Problemstatement
Creditisalargeindustryintheworldwideeconomyandplaysanimportantroleint he gro wtho f firmsa n d c o u n t r i e s Ino n e hand,manyenterpriseshaveutilizedc r e d i t lin estomakeprofitorspurtheirsales.Inanotherhand,creditinjectscapitaltotheeconomy ,allowproductionandexpansiontobringdevelopmenttofirmsinp ar ti cu lar and countryingeneral.Bankcreditcanbecategorizedintofourprimaryt y p es whichareloan s,discounts,financeleasingandwarranties.
Thegrantingofcreditplaysacrucialroleintheeconomicdevelopmentbecausethef a c t that noteverybusinesseshaveenoughmoneyorusealltheirmoneytofinanceth ei r p r o j e c t s However,c r e d i t i n s t i t u t i o n s don o t grantc r e d i t s o n i t s o w n t o a l l applican tsbutitshouldcomethroughaprocedureinwhichtheydecidewhetherorn o t toprovidec redittoaparticularapplicant Thereasonforthatistheyneedtoavoidtherisko facceptingbadloan(highprobabilityofdefault)orrejectinggoodl o an s (profitable loans).Thereforecreditriskmanagementholdsanimportantroleinbankingindustr y.T h e banksmanagecreditriskexposurethroughacreditriskpoliciessystemin whichevaluatethecreditriskofapplicant, sotominimizethed e f a u l t r i s k toget herw i t h maximizep r o f i t
T h e p r o c e s s o f i d e n t i f y c r e d i t r i s k includescollectingpreviousborrowe rs’information,classifying,a n a l y z i n g m u l t i elementsandvariablestoassesstheabilit yofclients’repayment.
Theincreasingdemandofcreditandindustrialintensecompetitionforcebankstoim plementnewschemestorefinedstatisticalmethodstofacilitatetheprocedureofmakingd ecisions thatisalsothestandardrequiredbyBaselCommitteeonBankingSupervision(BCBS)throu ghtheirsophisticateddocumentsandpapersleadingthebankingglobaltoanewcreditr ating/gradinggeneration.
Thecreditmodelsmakeprocesssystematicallyandalsoshortentimespentinthel o a n grantingp r o c e s s , t h u s r e d u c e t h e c o s t o f banksi n approving,m i n i m i z i n g objectivenessandinaccuratedecisionsbyusings t a t i s t i c a l techniquesand predictingt h e performanceo f c u s t o m e r s a s w e l l a s h e l p banksi n m e a s u r i n g th e risksandprofits.
Banks can enhance their credit risk management by utilizing models that streamline the pre-issuing and approval processes, as well as manage situations post-loan approval for borrowers Information technologies play a crucial role in aggregating vast amounts of data on bank customers throughout their loan lifecycles These records, which capture specific characteristics of customers at various points in time, are essential for credit rating and default prediction, ultimately determining whether customers fulfill their payment obligations.
Whatcouldbetheirp r o bl ems andwhethertheirproblemsrelatetotheirability ofpayingback?
Ifweassumet ha tf u t u r e w i l l fo ll ow t h e t re nd t h a t c r e a t e d i n t h e pa st ,t h e n cus tomer’sh is t o r i c al datacanbeusedtopredictanestimatedfuture.
Wen e e d t o keepi n mindt h a t , t h e s e a r e o n l y h e u r i s t i c models( a s o p p o s e d todeterministic),that canonlyprovideguesses,not absolute answers(andifallthingsa r e correctlypredictable,lifeissoboring).
Formanycountriesovertheworld,SmallandMedium- sizedEnterprises (SMEs)ar e c o n s i d e r e d a s backboneo f t h e e c o n o m y InO E
In Vietnam, small and medium-sized enterprises (SMEs) comprise over 97% of all businesses, highlighting their dominance in the economy According to 2015 statistics from the General Statistics Office, SMEs play a crucial role by creating over half a million jobs annually, employing more than 51% of the labor force, and contributing nearly 40% to the country's GDP They provide significant benefits, including job creation, attracting investment, reducing poverty, and increasing worker incomes, while also positively impacting the development of larger enterprises However, despite their undeniable importance, SMEs face challenges such as low-quality labor, weak management, limited technology, and difficulties in capital financing.
SISME’ssurveyr e l e a s e d o n J a n u a r y 1 3 , 2 0 1 6 , o n l y 3 0 % o f SMEsc o u l d a c c e s s t o bankl o a n i n t h e i r neededw h i l e t h e r e s t havet o s e e k t h e finan cial aidfrom othersourcesathighinterestrate.
ThereasonfordifficultiesincapitalfinancingforSMEscouldbethenon- widelyu se o f modelsf o r banksa n d o t h e r f i n a n c i a l i n s t i t u t i o n s t o giver a t e d f o r them.T h e r e areseveralreasonscouldbecountedonthisincompletion,which arelackofguidelinesfromtheofficials,ineffectivetoolstoevaluatefirm’screditwort hinesso r nothavingtheefficientmanagementsystem.
Thisseemst o b e a viciousc y c l e , SMEsdon o t haveenoughinformationa n d manag ements k i l l s tohavea g o o d d a t a , h e n c e t h e y f a c e d i f f i c u l t i e s i n r e c e i v i n g c r e d i t fromfinancialinstitutions,thenthereforeitcausestheincapacitytogrowandimplemen tnewmanagementskillsandinformationsystem.
Acknowledget he im po rt an t o f measuringp o t e n t i a l l o s s e s i nl o a n p or tf ol io , f i n a n c i a l institutionshaveinvestedalotofresourcestobuildstatisticalmodelstod o s o Howeveramongmanyc u r r e n t d e v e l o p e d modelst h a t mainlyf o c u s o n m easur in g c r e d i t riskf o r w h o l e s a l e c o m m e r c i a l l o a n s o r p o r t f o l i o , f e w attemptsthatattentiononsmallcommercialloansdespitethefactthatloansfromSM Esaredominantinbank’sassetportfolio.
Althoughalinearregressionmodelisfrequentlyusedina s c o r i n g modelbyestimati ngthecoefficientofcharacteristicvariables andgivingtheweightofscore,it hassomedrawbackssuchasimplicationassumingth attheattributemeasurementsarisefrommultivariatenormalpopulationssuchthattheclas seshaveidenticalcovariancemetrics,differingonlyinthevalueoftheirmeanvectors.Also, p o t e n t i a l l y mored a m a g i n g p r o b l e m i s t h a t manyo f t h e a t t r i b u t e s u s e d a s independentvariablesarediscreteandthuswouldtheninherentlytofollowa multinomial distribution.Theassumptionofthelinearregressionmodelinwhichvariableshavel inearrelationships,however,usuallydoesnotholdandisdeviatedfromthemultivar iatenormality assumption.Therearemanystudieshavingshowedt h a t mostoftheconsumercreditrati ngdatasetsareonlyweaklynonlinearandthat linearregressionandlogisticregressionbothgavegoodperformances.Thoseargumentsh aveencouragedmetouselinearandlogistictheoriesinthestudy.
Ingeneral,itishardtofindabeststatisticaltechnique tobuildupacreditratingmo del.Itdependsoncustomercharacteristic, structure ofdataset,particular p r o bl emsarisen,thepossibilityofgroupingapplicantsbasedontheircharacteristicsandobject iveofclassification,andthebanks’purposesforaspecificclientssegmentso r p r o d u c t s R e c e n t w o r l d w i d e s t u d i e s , w h i c h comparedt h e n e w advancedt e c h n i q u e s s u c h a s n e u r a l networksa n d s o p h i s t i c a t e algorithmsw i t h class icalstatisticalmethods,noevidencerevealsthatthesimpleclassicalmethodsmostwid elyusedinpracticedonotperformwell.
Researchobjectivesandresearchquestions
In Vietnam, Ernst & Young developed a credit risk assessment method in 2009 that includes over 100 indicators for evaluating enterprises This method has been applied universally across businesses of all sizes, despite the fact that a firm's size significantly influences credit risk Analysis of German and French SMEs indicates that small and medium-sized enterprises (SMEs) are generally at a higher risk compared to larger enterprises Additionally, extensive literature on credit ratings reveals that there is no definitive answer for banks regarding the best approach, as no single technique is universally applicable in all situations.
In this thesis, I aim to explore the transformation techniques used to convert data characteristics into variables that enhance model creation, particularly in the context of credit rating, which is data-rich and allows for user-defined groupings I will demonstrate the significant impact these transformation techniques have on expected outcomes, and I will delve deeper into these methods in subsequent chapters The primary objectives are to validate the effectiveness of classical statistical approaches in distinguishing customers, identify the determinants of SMEs' creditworthiness—including Probability of Default (PD), Expected Loss Given (ELG), and Probability of Loss (PDL)—and introduce a novel measurement of creditworthiness based on the number of late payment days to assess its effectiveness Lastly, I will examine how model settings influence banks' business management results.
Especiallyfromstatisticalandp r a c t i c a l perspectives,t o derives u i t a b l e m o d e l s basingonbinominalandmultinomialtargetsbutkeepingitsviabilityatthesametime Thisworkisalsoarealapplicationoftheelaborationofamodelfordecisionmaking.M o r e e x p l i c i t l y , I havet r i e d t o p r e s e n t a methodologicala p p r o a c h t h a t l e a d s totheelaborationofmodelshavingpowerinclassifyingclients.Bysolvingthem ainpurposes,Itrytofindsyntheticrelevantindicatorsthatsuittoarandomp o p u l a ti o n o fSMEsclients.Atthemomentthereisnouniversalscoringmodelthatc o u l d beusedby allthefinancialinstitutions,due tothefactthateachinstitutionpreservesi t s str ategyi n d e a l i n g w i t h t h e customers.T h e r e s u l t s s h o u l d h e l p t h e financiali nstitutionstofindvariablesthataremostaccuratelyusedincreditratings y s t e m andu nderstandhowlogistic/linearregressionworksinthisarea.
Payinge f f o r t s i n s o l v i n g t h e c u r r e n t l i m i t a t i o n , I haves h o w n t h a t b o t h logisticregressionandlinearregressioncanbesuccessfullyappliedtosolvecredi tratingp r o b l e m s Ialsohaveidentifiedpredictorvariablesofloanapplicantswhichpl ayanimportantroleindividingapplicantsintogroupsfordifferentcreditingpolicies.
Thepaperwillalsocomparebinomialandmultinomialapproachesbytheoretical a n d practicalevidences.Applyinginthespecificmarketcase,thepaperwillsuggestthe mostappropriatemodelforcreditratingfrombothpointofviewwhicha r e marketandstat istical.Hencewewanttoaddressthefollowingquestions.
Question1 : In t h e groupo f f i n a n c i a l a n d non- financialvariables,w h i c h o n e s becometheimportantindicatorsto predictthecreditworthinessofapplicants?
Thisq u e s t i o n w i l l b e a n s w e r e d b y regressingt h e groupo f f i n a n c i a l a n d non- financialvariablesaboutapplicantinformationagainstcharacteristicofdefau ltor not,loangroupsornumberoflatepayment daysusing binominallogistic,multino miallogisticandlinearregressionrespectively.
Question 2: Dothe cut-off pointsareimportant factors in implementingriskappetiteforbanks?Severalcut- offpointsrepresentforbanks’riskappetitewillbechosen todothebacktestsforabove ofregressiontocheckwhethertheymattertoc r ed i t worthinessassessment.”
Scopeofthestudy
Thisst ud yf oc uses on the c r e d i t ra ti ng of0 2 Vietnamesec o m m e r c ia l ban ks.
W e attempttoelaborateastatisticalframeworkforVietnamesebankstoimplement an e w creditratingmodelaswellashelpthemtogetasystematicandscientificwayto b u i l d c r e d i t riskmanagementp o l i c i e s o n d a i l y b a n k i n g b u s i n e s s a n d c r e d i t approval.
Thed a t a s e t w a s u s e d i n t h i s p a p e r comesf r o m 0 2 c o m m e r c i a l ban ksw h i c h providesmallloansneedtoberepaidinshortterm.Bothbankswithmoretha n25yearsoldandthecreditexposureforSMEsaccountsforaround10%oftotalcreditex posure.BothbanksarealsointhetimeforastrategytoretargettowardtheSMEsmarketf o r n ext5 y e a r s ( a s manyo t h e r Vietnameseb a n k s d i s c l o s u r e t h e sametargetsom erecentyears).
Datasetc o n s i s t s o f a l l a p p l i c a t i o n s receivedinJ u n e 2 0 1 5 , summingu p 2 , 3 4 2 observations.Alltheseapplicationsareformerclients,assuchtheirbehaviorin formationfromt he ir previousl o a n s a r e a v a i l a b l e a s w e l l a s informationa b o u t l o a n performance.ThereforeinformationabouthowlateClienthaspaidbacktheirloa nsandtheirowneconomic-socialcircumstanceatthetimeof grantingloansalsoa r e possibletocollect.
Atthetimethedatasetwascollectedforthisstudy,bothbankshaveusedthesameexpert method(createdbyEarn&Young-
2009)forallclientsincludeSMEs(andprobablytheystillusethatapproachfornow),withahugenumberofmorethan80indicators(butitisnowknowniftheyareallmeaningf ulinthepowerofevaluating,predictingandgradingcreditworthiness).
ContributionsandImplications
If these questions can be resolved, banks may find a reliable way to establish their own SME businesses and assist SMEs in overcoming capital challenges Once this is achieved, the outdated "subjective expert technique" introduced by Earn & Young in 2009 can be replaced with a more effective statistical methodology recommended by international standards and BCBS/Basel II Current statistical methods primarily calculate the probability of a client defaulting based on their characteristics, but they often overlook the likelihood of a loan falling into specific groups or which factors might predict a client's loan classification This issue is prevalent not only in Vietnam but also in many developed ASEAN countries, such as Singapore, which still categorize loans into groups for regulatory provisions The classification of loan groups is crucial as it directly impacts banks' provisions and net income, yet most existing methods fail to accurately predict loan group classifications.
PnL requirements (Shareholders’ requirements) Capital adequacy requirement
Policies and IT data mart
Loan group Exposure and Provisions (Accounting Model) Credit Scoring Model
Procedures: Credit Assessing, Credit Approving, Credit Portfolio Monitoring, Risk balancing, Credit Portfolio Planning, PnL procedures, Capital Adequacy Monitoring, Optimizing… Figure2-Creditriskmanagementtechniquesinbankingmanagement
Financialinstitutionscanuseourstudiedplatformtocreatetheirownsolutionson:Selectthemo stimportantindicatorsincreditmodelingprocess:selectingvariablesa n d informationt h a t a r e t h e mostimportant( l e s s informationrequirement,moresavingoncostandtimeandi nfrastructure,lesscollectingtimeforclients);
Use the researchlikea closesamplefor the wholemodelingprocessimplementing:i m p l e m e n t i n g newstatisticalmodelswhi chshowhowtocreatetherulestoaccept/ denytheapplicants.Buildingacreditratingstatisticalmodel,determiningw h i c h mode lst h e bankss h o u l d implementa n d w h i c h templatest h e b a n k c a n f ollo w t oidentifythemodels;
Beawareandbeablewiththeperiodicmodel’smodifyingtoreflectrisk– returnmanagement,decidingandmonitoringriskappetite:understandabouttheeffectsof themodels’characteristicsandfiguresonrealizingbusinessprofile.
Activelydailymanagethebenefitbychoosinganaccurateprovisionmakingplan:determ ining,incaseofacceptance,whichistheexpectedloangroupoftheclient,andbased onthebanks’policies,howmuchcanthebanksprice(interestrate)ont h e clientloa nandhowmuchdothebankshavetospendontheclient’sloangroupprovisioning.Monitori ng situationinwhichthecurrentclients’characteristicsmayc a u s e a r i s k t h a t t h e l o a n w i l l got o a higherl o a n group(moree x p e n s e f o r t h e bank).
Organizationofthethesis
Chapter 1 introduces the motivation for studying the role of credit in the overall economy and the ability of SMEs to access the banking credit market, focusing on the significant challenges faced by the Vietnamese market in credit rating techniques Chapter 2 reviews the current state of theories related to credit rating and client discrimination, presenting key findings from previous literature Chapter 3 details the approach, framework, and theoretical techniques used in credit modeling Chapter 4 analyzes the results and compares the predictive power of various models in identifying credit issues for SME clients The final chapter concludes the study.
Conceptsofcreditrating
Int h e f i n a n c i a l market,c r e d i t r i s k c a n b e s e e n a s t h e o l d e s t f o r m o f risk Iti s definedastheriskthatacounterpartinacreditagreementfailstomeethisorher obligation.Ina f u r t h e r d e t a i l , r i s k o f c r e d i t i s a c h a n c e t h a t t h e e x p e c t a t i o n o f receivingacertainamountofmoneyinagivenperiodisnotfulfilled.
Grantingc r e d i t i s t h e b a s i c a n d t h e biggestf u n c t i o n o f bank.T h e r e f o r e r i s k o f credi t takesrelevantrolesof institutionalriskasa whole.Becausethefuturecannotbefullyanticipatedandrepaymentassurancedoesnotex ist,allcreditsgrantedtoc l i e n t involverisk.Thereforeanalyzingcredits isthe taskofquantify riskarisenthededecisionofgrantingcreditornot.Theprocessinwhichcreditriskisqua ntifiedmustb e c o n s i d e r e d variousf a c t o r s r e l a t e d toapplicant.T h e maximum riskt h a t f in an c i a l institutionsc o u l d a c c e p t d e p e n d s ont h e i r p o l i c y , w h i c h i s c a l l e d “riskappetite”.
Duringthelastfewdecades,themarketofcreditproductsincreasedenormously ,andmosto ft h e i n s t i t u t i o n s a n a l y z e consumers’d a t a togivec r e d i t ,h owever,grantingcreditwidelydoesnotmeanthatrandomlydistributingcredittoallt hosew h o requestit.Itisnecessarytoconsiderafactorassociatedwithcreditw hichiscrucialincreditgrantingdecision,whichisrisk.
Creditisthemostimportantproductofbanking andfinancialinstitutionsbutloansa r e alwaysaccompaniedbyrisks.Theriskfor financialinstitutionsdependsonhoww e l l theycanidentifyitindiscriminatingbadapp licantsfromothers.Forsolvingt h i s problem,o n e way f o r banksand f i n a nc i a l i ns ti tu ti on s i s t o st a r t u s i n g creditrating.
Creditratingofclient’screditworthinesscanbeusedfordeterminingthedecision o fgratingc r e d i t o r n o t , i d e n t i f y i n g t h e i n t e r e s t r a t e s , s e t t i n g t h e l o a n limitandmitigatinglossesduetobaddebts.Alenderusuallymakestwotypesofdec isions: first,whethertograntcredittoanewapplicationornot,andsecond,howtodealwit hexistingapplication,includingwhethertoincreasetheircreditlimitornot.
Lendersusehistoricaldatagatheredfromobservedofapplicanttobuildapplicantsscor ecard.Theygatherdataaboutapplicants,suchasapplicant’sincome,financialas set s , typeofwork,andmanyothernon-financialinformation.
Professionalfinancial institutionstrytouse themoste f f e c t i v e av ai lab le meth odsf o r creditriskassessmenttomaketheirmostpossibleaccuratedecisions.Actually,th ecreditratingcanbefoundundermanyforms.AcreditscoreintheUSisa3- digitnumberthatrepresentsasnapshotofanindividualcreditrisklevelatapointo f ti me.TheMoody,awell-knownratingagencyusessystemofAaa,Aa1,Aa2,Aa3… whileTheFitchandS&PuseAAA,AA+,AA,AA-…fortheirranking.
Themodelso f c r e d i t r a t i n g a r e a p p l i e d i n t o f i n a n c i a l i n s t i t u t i o n s w h i c h a l l o w onlinecreditevaluation.Wecan,actually,befamiliarwiththeconc eptLOS(LoanOr ig i n at i n g System)whereclientsarerankedandcreditgratingdecisionsa remade.
Theavailablel i t e r a t u r e s o n c r e d i t r a t i n g aremainlyf o c u s o n t h e evolut iono f fi n a n ci a l i n d i c a t o r s f o r sampleo f e n t e r p r i s e s d u r i n g t h e i r o p e r a t i o n i n a n a l y z e d period.Themodelto quantifyfailureorsuccessofclient’smanagementstructureisbeingassessed,which canbe expressed as a linearcombination ofseveralfinancialindicatorstogetherwithnon- financialindicatorsandtheirowncharacteristics.
ThepioneerofcreditgrantingcanbeconsideredtoHenryWells,theexecutiveofSpi egelInc.w h o havedevelopeda cre di tr at in g modeldu ri ng theSecondWo rl d W a r
T h e reasont h a t h e cameu p w i t h t h a t modeli s h e n e e d e d a toolt h a t allowinexp eriencedanalystsevaluatec r e d i t w o r t h i n e s s f o r t h e f a c t t h a t manyo f t h e c ompany’squalifiedemployeeshasbeencalledfortheWar.
Let’shavea q u i c k viewo n c r e d i t r i s k managementf r a m e w o r k t h a t a f i n a n c i a l institutionoftenimplementsi n t h e i r system.B y h i e r a r c h y , w e haver i s k managementframeworkontop(whichincludesallkindsofrisk:creditrisk,o p e r a t i o n a l risk,marketrisk,strategicrisk….),followingbyacreditriskpolicies wherec a p t u r e a l l p r o c e s s i n c r e d i t r i s k identifying/managing/mitigating/ a p p e t i t e ….,thishugenumberofpoliciesisfollowedbymanyofproceduretoapplyi n dail ybankingactivities:creditapplicationgrading,loanissuing,loanmanagement,loanmon itoring,loanwork-out,loanprovisioningprocedures….
Intermofcreditriskevaluatingprocedure,twotechniquescanbelistedwhichareCre ditrating(whenissuingloan,bankstrytocapturetheProbabilityofDefault–
P D ) andLoangrouping(inwhichbankstrytoestimatetheabilitythatthisloanisgoingt ofallinacertaingroupthenmakecorrespondingloans’provisioning,oncet h e creditis granted).
Financialinstitutionsarerequiredtoh a v e toolt o a s s e s s t h e riska ss o c i a te d w i t h ea choftheirloanbeforegrantingtotheirclients,suchofthattoolhelpstosupportthemin thedecisionmaking process.Duringtheloan’slife,financialinstitutionsh avetomakecorrespondingprovision(makingcostintheincomestatement)andt h e s e costsaredecidedbythemselvesbasedonthelevelofriskdetected.
Day8 Day14 Day18 Day24 Day30 Day60
ACTION SMS Call Letter Call SMS Collection letter Callsome customers
Infact, t h e financial i nst it ut io ns havet omakecorrespondingprovision,w h i c h is c o s t intheincomestatementduringtheloan’slife.
Financiali n s t i t u t i o n s w h o gott h e moree f f i c i e n t a n d a c c u r a t e t o o l o f a s s e s s i n g cr edit riskhaveanadvantageoverothersastheywouldlesslikelytomake wrongdecisioningrantingcreditthenlessvulnerabletocreditriskascomparedt otheircompetitors.Cu rr en tl y, almostoflargefinancialinstitutionconstructt heirownedtool to assess creditusethecombinationofboth
CreditgrantingandLoangroupingthroughw h i c h providet h e measuremento f c r e d i t a p p l i c a n t ’ s risk,a s s u c h supportingthedecisionofgratingcreditornot.
Almostthefinalquestionsforacreditworthinessevaluationare:whatisthep r o b a b i l i t y o f d e f a u l t ( o r d i s c r i m i n a t i n g t h e c l i e n t t o bado r good),w h a t i s thepr obabilityth at the creditworthinessw il l comew or se o r bet te r( or i d e n t i f y if th e f u t u r e creditloangroupwillbehigherorlower).
Worldwideapproachesforcreditrating
Since the 1950s, credit ratings have been predominantly utilized in mortgage lending and credit cards Traditionally, banks have relied on the experience and judgment of credit approval experts to decide on loan issuance This expert-based method involves comparing current applicants' information with that of past borrowers who successfully repaid their loans, assuming similar repayment capabilities While this approach allows for the consideration of qualitative factors and historical credit records, it can be subjective and potentially inaccurate, as it may be influenced by the personal biases of the experts and lacks a comprehensive sample size Moreover, as population characteristics evolve over time, this method requires continuous updates without the establishment of new benchmarks.
Secondapproach,experts’methodsdeveloptoclassicalstatisticalmethodsusedind e v e l o p i n g scoringincludel i n e a r regression,l o a n s discriminatea n a l y s i s , logistic regressionprofitanalysisandMarkovChain.Statisticalmodelisabetterwayford e c i s i o n m a k i n g thanatraditionaljudgmentalmethod,butthemodelisnotperfect,sometimesa badapplicantswillreceivehighscore/ratingand willbe accepted,andviceversa,agoodapplicantcangetlowoneandberejected.
Besides,advancen o n s t a t i s t i c a l methodss u c h a s n e u r a l networks,e x p e r t s y s t e m andmathematicalp r o g r a m m i n g havea l s o b e e n a p p l i e d i n c r e d i t r a t i n g p r o c e s s duringrecentyears.
Oneo f t h e mostcommonlyu s e d a n d s u c c e s s f u l methodsi n t h e i n d u s t r y i s logit model.Partofthelogisticalmethodologies,logitmodelspopularizedinamajorityo f dev elopedcountriessincetheywereintroducedfrom1980s.
With the advancement of information technology, automatic decision-making applications have transformed traditional models In a typical automated credit application process, customers submit their information via an application form through the bank's platform, such as a website or mobile app Once the bank receives this information, it is processed quickly If the application is approved, funds can be transferred immediately to the applicant's bank account This process utilizes a Loan Originating System (LOS), where both internal and external information is displayed to the credit officer, who then decides whether to grant credit If approved, the applicant receives an automatic notification and must confirm their acceptance of the loan before the funds are directly transferred to their account.
Creditr a t i n g p l a y s a c r u c i a l r o l e i n p r o c e s s o f grantingc r e d i t Ith e l p s r e d u c e uncertainty,informationcostswhilehelppeopleinvestingt h e i r moneyi nthemarketbetteracknowledgethequalityofborrowers.Moreover,ratingsystemal soh e l p s t o improvet h e t r a n s p a r e n c y o f f i n a n c i a l marketb e c a u s e i t providest h e commonlanguagetothecreditriskevaluation.This transparencyisveryimportanti n thefinancialmarketforaiminginmanypurposes.
Empiricalstudiesoncreditrating
Alargenu mb er ofst ud ies onper so nal andent er pr is es c r e d i t rating/ grading havebeen c o n d u c t e d b y u s i n g variousmethods.M a n y o f t h e m a p p l i e d t w o o r t h r e e scoringmethodstoonepracticalissue.
Wiginton (1980) proposed the maximum likelihood estimation of the logit model as an alternative to the linear regression model for credit rating development After conducting a comparative analysis using actual data, the study concluded that the logit function is preferred over linear regression Similarly, Abdoue et al (2008) evaluated an Egyptian bank's personal loan data using three statistical techniques: linear regression, probit analysis, and logistic regression Their findings revealed that the ranking of these models varies depending on the bank's decision criteria.
(2009)comparedn e u t r a l networkandtheotherthreeconventionalstatisticaltech niquesinanEgyptianbank’spersonalloandata- set.Theconclusionisthesamethatthec h o o s i n g ofmodelsdependsonthebank’svi ew.Worth,A.P.andCronin,M.T.D.
In 2003, a study compared the performance of linear regression models with logit models using three sets of simulated population data, concluding that no single solution for credit rating could be established after analyzing various cutoff points Abdou and Pointon (2011) reviewed 214 studies on credit rating techniques and found no universally superior method applicable to all scenarios Additionally, Desai et al (1996) examined the effectiveness of neural networks against classical statistical methods, determining that when performance is measured by the percentage of correctly classified good and bad loans, logistic regression models are comparable to neural network approaches.
Ontheindicatorsside:Barakova,Irina,DennisGlennon,andAjayA.Palvia(20013)devel opeda n e w c r e d i t r a t i n g m o d e l , validateda n d comparedt h e performanceoftraditi onalparametric,semi-parametricandnon- parametricmodels.T h e y foundlittledifferencebetweenthesemodels,andconcl udedthattoranktheindividualsb y c r e d i t w o r t h i n e s s i s e a s i e r thant o p r e d i c t a c t u a l d e f a u l t r a t e s b y models.Forindependentvariables,in general,thefollowingriskfactorswillalmostt h e muststudiedwhenassessingaclient’scredi tranking:
(1)Character:characterisveryimportantf a c t o r b u t n o t s o e a s y t o b e measured. Expertmusti d e n t i f y i f a p p l i c a n t r e a l l y r e p a y l o a n o r n o t F o r t h i s therea r e somef a c t o r s t h a t w i l l givemoreclearlyinformationaboutapplicant:income, historyorpersonalcharacteroftheo w n e r s , humanr e s o u r c e s …
( 2 ) C a p a c i t y : i t i s a l s o veryi m p o r t a n t f a c t o r I t includesinformationaboutap plicant’sfinancialcapability,iftheclientcanrepayl o a n Revenues,N e t income,f i n a n c i a l statements…
Steenacker,A.,andJ.J.Goovaerts,( 1 9 8 9 ) ( 1 9 8 9 ) u s e d datafromBelgianbanksto buildupanumericalscoringsystemf o r p e r s o n a l l o a n s b y u s i n g logisticr e g r e s s i o n T h e r e s u l t s h o w s t h a t banksc a n adjust t h e c u t - o f f p o i n t w h i c h d e p e n d s o n thepercentageo f l o a n s t h e y w a n t t o accept.
(1):Financialinformation:Industry,Currentsolvencyratio,Quicksolvencyratio,Insta ntsolvencyratio,Workingcapitalturnover,Inventoryturnover,Receivableturnove r,Sales-to-fixeda s s e t s ratio,Totalliabilities/Totalasset,Long-termliabilities/ Ownerequity,Grossprofit/
N etsa lesa nd servicesrevenue,ROS,ROE,ROA,EBIT/loani n t e r es t expenses;
Non-financial information plays a crucial role in assessing a company's medium and long-term principal repayment capacity Key factors include an analysis of the cash flow statement from the most recent fiscal year, the professional experience of management personnel, and the years of operation since product launch Recent changes in business leadership and the percentage of transaction turnover via bank transfers or other credit institutions also impact financial stability Furthermore, industry prospects, the education level of management, and the status of debts with banks and other credit institutions over the past 12 months are essential considerations Additionally, understanding the situation of related customer groups and the number of banking services utilized by customers, such as deposit services, payments, and foreign exchange, is vital for a comprehensive financial analysis.
C, ),Informations h a r i n g a s r e q u e s t e d byBANKinthepast12months/ oratthetimeofmakingl o a n transaction,Easeofmarketentry(ofsamesector/ businesssector)ofnewbusinessa c co r d i n g t oc r e d i t o f f i c e r ’ s assessment,Levelo f p r o d u c t i o n technologyo f enterprise.
The literature on credit rating highlights its significant economic benefits for development According to Longenecker, Moore, and Petty (1997), credit ratings are encouraged in the U.S mortgage market to ensure underwriting consistency and cost-effectiveness Furthermore, credit ratings enhance small business lending by facilitating loan approvals without increasing default risks and making loan securitization more feasible TransUnion (2007) summarized that credit ratings lower decision-making costs, reduce moral hazard rates, and expand access to credit for applicants.
P a r i s i (2005)claimedthatoneofthebiggestbarriersformanycompaniesistheirc reditconstraints.He proposed thattoday’stechniquescansolvethethreemainproblems( a c c u r a c y , costandtechnol ogy)inpromotingcreditrating,thusincreasingthesalesa n d attractingmorecustomersby makingtheloanapprovalprocessefficientandbylooseningthecreditlimit.
Witht h e abovea n a l y s i s , t h e t h es i s h e l p s in identifyi f i t ex i s t s u i t a b l e s t a t i s t i c a l modelsforVietnamesebankandespeciallyforSMEsegmentandbasedonVietn amesebanksp o i n t o f viewo n t h e i r clients.W e w i l l a l s o t e s t i f t h e c u t - o f f p o in t s playimportantroleinmodelbuilding.
Thefollowing methodsusedinthis paperaretraditionalbinominallogisticregressio n,multinomiallogistic regressionandlinearregression.Alltechniquesaresuitablefortheproblemofcreditrating prediction.Thedetaildescriptionofthesemethodsisinthechaptersbelow.
Client Discriminating Output Client information Indicator group n1
Indicator group 1 Indicator group 2 Indicator group n2
Case 2: Expected future loan group Case 3:
Any argument on potential loan downgrading Case 1: Expected probability of default
Analyticalframeworkandhypotheses
Inthissection,we willpresentthegeneralpicturesofcreatingcreditratingmodels,st art f r o m c l i e n t i n f o r m a t i o n i n p u t ( c l i e n t information),t o p r o c e s s o f variablestreatmentt o c l i e n t d i s c r i m i n a t i n g o u t p u t ( c l i e n t grading.W e d e s c r i b e t h e r elationshipamongindicatorsbyaconceptualframeworkasfollow:
TheC l i e n t Informationi n c l u d e s F i n a n c i a l I n f o r m a t i o n groupa n d Non- FinancialInformationgroup,e a c h groupi n c l u d e s sub- groupsw h i c h a r e dividedb y t h e i r theoreticalandpracticalroleincorporategove rnanceaswellaseconomic–social
–political– operatingenvironmentoftheclient.ClientInformationisexpectedtob e putinablac kboxofcreditratingmodelingtogettheoutputofclientdiscriminating,w h e r e c l i e n t s a r e classifiedintot h r e e c a s e s : c a s e 1 showst h e e x p e c t e d probabil ityofdefault,case2givesustheexpectedfutureloangroupandcase3showsiftheclient haspotentialdowngradingprobability.
Thefinancialandnon- financialvariableswillbechosenindetailinnextsectionsw h il e forsummaryher e,pleaseseethefollowingtable:
Bachelor/ AboveAve rage/ Active/ VeryActiv eChange/ Nochange
Owners’abilities(equity,management,experience)asassessedbycredi tofficer
TotalC F>=0 Doubt/ NoAvera ge/ GoodBa d/Good
Clear/NotEstablishmentofoperationalprocessesandinternalcontrolprocess Exist/Not Internalpersonnelenvironmentofanenterpriseasassessedbycredit officers Good/Not
Planning Objectives,businessplaninnext1to3years
Historicalre lationshipw ithbanks principalandinterest)withinthepast12months RatioofsavingdepositsatBANKSonaverageoutstandingbalanceint h e pa st12months
RatioofrevenuestransferredthroughBANKSontotalrevenues(inthelast12 months)comparedwithratioof averagedebitbalanceatBANKSontotaldebitbalanceof CM(inthepast12months).
YearsStatusofdebtstoothercreditinstitutionsinthelast12months Ever/NoProspectsofbusinesssectoratthetimeofevaluation Incrisis/NoStabilityofinputswithmainimpactsonbusinesssectorofanenterprise
Dependenceonanumberofcustomers(outputmarket) Depend/No Relationsofexecutiveboardwithgoverningagenciesandrelevant ministries,departments(excludingcreditinstitutions) Good/Not
Estimationmethods
Attheearly stageof cre di t rating technologyinVietnam, c l ie n t ratingbydiscri minationweretheuniqueaccuratemethod,theystilloneofthemostpopularmethodno w(asalmostdevelopingcountriesuseexpertmethod).Byfar,themost accurateand po pu la r statisticalmethodislogistica p p r o a c h , w h i c h has less strictas sumptionsbutcanalwaysleadtoalinearresults.
Forthedetails,manyresearchpointoutsomanybenefitsofusinglogisticmodelsfo r t h e c o n s t r u c t i o n o f r a t i n g models:ther e s u l t s c o u n t f o r t h e variables’corr elations,w h i c h h e l p identifyinga l l i a n c e s w h i c h mayn o t b e d i s c o v e r e d ; t h e modelalsoi n d i v i d u a l l y a n d si mu lt an eo usl y c o u n t s fo r allvariables;theus e rc a n simplyevaluatetheerrors,mitigatetheproblematicissuesandoptimizethemodel.
Ingeneral,followlogisticregressionmodels,thedependentvariableisunderb i n a r y typ e,independentvariablescouldbeundercategoricalorcontinuoustype.
Alinearregressionisnotlimitedinrangewhilethebinominalvariablehasamust expectedw o r t h b e t w e e n 0 a n d 1 Logisticr e g r e s s i o n s (logita n d p r o b i t i n t h i s sense)allowpredictingtheprobabilityofaneventthatoccursornot.Theoutco me is1iftheclientdefaults(withprobabilityvalueis𝑝)andis0iftheclientdoesn o t default( withprobabilityiscorrespondingly1–𝑝).
Themultinomiallogisticfunctionsarea modificationofbinarylogistic(whereonlyt w o possibleoutcomescanoccur).Themodel fordichotomousoutcomevariableisb a s e d onlogisticdistribution. exp(� � �−� 1 )
Int h e p r o c e s s o f d i r e c t logisticregression,a l l explanatoryd e p e n d e n t variabl es comei n t o thef u n c t i o n simultaneously,w h i l e t h e s t e p w i s e methoda l l o w s o n e t o identifythemostimportantvariablesforclassificationthatshouldbeincludedint h e final model.
Forth is pape r, the usedlogisticregressioni s s t u d i e d withtheor de re d d e p e n de n t variable.Theoutcomeofordered logisticcanbeidentifiedinmanyways,inthisp a p e r , forabestfittingwiththeVietnameseconditions;theorderedlogisticoutcomew a s a d d r e s s e d b y c a t e g o r i z i n g bothcontinuousvariablesa n d d i s c r e t e variabl es.
Inthissection,wearegoingtointroduceinshortpresentationaboutlinear regression.Manypeopleaskwhywedon’tusealinearregression;itcansimplifytheap proachandcould bealsomoreunderstandable.Onecritical reasonisthat,themodelwouldgenerateindefinite expectedvalues(whichcanbe greaterthan1orsmallerthan0).Aswefocusheretheuseofnumberofdaysinlatepa ymentasasymbolforloan’sfuturegroup,bytheway,wecantransfertheindicatorandm akeameaningfuldependentvariablebyaflexibleconvenienttransformation.
Modelspecification
Tosolvethe research’sobjectives,wewill trytoformulatestatistical functions thatc a n forecastthePDofaloan,theabilitythataclientwillfallinoneoffivegroupsofl o a n b a s e a n d t h e f u n c t i o n w h i c h c a n supportf o r t h e l a s t findingso f l i n e a r regression.
Functionalform1:PDpredictionin whichweusebinominalformfordependentv ariablenameddefault01(1:default,0:notdefault)andusealogisticfunctiontop r e d i c t PD;
Functionalform2:Predictingexpectedfutureloangroupofanyclientinwhichw e u s e orde redlogisticfunctionwiththedependentvariableunderkindofloangroup( f 2 : from1to 5);
Functionalform3:Predictingthenumberoflatedaysinpayment,wherewetrytotransfe rcurrentloangroupintothenumberofdaysinlatepaymentandgivethisn e w de pe nde n tvariablelikea symbolfo rc or re sp on di ng loangroup.W e use f 2 n (transformedfr omf2)orlogarithmformofit:lnf2n(thatmakessenselikelnmustn o t anegativenumber)
Anyindicator/ categoricalvariablewillbetestedunderquantitativeformandqualitativeform(whichre flectbothcaseswherecreditofficercangiveclientpointso r categorizethem byorderingqualitativeranks,andfor that,thedummy/dummiesw i l l beused);
Regressiondiagnosticswillbeusedtodetectandremove multi-co- linearity(infact,p e r f e c t co- linearity),heteroskedasticity(notequalerrorterm’svariance),autocorrelation(correla tederrorterms)asifthosediagnosticscanbeimplementedf o r chosenmodels.
Inordertotestthestabilityofthemodels,weusesomewaystobacktest,wherethe coefficientsareestimated,thedependentvariablesarepost- estimatedandr esu lt s arecomparedwiththerealvalueofobservations.
Datasourcesanddatatreatment
Allo f t h e availablei n d e p e n d e n t variables,w h i c h w e r e u s e d b y t h e bank,w e r e co l l ec t e d t o apossiblyexplainingofexpectedfutureloangroupandthePD.Alltheexplanatorywerep re-analyzedtoidentifyanycorrelationwiththedependent.
Asw e c a n s e e i n d e t a i l f o r n e x t s e c t i o n s ( a l s o i n C h a p t e r 4 ) , thevariabl esa r e chosenandcategorizedintogroups.Also,somegroupswhichhavenotenoughdata o r nomeaningsdataweredropped.
Forabriefdescriptivestatistics andfiguresw h i c h arepresented below, we goto d at atreatmentforvariablestobetestedinthemodels
2342observations.Forquantitativevariables,I usedoutliertechniquetofindoutthe onesthatneedtobedetectedanddeleted.IusedQuartilefunctions(forabriefexplana tion,quartile1and3wereusedtobuildanupperandalowerboundofthevariables),we thanobtainthefollowingresults:X1with196outlierobservations,X2 with270 outlierobservations,X3with153outlierobservations,X4with201o u t l i e r obse rvations,X5with287outlierobservations,X6 with296outlierobservations,X7 with319outlierobservations,X8with189outlierobservations,X 9 w i t h
2 8 7 outlier observations,X10 with293outlier observations,X11 with310outlierobservations,X12with5outlierobservations,X1 3with465outlier observations,X14with153outlierobservations,X15with133outlierobservati ons,X16with150outlierobservations,X17with168outlierobservations,X18 with282outlierobservations.
9 ; i f w e keept h e o b s e r v a t i o n s w i t h o n e indicatoro u t l i e r , t h e sampler e m a i n s 1,25 2 whicharegoodforthestudy.WecalltheoriginalasFulldataandthesecondone( wherewekeepnooutlierobservationsand1indicatoroutlier)asSecondarydat a T h e d a t a w a s a l s o c l e a r e d o f meaninglessobservations;w e c l o s e t h e d a t a si mpletreatmentat1133observations.
BinominaltypenamedDefault01:istransformedfromB A N K S T 2 4 Loanclassification ,whereloangroup1or2areconsideredtobenotdefault,andloangroup3,4,5areconsi deredtobedefault;
MultinomialtypenamedF2:istheT24Loanclassification- attheirownrealgroup(category valuesare1,2,3,4,5withstrictlyordered);
Numericalt y p e namedF 2 n : i s t h e numbero f d a y s i n l a t e payment,w h i c h w a s transformedfromF2bytherule:ifF2=1,thenF2n(days);ifF2=2,thenF 2 n 0(days)
;ifF2=3,thenF2n`(days);ifF2=4,thenF2n0(days);ifF2=5,thenF2n60(da ys).Infact,thisruleisbasedontherealregulationofStateBankofVietnam.
Intheprocess,newdependentvariableswerealsocreatedforthepurposeofbettersignif icantandrealisticmodels,buttheywerenotinterpretedinthemodelswhiletheycouldnot comewithagoodresult.
Category Economic rationale Financial Variables
Therearemanytechniquesforselectingexplanatoryvariables,theycouldbealsou s e d as the y are,or inanotherf or m ofcategorized.H i s t o r i c a l l y , somanyp a p e rs a boutcreditratingaresupporting fortheuseofaccuratelycategorizingtechniques,a n d thiswillbethechoiceforthispaper.
TotalAssetsProfitability Abilitytoearnasatisfactoryreturns ROA,ROE Capitalization Measure of capital structure and leverage
Log Long term debt/Totalassets
Themodelwhereallexplanatorydependentvariableswereincludedisnamedthe“ful lmodel”.Thanksto STATAthatprovidestheWald test,wherethesignificanceo f everyexplanatoryindependentvariablewastestedtocom etoa“reducedmodel”,w itho n l y significantvariablesw e r e kept.However,o n c e w e e x c l u d e d variables,theomittedvariablescancausebiasandwe needtocheckwhe therthe reducedmodelcausesenhancementoverthefullmodel.Withthelimitedresource,Ichos eWaldtest,bythatway,IalsointroducedmyownideasbeforemakingsuchWa l d test. ForE.Altman’s2research,theexplanatoryvariableswerechosenasfollowing:
Initially,wehave84 independentvariablesXi(X1-X84)(which were used by
Clear unfiled fields or unique value for qualitative variable/no meaningobservations.Followingarethesubjecttobedeletedvariables:
History of off-balance sheet commitments (letters of credit,guarantees)
Ratioofrestructureddebt(principal)t o t o t a l outstanding(principal)atBA NKSatthetimeofassessment
Statusofinformationprovisionb y customersatt h e requestofBANKSw ithinthepast12months
Category Financial Variables last12months)
UtilizationlevelofBANKSservices(depositsandotherservices)ascompar edwitht h o s e ofothercreditinstitutions(excludingcreditservices)
Assets Averageageoftransportvehicles(trucks,passengercars, others)Extentofpropertyinsurance
Operationextentofanenterprise(Extentofp ro du ct consumption- onlyreviewingmainconsumptionoutlets)
Andp u t 4 1 i n d e p e n d e n t variablesi n correspondinggroupf o l l o w i n g theoretic alviews.
Att h i s s t e p , I w i l l t r y t o t e s t a l l t h e observationvaluesi f t h e valuesa r e n o t appropriated,andalsotransferthedataintocorrespondingranking
>10y 6 underbachelor 1 bachelorand above 2 Activenessandsensitivityofente rpriseleaderstomarketchangesa sassessedbyscoringofficers Qualitative
Leverage Totalliabilities/Totalassets Numeric NON
Owners’abilities(equity,man agement,experience)asasses sedbycreditofficer
Establishmentofoperationalpr ocessesandinternalcontrolp ro cesses Qualitative audit 1
Objectives,businessplaninn ext1to3years
Atleast1 times 1 (includingprincipaland interest)withinthepast12 Qualitative Maybe 2
Ratioofrevenuestransferredthr oughBANKSontotalrevenues(i nthelast12months)comparedwi thratioofaveragedebitbalanceat BANKSontotaldebitbalanceof CM(inthepast12months).
Relationsofexecutiveboardwit hgoverningagenciesandreleva ntministries,departments(excl udingcreditinstitutions)
Aftertransformation,weuse16quantitativevariablesand25qualitativevariables(ma kea t o t a l o f 4 1 i n d i c a t o r s i n u s e , comparet o 8 4 previousu s e d b y Earn& You ng).
Bythisstep,thequantitativevariableswerecategorized,theyarethenwereanalyzedwit hdependentvariables.Theinitialdistributionofgood/ badclientswasredistributedbydecilesandthenwasevaluatedthemostaccuraterelati veriskforallocating.Therelativerisksw e r e a l s o c o n s i d e r e d f o r e a c h categoryo f everyqualitativevariable,wheneverp o s s i b l e T h e r e a r e t w o r e a s o n s f o r establishinga n e w categorizationforqualitativeexpl anatory dependentvariables.Firstly,foravoidingsmallsizecategories(witharelativelyverysmal lnumberofobservations),whichmaycauselessrobustestimations.Secondly,forel iminatingc l o s e categoriesw h i c h s h o u l d b e combineda s o n e ?
Actually,t h e s e o b s e r v a t i o n s w i l l becapturedaswegothroughthenextstepsinb uildingmodel,whichensuretheobjectivenessoftheauthorinchoosingthebestqu alitativecategorization Asw e canalsoseeinnextsteps,theaccuracyofcategorizationh asbeendirectlytestedthrought h e multicollinearityo f b e i n g c a t e g o r i z e d a n d t h e significanceo f t h e m w h e n beingusedintheregression).
Owners’abilities(equity,management,experience)asassessedbycredit
No Name sign last12months)comparedwithratioof averagedebitbalanceatBANKSo n totaldebitbalanceofCM(inthepast12mont hs).
Ina b r i e f explanation,t h e signo f qualitativevariablesi s a r e s u l t o f previouso r d e r e d arrangementofthequalitativevariablestechniquesthatwerenotshownint h e paragraph.
Withaconsiderableeffort,duetothedifficultiesinthedatacollectionprocess, itwaspossibletoobtain1133observations,andthevariablesofinterestwereclassifie din12areas:(i)C a p i t a l resource( 2 ),(ii)HumanResource(5),
(iii)P erf o r man c e E ff ic ie nc y( 6 ),(iv)Leverage(1),(v)Margin( 2),
(vii)Payable( 4 ),(viii)Jurisdiction(1 ),( i x ) Organizationand procedures( 5 ), ( x ) Planning(1),(xi)Historicalrelationshipwithbanks(5),
(xii)Competitivefactors(5 ).Withinthe12interestedareas,wefocusonthe6ones:Resources,Efficiency,Liquidity,Organization,HistoricalrelationshipwithbanksandC ompetitivefactors.Itisimportanttonotethatthesearethemostimportantforacompany tosurvivea n d havegoodpotentialpaymentability;wehavetriedtofindanyargume ntsforthistargetb y r u n n i n g t h e modelsa n d observet h e r e s u l t i n C h a p t e r 4
Descriptivestatistics
DependentVariables Obs Mean Std.Dev Min Max
IndependentVariables Obs Mean Std.Dev Min Max
Capitalresource Equity 1133 6.78bils 6.32bils 10mils 122bils
Totalassets 1133 17.8bils 13.2bils 26mils 122bils
Netrevenues 1133 25.7bils 21.7bils 10mils 189bils
Estimated annual ROE based on ROE 1133 2.613 0.892 1 5 accumulatedfrombeginningofyeartothetimeofev aluation
IndependentVariables Obs Mean Std.Dev Min Max
Analysisor ecentfisc fcashflowa lyear statementofthemost 1133 2.860 0.501 1 3
Owners’ abilities (equity, management, 1133 1.811 0.392 1 2 experience)asassessedbycreditofficer
Organization and Book-keeping 1133 2.770 0.444 1 3 procedures Organizationofdepartments,divisions 1133 2.643 0.609 1 3
Planning Objectives,businessplaninnext1to3years 1133 2.783 0.431 1 3
IndependentVariables Obs Mean Std.Dev Min Max relationshipwith “BANKS”(includingprinc banks withinthepast12months ipalandinterest)
Ratioofsavingdeposits at“BANKS”on 1133 3.006 1.665 1 5 averageoutstandingbalanceint h e past1 2 months
Ratioofrevenuestransferredt h r o u g h “BANKS” ontotalrevenues(inthelast1 2 months)
Time period of credit relationship with“BANKS”
Relationsofexecutiveboardwithgoverningagencie sandrelevantministries,departments(excludingcre ditinstitutions)
Regressionresults
Int h i s c h a p t e r , sixmodelsa r e establishedusingthreed i f f e r e n t approachesw i t h t hreet y p e s o f D e p e n d e n t variabled e s c r i b e d i n C h a p t e r 3 , w i t h t h e p u r p o s e t o comparet h e i r stability,performance,e f f i c i e n c y a n d a c c u r a c y i n d istinguishcustomers.
To understand the relationship between the dependent variable (default status) and independent variables, we compiled a table illustrating these connections For instance, in the case of payable medium and long-term principal repayment capacity, our analysis revealed that among non-default samples, approximately 66% received the highest ratings from credit officers (indicating a repayment capacity greater than 1.5 times), while only 1.07% were rated the lowest (with a repayment capacity of less than 0.5 times) This stark contrast highlights the significant correlation between credit officer ratings—a qualitative factor—and the default probability of clients.
Default01 and Payable - medium and long term principal repayment capacity.
Werunafullmodelforbothcases,thefirstcasecalledModel1.1a,whereX1toX18 arequantitativevariablesand X19-X44arequalitative variables but wereusedlikequantitativeonesandthesecondcasecalledModel1.1b,whereX1toX18 arequantitativevariablesandX19-
Numberofdebtrestructuringoroverduedebtat“BANKS”(including principalandinterest)withinthepast12monthsRatioofrevenuestrans ferredt h r o u g h “BANKS”ont o t a l revenues(inthelast12month s)comparedwithratioofaveragedebitbalanceat“BANKS”ontotalde bitbalanceofCM(inthepast12months).
Numberofdebtrestructuringoroverduedebtat“BANKS”(including principalandinterest)withinthepast12monthsRatioofrevenuestrans ferredt h r o u g h “BANKS”ont o t a l revenues(inthelast12month s)comparedwithratioofaveragedebitbalanceat“BANKS”ontotalde bitbalanceofCM(inthepast12months).
Thelikelihoodratiochi-squareof52.35withap- valueof0.0001showsusthatourmodelasawholesignificantlyfits.
All coefficients are significant, with 7 out of 10 expected signs aligning with the model's results The remaining 10 variables are categorized into six main areas: Leverage, Margin, Payable, Competitive Factors, Organization and Procedures, and Historical Relationships with Banks This classification is highly logical.
For eachunitincreaseinGross profit/Netrevenues,thelogoddsofdefault(versusn o n- d ef a u l t ) decreasesby5.71;
Fore a c h p o i n t i n c r e a s e i n M e d i u m a n d l o n g termp r i n c i p a l repaymentc a p a c i t y (givenbythecreditofficer),thelogoddsofdefault(versusnon- default)decreasesb y 0.27;
Netrevenue,thelogoddsofdefault(versusnon-default)increasesby9.44;
Fore a c h p o i n t i nc rease i n Relations o f executiveb o a r d w i t h governingagencie sa n d relevantministries,departments(excludingcreditinstitutions)asassessed byc r e d i t officer,thelogoddsofdefault(versusnon-default)increasesby0.43;
Fore a c h p o i n t i n c r e a s e i n E s t a b l i s h m e n t o f o p e r a t i o n a l p r o c e s s e s a n d i n t e r n a l controlprocesses(givenbythecreditofficer),thelogoddsof default(versusnon-default)decreasesby0.24;
For eachpointincreaseinNumberofdebtrestructuringoroverduedebtatBANKS(includin g principalandinterest)withinthepast12months,thelogoddsofdefault(versusnon- default)decreasesby0.99;
ForeachpointincreaseinRatioofrevenuestransferredthroughBANKSontotalr evenues(inthelast12months)comparedwithratioofaveragedebitbalanceatBAN KSontotaldebitbalance,thelogoddsofdefault(versusnon- default)decreases by0.11;
ForeachpointincreaseinStatusofdebtstoothercreditinstitutionsinthelast12m onths,thelogoddsofdefault(versusnon-default)decreasesby0.62;
ForeachpointincreaseinStabilityofinputswithmainimpactsonbusinesssectorofan enterprise(givenbythecreditofficer),the logoddsofdefault(versusnon- d ef a u l t ) increasesby0.41;
Theconstant(intercept)meansthateveninbestcase,thereisalwaysachancethataclient becomesdefault.That’scompletelylogical.
Foranotherviewonclientgrading,bankscanchoosetocalculatenumericalnumbers/ ratiosfromthebusinessstatementand/ orfinancialstatementthendirectlyinputtheminthemodelforclientclassification.The bankscanalsorateclientsbypointsbasedonafixedrangenominatedforclientcharact eristics,whichiseasytoapplya n d a d d morevaluesf o r c a l c u l a t i n g t h e p r o b a b i l i t y o f p r e d i c t i n g c l i e n t default.
Allt h e c o e f f i c i e n t s a r e significant,6 / 7 e x p e c t e d signsa r e c o r r e c t followin gt h e model’sr e s u l t T h e 7 remainvariablesa r e classifiedi n 6 maina r e a s :L e v e r a g e , Margin,PerformanceEfficiency,Payable,Organizationandprocedures,Histor icalrelationshipwithbanks.Thattotallymakessense.
Test the discriminate power in classifying Independent variable: Book-keeping
All1133observationsinourdatasetwereusedintheanalysis;thelikelihoodratiochi- squareof59.62withap-valueof0.0001tellsusthatourmodelasawholefits,significantly.
Actually,tocometoanewselectionandtransformationoftheRepaymentCapacityvariable( mediumandlongtermprincipal repaymentc a p a c i t y ) a n d O r g a n i z a t i o n variabl e(Book- keeping),aquiteheavytaskhasbeendonetocheckiftheclassificationoftheclient sintogroupsisaccurate.FollowingisthecaseofBook-keeping:
Accordingtoth e table, f o r the f ir st proposedrankingsolutionforB oo k- ke ep in g, the3 rdand the2nds h o u l dbecombinedastheyarenotstatisticallydifferent Thatcan makea n o t h e r b e t t e r variablew h e n w e t r a n s f o r m X 2 8 i n t o X 2 8 n (i.e.X 2 8 :B o o k - k e e p i n g i s r a n k e d a s ( 1 ) N o t C l e a r , ( 2 ) Average,
( 2 ) C l e a r T h a t a l s o c a n makea b e t t e r variablew h e n weconsidera b o u t m u l t i c o l l o n i a n i s m w h i c h i s introducedi n t h e n e x t s e s s i o n s T h i s process wa sdoneforallqualitativevariables.
For eachunitincreaseinGross profit/Netrevenues,thelogoddsofdefault(versusn o n- d ef a u l t ) decreasesby6.99;
Netrevenue,thelogoddsofdefault(versusnon-default)increasesby14.11;
ForeachunitincreaseinReturnontotalassets,thelogoddsofdefault(versusnon- default)decreasesby6.10;
WhenM e d i u m a n d longtermp r i n c i p a l repaymentc a p a c i t y (givenb y t h e c r e d i t o f f i c e r ) turnsfrom0.5time.This isextremelylogicandmeaningful.
Statusofdebtstoothercreditinstitutionsint h e last12months,whentheStatusofdebtstoot hercreditinstitutionsinthelast12monthsisatBaddebtever,thelogoddsofdefaultis0.41,decreaseto0.09whenthevariable’svaluemovetoBaddebtbutnotnow,andcontinue todecreaseto0.07w h en thevariable’svaluemovetoNoBaddebtever.Thatisextremel ylogicandmeaningful.
Allo t h e r variablesa t means,f o r X 2 8 - Book-keeping,w h e n t h e B o o k - k e e p i n g situationisatUnclear,thelogoddsofdefaultis0.38,decreaseto0.07whenthe variable’svaluemovetoClear.Thatisextremelylogicandmeaningful.
Baseonthismethod,wecansuggestbuildinganautomatictooltocalculateanyPDf o r agiven client.Letuseaclientwithfollowinginformation:
Runningt h e model/ command(margin,at(x21=0.567x14=0.196x15=0.039x17=0.045x19n=2x 2 8 n = 2 x
N G U Y E N T H I A’sf u t u r e l o a n i s 3.6%.Andt h e a c t u a l l o a n ’ s statusi s n o t d e f a u l t Marginc o m m a n d i n thisc a s e c a n s h o w o u t t h e a b i l i t y t o become ausefulITtool.
Fortheorderedlogistic,the authortrytolookatfactorsthatinfluencethefutur elo an group(1to5).Forthispurpose,dependentF2-
8 1 6 % o f group1,9.2%o f group2,2.1%ofgroup3,1.6%ofgroup4and5.5%ofgroup5.
Werunafullmodelforboth2cases,thefirstcases(calledModel2.1a,whereX1to X1 8a re quantitativevariablesandX 19 -
X 44 are qualitativevariablesb ut were u s e d a s quantitativeo n e s ) a n d t h e s e c o n d case( c a l l e d M o d e l 2.1b,w h e r e X 1 t o X18arequantitativeandX19-
Numberofdebtrestructuringoroverduedebtat“BANKS”(including principalandinterest)withinthepast12monthsRatioofrevenuestrans ferredt h r o u g h “BANKS”ont o t a l revenues(inthelast12month s)comparedwithratioofaveragedebitbalanceat“BANKS”ontotalde bitbalanceofCM(inthepast12months).
Numberofdebtrestructuringoroverduedebtat“BANKS”(including principalandinterest)withinthepast12months
Ratioofrevenuestransferredt h r o u g h “BANKS”ont o t a l reven ues(inthelast12months)comparedwithratioofaveragedebitbalance at“BANKS”ontotaldebitbalanceofCM(inthepast12months).
Theinterestingobservationcanbewithdrawnatthispointisthat,comparingwitht h ebinominal models,thefollowing7independent variables(which are included in5 groups)stillhavesignificantpowerinexplainingthe creditrating:
Margin,Retu rn ontotalassets-PerformanceEfficiency,
Numberofdebtrestructuringoroverduedebtat“BANKS”(includingprincipalandinterest) withinthepast12months,R a t i o ofrevenuestransferredthrough“BANKS”ontotalreven ues(in thelast12months)comparedwithratio ofaveraged e b i t balanceat“BANKS”ontotaldebitbalanceofCM(inthepast12 months)-Historicalrelationshipwithbanks.
The4 u s e d independentvariablesi n previouslogitm o d e l i n g w h i c h a r e n o w n o t si gnificantforologitare:
Relations ofexecutiveboardwith governingagenciesandrelevantministries,departments(excludingcreditinstitutions)-
Book- keeping,Establishmentofoperationalprocessesandinternalcontrolp r o c e s s e s , Sta bilityofinputswithmainimpactsonbusinesssectorofanenterprise
Historicalr ela t io n sh i p withbanks.
Forthedetailinterpretationofthemodels,wefindthatqualitativemodelismoreacc urateinintroducingf o r ologitmodelintheresearch.Forthatreason, thefollowingwillquicklystatisticallyinterpretModel2.2anddetailsinModel2.3.
Allt h e c o e f f i c i e n t s a r e significant,5 / 6 e x p e c t e d signsa r e c o r r e c t f o l l o w i n g themodel’sr e s u l t T h e 6 remainvariablesa r e classifiedi n 5 maina r e a s :L e v e r a g e , Margin,Payable,PerformanceEfficiency,Historicalrelationshipwithba nks.T h a t trulymakessense.
Thelikelihoodratiochi-square of57.81witha p- valueof0.0001showsthato ur modelasawholefitssignificantly.
Fromthet a b l e , w e s ee the coefficients, thestandarderrors,thez- statistic,asso ciated p- values,a n d t h e 9 5 % c o n f i d e n c e intervalo f t h e c o e f f i c i e n t s Allofthemares tatisticallysignificantat90%confidence.
Thecutpointsshownatthebottomoftheoutputindicatewherethelatentvariableiscutto makethegivegroupsthatweobserveinourdata.Ingeneral,thesearenotusedi n t h e i n t e r p r e t a t i o n o f t h e r e s u l t s T h e c u t p o i n t s a r e c l o s e l y r e l a t e d t o thresholds,whicharereportedb y otherstatistical packages.F o r Hamilton,2 0 0 6 , p 2 7 9 , ologitestimateascore(Z)ofXvector:Z=-1.35X12–4.18X14+11.14X15
6/7expectedsignsarecorrectfollowingthemodel’sresult.Allthecoefficientsaresigni ficant,6 / 7 e x p e c t e d signsa r e c o r r e c t f o l l o w i n g themodel’sr e s u l t T h e 7 remainvariablesareclassifiedin6mainareas:Leverage,Margin,Payable,Perfor manceEfficiency,Organizationandprocedures,Historicalrelationshipwithban ks.Thatextremelymakessense.
All1133observationsinourdatasetwereusedintheanalysis,thelikelihoodratiochi- squareof75.9withap- valueof0.0001showsthatourmodelasawholefitssignificantly.Int h e t a b l e w e s e e t h e coefficients,theirs t a n d a r d e r r o r s , t h e z - statistic,associatedp- values,andthe95%confidenceintervalofthecoefficients.Allofthemarestatisticallysi gnificantat95%confidence.
(i.e.,goingfrom0to1),weexpecta1.3decreaseinthelogoddsofbeinginahigherlo angroup, givenalloftheothervariablesinthemodelareheldconstant.
Netrevenues(i.e.,goingfrom0to1),weexpecta4.03decreaseinthelogoddsofbe inginahigherloangroup,givenalloftheothervariablesinthemodelareheldconstant.
Netrevenue(i.e.,goingfrom0to1),we expecta11.31increaseinthelogoddsofbeinginahig herloangroup,givenallo f theothervariablesinthemodelareheldconstant.
Whenthereturnincreaseoneunit,thelogoddsofbeinginahigherloangroupisexpe ctedtodecreaseby8.05givenalloftheothervariablesinthemodelareheldconstant.Whent h e M e d i u m a n d longtermp r i n c i p a l repaymentc a p a c i t y , w h e n t h e re paymentcapacitychangesfrom0.5time,thelogoddsofbeingina higherloa ngroupisexpectedtodecreaseby2.1givenalloftheothervariablesinthemodelareheldcon stant.
ForX28-Book- keepinga s s e s s e d b y c r e d i t o f f i c e r , w e w o u l d s a y t h a t w h e n t h e B o o k - k e ep i n g turnsfromUncleartoClear,weexpecta1.9decreaseinthelogoddsofbeinginahi gherloangroup,givenalloft h e othervariablesinthemodelareheldconstant.
RatioofrevenuestransferredthroughBANKSontotalrevenues(inthel a s t 12mo nths)comparedwithratioofaveragedebitbalanceatBANKSontotald e b i t balanc e,wewouldsaythatwhentheRatioturnsfromlessthan30%to51-
70%,weexpecta0.77decrease inthelogoddsofbeinginahigherloangroup,gi venalloftheothervariablesinthemodelareheldconstant.WhentheRatioturnsf r o m lessth an30%to>70%,weexpecta0.32decreaseinthelogoddsofbeinginahigherloangroup,gi venalloftheothervariablesinthemodelareheldconstant.
Thecutpointsshownatthebottomoftheoutputindicatewherethelatentvariableiscutto makethegivegroupsthatweobserveinourdata.Ingeneral,thesearenotusedi n t h e i n t e r p r e t a t i o n o f t h e r e s u l t s T h e c u t p o i n t s a r e c l o s e l y r e l a t e d t o thresholds,whicharereportedbyotherstatisticalpackages.ForHamilton,2006,p 2 7 9 , ologitestimateascore(Z)ofXvector:Z=-1.3X12–4.03X14+11.31X15
Theb e n c h m a r k f or makingclass if ic at io n playsa n importantr o l e i n t h e model’s r e s u l t s ofpredicting.
BelowIusedthemarginscommandtoestimatethepredictedloangroupofclientate a c h leve lof Mediuma n d longtermp r i n c i p a l repaymentcapacityandBook- keeping,holdingallothervariablesattheirmeans.
Margin,at(x19n=(1)) predict(outcome()) Margin P
Thatisanextremelyhelpfultoolforthecreditofficertohaveagoodimageonthee x p e c t e d futureloangroupoftheclient.GiventhesetofclientatMediumandlongtermprincipal repaymentcapacity =1(meanst h i s setofclienthasthe repaymentcapacity 360days group5,lessthan180days group2,etc…),wefindthat,byhazard ,bothmodelshavepredictedcorrectly925cases,correspondingto8 1 6 4 % which canpredict thenon-defaultclients,while0% of default client (loansfromgroup3-5)ispredicted.
Mainfindings
Theestimationo f PDo r e x pe c t e d f u t u r e l oa ngroupc r e a t e mainactivitiesint h e r i s k management, incorrectestimationofthePDorfutureloangroupofclients leadt o t h e w r o n g e s t i m a t i n g o f riska n d return.N o w a d a y s , t h e evaluatio no f c r e d i t w o r t h i n e s s issupportedbymanystatisticalquantitativetools.The study showsthatbankscansavetimeandcostandmitigatetheriskwithinthewholecreditis suingprocess b y a goodmodelingo f c r e d i t r a t i n g basedo n t r a d i t i o n a l s t a t i s t i c methodologies:logit,linearandologitwithasortedlistofcollectedvariables.
The paper confirms the effectiveness of classical credit rating models in the Vietnamese market, with the logit model achieving over 84% accuracy in predicting good clients and 35% for bad clients by utilizing both quantitative and qualitative variables Linear regression also demonstrates strong predictive capabilities for late payment days, forecasting around 82% of good clients for future loan groups, although it shows limitations in predicting defaults The logit model further supports future loan downgrading predictions, with an accuracy of approximately 81% for future loan groups Additionally, models have been developed specifically for SMEs, addressing their challenges in obtaining credit from banks Overall, the research successfully identifies applicable models for modern banking management using traditional methodologies, which can be integrated into automated credit rating tools for loan issuance and review.
The study reveals that only specific client information is crucial for assessing creditworthiness, allowing for a streamlined list that reduces costs and saves time Consequently, banks can avoid the need to expand their data warehouses or manage large teams for data collection and analysis The research indicates that nearly all financial and non-financial characteristics of clients can be captured by selecting relevant indicators Specifically, only 13 indicators across 8 areas are necessary for credit rating, and fortunately, most of these indicators are applicable for modeling all three measurements of default probability.
(2) Grossprofit/Netrevenues,(3)Netoperatingprofit/Netrevenue– inMarginarea
(6) Relationsofexecutiveboardwithgoverningagenciesandrelevantministries,depar tments( e x c l u d i n g c r e d i t i n s t i t u t i o n s ) , a n d ( 7 ) Stabilityo f i n p u t s w i t h mai nimpactsonbusinesssectorofanenterprise–inCompetitive factorsarea,
(10) Objectives,businessplaninnext1to3years–inPlanningarea
(11) NumberofdebtrestructuringoroverduedebtatBANK(includingprincipa la n d interest)withinthepast12months,
(12)RatioofrevenuestransferredthroughBANKo nt o t a l revenues(i n t h e la s t 12 m onths)comparedw i t h r a t i o of averaged e b i t balanceatBANKontotaldebitbalanc eofCM(inthepast12months),
(13)Statuso f d e b t s t o o t h e r c r e d i t i n s t i t u t i o n s int h e l a s t 1 2 month s,-Historicalrel a ti o n sh ip withbankarea
Thepurposeo f findingthemostsignificantexplanatoryv a r i a b l e s hasbeenaddre ssed.Theresultshowsanenoughlargeareaoffactorsthatcouldbechosenwhenb u i l d i n g e ac h model,thankst o t h e f a c t t h a t a l l t h o s e i n d i c a t o r s havet h e i r ownval uest o e a c h modelevenb e e n u s e d i n d i f f e r e n t manners.Bankingmanagement’sc hallengei s n o t o n l y t o e s t a b l i s h i n t e r n a t i o n a l s t a n d a r d i n t e r n a l regulations,butalsotobuildtheirbusinessprocedurestomaximizeprofits.
Thirdly,thepaperaddressestheimportantrelationshipbetweentheoreticalmo delbuilding,choosingmodelc u t - o f f p o i n t a n d p r a c t i c a l i m p l e m e n t i n g riskmanagementsol ut io ns Infac t, t h e ban ks ca n ch ose t h ew a y h ow t h e y id en ti fy a goodclientorbadclientbyjustasim pleinputinthemodel.
Fourthly,thestudyalsosuggestsmanywaystogiveclientsanappropriateap p r o a c h f o r cr e di t riskidentification,whichistheproxy ofdatatreatmentand dataanalysis’sresults.Themodeliseasytomaintain(forstatisticalexperts).Arema rkablecharacteristicofthesystemisthefacilitytoadjustfornewestimationso fthem odelthatmayvaryalongtimeassocialandeconomicconditionschange.T h i s allows bankstoquicklyadapttheirriskmanagementpoliciesandproceduresto market’scirc umstance.Thepaperhasalsoillustratedtheproceduresthatshouldb e implementedbya bankinordertoestablishthemostfittingmodels,withdetailtreatmentsofdatathatwered etailsdescribed.
Limitationsofthestudy
First,thepaperhasbeenbuilt forVietnamesemarketwherethesixmodelshavepr esentedsuitableresults.However,itcancauseinvaliditywhenbeingappliedforo t h e r m a r k e t s w i t h i t s o w n e c o n o m i c - s o c i a l environment.A l a r g e r s t u d y w h i c h coversothercountriesfactorscanhe lptodeveloptheresearchtoabroadermethodology/literature.
Thepaper,asinitsfirstnatureistoanswerforSMEsclient,shouldnotbeimpliedforoth erclientsegments.Abroaderclientdatacanhelptofindoutanycommonpointsin applyingcreditratingforthegreaterclients’diversity.
The methods presented in this paper are straightforward combinations of two established transformation methodologies that yield superior results More importantly, these approaches empower scorecard developers to create scorecards that are easier to interpret, facilitate clearer explanations for end-users, and offer deeper insights into the portfolios they aim to clarify.
Presentedstatisticalmodelscanshowthestatisticalcontribution ofeachvariable. However,wecanchoseothernon- statisticaloptionslikeneuralnetworks,decisiontrees,geneticalgorithms,ex pe rt syst ems withtheirowncharacteristics inusing,example,theycanbewellinthein-sampleexplaining,butnotwellinout-samplepredicting.
Implications
Thel o a n p r o c e s s i n g h a s r a p i d l y i n c r e a s e d ins p e e d , thankst o scoringsy stems.R a t h e r t h a n p e r f o r m lengthyc r e d i t i n v e s t i g a t i o n , c r e d i t o r s a n d o t h e r l e n d e r s a r e abletoaccesscreditratingt o determinecreditrisks.Moreover,ma nyoftheexplanatorydependentvariablesstatisticallyshow ahighpower inpredictingcreditw o r t h i n e s s includingtheCapitalresource,HumanResource,
PerformanceEfficiency,Leverage,Margin,Liquidity,Payable,Organizationandp rocedures,Planning,Historicalr e l a t i o n s h i p withbanks.Bankss h o u l d f o c u s o n thisareaandsuggestedindicatorsforbuildingtheirowncreditratingmodels.
Ther e s u l t s o f t h i s a n a l y s i s i n d i c a t e t h a t betterc r e d i t r a t i n g modelsc a n b e b u i l t usingpiecewiselogisticregressionrather thanthesameregressionusingasi nglevariablep e r c h a r a c t e r i s t i c o r a t t r i b u t e Atthesametime,t h e m o d e l s a r e morer o b u s t , beingabletobetterhandlesachangingriskenvironment.
Ononehand,wemusttakeintoaccountthattheremustbeacontinuousreviewingo f t h e evolutiono f t h e p o r t f o l i o , a n a l y z i n g t h e behaviorandi t s changes.T h e p e r c e n t a g e o f defaultersshould befrequentlysupervisedtosee whether thepoliciesofgrantingmustbealteredsothattheglobalriskacquiredbytheinstitutioni snotincreased.
We can design a new autonomous system for evaluating applications based on the selected classification model This software maintains the final score calculation as a black box, ensuring that neither the credit manager nor the applicant knows the specific weights of the various information items and their contributions to the final result This approach minimizes manipulation and prevents biased scores Currently, the IT systems that support each service require review and extension.
To remain competitive, banks must implement future-proof technology that integrates seamlessly with existing legacy systems, preserving vital components that underpin their operations These systems, which handle billions of transactions daily, require rationalization and documentation to enhance their value without necessitating complete replacements By identifying and reusing core services, banks can streamline system development, significantly reducing construction and maintenance costs Additionally, leveraging IT technology enables banks to gather better information and generate insights for model improvements, ultimately saving clients time and costs by allowing them to provide minimal information and receive quick responses regarding loan eligibility.
Knowingtheprobabilityof default,potentialfutureloansgroupisalsoknowingtheexpectedrisk,banksc a n dynamic allyc h o s e t h e i r c l i e n t segments( w i t h s p e c i f i c a c c e p t e d levelofgoodness, pricing- riskandreturnbalancing,methodologytoriskmitigateandhelptheclienttoavoidbadsit uations,specificpositive/negativecovenantsbyeachclients…).
Ift h e b a n k h a s p r e f e r e n c e t o f o c u s o n predictiono f d e f a u l t ( f o r conservativebusiness),logitmodelssho uldb e th e m o s t acc ur ate option.F o r l i n e a r an dologitmodelsw h e r e c l i e n t s a r e graded(ons p e c i f i c range:1-2o r 1 -
6 ) , i t showmorepo wer inidentifygoodcustomers.Anessentialpointinlogitmode listhecut-offvalue,whichdirectlyimpacttheconceptofbankrelatingtobad/ goodclients.Thebanks’purposesandcorrespondingpoliciesdecidethechosen ofthebusiness;theya l s o influencethechoiceoftheirowncreditratingmodel.
WecanusePDforBCBS'scap ita la deq uac yr at io (CAR)calc ul at in g andcap it a lmanagement,decidetoacceptorrefusetheapplication.
Wec a n u s e e x p e c t e d l o a n groupf o r a flexiblec r e d i t g r a n t i n g p o l i c y , f o r l o a n dynamicp r o v i s i o n i n g p o l i c y a n d a l s o u s e theologitmodelf o r b e t t e r p r e d i c t i n g downgrading andbeingactiveinrisk-returnmanagement
Suggestionforfurtherstudies
Thisisnottosaythattheseapproachesshouldbeusedforeveryscorecarddevelopmentg oingforward,butthatshouldatleastbepartoftheavailabletools.T h e studydidn ot,either,aimatamoredetailedapproachofthefocusedtechniques.
The research acknowledges existing limitations that can be addressed through a more diverse methodological approach and enhanced data management Future steps include utilizing the Multi Discriminant Analysis (MDA) method and updating current Z-score functions used in banks Implementing neural networks and genetic algorithms may better mitigate the paper's limitations Additionally, there is a delay between collecting relevant information and bankruptcy dates, which restricts the predictive models to one-period predictions To improve the study, it is essential to extend the research beyond SMEs, consider various loan types (long/short, with collateral/non-collateral), and narrow the focus by industry to identify relevant models for different client types Furthermore, incorporating macroeconomic and industrial factors as independent variables will better reflect the business environment in the predictive models.
Letmefinish withacommentfromacreditprofessionalabouttheroleofstatisticalmodelsincreditratin g:“Althoughcreditriskassessmentisoneofthemostsuccessfulapplicationsofappliedstat istics,thebeststatisticalmodeldoesn’tpromisec r e d i t r a t i n g success,itdependso ntheexperiencedriskmanagementp r a c t i c e s , thewaymodelsaredevelopedandappli ed,andproperuseofthemanagementinformationsystems”(Mays1998).“Andatthesa metime,theselectionoftheindependentvariablesisveryimportantinthemodeldevelo pmentp h a s e becausetheydeterminetheattributesthatdecidethevalueofthecreditscore,a ndthevaluesoftheindependentvariablesarenormallycollectedfromtheapplicationf orm.Itissignificanttoidentifywhichvariableswillbeselectedandincludedint hefinalscoringmodels.”
Abdou,H.,Pointon,J.2009.Creditscoringanddecision- makinginEgyptianpublicsectorbanks.InternationalJournalofManageri alFinance5
( 2 0 1 1 ) C r e d i t scoring,s t a t i s t i c a l t e c h n i q u e s a n d evalua tioncriteria:areviewoftheliterature.IntelligentSystemsinA c c o u n t i n g , Finance&Management
Altman,E.I.(2000).Predictingfinancialdistressofcompanies:revisitingtheZ - s c o r e andZETAmodels.SternSchoolofBusiness,NewYorkUniversity,9- 12.
(2002).TheUseofCreditratingModelsandtheImportanceofac r e d i t cultu re.linha].NewYork.[15-09-2011]Disponívelemh tt p: // p eo ple stern.nyu.edu/ealtman/Presentations.htm.
(2007).ModellingcreditriskforSMEs:Evidencef r o m theUSmarket.Aba cus,43(3),332-357.
( 2 0 1 3 ) SampleSelectionB i a s i n AcquisitionC r e d i t ScoringM o d e l s : AnEvaluationo f t h e Supplemental-
Desai,V.S.,Crook,J.N.,Overstreet,Jr., G.A.,1996.Acomparisonofneuraln e t w o r k s andlinearscoringmodelsinthec r e d i t unionenvironment.Europ eanJ.Oper.
Gordy,M B ( 2 0 0 3 ) A ri sk -f ac to r modelf o u n d a t i o n f o r r a t i n g s - b a s e d bankcapi t al rules.Journaloffinancialintermediation,12(3),199-232.
(2007,May).Creditriskanalysisapplyinglogisticregression,n e u r a l networ ksa n d g e n e t i c algorithmsmodels.InProccedingso f t h e POMS1 8 t h A n n u a l C o n f er e n ce , D a l l a s , TX,USA(Vol.47,p.47).
( 2 0 0 4 ) C r e d i t R i s k Modelingo f M i d d l e Markets.NewYork:Z ic k l in SchoolofBusiness,BaruchCollege,CUNY.
Longenecker,JustinG.,CarlosW.Moore,and J.William Petty. (1997).Creditscoringa n d t h e smallb u s i n e s s : A reviewa n d t h e n e e d f o r r e s e a r c h A p a p erpresentedinUSASBE1997NationalConf erence, Sa nFrancisco,California.
VIETNAM–NETHERLANDSPROGRAMMEFORM.AINDEVELOPMENTECONOMICS ultyo f E c o n o m i c s andB usin ess AdministrationWestUniversityofTimisoa ra,Romania.
( 2 0 1 2 ) T h e u s e o f d i f f e r e n t a p p r o a c h e s f o r c r e d i t r a t i n g pre dictiona n d t h e i r comparison.In6thInternationalS c i e n t i f i c Conf erenceM a n a g i n g a n d M o d e l l i n g o f FinancialR i s k s (pp.4 4 8 -
( 2 0 1 1 ) H o w t o measuret h e q u a l i t y o f c r e d i t r a t i n g models.Financ ea úvěr: Czech Journal ofEconomicsandFinance,61(5),4 8 6 - 5 0 7
(2012) Designanddevelopmentofc r e d i t ratingmodelforthecommercial banksofPakistan:Forecastingcreditworthinessofindividualborrowers.Intern ationalJ o u r n a l o f B u s i n e s s a n d
( 2 0 0 3 ) : T h e u s e o f discriminanta n a l y s i s , logisticregressionand classificationtreeanalysisinthe developmentofclassificationmodelsfo rhumanhealtheffects.Theochem
(1980)ANoteontheComparisonofLogitandDiscriminantModelsof ConsumerCreditBehavior.TheJ o u r n a l ofFinancialandQuantitativeAnal ysis.
VIETNAM–NETHERLANDSPROGRAMMEFORM.A EVELOPMENTECONOMICS Page97of102
1 Credit :acontractualagreementinwhichaborrowerreceivessomethingofval uenowandagreestorepaythelenderatsomedateinthefuture,generallywithint erest.Thetermalsoreferstotheborrowingcapacityofanindividualorcompany (http://www investopedia.com/terms/c/credit.asp )
2 CreditRisksinclude3typeofrisk :DefaultRisk,CreditSpreadRisk,and
DowngradeRisk.Inthispaper,wewillexaminetheDefaultrisk,whichism orerelevantincreditissuingareaandisusedforbankinclientratingandcrediti ssuingp r o c e s s TheCreditSpreadriska n d Downgraderiska r e almostbeu sedinbondmarket.
Defaultrisk :Thefailuretopromptlypayinterestorprincipalwhendue.D ef au lt occurs whenadebtorisunabletomeetthelegalobligationofdebt repayment.Borr owersmayd e f a u l t whent h e y a r e u n a b l e t o maketherequired paymentorareunwillingtohonorthedebt.( http:// www.investopedia.com/terms/d/default2.asp)
3.1 Probabilityo f d e f a u l t ( P D ) : t h e degreeo f likelihoodt h a t t h e b orro wer ofaloanordebtwillnotbeabletomake
3.2 Thenecessaryscheduled repayments.Shouldthe borrower beu nabl et o p a y , t he ya r e thens a i d t o bein defaulto f t h e debt,a t whichpointthelendersofthedebthavelegalavenuestoattemptobt ainingatleastpartial repayment.Generally speaking,thehigher t h e d e f a u l t p r o b a b i l i t y a l e n d e r estimatesa b o r r o w e r t o have,thehighertheinterestratethelenderwillchargetheborr ower( a s compensationf o r b e a r i n g higherd e f a u l t risk).( http:// www.investopedia.com/terms/d/defaultprobability.asp)