美聯(lián)儲-債務(wù)靈活性 Debt Flexibility

FinanceandEconomicsDiscussionSeries

FederalReserveBoard,Washington,D.C.

ISSN1936-2854(Print)

ISSN2767-3898(Online)

Debtlexibility

RhysBidder,NicolasCrouzet,MargaretM.Jacobson,MichaelSiemer

2023-076

Pleasecitethispaperas:

Bidder,Rhys,NicolasCrouzet,MargaretM.Jacobson,andMichaelSiemer(2023).“Debtlexibility,”FinanceandEconomicsDiscussionSeries2023-076.Washington:BoardofGov-ernorsoftheFederalReserveSystem,

/10.17016/FEDS.2023.076

.

NOTE:StafworkingpapersintheFinanceandEconomicsDiscussionSeries(FEDS)arepreliminarymaterialscirculatedtostimulatediscussionandcriticalcomment.TheanalysisandconclusionssetfortharethoseoftheauthorsanddonotindicateconcurrencebyothermembersoftheresearchstafortheBoardofGovernors.ReferencesinpublicationstotheFinanceandEconomicsDiscussionSeries(otherthanacknowledgement)shouldbeclearedwiththeauthor(s)toprotectthetentativecharacterofthesepapers.

Debtflexibility*

RhysBidder?,NicolasCrouzet?,MargaretM.Jacobson§,andMichaelSiemer?

November2023

Abstract

ThispaperdocumentsnewfactsonthemodificationofbankloansusingFRY-14QregulatorydataonC&Iloans.Wefindthatloan-levelmodificationsofkeycontractualterms,suchasinterestandmaturity,occuratleastoncefor41%ofloans.Crosssectionaldifferencesinmodificationsaresubstantialandamplifiedbyborrowerdistress.Relativetosingle-lenderloans,syndicatedloansare1.5timesmorelikelytobemodifiedandin-terestratechangesaretwiceaslikely.Ourfindingscallintoquestionwhether1)creditordispersionmakesloanmodificationsmorechallengingand2)relationshiplendingbe-tweenbanksandsmallborrowerscreatesmorescopeforflexibilitywhenborrower-levelconditionschange.

Keywords:Corporatedebt,Renegotiation,SMElending,Relationshiplending.

JELcodes:G21,G32,G33.

*ThismaterialreflectstheviewsoftheauthorsandnotthoseoftheFederalReserveBoardofGovernors.TheauthorsthankJakeScottandConnorBrennanforexcellentresearchassistance.Additionally,theauthorsthankparticipantsataseminarattheFederalReserveBoardforusefulcommentsandfeedback.

?King’sCollegeLondon,

rhys.m.bidder@kcl.ac.uk.

?NorthwesternUniversity,

n-crouzet@.

§FederalReserveBoard,

margaret.m.jacobson@.

?FederalReserveBoard,

michael.siemer@.

1

1Introduction

Theoriesrationalizingdebtasanoptimalfinancialcontractoftenemphasizetheimportanceofstate-contingentshiftsincontrolandcash-flowrightsbetweenlendersandborrowersaswaystominimizethecostsofstateverificationortostoptheborrowerfromdivertingfunds

(Townsend,

1979;

AghionandBolton,

1992;

HartandMoore,

1994,

1998)

.Consistentwiththesetheories,evidencefromrecentempiricalworkshowsthatdebtcontractsareindeedhighlystate-contingentsecurities,withpayoffsandcontrolrightsthatadjustinaflexiblemannertofirmoutcomes.Thisstate-contingencycanbepartoftheexantecontractual

design,asempiricalworkontheprevalenceofdebtcovenantshasshown(SmithJrand

Warner,

1979;

ChavaandRoberts,

2008;

Nini,SmithandSufi,

2009;

Chodorow-Reichand

Falato,

2022),butitcanalsoresultfromexpostrenegotiationoutsidethestipulationsofthe

initialcreditagreement(RobertsandSufi,

2009;

Roberts,

2015).Mostofthisevidencecomes

fromthesyndicatedloanmarketwheregroupsoflenders,oftenincludingbothbanksand

non-banks,providecredittoasingleborrowerunderacommonloanagreement.1

Whileimportantintermsofvalue,thesyndicationmarketistypicallyrestrictedtorela-tivelylargeandmaturefirmswhosecapitalneedsexceedthecapacityofanysinglelender.Bycontrast,thevastmajorityofcorporateloancontractsinvolveonebanklendingtooneborrower,particularlyforsmallerorlessmatureborrowers.Theliteratureonrelationshiplendinghaslonghypothesizedthatabenefitofsuchlendingarrangementsistheflexibility

bankscanofferborrowersasthefinancialconditionsofthelatterevolve(Boot,Greenbaum

andThakor,

1993;

BootandThakor,

2000;

Boot,

2000)

.Thisflexibilityisfacilitatedbythefactthatthereisasinglecreditorintherelationship,thuseliminatingfree-riderproblems

inherentinadjustingdebtcontractswithmultiplecreditors(BoltonandScharfstein,

1996)

.2

However,thereislittledirectempiricalevidenceonwhetherloansintheserelationshipsareactivelymodifiedorrenegotiatedafterorigination,assyndicatedloansappeartobe.

Inthispaper,weofferaunifiedperspectiveonlenderflexibilitybycomparinglargeandsmallborrowersalongwithawidevarietyofloans.Inparticular,wefillanimportantgapintheliteraturebyprovidingsomeofthefirstempiricalfactsabouttheflexibilityofsinglelenderloanstosmallerfirms.Insodoing,weshowthattherearenotablydifferentpatternsamongtheseloansthanamongthemoreheavilyresearchedsyndicatedloanpopulation.

OurevidencecomesfromtheFRY14-Q,aregulatorydatasetofcorporateloanportfolios

oflargeUSbanks.3

Forourpurposes,theFRY14-Qhastwomainadvantages.First,after

1ForanearlyanalysisoftheUSsyndicatedloanmarket,see

DennisandMullineaux

(2000).Formorerecent

evidenceonitsevolution,see

Berg,SaundersandSteffen

(2021)

.

2Indeed,

DennisandMullineaux

(2000)arguethatsyndicatedloanshavelowerrelationshipintensity,and

aregenerallymore"transaction-oriented"thansingle-lenderbankloans.

3MoreinformationontheFRY-14Q,alongwithreportingguidelinesandforms,isavailableat

https:

///publications/fr-y-14-qas/y-14-qas.htm.

2

theissuanceofaloan,thedatasettrackstheevolutionofkeyloantermsandrenegotiationeventsataquarterlyfrequency.Second,thedatasetcontainsbothsyndicatedandsingle-lenderloans,allowingustocomparetheflexibilityoftheselendingrelationships.

Wehavefivemainempiricalfindings.First,consistentwithexistingevidenceforthesyndicatedloanmarket,wedocumenthighloanmodificationratesoverall:afterorigination,41%ofallloansexperienceatleastoneexpostmodificationtoeithertheinterestrate(orspread,forfloatingrateloans)orthematurity.Aroundaquarterofloansexperiencetwoormoremodificationevents.

Second,syndicatedandsingle-lenderloansexhibitsharpdifferencesinthefrequency,type,andtimingofloanmodifications.Unconditionally,55%ofsyndicatedloansaremodi-fiedatleastonce,and37%aremodifiedmorethanonce,whiletheanalogoussharesare37%and20%,respectively,forsingle-lenderloans.Inaddition,modificationstosingle-lenderloansoverwhelminglyconsistofmaturityextensions.Thesetendtooccuronlyonce,andtypicallytowardstheendoftheinitiallycontractedmaturityperiod.Incontrast,syndicatedloanstendtoexperienceinterestrateadjustmentsoccurringthroughoutthelifeoftheloan.

Third,weshowthatmodificationratesrespondtotheonsetoffinancialdistressaslenderstendtogivefinancialslacktoborrowerswhenfinancialconditionsdeteriorate.Butthispropensitytomodifyloansinresponsetofinancialdistressissubstantiallylargerforsyn-dicatedthanforsingle-lenderloans.Forthesamechangeintheinternalratingoftheloan,andcontrollingforobservablecharacteristics,theincrementalpropensityforabankinoursampletomodifyinterestratesisapproximatelytwiceaslargeforasyndicationasforasingle-lenderloan.

Fourth,notwithstandingthehigherratesofmodificationsintimesofdistress,thelion’sshareofadjustmentsoccuroutsidesuchepisodes.Diggingdeeper,wediscusswhetherthefrequentmaturityextensionsthatoccurforsingle-lenderloansarerelatedto"evergreening",i.e.banksrollingovertroubledloanstoavoidacknowledgingrealizedlossesontheoriginalcredits.Whileloansindistressareindeedmorelikelytobeextendedthanaverage,themajorityofextensionsoccurforloansthatdonotshowsignsofdeterioration.

Finally,weexplorewhethercovenantviolationsaccountforthehighrateofloanmodifi-cationsobservedamongsyndicatedloans.Assuch,wemergethesyndicatedloansubsamplewithinformationoncovenantsfromtheSharedNationalCredit(SNC)register.Eventhoughmostcovenantviolationsrecordedinthemergedsampleleadtoaloanmodification,weshowthatmostmodificationsoccurwithoutanycovenantbeingviolated.Thus,themodifi-cationswedocumentforsyndicationsarenotalltheresultofexantecontractualprovisions.

Asimplebutpotentiallypowerfulexplanationforthefactthatsingle-lenderloansaremodifiedfarlessoftenthansyndicationsisthatexpostmodificationsinvolvecoststhatdonotscalewiththesizeoftheloan.Weshowthatthisexplanationhassomeempiricalpower,

3

inthatthereisaclearupwardslopingrelationshipbetweenloanmodificationratesandloansize,evenaftercontrollingforotherobservablecharacteristicsoftheborrower,andforlenderandsectorfixedeffects.However,wealsoshowthatthispatterndoesnotentirelyexplainwhysingle-lenderloansarelesslikelytobemodified.Therefore,factorsbeyondfixedcostsmayaffectthedegreeofflexibilitythatbanksarewillingtoprovidetheirborrowers.Wehypothesizethatforsmallbusinesslending,banksinoursampleprimarilyrelyonstan-dardizedproductsandhardinformation—astheydoinconsumercreditmarkets—whereas

syndicatedloansmakegreateruseofsoftinformationandtailoredcreditagreements.4

Thebroaderimplicationofourfindingsisthatflexibilityanddiscretionarelessfrequentlyexercisedinsingle-lenderborrowingsituationsthaninsyndications.Thatis,thesyndicatedmarketmaybebestdescribedasinvolvingaformofrelationshiplendingwithflexibility,whereasthesingle-lenderloanmarketmaybestbedescribedasinvolvingaformofarm’s-lengthlendingwithoptionstoextendtheloanuponmaturity.Thus,inthedatawestudy,theflexibilitybenefitsthatsingle-banklendingrelationshipshaveoftenbeenhypothesizedtogenerateforsmallbusinessesappeartobelimited,particularlywhencomparedtothoseaffordedtoborrowerswithsyndicatedloans.

RelationtotheliteratureWhiletherehasbeenextensivetheoreticalworkonfactorsthat

facilitaterenegotiationanditsimpactsonexanteincentives(BerlinandMester,

1992;

Bolton

andScharfstein,

1996;

KiyotakiandMoore,

1997;

DiamondandRajan,

2001;

Garleanuand

Zwiebel,

2009;

Zhong,

2021;

GlodeandOpp,

2023),empiricalevidenceontheincidence

ofloanflexibilityismorescarceandrestrictedtocertaintypesofdebtcontracts,either

bonds(Asquith,GertnerandScharfstein,

1994)orsyndicatedloans(RobertsandSufi,

2009)

.Thepaperclosesttoourworkis

Roberts

(2015),whichusesasampleofcreditagreements

collectedfrompublicfilingsofsyndicatedlenderstotraceouttheevolutionofcontractualtermswithinindividualloans.

Relativetothisliteratureontheflexibilityofdebtcontracts,wemakeseveralcontribu-tions.First,weprovidesystematicevidenceonthenatureandfrequencyofloanmodifica-tionsforalargecross-sectionofborrowers,notlimitedsimplytoparticipantsinthesyndi-catedloanmarket.Ourextensionbeyondthesyndicatedmarketisimportant,asweshowthatsyndicationsseemtoprovidemoreflexibilitytoborrowersthansingle-lenderloans.Thissuggeststhatcreditordispersionmaynotnecessarilyimplyalowerlikelihoodofsuc-cesswhenattemptingtomodifyloans.Furthermore,thenatureofadjustmentstoloansalsodiffersbetweensyndicatedandsinglelenderfacilities.

Inaddition,wedemonstratethatflexibilityismanifestedinabroadervarietyandawider

4Consistentwiththishypothesis,weconfirmthatinternalratingsaremoreresponsivetochangesinfinancialinformationaboutborrowersforSDthanforSLloans.

4

setofcontingenciesthanhithertoconsidered.Ourworkrevealsahierarchyofflexibility:

modificationsthatdonotrequiretheexecutionofanewcreditagreementversusthosethatdo.Inbothcases,asubstantialfractionofflexibilityoccurswithoutreferencetocovenantviolationsandcanthereforeberegardedasexpostrenegotiationorincompletenessintheoriginalcontract.Furthermore,loanmodificationsinourdataarenotnecessarilycompelledbyfinancialdistressandtypicallydonotoccurintimesofbankruptcyorwherebankruptcyconcernsarehigh(unsurprisingly,theyaremorecommoninsuchsituations).

Wecontributetotheimportant—andrecentlyactive—literatureoncovenantsindebtcontractsbyshowingthatloanmodificationsareacommonoccurrenceevenoutsideepisodes

wherecovenantsare(orlikelytobe)triggered.Analysesofcovenantsinclude(SmithJrand

Warner,

1979;

BeneishandPress,

1993;

ChavaandRoberts,

2008;

Nini,SmithandSufi,

2009,

2012;

Murfin,

2012;

BradleyandRoberts,

2015)andarecentreinvigorationby

Chodorow-

ReichandFalato

(2022).Thisextantworkhasshownthatcovenantviolations,orthethreat

thereof,arecommoneventsoverthelifeofsyndicatedloansandoftenleadtomodificationsintheloan’scontractualtermsthatinturnhaveconsequencesforbothfinancialandrealborroweroutcomes.Theprevalenceofmodificationsamongsingle-lenderloans,whicharelesslikelytobegovernedbycomplexcreditarrangementssuchascovenants,alsosuggeststhatthecovenantchannel,whileimportant,isnotthefinalword.

Furthermore,wedemonstratethatforlargesyndications,themagnitudesofadjustmentstoloantermsinmodificationsoutsidecovenantviolationsaresubstantialandcomparabletothoseimplementedaroundthetimeofviolations.Thus,totheextentthatcovenantsareregardedasimportantsourcesofadjustmentsinloanterms—asemphasizedby

Chodorow-

ReichandFalato

(2022)—theadditionalsourcesofflexibilityweidentifyarealsoworthyof

notesinceloanmodificationsarealsomorefrequentthancovenantviolations.Together,thisevidenceoncovenantsimpliesalowerboundonthestate-contingencyofloans.

Finally,wecontributetotheliteratureonrelationshiplending(LelandandPyle,

1977;

Diamond,

1984,

1991;

Boot,GreenbaumandThakor,

1993;

BergerandUdell,

1995,

1998;

Boot

andThakor,

2000;

BergerandUdell,

2002,

2006;

Boltonetal.,

2016;

Papoutsi,

2021;

Faria-e

Castro,PaulandSánchez,

2021)

.Ourcontributionistochallengethenotionthatexpostdiscretionisanimportantbenefitofrelationshiplendingcontractsforsmallbusinesses,whodisproportionatelyarerepresentedinoursingle-lenderloandata.Ifanything,ourevidencesuggeststhatsingle-lenderrelationshiplendingislessconducivetoexpostdiscretionthanlendingthroughorganizedsyndicates.Inthisrespect,thepapersclosesttooursare

Cole,

GoldbergandWhite

(2004)and

Bergeretal.

(2005),whichbothhighlighttheorganizational

difficultiesthatlargebanksfaceinscalinguptraditionalrelationshiplendingmodels,andwhichmayresultinreducedflexibilityforborrowers.

5

2Dataconstruction

SourcesOurmainsourceistheFederalReserve’sFRY-14Q,whichcontainsdetailedquar-terlydataontheloanportfoliosofbankholdingcompanies,savingsandloanscompanies,andintermediateholdingcompanies.Becausethesedataarecollectedprimarilyforstress-testing,onlyinstitutionswith$100billionormoreintotalconsolidatedassetsarerequiredtoreporttheFRY-14Qschedules.Thus,ouranalysisspeakstolendingbylargebanks.

WeuseScheduleH.1whichcontainsfacility-leveldataoncorporateloansandleases.Thepopulationislimitedtoloanswithacommittedbalancehigherthan$1millionasofthe

reportingquarter.5

Ourbasicunitofobservationisaloan-quarter.Weobserveinformationoncommittedexposure,interestrate(orspread),collateral,maturitydate,andwhetherthecreditagreementwasrenewedorrestatedintheprecedingquarter.Additionally,summariesofborrowerfinancialinformationandinternalloanratingsareavailable.MoreinformationonkeyvariablesinScheduleH.1isreportedinAppendix

A.1.1.

Weusetwoadditionaldatasources,whichwemergewiththeFRY-14QH.1:dataonthepopulationofchapter11andchapter7bankruptciesintheUS;anddataoncovenantsandcovenantviolationsforthesubsampleofsyndicatedloanswithinoursample,fromtheSharedNationalCredit(SNC)program.MoreinformationonthesedatasourcesandourmergewithScheduleH.1arereportedinAppendices

A.1.2

and

A.1.3.

Otherdatasourceshavebeenusedtostudyhowloansaremodifiedafterorigination,particularlyinthewakeofcovenantviolations.

ThesesourcesincludeDealscan(Roberts

andSufi,

2009),hand-collecteddatafromSECfilings(Roberts,

2015),andtheSNCdata

(Chodorow-ReichandFalato,

2022)

.Relativetothesesources,theFRY-14Qhastwomainadvantagesforstudyingloanmodifications:itcoversawiderswathofthefirmpopulation(includingnon-syndicatedborrowers)anditallowresearcherstotrackloantermscontinu-ouslyafterorigination,makingitstraightforwardtoobservemodificationstokeyloantermsthroughthelifeoftheloan.OnedrawbacktotheFRY-14Qisthatitcontainsnoinformationoncovenants.However,bymergingtheFRY-14QwiththecovenantreviewsubsampleoftheSNCdata,weobservecovenantsdirectlyforasubsetofsyndicatedloans.Asdiscussedin

Chodorow-ReichandFalato

(2022),loansinthissubsetareaccompaniedbyadditional

informationoncovenantcomplianceandbankresponsestoanyviolations.

SampleselectionandloanclassificationOurloan-quarterleveldatasetrunsfrom2012:Q3to2022:Q1.Weapplyanumberofselectioncriteriatotherawdata.Inparticular,wedrop

5ReportinginScheduleH.1isaggregatedatthecreditfacilitylevel,definedasacreditextensionunderaparticularcreditagreement.Acreditfacilitymaybedrawnorundrawn,itmayconsistofatermloanoracreditline,itmaybesecuredorunsecured,andsoon.Whenthesamefacilityprovidesformultipleextensionsofcredit(forinstance,creditlineswithdifferentinterestratesthatmaybedrawnatdifferentdates),thesemultipleextensionsareaggregated.Anobligormayhavemultiplefacilitieswiththesamebank.

6

anyloanstoborrowersinthefinancialandrealestate,construction,andutilitiessectors;

loanscarriedoff-balancesheetbyreportingbanks;loansforwhichthereisagapinthesequencebetweenthefirstandlastobservations;andloanswithmissingordifficulttointer-

pretinformationonkeyloanterms.6

Thefinalsamplecontains3.3mobservationsfor337kloansto136kdistinctborrowers.AppendixTable

A-1

describessampleselectionstepsindetailandreportsborrower-levelsummarystatisticsinthefinalsample.

Inwhatfollows,wewillfrequentlycontrasttwogroupsofloans:single-lender(SL)loans,inwhichasinglebankisprovidingcredittoaborrower,andsyndicated(SD)loans,inwhichagroupoflenders,oftenincludingbothbanksandnon-bankinstitutions,providecredittoaborrowerunderacommonloanagreement.ScheduleH.1asksbankstoreportwhethertheloanisaparticipationinasyndicatedfacility,andweusethisflagtopartitionourloans

intothesetwogroups.7

Becausedifferenttranchesofthesamesyndicatedfacility(say,creditlineandtermloan)mayhavedifferentcontractualterms,weincludetranchesasseparate

observationsinoursample.8

Wenotethatourdatawillgenerallynotcontaininformationonalltranchesofagivensyndication,sincetheymaybeheldbyinstitutionsnotsubjecttotheFRY-14Qreportingrequirement.Appendix

A.1.1.3

providesadditionaldetails.

SummarystatisticsPanelAofTable

1

reportssummarystatisticsontheloansinoursam-pleatthetimeoftheirorigination,forallloansjointly,andforSLandSDloansseparately.SLloansmakeupabout74%ofthetotalnumberofloansinoursample,thoughasof2019:Q4,theyonlyaccountedforabouttwo-thirdsoftotalcommittedexposurebyvalue.

PanelAofTable

1

showsthreemaindifferencesincontractualtermsbetweenSLandSDloans.First,SLloansaresmaller:themedianSLloanhasacommittedexposureatoriginationapproximatelyone-ninthofthemedianSDloan.Second,SLloansaremorelikelytobesecured,thoughthemajority(72%)ofSDloansarebackedbysomecollateral.Third,39%ofSLloansarefixed-rate,whileonly6%ofSDloansare.

WealsonotethattheaveragematurityatoriginationishigherforSLloans.Here,thesummarystatisticsreportedinTable

1

masksubstantialvariationbetweenthetwoloangroups.AppendixFigure

A-2

showsthatthedistributionofloanmaturityatoriginationhasalargemassatfiveyearsforSDloans.Incontrast,forSLloans,short-term(one-year)creditlinesareveryprevalent,asaretermloanswithmaturitybeyondfiveyearsatorigination.Wewilltakethisheterogeneityintoaccountinouranalysisbycomparingoutcomesofloans

6Forinstance,aloanforwhichinformationonmaturityisalwaysmissing,orforwhichtheoriginationdatedeclinesacrosstwoconsecutivereportingperiods.

7Specifically,weusetheFRY-14QvariableparticipationflagtoclassifyaloanassyndicatedifitslastobservationintheFRY-14Qisnot“No,notparticipatedorsyndicated".Weexcludeloansthatneverexitthesyndicationpipeline.Appendix

A.1.1.3

providesmoredetails.

8However,wehaveconductedrobustnesschecksthatrestrictthesyndicatedsampletotranchesidentifiedinScheduleH.1asbeingheldbytheagentorleadbankofthesyndication.

7

ofsimilarmaturitiesatorigination.

Finally,AppendixTable

A-2

reportssummarystatisticsattheborrowerlevel,andcom-paresthesesummarystatisticswiththeirequivalentinCompustatdataforthesubsampleofpublicborrowersinourdataset.LoansofCompustatborrowersare,onaverage,morelikelytobesyndicated,creditlines,floatingrate,unsecured,andhavemorefavorablefinancingtermsthanloansofnon-Compustatborrowers.

3Empiricalfindings

Thissectiondescribesourmainempiricalfindings.Westartwithunconditionalfactsaboutloanmodifications.Wethendiscusshowmodificationsrelatetoborrowerfinancialdistress.

3.1Unconditionalfactsonloanmodifications

DefinitionWedefinealoanmodificationasachangeintheloan’sreportedmaturitydate,ortheloan’sinterestrate,orboth,fromreportingquartert?1toreportingquartert.

Twothingsareworthnotingaboutthisdefinition.First,oursamplecontainsbothfloating-rateandfixed-rateloans.Forfloatingrateloans,theinterestratespreadisreported

explicitlywhichallowsustodetectinterestratemodifications.9

Forfixedrateloans,weusechangesintheoverallreportedinterestrate,asopposedtoachangeinthespreadmeasurereportedinTable

1.

Second,ourdefinitionofloanmodificationsonlyinvolvestwoofthefourkeyloantermsthatareobservableinthedata(interestrate,maturity,collateral,andcommitment).Thisisbecausechangesincommitmentorcollateralarerareandgenerallyoccuralongsidechangesineitherinterestrateormaturity,aswedocumentbelow.Focusingoninterestrateormaturitychangeshelpsstreamlinethediscussion.

Ourdefinitionofloanmodificationsisbasedonchangesinloantermsoverthelifeoftheloan.However,notallofthesechangesarenecessarilyassociatedwithamajoramend-menttotheexistingcreditagreement.Majoramendmentstoexistingcreditagreementsare

flaggedbyreportingbanks.10

Werefertotheseeventsasre-originationsanddiscusstheirrelationshiptoloanmodificationsinSection

4.1.2.

ModificationfrequencyOurfirstsetoffindings,onthefrequencyofloanmodifications,isreportedinPanelBofTable

1.Thistableisconstructedusingthedatacollapsedtotheloan

levelandreportsthedistributionofthenumberofmodificationsperloan.

9Incaseswheretheinterestrateonafloatingrateloan–––thesumofthebaseplusthespread–––issubjecttoaceilingorfloor,wedonotcountanadjustmentofthespreadinordertomeetthefloorasamodification.

10Bankswillexplicitlyupdatetheinitiallyreportedoriginationdateoftheloaninthesecases,acknowledgingthesubstantivenatureofthechange.

8

TherearetwomainpointstonoteinPanelBofTable

1.

First,unconditionally,41%ofloansexperienceamodification–––achangeineithermaturitydateorinterestrate–––overthecourseoftheirlife.Themostcommonpatternisfortheloantobemodifiedonlyonce

afterorigination.11

Second,SDloansareapproximately1.5timesmorelikelytobemodifiedthanSLloans,at55%and37%,respectively.Table

1

alsoshowsthatthebulkofthedifferenceisaccountedforbyloansthatundergoseveralmodifications:26%ofSDloansundergothreeormoremodifications,whileonly13%ofSLloansdo.

AppendixFigure

A-1

reportsatime-seriescorrespondingtoPanelBofTable

1.Here,we

usethefullloan-quartersampleandcomputethefractionofloansundergoingamodifica-tionineachreportingquarter.ConsistentwithTable

1,inanygivenquarter,thefractionof

SDloansundergoingamodificationissubstantiallyhigher(about20%,onaverage)thanthe

fractionofSDloansundergoingamodification(about10%,onaverage).12

Thesemodifica-tionratesarestableoveroursampleperiod,exceptforaspikeinmodificationratesin2020,

particularlyforSDloans,whichwasfollowedbyaslightincreasesincethen.13

Thus,overall,modificationsinkeytermsafteroriginationisaverycommonoccurrence.Moreover,relativetoSLloans,SDloansaremorelikelytoexperiencemodifications—andrepeatedmodi