美聯(lián)儲(chǔ)-2023年銀行業(yè)動(dòng)蕩與銀行定期融資計(jì)劃

FinanceandEconomicsDiscussionSeriesFederalReserveBoard,Washington,The2023BankingTurmoilandtheBankTermFundingProgramDavidGlancy,FeliciaIonescu,ElizabethKlee,AntonisKotidis,MichaelSiemer,AndreiZlate2024-045Glancy,David,FeliciaIonescu,ElizabethKlee,AntonisKotidis,MichaelSiemer,andAndreiZlate(2024).“The2023BankingTurmoilandtheBankTermFundingProgram,”FinanceReserveSystem,/10.17016/FEDS.2024.045.NOTE:StafworkingpapersintheFinanceandEconomicsDiscussionSeries(FEDS)arepreliminarymaterialscirculatedtostimulatediscussionandcriticalcomment.TheanalysisandconclusionssetfortharethoseoftheauthorsanddonotindicateconcurrencebyothermembersoftheresearchstafortheBoardofGovernors.ReferencesinpublicationstotheFinanceandEconomicsDiscussionSeries(otherthanacknowledgement)shouldbeclearedwiththeauthor(s)toprotectthetentativecharacterofthesepapers.1*DavidGlancyAntonisKotidisAndreiZlateAbstract(BTFP)insupportingtheliquiditypositionsofvulnerablebanksduringtheMarch2023bankingturmoil.Weuncoverthreekey?ndings.First,ourhigh-frequencydatacon?rmthatbankswithofsecuritiescollateralattheBTFP(banksatthe90thpercentileinsecuritieslossesreplaced26centsofeverydollarofout且owswithBTFPborrowingThird,inadditiontofundingloangrowthanddepositout且ows,banksusedbuildcashholdings,indicatingthattheprogramenabledbankstopospotentialfuturefundingneeds.Overall,wedemonstratethattheBTFPenabledbankstomeetfundingneedsandpresKeywords:2023bankingturmoil,depositout且ows,uninsureddeposits,securitieslosses,BankTermFundingProgram(BTFP),emergencyliquidityfacilitiesJELClassi?cations:E52,E58,G01,G21*AllauthorsareattheFederalReserveBoard.WethankMarkCarlson,VirginiaLewis,DavidLowe,MattMalloy,FrancisMartinezandAlexandrosVardoulakis,aswellasseminarparticipantsattheFederalReserveBoardforhelpfulcomments.BohanWangprovidedoutstandingresearchassistance.TheviewsexpressedinthispaperarethoseoftheauthorsanddonotnecessarilyrepresentthoseoftheFederalReserveBoardofGovernorsoranyoneintheFederalReserveSystem.21IntroductionAkeyfunctionofbanksismaturitytransformation.Theyborrowshrisk(Drechsleretal.(2021)).Depositsusuallybearbelow-marketrateswheninterestratesrise,termassetswheninterestratesrise,uninsureddepositorsmayrun(Jiangetal.(2023),etal.(2023),Haddadetal.(2023)).ThefailureoftheBankoftheCommonwealthin1973isanunstablefundingbase(Prescott(2024)).ThebankingturmoilinMarch20franchisewheninterestratesfall(Drechsleretal.(2023)).Bankscouldalsoinvestinswapsthatfbankassetsusinginterestrateswappositions(e.g.McPhailetal.(2023);Granjaetal.(2024)).Thispaperestablishesthatanex-postlenderoflastresort(LOLR)interventioncalibratedtopreservebanks’adocumentanddiscusstheefectivenessoftheBankTermFundingProgram(BTFP)introducedin(Figure1,leftpanel).FollowingthefailuresofSiliconValleyBank(SVB)andSignatureBankinMarch2023,concernpanel).Inresponsetotheturmoil,theFederalReserveBoardestablishedtheBTFP,withapprovalbytheSecretaryoftheTreasury,pursuanttosection13(3)oftheFederalReserveAct.TheBTFPindexedswap(OIS)rate.Importantly,BTFPcollateralwasvaluedatparratherthanmarketvalue.310AprAugAprAugAprAvailable?for?SaleSecuritiesHeld?to?MaturitySecuritiesCoreDeposits($theefectivenessoftheBTFP.banks’assetsandliabilitiesfromfromtheBTFPandthediscountwindow.Theanalysiscentersaroundthreemainquestions.First,fundingneeds.However,BTFPwasanimportantsourc4fat-parcollateralvaluationattheBTFP.Furthermore,relianceonBTFPborrowingswasespeciallyftheysubsequentlyencumberedwithBTFPloans.Thiscollateralpledgingwasconcentratedinmortgage-backedsecurities(MBS)andcollateralizedisproportionatelylargefairvaluedecliThird,tounderstandhowbanksusedwholesaleborrowingsfromtheBTFPandotandloansovertime.Startingwithshiftsinthecompositionofbankliabilities,we?ndthatbankscashbufers,suggestingaprecautionaryreasonforwholesaleborrowing.Additionally,banksthatOur?ndingshaveseveralimportantimplications.First,BTFPwassuccessfulintargetingthemorevulnerablebanks,whoseBTFPborrowiwithlargesecuritylossesanthanotherbanks.VulnerablebanksalsopledgedsubstantiallymorecollateralattheBTFPthansecuritylosses,giventhatsuchbanksusedmorwithotherbanks.Theat-parcollateralvaluationadvantageofBTFPwasespeciallyimportantforvulnerablebanks,whichpledgedlargeramountsofMBSandCMOcollateral—securitieswithtorisinginterestrates—the?ndingthatbanksborrowingfromeitherBTFPornon-Fedsourcesexperiencedsimilartrendweremorelikelytoborrowfrom5Relatedliterature:Our?ndingscontributetotheliteratureontheinterplaybetweenbanks’liquidityandinterestraterisk.Intheory,theriskofrunsonabank’sdepositfranchiseincreases(2023);Jiangetal.(2023);Haddadetal.(2023)).Ourresultshighlighttheroleofbanks’thatbankswithlargerlossestappedmoreintoBTFPfunding—obtainingfundingaRelatedly,our?ndingscontributetotheliteratureonLOLRbyanalyzingtheBTFPintwoimportantways.First,alargetheoreticalliteraturediscussesthebene?tsofLOLRinterventionsduring?nancialcrises(DiamondandDybvig(1983);Freixasetal.(2004);RochetandVives(2004)).Bysupplyingtermfunding(Jasovaetal.(2021)),expandingcollateraleligibilityrequirements(VanBekkumetal.(2018)),ordirectlypurchasingcommercialloa(2024)),LOLRcanhelpbankssupportthesupplyofcreditandpreventacrunchwhenprivateliquidityfreezes.Our?ndingthatBTFPsupportedbanks’loanbalancesisconsistentwiththisliterature.ThisislargelyduetoBTFP’sone-year?xed-ratefundingat10basispointssprbank(e.g.,Drechsleretal.(2021)).Second,our?ndingthatbankswithlargesecuritieslossesdisproportionatelyusedtheBTFPprogram’sabilitytoencourageutiagainstcollateral,theBTFPhelpedpreventthemfrombecomingundercapitalized,whichwouldhavefurtherexacerbatedthebankrun(Eganetal.(2017);Jiangetal.(2023)).Inrisk.This?ndingmarksanimportantcontributiontotheliteratureonthedesignandtransmis-sionofcentralbankliquidityprovision(Drechsleretal.(2016);CarlsonandMacchiavelli(2020);CarpinelliandCrosignani(2021);Kotidisetal.(2022)).Moregenerally,ourpaperrelatestotheliteincreasethelikelihoodofarun(e.g.,DiamondandRajan(2011);BrunnermeierandPedersen(2009)).However,ourpaperoccurevenwhenassetsarefullyliquidandhigh-quality(e.g.,6ingspromptedruns,whichtheBTFPhelpedcontain.Ourevidencethatdepositorbehaviorduringbankruns(e.g.,IyerandPuri(2012);Iyeretal.(2016)).bankingturmoilevents.Forexample,Choietal.(2023)studythecontagionefWhileCiprianietal.(2024)studytheroleoffundamentalsandpublicsignalsonfunding,aswellashighlightingtheroleofparcollateralvaluationattheBTFPincontainingruns.2InstitutionalBackgroundSiliconValleyBank(SVB)wasa$200billionbankwithaconcentratedcustomerbasamidinvestors’concernsabouttheircryptocurrencyindustryexposures.WhileSilvergateBankdecidedtowinddownitsoperationsvoluntarily,SVBandSigftosizabledeclinesinbankstockprices.TheKBWBankIndexandKBWRegionalBankIndexsector.OnMarch12,theFederalReserveannouncedtheestablishmentoftheBTFP.TheBTFP7depositoryinstitutions.Inaddition,BTFPcollateralwasvaluedatpar,ratherthanmarketvalue.Eligiblecollateralwaslimitedtoopen-marketoperationeligiblecollateral,andincludedTreasuryTheleftpanelofFigure2showsthattheinterestrateonBTFPloansincreasedsteadilybetweeninterestratesoverthenextyear.TheinterestrateapplicabletonewBTFPloanswasadjustedonJanuary24,2024,tobenolowerthantheintereTherightpanelofFigure2plotstheoutstandingamountofBTFPborrowing(inorange)andofcollateralpledged(inblue)overtheprogram.TheoutstandingamountofBTFPloansandthecollateralpledgedattheBTFPincreasedrapidlyinthe?rstfewmonthsoftheprogram,motivetoaccesstheBTFPifandwhentheyneedto.AfterJanuary24,2024,whentheFederalFigure2:BTFPrateanduptakeAprAug5.6AprAug0AprAugAprAug4000Note:Theleftpanelofthe83MethodologyandData3.1MethodologyOuranalysiscentersaroundthreequestions.First,whatriskfactorscontributedtocoredepositThird,howdidbanksuseBTFPandotherfundstofundloansandaccumulatereservebalances?=β1+β2+√,Xi;t+εi;t(1)sideration,anduninsureddepositsandHTMsecuritieslossinteractingthetworiskfactorstocapturethefactthatHTMsecuritieslossesmaypredominantlyafecttheincentivesofuninsureddepositorstowithdrawfunds(Jiangetal.,2023).theHTMsecurityportfoliosincebanksarerestrictedinhedginginterestrateriskforHTMsecu-broadlysimilarwhenconsideringsecuritylossescombinedbetweisavectorofcontrolsthatincludesloangrowth,△Loansi;t;t,/Assetsi;t,toaccountforthefactthat liabilities),andFHLBadvancesasashareoftotalassets.2Finally,weincludebanksizeindicators 1AppendixTableB1presentsresultsfrom9sidevariableofspeci?cation(1)withchangesineitherBTFPborrosecuritiessales),andtestwhetherbanksthathad(2)(2)finterestand△Depositsi,t,t,/Assetsi,tisthechangeincoredepositsduringthistime.β1todepositout且owsdifersacrossbanks.Finally,Xi,tisavectorocontrolsinequatioinBTFPbalancesrelativetootherbanksthatdidnotchangeBTFPusageinaparticularweek.h∈[-10,-9,...,10]),sotheestimationdfollowingweeks(Bassettetal.,2014;Castroetal.,2022).3Theomittedcategoryisbankswithtotalassetsthatexceed$250billion.4Forrobustness,wealsoestimateequation(2)withExcessLoanGrowth—de?nedas(△Loansi;t;t,-△Depositsi;t;t,)/Assetsi;t—asthemainindependentvariable.Thismeasurere丑ectstheadditionalfundingneedthebankneedstoraise(besidesfromcoredeposits)tofundthegrowthintheirloanportfolio.October4th.3.2DataOuranalysisprimarilydrawsonformFR2644,whichprovidliabilitiesataweeklyfrequency.Tofocusattentiononinstitutionsparticipatinginconventionalbankingactivities,werestrictthesampletodomesticallycharteredbanks,andomitbanksincompaniesprincipallyinvolvedininvestmentbankingorcustodianactivities.5Ontheassetside,thedataprovideadetailedclassi?cationofdiferentholdings.Ontheliabilitfunds,discountwindowandBTFPborrowing.WeisolateFederalReserveborrowingfromotherwholesaleborrowingsbyseparatelyobtainingdiscountwindowandBTFPborrowingsdatafrommarkets,theremainingpartofwholesaleborrowingsismostlycomposedofFHLBadvances.Finally,3.3SummarystatisticsbalancesheetcharacteristicsofFR2644?lerstotheuniverseofbankThemostsigni?cantdiferencesaretotalassetdepositsthantheaveragecommerisrelevanttostudytheregionalbanksandbankswithhighrelianceonuninsureddeposits,whichweredisproportionatelyafectedbythebankingepisodeinMarch2023.Table1providesadditionalbackgroundinformationforthebanksinoursample.Itdistin-guishesbetweenborrowersandnon-borrowersfromtheBTFP sheetcharacteristicsforthebanksthataccessedtheprogram.Threeobservationsstandout.First, 5ThoughcreditunionsalsoparticipatedmeaningfullyintheBTFP,werestrictourattentiontobanksduetothelackofhighfrequencycredituniondata.6“Core”depositsarede?nedasliquid(non-CD)depositsplussmall(under$100,000)timedeposits.Thisde?nitionincludesbrokereddepositsthathavebeenparticipatedoutinsharesoflessthan$100,000,sowemayunderestimatedepositout丑owstotheextentthatbanksofsetsomeofthedepositout丑owsbyraisi7Thesearedailydata,whichweaggregateattheweeklylevelfollowingthetimingconventioninFR2644forms(i.e,Wednesdaycloseofbusinesslevels).BTFPborrowersaresubstantiallysmallerthannon-borrowers.Ingeneral,smallerbanksarelessandlargerunrealizedHTMsecuritieslossesrelativetoassets.Andthird,possiblyrelatedriskfactors,BTFPFHLBadvancesintherun-uptobankingturmoil.Table1:SummaryStatistics#Institutions($B)CoreDepositGrowth(%)UninsuredDeposittoAssets(%)UnrealizedHTM(%)FHLBadv(%)FHLB(%)FR2644Banks-4.52Borrowers-7.57-4.09B.2023Q1FR2644Banks-9.55Borrowers-16.13-8.63FR2644Banks-9.25Borrowers-8.25-9.39FR2644Banks-0.78Borrowers-1.02Notes:Thetableshowssummarystatisticsfor2022Q4CallReportdataforbanksandthesubsetofbanksincludedintheFR2644,brokendownbytheirrespectiveBTFPborrowerstatus(asof5/3/2023).Theuninsureddepositsshareisde?nedasuninsureddepositsrelativetototaldeposits;unrealizedsecuritieslosses,FHLBadvancesarerelativetototalassets.4Results4.1Whichbanksexperiencedcoredepositoutows?inthreeintervalsoflessacuteforeverydollarofuninsureddepositsatabank.Thisefectissizable:aonestandarddeviationisnotstatisticallysigni?cant.theriskoffailureattheidepositsharewithsecuritieslossesratesensitive.Incontrast,held-to-maturitysecuritieslossesanditsinteractionwithuninsuredHTMSecurityLosses△LoansLiquidAssetsFHLBAdvancesCommunityBankR△Deposits(5)0.00(0.01)(0.11)0.26**(0.08)-0.03**(0.01)0.55**(0.20)(0.01)(0.03)(0.72)(0.74)(0.82)(6)0.00(0.01)(0.01)(0.33)0.26**(0.08)-0.04**(0.01)0.55**(0.20)(0.01)(0.03)(0.72)(0.74)(0.83)(7)(0.01)(0.11)0.22**(0.08)-0.02**(0.01)(0.19)(0.01)(0.02)(0.67)(0.69)(0.77)(8)(0.01)(0.01)(0.31)0.22**(0.08)-0.02**(0.01)(0.19)(0.01)(0.02)(0.67)(0.69)(0.77)(1)-0.04**(0.01)(0.13)0.27**(0.07)-0.02**(0.01)0.60**(0.23)0.04**(0.01)(0.03)(0.79)(0.82)(0.91)(2)-0.03**(0.01)-0.03**(0.01)(0.37)0.26**(0.07)-0.02**(0.01)0.60**(0.23)0.04**(0.01)(0.03)(0.78)(0.81)(0.91)(3)0.01(0.01)(0.11)0.35**(0.07)-0.05**(0.01)(0.21)(0.01)(0.03)(0.76)(0.78)(0.85)(4)0.01(0.01)(0.01)(0.32)0.35**(0.07)-0.05**(0.01)(0.21)(0.01)(0.03)(0.76)(0.78)(0.86)Notes:Thetableshowsresultsfromcross-sectionalregressionsofchangesinbanks’coredepositsrelativetobeginningonperiodtotalassets(△Depositsi;t;t,/Assetsi;t)overthreeeight-weekintervalsonuninsureddeposits,HTMsecuritylosses,contemporaneousloangrowth(△Loansi;t;t,/Assetsi;t),liquidassetholdings(thesumofcash,federalfunds,reverserepoandsecuritiesholdings),averagefundingcost(annualizedinterestexpenseasapercentoftotalliabilities),depositgrowthin2022,FHLBadvances,andbanksizeindicators(i.e.,communitybanksareoneswithassetsunder$10billioninassetsandregionalbanksoneswithassetsbetween$10and$250billioninassets)measuredasof2022:Q4.Coredepositandloangrowtharewinsorizedatthe1and99percentlevels.Allvariablesexceptindicators,depositgrowthin2022,andaveragefundingcostarerelativetototalassets.+,*,**indicatesigni?canceat10%,5%,and1%,respectively.4.2Howdidbanksmanagecoredepositoutows?WholesaleBorrowingfromtheBTFPandPrivateSources:Next,wefocusonhowbanksftotalassets.8inaboutan8centincreaseinBTFPborrowing.RiskfactorsthatwereassociatedwithdepositAnotherreasonthatbankswithgreaterHTMsecuritieslossesmightaccessBTFPfundingisthatsuchlossesmakeBTFPfundingpreferabletootheralternatives,suchasFHLBborrowing.First,highHTMlossesmayindicatethatabank’sassetholdingsarelesincreaseborrowingrathersheddingliquidassetbyvaluingsecuritiesatpar,theBTFPincreasedborrowingcapacitymoreforbankswithgreaterexpectedtomeetmoreoftheirmarginalfundingneedswithBTFP.assesshowmuchbanksdiferintheirpropensitytofunddepositout且owswith(2.4%)wouldreplaceabout26centsoAnimportantquestionremains:Ist 8Variationindiscountwindowborrowingwasminimal.Whilediscountwindowborrowingspikedattheonsetofthestress,byMay3rdthemostseriouslystrainedbankswereclosed,andborrowingreturnedtonearpre-SVBlevels.9Indeed,bankswithgreaterHTMsecuritieslossesdisproportionatelyusedprivateborrowingevenbeforethebankingstressinMarch.borrowing,consistingpredominantlyofFHLBadvances.Thisborrowingwasamofsigni?cant.ThisresultindicatesthatthehighutilizationofBTFPforbankswithgreatersecurityoptingtoborrowfromtheBTFP.△DepositsAssetsHTMSecurityLossesAssetsUninsuredDepositsAssets△LoansAssetsLiquidAssetsAssetsAvg.FundingCostFHLBAdvancesAssetsCommunityBankRegionalBankConstantR△BTFPBorrowingsAssets△Non-FedBorrowingsAssets(1)-0.08**0.43**(2)-0.07**0.02**-0.02*-1.51*(3)-0.09**-0.02*-1.27*(4)-0.26**(5)-0.32**0.25**-0.04**-2.00**3.13**(6)-0.33**0.25**-0.04**-1.98**3.08**Notes:Thetableshowsresultsfromcross-sectionalregressionsofchangesinbanks’BTFPborrowingsandnon-Fedborrowings(wholesaleborrowingsthatexcludeBTFPanddiscountwindow)overtheintervalMarch8—May3oncontemporaneousdepositgrowth(△Depositsi;t;t,/Assetsi;t)uninsureddeposits,HTMsecuritylosses,contempora-neousloangrowth(△Loansi;t;t,/Assetsi;t),liquidassetholdings(thesumofcash,federalfunds,reverserepoandsecuritiesholdings),averagefundingcost(annualizedinterestexpenseasapercentoftotalliabilities),depositgrowthin2022,FHLBadvances,andbanksizeindicators(i.e.,communitybanksareoneswithassetsunder$10billioninassetsandregionalbanksoneswithassetsbetween$10and$250billioninassets)measuredasof2022:Q4.Coredepositandcontemporaneousloangrowtharewinsorizedatthe1and99percentlevels.Allvariablesexceptindi-cators,depositgrowthin2022,andaveragefundingcostarerelativetototalassets.+,*,**indicatesigni?canceat10%,5%,and1%,respectively.Overall,the?ndingsinTable3suggestthatBTFPborrowingwasmotivatedbybothprecau-demandforBTFPloans.pledgingofcollateraltotheBTFP.TableB3presentssimilarestimates,butwiththedependentvariablebeingtheamountcollateralpledgedtoBTFP(asashareofapledgedtotheBTFPbyalmosttwiceasmuchastheyactuallyborrow.10ThisresultprovidesfurtherevidenceofprecautionarymotivesforBTFPusage,asitshowsthatbanksatmoreriskadditionalpledgingwaspredominantlyaccountedforbyMBSandCMOsecurities.11LargeTimeDepositsandLiquidAssets:Howdidbanksadjustbalancesheetitemsbesidesfabankatthe90thpercentileinHTMsecuritylossessourcesaredisplayedinTableB2.Aspreviouslydiscussed,BTFPloansplayedasmallroleforbankswithoutHTMlosses.BankswithoutHTMlossesonlyincreasedBTFPloansplayedalargerroleforbankswithgreaterHTMlosses.BankswithgreaterHTMincreasedsecurityholdingsby4cents).Inshort,privateborrowingorreducingreserves,withBTFPborrowingshavingarelativelylimitedroleintothefundingmix.However,theshifttowardsBTFPfundingforhigh-HTMlossbanksismostlydrivenbyregionalbanks.TheestimatesimplythataregionalbankinthextrapolationtoarelativelyhighlevelofHTM-losses,thisresultprovidessomeusefulclarityonwhichbanksutilizedBTFPmostheavily,namelyregionalbankswithhighHTMlTosummarizetheresultsofthissection,mouseBTFPborrowingappeartodoscashholdings,bankswithhighHTMsecuritylossesreliedmoreheavilyonBTFPfunding,especiallygreaterbene?tfromat-parcollateralvaluation—themaintakeawayisthesame:BTFPprovidedfFigure3:HowBanksFundCoreDepositOut且ows:IncreaseLiabilitiesandReduceAssetsFigure4:HowBanksFundDepositOut且ows,By0(a)RegionalBanks0(b)CommunityBanksRobustnessDeposit且owsarenottheonlydevelopmentafectingbteresting,thecoe伍cientsonLoanGrowthinTabaroughlysymmetricefectonbanks’fundingdecisions:banksincreasedBTFPborrowingby10Growth(thediferencebetweenloangrowthanddepositgrowthasashareofinitialout且ows.FigureB1andTableB4showthatthemainresultsarerobusttothisalternativemeasitPricingMostoftheanalysisofthissectiontreateddepositout且owfpartsoftheportfolio.TableB5estimatesfromthespeci?cationinequation(2)fraise3-monthCDratesbyabout2and7basispoints,respectively.HoonCDsatothertimehorizons.Additionally,theseefectsaresecond-orderrelativetodiferences4.3UsesofWholesaleBorrowings:DynamicsofDeposits,Cash,andLoansFinally,weshiftourattentionfrominvestigatingwhatpromptsbankstoborrowfromBTFPtoaskinghowBTFPfun(3),showinghowcoredeposits(Figure5),cashholdings(Figure6)andloanbalances(Figure7)Topreviewresults,we?ndthatbanksuseBTFPtobuildupcashholdings(consistentwithBTFPhelpingbankstosatisfyprecautionarydemandforliquidity)narrativeproposedinSection4.2,butwithmoredetailonthetiminginvolved.DynamicsofCoreDeposits:Startingwithdepositdynamics,Figure5showsthatbankstendtouseBTFPafterasustainedperiodofweakcoredepositgrowth.Bankswithonedollarhigher10weeks.Theweek0efectgivesthissmallandstatisticallyinsigni?cant.withBTFPhavingsome?xedcostthatinhibitsusedespitefeaturesthatmakeitdesirablerelaFigure5:RelationshipBetweenCoreDepositsGrowthandBorrowing0(a)ChangearoundBTFPBorrowing0(b)Changearoundnon-FedBoDynamicsofCashHoldings:Figure6providessimilaranalysisforcashholdings.Onepoten-usedforthisfunction.IntheweekofBTFPusage,cashholdingsincreasebyaboutathirdoftheFigure6:RelationshipBetweenCashHoldingsandBorrowing0(a)ChangearoundBTFPBorrowing0(b)Changearoundnon-FedBoDynamicsofLoanBalances:Figure7plotschangesinloanbalancessurroundingachangeinborrowings.Overall,thepatternslooksimilarforthetwotypesofborrowings.Banksexperestimatesareforthemostpartstatisticallyinsigni?cant.ThoughbanksarelikelyfundingsopressuresthatwouldcausebdampentherelationshipbetweenBTFPborrowingandloangrowth.Figure7:RelationshipBetweenLoanGrowthandBorrowing0(a)ChangearoundBTFPBorrowing0(b)Changearoundnon-FedBorrowingtheefectivenessoftheBTFP.First,we?ndthatbankswithhighrelianceonuninsureddepositholdingsandborrowedfromnon-Fedsources(e.g.,FHLBadvances).BTFPborrowingplayedaprecautionarymotivefor