SScceennaarriioo--bbaasseedd AAMMAA Presentation for RMG-Conference 29./30. May 2003 May 2003 (final version 1.0) The following banks have contributed to the content and draftingof the presentation. Such contributions do not imply that the institutions will implement the approach set out in thepaper, but rather that they believe a Scenario-based AMA is conceptually sound and, if implemented with integrity, shouldberecognisedas qualifying for AMA status. The views expressed do not necessarily reflect the overall view of each individual institution. Banca Intesa Barclays Bank CreditSuisseFirst Boston Dresdner Bank FortisBank Halifax Bank of Scotland Lloyds TSB The Royal Bank of Scotland Group UFJ Holdings, Inc Euroclear RCO OR Group Risk Control CC Risikocontrolling Operational Risk © Scenario-based AMA working group OOuuttlliinnee ooff pprreesseennttaattiioonn Objectives of presentation Introduction Overview of concept How to determine appropriate scenarios for OR? Howtoensure that scenarios are consistent, relevant andcaptureall material OR? How to evaluate scenarios in anorganisation? How to use scenarios formodellingpurposes? Why a scenario based AMA improves OR management (example application) Key benefits of a scenario based AMA How other building blocks link into a scenario based AMA Overlap / similarities with LDA / Scorecard Illustrations © Scenario-based AMA working group Page 2 OObbjjeeccttiivveess ooff tthhee pprreesseennttaattiioonn To demonstrate that a scenario based approach is conceptually sound To define the key features of implemented scenario-based AMA approaches To prove that a scenario based approach is a valuable basis for managing risks To indicate that banks are fairly aligned in terms of thinking about scenarios To describe how scenarios can form an integral part of economic and regulatory capital calculations To illustrate how scenarios can be constructed in the most useful way © Scenario-based AMA working group Page 3 EExxeeccuuttiivvee SSuummmmaarryy A scenario based AMA is focused on a forward looking assessment of the key operational risks in an organisation taking into account both the internal control environment and external threats is an approach that employs the technique of individual scenarioevaluation in a similar fashion to market and credit risk is based on all available information (expert experience, internal / external losses, KRIs, quality of control environment) leads via a model to a sound economic capital number that helps to incentivise prudent and pro-active OR management bridges the gap between LDA and Scorecard approach (or at least has a considerable overlap with each) has been or is being successfully implemented in a number of international banks © Scenario-based AMA working group Page 4 IInnttrroodduuccttiioonn Scenarios are defined aspotentialevents(i.e.events that could happeninthe future). Risk is inextricably linkedtothe evaluationof „what-if“certain scenarios occur.The evaluation process involves providing answerstotwofundamentalquestions: How likely are certain scenariostohappen? How severe could their impact be? Scenarios are alreadyanimportant techniqueinthe evaluationofmarketandcredit risk as illustratedinthe following questions: What is the impact if the yield curve shifts by20bp?How likely is this? What is the impact if this customer defaults?How likely is this? Scenarios arealso an essentialcomponentinthe assessmentof operationalrisks anddeterminationofcapital. © Scenario-based AMA working group Page 5 OOvveerrvviieeww ooff ccoonncceepptt Scenario Scenario Data Determination of Model & Model Generation Assessment Quality Parameter Values Parameters Output How to determine How to evaluate How to use scenarios for modelling purposes? appropriate scenarios in an scenarios for OR? organisation? high CA 275,26 UA Potential Loss Severity medium highmedium low EINSEXTEPITEMAIFOGPR 127,69 EC low RR 574,68 Mean, low medium low medium high high Potential Loss Frequency Scenario Classes 892,72 Std. deviation, O etc. r 449,62 g a n i 688,35 s a t i 694,28 o n a l P a r t How to ensure s that scenarios are consistent, relevant and capture all material OR? © Scenario-based AMA working group Page 6 HHooww ttoo ddeetteerrmmiinnee aapppprroopprriiaattee sscceennaarriiooss?? The determination of a appropriate scenarios that are Scenario Classes representative for OR is key to the scenario based AMA O r g a Scenario classes –one or more–are classes that are n i s a derived from loss event types or risk factorsandthat t i o n contain scenarios. (Example:Scenario class= IT-breakdown.) a l P a Organisational parts are the partsofthe organisation for r t s whichORis separately evaluated. Generate scenario(s) perscenario classandorganisational part basedon:- guided discussions with thebusiness inworkshops(eg. Dresdner;Fortis, HBOS) amatrixof „critical resourcesandstatesofrisk“ (eg.Banca Intesa) weaknessesperrisk factor(eg. UFJ Holdings) aspecificationofcritical resourcesandfailure periods by the organisational parts (eg. Dresdner) addressing particular management concerns(all) All scenarios are fully documented to allow independent review and assessment © Scenario-based AMA working group Page 7 HHooww ttoo eennssuurree tthhaatt sscceennaarriiooss aarree ccoonnssiisstteenntt,, rreelleevvaanntt aanndd ccaappttuurree aallll mmaatteerriiaall ooppeerraattiioonnaall rriisskkss?? Consistent? Every organisational part must consider as aminimum eachofthe common setof scenario classes,thereby achieving consistency of the overall framework. Techniques such as workshops assist to achieve consistency of scenarios across organisational units Review by Internal Audit and Risk Functions provides further consistency between organisational parts. Relevant? Every organisational part assesses the relevanceof allscenariostoitsbusinessthereby ensuring relevancetothem(e.g.if aorg-unit is not dependenton IT,it does not make sense toevaluate the risk dueto ITbreak-down). Captureall material operationalrisks? The techniquestodetermine the scenario classes(eg useofexpert judgementand historical loss data)maximise coverageofknownandforeseen risks. This coverage is further enhanced when applied to the organisational parts and discussed with them to ensure that their specific risks are covered. © Scenario-based AMA working group Page 8 HHooww ttoo eevvaalluuaattee sscceennaarriiooss iinn aann oorrggaanniissaattiioonn?? Foreach scenario banks assesspotentialloss frequenciesand potentialloss severities. The organisational processofevaluation is basedon: questionnaires guided workshops central(management)expertise In order to come upwithplausibleanswers banks base their evaluationon informationrelevant tothe scenario, suchas anassessmentof operationalquality/qualityofcontrol environment past losses,key risk indicators,insurance industryandmanagerial experience Tovalidate the scenario evaluations banks apply techniquessuchas 4-eye principle Internal auditsofassessment processandresulting quality Comparison between actual lossesandexpert expectations Consistency checks(e.g.psychometric analysis, Groupfunctions challenge profilesof organisational parts,comparisontointernal audit findingsandvalidation.) © Scenario-based AMA working group Page 9 HHooww ttoo uussee eevvaalluuaatteedd sscceennaarriiooss ffoorr mmooddeelllliinngg ppuurrppoosseess?? A goodrisk model must be consistent, robust, and stable over time, so that changes in economic capital result from changes in the underlying risk profile and not from changes in the model. A model on risk requires plausible assumptions where distributions or analytical solutions are used The model requires the estimation of its parameters. The data required for the parameter estimation must go through arigorous data quality assurance. Raw Data AFTER data cleansing Parameter Estimates for Distribution Scenario Typical Frequency Typical Severity Mean Standard Mean Standard Class Frequency Upper Bound Severity Upper Bound Frequency Deviation Severity Deviation IT 4 2 200 233 * Scenario 11650300 * Scenario 26103001000 * Scenario 357250700 Control 6 3 250 125 * Scenario 136100500 * Scenario 2710400100 * Scenario 3820250900 ... ... ... ... ... © Scenario-based AMA working group Page 10
Description: