Private999ddd.com

BASELCOMMITTEEONBANKINGSUPERVISIONWORKINGPAPERSNo.
3–August2000CREDITRATINGSANDCOMPLEMENTARYSOURCESOFCREDITQUALITYINFORMATIONbyaworkinggroupledby:ArturoEstrellaandparticipationfrom:JohnAmmerJohnCarrollWilliamCoenAntonellaFogliaCraigFurfineDavidGreelyPatrickGuerchonovitchGeorgeHancHirotakaHideshimaTorJacobsonThiloLiebigJesperLindéAndrewLoganSylvieMathératHitoshiNaganoDavidNebhutPeterNigroFrankPackerJackReidhillSusanSzarkowitzBANKFORINTERNATIONALSETTLEMENTSBasel,SwitzerlandBASELCOMMITTEEONBANKINGSUPERVISIONWORKINGPAPERSNo.
3–August2000CREDITRATINGSANDCOMPLEMENTARYSOURCESOFCREDITQUALITYINFORMATIONbyaworkinggroupledby:ArturoEstrellaAbstractBankregulationhasmadeincreasinguseofexternalcreditratingsinrecentyearsand,inJune1999,theBaselCommitteeonBankingSupervisionproposedarevisiontotheinternationalcapitalaccordthatincludedamoreprominentroleforcreditratings.
Thisstudycontainsbackgroundinformationaboutcreditratingsandcomplementarysourcesofcreditqualityinformation,whichmaybehelpfultothoserevisingtheproposedaccordandmoregenerallytothoseinterestedintheuseofcreditratingsinregulation.
Thefocusofthestudyisonfactualinformation,ratherthanonsubjectiveassessmentsorexplicitpolicyrecommendations.
Thematerialmaybeclassifiedintofourbasiccategories:(1)factsaboutthecreditratingsindustry,suchaslistsofratingagencies,theextentoftheiractivities,marketpractices,etc.
,(2)factualinformationaboutalternativesourcesofcreditqualityinformation,(3)summariesoftheresultsofearlierresearchoncreditratingsfromvarioussources,includingacademics,supervisoryinstitutions,andratingagencies,and(4)empiricalworkperformedspecificallyforthisstudy,intendedtofillinafewgapsintheexistingempiricalliterature.
TheWorkingPapersoftheBaselCommitteeonBankingSupervisioncontainanalysiscarriedoutbyexpertsoftheCommitteeoritsworkinggroups.
TheymayalsoreflectworkcarriedoutbyoneormorememberinstitutionsorbyitsSecretariat.
ThesubjectsoftheWorkingPapersareoftopicalinteresttosupervisorsandaretechnicalincharacter.
TheviewsexpressedintheWorkingPapersarethoseoftheirauthorsanddonotrepresenttheofficialviewsoftheBaselCommittee,itsmemberinstitutionsortheBIS.
Copiesofpublicationsareavailablefrom:BankforInternationalSettlementsInformation,Press&LibraryServicesCH-4002Basel,SwitzerlandFax:+4161/2809100and+4161/2808100ThispublicationisavailableontheBISwebsite(www.
bis.
org).
BankforInternationalSettlements2000.
Allrightsreserved.
Briefexcerptsmaybereproducedortranslatedprovidedthesourceisstated.
TableofContentsSummaryofProjectandFindings1Preface:TheEconomicRoleofCreditRatingAgencies11PartI.
OverviewandSurveyofCurrentInstitutions14SectionA.
RatingAgencies14SectionB.
MethodologyandRatingsDefinitions15SectionC.
MarketPractices16SectionD.
RegulatoryCertificationProcedures40PartII.
ComplementarySourcesofCreditInformation55SectionA.
RegionallyBasedCreditScoring55SectionB.
ExportCreditRatingsforSovereigns87SectionC.
NewProductsfromRatingAgencies97SectionD.
PublishedSurveys101SectionE.
MeasurementsofProbabilityofDefaultBasedonAccountingData.
.
.
.
.
107SectionF.
MeasuresBasedonMarketPrices111PartIII.
DefaultandRecoveryStudies:QuantifyingtheRisk126SectionA.
DefaultStudies126SectionB.
StabilityofTransitionMatrices132SectionC.
ConsistencyAcrossSectors153SectionD.
RatingsDifferencesAcrossAgencies1741SummaryofProjectandFindingsBankregulationhasmadeincreasinguseofexternalcreditratingsinrecentyears.
Oneofthekeyexamplesofsuchapplicationsisthepackageofrulesfordeterminingtherequiredcapitalwithrespecttomarketriskinthetradingbook,issuedbytheBaselCommitteeonBankingSupervision(BCBS)in1996.
Moregenerally,externalcreditratingsareincorporatedinfinancialregulation,withexplicitrecognitionofselectedcreditratingagenciesbysomeregulatorsinsomejurisdictions.
ThediscussionprocessthatledtotheJune1999BCBSproposalforarevisedinternationalcapitalaccordfollowedthistrend,culminatinginamoreprominentproposedroleforcreditratingsinthedeterminationofoverallcapitalforbankinginstitutions.
Akeypartofthisdiscussionprocesswastheinitiationofresearchinthefieldofcreditratings,whichisofprimaryimportanceinordertoarriveatinformeddecisions.
However,thevolumeofinformationandresearchinthisfieldissolarge,andtheprocessofrevisingthecapitalaccordissodynamic,thattheBCBSidentifiedaneedtoapproachthisresearchprobleminamoresystematicway.
Thus,inSeptember1999,theBCBS'sResearchTaskForceformedaworkinggrouptocollecttheavailableandrelevantinformationoncreditratingsinasingledocument,tobemadeavailabletoallBCBSgroupsconsideringtheincorporationoftheseratingsinthenewaccord.
Giventheprogressivetimetablealreadyinplaceforrevisionoftheaccord,theworkinggrouphasmovedveryquicklytoaddresstheneedsofvariousdraftingandreviewgroups.
Theresultisalengthydetailedstudy,whichshouldbehelpfultothoserevisingtheproposedaccord,andmoregenerallytothoseinterestedintheuseofcreditratingsinregulation.
Animportantprincipleguidingtheapproachoftheworkinggroupisthefocusonfactualinformation,ratherthansubjectiveassessmentsorexplicitpolicyrecommendations.
Insomecases,theevidencemaypointclearlytoaparticularpolicyoption,inwhichcaseitmaybehardtodistinguishbetweenfactualresultandpolicyrecommendation.
Ingeneral,however,therewillberealpolicychoices,andithasbeenthegoaloftheworkinggroupinthosecasestoprovidedirectlyrelevantbackgroundmaterialthroughthestudy.
Thefactualmaterialmaybeclassifiedintofourbasiccategories.
First,therearedirectfactsaboutthecreditratingsindustry,suchaslistsofratingagencies,theextentoftheiractivities,marketpractices,etc.
Manyofthesefactswerepreviouslyavailabletothepublic,butwerenotcollectedinasinglereferencesource.
Second,thestudycontainsfactualinformationaboutalternativesourcesofcreditqualityinformation.
Thesesourcesincludecreditregisters,exportratings,accounting-basedandmarket-basedscores,publishedsurveys,andnewratingagencyproducts.
Third,thestudysummarisestheresultsofearlierresearchoncreditratingsfromvarioussources,includingacademics,supervisoryinstitutions,andratingagencies.
Thevastmajorityofthisresearchisempirical.
Ithasfocusedonissuesthatarequiterelevanttothepurposesofthepresentstudy,includingthepredictiveperformanceofcreditratingsandotherindicators,cross-sectionaldifferencesinperformanceacrossvariousdimensions,andthetrendandcyclicalbehaviourofcreditratings.
Finally,afourthcategoryoffactualmaterialconsistsofempiricalworkperformedspecificallyforthisstudy.
Thepurposeofthisworkisprimarilyto2fillinafewgapsintheexistingempiricalliterature.
Overall,itisdifficulttospanthewholebreadthofafieldlikethisinasinglestudy,buttheworkinggrouphasattemptedtomakeawiderangeofrelevantmaterialreadilyavailabletoparticipantsinthereviewofthecapitalaccord.
Thestudycontainsthreemainparts,namely,(I)anoverviewoftheratingsindustryandofitsuseinbankingregulation,(II)asurveyofcomplementarysourcesofcreditqualityinformation,and(III)areviewofstudiesoftheperformanceofcreditratings.
Asnoted,thefocusofthestudyisonfactualinformation.
However,abriefreviewoftheconceptualissuesregardingcreditratingsispresentedintheprefacethatfollowsthissummary.
Theprefaceofferssomepreliminarythoughtsaboutapossibleframeworkforapproachingthefactualresults.
Theremainderofthesummaryreportssomeofthekeyfindingsoftheindividualsections.
SectionsI.
A.
-I.
C.
OverviewandSurveyofCurrentInstitutions:RatingAgencies,MethodologyandRatingsDefinitions,andMarketPracticesInformationonthecoverageofbondissuancebyratingagenciesisgenerallyavailabletothepublicfromawidevarietyofsources,includingtheratingagenciesthemselves.
Inaddition,informationaboutthepracticesfollowedinthemarketisalsofairlyaccessible.
However,whileitmaybeeasytofindtheanswertoaspecificnarrowquestion,thedispersionoftheinformationmakesitverydifficulttogetanaccurateglobalviewofthisimportantmarketsector.
SectionsAthroughCofpartIofthestudyattempttoprovidesuchaglobalviewtotheextentthatitispossibleinasectorthathasbecomemuchmoredynamicinrecenttimes.
SectionI.
A.
focusesonfairlydirectinformationaboutthesizeandnatureoftheoperationsofratingagencies.
Itcontainslistsofagenciesthatwereidentifiedinthestudy,withinformationaboutsize(numberofemployees,capital,revenue,ratingsassigned),ownership,andgeographicdistributionofratings.
Thelistincludesmajorglobalagenciesaswellasarangeofregionalfirmsaswell.
1Preliminaryinformationastothecoverageoftheratingagenciesconsistsprimarilyofanindicationoftheproportionofthetotalofratedissuersthatiscoveredbyeachoffourmajoragencies.
Thefiguressuggestthateventheglobalagenciestendtospecialisetosomeextent.
Forinstance,inthecoverageofindustrialcompanies,bothMoody'sInvestorServiceandStandard&Poor'scoverrelativelylargeproportionsofUSandEuropeanissuers.
However,Moody'sseemstohavethelargestshareofratingsinAsia,whileS&PhasmoreextensivecoverageinLatinAmerica.
SectionI.
B.
reviewsthemethodologiesusedbytheagencies.
Ratingschemesdominatethemodelsusedbythelargestcompanies,Moody'sandS&P,butsomeoftheregionalagenciesdeviatefromthesestandards,principallyinthedirectionofsimplification.
Mostfirmsreport1ThelargeagenciesincludeFitch,Moody's,andS&P.
Duff&Phelps,anotherlargeratingagency,mergedwithFitchduringthepreparationofthisstudy.
CountriesrepresentedinthelistofregionalagenciesincludeAustria,Canada,Germany,Japan,Malaysia,Sweden,andtheUnitedStates.
3thattheyrateriskonarelative–ratherthanabsolute–scale,andmostindicatethattheyrate"acrossthebusinesscycle,"suggestingthatratingsshouldinprinciplenotbesignificantlyaffectedbypurelycyclicalinfluences.
SectionI.
C.
containsabriefreviewofselectedmarketpractices,primarilyintheareasoffeesandunsolicitedratings.
Withregardtofees,thereisacleardistinctionbetweenthelargerglobalfirmsandtheregionalfirmsastothedirectsourceoffeeincome.
Largeagenciescustomarilychargeafeetotheratedentityforissuingarating.
Regionalagencies,incontrast,predominantlyobtaintheirfeeincomefromsubscriberstotheirratinginformation.
Thisdifferenceprobablyarisesfromavarietyoffactors,suchasglobalreputation,regulatorycertification,andgeneralavailabilityofratinginformation.
Sizeappearstohavenoroleindeterminingwhichagenciesissueunsolicitedratingsandwhichdonot.
Amongsomeofthelargerfirms,Moody'sandFitchissueunsolicitedratings.
Duff&Phelps,beforeitsmergerwithFitch,didnot.
2SectionI.
D.
RegulatoryCertificationProceduresThissectionreviewstheregulatoryuseofexternalcreditratings,particularlybybankingregulators,intheG10countriesaswellasinafewselectednon-G10countries.
Someclearpatternsemergefromtheanalysis.
First,virtuallyallthecountriesexaminedusecreditratingsinfinancialregulation.
Second,amongtheG10countries,theprimaryuseinbankingregulationisinconnectionwiththe1996Baselrulesforcapitalwithregardtomarketrisk.
AnextensiontothisusageoccursinSwitzerland,inwhichsovereignratingsareusedtosupplementtheOECD/non-OECDdistinctionintheBaselAccord.
ThecriteriausedforrecognitionofratingagenciesaregenerallyconsistentinspiritwiththeJune1999BCBSproposal,thoughtheprecisewordingofthecriteriadiffersfromcountrytocountry.
Thecountriessurveyedtendtorecognisethelargeglobalratingagencies,withveryfewexceptions.
Notsurprisingly,theregionalagenciesaremorelikelytoberecognisedbyauthoritiesintheirlocalregions.
AcounterexampleisthatofaJapaneseratingagencythatisrecognisedbyregulatorsinthreecountries,butisnotofficiallyrecognisedinJapan.
Lackofofficialrecognitiondoesnotnecessarilysignifyanegativeviewoftheagency,whichhasamarketfollowing,butthephenomenonseemsinterestingfromthepointofviewofthedesignofinternationalrulesforrecognitionofagencies.
Thecountriessurveyedgenerallyhaveproceduresforreviewingtherecognitionofindividualagencies,andthesectionexaminesrecentchanges.
Primarilybecauseregulatoryrecognitionisarelativelyrecentphenomenon,allreportedchangesinvolveadditionsofnewagencies,ratherthande-recognitionofexistingchoices.
2S&Pissues"pi"ratings,whichareineffectunsolicited,butareexplicitlylabelledasbeingderivedfrompublicinformationonly.
4SectionII.
A.
RegionallyBasedCreditScoringThemainfocusofthissectionofthestudyisonentitiesgenerallyknownas"creditregisters,"whichhavebeensuggestedasasourceofcreditqualityinformationthatmayusefullycomplementexternalcreditratingsinthecontextofastandardisedapproachtocapital.
Thesectionexamines,amongotherthings,centralcreditregisters(CCRs)andscoringsystemsbasedoncentralfinancialstatementsdatabases(CFSDs).
Ingeneral,CCRsandCFSDsareservicesprovidedbyseveralEuropeancentralbanks,sometimesjointlywiththeprivatebankingsector.
Theycollect,process,manage,andreleaseinformationonbanks'creditexposures(CCRs)andonthecreditqualityofbankcounterparties(CFSDs).
CCRsandCFSDsinthecountriesunderreview3showcommonfeatureswithrespecttocoreinformationservices,butexhibitsomedifferenceswithregardtonon-coreinformation.
Forexample,CCRsallcoverexposurestocorporateandsovereigncounterparties,buttheydifferastotheinclusionofdataonprivatecustomersorfinancialinstitutions.
Incontrast,CFSDsfocusoncorporateentities.
Databasesizesvarysubstantiallybut,ingeneral,datacollectionisdesignedtocapturethelargest10%to20%ofcounterparties,whichaccountfor80%to90%ofthetotalexposureorbusinessactivity.
Withregardtodatacollectionandprocessing,centralbanksoftenusetheirownbranchnetworksandsystems,butmayoperatejointventureswiththeprivatebankingsector.
BanksandreportingfirmsarethemainrecipientsofCCRsandCFSDsoutputs,andgeneralortailoredstudiesusingCFSDdatamaybepubliclyreleased.
Theunderlyingratingmethodologyisacriticalpointinassessingthereliabilityofacreditinformationsystem.
Thesystemsunderreviewfocusontheassessmentofcounterparties'abilitytomeettheirfinancialobligations,andwereoftenoriginallydesignedtoappraisethequalityofbillsdiscountedbycentralbanks.
However,insomecasestheanalysismayalsoencompassabroaderviewoftheoverallsituationofratedentities.
Theanalyticalapproachiseitherjudgmentalorquantitative,butpresenttrendsseemtofavourquantitativemodels.
Withinmodels,formsofdiscriminantanalysisaremostcommonlyused,butalternativeapproacheslikeexpertsystemsmaybeobserved.
SectionII.
B.
ExportCreditRatingsforSovereignsVariouscountrieshaveestablishednationalschemes,generallyknownas"exportcreditagencies,"toprovidecreditinsuranceortoassistinthefundingofcreditforexports.
Thoughoperatedindependentlyineachcountry,manyoftheseagenciesformpartofanagreement,negotiatedattheOECD,tointroducealevelofconsistencyinthedeterminationofthenationalentities'ratings.
Exportcreditratingscouldpotentiallybeusedtosupplementexternalcreditratingsforregulatorypurposes.
Inpractice,however,boththeindividualagenciesandtheOECDhaveconcernsabouttheuseoftheseratingsforpurposesotherthantheonesforwhichtheywere3Austria,Belgium,France,Germany,Italy,PortugalandSpain.
5developed.
Muchoftheinformationfromtheagenciesisonlyavailabletotheintendedusersonaconfidentialbasis.
Thissectionofthestudydescribesthesearrangementsinsomedetail,andpresentsstatisticalanalysisoftheratingsfromsevenoftheagencies.
TheagenciesprovidetheseratingstothepublicthroughtheInternet.
Thestatisticalresultsfocusontheconsistencyoftheratingsacrossagenciesindifferentcountriesandontheconsistencyoftheseratingswithexternalcreditratings.
Withregardtocross-countryconsistency,rankcorrelationswerecomputedforthesovereignborrowersratedbyallsevenagencies.
Theserankcorrelationswereveryhigh,rangingfrom93%to100%.
Thus,althoughtheagenciesindifferentcountriesoperatesomewhatindependently,theirdeterminationsarelargelyconsistent.
Whenrankcorrelationsarecomputedbetweentheaverageexportcreditratingsandtheexternalratingsofthreelargecreditratingagencies,theresultsaresimilarlyhigh.
Inthiscase,rankcorrelationswithtwoofthecreditratingagencieswere94%.
Inthethirdcase,therankcorrelationis87%,butthislowernumberseemstobeexplainedmostlybyasingleoutlierobservation.
Thus,theoverallconclusionisthatexportcreditratings,thoughnotidenticalacrosscountriesortoexternalratings,arequiteconsistentinbothdimensions.
SectionII.
C.
NewProductsfromRatingAgenciesThetraditionalratingagencyproductisanassessmentofthecreditqualityofindividualdebtissuesofafirm.
Inrecentyears,ratingagencieshaveexpandedtheircoveragetootherdebtproductsandhaveintroducedvariantsorrefinementsoftheirtraditionalproducts.
Insomecases,suchasratingsonstructureddebt,theconceptofcreditratingisessentiallythesameasbefore,althoughthedebtproductmaybemorecomplex.
Thissectiondescribesfourproductsofratingagenciesthatmightpossiblyserveascomplementstomoretraditionalratingsinaregulatoryregimethatreliedonexternalcreditratings.
First,issuerratingsmakeitpossibletoexpandtheuniverseoffirmswithcreditratingsbeyondthosethathaveissuedpublicdebt.
Second,bankloanratingsadjustfordifferencesinexpectedrecoveriesoftenobservedforbankloansindefaultrelativetobondsindefault.
Third,bankfinancialstrengthratings,bymeasuringstand-alonecreditquality,allowanassessmentofthedependenceonthesafetynetforanyparticularsetofbanks.
Finally,sovereignceilings,whichreflectcountryrisk,denotethemaximumforeign-currencyratingthatanentitydomiciledinaparticularnationcanreceive,withveryfewexceptions.
SectionII.
D.
PublishedSurveysBecauseoftherelativelackofpublicdataonthecreditqualityofsovereigndebt,severalpublicationsproduceperiodicsurveysofthissector.
ThissectionofthestudyreviewsthesovereigndebtsurveysofInstitutionalInvestor,Euromoney,andtheEconomicIntelligenceUnitofTheEconomist.
Thesurveysarebased,respectively,onstaffassessmentsatabout100largebanks,onfindingsofapanelofexternalexperts,andoninternalstaffanalysis.
6Publishedresearchpapershavecomparedtheratingsissuedbythesepublicationstoothertypesofcreditqualityinformation.
Forexample,onestudyreviewedtherelationshipoftheseratingstostandardquantitativemeasuresofcountryrisk(e.
g.
macroeconomicandexternaldebtmeasures).
Theexplanatorypowerofthestatisticalequationswasfairlyhigh,rangingfrom77%forEuromoneyto97%forInstitutionalInvestor.
Otherresearchhasinvestigatedcorrelationsbetweenthepublishedratings,ononehand,andmarketspreadsorexternalcreditratings,ontheother.
Theseempiricalcorrelationsalsotendtobequitehigh.
SectionII.
E.
MeasurementsofProbabilityofDefaultBasedonAccountingDataStatisticalscoringmethodscombineandweightindividualaccountingratiostoproduceameasure–acreditriskscore–thatdiscriminatesbetweenhealthyandproblemfirms.
Themostwidelyusedstatisticalmethodsarediscriminantlinearanalysisandprobit/logitregression.
TheclassicFisherlineardiscriminantanalysisseekstofindalinearfunctionofaccountingvariablesthatmaximisesthedifferences(variance)betweenthetwogroupsoffirmswhileminimisingthedifferenceswithineachgroup.
Thevariablesofthescoringfunctionaregenerallyselectedamongalargesetofaccountingratiosonthebasisoftheirstatisticalsignificance.
Thecoefficientsofthescoringfunctionsrepresentthecontributions(weights)ofeachratiototheoverallscore.
Logitanalysisusesasetofaccountingratiostopredicttheprobabilityofborrowerdefault,assumingthattheprobabilityofdefaulttakesalogisticfunctionalformandis,bydefinition,constrainedtofallbetweenzeroandone.
Allinall,multivariateaccounting-basedcredit-scoringmodelshavebeenshowntoperformquitewell.
Inparticular,lineardiscriminantanalysisseemsrobustevenwhentheunderlyingstatisticalhypothesesdonotholdexactly,especiallywhenusedwithlargesamples.
Logitanalysishasproducedsimilarresults.
Somerecentstudiesusebothmethodsandchoosetheonewiththebestout-of-sampleperformance,toavoidproblemsofsample-specificbiasandoverfitting.
Arelativelynew–andlessthoroughlytested–approachtotheproblemofcreditriskclassificationisbasedonartificialintelligencemethods,suchasexpertsystemsandautomatedlearning(neuralnetworks,decisiontreesandgeneticalgorithms).
Thesemethodsdispensewithsomeoftherestrictiveassumptionsoftheearlierstatisticalmodels.
SectionII.
F.
MeasuresBasedonMarketPricesThissectiondiscussesmeasuresofcreditqualitybasedonequitypricedata.
Thesemeasuresusuallyalsoincorporateinformationfromfinancialstatements.
Spreadsondebtinstruments,whicharebasedonlyonmarketpricesofafirm'sdebt,arealsodiscussedinthesection.
Themainadvantageofmarketprice-basedmeasures,ascomparedwiththosebasedonaccountingdata,isthattheymaypickupmoresubtleandfast-movingchangesinborrower7conditions,whichmaybereflectedincapitalmarketvalues.
Inaddition,measuresbasedonaccountingdataareoftenonlytenuouslylinkedtoanunderlyingtheoreticalmodel.
Thereareessentiallythreebasictypesofinformationthatarerelevanttoestimatetheprobabilityofdefault:financialstatements,marketpricesofafirm'sdebtandequity,andsubjectiveappraisalsofthefirm'sprospectsandrisk.
Financialstatementsarereflectionsofwhathappenedinthepast,whereasmarketpricesareforwardlooking.
Atleastinprinciple,pricesembodythesynthesisedviewsandforecastsofmanyinvestors.
Afterdescribingindetailvariousprice-basedmeasures,thesectionprovidesadiscussionofthemeasures'performance.
Onestatisticprovidedistheso-calledpowercurvetest,whichmeasuresthemodel'sabilitytoidentifythefirmsthataregoingtodefaultforagivenleadtime(12monthshere).
Thepowercurvetestdoesnotrequirethatamodeldeterminedefaultprobabilitiesforcompanies,sinceonlyarankingofcompaniesisnecessary.
IllustrativeresultsforasampleofcompaniesfromsixEuropeancountries4showthatthedefaultrateforthelowest-rated10%offirmsrangesfrom30%(France)to49%(Norway).
SectionIII.
A.
DefaultStudiesThissectionlooksatthepowerofexternalcreditratingstopredictdefaults.
Itdrawsonvarioussources,includingreportsoftheratingagenciesaswellasacademicstudiesofdefaults.
Someofthestudiescitedarequiterecent,includingpaperspublishedbothbyMoody'sandS&Pin1999.
Inbroadterms,theresultsofallthesestudiessuggestthatcreditratingsconstituteusefulpredictorsofdefaultsatvarioustimehorizons,particularlyfornon-financialcompaniesintheUnitedStates,forwhichthemostextensivedataareavailable.
Onetypeofevidencefrequentlycitedinthisregardisthatcreditratingsareveryhighlyrank-correlatedwithsubsequentdefaultfrequencies.
Animportantdeterminantofthepredictivepowerofcreditratingsisthedateoftheissuanceofthecreditinstrument.
Thepatternuncoveredbyseveralresearchersisthat,foranygiveninitialrating,defaultstendtoincreaseinthefirstfewyears,leveloffafterthreetofouryears,andthentendtodecline.
Thispattern,whichisclearestforfirmswithlowinitialratings,isknownasthe"seasoningeffect.
"Thesectionalsoexaminestheevidencewithregardtorecoverygivendefault,thatis,oftheproportionofthevalueofthedebtinstrumentthatisrecoveredbytheinvestorincaseofdefault.
Notsurprisingly,therateofrecoveryincreaseswiththeseniorityofthedebt.
However,thesectionreportsevidenceastothenumericalmagnitudeofthiseffect,whichtendstobequitedramatic.
Forinstance,juniorsubordinateddebttendstohaverecoveryratesofonly14%to20%.
Incontrast,seniorsecureddebtshowsratesof55%to66%andtheratesforbankloans,whicharegenerallyverysenior,rangefrom70%to85%.
4France,Germany,Norway,Spain,Sweden,UnitedKingdom.
8SectionIII.
B.
StabilityofTransitionMatricesThissectionconsidersthreewaysinwhichratingtransitionmatricesmaychangeovertime(ordepartfrom"stability"):populationtrends,cyclicalchanges,andratingsdrift.
5Withregardtotrends,itisreadilyapparentthatthegeneralpattern,atleastsincetheearly1980s,isthataproportionoflower-ratedfirmshastendedtoincrease.
Superficially,thiscouldmeanthattheaveragequalityofratedfirmsislowernoworthatratingagenciesaretougheningtheirstandards.
Infact,itisdifficulttoidentifytheindividualimportanceofvariousspecificfactors,includingthoserelatedtothedemographicsoftherateduniverse.
Therehasbeenadramaticincreaseinthenumber,types,geographicaldispersion,creditquality,andindustrialclassificationoffirmsseekingratings.
Inaddition,moreattentionisdevotednowtosectorssuchassovereigns,emergingmarkets,andspeculativegradeissuers.
Thesechangesintherateduniverse,togetherwithfactorssuchasthe"seasoningeffect"discussedintheprevioussection,combinetocreatetheobserveddecliningoveralltrendincreditquality.
Thelargerratingagenciesreportthattheyrate"acrossthecycle,"thatis,thattheytaketheperspectiveofafullbusinesscycleandthusfactorintheexpectationthattheconditionofafirmislikelytodeteriorateatthetroughofthecycle.
Nevertheless,thereseemtobesystematicchangesinratingsoverthecourseofthebusinesscycle.
Inparticular,ratingsoflower-ratedfirmstendtofluctuatemoreoverthecycle.
Ratingsdriftreferstoapatternofcontinuingchangesinaratinginagivendirection.
Ifafirmisdowngraded,dofurtherdowngradesbecomemorelikelyorisitmorelikelythatitwillrecoverTheempiricalevidencesuggeststhattheresultsareasymmetricalforupgradesanddowngrades.
Specifically,downgradesaretypicallyassociatedwithfurtherdowngrades(thereisasortofpositiveserialcorrelation),butupgradesarenotnecessarilyassociatedwithfurtherupgrades(expectationsofchangesremainroughlythesameasbeforetheupgrade).
SectionIII.
C.
ConsistencyAcrossSectorsIftwoissuersindifferentsectorshavethesamerating(symbol)fromagivenratingagency,istheirlikelihoodofdefaultaboutthesameThemajorratingagenciesgenerallyassertthatthisistheirobjectiveinformulatingtheirratings.
However,lookingatpreviousdefaultexperiencebyratingcategoryfordifferentsectors,thisbecomesanempiricalquestion.
Thissectionreportstheresultsofaseriesofnewstatisticaltestsdesignedtogaugethedirection,magnitude,andsignificanceofpotentialsectoraldifferences.
Becauseofdatalimitations,itisonlypossibletocompareafewsectors.
Ofthese,theclearestresultsarewithregardtoUSfinancialversusnon-financialfirms,andwithregardtofirmsdomiciledintheUnitedStatesversusfirmsdomiciledelsewhere.
Inthefirstinstance,thereisstatisticalevidencethatUSfinancialfirmshaveahigherdefaultratethanUSnon-financial5Transitionmatricesmeasuretheprobabilitiesofmigratingfromacurrentcreditratingtoanothercreditratingwithinaspecifictimeperiod.
9firmswiththesameratings.
Thedifferencesarestatisticallysignificant,though,likethedefaultprobabilitiesthemselvesarenotlargeinabsoluteterms.
Forexample,whenallratinglevelsareconsidered,thedifferenceinone-yeardefaultratesis0.
77percentagepointsorlessonaverage.
Othersectoraldifferencesdonotappeartoaffectdefaultratessignificantly.
Astoconsistencyinratingsacrossgeographicallydistinctissuers,statisticallyrobustcomparisonisdifficultduetothelimitedamountofdataontheratingsofnon-USfirms,particularlyatlowerratinglevels.
SectionIII.
D.
RatingDifferencesAcrossAgenciesTheissueexaminedinthissectionisdifficultbecausethereisnoperfectlyuniformstandardsetofsymbolsinusebyallratingagencies.
Thus,thereisgenerallyaneedtoeffectsometranslationofsymbolicratingframeworksbeforeevenattemptingtoaddressthequestion.
Thissectionsummarisestheresultsofearlierresearchonthetopic.
Thereismuchdataavailableforthetwolargestglobalcompanies,Moody'sandS&P,bothofwhichtendtoratealmostallissuersintheUnitedStates.
Furthermore,thereisacommonly-usedtranslationoftheirtwosetsofratingscategories.
Forthesereasons,variousresearchershavecomparedtheratingsofthesetwofirms,withthegeneralconclusionthattheyarefairlycomparable.
MorenoticeabledifferenceshavebeenfoundbetweentheseandotheragenciesactiveintheUSmarkets,andamongagenciesinothermarkets,notablyJapan.
Researchershavefoundevidencethatratingsforagivenissuertendtobelowerfromthelargesttwoagenciesthanfromotheragencies.
Evenwithadjustmentsforsample-selectionbias(firmsratedbydifferentagenciesmaybedifferent),theresultstendtobeconfirmed.
Itshouldbenoted,however,thattherankcorrelationsbetweenratingsofeachofthelargesttwoagenciesandsomeotherUSagenciesareaboutashighastherankcorrelationsbetweenthetwolargeagenciesthemselves.
Thisresultsuggeststhattheagenciesarelikelytobeinagreementovertherelativeriskofborrowers.
OutsideoftheUnitedStates,Japaneseratingagenciesareamongtheoldestandmostactive.
Dataavailabilityhasthusattractedtheattentionofresearchers.
Inthiscase,analysishasuncoveredsomefairlylargedifferencesbetweenJapaneseagenciesandnon-Japaneseagencies,whichseemtobetougheronthelocalissuers.
Nevertheless,theremaybefewerdifferencesacrossagenciesaboutrelativeriskiness,asthereisevidencethatbothJapaneseandnon-Japaneseratingsarehighlycorrelatedwithmarket-determinedcreditspreads.
10MembersoftheWorkingGrouponCreditRatingsArturoEstrella,ChairFederalReserveBankofNewYorkPatrickGuerchonovitchCommissionBancaire,FranceSylvieMathératCommissionBancaire,FranceThiloLiebigDeutscheBundesbankAntonellaFogliaBancad'ItaliaHirotakaHideshimaBankofJapanHitoshiNaganoFinancialServicesAgency,JapanTorJacobsonSverigesRiksbankJesperLindéSverigesRiksbankAndrewLoganBankofEnglandJohnCarrollFinancialServicesAuthority,UnitedKingdomJohnAmmerBoardofGovernorsoftheFederalReserveSystemFrankPackerFederalReserveBankofNewYorkSusanSzarkowitzFederalReserveBankofNewYorkDavidGreelyFederalDepositInsuranceCorporation,UnitedStatesGeorgeHancFederalDepositInsuranceCorporation,UnitedStatesJackReidhillFederalDepositInsuranceCorporation,UnitedStatesDavidNebhutOfficeoftheComptrolleroftheCurrency,UnitedStatesPeterNigroOfficeoftheComptrolleroftheCurrency,UnitedStatesCraigFurfineBankforInternationalSettlementsWilliamCoenSecretariatoftheBaselCommitteeonBankingSupervision,BankforInternationalSettlements11Preface:TheEconomicRoleofCreditRatingAgenciesAprospectiveborrowerknowsmorethanitspotentiallendersaboutitsowncreditworthiness,andthusisinapositiontodiscloseinformationselectivelyinawaythatwouldfavourablybiasanoutsider'sopinion.
Becausepotentiallendersknowthataborrowerhasanincentivetoprovideadistortedpictureofitsprospects,itcanbeachallenge,evenforalow-riskborrower,toconvincelendersthatitwouldbeunlikelytodefault.
Withoutsomemeansforreliablytransmittingrelevantinformation,therecanbeamarketfailure,inthesensethatworthyinvestmentprojectsfailtobefinanced.
Whenabankactsasanintermediarybetweenaborrowerandtheultimatesuppliersoffunding,thebankcanalleviatethisinformationproblembyconductingathoroughinspectionofafirm'sfinancialconditionandfutureprospectsbeforedecidingwhether,andonwhatterms,creditwillbeextended.
Substantialevaluationandmonitoringcostsmightbeworthbearingforabankthatexpectstobefullyexposedtotheriskthattheborrowerwoulddefault.
Incontrast,ifthefinancinginsteadwerebeingprovideddirectlybyalargenumberofsmalllenders,itmightbethecasethatnosingleoneofthemwouldhaveenoughatstaketoprovideanincentiveforanadequateinformation-gatheringeffort.
Whenafirmborrowsinpublicdebtmarkets,itcanbedifficultforpotentialbondholdersindividuallytoassesstheriskthattheborrowerwilldefault,becausethecostofanadequatecreditanalysismaybeprohibitive.
Underthesecircumstances,creditratingagenciescanplayausefulrolebycollectinginformationaboutafirmandsharingitwithalargenumberofinvestors.
Similarly,aregulatoryenvironmentthatenforcesaccuratefinancialdisclosurebyfirmswishingtoissuesecuritiesalsohelpstoenhancetheflowofreliableinformationtoinvestors.
Suchdisclosurerequirementscouldonlytendtoimprovethequalityofinformationavailabletoratingagencies,evenwhentheyalsohaveaccesstonon-publicdata.
Thus,banks,creditratingagencies,anddisclosurerequirementsallservetoreducetheextenttowhichprofitableopportunitiesareleftunfunded.
Inorderforabankorcreditratingagencytohaveanincentivetoexpendsignificantresourcesinassessingaborrower'screditworthiness,itmustbeabletocaptureaportionofthebenefitoftheinformationthatisuncovered.
Forexample,whenabankbasesalendingdecisiononproprietaryinformation,itmayprefertokeepthedetailsofthetransactionprivatetopreventothersfromcapitalisingonitsknowledge.
Ifabankcouldnotkeepitscreditevaluationssecretfromothermarketparticipants,competitorsmightbeabletomimicthebanks'lendingdecisionswithouttheexpenseofperformingindependentassessments.
6Such"free-riding"woulderodetheincentiveforthebanktoundertakecreditanalysisinthefirstplace.
Notethatpotentialcompetitorsdonotnecessarilyhavetoknowthedetailsofacommercialbank'sassessmentofborrowerrisktobeableto"free-ride"–iftheycanobservethetermsofthebank'sloancontract,thentheycansimplyofferthesamedeal.
6Syndicatedloanarrangementsinwhichtheleadbankispaidanadditionalfeemayrepresentanalternativemeansforabanktocapturepartofthebenefitofundertakinganindependentcreditassessment.
Forlargerborrowers,otherbanksparticipatinginthesyndicatewouldlikelyhavesomepriorknowledgepertinenttotheriskofdefault.
12Thus,theabilitytoexcludeothersfromone'sownassessmentprocessincreasesthepayofftocommercialbanksfromundertakingassessmentsofcreditworthiness.
However,acreditmarketwithindependentinformation-gatheringandanalysisbyeachinvestorhassomedrawbacks,withrespecttoefficiency.
Itmayrequireeitherconcentrationofcreditrisk(ifthereisonlyonebankorasmallnumberofbanksvyingforaborrower'sbusiness)orwastefulduplication(withmanycompetingbanks).
Creditratingagencies,unlikecommercialbanks,donotrisktheirownmoneyonthebasisoftheirdefaultriskassessments.
Thus,incontrasttobankfinancing,withcreditratings,thefundingactivitycanbepotentiallyseparatedfromcreditanalysis,whichmayenablebothfunctionstobeprovidedinmorecompetitivemarkets.
Nevertheless,theviabilityofacreditratingagency,likeabank,dependsonitsabilitytoextractaprivatebenefitfromitscreditassessment.
Whenratingsarepubliclydisclosed,ashasbeenthepredominantpracticeforthepastseveraldecades,theratingagencyobviouslycannotexclude"free-riders"fromlearningtheirratingsoncetheyareannounced.
Mostratingagenciesensurethattheywillcaptureaportionofthebenefitoftheircreditanalysisbychargingtheborrowerfortheirserviceratherthantheinvestorcommunity,whicheffectivelyspreadsthecostoftheratingsevenlyacrossinvestors.
(Itisworthnoting,however,thatsomeofthemorerecentdebtratingindustryentrantsmaketheirassessmentsavailableonlybysubscription.
Itisnotcleartowhatextentthesefirmsarelosingrevenuefromcustomerssharinginformation,butmanyothertypesofinformationandpublishingbusinessesaresimilarlyvulnerable.
)Inorderforaborrowertobewillingtopayforaratingfromaparticularratingagency,itmustbelievetheratingislikelytoimprovethetermsunderwhichitcouldofferdebtsecuritiesinthepublicmarket.
Accordingly,fortheagency'sratingtohavevalue,potentialinvestorsmustbelievetheratinghasusefulinformationaboutcreditworthiness.
Thus,areputationforbeingunbiasedwouldbeavaluableassettoaratingagency.
Suchareputationmightbegainedbyalongtrackrecordofsuccessfulratingsinvariousmarkets.
Somecommentatorshavesuggestedthatestablishedratingagencieshaveenoughmarketpowertoearneconomicrents,asaresultofnaturalbarrierstoentry.
However,unlikesomeotherindustries,creditratingswouldnotnecessarilybeinefficientlyunder-producedinthiscontextbecause,totheextentthatratingfeesareconfidentialandindividuallynegotiated,pricediscriminationmaybefeasible.
Aratingagency'sreputationwouldalsotendtobebolsteredifitavoidedconflictsofinterestcreated,forexample,whentheratingagencyisowned,managed,orotherwiseinfluencedbythoseinstitutionsbeingrated.
However,marketforcescanalsohelpkeepratershonest.
Inthelongrun,asinvestorsroutinelycompareratingsacrossagenciesandthecorrespondencebetweenratingsanddefault,adeliberatesystematicbiasonthepartofanygivenagencywouldriskeventualdiscovery.
Intheshorterterm,unsolicitedratingscanalsoserveasausefulformofmarketdisciplinewhenthehiredraterhasbeentoogenerous.
Ratingagencies'economicviabilitymaybeenhancedbyproprietarysystemsfortransforminginformationintoratings.
Inmostcases,theseincludeacombinationofqualitativeandquantitativejudgements,althoughsomeofthenewersubscription-basedratersareusingmethodsthatarealmostentirelyquantitative.
Onemightworrythat,givenemployeemobilityinthecreditanalysisprofession,ratingagencieswouldhaveplentyofformeranalystswhoarewellinformedabouttheirratingpractices.
However,thisdoesnotnecessarilyeliminatethevalueofproprietarysystemsifresourcesarerequiredtoimplementaratingmethodology.
13Ratings-basedfinancialregulationcanpotentiallyaltertheincentivesthatcreditratingagenciesface.
Intheabsenceofregulatoryuseofratings,theonlyvalueofratingstoanissuerliesinthecredibilityofthesignalitsendstopotentialinvestorsaboutcreditquality.
However,thesituationchangeswhencreditratingsdeterminetheconditions,ifany,onwhichaninvestormaybuyaparticularbond.
Aregulatedinvestormightpreferthatacreditratingonabondsimplybehighenoughsothatitcanbeincludedinitsportfolio,ratherthanaccuratelyreflecttheissuer'sdefaultrisk.
Undersuchascenario,itisatleastconceivablethatanunprincipledratingagencywouldimplicitlycolludewithariskyissuerandinvestorswishingtoskirtportfoliorestrictionsbyprovidinganinflatedrating.
Suchadistortionmightbeavoidedifregulatorsapplyaprocessforcertifyingratingagencies'ratingsforregulatoryusethatusessimilarcriteriatowhatinvestorsusewhendeterminingwhichratingagenciesprovidethemwithcrediblesignalsaboutcreditquality.
Inaninternationalcontext,ratings-basedcapitalrequirementsthatrelyonthissortofsyntheticmarketdisciplinewouldbemosteffectiveifregulatorscooperateclosely,sothatthecertificationcriteriaareharmonisedacrossborders.
Absentsuchcoordination,internationalbankswouldhaveanincentivetobookaratedassetinthecountrythatcertifiestheratingagencywiththemostbenignviewoftheunderlyingcreditrisk.
Inadditiontobondratingspublishedbycreditratingagencies,thereareothermechanismsthroughwhichcreditinformationcanbedisseminated.
Insomecountries,creditinformationaboutbankloansandborrowersissharedthroughprivatecreditbureaus.
Withparticipationvoluntary,memberbanksimplicitlyarejudgingthatitisworthrevealinginformationabouttheirowncustomersinexchangeforaccesstoinformationaboutotherpotentialcustomers.
Insomeothercountries,banksarerequiredtoshareinformationabouttheirborrowersthroughpubliclyadministerednationalcreditregisters.
Interestingly,thereisevidencethattheexistenceofacreditregisterorcreditbureauisadeterrenttoborrowerdefaultandconsequentlyastimulustoaggregatebankcredit(seeJappelliandPagano,1999),despitetheclearimpedimenttoindividualbanksbenefitingfromproprietarycreditinformation.
Forbothcreditbureausandcreditregisters,nomoreinformationisshared,generally,thanwouldhavetobeprovidedinaprospectusorofferingcircularforapublicbondissue.
Thus,disclosurerequirementswouldseemtobeacloserinstitutionalanaloguetothesemechanismsthanbondcreditratings.
Nevertheless,inseveralcountrieswithpubliccreditregisters,thegovernmentproducesratingsfortheborrowersonthebasisofthesubmittedinformation.
Althoughtheprincipalmotivationfortheseassessmentshasbeenfordirectuseinbanksupervision,onecouldimagineanexpansionoftheirapplicationtoratings-basedcapitalrequirements.
Nevertheless,itdoesnotseemthatcreditassessmentisanactivitythatwouldbeperformedmosteffectivelyinthepublicsector,giventhatanumberofviablecompetitorshavearisenintheprivatesector.
Theprevalenteconomicviewsuggeststhatgovernmentsshouldonlyprovidegoodsandservicesthatcannotbeproducedprofitablybyprivatefirmsbecausenon-payingcustomerscannotbepreventedfromenjoyingthebenefits.
(Anoft-citedexampleofaso-calledpublicgoodisnationaldefence.
)However,eventhoughitmaybedifficulttopreventinvestorsfromusingpublishedcreditratingswithoutexplicitlypayingforthem,ratingagencieshavefoundanalternativewaytomaketheirbusinessprofitable–chargingratingfeestobondissuers.
14PartI.
OverviewandSurveyofCurrentInstitutionsSectionA.
RatingAgencies:Size,Ownership,GeographicDistributionofRatingsAssigned,GlobalversusRegionalFocus(AnnexI.
A,pp.
21-22)IntroductionIntheconsultativepaperpublishedbytheBaselCommitteeonBankingSupervisioninJune1999("ConsultativePaper"7),theproposaltoassignriskweightsderivedfromanexternalcreditassessmenthasresultedinanexaminationoftheroleplayedbyexternalcreditassessmentinstitutions.
Theconsultativepaperreferstoratingsinthepublicdomainand,forthatreason,thissectionfocusesonratingagencies,8specificallythosethatmaketheirratingsavailablebysubscriptionorotherwise.
TheratingagenciessurveyedwereidentifiedfromreturnsprovidedbythenationalsupervisorsandcentralbanksfromtheG10countries9(exceptLuxembourg).
Thesurveyproducedalistof26uniqueentities,andtwosignificantothers,RatingAgencyMalaysiaBerhad[RAM]andCapitalIntelligence,whichcametolightduringthecourseofthestudy.
Theoriginalfigureof26includedbothDuff&PhelpsandFitchIBCAalthough,followingtheinceptionofthisstudy,themergeroftheseagencieswasannouncedandlegallycompletedon1June2000.
Thismergerwillseethenumberof"major"establishedagenciesdecreasebut,forthesakeofcomparison,dataonthetwocomponentsandthemergedentity(availablefromtheFitchIBCAInternetsite)areincluded.
RatingAgenciesAttheoutsetofthisexerciseinSeptember1999,itwasbelievedthattheremightbesome130agenciesworld-wide,althoughindustrysourcesindicatedthisnumberwascloserto150.
Althoughthemergermentionedabovewillreducethenumberof"major"establishedagencies,thereisanecdotalevidencetosuggestthatthetotalnumberofagencieswillincreaseinfuture,mostnoticeablyinthelessdevelopedmarkets.
Basedonthedatagathered,itappearsthereisawidedisparityinsizeamongratingagencies,asmeasuredbythenumberofemployeesornumberofratingsassigned(seeAnnexI.
A).
Thelattermeasure–thenumberofratingsassigned–isambiguousinthatitisdependentonwhetherthereisintensiveanalyticalworkdoneontheinstitutionbeingevaluated,orwhethertheratingexerciseconsistsofusingpublisheddataasinputtoastatisticalmodel.
Itshouldbe7"ANewCapitalAdequacyFramework",BaselCommitteeonBankingSupervision(June1999).
8Theterm"ratingagency"isusedheretodenotecommercialorganisationswhichassessthecreditworthinessofobligorsandexcludesCreditRegisters,IndustryCo-operativeActivities,GuaranteePrograms,ExportCreditRatingsforSovereigns,andotherresearchbodies.
9Belgium,Canada,France,Germany,Italy,Japan,Luxembourg,theNetherlands,Sweden,Switzerland,theUnitedKingdom,theUnitedStates.
15notedthatbothoftheseapproachesmightsatisfytherequirementof"ongoingreviewand[responsiveness]tochangesinfinancialcondition"asrequiredbythe"objectivity"criterionintheConsultativePaper.
Furthermore,whereassomeagenciesstrivetoexertaglobalreach(mostprominentintheUS-basedagencies),theremaybeanapparent"cultural"effectwhereotheragenciesseektofindnichemarkets,eitherintermsofsectorand/orgeographicspecialisation(perhapsmostnoticeableintheSwedishagencies).
Adistinctionmustbemadebetweenagenciesthatprovideratings(eithersolicitedorunsolicited)onalimitednumberoffirms,andthosethathavethecapabilitytorateallofthecompaniesinagivenmarketplaceusingstatisticalmodels.
Thesecondgroupofagencies,mostevidentinSwedenandfocussingontheSwedishorNordicmarketplace,useobjectivefinancialdatathatcanbeeasilycapturedbymechanicalmeans,therebydispensingwiththeneedforlargenumbersoftrainedanalysts.
Thescopeexists,therefore,forthisgroupofagenciestoassignratingstoallfirmswithintheirchosensphereofoperations,effectivelyachieving100%coverageofthemarketplace.
Itmaybeseenthatasignificantnumberoftheagenciesaresubsidiariesoflargerentities.
Thismightbeseentobeofconcernwheretheparentisinvolvedineithertheratingorfinancialservicesindustries(forexample,ItalratingDCR),butbeingoflessconcernwheretherearealargenumberofshareholderswithnosingleentityhavingdominantcontrol(suchasRAM).
Ashasbeenmentionedbyanumberofpractitionersintheratingagencyindustry,however,theirintegrityandcredibilityisseentobeparamounttotheirstandingintheindustryandstemfromtheirperceivedfreedomfrominterferencefromexternalsources.
Agenciesmightbedefinedasbelongingtooneofthreecategories:NationalRegionallytargetedGlobalMembersofthefirstofthesecategoriesareespeciallyevidentinSweden,whereanumberofagenciesusestatisticalmodelstodeterminethecreditworthinessofpracticallytheentirenationalmarket.
Thesecondcategoryincludesagencies,suchasCapitalIntelligence,whichthroughorganicgrowthandacquisitionhavechosentorestricttheirfocustospecificregions(suchasGulf/Mediterranean,Asia/Pacific,andCentral/EasternEuropeinthecaseofCapitalIntelligence).
Finally,therearethetrulyglobalagencies.
Membershipofthiscategoryseemstobedependentuponreachingsomecriticalmass.
SectionB.
MethodologyandRatingsDefinitions(AnnexI.
B,pp.
23-24)Thereisapreponderanceofscales,ingeneralbaseduponanalphabeticaltaxonomy.
Thismayberelatedtothedesiretoavoidtheconnotationsofcardinalitythatmightbeimmediatelyassociatedwithanumericalratingscale,althoughanordinalnumericalscaleisusedbysomeagencies.
16Inthetable(AnnexI.
B),anagency'sdesignationof"probabilityofdefault"(PD)doesnotnecessarilydenotethatanexplicitPDiscalculatedfortheissuer/issue.
Rather,thisindicatesthattheratingscaleisbaseduponthelikelihoodofdefault.
Onlytwooftheagencies,namelyKMVCorporation(US)andUpplysningscentralenAB(Sweden),undertakethecalculation/derivationofanexplicitPD.
Therestoftheagenciesbasetheirratingsontherelativelikelihoodofdefault,pointingoutthattheyarenotinthebusinessofassigningabsoluteprobabilitiesofdefaulttotheissuer/issue,butrathertheyseektoconstructanordinal,relativerankingoftheabilitytoservicedebt.
WhilesomeagenciesdoproducedefaultstudieswithcalculatedPDsbyratingclass,thisisanexpostanalysis.
Incontrast,Moody'sconsidersthe"lossgivendefault"dimensionwhenassigningarating(seeAnnexIII-RatingAgencies:Notes,p.
28).
SomeoftheSwedishagenciesmaygiveimplicitconsiderationtothiselementwhenderivingsuggestedupperlimitsforobligors.
SectionC.
MarketPractices(AnnexI.
C,p.
25)Twotypesoffeesmightbedistinguished.
Firstly,thosepaidtoaccesstheratingofanobligorfromaprivatedatabase–ineffectasubscriptionfee,usuallyoftheorderoftensofdollars–andsecondly,thosepaidto"commission"aratingthatisthenmadepubliclyavailable.
Thefeeschargedforthelatterservicemayvarybothacrossagenciesandacrossthemarketplacesinwhichtheratingisassigned.
Theagenciesthatusestatisticalmodelsreliantuponpubliclyavailabledata(generallysubscription-basedservices)havenooverridingneedtoobtainaccesstoproprietaryinformation.
Theyareabletoproducearating,whichmightbetermed"unsolicited",withoutarequestfrom,orthecollaborationof,theentitybeingrated.
Theabilityofafirmtochallengearatingis,therefore,limitedtotheextentthatrelevantinformationisinthepublicdomain,andwhereitismadeawareoftheratingassignedtoit.
Asanexample,fortheSwedishagenciesthatusestatisticalmodels,thelevelofnotificationreceivedbythefirmbeingratedisdependentuponthenatureofthefirm:Ifthecompanyisanunregisteredfirm,theratedfirmwillalwaysbenotifiedthatsomeonehassoughtaratingIfthecompanyislimitedbysharesorisalimitedpartnership,itmaybenotifieddependingontheratinginformationcontent:–Whenthereportcontainsonlytheratingclass,theratedfirmwillnotbenotified.
–Whenthereportcontainsinformationaboutowners,boardmembersorofficersofthecompany,theratedfirmwillbenotified.
Ifafirmhasbeenadvisedofanassignedratingwithwhichitdisagrees,itispossibletoraisethiswiththeagency.
Itispresumedthattheagencyhasasufficientlytransparentmethod/modeltobeabletoexplainthereason(s)forassigningtherating.
(Forthisreason,itappearsthatthemodelsbeingusedarebecomingmoretransparentandless"blackbox"innature.
)17Unsolicitedratingsmaybemoreofanissuewheretheentitybeingratedwouldotherwisehavetopayafeetoobtainasolicitedrating.
Iftheratingisatthelowerendofthescale,doesthis"drive"theentitytoobtainasolicitedratingwithconsequentpaymentofafeeWhilstthelargeragencieshavegenerallysoughttodistinguishclearlybetweenthetwocategoriesofratings,anotablerecentdevelopmentisevidentinthedeclarationbyMoody'sthatunsolicitedratingswillbedeclaredassuchinanyinitialratingassignmentpressrelease(seeAnnexIII-RatingAgencies:Notes).
CoverageParameters(AnnexesIV.
A-F,pp.
34-39)Datathatmightbeusedtoderive"coverageparameters"(boththeabsolutenumberofratingsassignedandthesizeoftheuniverseofpossiblefirmstorate)hasprovedpredictablydifficulttogather.
Thisisduetotheproblemsinherentindefiningthenumerator(issuersorissues)andthedenominator(public/privatecompanies,eventhedefinitionofacompanyunderdifferentjurisdictions)inthisratio.
ThesuggestionimplicitintheConsultativePaperisthatissuersshouldbeconsideredratherthanissues.
Inlookingatthecoverageofratingsindifferentmarkets,adistinctionmustbemadebetweenthoseagenciesthatprovideratings,solicitedorunsolicited,onalimitednumberoffirms;andagenciesthathavethecapabilitytorateallcompaniesinagivenmarketusingstatisticalmodels.
Inthelattergroup,theseagenciescaneffectivelyproduce100%coverageasaratingcanbeproduceduponrequestandpayment.
Banks,Industrials,CorporateRatingsbyAgencyandCountry/NumberofBusinesses(AnnexIV.
G,p.
39)ThecoverageparametersinAnnexesIV.
AandIV.
Barederivedbyusingasthedenominatorthenumberofentitiesratedbyatleastoneofthefourratingagenciesinthetables.
Amoreinformativestatisticofthe"penetration"ofratingswouldbeifthedenominatorweretorepresenttheuniverseofentities"eligible"toberated,butthisisaproblematicsettodetermineandhasfuzzyboundaries.
10Topresentanadmittedlysketchyviewaboutthecomparativesizeofthepossibleuniverseofsuchentitiesacrosscountries,Dun&BradstreethaveprovideddataonthenumberofbusinesseswithturnoversinexcessofUSD10million,USD50million,USD250millionandUSD500million.
However,thecompanyhasnotprovidedtheratingsassociatedwiththesebusinesses(AnnexIV.
G).
Assuch,thefiguresshouldbeinterpretedonlyasacrudeindicatoroftherelativesizebetweencountriesofthepossiblepoolofentitiesthatmightberated.
InthecontextoftherevisionoftheBaselAccord,itisalsoimportanttorealisethatthereisasecondmajordrawbacktothesefigures.
Thedataconsidersonlytheabsolutenumberofratedentities/businesses,andnottheirrelatedexposuretothebankingindustry(i.
e.
itisnot"weighted"bythebankborrowingoftheentities/businesses).
10"Eligibility"isnotdefinedbyhardandfastcriteria,butmayalsobedependentuponthedesireofanobligortoobtainaratingforanumberofsubjectivereasons,suchaspeergroupcomparison.
18TradeOrganisationsCurrentlythereappearstobenoglobaltradeorganisationfortheratingindustry,althoughtheASEANForumofCreditRatingAgencies(AFCRA)wasformedin1995byratingagenciesfromIndonesia,Malaysia,thePhilippinesandThailand.
(ItisbelievedthatatradeorganisationmayhavebeensetuprecentlyinGermany.
)19AcknowledgementsTheworkinthispaperhasbeengreatlyhelpedbythecontributionsandresponsesofthefollowingindividualsandothersintheirorganisations:JohnAmmer(BoardofGovernorsoftheFederalReserveSystem);JimArmstrong(BankofCanada);JaapBikker(DeNederlandscheBank);RudiBonte(BankingandFinanceCommission,Belgium);MaurizioCerratti,BertrandRime(SwissNationalBank);ArturoEstrella,FrankPacker(FederalReserveBankofNewYork);CraigFurfine,WilliamCoen(BIS);AndrewLogan(BankofEngland);AntonellaFoglia(Bancad'Italia);PatrickGuerchonovitch(BanquedeFrance);TorJacobson,JesperLindé(SverigesRiksbank);ThiloLiebig(Bundesbank);SylvieMathérat(CommissionBancaire);MasaoYoneyama(BankofJapan).
PaulTaylor,AnitaHazzard(Duff&Phelps);MilesNicholls(Dun&Bradstreet);DavidAndrews,CharlesPrescott,LionelPrice(FitchIBCA);AndrewChmaj,Pierre-GuyCotin,DonaldSelzer(Moody's);BarryHancock,BarbaraRidpath(Standard&Poor's);DonaldJones,PaulJennings(ThomsonFinancialBankwatch).
GeneralnotestoAnnexes1.
DatainthetablescontainedinAnnexesI.
A-C,AnnexIIandthenotesinAnnexIIIhavebeenobtainedfromresponsesprovidedbyG10supervisorsandcentralbanks(exceptLuxembourg),orobtaineddirectlyfrompubliclyavailablesources,includingreportsandInternetsites.
2.
DatainAnnexesIV.
A-EhasbeenobtainedfromtheFTCreditRatingsInternationaldatabaseforJanuary2000,thissourcealsobeingusedforAnnexIV.
GinconjunctionwithdatafromDun&BradstreetonthenumberofbusinessesasatMarch2000.
Itisimportanttonotethatthecomparabilityoffiguresinthesetablesisultimatelydependentupontherebeingaconsistentdefinitionofthe"rating"acrosstheagenciesbythecompilerofthedata;inthiscase,thisisdefinedas:"theratingassignedtolong-termseniorunsecuredorseniorsubordinateddebt.
Inarelativelysmallnumberofcasestherepresentativeratingreferstosecureddebt(e.
g.
particularlyforUSutilities)orconvertibles(e.
g.
particularlyJapaneseissuers)ormaybesupportedbyathird-partyorparentguarantee.
Inthecaseofbanks,thelong-termratingoftenreferstodeposits".
3.
DatainAnnexIV.
FissourcedfromanoteproducedbySalomonSmithBarneyinJune1999.
4.
WherethereisablankentryinAnnexesI.
A-CandAnnexIIthisindicatesthattheinformationwasnotavailable.
5.
FollowingthemergerofDuff&PhelpsandFitchIBCAinJune2000,themergedentityisdenotedhereby"Fitch(2000)"andtheentriesforthisagencyandtheconstituentparts,priortoJune2000,appearneartheendofthetablesinAnnexesI.
AtoI.
C,AnnexIIandthenotesinAnnexIII.
Intheremainingtables,thefigureson20numberofratingsfortheconstituentagencieshavenotbeenaggregatedduetothepossibilityofdouble-counting.
6.
Theentriesforthetwonon-G10agencieshavebeenplacedattheendoftheAnnexes.
7.
Collectionofthisinformationhasbeenongoingsincethelastquarterof1999andsomeofthematerialcontainedintheAnnexeshasbeentakendirectlyfromtheInternetsiteoftherelevantorganisationorfrompubliclyavailabledocuments:therefore,somenotesmayderivefrominformationpreviouslypostedbutnolongerpresentontheInternetsiteoftheorganisation.
21AnnexI.
A–RatingAgencies:Size,Ownership,GeographicDistributionofRatingsAssigned,GlobalversusRegionalFocusRatingAgencyEmployeesRatingsAssignedOwnershipGeoDistofRatingsGlobal/RegionalA.
M.
BestCo.
>400analysts,statisticians,andeditorialpersonnel5,400Independent65countriesGlobalBonniersKreditfaktaINordenAB20Allcompanies(780,000)TheBonnierGroupSwedenRegional(Sweden)CanadianBondRatingService35>500corporateandpublicsectorissuersPrivateRegional(Canadian)CreditSafeAB21690,000outof770,000(inSweden)NorwegiancompanySwedenRegional(Sweden)DominionBondRatingService30>500corporateandgovernmentissuersPrivateRegional(Canadian)Dun&Bradstreet11,000Databaseof53mcompaniesIndependent230countriesGlobalEgan-JonesCreditRatingCo.
2,000companiesUSRegional(US)EuroRatingsAG7analystsIndependentshareholdersGermanyRegional(GermanyandAustria)InstantiaCreditsystemABInternational5Allcompanies(780,000)viaKreditFaktadatabasesPrivate(TheKsterfamily)SwedenRegional(Sweden)ItalratingDCRSpA53OneItalianinvestmentbank(50%ofthecapital)andDuff&Phelps(15%ofthecapital)ItalyRegional(Italy)JapanCreditRatingAgency,Ltd(JCR)74600Leadinginstitutionalinvestors,includingmajorinsurancecompaniesandbanksGlobalJapanRatingAndInvestmentInformation,Inc.
(R&I)1401,100NikkeiNewspaper35countriesGlobalKMVCorporation25,000firmsIndependentGlobalLaceFinancialCorp.
1,000largestUSbanks;250foreignbanks;2,500largestUScreditunions;35largesttitleinsurancecompaniesMikuni&Co4,000issues/1,600firmsIndependentJapanRegional(Japan)Moody'sInvestorsService1,500>9,000Dun&Bradstreet70countriesGlobalNeufeld'sCreditInformationAB2AfewcompaniesPrivate(RobertNeufeld)SwedenRegional(Sweden)R@SRatingServicesAG8analystsIndependent,majorshareholderBavarianemployer'sassociationGermanyRegional(Germany)22RatingAgencyEmployeesRatingsAssignedOwnershipGeoDistofRatingsGlobal/RegionalStandard&Poor's1,000analysts3,478globalcorporateissuers(1997);2,614UScorporateissuers(1997)McGraw-Hill(publishingandmedia)>70countriesGlobal(22offices)SVEAKredit-InformationAB3Allcompanies(780,000)viaKreditFaktadatabasesPrivate(Lennartgren)SwedenRegional(Sweden)SVEFOSverigeAB30Allcompanies(780,000)viaKreditFaktadatabasesTeliaAB(largestphonecompany)SwedenRegional(Sweden)ThomsonFinancialBankwatch69analysts>1,000(650issuers,400issues)ThomsonCorporation85countriesGlobal(6offices)UnternehmensratingagenturAG(URA)12analystsIndependentGermanyRegional(Germany)UpplysningscentralenAB(UCAB)160Allcompanies(780,000)4largestprivateSwedishbanksSweden&NorwayRegional(Sweden&Norway)MergedagenciesDuff&PhelpsCreditRatingCo.
>600employees68%ofLatinAmericandebtissues,64%Chile,77%CostaRica,45%Mexico,75%Peru,100%Colombia,70%VenezuelaIndependent>50countriesGlobalFitchIBCA400analysts10,163globalissuers;9,033USissuersFIMALAC(Frenchconglomerate)70countriesGlobal(29offices)Fitch(2000)1,100employees1,600financialinstitutions,over800corporatesand700insurancecompanies,67sovereigns,3,300structuredfinancingsand17,000municipalbondsratings(UStax-exemptmarket)FIMALAC(Frenchconglomerate)75countriesGlobal(40offices)OutsideG10RatingAgencyMalaysiaBerhad(RAM)Ownedbycommercial,merchantbanks,financecompanies,AsianDevelopmentBank,FitchIBCACapitalIntelligence11analysts>400banksIndependent37countriesRegional(Gulf/Mediterranean,Asia/Pacific,Central/EasternEuropean)Notes1.
Ratingsassignedmaybeeithertoissuersortoissues.
(Furtherworkisneededtoconsiderhowmultipleissueratingsmightbebroughttogethertogivearepresentativeissuerrating.
)23AnnexI.
B–MethodologyandRatingsDefinitionsRatingAgencySymbolsPDvsELRelativevsAbsoluteRiskCyclevsPointintimeA.
M.
BestCo.
A++,A+(Superior),A,A-(Excellent),B++,B+(VeryGood),B,B-(Fair),C++,C+(Marginal),C,C-(Weak),D(Poor),E(UnderRegulatorySupervision),F(InLiquidation),S(RatingSuspended)BonniersKreditfaktaINordenAB1(worst)to5(best)PDRelativePointCanadianBondRatingServiceA++,A+,A,B++,B+,B,C,D,SuspendedCreditSafeAB0(worst)to100(best)PDRelativePointDominionBondRatingServiceAAA,AA,A,BBB,BB,B,CCC,CC,C,DPDDun&BradstreetAAA,AA,A,B,C(AAAbest,Cworst)PDRelativePointEgan-JonesCreditRatingCo.
EuroRatingsAGAAA,AA+,AA,AA-,A+,A,A-,BBB+,BBB,BBB-,BB+,BB,BB-,B+,B,B-,CCC+,CCC,CCC-,CC,DPDRelativeCycleInstantiaCreditsystemABInternational1(worst)to5(best)PDRelativePointItalratingDCRSpAAAA,AA+,AA,AA-,A+,A,A-,BBB+,BBB,BBB-,BB+,BB,BB-,B+,B,B-,CCC,DD(seenote)JapanCreditRatingAgency,Ltd(JCR)AAA,AA+,AA,AA-,A+,A,A-,BBB+,BBB,BBB-,BB+,BB,BB-,B+,B,B-,CCC,CC,C,DCycleJapanRatingAndInvestmentInformation,Inc.
(R&I)AAA,AA+,AA,AA-,A+,A,A-,BBB+,BBB,BBB-,BB+,BB,BB-,B+,B,B-,CCC,CCKMVCorporationn/aPDAbsolute(computePDs)PointLaceFinancialCorp.
A+,A,B+,B,B-,C+,C,D,EMikuni&CoAAA,AA,A,BBB,BB,B,CCC,CC,DDDMoody'sInvestorsServiceAaa,Aa1,Aa2,Aa3,A1,A2,A3,Baa1,Baa2,Baa3,Ba1,Ba2,Ba3,B1,B2,B3,Caa1,Caa2,Caa3,Ca,C(seenote)ELRelativeCycleNeufeld'sCreditInformationABNoformalgradingschemePDandELRelativePointR@SRatingServicesAGAAA,AA+,AA,AA-,A+,A,A-,BBB+,BBB,BBB-,BB+,BB,BB-,B+,B,B-,CCC+,CCC,CCC-,CC,DPDRelativeCycleStandard&Poor'sAAA,AA+,AA,AA-,A+,A,A-,BBB+,BBB,BBB-,BB+,BB,BB-,B+,B,B-,CCC+,CCC,CCC-,CCPDRelativeCycleSVEAKredit-InformationAB1(worst)to5(best)PDRelativePointSVEFOSverigeAB1(worst)to5(best)PDRelativePointThomsonFinancialBankwatch(Issuer)A,A/B,B,B/C,C,C/D,D,D/E,EIntermediateTimeHorizonThomsonFinancialBankwatch(LongTerm)AAA,AA+,AA,AA-,A+,A,A-,BBB+,BBB,BBB-,BB+,BB,BB-,B+,B,B-,CCC,CC,DCycle24RatingAgencySymbolsPDvsELRelativevsAbsoluteRiskCyclevsPointintimeThomsonFinancialBankwatch(Sovereign)AAA,AA+,AA,AA-,A+,A,A-,BBB+,BBB,BBB-,BB+,BB,BB-,B+,B,B-,CCC,CC,C,DLongTermUnternehmensratingagenturAG(URA)AAA,AA+,AA,AA-,A+,A,A-,BBB+,BBB,BBB-,BB+,BB,BB-,B+,B,B-,CCC+,CCC,CCC-,CC,DPDRelativeCycleUpplysningscentralenAB(UCAB)1(worst)to5(best)[Grade1PD>26%,grade5PD26%,grade5PD=$10M>=$50M>=$250M>=$500MUnitedStates5544153,3132,544155,50037,8269,0354,819France2611125625726,3095,5811,280684Germany3413167842,8519,7002,1011,100Italy1431917020,3723,497493228UnitedKingdom639906527,3298,5472,1821,148Japan219974586,40922,0154,5362,258Netherlands41136333,8311,851531341Sweden-723245,2141,105231106Luxembourg-4221123083157Belgium27613106,0081,354232108Switzerland141974,4131,358450252Canadan/an/an/an/a14,1723,325956534Non-G10(Selected)Argentina14947443,14163811247Austria-24683,7511,081262150Chile291582333041588058Korea13367789488184108Spain356412910,6341,958350159Source:Duff&Phelps/FTCreditRatings(January2000)Dun&Bradstreet(March2000)40SectionD.
RegulatoryCertificationProceduresIntroductionThissectionsummarisestheresultsofasurveyintothecurrentuseofcreditratingagencies'ratingsinfinancialregulation,particularlybankingsupervision,acrosseighteencountries.
11Participantswereaskedforinformationonhowratingswereusedintheircountries'regulations,whichagencies'ratingswereeligibleforuse(andhowthishadchangedovertime),therecognitioncriteriausedbythesupervisoryauthoritiesandwhethertheauthoritiescarriedoutongoingmonitoringofagencies'performance.
Theywerealsoaskedforinformationonhowtheirsupervisoryauthoritiestreatedsplitandunsolicitedratings.
ThecountriesselectedwerethetwelvemembersoftheBaselCommitteeonBankingSupervision(BCBS)andsixinterestingnon-members:Australia,Argentina,Chile,HongKong,MexicoandNewZealand.
Thelatterwereselectedonthebasisoffulfillingoneormoreofthefollowingcriteria:havinganinnovativeapproachtofinancialregulation;arelativelyrecentexperienceoffinancialinstability;aknowninterestinguseofcreditratingsintheirregulationsand/ortheirgeneralimportancetotheworldfinancialsystem.
Responseswerereceivedfromalleighteencountries.
Inanysurveythereisthepotentialproblemofrespondentsinterpretingthequestionsindifferentways.
Itisanissueherebecausesomerespondents'repliespertaintojusttheprudentialregulationofbanks,whileothersrefertoallformsoffinancialregulationintheircountry.
Forthesakeofcomparability(andalsobecausethemotivationbehindthisstudyisthereformofthe1988BaselAccord)thediscussioninthetextreferstojusttheregulationofbanks.
Whererespondentshaveprovidedinformationontheuseofratingsintheregulationofothertypesoffinancialfirms,thatinformationisincludedinthetables,butitisnotactivelydiscussed.
Thesurveyquestionsandanswersaresummarisedinthefollowingseparatesections.
InformationonBCBS-membercountriesispresentedfirst,followedbyinformationonthesixothercountries.
11WiththankstoFrankPacker(formerlyoftheFederalReserveBankofNewYork)forhisassistanceinthedesignofandtheresponsetothesurveyquestionsandthefollowingindividualsfortheirresponsetothesurvey:JaapBikker(DeNederlandscheBank);RudiBonte(CommissionBancaireetFinancière);JoseLuisBracho(ComisionNacionalBancariaydeValores);DenysBruce(ReserveBankofNewZealand);JohnCarroll(FinancialServicesAuthority);AntonellaFoglia(Bancad'Italia);BoGreborn(Finansinspektionen);PatrickGuerchonovitch(BanquedeFrance);WilliamJones(AustralianPrudentialRegulationAuthority);FredericLau(HongKongMonetaryAuthority);ThiloLiebig(DeutscheBundesbank);EricOsch(CommissiondeSurveillanceduSecteurFinancier);BradShinn(OfficeoftheSuperintendentofFinancialInstitutions);UweSteinhauser(Eidg.
Bankenkommission);MaríaInésUrbina(BancoCentraldeChile);AgustinVillar(FinancialRepresentativeOfficeinEurope,RepublicofArgentina);MasaoYoneyama(BankofJapan).
41Dothefinancialregulatoryauthoritiesinyourcountryuseratingagencies'ratingsinregulationIfso,pleaseindicatetheparticularregulation.
Table1summariseshowcreditratingagencies'ratingsarecurrentlyusedinthefinancialregulationundertakeninthecountriespolled.
ItisevidentthatratingsareusedinelevenoutofthetwelveBCBS-membercountries.
TheexceptionisGermany.
Ofthesixnon-members,fiveuseratingsintheirfinancialregulation,butMexicodoesnot.
SevenoutoftheelevenBCBSmembersthatuseratingsintheirprudentialsupervisionofbanksdososolelytodeterminewhatisaqualifyingdebtsecurityorotherinterestraterelatedinstrumentforthecalculationofthecapitalrequirementforspecificinterestraterisk.
ThisissetoutinthestandardisedmethodologyofthemarketriskamendmenttotheoriginalBaselAccord.
ForthemembersoftheEU/EEA,thisisencapsulatedintheCapitalAdequacyDirective(CAD).
12MembercountriesareallowedtowaivethispartoftheCAD,ifthey"judgeitinappropriateinthelightof,forexample,thecharacteristicsofthemarket,theissuer,theissue,orsomecombinationofthosecharacteristics".
ItisthiswaiverwhichGermanyoptstoexercise.
Twooutofthesixnon-membercountries(AustraliaandHongKong)alsouseratingstojudgeaqualifyingdebtsecurityformarketrisk.
OnlyfourBCBSmembersuseagencies'ratingsintheirprudentialregulationofbanksforpurposesotherthanmarketrisk.
InBelgium,banksarerequiredtoprovideinformationonthecompositionofpartsoftheirsecuritiesportfolio,splitbyrating.
InSwitzerland,ratingsareusedinthedefinitionoftheriskweightsforcreditrisk,wheretheappropriateweightdependsonthecounterparty'slocation.
ThedefinitionofwhatconstitutesanOECDcountryhasanadditionalcriterion,whichexcludescountrieswithalower-than-investment-graderating(byarecognisedagency)ontheirlong-termforeigncurrencyliabilitiesorwheretheyareunrated,theiryieldtomaturityandremainingdurationarenotcomparablewiththoseoflong-termliabilitieswithaninvestmentgraderating.
IntheUnitedKingdom,thosebanksusingthemismatchapproachtoliquiditymonitoringareabletoinsert"marketableassets"inanearliertimeband(atadiscounttotheirrecordedvalue)ratherthantheonethatcorrespondstotheirlatestcontractualmaturity.
ThedefinitionofamarketableassetusesthesamecriteriaforqualifyingitemsasisusedintheCAD.
12TheCADappliestobothcreditinstitutionsandinvestmentfirms.
42Table1:Useofratingagencies'ratingsinfinancialregulationCountryDetailsoftheregulationBCBSMembersBelgiumCAD/Marketriskamendment.
Prudentialreporting:thedescriptivetablesrelatingtothecompositionofabank'ssecuritiesportfoliorequireinformationonsecurities'ratingsandtheagencieswhichissuedtheratings.
CanadaMarketriskamendmentFranceCAD/MarketriskamendmentGermanyNoItalyCAD/MarketriskamendmentJapanMarketriskamendmentLuxembourgCAD/MarketriskamendmentNetherlandsCAD/MarketriskamendmentSwedenCAD/MarketriskamendmentSwitzerlandMarketriskamendment.
Creditrisk:somerisk-weightsdependonwhetherthecounterpartyislocatedwithinanOECDcountry.
WhereOECDcountriesaredefinedasfullmembersoftheOECD,orcountriesthathaveconcludedspecialcreditagreementswiththeIMFinconnectionwiththeGeneralAgreementsonCreditofthelatter,excludingthosewhichhavere-scheduledtheirexternaldebtsduringtheprevious5years,orhavealowerratingthaninvestmentgradeonitslong-termforeigncurrencydebt(whereithasnorating,itsyieldtomaturityandremainingdurationmustnotbeincomparablewiththoseoflong-termliabilitieswithinvestmentgraderatings).
Investmentfunds:fundmanagersarerestrictedwithwhomtheymayconcludecertainderivativetransactions,dependentonthecounterparty'screditrating.
UKCAD/Marketriskamendment.
Liquidityreportingguidelinesfornon-clearingbanks.
USASeeTable6inAppendix1(p.
54)Sixnon-membersoftheBCBSArgentinaBanksandfinancialcompaniesmustseekaratingfromanauthorisedratingagency.
Theratingreflectstheabilityofthefinancialinstitutiontorepayitsmedium-andlong-termliabilities.
Althoughtheratingscalesareidenticaltothoseusedbyinternationalratingagencies,theratingsdonotencompassthecountryriskanalysis.
Inthecaseofbranchesofforeignbanksorsubsidiarieswhollyownedbyforeignbankswhoseheadquartersguaranteetheobligationsoftheirsubsidiariesirrevocably,thereisanalternativeratingsystem.
Financialinstitutionsmustprovidecopiesofthereportstocustomerswhorequestthemfreeofcharge.
However,theycannotbeusedinadvertisingcampaignsorprinteddocuments.
Thecentralbankpreparesalistofbanksthatcanreceivetimedepositsfrominstitutionalinvestors(pensionfunds).
Bankswithweakratingsareexcludedfromthislist.
TheComisionNacionaldeValores(CNV),thestock-marketwatchdog,doesnotextendauthorisationforthepublicofferofasecurityunlessitsissuerhassoughttworatings.
Inaddition,pensionfundsarenotallowedtoinvestinassetsthatdonotexceedacertainratingthreshold,whichissetatBBBfordomesticcreditratingsandBforratingsissuedbyinternationalagenciesonsecuritiesofresidentissuers.
Thesameprovisionsareextendedtotheinsuranceindustry.
Inthiscase,theinsuranceindustryisbeingaskedtoinvestinratedsecuritieswithaminimumrating,andalsoaskedtoseekaratingaspoliciesissuedbythemincreasinglyarebeingsoldtopensionfunds.
43AustraliaPrudentialStatementC3–CapitalAdequacyforBanks,ratingsareusedtodeterminethecapitalrequirementforspecificriskforinterestrateriskinthetradingbook.
PrudentialStatementC1–Recognisesmortgageinsurance(aspartofloantovaluationcalculations)forrisk-weightingloanssecuredbyresidentialmortgageswherethelenders'mortgageinsurercarriesacreditratingofAorhigherfromanapprovedcreditratingagency.
PrudentialStatementC2–Coverssecuritisationandfundsmanagement,andalsomakesreferencestocreditratings(forexample,indeterminingtheadequacyofcreditenhancementsprovidedtosecuritisationschemes).
ChileCompaniesthatissueequitiesinADRsmusthaveaminimumrating.
Chileaninstitutionsareonlypermittedtoinvestinoverseassecuritiesthathaveaminimumrating.
HongKongLiquidityregime:authorisedinstitutions'(AIs)holdingsofmarketabledebtsecuritiesmayberegardedasliquefiableassetsforthecalculationoftheliquidityratioifthedebtsecuritiessatisfythequalifyingcreditrating.
Thestatutoryminimumliquidityratio,expressedasapercentageofliquefiableassetstoqualifyingliabilities,is25%forallAIs.
Capitaladequacyregime:underthemarketriskcapitaladequacyframework,debtsecuritiesinthetradingbookthatsatisfytheminimumratingsmaybeincludedina"rated"category.
Ratedsecuritiescarrylowerriskweightingscomparedwith"unrated"securities.
DiscountWindow:threetypesofsecuritiesareeligibleforovernightrepoundertheDiscountWindowoperatedbytheHKMA,namely,theExchangeFundpaper,theexistingSpecifiedInstruments,andotherHongKongdollarsecuritieswithlong-termratingshigherthantheminimumacceptableratings.
Lenderoflastresort:inadditiontotheeligiblepaperfortheDiscountWindow,otherHongKongdollarsecuritieswithaninvestmentgraderatingassignedbyarecognisedcreditratingagencyarealsoeligibleforrepotransactionundertheLOLRfunction.
MexicoNoNewZealandAregisteredbankthathasacreditratingonitsseniorunsecuredlong-termNewZealanddollardebtpayableinNewZealandisrequiredtodisclosethatratinginitsquarterlydisclosurestatements.
Thedisclosuresincludeinformationonthenameoftheratingagency,thedateofrating,thenatureoftheratingsnomenclatureused,andanychangestotheratingovertheprevioustwoyears.
Abankthatdoesnothavearatingofthespecifieddebtobligationalsoisrequiredtodisclosethatfactinitsquarterlydisclosurestatements.
44TheUnitedStateshasalongerhistoryofusingratingagencies'ratingsinfinancialregulation.
AnoverviewisshowninAppendix1(p.
54).
In1931,bankswererequiredtomarktomarketlowerratedbonds.
In1936,theywereprohibitedfrompurchasing"speculativesecurities".
In1994,thecapitalrequirementsonbanks'holdingsofdifferenttranchesofasset-backedsecuritiesweremadeafunctionoftheirrating.
In1999,theabilityofnationalbankstoestablishfinancialsubsidiarieswasrestrictedbasedontheirrating.
Threeofthenon-BCBSmembercountriesuseratingagencies'ratingsintheirprudentialregulationofbanks(apartfrominthecontextofmarketrisk).
InArgentina,thecentralbankconstructsalistofbankspermittedtoreceivetimedepositsfrominstitutionalinvestors(pensionfunds)andabank'sratingisoneofthefactorsonwhichthecentralbankbasesitsdecisiononmembership.
InAustralia,ratingshavetwoadditionalrolesindeterminingbank'scapitalrequirement.
First,theAustralianPrudentialRegulationAuthority(APRA)recognisesmortgageinsurancebyinsurersthathavearatingofAorhigherbyarecognisedratingagency.
Second,ratingshavearoleindeterminingtheadequacyofcreditenhancementsprovidedtosecuritisationschemes.
InHongKong,ratingsareusedtodeterminewhatisaliquefiableassetintheliquidityregime.
InArgentinaandNewZealand,theauthoritiesmakeuseofagencies'ratingsofthebanks(asopposedtotheircounterparties)intheirregulation.
Theyareusedtoprovideinformationtothebanks'creditorsandtherebyfacilitatemarketdiscipline.
Intheformer,allbanksandfinancialinstitutionsmustobtainaratingfromarecognisedagency.
Eachbankisrequiredtoprovideacopyoftheassociatedreporttoanycustomerrequestingonefreeofcharge.
InNewZealand,aregisteredbankisrequiredtoincludeinitsquarterlydisclosurestatementsinformationonwhetherithasarating(andspecificdetails,ifitdoes)onitsseniorunsecuredlong-termdebtdenominatedinNewZealanddollarspayableinNewZealand.
ThisincludesboththeGeneralDisclosureStatement,whichistargetedatsophisticatedreadersandtheKeyInformationStatement(KIS),whichistargetedatthe"prudentbutnotexpertinvestor".
Ratingsareregardedasbeingaparticularlyusefulindicatorofrelativebankcreditqualityforthelattergroupbecausetheyarerelativelyeasytounderstand.
EachbankisrequiredtodisplayitsKISstatementandbeabletomakeitavailableimmediatelyinallitsbranches.
Italsomustbedisplayedonitsinternetsite.
ThemotivationbehindtheNewZealandrequirementsisthreefold.
First,itprovidescreditorsofbankswithadditionalinformationwithwhichtoassessthesoundnessofabankandcompareonebankwithanother.
Second,itreinforcesincentivesfortheprudentmanagementofbanksandisconsistentwiththeaimofplacinggreateremphasisonmarketdisciplineasameansofpromotingasoundbanksystem.
Third,itassistsinreducingtheperceptionthatbanksaresomehowunderwrittenbythegovernmentorcentralbank.
WhichratingagenciesareusedComparingthenumberofratingagencieseachcountry'sbankingsupervisorrecognisesiscomplicatedbecausesomesupervisorsrecognisedifferentpartsofthesameratingagency,whileothersrecognisetheagencyinitsentirety.
Thisoccurswhentwoagencieshave45merged13orwhenthereisalocalsubsidiaryofaninternationalagencyandtheinternationalagencyitselfoperatinginthesamecountry.
Itseemslikelytoreflecttheratingagencies'varyingoperatingstructuresacrossdifferentcountries,asmuchassupervisorsuseofdifferentrecognitionpolicies.
Regardlessofthecause,itcomplicatesthecomparison.
Table2showstheratingagenciesrecognisedbythebankingsupervisorsineachoftheBCBS-membercountriesandtheselectednon-members.
Itisconstructedonacomparablebasis,inthatitignoresasupervisorissuingmultiplerecognitionstothesameagency(oritsvariousparts).
Thetotalnumberofagenciesrecognisedineachcountryisshownintheright-handcolumn.
Itisevidentthereisconsiderabledisparityinthenumberofrecognitionsgrantedbythesupervisorsinthevariousmembercountries.
TheUnitedKingdomrecognisesthemostagenciesatten.
ItisfollowedbyFrance14andtheNetherlands,whicheachrecognisenine.
ThebankingsupervisorsinLuxembourgandSwedenrecognisetheleastatthree.
Analysisofthefournon-BCBSmembers(showninthelowerhalfofTable2)thatrecogniseagenciessuggeststheyrecogniseaboutthesamenumberastheircounterpartsintheBCBS.
ThemeannumberofagenciesrecognisedinArgentina,Australia,ChileandHongKongis6.
Thiscomparestoafigureof6.
3fortheelevenmembercountrieswhichuseratings.
However,thisresultisheavilydependentontheinclusionofAustralia.
Therowsattheendofeachsection(withthefiguresinitalics)showthenumberofcountriesinwhicheachagencyisrecognised.
FitchIBCA,Moody's,andS&Parerecognisedinallelevenmembercountriesthatusecreditratingsintheirbankingsupervision.
Thereafter,ThompsonBankwatchisrecognisedineight,DuffandPhelpsinsevenandDominionBondRatingServiceandJapanCreditRatingAgencyinfive.
Theremainderarerecognisedinfourorfewercountries.
Thedominanceoftheinternationalagenciesisalsoapparentinthenon-BCBSmembercountries.
FitchIBCAandS&Parerecognisedbyallfourcountriesthatgrantrecognition(asisThompsonBankwatch).
Moody'sisrecognisedinthree.
Moody'sdoesnothaveasubsidiaryinArgentina,soitdoesnotissueratingswhichfollowthecentralbank's(whichisresponsibleforbankingsupervision)guidelines.
However,itisrecognisedbytheArgentinesecuritiesregulator.
13ThiswascompiledpriortotheannouncementofthemergerbetweenFitchIBCAandDuff&PhelpsCreditRatingCo(DCR).
14Ifmultiplerecognitionsofthesameagencyarepermitted,Franceissuesthemostrecognitionsatthirteen.
46Table2:RatingagenciesrecognisedinvariouscountriesCanadianBondRatingService(CBRS)DominionBondRatingService(DBRS)Duff&PhelpsCreditRatingAgencyFitchIBCA1ItalRatingDCRSPAJapanCreditRatingAgency(JCR)JapanRatingandInvestmentInformation1Mikuni&CoMoody'sInvestorServiceStandard&Poors'RatingServicesThompsonBankWatchTotalnumberofratingagenciesrecognisedMembersoftheBaselCommitteeonBankingSupervisionBelgiumΤΤΤΤΤ5CanadaΤΤΤΤΤΤ6FranceΤΤΤΤΤΤΤΤΤ9ItalyΤΤΤ2ΤΤΤ6JapanΤΤΤΤΤΤΤ7LuxembourgΤΤΤ3NetherlandsΤΤΤΤΤΤΤΤΤ9SwedenΤΤΤ3SwitzerlandΤΤΤΤΤΤ6UKΤ2Τ2Τ2ΤΤ2Τ2Τ2ΤΤΤ210USAΤΤΤΤΤ5TotalBCBS4571115421111869Interestingnon-membersoftheBCBS3ArgentinaΤΤΤΤ4AustraliaΤ2Τ2Τ2ΤΤ2Τ2Τ2ΤΤΤ210ChileΤΤΤΤΤ5HongKongΤΤΤΤΤ5Totalnon-BCBS1224012134427Total579161663151612Notes1.
Seecommentsinthetextontreatmentofwhereacountry'sregulatorrecognisesmorethanonepartofaratingagency.
2.
Indicateswheretheregulatorstatedthatthisagencyisonlyrecognisedforasubsetofissues.
3.
NewZealandisexcludedasitdoesnotrecogniseagencies.
47TheremainingnoteworthyfeatureinTable2istheAustralian,SwissandtheUKbankingsupervisorsuseofMikuno&Co,aJapaneseratingagencythatisnotrecognisedbytheJapanesesupervisors.
Alltheotherratingagenciesrecognisedbyoverseassupervisorsarerecognisedbytheirdomesticcounterpartsaswell.
HowhasthelistofeligibleagencieschangedovertimeTable3showsthechangestothelistofrecognisedratingagenciesforeachcountrysinceitscreation.
EightoftheelevenBCBSmembercountriesthatuseratingshaveeithernevermadechangestotheirlists,orthechangesreflectmergersandtake-overactivitybetweenexistingmembersofthelist.
TheeightcountriesareCanada,France,Japan,Luxembourg,theNetherlands,Sweden,Switzerland,andtheUnitedKingdom.
ThesameistrueforArgentina,Australia,andChilewhenlookingatthefournon-BSBCmembercountriesthatrecogniseratingagencies.
Thelackofanychangetothelistsofrecognisedagenciesinthesecountriesprobablyreflectstheirrelativelyrecentconstruction.
Informationontheirdateoffirstconstruction(wheresuppliedbytherespondent)iscontainedinthesecondcolumnofTable3.
Mostwerecompiledin1996or1997toimplementeithertheCADorthemarketriskamendment.
Wheretheregulatoryauthoritieshavealteredtheirlistsofrecognisedagencies,theirmainactionhasbeentoaddnewagencies.
InBelgiumandItaly,twoagencieswereadded.
IntheUnitedStates,fouragencieshavebeenadded.
InHongKong,threeadditionalagencieshavebeenrecognised.
Wheredeletionshaveoccurredtheydidsoforreasonsofmergerandtake-overofonerecognisedagencybyanother.
NorespondentfromaBCBSmembercountryreportedacasewhereanagencywasremovedfromitslistforanotherreason,suchasincompetence.
Likewise,noneoftherespondentsfromnon-membersoftheBCBSreportsuchanexclusion,althoughtheArgentineanreplycitedoneexclusionbytheirstockexchangeregulators,noexplanationforthemovewasgiven.
Theabsenceofdeletionsforreasonsotherthantake-overormergerisofinterest.
Withoutgreaterknowledgeitisnotpossibletoknowwhetherthisreflectstheconsistentlyhighqualityoftherecognisedratingagenciesorthelackofmonitoringbytheregulatoryauthorities.
Thelatteroralternativestoit(suchasregularlyrepeatedsurveysofwhichagenciesmarketparticipantsuse)wouldseemapre-requisiteforforceddeletions.
48Table3:Alterationstoregulators'listofrecognisedratingagenciesCountryListfirstcreatedSubsequentchangestolistBCBSmembersBelgiumJune1996AgenciesinitiallyrecognisedwereMoody'sInvestorService,Standard&Poor'sCorporation,IBCALtdandThomsonBankwatch.
Thefollowingwereaddedtothelist:FitchInvestorsServicesLPandDuff&PhelpsCreditRatingCompany.
CanadaNovember1997NochangeFrance1995NochangeItaly1996Listfirstpublishedin1996containedItalRatingDCRSpa,Moody'sInvestorServices,StandardandPoor'sandFitchInvestorsService-IBCA.
In1999,Duff&PhelpsCreditRatingCoandThomsonBankwatch,Incwereadded.
Japan1997NochangeexceptformergersbetweenagenciesLuxembourgSinceimplementationoftheCADNochangeNetherlands1996NochangeSweden1January1996NochangeSwitzerland1997NochangeUK1995Nochangeexceptmergers/take-oversUSA1975WhenthetermNRSROwasinitiallyadoptedin1975,threeagencieswererecognised:Moody'sInvestorService,Standard&Poor'sandFitchInvestorServices.
In1982,DuffandPhelpswasadded.
In1983,McCarthy,Crisanti&Maffeiwasadded.
In1991,IBCAwasadded,andin1992,ThomsonBankwatchwasadded.
In1991,Duff&PhelpsacquiredtheratingfranchiseofMcCarthy,Crisanti&Maffeiin1991,sothiswasdeleted.
InNovember1997,FitchIBCAmerged,soIBCAnolongerhasNRSROstatus.
Non-membersoftheBCBSArgentinaMid1997Nochangetothecentralbankregistry.
TheCNVregistryopenedin1992,sincethen,onecompanyhasbeenexcludedandtwotakenoverbytheirforeignshareholders.
AustraliaJanuary1997NochangeChileN/ANochangeHongKong1994In1994,Moody'sandS&PfirstrecognisedfortheLiquidityRegimeandtheLiquidityAdjustmentFacility.
In1996,IBCA,ThomsonBankwatchandtheJapanBondResearchInstitute(whichbecametheJapanRatingandInvestmentInformationInc)wereaddedtothelist.
In1997,thefivewereacceptedforthereportingofmarketrisk,in1998forassessingtheeligibilityofdebtsecuritiesforDiscountWindow,andin1999forthelenderoflastresortfunction.
NewZealandN/ANotapplicable49WhatarethecriteriaforeligibleagenciesTable4showsthecriteriausedbybankingsupervisorsinthevariouscountriestorecogniseratingagencies.
Theoriginalwordingfromtheirreplies15hasbeenkepttoavoidanypossiblemisrepresentation.
Theexistingnationalrecognitioncriteriaarecomparedwiththoseproposedforeligibleexternalcredit-assessmentinstitutionscontainedwithintheconsultativepaper.
Attheoutset,itshouldbesaidthatthiscomparisonshouldbetreatedwithsomecaution,asthecomparabilityoflanguageismoresubjectivethantherelativitiesbetweennumbersandtheinterpretationmaynotcoincidewiththemeaningintendedbytherespondents.
Theobjectivitycriteriaintheconsultativepaperspecifiesthat"themethodologyforassigningcreditassessmentsmustberigorous,systematic,continuousandsubjecttosomeformofvalidationbasedonhistoricalexperience".
VirtuallyallmembersoftheBCBSspecifyobjectivityorsomethingakintoitaspartoftheircriteria.
Thetwothatdonot(LuxembourgandtheUnitedKingdom)haverecognitioncriteriabasedmainlyonmarketusage.
TheUnitedKingdom'seligibleagencieslistisbasedonasurveyofwhichagenciesbanksuse.
Ofthenon-BCBSmembers,therespondentsfromAustraliaandHongKongreporttheuseofobjectivity.
Thesecondcriteriaproposedintheconsultativepaperisindependence,wherethemethodologymustbe"asfreeaspossiblefromanyexternalpoliticalinfluenceorconstraints,oreconomicpressuresfromassessedentities".
OnlyfourBCBSmembercountriesexplicitlystatedthewordindependenceorusedaphrasewithsimilarmeaning.
ThesewereBelgium,Japan,SwitzerlandandtheUnitedStates.
OnlytheItalianrespondentexplicitlycitedtransparencyasoneoftherecognitioncriteriausedbythebankingsupervisorwithinhercountry.
TherespondentsfromfouroftheBCBSmembercountriesexplicitlycitecredibilityasoneofthecriteriatheirbankingsupervisorsuse.
ThecountriesareBelgium,Italy,SwitzerlandandtheUnitedStates.
Ofthenon-members,AustraliaandHongKongwouldalsoseemtoplacerelianceoncredibility.
Noneoftherespondentscitedinternationalaccessasoneofthecriteriathattheycurrentlyusetodeterminewhetheranagencyshouldberecognised.
OftheBCBSmembers,onlytheUnitedStateshasanexplicitresourcecriterion-theemploymentofadequatemembersofstaffwiththenecessaryeducationandexperience.
However,theBelgianandJapaneserespondentsspecifiedanexaminationoftheorganisationwhichpresumablyincludesitsresources.
Ofthenon-members,theHKMAalsotakeintoconsiderationthenumberofanalysts.
AnalysisoftheresponsesshowthattwoBCBSmembersciteinternationalrecognitionasonecriteriontheyuse.
TheseareCanadaandLuxembourg.
Itisunclearwhethertheymean15TheChileananswerhasbeentranslatedfromSpanish,theprecisewordingmaythereforenotbeidenticaltotheoriginalresponse.
50overseasmarketusageorregulatoryrecognition.
Ifitistheformer,theSwissalsousesuchacriterion.
Non-membersArgentina,AustraliaandHongKongalsociteinternationalreputation.
Someofthenon-membersalsoexplicitlyuseinternationalregulatoryrecognition.
ThismayraiseapotentialissuefortheimplementationoftheproposedreformstotheAccord,ifnon-membersarelikelytofollowthelistofagenciesrecognisedbyBCBSmembers,withoutcheckingtheircompetencyintheirownmarket.
IsthereongoingmonitoringoftheperformanceofagenciesThreeoftherespondentsfromBCBSmembercountries(France,ItalyandJapan)reportedtheirprudentialsupervisorsundertakeongoingmonitoringoftheratingagenciestheyrecognise.
InItaly,thisisundertakenonthebasisofpubliclyavailableinformation.
InJapan,ratingperformanceisoneofthelegal-basedcriteriaforrecognition.
Thelawdoesnotdetailhowthisshouldbeundertaken,soitisatthesupervisor'sdiscretion.
Oftheeightmembercountries(Belgium,Canada,Luxembourg,theNetherlands,Sweden,Switzerland,theUnitedKingdomandtheUnitedStates)whichdonotmonitorratingagencies'performances,fourreportedcaveatstotheirnegativeresponse.
InBelgium,filesarekeptup-to-datewhichcontainsignificantinformationondifferentratingagencies.
IntheUnitedStates,responsibilitylieswithagenciesthemselvestoreportanychangesintheirorganisationalstructuresorratingpractices.
Ifanychangesaffectanagency'scredibility,NRSROdesignationmaybewithdrawn.
InCanadaandSwitzerland,theauthoritieswouldinvestigateifevidencebecameavailablethatanagencywasmakinganincreasingnumberofratingerrors(inthecaseoftheformer),oritsperformancewastrulyquestionable(inthecaseofthelatter),whichwoulddeterminethefutureoftheagency'srecognition.
Insomecases,thedifferencebetweenthosereportingyesandthosereportingnowithacaveatmaybelittlemorethansemantic.
Noneofthenon-BCBSmembersproactivelymonitorratingagencies'performance.
TheHKMApointoutthatalthoughtheydonotundertakeongoingmonitoring,theywillreviewtherecognisedstatusofanagencyifitcomestotheirattentionthatthereisamarkeddeteriorationintheperformanceand/ormarketacceptanceoftheagency.
TheAustralianstakeasimilarposition.
Theuseofmarketacceptanceasarecognitioncriterionmayactasasubstituteforongoingmonitoring:ifanagency'sperformancedeterioratesmarkedlyitsmarketacceptancewilldeclineandatsomepointtheagencywouldfailtomeetthecriteriaandloseitsrecognition.
Althoughtheeffectivenessofthismaydependonhowoftenbankingsupervisorsrevisitwhichagenciesmarketparticipantsuse.
51Table4:TherecognitioncriteriainvariouscountriesCountryWhataretheCriteriaforEligibleAgenciesBCBSMembersBelgium"Theratingagencieswererecognisedonthebasisofageneralassessmentoftheircredibility(interaliatheireffectiveusebyBelgiancreditinstitutions),anexaminationoftheirorganisationandmethodology,withparticularattentionfortheirintegrityandindependence.
"Canada"Althoughnoratingagencycriteriaissetinourguidelinesthelistchosenrepresentsthosefirmsthatareinternationallyrecognisedforhavingaccurateandfairratingswhilemaintainingalonghistory.
"FranceInformationnotinthepublicdomain.
Italy"Theagenciesarechosenonthebasisoftheircredibility,objectivity,transparencyandroleplayedintheItalianmarket.
"Japan"Ratingperformance,managementstructure,organisation,ratingmethodologyandindependencefromcapitalstructure.
"Luxembourg"Themaincriteriaaretheagencies'internationalrecognitionandtheirmarketpresence.
"Sweden"Nospecificcriteriahasbeendevelopedregardingwhichratingagenciesareeligible.
However,wewouldonlyallowanyofthemajoragenciesthathasalongstandingtrackrecordandwhoseperformancecouldbecheckedifneeded.
ThethreeratingagenciesthatareconsideredtobeeligiblearetheonesthattheSwedishbanksusewhentheratingtheirdebtinstruments.
"Switzerland"Objectivity/experience/reputation/independence/coverageofcounterpartieslocatedinEurope,NorthAmerica,andJapan.
"UK"Marketrecognition.
"USA"PotentialNRSROsmustmeetthefollowingcriteria:RecognitionasbeingcredibleandreliablefromthepredominantusersofsecurityratingsintheUS.
Abilitytooperateindependentlyofeconomicpressuresorcontrolofthecompaniesbeingrated(e.
g.
havingsufficientfinancialresources)Employinganadequatenumberofstaffmemberswiththeeducationandexperiencenecessarytocompetentlyevaluateanissuer'screditUtilisingsystematicratingproceduresdesignedtoproducecredibleandaccurateratingsPractisinginternalcomplianceprocedurestopreventthemisuseofnon-publicinformation"Non-membersoftheBCBSArgentina"Inthecaseofthecentralbankregistry,therearetwoconditionsforparticipation:(a)theratingagencymustprovideratingsinatleasttencountriesandfiveofthemmustbeinLatinAmerica;(b)theirreportmustfollowguidelinessetbythecentralbank.
Incaseoftheregistryofthestockmarketwatchdog,therequirementsare:(1)tobeorganisedasalimitedcompany;(2)thesoleobjectofwhichiscreditratingactivities;(3)aminimumcapitalofUSD250,000;(4)disclosureofshareholders'names;(5)getapprovalforratingproceduresandmanualsfromthewatchdog;(6)itsdirectorscannotbeshareholders,directorsoremployeesofthosecompaniestheyrated.
"Australia"Wehavenoformalcriteriafordeterminingacceptanceofratingagencies.
Basically,welookatstatusandacceptanceofagenciesinthemarketandbyotherregulators(thisisnotedaboveespeciallyrelevantwherearatingagencydoesnotoperateinAustralia)andtheperformanceoftheagency.
PerformanceismorerelevantintheAustraliancontextwherewecansometimescompareassessmentsandmonitorperformanceofinstitutionsrelativetotheirratings.
"ChileCoverage,reputation,historyandpositioninthemarket.
HongKong"Thecriteriaforassessingtheeligibilityoftheregulatoryagenciesinclude;(i)ownership,historyandbackgroundfortheagency;(ii)methodologyandcoverageoftheratings;(iii)marketacceptance;and(iv)regulatoryacceptanceanditspurpose.
"52Table5:HKMA'sminimumacceptableratingsforthreepurposesRecognisedagencyPurposeLiquidityratioDiscountwindowLenderoflastresortLong-termShort-termBankissuerNon-bankissuerMoody'sInvestorService,IncA3Prime1A3A2Baa3Standard&Poor'sCorporationA-A-1A-ABBB-IBCALtdA-A1A-ABBB-ThompsonBankwatchA+TBW-1A+AA-BBB+R&IA+a-1+A+AA-BBB+Howaredifferencesinratingsbetweentheratingagencies(i.
e.
splitratings)handledPerhapsunsurprisinglygiventhatsevenoutofelevenBCBSmembercountriesuseratingagencies'ratingsintheirbankingsupervisionforjustthestandardisedmethodologyofthemarketriskamendment(orCADiftheirEuropean),virtuallyallrespondentsfromBCBSmembercountriesreportedthecriteriaforaqualifyingitem.
Thisstatesthattheinstrumentmustbe"ratedinvestmentgradebyatleasttwocreditratingagenciesspecifiedbythenationalauthority;orratedinvestment-gradebyoneratingagencyandnotlessthaninvestmentgradebyanyotherratingagencyspecifiedbythenationalauthority".
16TheexceptionwastheUS,wheremostregulatorsadoptanexplicitpolicy,acceptingeitherthehighestorsecondhighestrating.
17Theresponsesonthetreatmentofsplitratingsweremorevariedamongnon-BCBSmembers.
TheAustraliansreportedthatmostoftheiruseofratingswassetintermsofabenchmark.
Itwasonlyinthecasewhereasplitdisagreedastowhetheranentityisaboveorbelowthebenchmarkthatitwouldcauseaproblem.
Inthesecircumstances,theirprudentialguidelinesallowscopeforabanktomakeacasetoAPRAtodisregardthelowerrating.
InHongKong,aparticularsecurityonlyneedstomeettheminimumacceptableratingofoneofthefiverecognisedcreditratingagenciestobecomeeligiblefortheliquidityregime,thediscountwindowandthelenderoflastresortfunction.
ThelatteraredisplayedinTable5.
Theyareof16TakendirectlyfromtheBCBSmarketriskamendment.
17Incontrast,theNationalAssociationofInsuranceCommissioners(NAIC)usesindependentanalysis.
TheNAIC'sSecurityValuationOffice(SVO)usesitsownjudgementinchoosingeitherthehigherorlowerrating.
53interestbecausetheHKMAdiscountssomeagencies'ratingsrelativetoothers.
Theymapthedifferentratingsassignedbytherecognisedcreditratingagenciesbylookingatthedefinitionstheyuseforeachratingscategoryandbycomparingtheratingstheyassigntosomeselectedcorporations.
Theyfollowthestandardtreatmentformarketrisk.
InthedisclosureregimeinNewZealand,abankisrequiredtorevealallratingssothatanydifferencesinviewsbetweenratingagenciesaretransparenttoreaders.
Theyfurtherreport"(t)heReserveBankhasdeliberatelyattemptedtoencouragecompetitioninbankanalysis(andnotonlyfromratingagencies…),intheexpectationthatthiswillpubliclyraisediscussiononkeyissuesrelatingtobankfinancialperformanceandposition".
Splitratingsarethereforenotviewedinanegativelight.
IsthereadistinctionbetweenthetreatmentofsolicitedandunsolicitedratingsIntenoutoftheelevenBCBSmembercountriesthatuseratings,respondentsreportthattheirbankingsupervisorsdonotdistinguishbetweensolicitedandunsolicitedratings.
Anumberofcountriesexpressuneaseabouttheuseofratingsbasedsolelyonpublicinformation.
However,onlytheUSrespondentreportsthatsupervisorsarecurrentlyconsideringmakingadistinction.
RespondentsfromLuxembourg,SwedenandtheUnitedKingdomacknowledgeconcernabouttheuseofunsolicitedratings,butreportitisnotaparticularlyactiveissuebecauseoftheabsenceorinfrequencywithwhichunsolicitedratingsareissuedintheircountries.
TheCanadianresponsereportsthat"althoughnotexplicitinourguidance,OSFIbelievesthatonlysolicitedratingsshouldbeacceptable".
Itdoesnothoweverstatehowthisprincipleisimplementedinpractice,giventhatnotallratingagenciessignalwhetheraratingisunsolicitedandthereforebasedonpublicinformation.
Thenon-BCBSmembersechothesentimentsoftheBCBSmembers.
Theydonotdrawdistinctionbetweensolicitedandunsolicitedratings,buthavereservationsabouttheuseofthelatter.
ConclusionOfthetwelveBCBSmembercountriessurveyed,onlyGermanydoesnotusecreditratingagencies'ratingsinitsbankingsupervision.
Oftheremainingeleven,sevenusethemonlyforthemarketriskamendment(orCAD).
Onaverage,BCBSmemberscurrentlyrecognisesixagencies.
Giventherelativelyrecentadoptionofthemarketriskamendment(orCAD),therehasnotbeenmuchchangetothelistofeligibleagenciesovertime.
Mostoftheelevensupervisorsuseobjectivityasoneoftheirrecognitioncriteria.
Theiruseoftheotherfivecriteriaismorepatchy,withtwo–transparencyandinternationalaccess–beingusedrarely,ifatall.
OnlythreeBCBSmembersundertakeongoingmonitoringofagencies.
Splitratingsaregenerallydealtwithusingtheconditionsunderthemarketriskamendmentforaqualifyingitem.
Theredoesnotappeartobeadistinctionbetweenthetreatmentofsolicitedandunsolicitedratings.
54Appendix1Table6:Apartiallistofratings-dependentregulationintheUnitedStates18YearAdoptedRatings-DependentRegulationMinimumRatingHowmanyRatingsRegulator/Regulation1931Requiredbankstomark-to-marketlowerratedbondsBBB2OCCandFederalReserveexaminationrules1936Prohibitedbanksfrompurchasing"speculativesecurities"BBBUnspecifiedOCC,FDIC,andFederalReservejointstatement1951Imposedhighercapitalrequirementsoninsurers'lowerratedbondsVariousN.
A.
NAICmandatoryreserverequirements1975Imposedhighercapitalhaircutsonbroker/dealersbelow-investment-gradebondsBBB2SECamendmenttoRule15c3-1:theuniformnetcapitalrule1982Easeddisclosurerequirementsforinvestment-gradebondsBBB1SECadoptionofIntegratedDisclosureSystem(Release#6383)1984Easedissuanceofnon-agencymortgage-backedsecurities(MBSs)AA1CongressionalpromulgationoftheSecondaryMortgageMarketEnhancementActof19841987PermittedmarginlendingagainstMBSsand(later)foreignbondsAA1FederalReserveRegulationT1989Allowedpensionfundstoinvestinhigh-ratedasset-backedsecuritiesA1DepartmentofLaborrelaxationofERISARestriction(PTE89-88)1989ProhibitedSavings&Loansfrominvestinginbelow-investment-gradebondsBBB1CongressionalpromulgationoftheFinancialInstitutionsRecoveryandReformActof19401991Requiredmoneymarketmutualfundstolimitholdingsoflow-ratedpaperA119120SECamendmenttoRule2a-7undertheInvestmentCompanyActof19401992Exemptedissuersofcertainasset-backedsecuritiesfromregistrationasamutualfundBBB1SECadoptionofRule3a-7undertheInvestmentCompanyActof19401994Imposesvaryingcapitalchargesonbanks'andS&Ls'holdingsofdifferenttranchesofasset-backedsecuritiesAAA&BBB1FederalReserve,OCC,FDIC,OTSProposedRuleonRecourseandDirectCreditSubstitutes1998DepartmentofTransportationcanonlyextendcreditassistancetoprojectswithaninvestmentgraderatingBBB1TransportInfrastructureFinanceandInnovationAct1998199921RestrictstheabilityofnationalbankstoestablishfinancialsubsidiariesA1Gramm-Leach-BileyActof199918Unlessotherwisenoted,theitemsinthetablearereproducedfromthefollowingpaper:Cantor,RichardandFrankPacker.
"TheCreditRatingIndustry"inFRBNYQuarterlyReview,Fall1994;6.
Forotherratings-dependentregulation,refertoSECReleaseNo.
34-39457,FileNo.
S7-33-97;pp.
1-8.
19Highestratingsonshort-termdebt,generallyimplyinganA-long-termdebtratingorbetter.
20IfissueisratedbyonlyoneNRSRO,itsratingisadequate:otherwisetworatingsarerequired.
21Gramm-Leach-BilleyActof1999,TitleI,p.
91.
55PartII.
ComplementarySourcesofCreditInformationSectionA.
RegionallyBasedCreditScoringIntroductionAccordingtotheBaselCommittee'sJune1999ConsultativePaper,therationaleoftheFirstPillarofthenewcapitalframeworkisdrivenmostlybytheperceivedneedtoreplacethecurrentfixedriskweightingswithexternalcreditassessmentorinternalratings,asfaras"sophisticatedbanks"areconcerned.
TheConsultativePapersuggestsintroducingexternalcreditassessmentspublishedbylargeinternationalratingagencies,suchasStandard&Poor's,Moody'sInvestorService,andFitch-IBCA.
Whatevertheadvantagesofsuchdata,theseratingscoulddisplayanexcessivelytightcoverageofcounterpartiesinsomebanks'portfolios,andhencetheywouldnotmuchimprovetheriskweightingframework.
22Itisthereforeworthlookingatothersourcesofcreditriskassessment,suchascentralcreditregisters(CCRs),andscoringbasedoncentralfinancialstatementsdatabases(CFSDs).
CCRsandCFSDsareakindofserviceprovidedbyseveralEuropeancentralbanks,sometimesoperatingjointlywiththebankingsector.
TheCCRserviceinvolvescollecting,processing,managingandreleasinginformationonbanks'creditexposures,whileCFSDsdothesameonthecreditqualityofcounterpartiestowhichbanksareexposed.
Thefirsttwopartsofthissectionanalysesbothservices.
Theanalysiscoverstheinstitutionalschemegoverningthesystems(ownershipandmaintenance);thecomponentsofthesedatabases(datasubjecttoreporting,reportingthresholds,etc.
);andcoverage,ortheshareofcreditexposureonwhichacreditriskassessmentisissued.
Thethirdpartdealswithtechnicalaspectsofdatacollectionandpublishing,andthefourthpartdiscussestheunderlyingratingmethodologiesandmainissuesraisedbylinkingratingsanddefaultstatistics.
ThemainfindingsregardingCCRsandCFSDsinthecountriesstudied23showthatthetwosystemshavecommonfeaturesregardingcoreinformationservices,butlargerdifferencesfornon-coreinformation.
CCRsbasicallyincludeexposuresoncorporateandsovereigncounterparties,buttheydifferastowhetherornottheyincludedataonprivatecustomersorfinancialinstitutions.
Banksarereportinginstitutionsregardlessofthecountryinwhichtheyarebased,butinvestmentservicescompaniesmaybeassociated.
CFSDsfocusoncorporateentities.
Thesizesofdatabasesmayvarygreatly,butingeneral,datacollectionisdesignedtocapture10-20%ofthelargestcounterpartiesweightingfor80-90%ofthetotalexposureoroftotalbusinessactivity.
Intermsofcoverage,CCRs22Thiscouldbeaconcernespeciallyfornon-USandnon-UKbanksgiventheprominentweightofUS-andUK-basedcounterpartiesreportedinratingagenciessurveys,despiteasharpbutrecentriseintheshareofnon-USissuersinthetotalnumberofratedissuers.
23Austria,Belgium,France,Germany,Italy,PortugalandSpain.
56generallycapturethebulkofexposurescarryingcreditrisk(exceptexposureresultingfromfuturesandderivatives)atleastonnon-bankingcounterparties.
Centralbanksoftenusetheirownbranchnetworksandsystemsfordatacollectionandprocessing,butmayoperateinjointventurewiththebankingsector.
BanksarethemainrecipientsofCCRsandCFSDsservices.
Dependingontheservicesprovided,generalortailoredstudiesusingCFSDsdatamaybepubliclyreleased.
Theunderlyingratingmethodologyisacriticalpointinassessingthereliabilityofacreditinformationsystem.
Thesystemsunderreviewincludetheassessmentofcounterparties'abilitiestomeettheirfinancialobligationsinsofarastheyhavebeenoftenoriginallydesignedforappraisingthequalityofbillsdiscountedbycentralbanks.
However,theanalysismayencompassabroaderviewontheoverallsituationofratedentities.
Theanalyticalapproachiseitherjudgmentalorquantitative(dependingonthemodel)butthedevelopmentofquantitativemodelsseemstobeacommontrend,althoughthesemodelsarenotembeddedthesamewayintheratingprocess.
Discriminantanalysisorcomparableapproachesarethemostcommonones,butalternativeapproacheslikeexpertsystemsmaybeobserved.
Whateverthequalityoftheratingapproach,theeffectivenessofthisapproachinacreditriskassessmentprocessdependsontheabilitytolinktheratingsclassestoadefaultprobabilityfunction.
Oncethislinkisestablished,thecreditassessmentsystemhastobesufficientlystable(therelationshipbetweenratingclassesanddefaultprobabilityshouldnotfluctuaterandomly)andsensitive(ratingsmustresponsepromptlytocreditevents).
OnedifficultyencounteredwhencomparingthedifferentratingsystemsderivedfromCFSDsisinthediscrepancyinthelegalandeconomicdefinitionof"default"usedineachsysteminvariouscountries.
ResearchisbeingconductedattheEuropeanCentralBank(ECB)andnationalcentralbanks(NCB)leveltoaddressthisissue.
RegionallyBasedCreditScoringThissectionofthestudyfocusesonsevenEuropeancountriesthathaveimplementedCCRsandCFSDs.
ThecountriescoveredareAustria,Belgium,France,Germany,Italy,PortugalandSpain.
CCRsareownedandmanagedbyEuropeancentralbanks.
CCRscollectdatafrombanksdescribingtheircreditexposuresanddetailsoneachborrower.
Thescopeofthedatadependsonthereportingcountry.
Ataminimum,CCRsincludecorporateborrowersandpublicadministrations.
Thecollecteddataisusedforbankingsupervisionandmonetarypurposesandisfedbacktothereportingbanks.
CFSDsaregenerallyownedandmanagedbycentralbanks.
Insomecases,ownershipandmanagementissharedwiththebankingindustry.
Originallyempoweredtodiscounttradebillstorefinancebanks,centralbanksbuiltupcorporateinformationfilestoappraisethecreditworthinessofsignaturespresentedfordiscounting.
Thebankingindustrywaslatergrantedaccesstothesekindoffiles,whichincludecorporatefinancialstatementsandtheratingsassignedtoeachcorporaterecordedinthedatabase.
ThecoverageofdifferentsectorsvarieswitheachCFSD.
Ratingsresultfromeitherjudgmentalorstatisticalmethods.
CFSDscanbeusedtodetermineeligibilityofcorporatedebtinstrumentsincentralbanks'bankingrefinancingprocedures,bankingsupervision,andeconomicresearch.
Banksusetheseratingsforassessingcreditrisk.
57CCRsandCFSDsarenationalorshareddatabases.
Eachcountries'coveragedoesnotgobeyondtheirnationalborders.
Withintheirborders,however,thesedatabasesaimatbeingasexhaustiveaspossible.
Inmostcases,theyaremoreexhaustivethanratingagencieswithregardstospecificgeographicalcoverage.
Asshowninthetables,thesecentraldatabasesareoftenmonopolies,eitherownedbygovernmentagencies/centralbanks,orbyprivateoperators.
Inthisrespect,centraldatabasesworkunderamoreorlesscompulsoryregime,bywhichcontributorsmustabide.
Centraldatabaseshaveadualrole-forthecentralbanksandthebankingsystemasawhole,andsometimesforadditionalclaimholderslikeinsurancecompaniesorothernon-bankingfinancialcompanies.
Furthermore,theydifferfromratingagenciessincetheyarenotfuelledbydataprovidedbyborrowersorissuers,butbypublicinformationorprivatedataspecificallyreleasedtothecentraldatabasesbycreditorsthemselves.
ItisalsoworthwhiletonoteCreditAssessmentAgencies(CAAs),whicharediscussedinmoredetaillaterinthissection.
CAAshavemuchincommonwithCFSDsandratingagenciesinthattheirbusinessisbasedoncollectingfinancialinformationoncompanies.
Beyondthis,thepurposesandorganisationofCAAsaredifferentfromthoseofCFSDs.
CAAsareprivatebusinessesthatsellinformationorprovideconsultancyservicesregardingcreditmanagementandcashcollection.
Generally,theyassignratingsbasedonproprietarymethods,buttheseareonlyapartofawholesetofcommercialservices.
Manyoftheirclientsarecommercialandindustrialcompaniesthatneedtoassessthecreditworthinessoftheirownclientsandvendors.
CAAsdonotoperateaspublicormutualinformationprovidersascentralbanksdo,especiallyvis-à-visthebankingindustry.
ThisstudypartlyaddressestheissueofCAAs.
Asprofit-drivenorganisations,CAAscannotoperateasofficialormutualinformationproviders,anddonotbenefitfromalegallybindingstatus.
Asaresult,theyonlyrelyonpublicdata,whichcanbeboughtonthemarket.
Theydonothaveaccesstonon-publicdata,whichmaybereleasedtoacentralregister,especiallybankingexposuretonon-financialcounterparties.
Inthisrespect,centralregistershaveaclearadvantageintermsofindependenceaspublicormutualorganisations.
CentralCreditRegistersOwnershipandmaintenanceofCCRdatabasesCCRsaremostoftenownedandmanagedbycentralbanks.
Thesesystemshavebeenimplementedforalongtime(establishedin1946inFrance,1963inSpain,1964inItaly,1967inBelgium)andhavelegallybindingstatuses.
ThegeneralgoalofCCRsistoenhancethetransparencyofbankingactivitiesandmaketheseactivitiessecure.
Giventhenecessarycloserelationshipwithcreditinstitutions,theconcernedEuropeancentralbankswererequiredbylaw(orprovisionsofspecificbankingregulations)tocollect,processanddisseminateinformationregardingcreditexposure.
France'sCCR(theServiceCentraldesRisques)wasestablishedbytheBankingRegulatoryCommittee24inMarch1946.
InGermany,theprovisionsgoverningtheCCRarefoundinsections2(2),14,19and20ofthe24Thetitleofthebankingregulatorwas"NationalCreditCounsel"atthattime.
58BankingAct.
TheSpanishCCRwasestablishedbyaDecree-Law(specialgovernment'srulingenforcedasalaw)inJune1962.
InBelgium,thelawof1967establishingtheCCRisintegratedinsectionVI,article91,oftheBankingSupervisionLawofMarch1993,andthepracticalaspectsaregovernedbytheRoyalDecreedatedDecember1994.
TheItalianCCRwasestablishedin1964underadecreeissuedbytheInterministerialCommitteeforCreditandSavings.
Thelegalbasisofthatdecreewasthe1936BankingLaw.
Themovewasconfirmedbya1994decreebasedonthe1993BankingLaw.
ComponentsThecomponentsofCCRdatabases(systemcontributorsandparticipants)aredetailedinTablesItoIV(pp.
77-80).
TheCCRsofthecountriesreviewedinthissectionallhandlereportsonfinancinggrantedtocorporates,andtoatleastashareofpublicadministrations.
Beyondthiscommonpurpose,theEuropeanCCRsdisplayseveraldifferences.
CCRs'elementarydataincludethenecessaryinformationforidentifyingborrowerssuchasnationalidentificationnumber,nameand/ortrademark,address,economicsector,andgeographicalzoneforforeignborrowers.
Asfordataoncategoriesofborrowers,corporatesareincludedinalltheCCRswithoutexception.
Publicadministrationsarealsoincluded,butsomeCCRshaveintroducedlimitstothescopeofreportingonsuchorganisations.
ThisisthecaseinGermanywheretheCCRonlyincludesforeignpublicauthorities.
InAustriaandBelgium,centralgovernmentbodies'exposureisnotreported.
However,thegapthatresultsfromtheselimitsisnotmaterialsincethemajorityofCCRsdonotincludenegotiabledebtinstruments,whicharetheonlyfinancinginstrumentsusedforcentralgovernmentfunding.
Nonetheless,theabsenceofthelocalgovernmentbodiesintheGermanCCRmayimplystrongerconsequencesinthereportinggapbetweenthesystems.
Exceptforthesetwocategoriesofborrowers,theCCRsshowsomediversity:householdborrowingisreportedinBelgium,Italy,andPortugal.
SomeindividualborrowersarereportedintheFrenchCCR,butonlyintheeventofindividualentrepreneurship.
Includinghouseholdsmaycreatesomedifficulties:thenumberofreportedloanscouldmakeupthebulkofallthereportedfinancingsintheeventofalowreportingthresholdintermsofloanamount.
Insuchcases,themainproblemsare,ontheonehand,handlingahugedatabaseand,ontheother,findingreliableandcomputablepublicinformationtoappraisetheseloans.
OtherwisetheCCRisonlyusedtoassessglobalamountsofloanstoprivatecustomers.
Ingeneral,suchheavyreportingasaCCRisnotnecessaryonlytoassesssectoralamountsofcreditbutitshouldalsosupportcreditqualityappraisal.
Thelatterisdifficulttoachieveregardingprivatecustomersbecauseofthelackofrelevantdataandoftheregulationprotectingprivacy,whichismorestringentwhenhouseholdsareconcernedthanprofessionalsandinstitutions.
Forexample,inFrance,collectingdataonindividualsissubjecttoaformalapprovaloftheNationalCommissionofInformationTechnologyandFreedom(CommissionNationaleInformatiqueetLiberté)whichverifiesthattheprivacyprotectionregulationisproperlyenforced.
ReportingoncreditinstitutionsalsovariesamongEuropeanCCRs.
SomedatarelatedtoexposureoncreditinstitutionsisavailableinAustria,Germany,France,andItaly,butthescopeoftheinstitutionsreportedonisnotfullycomparable.
InFrance,theCCRreportsonforeigncreditinstitutionswithexposurebeyondoneyear.
TheGermanCCRisthemost59expansivedatabaseintermsofreportingoncreditinstitutions.
Itincludesexposureofbanks'generalmanagersandtheirfamilies,theexposureofpubliccreditinstitutions,andinterbankloanswithmaturityofupto90days.
TheGermanCCRbeganreportingoninterbankexposurein1996(FifthAmendmenttotheBankingLaw).
ReportingthresholdswiththecategoriesofborrowersdifferamongexistingCCRs.
Whendenominatedineuros,standardreportingthresholdsrangefromzeroinPortugaltoEUR1,533,876inGermany.
TheaveragethresholdvalueisaboutEUR300,000.
ExcludingGermany,whichhasanaveragethresholdfivetimestheoverallaverage,theaveragethresholdismorerealisticallyataboutEUR90,000.
ThesedifferenceshavebigconsequencesontherangeofthenumberofreportedentitiesintheCCRs.
However,thegapbetweentheGermanCCRandtheothersispartlyfilledsincethethresholdofEUR1,533,876coversgroupsof"combinedborrowers",definedasnotonlyborrowersconsolidatedbythesameparentcompany,butalsothosesubjecttoextendedfinancialortraderelationshipthatentailsamutualorone-sideddependence.
Thisextendeddefinitionof"borrower"isconsistentwiththatoftheprudentialregulationof"largeexposure".
Moreover,somecountrieshaveintroducedspecificreportingthresholdsforeitherforeigncounterparties(Spain),impairedloans(Italy),orexposureofforeignbranchesofdomesticbanks(Belgium).
(SeeTableIV,p.
81.
)ThemostcommonfrequencyforreportingamongtheCCRsunderreviewismonthly,exceptfortheGermanCCR,whichisonaquarterlybasis.
ThetypesoffinancinginstrumentsreportedoninalltheCCRsunderreviewincludeloans(drawnamount),committedcreditlines(undrawnportion),andgrantedguarantees,exceptforBelgiumandPortugal,whichdonotcollectdataonguarantees.
Notsurprisingly,gapsbetweenCCRswidenasthedistancebetweentheinstrumentcategoriesofthecorecreditportfolioincreases.
TheGermanandtheItalianCCRshavethemostcomprehensivesetofreportedinstrumentsandadditionalinformation:creditequivalentamountofderivatives,eventsofborrowers'bankruptciesandmaturityofmortgageloans(Germany),impaired/delinquentloans25andrisk-mitigatingguarantees(Italy).
ImpairedloansarealsoreportedinSpain,whereastheFrenchCCR'sdefaultreportingframeworkincludesbankruptcycasesandeventsofpaymentdefault.
TheSpanishCCRincludesfixed-incomesecuritiesandtheFrenchsystemhasdataonsecuritisedloans.
26AlltheCCRsstudiedherehavecommonfeaturesasfarasthecorereportinginstitutionsareconcerned,namelythedomesticcreditinstitutionsheadoffices.
TheremainingcontributorsdifferfromoneCCRtoanother.
Thedomesticcreditinstitutions'headofficesmakeupthebulkofthecreditexposurereports.
Inaddition,theforeignbranchesofthedomesticbanksaregenerallyincludedwithinthescopeoftheregisters.
Moreover,allEuropeanCCRscollectdatafromEuropeanbranchesofforeignbankinggroups.
TheGermanCCRisprobablythemostextensiveregisterintermsofreportinginstitutionsduetotheFifthAmendmenttotheGermanBankingLaw,which25Reportingthresholdsarenotapplicablewithrespecttoimpairedloans,whichhavetobecompletelyreported.
26Itisanindicatoroftheexistenceofsecuritisedloansbutnotacomprehensivereport.
60extendsthecompulsoryreportingtoinvestmentservicesinstitutions.
LeasingcompaniesreporttotheCCRonlyifthesecompaniesarebankingsubsidiaries.
Additionally,theGermanCCRincludesreportsfromdomesticbankingsubsidiariesincorporatedabroad,andthosefrominsurancecompanies.
TheSpanishCCRistheonlyregisterwhichextendsreportingobligationstotheCentralBankandtodepositinsuranceandmutualguaranteefunds.
From1997,theItalianCreditRegisteralsoincludesloansfromleasingandconsumercreditcompanies.
CoverageratioThenumberofreportedborrowersvaryfromoneregistertoanother(seeTableII)asaresultofthedifferencesinreportingthresholdsandstatusofconcernedborrowers.
However,someindicatorsareusefulinapproximatingtheregisters'coverage.
Hence,giventhereportingthresholdsofeachcountry,onecanassumethatthemajorityofthemostimportantborrowers(carryingthemajorpartofbankingfinancing)areproperlydescribedintheregisters.
Furthermore,thepercentageofthereportedfinancingtotheamountofassetsheldbythebankingsystem(forthesamecategoriesofassetsasintheregisters)generallyexceeds75%,dependingonthecountry.
InFrance,theexposuresoncorporatesreportedintheregistercoverabout80%ofthetotalcorporatefinancing(90%forindustrialandcommercialcompanies,seeTableV,%ofoutstandingcredit);theregisterdoesnotcoverthe"verysmallbusinesses"(individualentrepreneurs,smallcorporates)ofwhichindividualloansmaynotexceedthereportingthreshold.
ThetotalvalueofexposurereportedintheFrenchCCRrepresentsaboutEUR550billion,27whichmakesupthebulkofcorporatefinancing.
Hence,whenaddedtotheapproximatelyEUR380billioninhouseholdfinancing,thetotalcorrespondsto90%oftotalbanks'loanstonon-bankingcounterparties.
InGermany,thetotalreportedvalueofexposureisaboutEUR4,000billion(derivativescreditequivalentexcluded),orabout90%ofGermanbanks'totalinterbankfundingandcorporate/privatecustomersfinancing.
TheSpanishregisterreportsatotalamountofEUR900billion,whichrepresentanequivalentproportionand80%ofthetotalfinancingisidentifiedintheItalianregister.
TheAustrianCCRcoversabout70%ofoverallbanklending.
TheCCRsdisplayasignificantshareofthetotalbankingloans,atleasttonon-financialcounterpartiesinmostcountries.
Asamatteroffact,coverageratiosseemsufficientenoughforusingcreditregisters.
Moreover,itiscrucialtolinkeachidentifiedcounterparty(orgroupsofsimilarcounterparties)toanavailableandreliablecreditriskassessment.
Itdependsmainlyonthequalityofthefinancialstatementsdatabasesandonthatoftheprevailingcreditriskassessmentmethodologies.
CentralFinancialStatementsDatabasesTheCFSDsshowadifferentoverallpicturefromtheCCRs.
Indeed,collecting,processingandanalysingfinancialstatementsarenotautomaticallysubjecttocentralbanks'monopolies.
27EUR920billion,includingoffbalance-sheetitems.
61(Germanyisamajorexceptiontocentralbanks'monopoliesoncreditregisterswiththeSCHUFAdatabase.
28)ContrarytocreditexposurereportedinCCRs,financialstatementsaregenerallypublicinformation,althoughinpracticetheymaybedifficulttoobtain.
Otherinstitutions,eitherpublicorprivate,maysharefinancialdatacollection/analysiswiththecentralbanks.
Ingeneral,thesituationregardingCFSDsisclosetotheoppositeofthatregardingCCRs.
CCRscollectrelativelysimplebutproprietarydata(lowcosteitheronthecollectingorontheprocessingside)releasedbyalimitednumberofcreditinstitutions.
However,thetheoreticalscopeofreportingentitiestoCFSDsismuchlargersinceitincludesfinancialstatementsissuedbyallentitiessubjecttofinancialreporting(corporates,individualentrepreneurs,independentgovernmentagencies,specialpurposeentities,etc.
).
Allofthesemayencompasshundredofthousandstomillionsofindividuals,dependingonthesizeoftheeconomy,andhundredstothousandsofdifferentsectors.
Asaresult,thepracticalcostofdatacollection,processingandanalysisisgreaterthanforcreditriskexposure.
Thismayexplainwhysomeinstitutionsfocusonlyondatacollectionorprocessing,orlimittheirdatacollectionandanalysistospecifickindsoffinancialreportsissuers.
InthefieldofCFSDs,centralbankscompetewithprivatedatabases,atleastinsomeeconomicsegments.
Thenextsectionwilllookmainlyatthecentralbanks'databases.
OwnershipandMaintenanceofCFSDdatabasesInallthecountriesunderreview,thecentralbanksholdandmanagetheirowndatabases,exceptinItaly,despitedifferentdatacollectionanddistributionprocesses.
TheorganisationoftheItalianCFSDisunique:ratherthanimplementingitsowndatabase,theBankofItalysetupajointpartnershipwiththeItalianBankingAssociationtoestablishthe"CentraledeiBilanci"(CDB).
CDBisincorporatedasaprivatecompany.
Exceptforthecentralbanks'CFSDs,privateCAAsorlocalchambersofcommerceprovidesimilarservices,althoughtargetsmaydiffer.
Theworld-widebusinessinformationgroupDun&Bradstreet(Moody'smainshareholder)hassetupsubsidiariesineachEuropeancountryanditholdsgenerallyalargestakeinthelocalmarketsofbusinessinformation.
InFrance,Dun&BradstreetencountersfiercecompetitionfromCOFACE-SCRL(subsidiaryoftheFrenchexportinsurancegroupCOFACE)andORT.
Eachofthesethreecompaniesdisplaycomparabledataresources(15to20millionfinancialstatementsofEuropeancorporates.
ComponentsandcoverageratioThebasiccomponentsofthesedatabasesarefinancialstatements,enteredeitheronaspecificallystandardisedformatdedicatedtotheCFSDs,oraccordingtothegeneralaccountingrulesandinternalrevenuereportingformat.
Thefinancialstatementissuersareidentifiedinthedatabasewiththeirnationalidentificationnumber,incorporationandtrade28SCHUFAisashareddatabaseestablishedbyretailerstoappraiseprivatecustomers'creditqualityforconsumercredit.
62names,andeconomicsector,whichcanbematchedwiththeCCR'sidentificationdata.
TheCFSDsincludeonlycorporatefinancialstatements.
Theysometimesincludeotherbusinessesunderstatusofpartnershiporspecialpurposevehicles.
However,giventhelackofcomparabilitywiththereportsissuedbypublicadministrationsorinsurancecompanies,thesearegenerallynotincluded.
ThemajorfindingregardingthecomponentsandcoverageofCFDSsisthatthe"20/80"or"10/90"rulegenerallyapplies:fewmediumandlargefirmsaccountforthemajorityoftotalcorporatebankingindebtedness,oftotalstaffemployed,etc.
France'sCFSD,theFIBEN(FIchierBancairedesENtreprises),coversasignificantshareoftheFrenchcorporates.
AccordingtotheFrenchNationalInstituteofStatistics(INSEE),thereareabout2.
1millionFrenchcorporates.
Thesecorporatesarelegallyboundtosubmittheirfinancialstatementstothetradecourtsandthetaxoffice,whichhassetthestandardreportingformatnamedthe"taxwad".
Ofthistotal,corporatesaredividedintotwogroupstax-wise:1.
7millionreportindustrialandcommercialprofits,and400,000reportnon-commercialprofits.
Oftheindustrialandcommercialcorporates,64%,or1.
1million,releasesimplifiedfinancialreportsand36%(640,000)releasecomprehensivedetailedreports.
ComprehensivereportscorrespondmoreorlesstothelargestcompaniesthathaveaprominentshareintheGDP,withthehighestnumberoftotalstaffemployed,andcorporatebankingindebtedness.
FIBENmainlycollectsdataonthecomprehensivereportsforcorporateswithturnovermorethanEUR760,000.
TheFIBENsampleincludesaboutone-thirdofthefirmsissuingcomprehensivereportsandwhichaccountfornine-tenthsofthecorporateindebtednessandoftotalstaffemployed.
Withinthissampleof230,000entities,about50,000release"non-standard"financialstatementsduetothespecificityoftheireconomicsectors,suchaseducation,health,governmentagencies,financialservicesexceptbanking,andrealestate.
Thedatabaseshowsabiasoverweightinginthemanufacturingandtradesectors.
The180,000"assessable"reports(28%ofthecomprehensivereports,9%ofthetotalnumberoffirms)cover45%oftherecordedamountsintheCCR(seeTableV).
FIBENhasalongtrackrecordofmorethan25years.
Itincludesindividualandconsolidatedfinancialstatements.
Thelattergroupoffinancialstatementsisnotincludedintheabovementionedfigures.
InGermany,theBundesbank'sCFSDcoversabout70,000financialstatements(numberrangingfrom50,000to80,000dependingontheyear)withatrackrecordofmorethan25years.
ThetotalnumberoffirmsinGermanyis1.
75million.
TheGermanCFSDcovers60%ofthetotaltaxableturnoverand60%oftheaggregateamountofcorporatebalancesheets.
However,asshowninTableV,thecoverageratioismuchbetterforlargeindustrialfirmsincorporatedintheformerWestGermany.
Consolidatedfinancialstatementsarenotrecordedinthedatabase.
Financialstatementsarereleasedaccordingeithertoa"taxformat"ora"commercialformat".
Thelatterismorebusinessorientedandisreleasedthreemonthsafteryear-end.
Thetaxformatisdelayedbyoneyear,butisasimplerreportandrepresentsthebulkoffinancialstatements.
ContrarytotheFIBENsystem,thesizeoftheGermanCFSD'ssampleisnotdrivenbyafixedreportingthresholdbutdependsondebtandcollateralrulesintheBundesbank'srefinancingprocedures.
Indeed,theeligibilityofcollateralisconditionedbybanks'reportingofcorporatedebtissuers'financialstatementstotheBundesbank.
Otherwise,financialstatementcollectionbytheBundesbankisnotcompulsory.
Asaresult,thesamplesize63fluctuatesaccordinglyandthelargesthighcreditqualitycorporatesareusuallyoverweightedinthissample.
TheorganisationofAustria'sCFSDissimilartoGermany's.
Datacollectiondependsontheeligibilityofcorporatedebtinstrumentsascollateralforthecentralbank'sliquidityinjections.
Asinthepreviousexamples,thelargestmanufacturingfirmsareoverweightedintheAustriandatabase.
However,anadditionalsetoffinancialstatementsreleasedbysmallbusinessesiscollectedbyaspecialisedresearchinstitute,IGH,29andisintegratedintheNationalBankofAustria'sinformationsystem.
ThecoverageratioisnotedinTableV.
InItaly,theCentraledeiBilanci(CDB)includes40,000individualfinancialstatements,lessthan4%ofthetotalnumberofcompanies.
Ofthose,1,300aredetailedreportsfromlargecompanies,and700areconsolidatedfinancialstatements.
Despitethescope,whichexpandsbeyondthemanufacturingsectortorealestate,leasing,andholdingcompanies,themajorityofthestatementsarefromlargemanufacturingfirmsintheindustrialisednorthofItaly.
ThereportingturnoverthresholdoftheCDBisEUR100million,comparedwiththeEUR760,000thresholdoftheFrenchCFSD.
However,thereportsavailablehavemanydetails,especiallytheconsolidationperimetersandshareholdinglinks.
TheCDBwassetupin1983.
AsinAustria,otherinstitutionshavesetupadatabasethatcapturesmorefullythesmall-andmedium-sizedcompanies.
ThelocalChambersofCommerceandlocalcompanieshavesetuptheCERVEDdatabaseasajointstockcompany.
Ithascollectedthefinancialstatementsofabout500,000privatelimitedcompaniessince1993.
CERVED'sscopeismorerepresentativeofItaliancorporatesthantheCDB's,butthereportingframeworkislessdetailed.
AccordingtoanagreementwithCERVED,theCDBalsoprovidesthefinancialstatementsofthecompaniesincludedintheCERVEDdatabase.
InSpain,acomprehensivesetofdatabaseshasbeensetuptocompletetheoriginalCFSD(seeTableV).
Thefirstdatabase,CentraldeBalancesAnnual(CBA),wasestablishedin1983.
Itincludes8,000detailedfinancialreportsfromlargecorporates,whichaccountfor36%ofthenationalaggregateturnover.
Thesereportsareconfidential.
Theframeworkresultsfromadetailedquestionnaireofabout600items(forlargecorporates),or300items(forothercorporates).
TheBankofSpainhasintroducedcollectingsomequalitativeinformation.
Itrequestsfirmstoreleaseastatementrelatedtothesourcesandapplicationoffunds(detailedcash-flowanalysis).
TheCBAisupdatedonayearlybasis.
Anadditionaldatabase,CentraldeBalancesTrimestral(CBT),wassetupin1993.
Itcoversasampleof1,000largefirmsandincludesaquarterlysimplifiedquestionnairemainlyfocusedonP/Litems.
Inordertoexpanditscoveragetowardsmall-tomedium-sizedcompanies,theBankofSpainhascooperatedfor10yearswiththelocalRegistersofCommerce.
Theseregisterscollect300,000financialstatements.
Ofthesestatements,about170,000areusefulforanalysisthroughtheCentraldeBalancesdelBancodeEspaa-RegistrosMercantiles(CBB).
Thefinancialreportsarebasedona300-itemsimplifiedquestionnaire.
29InstitutfürGewerbe-undHandwerksforschung.
64TheNationalBankofBelgium's(NBB)CFSDisuniqueinthattheNBBistheonlycentralbankempoweredbylawsince1976tocollectallfinancialstatements.
Ingeneral,incountrieswheresuchalegalobligationisenforced,thenationalinstituteofstatistics,tradecourts,orchambersofcommercecollectthestatements.
Thus,theNBBholdsanexhaustivedatabasewhichcoversabout200,000firms.
Firmsthatreleasea"simplifiedreport"havethefollowingfeatures:annualturnoveroflessthanEUR5million,balancestatementsofnotmorethanEUR2.
5million,oranaverageoflessthan50staffmembers.
Thenumberofthesesmall-tomedium-sizedcompaniesis190,000andaccountfor95%ofthetotal,butonly15%intermsofbalancestatementsvalue,and20%intermsofvalueadded.
(SeeTableVforbreakdownbysector.
)NoCFSDhasbeenestablishedintheUK,buttheBankofEngland(BOE)performssomespecificstudiesonthecorporatesectoreithertoappraisethesituationofpossiblylargeailingfirmsortoassesssmallbusinesses'accesstobankingfinancing.
TheEUmember-statessetuptheEuropeanFinancialDatabaseCommitteein1985.
Thecommittee'sworkhascreatedtheHarmonisedEuropeanAccountsDatabase(knownbyitsFrenchacronymBACH).
BACHincludescorporatefinancialstatementsfrom13countries30underharmonisedreportingstandards.
ThisdatabaseismanagedbytheEuropeanCommission(GeneralDirectorateII-FinancialInstitutionsandBusinessLaw).
ThecountrieswhichdonothaveaCFSDusenationalinstitute'sstatistics.
ExceptforBelgium,thesamplesincludelimitednumbersofcorporates(seeTableV).
Thedatabaseresultsfromthecommittee'stargettofocusonmethodologicalharmonisationandmappingthecriteriaofeligibilityofcorporatedebtinstrumentsascollateralforcentralbanks'liquidityfunding.
ItisworthwhiletonotethatBACHwouldprobablynotbesufficienttoconductaEuropean-wideindividualcorporateanalysissinceithasnotbeendesignedforsuchapurpose.
However,inthecourseofdefiningcomparableriskassessmentmethodologies,theresultsmaybeuseful.
OrganisationofdatacollectionDatacollectionisdonethroughtwopossiblechannels,non-bankingfirmstoberatedandspecialisedcollectorsoffinancialdata.
TheBundesbankandtheNationalBankofAustriacollectcorporatefinancialstatementsonacompulsorybasiswhenbanksintendtodiscounttradebillsortocollateralisecorporatedebtinstruments.
Inthisevent,centralbankfundingisconditionedbyreportingthefinancialstatementsofthecorporatedebtissuerstoCFSDs.
Bothcentralbanksreviewthedataqualityandanalysis.
Inordertoenlargetheirdatacollectionbases,theyhavetiedlinkswithexternaldataprovidersspecialisedinsmall-tomedium-sizedbusinesses.
TheBundesbank'snetworkisinvolvedindatacollection,theStatisticsDirectorateperformsdataprocessing,andtheCreditDirectorateisinchargeofanalysingthefinancialstatements(seeratingmethodology).
TheBankofFrance(BOF)andtheNationalBankofBelgium(NBB)bothcollectfinancialstatementsdirectlyfromtheirissuers.
Asmentionedabove,theNBBistheonlycentralbank30Austria,Belgium,Denmark,Finland,France,Germany,Italy,Japan,Netherlands,Portugal,Spain,SwedenandUnitedStates.
65legallyempoweredtocollectallcorporatefinancialstatements.
TheBOFusesitsnetworkof211branchestocollectdata.
Thisapproachhastheadvantageofkeepingadirectlinkwithreportingfirms,andthereforeitenhanceslocalanalysts'knowledgenotonlyofthefirmsthemselvesbutalsooftheirspecificlocalenvironment.
Italsoinfluencesfirms'willingnesstocooperate,sinceBOFbranchescantailorthecontentsthroughselling"value-addedanalysis"fromtheCFSD.
However,thedrawbackofthisapproachisthevolumeofresourcesthatwouldberequiredtoexpandthescopeofreportingcorporatestothosewithlessthanFRF5millioninturnover.
TheBOFdatabasemanages200,000firmswithabout1,000staffmembers.
Datacollectionandanalysisaredecentralisedactivitiesperformedbylocalbranches.
TheglobaldataprocessingandtheratingmethodologyguidelinesarecentralisedattheCorporateDirectorateoftheBOF.
TheBankofSpain(BOS)andtheBankofItaly(BOI)havedelegateddatacollectiontoajointsubsidiary(BOI)ortodistinctinstitutions(BOS)properlyequippedforthispurpose.
TheBOS'totalstaffis3,200anditslocalbranchesarelimitedto53.
Thus,thestaffdedicatedtocorporateanalysisislimitedtoabout50peoplewhofocusondataprocessingandanalysis.
TheBOI'schoiceofajointsubsidiary(CDB)withtheItalianBankingAssociationfostersatransparentpricingpolicyofdatadiffusion.
However,theseinstitutionsdependontheiragreementswithotherpublicadministrationsorprivateoperators.
TechnicalaspectsofdatacollectionDatacollectionisgenerallyperformedthroughdigitaldevicesbutsometimesnotcompletely.
TheBOSandtheBOIhaveimplementedafullydigitaldataflowwiththeirinformationproviders.
Itisonepossibleadvantageofusingaproxyforcollectingdata:itmaybemoreconvenienttosetadigitalflowwithonecounterpartythanwiththousands(ofbanks)orhundredsofthousands(ofcorporates).
Forexample,theBOSusesanopticalcharacterrecognitionto"digitalise"theoriginalRegistersofCommerce'sreports.
AttheBundesbank,two-thirdsoftherawinformationiscollectedthroughdigitaldevicesanditplanstoimplementremotedatatransferfrom"machinetomachine"oron-lineprocessing.
TheBOFhasalreadyimplementedsucharemotedatacollectionsystem.
DatabaseUsersCentralBanksarethemainusersoftheirowndatabases.
Theirprimarypurposeforalongtimehasbeenassessingtheeligibilityofbillsandcorporatedebtinstrumentsincentralbanks'fundingprocedures.
ItremainsthemainpurposeoftheGermanandAustriansystemssincetheseinstitutionsdonothaveanydirectsupervisoryresponsibilitiesforbanks,asinBelgium.
However,thesecentralbanksshareinformationwiththeBankingSupervisionAuthorities(BSA)31toprepareon-siteexaminationsandperformbankinganalyses.
InFranceandSpain,thepurposesaretheeligibilityassessmentofbillsanddebtinstruments,riskanalysisforbankingsupervisionandresearchpurposes.
InItaly,CCRdataandfinancialstatementdataareusedforresearchandsupervisorypurposes.
Forsupervisorypurposes,BSAshaveto31TheFederalOfficeofBankingSupervision(BAK)inGermany,theBankingandFinancialCommissioninBelgiumandtheMinistryofFinanceinAustria.
66matchCCRsandCFSDstoinvestigatebankingportfolios.
Theydothistoeitherlookattheglobalsituationofportfolios,ortolookatindividualcounterparties,especiallyforpreparingon-siteexaminations.
Moreover,theBOFandtheBOIhaveimplementedautomatedtools32forbankinganalysestoprovideexaminerswithhighlydata-intensiveanalysesofbanks'financialandprudentialsituations.
Allcountriesunderreviewperformmacro-ormicro-economicanalysisandresearchusingCCRsandCFSDs.
Banksarethesecondmainusers.
CCRsredistributeinformationonexposuretoenableeachbanktoappraiseitsshareinthetotalbankingindebtednessofitscustomers.
Hence,eachonecanaccessits"marketshare"inlendingactivities.
CFSDsdisclosecorporateratingsandfinancialstatements.
Thecombinationofbothsourcesofdataoffersbanksausefulinstrumentforcustomers'exposuremonitoringandcomparingtheirlendingpolicywiththatofcompetitors.
Moreover,theBOFhasmadeinformationoneligiblehighcreditqualitycorporatesavailableontheInternet.
Allbanksoperatingintheeurozonecanretrievethisinformation,accordingtotheBOF'sandESCB'sliquidityprocedure.
CorporatecontributorstoCFSDs-eitherdirectlyorthroughbanks'reportstoCFSDs-receivetheirownratings.
InFrance,companiesratedbytheBOFhavethelegalrighttoaccessandtomodifyinformationreportedintheFIBENdatabase.
CFSDsmayoffer"valueadded"servicesbeyondtheseminimumrequirements.
Somecentralbankscarryoutextensiveinformationstrategiestowardcorporates,especiallytheBOFandtheBOS.
Excepttheretrocessionoftheirrating,theseCFSDsaddcommercialinformationservicestocompanies.
Theseservicesrangefromdisclosureofdetailedratiostoquasi-consultancyandcontributetocoverdata-collectingandprocessingcosts.
TheBOFhasintensivelyprocessedFIBEN'sdataintocommercialproductssuchasrules-basedexpertsystemdiagnosis("GEODE")andfinancialanalysispackages("ModulesFIBEN").
TheBOSreleasesdetailedratiosanalysisincludingpeer-groupcomparisons.
TheinformationcontainedintheItalianCDBaredistributedonlytomembers'banksandtotheBOI.
RatingmethodologyandstatisticalrelationshipbetweenratingsandcorporatedefaultsPurposesofratingmethodologiesandbestpracticesFirst,beforeanalysingtheratingmethodologies,itisnecessarytopointoutthedifferentpurposesofthesemethodologies.
TheBundesbankandBankofAustriarestrictthepurposesoftheirrespectivecorporateratingstoassigningeligibility/non-eligibilitystatustotradebillsandothercorporatedebtinstruments,asdoestheBOS.
TheBOFaimstomaintainacorporateclassificationsystemthatwouldbeusefuleitherforeligibilityassignment,theprudentialreviewofbankingportfolios,orforinformationonbanksforinternalriskmanagementpurposes.
TheBOI'scorporateclassificationsystemhasbeendevelopedbytheSupervisionDepartmentforresearchpurposes.
Itisalsousedforprudentialanalysis.
Nonetheless,whatevertheeffectiveuseofin-housecorporateratingsassignedbycentralbanksthrough32TheBOFhasimplementedtheSAABAsystem("BankingAnalysisSupportSystem")andtheBOIthePATROLsystem.
67theirCCRsandCFDSs,thehistoricalbaseofallsystemsistheeligibilityclassification.
ThecentralbankswhichparticipateintheEuropeanSystemofCentralBanks(ESCB)arediscussingwaystheeligibilityproceduresshouldbeharmonised,oratleastproperlycomparedwithintheESCB.
Creditorcorporatedirectoratesoftheparticipatingcentralbankshavecarriedoutguidelinesforbestpracticesincorporateriskassessment.
Hence,priortodescribingtheeffectivemethodologies,weshallpresentasummaryoftheseguidelines33thatcouldpossiblyinfluencetheplannedsupervisoryframeworkofexternalratings.
Thein-houseriskassessmentpracticesintheESCBaimatselectingthehighestcreditqualityassets,or"eligibilityfortier-twocollateral".
Inaddition,centralbanks'in-houseratingdepartmentsdonothavecontractualrelationshipwithratedfirms.
Objectivesoftheanalysisandmaincharacteristicsof"soundenterprises":ratingsresultfromtheassessmentoffirms'abilitiestomeetfinancialobligations,takingfirms'individualandsectoralrisksintoaccount.
Accordingtothisobjective,thescopeofanalysisincludesthefollowing:businesscharacteristics,financialanalysis,managementevaluation,industryandeconomicenvironment.
Thisgeneralpatternisconsistentwiththatofratingagencies.
Themaincriteriaforeligibilityarethefollowing:afavourablemarketpositionthatentailslargecash/earningsgenerationinproportiontofinancialexpenses,supportedbyaconservativecapitalstructureandsecuredaccesstoexternalfinancing.
Howtheanalysisshouldbeconductedinpractice:NCBs'guidelinessuggestthefollowingstepsofanalysis:–marketpositionandbusinesscharacteristics:strengthsandweaknessesincompetitiveposition(marketshare,businessstrategy,technologyetc.
),legalstatus,ownershipstructureandrelationshipwithaffiliatedcompanies,–financialanalysis:qualityoffinancialstatements,ratiosanalysis(intrinsicanalysisandpeergroupcomparison)regardingthefollowingitems:operatingcash-flow/externalfinancing,operatingcash-flow/financialexpenses,shorttermdebts/liquidassets,longtermdebts/fixedassets,levelofprofitability(margins,productivity),attractivenessofbusinessforequityinvestors(costofcapital),financialflexibility(paymentbehaviour,accesstobankingandmarketsfunding),–generalmanagementevaluationandtrackrecord,–industryandeconomicenvironment:firm'sabilitytosustainbusinesscyclesandchangesintechnologicalenvironment.
Thisanalyticalframeworkisprocessedthroughcomputationaltools.
Theresultsarethensubjecttoafinalandconclusivecreditassessmentthatmustbeissuedbyanindependent33Thefinalversionoftheseguidelinesisstillbeingdiscussedandaconclusivereportisexpectedtobedisclosedinlate2000.
68creditcommittee(orequivalent)composedofthebank'sseniorexecutives.
Theassessmentmustensuretheobjectivityoftheanalysisandthesoundnessoftheanalyticaltools.
TheexistingNCBs'ratingsystemsaregenerallyconsistentwiththeabovementionedguidelinesasshowninthefollowingexamplesofFrance,Germany,andSpain.
BankofFrance'sratingsystemandmethodologyAmongthefirmsrecordedintheBOF'sFIBENdatabase,180,000companiesreleaseacompletesetofinformation(financialstatements,identificationofshareholdersandmanagers,etc.
),enablinglocalbranchestoassignacomprehensiverating.
Theshareofthese180,000corporatesinthetotaloutstandingbankingcredittocorporatesisabout45%ofthetotaloutstandingcreditidentifiedintheBOF'sCreditRegister.
BOFanalystsalsouseconsolidatedfinancialstatementscovering2,500groups,informationonownershipstructures,andreportsonpaymentdefaults.
TheBOF'sratingsystemisbasedonacomprehensiveassessmentwhichisdisclosedthroughathree-positioncode.
Eachpositionstandsforthesize,theoverallcreditquality,andtheregularityofpaymentstovendorsandbanks.
SeeTableVI(p.
83)fortherating"notches"andrationaleforeachone.
Thecreditqualityratingiscompletedbyamanagementqualityratingassignedtoindividualsactingasgeneralmanagersandbasedonpublicinformation(seeTableVI).
Asupplementary"transparencyindicator"("T")isawardedtocompaniesthathaveagreedtoanexchangeofinformationwiththeirbankcreditorsandthat,accordingtothatagreement,havegiventhemupdateddataontheirfinancialpositionandfutureprospects.
Thetransparencyindicatorisindependentfromthecreditquality,butitisnotcompatiblewithan"unknowncompany"rating("0"rating).
Anotherindicator("R")pointsoutcompaniesthatdonotcomplywiththelegalfinancialdisclosureframework,suchaslatefillingoffinancialstatements.
TheBOF'sratingsresultfromastrictlyjudgmentalapproachfollowingtheabovementionedrationale.
Therationaleisdetailedandstandardisedinareferencemanualof200pages,with5updatescarriedoutsince1994.
Thecomputationaltoolsareusedasdecision-makingsystems,butthesetoolsdonotreplacetheanalyticalframeworkitself.
Theratingoffersanoverallindicationofallthecollecteddataonacompany.
Theanalysisisperformedsoastoelaborateanappraisalofthecompany'sshort-termsoundnessoveraoneortwoyearperiod.
Thequalityandtheobjectivityoftheratingisensuredbythesignificantvolumeofhumanandtechnicalresourcesallocatedtocompanies'ratings:1,000peopleonstaffdedicatedtodatacollectionandprocessing,andratinganalysis;theBOF'smainframeresources,includingautomaticcontrolandwarningdevices;controlofratinganalystsbybranchmanagersandinternalaudits,andsystematictrackrecords.
Inaddition,theBOF'sindependencevis-à-visanythirdpartyisguaranteedbylaw,accordingtotheEMUstandards.
Aratingmanualisalsoavailable.
69Bundesbank'sratingmethodologyTheGermansystemreliesmoreonautomatedprocessesthantheFrenchone.
Thisaspectwillbeanalysedinmoredetaillaterandonlythestandardisedanalyticalapproachispresentedatthatstage.
Thefirststepoftheratingprocedureconsistsofdrawingupafirm'sfundsstatement,whichisderivedfromthepublicfinancialstatements.
Cash-flowisbrokendowntotracktheuseofturnoverflows,capitalexpenditureanddisinvestment,andtheresultingsurplus/deficit.
Theresultisakeyindicatortoassesssoundnessoffirms.
Thesecondstepispeer-groupcomparison.
Eachpeergroupissub-dividedbytranchesaccordingtovolumeofturnover.
Then,theanalysisisconductedthroughanautomatedprocedureusingalistofratiosreportedinTableVI.
Despiteanextensiveuseofcomputerisedanalysis,thefinalcreditassessmentismadebytheBundesbank'sbranchmanagers,whoarerequestedtoaddanydatanotavailableintheautomatedsystem.
Theyarealsorequestedtocheckthattheresultsareplausibleandtakeanyspecificityintoaccount.
Theseratingsarethendisclosedtosubscribingbanksandtotheratedfirms.
BankofSpain'sratingmethodologyTheBOSstartedperformingin-houseriskassessmentin1997.
Themainobjectivewastoenlargethesetofeligibleassets.
Asamatteroffact,internationalratingagenciesassignratingstofewSpanishcompanies,evenamongthemostimportantones.
TheBOSholdsanextensivedatabase("CentraldeBalances")including170,000comprehensivefinancialstatementsandhasperformeddetailedanalysisforalongtimebutwithoutsettingaformalassessmentsystemuntil1997.
Atthepresentstage,theBOS'screditanalysisdepartment(OperationsDepartment)hasrestrictedthescopeofanalysisto48largefirms(mostofthemissuingthemostliquidsharesandsecuritiesontheSpanishmarkets).
Theanalyticalapproachisonlyjudgmentalanditisbasedonaready-to-assessreportwhichsummarisestherelevantinformationcombiningdatafromtheCCRandtheCFSD.
Thissummaryincludesthefollowingitems:Financialstatementswithapreliminaryratiosanalysis:balancesheetstructureandpeergroupscomparison,estimatedcurrentlevelofprofits.
Lastyearsmaterialchangesinthecompany'ssituation(capitalstructure,businesslines,mergers,strategicalliances).
Ratiosoverthelastfouryears(profit/netassets,capitalrecoveryratio,financingcost,structureofindebtedness).
Risksdata:concentrationofbankingexposuresonthecompany,eventofbankruptcy.
Auditors'reportsandexternalratings.
Theanalysisisbasedonalong-termperspective(rather"throughthebusinesscycle"than"pointintime")andisconcludedbyafourpossiblecreditqualityassignment:levels1and2correspondto"highcreditquality"and"goodcreditquality",level3correspondstoa70"dubiouszone"whichwarnsaboutlackofrelevantinformation,andlevel4correspondstoriskyfirms.
Thefinalratingisdecidedbyacreditcommitteeboundtoafullconsensus.
TheratingisthenendorsedbytheBOS'sExecutiveCouncil.
TheBOSintendstoextenditsworktowardstatisticalanalysisofbankruptcyriskoverasampleofabout1,000companies.
AnalyticaltoolsandcomputationtechniquesTheratingmethodologiesdescribedabovearecarriedoutwiththeextensiveuseofautomatedcomputationtools.
Thiscanbeseenasleadingtoatrade-offbetweenhumanthinkingandcomputerprocessing.
Alltheinstitutionsunderreviewuseorintendtouseautomatedcomputationatleastforresearch,andpossiblyaseffectivedecision-makingsystems.
Differentmathematicaltechniquesarepotentiallyusefultoimplementautomatedratinganalysis(discriminantlinearanalysis,probit-logitmodels,rules-basedexpertsystems(RBES),neuralnetworks,decision-treesetc.
).
ThemostcommonlyusedtechniquesbytheinstitutionsunderreviewarediscriminantanalysisandRBES.
WeshallfocusonthediscriminantanalysisperformedbytheBundesbank,theBOFandtheBOI,andontheRBESimplementedbythetwoformerinstitutions.
34TheBankofFrance'sscoringmodelTheBOF'sCorporateDirectorate(CompaniesObservatoryDept.
)hasrealisedaprobabilisticcorporatescoringmodel,the"BDFSCORE",35basedonastatisticalanalysisoftheBOF'sCFSDFIBEN.
Thisscoringisneitherasubstituteforthejudgmentalratingnorafullyintegratedstagewithinthecourseoftheratingprocess.
Itisratheracomponentoftheanalysts'toolboxforcorporateriskassessment.
ThemodelisbasedonCFSD'sfinancialdata(thesamebasethatisusedforratingprocess)combinedwithpublicdefaultsstatistics.
ThemodelwaselaboratedusingtheFisherdiscriminantanalysisbecauseofitsrobustness,easyinterpretability,probabilisticutilisation,andeasymaintenance.
Thetargetofsuchananalysisistoreachtheoptimumfrontiersplittingfailedandnon-failedcompanies.
Theoptimumfrontierresultsfromthebest-fittedweightedcombinationofrisk-meaningfulratios(theratiosaredefinedinTableVI).
The"rawresults"ofthescoringconsistofanotemoreorlessclosetothethresholddistinguishingthefailedfromthenon-failedcompanies.
Theriskassessment-orientednatureofthemodelisenhancedbycalculatingtheposteriorprobabilityoffailureaccordingtotheBayesiantheorem.
Itenablestheclassificationofcompaniesinriskclassesrelatedtoaprobabilityoffailureexpressedinpercentage.
Thechosentimehorizonisthreeyears.
Themodelisusedtoperformindividualriskassessment,sector-basedriskassessment,risklevelindividualtrajectoriestrackedyearbyyear,andtransitionmatricesinriskclasses.
Thescoringfunctionisavailableinseveralversionsfittedtoeconomicsectors(manufacturing,trade,transportsetc.
).
34Othertechniques,especiallyneuralnetworks,havebeentriedbutarenotintegratedintheeffectiveratingprocess.
35Foratechnicaldescriptionsee"DetectingtheRiskofCompanyFailureattheBankofFrance",M.
Bardos,JournalofBankingandFinance#22,1998.
71GiventheoptimalsizeoftheFIBENdatabaseandthereliabilityofdata,ithasbeenseenthatthescoringmodelisefficientandseveralnewrelatedtoolsareindevelopment.
Inadditiontoindividualcorporateassessmentandeconomicresearch,thescoreshavebeensuccessfullycombinedwithCCR'sriskexposureandintegratedwithinthesupervisoryriskappraisalsystemSAABA.
Inadditiontothescoringmodel,theBOFhasrealisedaRBES("GEODE")usingthestandardjudgmentalrationaleandstatisticallyprovenquantitativeelements.
Theresultsareusedasaninternalcomplementarydecision-makingtoolandaresoldtocompanies'managers.
Bundesbank'sautomatedcreditworthinesssystemTheBundesbank'stoolsaresimilartothoseusedbytheBankofFrance(fordiscriminantratios,seeTableVI).
Nonetheless,theGermansystemisparticularsincethescoringandRBESareformallyintegratedintheratingprocessitselfandconstituteakindof"analyticalpackage".
Thescoringiscomputedonasetofquantitativeandqualitativedata;aspecificfunctionisdesignedforeachofthefollowingsectors:manufacturing,trade,andothersectors.
Theprobabilityoffailureismeasuredbythedistanceofeachcompanytothecut-offpointbetweenfairandriskycompanies.
Thescoringresultsshow20%ofthecompaniesscoredare'neutral'.
StatisticalresultsarethenrefinedthroughtheRBES,whichreducestheshareoftheneutralgroup(i.
e.
undefinedintermsofrisk)downto6%.
TheGermanRBESisdesignedtocopewith"fuzzylogic"andcontributestostandardisequalitativedata.
TheRBESenhancesormitigatesoriginalstatisticalscoresbutcannotcontradicttheinitialresults.
Thefinalratingisassignedonajudgmentalbasis.
BankofItaly'sautomatedcreditworthinesssystemTheBOI'ssupervisiondepartmenthasdevelopedacorporatescoringsystembasedonalogitregression,whichisusedforresearchandforoff-siteprudentialreviewofbankingportfolios.
Itsuseforassessingbanks'internalratingsystemsisunderconsideration.
Centraledeibilancihasalsosetitsowncorporatescoringmodel.
ThisscoreisprovidedtogetherwiththedataonthefinancialstatementstoassociatedbanksandtotheBOI.
ExamplesofSimilarRatingandScoringServicesProvidedbyPrivateCreditAssessmentAgenciesTheseexamplesdonotcoverexhaustivelytheavailableprivateriskinformationservicesinEurope.
Indeed,thistopicwouldrequireitselfaspecificandextendedstudy.
However,itisinterestingtopresentafewrepresentativeexamples.
Thefollowingdiscussionwillfocusonthreecases:apublicscoringcarriedoutbyaFrenchbusinessassociation,andtwomajorCAAs.
OneoftheCAAsisaffiliatedwithaleadinginternationalinformationgroup;theotherisaffiliatedwithaleadingEuropeangroup.
72TheAFDCCscoringTheFrenchAssociationofCreditManagers(AFDCC)isanassociationofprofessionalsinvolvedincreditmanagement,especiallyincompaniesthatmanagetheirfinancialrelationshipswiththeircustomersandvendors.
TheAFDCChasdevelopedascoringfunctionaimedatclassifyingsmall-andmedium-sizecompanies.
Itisfreeandpubliclyreleased.
Itisusedextensivelybycorporatecreditmanagerswhoseavailableresourcesforcreditriskmodellingaregenerallylimited.
ThefirstAFDCC'sscoringwasdisclosedin1992.
Itwasbasedonalimitedsampleof2,000corporatesanddidnottakesectoralandsizedifferencesintoaccount.
AnewscoringfunctionwasreleasedinFebruary1999.
36Thedatabasehasbeenenlargedtocover220,000companiesandupdatedwithdatafrom1994to1996.
Thescoringnowincludeselevenspecificfunctionstakingsectorsandbusinessvolumespeculiaritiesintoaccount.
Asthepreviouslymentionedscorings,scorescanbeexpressedintermsofdefaultprobabilityexpressingabankruptcyriskata1.
5-yeartimehorizon.
Hence,itcouldbeconsideredausefulbenchmarkofcorporatecreditriskassessment.
Butoneshouldremaincautiousatthepresentstagebeforeconcludingonthistopic.
Indeed,someaspectshavetobediscussed:ThesampleonwhichthescoringisbasedwasprovidedbySCRL,amajorCAA.
Thedatabaseincludes250,000companiesonwhichSCRLhascollectedcomprehensiveinformationandconductedarating.
AlthoughthisbaseseemstobeaslargeasthatoftheBOFonthesamepopulation(Frenchcorporates),theeffectivecoverageoftheSCRL'scouldbemuchmorenarrow:amongthe250,000firms,220,000constitutethescoringsampleofwhich150,000companieshaveaturnoverrangingfromFRF1millionto10million;and70,000rangingfromFRF10millionto100million.
ThecoverageisnotstrictlycomparablewiththatoftheBOF(companieswithmorethanFRF5millioninturnover)butitseemsthatthe"1-10"groupof150,000companiesisfarfromcoveringthewholeeffectivenumberofcompaniesofthatsize.
Thegoodness-of-fitofthefunctionseemsequivalenttothatoftheBOF'sscoring(about80%ofsuccessfulback-testing).
Butthe"real"qualityoftheAFDCC'sscoringisdifficulttoassesssincethescoringfunctionswerecomputedonrandomlyselectedsub-samplesofabout10,000companies.
Inaddition,thegoodness-of-fitwasonlydeterminedonthesesamplesandnotonabroaderscope.
Therefore,thescoringfunctionmodelizessatisfactorilythescoringsampleitself,butthereisnofurtherevidenceforthebulkofthesmall-andmedium-sizedcompanies.
COFACE-SCRLCOFACE-SCRLclaimstobethefirstFrenchCAAwithadatabaseincludinginformationon4millionFrenchbusinesses37and13millionEuropeanfirms.
However,asmentionedabout36"LescoreAFDCC2,principesdeconstructionetguided'utilisation",M.
Dietsch,Centred'EtudedesPolitiquesFinancières,Feb.
1999.
37Asthereareabout2.
5millionFrenchcorporates,thisfigureshouldalsoincludeindividualentrepreneurswhicharenot"incorporated"i.
e.
theirbusinessisnotlegallyembodiedinacompanystatus.
73theAFDCC'sscoring,theSCRLdatabaseofcompaniesratedonacomprehensivebasisseemstobelimitedto250,000firms.
SCRLhascommercialrelationshipswith14,000customers,mainlyfirmsseekinginformationontheirowncustomersandvendors.
SCRLreleasestwodifferentcreditqualityanalysestoitscustomers.
Oneanalysisisacomputer-basedratinggeneratedbyaRBES,whichreplicatestheSCRL'sratingrationale.
Theotherisanefficiencynote,whichmeasuresthedistancebetweenacompanyandthe"efficiencyfrontier".
Theefficiencyfrontieristheaxisorthecurveofthebusinesses'input-outputrelationships,expressingthebestobservedoutputmaximisationgivenacertaininput,orinputminimisationforagivenoutput.
Ratingsdisclosureisrestrictedtothesubscribers.
ThecoverageratiooftheRBESratingisnotdisclosedandthatoftheefficiencynoteencompasses350,000FrenchcompaniesofwhichturnoverrangesfromFRF1millionto1billion.
InsofarasSCRLencounterscompetition,itsin-housemethodologyisnotpubliclydisclosed.
DUN&BRADSTREETD&Bcovers4millionFrenchcompaniesand19millionEuropeancompanies,buttheeffectivecoverage(comprehensiveinformation)hasnotbeendisclosed.
D&BissupportedbyaninternationalnetworkheadquarteredintheUnitedStates.
Oneofthegroup'saffiliatesistheratingagencyMoody'sInvestorService,butMoody'sofficialssayD&BandMoody'sdonotexchangedataormethodologicaladvices.
D&BsellsratingsgeneratedbyascoringbasedonadiscriminantanalysisandaRBES,andsellsfinancedecision-makingandmanagementproducts.
Themodels'parameters,databasesandqualitytestshavenotbeendisclosed.
RatingsandCorporateDefaultsThequantitativereviewofthestatisticalrelationshipbetweenthedifferentabovementionedratingsystems,especiallythosebasedonCFSDs,isbeyondthescopeofthisstudy.
Somequantitativeanalysishasalreadybeendonewithinthecourseofbuildingprobabilisticscoringsystems.
However,atthisstage,themainquestioniswhetheritispossibletoestablishamappingofthemultipleratingframeworks,thatis,tocompareandharmonisethesedifferentframeworksintermsofdefaultpredictability.
ThisissueisbeingtackledbyCreditDirectoratesoftheCentralBanksmembersoftheESCBtoimplementcompatiblestandardsfordeterminingtheeligibilityofcorporatedebtinstrumentsascollateralofESCB'sfunding.
DefinitionsofdefaultacrosscountriesSomediscrepanciesmayexistacrosscountriesregardingthemeaningoftheterm"default"whileotherdiscrepanciesmayexistamongreportingframeworksanddefaultdatabases.
Thenotionof"default"mayrefertopaymentdefaultoritmayrefertothevariouslegaldefinitionsofdefault.
DefinitionsofpaymentdefaultUnpaidobligationsonduedate:thiscouldresultfromatechnicalevent(e.
g.
disruptionintransferofdigitaldata,administrativemistakeetc.
),atradeconflict(acorporatecustomerwhichrefusestopayanunsatisfactoryproductorserviceora74delayindelivery),atemporaryliquidityshortageetc.
Itrefersingeneraltolaggedpaymentsnolongerthantheminimumprudentialregulatorydurationofadelinquentordoubtfulasset(generallythreemonths,unlessalitigationprocessisopenedmeanwhile),Doubtfulordelinquentasset:afteraregulatoryduration(threemonths)orafterbeginninglitigationofanykindatanytimeeitheronacontractualbasis(declarationofaneventofdefaultasdefinedinthefinancingcontract,accelerationoftheloan)oronthebasisofacourt'sdecision.
Itcorrespondstotheprudentialdefinitionofdefaultwhichisnotstrictlylitigation-based,albeitlitigationcasesareincludedtherein.
Insolvency:thedebtorisgloballyunabletocoveritsfinancialobligations.
Itdoesnotnecessarilyresultinlegallystatedbankruptcy.
Insomecountries,likeGermany,insolvencydoesnotleadtolegalbankruptcythreetimesoutoffourbecauseoftheprioritygiventonegotiatedsettlements/treatmentofdefaultsovercourtaction.
Nevertheless,insolvencyispubliclyknownandlegallydefined.
InFrance,ifitispossible,theCommercialCourtsconductamicablenegotiationsunderstrictconfidentiality.
Thismayconcerndoubtfulassetsaswellaslatepayments.
DefinitionsoflegalbankruptcyThedefinitionoflegalbankruptcydependsontheprovisionsofthetradeorbankruptcylawineachcountry.
Thecommoncharacteristicofalllegalenvironmentsisthateventofbankruptcyaretriggeredbyaformaldeclarationofsuchaneventtoacourt.
InFrance,bankruptcyisatwo-stepprocedure:thecompanyhastodeclareitselfunabletomeetanyfinancialobligations(completebreachofpaymentobligations)beforethecourtopensformallyajudiciarybankruptcyprocedure.
Afterlookingatthecompany'ssituation,thecourtisempoweredeithertoenforceajudiciaryrecoveryplan,ortodeclarethewindingupofoperations.
Thewindingupofoperationsoccursinthemajorityofcases.
Allcourts'decisionsarepubliclydisclosedandcollectedbystatisticalinstitutes,eitherstate-ownedorprivate,dependingonthecountry.
Whateverthedefinitionofdefault,itisworthwhileunderliningthateventsofbankruptcydonotresultfromstrictlymechanicalprocesses.
Theyareinfluencedbythedebtorcompany'sbehaviourandbythatofthecourt,especiallyregardingthedurationoftheprocedureandtheassessmentofthedebtor'ssituationvis-à-visthecreditors.
Inaddition,creditorsarestronglyencouragedordeterredfromdeclaringtheaccelerationofloansandalitigationprocessnotonlybytheeffectivesituationofthedebtorbutalsobythesizeofthecompany(largefirmsgenerallyavoidcourtaction),andprovisionsofthebankruptcylaw.
Asaresult,defaultsreportedinCCRsorCFSDsdonotexpressthesameeconomicinformation.
InAustria,FranceandGermany,CCRsandCFSDsincludeeventsofbankruptcy(legalprocedures).
Inaddition,theFrenchCFSDreportspaymentdefaults(firstdefinition)throughthepaymentqualitygradeincludedintheratings.
Thisnotionofdefaultincludestechnicaleventsandtradeconflicts.
TheBOF'sanalystscollecttrackrecordsonpaymentdefaultandappraisethequalityofpaymentonacasebycasebasistodistinguishtechnicaleventsfromtemporaryliquidityshortagesandinsolvencycases.
TheItalianandtheSpanish75CCRsrefertodoubtful/impairedloansandthustheysticktotheprudentialapproachofdefault.
Inaddition,legalbankruptciesandinsolvencysituationsareincludedintheCCRs.
Bankruptcydataaremoresuitableforissuer/borrowerratingsthanforissue/loanratings,whereasdefaultdata(broaddefinition,includingeventsdifferentfrombankruptcy)aremoresuitableforissue/loanratings.
Allofthereviewedratingsystemsarebasedonborrowerratings,whichhavestraightforwardassociationswithbankruptcyrisklevels.
Inaddition,itisconsistentwiththenatureofthebulkofcompaniesratedthroughCFSDsandCCRsi.
e.
small-andmedium-sizedcompaniesfundedbybankingloansorcreditlinesthatareheldbybanksuntilrepaymentdates.
Inthisrespect,bankruptcycasesprovidecriticalinformationgiventhenatureoftheborrowers(moreexposedtobankruptcyriskthanlargecorporates)andoftheassets(non-negotiable).
Thisdoesnotunderminetheusefulnessofdefaultdata,whichcanhelpspotstructuralcorporateweaknesses.
PerimetersofreferencesamplesTheoretically,thebestreferencesampleismostorallcompanies.
Althoughitshouldbestraightforward,thesampleisnotsimplyidentifyingandcollectingdataonallexistingfirms.
First,thenotionofcompanymaydifferacrosscountriesandacrossdatabases.
Thereisalsothequestionofhowtotreatpartnerships,single-shareholdercompanies,specialpurposevehicles,governmentagenciesestablishedundera"quasi-corporate"status,andindividualentrepreneurs,etc.
Moreover,statisticalinstitutessometimescannotidentifypreciselytheincorporatedcompaniesthathaveceasedanyeffectivebusinessbutarestillformallyregistered.
Itisnecessarytoensurethatreferencesamplesarecomparablefromonecountrytoanother.
RelevanttimehorizonsforcomparingdefaultstatisticsDefaultratesareestimatedthroughoutacertaintimehorizonthatshouldbeconsistentwiththeratingtimehorizon.
Theoretically,thedefaultriskincreaseswiththetimehorizon.
Thedefinitionofthetimeperiodmaybearrivedatthroughdifferentapproaches:Thestartingdateisfixedatanytimeforasetofmultiplecompaniesclassifiedaccordingtotheirrating.
Therefore,thestartingdateisthesameforallthecompaniesandthecumulateddefaultratesaredeterminedforthefollowingyears.
Foreachclassofrating,withastartingdateonyearn,thecumulateddefaultsatyearendn+1aredeductedfromtheoriginalpopulationofcompanies,38thenthesameprocessisrepeatedatyearn+2.
etc.
Thestartingdateistheratingassignmentdate,associatedtothefinancialstatementsasofthesamedate.
Itisobservedwhethertheratedcompaniesdefaultwithinthefollowingyear,thesecondfollowingyear,etc.
,dependingonthetimehorizon.
Accordingtothisapproach,thestartingdateandthefollowingscrutinisedperiod38Itcorrespondstoabankruptcy-riskapproach.
76maybedifferentforeachcompany.
Thisapproachissuitableformatchingthefinancialstatementsandtheratingand,withafastpacedratingupdates,forassessingtherelationshipbetweenshiftsinratinglevelsandthedisclosureoffinancialstatements.
Thestartingdateisthedefault(bankruptcy)date:itisabackwardanalysistoestimatetherelationshipbetweentheeventofdefaultandthepreviousrating(thepreviousyearforaone-yeartimehorizonetc.
).
Thenon-defaultedcompanieshavetobeintroducedtothesampletoestimateaglobaldefaultrate.
Thesethreeapproachesarepossiblewaystoevaluatingtheratingsystemsperformancewithregardstodefaultpredictability.
StudiesperformedbyworkinggroupsonharmonisationofeligibilitycriteriaintheESCBhavesimilarresultsforthefirsttwoapproaches.
CriteriaofanefficientratingsystemAnefficientratingsystemshouldhighlightanegativecorrelationbetweenbestqualityratingsandthedefault/bankruptcyrate.
Thedifficultyisinestablishingacommonstandardvaluableacrossmultiplecountries.
Moreover,theshareofsmallbusinessesinfluencesthebankruptcyrate.
InAustria,France,andGermany,somestudiesshowthatthelargerthecompanies,thelowerthebankruptcyrate.
39Ifthelegalenvironmentdoescontributesometimestolegalbankruptcyincasesofdefault,thensmallcompaniesaremorevulnerableinthiscontext.
40Whentherateofdoubtfulassetsistakenintoaccount,asinSpain,thesizeeffectcouldbelessclearbecausemediumandlargecompaniesaremoreexposedtolaggedpaymentsthantolegalbankruptcy.
Thatnarrowsthegapbetweensmallandlargecompanies.
Theefficiencyofratingsystemsalsodependsontransitionmatriceswhichshouldbelookedataccordingtothefollowingcriteria:Stability:aratingshouldbestablethroughoutthetimehorizononwhichitissupposedtogiveariskassessment.
Forexample,ifaratingisvaluableforaone-yeartimehorizon,fewratingchangesshouldoccurwithinthisperiod.
Sensitivity:itisnotcontradictorywiththepreviouscriteria;ratingsshouldrespondtoanychangeinthecompany'ssituation.
Highorgoodqualityratingsshouldbedowngradedsufficientlyearlybeforeanyeventofdefault.
Therefore,anunstabletransitionmatrixshouldbejointlyobservedwithahighaveragedefaultrate.
Interpretingtransitionmatricesrequirelongtrackrecordsandhighfrequencyinformationontimeperiodsprecedingdefaults.
39"Risqueettailledesentreprisesindustrielles",M.
Bardos,1998.
40Foradifferentcontext,seetheChapter11issueintheUS.
77EuropeanCentralCreditRegistersandCentralFinancialStatementsDatabases:ComparisonofTheDifferentExistingFrameworksTablesItoIV:CentralCreditRegistersTableI:ReportingbanksandotherfinancialinstitutionsReportinginstitutionsAustriaBelgiumFranceGermanyItalyPortugalSpainReportinginstitutionsincorporatedundernationalregulationsCreditInstitutions:exposuresdeclaredbyHeadOfficeXXXXXXXCreditInstitutions:exposuresdeclaredbyforeignbranchesXXXX41XLeasingandfactoringcompanies42XXX43XXInsurancecompanies:exposuresdeclaredbyHeadOfficeXXInsurancecompanies:exposuresdeclaredbyforeignbranchesXXOtherreportinginstitutions:SocialWelfareandFederalLabourOfficeXVenturecapital,nationaleconomicrecoveryinvest.
Fund44XCreditcardsissuedbynon-financ.
corp.
XCentralBankXDepositsinsuranceorguaranteefund,mutualguaranteefundsXEuropeanbranchesofforeignbankinggroupsCreditinstitutionsXXXXXXXLeasingandfactoringcosXXCentralBank,depositsinsuranceorguaranteefund,mutualguaranteefundsXForeigninstitutionsundertheregimeoffreedomofservicesoffering45CreditinstitutionsXXLeasingandfactoringcosXForeignsubsidiariesofdomesticinstitutionsCreditinstitutionsX41Italy:onlyloansgrantedfromabroadtoItaliancounterparties.
42Austria:includingloansgrantedbydomesticbanks'foreignbranches.
SpainandPortugal:leasingandfactoringcompaniesareconsideredcreditinstitutions.
43FactoringcompaniesreportdirectlytotheCCRwhateverthenatureoftheirmainshareholder(s).
Onthecontrary,leasingcompaniesaresubjecttotheGermanCCRreportingonlyifthesecompaniesarebankingsubsidiaries.
44Germany:includingloansgrantedbydomesticbanks'foreignbranches.
45Offeringservicesisallowediftheproviderisincorporatedasacreditinstitutioninitshome-country.
(e.
g.
someGermanleasingcompaniesarenotconsideredcreditinstitutions).
78TableII:CategoriesofborrowersAustriaBelgiumFranceGermanyItalyPortugalSpainDatabasecoverageNumberofborrowers95,000425,000600,000270,0001,600,0003,400,00010,000,000Ofwhichcorporates(opposedtoindiv.
entrep.
orhouseholds)35,000160,000580,000170,000N.
A400,000620,000Categoriesofborrowers(domesticandforeign)CorporatesXXXXX46XXCreditinstitutionsXX(onlyforeignones)XXPublicadministrationsX47X48XX(onlyforeignones)XXXHouseholdsXXXCreditinstitutionsgeneralmanagersandtheirfamilyXNon-financialguarantorsXX46Italy:onlyloansgrantedfromabroadtoItaliancounterparties.
47Onlylocalauthorities.
48ExceptfinancingstotheKingdomofBelgium.
79TableIII:CategoriesofreportedfinancinginstrumentsandreportsperiodicityAustriaBelgiumFranceGermanyItalyPortugalSpainLoans(drawnamount)MonthlyQuarterly49MonthlyQuarterlyMonthlyMonthlyMonthlyTotalfinancing(includingundrawnamount)MonthlyMonthlyMonthlyMonthlyMonthlyGuarantees(granted)MonthlyMonthlyQuarterlyMonthlyMonthlyGuarantees(received)50MonthlyMonthlyCollateralandpledge(risk-mitigating)MonthlyDerivativesMonthly51Monthly52Quarterly53TobeimplementedSecuritisedloansMonthlyFixed-incomesecuritiesMonthlyAdditionalinformationEventofbankruptcyBoF'srating,F/S,eventofpaymentdefault,eventofbankruptcyEventofbankruptcy,maturityofmortgageloansandofloansgrantedbypublicauthoritiesImpairedloans,unauthorisedoverdraftsInitialmaturity,impairedloans49Exposuresonforeigncounterparties:thedrawnamountisnotreportedandtheonlyinstrumentcategoriesofcashadvancesandconfirmedcreditlinesarereported.
50Stand-aloneguaranteesorrisk-mitigatingguaranteesattachedtoacreditexposure.
51Indicationofexistenceofswaptransactionswithoutanyvaluation.
52Indicationofexistenceofswaptransactionswithoutanyvaluation.
53Valuation(creditequivalent)included.
80TableIV:ReportingthresholdsAustriaBelgiumFranceGermanyItalyPortugalSpainGeneralthresholdATS5,000,000;=EUR363,364BEF1,000,000=EUR24,789FRF500,000=EUR76,225DEM3,000,000=EUR1,533,876ITL150,000,000=EUR77,469NILESP1,000,000=EUR6,010fordomesticcounterparties,ESP50,000,000=EUR300,506forforeigncounterpartiesSpecialthresholdsBEF100,000,000=EUR2,478,935forexposuresreportedbythedomesticbanks'foreignbranchesDEM250,000=EUR127,823forloanstobanks'generalmanagersandtheirfamiliesNothresholdforimpairedloansPTE5,000,000=EUR24,940forhouseholdsESP10,000,000=EUR60,101forfundingofSpanishcorp.
businessabroadandguarantors,ESP1,000,000=EUR6,010forailingforeigncounterpartiesThresholdsarerelatedto:IndividualborrowersIndividualborrowersIndividualborrowersGroupsofborrowerswithcommoninterestsorlinkedtoeachotherIndividualborrowersIndividualborrowersIndividualborrowers81TableV:Financialstatementsdatabases/coverageratioFranceTurnoverinFRFmn%intotalnumberoffirms%ofoutstandingcredit%oftotalemployedstaff2501.
049.
737.
0)TotalFIBEN37.
290.
587.
6)France(FIBEN)Sectorsbreakdown%inno.
ofcosManufacturing28%Trade36%Building11%Transport5%Services20%GermanyCoveragein%oftotalB/SamountAverageratio=60%Averageratio"WestGermany"=60%Averageratio"EastGermany"=25%Accordingtolegalstatus%ineco.
sectorstotalturnoverBysizeofB/Smanufacturing70%Ltdcompanies75%trade50%uptoDEM10mn10%Partnerships50%construction25%10-10060%Individualentrepreneurs10%servicesN.
S>10080%AustriaSectorNumberofcompaniesBreakdown%innumberofcosCoverageasa%ofturnoverTotal6890100%33%Energy451%81%Manufacturing255437%54%Trade235834%27%Construction85612%25%Transport,communications4106%26%Otherservices66710%6%82Spain(CBA)Largecosdatabase=8,000cosi.
e.
1%totalnumberofcosSectorCoveragein%ofGDPCoverageasa%ofturnoverTotalN.
A.
36%Energy80%90%Manufacturing35%45%Trade10%25%Construction15%30%Transport,communications60%20%Otherservices5%10%SpainSamplebreakdown(sizecriteria)Netturnover%ofcompanies%ofnetturnover>EUR40mn11847-402512FRF5bnB:>FRF1bnFRF500mnFRF200mnFRF100mnFRF50mnFRF10mnFRF5mnCopyrightKMV199955bpSpreadoverUST90%CreditSpreadCreditSpreadNon-CreditComponentDecompositionofSpreadsoverU.
S.
TreasuryAAACreditsBBCredits200bpSpreadoverUST10%25%65%Non-CreditComponentCallOption10%121Reduced-formapproachDuffieandSingleton(1998)definedasecondapproachcalled"reducedform".
Inthisapproach,thetimeofdefaultismodelledasanexogenouslydefinedintensityprocess.
Defaultdoesnotneedtodependexplicitlyontheissuer'scapitalstructure.
Sincethedefaultprocesscanbeendogenouslyderived,thestructuralapproachisaspecialcaseofthereduced-formapproach.
Divorcingtheissuerfromtheintensityprocessenablesmodellingdefaultwithmuchinformationaboutwhytheissuerdefaults.
However,modellingdefaultwithouttheoreticalguidancerunstheriskofignoringbothmarketinformationanddrawingerroneousconclusions.
Asampleofreduced-formmodelscanbefoundinCathcartandEl-Jahel(1998),DuffieandSingleton(1998),JarrowandTurnball(1995)andJarrow,LandoandTurnball(1997).
PerformanceDeterminingtheperformanceofadefaultmeasureisbothatheoreticalandanempiricalproblem.
ThequestioniswhatexactlydowemeanbyperformanceorpredictivepowerInpractice,wearenotabletoclassifyfirmsinto"willdefault"and"willnotdefault"categories;wecanonlyhopetoestimateprobabilitiesofdefault.
Therefore,testingtheperformanceofadefaultmodelmeanstoinvestigateitsabilitytodiscriminatebetweendifferentlevelsofdefaultrisk.
Performanceshouldbemeasuredalongseveraldimensionsincludingpowerofdiscrimination,abilitytoadjusttothecreditcycle,andtheabilitytoquicklyreflectanydeteriorationincreditquality.
Animportantissueisthedefinitionofdefault.
Defaultshavedifferentdefinitionsacrosscountries,butalldefinitionsindicateeconomicdistress.
Therefore,whencomparingtheperformancesofdifferentmeasures,itisimportanttousethesamedefaultdefinitionandthesamesampleofcompanies.
Onewaytodetermineamodel'spowerofdiscriminationisusingtheso-calledpowercurvestest.
Thistestmeasuresamodel'sabilitytoidentifythefirmsthataregoingtodefaultforchosenleadtimes(forexample,12months).
Theabilitytodiscriminateismeasuredusinga"hit-miss-ratio".
Thepowercurvetestdoesnotrequirethatamodeldeterminedefaultprobabilitiesforcompanies,butitdoesrequirearankingbetweencompanies.
Inthetesting,percentiles(every10%mark)arecalculatedafterrankingtheEDF's(ratings)gainedfromthewholepopulationinascendingorder.
WiththedistributionofEDF's(ratings)established,ahit-missratioisthencalculatedforeachpercentile.
IftheEDF(rating)ofeachdefaultedcompanyisselectedatthedefaultdateminusleadtime,(i.
e.
ifwechoosetheleadtime12months),wetaketheEDF(rating)12monthsbeforethedefaultdate.
Theresultsofthiscalculation(carriedoutbyKMV)forthePrivateFirmModelfromKMVareshowninFigure12foraleadtimeof12monthsfordifferentcountries.
Figure11sumsupthedataandtestingenvironment.
122Figure11Figure12EuropeanPowerTestingPercentofdefaultsexcludedwhenexcludinglowerpercentofthepopulationUKPrivates*FrancePrivates*SpainPrivates*NorwayPrivates^SwedenPrivates*10%38%30%41%49%38%20%58%56%57%65%55%30%75%68%64%74%68%40%85%76%74%79%77%(^Bankruptciesonly,*DefaultsandBankruptcies)EuropeanTestDataUKFranceSpainNorwaySwedenPopulation(company-years)40700755001950080817118Population1996901316506582332485447Yearsofdata55332Defaultsinsample52170726610587Meanprobability1.
28%0.
94%1.
36%1.
3%1.
22%123InterpretationofpowercurvesWetaketheUKtestingresultsasanexample.
Ifabankdoesnotlendtothebottom10%companiesrankedbyEDF,thenitcanavoid38%oftheentiredefaultingfirmsinthesample.
ForGermanyandtheUnitedStates,thereare,sofar,onlytestingresultsforCreditmonitor(KMVmodelforlistedcompanies).
TheresultsandthemethodologyfortheGermantestingisdescribedinLiebigandNyberg(1999).
GermanyTestDataPopulation(company-years)AveragepopulationYearsofdataDefaultsinsampleMeanprobability41375407-8180.
19Figure13GermanyPowerTestingResultsFigure14showsthetestingresultsforaleadtimeof12months.
PercentofpopulationexcludedPercentofdefaultsexcluded10%47%20%82%30%94%Figure14124ReferencesAltman,E.
I.
/Caouette,J.
B.
andNarayanan,P.
(1998),"ManagingCreditRisk",JohnWiley&SonsBlack,F.
andCox,J.
C.
(1976),"ValuingCorporateSecurities:SomeEffectsofBondIndentureProvisions",JournalofFinance,31(2),351-367Black,F.
andScholes,M.
(1973),"Thepricingofoptionsandcorporateliabilities",JournalofPoliticalEconomy,81,637-659Bohn,J.
R.
(1999),"Asurveyofcontingent-claimsapproachestoriskydebtvaluation",UniversityofCalifornia,BerkeleyBrennan,M.
J.
andSchwartz(1977),"ConvertibleBonds:ValuationandOptimalStrategiesforCallandConversions",JournalofFinance32(3),1699-1715Brennan,M.
J.
andSchwartz(1978),"CorporateIncomeTaxes,Valuation,andtheProblemofOptimalCapitalStructure",JournalofBusiness,51(1),103-114Brennan,M.
J.
andSchwartz(1980),"AnalyzingConvertibleBonds",JournalofFinancialandQuantitativeAnalysis,15(4),907-929Cathcart,L.
andEl-Jahel,L.
(1998),"ValuationofDefaultableBonds",JournalofFixedIncome,8(1),65-78Crosbie,P.
J.
(1999),"ModelingDefaultRisk",KMVCorporationDuffie,D.
andSingelton,K.
J.
(1997),"AnEconometricModeloftheTermStructureofInterest-RateSwapYields",JournalofFinance,52(4),1287-1321Duffie,D.
andSingelton,K.
J.
(1998),"ModelingTermStructureofDefaultableBonds",StanfordUniversityGeske,R.
(1977),"TheValuationofCorporateLiabilitiesasCompoundOptions",JournalofFinancialandQuantitativeAnalysis,12,541-552Ingersoll,J.
E.
(1976),"ATheoreticalandEmpiricalInvestigationoftheDualPurposeFunds",JournalofFinancialEconomies,3,83-123Ingersoll,J.
E.
(1977a),"AContingentClaimsValuationofConvertibleSecurities",JournalofFinancialEconomies,4,289-321Ingersoll,J.
E.
(1977b),"AnExaminationofCorporateCallPoliciesonConvertibleSecurities",JournalofFinance,32(2),463-478Jarrow,R.
A.
andTurnball,S.
M.
(1995),"PricingDerivativesonFinancialSecuritiessubjecttoCreditRisk",JournalofFinance,50(1),53-85Jarrow,R.
A.
/Lando,D.
andTurnball,S.
(1997),"AMarkovModelfortheTermStructureofCreditRiskSpreads",ReviewofFinancialStudies,10(2),481-523KMVCorporation(1997),"ModelingDefaultRisk",Mimeo125KMVCorporation(1999),"Valuation,EDFsandValuingCashFlows",MimeoLeland,H.
E.
(1994),"CorporateDebtvalue,BondCovenants,andOptimalCapitalStructure",JournalofFinance,49(4),1213-1252Leland,H.
E.
andToft,K.
B.
(1996),"OptimalCapitalStructure,EndogenousBankruptcy,andtheTermStructureofCreditSpreads",JournalofFinance,51(3),987-1019Liebig,T.
andNyberg,M.
(1999),"TestingResultsofCreditMonitor(KMV)forlistedGermanCompanies",DeutscheBundesbankLongstaff,F.
A.
andSchwartz(1995),"ASimpleApproachtoValuingRiskyFixedandFloatingRateDebt",JournalofFinance,50,789-820Merton,R.
C.
(1974),"Onthepricingofcorporatedebt:theriskstructureofinterestrates",JournalofFinance,29,449-470Zhou,C.
(1997),"AJump-DiffusionApproachtoModelingCreditRiskandValuingDefaultableSecurities",FederalReserveBoard126PartIII.
DefaultandRecoveryStudies:QuantifyingtheRiskSectionA.
DefaultstudiesIntroductionCreditratingsdistilavastamountofqualitativeandquantitativeinformationoncreditqualityintoasimpleordinalranking.
Thesimplicityofthisrankingallowsevenarelativelyunsophisticatedinvestorthemeanstocomparetherelativecreditriskofawidevarietyofdebtinstruments.
Itisquitesimpletounderstandthatabondwitha"B"ratingcarriesrelativelygreaterriskthanonewithan"A"rating.
However,amoresophisticatedinvestorwilloftenrequirethemeanstoassesstheabsolutecreditriskofadebtinstrument.
Forexample,thesophisticatedinvestormaywanttoknowtheprobabilitywithwhichthe"B"ratedbondwilldefaultinthenextfiveyearsandtheamounttheinvestorcanexpecttorecoverintheeventofdefault.
DefaultstudiesInordertomeettheneedsoftheincreasinglysophisticatedinvestorandtogaugethehistoricalconsistencyoftheirratingprocesses,thetwomajorcreditratingagencies,Standard&Poor'sCorporationandMoody'sInvestorService,conductongoingresearchintothedefaultexperiencesoftheirratedissuers.
S&P'slong-termcorporatedefaultandratingtransitionstudy(BrandandBahar1999)analysesthecredithistoryofmorethan7,300obligationsratedbyS&P'sbetween1981andthepresent.
Structuredfinancings,publicsectorissues,ratingsbasedonpublicinformation(the"pi"ratings),andsovereignissuersareexcluded.
Moody'slong-termcorporatebonddefaultstudy(Keenan1999)analysesthecredithistoryofatotalof15,200issuersand2,200defaultingissuesovera79-yearperiodwithspecialemphasisontheperiodof1970tothepresent.
Moody'sdefaultstudyalsoexcludesstructuredfinancings,municipaldebtissuers,aswellasprivateplacements,andissuerswithonlyshort-termdebt.
Sovereignsissuinglong-termdebttothepublic,however,areincluded.
Non-USissuerscompriseoverone-thirdofthecurrentMoody'srateduniverse.
Bothstudiesemploysimilarmethodologies.
Theissuer,nottheissue,istheunitofstudy.
Subordinationiscontrolledforthroughthecalculationofanimpliedseniorunsecuredratingforissuershavingonlysubordinatedrateddebt(S&Preferstotheseas"issuercreditratings").
Therule-of-thumbisthatanimpliedseniorratingisonenotchhigherthanthesubordinatedratingifinvestment-grade,andtwonotcheshigherifspeculative-grade.
S&Pdefinesdefaultasthefirstmissedpaymentofprincipleorinterestondebt,ratedorunrated.
Moody'sdefinesdefaultasanymissedordelayedpayment,ordistressedexchangediminishingtheobligationorhavingtheapparentpurposeofhelpingtheissueravoiddefault.
Inordertotrackaratingthroughtime,bothagenciesgroupallissuersholdingagivenimpliedseniorratingtogetheratthestartofeachyear,thentrackthegroupfromthatyeartotheendofthestudy.
S&Pcallsthesegroupings"staticpools"whileMoody'sreferstothemas"cohorts".
Bytrackingthesegroupingsthroughthesubsequentyearstheagenciesareabletodeterminecumulativedefaultratesovermultipleyearhorizons.
Thesestudiesfindawell-definedcorrelationbetweenthecreditratingandtheprobabilityofdefault:alowercreditratingcorrespondingtoahigherprobabilityofdefault.
Furthermore,127thesestudiesfindthatlowercreditratingsarelessstable.
TheS&P'sstudy,forexample,findsthathigherratingsarelongerlivedwithloweroriginalratingscorrespondingtoashortertimetoaneventualdefault.
Moody'ssimilarlyfindsaconsistentcorrelationbetweendefaultratevolatilityandalowcreditratingwhichremainsevidentoverhorizonsaslongas20years.
Defaultratesexhibitacyclicalcharacterroughlyconsistentwiththecreditcycle(CantorandFons1999),andadversebusinessconditionscoincidewithdefaultupswingsforallratingspools(BrandandBahar1999).
Thisisnotsurprisingasbothagenciesstatethattheyseekto"aimthroughthecycle"whenratinganissuer.
Table1:Defaultratesat1and5yearhorizonsbyagency(percent)BondRating1YearDefaultRate5YearDefaultRateS&PMoody'sS&PMoody'sS&PMoody'sAAAAaa0.
000.
000.
150.
22AA+Aa10.
000.
000.
270.
25AAAa20.
000.
000.
110.
50AA-Aa30.
000.
070.
400.
45A+A10.
030.
000.
480.
75AA20.
040.
000.
320.
66A-A30.
070.
000.
820.
45BBB+Baa10.
200.
041.
151.
45BBBBaa20.
190.
081.
361.
29BBB-Baa30.
300.
313.
212.
79BB+Ba10.
620.
645.
798.
45BBBa20.
780.
596.
889.
66BB-Ba31.
192.
5512.
2320.
76B+B12.
423.
5616.
1825.
56BB27.
936.
8524.
6628.
52B-B39.
8412.
4129.
1637.
49CCCCaa1-C20.
3918.
3141.
2938.
30Investment-Grade0.
080.
040.
710.
82Speculative-Grade3.
833.
6716.
0820.
26Source:Moody'sdatafromKeenan(1999).
S&PdatafromBrandandBahar(1999).
Note:Moody'sdatacoversperiodfrom1983-1998.
S&Pdatacoversperiodfrom1981-1998.
Table1liststhehistoricaldefaultfrequenciesforbothagencies(aspercentages)forone-andfive-yearhorizons.
GiventhesimilarityintheratingsmethodologiesbetweenS&P'sandMoody's,historicaldefaultratesprovideameansofassessingthecomparabilityoftheircreditratings.
SeasoningeffectAltman(1989)raisesanimportantpointregardingtheeffectoftheseasoningofabondissueonmeasureddefaultrates.
Unliketheagencystudies,Altmantakestheindividualbondissueastheunitofstudy,andsocanassesstheeffectofabond'sageontheprobabilityofdefault.
Typically,themarginalprobabilityofdefaultrisesquicklyinthefirstthreeyears,afterwhichitlevelsoff.
Therefore,theseasoningeffectismostimportantinthefirstthreetofouryearsofabond'slife(Caouetteet.
al.
1998).
Thisfitswiththeideathatasmostissuershaveagreatdealofcashonhandafterabondissue,theycanmeetseveralyearsofinterestpaymentseven128ifoperatingcashflowsarepoor.
Therefore,anissuerwithpoorcashflowswillstillhavealowinitialprobabilityofdefault,andtheprobabilityofdefaultonlyrisesovertimeasthecashinfusionfromthedebtissuebecomesdepleted.
Usingtheissuerastheunitofstudy,Moody's(1999)calculatesthehazardrateofdefaulttoassesstheeffectofthelengthofacorporatebondissuer'scredithistoryontheprobabilityofdefault.
Thesehazardratesofdefaultarequitelowinthefirsttwoyears,risesharplyinthethirdyearandpeakinthefourthyearafterwhichtheytaperoff.
ThisisinapparentagreementwithAltman'sstudy.
SectorcomparisonsAlthoughtheagencydefaultstudiesdonotcategorisedefaultratesbyratingswithinindustries,CantorandFons(1999)givesomeinsightintothecomparabilityofratingsacrossindustriesforMoody'sratings.
Traditionally,Moody'shastailoreditsratingstothevariousneedsoftheinvestorsinthefourmainsegmentsofthebondmarket:corporates,municipals,investor-ownedutilities,andstructuredfinance.
Forexample,Moody'sratingsofindustrialandfinancialcompanieshaveemphasisedtherelativeprobabilityofdefault.
Forspeculative-gradeissuers,Moody'semphasisedexpectedlossratherthantheprobabilityofdefault.
ForthestructuredfinancesectorMoody'salsoemphasisedexpectedlosswhileforgovernmentandmunicipalsithasemphasised"financialstrength".
Furthermore,thetraditionalneedtoprovideausefuldispersionofratingswithinasectorcompromisesthecomparabilityofratingsacrosssectors.
Asthesemarketshavebecomeincreasinglyintegrated,however,Moody'shasmovedtoadopt"expectedloss"asitsunifyingconceptofcreditrisk.
CaveatsDefaultstudiesofbondratingsprovideawealthofquantitativeinformationonthehistoricaldefaultexperienceofrateddebt.
However,thisquantitativedetailmustbeusedwithsomecaution.
Caouetteetal(1998)concludethattheagenciesremainbetteratmeasuringrelativeratherthanabsolutecreditrisk.
Despitetheneedsofinvestorsformorequantitativedetail,thesimplicityoftheordinalrankingsysteminherentlyobscuresagreatdealofinformationabouttheindividualissuer.
ArecentstudybyKMVCorporation(Kealhofer,Kwok,andWeng1998)emphasisesthefactthatthedegreeofheterogeneitywithineachcreditratingislarge.
Itisnotuncommonforbondswithhighercreditratingstohavehigherdefaultratesthanbondswithlowercreditratings(thatis,theratingsoverlap).
Forexample,thereareBBBbondswithAAdefaultratesdueperhapstoalackoftimelinessinupgradinganddowngradingissuers.
Thiswiderangeofdefaultrateswithinagroupcancausethemeandefaultratetovarysignificantlyfromthemedian.
Themeanmaybealmosttwiceaslargeasthemedianwithasmanyas75%ofthebondissuershavingdefaultratesbelowthemean.
Therefore,itcanbemisleadingtousethehistoricalaveragedefaultrateofaratingcategoryastheexpecteddefaultrateforagivenratedissuer.
RecoverystudiesAninvestortypicallywantsmoreinformationthansimplytheprobabilityofdefaultassociatedwithacreditrating.
Theinvestoralsowantstoknowtheexpectedlossintheeventofdefault.
Theratingagencieshaveproducedstudiestoaddressthisneed.
VandeCastleandKeisman(1999)analyserecoveriesof829debtinstrumentsfromover219defaults.
129Recoveriesaregiveninpresentvaluetermsdiscountedfromthedateofemergenceorliquidationtothelastdateinterestwaspaid.
Increasingtheamountdebtsubordinatedtoagivenissueorincreasingtheamountofcollateraltendstoincreasetherecoveryratefortheissueintheeventofdefault.
Table2:RecoveryEstimatesforS&P'sRatedDebt(PercentofParValue)DebtClassRecoveryStdDeviationBankLoans84.
524.
9Seniorsecured65.
728.
4Seniorunsecured49.
335.
8Seniorsubordinated36.
831.
0Subordinated26.
130.
3Juniorsubordinated13.
624.
4Source:VandeCastleandKeisman(1999).
Note:Datacoversperiodfrom1987-1998.
AspartofMoody'sdefaultratestudy,Keenan(1999)usesthetradingpriceofdefaultedsecuritiestocalculaterecoveryrates(asapercentofpar)ondefaultedsecurities.
Thisstudyalsoshowsthatrecoveriesincreasewiththeseniorityoftheclaim.
Table3:RecoveryEstimatesforMoody'sRatedDebt(PercentofParValue)DebtClassRecoveryStdDeviationBankLoans70.
2621.
33Seniorsecured55.
1524.
31Seniorunsecured51.
3126.
30Seniorsubordinated39.
0524.
39Subordinated31.
6620.
58Juniorsubordinated20.
3915.
36Source:Keenan(1999).
Note:Datacoversperiodfrom1977-1998withtheexceptionofbankloanswhichcovertheperiodfrom1989-1996.
Therecovery(orloss)studiesconsistentlyshowthatseniorityandcollateraldecreasetheexpectedlossondefaulteddebt.
Thissupportstheconventionofratingsubordinateddebtanotchortwobelowseniordebt.
Furthermore,onceseniorityisaccountedfor,thedistinctionbetweeninvestmentgradeandnon-investmentgraderatingshasnoeffectonrecoveries(Caouetteetal1998).
Avarietyofstudieslendfurthercredencetotherule-of-thumbthatdefaultedsecuritiesrecoverabout40centsonthedollar.
DefaultandrecoverystudiesonsyndicatedbankloansTraditionallytherehasbeenadearthofinformationonthedefaultandrecoveryexperienceofratedbankloanscomparedtothatavailableoncorporatebonds.
Recently,however,the130growthofthesyndicatedbankloanmarkethasallowedforapairofstudiesonthedefaultandrecoveryexperienceofratedsyndicatedbankloans.
AltmanandSuggitt(1997)conductedthefirstdefaultstudyofsyndicatedbankloans.
ThestudyusesadatasetoncommercialloansofatleastUSD100millioninsizeforwhichtherewaseitheranoriginalratingassignedbyoneormoreofthemajorratingsagenciesorabondratingfromwhichanimpliedseniorratingcouldbeassigned.
Theauthor'snotethatMoody'sdoesnotusesuchanassignmentinitstreatmentoftheloansofissuerswithbondsoutstanding.
Byusingassignedratings,however,theauthorswereabletosubstantiallyincreasethesamplesize.
Thesamplecoversthetimeperiodfrom1991through1996,involves4,069ratingsonover2,184borrowersforatotalofUSD2.
4trillioninaggregatefacilities.
Assignedratingsaremorepredominantbefore1995whenMoody'sbeganratinglargecorporateloans.
ThestudyemploysAltman'sissue-basedmortalitymethodology(Altman1988)thatdiffersfromtheissuer-basedmethodologyoftheagencystudies.
Table4:SyndicatedBankLoanandCorporateBondDefaultRates(Percent)1YearDefaultRate5YearDefaultRateRatingBankBondBankBondAaa0.
000.
000.
000.
00Aa0.
000.
000.
000.
00A0.
000.
000.
120.
05Baa0.
040.
000.
040.
54Ba0.
170.
007.
104.
42B2.
300.
819.
979.
24Caa15.
242.
6531.
7729.
51Source:AltmanandSuggitt(1997).
Note:Datacoversperiodfrom1991-1996.
Thedefaultratesofthesyndicatedbankloansarequitesimilartothoseofthebonds,particularlyatthefive-yearhorizon.
Thismay,however,beinfluencedbytheuseofassignedratings.
ArecentstudybyFitchResearch(Grossman,Brennan,andVento1997)ontherecoveryexperienceofdefaultedsyndicatedbankloanscomplementstheAltmandefaultstudy.
Fitchexamines60distressedsyndicatedsecuredbankloansworthatotalofUSD25billionoverasix-yearperiodendinginJune1997.
Unliketheagencystudies,Fitchusesthemarketpriceofthedefaultedloanattheendoftherestructuringorbankruptcyperiodasaproxyforthefinalrealisationofvaluewithoutdiscountingthevaluebacktothepointofinitialdistress.
Distressedloansincludenotonlythoseindefault,butalsothosemakinginterestpaymentsbutthatthemarketvaluesatlessthan80centsonthedollar.
Thestudyfindsthatrecoveriesondistressedbankloansarehigh(82%)comparedtorecoveriesonseniorsubordinateddebt(42%)andsubordinateddebt(39%).
Furthermore,thestudyfindsthattheamountoftherecoveryisrelatedtotheindustryoftheissuer.
Asexpected,lowlevelsofhardassets,highlevelsofpledgedassets,orobsolescence-riskreducerecoverieswhileasolidbusinessfranchiseandhighqualityplantandpropertyincreaserecoveries.
Thestudyalsofindsthattheaveragedistressedbankloanremainsindistressfor19monthsduringwhichtimeitsreturnsexhibitamoderatecorrelationwiththestockmarket.
131ConclusionCreditratingssimplifyavastamountofinformationpertainingtocreditqualityintoasimpleordinalranking.
Thedefaultandrecoverystudiesconductedbytheagenciesandotherresearchersrelatethesecreditratingstoquantitativecreditriskinformationsuchasdefaultratesandexpectedlossratesondefaulteddebt.
Whilethisquantitativedetailisusefulforgaugingtheabsolutecreditriskassociatedwithadebtinstrument,thecreditratingremainsamoreaccurateassessmentoftherelativeratherthantheabsolutecreditriskofanissuer.
ReferencesAltman,EdwardI.
1989.
MeasuringCorporateBondMortalityandPerformance.
JournalofFinance44,no.
4:909-922.
Altman,EdwardI.
1993.
DefaultedBonds:Supply,DemandandInvestmentPerformance,1987-1992.
FinancialAnalystsJournalMay/June:55-60.
Altman,EdwardI.
,andHeatherJ.
Suggitt.
1997.
DefaultRatesintheSyndicatedBankLoanMarket:AMortalityAnalysis.
WorkingPaperSeries,no.
S-97-39.
NYUSalomonCenter.
Brand,Leo,andRezaBahar.
1999.
RatingsPerformance1998.
Standard&Poor'sCorporation.
Cantor,Richard,andJeromeS.
Fons.
1999.
RatingMethodology:TheEvolvingMeaningofMoody'sBondRatings.
Moody'sInvestorsService.
Caouette,John,EdwardAltman,andPaulNarayanan.
1998.
ManagingCreditRisk.
WileyPublishers.
Grossman,RobertJ.
,WilliamT.
Brennan,andJenniferVento.
1997.
SyndicatedBankLoanRecoveryStudy.
FitchInvestorsService,L.
P.
Kealhofer,Stephen,SherryKwok,andWenlongWeng.
1998.
UsesandAbusesofBondDefaultData.
KMVCorporation.
Keenan,SeanC.
1999.
SpecialComment:HistoricalDefaultRatesofCorporateBondIssuers,1920-1998.
Moody'sInvestorsService.
VandeCastle,Karen,andDavidKeisman.
1999.
RecoveringYourMoney:InsightsIntoLossesFromDefaults.
Standard&Poor'sCreditWeekJune16:29-34.
Ward,David,andGaryGriepentrog.
RiskandRewardinDefaultedBonds.
1993.
FinancialAnalystsJournalMay/June:61-65.
132SectionB.
StabilityofTransitionMatricesIntroductionBondratingshavebeenassignedtopubliclytradedindebtednesseversincetheestablishmentoftheratingsindustrybyMoody'sInvestorServicein1909.
66Thecreditopinionassignedbyaratingcompanyassessesanissuer'sabilitytomeetitsdebtobligationsinatimelyfashion.
Theseratingsaretypicallyfirstassignedatthetimeofinitialissuance.
Theagenciesalsoperformbothperiodicandmarket-basedreviewsofthoseratingsovertime.
Ratingschangesreflectanagency'sassessmentthatacompany'screditqualityhasimproved(upgrade)ordeteriorated(downgrade).
Theratingagencieshavebeensuccessfulinassessingtherelativelevelsofrisk,butthedegreeofthosechanges,orratingsmigration,andtheabsolutelevelofriskthatindividualratingsrepresenthavevariedovertimeandacrossratingscategories.
Theissueofcreditqualitymigrationisveryimportantforfixed-incomeinvestors,institutions,regulators,andmanagersofcreditrisk.
Investorsareconcernedwithmigrationofratingsbecauseitinfluencesthepriceofabond.
Institutionsareconcernedwithratingschangesbecauseofinternalpolicieslimitingthepercentageofbelow-investment-gradeloansthatbanksdesiretohold.
Regulatorsareconcernedwithratingssinceinsomecasestheydetermineinvestmenteligibilityofassetsandvaluationforcapitaldetermination.
Creditratingsmigrationalsoplaysamajorroleincreditriskmanagement.
67Finally,riskmanagementtools,suchasCreditMetrics,specificallyutilisecreditmigrationmeasuresasoneofitsprimaryinputs.
68Theratingagenciestracktheperformanceandstabilityoftheirratingsinwhatarecommonlyreferredtoas"transition"matrices.
Thesematricesassesstheissuers'"creditpaths",orratingshistory.
Theyareconstructedmainlyforriskmanagersandinvestorstoprovideinsightintoexpectedfuturepaths.
BothMoody'sandS&Preleaseannualorperiodicstudiesontheperformanceandstabilityoftheirratingsandpaths.
69Theprimarypurposeofthissectionistoprovideadetailedexaminationofthesetransitionmatricesandtheirstability.
TransitionmatricesTransitionmatricesmeasuretheprobabilityofacreditratingbeingupgradedordowngradedwithinaspecifictimeperiod.
S&PandMoody'sbothlookattheratingmigrationofcredit66In1922,thePoor'sCorporation(S&P'spredecessor)wasestablishedandbecameanactiveplayer.
BoththemajorraterscurrentlyratenearlyeverytaxablesecurityintheUSmarketthathasbeenregisteredwiththeSecurityandExchangeCommission.
67SeeLucas(1995)foradiscussiononhowOTCmarketparticipantsusedowngradeprovisionstoreducecounterpartyrisk.
68SeeGupton,FingerandBhatia(1997).
69Forexample,seeMoody'sInvestorService,GlobalCreditResearch,"HistoricalDefaultRatesofCorporateBondIssuers",January1999;Standard&Poor's,SpecialReport,"RatingsReport1998:CorporateDefaultsRiseSharplyin1999",(March1999).
133qualityinallratingscategoriesforvarioustimehorizons,includingone,five,and10years,andinsomecaseslonger.
Thetransitionmatricesissuedbythemajorratingscompaniesincludeallindustrialandtransportationcompanies,utilities,financialinstitutionsandsovereignsthathaveissuedlong-termdebttothepublic.
Transitionmatricesarecalculatedbycomparingbeginning-of-periodratingstoend-of-periodratings.
Transitionmatricesfocusontwodistinctpointsintime,typicallythefirstandlastdaysofayear,andignoreanyinterveningchanges.
Table1providesagenericexampleofatransitionmatrix.
Theverticalaxisshowsratingsatthebeginningofayear,whilethehorizontalaxisshowsratingsattheend.
Fromthesetwopoints,thematrixmeasurestheprobabilitythatanissuerwillexperienceanetchangefromtherowratingtothecolumnstatusattheendofaspecifiedtimeperiod.
Thevaluesalongthediagonals,inbold,representtheprobabilitythatanissuerwillhavethesameratingatthebeginningandendoftheperiod.
Theoff-diagonalelementsrepresentthepercentagesofupgradesordowngrades.
Movementtotherightofeachdiagonalelementrepresentsdowngrades,whilemovementstotheleftrepresentupgrades.
Forexample,ofalltheissuersratedAaaatthebeginningoftheyear,85.
44%wereratedAaaattheendoftheyear;9.
92%wereratedAa,and0.
98%wereratedA.
Table1:Moody'sOne-YearTransitionMatricesCorporateAverageOne-YearRatingTransitionMatrix,1980-1998Ratingto(%)InitialRatingAaaAaABaaBaBCaa-CDefaultWRAaa85.
449.
920.
980.
000.
030.
000.
000.
003.
63Aa1.
0485.
529.
210.
330.
140.
140.
000.
033.
59A0.
062.
7686.
575.
680.
710.
170.
010.
014.
03Baa0.
050.
326.
6880.
555.
720.
950.
080.
155.
49Ba0.
030.
070.
515.
2076.
517.
400.
491.
348.
46B0.
010.
040.
160.
606.
0776.
122.
546.
507.
96Caa-C0.
000.
000.
661.
053.
056.
1162.
9725.
160.
00WR-WithdrawnRatingSource:Moody'sInvestor'sService,January1999,"HistoricalDefaultRatesofCorporateBondIssuers,1920-1998".
RatingsuniverseTheMoody'sratingsuniverseincludesover15,200issuersthatsolddebtbetween1919and1999.
AsofJanuary1999,over4,600issuersheldcurrentMoody'sratings.
70S&Pissued7,328long-termratingsfrom1981throughtheendof1998.
AsofJanuary1999,around4,00070Moody'sSpecialComment"HistoricalDefaultRatesofCorporateBondIssuers,1920-1998",January1999.
134issuersheldS&Pratings.
71Theconstructionoftheratingagencies'transitionmatrices,andactuarialmortalityapproachtotransitionsemployedbyAltmanandKao,areoutlinedbrieflybelow.
72TransitiontrackingmethodsS&Ptransitionmatricesusetheimpliedseniorunsecuredratingofeachissuer,regardlessofthesizeofaparticularissueorthenumberofsharesoutstandingfromthatparticularissue.
S&Pproducetransitionmatricesbywhattheyrefertoas"staticpools".
Thesepoolsareformedonthefirstdayofeachyearandarefollowedfromthatpointforward.
Membershipinapoolremainsstatic,exceptforwithdrawnratings,called,ordefaultedissues.
73Theratingsarecomparedonthefirstandlastdayofeachyeartoconstructthetransitionmatrixforthatpool.
Everyyearanewstaticpoolisformedwithnewissuersfromthatyearandtheactiveissuersofthepreviouspool.
Moody's,similartoS&P,reliesuponanimpliedseniorunsecuredratingoftheissuer,ratherthantheratingsofindividualdebtinstruments.
Theytakeasimilarapproachtoconstructingthetransitionmatrices,buttheyrefertothepoolsas"dynamiccohorts".
Thecohortsaredynamicinthattheychangewhenratingsarewithdrawn,defaultedorcalled.
LikeS&Ptheyusethefirstandlastdaysoftheyearinconstructingcohorts.
Foreachperiod,thetransitionsexpresstheratioofissuerswhochangedtoissuerswhocouldhavechanged.
74Severalacademicstudieshavetakenaslightlydifferentapproachtomeasuringandreportingratingstransitions.
Altman(1989)andAltmanandKao(1992)werethefirsttotakeanapproachtoconstructingtransitionmatricesassessingthechangesfromaninitialbondrating,usuallyattimeofissuance.
Theyarguethatthisdistinctionisimportantbecauseofanageingorseasoningeffectthatisobservableintheearlyyearsafterissuanceandthatsuchaneffectgenerallydisappearswithinfourtofiveyears.
ThisresultisintuitivelyappealingbecauseasAltman(1992)notes,thatastimepassesstrongcompaniesareabletocallorrepurchasetheirdebtandrefinanceitwithlowercouponissues.
Thus,theremainingpoolsofissuersnaturallydisplayhigherdefault/transitionrates.
Asquithetal(1989),Altman(1993),JonsosonandFridson(1996)andCarty(1997)havedocumentedtheageingeffectelsewhereintheliterature.
Severalresearchersfocusingonhigh-yieldbondshavefoundalsofoundsimilarbondageingeffects.
Forexample,MoellerandMolina(1999)findthathigh-yieldbondsfaceanincreasingdefaultriskintime.
Besidescohortorpoolconstruction,therearetwootherimportantdifferencesinhowAltmanandKao(1992)constructtransitionmatrices.
First,AltmanandKaotransitionsarebasedon71StandardandPoor's,"RatingsPerformance1999,"p.
5and13.
72NeitherFitchnorDuffandPhelpsreleasetransitionmatrices.
IfonetriedtoconstructatransitionmatrixforFitch,however,itwouldbebiasedsincetheyonlyrevealratingsofborrowerselectingtomakethisinformationpublic.
73Mostofthewithdrawnratingsoccurwhenanentity'sonlyoutstandingissueispaidofforwhentheprogramratedisterminatedandthedebteliminated.
Somearealsoeliminatedduetolackofinformationormergersandacquisitions.
74Thisisidenticaltoa"hazard"rate.
135theratingsofspecificissues,ratherthantheimpliedseniorunsecuredratingofissuers.
Second,unliketheratingcompanies,AltmanandKao(1992)donotincludetheratingscategory"withdrawn"whenreportingtheirtransitionmatrices.
Awithdrawnratingcanmeanoneoftwothings:abondhasbeencalledorredeemedinanacquisition,oralternatively,thereisinsufficientinformationtoratethebond.
75AltmanandKao'smortality-basedapproachassumesthattheseissuerswerenotatriskofhavingtheirratingschangedandaredroppedfromtheanalysis.
76Theprimarydifferencethatariseswhencomparingthetwowaystoconstructtransitionmatricesisthatthepoolsorcohortstrackedbythemajorratingcompaniescontainportfoliosofbothseasonedandnew-issuebonds.
Theratersbelievethatthisapproachismoreappropriatefortworeasons.
First,theyarguethatifaninvestorfeelsthatthebusinessconditionsofthecurrentyeararesimilartothoseofapreviousyear,theymayconsultdirectlythatyear'scohorttoassesswhattransitionpatternstoexpect.
(Essentially,thisisanargumentthatmacro-phenomenaaremoreimportantthanthevintageeffectsthatthemortality-basedapproachpicksup.
)Second,theyarguethatbecausefewissuersdefaultearlyintheirratingshistory,theirapproachavoidsgivinglowerdefaultratesinperiodsofhighorincreasingratingsactivity.
Issuesofconstructionaside,comparingtransitionmatricesisproblematicbecauseofthedifferenttimeperiodsthattheraters'datacover.
Moody'sdatacover1920-1998;publishedS&Pdatacover1981-1998;andtheAltman/Kaostudiesusedatafromthe1970'sto1989.
Moreover,asdescribedbelow,changesinthenumberandtypesofdebtissues,theindustriesrated,andinitialcreditqualityoverthosetimeperiodsexacerbatethedifficultiesinmakingdirectcomparisonsoftransitionmatrices.
Theremainderofthissectionisoutlinedasfollows.
First,wewillreviewtheliteratureonseculartrendsassociatedwithcreditratingsthatmayaffecttheobservedtransitionmatrices.
77Forexample,weexaminegeneraltrendsinthenumberandtypesofissuerratingsandtheirpossibleimpactontransitions.
Thenextsectionexaminestheroleofcyclicalconsiderationstakenbyagencieswhenassigningtheseratingsandtheimplicationsofthesefactorsonobservedtransitionmatrices.
Finally,wewillreviewtheliteratureonratingschanges,ordrift,foracomparisonofthechangesmadebyMoody'sandS&Povervarioustimehorizons.
Thefinalsectionalsoexplorestheimpactofthedifferentmethodologiesonobserveddrift,andincludesadiscussionofthelimitationsinherentincomparingagencies'publishedtransitionmatrices.
75BothMoody'sandS&Psayratingsareneverwithdrawnduetopoorcreditquality.
76Theratingagenciesofferusersoftheirdatabasestheoptionofincludingorexcludingwithdrawnratings.
SeeCarty(1997)forafullerdiscussionofthetreatmentofwithdrawnratings.
77Thesecond-tieragencies,suchasFitch,andDuffandPhelps,haveproducedverylittleliteratureontheperformancesoftheirratings.
136SecularTrends:EvidencefromtheRatingsCompaniesChangesincreditqualityareofobviousinteresttoinvestorsandregulatorsalike.
Beforeexaminingchangesincreditqualityasmeasuredbyratings,however,itisimportanttoassesswhatthesechangesaremeasuringandhowthishaschangedovertime.
Oneofthemostfundamentalchangesincapitalmarketsoverthepastfewdecadesisthelargeincreaseinthenumberandtypesoffirmsacquiringratings.
Thissectionexaminessomeoftheseculartrendsthatmayimpactratingsperformanceandthestabilityoftransitionmatrices.
78NumberofratedissuesTable2showsthatthenumberofissuersratedbyS&Phasalmosttripledfromjustover1,300in1981,tomorethan4,000in1998.
Thisisreflectiveoftheincreaseduseofleveragebycorporationsandtheincreasedattractivenessoffixedincomesecuritiestoinvestors.
Table2alsoprovidesthebreakdownbetweeninvestmentgradeandspeculativegradeissuedratingsforthattimeperiod.
First,thereisanobviouscyclicalcomponenttoissuerratings,whichwillbeanalysedinthenextsection.
Moregenerally,speculativegradeissuershavebecomeamoreimportantcomponentoftherateduniverse.
ChangesinfirmsobtaininginitialratingsTable3separatesoutfromtheS&Pdatafirmsthatareobtaininganinitialrating,ratherthanexaminingtheoveralldistributionofallfirmsreceivingratings.
ThetableshowsthatS&Passignedinitialratingstoalmosttentimesasmanyfirmsin1998(720)asitdidin1981(75).
Italsoshowsthatinrecentyearsmorefinanciallyweakercompanies,bothintermsofnumberandpercentofratedissuers,seekfundinginthebondmarket.
7978Moody'sandS&Phaveverysimilarratingsuniverses.
Thus,thefiguresdescribedbelowfromeitheroftheuniversesarecloselyinterchangeable.
79Moody'shasexperiencedsimilartrends.
In1999over4,600issuersheldMoody'sratings,whilecomparablefiguresareabout2,500in1990,1500in1988,1,000in1970.
(Moody's(1999)p.
8).
137Table2:S&P:NumberofIssuersbyYearYearNumberofIssuersInvestmentGrade(percent)SpeculativeGrade(percent)19811,35477.
30%22.
70%19821,40876.
6023.
4019831,42976.
5023.
5019841,52375.
6024.
6019851,61673.
6026.
4019861,86771.
0029.
0019872,02765.
7034.
3019882,11064.
1035.
9019892,17665.
6034.
4019902,18667.
5032.
5019912,12772.
0028.
0019922,21776.
4023.
6019932,40076.
3023.
7019942,72673.
2026.
8019953,09277.
3022.
7019963,26472.
5027.
5019973,51671.
6028.
4019984,01467.
7032.
30Source:StandardandPoor's,"RatingsPerformance1998",(1999)Table3:S&PInitialRatings1981-1998YearTotalNewIssuesRatedInvestmentGrade(percent)SpeculativeGrade(percent)19817554.
67%45.
33%198210770.
1029.
90198317862.
9237.
08198415050.
0050.
00198532360.
6839.
32198632937.
6962.
31198731541.
2758.
73198825857.
2542.
25198921753.
9246.
08199014977.
1822.
82199124883.
8716.
13199231958.
6241.
38199353049.
8150.
19199451663.
7636.
24199535149.
2950.
71199649750.
3049.
70199773539.
3260.
68199872033.
1766.
83Source:StandardandPoor's,"RatingsPerformance:1998"p.
4(March1999).
138IndustrialcompositionTheindustrialcompositionoffirmsobtainingratingshaschangedovertime.
Forexample,intheearlypartofthecentury,railroadswerethemajortypeofentityseekingratings.
In1920,morethanhalfoftheissuersMoody'shadratedwererailroadcompanies,followedbyutilities,industrial,andfinancialcompanies.
80In1999,industrialcompaniesmakeupalmost39%ofthetotalnumberofratedfirms;non-bankfinancialcompaniesholdthesecondspotat17%;banksarethird,makingup14%ofratedfirms.
Otherchangesincludethetypesofratingsincludedincalculatingtransitions.
Forexample,beginningin1995,S&Pwideneditsselectioncriteriaholdersofotherlong-termcreditratings,includingcorporateratingsandcounterpartyratings.
81GeographictrendsThetypeoffirmsincludedintherateduniversehasalsochangedgeographically.
In1998,62%ofMoody'sratedissuerswerefromtheUnitedStates,followedbytheUnitedKingdom(5.
6%oftherateduniverse),Japan,(5.
5%)andCanada(3.
8%).
82Inrecentyears,however,membersoftheEuropeanCommunity(includingtheUK)andemergingmarkethavebeethefastestgrowingcomponentsoftherateduniverses.
Overtime,thathasdramaticallychangedthecompositionoftheratingsuniverse.
AccordingtoMoody'sthegrowthinthenumberofEC-domiciledissuerssince1980hasaveragedaround20%peryear,whiletheaverageforallothercountriesisunder6%.
Emergingmarketratedissuers,althoughtheystillmakeuparelativelysmallpercentageoftherateduniverse,grew45%annuallyfrom1993to1998.
Mostofthoseissuersreceivespeculativegrades,sotheyhavehadasignificantimpactonthebelow-investmentgradeuniverse.
Trendsinlong-termletterratingschanges/ratingsstabilityovertimeSeveralacademicstudieshaveexaminedthenumberandpercentagesofcreditupgradesanddowngradesovertime.
CartyandFons(1994)employMoody'sdatabaseofover4,700long-termissuesand2,400short-termissuescalculatetheannualpercentageofissuersupgradedordowngradedoneormoreletterratingsovervariousdecades.
Thenumeratorintheiranalysisisthenumberofissuersupgradedordowngraded;thedenominatoristhenumberofratingsatriskofchangeoverthattimeperiod.
Theyfindthattrendsinoverallcorporatecreditquality,asmeasuredbythepercentageofupgradesanddowngrades,havechangedovertime.
Forexample,duringthe1950-1980timeperiodtheyfindthatonaverage4.
77issueschangedratings,whiledriftaveragedamere-0.
07%.
Duringthe1980-1993timeperiod,however,theaveragenumberofissuersexperiencingratingschangesroseto12.
43%,whiledriftturnedmorenegativeat-4.
97%.
80SeeMoody'sSpecialReport,"HistoricalDefaultRateofCorporateBondIssuers",(January1999),p.
8.
81SeeS&PCreditWeek,"CorporateDefaultRiseSharply",March10,1999,p.
10.
82SeeMoody'sSpecialReport,"HistoricalDefaultRateofCorporateBondIssuers",(January1999),p.
8.
139LucasandLonski(1992)examinethecreditratingsofmorethan4,000ratedUSandinternationaldebtissuesfrom1970to1990.
Theyfindthatcorporatecredit-worthinessbecamemorevolatileoverthetimeperiodandthattheincreasedvolatilitywasaccompaniedbyadownwardtrendinratings.
Forexample,theyfindthat1%ofAAA-ratedcompanieswasdowngradedtospeculativegradewithinfiveyears,andtwo2%intenyears.
ForBaa-ratedcompaniesthecomparablefigureswere9%withinfiveyearsand15%withintenyears.
CantorandPacker(1994)findthattheexperiencesincethe1970sindicatedthatthecorrespondencetodefaultprobabilitiesissubjecttosubstantialchangeovertime.
Finally,arecentlypublishedpaperbyNickell,PerraudinandVarotto(2000)reportsdifferencesintransitionmatriceswhensubsettingoutsomecountries,suchasJapanandindustries,suchasbanking.
AltmanandKao(1992)analysethestabilityofnewlyissuedS&Pratingsfortwosub-periods-1970to1979and1980to1988.
Table4reproducestheirresults.
Thetableshowsthatforeveryratingandtimehorizon(1-5years)newlyratedissuesfromtheearlierperiodexhibitmorestability.
Asdiscussedearlier,however,thecompositionandsizeoftheratingsuniversechangedoverthistimeperiod.
140Table4:StabilityofNewlyIssuedRatingsRatingYearsAfterIssuance1970-79ProportionUnchanged1980-89ProportionUnchangedAAA197.
4%91.
6%294.
779.
2392.
166.
9487.
757.
9580.
049.
3AA195.
3%91.
1%287.
583.
1380.
375.
8473.
171.
8567.
668.
6A196.
3%88.
6292.
376.
9387.
968.
0483.
162.
2577.
758.
9BBB196.
3%87.
4293.
676.
8387.
963.
6480.
956.
8574.
352.
3BB198.
6%81.
0%288.
267.
1370.
158.
3462.
750.
0546.
930.
8B197.
5%93.
3%297.
581.
6382.
572.
7475.
857.
8568.
750.
5CCC1100%91.
6%281.
380.
4346.
771.
0440.
050.
0535.
721.
4N=2603N=-4592ChangesinratingsseverityThereissomenewliteratureonwhethertheseverityoftheratingsprocesshaschangedovertime.
Specifically,Blume,LimandMackinlay(1998)addresswhethertherecenttrendsofdowngradesversusupgradesincorporatebondratingsistheresultofdecliningcreditqualityofUScorporatedebtoralternativelytheratingsstandardshaveevolvedovertime.
Using141ratingsdatafrom1978-1995,theyarguethatratingagencieshavebecomemorestringent,implyinginpartthedownwardtrendinratingsisaresultofchangingstandards.
83ConclusionTherehasbeenadramaticincreaseinthenumber,types,geography,creditqualityandindustrialclassificationoffirmsobtainingratingsfromthemajorratingagencies.
Theratingsuniverseismuchdifferenttodaythanitwasevenadecadeago,withmoresovereigns,emerging-market,andspeculativegradeissuers,inadditiontoachangingindustrialcomposition.
Thelargeincreaseinthenumberofissuersandthewelldocumented"ageing"orseasoningeffect,suggestthattheratingsagenciestransitionmatricesneedtobeinterpretedwithcautionsincehavingalargepercentageofunseasonedcreditsordifferingpercentagesacrosspoolsorcohortsmaydisguiseothercreditqualitytrends.
Theoveralldownwardtrendininitialratingsovertimealsoimpactstheobservedtransitionprobabilities,sincetheseissueshavehistoricallybeenmorevolatile.
Finally,becauseratingscaptureoverallcreditqualitythatincludesdefaultprobability,lossseverity,financialstrengthandtransitionrisk-issuerswithidenticalratingsmayinfactexhibitvastlydifferenttransitionpaths.
84Giventhevarietyofchangesintherateduniverseovertime,itispossiblethattypesofissuerswithidenticalratingsmayhavevastlydifferenttransitionmatrices.
Anissuethattheliteratureexaminesiswhether,giventheenormousamountofactivityandincreasedrelianceoncreditratings,theratingscompanieshavealteredtheseverityoftheirassessmentsovertime.
Blume,LimandMcKinlay(1999)findsupportthishypothesis.
Ifthisisthecase,changesinratingsseverityovertimewillalsoinfluencetheobservedtransitionmatrices.
Finally,afullexplorationofthepropertiesoftransitionmatricesshouldincludemorethananassessmentoftheinfluenceofseculartrends.
Otherfactorsthatshouldbereviewedcarefullyincludethediscretenatureoftransitionmatrices,thepossiblelackofratingstimeliness,andthepossiblewidespreaddifferencesinissuerdefaultprobabilitiesofidenticallyratedissues.
85CyclicalConsiderationsAnimportantissueinanalysingratingsandtheirtransitionsishowtobestincorporatemacroeconomiccycles.
Cyclesareforcesexogenoustotheissuersuchaschangesininterestratesandincome,orfluctuationsinconsumerorbusinessconfidence.
Companiescannotcontroltheseevents,andsomecompaniesorcountriesareinfactmoresusceptiblethanotherstotheirinfluences.
Thetiming,lengthandseverityofcyclesareatbestdifficulttopredict,whichcomplicatesfurthertheirincorporationintocreditratings.
Finally,incorporatingcyclicalconditionsintoissuerratingsismadeevenmoredauntingfortheratingcompanies83Moody'sandS&Pratingshavebeenfairlycomparableovertheyears.
84AccordingtoMoody'ssectorssuchasbanks,sovereigns,investorownedutilities,localgovernmentsandsecuritiesfirmsaresubjecttohigher"transition"risk.
SeeMoody'sRatingsMethodology,"TheEvolvingMeaningofMoody'sBondRatings",p.
3(August1999).
85Kealhofer,KwokandWeng(1998)containadiscussionofsomeoftheseissues.
142becauseastheyareactiveinmorethan100countries.
Ideally,theywouldneedtoproduceaccurateforecastsforeachcountrytofullyincorporatetheseconditionsintotheirratings.
Giventhecomplexityandeffortsinvolvedinmacroeconomicforecasting,theratingsagenciesgiveonlymodestweighttocyclicaleconomicconditions.
Theyare,ingeneral,reluctanttomakeratingschangesbasedsimplyoncyclicalconsiderations.
86Still,giventhecurrentstateoftheeconomy,ratersaresupposedtoseparateoutequilibriumandcyclicalcomponentsofcompanieswhenassigningarating.
Thisismadeevenmorecomplexwhenwelookoverlong-termhorizonswherebystructuralfactors,suchastechnologyadvancesandchangesintastesandpreferences,alsooccur.
Allthesefactorsmakeitevenharderto"rateoverthecycle"sincetheyalsohaveoccurredatirregularintervals.
Asaconsequenceofratingacrossthecycle,theratingsagenciesexpectthattheactualmeasureddefaultratesbyratingcategorytoexhibitpronouncedcyclicalpatterns,risinginrecessionandfallinginrecoveries.
BothMoody'sandS&Pupgradesanddowngradesarefairlyconsistentwiththispattern.
Table2showsthatthecyclicalpatternsinthedistributionofratingsnumberofspeculativegradeissuersrisesduringrecessions.
Althoughtheratingsagenciesrateacrossthecycle,thismayhavedifferentimplicationsforinvestment-gradeandbelow-investment-gradeborrowers,whoaremorelikelytobesusceptibletocyclicalconditions.
Severalacademicpieceshaveexaminedtheimpactofeconomicconditionsonhigh-yieldborrowers.
HelwegeandKleiman(1997)examinetheimpactofeconomicshiftsonthedefaultratesofhigh-yieldcorporatebonds.
Theyfindthattheriskinessofbondsoutstanding,thetimethathaspassedsinceabondwasissued,aswellaseconomicconditionsinfluencethetransitiontodefault.
Wecansurmisethattheresultscanbegeneralisedtoforothernon-terminatingtransitions.
McDonaldandVandeGucht(1999)examinehigh-yieldbondsandfindtheseissuersaremorelikelytodefaultwheneconomicconditionshaveworsenedandnoimprovementisanticipated.
ArecentlypublishedpaperbyNickell,PerraudinandVarotto(2000)investigatestheimpactofthemacroeconomyonratingstransitions.
Theauthorsbreakupthestateoftheeconomyintothreedistinctstates:peak,normalandtroughdependingonwhetherrealGDPwasintheupper-,middle-orlower-thirdofgrowthratesofthesampleperiod(1970-1997).
Theyfindseveralinterestingresults.
First,theyfindthatlower-ratedbondshavelessratingsvolatility,andarelesspronetodowngrades.
Second,theyfindthatdefaultprobabilitiesareespeciallysensitivetothebusinesscycle.
Finally,theyfindthatsometransitionprobabilitiesarecounterintuitive,suchasthetransitionprobabilityfromCaatoCabeinghighestinnormaltimes.
Examiningtwostaticpoolsatdifferentpointsinthebusinesscyclecangivesomeideaofhowtransitionstootherstatesbesidesdefaultareinfluencedbycyclicalconditions.
Table5showsS&Pstaticpoolsfromthemostrecenttroughandpeak(tothatpointinthebusinesscycle)years,1991and1998.
Asshowninthetable,thepercentageofdowngradesishigherinthetroughperiodthanatthepeak.
Forexample,9.
76%ofAAAratingsweredowngradedtoAAcomparedto6.
06%in1998.
Similarresultsholdfortheotherinitialratings.
Another86Moody'sRatingMethodology,"TheEvolvingMeaningofMoody'sBondRatings",(August1999),p.
7.
143intuitivelyappealingresultinthetableisthatthepercentageofwithdrawnratingsarehigherduringtheboomforhigherratedissuers,mostlikelyduetocallsandrefinancing.
Thesecomparisonsmustbetemperedbythefactthatthetwopoolsmaydiffergreatlybythefactorsdiscussedinthefirstsection,including"ageing",industrialcompositionandgeography.
ConclusionThetransitionmatricesprovidedbytheratingscompaniesattempttorateacrossthecycle.
Cyclicaleconomicconditions,however,affecttheratingsandtheirstability.
Infact,theswingsinratingsaremorepronouncedduringrecessionsandboomperiodswithmoreissuersdowngradedduringrecessionsandmoreupgradedduringexpansionaryperiods.
Asonewouldsurmise,speculative-gradeissuersaremoresusceptibletoeconomiccyclesthaninvestment-gradeissuers.
Nickell,PerraudinandVarotto(2000)findthatlow-ratedbondshavelessratingsvolatilityduringbusinesspeaksandthatdefaultprobabilitiesareespeciallysensitivetothebusinesscycle.
Thegreatersensitivityofthemigrationtodefaulttothebusinesscycleisnotsurprisinggiventhattheratingscompaniesinitiateotherratingschanges,whiledefaultsarebeyondtheircontrol.
Theacademicliteraturehasfocusedontheimpactofeconomicconditionsonhigh-yieldbondstothefinalmigration-default.
Twoseparatestudies,HelwegeandKleiman(1997)andMcDonaldandVandeGucht(1999)findthateconomicconditionsareanimportantdeterminantofthetransitiontodefaultforhigh-yieldissues.
AquickcomparisonofS&Ppoolsoverbothboomandbustyearsshowsthatnotonlyspeculativegradeissuesexhibithigherpropensitytobeingdowngraded,butinvestmentgradeissuersdoaswell.
144Table5:One-YearTransitionMatricesPanelA.
S&P1991StaticPoolOne-YearTransitionMatrixTroughoftheBusinessCycleRatingto(%)InitialRatingAAAAAABBBBBBCCCDN.
R.
AAA89.
029.
760.
610.
000.
000.
000.
000.
000.
61AA0.
0089.
957.
790.
000.
000.
000.
000.
002.
26A0.
170.
6790.
526.
660.
170.
000.
000.
001.
83BBB0.
000.
825.
1682.
075.
710.
540.
540.
544.
62BB0.
000.
000.
007.
1473.
957.
141.
682.
527.
56B0.
000.
340.
000.
344.
8068.
153.
7713.
019.
59CCC0.
000.
000.
001.
524.
556.
0645.
4630.
3012.
12PanelB.
S&P1998StaticPoolOne-YearTransitionMatrixPeakoftheBusinessCycleInitialRatingAAAAAABBBBBBCCCDN.
R.
AAA90.
306.
060.
000.
610.
000.
000.
000.
003.
03AA0.
1889.
645.
710.
180.
000.
000.
000.
004.
29A0.
091.
4687.
225.
110.
180.
000.
000.
005.
94BBB0.
320.
002.
7984.
934.
460.
670.
220.
346.
59BB0.
000.
160.
165.
3375.
445.
982.
750.
659.
21B0.
000.
000.
150.
626.
1676.
275.
094.
477.
24CCC0.
000.
003.
330.
000.
0020.
0033.
3336.
676.
67N.
R.
Indicatesnoratingattheterminaldate.
RatingsDriftIntroductionAfterinitialissuanceandassignmentofratings,theratingsagenciesperformbothperiodicandmarket-basedreviewsofthoseratings.
Oftenthesereviewsresultinchange,ordrift,intheirratingssignifyingachangeinoverallcreditquality.
Therearethreesourcesofinformationonratingsdrift.
Thefirsttwoarefromtheacademicarena,includingstudiesbyAltmanandKao,(1992a,1992b)whichusedataforallS&Pratingsfrom1970to1989.
ThesecondsetofstudiesusesMoody'sdata,whichspansalongertimeperiod.
ThesestudiesincludeLucasandLonski(1992),CartyandFons(1993)andCarty(1997).
ThethirdsourceincludesisthesetofperiodicstudiesthatMoody'sandS&Prelease.
Inlightofthedramaticchangestotheratingsuniversediscussedintheprevioussection,thetimeperiodchosenisasimportantasthestudyoneexamines.
Theremainderofthissectionbrieflysummarisestheliteratureonratingsdriftovervarioustimehorizons,ratingsissuersandtransitionmethodology.
145Ratingsdrift:acomparisonTheobservedstabilityandperformanceofthetransitionmatricesvariesacrosstheratingsagenciesandindependentresearchers.
Caouetteetal(1998)attempttoputthisissueinperspectivebyexaminingtransitionmatricesforthreedifferenttimehorizons.
Thesetransitionmatrices,reproducedinTables6,7,and8,comparethetwomajorratersandtheAltmanandKaoapproach,whichincludeswithdrawnratings.
Theprincipallimitationwhencomparingthemisthatthetransitionmatricescoverdifferenttimeperiods.
TheMoody'stransitionmatricesareforthe1938-1996cohorts,theS&Paverageratingstransitionsarefortheirstaticpoolofbondsfortheyears1981-1996,andtheAltmanandKao(1992)resultsarefor1971-1989timeperiod.
Giventhesecularchangesthatoccurredoverthistimeperiod,thecomparisonsshouldbemadewithcaution.
StillthetablesyieldsomeinterestingfindingsabouttheMoody'sandS&PapproachascomparedtotheAltmanandKaoapproach,aswellassomedifferencesinratingsdriftbetweenthetwomajorratingcompanies.
Whatevermethodologyischosen,thelongerthetimehorizon,thelessstabletheratings.
Forexample,asdisplayedinTable6,AltmanandKaofindthat94.
3%oftheAAAissuesretaintheratingafteroneyear;Moody'sreportsthisfigureat88.
3%,andS&Preportsitat88.
5%.
Table7showsthatthosefiguresare69.
8%,60.
8%,and54.
0%overfiveyears,forAltmanandKao,Moody's,andS&P,respectively.
Asthetimehorizonlengthens,Caouetteetalreportthatthedifferencesbetweentheratingagencies'resultsandtheAltmanandKaoresultsbegintowiden.
Astheyexplain,thisisduetotwofactors:the"ageingeffect"andtheomissionofthewithdrawnratingsfromAltmanandKao'sobservations.
BecausetheAltmanandKaotabulationsincorporatethefactthatissuersfaceachangingriskorhazardofaratingschangesovertime,themorestableresultsusingtheirmethodologyprovidesstrongevidenceofthe"ageing"phenomenon.
Carty(1997)analysestheimpactofthewithdrawncategory,andfindsthataftercorrectingforthisdifferencethattheone-yeartransitionmatricesfortheratingagenciesbegintoconvergeontheAltmanandKaoresults.
CaouetteetalalsonotethattheAltmanandKaoresultsdivergefromtheratings'companiesresultsandthatthesedifferencesaremostpronouncedatthelowerendofthequalityspectrum.
Foreverycreditratingandtimehorizontheyfindahigherpercentageofissuerretainingtheirratings.
WhencomparingMoody'sandS&Pratings,Caouetteetalfindthatthematricesdivergeafterfiveyearsbuttheydonotofferanyexplanationexceptthedifferenttimeperiodsunderinvestigation.
AltmanandKao(1992),usingnewlyissuedratingsfromS&Pdataanalysetheratingsmigrationinfourseparatesub-periods–1970-79,1980-89,1977-82and1983-89.
Theyfindthatoveraone-tofive-yearhorizonthatAAA-ratedissueshadthegreateststability.
AsAltman(1992)arguesthisisnotasurprisingresultgiventhattheyonlycangoinonedirection–downward.
Whenlookingpastafive-yearhorizon,however,theyfindthatAratingswerethemoststableratingscategoriesforyears6through10.
TheauthorsalsofindthatBB-ratedissuesaretheleaststable–only86.
1%retainedtheirratingsafteroneyear,40.
8%over5years,and21.
6%overtenyears.
AltmanandKao(1992)alsoexaminemovementsinratingsgiventhedirectionofthefirstchange.
Astheynote,ratingschangesarethenormratherthantheexception,sotheyposethequestionofwhetherinitialchangesrevealanythingaboutfutureratingschanges.
Theyfindsignificantserialcorrelationwhentheinitialchangeisadowngradeandnonewhentheinitialchangeisanupgrade.
Theytesttheseresultsforvarioussub-periodsandfindthemfairlyrobusttotimeperiodsandissuerindustrialclassification.
146Table6:One-YearTransitionMatrixAaaAAAAaAAAABaaBBBBaBBBBBCaaCCCDef.
RWAAA(A/K)Aaa(M)AAA(S&P)94.
388.
388.
55.
56.
28.
10.
11.
00.
70.
00.
20.
10.
00.
00.
10.
00.
00.
00.
00.
00.
00.
00.
00.
0--4.
32.
6AAAaAA0.
71.
20.
692.
686.
888.
56.
45.
87.
60.
20.
70.
60.
10.
20.
10.
10.
00.
10.
00.
00.
00.
00.
00.
0--5.
42.
4AAA0.
00.
70.
12.
62.
32.
392.
186.
187.
64.
74.
75.
00.
30.
60.
70.
20.
10.
20.
00.
00.
00.
00.
00.
4--6.
03.
6BBBBaaBBB0.
00.
00.
00.
00.
30.
35.
53.
95.
590.
082.
582.
52.
84.
74.
71.
00.
61.
00.
10.
10.
10.
30.
30.
2--7.
75.
7BBBaBB0.
00.
00.
00.
00.
10.
10.
00.
40.
66.
84.
67.
086.
179.
073.
86.
35.
07.
60.
90.
40.
90.
01.
11.
0--9.
48.
9BBB0.
00.
00.
00.
00.
00.
10.
20.
10.
21.
60.
60.
41.
75.
86.
093.
756.
372.
81.
73.
53.
41.
13.
54.
9--10.
512.
2CCCCaaCCC0.
00.
00.
20.
00.
00.
00.
00.
00.
30.
00.
31.
00.
01.
32.
22.
85.
39.
692.
571.
953.
14.
612.
419.
3--8.
814.
2SourceandKeyA/K=AltmanandKao(1992a)newlyissuedbonds.
M=Moody's(1983-1998)fromCarty(1997)-cohortofbonds.
S&P=S&P(1981-1996)fromS&P(1997a)-staticpoolofbonds.
RW-ratingswithdrawn.
147Table7:Five-YearTransitionMatrixAaaAAAAaAAAABaaBBBBaBBBBBCaaCCCDef.
RWAAA(A/K)Aaa(M)AAA(S&P)69.
860.
854.
023.
515.
223.
52.
94.
36.
03.
61.
01.
70.
10.
50.
50.
00.
10.
20.
10.
00.
20.
10.
20.
2--18.
013.
8AAAaAA2.
43.
42.
467.
154.
153.
422.
515.
924.
45.
03.
44.
01.
01.
20.
90.
30.
20.
80.
10.
00.
11.
70.
60.
4--21.
113.
7AAA0.
40.
30.
29.
25.
97.
072.
055.
753.
415.
110.
314.
91.
92.
63.
00.
70.
71.
80.
00.
10.
30.
70.
60.
7--23.
418.
7BBBBaaBBB0.
40.
10.
21.
60.
91.
319.
610.
015.
565.
447.
141.
57.
68.
09.
61.
72.
03.
41.
90.
30.
81.
82.
32.
0--29.
325.
7BBBaBB0.
00.
10.
10.
00.
30.
47.
71.
92.
920.
410.
413.
440.
836.
519.
916.
58.
19.
67.
81.
31.
76.
85.
911.
4--35.
640.
2BBB0.
10.
00.
00.
00.
10.
12.
30.
50.
64.
02.
42.
87.
710.
39.
553.
332.
216.
611.
83.
52.
520.
812.
922.
5--38.
245.
4CCCCaaCCC0.
00.
00.
20.
00.
00.
02.
60.
00.
93.
61.
62.
42.
04.
03.
530.
77.
86.
126.
529.
65.
434.
028.
042.
0--29.
039.
5SourceandKeyA/K=AltmanandKao(1992a)newlyissuedbonds.
M=Moody's(1983-1998)fromCarty(1997)-cohortofbonds.
S&P=S&P(1981-1996)fromS&P(1997a)-staticpoolofbonds.
RW-ratingswithdrawn.
148Table8:Ten-YearTransitionMatrixAaaAAAAaAAAABaaBBBBaBBBBBCaaCCCDef.
RWAAA(A/K)Aaa(M)AAA(P)52.
141.
630.
735.
619.
024.
77.
16.
112.
04.
62.
16.
50.
00.
80.
50.
40.
20.
00.
00.
00.
00.
20.
61.
1--29.
624.
6AA(A/K)AaAA3.
54.
53.
145.
733.
430.
327.
119.
829.
219.
04.
96.
42.
42.
21.
50.
20.
60.
40.
00.
10.
22.
11.
51.
8--33.
127.
1A(A/K)AA0.
80.
30.
517.
36.
57.
560.
938.
633.
520.
011.
115.
83.
43.
33.
90.
91.
11.
70.
60.
20.
21.
12.
22.
4--36.
734.
5BBB(A/K)BaaBBB0.
00.
80.
32.
81.
31.
736.
111.
714.
642.
330.
024.
58.
26.
46.
14.
62.
11.
91.
90.
30.
14.
14.
15.
2--43.
845.
7BB(A/K)BaBB0.
00.
10.
10.
00.
30.
110.
33.
03.
725.
510.
110.
320.
617.
16.
612.
56.
13.
817.
21.
30.
513.
99.
920.
0--52.
154.
8B(A/K)BB0.
00.
00.
00.
00.
10.
15.
70.
80.
78.
62.
72.
96.
77.
84.
740.
913.
82.
26.
62.
10.
631.
518.
628.
6--54.
161.
2CCC(A/K)CaaCCC--0.
00.
0--0.
00.
0--0.
00.
6--2.
12.
3--3.
24.
6--5.
12.
9--13.
50.
0--36.
936.
2--39.
253.
5SourceandKeyA/K=AltmanandKao(1992a)newlyissuedbonds.
M=Moody's(1983-1998)fromCarty(1997)-cohortofbonds.
S&P=S&P(1981-1996)fromS&P(1997a)-staticpoolofbonds.
RW-ratingswithdrawn.
GiventhesimilaritiesintheapproachesS&PandMoody'susetocreatepoolsandthatthesimilaritiesintheirratingsuniverses,itisusefultolookattransitionmatricesforsimilartimeperiodsandhorizons.
Table9showsbothagencies'averageone-yeartransitionsoversimilartimeperiods.
ForissuerswitharatingaboveaBB,S&Pratingsappearmorestable.
Forexample,S&PissuesratedAAA,retaintheirrating89.
48%ofthetimeforaone-yeartimehorizon,whilethissamefigureis85.
44%forMoody's.
SimilarresultsholduntilwereachtheBBratedissues.
ForissuesratedBBandbelow,Moody'sissuersaremorelikelytoretaintheirratingsoveraone-yeartimehorizon.
Thisappearstobedue,atleastinpart,tothefactthatS&Pismorelikelytowithdrawratingsforissuersbelowthatgrade.
149Table9:CorporateAverageOne-YearRatingsTransitionMatrix:Moody'sandS&PMoody'sS&PAAAAaaAaAAAABaaBBBBaBBBBBCaaCCCDef.
RWAaaAAA85.
4489.
489.
927.
260.
980.
470.
000.
080.
030.
040.
000.
000.
000.
000.
00.
003.
632.
67AaAA1.
040.
6285.
5288.
999.
216.
550.
330.
580.
140.
060.
140.
030.
000.
000.
030.
003.
593.
07AA0.
060.
052.
762.
1886.
5787.
955.
684.
910.
710.
540.
170.
010.
010.
040.
010.
044.
034.
07BaaBBB0.
050.
040.
320.
256.
685.
2380.
5582.
665.
724.
540.
950.
160.
080.
220.
150.
225.
495.
95BaBB0.
030.
040.
070.
090.
510.
555.
207.
0476.
5173.
987.
407.
170.
490.
921.
340.
928.
469.
21BB0.
010.
000.
040.
090.
160.
250.
600.
416.
076.
1476.
1273.
152.
544.
826.
504.
827.
9611.
65CaaCCC0.
000.
160.
000.
000.
660.
321.
050.
643.
052.
096.
1110.
4362.
9752.
0126.
1620.
390.
0013.
97M=Moody's1980-1998fromMoody'sInvestorServices(1999)-cohortaverageS&P=S&P1981-1998fromS&P(1999)-staticpoolofbonds.
RW=ratingswithdrawnConclusionTheobservedmigrationpatternsintransitionmatriceshingecriticallyonthetimeperiodexamined,themethodologyemployed,andtheratingssystems.
Drift,nomatterhowitismeasured,however,haschangedovertime.
Inrecentyears,therehasbeenadownwardtrendinratingsdrift.
Thisisdriveninpartbythetypesoffirmsobtainingratingsandthefactthatlower-ratedfirmsaremorevolatile.
Severaltransitionmatricesfocusingondifferenttimeperiods,ratersandmethodologiesfromdifferentstudieswereexaminedtoassistinexaminingtheseissues.
Giventhelimitedamountofempiricalresearchontheimpactofeachoftheseissues,moreresearchneedstobedone.
AppendixSeveralresearchershavearguedthattheratingstransitionsprovidedbytheratingscompaniesarelimitedintheirusesincethe"averages"reportedaremisleadingandtimelinessofratingschangesdistortstheirreportedstability.
Forexample,Kealhofer,KwokandWeng(1998)arguethatthehistoricaltransition(default)ratescandeviatesignificantlyfromtheactualtransition(default)ratesandthattherearesubstantialdifferencesofdefaultwithinbondratinggrades.
Theybelievethatthisistruefortworeasons.
First,theyarguethatthisbiasisinherentbecauseofthediscretenatureofratings.
Second,theyarguethatthelackoftimelinessinratingscausesasignificantbiasintransitionprobabilities–theprobabilityofremainingatthesamequalityisoverstatedbyaboutdoubleformostgrades,whereasotherprobabilitiesareunderstated.
150Second,theissueofratingstimelinesscametoaheadduringtheAsiancrisiswhentheratingsagenciesdowngradedthebondsnotlongbeforethetimethattheinformationwasreleasedpublicly.
87Kealhofer,KwokandWeng(1998)examinesomeoftheseissuesbyconstructingadefaultmatrixbasedonEDFdefaultratesratherthanratingsgrades.
88Theyclassifyfirmsintodifferentdefaultrangesthataretheratingsthatcorrespondtorangesthataretypicalforaratingsgrade(e.
g.
AAA:lessthan.
02%,AA:.
03%to.
06%andsoforth).
TheyconstructaOne-YearTransitionMatricesastabulatedfromexpecteddefaultfrequenciesforthe1990-1995bymappingthestateddefaultratesforeachgradeandthenfindingthestabilityoftheKMVdefaultequivalent.
TheyfindevidencethattheEDF'sarelessstablethanratings,andarguethattheagencyratingscanbeconsideredrelativelysticky.
Table10:One-YearTransitionRatesBasedonKMVEDFRangesRatingto(%)InitialRating1(AAA)2(AA)3(A)4(BBB)5(BB)6(B)7(CCC)8(Default)1(AAA66.
2622.
227.
372.
450.
860.
670.
140.
022(AA)21.
6643.
0425.
836.
561.
990.
680.
200.
043(A)2.
7620.
3444.
1922.
947.
421.
970.
280.
104(BBB)0.
302.
8022.
6342.
5423.
526.
951.
000.
265(BB)0.
080.
243.
6922.
9344.
4124.
533.
410.
716(B)0.
010.
050.
393.
4820.
4753.
0020.
582.
017(CCC)0.
000.
010.
090.
261.
7917.
7769.
9410.
1387SeeCaouetteetal(1998)p.
79ortheEconomist(1997).
88Theusealsopubliclytradedcompanies,notonlythosewithimpliedseniordefaultratings.
151ReferencesAltman,Edward(1989).
"MeasuringCorporatebondMortalityandPerformance",JournalofFinance(Sept.
),909-922.
Altman,Edward(1998).
"Theimportanceandsubtletyofcreditratingmigration",JournalofBankingandFinance,1231-1247.
AltmanEdwardandDuenLiKao(1992).
"RatingsDriftofHighYieldBonds",JournalofFixedIncome,15-20.
AltmanEdwardandDuenLiKao(1992).
"TheImplicationsofCorporateBondRatingsDrift",FinancialAnalystsJournal,(May/June)64-75.
Asquith,Paul,DavidMullinsandEricWolff(1989)"OriginalIssueHighYieldBonds:AgeingAnalysesofDefault,Exchanges,andCalls",JournalofFinance,(September)923-952.
Blume,Marshall,FelixLim,andA.
CraigMackinlay(1998).
"TheDecliningCreditQualityofUSCorporateDebt",JournalofFinance(August)1389-1413.
Caouette,John,EdwardAltmanandPaulNarayanan(1998).
ManagingCreditRisk:TheNextGreatFinancialChallenge,JohnWileyandSons.
CantorandPacker(1994).
"TheCreditRatingsIndustry",FederalReserveBankofNewYorkQuarterlyReview,1-26.
Carty,Lea(1997).
"Moody'sRatingMigrationandCreditQualityCorrelation,1920-1996",Moody'sSpecialReport(July).
Carty,LeaandFons(1994).
"MeasuringChangesinCorporateCreditQuality",JournalofFixedIncome,(June)27-41.
Gupton,G.
M.
,C.
C.
Finger,andN.
Bhatia(1997),CreditMetrics:TheBenchmarkforUnderstandingCreditRisk.
Technicaldocument.
NewYork:J.
P.
Morgan,Inc.
Helwege,JeanandPaulKlieman(1997).
"UnderstandingAggregateDefaultRatesofHighYieldBonds",JournalofFixedIncome,(June)55-61.
JonsonnJ.
G.
andM.
S.
Fridson(1996).
"ForecastingDefaultRatesonHighYieldbonds",JournalofFixedIncome,69-77.
Kealhofer,Stephen,SherryKwok,andWenlongWeng(1998).
"UsesandAbusesofBondDefaultRates",KMVProprietaryDocument.
Lucas,D.
(1995).
"TheEffectivenessofReducingDowngradeProvisionsinReducingCounterpartyRisk",JournalofFixedIncome,7-14(March).
LucasandLonski(1992).
"ChangesinCorporateCreditQuality1970-1990",JournalofFixedIncome,7-14.
152McDonald,CynthiaandLindaM.
VandeGucht(1999).
"HighYieldBondDefaultandCallRisk",ReviewofEconomicsandStatistics81(3),409-419.
Moeller,ThomasandCarlosMolina(1999).
"LiveandLetDie:TheSurvivalandDefaultofOriginalIssueHigh-YieldBonds",UniversityofTexas,mimeo.
Moody'sRatingsMethodology(1999)"TheEvolvingMeaningofMoody'sBondRatings",(August).
Moody'sSpecialReport(1993).
"CorporateBondDefaultsandDefaultRates",Moody's(January).
Moody'sSpecialComment(1999)"HistoricalDefaultRatesofCorporateBondIssuers,1920-1998",(January).
Moody'sInvestorService,GlobalCreditResearch(1999).
"HistoricalDefaultRatesofCorporateBondIssuers",(January).
Nickell,Pamela,WilliamPerraudinandSimoneVarotto(2000).
"StabilityofTransitionMatrices",JournalofBanking&Finance,24(1&2),203-227.
StandardandPoor's(1999).
"RatingsPerformance1998"(March).
StandardandPoor's(1999).
"CorporateDefaultsRiseSharply",CreditWeek,(March10),9-16.
153SectionC.
ConsistencyAcrossSectorsIntroductionTheBaselCommitteehasproposedagreaterrelianceonexternalratingstodeterminetheriskweightingofbankingbookassets.
Accordingtotheproposednewstandardisedapproach,theriskweightingmaydependbothoncreditratingandsector(i.
e.
sovereigns,banks,orcorporates).
SovereigncreditshavereducedriskweightingsrelativetobanksatallratingsAandhigher,89andatallratingsrelativetocorporatesatBBBandhigher.
BankcreditshavelowerriskweightingsthancorporatecreditsatAandBBBratings.
Atotherratings,thechargesareidenticalacrosssectors.
Thispaperexaminestheextenttowhichsectoraldifferencesexistfordefaultratesandrecoveriesincasesofdefault.
Consistentwiththedistinctionsofthenewapproach,wefocusonthedifferencesamongsovereign,bank,andcorporateratings.
Whenbiasesareapparent,aretheyconsistentwiththeasymmetriesintheweightsproposedbytheBaselCommitteeIdenticalriskweightsacrosssectorsmaybemosteffectivewhenratingsindifferentsectorsreflectsimilarlevelsofexpectedloss;differentweightsmaybemosteffectivewhenratingsreflectdifferentlevelsofexpectedloss.
Boththeliteratureonthistopic,andourownanalysistofollow,arebasedonthehistoricalexperienceoftheratingsofthetwomajoragencies,Moody'sInvestorsServiceandStandardandPoor's.
Sinceadowngradetoalowerratingimpliesthataneventualdefaultismorelikely,wealsoexaminesectoraldifferencesinthedegreeofdowngraderisk.
Wealsoinvestigatetheextenttowhichobligordomicilehasaffectedtheassociationofcreditratingswithdefaultprobabilities,recoveries,anddowngrades.
MosthistoricalcreditratinganddefaultdataaretheratingsofUSentities,andissuedbyMoody'sandS&P,agenciesheadquarteredintheUnitedStates.
If,assomesuggest,thenewstandardisedapproachgreatlyincreasestheimportanceofMoody'sandS&P'sratingsofnon-USentities,thenthedegreetowhichtheseratingsareconsistentacrossobligordomicileisofinterest.
Internationaldifferencesinaccountingandlegalsystems,businesspractices,andtheroleofgovernmentintheeconomymaymakeitdifficulttocomparethedefaultriskofbondissuersdomiciledindifferentcountries.
OurempiricalresultssuggestthatMoody'sandS&Phavenotbeencompletelysuccessfulincalibratingtheirratingssothatlike-ratedbondsofdifferentsectorsareexposedtosimilarlevelsofdefaultrisk.
USbanksexperiencedsignificantlymoredefaultsthanUSindustrialfirmsoverthisperiod,takingtheyearandMoody'sratingasgiven.
USnon-bankfinancialfirmshavehadgreaterdefaultsthanUSindustrialfirms,takingtheyearandS&Pratingasgiven.
Theseresultsareatoddswiththeproposalthat,forsomeratinglevels,bankandsecuritiesfirmobligationscarryalowercapitalrequirementthananotherwiseidenticalliabilityofanindustrialfirm.
Nevertheless,itisworthnotingthatoursampleperiodmaybesomewhatatypicalfortheUSinthatitincludedanintervalinwhichanunusuallylargenumberofdepositoryinstitutionsfailed.
89Hereweignoretheproposedfirstoptionforweightingclaimsonbanks,basedontheratingofthesovereigninwhichitisdomiciled,andconsideronlythesecondoptionbasedonthebankratingitself.
154Limitedavailabilityofdataontheratingsofnon-USfirmsmadeitdifficulttoreachstatisticallyrobustconclusionsastowhethertheratingagencieshavebeenmoresuccessfulincalibratingtheirratingsacrossgeographicallydistinctissuers.
Apparentgeographicaldiscrepanciesindefaultratesarenolongersignificantoncewesimultaneouslycontrolforthecreditratingandtimeperiod.
OverviewofliteratureTheliteratureonsectoraldifferencesinthemeasurementofcreditriskisfairlylimited.
Oneareaoffocusissplitratings,thatis,whenthesameissuerreceivesdifferentratingsfromthevariousagencies.
Theliteraturealsocoverstheassociationofratingsandspreads,andtheassociationofratingswithsubsequentoutcomes,includingratingtransitionanddefault.
Anumberofthestudiesreportevidenceofsectoraldifferencesinindirectmeasuresoftheriskofdefaultloss,suchasyieldspreads,therelativelikelihoodofsubsequentupgradesanddowngrades,recoveryrates,andthefrequencyofsplitratings,whichareunderstoodtoreflectuncertaintyabouttheissuer'screditquality.
Inparticular,resultsinthisliteraturesuggestthat,holdingratingsconstant,bankbondsmayberiskierthanindustrialbonds,buttheresultsaremixedforcomparisonsbetweensovereignandcorporatebondsandforgeographicaldistinctions.
Fewpapersaddressdefaultratesdirectly,andweknowofnopreviousworkthatundertakesasystematicstatisticalanalysisofsectoraldifferencesindefaultrates.
Turningtosplitratings,DonaldMorgan(1998)measuredthefrequencyofcreditratingagencydisagreementinthebankingversusothersectors.
ConsistentwithCantorandPacker(1994),hefindsthatsplitratingstendtobemorefrequentinbankingthaninothersectors.
CantorandPacker(1995)findthatsplitratingsaremorecommonforlower-ratedsovereignsthanlower-ratedUScorporatesandlesscommonforhigherrated(AAA/AA)ratedsovereignsthanhigher-ratedUScorporates.
Theseresultssuggestgreateropacityinthemeasurementofcreditriskforbanksrelativetocorporates,forlower-ratedsovereignsrelativetocorporates,andlessopacityforhigher-ratedsovereignsrelativetocorporates.
Thefirstresultmaybeinconsistentwithoneoftheasymmetriesofthenewproposedcapitalcharges(thelowercapitalchargeforbanksrelativetocorporatecreditsatsomeratinglevels).
Thelastresultmaybeconsistentwithanotherasymmetry(thelowercapitalchargeforsovereignsrelativetocorporatecreditsatsomeratinglevels).
Withregardtoobligordomicile,twopapersmeritmention.
AnexaminationofsplitratingsbyBeattieandSearle(1992)suggeststhatagenciesjudgeissuersfromtheirowncountrymoreleniently.
However,CantorandPacker(1994)foundthat,forratingsofinternationalbanks,observeddifferencesbetweenhomeandforeignratingsreflectedprincipallydifferencesinthescalesofindividualratingsagencies,ratherthanhome-countrybias.
Therelationshipofratingsandspreadsalsoappearstodifferbysector.
CantorandPacker(1995),andJacksonandPerraudin(1999)documentatendencyforspreadstobehigherforsovereigncreditsatlowercreditratings(BBBandlower)thansimilarlyratedcorporatecredits.
Whilethismaybeduetolowerexpectedrecoveryondefaultedsovereignbondsthancorporatebonds,thereistoolittlerecenthistoryofdefaultedsovereignbondstotestthisexplanation.
Spreadsonbankdebenturesalsoappeartohavebeengreaterthroughoutthe1990sthanspreadsoncomparablyratedcorporatebonds(JacksonandPerraudin,1999).
Recoveriesmightaccountforsomeofthisdifference,asAltmanandKishore(1996)reportUSindustrydifferencesin"recovery"(i.
e.
salvage)ratesondefaultedbonds,andfindsthatfinancialinstitutionstendtohavelowerrecoveryratesthanindustrials.
155Nickell,Perraudin,andVarotto(2000)focusonrating"transitions"(alsoknownasrating"migrations"),andfindthatbankstendtohavelessstableratingsthanindustrials.
Higherratedbankshavemoredowngrades,butlower-ratedbanksareupgradedmoreoftenthanlower-ratedindustrialissuersare.
JacksonandPerraudin(1999),drawingonatableintheabovestudy,reportthatover1-yearhorizons,banksratedBsufferfewerbonddefaultsthanB-ratedindustrialissuers,althoughthedifferenceisnotstatisticallysignificant.
Comparingsovereignswithotherissuers,S&P(1999)hasreportedgreaterstabilityforsovereignratingsthancorporateratings.
However,S&P'sbriefdoesnottakeaccountofthefargreaterfrequencyofwithdrawnratingsinthecorporatesector,nordoesitaddressissuesofstatisticalsignificanceraisedbytherelativelysmallnumberofobservationsforsub-primesovereigns.
WithregardtodistinctionsbetweenUSandnon-UScompanies,Nickelletal.
(1998)findthathigher-ratedJapanesefirmsaremorelikelytobedowngradedbyMoody'sandthatlower-ratedJapanesefirmswerelesslikelytobeupgraded.
AnotheranalysisofratingsinJapansuggeststhattheJapaneseratingsofMoody'sInvestorsServicemayberelativelytough,sincefewerdefaultshavebeenobservedovertimeinJapanthanwouldhavebeenpredictedbyMoody'sratings,despiteJapan'sstagnanteconomicconditionsinthe1990s(JCIF,1999).
LimitationsoftheanalysisOurempiricalanalysisusestheratingshistoryofratedissuersfromthetwolargestcreditratingagenciesintheworld,Moody's(forissuerswithratedbondsoutstandingduringtheperiod1970-1998),andS&P(fortheperiod1981-1998).
Thefocusonthesetwoagenciesisanunavoidablelimitationoftheanalysis.
Clearlytheproposedframeworkdoesnotenvisiontheexternalcreditratingsthatapplytobeonlythoseofthelargesttwoagencies;however,thesearetheonlytworatingagenciesthatprovidesufficientdataforasectoralcomparisonoftheassociationofratingswithdefaults,recoveries,anddowngrades.
AnotherlimitationoftheanalysisisthatevenMoody'sandS&Pdonotprovidehistoricaldataontheirratingsinallsectors.
TheperformanceofsecuritiesratedinthelucrativeandextensiveareasofmunicipalfinanceandstructuredfinancearenotpartofthepubliclyavailabledatabasesofMoody'sandS&P.
Somequestionshavebeenraisedwithregardtothestabilityofratingsstandardsintheareaofstructuredfinance(CantorandPacker,1994).
Themunicipalfinanceareaisoneinwhichratingsaresuggestedtobemuchtougherandassociatedwithlowerdefaultprobabilitiesthanothersectorsevenbytheratingagenciesthemselves(forexample,Moody's,1999).
However,thedataaresimplynotavailabletoinvestigatetheperformanceofratingsinthoseindustrysectorsrelativetoothersectors.
Totheextentthatthedisclosureofratingshistoriesisasignalofthestabilityanddependabilityofthosehistories,weshouldexpecttheratingsinconsistenciesobservedacrosssectorsandacrosscountriestobealowerboundoftheunobservedinconsistenciesthatmayexistmoregenerally.
Descriptivestatistics:defaultratesbysectorsTable1reportstheone-yeardefaultratesbyinitialratingandsectorofissuer,accordingtotheMoody'sdatabase,whichcoversdefaultsbetween1970and1998.
Thedefaultratesarecalculatedusingestimatesofmid-perioddenominators,constructedbysubtractinghalfofthenumberofratingswithdrawn(overthewholeperiod)fromthenumberofratedissuersatthe156beginningoftheperiod.
Ratingwithdrawalsaregenerallynotadversecreditevents(Carty,1997).
Defaultratesaremeasuredacrosssevensectors-USbanks,otherUSfinancialfirms,USnon-financialfirms,non-USbanks,othernon-USfinancialfirms,non-USnon-financialfirms,andsovereigns.
Intermsofoveralldefaultrates,thedefaultrateforUSbanksof1.
43%isaboutthesameasthatofUSnon-financialfirms.
Botharewellabovethedefaultrateof0.
50%fornon-USnon-financialfirms,whichinturnishigherthanthe0.
08%defaultratefornon-USbanks.
Basedonthesenumbersalone,itappearsthatUSfirmsareriskierthannon-USfirmsare,andnon-USbanksareparticularlysafe.
Sovereignshadincurredjustonebonddefaultsince1983onaforeign-currencyobligationratedbyMoody's.
90However,overallsectoraldefaultratesindicatelittleaboutsectoraldifferencesinthecorrespondenceofratingstodefault,sincetheunderlyingratingscompositionofeachsectoralpoolofborrowersislikelytodiffer.
Forinstance,iftheaverageratingsofUSbanksweremuchhigherthanthoseofUSnon-financialfirms,thesimilardefaultrateswouldimplythatratingswererelativelylenientforbanks.
Or,iftheaverageratingsofnon-USfirmsweremuchhigherthanforUSfirms,theobserveddifferenceinthedefaultratesmayresultpurelyfromthisdifferenceofratingscomposition.
InTable2,wecontrolfortheratingscompositionandreportthedefaultratesforUSandnon-USfirmsateachMoody'sratinglevel.
Wereportbothone-yearandfive-yeardefaultrates.
Inbothcases,estimatesofmid-perioddenominatorsareconstructedbysubtractinghalfthenumberofwithdrawnratings,asbefore.
91Thefive-yeardefaultratesaretheaverageoutcomesforannualcohortsfromJanuary1970toJanuary1994.
Ataone-yearhorizon,USfirmsratedBashowaslightlyhigherpropensitytodefault(1.
3%vs.
0.
8%);thoseratedBandintheCaa/Ca/Crangealsoshowahigherpropensitytodefault(6.
9%vs.
2.
4%forB's,and20%versus15%forCCCs).
Notsurprisingly,measuringdefaultsatafive-yearhorizonresultsinhigherdefaultratesateachratinglevel.
AteachratingcategoryotherthanAaa,thelikelihoodofdefaultoverafive-yearhorizonisnoticeablygreaterforUSfirms.
Thesefiguressuggestthat,duringthesampleperiod,Moody'swasmoreconservativeinitsratingofnon-USthanUSfirms,butwithverylimiteddatafornon-UScompaniesatthelowerratinglevels,itisdifficulttoreachadefinitiveconclusionfromthiscomparison.
InTable3,weagaincontrolfortheratingscompositionandthistimereporttheoneandfive-yeardefaultratesforUSbanksandnon-financialfirmsateachMoody'sratinglevel.
Ataone-yearhorizon,USbanksratedBashowaslightlyhigherpropensitytodefaultthanUSnon-financialfirms(1.
9%vs.
1.
3%),andthedefaultratesarehigherintheBandCaa-Crangesaswell(13.
8%vs.
6.
6%forB's,and56.
4versus18.
8%forCaa-C).
Again,measuringdefaultsatafive-yearhorizonresultsinmoredefaultsateachratinglevel,andmorestriking,butconsistent,sectoraldifferences.
AteachratingcategoryotherthanAaa,thelikelihoodofdefaultoverafive-yearhorizonissignificantlygreaterforUSbanks.
Atleastduringthe90ThedefaultwasalatepaymentonaEurobondbyPakistanin1998,whichwasnotconsideredadefaultunderS&P'sdefinitionbecausethecouponwasdisbursedwithinthecontractualgraceperiod.
91Theimplicitassumptioninthisconstructionisthatratingswithdrawalsaredistributedevenlythroughtheperiod.
Themid-perioddenominatorproducesameasureofthedefaultratethatisclosetotheaverage"hazardrate"ofdefault.
157sampleperiod,Moody'swasmoreconservativeinitsratingofUSnon-financialfirmsthanUSbanks.
92ProbitregressionanalysisOfcourse,thedistinctionsnotedinTables2and3couldbeareflectionoffactorsotherthangenuinesectoraldifferences.
Differencesovertimeinthefrequencyofsectoralratings,combinedwithdifferentdefaultpatternsacrosstime,couldaccountforthedifferences.
Forinstance,therehasbeenamuchgreaterpercentageofMoody'sratings,particularlyatthelowerratinglevels,assignedtonon-USborrowerssincethemid-1990sthanpreviously.
Itmayalsobestatedthat,whiletheconditionofUSfirmsimprovedfromthelate1980s-early1990stothemid-1990s,non-USfirmsweregoingthroughdifferenteconomicconditionsintheirrespectivecountries.
Table4presentssummaryresultsforfourmultivariateprobitmodelsthatweestimated,wherethedependentvariable,estimatedoverthousandsof"issuer-years",istheprobabilitythattheissuerdefaultedthatyear.
(FulldetailsoftheestimatedmodelcoefficientsfromwhichthesestatisticsarederivedappearinAppendixTableA1.
)Theexplanatoryvariablesincludeanindicatorforeachyeartocontrolfortime-varyingeffects,anindicatorvariableforeachratinglevelbelowA3orA-,andindicatorvariablesforfourbroadissuerclasses:USnon-financialfirms,USbanks,otherUSfinancialfirms,andallnon-USfirms.
93(Sovereignswereexcludedfromthisexercisebecause,withonlyafewdozenlow-ratedissuer-yearsskewedtowardthelate1990s,boththeactualandexpectednumberofdefaultswereclearlytoolowtomakereliableinferences.
)Becausetheprobitrepresentationisbasedonanon-linearmultivariatefunction,individualparameterestimatesaredifficulttointerpretoutofcontext.
Tofacilitateinterpretationoftheresults,wecomputeanestimatedprobabilityofdefaultforeachsectoroveralloftheissuer-yearobservationsinthesample,usingtheestimatedparametersforyear,ratingandsector,andforeachcomputationassumingthatalloftheobservationscamefromasinglesector.
ThesectoralstatisticspresentedinTable4representthedifferenceinthisestimatedprobabilityofdefaultfortheindicatedsectorrelativetothatfortheUSnon-financialsector.
Forexample,fortherestrictedmodelinthefirstcolumn,inwhichthethreeclassesofUSfirmsaretreatedasasinglegroup,theaverage"fitted"one-yeardefaultprobabilityisestimatedtobeahairhigher(by0.
09%)fornon-USfirmsthanUSfirms.
(Note,however,thatthedifferenceisnotdistinguishablefromzerowith95%confidence.
)Sincetheamountofnon-USdefaultdataislimited,itisimpossibletoderivedefinitiveconclusionsfromtheanalysis.
However,itmaybethattheapparent"homebias"inratingsinTable2isaresultoftimeeffects.
92HigherdefaultratesforUSbanksrelativetoUSnon-financialfirmsarealsoapparentfromS&Pdata(availableforyears1981-1998).
93Thedearthofsame-yeardefaultsofAandAa-ratedcredits(atotalof2intheMoody'ssampleand6forS&P,inbothcasesoutofthousandsofissuers)posedpracticalobstaclesforincludingdummyvariablesforhighratingsintheprobitspecification.
158Theresultsinthenextcolumn,however,showthatthehigherfrequencyofUSbankdefaultersatagivenMoody'srating(comparedtoUSnon-financialfirms,asdocumentedinTable3),isbothrobusttotimeeffectsandstatisticallysignificant.
Theaveragedefaultrateis0.
77%higherforthefullsample,andastriking2.
29%higherforthe"junkbond"portionofthesample.
TheresultsareslightlydifferentforS&Pratings,withotherUSfinancialfirmsshowingastatisticallysignificantelevationindefaultrate.
CaveatWhiletheprobitresultsareindicativeofstatisticalsignificance,itisimportanttonotethatthedefault-ratediscrepancies,withregardtoMoody'sratingsofUSbanksandUSnon-financialfirms,resultmainlyfromonehistoricalepisode–thethriftcrisisofthelate1980sandearly1990s.
Table5indicatesthat21ofthe33bankdefaultsforthewholeperiodof1970-98wereofUSthriftsin1989-91.
Morethan40%ofthe49ratedthriftsatthebeginningof1989defaulted.
Inretrospect,Moody'sgreatlyoverestimatedtheabilityofthriftstomakeitthroughtheyears1989-91withoutdefault.
TotheextentthattherehavebeendramaticchangesintheUSbankregulatoryregime,andthemethodologyforratingbankshasbeenadjustedtoaccountforthem,bankratingshavenotnecessarilybeenmorelenient,thatis,associatedwithhigherdefaultratesatagivencreditrating,subsequently.
Interestingly,thethriftcrisisdidnothavethesameimpactonestimatedsectoraldefaultratesbyS&Prating.
Only10issuersintheUSbankcategory(whichincludesthriftinstitutions)ratedbyS&Pdefaultedduring1989-91.
ThedisparateexperiencesaroseatleastinpartbecauseS&Pratedsubstantiallyfewerofthespeculative-gradedepositoryinstitutionsthanMoody'satthattime.
RecoveryratesSinceexpectedlossesareafunctionofboththeexpectedprobabilityofdefaultandtheexpectedseverityoflossgivendefault,sectoraldifferencesintheprobabilityofdefaultatgivenratingsdonotnecessarilyimplysectoraldifferencesinexpectedlosses.
Iftheywerecounter-balancedbydifferencesinrecoveryrates,thentheexpectedlossescouldbethesameacrosssectors.
AndincontrasttoS&P,whichsaysthatitsratingsaremeanttoranktherelativelikelihoodofcorporatedefault,Moody'sexplicitlyindicatesthatitincludesconsiderationsofrecoveryinitscorporateratings.
ThemostdetailedpapertodateonrecoveriesbyindustryisbyAltmanandKishore(1996),theresultsofwhicharereproducedinCaouetteetal(1996).
AltmanandKishoreuseS&P'sconventionofmeasuringrecoveriesasthemarketpriceofthebondsasapercentageoffacevalueshortlyafterdefault.
Therecoveriesonthedefaultsofthe66financialinstitutionsaveraged35.
7%,belowthe41%averageonall696defaults.
Thelowerdefaultratedidnotappeartobeafunctionoflowersenioritysincearoundtwo-thirdsofthefinancialinstitutionissueswereseniorsecuredorseniorunsecuredobligations,relativetolessthanone-thirdfortheentiresample.
Thus,recoverieshavebeenlowerforfinancialinstitutions,oppositetowhatweshouldexpectifrecoverieswerecounterbalancingthehigherdefaultratesofbanks.
Moody'sdatabaseallowsforamorerecentexaminationofthedegreetowhichrecoveries,andbyextension,expectedlosses,coulddifferbyindustry.
Moody'smeasuresarecoveryrateasthesecondarymarketpricesofabond30daysafterdefault.
Thedatabasereports159recoveriesonthebondsof595issuersthatwereratedbyMoody'sanddefaultedbetween1970and1998.
Inthecaseofmultipleclassesofbondsoutstandingforanyonefirm,wetaketheweightedaverageofrecoveriesforthatfirm.
InTable6,wereporttherecoveryrateonthedefaultedcorporatebonds,brokenoutinthefirsttworowsbyUSbankversusUSnon-financialfirms,andtheninthenexttworows,byUSversusnon-USfirms.
(TheabsenceofaseparatecategoryofUSnon-bankfinancialfirmsexplainswhythefirsttworowsdonotquiteadduptothethird.
)TheUSbankrecoveriesarestarklylower,withanaveragerecoveryof22%versusaround40%fornon-financialfirms.
Thestatisticresultingfromthet-testonthedifferencesis3.
8,indicatingstatisticalsignificancebeyondthe95%confidencelevel.
Bycontrast,theaveragerecoveriesforthe31ratedbondsofnon-USfirmsthatdefaultedis42%,whichdoesnotdiffersignificantlyfromthesampleof565ratedUSfirmsalone.
Itispossiblethatdifferencesintheaverageseniorityofthebondsissuedbyfirmsindifferentsectors,orofthebondsissuedbyfirmsdomiciledincountriesoutsidetheUScouldbedistortingtheresults.
Ifbankobligationstendedtobelessseniorthannon-financialobligations,thatcouldexplainsomeofthedifferencesinrecoveriesthatwesee.
Conversely,ifUSfirmobligationstendedtobemorejuniorthanthoseofnon-USfirms,thatcouldbemaskingdifferencesinrecoveriesthatarenotapparentintheaggregatesample.
Topartiallycontrolforsectoraldifferencesinthelevelofseniority,werecalculaterecoveryratesbysectorforsubordinatedbondsonly,reportedinTable7.
Thesamplesizedecreasesfrom596to358issuers.
ThedifferenceinaveragerecoverybetweenUSbanksandnon-financialcompaniesremainslarge(19.
9%vs.
35.
5%),andathighlevelsofstatisticalsignificance(t-statisticequals3.
1).
94ThedifferenceinrecoveriesbetweenUSandnon-USfirmshasincreasedsomewhat(from3.
1%to6.
3%),butremainsstatisticallyinsignificant.
Onlyfiveofthethirty-onenon-USdefaultedbondissuerswithavailabledataonrecoverieshadsubordinatedbondsoutstanding.
Inconclusion,itappearsthatthedifferencesindefaultratesbetweenUSbanksandUSnon-financialswerenotcounterbalancedbythedifferencesinrecoveriesonthosedefaults.
Ifanything,therecoveriestendedtobemuchlowerforUSbanksthanthoseforUSnon-financialfirms.
However,keepinginmindthatneitherresultisstatisticallysignificant,theslightlylowerdefaultratesforUSrelativetonon-USfirmsmayhavebeenaccompaniedbysomewhatloweraveragerecoveryrates.
DowngraderatesDowngradesreflectanincreasedlikelihoodofdefaultofanobligationinfutureperiods.
Thus,iftwobondsaresubjecttoequalnear-termdefaultrisk,theinstrumentwithgreaterdowngraderiskwouldlikelyhavemorelonger-termdefaultrisk.
Similarly,itmaybeappropriatefortwobondswithdifferingshort-termdefaultprospectstocarrythesamecreditratingifthebondthatislesslikelytodefaultintheneartermhas,atthesametime,morevulnerabilitytoagradualdeteriorationincreditquality–i.
e.
greaterdowngraderisk.
94Onecaveatwithrespecttothiscomparisonisthatsubordinatedbankbondsmaynotbeofequivalentsenioritytosubordinatedindustrialbonds,giventhatdepositliabilitiesareseniortoallbankdebenturesintheUnitedStates.
160Whiledefaultratesoveralonghorizon,suchas10years,alsomeasurelong-termcreditrisk,analysisusingsuchmeasuresmustimposeacut-offyearforinitialratingsintherelativelydistantpast.
Thus,animportantcomplementaryquestioninassessingtheconsistencyofratingsacrosssectorsmaybewhetherthelikelihoodofadefaultordowngrade,controllingforinitialratinglevel,isconsistentacrosssectors–oratleastwhetheranydifferencesareoffsetting.
Table8showsdowngraderatesforsovereignandotherissuesbythebeginning-of-periodMoody'sandS&P'screditratingsfortheperiod1981-1998.
TheshorterperiodisutilisedsincetheS&Pdataareavailablefromthatyear.
Forthepurposesofouranalysis,anissuerisconsidereddowngradedifitsratingonunsecuredseniordebtmovesfromoneletter-gradecategorytoalowerletter-gradecategory,orifitdefaulted.
(Thus,downgradeswithinlettergrades-e.
g.
fromA1toA3-arenotcounted.
)Again,one-halfofthewithdrawnissuersareremovedfromthedenominator,andratesatbothone-yearandfive-yearhorizonsarereported.
Table8indicatesthatsovereignshaveshownstrikinglylowerdowngraderatesatboththeone-yearandfive-yearhorizonsatthehighestratinglevelsofAaa(AAA),Aa(AA),andAthanotherentitiesratedbyMoody'sandS&P.
Forexample,38%ofallfirmsratedAabyMoody'sweredowngradedoverafive-yearhorizonversus5%ofallsovereigns.
Atleastintermsofthedirectionofthebias,theseareconsistentwiththelowerriskweightsplacedonsovereigncreditsintheproposedstandardisedapproach.
However,inthecategoriesofBaa,wheresovereigncreditsalsohavelowerriskweightsrelativetocorporates,thedowngraderatesarehigherforsovereignsthancorporates.
Downgraderatesforsovereignsinthenon-investmentgraderanges[Ba(BB),B]appearsimilartootherfirmsattheone-yeartimehorizon,butloweratthefive-yearhorizon.
Sinceourcalculationofdowngraderatesatafive-yearhorizonutilisesannualcohortsbasedoninitialratingsonlythrough1994,therewereveryfewsovereignobservationsattheBaandBgradelevels,andthelowerdowngraderatesshouldnotbetakentooseriously.
Table9againreportsdowngraderates,butthistimecomparingUSandnon-USfirms.
Heretheresultsaremixed,andMoody'sandS&Pratingsprovidecontrasts.
AmongthedifferencesinMoody'sratingsbetweenUSandnon-USfirms,thedifferencesinone-yeardowngraderatesarenoticeableonlyatBaa(6.
3%forUSvs.
8.
4%fornon-US)andB(9.
7%forUSvs.
6.
8%fornon-US)andCaa-C(18.
9%forUSvs.
15.
2%fornon-US).
GivenaMoody'slowrating,non-USfirmsarelesslikelytobedowngradedordefaultwithinayear.
However,givenalowS&Prating,non-USfirmsaremorelikelytobedowngraded.
Atafive-yearhorizon,Moody'sandS&Pdowngraderatesaremorecloselyinlinewitheachother.
Moody'sdataindicatethatnon-USfirmsweremorelikelytobedowngradedfromAaathanUSfirms,butthatUSfirmsateverysubsequentratingcategoryweremorelikelytobedowngradedthannon-USfirms.
Thedifferentialindowngraderateswereespeciallylargeinthelowerratinglevels:30.
5%vs.
18.
8%forBa's,35.
3%vs.
28.
9%forB's,and47.
1%vs.
15.
4%forCaa-Cs.
S&PresultsalsoshowhigherdowngraderatesforUSfirms,thoughinasomewhatmorelimitedratingarea-A,BBB,andBB,andB.
However,thedifferenceindowngraderatesatS&P'sBB(26.
6%forUSfirmsvs.
9,4%fornon-USfirms)andB(30.
9%vs.
9.
6%)ratinglevelsareparticularlylarge.
Atthesameratinglevel,USfirmshavefacedgreaterdowngraderiskthannon-USfirms.
161Table10reportsthedowngraderatesforUSbanksversusUSnon-financials.
WithMoody'sdata,forhigher-gradecreditsratedAaaandAa,USbankshavebeenfarmorelikelytobedowngradedthanUSnon-financials,bothatoneandfive-yearterms.
Whileflatteningforthemediumratinglevels(A,Baa),thedifferenceagainisapparentatbothoneandfiveyear-termsfornon-investmentgraderatings-forexample,Ba,BandCaa-C.
TheresultsusingS&Pdata,thoughsomewhatweaker,arebasicallyconsistentwiththoseusingMoody's.
WhileAAA/AAratinglevelsareassociatedwithmoreUSbankdowngradesthanUSnon-financials,non-financialstendtohavehigherdowngraderatesfortheA/BBBmiddleregion.
BankshavesignificantlylowerdowngraderatesintheBratedarea.
Aswiththedefaultrateanalysis,wetestfortheindependentinfluenceofsectoraldistinctionsonthelikelihoodofadowngradebyestimatingmultivariateprobitmodels,andusingtheestimatedcoefficientstocomputeimpliedsectoraldifferences(Table11).
Thedependentvariableistheprobabilityofaletter-gradedemotionordefault.
TheexplanatoryvariablesarethesameasforthedefaultprobitspresentedinTable4,exceptthatasovereignissuerindicatordummy,aswellasdummyvariablesforAaandAratinglevels,cannowbeincluded.
(FulldetailsoftheestimatedmodelcoefficientsfromwhichthereportedstatisticsarederivedappearinAppendixTableA2.
)S&P'ssomewhatlowerdowngraderatesforsovereignsrelativetocorporateissuers(documentedinTable8)turnouttobebothrobusttocalendardummiesandstatisticallysignificant.
ThemixedresultsoftheUSversusnon-UScomparisons(fromTable9)nowresultinsignificantlyhigherdowngraderatesfornon-USfirms,oncetherelativeabundanceofnon-USratingsinmorerecentyears(whichsawalower-than-averageoverallrateofdowngrades)istakenintoaccount.
95Finally,consistentwiththefiguresinTable10,USbankshavebeensignificantlymorepronetodowngradesthanUSnon-financialfirms.
ConclusionInsummary,wefindlimitedevidencethatMoody'sandS&P'sratingshavebeenimperfectlycalibratedacrossissuersectorsinthepast.
Inparticular,foragivencreditrating,defaultratesseemtobehigherforUSfinancialfirmsthanforUScorporates,butthedifferencebetweenUSfirmsandthosedomiciledinothercountriesseemstobeinsignificant.
However,itmustbestressedthattakingintoaccountthelimitedamountofnon-USdataavailableforthecurrentstudy,furtheranalysisisnecessary.
SectoraldifferencesindowngraderatesandrecoveryratesdonotoffsetthehigherdefaultrateforUSbanks,intermsofoverallcreditrisk.
However,particularlyifbanksupervisionhasbecomemoreeffectiveorcreditratingshaveadjusted,onewouldnotnecessarilyexpectUSbankdefaultratestobehighergoingforward.
Thetopicsofafewrecentreports(forexample,S&P,1999;andMoody's,1999a)publishedbyratingagenciessuggestthattheyarepayingincreasingattentiontosectoralcomparisons.
95ItispossiblethatUSissuershavebeenlesspronetodowngradesoverthelastfewyearsduetounexpectedlyfavourablemacroeconomicconditions.
TheeconomicclimateintheUnitedStatessincethemid-1990shasbeenparticularlyoutstanding,whileJapanandsomeoftheemergingmarketeconomies(whichaccountforanincreasingproportionofcreditratings)haveexperiencedamuchbumpierride.
162ReferencesAltman,Edward,andV.
M.
Kishore.
"AlmostEverythingYouAlwaysWantedtoKnowAboutRecoveriesonDefaultedBonds",FinancialAnalystsJournal,Nov/Dec1996,57-63.
BaselCommitteeonBankingSupervision(1999).
"CapitalAdequacyFramework",Basel,June.
Beattie,Vivien,andSusanSearle(1992).
"CreditRatingAgencies:TheRelationshipbetweenRaterAgreementandIssuer/RaterCharacteristics".
JournalofInternationalSecuritiesMarkets(Winter):371-5.
Cantor,Richard,andFrankPacker(1994).
"TheCreditRatingsIndustry",FederalReserveBankofNewYorkQuarterlyReview,1-26.
CantorRichard,andFrankPacker(1995).
"SovereignCreditRatings",CurrentIssuesinEconomicsandFinance,FederalReserveBankofNewYork,June.
Cantor,Richard,andFrankPacker(1996).
"TheDeterminantsandImpactofSovereignCreditRatings"EconomicPolicyReview.
Caouette,John,andEdwardAltman,andPaulNarayanan.
(1999)ManagingCreditRisk.
NewYork:JohnWileyandSons,1999.
Carty,Lea(1997).
"Moody'sRatingMigrationandCreditQualityCorrelation,1920-1996",Moody'sSpecialReport(July).
Jackson,Patricia,andWilliamPerraudin.
"TheNatureofCreditRisk:TheEffectofMaturity,TypeofObligor,andCountryofDomicile",FinancialStabilityReview,November1999.
JapanCenterforInternationalFinance(JCIF).
1999.
ΑCharacteristicsandAppraisalofMajorRatingCompanies(1999)-FocusingonRatingsinJapanandAsia,April.
Moody'sInvestorsService(1999a)"TheEvolvingMeaningofMoody'sBondRatings".
(August).
Moody'sInvestorService(1999b)"HistoricalDefaultRatesofCorporateBondIssuers",(January).
Morgan,Don.
"JudgingtheRisksofBanks:WhatMakesBanksOpaque"FederalReserveBankofNewYorkWorkingPaper,March1998.
Nickell,Pamela,WilliamPerraudinandSimoneVarotto(2000),"StabilityofRatingTransitions",JournalofBankingandFinance,24(1-2):203-227.
StandardandPoor's(1999).
"SovereignversusCorporateRatingsStability",CreditWeek,September9-13.
163Table1One-YearDefaultRatesbyTypeofIssuer(1970-1998)TypeofIssuerDefaultRate(%)NumberofDefaultsEffectiveDenominatorUSBanks1.
43332300OtherUSFinancialFirms0.
78354469USNon-FinancialFirms1.
4251636252Non-USBanks0.
0822400OtherNon-USFinancialFirms0.
1232430Non-USNon-FinancialFirms0.
50204027Sovereigns0.
141698Total1.
1661052576Notes:Defaultratesarecalculatedusingestimatesofmid-perioddenominators,constructedbysubtractinghalfofthenumberofratingswithdrawn(overthewholeperiod)fromthenumberofratedissuersatthebeginningperiod.
Ratingwithdrawalsaregenerallynotadversecreditevents.
Moody'scalculatesone-yeardefaultratesthesamewayaswedo.
Table2DefaultRatesbyInitialMoody'sCreditRating:USvs.
Non-USFirms(1970-1998)One-YearHorizonFive-YearHorizonCreditRatingUSFirmsNon-USFirmsUSFirmsNon-USFirmsAaa0/1732.
0=0.
0%0/1032.
5=0.
0%0.
1%0.
3%Aa0/4499.
5=0.
0%2/2577.
0=0.
1%0.
4%0.
2%A1/12306.
0=0.
0%0/2778.
0=0.
0%0.
6%0.
2%Baa13/10389.
0=0.
1%1/1214.
0=0.
1%1.
8%0.
9%Ba114/8487.
0=1.
3%6/754.
5=0.
8%11.
6%7.
8%B350/5082.
5=6.
9%11/468.
0=2.
4%31.
4%23.
5%Caa,Ca,C106/524.
5=20.
2%5/33.
0=15.
2%50.
2%15.
4%Notes:Defaultratesarecalculatedusingestimatesofmid-perioddenominators,constructedbysubtractinghalfofthenumberofratingswithdrawn(overthewholeperiod)fromthenumberofratedissuersatthebeginningoftheperiod.
Ratingwithdrawalsaregenerallynotadversecreditevents.
Moody'scalculatesone-yeardefaultratesthesamewayaswedo,buttheyuseadifferentmethodologyformulti-yeardefaultrates.
Five-yeardefaultratesareforannualcohortsfromJanuary1970toJanuary1994.
164Table3DefaultRatesbyInitialMoody'sCreditRating:USBanksvs.
USNon-FinancialFirms(1970-1998)One-YearHorizonFive-YearHorizonCreditRatingBanksNon-FinancialBanksNon-Fin.
Aaa0/112.
5=0.
0%0/1288.
5=0.
0%0.
0%0.
1%Aa0/371.
0=0.
0%0/3253.
5=0.
0%2.
5%0.
3%A0/1093.
5=0.
0%1/9595.
0=0.
0%2.
2%0.
4%Baa0/322.
0=0.
0%11/9178.
5=0.
1%5.
5%1.
6%Ba5/258.
0=1.
9%102/7717.
0=1.
3%28.
2%11.
0%B17/123.
0=13.
8%315/4755.
5=6.
6%47.
6%30.
8%Caa,Ca,C11/19.
5=56.
4%87/464.
0=18.
8%70.
6%51.
7%Notes:Defaultratesarecalculatedusingestimatesofmid-perioddenominators,constructedbysubtractinghalfofthenumberofratingswithdrawn(overthewholeperiod)fromthenumberofratedissuersatthebeginningoftheperiod.
Ratingwithdrawalsaregenerallynotadversecreditevents.
Moody'scalculatesone-yeardefaultratesthesamewayaswedo,buttheyuseadifferentmethodologyformulti-yeardefaultrates.
Five-yeardefaultratesareforannualcohortsfromJanuary1970toJanuary1994.
165Table4AverageImpliedSectoralDifferencesinOne-YearDefaultRatesPercent,RelativetoUSNon-FinancialFirms,fromProbitModelVariablesControlledforinEstimationMoody'sRating(Alphanumeric);YearDummiesStandard&Poor'sRating(LetterGradeonly);YearDummiesSamplePeriod1983-19981981-1998USsectorshavesameexpecteddefaultratesRestrictedNoRestrictedNoAverageImpliedDifferenceOverAllRatingLevels:(t-statisticforsectorparameterinparentheses)USBanks0.
77(2.
37)0.
06(0.
23)OtherUSFinancialFirms0.
34(1.
36)0.
63(2.
03)AllNon-USFirms0.
09(0.
36)0.
14(0.
55)0.
14(0.
58)0.
17(0.
71)AverageImpliedDifferenceOverSpeculativeGradeRatingLevelsOnly:USBanks2.
290.
21OtherUSFinancialFirms1.
032.
07AllNon-USFirms0.
280.
410.
470.
57Notes:Differencesinboldarestatisticallysignificantwith95%confidence.
RatingleveldummyvariablesareincludedonlyforBaa1(Moody's)andBBB(S&P)andlower,becauseofscarcityofdefaultsathigherratings.
SeeAppendixTableA1forestimatedcoefficients.
166Table5USThriftDefaultsfromMoody'sJanuary1989CohortDefaultsCreditRatingNumberofIssuers1989199019911989-1991Aaa20000Aa00000A60000Baa90112Ba191438B115319Caa-C22002Total4988521Table6RecoveryRatesonDefaultedCorporateBonds(1970-1998)TypeofIssuerNumberofCasesFaceValueUSDBillionAverageRecovery(percent)DifferenceT-statisticUSBank262.
722.
0USNon-Financial51588.
539.
917.
93.
8USFirm56596.
9Non-USFirm317.
642.
03.
10.
7Total596104.
439.
0Note:Comparisonsinboldarestatisticallysignificantwith95%confidence.
Moody'smeasuresrecoveryrateasthesecondarymarketpriceofabond30daysafterdefault,withoutregardtoearlierpricesorotherfeaturesofthebond.
167Table7RecoveryRatesonDefaultedCorporateBondsSubordinatedBondsOnly(1970-1998)TypeofIssuerNumberofCasesFaceValueUSDBillionAverageRecovery(percent)DifferenceT-statisticUSBank201.
719.
9USNon-Financial31836.
035.
515.
63.
1USFirm35340.
534.
1Non-USFirm51.
440.
46.
30.
6Total35841.
834.
2Note:Comparisonsinboldarestatisticallysignificantwith95%confidence.
Moody'smeasuresrecoveryrateasthesecondarymarketpriceofabond30daysafterdefault,withoutregardtoearlierpricesorotherfeaturesofthebond.
168Table8DowngradeRates(%)byInitialMoody'sandStandardandPoor'sCreditRating:Sovereignvs.
AllFirms(1981-1998)One-YearHorizonFive-YearHorizonCreditRatingSovereignAllFirmsSovereignAllFirms(Moody's)Aaa(S&P)AAA7.
03.
211.
78.
027.
919.
139.
330.
6AaAA1.
21.
610.
27.
45.
21.
538.
030.
7AA2.
51.
26.
65.
94.
912.
120.
921.
1BaaBBB10.
96.
76.
76.
126.
121.
715.
116.
2BaBB15.
010.
010.
09.
512.
527.
830.
026.
1BB3.
610.
59.
48.
80.
00.
035.
230.
6Caa,Ca,CCCC18.
721.
945.
951.
9Note:Downgradesincludedefaults.
Downgraderatesarecalculatedusingestimatesofmid-perioddenominators,denominators,constructedbysubtractinghalfofthenumberofratingswithdrawn(overthewholeperiod)fromthenumberofratedissuersatthebeginningoftheperiod.
Five-yeardowngraderatesareforannualcohortsfromJanuary1981toJanuary1994.
169Table9DowngradeRates(%)byInitialMoody'sandStandardandPoor'sCreditRating:USvs.
Non-USFirms(1981-1998)One-YearHorizonFive-YearHorizonCreditRatingUSFirmsNon-USFirmsUSFirmsNon-USFirms(Moody's)Aaa(S&P)AAA11.
17.
112.
410.
534.
028.
845.
835.
8AaAA10.
67.
49.
67.
638.
630.
136.
733.
1AA6.
55.
96.
85.
721.
421.
317.
719.
8BaaBBB6.
35.
98.
47.
415.
316.
311.
614.
5BaBB9.
99.
710.
97.
630.
526.
618.
89.
4BB9.
78.
86.
814.
935.
330.
928.
99.
6Caa,Ca,CCCC18.
921.
815.
240.
047.
151.
915.
466.
7Note:Downgradesincludedefaults.
Downgraderatesarecalculatedusingestimatesofmid-perioddenominators,denominators,constructedbysubtractinghalfofthenumberofratingswithdrawn(overthewholeperiod)fromthenumberofratedissuersatthebeginningoftheperiod.
Five-yeardowngraderatesareforannualcohortsfromJanuary1981toJanuary1994.
170Table10DowngradeRates(%)byInitialMoody'sandStandardandPoor'sCreditRating:USBankvs.
USNon-FinancialFirms(1981-1998)One-YearHorizonFive-YearHorizonCreditRatingUSBanksUSNon-FinancialsUSBanksUSNon-Financials(Moody's)Aaa(S&P)AAA27.
614.
410.
67.
577.
646.
932.
130.
6AaAA15.
010.
810.
16.
866.
844.
235.
827.
9AA6.
54.
26.
66.
222.
917.
421.
922.
3BaaBBB8.
55.
56.
35.
915.
717.
515.
916.
4BaBB13.
710.
79.
89.
538.
927.
230.
926.
5BB23.
619.
49.
18.
450.
041.
134.
730.
3Caa,Ca,CCCC56.
421.
116.
821.
172.
750.
044.
252.
4Note:Downgradesincludedefaults.
Downgraderatesarecalculatedusingestimatesofmid-perioddenominators,denominators,constructedbysubtractinghalfofthenumberofratingswithdrawn(overthewholeperiod)fromthenumberofratedissuersatthebeginningoftheperiod.
Five-yeardowngraderatesareforannualcohortsfromJanuary1981toJanuary1994.
171Table11AverageImpliedSectoralDifferencesinOne-YearDowngradeRatesPercent,RelativetoUSNon-FinancialFirms,fromProbitEstimatesVariablesControlledforinEstimationMoody'sRating(Alphanumeric);YearDummiesStandard&Poor'sRating(LetterGradeonly);YearDummiesSamplePeriod1983-19981981-1998USsectorshavesameexpecteddefaultratesRestrictedRestrictedNoRestrictedRestrictedNoUS&non-USfirmshavesameexpecteddefaultratesRestrictedNoNoRestrictedNoNoAverageImpliedDifferenceOverAllRatingLevels:(t-statisticforsectorparameterinparentheses)USBanks3.
90(6.
10)0.
93(1.
73)OtherUSFinancialFirms0.
73(1.
60)0.
38(0.
63)AllNon-USFirms2.
66(6.
78)3.
10(7.
64)2.
40(5.
62)2.
52(5.
81)SovereignGovernments-1.
65(-1.
43)-0.
69(-0.
59)-0.
29(-0.
26)-2.
76(-2.
44)-2.
28(-2.
04)-2.
18(-1.
94)Note:Differencesinboldarestatisticallysignificantwith95%confidence.
SeeAppendixTableA2forestimatedcoefficients.
172Appendix:TableA1ProbitModelEstimatesforOne-YearDefaultRatesMoody's(1983-1998)S&P(1981-1998)Intercept-4.
08(0.
21)-4.
12(0.
20)-3.
51(0.
14)-3.
54(0.
14)1981Dummy-0.
64(0.
28)-0.
64(0.
28)1982Dummy0.
20(0.
13)0.
20(0.
13)1983Dummy0.
21(0.
16)0.
21(0.
16)-0.
07(0.
15)-0.
07(0.
15)1984Dummy0.
32(0.
15)0.
32(0.
15)-0.
04(0.
14)-0.
04(0.
14)1985Dummy0.
30(0.
14)0.
29(0.
14)0.
05(0.
13)0.
05(0.
13)1986Dummy0.
57(0.
11)0.
57(0.
11)0.
28(0.
11)0.
29(0.
11)1987Dummy0.
36(0.
11)0.
36(0.
11)-0.
30(0.
13)-0.
29(0.
13)1988Dummy0.
23(0.
11)0.
21(0.
11)-0.
03(0.
11)-0.
03(0.
11)1989Dummy0.
57(0.
10)0.
56(0.
10)0.
05(0.
11)0.
05(0.
10)1990Dummy0.
74(0.
10)0.
72(0.
09)0.
36(0.
10)0.
36(0.
10)1991Dummy0.
71(0.
10)0.
69(0.
10)0.
45(0.
10)0.
45(0.
10)1992Dummy0.
22(0.
12)0.
20(0.
12)0.
10(0.
12)0.
10(0.
11)1993Dummy0.
14(0.
13)0.
13(0.
12)-0.
34(0.
15)-0.
34(0.
15)1994Dummy-0.
19(0.
13)-0.
20(0.
13)-0.
26(0.
13)-0.
26(0.
13)1995Dummy0.
03(0.
10)0.
03(0.
10)-0.
01(0.
11)-0.
00(0.
09)1996Dummy-0.
39(0.
13)-0.
39(0.
13)-0.
34(0.
13)-0.
33(0.
13)1997Dummy-0.
28(0.
12)-0.
28(0.
12)-0.
20(0.
12)-0.
20(0.
11)Baa1Dummy0.
40(0.
35)0.
40(0.
35)Baa2Dummy0.
63(0.
28)0.
64(0.
28)Baa3Dummy1.
06(0.
24)1.
07(0.
24)0.
65(0.
14)0.
66(0.
14)Ba1Dummy1.
28(0.
22)1.
28(0.
21)Ba2Dummy1.
25(0.
22)1.
27(0.
22)Ba3Dummy1.
80(0.
20)1.
83(0.
20)1.
16(0.
13)1.
18(0.
13)B1Dummy2.
02(0.
20)2.
05(0.
20)B2Dummy2.
37(0.
20)2.
40(0.
20)B3Dummy2.
70(0.
20)2.
73(0.
20)1.
86(0.
12)1.
88(0.
12)Caa-CDummy3.
14(0.
21)3.
16(0.
20)2.
70(0.
13)2.
72(0.
13)Non-USFirmDummy0.
04(0.
10)0.
05(0.
09)0.
06(0.
10)0.
07(0.
10)USBankDummy0.
25(0.
11)0.
03(0.
12)Non-BankFinancialUSFirmDummy0.
12(0.
09)0.
23(0.
11)Notes:Standarderrorsinparentheses.
ThecumulativenormaldensityfunctionappliedtotheinterceptgivesthefittedprobabilitythataUSnon-financialfirmratedhigherthanBaa1atthebeginningof1998defaultedbytheendofthatyear.
Forotherfitteddefaultprobabilities,addanyappropriatedummyvariablecoefficient(s)totheinterceptbeforeapplyingthenormalcdf.
ForS&Psamples,thereisonlyonedummyperratinglettergrade(i.
e.
BBB,BB,B,andCCC-C).
WeomitdummiesaboveBaa1becausenear-termdefaultsareveryrareathighratinglevels.
173Appendix:TableA2ProbitModelEstimatesforOne-YearDowngradeRatesMoody's(1983-1998)S&P(1981-1998)Intercept-1.
09(0.
05)-1.
22(0.
05)-1.
25(0.
05)-1.
42(0.
05)-1.
50(0.
04)-1.
51(0.
05)1981Dummy0.
20(0.
05)0.
25(0.
06)0.
26(0.
06)1982Dummy0.
18(0.
06)0.
23(0.
06)0.
23(0.
06)1983Dummy-0.
15(0.
06)-0.
08(0.
06)-0.
07(0.
06)0.
03(0.
06)0.
08(0.
06)0.
08(0.
06)1984Dummy-0.
17(0.
06)-0.
10(0.
06)-0.
09(0.
06)-0.
01(0.
05)0.
03(0.
06)0.
04(0.
06)1985Dummy-0.
02(0.
06)0.
05(0.
05)0.
06(0.
05)0.
16(0.
05)0.
20(0.
05)0.
21(0.
05)1986Dummy0.
05(0.
05)0.
11(0.
05)0.
12(0.
05)0.
26(0.
05)0.
31(0.
05)0.
31(0.
05)1987Dummy-0.
16(0.
05)-0.
10(0.
05)-0.
11(0.
05)-0.
06(0.
05)-0.
02(0.
06)-0.
02(0.
06)1988Dummy-0.
11(0.
05)-0.
05(0.
05)-0.
05(0.
05)0.
07(0.
05)0.
12(0.
05)0.
12(0.
05)1989Dummy-0.
02(0.
05)0.
03(0.
05)0.
03(0.
05)0.
07(0.
05)0.
12(0.
05)0.
12(0.
05)1990Dummy0.
08(0.
05)0.
13(0.
04)0.
13(0.
05)0.
21(0.
05)0.
25(0.
05)0.
25(0.
05)1991Dummy0.
03(0.
05)0.
08(0.
04)0.
07(0.
05)0.
15(0.
05)0.
19(0.
05)0.
19(0.
05)1992Dummy-0.
03(0.
06)0.
01(0.
04)0.
01(0.
05)-0.
01(0.
05)0.
02(0.
05)0.
02(0.
06)1993Dummy-0.
19(0.
05)-0.
16(0.
05)-0.
16(0.
05)-0.
16(0.
05)-0.
13(0.
05)-0.
13(0.
05)1994Dummy-0.
51(0.
06)-0.
49(0.
05)-0.
49(0.
05)-0.
20(0.
05)-0.
18(0.
05)-0.
18(0.
05)1995Dummy-0.
36(0.
05)-0.
34(0.
05)-0.
34(0.
05)-0.
19(0.
05)-0.
17(0.
05)-0.
17(0.
05)1996Dummy-0.
52(0.
05)-0.
51(0.
05)-0.
50(0.
05)-0.
36(0.
05)-0.
34(0.
05)-0.
34(0.
05)1997Dummy-0.
36(0.
05)-0.
35(0.
04)-0.
35(0.
05)-0.
19(0.
05)-0.
18(0.
05)-0.
18(0.
05)Aa1Dummy-0.
80(0.
09)-0.
80(0.
09)-0.
81(0.
09)Aa2Dummy-0.
33(0.
06)-0.
32(0.
06)-0.
32(0.
06)Aa3Dummy0.
23(0.
05)0.
26(0.
05)0.
25(0.
05)-0.
02(0.
04)-0.
01(0.
03)-0.
01(0.
04)A1Dummy-0.
81(0.
06)-0.
77(0.
06)-0.
78(0.
06)A2Dummy-0.
42(0.
05)-0.
36(0.
05)-0.
37(0.
05)A3Dummy0.
12(0.
05)0.
18(0.
05)0.
17(0.
05)-0.
14(0.
04)-0.
12(0.
03)-0.
12(0.
04)Baa1Dummy-0.
73(0.
07)-0.
66(0.
07)-0.
67(0.
07)Baa2Dummy-0.
51(0.
06)-0.
44(0.
06)-0.
44(0.
06)Baa3Dummy0.
11(0.
05)0.
18(0.
05)0.
19(0.
05)-0.
12(0.
04)-0.
08(0.
04)-0.
08(0.
04)Ba1Dummy-0.
48(0.
06)-0.
41(0.
06)-0.
41(0.
06)Ba2Dummy-0.
18(0.
06)-0.
11(0.
06)-0.
10(0.
06)Ba3Dummy0.
15(0.
05)0.
23(0.
05)0.
25(0.
05)0.
12(0.
04)0.
16(0.
04)0.
16(0.
05)B1Dummy-0.
43(0.
06)-0.
35(0.
06)-0.
33(0.
06)B2Dummy-0.
06(0.
06)0.
03(0.
06)0.
05(0.
06)B3Dummy0.
34(0.
06)0.
43(0.
06)0.
45(0.
06)0.
06(0.
04)0.
11(0.
04)0.
11(0.
05)Caa-CDummy0.
40(0.
08)0.
50(0.
08)0.
51(0.
08)0.
61(0.
07)0.
66(0.
07)0.
66(0.
07)Non-USSovereignDummy-0.
13(0.
09)-0.
05(0.
09)-0.
02(0.
09)-0.
24(0.
10)-0.
20(0.
10)-0.
19(0.
10)Non-USFirmDummy0.
18(0.
03)0.
21(0.
03)0.
16(0.
03)0.
17(0.
03)USBankDummy0.
26(0.
04)0.
07(0.
04)Non-BankFinancialUSFirmDummy0.
05(0.
03)0.
03(0.
04)Notes:Standarderrorsinparentheses.
Forfitteddowngradeprobabilities,addanyappropriatedummyvariablecoefficient(s)totheinterceptandapplythenormalcdf.
ForS&Psamples,thereisonlyonedummyperratinglettergrade(i.
e.
AA,A,BBB,BB,B,andCCC-C).
174SectionD.
RatingsDifferencesAcrossAgenciesIntroductionAmajorissueraisedbytheextensiveuseofratingsbyregulatoryauthoritiesisthedegreetowhichratingsmaydifferacrossagencies.
Ifthecorrespondenceofparticularletter-gradestodefaultand/orexpectedlossdifferssignificantlyacrossagencies,thatcouldhaveseriousimplicationsfortheconsistencyandoptimalityofregulatoryrulesthatrelyonratings.
Inparticular,unlesshigherratingswerediscountedintheeventofsplitratings,theratingsofthemorelenientagencieswouldbetheonesthatcountedwhenmultipleratingswereobtained.
Inaddition,ashasbeennotedelsewhere(forexample,CantorandPacker,1994),borrowersmaybeencouragedtoshopforfavourableratings,andagenciestoshadetheirratingsforcompetitivereasons.
Theincentivesofratingagenciestobelenientshouldbeatleastsomewhatmitigatedbythelong-termdamagetotheirreputationsthatgivingconsistentlyeasyratingswouldentail.
Ultimately,thedegreetowhichreputationmayactasaconstraintonratingsdifferencesisanempiricalissue.
MeanRatingsDifferencesOneofthefirstacademicpaperstocollatedataonratingsdifferencesofmanyagencieswasthatofBeattieandSearle(1992a).
Utilisingalargesampleoflong-termcreditratingsreportedbytwelveoftheleadinginternationalratingagenciesandrecordedinthe1990publicationofCreditRatingsInternational,theyfoundmorethan5,000caseswhereapairofratingagencieshadratingsoutstandingforthesameborrower.
Lessthanone-halfoftheratingspairsagreedprecisely,andmorethan20%differedbytwonotchesormore.
(One"ratingnotch"is,forexample,thedifferencebetweenanAandA+rating;tworatingsnotchesthedifferencebetweenAandAA-).
Ofcourse,differencesofopinionamongagenciesaretobeexpected.
Alternativeratingmethodologiesmaycoexistamongagencies,ascandifferentsubjectivejudgementsregardingqualitativeriskfactors.
Thekeyquestionistowhatdegreetheobserveddifferentialsreflectsystematicdifferencesintheratingsscalesofagencies.
BeattieandSearleaddressthisquestionbycomputingthemeanratingsdifferencesacrossjointlyratedcompaniesforeverypossiblepairofratingagencies.
TheagencypairwiththelargestnumberofjointlyratedcompaniesisthatofMoody'sInvestorsServiceandStandard&Poor's,thetwolargestratingagencies.
Theaveragedifferenceintheirratingsforthe1,398jointlyratedcompanieswasonlyfive-one-hundredthsofanotch,suggestingthattheyassignverysimilaraverageratings.
TheroughequivalenceintheratingsonjointlyratedissuesofMoody'sandS&P'shasbeennotedinmanyotherpapersaswell(Perry,1985;Ederington,1986;EderingtonandYawitz,1987;CantorandPacker1994,1995,1997a;JewellandLivingston,1998).
However,theroughequivalenceinratingscalesdoesnotnecessarilyextendtootherratingagencies.
Forexample,whentheratingsofeightotherratingagenciesarecomparedtothosegivenbyMoody'stothesameborrowers,theratingsoffiveothersweresignificantlyhigherthanMoody's.
TheratingsofthethirdandfourthlargestUSagencieseachratedaboutathird175ofanotchhigherthanMoody's,andtwooftheJapaneseratingagenciesratedonaveragebetweenonetotwonotcheshigherthanMoody's.
SectoralPatternsinMeanRatingsDifferencesInexercisessimilartothoseperformedbyBeattieandSearle,CantorandPacker(1994)examinedtheinitialratingsbythefourlargestUSagenciesofspeculative-grade(or"junk")bondsissuedbetween1989and1993.
WhiletheyfoundtheratingsofMoody'sandS&Ptobenearlyidenticalonaverage,theyfoundthethirdandfourthlargestagenciestodisagreewithMoody'swithgreaterregularityandonagreaterscaleinthejunk-bondsamplethanintheaforementionedsampleofBeattieandSearle.
Theratingsofthesmalleragencieswerebetweenoneandone-and-a-halfratingnotcheshigherthanthoseofMoody'sandS&P.
CantorandPacker(1994)alsocomparedtheinternationalbankratingsofnineotherratingagencieswiththoseofMoody's.
Aswiththejunk-bondsample,ratingsagreementwaslessfrequentandratingsdifferentialshigherforbanksthanfortheuniverseofratingsexaminedbyBeattieandSearle.
Notably,thethreeJapaneseagenciesratedbankstwotothreenotcheshigheronaveragethantheUSagencies.
FrequencyofHigherandLowerRatingsOtherstatisticssupporttheviewthatthereissomedifferenceinratingscalesbetweenthelargerandsmalleragencies.
OnestudyofratingsofUSissuersfrom1989to1993foundthatbondratingsfromthesmallerratingagenciesweremuchmorelikelytobehigherthanlowerthanthoseofMoody'sandS&P(CantorandPacker,1994).
Forexample,FitchratedhigherthanMoody's58%ofthetimeandhigherthanS&P50%ofthetime.
FitchratingswerelowerthanMoody'sonlyabout6%ofthetimeandlowerthanS&P7%ofthetime.
AsimilarpatternarosewhencomparingDCR(Duff&Phelps)ratingswithMoody'sandS&P,asillustratedbyTable1.
Bycontrast,measuresofcorrelationbetweentheratingsofeachofthelargesttwoagenciesandtheotherUSagenciesareaboutashighastherankcorrelationbetweenthetwolargeagenciesthemselves.
Thisresultsuggeststhatdifferencesinagencyratingsasmeasuresofabsoluteriskdonotimplydifferencesinratingsasmeasuresofrelativerisk.
176Table1:RatingdifferencesbetweenagenciesDistributionofDCR'sratingsrelativeto:DistributionofFitch'sratingsrelativeto:Moody'sStandardandPoor'sMoody'sStandardandPoor'sPercentratedhigher49.
743.
258.
749.
7Percentratedsame39.
64435.
543.
2Percentratedlower10.
712.
85.
87.
1Averagedifferenceinratingnotches0.
600.
460.
740.
56Note:Thetablecompares363firmsratedjointlybyMoody's,StandardandPoor's,andDCR,and157firmsratedjointlybyMoody's,S&P,andFitchatyear-end1993.
Source:CantorandPacker(1997a)SampleSelectionBiasOnepossiblepitfallofsimplecomparisonsofaverageratinglevels(oroftheobservedfrequencyofhigherorlowerratings)arisesfromdifferencesinratingspoliciesoftheratingagencies.
Moody'sandS&PratealltaxablecorporatebondspubliclyissuedintheUnitedStates,regardlessofwhetheraratinghasbeensolicitedbytheissuer.
BothMoody'sandS&PalsofrequentlyissueunsolicitedratingstoissuersfromoutsidetheUnitedStatesaswell(althoughS&Pmarksthoseratingswitha"pi",whichdenotesthattheratingisbasedentirelyonpublicinformation).
MostoftheotherratingagenciesintheUnitedStateshavealongstandingpolicyofratingbondsonlyontherequestoftheissuer,whichinvolvesafeebeingpaidfortheratings.
Itispossiblethatthesmalleragencies'ratingsareonlypurchased(andthusreported)whenthereisastrongexpectationofimprovementuponMoody'sandS&Pratings,whilewhenthesmalleragenciesmight,infact,ratelower,theirratingsarenotpurchased.
Thisimpliesapotentialbiasinthemeanratingandinthefrequencycomparisons,whichisknownintheeconometricliteratureassampleselectionbias.
CantorandPacker(1997a)controlfortheexistenceofpotentialsampleselectionbiasusinganapproachpioneeredbyHeckman(1979).
Theyfindlimitedevidenceforsignificantsampleselectionbiasandthusmuchstrongerevidencefordifferencesinratingsscales.
Whilesampleselectionbiasmayexplainsomepair-wiseratingsdifferentials,mostisattributabletoratingsscaledifferentials.
ImpactofThirdRatingsInthecaseofsplitratingsbyMoody'sandS&P,bothratingsaffectbondyields,andthebestforecastofyieldsisobtainedwhenyieldsareinferredfromtheaverageofthetworatings(CantorandPacker,1997b).
OtherworksuggeststhatthethirdratingcanoftenserveasatiebreakerwhenMoody'sandS&Pdisagree.
IntheUS,theSecuritiesValuationOffice(SVO)oftheNationalAssociationofInsuranceCommissioners(NAIC),conductsitsownanalysistodetermineaqualitycategoryforcapitalchargepurposeswhenratingagenciesratingsaresplitacrossqualitycategories.
Analysisofasampleof305splitratedbondsasof177year-end1994indicatesthattheratingsofallthefourlargestagenciescorrelatedwiththeSVO'sdeterminations.
Inparticular,whenMoody'sandS&Pdisagree,theSVOismuchmorelikelytoassignthehigherqualitycategorywhenahigherthirdratinghasbeenassignedthanwhennothirdratingisassigned(CantorandPacker,1996).
Reinebach(1998)reportstheresultsofastudybyJewellandLivingston.
Theauthorsanalysedthebehaviourof235bondsratedbyFitch,Moody's,andS&PbetweenJanuary1991andMarch1995,andfoundthatforcorporatebondssimilarlyratedbyMoody'sandS&P,ahigherFitchratingresultedinalowerspread.
TheyielddifferentialwidenedincaseswheretheMoody'sandS&PratingsdifferedandtheFitchratingcoincidedwiththehigherofthetwo.
Regardlessofratingsdifferences,themarketappearstorewardissuerswithaloweryieldwhenathirdratingisassigned,especiallywhentheratingishigher.
ComparingtheRatingsofUSandNon-USAgenciesMosthistoricalcreditratinganddefaultdatapertaintoratingsofUSentitiesassignedbyMoody'sandS&P,agenciesheadquarteredintheUnitedStates.
Nevertheless,inrecentyears,bothagencieshavedramaticallyexpandedtheiroverseasoperations.
ThedegreetowhichtheoverseasratingsofestablishedUSagenciesdifferfromlocalratingagenciesisofinterest.
Thereisampleroomforvaryinginterpretationsofdistinctivebusinesspractices,accountingandlegalsystems,whenassessingtherisksofissuersindifferentcountries.
OutsideoftheUnitedStates,Japaneseratingagenciesareamongtheoldestandmostactive.
Dataavailabilityhasthusattractedtheattentionofresearchers.
NumerousauthorshavedocumentedthatJapaneseratingagenciesconsistentlygivehigherratingstothesamebondissuesthandoMoody'sandS&P(e.
g.
HiraiandTomita,1996;Watanabe,1996;PackerandReynolds,1997).
AtablefromoneofthesearticlesispresentedhereasTable2.
ItcomparestheUSandJapanesecreditratingsof60samuraibondissuersduring1995and1996,and236JapaneseissuersinthedomesticbondmarketonNovember1,1996.
Samuraibondsareyen-denominatedbondsissuedinJapanbynon-Japaneseobligors.
Bondsarebrokenintotwoadditionalcategories,investment-gradeandnon-investmentgrade.
Table2:RatingDifferentialsbetweenJapaneseandUSAgenciesSamuraiIssuersDomesticIssuersGradeNumberofJointlyRatedIssuesAverageNotchDifferentialNumberofJointlyRatedIssuesAverageNotchDifferentialInvestment46+2.
4197+2.
6Non-investment14+2.
539+4.
8Sources:PackerandReynolds(1997).
BasedonratingsfromMoody's,StandardandPoor's,NipponInvestorsService,theJapanBondRatingInstitute,andtheJapanCreditRatingAgency.
TheevidencefromTable2suggeststhatJapaneseratingsareconsistentlyhigherthantheUSratings,approximately2.
5notcheshigherforallcategoriesexceptspeculative-gradedomesticissues,whichare4.
8notcheshigher.
SincenearlyallborrowersintheJapanesemarkethave178atleastoneJapaneseagencyrating,sampleselectionbiasisanextremelyunlikelyexplanationfortheseconsistentlylargeratingdifferentials.
Becausetheintersectionacrossagenciesofratedissuersisrelativelysmall,thetableabovebyitselfcannotresolvethequestionofwhetherJapaneseratingscalesaremorelenientthanUSagencyscales,orwhetherUSagenciesrateJapanesecorporationswithatougherscalethantheyrateUScorporations.
However,furtherinsightcanbeobtainedfromthecorrespondencebetweenratingsandsubsequentdefaults.
SomepreliminaryevidenceonthisfrontisavailablefromJCIF(1999).
ThisanalysissuggeststhattheJapaneseratingsofMoody'sInvestorsServicemayberelativelytough,sincefewerdefaultshavebeenobservedovertimeinJapanthanwouldhavebeenpredictedbyMoody'slowratings(givenMoody'sUSdefaultexperience),despiteJapan'sstagnanteconomicconditionsinthe1990s.
However,theJCIFsampleistoosmalltoestablishstatisticalsignificance.
ConclusionInsummary,theliteraturefindsclearevidenceofdifferencesinratingsscalesoncewemovebeyondthetwolargestagencies.
Someofthesedifferencesmightbeexplainedbysampleselectionbias(derivingfromthefactthatMoody'sandS&PratealmostallissuersintheUnitedStateswhileotheragenciesrateonlyuponrequest).
However,theonestudythatempiricallyinvestigatestheinfluenceofsampleselectionbiasfoundthatmostoftheratingdifferencescouldnotbeexplainedbysampleselectionbias.
Yet,differencesinratingscalesnotwithstanding,themarketappearstovaluetheadditionalinformationofferedbyadditionalratings,withreducedspreadsonissueswithmultipleratings.
Inaddition,therankcorrelationsbetweenratingsofeachofthelargesttwoagenciesandotherUSagenciesareaboutashighastherankcorrelationsbetweenthetwolargeagenciesthemselves.
Thisresultsuggeststhattheagenciesarelikelytobeinagreementovertherelativeriskofborrowers.
BecausetheyareamongtheoldestandmostactiveofratingagenciesoutsideoftheUnitedStates,Japaneseratingagencieshaveattractedtheattentionofresearchers.
LargeratingdifferencesareapparentbetweenJapaneseratingagenciesandnon-JapaneseratingagenciesthattendtorateJapanesecreditslower.
Itisunlikelythatsampleselectionbiasisamajordeterminantoftheselargeratingdifferentials.
However,itisalsonotclearwhetherJapaneseratingagencieshaverateddomesticcreditsusingahigherratingscalethanthatusedbyMoody'sandS&P,orwhetherUSratingagencieshaveratedJapanesecreditsusingatougherratingscalecomparedtocreditsintheUnitedStates.
(Thetwoexplanationsarenotmutuallyexclusive.
)TherelativelylowdefaultratesonJapanesecreditsgivenspeculative-graderatingsbyMoody'slendssomecredencetothelatterview,thoughthehistoryofratedissuersisstilltooshorttoreachanydefinitiveconclusion.
179ReferencesBeattie,VivienandSusanSearle(1992a),"BondRatingsandInter-RaterAgreement",JournalofInternationalSecuritiesMarkets167-172(Summer)Beattie,Vivien,andSusanSearle(1992b),"CreditRatingAgencies:TheRelationshipbetweenRaterAgreementandIssuer/RaterCharacteristics".
Cantor,RichardandFrankPacker(1994),"TheCreditRatingIndustry",FRBNYEconomicPolicyReview1-26(Summer-Fall).
Cantor,Richard,andFrankPacker(1996),"DiscretionintheUseofRatings:TheCaseoftheNAIC",JournalofInsuranceRegulation.
Cantor,Richard,andFrankPacker(1997a),"DifferencesofOpinionintheCreditRatingIndustry",JournalofBankingandFinance.
Cantor,Richard,FrankPacker,andKevinCole(1997b),"SplitRatingsandthePricingofCreditRisk",JournalofFixedIncome,December.
CantwellandCompany(1998),InternationalSurveyofCreditRatings,SurveyResults.
Ederington,Louis(1986),"WhySplitRatingsOccur".
FinancialManagement(Spring):37-47.
Ederington,Louis,andJessYawitz(1987),'TheBondRatingProcess".
InEdwardAltman,ed.
,HandbookofFinancialMarkets.
NewYork:JohnWileyandSons.
Euromoney(1996),"RatingAgencies,HowManyCanPlaytheRatingsGame"(May).
Heckman,James(1979),"SampleSelectionBiasasaSpecificationError",Econometrica(January)153-161.
Hirai,Naoki,andHiroshiTomita.
1996.
"CreditRatingsinJapan:AProgressReport".
NRIQuarterly5,no.
4(winter).
House,Richard(1995),"RatingtheRaters",InstitutionalInvestor,pp.
53-66.
Hayakawa,S.
JapaneseFinancialMarkets102-109.
JapanCenterforInternationalFinance(1999),"CharacteristicsandAppraisalofMajorRatingCompanies–FocusingonRatingsinJapanandAsia",Tokyo.
Jewell,Jeff,andMilesLivingston(1998),"SplitRatings,BondYields,andUnderwriterSpreads",JournalofFinancialResearch,Summer.
Packer,Frank,andElizabethReynolds(1997),"TheSamuraiBondMarket",CurrentIssuesinEconomicandFinance,June.
Perry,LarryG.
(1985),"TheEffectofBondRatingAgenciesonBondRatingModels".
JournalofFinancialResearch(Winter):307-315.
180Reinebach,Adam(1998),"StudyShowsThirdRatingShrinksSpreads",InvestmentDealer'sDigest,April13.
Watanabe,Shigeru(1996),"CorporateFinance",inJapaneseFinancialMarkets,editedbyS.
Hayakawa.