock. Shadowbankssuddenlyhavetroublefloggingtheirsecuritiesandmustcutbackon activity.Asthevolumeofsecuritizationfalls,bankscompensatebytighteningstandards, increasingcollateralrequirementsandinterestrates.Thus,ourtheoryprovidesasimple mechanismthatlinksmarketsentimenttobankcreditspreadsviathenonbankfinancial sector.

Thenextstepofouranalysisconsistsinmeasuringthemacroeconomiceffectsof thispropagationmechanism.Weembedtheframeworkintoarichdynamicstochastic generalequilibriummodeldesignedtofittheEuropeandata(Section3).Borrow- ersincludeasubsetofhouseholds,whichemployloanstopurchasehousing,and entrepreneurs,whichemployloanstopurchasecapital.Afractionoftheseagentsde- faultsinequilibrium,pushingbankstochargeaspreadovertherisk-freerate.Weuse macroeconomicandfinancialdatafortheeuroareafrom1999Q1to2019Q4alongwith standardBayesiantechniquestoestimatetheparametersofourmodel.

Ourmainfindingisthatamarketsentimentshocktriggersdynamicsthatmirror actualbusinesscycles(Section4).Higherspreadsforcehouseholdsandfirmsto takeonfewerloans.Housepricesandhouseholdnetworthdrop,causingafallin consumption.Capitalpricesandfirmnetworthdrop,causingafallininvestmentand employment.Arecessionensues.Thepresenceofthetwochannels—onhouseholds andfirms—iskeytoreplicatingthejointbehaviorofthedata.Weestimatethatthe marketsentimentshockisresponsiblefor49to55percentofthevarianceinoutput, consumption,investment,andhoursworkedoverthepasttwodecades.Theshockalso drivesmostofthemovementsinthetwocreditspreads,thenominalinterestrate,and isabigforcebehindcreditquantitiesandhouseprices.Asfarasweknow,wearethe firsttoassignsuchalargeroleinEuropeanbusinessandfinancialcyclestoasingle disturbance.

Tobuildtrustinthisstory,weperformtwoexternalvalidationexercises(Section 5).First,wecomparethetimeseriesofthemarketsentimentshockcomingfromour estimatedmodeltoameasureofsystemicfinancialstressinEurope,whichwedonotuse intheestimation.Thetwoseriescorrelatewell,spikingrightbeforethetworecessions ofthesample.Second,followingarecentpaperbyAngeletos,Collard,andDellas (2020),weestimateastructuralvectorautoregression(VAR)whereweidentifyashock asonethatmaximizesthevolatilityinoutputatbusiness-cyclefrequency.Werepeat

3

theprocedurebytargetingthefinancialstressindex.Weshowthatthesetwoimpulses producevirtuallyidenticalresponsesforallendogenousvariables,andhencerepresent twofacetsofthesame"mainbusiness-cycleshock".Whatismore,thedynamicsclosely resemblethoseofourstructuralmodelwhenhitbyamarketsentimentshock.These independentexperimentsreinforcethecredibilityofsentimentshocks,orchangesin aggregateconfidence,asmajordriversofbusinessandfinancialcycles.

Ouranalysiscontributestoavastliteraturewhichestimatesstructuralmodelstoun- derstandeconomicfluctuations.4Asnotedabove,nostudyexplainsthejointdynamics ofhouseholdandfirmcreditspreadsandquantitieswithauniqueforce.Oneexception isBecardandGauthier(2020),wherewefindthatashocktotheabilityofbanksto redeploycollateraltiesthespreadstogetherandhelpsaccountforthecomovementsbe- tweenconsumptionandinvestmentseeninUSdata.Thatshock,however,isexogenous, andshouldbeviewedasareducedformfordeepereventsoccurringinthefinancial system.Thispapergoesastepfurtherandmicrofoundstheseveryeventsbymodeling themarketforsecuritizeddebt—hencemakingthestoryendogenous.Inourmodel thefinancialsectorbecomesaconduitthroughwhichwavesofoptimismorpessimism propagateandamplify.

Thisarticlealsorelatestotwoactivelinesofresearch.Thefirstlineemphasizes theroleofconfidenceandexpectationsonbusinesscycles(Lorenzoni2009,Angeletos andLa’O2013,BeaudryandPortier2014,Benhabib,Wang,andWen2015,Angeletos, Collard,andDellas2018).Wedonotattempttomodelthecauseofextrinsicshocks(eg coordinationfailure,imperfectinformation,departurefromrationality)butrathertake theseshocksforgrantedandstudytheirconsequences.Ourresultsrevealthatsentiment swingscanaccountforthecyclicalbehaviorofabroadrangeofmacroeconomicand financialaggregates.Theotherlineofresearchfocusesonshadowbanking.Shadow banksexistbecausei)theyprovideliquidityandfundingtofinancialintermediaries (Pozsaretal.2013,Gennaioli,Shleifer,andVishny2013,MoreiraandSavov2017); ii)theybenefitfromlightregulation(GortonandMetrick2010,Acharya,Schnabl,and Suarez2013,Plantin2015,BegenauandLandvoigt2020).Ourmodelbelongsinthe firstcategory,butsharesthecommoninsightinthisliteraturethatshadowbanksmake thefinancialsystemandthewholeeconomymorevulnerabletoreversalsinconfidence.