THREEESSAYSINFINANCEByJaeminBaikADISSERTATIONSubmittedtoMichiganStateUniversityinpartialoftherequirementsforthedegreeofEconomicsŒDoctorofPhilosophy2016ABSTRACTTHREEESSAYSINFINANCEByJaeminBaikChapter1,usingindividualKoreanbankloansoutstandingdatabyloanssector,cal-culatestheintensityofherdingforlendingtosectoraswellasoveralllendingmarketbyawayoftwomethods.Bylookingatthetimeseriespatternoftwomeasures,Iexaminethechangeinbanks'lendingbehavior.Withtheresultsfromtwomeasures,Itrytoempiricalevidenceofherdinginthepastandsomefeaturesrelatedtotheherdbehavior.Also,usingtheloansdatabyindustry,IinvestigatefurtherrationalityofherdingbycalculatingafteradjustmentLSVmeasuretoexcludeindustry-rationalfactorfrombeforeadjustmentLSVmeasure.Finally,usingoneofmea-sureswhichallowsustogettheintensityofindividualeconomicagents,thispaperem-piricallyanalyzethefactorsthatintensifyKoreanbank'sherding.TheresultimpliesthatextentofherdingwasfollowedbyKoreanbanksforlendingtobothoveralllendingmarketandhouseholdsectorduring2002-2003,andinlendingtoSME,theherd-ingisconsideredtohavebeengreatlyaround2003andduring2006and2008.Inaddition,herdingtendstogetmostduringtimeswhenbankschangetheirlendingdecisionsundertheincreaseduncertaintyaboutexpectationforeconomiccondi-tions.Theextenttowhichthebankcharacteristicsandstructural/macroeconomicfactorsaffecttheherdingintensityseemstovarybythelendingsectors.Chapter2investigatewhetherthebankhealthandthecreditshocktothebankaffectsitsclientexportusingdataforKoreanbank-leveldataonKoreanbanks,andmatchedloansdataduringtheperiod2010-2013.Also,byaddingtheinteractiontermbetweenthebankhealthandtheownershipconcentrationmeasure,theanalysisisextendedtotesttheroleoftheownershipstructurefortheIthatthemoresharesoftheareheldbythefamilymembers,themorelikelythatistoparticipateintheexportsactivity.Italsoimpliesthatthemoreshareholdingsbythefamilymembershelpstheexportactivitymoreresilientwhenhealthofbanksfromwhichtheborrowsgetworse.Itisnotonlyhealthofitsmainbank,butalsothatofallothercreditorbanksthataffectstheexportactivity.Also,whenseparatingtheeffectstransmittedbymainbankandotherbanksatthesametime,themainbankplaysmorerolethanothercreditorbanks.Chapter3explorestheextensiveliteraturesontheeffectsoftheshocksandthecorporategovernanceontherealactivities.Withthegrowinginterestintheglobaltradecollapseafterthecrisis,Iamfocusingontheimpactontheexports.Therehavebeenmanystudiesregardingtransmissionofshockstobankstotherealactivityandhowthequalityoftheownershipstructureaffectsthedeci-sionanddynamics.Iexplorethemandsummarizetheinthispaper.Tomybelovedwife,Eunyoung,andmypreciousthreedaughters,Seohyeon(Zion),Sevin(Sharon),andYerin(Jordan)ivACKNOWLEDGEMENTSSixyearsinEastLansing,Michigannowfeellikeashortsummertrip.WhileIwasthere,however,itfeltlikealongandriskyjourney.ThoughittookmuchlongerthanIplannedtomydegree,Istandamazedlookingbackthosetimes.Alongtheway,Iwasneveralone.Iamhonoredtorecallallthenamesthatmademyachievementpossible.First,Iwouldliketoexpressmysinceregratitudetomyadvisor,RaoulMinetti,forthecontinuoussupportformyPh.Dstudyandtheresearch.Hisguidancehelpedmeinallthetimeofresearchandwritingofthisdissertation.Withouthissupport,itwouldneverbepossibletomydegree.Besidesmyadvisor,Iwouldliketothanktherestofmydissertationcommitteemembers:ChunZhu,LuisDeAraujo,andKirtC.Butlerfortheirinsightfulcommentsandencouragement.Ithankmyfriendsandco-workersatthedepartmentofEconomics.Inparticular,IamgratefultoCheol-KeunChoandSoobinKimforbeingsuchtrustfulandthoughtfulfriendsandfellowresearchers.SpecialthanksgotoPastorJohnWonfortheconstantsupportofprayer,andtomyneighbors,Mr.Cooper,WakaiandMisatofortheirkindnessandhelpinmanyways.IamalsoverygratefultomyparentsinKoreafortheirnever-endingtrust,helpandlove.IcanneverforgettheloveIgotfrommyuncleandaunt,KyunghwanandHannahBaikinIllinois.Theirhospitalityandsupporttouchedmyhearteverytimewewentvisitthem.Mythankfulheartgoestomyfamily:myamazingwife,EunyoungLeeforherjoyfulsoulandthousandsoflunchboxes,andmythreelovelydaughters,Seohyeon,SevinandYerin.Thankyouallforgivingmereasonstolive,workandlove.Lastbutnotleast,IgivethankstomyGodandsavior,whoismyallinall.Maythishumbleresearchgiveyougloryandhonor."Everygoodandperfectgiftisfromabove(James1:17).vTABLEOFCONTENTSLISTOFTABLES.......................................viiLISTOFFIGURES......................................viiiCHAPTER1HerdBehaviorintheKoreanBankLendingMarket........11.1Introduction.....................................11.2LiteratureReview..................................41.3MeasurementofHerdingIntensity........................81.3.1Methodology................................81.3.2Data......................................131.3.3AnalysisbyHerdingMeasures......................151.3.4RationalityinHerding...........................291.4EmpiricalAnalysisofFactorsthatAffectBank'sHerdinginLendingMarkets311.4.1MeasuresandVariables..........................331.4.2Estimation..................................351.4.3Results....................................361.5Conclusion......................................42REFERENCES.......................................44CHAPTER2TheExtensiveMarginEffectsofBankHealth,CreditShockontheFirm'sExportandtheRoleofCorporateGovernance.....472.1Introduction.....................................472.2DataandEstimation................................472.2.1DataSource.................................472.2.2Measurement................................492.2.3Model.....................................512.3RegressionResults.................................522.3.1TheLaggedLeveloftheCredit......................522.3.2TheLaggedChangeoftheCredit.....................662.4SummaryandConclusion.............................75REFERENCES.......................................76CHAPTER3TheSurveyofEffectsofFinancialShocksandCorporateGover-nanceontheFirms............................783.1Introduction.....................................783.2EmpiricalWorks...................................783.2.1FinancialShocks...............................783.2.2CorporateGovernance...........................80REFERENCES.......................................82viLISTOFTABLESTable1.1LSVmeasuresbyloanssector.......................24Table1.2SimplemeanofLSVforoverallloansmarketusingdatabyindustry(beforeadjustment).............................30Table1.3SimplemeanofLSVforoverallloansmarketusingdatabyindustry(afteradjustmentforrationalfactors)...................32Table1.4Thesummaryofstatistics..........................35Table1.5Estimationresultsofherdingintensityinoverallloansmarket....37Table1.6EstimationresultsofherdingintensityinlendingtoSME.......38Table1.7Estimationresultsofherdingintensityinlendingtohousehold....38Table2.1Exportdummyandlaggedlevelofthecredit(lagby1year)......54Table2.2Exportdummyandlaggedlevelofthecredit(lagby2year)......57Table2.3Exportdummyandlaggedlevelofthecredit(lagby1&2year)...60Table2.4Exportdummyandlaggedlevelofthecredit(averageofcredit)...63Table2.5Exportdummyandlaggedchangeofthecredit(lagby1year)....66Table2.6Exportdummyandlaggedchangeofthecredit(lagby2year)....69Table2.7Exportdummyandlaggedchangeofthecredit(lagby1&2year)..72viiLISTOFFIGURESFigure1.1Sharesofloansoutstandingbyallbanks.................16Figure1.2Sharesofloansoutstandingbyallbanks(detail).............16Figure1.3KHIit,t+kforoverallloansmarkets,k=1..................18Figure1.4KHIit,t+kforoverallloansmarkets,k=3..................18Figure1.5KHIit,t+kforoverallloansmarkets,k=5..................19Figure1.6KHIit,t+kforoverallloansmarkets,k=7..................19Figure1.7Herdingintheoverallloansmarket(KHI)................20Figure1.8Herdingbyborrowersector(KHI).....................21Figure1.9Herdingintheoverallloansmarket(LSV)................23Figure1.10Loansoutstandingbyeachbank......................27Figure1.11Simpleandweightedmeanforoverallloansmarket(KHI).......28Figure1.12Simpleandweightedmeanforoverallloansmarket(LSV).......28Figure1.13SimplemeanofLSVforoverallloansmarketcalculatedusingindus-trydata....................................31Figure1.14SimplemeanofLSVforoverallloansmarketbeforeandafteradjust-ment......................................32viiiCHAPTER1HerdBehaviorintheKoreanBankLendingMarketfiWhenthemusicstops,intermsofliquidity,thingswillbecomplicated.Butaslongasthemusicisplaying,you'vegottogetupanddance.We'restilldancing.fl(CiitgroupCEOCharlesPrince,inanexclusiveinterviewwiththeFinancialTimes,July2007)1.1IntroductionHerdbehaviorisfrequentlyobservedinmarketsnowadayscausingaturmoilinthemarketandinstabilityevenintherealeconomy.Therehasbeenagrowinginterestintheliteratureonherdingbehavior.Theoreticworkstrytoaddressthemotiva-tionsofherdbehaviorandmostofempiricalstudiesarefocusedontheherdinginthecapitalmarkets.However,consideringthatthebanksplayanimportantroleasintermediariesandtheimpactofherdingsideeffectsismoreattentionneedstobepaidtobank'sherding.Nakagawaetal.(2012)indicatesthatthedeteriorationoftherealeconomyinthe1990smayhavebeenattributablepartlytotheherdbehaviorintheJapaneseloanmarketduringtheperiodoftheeconomicbubbleinthelate1980s.More-over,theintensityofsuchherdingseemtogetmorepronouncedthaninthepastinlinewiththeaccelerationofglobalizationandliberalization.1Untilthetimejustbe-forethesubprimemortgagecrisiserupted,almosteveryinstitutionallovertheglobehadbeentradingonsubprimemortgagebackedsecurities,andthisherdingwas1Liu(2014)suggeststhatcontrollingformacroeconomicandsomecharacteristicsofbankingindustryinU.S.,herdinghasatrendofincreasingovertimeandherdinglevelinthe2000sishigherthaninpreviousdecades.1followedbyworldwidemeltdown.AssuggestedbyFreixas(2010),bankherdingbehaviorisoneofthefactorsthatledbanks'underestimateofriskandoverin-vestmentinmortgagelending.That'swhybankherdingitselfisimportantandneedtoreceiveexplicittreatmentinregulations.Indeed,herdbehaviorhasbeenconsideredastypicalbehaviorofKoreannancialinstitutionsincludingbanksforalongtime.IntheKoreanstockmarket,itiswellknownthatprivateinvestorswhohadlessinformationthaninstitutionalinvestorsshowedherdbehaviorespeciallyduringbullmarkets.Herdbehavioralsooccurredfrequentlyintheloanmarketsevenamonginstitutionsthatwasknowntohavemoreinformation.Evidencecanbeobtained,forexample,fromalotofarticlesinnewspaperssince2000.Thedramaticincreaseinthehouseholdloansoutstandingandthecompetitionamongcreditcardcompaniestoincreasetheircreditcardlendingin2001-2002isbelievedtohaveoriginatedfromherding.During2002-2003and2006-2007,theriseinlendingtosmallandmediumsizedenterprises(SME)isknowntobealsoattributabletobanks'herdbehavior.However,theinformationinthosearticlesisbasedmoreonanecdotalevidencethanempiricalstudy.That'swhymoreempiricalinvestigationshouldbeconducted.Furthermore,asthesysteminKoreahaslongbeenbank-oriented,andbankshavebeenplayingamajorroleintheeconomy,itisimportanttoexaminetheexistenceandcauseofherdbehavioramongKoreanbanks.Thepurposeofthispaperisasfollows.First,ImeasuretheextenttowhichbankscollectivelyincreaseordecreaseloanstocategoriesaswellasintheoverallloansmarketineachperiodbyusingtwoherdingindexescalculatedbasedonindividualKo-reanbank'sloansoutstandingdatabyborrowersector.Herdingbehaviorhasbeenmod-eledintheoreticalworksandwelldocumentedforcapitalmarketherding.However,onlyafewstudieshaveempiricallyinvestigatedherdingamongbanks,especiallyinbanklendingdecision.Thispaperisoneofthefewstudiesthatexaminesherdbehaviorindo-mesticlendingmarketforKoreancommercialbanksusingmorerecentandlongperiod2ofloansoutstandingdataforindividualbank.Toobtainherdingindexes,Ibasicallylookatloanssectors:largeSME(corporateloans),mortgage,housingfunds,andothers(householdloans).AndIapplythestandardmethodofLakonishoketal.(1992)(henceforthLSV)andamorerecentmeasurepresentedbyKang(2010)(henceforthKHI)toeachsector.Second,usingandcomparingtwomeasures,Iexaminethechangeinbanks'lendingbehavioroverthesampleperiodsbylookingatthetimeseriespatternofmeasures.andseeifwecanempiricalevidenceaboutexistenceofherdingforlendingtoasectorandinoverallloansmarketaswell.Also,Isomefea-turesrelatedtoherdbehaviorinKoreanlendingmarket.ThisstudymayserveasanempiricalcomparisonoftwomeasureindicatorsinthesensethatthispaperisthepaperusingbothtodetectKoreanbanksherding.Third,Iinvestigatefurtherindetailrationalityofherdbehavior.Rational(orspurious)herdingdenotessimilarresponsesbyinstitutionstocommoneconomicconditionsthataffecttheirlendingdecisions.Intheempiricalanalysisonherding,evenifastatisticalrelationshipbetweenloansofdifferentancialinstitutionsisdetected,therelationshipdoesnotautomaticallyimplytheexistenceofirrational(orintentional)herding.So,followingUchidaandNakagawa(2007),Iobtain(before-adjustment)LSVmeasurebasedonKoreanbanks'loansdatabyindustry,whichonlyexcludesrationalherdingbasedonmacroeconomicconditions.Then,adjustingfurtherforherdingresultingfromrationalbehaviorbycontrollingforrationalfactors,IexaminetheevidenceaboutirrationalherdinginKo-reanloansmarket.Finally,usingthemeasurebyKang(2010)whichallowsustogettheherdingintensityindexforindividualbank,Iempiricallyanalyzethefactorsthatinten-sifyherdingofKoreanbanksinthedomesticloanmarket.Liu(2014)alsoinvestigatesifthechangeinthepatternsofherdingmeasuresarerelatedwithmacroeconomicvariablesandbanksconditionalongtime.However,inordertoobtainexplanatoryvariablesforestimationrepresentingcharacteristicsofbanks,heaggregatesalldataonbalancesheetandincomestatementacrossallbanks,becauseLSVindexusedasdependantvariable3measuresonlyherdingofoverallbankingindustryforeachcategoryofloans.Inthispaper,however,Iinvestigateeffectsofcharacteristicsofindividualbankaswellasmar-ketstructural/macroeconomicfactorsontheintensityofindividualbank'sherding.Forthisreason,theherdingmeasurebyKang(2010)isusedasdependentvariablebecauseitgivesusherdingindexforindividualbank.1.2LiteratureReviewInthissection,Ireviewrelatedliteraturethatprovidesthreebasicexplanationsforherding:ofinformation,reputation/compensationmodel,andinstitutionalincen-tives.DevenowandWelch(1996)itasabehavioralpatterncorrelatedacrossindi-vidualagents,andBikhchandaniandSharma(2000)pointoutitisherdbehaviorwhenindividualeconomicagentsmakesimilarchoicesafterobservingothers'behaviors.Hir-shleiferandHongTeoh(2003)herdingasthesimilarityofbehaviorsresultingfrominteractionamongeconomicagents.Oneoftheprimaryreasonsforherdingistheofinformation.HerdingrelatedtoinformationamongeconomicagentswasanalyzedbyBikhchandanietal.(1992)andBanerjee(1992).Wheneconomicagentshaveinsufinformationorarenotcapableofanalyzinginformation,theyrefertoandimitatetheprecedingbehaviorsofothereconomicagentstomakedecision.Sias(2004)showssomeevidencethatimpliesinstitutionalinvestorsinstockmarketherdasaresultofdeducinginformationfromeachother'strades.Forbanks,BarronandValev(2000)showthatbankswithlesswealthtendtoavoidinformationgatheringandfollowbankswithmorewealththatcaneasilygetinformation.Therefore,smallerbankshaveanincentivetofollowthebehaviorof(presumablybetterinformed)largerbanks,andbanksareendogenouslyseparatedintoleadersandfollowers.Analternativeexplanationisthereputation/compensationmodeldevelopedby4ScharfsteinandStein(1990).Inthereputationmodel,abankmanagerhasincentivestomimicotherbanks'managersbecausemanagerswanttoprotecttheirreputation.Iftheactionresultsinagoodoutcome,he/sheEveniftheactionyieldsabadoutcome,thefactthatothermanagersmadeasimilarchoiceshieldsthemanager.Itissaferforthemtoerrtogetherthanindividually.Borioetal.(2001)arguethattherewardstructurethatlimitsblameinthecaseofcollective,asopposedtoindividualfailure,isthemostcommonfactorbehindbankherdingbehavior.Accordingtothecompensationmodel,itisdifforaprincipaltoevaluatetheabilityofanagent,he/sheevaluatestheagentbasedontheperformanceofotheragentswhoareengagedinthesamebusiness.Andthisleadstheagenthavingarisk-averseutilityfunctiontofollowtheotheragents'behaviorinordertonarrowthegapsintheirrewardsacrossdifferentcircumstances.Rajan(1994)alsopointsoutanexampleofbankmanagerswhotendtoimitateotherbanks'behaviorbyaccumulatingalesseramountofloanlossprovisionsinordertoavoidcriticism.Thepapersthatlookatinstitutionalincentivesareconcernedmostlywiththefollow-ingproblem:Ifonesmallbankfails,thedepositinsurancecanguaranteeallthedepos-itorsreceivetheirsavingsbackandtheeconomywillnotsufferfromthebankruptcy.Ifmanysmallbanksfailatthesametime,theconsequencesaredifferent.Thebankingsec-toriscrucialfortheexistenceofanymoderneconomy,sothefailureofseveralbankscandisruptthenormalfunctioningoftherealeconomy.Asaconsequence,thegovernmentiscompelledtorescuethebanks.Inotherwords,ifmanysmallbanksfailtogether,itisequivalenttowhenonebigbankfails.Thisisthemainreasonwhybankshaveincentivestocopytheactionsofitscompetitors.Ifabankisgoingtofail,itpreferstodosowithcompanysothechancesofbeingrescuedarehigher.AcharyaandYorulmazer(2007)carefullyevaluatethegovernment'stimeinconsistencyproblemthatleadsittobailoutbanksthatfail.Thepaper'smainargumentisthatwhenmanybanksfail,itisex-postop-timaltobailoutthebanks,whereaswhenfewbanksfail,thebanksthatsurviveacquirethosethatfail.Therefore,banksitoptimaltoherdsotheycansurviveorfailtogether5sincesurvivingalonemeanshavingtotaketheriskstoacquirefailedbanks.Thispaperalsoshowsthatsmallbankshavestrongerincentivestoherdthanlargebanks.Whenasmallbankimitatesalargebank,theprobabilityofbeingbailedoutishigherwhenthereisfailurethanifitdifferentiateditself.However,bigbankshavelittleextrafromcopyingasmallbank'sbehavior.Somerelatedstudiestendtointerpretanycollectivebehaviourofbanksasherdingthatcausesinefoutcomes.However,thiscollectivebehaviourdoesnotnecessarilycorrespondtotheexistenceofbankirrationalherdinganditsinef.Thecollectivebehaviourbybankscanstemnotonlyfromtheexistenceoftheirherding,butalsofrombanks'responsestochangesineconomicenvironmentsthatuniformlyaffecttheloande-mandbyborrowersandtheloansupplybybanks.However,asmentionedinBikhchan-daniandSharma(2000),itneedspointingoutthatempiricallydistinguishingrational(orspurious)herdingfromirrational(orintentional)herdingiseasiersaidthandoneandmayevenbeimpossible,sincetypicallyamultitudeoffactorshavethepotentialtoaffectaninvestmentdecision.UchidaandNakagawa(2007)excludefurtherrationalfactorsfromLSVmeasure,becauseevenifLSVmeasuredoesnotincludebanks'rationaldecisionbasedonoverallmacroeconomicconditions,thisdoesnotnecessarilyimplythatbanksareirrationalwhenLSVmeasureindicatestheexistenceofherding.Fewstudieshaveempiricallyinvestigatedwhetherherdingisobservedamongbanks,especiallyinbank'slendingdecision.SincetheseminalworkrelatedtobanksherdingintheloanmarketbyJainandGupta(1987),mostofstudieshasfocusedontheJapanesebanks,becauseithasbeenarguedthatherdbehaviorisatypicalactionofJapanesebanks,andtheirbehaviourhasbeenoftencriticizedasasymboloftheinefoftheJapanesemarket.JainandGupta(1987)andBarronandValev(2000)exam-inecausalityinloansoutstandingtoLatinAmericancountriesaggregatedbybanktypetoherdingamonggroupsofbanksintheU.S.in1980s.TheythatsmallU.S.banksreplicatethelendingbehavioroflargeU.S.banksinlendingtodevelopingcoun-6tries.Basedonthis,NakagawaandUchida(2011)investigateswhetherJapanesebanksfollowedherdbehaviorduring1980sand1990susingdataofloansoutstandingbytypesofbanks.Theirempiricalresultsshowtheevidenceforherdbehaviourinlendingbe-tweendifferenttypesofJapanesebanksin1980s.First,herdingisobservedintheearlytomid1980s,whichisimmediatelyafterthebeginningofderegulation.Sec-ond,herdingisobservedinloanstonewborrowerswhobecamenewcustomersafterderegulation.Finally,bankswereinclinedtofollowthelendingbehaviourofbanksthatwereconsideredtobemoreinformedaboutthenewborrowers.Nakagawa(2008)leader-followerrelationshipsbetweenlendingbehaviorbydifferenttypesofJapanesebanksinurbanandregionalcitiesrespectively.Theyimplythatlocalbanksfol-lowmajorbanksinurbancities,whilelocalbanksfolloweachotherinregionalcities.Inaddition,herdingismostlycausedbylocalbanksandmorefrequentinregionalcitiesthaninurbancities.However,thesestudiesinvestigatetherelationshipbetweenloansofdifferentgroupsofbanks.UchidaandNakagawa(2007)isoneoftheempiricalstudiesthatusetheherd-ingmeasurebyLakonishoketal.(1992)(LSV)2,calculatedusingthedataofindividualbanks'loansoutstandingtodetectherdingofJapanesebanks.Theyidentifythegreatextentofherdingofmajorbanksinlendingaroundthelate1970s,whenJapanwashitbythesecondoilcrisis,duringthebubbleperiodofthelate1980s,andinthe1990swhenthebubblecollapsedandstagnationfollowed.Also,adjustingfurtherforherdingresult-ingfromrationalbehavior,theyobservedevidenceontheexistenceofirrationalherdingonlyinthebubbleperiod.Liu(2014),usingCallReportdata,thatU.S.commercialbankshavebeenfollowingherdingbehaviorinthedomesticloanmarketfrom1976to2008.,herdingoccurredinfouroutofNBER-datedrecessionsduringthesampleperiodsandtherehasbeenasharpincreaseinherdingsinceearly2000s.Inspiteofmanystudiesexaminingtheexistenceofherdingbyinstitutions,2Seethedetailsinthenextsection.7itisdifttoanypreviousstudiesthatempiricallyanalyzethefactorsaffectingherdingintensityofbankespeciallyinlendingmarket.Liu(2014)istheonethatexploresexplicitlyrelationshipsbetweenherdingmeasuresandmacroeconomic/bankconditions.First,regressionresultsindicatesthatbankstendtoherdmorewhentherearemoreun-certaintiesinthemarketandsupportstheofinformationhypothesis.Second,itindicatesthatbankstendtofolloweachotherwhentheconditionofthebankingsystemsarelessfavorable.Third,hecalculatesherdingmeasuresforbanksofdifferentsizesandahigherlevelofherdingamongsmallbanksduringmostofthesampleperiods,suggestingthatherdinginlendingdecisionsismorelikelytobeinducedbyinformationconcernssincesmallbankshavelessinformationandincurhighercostofacquiringinfor-mation.Beforethat,Graham(1999)investigatedthefactorsaffectingherding.However,itisbasedontheherdingofinvestornewsletters.TheleadingnewsletterisperceivedinadvanceinGraham(1999)'sstudyandtheoptionalbehaviorofthenewslettersarejustastwo:tofollowornottofollowtheleader.Forthiscase,itiseasytograsptheintensitythattheothernewslettersfollowtheleadingonewithoutusingtheherdingindex.ForthelendingmarketofKoreanbank,ontheotherhand,nosurelyleadingbankseemstoexistandevenifthereis,thatbankcouldbereplacedwithothersovertimes.Notonlythat,asbanksdecidethesizeoftheirlending,theirnumbersofoptionsthebankscouldhaveareunlimited.Forthisreason,IusetheherdingindexwhichwillbederivedinSection3forempiricalanalysisofthefactorswhichcauseherdinginbank'slendingmarketinKorea.1.3MeasurementofHerdingIntensity1.3.1MethodologyTodemonstratetheexistenceandtrendofherdbehaviorandanalyzethefactorsinten-sifyingherdinginbanklendingamongKoreanbanksduringthesampleperiods,itis8necessaryforustomeasuretheherdingintensity.LSVmeasurestheintensityofherdbehaviorofmoneymanagersbycapturingdevi-ationofdecisionbyindividualmoneymanageronbuyingorsellingagivenstockfromoverallmarket'saverageinvestmentdecision.InLakonishoketal.(1992),theyuseequity-typefunddatafortheyearsbetween1985and1989inordertoestimatewhetherherdingtookplaceamongfundmanagersandshowthattherehadbeennoobviousherdbehav-ior.Sincethisstudy,theLSVmeasurehasbecomethemostfrequently-usedmeasuretoquantifyherdbehavioramongequityinvestors.Themeasureisastandardinthelitera-tureandappliedinnumerousstudiesmainlyforstockmarket.3.UchidaandNakagawa(2007)usedLSVmethodtomeasuretheherdingintensityinbanklendingmarket.TheLSVindexforanalysisofherdbehaviorinbanklendingtoaparticularcategoryiintimetisas:LSVit=jPitPtjEjPitPtj(1.1)wherePitistheproportionofbanksthatactuallyincreasedtheirloansoutstandingtocategoryiintimetandPtistheexpectedproportionofbanksthatincreasedtheirloansoutstandingint,whichiscalculatedasameanofalltheobservedPitsintimet.4Thiscanbeconsideredasanindicatorofthebanks'overalllendingpolicy.Ifeverybankindepen-dentlyexpands(orreduces)itsloansoutstandingtosectoriintimetwithprobabilityPt(or1Pt),theobservedvalueofPitbecomesclosetoPtandthetermwillbecomezero.If,ontheotherhand,bankscollectivelyincreaseordecreaseloanstocategoryiintimet,theobservedvalueofPitdepartsfromPt.Thetermofequation(1.1)thustheextenttowhichbanks'lendingpoliciestothesectoriintimetdeviatefromtheoveralllendingpolicyintimet.3Forexample,seeGrinblattetal.(1995),Wermers(1999),andChoeetal.(1999).SeeBikhchan-daniandSharma(2000)forsurveyofsuchstudies.MorerecentstudiesincludeWylie(2005).4Pit=XitNit,whereNitandXitarethenumbersofbanksthatwereactiveinissuingloansandthatincreasedloansoutstandingtocategoryiintimetrespectively.9Notethatevenifthereexistsnoherding,theexpectedvalueoftheabsolutetermofLSVitispositive.ThelasttermEjPitPtjissubtractedsoastonormalizethemeasureandmakeitsmeanzerounderthenullhypothesisofnoherding.BecausetheLSViscalculatedbasedontheassumptionofbinarydistributionwhere,ifthereisnoherding,banksincreaseloansoutstandingtocategoryiintimetwithprobabilityofPtanddecreasewithprobabilityof(1Pt),EjPitPtjrepresentsexpectationofbinarydistribution.5ValuesoftheLSVherdingmeasuresthuscanbeinterpretedasthetendencyofbankstoincreaseloanstoagivencategoryinagiventimeabovetherandomdistributionoflendingdecisions.PositivevaluesoftheLSVmeasurethatdifferfromzeroprovideevidenceofherdingbehavior.ThehighertheLSVmeasure,thestrongertheherding.ThentheoverallLSVindexforeachtimet(LSVt)isderivedbyobtainingasamplemeanofLSVmeasureamongallcategories:LSVt=1IIåi=1LSVit(1.2)where,Iisthenumberofloancategories.SincePtissuctractedfromPit,themeasuredoesnotincludebanks'rationaldecisionsbasedonoverallmacroeconomicconditions.However,thisdoesnotnecessarilymeanthatbanksareirrationalinthesensethatra-tionalbanksmakelendingdecisionsnotonlybasedonoverallmacroeconomicconditionsbutalsoonconditionstoindividualloansector,i.EvenifPtissubtractedfromPitinthemeasure,itcouldstillranincreaseordecreaseinloansoutstadingbasedonsectorrationalfactors.InordertogettheadjustedLSVherdingmeasurewhicheliminatessectorreasonsforherding,followingUchidaandNakagawa(2007),weestimatethefollowingequationbytheordinaryleastsquares:PitPt=a+bXi+eit,(1.3)5NitåXit=0hjXitNitPtjnNitCXitPXitt(1Pt)1Xitoi10whereXiisavectorofsectorcontrolvariables.eitrepresentstheportionofherd-ingwhichcannotbeexplainedbyrationalfactors.Therefore,wecanobtaintheadjustedLSVherdingmeasurethattheextentofherdingafteradjustmentasfollows:LSVAit=jeitjEjeitj.(1.4)LSVindexhassomeadvantagesoverothermethods:itmeasurestheherdinginten-sitymoreeasilythanothermeasuresdoandwecantestthestatisticalofherding.As,byPitfollowsabinomialdistributionwithmeanPtandvariancePt(1Pt)/Ni,wecangettheChi-squaredvaluesforeachloancategorybyastatistic,Zit=PitPtpPt(1Pt)/Nitwhichisknowntoapproximatethestandardnormaldistri-butionundersomeconditions.6Usingthisfeature,wecanthestatisticalevidenceofwhetherherdingexistsornotforthelendingtotheloancategoryatsomeperiod.Nevertheless,ithasthefollowinglimitations(BikhchandaniandSharma(2000)).First,inLSVmethod,theherdingintensityisdeterminedbythenumbersofinvestorsbuyingorsellingstocksforacertainperiod.Inreality,however,theinstockmarketsisdeterminedbythesizeofbuyingorsellingratherthannumbersofbuyersorsellers,andthiscouldmakeLSVindexdifferfromtheinstockmarkets.Second,us-ingLSVindex,itisimpossibletoidentifyintertemporalbehavioralpattern.measuretheherdingresultingfromtheinteractionamongeconomicagents.ThisisbecauseLSVindexappearshigherevenifinvestorsrespondtocommoninformationatthesametime.Third,althoughisispossibletomeasuretheaverageherdingintensityforacertaingroupbyus-ingthismethod,itisimpossibletomeasuretheherdingintensityofindividualagents.Fourth,althoughitispossibletomeasuretheherdingintensityamonginvestors/bankswithinacertaingroupbyusingthisindex,itisimpossibletomeasuretheherdingin-tensityforanentiremarket.Inrespectofanentirestockmarket,sellersexistwhenever6Fromequation(1.1),thisequals(thenon-absolutevalueof)thetermoftheLSVmeasurewhichisnormalizedforitsvariancetotakethevalueof1underthenullhypothesisofnoherding.SeetheUchidaandNakagawa(2007)formoredetailsonthetest(p568-569).11therearebuyers,whichmakesitimpossibletoeouttowardwhichdirectionherdingoccurredonlybyusingLSVmethod.AstheLSVmethodforanalyzingherdbehaviorinbanklendingmarketmeasurestheherdingintensityonthebasisofnumbersofbanksthatincreasesloansduringoneperiodonly,ithasthelimitationstheLSVpresentswiththeexceptionofthefourth.However,LSVindexisstillusedasstandardmeasureinmanyempiricalstudiestoinvestigatewhetherherdbehaviorisfollowedbyinstitutesincludingbanks.7Severalalternativemeasureshavebeenproposed.However,theyarelessusefulornotappropriatetocaptureherdbehavioramongbanks.Wermers(1994)developsamethodofmeasuringherdingintensitybyusingatimelaggedcorrelation,whichisknownasthePortfolioChangeMeasure(PCM)(BikhchandaniandSharma(2000))allowingustotakeintoaccounttheintertemporalbehavioralpatternbetweeneconomicagents.However,itisstillimpossibletoidentifytheherdingforanindividualagents.Forthisreason,fewstudiesuseitfortheirempiricalanalysis(UchidaandNakagawa(2007)).JainandGupta(1987)conducttheanalysisofthelendingbehaviorofU.S.banksusingtheGrangercausalitycoefItsolvesthesecondproblemwithLSV,butitisalsoimpossibletodetecttheintensityofindividualagents'herding.Kang(2010)suggestsanewherdingindexwhichmakesitpossibleforustomeasuretheherdingintensityofindividualbankaswellasthatofoverallbankingindustry.Wecanmeasuretheintensityofherdingforindividualbanksusingtimelagcorrelation.Thatis,theherdingindexofbanki(KHIit,t+k)forthek+1periods,whichisbetweentandt+k,isderivedbythefollowingtimelagcorrelationcoefKHIit,t+k=kåj=0(4Lmt+j14LmAt1,t+k1)(4Lit+j4LiAt,t+k)kdmt1,t+k1dit,t+k(1.5)7Forexample,seeWermers(1999),Choeetal.(1999),Wylie(2005),UchidaandNakagawa(2007),Liu(2014),12where,4Lmt+j1denotestheaveragegrowthrateofloansoutstandingofwholebanksexceptbankiatperiodt+j1,4LmAt1,t+k1,dmt1,t+k1theaverageandstandardde-viationof4Lmt+j1respectivelybetweent1andt+k1,4Lit+jgrowthrateofbanki'slendingatperiodt+jand,4LiAt,t+kanddit,t+ktheaverageandstandarddeviationof4Lit+j,respectivelybetweentandt+k.Byusingtheherdingindexobtainedabove,Iderivetheherdingindexesforallindividualbanks'lendingactivity,andtheaverageherdingindexofwholebankingsectorforeachofloanscategorybyaveragingthein-dexes.KHIt,t+k=1NNåi=1KHIit,t+k(1.6)whereNisthenumberofbanks.KHIisdeterminedbythesizeoflending,notjustthenumberofbanksthatexpandedlendingasinLSV.Also,itisderivedusingtimelagcorrelationcoefwhichcap-turestheherdingcausedbyinteractionamongbanks.AndunliketheLSVandPCMthatmeasureaverageherdingofacertaingroup,KHImeasuresherdingofindividualbank.Finally,thisindexderivescorrelationacrossseveralperiodsratherthanbetweentwope-riodsinordertosubstantiallyreducethenoisecausedbyirregularityoftimesseries.Inthispaper,IusebothLSVindexandKHItomeasureherdingintensity.Theformerisnotonlystandardoneintheexistingliteratures,butalsocapableofexcludingrationalherding.Also,thelatterisusefulinthesensethatitimprovesthepreviousoneswithmoretheoreticalconformityespeciallyforanalyzingtheherdingbybanksinthelendingmarket.1.3.2DataTheprimarydatasourceIuseforconstructingtheherdingindexesandcharacteristicvari-ablesofindividualbankusedintheestimationistheFinancialStatisticsInformationSys-tem(FISIS)andFinancialInstitutesManagementAnalysisSystemdatabasefromFinan-13cialSupervisoryService(FSS).Thesedatabasecontainapartofdataonbanks'individual-entitylendingactivityloanstheyreporttoFSSmostlyonquarterlybasis.Tocalculatetheherdingindexes,Iusetwotypesofloansoutstandingdatabyindividualbank.First,IuseloansdatabyborrowersectortoinvestigateextistenceofherdinginKoreanloansmarket,asitisusuallyclaimedthatbanksfollowherdinginlendingtosector.Theloansoutstandingdatabyborrowersectorarelargelycategorizedintothreeborrowergroups:household,andpublicsector.Theloanstothearedividedintotwosectors-thelargeandSME-andthehouseholdloansintothreesub-categories-mortgageloans,housingfunds8,andothers9.Thesecategoriesforloansdatabyborrowersec-tor-largeloans,SMEloans,mortgageloans,housingfunds,andothers-areusedtomeasuretheherdingbyLSVandKHIindexforthesampleperiodof2000:1Q-2012:2Q.10Second,theloansoutstandingdatabyindustryareusedtofurtherexplorerationalityofherding.Thedatatothefollowingnineindustriesbyindividualbankareavailablebe-tween2003:3Qand2012:2Q:(1)Agriculture,ForestandFisheries,(2)Manufacturing,(3)Construction,(4)WholesaleandRetail,(5)AccommodationandFood,(6)Telecommuni-cation,(7)RealEstate,(8)Serviceandothers,and(9)Household.LoansoutstandingtotheseninetypesofindustriesareusedtocalculateLSVbeforeandafteradjustment.MostofmacroeconomicfactorsusedasexplanatoryvariablesforestimationlaterarefromEco-nomicStatisticsSystem(ECOS)databasefromtheBankofKorea.Iusethequarterlydatabetween2000:Q1and2012:Q2.InKorea,commercialbanksconsistsofnationwidebanks,localbanks,andbranchesofforeignbanks.Inthispaper,banks'herdbehaviorinlendingmarketisfocusedonthedomesticloansinKoreanWon.Tothisend,branchesofforeignbanksareexcludedfromthesampleastheirproportionofloansoutstandingindomesticloansisverylow.8Housingfundsareloansmadeonlyforthepurposeofpurchasingorrentingthehouse.9Mostofthemare(unsecured)individualcreditloans.10Outofloanscategories,dataformortgage,housingfunds,andothersareavailableonlyfrom2003:3Q,whileloanstoSME,largeandhouseholdareavailablefrom2000:1Q.Besides,loanstopublicsectorareexcludedintheanalysisastheyaredeterminedmainlybydemand.1413commercialbanks(7nationalwideand6localbanks)thatarecurrentlyinbusinessarethesubjectsoftheanalysis,whilethosethatweretakenoverduringsampleperiodareincludedforourpurposeinmergedbanks.11Figure1.1presentssharesofloansoutstandingheldby13commercialbanksforeachsector.In2000,thecorporationloans(SMEandlargeoccupyover60%oftotalloansoutstanding.However,sincethen,theproportionofhouseholdloansincreasesfastuntilthelate2005,whentheyaccountforabout56%oftotalloansoutstanding.Asof2012:2Q,theshareofcorporationloansis50%andhouseholdloanshave48%ofshares.Amongcorporationloans,theproportionofloanstolargereducedfrom18%in2000:1Qto4%in2007:1Q,butsincethenitkeptincreasingtoabout12%in2012:2Q.LoanstoSMEshowrelativelystableshareof40%onaverageduring2000:1Q-2012:Q2.Loansoutstand-ingofloanstopublicandothersectorssuchaslocalgovernment,publiccorporationandnon-prorganizationhaveverylowshare.Figure1.2showsindetailtheproportionofeachloanscategoryforhouseholdloans.Mortgageloanstakesmostpartofhouseholdloansanditsportionhasincreasedovertime,whiletheshareofbothhousingfundsandothershasdiminished.Consideringthis,theintensiveherdinginthelendingtoSMEand/orinmortgageloanmarketcouldbemoretotheeconomy.1.3.3AnalysisbyHerdingMeasuresNow,Iexaminethechangeinbanks'lendingbehavioroverthesampleperiodsbylookingatthetimeseriespatternofmeasures.andseeifwecanempiricalevidenceaboutexistenceofherdingforlendingtoasector.Also,IsomefeaturesrelatedtoherdbehaviorinKoreanlendingmarketandinvestigateifsomecharacteristicofherdingisconsistentwiththerelevanttheories.11Theamountoflending,non-interestincome(commissionincome+otheroperatingincome),totalrevenue,equityandriskweightedassetareaggregatedtoderivethetotalloan,herdingintensityandBIScapital.15Figure1.1SharesofloansoutstandingbyallbanksFigure1.2Sharesofloansoutstandingbyallbanks(detail)16Here,Ilookattheherdingpatternusingtheloansoutstandingdatabyborrowersec-tor.Byusingtheequation(1.5),IderiveKHI,theherdingindexesforallindividualbanks'lendingactivity.Tothisend,thevalueofkneedstobedeterminedAsthevalueofkcanbeseentorepresenthowlongonebank'sinternaldecisionmakingprocessandfundamentallendingbehaviormaypersist,itisdeterminedbasedonthefollowingreasons.First,itisknownthattheKoreanbanks'herdbehaviorinthelendingtothespe-sectorhappenedeverytwoyearsandlasteduptotwoyearssince2000.12Itmeansthatthebanksarelikelytokeepthesamelendingbehaviorforthoseperiodsinteractingwithotherbanks.Second,itisassumedthatbanksinternaldecisionmakingandtheirlendingbehaviormaybeaffectedbythechangeintheCEO.BecausethetermofoffortheKoreancommercialbanks'CEOisusuallytwoorthreeyears,itislikelytochangethebank'slendingdecisionfundamentally.Consideringallthese,Iusetwoyears(eightquarters,k=7)averageherdingindexthroughouttheanalysis.NoticethatKHIderivesthecorrelationacrossk+1quarters,so,asexpectedintheequation(1.5),highervalueofkreducesthenoisemorecausedbyirregularityoftimeseries.(SeeFigure1.3-1.6)13Figure1.7representstheherdingbyallbanksinoverallloansmarket14.LookingatthetimeseriespatternoftheKHI,forthesimplicityofanalysis,Ifocusonthelocalpeakpointassumingthatbanks'herdingismostaroundthatperiod.Fortheoverallloansmarket,Koreanbanksseemtofollowacyclicalherdingpatternoverthesampleperiods.Theherdingmeasurereachedhistorichighof0.65around2003:1Q.Afterthat,twomorelocalpeaksappearin2005:4Qandthe2008:2Q,eventhoughtheydonotseemtobeasintensiveasonearound2003.12Householdcreditloan(2001-2002),SMEloan(2003-2004,2006-2007),householdmortgageloan(2005-2006)13Notethattheherdingindex,KHIit,t+7,atacertainperiodimpliestheaverageherdinginten-sity,whichconsistsofcurrentperiodandthefollowing7quarters(2years).So,inthegraphforKHIit,t+7,thehorizontalaxisist+2,whichisthemedianofperiodsoverwhichthetimelagcorrelationisderived.14NotethatwhenusingKHItocalculatetheherdingmeasureforoverallloansmarket,itisdirectlyobtainedbyusingequation(1.5)and(1.6)withtotalloansoutstandingdata.Ontheotherhand,theLSVmeasureforoverallloansmarketisobtainedbycalculatingthemeasureforeachloanscategoryinequation(1.1)andaveragingtheminequation(1.2).17Figure1.3KHIit,t+kforoverallloansmarkets,k=1Figure1.4KHIit,t+kforoverallloansmarkets,k=318Figure1.5KHIit,t+kforoverallloansmarkets,k=5Figure1.6KHIit,t+kforoverallloansmarkets,k=719Figure1.7Herdingintheoverallloansmarket(KHI)Figure1.8representstheherdingpatternforeachsector.15Theherdingindexforhouseholdloans,increasingsteadilyfromthebeginningofthesampleperiod,exceeded0.6duringtheperiodbetween2002:2Qand2003:1Q.ThisseemstostemfromthefactthatinstitutionsinKoreasharplyexpandedtheirlendingtohouseholdsasKoreangovernmenthaseasedregulationonhousingloanandtakenpolicytoencouragecreditcarduseinordertostimulatedomesticeconomicactivityin2001.Ontheotherhand,theoccurrenceofhighlyherdingexceeding0.6intermsofitsindexinthelend-ingtohouseholdmarketbroughtaboutsideeffectsofradicallyincreasingdelinquencyinhouseholdlending.Thus,thehistorichighlevelaround2003inoveralllendingmar-ketseemstobeledmainlybytheherdinginthelendingtothehouseholdsector.Also,herdingforthelendingtoSMEseemstofollowthecyclicalherdingpattern,whichlookssimilartothatfortheoverallloansmarket.ItcouldbeinferredthatKoreanbanks'herdbehaviorintheoverallloansmarketismainlycausedbyherdingforthelendingtoSME.15Sincetheloansoutstandingdataforthesub-categoriesofhouseholdareavailableonlyfrom2003:3Qasmentionedbefore,the(total)householddataareusedheretoshowwhichsectorcon-tributedtothehighlevelofherdingaround2003.20Figure1.8Herdingbyborrowersector(KHI)TheherdingmeasureforlendingtoSMEremainhighbetweenearly2003andearly2004.Theindexalsoreachedhigharound2006and2008.Whenitcomestothelendingtomortgagemarket,herdingindexwasatitshighestlevelaround2007andlate2008,buttheyseemtoberelativelylowerthanthatforlendingtohouseholdandSME.Also,wecanseethatwhiletheherdingindexesoflendingtoSMEshowedthethreelocalpeakswithrelativelyhighlevelofindexeseverytwoorthreeyears,theherdingintensityforlendingtohousehold,reachingthehistoricpeakof0.68in2003:1Qdroppedsharplyuntil2005:2Qandstayedbelowuptonowexceptduringacoupleofquarters.Thiscanbeexplainedbytwofactors.First,beforethecurrencycrisisin1997,Koreanbankswerestrictlyregulatedbothexplicitlyandimplicitlyinallbusinessareas(e.g.loanamounts,interestrates,borrowers,fees).Inparticular,thebankswereforcedtoallocatemorecredittotheratherthanhousehold.However,afteraseriesofialdereg-ulationmeasureswereintroducedinthelate1990salongwithrestructuringthesystemaswellasthebanksbegantoincreasetheloansoutstandingtohouseholdwhichwasconsideredtobelessriskybutmoreprthantheloanstothe21Deregulationhelpedtoboosttheloansoutstandingandloansharesofhouseholdbegin-ninginthelate1990s.16Inthoseperiods,herdingisobservedparticularlyinlendingtohouseholdpossiblyduetothelackofinformationaboutthenewcustomersector,whichisalsoconsistentwithJainandGupta(1987)andBarronandValev(2000).TheythatherdingbyU.Sbanksoccurredinthe1980s,whentheyincreasedlendingtoLatinAmer-icancountriesthatwerenewborrowersforthem.NakagawaandUchida(2011)alsoarguethatJapanesebanksfollowedtheherdbehaviorintheearlytomid1980s,whenbanksmadeloanstothenewemergingindustriesimmediatelyafterderegula-tionbegan.Second,afterthecreditcardcrisisin2003inKorea,everyeffortsweremadebygovernmentandbankregulatortodevelopthecentralizedsystemthatcollectsthepersonalcreditinformationfromeachbankandsharethemwithallthebanks.Ithelpedtolessentheasymmetricinformationprobleminthehouseholdloansmarketandcon-tributedtoreductionintheherdingbehaviorbybanksforlendingtohousehold.Finally,wecanseethattheintensityofherdinghasweakenedinrecentyears.TheLSVmeasureforKoreanbanksduring2003:4Q-2012:2QaredepictedinFigure1.9.Itlooksveryvolatile,soitisdiftodetectthetrendofherdingintensityrelativetoKHI.ThatisbecausetheonlycorrespondingperiodofloansoutstandingdataistakenintoaccounttocalculatetheLSVmeasureforthattime,whilethetimeseriescurveforKHIissmoothedoutbycapturingtheaverageintensityofherdingforthefollowingeightquarters.However,itcanbesaidthatthetimeseriespatternforhedingbyLSVispartlyconsistentwiththatofKHIinthesensethattheintensityofherdingdecreasesby2005anditreacheshighlevellate2006.Inaddition,herdingisinthequarterof2008and2009eachasisalsocasewithKHI.Ifweexistenceofherdingbasedonthe5%cancelevelandassumptionthattheherdbehaviorbybanksinlendingmarketpersistoversomequarters,wemayidentifywhenstatistically16Inthelate1996,theloansoutstandingbyKoreancommercialbankstohouseholdamountedto51trillionKoreanWon,whichaccountedforonly41%ofthattotheHowever,thehouseholdloansoutstanding,increasingover40%peryearfor4yearsfrom1999to2002,reached300trillionWoninthelate2005,whichsurpassedthe299trillionWonofloansoutstandingtothe22Figure1.9Herdingintheoverallloansmarket(LSV)herdingisfollowed.TheresultwiththeChi-squaredtestispresentedinTable1.1.Forherdingforoverallloansmarket(LSVt),whichisasthemeanofherdingmea-sureforeachloanscategoryintimet(LSVit),theevidenceofherdingcanbefoundfrom2006:1Qto2008:2Qexceptin2007:4Q.Also,theherdingseemstobefollowedinthehalfof2008,2009.and2010.WhenitcomestothelendingtoSME,theevidenceofherdingisobservedfrom2006:1Qtothehalfof2007.Inthehalfof2009andthelate2010,thereareevidencesofherding,buttheydidnotlastlong.Theherdinginthelendingtothelargeoccurredin2006,whenthedecreaseinloansoutstandingbyabout4banks(=30%of13banks)wascausedbyherdbehavior.Incaseofmortgageloan,thereseemstobenostrongstatisticalevidenceofherdbehavior.WiththeseresultsfromKHIandLSVmeasureobtainedusingthesameloansout-standingdatabysector,wecangetsomeKHIandLSVdoesnotnecessarilygivetheconsistentevidenceofherding.Thatisprobablyduetothedifferenceinthewayhowtheindexesarecalculated.However,itcanbeconcludedthatsince2000,acantextentofherdingwasfollowedbyKoreanbanksforlendingtotheoveralllending23Table1.1LSVmeasuresbyloanssectorTotalloanSMEloanHouseholdloanLSVtc2LSVitc2LSVitc22000.1Q0.08**10.4920.0431.4120.120*3.4882Q0.046.803-0.0340.2590.120*3.4883Q-0.021.9550.0150.880-0.0620.0984Q0.07**12.557-0.0920.0190.215**5.5662001.1Q0.057.4290.0461.2380.199**4.9522Q0.22***26.7880.181**4.6650.258***7.4853Q0.17***19.5000.131*3.2500.208**5.7784Q0.11***15.202-0.0920.0190.331***10.3102002.1Q0.12***14.1820.120*3.4880.120*3.4882Q0.11**12.6150.0762.0770.153**4.3333Q0.15***17.5380.153**4.3330.153**4.3334Q0.20***22.1000.125*2.9250.279***8.1252003.1Q0.11**12.1480.222**6.314-0.0090.5462Q0.12***14.1820.120*3.4880.120*3.4883Q0.19***21.2680.187**5.2700.187**5.2704Q0.15***21.327-0.0900.0200.140*3.3752004.1Q0.07*9.3510.145*3.692-0.0470.1972Q0.12***13.6190.123*2.7860.123*2.7863Q0.037.027-0.0820.0230.110*2.8344Q0.09**12.3500.125*2.9250.279***8.1252005.1Q-0.070.650-0.1050.000-0.0290.3252Q0.10**11.103-0.0130.4990.217**5.7713Q0.013.5190.0150.8800.0150.8804Q-0.041.6610.0301.071-0.0090.546Note:***,**,*representstheresultsinwhichthenullhypothesisofnoherdingisrejectedat1%,5%and10%levelrespectively.24Table1.1(cont'd)TotalloanSMEloanHouseholdloanLSVtc2LSVitc2LSVitc22006.1Q0.13***14.9500.279***8.125-0.0290.3252Q0.12***16.6540.0852.1240.162**4.1633Q0.15***14.9450.145*3.6920.145*3.6924Q0.13***19.1580.169**4.7890.0922.4442007.1Q0.05*8.4970.162**4.163-0.0690.0852Q0.04*8.6470.120*3.488-0.0340.2593Q-0.013.4670.0631.733-0.0910.0004Q0.003.519-0.0620.0980.0922.4442008.1Q0.11***13.4360.153**4.3330.192**5.7692Q-0.031.486-0.0510.124-0.0510.1243Q-0.013.2790.0331.148-0.0820.0234Q-0.012.105-0.0150.487-0.0150.4872009.1Q0.08**12.6150.233**6.2310.233**6.2312Q0.11***16.0300.0922.4440.0150.8803Q0.08**10.750-0.0090.5460.0681.7704Q0.12**12.6150.0030.6920.233**6.2312010.1Q-0.031.950-0.1050.000-0.0290.3252Q0.05*9.3890.208**5.7780.131*3.2503Q-0.060.391-0.0620.098-0.0620.0984Q0.09***13.6590.179**4.4930.294***8.8062011.1Q0.015.8800.1203.488-0.0720.0292Q0.014.3010.0150.8800.0922.4443Q0.14***19.7690.0331.1480.110*2.8344Q0.05*8.4500.202**5.2000.125*2.925Note:***,**,*representstheresultsinwhichthenullhypothesisofnoherdingisrejectedat1%,5%and10%levelrespectively.25marketduring2002-2003,whichismainlyattributabletoseverelyherdinginthelendingtohousehold,possiblyduetotheinformationproblem.causedbythedereg-ulation.Sincethen,herdingintheoveralllendingmarketandlendingtoSMEislikelytohavebeenintensitwomoretimesaround2006and2008-2009.Ontheotherhand,theherdingmeasureinthelendingtohouseholddroppedafterseverherdinginearly2000s,andstayedbelow0.Incontrasttotheanecdotalevidence,herdingbasedonKHIandLSVmeasuredoesnotseemtohavebeenfollowedinthemortgagemarketduringthesampleperiodHerdingtendstogetmostduringtimeswhenbankschangetheirlendingdecisionsintermsofscaleofincreaseinloansoutstanding.Thoseperiodsalsotendstocoincideswiththepeakandtroughinthebusinesscycles.Thatim-pliesthatbanksfollowherdbehaviormostwhenthereisuncertaintyaboutexpectationforeconomicconditions.TheanalysissofarisbasedonasimplemeanoftheKHIandLSVmeasureover13banksand9industriesineachquarterofthesampleperiod.Thereare,however,greatdiscrepanciesinloansoutstandingbyloanssectorandbankasdepictedinFigure1.1andFigure1.10.Therefore,itmightnotbeappropriatetouseasimplemeanoftheKHIandLSVmeasure.Tothisend,IcalculateaweightedmeanofKHIandLSVmeasureeachasfollows;WeightedKHIt,t+k=Nåi=1wiKHIit,t+k(1.7)WeightedLSVt=Iåi=1wiLSVit(1.8)wheretheweightforeachi,Wi,istheloansoutstandingtoaindividualbankidividedbytotalloansoutstandingineachsectorintimetforKHI,andtheloansoutstandingtoasectoridividedbytotalloansoutstandingforLSVmeasure.1717Notethatidenotesabankinequation(1.5)forKHIandloanscategoryinequation(1.1)forLSV.26Figure1.10LoansoutstandingbyeachbankFigure1.11and1.12plotstheweightedmeanofKHIandLSValongwiththesimplemeanmeasures.ComparingtwomeasuresforKHIinPanelA,wecanthatqual-itativeresultsfromsimplemeanmeasurestillholdsfortheweightedmeanmeasure.Thegeneraltrenddoesnotchangeevenifwefocusontheweightedmeanmeasure.How-ever,theweightedmeantoagreaterextentthansimplemeans.Inparticular,theweightedmeanmeasureshowshigherlevelofcorrelationduringtheperiodswhentheherdingishighlywhichimpliesthatbigbanksfollowtheherdingmoreintensively.Itcanbesaidthatseverecompetitionbetweenafewbigbankstoexpandloansinboomingand/orcontractloansincreaseinrecessionmaycauseinsta-bilityandofbusinesscycle.WhenitcomestoLSVmeasure,exceptacoupleofperiods,theweightedmeanLSVmeasureislowerthanthesimplemeanLSVduringsam-pleperiods,whichimpliesthatherdinginoveralllendingmarketismainlyattributabletotheherdingintheloanssectorsthathasrelativelylowportionofloansoutstanding.However,in2006:1Q,asherdingintensityinlendingtoSMEwasveryhighwhilethemeasureforlendingtootherloanssectorstayedlow,theweightedmeanLSVmeasure27Figure1.11Simpleandweightedmeanforoverallloansmarket(KHI)Figure1.12Simpleandweightedmeanforoverallloansmarket(LSV)28showshigherlevelthanthesimplemeanmeasure.Inaddition,comparingLSVtoKHIinFigure1.11and1.12,thetimeseriespatternofweightedmeanLSVlooksmoresimilartothatofKHIthanthesimplemeanmeasure.1.3.4RationalityinHerdingInSection3.3,Icalculatedtwoherdingmeasures,KHIandLSV,andtriedtodetecttheex-istenceofherdingintheKoreanloansmarketusingloansoutstandingdatabyloanssec-tor.EveniftheKHImeasureimprovedthepreviousonesincludingLSVwithmorethe-oreticalconformity,ithaslimitstooutirrationalherdbehaviorwhereindividualbankdoesnotmakelendingdecisionbasedonitsowninformation,butjustfollowotherbank'sdecision.Here,followingUchidaandNakagawa(2007),IinvestigatemorecloselytherationalityofthedetectedherdingbycalculatingtheadjustedLSVmeasureusingKo-reanbanks'loansoutstandingdatabyindustry.theadjustedLSVmeasureexcludestherationalherdingbasedonnotonlymacroeconomicconditions,butalsofactors.ThesimplemeanoftheLSVmeasureobtainedfor2003:4Qthrough2012:2Qarerep-resentedinTable1.2anddepictedinFigure1.13.Wecanthatsince2005:3Q,timeseriespatternofherdingissomewhatsimilartooneobtainedusingtheloansdatabysector,while,beforethat,itlooksdifferentfromeachother.ItcouldbeduetothefactthattwoLSVmeasuresarecalculatedusingdifferentcategoryofloansoutstandingdata.AsBikhchandaniandSharma(2000)pointedout,theLSVmeasuremaynotbeagoodmeasureunlessthedatacategoryoverwhichthemeasureisaveragedareappropriatelychosen.However,theempiricalevidenceaboutherdingisobservedinmanyofsampleperiods.Inaddition,notethat,asmentionedinSection3.1,LSVmeasuresobtainedheredoesnotincludebanks'rationaldecisionbasedonoverallmacroeconomicconditions.Now,inequation(1.3),Iusethefollowingtwovariablestofurthercontrolforrationalfactors.First,Iuseavariablerepresentingtherela-29Table1.2SimplemeanofLSVforoverallloansmarketusingdatabyindustry(beforeadjustment)LSVtc2(9)LSVtc2(9)2003:4Q0.069***26.1852008:1Q0.095***31.6222004:1Q0.048**18.147-0.0027.8290.0089.0190.052**18.0510.071**21.0870.047**19.1470.105***36.1692009:1Q0.081***22.8242005:1Q0.054*16.5710.03212.7170.086***25.9070.02411.3960.01511.3980.121***32.6370.045**20.0452010:1Q-0.0096.2312006:1Q0.028**16.9670.037*15.0160.077**19.675-0.0058.6910.063**21.6060.088***26.9040.048**20.7692011:1Q0.03014.4472007:1Q0.00513.8970.03314.2680.038*15.5270.0169.6830.0017.3580.043**17.8040.055***24.0392012:1Q0.077***23.3240.00911.333Note:***,**,*representstheresultsinwhichthenullhypothesisofnoherdingisrejectedat1%,5%and10%levelrespectively.tiverealGDPgrowthratebyindustryfortheeightindustriesexcept'Household,'andthegrowthrateofrealconsumptionexpenditureofhouseholdfor'Household.'Thesevariablesaretocontrolfortherelativeprofeachindustry.ThedataareavailablefromReportonNationalAccounts.Second,housingpricesareusedasacontrolvariable.InKorea,housingishighlyvaluedascollateral.Itakethenationalhousingpriceasanexplanatoryvariable,whichisavailablefromSurveyonNationalHousingPricebyKook-minBank.Iapplythepriceindexonlytothreeindustries:Construction,RealEstate,andHousehold.Iconstructavariablewhichisaproductoftherateofincreaseinindexanddummyvariablethattakesthevalueof1forthethreeindustriesand0fortheothers.30Figure1.13SimplemeanofLSVforoverallloansmarketcalculatedusingindustrydataTable1.3presentsthesimplemeanofherdingmeasuresadjustedforrationalfactors,andFigure1.14depictstheadjustedmeasuresalongwiththeunadjustedmeasures.Wecanthatevenafteradjustmentforrationalherding,theevidenceaboutherdingstillholdsinalmostsameperiodsasbeforeadjustment.Thedifferencebetweensolidanddottedlinesrepresentstheportionofherdingwhichwasadjustedbyrationalfactors.Duringmostofthesampleperiods,fewofherdbehaviorcanbeexplainedonrationalgrounds.ItimpliesthatpurelyirrationalherdingwasdominantinKoreanloansmarket.1.4EmpiricalAnalysisofFactorsthatAffectBank'sHerd-inginLendingMarketsInthelastsection,Ifoundevidenceofherdinginbanks'lendingdecisiontodifferentloancategories.Inthissection,Iinvestigateifthetimeseriespatternsofindividualbank's31Table1.3SimplemeanofLSVforoverallloansmarketusingdatabyindustry(afterad-justmentforrationalfactors)LSVtc2(7)LSVtc2(7)2003:4Q0.080***29.1762008:1Q0.087***30.0002004:1Q0.051***20.349-0.0186.4740.02411.4500.032**15.8110.080***23.2790.038***19.3390.105***36.4112009:1Q0.093***23.0802005:1Q0.043**15.6250.035*13.7250.063***20.0170.02511.715-0.0088.1880.132***35.4250.047***19.4162010:1Q-0.0136.2992006:1Q0.010*13.7460.01211.4100.056**15.517-0.0205.7780.053**17.9350.077***23.9330.031***19.4362011:1Q0.024*12.2802007:1Q0.024**15.6740.02111.9690.023*12.756-0.0057.096-0.0056.2740.053***18.9320.056***23.7572012:1Q0.080***24.2730.0079.440Note:***,**,*representstheresultsinwhichthenullhypothesisofnoherdingisrejectedat1%,5%and10%levelrespectively.Figure1.14SimplemeanofLSVforoverallloansmarketbeforeandafteradjustment32herdingmeasuresarerelatedwithindividualbank'scharacteristicsand/ormarketstruc-turalandmacroeconomicfactors.1.4.1MeasuresandVariablesFollowingKang(2010),weusenon-interestincome/totalrevenueandunsecuredlend-ing/totallendingratio,theBIScapitaladequacyratioasthecharacteristicsofindividualbank.Also,theratioofM1/Lf,thespreadbetweendepositinterestrateandloaninter-estrate,theIndex(HHI),andthegapbetweentheCCI(Coinci-dentCompositeIndex)anditsbenchmarkvalueof100areusedasthemarketstructuralandmacroeconomicfactors.Thenon-interestincome/totalrevenueandunsecuredlend-ing/totallendingratioareproxyofinformationamountthateachbankholds.Therefore,thelowerthoseratio,thelowerpossibilityofherding.TheBIScapitaladequacyratioisusedtomeasuretheextenttowhichbanksemphasizetheirlong-termprAsthebankswithlowBIScapitalratiotendtopursuemoreonshort-termprtheyareex-pectedtofollowherdingmore.Asmarketstructuralandmacroeconomicfactors,TheM1/Lfratioisusedasaproxythatmeasurestheratioofshort-termliabilitytototalliabil-ity.Thus,thehigherthisratio,themoreintenseherdingisexpected.Fortheloan-depositinterestratespread,themoreintensethecompetitionamongbankstomakeloansis,thefurtherthespreadbetweendepositandlendinginterestratenarrows.Asaconsequence,itisexpectedthataincrease(decrease)inthespreadbetweendepositinterestrateandloaninterestratewillgenerate(reduced)herdinginbanklending.TheHHIisaddedtomeasurethebankmarketconcentrationasaresultofM&Abetweenbanks.TheincreaseoftheHHIisalsoexpectedtobeaccompaniedbyariseintheherdingindex.Takingintoaccountthebusinesscyclewheretheeconomicagents'expectationschange,theCCIgapisusedasaindependentvariable.Consideringthatherdingcanbeintensi-whenthebusinesscyclemoresharply,herdingisfollowedmoreintenselywhentheCCIgapwidens.33Besides,asarobustcheck,Iaddthebankbalancesheet-depositratio,andliquid-ityratio-andincomestatementvariables-returnonassets(ROA),netinterestmargin(NIM),andtheratioofnon-performingloanstototalbankloans-tothebaselineoneatatimeasindependentvariable,followingLiu(2014).Iusethefractionofassetsfundedwithdepositsasanindicatorofthebank'sabilitytoprovideliquiditytoborrowingWhendepositratiosarehigher,banksmaybemoreabletoabsorbshockstothepricingofotherliabilitiesandthereforechangelendingstandardslessfrequentlyormoreslowly.Thebanks'liquidityratio,calculatedastheratiooftheaggregatedliquidassetstotheto-talassets,isexpectedtoshowthenegativerelationshipwiththeherdingindex,becausewhenthedepositratioislow,banksarelessabletocovertheirloanlossesorlossesfromotherinvestments.TheROAandNIM,asaproxyforbanks'overallpr,arealsoexpectedtoshowanegativerelationshipwithherdingintensity,whilethecoefontheratioofnon-performingloans,asameasureofbanks'expectedlossfrombadloans,isexpectedtobepositive.Lastly,dummyvariablesforbanksunderforeignmanagementandtimetrendarein-cludedamongexplanatoryvariablestocontroltheotherfactors.ThisisbecausebanksunderforeignmanagementinKoreaaccountforalargepartoftheKoreanbankingin-dustry18andsomepointoutthattheirmanagementpracticesdiffergreatlyfromthoseofbanksunderdomesticmanagement.Iaddatimetrendvariableinordertocontrolotherchangesinstructuralfactorsbesidemarketstructureandmacrofactorswhichcouldhaveeffectsonherding.Thebasicstatisticsofexplanatoryvariablesinthisanalysisarepre-sentedinTable1.4.18BanksmanagedbyforeignerssuchasKoreaExchangeBank,StandardCharteredBank,andCitiBankamountto15.8%outoftotalloansheldbyallcommercialbanksonaverageduringthesampleperiod.34Table1.4ThesummaryofstatisticsExplanatoryVariablesMeanStd.DevInformationCapacityUnsecuredLoan/0.40600.0779TotalLoanUnsecuredSMELoan/0.446800.0938TotalSMELoanUnsecuredHouseholdLoan/0.33820.1033TotalHouseholdLoanNon-interestIncomeRatio34.565322.6808Long-termPrConcernBISCapitalAdquacyRatio11.97221.9485Short-termLiquidityRatioM1/Lf0.213400.0261IntensityofCompetitionDepositRate-LoanRate-1.95000.3382DepositRate-SMELoanRate-1.96040.4516DepositRate--2.09380.4606HouseholdLoanRateMarketConcentrationHHI0.15130.0067BusinessFluctuationAbsolutevalueof1.08130.8390(CoincidentCompositeIndex-100)1.4.2EstimationTheempiricalanalysisisconductedbasedontheperiodbetween2000:1Qand2011:4Q.19andthequarterlydataareusedfortheanalysis.Forthedependentvariable,theKHIfortotallending,lendingtoSMEandlendingtohouseholdareusedherdinginlendingtoSMEandhouseholdcouldbemorelikelytoresultinbankinsolvency.Asdependentvariables,IuseKHIit,t+7representingtheaverageherdingintensitiesofindividualbanks'totallending,lendingtoSMEandlendingtohouseholdduringtheeightquarterssinceperiodt.Thepanelanalysisestimationequationisasfollows:KHIit,t+7=ai+bXit+gIt+eit,(1.9)whereKHIit,t+7denotesherdingintensityofbankiduringthe8quartersbetweent19Asdependentvariablesaretheaverageintensityofherdingduring8quartersbeginningperiodt,theherdingintensitybetweenthequarterof2010andthefourthquarterof2011becomesthelastdependentvariables.Therefore,thenumbersofdependentvariablesusedinempiricalanalysisbecome533(13commercialbanks*41quarters)35andt+7,aidummyvariablesofindividualbanks,Xitcharacteristicvectorofbankiatperiodt,Ittimeseriesvariablevector,andbandgcoefvectors.1.4.3ResultsTheestimationresultsforherdingintensityforthetotallendingarepresentedinTable1.5.Itdemonstratesthattheonly'unsecuredlending/totallending'variableiscantamongthecharacteristicsofbanks.Italsorevealsthat,amongtimeseriesvariables,the'depositinterestrate-loaninterestratedifferential'andthelendingmarketHHIareallandthatthesignsoftheestimatedcoefareconsistentwiththeex-pectations.TheresultseemsthatmanyofthecoefofexplanatoryvariablesarenotHowever,itdoesnotnecessarilyimplythatthesevariableswillnotaffecttheherdingintensityofindividuallendingmarketssuchaslendingtoSMEorhousehold.Thisisbecausetheeffecttheexplanatoryvariableshaveonlendingtoonesectorcoulddifferfromthatonlendingtoothersectors.Evenifherdingdoesnotariseintotallending,herdinginlendingcouldbringaboutexcessiveoffundsintooroutofasec-tor,whichcausessideeffectssuchasgrowinglossofbanksorinassetprices.Lookingintotheestimationresultsindetail,thecoefof'depositinterestrate-av-eragelendinginterestratedifferential'havepositivevalues,andthisimpliesthat,whencompetitionamongbanksbecomesce,theherdbehaviorofbankscouldbeGiventhatcoefoflendingmarketHHIisestimatedtobeavalue,itisfoundthattheherdingintensityofbanksdeepensinacasethemarketishighlyconcentratedbyreducednumbersofbanksduetoconsolidationamongbanks.However,thevariablesrepresentingtheofbusinesscycle,arenotshowingthatthereisnoempiricalevidencethattheherdingofbank'stotallendinghascorrelationwithbusinesscycle.TheresultsofempiricalanalysisoftheherdingintensityoflendingtoSMEandhouse-holdaredisplayedinTable1.6andTable1.7respectively.Incontrasttotheestimation36Table1.5EstimationresultsofherdingintensityinoverallloansmarketTotalLoanCoeff.Std.ErrorUnsecuredLoanRatio0.656740.2735**Non-interestIncomeRatio0.000050.00138BISCapitalAdquacyRatio0.0150.0132ForeignBankDummy-0.011020.1168M1/Lf0.691861.284DepositRate-LoanRate0.1410.08724*HHI25.107227.02556***BusinessFluctuation0.001960.02186TimeTrend0.01610.01939Observation533R20.1254oftotallending,fortheestimationoflendingtoSME,Iusethe'unsecuredlendingtoSME/totallending'variablesasaproxyforcapacitytoproduceinformation,andthe'de-positinterestrate-SMEloaninterestratedifferential'asatimeseriesvariableofthein-terestratedifferentialrepresentingtheintensityofcompetitionamongbanksintheSMEloanmarket.Similarly,Iusethe'unsecuredlendingtohousehold/totallending'vari-ables,andthe'depositinterestrate-householdloaninterestratedifferential'asatimeseriesvariablewhenherdingintensityforhouseholdloanmarketisusedasdependantvariable.IcomparetheestimationresultswhichusetheherdingintensityoflendingtoSMEswiththoseusingtheherdingintensityoflendingtohouseholdasadependantvariable.Amongexplanatoryvariablesrepresentingthecapacitytoproduceinformation,thees-timatedcoefofunsecuredlendingratioappearsnegativewhentheherdingintensityoflendingtohouseholdsisusedasadependentvariable(Table1.7),whileitiswhentheherdingintensityoflendingtoSMEisusedasadepen-dentvariable(Table1.6).Thisisbecausetheextenttowhichtheproportionofunsecuredlendingrthecapacitytoproduceinformationcandifferbetweenlendingtohouse-holdandSME.Forinstance,asmortgagelendingtohouseholdagainstthecollateralvalue37Table1.6EstimationresultsofherdingintensityinlendingtoSMEDep.VarSMELoanCoeff.Std.ErrorUnsecuredLoanRatio-0.202410.26501Non-interestIncomeRatio-0.001650.00131BISCapitalAdquacyRatio-0.044180.01188***ForeignBankDummy0.304240.11061***M1/Lf4.1763441.70277**DepositRate-LoanRate0.072220.09434HHI-9.860896.82077BusinessFluctuation0.052850.021184**TimeTrend-0.0012980.02679Observation533R20.1692Table1.7EstimationresultsofherdingintensityinlendingtohouseholdHouseholdLoanCoeff.Std.ErrorUnsecuredLoanRatio-1.270220.18934***Non-interestIncomeRatio0.002180.00123*BISCapitalAdquacyRatio0.022810.01167*ForeignBankDummy-0.011360.10432M1/Lf0.427810.7666DepositRate-LoanRate0.286820.04692***HHI37.587374.85783***BusinessFluctuation0.005030.01943TimeTrend-0.041360.01334***Observation533R20.387838ofhousing,ishighlyliquidsothatbanksinKoreacaneasilyforecloseonthemortgageandseizethecollateraltoretrievetheirfunds.However,thevalueofproductionfacili-tiesorfactorysites,whicharemortgagecollateralforlendingtoSME,notonlydeclinessharplyinlinewiththedeterioratedoperatingperformanceofthecorrespondingbutalsoassetsarealsolessliquid,whichmakesitdiftoovercometheproblemofin-formationinsufbysecuringsuchSMEloans.Forthesereasons,itishighlylikelythattheproportionofunsecuredlendingtohouseholdiscloselylinkedtobanks'capac-itytoproduceinformation.However,whileitisrelativelylesslikelythattheratioofunsecuredlendingtoSMEiscloselylinkedwithbanks'capacitytoproduceinformation.Therefore,theeffectoftheproportionofunsecuredlendingonherdingintensityishighlylikelytobecomemoreobviousinlendingtohouseholdsratherthaninlendingtoSME.Fortheproportionofnon-interestincome,theestimatedcoefiswhentheherdingintensityoflendingtohouseholdisusedasadependentvariable,However,itisntwhentheherdingintensityoflendingtoSMEisusedasadependentvari-able.Amongthetypesofnon-interestincomeusedasexplanatoryvariables,thecommis-sionincomesandotheroperatingprareincluded.Andcommissionincomegainedusuallythroughretailbankingandtransactionswithhouseholdsaccountsforarelativelylowproportionofnon-interestincome.Therefore,itcanbeassumedthatthelargerpro-portionofnon-interestincome(thelowertheproportionofinterestincome),thelowerweightofretailbankingandhouseholdrelatedbusinessandthelargerproportionofcor-poratebusiness.Thisimpliesthatalthoughalargerweightofnon-interestincomemayweakenabank'scapacitytoproduceinformationrelatedtohousehold,itmaynotweakentheonerelatedtoSME.Itfollowsthatitispossiblethattheestimatedcoefoftheproportionofnon-interestincomemaybeforlendingtohouseholdwhereasitisforlendingtoSME.Second,whenweusethecapitaladequacyratioasanexplanatoryvariable,thesignsofestimatedcoefinlendingtoSMEarenegativeasexpectedanditsis39alsohigh.Highcapitaladequacyratioscouldresultinlowherdingintensityasincentivesarisetomanagebanksfromalong-termperspective.Thisseemstostemfromthefactthatbankswhichpursuethelong-termprbyplacingtheirpriorityonmanagementstabilitytendtoshowlessherdbehaviorintheirlendingtoSMEthanintheirlendingtohousehold.Inpractice,theriskofdelinquencyinlendingtoSMEishigherthanthatinlendingtohouseholds.20Third,whenthedummyvariableforcommercialbanksmanagedbyforeignersisusedasanexplanatoryvariable,theestimatedcoefisnotwheretheherdingintensityoflendingtohouseholdisusedasthedependentvariable.However,itexhibitsapositivevaluewhentheherdingintensityoflendingtoSMEisusedasadependentvariable.ThiscanbeinterpretedthatbanksunderforeignmanagementaremorelikelytoherdinthemarketoflendingtoSME,astheyhaveinsufinformationonthedomesticmarket.Althoughtheresultsfortheregressionasarobustcheckarenotreportedinthetables,theliquidityratioandROAshowsarelationshipwiththeindividualbank'sherdingindex,especiallyforthecasewheretheherdingforthelendingtoSMEisusedasadependentvariable.Itimpliesthatthebank'sherdingintensityismoresubjecttothebank'shealthorprabilityintheSMElendingmarket,ratherthaninthehouseholdloansmarket.Next,Iexaminetheestimatedcoefsofmarketstructureandmacroeconomicvariables.First,M1/Lfratioisshowntohaveapositivevalueintheesti-mationofherdingofthelendingtoSME.Thisimpliesthatwhencontrollingforothervariables,theincreaseinshort-termliquidityhasaneffectonmoreherding20ThedelinquencyrateofloanstoSMEremainshigherthanthatofloanstohouseholdonabasisofKoreancommercialbanklending.Thegapofdelinquencyratesbetweenlendingto(large+SME)andhouseholdrecordedaverage0.8%pbetweenthesecondquarterof2003andthefourthquarterof2011.Notablyithasexpandedto1.6%pduringtheandsecondquarterof2009whenthecreditcrunchwasseverelyraging.ConsideringthatthedelinquencyrateofSMEisusuallymuchlargerthanthatofthelargethegapofdelinquencyratesbetweenlendingtoSMEandhouseholdwouldbelarger.40ofthelendingtoSME.However,itisnotwhentheherdingofthelendingforhouseholdisusedasadependantvariable.Second,Iexaminetheestimatedcoefoftheexplanatoryvariableofdepositin-terestrate-loaninterestratedifferentialwhichrtheintensityofcompetitionamongbanks.WhenlendingtoSMEisusedasadependentvariable,thecoefispositiveasexpected,butnotInthecasewherelendingtohouseholdisusedasade-pendentvariable,thecoefhasapositivevalue.Itindicatesthatthemorececompetitiontomakeloansintetheherdinginthehouseholdlendingmarket.Third,theestimatedcoefofHHIrepresentingthedegreeofmarketconcentra-tion,hasapositivevaluewhenthelendingtototalhouseholdareusedasdependentvariables.However,ithasavalueinthecasethelendingtoSMEisusedasadependentvariable.Thetransactionallendingwhichbanksdealwithinastandardizedmethodbyusingpubliclyavailableinformation,ismostlysmallsizedlend-ingtohousehold.Therefore,theincreaseintransactionallendinginlinewithconsolidationofbanksishighlylikelytooccurinthelendingtohousehold.Fourth,IlookintotheestimatedcoefofexplanatoryvariableoftheCCIgap,whichrepresentsthesizeofbusinessWhentheherdingintensityoflendingtoSMEisusedasadependentvariable,thecoefofthecorrespondingexplanatoryvariableshasapositivevalue.ThisimpliestheherdingoflendingtoSMEisgettingastheeconomicboomingorrecessdeepens.AslendingtoSMEtendstorespondmoresensitivelytoeconomicconditionthanhouseholdloans,itisexpectedthattheherdingwouldbemostaroundthepeakortroughofbusinesscycle.However,theestimatedcoefofthesizeofbusinessisnotinthecasethelendingtohouseholdisusedasdependentvariable.Finally,theestimatedcoefoftimetrendintheestimationofhouseholdloanisnegative,whichsuggeststhatintensityofherdingforhouseholdloanlessensovertime.Itisconsistentwithourinthelastsection.411.5ConclusionThisstudymeasurestheherdingintensityofbanks'herdingforsomeparticularloancategoriesaswellastheoveralllendingdecisionintheKoreanloanmarket,usingtwoherdingindexes:LSVwhichisastandardintheexistingliteraturesandKHIwhichismorerecentoneandallowsustoobtaintheherdingindexofindividualagents.Withherdingmeasurescalculatedusingloansoutstandingdatabysector,thispaperlooksatthetimeseriespatternofmeasurestoexaminethechangeinbanks'lendingbehavioroverthesampleperiodandempiricalevidenceaboutexistenceofherdingforlendingtoasectorbycomparingtheresultsfromKHIandLSV.Afterthat,usingtheloansdatabyindustry,IcalculateLSVherdingmeasuresbeforeandafteradjustmenttofurtherinvestigatetherationalityofherding.Additionally,usingKHI,thisstudyconductedanempiricalanalysisaboutthecausesofherdinginbanklendingbasedonKoreanbankpaneldata.KHIandLSVdoesnotnecessarilygivetheconsistentevidenceofherding.Thatisprobablyduetothedifferenceinthewayhowtheindexesarecalculated.However,itcanbeconcludedthatsince2000,aextentofherdingwasfollowedbyKo-reanbanksforlendingtotheoveralllendingmarketduring2002-2003,whichismainlyattributabletoseverelyherdinginthelendingtohousehold,possiblyduetotheinformationproblem.causedbythederegulation.Sincethen,herdingintheoveralllendingmarketandlendingtoSMEislikelytohavebeentwomoretimesaround2006and2008-2009.Ontheotherhand,theherdingmeasureinthelendingtohouseholddroppedafterseverherdinginearly2000s,andstayedbelow0.Incontrasttotheanecdotalevidence,herdingdoesnotseemtohavebeenfollowedbyKoreanbanksinthemortgagemarketduringthesampleperiod.Also,Herdingtendstogetmostduringtimeswhenbankschangetheirlendingdecisionsintermsofscaleofincreaseinloansoutstandingundertheincreaseduncertaintyaboutexpectation42foreconomicconditions.Anempiricalanalysiswasconductedintothefactorscausingtheherdbehaviorofbanks.Theresultsmaybesummarizedasfollows.Intermsofmacroeconomicfactors,whenlendingmarketisconcentratedandorcompetitionamongbanksbecomescer,herdingintotallendingandlendingtohouseholdsbecomesfurtherHowever,itseemsthatherdinginlendingtoSMEisfurtherwhenfundsintoshort-termproductsandgreatlywhenthebusinesscycleisclosetoatroughorapeak.Also,controllingforotherfactors,intensityofherdingforlendingtohouseholddecreasedduringthesampleperiod.Amongbankcharacteristicvariables,anincreaseintheunsecuredloanratiorepresentingthecapacitytoproduceinformationresultslargelyinherdinginhouseholdlendingandtotalloanlending.ItislikelythatherdingforSMEandhouseholdismorewhenbankpursuestheshort-termprMeanwhile,thebankunderforeignmanagementfollowsherdbehaviorinthelendingmarkettoSME.43REFERENCES44REFERENCESAcharya,V.V.andT.Yorulmazer:2007,`Toomanytofail?Ananalysisoftime-inconsistencyinbankclosurepolicies'.Journalofintermediation16(1),1Œ31.Banerjee,A.V.:1992,`Asimplemodelofherdbehavior'.TheQuarterlyJournalofEco-nomicspp.797Œ817.Barron,J.M.andN.T.Valev:2000,`InternationallendingbyUSbanks'.JournalofMoney,CreditandBankingpp.357Œ381.Bikhchandani,S.,D.Hirshleifer,andI.Welch:1992,`Atheoryoffads,fashion,custom,andculturalchangeasinformationalcascades'.JournalofpoliticalEconomypp.992Œ1026.Bikhchandani,S.andS.Sharma:2000,`Herdbehaviorinmarkets'.IMFEco-nomicReview47(3),279Œ310.Borio,C.,C.P.Lowe,etal.:2001,`Procyclicalityofthesystemandstability:issuesandpolicyoptions'.BISpapers1,1Œ57.Choe,H.,B.-C.Kho,andR.M.Stulz:1999,`Doforeigninvestorsdestabilizestockmar-kets?TheKoreanexperiencein1997'.JournalofFinancialEconomics54(2),227Œ264.Devenow,A.andI.Welch:1996,`Rationalherdingineconomics'.EuropeanEconomicReview40(3),603Œ615.Freixas,X.:2010,`Post-crisischallengestobankregulation'.EconomicPolicy25(62),375Œ399.Graham,J.R.:1999,`Herdingamonginvestmentnewsletters:Theoryandevidence'.TheJournalofFinance54(1),237Œ268.Grinblatt,M.,S.Titman,andR.Wermers:1995,`Momentuminvestmentstrategies,port-folioperformance,andherding:Astudyofmutualfundbehavior'.TheAmericaneco-nomicreviewpp.1088Œ1105.Hirshleifer,D.andS.HongTeoh:2003,`Herdbehaviourandcascadingincapitalmar-kets:Areviewandsynthesis'.EuropeanFinancialManagement9(1),25Œ66.Jain,A.K.andS.Gupta:1987,`SomeevidenceonherdingbehaviorofUSbanks'.JournalofMoney,CreditandBanking19(1),78Œ89.Kang,J.-K.:2010,`Eunghaengdaechuleuissolimhyeonsangegwanhansiljeungbunseok'.EconomicAnalysis16(1),1Œ35.Lakonishok,J.,A.Shleifer,andR.W.Vishny:1992,`Theimpactofinstitutionaltradingonstockprices'.Journalofeconomics32(1),23Œ43.45Liu,C.:2014,`Herdingbehaviorinbanklending:Evidencefromuscommercialbanks'.AvailableatSSRN1917552.Nakagawa,R.:2008,`HerdbehaviorbyJapanesebanksinlocalmarkets'.Nakagawa,R.,H.Oiwa,andF.Takeda:2012,`TheeconomicimpactofherdbehaviorintheJapaneseloanmarket'.FinanceJournal20(4),600Œ613.Nakagawa,R.andH.Uchida:2011,`HerdbehaviourbyJapanesebanksafterderegulation'.Economica78(312),618Œ636.Rajan,R.G.:1994,`WhybankcreditpoliciesAtheoryandsomeevidence'.TheQuarterlyJournalofEconomicspp.399Œ441.Scharfstein,D.S.andJ.C.Stein:1990,`Herdbehaviorandinvestment'.TheAmericanEconomicReviewpp.465Œ479.Sias,R.W.:2004,`Institutionalherding'.Reviewofstudies17(1),165Œ206.Uchida,H.andR.Nakagawa:2007,`HerdbehaviorintheJapaneseloanmarket:Evi-dencefrombankpaneldata'.JournalofFinancialIntermediation16(4),555Œ583.Wermers,R.:1994,`Herding,tradereversals,andcascadingbyinstitutionalinvestors'.AvailableatSSRN5644.Wermers,R.:1999,`Mutualfundherdingandtheimpactonstockprices'.TheJournalofFinance54(2),581Œ622.Wylie,S.:2005,`Fundmanagerherding:AtestoftheaccuracyofempiricalresultsusingUKdata'.TheJournalofBusiness78(1),381Œ403.46CHAPTER2TheExtensiveMarginEffectsofBankHealth,CreditShockontheFirm'sExportandtheRoleofCorporateGovernance2.1IntroductionItisknownthatthehealthofbanksthatprovidesancetotheisanim-portantdeterminantofexports(AmitiandWeinstein(2011)).Inthischapter,Iinvestigatewhetherthebankhealthandthecreditshocktothebankaffectsitsclientexportandwhethermoreconcentratedownershipstructureforthemakesitsexportactivitiesresilientornot.2.2DataandEstimationInthissection,Iprovidedetailsonthedataandmeasurementofthevariables.Fortheanalysis,Iusethreemaindatasets:dataforKoreanbank-leveldataonKoreanbanks,andmatchedloansdata.2.2.1DataSourceThesourceofdataistheKISVALUEandKISLINEdatabase,thebusinessin-formationsourceofKoreancreditratingcompany,KoreanInvestorsService(KIS).Thisdatasetprovidesinformationonwiderangeofcharacteristicsandstate-mentsnotonlyonthepubliclytradedbutalsoontheprivatelyheldthatare47subjecttoexternalauditinKorea1.Theyareallrequiredtoreportannuallytheirauditstatementincludingstatementstotheauthorities.Thesamplecoversallindustriesexcepttheandinsuranceindustriestoavoidendogeneityproblem.Ingeneral,theKoreanyearrunsfromJanuarytoDecemberandmorethan90percentofthereportstheirstatementincludingloansdatafortheyearendinginDecember.Forthisanalysis,thewhoseyearendsinotherthanDecemberareexcludedfromthesampleforconsistency.Thebank-leveldatacomesfromtheFinancialStatisticsInformationSystem(FISIS)databasemaintainedbytheFinancialSupervisoryService,Korea'sintegratedsupervi-soryauthority.TheFISIScontainsextensiverangeofinformationoninstitu-tionsincludingstatementsandsoundnessindicators.Mytionofabank,inthispaper,coversKoreannationalwide,regionalandspecialbank2.Theforeignbanksbranchesareexcludedbecausetheshareoftheloanstotheisverylowandrelativelysmallnumberofborrowfromtheforeignbanks.IusethematchedloansdatafromDataAnalysis,RetrievalandTransfer(DART)System,theKoreanelectronicdisclosuresystem.DARTprovidesloansdataforeachthatissubjecttoexternalauditintheannotationpartoftheannualstatementofaudit.Thematcheddatabasicallycontainsthetotalamountoftheshort-termandlong-termloansrespectivelyattheendofyearandthenameofeverybankfromwhichthemborrows.Icollectedloansinformationbyhandbysearchingthestatementofauditforeachfor2012and2013.Someofborrowingfromthemultiplebanks1AccordingtotheArticle2oftheActonExternalAuditofStockCompaniesandtheArticle2ofEnforcementofDecree,astockcompanyfallingunderanyofthefollowingshallbesubjecttoaccountingauditconductedbyanexternalauditor;1.Astockcompanywhosetotalassetsasoftheendoftheimmediatelyprecedingbusinessyearamounttonolessthantenbillionwon,2.Astock-listedcorporationorastockcompanythatintendstobecomealistedcorporationduringthepertinentbusinessyearorthefollowingyear,3.Asockcompanywhosetotalliabilitiesandtotalassetsamountstonolessthansevenbillionwonrespectivelyasoftheendoftheimmediatelyprecedingbusinessyear,4.Astockcompanywhosenumberofemployeesexceeds300asoftheendoftheimmediatelyprecedingbusinessyearandwhosetotalassetsamounttonolessthansevenbillionwon.2Duringthesampleperiods,2010to2013,thereweresevennationalwidecommercialbanks,sixregionalbanks,andspecialbanksinKorea.48reportsthenameofonlyapartofthebank3.TheseareexcludedfromthesampleaseverybankthathastheloansoutstandingtothecannotbeAlso,Iincludeonlythatreporttheirloaninformationinbothyears.2.2.2MeasurementTocapturetheeffectofthebanks'healthandthecreditshocktothebankontheexportactivityforeachIneedtoidentifywhichbank(s)withwhichahasrelationshipcouldpotentiallyaffecttheactivity.First,Ifocusonthemainbankonly.Inabank-centeredsystemlikeKorea,obtainmostoftheirexternalfromtheirmainbanks,althoughtheymaintainbankingrelationshipswithseveralbanks(Baeetal.(2002)).Themainbankforeachforouranalysis,isasthebankthathasthemostloansoutstandingtotheattheendofyear2012and2013.ThematchedrmloansdataallowsmetoidentifythemainbankforeachConsideringthefactthatthemaintainscontinuingrelationshipwithitsmainbankfortheextendedtime,itisassumedherethatthehadthesamemainbankinthehalfofthesampleperiodyear,2010and2011,whenthereisnomatchedloansdataavailableyet.Second,asmoreandmoreborrowfromseveralbanksratherthanfromitsmainbankonly,thechancesarethataactivitymaybeaffectednotonlybythehealthofitsmainbank,butalsobythatofotherbanksfromwhichitborrows.Tocapturethis,Ialsousetheothermeasurethatisaweightedaverageofcredit,wheretheweightsaretheshareofbankloans.Icomputetheshareofbankloansforeverybankfromwhichaborrowsin2012and2013,andtaketheaverageofthemastheweights.Also,oneothercomplicationregardingthestudyofbank-relationshipismergerinbankingindustrybecauseitmakeshardertomatchbanksandFortunately,how-ever,therewasnomergebetweenKoreandomesticbanksbetween2010and2013.3Usually,alargewhichborrowsalargeamountfrommultiplebankstendstoreportonlyonerepresentativebankname.49Thebankhealthandthecreditshocktothebankismeasuredastheratioofthebankcapitaltotheassets,capitaladequacyratio,non-performingloansratio,returnonasset(ROA),andreturnonequity(ROE)foreachbank.FollowingJoh(2003)whoshowsthatthecontrollingfamily'sownershipconcentrationcanserveasabettercontrol-ownershipdisparitymeasurefortheKoreandata,Imeasuretheownershipconcentrationtoexaminethecorporategovernanceeffects.TheKISLINEprovidestheownershipdatasuchasthenamesandshareholdingsofthelargestshareholderanditsspeciallyrelatedpersonsincludingfamilymembers.Inthispaper,themeasureofownershipconcentrationiscomputedasthesumofshareholdingsofthelargestindividualshareholderandhis/herfamilymembersattheendofyear2010to2013.Thewhoselargestshareholderisacorporationoranorganization,notaprivateindividual,isexcludedasthefamilyrelationshipcannotbeTheownershipconcentrationforthesampleinthispaperis58.9%.ThiseishigherthanwhatisfoundbyJoh(2003),presumablyasmanyofthelarge-typicallypubliclytradedand/orafwiththebusinessgroup-areexcludedintheprocessofcollectingthematchedloansdata.Ownershipconcentrationinpubliclytradedislowerthaninprivatelyheldandthatofafwithbusinessgroupislowerthanthatinindependent(Joh(2003)).TheKISVALUEdatasetprovidesuswiththedomesticandforeignsalesaswellastotalsales.Isumupthedomesticandforeignsalesandcompareitwiththetotalsalesinordertocheckvalidityofreportedexportdataandtodistinguishthewithzeroforeignsalesfromtheonewhoseexportdataaremissing.Onlytheobservationswhosedomesticandforeignsalesdataadduptothetotalsalesareincludedinthesample.Itiswellknownthatexportersarebigger,morecapitalintensive,andmoreproductivethannon-exporters(BernardandJensen(2004)).Thus,Iaddthecharacteristicsascontrolvariables.Imeasuresizebytheassetsandthenumberofemployee,capitalintensitybycapital-laborratiowhichisastheassetoverdirectlabor,50laborproductivitybyvalueaddedperworkerornetsalesperworker.Ialsoaddage,whichistimevarying,usinginformationonthedateofestablishmentofthefromKISVALUEdataset.Furthermore,Iincludethefollowingdummyvariables:(i)industrydummiesforfour-digitindustryleveltocontrolforunobservedeffects,(ii)yeardummiestocontrolforunobservedmacroeconomiceffects.2.2.3ModelOurbasicestimatingequationis:expdumt=aOwnershipt1+bCreditt1+gOwnershipt1Creditt1+Zt1d+Di+Dt+ei,t,(2.1)whereexpdumtisanindicatorvariablethattakesthevalueof1iftheexportsisgreaterthanzeroattimet,Ownershipt1istheownershipconcentrationmeasurethatiscomputedasthesumofshareholdingsofthelargestindividualshareholderandhis/herfamilymembersattimet1,Creditt1isthebank'shealth4attimet1,Zt1isavectorofloggedvalueofcharacteristicsattimet1,andDiandDtisindustryandyeardummyrespectively.ForCreditt1,asitmaytakemorethanoneyearforthebankhealthandthecreditshocktothebanktoaffecttheexportsactivity,Iaddtwoyearlaggedvariableandtheaverageofmeasureofoneyearandtwoyearslaggedmeasureasarobusttest.Also,Ownershipt1andCreditt1weredemeanedbeforetheinteraction.DemeaningthosevariablesimpliesthatthecoefonOwnershipt1orCreditt1isequaltotheaveragepartialeffectforOwnershipt1orCreditt1.Ultimately,Iinvestigatewhetherafutureexportsperformanceisaffectedbythebank'scialhealthand/orthecreditshocktothebank,andwhetherthemore4IfItakethechangeinCreditb,t1,thatis(4Creditb,t1),itcanbeinterpretedasacreditshocktothebankfromwhichaborrows.51family-concentratedownershipstructuremakesitsexportsmoreresilientornot.Forthatpurpose,Iusetheownershipconcentration,thecharacteristics,andthebankhealthasthelaggedvariablesasabaseline.So,thecoefbandgareofmymaininterest.Iruntheregressionbasedonthefollowingthreeassumptionsastowhichbank'shealthorcreditshockaffectstheexportsactivity;(i)onlythemainbankforeachaffectsthelikelihoodthataexports(column(1)),(ii)thehealthorthecreditshockforallthebanksfromwhichaborrowsmayhavetheextensivemarginef-fectsonthefrimexports(column(2)),and(iii)themainbank'shealthorcreditshockhasitsownimpact,buttheotherbanksinrelationwithamayhavetheeffectontheexportaswell(column(3)).So,Iseparatethetotalcreditintocreditfromthemainbank(Credit_maint1)andweightedaverageofthecreditfromallotherbanks(Credit_otherst1).AndIalsoaddinteractiontermsbetweenOwnershipt1andCredit_maint1andbetweenOwnershipt1andCredit_otherst1.Theassumption(iii)standssomewherebetweentwoextremeones;Ifocusonthemainbankonlyin(i),whilethehealthorcreditshocksofallthebanksfromwhichaborrowspotentiallyaffectstheexportsinproportiontotheshareofloansin(ii).2.3RegressionResults2.3.1TheLaggedLeveloftheCreditIexaminetheextensivemargineffectsofthelaggedlevelofbank'sCreditasaproxyofbank'shealthinTable2.1through2.4.Asabaselinemodel,table2.1showstheresultsfortheregressionfortheextensivemargineffectswithoneyearlaggedlevelofcreditvari-ables(Creditt1).Amongthecharacteristicvariables,thecoefonthenumberofemployee,oneoftheproxyforthesize,isstatisticallyoveralltion.Itimpliesthatthebiggersizeofthethemorelikelytheistoexport.Also,52thecoefontheageshowsapositiverelationshipforsonemeasuresofcredit.FromtheCreditt1variable,wecanseethatthehealthierthebanksfromwhichaborrows,theismorelikelytoexport.Itistruewhetherwelookatthemainbankorweassumethatallthecreditorbanksmatter.TheinteractionbetweenOwnershipt1andCreditt1ispositiveandformostofcreditmeasures:thelikelihoodthattheentertheexportactivityincreasesathigherrateforthewithhigherfamilyownershipconcentration.ItindicatesthatthecorporategovernancewherethefamilymembersholdmoreshareshelpstheexportactivitymoreresilientwhenthenegativecreditshockoccurstotheWhenlookingattheeffectsbyseparatingthecreditsbymainbankfromotherbanks,thecoefonbothCreditt1anditsinteractionwithOwnershipt1ismoreforthemainbankthanforotherbanks:theexportisaffectedmorebythehealthofitsmainbankratherthanbyotherbankswithwhichithasrelationship.TheonlydifferenceofTable2.2isthatthelagofCreditvariableisnowtwoyears.Itassumesthatthehealthofbanksaffectstheexportsactivityfortheintwoyears.TheoverallresultsforTable2.2isthesame,butthemagnitudeofthecoefonCreditt2andinteractiontermtendstobelargerthaninTable2.1.InTable2.3,IincludedbothoneyearandtwoyearlaggedCreditvariablesatthesametime.WecanseethatthesignofsomeoftheCreditt2andinteractiontermchangesintothenegativeandtherelationshipdisappearsinmostcases,whilethenumberofemployeekeepspositiverelation.Finally,Table2.4showstheresultsfortheregressionwhereIusetheaverageofCreditvariablethatlagsoneyearandtwoyears.Themainresultsstillholdswhenitisassumedthattheaveragehealthforthebanksaffectstheexport.53Table2.1Exportdummyandlaggedlevelofthecredit(lagby1year)(1)CreditbymainbanksBankCapital/Tier1Non-PerformingROEROAAssetsCapitalLoansOwnershipt1-0.0535-0.0485-0.0334-0.0376-0.0541*(0.0327)(0.0326)(0.0326)(0.0327)(0.0327)Creditt10.0076***0.0056***0.03620.00240.0758**(0.0026)(0.0019)(0.0232)(0.0022)(0.0296)Ownershipt1Creditt10.0222***0.0121***0.00540.00750.2590***(0.0063)(0.0043)(0.0562)(0.0054)(0.0723)ln(Fixedassetst1)0.02850.02930.03290.03380.0316(0.0211)(0.0211)(0.0211)(0.0211)(0.0211)ln(Employeet1)0.0773***0.0804***0.0770***0.0771***0.0782***(0.0204)(0.0204)(0.0205)(0.0204)(0.0203)ln(Firmaget1)0.0366*0.0360*0.03060.03020.0335*(0.0204)(0.0204)(0.0205)(0.0204)(0.0203)ln(CapitalLaborratiot1)0.0013-0.0004-0.0028-0.00060.0020(0.0187)(0.0187)(0.0187)(0.0187)(0.0187)ln(Valueaddedperworkert1)0.00070.00080.00810.00710.0078(0.0181)(0.0181)(0.0181)(0.0182)(0.0181)Constant-0.596***-0.665***-0.618***-0.593***-0.608***(0.1950)(0.1960)(0.1970)(0.1960)(0.1950)Observation2,2822,2822,2822,2822,282R20.3600.3590.3560.3560.360Note:Coefontheindustryandyeardummiesarenotreported.Standarderrorsareinparentheses***,**,*indicatestatisticallyatthe1%,5%,and10%levelrespectively.54Table2.1(cont'd)(2)WeightedaverageofcreditBankCapital/Tier1Non-PerformingROEROAAssetsCapitalLoansOwnershipt1-0.0663**-0.0530-0.0349-0.0427-0.0574*(0.0332)(0.0330)(0.0328)(0.0330)(0.0330)Creditt10.0113***0.0063***0.0471*0.0068**0.1190***(0.0034)(0.0022)(0.0285)(0.0028)(0.0388)Ownershipt1Creditt10.0319***0.0123**-0.01110.00340.2390**(0.0087)(0.0056)(0.0728)(0.0072)(0.0975)ln(Fixedassetst1)0.02650.02970.03300.03440.0315(0.0211)(0.0211)(0.0211)(0.0211)(0.0211)ln(Employeet1)0.0782***0.0804***0.0782***0.0794***0.0790***(0.0204)(0.0205)(0.0205)(0.0206)(0.0205)ln(Firmaget1)0.0397*0.0379*0.03120.03140.0351*(0.0204)(0.0205)(0.0205)(0.0204)(0.0203)ln(CapitalLaborratiot1)0.0008-0.0030-0.0043-0.0047-0.0021(0.0187)(0.0188)(0.0188)(0.0188)(0.0187)ln(Valueaddedperworkert1)0.00930.00980.00990.01170.0104(0.0181)(0.0182)(0.0182)(0.0183)(0.0182)Constant-0.608***-0.664***-0.630***-0.612***-0.619***(0.1940)(0.1970)(0.1980)(0.1960)(0.1950)Observation2,2822,2822,2822,2822,282R20.3620.3590.3560.3570.360Note:Coefontheindustryandyeardummiesarenotreported.Standarderrorsareinparentheses.***,**,*indicatestatisticallyatthe1%,5%,and10%levelrespectively.55Table2.1(cont'd)(3)SeparatecreditsbymainbankfromotherbanksBankCapital/Tier1Non-PerformingROEROAAssetsCapitalLoansOwnershipt1-0.0684**-0.0555*-0.0419-0.0519-0.0624*(0.0332)(0.0331)(0.0330)(0.0332)(0.0332)Credit_maint10.0062**0.0053***0.03680.00110.0597*(0.0028)(0.0020)(0.0249)(0.0022)(0.0305)Ownershipt1Credit_maint10.0185***0.0097**-0.03740.00580.2470***(0.0068)(0.0048)(0.0619)(0.0057)(0.0766)Credit_othert10.0052*0.00058-0.00060.0068***0.0686**(0.0028)(0.0018)(0.0239)(0.0025)(0.0331)Ownershipt1Credit_othert10.01240.00570.1130*0.00650.0245(0.0080)(0.0050)(0.0680)(0.0069)(0.0921)ln(Fixedassetst1)0.02490.02890.03290.02950.0281(0.0212)(0.0212)(0.0212)(0.0211)(0.0211)ln(Employeet1)0.0772***0.0796***0.0763***0.0785***0.0795***(0.0205)(0.0205)(0.0205)(0.0205)(0.0205)ln(Firmaget1)0.0395*0.0379*0.03300.0354*0.0361*(0.0204)(0.0205)(0.0205)(0.0204)(0.0204)ln(CapitalLaborratiot1)0.0021-0.0003-0.0026-0.00120.0020(0.0187)(0.0187)(0.0187)(0.0187)(0.0187)ln(Valueaddedperworkert1)0.00870.00830.00840.00890.0089(0.0181)(0.0181)(0.0182)(0.0182)(0.0181)Constant-0.601***-0.664***-0.621***-0.615***-0.622***(0.1940)(0.1960)(0.1970)(0.1950)(0.1950)Observation2,2822,2822,2822,2822,282R20.3630.3600.3570.3590.362Note:Coefontheindustryandyeardummiesarenotreported.Standarderrorsareinparentheses.***,**,*indicatestatisticallyatthe1%,5%,and10%levelrespectively.56Table2.2Exportdummyandlaggedlevelofthecredit(lagby2year)(1)CreditbymainbanksBankCapital/Tier1Non-PerformingROEROAAssetsCapitalLoansOwnershipt1-0.0569-0.0567-0.0311-0.0460-0.0639(0.0390)(0.0389)(0.0391)(0.0392)(0.0391)Creditt20.0092***0.0070***0.0220.00270.0833**(0.0031)(0.0022)(0.0282)(0.0024)(0.0326)Ownershipt1Creditt20.0273***0.0190***0.03150.0150**0.3270***(0.0077)(0.0052)(0.0677)(0.0061)(0.0811)ln(Fixedassetst1)0.02390.02440.02990.03000.0262(0.0259)(0.0259)(0.0261)(0.0260)(0.0259)ln(Employeet1)0.0750***0.0790***0.0737***0.0742***0.0763***(0.0251)(0.0252)(0.0253)(0.0252)(0.0251)ln(Firmaget1)0.02340.02490.01610.01590.0205(0.0254)(0.0255)(0.0256)(0.0254)(0.0254)ln(CapitalLaborratiot1)0.00550.00410.00170.00440.0074(0.0229)(0.0229)(0.0230)(0.0229)(0.0229)ln(Valueaddedperworkert1)0.00370.00370.00450.00480.0058(0.0216)(0.0216)(0.0218)(0.0217)(0.0216)Constant-0.564***-0.676***-0.581***-0.569***-0.581***(0.2180)(0.2210)(0.2230)(0.2200)(0.2190)Observation1,5941,5941,5941,5941,594R20.3830.3830.3750.3780.384Note:Coefontheindustryandyeardummiesarenotreported.Standarderrorsareinparentheses.***,**,*indicatestatisticallyatthe1%,5%,and10%levelrespectively.57Table2.2(cont'd)(2)WeightedaverageofcreditBankCapital/Tier1Non-PerformingROEROAAssetsCapitalLoansOwnershipt1-0.0709*-0.0646-0.0362-0.0537-0.0715*(0.0395)(0.0395)(0.0396)(0.0397)(0.0396)Creditt20.0139***0.00791***0.04840.0053*0.1070**(0.0042)(0.0027)(0.0337)(0.0031)(0.0433)Ownershipt1Creditt20.0380***0.0200***0.02610.0170**0.4010***(0.0170)(0.0068)(0.0849)(0.0083)(0.1330)ln(Fixedassetst1)0.02230.02540.03030.03070.0269(0.0259)(0.0259)(0.0260)(0.0260)(0.0259)ln(Employeet1)0.0751***0.0777***0.0752***0.0737***0.0748***(0.0251)(0.0252)(0.0253)(0.0252)(0.0251)ln(Firmaget1)0.02590.02750.01770.01710.0218(0.0254)(0.0256)(0.0256)(0.0255)(0.0254)ln(CapitalLaborratiot1)0.00400.0005-0.00070.00190.0047(0.0229)(0.0230)(0.0231)(0.0231)(0.0230)ln(Valueaddedperworkert1)0.00510.00470.00680.00520.0044(0.0217)(0.0218)(0.0219)(0.0218)(0.0217)Constant-0.578***-0.676***-0.612***-0.578***-0.584***(0.2180)(0.2220)(0.2240)(0.2200)(0.2190)Observation1,5941,5941,5941,5941,594R20.3840.3820.3760.3780.383Note:Coefontheindustryandyeardummiesarenotreported.Standarderrorsareinparentheses.***,**,*indicatestatisticallyatthe1%,5%,and10%levelrespectively.58Table2.2(cont'd)(3)SeparatecreditsbymainbankfromotherbanksBankCapital/Tier1Non-PerformingROEROAAssetsCapitalLoansOwnershipt1-0.0744*-0.0668*-0.0499-0.0676*-0.0771*(0.0396)(0.0396)(0.0397)(0.0399)(0.0398)Credit_maint20.0084**0.0067***0.01340.00200.0766**(0.0033)(0.0024)(0.0301)(0.0024)(0.0332)Ownershipt1Credit_maint20.0218***0.0153***-0.03840.0119*0.2900***(0.0082)(0.0058)(0.0745)(0.0063)(0.0845)Credit_othert20.00410.00010.02310.0046*0.0294(0.0035)(0.0021)(0.0279)(0.0027)(0.0361)Ownershipt1Credit_othert20.0196**0.00900.1840**0.0155**0.1480(0.0100)(0.0059)(0.0794)(0.0076)(0.1010)ln(Fixedassetst1)0.02200.02510.02930.02790.0255(0.0261)(0.0260)(0.0261)(0.0260)(0.0260)ln(Employeet1)0.0730***0.0772***0.0728***0.0728***0.0747***(0.0252)(0.0252)(0.0253)(0.0252)(0.0252)ln(Firmaget1)0.02650.027270.02170.02230.0243(0.0254)(0.0255)(0.0256)(0.0255)(0.0254)ln(CapitalLaborratiot1)0.00470.00330.00020.00260.0059(0.0229)(0.0229)(0.0230)(0.0229)(0.0229)ln(Valueaddedperworkert1)0.00410.00370.00450.00410.0054(0.0216)(0.0217)(0.0217)(0.0217)(0.0216)Constant-0.567***-0.672***-0.591***-0.581***-0.577***(0.2180)(0.2210)(0.2220)(0.2200)(0.2190)Observation1,5941,5941,5941,5941,594R20.3860.3840.3790.3820.385Note:Coefontheindustryandyeardummiesarenotreported.Standarderrorsareinparentheses.***,**,*indicatestatisticallyatthe1%,5%,and10%levelrespectively.59Table2.3Exportdummyandlaggedlevelofthecredit(lagby1&2year)(1)CreditbymainbanksBankCapital/Tier1Non-PerformingROEROAAssetsCapitalLoansOwnershipt1-0.0563-0.0708*-0.0195-0.0473-0.0688*(0.0395)(0.0395)(0.0394)(0.0393)(0.0392)Creditt1-0.00610.00010.10400.00290.0535(0.0153)(0.0198)(0.0700)(0.0044)(0.0562)Creditt20.01500.0066-0.05530.00110.0538(0.0149)(0.0193)(0.0650)(0.0037)(0.0486)Ownershipt1Creditt10.0055-0.0824**0.3730**0.00600.1990(0.0449)(0.0413)(0.1740)(0.0112)(0.1330)Ownershipt1Creditt20.02150.0984**-0.26200.01140.2030*(0.0445)(0.0402)(0.1610)(0.0099)(0.1220)ln(Fixedassetst1)0.02400.02270.02630.03100.0279(0.0260)(0.0260)(0.0260)(0.0260)(0.0259)ln(Employeet1)0.0743***0.0805***0.0750***0.0742***0.0754***(0.0252)(0.0252)(0.0253)(0.0253)(0.0251)ln(Firmaget1)0.02330.02410.02210.01630.0219(0.0254)(0.0255)(0.0256)(0.0255)(0.0254)ln(CapitalLaborratiot1)0.00530.00600.00420.00340.0061(0.0229)(0.0229)(0.0230)(0.0230)(0.0229)ln(Valueaddedperworkert1)0.00320.00560.00270.00510.0051(0.0216)(0.0217)(0.0217)(0.0218)(0.0217)Constant-0.556**-0.655***-0.623***-0.578***-0.597***(0.2200)(0.2200)(0.2300)(0.2200)(0.2190)Observation1,5941,5941,5941,5941,594R20.3830.3840.3790.3780.385Note:Coefontheindustryandyeardummiesarenotreported.Standarderrorsareinparentheses.***,**,*indicatestatisticallyatthe1%,5%,and10%levelrespectively.60Table2.3(cont'd)(2)WeightedaverageofcreditBankCapital/Tier1Non-PerformingROEROAAssetsCapitalLoansOwnershipt1-0.0676*-0.0794**-0.0212-0.0557-0.0750*(0.0408)(0.0402)(0.0403)(0.0397)(0.0397)Creditt1-0.00260.00390.08220.0111*0.160**(0.0227)(0.0285)(0.1020)(0.0063)(0.0799)Creditt20.01640.0040-0.0152-0.0025-0.0053(0.0227)(0.0280)(0.0925)(0.0054)(0.0710)Ownershipt1Creditt1-0.0172-0.102**0.4590**-0.00720.0692(0.0641)(0.0521)(0.2310)(0.0147)(0.1730)Ownershipt1Creditt20.05480.1190**-0.33400.0229*0.3660**(0.0643)(0.0511)(0.2050)(0.0133)(0.1630)ln(Fixedassetst1)0.02230.02460.02770.03090.0277(0.0259)(0.0260)(0.0261)(0.0260)(0.0259)ln(Employeet1)0.0751***0.0794***0.0763***0.0765***0.0762***(0.0252)(0.0252)(0.0253)(0.0253)(0.0252)ln(Firmaget1)0.02560.02700.02430.01720.0225(0.0255)(0.0256)(0.0257)(0.0254)(0.0254)ln(CapitalLaborratiot1)0.00400.00130.00140.00010.0024(0.0230)(0.0230)(0.0231)(0.0231)(0.0230)ln(Valueaddedperworkert1)0.00520.00670.00480.00880.0065(0.0217)(0.0219)(0.0219)(0.0219)(0.0218)Constant-0.572***-0.660***-0.647***-0.600***-0.613***(0.2200)(0.2230)(0.2240)(0.2210)(0.2200)Observation1,5941,5941,5941,5941,594R20.3840.3840.3780.3800.385Note:Coefontheindustryandyeardummiesarenotreported.Standarderrorsareinparentheses.***,**,*indicatestatisticallyatthe1%,5%,and10%levelrespectively.61Table2.3(cont'd)(3)SeparatecreditsbymainbankfromotherbanksBankCapital/Tier1Non-PerformingROEROAAssetsCapitalLoansOwnershipt1-0.0675-0.0758*-0.0371-0.0697*-0.0815**(0.0410)(0.0403)(0.0403)(0.0399)(0.0399)Credit_maint1-0.00920.00060.08710.00130.0354(0.0154)(0.0198)(0.0709)(0.0045)(0.0575)Credit_maint20.01720.0058-0.04110.00070.0523(0.0151)(0.0193)(0.0655)(0.0037)(0.0488)Ownershipt1Credit_maint1-0.0033-0.1000**0.3470*0.00700.2440*(0.0452)(0.0459)(0.1830)(0.0117)(0.1430)Ownershipt1Credit_maint20.02470.1120**-0.3070*0.00800.1400(0.0447)(0.0447)(0.1700)(0.0099)(0.1260)Credit_othert1-0.00710.0237-0.2420**0.0117**0.1490**(0.0209)(0.0280)(0.1040)(0.0057)(0.0710)Credit_othert20.0114-0.02330.2300**-0.0042-0.0865(0.0216)(0.0276)(0.0968)(0.0050)(0.0644)Ownershipt1Credit_othert1-0.02760.0784-0.0974-0.0094-0.1690(0.0625)(0.0751)(0.2750)(0.0152)(0.1860)Ownershipt1Credit_othert20.0479-0.06870.26400.0227*0.2740(0.0648)(0.0742)(0.2560)(0.0133)(0.1690)ln(Fixedassetst1)0.02110.02200.02820.02910.0277(0.0261)(0.0261)(0.0260)(0.0260)(0.0260)ln(Employeet1)0.0072***0.0806***0.0727***0.0723***0.0734***(0.0253)(0.0253)(0.0252)(0.0253)(0.0252)ln(Firmaget1)0.02540.02530.02550.02390.0265(0.0255)(0.0256)(0.0256)(0.0255)(0.0254)ln(CapitalLaborratiot1)0.00450.00670.00170.00100.0040(0.0229)(0.0229)(0.0229)(0.0229)(0.0229)ln(Valueaddedperworkert1)0.00360.00650.00050.00340.0035(0.0216)(0.0217)(0.0217)(0.0218)(0.0216)Constant-0.546**-0.643***-0.638***-0.609***-0.613***(0.2200)(0.2200)(0.2200)(0.2200)(0.2190)Observation1,5941,5941,5941,5941,594R20.3860.3860.3840.3840.388Note:Coefontheindustryandyeardummiesarenotreported.Standarderrorsareinparentheses.***,**,*indicatestatisticallyatthe1%,5%,and10%levelrespectively.62Table2.4Exportdummyandlaggedlevelofthecredit(averageofcredit)(1)CreditsbymainbankBankCapital/Tier1Non-PerformingROEROAAssetsCapitalLoansOwnershipt1-0.0565-0.0554-0.0367-0.0491-0.0700*(0.0390)(0.0389)(0.0391)(0.0393)(0.0392)AverageLaggedCredit0.0093***0.0069***0.03920.00370.1070***(0.0032)(0.0023)(0.0303)(0.0028)(0.0376)Ownershipt1AverageLaggedCredit0.0278***0.0186***0.07810.0177**0.4010***(0.0078)(0.0053)(0.0728)(0.0069)(0.0915)ln(Fixedassetst1)0.02400.02490.02900.03130.0279(0.0260)(0.0260)(0.0260)(0.0260)(0.0259)ln(Employeet1)0.0753***0.0789***0.0747***0.0737***0.0754***(0.0251)(0.0252)(0.0253)(0.0252)(0.0251)ln(Firmaget1)0.02340.02420.01830.01650.0219(0.0254)(0.0255)(0.0256)(0.0254)(0.0253)ln(CapitalLaborratiot1)0.00560.00360.00210.00330.0060(0.0229)(0.0229)(0.0230)(0.0229)(0.0228)ln(Valueaddedperworkert1)0.00380.00390.00480.00460.0050(0.0216)(0.0217)(0.0217)(0.0218)(0.0216)Constant-0.575***-0.672***-0.605***-0.577***-0.596***(0.2190)(0.2210)(0.2230)(0.2200)(0.2180)Observation1,5941,5941,5941,5941,594R20.3830.3820.3760.3780.385Note:Coefontheindustryandyeardummiesarenotreported.Standarderrorsareinparentheses.***,**,*indicatestatisticallyatthe1%,5%,and10%levelrespectively.63Table2.4(cont'd)(2)WeightedaverageofcreditBankCapital/Tier1Non-PerformingROEROAAssetsCapitalLoansOwnershipt1-0.0709*-0.0632-0.0408-0.0561-0.0768*(0.0396)(0.0395)(0.0396)(0.0398)(0.0398)AverageLaggedCredit0.0139***0.0079***0.0594*0.0075**0.1450***(0.0042)(0.0027)(0.0361)(0.0036)(0.0484)Ownershipt1AverageLaggedCredit0.0377***0.0193***0.07020.0169*0.4420***(0.0108)(0.0069)(0.0924)(0.0092)(0.1250)ln(Fixedassetst1)0.02250.02580.02950.03210.0285(0.0259)(0.0260)(0.0260)(0.0260)(0.0259)ln(Employeet1)0.0753***0.0777***0.0755***0.0742***0.0745***(0.0251)(0.0252)(0.0253)(0.0253)(0.0251)ln(Firmaget1)0.02620.02680.01970.01810.0235(0.0254)(0.0256)(0.0257)(0.0254)(0.0254)ln(CapitalLaborratiot1)0.00380.00015-0.0004-0.00070.0017(0.0229)(0.0230)(0.0231)(0.02230)(0.0229)ln(Valueaddedperworkert1)0.00510.00500.00670.00670.0049(0.0217)(0.0218)(0.0219)(0.0219)(0.0218)Constant-0.591***-0.672***-0.624***-0.597***-0.610***(0.2180)(0.2220)(0.2240)(0.2210)(0.2200)Observation1,5941,5941,5941,5941,594R20.3840.3810.3760.3790.384Note:Coefontheindustryandyeardummiesarenotreported.Standarderrorsareinparentheses.***,**,*indicatestatisticallyatthe1%,5%,and10%levelrespectively.64Table2.4(cont'd)(3)SeparatecreditsbymainbankfromotherbanksBankCapital/Tier1Non-PerformingROEROAAssetsCapitalLoansOwnershipt1-0.0737*-0.0661*-0.0521-0.0705*-0.0827**(0.0396)(0.0396)(0.0397)(0.0400)(0.0399)AverageLaggedCredit_main0.0083**0.0067***0.03410.00230.0933**(0.0034)(0.0025)(0.0327)(0.0029)(0.0390)Ownershipt1AverageLaggedCredit_main0.0219***0.0146**0.00680.0137*0.362***(0.0083)(0.0059)(0.0801)(0.0072)(0.0967)AverageLaggedCredit_other0.003930.00020.01040.0064**0.0493(0.0035)(0.0022)(0.0295)(0.0030)(0.0401)Ownershipt1AverageLaggedCredit_other0.0186*0.00950.1770**0.0153*0.1280(0.0098)(0.0060)(0.0835)(0.0085)(0.1120)ln(Fixedassetst1)0.02220.02540.02920.02810.0261(0.0261)(0.0260)(0.0261)(0.0260)(0.0259)ln(Employeet1)0.0734***0.0771***0.0735***0.0731***0.0748***(0.0252)(0.0252)(0.0253)(0.0252)(0.0251)ln(Firmaget1)0.02670.02680.02230.02340.0256(0.0254)(0.0255)(0.0257)(0.0255)(0.0254)ln(CapitalLaborratiot1)0.00480.00290.00070.00140.0048(0.0229)(0.0229)(0.0230)(0.0229)(0.0228)ln(Valueaddedperworkert1)0.00420.00390.00480.00420.0050(0.0216)(0.0217)(0.0217)(0.0217)(0.0216)Constant-0.580***-0.667***-0.604***-0.602***-0.605***(0.2180)(0.2210)(0.2230)(0.2200)(0.2190)Observation1,5941,5941,5941,5941,594R20.3850.3830.3780.3820.386Note:Coefontheindustryandyeardummiesarenotreported.Standarderrorsareinparentheses.***,**,*indicatestatisticallyatthe1%,5%,and10%levelrespectively.652.3.2TheLaggedChangeoftheCreditTable2.5through2.7givesustheresultsfortheregressionwhereIusethelaggedchangeoftheCreditvariable(4Credit),insteadofthelevel.Now,wecaninterpretthepositivevalueof4Creditasthepositivecreditshocktothebanks.AsisthecaseswhereIusedthelaggedleveloftheCreditvariable,IaddtwoyearslaggedCredit,andbothoneyearandtwoyearslaggedvariablesinTable2.6through2.7.Aswecansee,thesignofthecoefon4CreditanditsinteractionwithOwnershipt1isnotconsistent.Also,wecanthatthereisnostatisticallyrelationshipinmostcases,whileonlythecoefonthenumberofemployeevariableis.Table2.5Exportdummyandlaggedchangeofthecredit(lagby1year)(1)CreditsbymainbankBankCapital/Tier1Non-PerformingROEROAAssetsCapitalLoansOwnershipt1-0.0253-0.0334-0.0218-0.0256-0.0263(0.0387)(0.0383)(0.0383)(0.0384)(0.0384)4Creditt1-0.0165-0.01290.0639-0.0002-0.0228(0.0150)(0.0191)(0.0649)(0.0036)(0.0481)Ownershipt14Creditt1-0.0212-0.110***0.2810*-0.0072-0.0610(0.0446)(0.0399)(0.1610)(0.0097)(0.1180)ln(Fixedassetst1)0.03000.02650.02850.02940.0294(0.0260)(0.0260)(0.0260)(0.0261)(0.0261)ln(Employeet1)0.0710***0.0742***0.0719***0.0731***0.0727***(0.0253)(0.0252)(0.0252)(0.0253)(0.0253)ln(Firmaget1)0.01460.01660.01660.01400.0146(0.0255)(0.0255)(0.0255)(0.0255)(0.0255)ln(CapitalLaborratiot1)0.00190.00590.00380.00320.0030(0.0229)(0.0229)(0.0229)(0.0230)(0.0230)ln(Valueaddedperworkert1)0.00270.00450.00160.00410.0039(0.0217)(0.0216)(0.0217)(0.0217)(0.0217)Constant-0.520**-0.558**-0.547**-0.550**-0.549**(0.2210)(0.2200)(0.2190)(0.2200)(0.2200)Observation1,5941,5941,5941,5941,594R20.3760.3790.3770.3750.375Note:Coefontheindustryandyeardummiesarenotreported.Standarderrorsareinparentheses.***,**,*indicatestatisticallyatthe1%,5%,and10%levelrespectively.66Table2.5(cont'd)(2)WeightedaverageofcreditBankCapital/Tier1Non-PerformingROEROAAssetsCapitalLoansOwnershipt1-0.0246-0.0319-0.0217-0.0244-0.0245(0.0386)(0.0383)(0.0384)(0.0384)(0.0384)4Creditt1-0.0111-0.01160.00710.00330.0378(0.0227)(0.0228)(0.0924)(0.0054)(0.0707)Ownershipt14Creditt1-0.0360-0.1260**0.3310-0.0185-0.1930(0.0643)(0.0510)(0.2050)(0.0131)(0.1560)ln(Fixedassetst1)0.03010.02810.02970.02850.0290(0.0260)(0.0260)(0.0261)(0.0261)(0.0261)ln(Employeet1)0.0719***0.0738***0.0729***0.0740***0.0738***(0.0253)(0.0252)(0.0252)(0.0253)(0.0253)ln(Firmaget1)0.01380.01680.01660.01330.0136(0.0255)(0.0256)(0.0255)(0.0255)(0.0255)ln(CapitalLaborratiot1)0.00240.00420.00290.00440.0038(0.0229)(0.0229)(0.0229)(0.0230)(0.0230)ln(Valueaddedperworkert1)0.00340.00410.00240.00450.0044(0.0217)(0.0217)(0.0217)(0.0217)(0.0217)Constant-0.530**-0.561**-0.556**-0.547**-0.547**(0.2210)(0.2210)(0.2200)(0.2200)(0.2200)Observation1,5941,5941,5941,5941,594R20.3750.3780.3760.3760.376Note:Coefontheindustryandyeardummiesarenotreported.Standarderrorsareinparentheses.***,**,*indicatestatisticallyatthe1%,5%,and10%levelrespectively.67Table2.5(cont'd)(3)SeparatecreditsbymainbankfromotherbanksBankCapital/Tier1Non-PerformingROEROAAssetsCapitalLoansOwnershipt1-0.0248-0.0305-0.0266-0.0261-0.0257(0.0388)(0.0384)(0.0384)(0.0384)(0.0384)4Credit_maint1-0.0170-0.01220.0458-0.0001-0.0242(0.0151)(0.0191)(0.0652)(0.0037)(0.0482)Ownershipt14Credit_maint1-0.0203-0.1260***0.3240**-0.00524-0.0227(0.0449)(0.0442)(0.1700)(0.0099)(0.1240)4Credit_othert1-0.00570.0209-0.2290**0.00460.0892(0.0206)(0.0275)(0.0959)(0.0050)(0.0641)Ownershipt14Credit_othert10.00870.0648-0.3220-0.0161-0.1710(0.0611)(0.0736)(0.2550)(0.0132)(0.1660)ln(Fixedassetst1)0.03020.02560.02870.03070.0313(0.0261)(0.0261)(0.0260)(0.0261)(0.0261)ln(Employeet1)0.0707***0.0759***0.0719***0.0717***0.0712***(0.0253)(0.0252)(0.0252)(0.0253)(0.0253)ln(Firmaget1)0.01460.01520.01840.01460.0152(0.0255)(0.0255)(0.0254)(0.0255)(0.0255)ln(CapitalLaborratiot1)0.00200.00730.00210.00250.0019(0.0230)(0.0230)(0.0229)(0.0230)(0.0230)ln(Valueaddedperworkert1)0.00250.0051-0.00010.00310.0028(0.0217)(0.0217)(0.0216)(0.0217)(0.0217)Constant-0.520**-0.549**-0.583***-0.547**-0.549**(0.2210)(0.2210)(0.2190)(0.2200)(0.2200)Observation1,5941,5941,5941,5941,594R20.3760.3800.3810.3760.376Note:Coefontheindustryandyeardummiesarenotreported.Standarderrorsareinparentheses.***,**,*indicatestatisticallyatthe1%,5%,and10%levelrespectively.68Table2.6Exportdummyandlaggedchangeofthecredit(lagby2year)(1)CreditsbymainbankBankCapital/Tier1Non-PerformingROEROAAssetsCapitalLoansOwnershipt1-0.0125-0.0336-0.0153-0.0141-0.0141(0.0597)(0.0604)(0.0600)(0.0593)(0.0593)4Creditt2-0.0211-0.02850.02210.002310.0373(0.0298)(0.0273)(0.1110)(0.0055)(0.0794)Ownershipt14Creditt20.0554-0.08330.3790-0.0154-0.1160(0.0839)(0.0725)(0.3090)(0.0172)(0.2490)ln(Fixedassetst1)0.02380.01690.02190.02220.0221(0.0415)(0.0416)(0.0415)(0.0415)(0.0415)ln(Employeet1)0.0727*0.0774*0.0780*0.0766*0.0766*(0.0402)(0.0400)(0.0400)(0.0400)(0.0400)ln(Firmaget1)-0.0296-0.0225-0.0303-0.0317-0.0305(0.0413)(0.0415)(0.0413)(0.0414)(0.0415)ln(CapitalLaborratiot1)0.00980.01360.01050.01080.0105(0.0364)(0.0366)(0.0364)(0.0363)(0.0364)ln(Valueaddedperworkert1)0.01210.01780.01600.01600.0169(0.0357)(0.0353)(0.0354)(0.0353)(0.0353)Constant-0.432-0.422-0.410-0.415-0.417(0.3630)(0.3630)(0.3620)(0.3630)(0.3630)Observation756756756756756R20.4020.4030.4030.4020.401Note:Coefontheindustryandyeardummiesarenotreported.Standarderrorsareinparentheses.***,**,*indicatestatisticallyatthe1%,5%,and10%levelrespectively.69Table2.6(cont'd)(2)WeightedaverageofcreditBankCapital/Tier1Non-PerformingROEROAAssetsCapitalLoansOwnershipt1-0.0153-0.0378-0.0156-0.0110-0.0114(0.0601)(0.0606)(0.0603)(0.0593)(0.0594)4Creditt2-0.0102-0.0145-0.02040.00680.0973(0.0433)(0.0423)(0.1550)(0.0081)(0.1160)Ownershipt14Creditt20.0392-0.1910*0.4400-0.0323-0.2910(0.1180)(0.1060)(0.4130)(0.0244)(0.3510)ln(Fixedassetst1)0.02290.01630.02310.02350.0229(0.0415)(0.0416)(0.0416)(0.0414)(0.0415)ln(Employeet1)0.0748*0.0793**0.0786**0.0755*0.0765*(0.0402)(0.0399)(0.0400)(0.0399)(0.0399)ln(Firmaget1)-0.0300-0.0229-0.0292-0.0350-0.0330(0.0414)(0.0417)(0.0414)(0.0414)(0.0415)ln(CapitalLaborratiot1)0.01090.01710.008520.01160.0108(0.0364)(0.0365)(0.0365)(0.0363)(0.0363)ln(Valueaddedperworkert1)0.01440.01920.01780.01720.0190(0.0359)(0.0353)(0.0355)(0.0353)(0.0354)Constant-0.420-0.405-0.420-0.433-0.435(0.3630)(0.3630)(0.3630)(0.3620)(0.3630)Observation756756756756756R20.4010.4040.4020.4030.402Note:Coefontheindustryandyeardummiesarenotreported.Standarderrorsareinparentheses.***,**,*indicatestatisticallyatthe1%,5%,and10%levelrespectively.70Table2.6(cont'd)(3)SeparatecreditsbymainbankfromotherbanksBankCapital/Tier1Non-PerformingROEROAAssetsCapitalLoansOwnershipt1-0.0228-0.0430-0.0265-0.0113-0.0105(0.0600)(0.0611)(0.0603)(0.0594)(0.0596)4Credit_maint2-0.0175-0.02900.00770.00230.0343(0.0301)(0.0273)(0.1120)(0.0055)(0.0797)Ownershipt14Credit_maint20.0667-0.06470.4060-0.0181-0.1310(0.0842)(0.0730)(0.3120)(0.0175)(0.2500)4Credit_othert20.03710.0550-0.01420.0004-0.0148(0.0449)(0.0453)(0.1430)(0.0076)(0.1070)Ownershipt14Credit_othert20.1530-0.2400*-0.7720*-0.0237-0.2280(0.1360)(0.1320)(0.4370)(0.0235)(0.3230)ln(Fixedassetst1)0.02310.02590.02390.02140.0215(0.0416)(0.0420)(0.0415)(0.0415)(0.0416)ln(Employeet1)0.0719*0.0683*0.0751*0.0755*0.0760*(0.0403)(0.0402)(0.0400)(0.0400)(0.0400)ln(Firmaget1)-0.0266-0.0227-0.0264-0.0334-0.0316(0.0414)(0.0416)(0.0414)(0.0415)(0.0416)ln(CapitalLaborratiot1)0.01120.01040.01020.01200.0115(0.0364)(0.0366)(0.0363)(0.0364)(0.0364)ln(Valueaddedperworkert1)0.01040.01500.01310.01540.0162(0.0360)(0.0353)(0.0355)(0.0355)(0.0355)Constant-0.473-0.370-0.430-0.426-0.422(0.3640)(0.3640)(0.3640)(0.3630)(0.3640)Observation756756756756756R20.4040.4070.4060.4030.402Note:Coefontheindustryandyeardummiesarenotreported.Standarderrorsareinparentheses.***,**,*indicatestatisticallyatthe1%,5%,and10%levelrespectively.71Table2.7Exportdummyandlaggedchangeofthecredit(lagby1&2year)(1)CreditsbymainbankBankCapital/Tier1Non-PerformingROEROAAssetsCapitalLoansOwnershipt1-0.04380.0603-0.0375-0.0485-0.0666(0.0631)(0.0795)(0.0692)(0.0669)(0.0702)4Creditt10.0079-0.00370.1160-0.0093-0.1390(0.0226)(0.0367)(0.1070)(0.0074)(0.0908)4Creditt2-0.0245-0.02910.0147-0.0013-0.0343(0.0299)(0.0276)(0.1110)(0.0061)(0.0917)Ownershipt14Creditt1-0.1080-0.1830*0.2180-0.0171-0.2920(0.0693)(0.1020)(0.3210)(0.0197)(0.2420)Ownershipt14Creditt20.0721-0.06310.3420-0.0243-0.3090(0.0850)(0.0734)(0.3120)(0.0193)(0.2900)ln(Fixedassetst1)0.02380.01330.01860.01950.0163(0.0415)(0.0416)(0.0416)(0.0415)(0.0416)ln(Employeet1)0.0739*0.0824**0.0765*0.0759*0.0764*(0.0402)(0.0401)(0.0400)(0.0400)(0.0401)ln(Firmaget1)-0.0297-0.0277-0.0258-0.0295-0.0267(0.0413)(0.0418)(0.0416)(0.0416)(0.0416)ln(CapitalLaborratiot1)0.01230.01560.01360.01360.0152(0.0364)(0.0366)(0.0364)(0.0365)(0.0366)ln(Valueaddedperworkert1)0.01400.02050.01160.02200.0224(0.0357)(0.0354)(0.0355)(0.0355)(0.0354)Constant-0.391-0.469-0.422-0.404-0.399(0.3680)(0.3640)(0.3630)(0.3630)(0.3640)Observation756756756756756R20.4040.4070.4050.4040.405Note:Coefontheindustryandyeardummiesarenotreported.Standarderrorsareinparentheses.***,**,*indicatestatisticallyatthe1%,5%,and10%levelrespectively.72Table2.7(cont'd)(2)WeightedaverageofcreditBankCapital/Tier1Non-PerformingROEROAAssetsCapitalLoansOwnershipt1-0.0448-0.0957-0.0562-0.0785-0.1180(0.0640)(0.0886)(0.0748)(0.0719)(0.0777)4Creditt10.01670.02890.0584-0.0078-0.0951(0.0357)(0.0535)(0.1580)(0.0108)(0.1360)4Creditt2-0.0164-0.0190-0.01680.00390.0479(0.0435)(0.0423)(0.1550)(0.0091)(0.1360)Ownershipt14Creditt1-0.1520-0.2950**0.4300-0.0427-0.7340**(0.1070)(0.1430)(0.4620)(0.0285)(0.3610)Ownershipt14Creditt20.0575-0.1780*0.3820-0.0521*-0.7380*(0.1190)(0.1070)(0.4170)(0.0278)(0.4150)ln(Fixedassetst1)0.02160.01430.02190.01980.0162(0.0416)(0.0416)(0.0417)(0.0415)(0.0415)ln(Employeet1)0.0755*0.0878**0.0780*0.0761*0.0793**(0.0403)(0.0400)(0.0400)(0.0399)(0.0400)ln(Firmaget1)-0.0354-0.0271-0.0229-0.0342-0.0316(0.0415)(0.0418)(0.0418)(0.0416)(0.0417)ln(CapitalLaborratiot1)0.01540.01430.01050.01610.0179(0.0366)(0.0366)(0.0366)(0.0365)(0.0365)ln(Valueaddedperworkert1)0.01670.02620.01410.02130.0237(0.0360)(0.0355)(0.0357)(0.0355)(0.0354)Constant-0.382-0.432-0.428-0.392-0.358(0.3690)(0.3640)(0.3640)(0.3660)(0.3670)Observation756756756756756R20.4030.4090.4030.4060.407Note:Coefontheindustryandyeardummiesarenotreported.Standarderrorsareinparentheses.***,**,*indicatestatisticallyatthe1%,5%,and10%levelrespectively.73Table2.7(cont'd)(3)SeparatecreditsbymainbankfromotherbanksBankCapital/Tier1Non-PerformingROEROAAssetsCapitalLoansOwnershipt1-0.04850.0272-0.0546-0.0981-0.129(0.0638)(0.0891)(0.0843)(0.0762)(0.0829)4Credit_maint10.00390.01630.0946-0.0072-0.1090(0.0229)(0.0372)(0.1110)(0.0075)(0.0923)4Credit_maint2-0.0202-0.0248-0.03660.00090.0076(0.0302)(0.0280)(0.1130)(0.0063)(0.0950)Ownershipt14Credit_maint1-0.1110-0.1960*0.1290-0.0190-0.3330(0.0699)(0.1020)(0.3290)(0.0199)(0.2440)Ownershipt14Credit_maint20.0817-0.03180.4050-0.0323-0.4290(0.0853)(0.0754)(0.3170)(0.0198)(0.2960)4Credit_othert1-0.01540.0772-0.3920**0.00980.1950(0.0284)(0.0489)(0.1990)(0.0102)(0.1310)4Credit_othert20.03900.0674-0.06570.00550.1110(0.0450)(0.0462)(0.1520)(0.0088)(0.1310)Ownershipt14Credit_othert1-0.05580.11600.2330-0.0463-0.6340(0.0856)(0.1520)(0.6220)(0.0306)(0.3950)Ownershipt14Credit_othert20.1490-0.2600*-0.6780-0.0473*-0.6680(0.1370)(0.1350)(0.4660)(0.0284)(0.4190)ln(Fixedassetst1)0.02360.02290.02530.02260.0206(0.0416)(0.0420)(0.0416)(0.0417)(0.0417)ln(Employeet1)0.0706*0.0760*0.0705*0.0713*0.0724*(0.0404)(0.0402)(0.0400)(0.0401)(0.0401)ln(Firmaget1)-0.0323-0.0330-0.0260-0.0267-0.0250(0.0418)(0.0419)(0.0417)(0.0419)(0.0419)ln(CapitalLaborratiot1)0.01430.01360.01170.01240.0138(0.0364)(0.0365)(0.0363)(0.0366)(0.0366)ln(Valueaddedperworkert1)0.01150.02180.00350.02000.0233(0.0361)(0.0355)(0.0357)(0.0357)(0.0356)Constant-0.400-0.399-0.377-0.308-0.263(0.3690)(0.3650)(0.3650)(0.3690)(0.3710)Observation756756756756756R20.4080.4160.4120.4080.410Note:Coefontheindustryandyeardummiesarenotreported.Standarderrorsareinparenttheses.***,**,*indicatestatisticallyatthe1%,5%,and10%levelrespectively.742.4SummaryandConclusionTherehavebeenmanyarticlesthatexaminetheeffectofbank'shealthontherealactivities.ThemaincontributionofthispaperisthatItestwhetherthecor-porategovernanceasmeasuredbyownershipconcentrationfollowingJoh(2003),com-putedasthesumoftheshareholdingsbyfamilymembers,makestheexportre-silientwhenthenegativeshockoccurstothebanks.Also,withthehelpofhand-collectedmatchedloansdata,Icouldidentifythemainforeachandcomputetheshareofloansfrommultiplecreditorbanks.Mainisthatthemoreconcentratedownershipstructurefortheintermsoffamilymembersshareholdingshelpstheexportactivitymoreresilientwhenhealthofbanksfromwhichtheborrowsgetworse.Itisnotonlyhealthofitsmainbank,butalsothatofallothercreditorbanksthataffectstheexportactivity.Also,whenseparatingtheCreditbymainbankandotherbanksatthesametime,themainbankplaysmorerolethanothercreditorbanks.Thispaperhaveanumberofimplicationsforthefutureresearch.First,thematchedloansdataforanotheryearscouldmakethetestmorerobust.Giventhismatcheddatafortwoyearsonly,Iassumedthattheshareofloansfromthebanksforeachisconsistentevenbeforethosetwoyears.So,ifwecollectthematcheddataforlongerperiodandmorecouldthisgap.Second,wecanaddthechaeboldummyvariabletoseeifthechaebol-afcanmakesanydifferenceintheextensivemargineffect.75REFERENCES76REFERENCESAmiti,M.andD.E.Weinstein:2011,`ExportsandFinancialShocks'.QuarterlyJournalofEconomics126(4),1841Œ1877.Bae,K.-H.,J.-K.Kang,andC.-W.Lim:2002,`Thevalueofdurablebankrelationships:evidencefromKoreanbankingshocks'.Journalofeconomics64(2),181Œ214.Bernard,A.B.andJ.B.Jensen:2004,`Whysomeexport'.ReviewofEconomicsandStatistics86(2),561Œ569.Joh,S.W.:2003,`CorporategovernanceandprevidencefromKoreabeforetheeconomiccrisis'.JournalofEconomics68(2),287Œ322.77CHAPTER3TheSurveyofEffectsofFinancialShocksandCorporateGovernanceontheFirms3.1IntroductionSincetheglobalcrisis,researchersandpolicymakersarepayingmoreattentiontotheimpactoftheshocksontheoverallrealeconomy.,therearegrowingnumberofliteraturesthatstudytheimportanceofinstitutions-thebankinparticular-asacreditsuppliertotheinthesensethattheshockstothebankhealthorthecreditshocktobanksmayaffecttheinmanyaspects.Also,thecorporateownershipstructureisknowntohaveaimpactonthedecisiononaswellasperformance.Inthischapter,Iexplainandexplorethepreviousliteraturesthatinvestigatetheeffectsofshocksandcorporategovenanceonthesideof3.2EmpiricalWorks3.2.1FinancialShocksTheshockshaveanimpactonrm'sdiverserealactivities.Forexam-ple,studiesshowthatdomesticshocksnegativelyaffectcorporateinvestments(Duchinetal.(2010)),(AmitiandWeinstein(2013))),techspending,employment,andcapitalspending(Campelloetal.(2010))aswellasexport(AmitiandWeinstein(2011)).Also,basedonPeekandRosengren(2000),CetorelliandGoldberg(2011),CullandPeria78(2013)foundtheevidencethattheshocksaretransmittedbyglobalbanksevenacrossthecountrythroughbankingsector.Especially,therapiddeclineininternationaltradeduringthe2007-2009globalcialcrisistriggeredrenewedinterestinthesequestionsagain,withrecentstudiesafingthatcredittighteningwasanimportantchannelthroughwhichthecrisisdistressedworldtrade.Thecrisisin2008hasledresearcherstoaskwhethercreditcon-straintsfacedbyexportersplayedaroleintheglobaltradeslowdown.Thereisawiderangeofanswers.AmitiandWeinstein(2011),whichmatchesJapaneseexporterstothebankswhichprovidethemwithtradecreditandastronglinkbetweenthehealthofthesebanksandexportgrowth.Thereisalsogrowingmicro-levelevidencethatcreditmarketimperfectionsseverelyrestrictexportcapacity.Muûlsetal.(2008)showsthatliquidity-constrainedinBelgiumarelesslikelytobecomeexportersbyusinganindicatorofcreditworthiness.SimilarresultsarereportedbyBermanandHéricourt(2010),whoproxyliquidityneedswithbalance-sheetvariablesfor5,000in9developingeconomies.Achallengeforthesestudieshasbeenestablishingacausaleffectofcreditcondi-tionsonexportperformancesincethemeasuresofconstraintstheyuseareendogenoustointernationaltradedecisions.Morerecently,AmitiandWein-stein(2011)haveexploredexogenousshockstoavailabilityofexternalandshownthatJapanesebankstransmittedshockstoexportersthroughthechanneloftradeduringthesystemiccrisesthatplaguedJapaninthe1990s.Similarly,Bricongneetal.(2010)havefoundthattheexportsofFrenchinmoreexternal-dependentsectorsweremoreadverselyhitduringtherecentglobalcrisis.MinettiandZhu(2011)usesurveydataoncreditrationinginItalytoesti-matetheimpactofcreditrationingonexport.Theythatthattheprobabilityofexportingislowerforrationedandthatcreditrationingreducesforeignsales.79Paravisinietal.(2015)thatthereversalofcapitalduringtheglobalcrisisnegativelyaffectedtheexportcapacityofPeruvianexporters.ChorandManova(2012)thatvulnerablesectorsinsourcecountriesdidindeedexperienceasharperdropinmonthlyexportstotheUnitedStates.Behrensetal.(2013)arguethatforBelgium,totheextentthatvariablesaffectedexports,theyalsoaffecteddomesticsalestothesameextent.Incontrast,thereisaliteraturethatnoevidencethattradecreditplayedaroleinrestrictingimportsorexportsfortheUnitedStates(Levchenkoetal.(2010))Understandingtheroleoffrictionsforexportparticipationhasim-portantpolicyimplications,particularlyforunderdevelopedcountriesthathavehighre-silienceonextensivecross-bordertradeforeconomicgrowth.3.2.2CorporateGovernanceOwnershipstructureisanimportantfactorindeterminingvalue(DemsetzandLehn(1985)).JensenandMeckling(1976)concentratedownershipbyowner-managersmini-mizestheagencyproblemthatarisesfromtheseparationofownershipandcontrol.Thefamilyownershipcanleadtobettermonitoringofmanagerialdiscretionandre-duceprincipal-agentcostsassociatedwithdiffusedshareownershipstructures.Accord-ingtothisargument,thewherethefamilymembersholdmoresharesmayprovideacompetitiveadvantageandimproveshort-andlong-termperformance(AndersonandReeb(2003)).Thereareanumberofempiricalstudies:Chang(2003)thatcontrol-lingfamilyownershipisassociatedwithbetterperformanceinSouthKorea.Joh(2003)examinestherelationshipbetweenownershipstructuresandtheaccountingper-formanceusingover5,800sampleinKorea.Shethatbeforethecurrencycrisisin1997,Koreanwhosecontrollingfamilyshareholdershadmoreownershipoutperformedthosewherethefamilymembersownershipislessconcentrated.Mitton(2002)alsoshowsthathigherblockownershipbythelargestshareholderisassociated80withhighercrisis-periodstockreturns.Thecorporateownershipstructurealsohaveanimpactonthedeci-sion.Linetal.(2011)thatthecostofdebtishigherforcom-panieswithawiderdivergencebetweenthelargestultimateowner'scontrolrightsandrights.Linetal.(2012)showthatthedivergencebetweenthecontrolrightsandrightsofaborrowinglargestultimateownerhasaimpactontheconcentrationandcompositionoftheloansyndicate.Theyalsoinvestigatetherelationbetweenaborrowingownershipstructureanditschoiceofdebtsourceandthatthedivergencebetweenthecontrolrightsandrightsofaborrow-inglargestultimateownerhasanegativeimpactontherelianceonbankdebt(Linetal.(2013)).MinettiandYun(2015)suggeststhatowner-shipconcentrationnegativelyaffectsthepercentageshareoftheloanheldbytheleadarrangerasameasureofthesyndicatedloanconcentration.,andtheseeffectsweakenafterthechaebolreformofthelateninetiesinKorea.Also,Minettietal.(2015)Minettietal.(2015)studytheeffectsoffamilyownershiponexportusingdataonItalianandshowsthatfamilyownershipincreasestheprobabilitythatexport.81REFERENCES82REFERENCESAmiti,M.andD.E.Weinstein:2011,`ExportsandFinancialShocks'.QuarterlyJournalofEconomics126(4),1841Œ1877.Amiti,M.andD.E.Weinstein:2013,`Howmuchdobankshocksaffectinvestment?Evidencefrommatchedloandata'.Technicalreport,NationalBureauofEco-nomicResearch.Anderson,R.C.andD.M.Reeb:2003,`Founding-familyownershipandperfor-mance:evidencefromtheS&P500'.Thejournalof58(3),1301Œ1328.Behrens,K.,G.Corcos,andG.Mion:2013,`Tradecrisis?Whattradecrisis?'.Reviewofeconomicsandstatistics95(2),702Œ709.Berman,N.andJ.Héricourt:2010,`Financialfactorsandthemarginsoftrade:Evidencefromcross-countrydata'.JournalofDevelopmentEconomics93(2),206Œ217.Bricongne,J.C.,L.Fontagné,G.Gaulier,D.Taglioni,andV.Vicard:2010,`Exportsandsectoraldependence:evidenceonFrenchduringthegreatglobalcrisis'.Campello,M.,J.R.Graham,andC.R.Harvey:2010,`Therealeffectsofcon-straints:Evidencefromacrisis'.JournalofFinancialEconomics97(3),470Œ487.Cetorelli,N.andL.S.Goldberg:2011,`Globalbanksandinternationalshocktransmis-sion:Evidencefromthecrisis'.IMFEconomicReview59(1),41Œ76.Chang,S.J.:2003,`Ownershipstructure,expropriation,andperformanceofgroup-afcompaniesinKorea'.AcademyofManagementJournal46(2),238Œ253.Chor,D.andK.Manova:2012,`Offthecliffandback?Creditconditionsandinterna-tionaltradeduringtheglobalcrisis'.Journalofinternationaleconomics87(1),117Œ133.Cull,R.andM.S.M.Peria:2013,`Bankownershipandlendingpatternsduringthe2008Œ2009crisis:evidencefromLatinAmericaandEasternEurope'.JournalofBanking&Finance37(12),4861Œ4878.Demsetz,H.andK.Lehn:1985,`Thestructureofcorporateownership:Causesandconsequences'.Journalofpoliticaleconomy93(6),1155Œ1177.Duchin,R.,O.Ozbas,andB.A.Sensoy:2010,`Costlyexternalcorporatein-vestment,andthesubprimemortgagecreditcrisis'.JournalofFinancialEconomics97(3),418Œ435.Jensen,M.C.andW.H.Meckling:1976,`TheoryoftheManagerialbehavior,agencycostsandownershipstructure'.Journalofeconomics3(4),305Œ360.83Joh,S.W.:2003,`CorporategovernanceandprevidencefromKoreabeforetheeconomiccrisis'.JournalofEconomics68(2),287Œ322.Levchenko,A.A.,L.T.Lewis,andL.L.Tesar:2010,`Thecollapseofinternationaltradeduringthe2008Œ09crisis:insearchofthesmokinggun'.IMFEconomicreview58(2),214Œ253.Lin,C.,Y.Ma,P.Malatesta,andY.Xuan:2011,`Ownershipstructureandthecostofcorporateborrowing'.JournalofFinancialEconomics100(1),1Œ23.Lin,C.,Y.Ma,P.Malatesta,andY.Xuan:2012,`Corporateownershipstructureandbankloansyndicatestructure'.JournalofFinancialEconomics104(1),1Œ22.Lin,C.,Y.Ma,P.Malatesta,andY.Xuan:2013,`Corporateownershipstructureandthechoicebetweenbankdebtandpublicdebt'.JournalofFinancialEconomics109(2),517Œ534.Minetti,R.,P.Murro,andS.C.Zhu:2015,`FamilyFirms,CorporateGovernanceandExport'.Economica82(s1),1177Œ1216.Minetti,R.andS.-G.Yun:2015,`Institutions,BailoutPolicies,andBankLoanContract-ing:EvidencefromKoreanChaebols'.ReviewofFinance19(6),2223Œ2275.Minetti,R.andS.C.Zhu:2011,`Creditconstraintsandexport:MicroeconomicevidencefromItaly'.JournalofInternationalEconomics83(2),109Œ125.Mitton,T.:2002,`AcranalysisoftheimpactofcorporategovernanceontheEastAsiancrisis'.Journalofeconomics64(2),215Œ241.Muûls,M.etal.:2008,`Exportersandcreditconstraints:Aapproach'.Techni-calreport,NationalBankofBelgiumBrussels.Paravisini,D.,V.Rappoport,P.Schnabl,andD.Wolfenzon:2015,`Dissectingtheeffectofcreditsupplyontrade:Evidencefrommatchedcredit-exportdata'.TheReviewofEconomicStudies82(1),333Œ359.Peek,J.andE.S.Rosengren:2000,`Collateraldamage:EffectsoftheJapanesebankcrisisonrealactivityintheUnitedStates'.AmericanEconomicReviewpp.30Œ45.84