ContentslistsavailableatScienceDirect
North
American
Journal
of
Economics
and
Finance
Does
the
bank
risk
concentration
freeze
the
interbank
system?
Marcella
Lucchetta
∗Universita’Ca’FoscaridiVenezia,DipartimentodiEconomia,Italy
a
r
t
i
c
l
e
i
n
f
o
Articlehistory:
Received19May2014
Receivedinrevisedform2April2015
Accepted2April2015
Availableonline17April2015
Keywords:
Interbanksystem
Risksconcentration
a
b
s
t
r
a
c
t
Probably,onetestofthestabilityofthebankingsystemisto evalu-atehowriskyassetsaredistributedacrossbanks’portfoliosandthe implicationsforthecontagionviainterbankrelations.Thispaper explorestheoreticallyabanksectorwithrisksconcentrationand thefunctioningofinterbankmarkets.Itemploysasimplemodel wherebanks areexposedtobothcreditand liquidityriskthat suddenlycorrelateoverthebusinesscycle.Weshowthatrisk con-centrationmakesinterbankmarketbreakdownsmorelikelyand welfaremonotonicallydecreasesinriskconcentration.
©2015ElsevierInc.Allrightsreserved.
1. Introduction
Akeyfeatureofthe2007–2008financialcrisishasbeenthedisruptionandprolonged malfunc-tioningofinterbankmarkets(see,e.g.Acharya&Merrouche,2013;Afonso,Kovner,&Schoar,2011; Ciccarelli,Maddaloni,&Peydró,2013;Heider,Hoerova,&Holthausen,20101)sometimesrelatedtothe interbanknetworkstructure(Georg,20132).Thishascomeasasurprisetomostobservers,since inter-bankmarketshavebeenfunctioningsmoothlyhistorically,eveninthefaceofseverestressepisodes
∗ Tel.:+393356955599.
E-mailaddress:[email protected]
1Ciccarellietal.(2013)findthatfinancialintermediariesareextremelyfragileintheEU.TheECBeffectivelypartlysubstituted
theinterbankmarketand,inturn,inducedasubsequentsofteningoflendingconditions.
2Georg(2013)showsthatnetworkswiththecentralbankthatintervene,solvingmarketincompleteness,aremorestable
thanrandomnetworks.
http://dx.doi.org/10.1016/j.najef.2015.04.002
suchastheLTCMfailure(Furfine,2000)anditattachesimportancetopolicyinterestrateforloan portfolioriskandbankliquidityasrecentfoundinGiulioni(2015).
Theoreticalresearchaddressingwhyinterbankmarketsmaynotfunctionproperlyhasprovided explanationsbasedoninformationalandmarketfrictionssuchas:asymmetricinformationor mar-ketincompleteness(e.g.Acharya&Skeie,2011;Flannery,1996;Freixas&Jorge,2008;Georg,2013; Gale&Yorulmazer,2012;Heideretal.,2010);informationcontagionwherethepoorperformance ofeachbankconveyspotentialbadnewsaboutthecommonfactoraffectingloanreturns(Acharya& Yorulmazer,2008);marketpower(e.g.Acharya,Gromb,&Yorulmazer,2012;Cai&Thakor,2008);and malfunctioningsecondaryassetmarkets(e.g.Diamond&Rajan,2005,2008;Gorton&Huang,2004, 2006).
Since theearly2000s,however, therehasbeenincreasingevidenceof a positiverelationship betweenmeasuresofsystemicriskinmajorbankingsystemsandbankconcentration.3Thisevidence raisesthequestionofwhetherthecorrelationofriskspossiblyinducedbyhigherbankrisk concen-trationcouldhaveasignificantimpactonthefunctioningofinterbankmarketsandwelfare,andifso, why.Toourknowledge,thisquestionhasnotbeenaddressedintheliterature.Arecentstudydiscuss theinternationalportfoliodiversificationofMiralles-Marcelo,delMarMiralles-Quirós,and Miralles-Quirós(2015),butasidefromtheinterbankmarketandthecorrelationbetweenliquidityriskand creditrisk.
Specifically,inthispaperweexploretheimplicationsofbankingsystemriskconcentrationforthe functioningofinterbankmarketsinamodelwherebanksareexposedtobothcreditandliquidityrisk. Indeed,asshowinCornett,McNutt,Strahan,andTehranian(2011),Brunnermeier(2009),Covitzand Downing(2007)andEricssonandRenault(2006)creditandliquidityriskmaydramaticallycorrelate overthebusinesscycleespeciallyduringfinancialcrisis.Moreover,DaSilvaandDivino(2013)show thatcreditriskispro-cyclicalanddefaultriskdependsonstructuralfeatures,underliningthebanking regulationroleinpresenceofcreditandliquidityshocks.
Thelineofinquiryofthispaperisrelatedto,andbuildson,thecontributionsbyIbragimov,Jaffee, andWalden(2011)andWagner(2010,2011),whoshowthatinefficienciesmayarisefromindividual bankdiversification,whichdoesnotnecessarilyresultinamoreresilientbankingsystem.However, thesepapersdo notconsiderinterbankmarketsand multiplesourcesofrisks,suchascreditand liquidityrisk,aswedo.4
WebuildasimplemodelalongthelinesofDiamondandDybvig(1983)andBhattacharyaandGale (1987)models,wherebanksareexposedtobothcreditandliquidityrisk,andtherearenoinformational ormarketfrictions.Themarketfailureinoureconomyisthatcontractsareincompleteandtherefore, notallriskscanbeinsured.Interbankmarketbreakdownsaredefinedasparameterconfigurations underwhichthereisnointerbankmarketequilibrium,andbanksimplementautarkicallocations.
Weshowthatanincreaseintheconcentrationofrisks,possiblyarisingfromconcentratedmarket structures,makesinterbankmarketsbreakdownsmorelikely.Differingfromthepreviousliterature, ourresultsarenotdrivenbyasymmetricinformation,marketpowerordysfunctionalsecondary mar-kets.Rather,theyareexplainedbycreditandliquidityriskscorrelationassuddenlyhappensoverthe businesscycle.Asanexample,theabilitytodiversifytheserisksmaybepreventedbyriskmanagement diseconomiesassociatedwithlargesizesoffinancialinstitutionsandthewidescopeandcomplexity oftheiractivities.Indeed,systemicallyimportantfinancialinstitutions(SIFIs)areseenasinstitutions
3GroupofTen(2001)concludedthat“Evidencesuggeststhat[risk]interdependenciesbetweenlargeandcomplexbanking
organizationshaveincreasedoverthelastdecadeintheUnitedStatesandJapan,andarebeginningtodosoinEurope.Although
acausallinkhasnotbeenestablished,theseincreasesarepositivelycorrelatedwithmeasuresofconsolidation.”Apositive
relationshipbetweenbankconcentrationandmeasuresofbanksystemicriskisfoundinBoyd,DeNicolò,andLoukoianova
(2009a)andBoyd,DeNicolò,andJalal(2009b).Duringtheperiodsofintenseconsolidationofthelastdecade,DeNicolòand Kwast(2002)foundincreasedriskinterdependenciesamongU.S.largeandcomplexbankingorganizations.Riskprofilesof
largeandcomplexU.S.andEuropeanbankswerealsofoundtohaveincreasedintheU.S.inEurope,andgloballyinDeNicolò,
Hayward,andBhatia(2004a),DeNicolò,Bartholomew,Zaman,andZephirin(2004b),Stiroh(2004),Hartmann,Straetsman, anddeVries(2005),StirohandRumble(2006),andHoustonandStiroh(2006).
4ThispapercanbeviewedasalsoindirectlyrelatedtothepapersbyGaietal.(2011)andAnandetal.(2012),astheyanalyze
numericallyhownetworkconnectionsanddifferentexogenousinterbankstructuresaffectbanks’shorttermfundingbutdo
tobecarefullymonitoredbyregulatorasapotentialsourceofsystemicrisk(seee.g.ECBFinancial StabilityReview,December2009).
InSection2wedescribethemodel.Themodelset-upcapturesinstylizedformtheimportant distinctionbetweendiversificationofabankingfirmanddiversificationofthefinancialsystem empha-sizedanddocumentedempiricallybyDeNicolòandKwast(2002).Bankscanbeperfectlydiversified individually,butaggregateriskinthebankingsystemcanbeeitherperfectlydiversifiedor concen-tratedacrossbanks.Asthesystembecomescomposedoffewerandlargerbanks,eachbankwillbemore diversifiedindividually,butthebankingsystemwillbelessdiversified,sincealargerfractionofbanks isexposedtothesameaggregateshocks.Inthemodel,thedegreeofbankingsystemdiversificationis parameterizedbetweenthetwoextremesofperfectdiversificationandmaximalconcentration.
Section3definestheinterbankequilibrium.Interbankmarketbreakdownsaresimplydefinedas situationsinwhichtheinterbankequilibriumdoesnotexistandisreplacedbytheautarkic equilib-rium,whereeachbankisdisconnectedfromeachother.
InSection4,theexistenceofinterbankequilibriaisestablishedforthetwoextremecasesofa perfectlydiversifiedandamaximallyconcentratedbankingsystem.Then,itisshownthatforalarge setofeconomies,thesizeofthesetofinterbankequilibriaunderadiversifiedbankingsystemisalways strictlylargerthanthatofaconcentratedbankingsystemforanylevelofcreditandliquidityrisk.This resultindicatesthatinthepresenceofaggregaterisk,adiversifiedbankingsystemislikelytobeless pronetointerbankmarketbreakdowns.
Section5definesthewelfarepropertiesofinterbankandautarkicallocations.Whenaninterbank equilibriumexists,itgivesahigherexpectedutilitytodepositorsthantheautarkicequilibrium,but importantly,depositors’welfareismonotonicallydecreasinginthedegreeofriskconcentration.
Section6providesanexampleextendingthemodeltocapturetheShehzadandDeHaan(2013) findings.Theirresultssuggestthatstockpricesoflargebankswereaffectedmoreduringthecrisis thanthoseofsmallbanks.Theyalsofindthatmanagerialefficiency,loanquality,leverage,andthe volumeofoutstandingloansaffectbankstockprices.Thus,inourextension,riskconcentrationand ahighprobabilityofinterbankmarketbreakdownsisduetoriskmanagementdiseconomiesofscale andscopeaffectinglargeandcomplexfinancialfirms.Specifically,whenourmodelismodifiedby introducingariskcontroltechnologywithdecreasingreturnsoveracertaininvestmentthreshold, bankswillchoosealevelofriskconcentrationthatincreasesinsize.Suchlevelishigherthelargeris thecostofriskcontrolarisingfromtheinternalorganizationoflargeandcomplexfirms.Thisresult suggestsimprovementofriskcontroltechnologiesinlargeandcomplexfinancialinstitutionsandin regulatorybodiesmaybepolicyconcernsasimportantasotherpolicesconsideredintheliteratureto minimizetheprobabilityofinterbankmarketbreakdowns.
Section7concludes.ProofsofallpropositionsareintheAppendix.
2. Themodel
Therearethreeperiods,t=0,1,2,andoneriskyassetthatyieldsarandomreturnatdate2perunit investedatdate0.Itcanassumetwovalues,R=Rh,andR=Rl,withtheprobabilitiesspecifiedbelow.
Ifaportionoftheinvestmentinthisassetisliquidatedatt=1,ityieldsacertainreturnofperunit invested.Thefractionoftheassetthatisliquidatedisdenoted˛.ItisassumedthatRh>R>≥1l,5so thatifstorageisavailable,thistechnologywillbedominatedinrateofreturninbothdatesandwill beneverused.Thisisnotastrongassumptionsinceinthespecialcase=1,thelongassetmimicsthe storagetechnologyandcanbeliquidatedtosatisfyearlyconsumers,aswedefineshortly,ortolend intheinterbankmarket.
Atdate0consumersareendowedwithoneunitofthedate0consumptiongood,whichisassumed tobeinvestedallinonebank.Consumersareuncertainabouttheirtimepreferences:withprobability theyareearlyconsumers,whowanttoconsumeatdate1only,andwithprobability1−they
5Thisimpliesthattheexpectedvalueofthelongassetisgreaterthanone(thereturnofthestoragetechnology).Furthermore,
arelateconsumers,whowanttoconsumeatdate2only.6Weassumethatdepositorshavelimited participationintheassetsmarkettherebytheyneedbankingintermediation.7Theirpreferencesare representedbyautilityfunctionU(c),twicecontinuouslydifferentiable,increasing,andstrictly con-cave.Thefractionofearlyconsumersisalsorandomandcanassumetwovalues:=h,and=l,
with1>h>l>0.
ThebankingsectoriscomposedbyNex-anteidenticalbanksthatinvestconsumer’sendowmentsat date0.Thebankingsectorisperfectlycompetitive,sothatbanks’objectiveistomaximizedepositors’ expectedutility.
Eachbankisexposedtoliquidityandcreditriskshocks.Therealizationofbothshocksisobserved byabankatdate1.Weassumethereisnoaggregateuncertainty,sothatthefractionofbankswhich isexposedtoagivencombinationofcreditandliquidityshocksisdeterministic.However,the prob-abilities(andhencethedistribution)ofthesebanksatdate0dependsonanexogenousparameter ∈(0,1)thatindexesthedegreeofconcentrationofliquidityandcreditriskinthebankingsystem.
Specifically,thefractionofbanksthatexperienceagivenpairofrealizationsofcreditandliquidity risk(R,)isgivenbythefollowingtable,withp∈(0,1)andq∈(0,1),wherepistheprobabilityofa banktohaveahighreturnandalowliquidityshockandqistheprobabilityofabanktohaveahigh returnandahighliquidityshock.Theseprobabilities,pandq,areindependent.
Thus,asofdate1,therearefourtypesofbanks:
Thereafterwerefertoparameterasanindexof“riskconcentration”.
Afractionofbanksp(type1)experiencesalowliquidityshockandahighfinaldatereturnonthe asset;bycontrast,afractionofbanks(1−p)(type4)experiencesthereverse,thatis,ahighliquidity shockandalowfinaldatereturnontheasset.Thus,creditrisksandliquidityrisksareperfectlypositively correlatedfortype1and4banks.
Conversely,afractionofbanks(1−)q(type2)experiencesahighliquidityshockbutalsoahigh finaldatereturnontheasset,whereasafractionofbanks(1−)(1−q)(type3)alowliquidityshock butalsoalowfinaldatereturnontheasset.Thus,creditriskandliquidityrisksareperfectlynegatively correlatedfortype2and3banks.
If=0,thenthebankingsystemismadeofbanksthatdiversifytheircreditandliquidityrisks.By contrast,if=1,banksconcentratetheircreditandliquidityrisks.Asincreases,thebankingsystem exhibitsahigherconcentrationofrisksforanygivenvaluesof(p,q).
3. Interbankequilibrium
Hereweexaminehowwellaninterbankmarketworksforanylevelofriskconcentration(i.e.any ).
Thereisaninterbankmarketwhereliquiditycanbetradedattheintermediatedate.Theamount offundsthateachbanktradeintheinterbankmarketisdenotedbandthegrossinterbankrateis denotedbyr.
6When=1,inthiscaseearlyconsumersreceive1suchthatinexpectationtheincentivecompatibilityconstraintissatisfied:
i.e.lateconsumersreceiveE[R]>1.
Atdate0,competitivebanksmaximizetheexpectedutilityofdepositors.Theychoosetheamount ofborrowingb(ifpositive,borrowing,ifnegative,lending)andtheamountofassettoliquidate,˛,to solve: Max ˛,b˘1=U(c1)+(1−)U(c2) (1) Subjectto c1=˛+b (2) (1−)c2=R(1−˛)−rb, (3)
whereristhe“interbank”rate,tobedeterminedinthet=1creditequilibrium. Assumeaninterbankequilibriumexists.Substituting(3)in(2)throughb,
c1+(1−)c2
r =˛+ R
r(1−˛). (4)
Thesolution˛*willbetheonethatmakesthebank’sbudgetconstraintthelargest,i.e.thatmakes
therighthandsideof(4)thelargest.Hence,thesolutionisgivenby: ˛∗=0 if R
r > and ˛∗=1 if R
r <. (5)
Observethattheexistenceofastoragetechnologywouldnotaffecttheconditionstoliquidatethe longasset.Indeed,aslongasthelongassetvalueatperiod1islowerthat,itisalwaysoptimalto liquidateit.
Anecessaryconditionfortheexistenceofaninterbankequilibriumisthat r∈
Rl , Rh . (6) Thisisbecauseifr≤Rl,by(4)allbankswillnotliquidatetheinvestmentintheriskytechnology,and
by(2)theywillwishtofinancealldate1consumptionbyborrowing.Thus,therewouldbenolenders, hencenointerbankequilibrium.Likewise,ifr≥Rh
,by(5)allbankswillliquidatetheinvestmentin
theriskytechnology,andby(3)theywillwishtofinancealldate2consumptionbytherepayments onlendingatdate2.Butatdate1therewouldbenoborrowers,hencenointerbankequilibrium.
Using(2),(3)and(5)in(1),bank(R,)solves: Max b ˘1=U
˛∗+b +(1−)UR(1−˛∗)−rb 1− , (1a)thefirstorderconditionwithrespecttobis U
˛∗+b =rUR(1−˛∗)−rb 1− . (7)Thus,thesolutionofthebankproblemisgivenby(5)and(7).Notethatbanksoptimalchoicesare theliquidationdecision˛*(R),whichdoesnotdependon,andb(R,).Theliquidationdecisionin responsetothecreditriskrealizationdoesnotdependontheliquidityshock(in(5)nothingdepends on),buttheborrowingdecisiondependsonbothshocks(by(7)).
Wecharacterizeequilibriaforlogutilitypreferences,i.e.U(c)=ln(c).Thechoiceofthisspecification ismotivatedbysimplicityandbythefactthatforthesepreferences,date1spotmarketallocationsand optimalbankingallocationsgenerallycoincideinliquiditypreferenceframeworkssuchasours(see Allen&Gale,2007).Ourresultsarerobusttomoregeneralconsumerutilityspecifications.Itiseasyto seeinEq.(7)thatitissufficienttoadoptaconsumer’sutilityfunctiontwicecontinuouslydifferentiable, increasing,andstrictlyconcave.Hence,thesepreferencescanbeviewedasausefulbenchmarkto judgedifferencesinequilibriumsandassociatedwelfarepropertiesindiversifiedandconcentrated riskeconomiesindependentlyofefficiencywedgesbetweenmarketandbankingallocations.
Eq.(7)yields: ˛∗+b= R(1r(1−˛−∗))−rb. (8) Solving(8),wehave b(R,)=1r(R(1−˛∗)−r(1−)˛∗) . (9) Sincer∈
Rl ,R h,by(5),optimalasset’sliquidationis˛*(Rl)=1and˛*(Rh)=0.
Insum,thefourbanktypeshavethefollowingborrowing/lendingpositions
b(Rl,l)=−(1−l) (9a) b(Rl,h)=−(1−h) (9b) b(Rh,l)=1 r
lRh (9c) b(Rh,h)=1 r hRh, (9d)andequilibriumintheinterbankmarketrequires p1rlRh+(1−)q1
rhRh−(1−)(1−q)(1−l)−(1−p)(1−h)=0. (10) Theaboveequation(10)islinearwithrespecttorandhastheuniquesolution:
r∗= Rh
pl+(1−)qh
[(1−)(1−q)(1−l)+(1−p)(1−h)]. (11)
Eq.(11)saysthattheinterbankequilibriumrateraisesastheliquidityneedsandtheopportunity costofholdingtheasset,Rh,increase.
4. Comparisonsofequilibria
Inthissectionweidentifyconditionsensuringexistenceofequilibriafortheextremevaluesof, andcomparethesetofparametersforwhichequilibriaexistsforsuchvalues.
Using(11)andr∗∈
Rl ,R h ,weobtain 1≥ pl+(1−)qh (1−)(1−q)(1−l)+(1−p)(1−h)≥ Rl Rh. (12)Weuse(12)toassesstheexistenceofinterbankequilibriaundertwoextremecases,thatofbanking system’sperfectdiversification(=0),andthatofmaximalriskconcentrationinthebankingsystem (=1).Themainresultissummarizedinthefollowingproposition.
Proposition1. Fortheperfectdiversifiedeconomy(=0)andfortheperfectconcentratedeconomy
(=1),thesetofinterbankequilibriaisnon-emptyforanyparameterconfigurationofcreditandliquidity risk(h,l,Rh,Rl).
Now,wewishtocomparethesizeofthesetofeconomies,indexedbypandq,forwhichequilibria existfor=0andfor=1.
Weassesstheequilibriadomainsdefiningtwosetsasbelowdetailed.Tothisscope,wemeasure thesesetscomputingthelinearextremes[
v
0,v
1]forthecase=0and[v
1,v
11]for=1.TheseextremesUsingtherightandthelefthandsideof(12),theequilibriumdomainsfor=0,[
v
0,v
10],andfor =1,[v
1,v
11],arerespectivelyv
0= Rl Rh(1−l) h+ Rl Rh(1−l) ,v
10= 1−l 1+h−l(13) and
v
1= Rl Rh(1−h) l+ Rl Rh(1−h) ,v
11= 1−h 1−h+l. (14)
Thelargeristheintervalforwhichequilibriaexistsunderdiversificationorconcentration,the largeristhesetofeconomiesthatmaybenefitfromtherisksharingopportunitiesofferedbythe interbankmarket.
Considerthedifferencebetweentheequilibriaintervalwhen=0andwhen=1,definedas:
G≡0−1=(
v
0−v
10)−(v
1−v
11). (15) Computationsgive: 0= h(1−l)1− Rl Rh (1+h−l)h+ Rl Rh(1−l) , (16) and 1= l(1−h)1− Rl Rh (1−h+l) l+Rl Rh(1−h) . (17)Ofparticularinterestisthecomparisonofthesetofequilibriaforrelativelylargecreditandliquidity shocks.Thiscomparisonismadeclearerbysubstitutingh=l,where 1
l ≥≥1,andRl=ˇRhwhere
ˇ≤1.Parameterisameasureoftheliquidityriskandˇisthecreditrisk.Thesmallerisˇ,thelargeris thedifferencebetweenhighandlowreturn.Thelargeris,thelargeristhedifferencebetweenlow andhighliquidityshock.
Thus,(15)canbeexpressedas
G(l,,ˇ)= l(1−l)
(1+l−l)
l+ˇ(1−l)−l(1−l)
(1−l+l)
l+ˇ(1−l). (18)Thefollowingpropositionestablishesarankingofthesizeofequilibriaunderadiversifiedeconomy andarisk-concentratedeconomy.
Proposition2. Thereexistsalsuchthatforl≥landany(ˇ,)thesetofinterbankequilibrium
underthediversifiedeconomyisalwaysstrictlylargerthanfortheconcentratedeconomy.
Whenliquidityshocksarelarge,thesetofinterbankequilibriaofthediversifiedeconomyislarger thanthecorrespondingsetoftherisk-concentratedeconomyandtheinterbankequilibriumislikely tobreakdownformoremodelparameterranges.Therefore,adiversifiedeconomyoffersabetter insuranceagainsthighliquidityandcreditrisk.Finallynotethatby(18),forˇ=␥=1,whichamounts toabsenceofcreditandliquidityrisk,G(1,1)=0,thenthetwosetsareequivalent.Thismeansthat bankmarketstructureisimportantespeciallywhentheshocksarehigh.
Fig.1showsgraphicallytheresultofProposition2forasetofeconomies.Thesurfacerepresenting thefunctionG(l,,ˇ)isincreasinginliquidityrisk.
AsFig.1highlights,thesetofequilibriaisincreasinginthedegreeofbanks’riskdiversificationfor aliquidityriskgreaterthat¯l.Thisindicatesthataninterbankislesslikelytobreak-downwhenthe
Fig.1. BehaviorofG(l,,ˇ).
5. Welfarecomparisons
Wewanttocompareagents’welfareineconomieswithbankingsystemsdifferingaccordingto marketstructureandriskconcentration.Todothat,wecomputedepositors’expectedutility.
Using(2)and(3),theconsumptioninstates1and2is c1= ˛+b
(19)
and
c2= R(1−˛)−rb
1− . (20)
Substitutingequilibriumvalues˛,bandrin(19)and(20),weobtaintheconsumptionallocation offeredbydifferentbanktypes,Ctype(i)=(ctype(i)
1 ,c2type(i))fori=1,2,3,4: Ctype(1)=
(1−)(1−q)(1−l)+(1−p)(1−h) pl+(1−)qh ,Rh(21a) Ctype(2)= (1−)(1−q)(1−l)+(1−p)(1−h) pl+(1−)qh ,Rh
(21b) Ctype(3)= , R hpl+(1−)qh (1−)(1−q)(1−l)+(1−p)(1−h)
(21c) Ctype(4)= , R hpl+(1−)qh (1−)(1−q)(1−l)+(1−p)(1−h)
. (21d)
Theex-antetheexpectedutilityofarepresentativeconsumeristherefore
W≡ 4
i=1 P(type(i))U(c1type(i),c type(i) 2 ) fori=1,2,3,4. (22) Equivalently:Fig.2.ThebehaviorofWasafunctionofandliquidityrisk. Unfolding(23)weget: W ≡p
llog (1−)(1−q)(1−l)+(1−p)(1−h) pl+(1−)qh+(1−l)log(Rh)
+(1−)q hlog (1−)(1−q)(1−l)+(1−p)(1−h) pl+(1−)qh
+(1−h)log(Rh)
+(1−)(1−q) llog()+(1−l)log Rhpl+(1−)qh (1−)(1−q)(1−l)+(1−p)(1−h)
+(1−p) hlog()+(1−h)log Rhpl+(1−)qh (1−)(1−q)(1−l)+(1−p)(1−h)
(24)
Fig.2showsWforgivenreturnparameters.NotethatWisdecreasingbothinandintheliquidity riskparameter.
Fig.2showsthatwelfaredecreaseswhenbankingsystemriskisconcentratedandliquidityriskis high.Thecornerrepresentinghighlevelsofandcanbeviewedasa“crisis”setofriskrealizations withhighwelfarelosses.
Whentheinterbankequilibriumdoesnotexist,theautarkicallocationwillprevail.Depositors’ expectedutilityunderautarkyisgivenbythesolutionofthefollowingproblem:
Max ˛ W A=U(c1)+(1−)U(c2) (1b) Subjectto c1=˛ (2b) (1−)c2=R(1−˛). (3b)
Theoptimalsolutionis˛*=.Therefore,substitutingin(2b)and(3b),theconsumptionallocations foreachbanktypeatanautarkicequilibriumare:
Ctype(1)=Ctype(2)=
,Rh (25a)Fig.3. BehaviorofWandWAwithdifferentliquidityshocks.
Correspondingly,theexpectedutilityofanagentattheinitialdateis:
WA≡p[llog()+(1−l)log(Rh)]+(1−)q[llog()+(1−l)log(Rh)]
+(1−)(1−q)[llog()+(1−l)log(Rl)]+(1−p)[llog()+(1−l)log(Rl)]
(26) Fig.3showsW,theexpectedutilitywhentheinterbankequilibriumexists,andWA,theexpected
utilityunderautarky,asfunctionsofand␥fortwosetsofeconomies.Forhighvaluesofrisk con-centrationandliquidityrisktheinterbankequilibriumdoesnotexist,asWA>W.Whentheinterbank
surfaceliesovertheautarkyplane,theinterbankequilibriumexistssinceW>WAforallparameter
values,withWstrictlydecreasinginthedegreeofriskconcentration.
Fig.3illustratesthegeneralfindingsummarizedinthefollowingproposition3.Thefiguredescribes theexistenceoftheinterbankmarketandthewelfarebehaviorfordifferentlevelsofriskconcentration andliquidityshocks.Thecornerwithhighliquidityshocksandhighconcentrationisacrisis.Inthis casetheinterbankmarketwillnotexistsincetheautarkyequilibrium(theplaneWA)giveshigher
welfarethantheinterbankequilibrium.
Proposition3. (a)Thereexists(,¯ )¯ suchthatWA>Wforall(,)>(,¯ ).¯
(b)WhenW>WAtheinterbankequilibriumexists,andWisstrictlydecreasinginforany.
Itisusefultoillustrateexamplesfordifferentsetofeconomies.Fig.4clearlyshowsthatwhen ishighandliquidityrisk,,issufficientlyhigh,theinterbankallocationisdominatedbytheautarky allocation.
Fig.5showsthatwithhighcreditrisk,i.e.lowprobabilityofrealizationofthehighreturn,the interbankmarketismorelikelytoexistthantheautarkyallocation.Thissuggeststhatinterbank marketsinsureagainstcreditriskalso,whileWremainsdecreasingwithrespectto.
Thepictures3,4and5highlightthatwelfarealwaysdecreaseswhentheeconomyexhibitsahigh risksconcentrationandhighliquidityexposure.
6. Anexampleofendogenousdegreeofriskconcentration
Sofarthedegreeofriskconcentrationinthebankingsystem(parameter)hasbeentreatedas exogenous.Wehaveshownthatwelfaredecreasesinthedegreeofmarketconcentration.Forall
Fig.4. WandWAwithrelativelyhighpandq.
Fig.5.WandWAwithrelativelylowpandq.
examplesshown,thehighestdepositors’expectedutilityisreachedattheminimumlevelofrisk concentration(=0).
Recallthatisaparameterthatdeterminestheprobabilitiesofcreditandliquidityshocksasof date0.Therefore,achoiceofcanbeviewedasabankchoiceofcreditandliquidityrisk.Ifachieving perfectdiversificationatasystemleveliscostless,theninaperfectlycompetitivebankingsystem, bankswouldchoosetheminimumlevelofriskconcentrationandtheprobabilityofinterbankmarket breakdownswouldbeminimized.
Inreality,ShehzadandDeHaan(2013)suggestthatstockpricesoflargebankswereaffected moreduringthecrisisthanthoseofsmallbanksandtheyalsofindthatmanagerialefficiency,loan quality,leverage,andthevolumeofoutstandingloansaffectbankstockprices.Literally,achieving diversificationandcontrollingriskiscostly,sinceriskmanagementcanbeviewedasatechnology availabletofirmssimilar,forexample,tothetechnologyunderlyingcreditriskmodels.Asthesize andscopeoftheoperationsofbanksexpands,financialfirms’spanofcontrolovertheirmanyunitscan becomelessefficientincontrollingrisk.Inthislight,alackofsufficientbankingsystemdiversification
andahigherprobabilityofinterbankmarketbreakdownsmaybeinpartduetoriskcontroldiseconomies ofscaleandscope.
Thepotentialforfirstordereffectsofriskmanagementdiseconomiesofscaleandscopeonrisk concentrationinthebankingsystem,andtheirrelationshipwithbanksize,canbeillustratedbythe followingmodificationofourmodel.
Supposeatdate0bankshaveasizeS≥0,andchooseemployingpartofdate0resources. Specif-ically,theyinvestafractionxofdateresourcesinthetechnologyandchooseincurringacostz()S
asafractionofdate0resources,where>0isthescalecostparameter.Theirresourceconstraintat date0istherefore:
x+z()S=S (27)
Assumez()=a0
+a1,witha0anda1aspositivecoefficients.Thisfunctioncanbeinterpreted
asacostfunctionofariskcontroltechnologythatexhibitsdecreasingreturnstoinvestmentovera certainthreshold.Itsparameterscoulddependonsizeandscopeoffinancialfirmsoperations,aswell asonincentivesarisingfromasymmetricinformation,marketpowerandotherfactorspointedoutin theliterature,whichmayinturnbeaffectedbyfirms’size.
TocomputeW(),wereplace1withS(1−z()S)intheconsumptionallocations,
c1= ˛S(1−z()S )+b (28) and c2= R(1−˛)S(1−z()S )−rb 1− . (29)
Then,optimalborrowing/lendingchoicesintheinterbankmarketare: b(R,)=S(1−z()S)
r (R(1−˛∗)−r(1−)˛∗) , (30)
Withtheliquidationchoices˛unchanged,theequilibriumintheinterbankmarketisthesolution ofthefollowingequation:
pS(1−z()S) r lRh+(1−)q S(1−z()S) r hRh−(1−)(1−q)S(1−z()S)(1−l) −(1−p)S(1−z()S)(1−h)=0, (31) whichyields r∗= Rh
pl+(1−)qh [(1−)(1−q)(1−l)+(1−p)(1−h)]. (32)Notethat(32)isequivalentto(11),meaningthattheequilibriuminterbankratedoesnotdepend onthefunctionS(1−z()S).Thismeansthatinthemarketforliquidityistheexcessofdemand
onsupplythatdeterminesinterbankrate.Multiplyingbothdemandandsupplyforthesizefunction wouldnotaffecttheequilibriuminterbankrate.
The consumption allocations for thefour bank typeswhich allowus tocomputedepositors’ expectedutilityare:
Ctype(1)=
(1−)(1−q)(1−l)+(1−p)(1−h)S(1−z()S) pl+(1−)qh ,RhS(1−z()S)(33a)
Fig.6. ExpectedutilityWwithcostsofriskcontrol=1andS=4. Ctype(2)=
(1−)(1−q)(1−l)+(1−p)(1−h)S(1−z()S) pl+(1−)qh ,RhS(1−z()S)(33b) Ctype(3)= S(1−z()S), Rh pl+(1−)qhS(1−z()S) (1−)(1−q)(1−l)+(1−p)(1−h)
(33c) Ctype(4)= S(1−z()S), Rh pl+(1−)qhS(1−z()S) (1−)(1−q)(1−l)+(1−p)(1−h)
. (33d)
Fig.6shows(withdifferentangles)anexampleoftheexpectedutilityfunctionWasafunctionof andtheliquidityriskparameter.ItisapparentthatfunctionWisstrictlyconcave,withamaximum forasaninteriorpoint.WeplotthesurfaceforsizeparameterS=4and=1.
Proposition4. Foranyvalueofandforagivenlevelofliquidityrisk,¯ theoptimalbankrisk
concen-trationlevel*isincreasinginS.
Proposition4statesthatthescaleparameteristhedeterminantofthedegreeofdiversification thatabankchooses,andbankdiversificationisinverselyrelatedtoitsassetssize.
Fig.6showstheconcavitypropertyofthewelfarefunctioninthecaseofcostlydiversification.This impliesthechoiceofadifferentfrom0.
Fig.7showstheoptimalfordifferentbanksize.Themaximumoftheexpectedutilityisdifferent accordingtoS.Theconcentrationisgreaterforabanksizelargerthanone.Bycontrast,forasmall banksizetheoptimalriskconcentrationapproacheszero.
Therefore,theoptimallevelofriskconcentrationmightbelargerthantheminimumfeasible.In turn,riskconcentrationislargerforalargerbanksize.Thelevelofa0anda1maydependfundamentally
Fig.7.ExpectedutilityWforS=1andS=4.
financialfirms’size,suchastoo-big-to-failincentives,maycarryhigherriskcontrolcosts,whichin turncouldresultinahigherlevelofriskconcentrationinthebankingsystem.
7. Conclusion
Ourmodelshowsthatriskconcentrationseemsoffirstorderimportanceforthesmoothfunctioning ofinterbankmarkets.Riskconcentrationmayincreaseinbanksizeunderreasonableassumptions abouttheriskmanagementtechnologyavailabletobanks.Differingfrommostliterature,ourresults areobtainedinamodelwithnoasymmetricinformation,nomarketpowerordysfunctionalsecondary markets.
Withregardtopolicy,theliteraturehasprominentlyfocusedontheimportantroleoftheCentral Bankaslenderoflastresortwheninterbankmarketbreakdownsoccur(seee.g.Freixas,Parigi,& Jean-Charles,2000;Freixas,Antoine,&Skeie,2009;Goodhart&Illing,2002;Repullo,2005).Ourmodel suggeststhattheimprovementofriskmanagementtechnologiesinlargeandcomplexfinancial insti-tutions,aswellasinregulatorybodies,maybeasimportanttominimizetheoccurrenceofinterbank
marketbreakdowns.
Acknowledgements
IacknowledgesfinancialsupportundertheprojectSYRTO,fundedbytheEuropeanUnionunder the7thFrameworkProgramme(FP7-SSH/2007-2013–GrantAgreementn◦320270)andtheproject MISURA,fundedbytheItalianMIUR.
The researchleadingtotheseresultshasreceivedfundingfromtheEuropeanUnion,Seventh
FrameworkProgrammeFP7/2007-2013undergrantagreementSYRTO-SSH-2012-320270.
IthankwithoutimplicationstheseminarInstituteandtheDeutscheBundesbankforcomments andsuggestions.
Appendix.
Proposition1. Fortheperfectlydiversifiedeconomy(=0)andfortheperfectlyconcentrated
econ-omy(=1),thesetofinterbankequilibriasisnon-emptyforanyparameterconfigurationofcreditand liquidityrisk(h,l,Rh,Rl).
Proof. Theleft-handsideinequalityof(12)canbeexpressedas:
(1−)(1+h−l)q≤(1−)(1−l)+(1−h)−p(1−h+l), (A1)
If=0, then (1+h−l)q≤(1−l) (A2)
If=1, then 0≤(1−h)−p(1−h+l) (A3)
Theright-handsideinequalityof(12)canbeexpressedas: (1−)(h+ Rl Rh(1−l))≥ Rl Rh[(1−h)+(1−)(1−l)]−p[(1−h) Rl Rh+l] (A4) If=0, then q
h+ Rl Rh(1−l) ≥ Rl Rh(1−l) (A5) If=1, then 0≥ Rl Rh(1−h)−p[(1−h) Rl Rh+l] (A6)Thus,thesetofeconomiesforwhichaninterbankequilibriumexistsisindexedbyq∈[0,1]if=0, andbyp∈[0,1]if=1.
If=0,using(A2)and(A5),weget:
Rl Rh(1−l) h+ Rl Rh(1−l) ≤q≤ 1−l 1+h−l. (A7)
Therefore,if=0,thesetofinterbankequilibriaisnon-emptyif 1−l 1+h−l≥ Rl Rh(1−l) h+ Rl Rh(1−l) . (A8)
Inequality(A8)impliesthat(1−l)(h+Rl
Rh(1−l))≥ R l
Rh(1−l)(1+h−l),whichcanbe
fur-thersimplifiedtoh≥ Rl
Rhh,whichinturnisalwaysverifiedsince R l
Rh <1.
If=1,using(A3)and(A6)weget: 1−h 1−h+l≥p≥ Rl Rh(1−h) l+ Rl Rh(1−h) . (A9)
Inequality(A9)isalwayssatisfied,sinceitreducestol≥ Rl
Rhl,whichholdsas R l
Rh <1.
Proposition2. Thereexistsalsuchthatforl≥landany(,ˇ)thesetofinterbankequilibrium
underthediversifiedeconomyisalwaysstrictlylargerthanfortheconcentratedeconomy.
Proof. Differentiating(18)withrespecttol,andevaluatingthederivativeatl=1:
Gl(0,,ˇ)
(−1)
4(1+ˇ)2+2ˇ(1+5ˇ)−3ˇ2−(1+7ˇ+8ˇ2)
(−2)2(ˇ(−1)−1)2 , (A10)
since(−1)>0and(−2)2(ˇ(−1)−1)2>0,(A10)ispositiveif
4(1+ˇ)2
Observethatforl→1,→1,then(A11)reducesto
3+2ˇ>0. (A12)
Therefore,thereexistvaluesof¯land¯suchthatforl≥landfor≥¯thefunctionG(l,,ˇ)
isincreasing.
Proposition3. (c)Thereexists(,¯ )¯ suchthatWA>Wforall(,)>(,¯ );¯
(d)WhenW>WAtheinterbankequilibriumexists,andW,isstrictlydecreasinginforany.
Proof. Weprovepart(a)and(b)separately.
(a)Thereisa(,)>(,¯ )¯ suchthattheautarkyallocationdominatestheinterbankequilibrium. Theexpectedutilityunderautarkyfor=0andfor=1are
WA0≡q[llog()+(1−l)log(Rh)]+(1−q)[llog()+(1−l)log(Rl)], (A13)
and
WA1≡p[llog()+(1−l)log(Rh)]+(1−p)[llog()+(1−l)log(Rl)]. (A14)
Hence,wecomparetheexpectedutilitywithinadiversifyeconomybetweeninterbankallocation andautarkyallocation
W0−WA0≡q
llog (1−q)(1−l) ql +(1−l)log(Rh) +(1−q) llog()+(1−l)log Rhql (1−q)(1−l) −q[llog()+(1−l)log(Rh)] +(1−q)[llog()+(1−l)log(Rl)]⇒q l log (1−q)(1−l) ql −log() +(1−q) (1−l) log Rhql (1−q)(1−l) −log(Rl) , (A15)Thedifferenceofexpectedutilityinariskconcentratedeconomybetweentheinterbankallocation andautarkyallocationis
W1−WA1≡p
llog (1−p)(1−l) pl +(1−l)log(Rh) +(1−p)llog() +(1−l)log Rhpl (1−p)(1−l) −p[llog()+(1−l)log(Rh)] +(1−p)[llog()+(1−l)log(Rl)]⇒p l log (1−p)(1−l) pl −log() +(1−p) (1−l) log Rhpl (1−p)(1−l) −log(Rl) . (A16)Computing(A15)and(A16)for→ 1
l,W0−WA0>0,andW1−WA1<0.Then,weconcludethat
thereexistacouple(,)>(,¯ )¯ suchthattheautarkyallocationgivesahigherexpectedutilitythan theinterbankmarketequilibrium.
(b)Whenwehaveaninterbankequilibrium,theexpectedutilitydecreasesin.Takethederivative of(24)withrespecttoandcomputeitfor→1andfor→ 1
l, W→1
→ 1 l →−<0. (A17)Computethederivativealsoforl→0
W→1
l→0<0. (A18)Therefore, as risk concentration increases, the expected utility in the interbank economy is decreasing.Moreover,theexpectedutilityfor=0andfor=1are
W0≡q
llog (1−q)(1−l) ql +(1−l)log(Rh) +(1−q) llog()+(1−l)log Rhql (1−q)(1−l) (A19) and W1≡p llog (1−p)(1−l) pl +(1−l)log(Rh) +(1−p) llog()+(1−l)log Rhpl (1−p)(1−l) . (A20) Let→ 1l,thedifferenceofexpectedutilitiesis
W0−W1=q
log (1−q)(1−l) q +(1−q) llog()+(1−l)log Rhq (1−q)(1−l)−
pllog( 0)+(1−l)log(Rh)+(1−p)log()>0, (A21)sinceW1
→ 1 l <0andW0→ 1 l >0. Letl→0, W0−W1→0. (A22)Weconcludethatforanytheexpectedutilitydecreasesinconcentration,.
Proposition4. Foranyvalueofandforagivenlevelofliquidityrisk,¯ theoptimalbankrisk
concen-trationlevel*isincreasinginS.
Proof. Taketheoptimal*solvingWforagivenlevelofliquidityrisk¯andforp=q=0.5.ForS→0,
wehavethat*→0andforS→∞*→1. References
Acharya,V.V.,&Merrouche,O.(2013).Precautionaryhoardingofliquidityandinterbankmarkets:Evidencefromthesubprime
crisis.ReviewofFinance,17(1),107–160.
Acharya,V.V.,&Skeie,D.(2011).Amodelofliquidityhoardingandtermpremiaininterbankmarkets.JournalofMonetary
Economics,58.
Acharya,V.V.,Gromb,D.,&Yorulmazer,T.(2012).Imperfectcompetitionintheinterbankmarketforliquidityasarationale
forcentralbanking.AmericanEconomicJournal:Macroeconomics,4(2).
Acharya,V.,&Yorulmazer,T.(2008).Informationcontagionandbankherding.JournalofMoneyCreditandBanking,40,215–231.
Afonso,G.,Kovner,X.,&Schoar,A.(2011).Stressed,notfrozen–Thefedfundsmarketandthecrisisof2007/2008.Journalof
Finance,60.
Allen,F.,&Gale,D.(2007).Understandingfinancialcrises.Oxford,UK:OxfordUniversityPress,ClarendonLecturesinFinance.
Anand,K.,Gai,P.,&Marsilli,M.(2012).Rolloverrisk,networkstructureandsystemicrisk.JournalofEconomicsDynamicsand
Control,36,1088–1100.
Bhattacharya,S.,&Gale,D.(1987).PreferenceShocks,LiquidityandCentralBankPolicy.Newapproachestomonetaryeconomics,
Boyd,J.H.,DeNicolò,G.,&Loukoianova,E.(2009).Bankingcrisesandcrisisdating:Theoryandevidence.InIMFWorkingPaper 09/141.
Boyd,J.H.,DeNicolò,G.,&Jalal,A.M.(2009).Bankcompetition,riskandassetallocations.InIMFWorkingPaper09/143.
Brunnermeier,M.K.(2009).Decipheringtheliquidityandcreditcrunch2007–2008.TheJournalofEconomicPerspectives,77–100.
Cai,J.,&Thakor,A.V.(2008).Liquidityrisk,creditrisk,andinterbankcompetition.InWorkingpaperOlinBusinessSchool,W.D.C.
Ciccarelli,M.,Maddaloni,A.,&Peydró,J.L.(2013).Heterogeneoustransmissionmechanism:Monetarypolicyandfinancial
fragilityintheeurozone.EconomicPolicy,28(75),459–512.
Cornett,M.M.,McNutt,J.J.,Strahan,P.E.,&Tehranian,H.(2011).Liquidityriskmanagementandcreditsupplyinthefinancial
crisis.JournalofFinancialEconomics,101(2),297–312.
Covitz,D.,&Downing,C.(2007).Liquidityorcreditrisk?Thedeterminantsofveryshort-termcorporateyieldspreads.The
JournalofFinance,62(5),2303–2328.
DaSilva,M.S.,&Divino,J.A.(2013).Theroleofbankingregulationinaneconomyundercreditriskandliquidityshock.The
NorthAmericanJournalofEconomicsandFinance,26,266–281.
DeNicolò,G.,&Kwast,M.L.(2002).Systemicriskandfinancialconsolidation:Aretheyrelated?JournalofBanking&Finance,
26,861–880.
DeNicolò,G.,Hayward,P.,&Bhatia,A.V.(2004).U.S.LargeComplexBankingGroups:Businessstrategies,risksandsurveillance
issues.InIMFCountryReport04/228,July.Washington:InternationalMonetaryFund.
DeNicolò,G.,Bartholomew,P.,Zaman,J.,&Zephirin,M.(2004).Bankconsolidation,internationalizationandconglomeration:
Trendsandimplicationsforfinancialrisk.FinancialMarketsInstitutions&Instruments,13(4),173–217.
Diamond,D.W.,&Dybvig,P.H.(1983).Bankruns,depositinsurance,andliquidity.TheJournalofPoliticalEconomy,91(3),
401–419.
Diamond,D.W.,&Rajan,R.G.(2005).Liquidityshortagesandbankingcrises.JournalofFinance,60,615–647.
Diamond,D.W.,&Rajan,R.G.(2008).Illiquidityandinterestratepolicy.InWorkingpaper.UniversityofChicago.
ECB.(2009December).Financialstabilityreview.EuropeanCentralBankFrankfurtAmMain.
Ericsson,J.,&Renault,O.(2006).Liquidityandcreditrisk.TheJournalofFinance,61(5),2219–2250.
Flannery,M.J.(1996).Financialcrises,paymentsystemproblemsanddiscountwindowlending.JournalofMoneyCreditand
Banking,28,804–824.
Freixas,X.,Parigi,B.M.,&Rochet,J.-C.(2000).Systemicrisk,interbankrelations,andliquidityprovisionbytheCentralBank.
JournalofMoney,CreditandBanking,32(3),611–638.
Freixas,X.,Martin,A.,&Skeie,D.(2009).Bankliquidityinterbankmarkets,andmonetarypolicy.FederalReserveBankofNew
YorkStaffReportN.371.
Freixas,X.,&Jorge,J.(2008).Theroleofinterbankmarketsinmonetarypolicy:Amodelwithrationing.JournalofMoney,Credit
andBanking,40(6),1151–1176.
Furfine,C.H.(2000).Interbankpaymentsandthedailyfederalfundsrate.JournalofMonetaryEconomics,46,535–553.
Gai,P.,Haldane,A.,&Kapadia,S.(2011).Complexityandcontagion.JournalofMonetaryEconomics,58,453–470.
Gale,D.,&Yorulmazer,T.(2012).Liquidityhoarding,staffreport488FederalReserveBankofNewYork.InForthcomingin
theoreticaleconomics.
Georg,C.P.(2013).Theeffectoftheinterbanknetworkstructureoncontagionandcommonshocks.JournalofBanking&Finance,
37(7),2216–2228.
Giulioni,G.(2015).Policyinterestrate,loanportfoliomanagementandbankliquidity.TheNorthAmericanJournalofEconomics
andFinance,31,52–74.
Gorton,G.,&Huang,L.L.(2004).Liquidity,efficiencyandbankbailouts.AmericanEconomicReview,94,455–483.
Gorton,G.,&Huang,L.L.(2006).Bankingpanicsandendogeneitycoalitionformation.JournalofMonetaryEconomics,53,
1613–1629.
Goodhart,C.,&Illing,G.(2002).Financialcrisescontagion,andthelenderoflastresort.OxfordUniversityPress.
GroupofTen.(2001).Reportonconsolidationinthefinancialsector,January,Basel.
Hartmann,P.,Straetsman,S.,&deVries,C.(2005).Bankingsystemstability:Across-Atlanticperspective,WorkingPaperNo.527
September.Frankfurt:EuropeanCentralBank.
Heider,F.,Hoerova,M.,&Holthausen,C.(2010).Liquidityhoardingandinterbankmarketsspreads:Theroleofcounterparty
risk.InCEPRDiscussionPapers:7762.
Houston,J.F.,&Stiroh,K.J.(2006).Threedecadesoffinancialsectorrisk,“U.S.FinancialHoldingCompanies,”July.NewYork:Federal
ReserveBankofNewYork(unpublished).
Ibragimov,R.,Jaffee,D.,&Walden,J.(2011).Diversificationdisasters.JournalofFinancialEconomics,99,333–348.
Miralles-Marcelo,J.L.,delMarMiralles-Quirós,M.,&Miralles-Quirós,J.L.(2015).Improvinginternationaldiversificationbenefits
forUSinvestors.TheNorthAmericanJournalofEconomicsandFinance,32,64–76.
Repullo,R.(2005).Liquidity,risktaking,andthelenderoflastresort.InternationalJournalofCentralBanking,2,47–80.
Shehzad,C.T.,&DeHaan,J.(2013).Wasthe2007crisisreallyaglobalbankingcrisis?TheNorthAmericanJournalofEconomics
andFinance,24,113–124.
Stiroh,K.J.(2004).Diversificationinbanking:Isnoninterestincometheanswer?JournalofMoneyCreditandBanking,36(5),
853–882.
Stiroh,K.J.,&Rumble,A.(2006).Thedarksideofdiversification:ThecaseofU.S.FinancialHoldingCompanies.JournalofBanking
andFinance,30(8),2131–2161.
Wagner,W.(2010).Diversificationatfinancialinstitutionsandsystemiccrises.JournalofFinancialIntermediation,19,333–354.