Orange county case.docx
《Orange county case.docx》由会员分享,可在线阅读,更多相关《Orange county case.docx(13页珍藏版)》请在冰豆网上搜索。
Orangecountycase
PhilippeJorion'sOrangeCountyCase:
UsingValueatRisktoControlFinancialRisk
ThisWebcasecanbeusedbyacademicinstitutionsfreeofcharge;
otherusersshouldcontactProfessorJorion.
Thecaseisalsosubjecttocontinuousimprovements.
©2004-PhilippeJorion
(Downloadedfromhttp:
//www.gsm.uci.edu/~jorion/oc/case.html)
Summary
Thepurposeofthiscaseistoexplainhowamunicipalitycanlose$1.6billioninfinancialmarkets.Thecasealsointroducestheconceptof"ValueatRisk"(VAR),whichisasimplemethodtoexpresstheriskofaportfolio.Afterthestringofrecentderivativesdisasters,financialinstitutions,end-users,regulators,andcentralbankersarenowturningtoVARasamethodtofosterstabilityinfinancialmarkets.ThecaseillustrateshowVARcouldhavebeenappliedtotheOrangeCountyportfoliotowarninvestorsoftheriskstheywereincurring.
(1)Introduction
InDecember1994,OrangeCountystunnedthemarketsbyannouncingthatitsinvestmentpoolhadsufferedalossof$1.6billion.Thiswasthelargestlosseverrecordedbyalocalgovernmentinvestmentpool,andledtothebankruptcyofthecountyshortlythereafter.
ThislosswastheresultofunsupervisedinvestmentactivityofBobCitron,theCountyTreasurer,whowasentrustedwitha$7.5billionportfoliobelongingtocountyschools,cities,specialdistrictsandthecountyitself.Intimesoffiscalrestraints,Citronwasviewedasawizardwhocouldpainlesslydelivergreaterreturnstoinvestors.Indeed,Citrondeliveredreturnsabout2%higherthanthecomparableStatepool.PlotCitron'strackrecord(Figure1).
Citronwasabletoincreasereturnsonthepoolbyinvestinginderivativessecuritiesandleveragingtheportfoliotothehilt.ThepoolwasinsuchdemandduetoitstrackrecordthatCitronhadtoturndowninvestmentsbyagenciesoutsideOrangeCounty.Somelocalschooldistrictsandcitiesevenissuedshort-termtaxablenotestoreinvestinthepool(therebyincreasingtheirleverageevenfurther).Thiswasinspiteofrepeatedpublicwarnings,notablybyJohnMoorlach,whoranforTreasurerin1994,thatthepoolwastoorisky.Unfortunately,hewaswidelyignoredandBobCitronwasre-elected.
Theinvestmentstrategyworkedexcellentlyuntil1994,whentheFedstartedaseriesofinterestratehikesthatcausedseverelossestothepool.Initially,thiswasannouncedasa``paper''loss.Shortlythereafter,thecountydeclaredbankruptcyanddecidedtoliquidatetheportfolio,therebyrealizingthepaperloss.Howcouldthisdisasterhavebeenavoided?
(2)ThePortfolio
Infact,BobCitronwasimplementingabigbetthatinterestrateswouldfallorstaylow.The$7.5billionofinvestorequitywasleveragedintoa$20.5billionportfolio.Throughreverserepurchaseagreements,Citronpledgedhissecuritiesascollateralandreinvestedthecashinnewsecurities,mostly5-yearnotesissuedbygovernment-sponsoredagencies.OnesuchagencyistheFederalNationalMortgageAssociation,affectionatelyknownas``FannieMae''.
Theportfolioleveragemagnifiedtheeffectofmovementsininterestrates.Thisinterestratesensitivityisalsoknownasduration.2.1Defineduration
Thedurationwasfurtheramplifiedbytheuseofstructurednotes.Thesearesecuritieswhosecoupon,insteadofbeingfixed,evolvesaccordingtosomepre-specifiedformula.Thesenotes,alsocalledderivatives,wereinitiallyblamedforthelossbutwereinfactconsistentwiththeoverallstrategy.
Citron'smainpurposewastoincreasecurrentincomebyexploitingthefactthatmedium-termmaturitieshadhigheryieldsthanshort-terminvestments.OnDecember1993,forinstance,short-termyieldswerelessthan3%,while5-yearyieldswerearound5.2%.Withsuchapositivelyslopedtermstructureofinterestrates,thetendencymaybetoincreasethedurationoftheinvestmenttopickupanextrayield.Thisboost,ofcourse,comesattheexpenseofgreaterrisk.PlotthetermstructureonDecember1993(Figure2).Displaytermstructureofinterestratesasoflastweek:
Bloomberg.
Thestrategyworkedfineaslongasinterestrateswentdown.InFebruary1994,however,theFederalReserveBankstartedaseriesofsixconsecutiveinterestrateincreases,whichledtoabloodbathinthebondmarket.Thelargedurationledtoa$1.6billionloss.PlotthepathofinterestratestoDecember1994(Figure3).Graphinterestratesoverrecentperiod:
MinneapolisFed.
(3)ValueatRisk
WhatisVAR?
VARisamethodofassessingriskthatusesstandardstatisticaltechniquesroutinelyusedinothertechnicalfields.Formally,
VARisthemaximumlossoveratargethorizonsuchthatthereisalow,prespecifiedprobabilitythattheactuallosswillbelarger.
Basedonfirmscientificfoundations,VARprovidesuserswithasummarymeasureofmarketrisk.Forinstance,abankmightsaythatthedailyVARofitstradingportfoliois$35millionatthe99%confidencelevel.Inotherwords,thereisonlyonechanceinahundred,undernormalmarketconditions,foralossgreaterthan$35milliontooccur.
Thissinglenumbersummarizesthebank'sexposuretomarketriskaswellastheprobabilityofanadversemove.Asimportantly,itmeasuresriskusingthesameunitsasthebank'sbottomline---dollars.Shareholdersandmanagerscanthendecidewhethertheyfeelcomfortablewiththislevelofrisk.Iftheanswerisno,theprocessthatledtothecomputationofVARcanbeusedtodecidewheretotrimrisk.
3.1IntroductiontoVAR
3.2MethodstomeasureVAR
3.3DurationandVAR
NodoubtthisiswhyregulatorsandindustrygroupsarenowadvocatingtheuseofVARsystems.Bankregulators,suchastheBasleCommitteeonBankingSupervision,theU.S.FederalReserve,andregulatorsintheEuropeanUnionsuchasBritain'sFinancialSupervisoryAuthorityhaveconvergedonVARasanacceptableriskmeasure.TheSecuritiesandExchangeCommissionhasissuedanewruletoenhancethedisclosureofmarketrisk.TherulerequirespubliclytradedU.S.corporationtodiscloseinformationaboutderivativesactivityusingaVARmeasureasoneofthreepossiblemethods.
SeethetextoftheEuropeanCapitalAdequacyDirective(98/31/EC)whichallowstheuseofVAR-basedinternalmodels.
OtherSiteswithVARInformation
Perhapsthemostnotableofprivate-sectorinitiativestowardbetterriskmanagementisthatofJ.P.Morgan,whichunveileditsRiskMetricssysteminOctober1994.Forecastsofriskandcorrelationsformorethan400assetsareposteddailyontheRiskMetricssite(nowwithasix-monthlagforfreedata).TheRiskMetricsdatabaseallowsuserstocomputeaportfolioVARusingtheDelta-Normalmethodbasedona95%confidenceleveloveradailyormonthlyhorizon.
AnotherinterestingsiteisthatofDeutscheBank,whohasacquiredBankersTrustanditsexpertiseinriskmanagement,includingtheRAROC2020system.Thebankprovidesanintegratedapproachtoriskmeasurement,whichisdescribedinitsRiskManagementServicessite.
ThereisagrowingarmyofvendorswhoprovidesoftwarerangingfromExceladd-onstomillion-dollarfirm-wideriskmanagementsystems.Forinstance,visitthesitesofAlgorithmics,BARRARiskManagement,KamakuraCorporation,MisysRiskVision,SungardTradingandRiskSystems.
Amongconsultants,CapitalMarketRiskAdvisorsarewellknown.TheNewYorkconsultingfirmwashiredNovember3,1994,todissecttheOrangeCountyportfolio.Withinaweek,CMRAwarnedthecountythatthepoolhadalreadylost$1.5billion.Thefirmnowspecializesinthevaluationofcomplexportfolios,andin"financialforensics"--analyzingsourcesoffinanciallosses.
Thereisevenanassociationofriskmanagementprofessionals,theGlobalAssociationofRiskProfessionals,whichprovidesaforumfortheexchangeofinformationandeducationintheareaoffinancialriskmanagement.GARPadministersthe"FinancialRiskManager"certificationuponsuccessfulcompletionofanexamination.Forlinkstoriskmanagementsites,visitthefollowingaddress:
BarrySchachter.
Finally,thedebateonVARisheatingup!
Readthe"con"side,advancedbyNassimTalebfollowedbythe"pro"side,advancedbyPhilippeJorion,inarebuttalinDerivativesStrategy,amagazinedevotedtoderivatives.
(4)Questions
LetusplaceourselvesinthepositionofthecountySupervisors,whohadtodecideinDecemberof1994whethertoliquidatetheportfolioormaintainthestrategy(obviously,basedonpastinformationonly).Atthattime,interestrateswerestillonanupwardpath.AFederalOpenMarketCommitteemeetingwasloomingonDecember20,anditwasfearedthattheFedwouldraiseratesfurther.Toassessthepossibilityoffuturegainsandlosses,VARprovidesasimplemeasureofriskintermsthatanybodycanunderstand--dollars.
(1)Durationapproximation.
Theeffectivedurationofthepoolwasreportedbythestateauditoras7.4yearsinDecember1994.Thishighdurationistheresultoftwofactors:
theaveragedurationofindividualsecuritiesof2.74years(mostofthesecuritieshadamaturitybelow5years),andtheleverageoftheportfolio,whichwas2.7atthetime.In1994,interestrateswentupbyabout3%.Computethelosspredictedbythedurationapproximationandcompareyourresultwiththeactuallossof$1.64billion.
(2)ComputationofportfolioVAR.
(2)Theyieldsdatafilecontains5-yearyieldsfrom1953to1994.Usingthisinformationandthedurationapproximation,computetheportfolioVARasofDecember1994.Riskshouldbemeasuredoveramonthatthe95%level.ReportthedistributionandcomputetheVAR:
-usinganormaldistributionforyieldchanges(Delta-Normalmethod),and
-usingtheactualdist