Default predictors in retail banking – an empirical study in vietnam

UNIVERSITYOFECONOMICS INSTITUTE VIETNAM THENETHERLANDS VIETNAM-NETHERLANDS PROGRAMMEFOR M.A.INDEVELOPMENTECONOMICS... DECLARATION...i ACKNOWLEDGEMENTS...ii ABSTRACT...iii TABLE OFCONTENT

UNIVERSITYOFECONOMICS INSTITUTE

VIETNAM THENETHERLANDS

VIETNAM-NETHERLANDS PROGRAMMEFOR M.A.INDEVELOPMENTECONOMICS

ACKNOWLEDGEMENTS

FirstandforemostIwouldliketooffermygratitudetomysupervisor,Dr.LeCongTru,forinvaluablecomments,remarksandengagementthrought h e l e a r n i n g processo f t h e t h e s i s ThenIhaveMr.LeDucAnhtothankforintroducingmetothetopic.Iamalsomuchobligedt o AssociateProf.Dr.NguyenTrongHoai,Dr.PhamKhanhNamandDr.LucaTasciottiforhelpfulremarksonmyTRDaswellaskeepingmeontherighttrack.Fortheavailabilityoft h e dataset,IamthankfultoMDE.TranThuTrangfromtheHeadOfficeofBIDV.Lastbutn o t least,Iamdeeplyindebtedtomyparents,mydearlybelovedwife,mybrothersandsistersf o r alltheunderstandingandspiritualassistance.Iwillwholeheartedlybegratefulforeverforyourlove

ABSTRACT

Duetointensecompetition,over-lendingandeconomicturmoil,bankingsysteminVietnami s performingloans.Giventheconsiderablegrowthofretailbankingmarket,anexplorationofriskpredictorsbecomescrucialmorethanever.Thispaperinvestigateskeyfactorst h a t influencel o a n repaymentperformanceamongi n d i v i d u a l customers.T h e s u r v e y coversa representativesampleo f personall o a n s fromo n e oft h e largestVietnamesecommercialbanks.A logistic regressiontechniquei s employedto evaluatetherelationshipbetweendelinquencyandborrowercharacteristicsandloanfeatures.Theregressionr e s u l t s revealthatborrowercharacteristics,e.g.b o r r o

sufferingahugeamountofnon-w i n g

history,bank-accountholdingandeducationlevel,ratherthanloanfactors,suchaspurposes,durationandcreditl i m i t , haves t r o n g e r effectso n t h e defaultoutcome.T h i s s u g g e s t s t h a t b a n k e r s a p p l y appropriateadjustmentstoborrowercharacteristicstominimize defaultrisk

Keywords:retailbanking,creditscoring,default,risk,logisticregression,probability.

DECLARATION i

ACKNOWLEDGEMENTS ii

ABSTRACT iii

TABLE OFCONTENTS iv

LISTOFTABLES vii

LISTOFFIGURES vii

LISTOFABBREVIATIONS viii

Chapter1 INTRODUCTION 1

1.1 Background 1

1.2 Problemstatement 2

1.3 Researchobjectives 3

1.4 Researchquestions 4

1.5 Justificationofthe study 4

1.6 Scopeof thestudy 4

1.7 Organizationofthe study 5

Chapter2LITERATUREREVIEW 6

2.1 Historyofcreditscoring 6

2.2 Conceptsofcreditscoring 7

2.3 Reviewsofeconomictheories 10

2.4 Reviewsofempiricalstudies 12

2.4.1 Defaultpredictorsinmarketsforcreditcards andinstantloans 12

2.4.2 Defaultpredictorsinmarketsforautomobiles,mortgagesandrealpropertyconstructio n 14

2.4.3 Defaultpredictorsinmarketsforindividualloans 16

2.5 Chaptersummary 17

2.5.1 Empiricalliteraturesummary 17

2.5.2 Problems andlimitationsofpreviousstudies 20

2.5.3 Conceptualframework 21

Chapter3DATAANDRESEARCHMETHODOLOGY 22

3.1 Datacollection 22

3.2 Variablesmeasurements 23

3.2.1 Responsevariable 23

3.2.2 Explanatoryvariables 24

3.2.2.1 Borrowercharacteristics 24

3.2.2.2 Loancharacteristics 27

3.3 Researchmethodology 28

3.3.1 Descriptive analysis 28

3.3.2 Econometricmodel 29

3.3.2.1 Methodologiesfor CSM 29

3.3.2.2 Logisticregression 30

3.4 Validation 32

3.4.1 Overallevaluations 33

3.4.2 Statisticaltestsofindividualpredictors 34

3.4.3 Goodness-of-fitstatistics 34

3.4.3.1 PseudoR-squaredstatistics 34

3.4.3.1.1 CoxandSnell'sR2 34

3.4.3.1.2 Nagelkerke'sR2 35

3.4.3.2 HosmerandLemeshowtest 35

3.4.4 Validationsofpredictedprobabilities 36

3.4.4.1 Classificationtable 36

3.4.4.2 Areaunder theROC curve 37

3.5 Analyticalframework 38

3.6 Chaptersummary 38

Chapter4DATAANALYSISANDRESULTS 39

4.1 Descriptive statistics 39

4.1.1 Personaltastesforloansbyages 41

4.1.2 Discretionaryincomes anddefault 42

4.1.3 Nexusbetweenloan amount andloanoutcomes 43

4.1.4 Loandurationandloan outcomes 43

4.1.5 Collateralvalueandloanoutcome 44

4.1.6 Differencesinvariablesbetweendefaultedandnon-defaultedloans 45

4.1.7 Correlationmatrixamongindependentvariables 45

4.2 Informationvalue 46

4.3 Empiricalresults 48

4.3.1 Modelestimation 48

4.3.2 Assumptionverification 50

4.3.3 Modelvalidation 51

4.3.3.1 Overallevaluationsandstatisticaltestsofindividualpredictors 51

4.3.3.2 Goodness-of-fitstatistics 52

4.3.3.3 Validationsofpredictedprobabilities 52

4.3.3.3.1 Classificationtable 52

4.3.3.3.2 Receiveroperatingcharacteristicandareaunder the ROCcurve 53

4.3.4 Resultinterpretation 53

4.3.4.1 Borrowercharacteristics 54

4.3.4.2 Loancharacteristics 56

4.4 Chaptersummary 57

Chapter5CONCLUSIONANDPOLICY IMPLICATIONS 58

5.1 Conclusion 58

5.2 Policyimplications 59

5.3 Limitationsandfurtherstudies 62

REFERENCES 63

APPENDIX 67

LISTOFTABLES

Table2.1.Creditscoringvs.Creditrating 9

Table2.2 Summaryofvariables 18

Table3.1.Overviewofvariables 31

Table3.2.Predictiveaccuracyof CSMs 36

Table4.1.Variablesinitiallyconsideredfor theCSM 40

Table4.2.Loantypestatistics 41

Table4.3.Loanduration 43

Table4.4.Differencesinvariablesbetweenthe loanoutcomes 45

Table4.5.Correlationcoefficientsamongcontinuousindependentvariables 46

Table4.6.Informationvaluesforexplanatoryvariables 47

Table4.7.Regressionresults 49

Table4.8.Classificationtable 52

Table4.9.Performanceofthemodels 53

LISTOFFIGURES Figure2.1.Processof creditscoring 8

Figure2.2 Conceptualframework 21

Figure3.1 ROCCurveandAUC 37

Figure3.2 Steps in binarylogisticregression 38

Figure4.1.Genderandloan sample 39

Figure4.2 Averageloansizevs.ageandpurposes 42

Figure4.3.Defaultfrequenciesamongdifferent groups ofdiscretionaryincomes 42

Figure4.4.Defaultfrequenciesamongdifferent groups ofloanamounts 43

Figure4.5.Defaultfrequenciesamongdifferent groups ofloanduration 44

Figure4.6.Defaultfrequenciesamongdifferent ratios ofcollateral-to-loan 44

Figure4.7 ROCcurvesandAUC 53

Chapter1 INTRODUCTION

Thischapterintroduces t h e thesis topic andidentifi esthe main issueswh ic h will

be coveredi n t h e f o l l o w i n g s e c t i o n s Theb a c k g r o u n d andm o t i v a t i o n t o t h e s t u d

y w i l l comefirst.Thentheresearchobjectives,researchquestionsandscopewillbeintroduced.Thenextw i l l bethemaincontributionofthestudyandthethesisstructureistobebrieflydisplayeda

tt h e endofthechapter

1.1 Background

Ins p i t e o f t h e w i d e v a r i e t y o f b a n k i n g b u s i n e s s e s , providingl o a n s

f o r corporatecustomersandindividualsconstitutesthemajorityofproceedsforcommercialbanksaswellasothercreditinstitutions.Asinformationasymmetriesprevail,lendersaretradingwitharisk

o f borrowersfallingindefault(Stiglitz&Weiss,1981).However,asymmetricinformationisn o

t theonlythreatsincesocialfactorsalongwitheffectsofbusinesscyclesmayalsoimpactu p o n t

h e d e l i n q u e n c y ( Allen,DeLong,& S a u n d e r s , 2 0 0 4 ).T o advocatel e n d i n g activities,measurementofcreditriskhasbeentakenseriouslyandthereforehasmadedramaticprogressovert w o pastdecades( Altman& Saunders,1 9 9 7 ).Theset w o scholarsp o i n t o u t severalforcesthatgiveimpulsetocredit-riskmeasurement Theyinvolve:

(1)aworldwideincreasei n caseso f b a n k r u p t c i e s ,

( 2 ) disintermediationtrendbyt h e largestborrowersandhighestquality,

(3)marginalcompetitivenessonloans,

(4)adecreasingvalueofproperty( a n d collateralasa result),a n d ( 5 ) a sharpr i s e i n

off-balancesheeti n s t r u m e n t s AfterBankf o r InternationalSettlements(BIS)haslaunchedtherevisedframeworkBaselII,banksareencouragedtopromotetheirapproachesoncredit-

riskmeasurement(Claessens,Krahnen,&Lang,2005)andvendorsstarttoofferimprovedmodelstobanksforcalculatingther e g u l a t o r y capitalrequirements

Togetherwiththerapidincreaseinbankloansforcorporatesandinstitutions,theneedo f i n d

i v i d u a l creditt o d a y i s ati t s highest( Brown,Taylor,& W h e a t l e y P r i c e , 2005)and" l e n

d i n g boomappearstob e pa rt ic ul a rl y strongi n t h e segmento f l oa ns t o h o u s e h o l d s , " asarguedbyBackéandWójcik(2008).Unlike thewholesalebankingwhich tradeswithlargeandt

y p i c a l l y r a t e d borrowers,t h e retailbankingd e a l s w i t h s m a l l l o a n s i z e s anda hug

en u m b e r ofpersonalclientswho,inmostcases,havenocreditratingsatall.Sinceeachloanisr e l a

t i v e l y notlargei n a m o u n t , i t i s impliedt h a t t h e r i s k o f defaulto n anypersonall o a n i s q u i t

e minimal.Traditionally,a l o a n approvali s basedo n t h e creditofficer'sjudgmento r

experiencefrompreviousdecisions.However,itiscostlyandtime-consumingforeachloanprofiletobeexaminedseparately.Infact,nolossonanyseparateretailloancanputabankcloset o insolvency.Hence,u n i t costf o r appraisingthe defaultr i s k o f a retaill o

a n m a y belargerthantherewardintermsoflosspreventionanditmightnotbeworthwhiledeterminingt h e riskonthebasisofanindividualloan.Asaresult,tomeasurethelevelofcreditriskasaw h

o l e f o r s u c h i n d i v i d u a l l o a n segmentation,b a n k s u s e l o a n defaultp r e d i c t i n g m o d

e l s orcredits c o r i n g m o d e l s ( C S M s ) w h o s e goali s t o forecastbadoutcomesandm a k e s u r

e t h a t goodloansarenot falselyrejectedandbadloansarenot wronglyacceptedeither

Accordingt o Yang,N i e , andZ h a n g ( 2 0 0 9 ) ,t h e creditr a t i n g ofb a n k s hast h

(2)Creditscoring,and(3)Probabilityofdefaultmodel.Sofartherehavebeenmanygoodremarksassociatedwiththoseapproachesasfollows.Brill(1998)arguesthatbuildingandrefiningaCSMcanhavecertainbenefitss u c h ascostsavingi n creditassessment,fastercreditanalysisandimprovementi n cashf l o w andcollections.ChenandH u a n g ( 2003)accountt h a t w i t h considerablel o a n portfolios,just

aslightenhancementincreditscoringauthenticitycanlowerthelenders'riskandtranslates i g n i f i c

a n t l y in to

latersavings.Fishelson-Holstine(2004)provest h a t C S Ms aredevisedtoaccommodatetheneedofincreasingloanvolume,mitigatingcreditrisksandtreatingcustomerimpartially.Thati s t h e reasonw h y s u c h t o o l sarebeneficialt o b o t h institutionalcreditorsandborrowers.Inthesameway,Allenetal

(2004)reckonthatbanksapplyingCSMstendtobemoreefficientatlowercosts.DinhandKleimeier(2007)insistt h a t ifagoodmodelisemployedwiththeavailabilityofreliabledata,scorin

th e risk.Noticeably,theBoardofGovernorsoftheFederalReserveSystem(2007)reportstothe Congressthat "creditscoringreducesthecostoflending orfacilitatesmoreeffectiverisk-

basedpricingofloans,increaseduseofcreditscoringmayexpandtherangeo f applicantstowhom lendersare ableto makeloansprofitably"(pp.42-43)

1.2 Problemstatement

BankingsectorinVietnamhasbeengrowingsignificantlyinthelastdecade.AreportbyMcKinseyGlobalInstitute(2012)revealsthattotalbankcredittoGDPi n nominallocalc u r r e n c yhasincreasedsharply fromapproximately 22%asof2000tomorethan120%tenyearslater,equivalentto33%annualgrowthwhichisthehighestamongneighboringcountries:India,C h i n a, Indonesia,Malaysia,Thailandandt h e Philippines.T h e statisticalfigurerevealsthatVietnamese

economytotallydependsonthebanks'sourcesnowjustafteronedecade.Therapidexpansioninbanks'lendingwillalsobringnon-performingloans(NPLs).

Whilethereportedlevelofbadloansappearstobeundercontrol,thetruevolumeislikelytob e muchhigherthanwhatispublicized.Inreality,NPLsclimbupto10%inMay2012,4%higherthanthatin2011andequalto10%oftheyear'sGDP.1Therefore,thesituationisanurgefor

astricterstandardofbaddebtrecognitioninordertomanagecreditrisk,especiallyint h e contextofglobalfinancialcrisis,Vietnameseeconomicdownturnandintensecompetitionoverthepastfewyears

AsVietnam'sb a n k i n g marketi s maturing,b a n k s havet o dealw i t h competitionn o t o n

l y

fromotherdomesticcreditinstitutionsbutfromwell-performingforeignbanksaswell.Despitethefactthatretailbankinghasbeengrowingrapidlyoverrecentyears,BIDVhasnottakenthistrendseriously

Anewchapteropensafteritsinitialpublicoffering(IPO),2t h ebankhappenst o perceivet h a t t h e wholesaleb a n k i n g a c t i v i t i e s cannotb e enoughf o r growtha n d prosperity,andthusnotguaranteeitsleadingroleinVietnam'sbankingsector.Consequently,recentchangesandd e v e l o p m e n t s i n t h

e fieldofb a n k i n g h a v e l e d t h e bankt o a renewed

interestinretailsegment.Sinceitisnoteconomicaltodevoteextensiveresourcestoanalyzepersonalcreditr i s k o f defaulto n a case-by-

casebasis,a CSMiso f importanceinstead.A s BIDVisoneofthelargestbanksinVietnamwithanationwidenetworkontheonehandandt h e standardo f baddebtclassificationhast o c o m p l y witht h

e s t a t e regulationo n t h e o t h e r hand,t h e implicationsint h e empiricalstudyatt h i s bankcanbegeneralizedf o r th e w h o l e Vietnamesebankingsystem

1.3 Researchobjectives

Thiss t u d y aimsatdistinguishinga goodr i s k f r o m a badr i s k , therefore,t h e keypurposei s t o e x a m i n e factorst h a t causedefaulti n personall o a n s andh o w t h e proposedmodelscanhelptomanagecreditrisk.Inpursuitofthis,theresearchattemptstofocusontherelationshipbetweentheloanoutcomesandborrowercharacteristics(e.g.housing,employmentstatusandincomelevel)andloancharacteristics(e.g.interestrate,collateraltypeandvalue).3M o r especifically,t h i s papermeasurest h e i m p a c t o f l o a n s i z e , durationandpurposeson

uponpersonaldefaultswoulddependonthewholeeconomiccycle.However,therearesomevariablesthatcanb e subje ctedtochangesinthemacroeconomy.Suchvariablesareincome,interestrateandloanduration.

1.4 Researchquestions

Toanswertheoverallaim,theresearchquestionsareasfollows

Mainquestion:Whatarethedeterminingfactorsthatimpactthedebtors'performancei n personalloans?

economicchangesafterthe2007USmortgagecrisis,whichsparkedaworldwidetougheconomicti

me.Third,am o d e l w i t h o u t t h e typicallysignificantl o a n characteristic,d i s c r e t i o n a r

y incomes,i s constructedtoprovidethebankwithpossiblecheckingformisinformationorpotenti

alfraud.Especially,theresearchtakesintoaccountthecreditprofileofclientstostudyhowthosewithdifferentb o r r o w i n g h i s t o r i e s i n c l i n e t o b e h a v e w i t h t h e i r currentl o a n s Finally,t

h e constructedmodelsaresubjecttoawiderangeofvalidationtestingtoinsuretheirreliabilityandgeneralization

1.6 Scopeofthestudy

Lendingprocessi s a r e l a t i v e l y straightforwardserieso f a c t i o n s i n v o l v i n g t w o principleparties.Theseactionsgofromthei n i t i a l l o a n applicationt o t h e successfulrepaymentoftheloanoritsdefault.Inlendingactivities,SummersandWilson(2000,p

38)summarizes i x functionalresponsibilitiesb a s i c a l l y connectedwithgrantingcredittoclients.Theyinvolve:(1)assessmentofclients'creditrisk,

(2)makingcreditgrantingdecisionw i t h regardt o termsandl i m i t s ,

( 3 ) collectingd u e receivables(debts)anda c t i n g againstdefaulters,

( 4 ) mo ni to ri ng clients'behavior andcompilingmanagementi n f o r m a t i o n ,

(5)bearingrisk ofdefault,and(6)financingtheinvestmentinreceivables

Thisresearchinvestigatest h e f o u r t h o f t h e s i x l e n d i n g responsibilities.Hence,t h ewritingconcentratesonstatisticaldatacollectionofcustomerbehaviorsandtriestodiscovermanagerialoutcomefromanalyzingthedatawhichiscollectedinthe2008-

2011period.Thescopeofthestudyisdefinedtoincludeconsumercredit(coveringautomobiles,recreationalvehicles,education,creditcards,etc.),homeimprovementloansormortgagesforpurchasinga house.Individualloanswith businesspurposesarealsoincludedin the research.4

1.7 Organizationofthestudy

Ther e m a i n i n g o f t h e e s s a y i s organizedi n t h e f o l l o w i n g way.T h e s e c o n d chapterbeginsbyl a y i n g o u t t h e h i s t o r y ofdefaultpredictingandt h e n l o o k s att h e theoreticalandempiricaldimensionsoftheresearch.Chapter3describesthedata,researchmethodologyandcriteriaf

o r a s s e s s i n g t h e q u a l i t y o f t h e econometricm o d e l s C h a p t e r 4 analysest h e data,reportsthefindingsandevaluatestheresults.Thelastchapterconcludesthepaperwithpolicyimplicationsanddiscussesthe limitationsanddirectionsforfurtherstudies

4 CSM's

roleisimportantinthe

BaselIIimplementationforretailportfolio.Smallloansforbusinesspurposeswithamountslessthan$1,000,000intheU.S.and managedonapooledbasisarecategorizedasretailcredits.

Chapter2 LITERATUREREVIEW

Thepurposeofthischapteristoreviewtheoreticalandempiricalliteratureondefaultpredictorsinretailbanking.Thechapterisdividedintosixsub-

sections Thefirsttwopartsd i s c u s s

t h e h i s t o r y aswellasconceptsofcreditscoring.The nexttwopartscovert h e economictheoriesandempiricalstudiesonthesubject.Thenextonesummarizesempiricalliterature;especiallyitrecallsthesignificanceofthemanyvariablesthatareusedtopredictcreditperformance.Thefinalpartprovidessummaryofdeterminantsofcreditdefaultintheempiricalliterature

2.1 Historyofcreditscoring

Mostacademicwriterscitea reportbyDurand(1941)releasedthroughtheNationalBureauofEconomic Research(NBER)astheearlieststatisticalapproachtothe issueofcreditapplicationselection.Inthe1940stheso-

calledcreditscorecardsystemswereimplementedf o r somemailorderfirmsandfinancialcreditsuppliers(Lawrence&Solomon,2002).However,itwasnotuntil1958thattheinitialcommercialratingsystemswerecreatedbyBillFairandE a r l Isaacf o r AmericanInvestment,a financecorporationl

o c a t e d i n S t Louis,M i s s o u r i

(Fishelson-Holstine,2004).Theirfirstprojectswereprovedtobesuccessfulasthosescoringsystemshelpedtoreducedefaultcasesupto20-30percentwhile sustainingsimilarvolumes;theycouldalsobeusedtomaximizelendingamountby20-30 percent with thesamedelinquentlevel

TheexerciseofCSMbecamemorewidespreadintheearly1960swhenthebusinesso f creditcardmaturedandtheneedfordecision-

makingspeedbecamenecessary(Anderson,2007).Aftert h a t , credits c o r i n g p r o c e s s i n g wasappliedt o o t h e r c u s t o m e r c l a s s e s Att h i s p o i n t , MyersandForgy(1963)madeacomparisonbetweendiscriminantanalysisandregressioni n r a t i n g applications.Later,a b a n k

r u p t c y p r e d i c t i o n m o d e l wasdescribedbyBeaver(1966).T h e s e w o r k s focusedo n t

w o a s p e c t s : failurep r e d i c t i o n andcreditq u a l i t y classification.AccordingtoAllenetal.(2004),themostusedtraditionalCSMwasthe1968m u l t i p l e discriminantanalysisforcorporationsbyAltman.Morethanonedecadelater(1980),t h e sameauthorintroducedth e basislendingprocessforbanksasanintegratedsystemandm a d e ananalysison howto setthecriteriaforcommercialloanassessment

The1960s,accordingtoRitter(2012),alsowitnessedadramaticallyincreasingattentiont o

a m a t t e r o f equality,s p e c i f i c a l l y i n t h e c o n t e x t o f t h e AmericanC i v i l RightsMovement.Inthosedays,singlewomen,divorcees,orwidowscouldnotgetamortgageloanw i t h o u t providingamalerelativeasaco-signer.Inaddition,minorityapplicants,particularlyAfrican-Americanb o r r o w e r s , ande l d e r l y customersexperiencedl i m i t e d o p t i o n s i n gainingcredit.Underthecircumstances,theEqualCreditOpportunityAct(ECOA)wasimplemented.T h e 1 9 7

4 E C O A makesi t illegali f creditorsdiscriminateagainstt h e i r applicantso n t h e groundso f "race,color,religion,nationalo r i g i n , s e x , maritals t a t u s , o r age(providedt h e applicanthast h e

c a p a c i t y t o c o n t r a c t ) " (ECOA,1 5 U S C 1 6 9 1 ets e q , Part2 0 2 , Section202.1).Sincethen,consumercreditdecisionshaveconsiderablyreliedmoreoncredithistoryprovidedbycreditreportingagenciesthathaveusedverifiabledatatobuilduptheircomputerizedfiles

Thankstothesurgeofautomatedstatisticalcreditratingandmortgagescoringasanapproachtoapprovingandunderwritingloans,housingfinanceevolvedinthe1990s.Automated

u n d e r w r i t i n g , whichwaspreviouslyemployedi n creditcardbusinessandautomobileloans,hasturnedintothemostimportantmortgageunderwritingapproache s p e c i a l l y s i n c e 1 9 9 5 (Straka,2 0 0 0 ).Besidest h e CSM'swiderandwideru s e i n m o r t g a g e origination,Mester(1997)demonstratedthatasmanyas97%ofbankshaveemployedCSMst o evaluatecreditc a r d applicationsw h i l e 7 0 % o f themhavet a k e n o n s u c h t o o l s f o r t h e i r s m a l l businessfinancing.SinceAltman'straditionalCSM,therehavebeenothermethodologiesformeasurementofcreditrisk.Inaccordancewith Al le n etal

(2004),therearef o u r approachesi n dealingw i t h multivariateC S M s :

( 1 ) l i n e a r p r o b a b i l i t y m o d e l , ( 2 ) logisticmodel,

(3)probitmodel,and(4)multiplediscriminantanalysismodel.Allofthefourm o d e l s canbesuccessfulinpickingoutvariablesthathavepredictivepowertodifferentiatedefaultersfromnon-defaulters

ng predicting

Scoring model

X 11 , X 12 , X 13 , Y 1 = ?

X 21 , X 22 , X 23 , Y 2 = ?

X n1 , X n2 , X n3 , Y n = ?

Source: Liu Yang ()

mount).Theseassessmentslaythefoundationonthecreditor'sownexperiencewithhistoricalinformationa n d a vi ew o f t h e a p p l i c a n t s ' prospectst a k e n i n t o account.Andersone x p l a i n s thatthisapproachissuitableforcommunitieswherethelendersandborrowershavep e r s o n a l l y knowneachother.However,itisnotefficientinmoderntimeofextendedbranchnetworksandcustomermobility.Regardingcreditdecisionbasedonsuchjudgmentalsystem,evenatitsbest,isstillinaccurate.Fishelson-Holstine(2004)reckonsthecreditor'srule-

basedsystemsasa serieso f h u r d l e s Eachapplicanth a s t o fulfillallt h e c r i t e r i a i n o r d e r t

o b e approvedwhereaseachfactorisregardedinisolation.Therefore,itisunlikelyforstrengthstooffs etweaknesses.

Onthecontrary,aCSMperformsarigorousanalysisofavailabledataandisbasedona thoroughknowledgeabouttherelationshipbetweenhistoricalbehaviorsandperformanceint h e future.Theresultofscoringisasinglescorewhichrepresentsabalancedsnapshotofaparticularapplicant'srisk.Apersonmayhaveweaknessesinacertainareabutgainstrengthsi n another.Thelinkbetweenallofthefactorsisexaminedandeveryfactorisgivenweightini t s relationshipw i t h others.A n

d e r s o n c o m m e n t s t h e credits c o r i n g u s e hass h i f t e d creditbusinessfromrelationshi

unsecuredlending.The factthat experiencehasbeenreplacedbydatahashinderedh u m a n judgementfromplayingmoreofarole.ThoughthefiveCsarestillapplied,byextractingthemaximumvaluefromavailablei n f o r m a t i o n , credits c o r i n g cancapturem u c h oft h e m Thisdoesn o t meant

h a t h u m a n j u d g e m e n t andcollateralt o t a l l y disappear;theyarethere, yetw i t h a le ss stresse

d role

Underscoringapproach,borrowersarerankedaccording totheprobability thatthey

w i l l

belateintheirpaymentsordefault.Creditorstypicallysetupacut-

offscoreforacertainacceptablelevelofrisk.Forinstance,acreditormightestablishacut-offscoreforhisorherportfoliosothattheminimumoddsofrepaymentare25to1.Theapplicantswhosescoresarebelowthe cut-

offwillberejectedwhile thosewho areratedaboveitwillbe accepted.T h e levelofjustifiedriskcanvarybetweenlenders,portfoliosandmaychangefromtimetotime.T h e c u t -

o f f v a l u e c a n alsobea goods t a n d a r d ofp r i c i n g l o a n s inaccordancewiththedelinquentrisk

RelevanttoCSM,credit ratingmethodisalsoaddressedtoriskmanagement.However,

t hi s a p p r o a c h considersa widerr a n g e o ffactorst o c l a s s i f y corporate andinstitutionalcreditrisksintogrades.Thefollowingtableshouldmakeclearabouttheblurringb o u n d a r y betweenthesetwoapproaches

businessloansandi n s t i t u

t i o n a l loanswithininstituti ons

worldwidecompanies,fina ncialinstrumentsandSover eigns

numberofgrades(withinc o m

m o n frameacrossmajorr a t

i n g agencies) Users institutionitself institutionitself,supervisor lenders,investors,regulators Mainp

r d l y bep o s s i b l e f o r a ju dg em en t al d e c i s i o n procedureto weightand

assessthatmuchcomplexity.Furthermore,aportionoftheavailabledataseemstocorrelateo r havenoreliableconnectionwithpaymentbehaviorinthefuture.Byselectingtheproperfactorsandanalyzingthecorrelation,CSMuserswouldbenefitbyminimizingtheinfluenceo f poorqualitydata.

ThemainfeaturethatdistinguishesmostCSMsiswhethertheydependonexhaustiveanalyticalstatisticso f realcreditexperiencet o ascertainwhichfactorsaret o b e p u t i n t o considerationaswellast h e weightthat eachonesh ou ld be assignedi n making finalcreditdecision.Byemployinga consistentdataset,a la rg e number ofstaffi n anorganizationcanreacht h e s a m e credit

d e c i s i o n Nonetheless,judgementalapproachesa n d C S M s m a y n o t alwaysproducesamedecisionswithrespecttosameapplicants.Underthecircumstances,itissuggestedthatsomefurtherexaminationsareneededinsteadofapprovingorrejectingthemd i r e c t l y (Dinh&Kleimeier,2007).Lastbutnotleast,aCSMisnotapanaceabutitshouldbetreatedasa refinementt o o l i n t h e processf o r creditgrantingi n ordert o formt h e ultimatecombinationo f b o t h humanbestpracticesandstatisticalapplications(VanG o o l , Verbeke,Sercu,&Baesens,2012)

incomehypothesiswhichpostulatest h a t spendinghabitsofpeopledonotchangeorarestableovertimewhenconsumerincomechanges.Keynesiantheoryrevealsthefactorshavingimpactonconsumptionareasfollows:

(1)individual'sreali n c o m e , (2)pastsavings,a n d (3)interestrate.However,o n e o f n o n

-i n c o m e determ-inantsthat-influencesMPC-isconsumercred-it.W-ithcred-itava-ilab-il-ity,consumersspendtheirincomemoreandotherwise,ifcreditiscostlyornoteasilyaffordable,consumerincomeisspentless.ItisnotedthatMPCismuchhigherforonewhoisnearertohis

orhercreditlimit(Gross&Souleles,2002)

Concerningconsumptionhabits,Duesenberry(1949)revealsthatpeopleusuallyobservebehaviorandconsumptionpatternsfromtheirpeerstodecideontheir consumptionandsavings.Ifaconsumerfeelsanurgeforconspicuous consumptionasawhole,ahigherstandardo f c o

n s u m i n g actionm a y bedisplayed.Namingt h i s p h e n o m e n o n demonstrationeffect,heargu

esthatitpromotesunhappinesswithpresentconsumptionlevels,whichimpactso n savingsratesandt

h u s affectsmacroeconomicgrowth.C l o s e t o Duesenberry,N u r k s e (1957)addsthatthesociety'sexposuretonewgoodsorwaysoflivingbringsaboutunhappinesswithpreviouslyacceptedconsumptionpractices.Theconsumers'preferencefunctionsarenotindependentbutinterdependentas"peoplecomeintocontactwithsuperiorgoodsors u p e r i o r p a t t e r n s o f consumption,w i t h newarticleso r newwayso f m e e t i n g o l d wants."Hence,heconcludesthatthesepeople"areapttofeelafterawhileacertainrestlessnessanddissatisfaction.Theirknowledgei s extended,t h e i r imaginationstimulated;newd e s i r e s a r e aroused,t h e p r o p e n s i t y t o c o n s u m e i s s h i f t e d upward

"( p 5 9 ) A s a r e s u l t , creditdebtwillbeinvolvedinconsumptionifthereisanincomeshortageandconsumptionist o happen.Jappelli,Pischke,andSouleles(1998)provethathouseholdswithcreditcardsarebetterequippedtoconsumethanthosewithoutbankcards.TherealityisthatAmericanshavegrown comfortablewithconsumercreditasawayto smoothincomeover thelastfewdecades(Durkin,2000)

rationingissaidtohappenwhenpeoplewouldliketoborrowmoreataquotedinterestrateb u tcreditorspreventthemfromdoingso.6Intheirstudies,thoseauthorsanalyzethatbanksm a y rejectsomedebtorsbecauseofthetwocomponentsofasymmetricinformation.Thefirstcomponent–

a d v e r s e selection–

happenswhengoodb o r r ow e r s mightn o t t a k e th e l o a n ast h e y findinterestratetoohighwhileriskyborrowerscaneasilyaccepttotaketheloanatallcost.Thehighinterestratewillraisetheaverageriskofborrowers,possibly decreasingthelenders'profits.Thesecondcomponent–moralhazard–

occursaftertheloanhasbeengiven.Borrowersmightchangetheirbehaviors,i.e.toinvestinriskierprojectsbecausetheyarenotdealingw i t h t h e i r o w n funds.Ifs u c h r i s k y i n v e s t m e n t doesn o t payo f f , l e n d e r s cannot

5 In

mostothermarkets,suchasituationwillsimplyleadtoapriceincrease,whichmakesdemandequatesupply.

6 Another

notionofcreditrationingis,forexample,Bester(1985,p.850):"Creditrationingissaidtooccurwhen someborrowersreceivea

loanandothersdonot,althoughthelatterwouldacceptevenhigherinterestpaymentso r anincreaseinthecollateral."

recovert h e f u n d s A s rationallenderscannotc o n t r o l allt h e behaviorso f t h e borrowersdirectly,theywillformulatethecontracttermsinamannerthatinducestheborrowerstotakeactionsforthe sakeof thelenders,aswellastoappeallow-riskborrowers.

Despiteinterestratea c t i n g asa s c r e e n i n g device f o r distinguishinggoodr i s k s frombadrisks,itisnotthe onlycontracttermthatisimportant.AccordingtoStiglitzandWeiss(1981),thecollateralamountdoesaffecttheborrowerintermsofbothbehaviorandd i s t r i b u t i o n Theoreticalliteratures h o w s t h r e e m a i n reasonsexplainingwhycollateralrequirementsarecommonlyquotedinloancontracts.Firstofall,collateralisrequiredbecausetherei s n o c e r t a i n t y t h a t

a borrower'sfuturei n c o m e w i l l b e asexpected.Ifanenforceableclaimagainstone'sfuturewealthandincomecouldbeissuedthentherewouldben o m o t i v a t i o n f o r hisorherbanktodemandsomekinds

ofassetsasaguarantee(Plaut,1985).Therefore,thebankwillresorttocollateraltoreduceloanlossincaseofdefault.Thisreasoni s intuitiveandindependentofasymmetricinformationbetweenthecreditor andthedebtor.Secondly,collateralmayreduceadverseselectionproblemastheborrowerownsbetterinformationthanthebankbeforeitslendingdecision.Suchprivateinformationcanresultincreditr a t i o n i n g s i n c e t h e banki s u n a b l e t o sett h e l o a n p r i c e a c c o r d i n g t o

i t s customer'sq u a l i t y (Stiglitz&Weiss,1981).Actingasasignalingdevice,collateralcanconveyborrower'svaluableinformationt o t h e lender,w h o w i l l t h e n screendebtorsbyofferingth

e

alternativebetweenalow-pricesecuredloanandahigh-priceunsecuredone.Ahigh-qualityborrowertendst o c h o o s e collateralt o signalh i s o r herq u a l i t y andt h u s securesa lowerinterestratef o r t h e l o a n (Bester,1985).Finally,collateralc a n helpt o reducemoralh a z a r d problemsoncet h e m o n e y h a s beendisbursed,byd e t e r r i n g t h e borrower'sm o t i v a t i o n s t

o i n v e s t t h e fundinriskierprojectsorto makelessefforttobringthefinancedprojecttosuccess(Boot,Thakor,&Udell,1991).Usingcollateral,indeed,thebankisabletoalignitsdebtor'sinterestwith its ownsincethedebtorwillendureagreaterloss ifhisdefaultshould happen

2.4 Reviewsofempiricalstudies

2.4.1 Defaultpredictorsinmarketsforcreditcardsandinstantloans

Manyo f t h e p r e v i o u s empiricalstudieso n r e t a i l credit,e s p e c i a l l y c o n s u m e rcreditmarket,focusonmost traditionall oa ns whichdifferfroml oa ns f o r credit cardsandinstantl o a n s inthe following keyrespects.W i t h traditionalloans, loan amountsarepredeterminedandpaymentschedulesarefixed;whileincreditcardandinstantloanmarket,theactualloanamou

ntsareatt h e clientele'sdiscretionaftera f i x e d creditl i n e i s m a d e available.Debtrepaymentf

o r t r a d i t i o n a l l o a n s i s oftenpaidi n installments,whereasa creditc a r d h o l d e r

flexiblypaysbackhisdebtintermsofarequiredmonthlyrepaymentasacertainpercentageo f theoutstandingbalance.Finally,contrarytomanytraditionalloans,creditcardandinstantl o a n customersmaybegrantedw i t h a s m a l l e r amountoff u n d s w i t h o u t a n y s t r i c t requirementsoncollateral.

JacobsonandRoszbach(2003)d o researcho n c r e d i t

-s c o r i n g i n whichtheyrecommenda probitapproacht o e v a l u a t e p o r t f o l i o creditr i -s k Int h e i r study,t h e d a t a i s collectedfromnotonlyapprovedloansbutalsorejectedapplicantsaswell.Inthisway,theyj u s t i f y thattheirmodelsdonotsufferfromtheso-calledsample-selectionbiaswhichc o m m o n l y existsintheliterature.Thedatasetconsistsof13,338loanapplicantsfromamainl e n d i n g institutioninSwedenfromSeptember1994toAugust1995.Atfirst,57variablesareobservedbutonly16arecultivated.Thereasonwhymostofthevariablesareignoredisthatt h o s e variableshavelittleexplanatorypowerortheyarebetterexplainedbyothervariables.T h e empiricalevidencerevealsthatage,marital status,changeinyearlyincomeandcollateral-

freecreditamountaresignificantlyinfluentialinthedefault.Surprisingly,loansizes h o w s nosignificantimpactonthedefaultandhigherincometurnsouttoconnectwithhigherdefaultrisk

Agarwal,Chomsisengphet,andLiu(2011)e m p l o y anothert e c h n i q u e t o s t u d y t h

e influenceo f

individual-social-capitalr o l e o n i n d i v i d u a l defaultandbankruptcyo u t c o m e s T h e y useamonthlysetofpaneldatacoveringatleast170,000cardholdersintheU.S.fromJ a n u a r y 1997toJune2000andconductthestudywithCoxproportionalhazardmodel.Withobservationsofallborrowers'defaultandbankruptcyfilingstatus,theycouldfindthosethataffectt h e defaultarefinancialdistressfactors(e.g.r i s k i n e s s , income,debt,spendingandwealth),economicconditionsandlegalenvironmentaswellassocio-

demographiccharacteristics(e.g.age,statusofmarriageandhomeownership).Aftercontrollingforfinancialdistressfactors,economicconditionsandlegalenvironment,the resultsrevealthatcardholdersaremorelikelytodefaultiftheymigratefromtheplaceofbirth.Regardingage,t h e riskrisesbutthenfallsandgroupsthathavethesmallestbankruptcyriskaretheyoungest( 3 0 yearsoldoryounger)andtheoldest(60orolder).Anotherfindingisthatborrowerswhoaremarriedandownahousearerespectively17%and24%lessprobabletofallintoarrears,and25%

and32%lessinclined tofile for bankruptcy

DifferentfromtheabovestudiesAutio,Wilska,Kaartinen,andLähteenmaa(2009)don o t a

p p l y anyregressiont o researcht h e p o s s i b i l i t y ofdefaultb u t theyanalyzet h e demographicbehaviorsoftakenloans.Theycarryoutacomprehensiveinvestigationontheu s e ofinstantloansinFinlandinthecontextoffinancialcrisis.Thesampledataconsistedof

1,610respondentsamong1,951youngadultsfilling outquestionnaireswhichinclude demographicprofilesasgender,age,income,householdstructure,occupationandemploymentstatus.Theintervieweesarealsodemandedtorevealwhichkindoftheircredit:s m a l l loans,studentloans,creditcardsormortgages.Basedonthecollecteddata,theresearchersexaminethepeople'sattitudestowardsborrowing.Theresultsuncoverthatyoungpeopleoftheages18to23usesmallinstantloansmorethanthe24-to-29-year-

oldones.Ont h e otherhand,thelattergrouphasmoreuseofconsumercreditastheyhavehigheroccupationalstatusandincome.Genderdoesnotappeartoaffectthenumberofloanstaken,b u t income,occupationalstatus andhouseholdstructuredo

2.4.2 Defaultpredictorsinmarketsforautomobiles,mortgagesandrealpropert

yconstruction

Asautomobiles7arethemostcommonconsumptiongoodsthatAmericansoftenpurchaseo n credit,Agarwal,A m b r o s e , andChomsisengphet(2008)s h o w t h e i r interesti n investigatingt h

e performanceo f automobilel o a n s i n termso f defaultr i s k andp r e p a y m e n t risk.8T h em a t u r i

t i e s a r e f o u r andf i v e yearsa n d t h e l o a n performancei s observedf r o m J a n u a r y 1 9

9 8 t o March2 0 0 3 i n a largefinancialinstitutioni n t h e NortheasternU S T h e y conductacompeting-

riskmodelwithadatasetof20,466directpersonalloans9f o rpurchaseso f newandusedautomobiles.Theirmainfindingsareasfollows.Itismorelikelytodefaulto n aloanforausedcarwhereasthereishigherlikelihoodofprepaymentonaloanforanewcar.Surprisingly,a c u s t o m e r w i t h lowercreditscoreshasl o w e r p r o b a b i l i t y ofdefaultb u t higherlikelihoodo f prepayment.A s expected,a r i s e i n t h e loan-to-

valueratioraisest h e p r o b a b i l i t y o f defaulta n d decreasest h e likelihoodo f prepayment.Incomehasa p o s i t i v e impacto n t h e p r e p a y m e n t w h i l e unemploymenthasa n e g a t

i v e effecto n t h e l o a n performance.Paradoxically,adecreaseinthebasicratewillincreaseboththeprobabilityofdefaultandprepayment.Andmostinterestingly,borrowersformostluxuriouscarsinclineto

repaytheloansinadvancewhilethosewhoborrowtofinancemosteconomicalcarstendtoperformtheloanswell

Concerningmortgages,PeterandPeter(2011)u s e i n c o m e ando t h e r t w e l v e factorsw i

t h 3 , 4 3 1 WesternAustralianhouseholdsto estimate thelikelihoodofloandefault.Theyaim

7 Automobiles

meancarsandlighttrucks.Besidesautomobiles,othertermssuchasautos,carsandvehiclescanb e used interchangeable.

8 Prepayment

riskoccurswhenautopurchaserspayofftheloansearly,whichreducesorviolatesthecreditors'f l o w ofinterestpayments.

9 Agarwal

etal.(2008,p.18)Agarwaletal.makeitclearthat"directloansareissueddirectlytotheborrower,

andindirectloansareissuedthroughthedealer."Inthelattercase,financialinstitutionshaveagreementswitha u t o

m o b i l e dealershipstograntloansatfixedinterestrates.

toformt h e relationshipbetweenh o m e o w n e r s ' r i s k defaultandt h e i r characteristicsw i t h a logisticregression.Thestatisticsimplythatonaverage93%ofthehouseholdsarepunctual.T h e regressionresultsrevealt h a t i n c o m e i s a h i g h l y predictivevariable,i e loweri n c o m e levelisoneofthemainfactorsthatsendthemortgagorstodefault.Another notablycriticalfactorisloan-to-valueratiowhichmeansthelikelihoodofdefaultwouldincreasewithhigherloan-to-

valueratio.Onemorefindingisthatinallprobabilitythedefaultwouldbehigherift h e heado

f t h e f a m i l y i s l e s s educated,youngeranddivorcedi n comparisonw i t h o t h e r households.Theageoftheheadofhouseholdsisalsoagoodindicatorwhichsignalsthattheperformanceofmortgagerepaymentstendstobenegativelyimpactedbyyoungerhouseholds

Otherindicators,forinstance,employmentstatus,migrantstatusandmacroeconomicfactor10shownosignificance.Thedevelopedmodelisprovedtobeausefultoolthatbenefitsprivatelenders,p o l i c y m a k e r s andgovernmentinassessingt h e defaultriskanddevelopingt h e appropriatestrategiestominimizethem

Asthereh a v e b e e n m a n y studieso n p r e d i c t i n g defaulti n t h e d e v e l o p e d countries,KočendaandVojtek(2011)carryoutaresearchinaEuropeanemergingmarket.TheygainaccesstoadatasetofaCzechbankwhichspecializesingrantingpersonalloansofsmallandm e d i u

m

sizeinthefieldofrealestatereconstructionandpurchase.Thedatacontainssocio-demographiccharacteristicsaswellasotherinformationon3,043individualswhogotloansbetween1999and2006.TheirmodelsaretestedwithlogisticregressionandCART11analysis

afterinformationvaluesofvariablecategoriesarecalculated.Theresultisthatsixoutof21variablescanhaveabilityt o discriminatebetweengoodandbadclients.Theya r e alsovariablesthathavehighinformationvalues.Itisnoticeablethatbothapproachescanproducesimilarefficiencyanddetectthesamefinancialfactorsandsocio-

demographicfactorsasthek e y predictorso f l o a n performance.Clientsw i t h higheramounto f

o w n resourcesreflecta d i s t i n c t l y lower defaultprobability.Largerl o a n s arerecognizeda

sr i s k i e r debts.Loansf o r p r o p e r t y renovationtendtoexposetomore

riskthatthoseforrealestatepurchases.Likeotherstudies'finding,higheducatedcustomerscanbemucheasierinrepayingtheirdebts.Clientsw i t h longerrelationshipw i t h t h e bankarel e s s r i s

k y t h a n t h o s e w i t h shortertime.A l s o , marriedclientsshowsignsofmoresafetytograntloanscomparedtothosewithnospouse.T h e twoauthorsalsomanagetobuildawell-

performedmodelexcludingthemostsignificantfactor–

client'sow n resources.Bythisway,t h e y re- confirmed t h e p r e d i c t i v e poweroft h e demographicvariables

2.4.3 Defaultpredictorsinmarketsforindividualloans

Someauthorsdonotnarrowdowntheirexperimentstoaspecificloanpurposeliketheaforementionedliteratureb u t incorporateallp u r p o s e s i n t o t h e samem o d e l Forinstance,ÖzdemirandBoran( 2 0 0 4 ) e m p l o y a l o g i s t i c b i n a r y regressiont o evaluatet h e relationshipbetweenconsumerloans'defaultriskandtheirdemographicandfinancialvariables.Theyuset h e datasetof500customer recordsmostly involvingindividualsupportloans, homeloansandcarloansprovidedbyanIstanbulprivatebankfrom1999to2001.Interestingly,theyfindn o notablerelationshipbetweendemographicvariablessuchassex,ageandoccupationsectorandthedefaultrisk.Theonlydemographicvariablethatimpactsonthecreditperformanceisresidentialstatus.Meanwhile,thefinancialvariables such asinterestrateandmaturity havea b i l i t y topredicttheoutcomes.Thelongerdurationorthehigherinterestrate,themoreriskf o r customersn o t p a y i n g b

a c k t h e i r debtsi n t i m e ÖzdemirandBoranconcludefinancialvariablesh a v e m o r e significanteffectso n t h e customers'performancet h a n demographicvariablesd o T h e f i n d i n g s u g

g e s t s t h a t b a n k e r s s h o u l d p u t appropriatefocuso n financialvariablesto

minimizedefaultrisk

InVietnam,Vuong,Dao,Nguyen,Tran,andLe(2006)becomethefirsttoexplorethel i n k betweenl o a n performanceandborrowercharacteristics.1 6 featureso f customersaretakeni n t o consideration;m o s t o f t h e m aresocial-

demographicalvariables.T h e dataseti s obtainedfromT e c h c o m b a n k ,12whichcovers1 , 7 2 7

i n d i v i d u a l customerrecordsi n c l u d i n g 1 , 3 7 4

goodloansand353non-performingloans.Withlogisticregression,theirmodelproducesahighlypredictiveaccuracyof99%.Theauthorsthenofferaclassificationsystem

ofC S M basedo n t h e o u t c o m e o f performanceprobability.Forexample,goodcustomersclassi

2005periodcoveringawidevarietyofloanpurposesfrombusinessloans,mortgagesandhomeloa

ns,mobile assetfinancingtogeneralcredit( l i v i n g e x p e n s e s o r consumptionw i t h o u t collateral)andcreditc a r d facilities.T h e y

12 Techcombank

stockBank,whichwasestablishedonS e p t e m b e r 27th,1993andtheHeadOfficeisinHanoi.

standsforVietnamTechnologicalandCommercialJoint-initiallyrelyontheloanofficer'sexpertknowledgetoname22pertinentvariables.Afterthat,t h e forwardstepwisem e t h o d h e l p s t h e m choose1 6 o u t o f t h e first2 2 i n t h e m o d e l W i t h logisticregressionagain,theydevelopaflexiblemethodthatisconstructedfromthestandardsoftransactionallending,yetrelationshiplendingisalsoleftroom

for.Theyfindthatregion,residentialstatus,maritalstatus,collateraltypeandeducationarenoteffectivepredictors.Them o s t importantvariablesaret i m e w i t h bank,bankaccountfollowedbygender,loannumberandduration.TheirCSMissuggestedtoberegularlyu p d a t e d toresponsetoeconomicchanges

s n o t sufficient,butt h e covariancerisk13o rt h e defaultbetas h o u l d bealsotakenintoaccount.Intheirstudy,MustoandSoulelesaccessarichandunique

sourceofcreditdatafromExperian,oneoftheleadingU.S.creditbureaus.Thepaneldatasetcontains1 0 0 , 0 0 0 r a n d o m sampleso f m o n t h l y customersbetweenMarch1 9 9 7 andMarch2 0 0 3 Therearedifferentcategoriesofgrantedloans,suchascreditcards,automobileloans,mortgages,etc.Themajorresultrevealsthatthereissignificantheterogeneityamongcustomersinthedefaultbeta.Highcovariance-riskconsumerstendtogetlowcreditscores,

i.e.highdefaultlikelihood.Thereisapositivecorrelationbetweenthesumofmoneyreceivedbyborrowersandtheircreditscoresandanegativecorrelationbetweenthefundlenttothemandt h e covariancer i s k Thosew h o e x p o s e t o highcovariancer i s k i n c l u d e t h e young,t h e single,h o m e renters,l o w e r - i n c o m e c o n s u m e r s , andr e s i d e n t s i n regionsw i t h lowerh e a l t h -

insurancecoveragebut ahigherdivorcerate

2.5 Chaptersummary

2.5.1 Empiricalliteraturesummary

Thefollowingtablesummarizesandcomparesthecommonvariablesusedinpreviousempiricalliteraturet o p r e d i c t creditd e f a u l t T h e t a b l e mayn o t expresse x a c t l y t h e s a m evariablesemployedbyt h e authorsb u t i t meanst o reviewt h e m o s t s i g n i f i c a n t ones.T h evariablesa r e dividedintot w o groups:

( 1 ) Borrowercharacteristicsand( 2 ) L o a n characteristics.Itisnoteworthythatdividingandn

amingsuchcharacteristicscanvary

13 See

Lusardi's(2006)commentonhowtocalculatethecovariancerisk.

betweenauthors.Forinstance,ÖzdemirandBoran(2004)nameDemographiccharacteristicsv s Financialcharacteristics,KočendaandVojtek (2011)sortSocio-

demographicvariablesfromBank-clientrelationshipvariables,VanGooletal.

(2012)useBorrowercharacteristicsandL o a n characteristics.O n t h e contrary,s o m e authorsd

o n o t groupt h e variablesatall(Jacobson&Roszbach,2003;Vuongetal.,2006;Dinh&Kleimeier,2007).Inthisstudy,webaseonVanGooletal (2012)to classifythevariablesintoborrowerandloancharacteristics

Dinh&Kl eimeier( 2

Maturity/Loanduration ** **

regression 13,338loan

Logitr eg r

e s s i o n 500

Logitr e g r

e s s i o n 1,727retail

Logitr e g r e

s s i o n 56,037retail

Sweden

consumer loansin Istanbul

loansfrom Techcombank, Vietnam

loansin

V i et n a m

Borrowercharacteristics

Agarwal,Ambro se&Chomsiseng phet(2008) Autioetal.,(2009)

Kočenda&

V o j t e k (2011)

Peter

&Peter(

2011)

Agarwal,Chomsi sengphet

Questionnaire

&Descriptive statistics 1,610

Logitregre ssion,CA

R T 3,043real

Logitregre ssion 1,303

Coxproportio nalhazardmo del 170,793

Samplesize autoloansin

Northeastern U.S.

young adultsin Finland

estateloansi nCzech

homebuyers inWestern Australia

cardholders inU.S.

N.Smeansnotsignificant;

*,** and***meansignificancelevelat10%,5%and1%respectively;