已阅读5页,还剩53页未读, 继续免费阅读
版权说明:本文档由用户提供并上传,收益归属内容提供方,若内容存在侵权,请进行举报或认领
文档简介
3 Measuringmarketrisk VaRapproach 3 1Introduction3 2UnderstandingVaR3 3Riskmetrics3 4Historicsimulation3 5Mote Carlosimulation3 6BISstandardizedmodel 3 1Introduction J P MorganG30BISInvolvingprobabilitycomponentalongwithlossseveritycomponentinriskmeasurement dealingwiththelimitationofsensitiveapproachesVaRRevolutionExtensiontoCreditandOperationalriskandthereforeintegratedriskmanagement AdvantagesofVaRapproach oversensitivityapproach completemeasureofrisk Measuringriskusingthesameunit dollarAggregateviewofaportfolioriskaccountingforleverageandcorrelationeffectsintegratednature notonlyderivativesbutalsoallotherfinancialinstruments andcanbebroadenedfrommarketrisktoothertypesoffinancialriskone numberindicator 3 2UnderstandingVaR QuestionsleadingustoVaRmeasureDefiningValueatRiskKeyelementsofcalculatingVaRWhatdoesPDFcurvetell Approachestoprobabilitydistribution typesofVaR WorkingoutVaRthroughariskfactor QuestionsleadingustoVaRmeasure Asaportfoliomanager youmaybeaskedbyyourbossfollowingquestions Q1 Givenamarketchangeorshock howmuchcouldyourportfoliosuffer Q2 Ifitturnsouttobeabaddaytomorrow whatistheworstlossofyourinvestment Thefirstisasensitivityquestion andyoucangiveaclearanswerafteryoudoasensitivitymeasureasweshowedinpreviouslectures Thesecondisnotaclearquestion Beforeyoutrytoanswerityouhavetoaskback Whatdoyoumeanpreciselyby abadday or Howbadthedayyousupposeittobe SensitivityQfollowedbyprobabilityQ Q1leadingtoVaRmeasure ExampleofsensitivityQ Givena25bpsyieldrise howmuchcouldyourinvestmentportfolio P 1m MD 2ys suffer A dP D dR 1 R P MD dR P 2 0 0025 1 000 000 5 000 Youmaybefurtherasked Howlikelycoulditbethecase Ormoreprecisely giventhenormalmarketcondition howlikelywouldyourportfoliosuffernotmorethanthatamountofmoneyoveratargetholdingperiod AVaRquestion Morefrequently thequestionisputinanotherway Whatistheworstlossyourportfoliocouldsufferoveratargetholdingperiodwithprobabilityofagivenlevel say1 undernormalmarketcondition AstandardVaRquestion Define badday andcompleteQ2 Q2leadingtoVaRmeasure Oneeasywaytodefinea badday istodefinethelosssizeorseverity Butthiswillmakethequestionmakenosense Itisreasonable andmeaningful todefinea badday insuchawaythatthedayissobad orthelossissoseverethatsuchaday orloss occursundernormalmarketconditiononlyonceoutofevery20 tobedefined tradingdays putanotherway thechance probability oftheoccurrence loss isonly5 tobedefined Ifitturnsouttobeabaddaytomorrow whatwillbetheworstlossofyourinvestment giventhatdayhappensonceoutofevery20tradingdays AstandardVaRquestion DefiningValueatRisk Definition TheworstlossoveragiventargetholdingperiodatagivenconfidencelevelundernormalmarketconditionTwopre specifiedvariables HoldingPeriod 1day 1weekormore ConfidenceLevel 95 99 orevenhigher VaRisananswerto Undernormalmarketcondition whatistheworstlosscouldmyinvestmentportfoliosufferwithaprobabilityof5 withinonetradingday Iam95 99 surethattheinvestmentportfoliowillsuffernotmorethan 5 000 20 000 lossover1dayundernormalmarketcondition KeyelementsofcalculatingVaR HoldingperiodConfidencelevelProbabilitydistributionofreturnProbabilitydensityfunction PDF inthecaseofcontinuousrandomvariableWorkingouttheprobabilitydistributionofreturnisthemostdifficultpartoftheVaRwork Holdingperiod DEARandMore than one dayVaR Holdingperiodisthetargettimehorizonduringwhichyouholdyourinvestmentposition alsotargetmeasuringperiod One dayVaR orDailyVaRisusuallycalculated especiallyinthecaseofRiskmetricsmodel ItisusuallytermedDailyEarningatRisk DEAR More than one dayVaRcanbederivedfromDEARfromfollowingformula Undertheassumptionthatmarketvalidityisconstantovertime N DayVaR DEAR NBISrequires10 dayVaR Confidencelevel ProbabilityofthelossDefiningthe badday frequencyofthebaddayLevelofconfidenceoveryourlossforecast riskestimate Confidenceinterval statisticallytermed 95 Riskmetricsvs 99 BISrequirementThehighertheCL thebiggertheVaRnumber WhatcouldhappenifCLissettoeither100 or0 Answersarerightstatementsbutnonsense WhatdoesPDFcurvetell aEb E EE ProbabilityDistributionofarandomvariableXX randomreturn orvalue L G ofinvestmentE expectedreturn theprobabilityweightedaveragevalueofallpossibleoutcomesoftherandomvariable SD variance measureofdispersion expectedsurprisesVarianceistheprobabilityweightedsumofthesquareddeviationsofalltheoutcomesfromtheirexpectedvalue orsimply theexpectedsquareddeviationoftherateofreturnfromitsexpectation PDF1 F a P X a PDF2 N a P X a X X Area P X a Area P a X b Area P X b 1 F b 68 26 WhatdoesPDFcurvetell aEb a E Eb E ProbabilityDistributionofarandomvariableXPDFtellstheprobabilitythattherandomvariable X doesnotexceedthespecifiedcriticalvaluea F a P X a whichcanbeillustratedbythesizeofthearealefttothecriticalvalue Thesizeoftheareaindicatestheprobabilitythatthevariablefallwithinthetwocriticalvalues ThewholeareaunderneaththePDFcurveis1 PDF1 F a P X a PDF2 N a P X a X X Area P X a Area P a X b Area P X b 1 F b 68 26 Normaldistribution Anormaldistributioncanbecompletelydescribedbytwofeatures expectedvalue E andstandarddeviation NDissymmetricandbell shaped smalltailoneitherside whichononehandmeansprobabilityofextremevalueoneithersideissmall ontheotherhandmeanschangesoftherandomvariablearecenteredarounditsmeanvalue 68 26 ofchangesfallwithin1SDoneitherside 95 44 within2SDs 99 74 within3SDs90 within1 65SDs 95 within1 96SDs 98 within2 33SDsNormaldistributionassumptioninfinanceandfattailproblem a E Eb E PDF N a P X a X 68 26 ConfidencelevelandcriticalvalueofaPDF SupposethePDFofyourinvestmentportfolioisgivenasfigureinnextslide Supposeyourcurrentpositionis 100 000 theoreticallyreflectingtheexpectedvalueofrandomlyfluctuatingvalueofyourportfolio SupposetheSD 5 000 1 65 8 250Thecriticalvalue1 65 awayfromEintheleftsideis 8 250inlossor 91 750invalue whichindicates With95 confidencelevel youbelieveyourlosswillnotexceed 8 250 G LofPortfolio ValueofPortfolio 8 2500 8 250 91 750 100 000 108 250 E 1 65 EE 1 65 90 5 5 PDFofyourinvestmentportfolio Approachestoprobabilitydistribution typesofVaR HistoricalsimulationVariance covarianceapproachMonte Carlosimulation Howaboutwhentheprobabilitydistributionofyourportfolioisnotdirectlyknown Dueto 1 Integratedcontributionfromdifferentriskfactors likeinterestrisk FXrisk etc 2 Aggregationofmultipletypesofassets likepositionsinbonds equity commodities 3 Moreimportantly shorthistoryanddynamicadjustmentofyourportfolioThen weturnto 1 decomposingtheriskintoriskfactors 2 workingoutthePDFofriskfactors 3 andlinkingtheriskfactorchangetoyourportfoliovaluechangeusingsensitivity WorkingoutVaRthroughariskfactor Marketrisk Estimatedpotentiallossunderadversecircumstances valuepricePotential VolatilityVaR ofthe sensitivityof adversemoveofriskpositionthepositioninyieldfactor VaR position PricevolatilityofthepositionPricesensitivity unit or 1sensitivity i e 1position ssensitivitytoariskfactor likeinterestrate stockprice changePricevolatility unit or 1volatilitye g 1position svolatility Marketrisk Estimatedpotentiallossunderadversecircumstances valuepricePotential VolatilityVaR ofthe sensitivityof adversemoveofriskpositionthepositioninyieldfactor VaR position Pricevolatilityofanasset valuepriceProbability VolatilityVaR ofthe sensitivityof Distributionofriskpositionthepositionofriskfactorfactor givenaunitchangeinriskfactorVaR 100 000 0 005 1P 1bp 16 5bps 95 CL 9 250 95 CL VaR P S DR valueofyourposition Unitsensitivityofyourpositiongivenaunitchangeofriskfactor BadChanges Volatility PDF ofriskfactor 1position svolatility Definethe badchange ofriskfactor Equivalenttodefininga badday Itiseasytodefineoridentifythe baddirection ofriskfactorchange e g riseininterestrate yield isadversechangedirectiontoyourlongpositioninbond Howtodefineamorespecific badchange ismoretroublesome youneedtospecify howbadthat changewillbe Oneeasywaytodefinea badchange istodefinethesizeorseverityofthechange Butthisdoesnotgeneratetoomuchsenseforriskmanagementsincetheprobabilityisnotspecified Everyday undernormalmarketcondition youexperienceadversechangeswhilethemarketfluctuates someadversechangesfrequentlyoccurbutaretolerabletoyouduetotheirrelativelysmallsize othersareseriousorevendisastroustoyou butdonothappenveryoften Asariskmanager yourfocusisonthelatter Itisreasonable andmeaningful todefineabadchangeinsuchawaythatabadchangeisanadversechangewhichhasatmost 5 chancetooccurundernormalmarketcondition putanotherway itoccursonceoutofevery 20 timesofexperimentorobservation Thisdefinitioncombinestogethertheseverityandprobabilityofthe badchange 3 3Riskmeticsmodel IntroductionDennisWeatherstone sOrderBasicmethodologyandprocessCalculatingVaRforanaiveinvestmentportfolioofanFIFixed incomesecuritiesFXEquitiesPortfolioaggregationCriticismsagainstRiskmetricsBISregulationonVaR basedinternalmodelsinlargebanks IntroductiontoRiskmetricsmodel InternalmodelofJ P Morgan 1994 NormaldistributionisassumedformarketchangeOne dayholdingperiod95 confidencelevelBenchmarkofmarketriskmanagement DennisWeatherstone sOrder Atcloseofbusinesseachdaytellmewhatthemarketrisksareacrossallbusinessesandlocations Inanutshell thechairmanofJ P Morganwantsasingledollarnumberat4 15pmNewYorktimethattellshimJ P Morgan smarketriskexposurethenextday especiallyifthatdayturnsouttobea bad day TherequiresingledollarnumberisDailyEarningsatRisk DEAR ordailyVaR Anontrivialjob ForeignFixedExchangeEmergencyincomeSTRIT CommoditiesDerivativesEquitiesMarketsProprietaryTotalNumberofactivelocations1412511871114Numberofindependentrisk takingunits3021816141119120ThousandsofTransactionsPerday 5 5520BillionsofdollarsindailyTradingvolume 10 301150 Basicmethodologyandprocess CalculatingDEARfiguresforeachofthebusinesslines riskfactors StandaloneriskFixed incomesecuritiesFXEquity PortfolioAggregation PortfolioRiskDifferenttradingpositionsaggregatedDifferentriskfactorsaggregatedCorrelationeffectconsidered CalculatingVaRforanaiveinvestmentportfolioofanFI SupposeanFIhasfollowinginvestmentpositions 1 a 1millionmarketvaluepositionin7 yearzero couponbonds 2 Swf1 6millioninspotSwissfrancs FXrateisWsf1 60 atthedailyclose 3 1milliontradingpositioninstocksthatreflectaU S stockmarketindex WhatistheDEAR VaR 95 confidencelevel Fixed incomesecurities Suppose 1 TheFIhasa 1millionmarketvaluepositionin7 yearzero couponbonds 2 Today syieldonthesebondsis7 243 ThenS MD D 1 R 7 1 7 243 6 527 3 Thedailychangeofyieldisnormallydistributedanditsvolatility is10bpsDEAR P S DR 95 CL 1million 6 527 1 65 10bpc 95 CL 1million 1 077 95 CL 10 770 95 CL Pricevolatility MD Potentialadversechangeinyield 6 527 0 00165 1 077 DEAR Marketvalueofposition Pricevolatility 1 000 000 01077 10 770 FromDEARtomore than one dayVaR Tocalculatethepotentiallossformorethanoneday N dayVAR DEAR NExample Forafive dayperiod VAR 10 770 5 24 082 FX InthecaseofForeignExchange DEARiscomputedinthesamefashionweemployedforinterestraterisk DEAR P S DR 95 CLDEAR Dollarvalueofposition Pricevolatility Suppose 1 theFIhadaSwf1 6millioninspotSwissfrancs FXrateisWsf1 60 atthedailyclose ThismeansP 1million 2 ChangesoftheFXratearenormallydistributedandthehistoricalvolatility ofdailychangesinthespotFXrateis56 5bps DEAR P S DR 95 CL 1million 1 1 65 56 5bps 95 CL 1million 93 2bps 95 CL 9 320 95 CL Equities Accordingtomodernportfoliotheory therearetwotypesofrisktoanequitypositioninanindividualstock Totalrisk Systematicrisk UnsystematicriskU riskcanbelargelydiversifiedawayinaverywell diversifiedportfolio S riskreflectsthecomovementofthatstockwiththemarketportfolio forwhichthestockmarketindexcanbeaproxy ThesensitivityofastocktothemarketportfolioisgivenbyCAPMmodel E ri rf i E rM rf i iM M2 Forawell diversifiedstockportfolio DEAR P S DR 95 CL wherethemarketreturnvolatilityistakenas1 65sM P DIndex 95 CLDEAR Dollarvalueofposition Pricevolatility Iftheportfolioreplicatesthestockmarketindex 1 Inlesswelldiversifiedportfolio theeffectofU riskshouldbeconsidered IfCAPMmodeldoesnotofferagoodexplanationofassetpricingcomparedto say multi indexAPTmodel adegreeoferrorshouldbebuiltintoDEARcalculation Suppose 1 TheFIholdsa 1milliontradingpositioninstocksthatreflectaU S stockmarketindex Then 1 2 Thedailyreturn changeofvalue onthestockmarketindexisnormallydistributedanditsvolatilityis2 DEAR P DR 95 CL 1million 1 1 65 2 95 CL 1million 3 3 95 CL 33 000 95 CL PortfolioAggregation individualDEARsoftheFI BondDEAR 10 770 95 CLFXDEAR 9 320 95 CLSUM 53 090EquityDEAR 33 000 95 CLSimplysummingupindividualDEARsdoesnotcomplywithmodernportfoliomanagementtheoryforcorrelationanddiversificationeffectisnottakenintoaccount InordertoaggregatetheDEARsfromindividualexposureswerequirethecorrelationmatrix Three assetcase DEARportfolio DEARa2 DEARb2 DEARc2 2rab DEARa DEARb 2rac DEARa DEARc 2rbc DEARb DEARc 1 2 correlationmatrix r Seven YearZeroSwf 1U SStockIndexSeven yearzero 2 4Swf 1 1U Sstockindex AggregatingindividualDEARintoportfolioDEAR DEARportfolio DEARz 2 DEARswf 2 DEARu s 21 2 2Xrz SwfXDEARzXDEARswf 2Xrz U S XDEARzXDEARU S 2XrU S SwfXDEARU SXDearswf DEARportfolio 10 77 2 9 32 2 33 2 2 2 10 77 9 32 2 4 10 77 33 2 1 9 32 33 39 969 53 090Portfolioeffect 53 090 39 969 13 121 PortfolioDEARSpreadsheet InterestRateRiskNotionalAmounts U S millionsequivalents FXRiskTotal112345710interestSpotFXPortfolioTotalmonthyearYearsYearsyearsyearsyearsyearsDEARFXDEARAustraliaAUDBelgiumBEFCanadaCADDenmarkDKKFrance19 301148FFR48Germany 1930 1127DEM27ItalyLIRJapanYENNetherlandsNLGSpainESBSwedenSEKSwitzerlandGBPUnitedKingdomGBPUnitedStates101076USD76Total1010151151Portfolioeffect 62 62 RISKDATAPRINTCLOSETotalDEAR 000S 8989 CriticismsandshortcomingsofRiskmetrics Assumptionofasymmetricnormaldistributionforallassetreturns Forsomeassets suchasoptionsandshort termsecurities bonds thisishighlyquestionable Limitationofnormalmarketconditionandsupplementarystresstestingorscenarioanalysis BISregulationonVaR basedinternalmodelsinlargebanks IncalculatingDEAR adversechangeinratesdefinedas99thpercentile ratherthan95thunderRiskMetrics Minimumholdingperiodis10days meansthatRiskMetrics dailyDEARmultipliedby 10 Capitalchargewillbethehigherof Previousday sVAR orDEAR 10 AverageDailyVARoverprevious60daystimesamultiplicationfactor 3 Subjecttoback testing 3 4HistoricorBackSimulation AdvantagesBasicideaProcessofHistoricSimulationWeaknesses Advantages SimplicityDoesnotrequirenormaldistributionofreturns whichisacriticalassumptionforRiskMetrics Doesnotneedcorrelationsorstandarddeviationsofindividualassetreturns Basicidea Revalueportfoliobasedonactualprices returns ontheassetsthatexistedyesterday thedaybefore etc usuallyprevious500days Thencalculate5 worst case 25thlowestvalueof500days outcomes Only5 oftheoutcomeswerelower ProcessofHistoricSimulation Converttoday sFXpositionsintodollarequivalentsattoday sFXrates MeasuresensitivityofeachpositionCalculateitsdelta MeasureriskActualpercentagechangesinFXratesforeachofpast500days Rankdaysbyriskfromworsttobest Examples Example1 TextbookPage244 246Example2 Page257 QuestionsandProblemsNO 16 Weaknesses Basicassumption therecentpastdistributionofexchangeratesisanaccuratereflectionofthelikelydistributionofFXratechangesinthefuture thatexchangeratechangeshavea stationary distribution Disadvantage 500observationsisnotverymanyfromstatisticalstandpoint Increasingnumberofobservationsbygoingbackfurtherintimeisnotdesirable Couldweightrecentobservationsmoreheavilyandgofurtherback 3 5MonteCarloSimulation Toovercomeproblemoflimitednumberofobservations synthesizeadditionalobservations Perhaps10 000realandsyntheticobservations Employhistoriccovariancematrixandrandomnumbergeneratortosynthesizeobservations Objectiveistoreplicatethedistributionofobservedoutcomeswithsyntheticdata 3 6RegulatoryModels BIS includingFederalReserve approach MarketriskmaybecalculatedusingstandardizedBISmodel Specificriskcharge Generalmarketriskcharge Offsets Subjecttoregulatorypermission largebanksmaybeallowedtousetheirinternalmodelsasthebasisfordeterminingcapitalrequirements BISStandardizedModel Specificriskcharge Riskweights absolutedollarvaluesoflongandshortpositionsGeneralmarketriskcharge reflectmodifieddurations expectedinterestrateshocksforeachmaturityVerticaloffsets AdjustforbasisriskHorizontaloffsetswithin betweentimezones Terms ValueatRisk Va
温馨提示
- 1. 本站所有资源如无特殊说明,都需要本地电脑安装OFFICE2007和PDF阅读器。图纸软件为CAD,CAXA,PROE,UG,SolidWorks等.压缩文件请下载最新的WinRAR软件解压。
- 2. 本站的文档不包含任何第三方提供的附件图纸等,如果需要附件,请联系上传者。文件的所有权益归上传用户所有。
- 3. 本站RAR压缩包中若带图纸,网页内容里面会有图纸预览,若没有图纸预览就没有图纸。
- 4. 未经权益所有人同意不得将文件中的内容挪作商业或盈利用途。
- 5. 人人文库网仅提供信息存储空间,仅对用户上传内容的表现方式做保护处理,对用户上传分享的文档内容本身不做任何修改或编辑,并不能对任何下载内容负责。
- 6. 下载文件中如有侵权或不适当内容,请与我们联系,我们立即纠正。
- 7. 本站不保证下载资源的准确性、安全性和完整性, 同时也不承担用户因使用这些下载资源对自己和他人造成任何形式的伤害或损失。
最新文档
- 橱窗展示空间设计案例解析
- 介绍礼仪动画版
- 日用电商产品介绍
- 电工年终工作总结个人总结(5篇范例)
- 杭州市富阳区2025年劳资专管员能力考核试题及答案
- 2025安全员-C证上岗证理论试题及答案
- 2025石油化工产品购销合同范文
- 2025广西河池市金城江区金碗贸易投资有限公司招聘工作人员笔试历年典型考点题库附带答案详解试卷2套
- 2025广东河源市连平县城乡投资有限公司连平县旅游实业投资有限公司招聘拟聘人员笔试历年典型考点题库附带答案详解试卷2套
- 2025年福建省晋江水务集团有限公司秋季招聘15人笔试历年难易错考点试卷带答案解析试卷2套
- 国企合规培训课件
- 天津市2021-2023年中考满分作文48篇
- 早教入股协议书范本合同
- 公司租海外仓库合同范本
- 2025至2030中国原生铼金属市场发展模式及未来前景预测报告
- 医院院企合作模式探索与实践
- 小区施工损坏赔偿合同范本
- 医院合作实施方案
- 口腔医学技术职业生涯
- 老年患者深静脉血栓护理
- 静脉采血操作并发症的预防与处理
评论
0/150
提交评论