Stress Testing & Scenario Analysis Contributing to greater financial resilience Securities and Investment Institute Risk Forum London, 11 December 2008 Risk & Capital Markets Advisory Welcome and Introductions Wolfgang Schwerdt Director Bancstreet Capital Partners Ltd formerly Senior Economist at the European Central Bank Pekka Luoma Chairman Bancstreet Capital Partners Ltd Copyright © 2008 Bancstreet Capital Partners Ltd. All Rights Reserved. www.bancstreet.com Evolution of Risk Management In the last three decades Risk Management has evolved rapidly from dark art to mechanised industry. Copyright © 2008 Bancstreet Capital Partners Ltd. All Rights Reserved. www.bancstreet.com Evolution of Risk Management Move from static and subjective approaches to dynamic statistical models. VAR, ETL and related approaches are now in near universal usage and enshrined in regulations. Quantitative risk models have become technically feasible for almost all aspects of risk including market-/ credit- and even certain aspects of liquidity and operational risk. Risk information has become more transparent, available and accessible. Risk Management has transformed from a “Dark Art” to a Mechanized Industry. Copyright © 2008 Bancstreet Capital Partners Ltd. All Rights Reserved. www.bancstreet.com Why do we need Stress Testing ? Risk Management means managing exceptions. Much of risk management practice is concerned with common and expected exceptions. Stress testing deals with uncommon, unexpected exceptions that can have large or devastating consequences. The higher the stakes the greater the importance of effective stress testing. Copyright © 2008 Bancstreet Capital Partners Ltd. All Rights Reserved. www.bancstreet.com Why Stress Testing ? A bit more detail... Historical statistical models like VAR/ETL are simplifications and cannot capture aspects of risk not directly observable in past experience. Expected Risk measurable “Un-Expected” Risk Using VAR/ETL remaining Residual Risk visualised through Stress Testing VaR / ETL measures cannot capture the effect of truly extreme, sudden & dramatic changes. These frequently lead to the breakdown of statistical relationships, like correlations, on which risk measures are based. Stress testing tries to evaluate these aspects, not covered by measurements of expected risk and measurable, “UnExpected” Risk. Copyright © 2008 Bancstreet Capital Partners Ltd. All Rights Reserved. www.bancstreet.com Evolution of Stress Testing From Sensitivities to Historical & Hypothetical Scenarios Copyright © 2008 Bancstreet Capital Partners Ltd. All Rights Reserved. www.bancstreet.com Stress Testing today… Stress testing today is performed through measuring VAR or ETL for a simulated future using selected historical and/or hypothetical scenarios. This simulation based approach is helped by the advances in computer hard and software, significant progress in Monte-Carlo and quasi random simulation approaches. This type of risk analysis is very different from the measurement of historical volatility or of a stock beta. It places enormous emphasis on skilful and informed judgement. However, many organisations have difficulties injecting the necessary level of skilful and informed judgement because: The convenience of historical data over appropriate hypothetical scenarios. Difficulties of getting senior executives directly involved. Difficulties of overcoming the entrenched fragmentation at a systems and organisational level. Copyright © 2008 Bancstreet Capital Partners Ltd. All Rights Reserved. www.bancstreet.com Stress Testing at major financial institutions How safe are decisions made on the basis of common stress testing practices ? Copyright © 2008 Bancstreet Capital Partners Ltd. All Rights Reserved. www.bancstreet.com Stress Testing at major financial institutions Consider the following points from the 2005 BIS survey: “Property price [hypothetical] scenarios [were] generally confined to markets experiencing [a boom or which had big historical fluctuations].” “At a very basic level [...] stress testing is often undertaken in very different areas of a firm, making internal consistency across integrated scenarios difficult.” “Treatment of market liquidity was identified as a key issue in the 2000 survey. Only limited progress has been made since then.” “An unanticipated exit of a major market player is treated as one of the risk sources [for] market turbulence. Not many institutions are running this type of scenario.” Quoted from a report by the Bank for International Settlements entitled “Stress Testing at Major Institutions: Survey Results and Practice”, published in January 2005. (Pages 11-14) Copyright © 2008 Bancstreet Capital Partners Ltd. All Rights Reserved. www.bancstreet.com Is current Stress Testing practice in good health? ??? Was poor Stress Testing & Scenario Analysis a cause for the credit crunch? Could more effective Stress Testing & Scenario Modelling have prevented the credit crunch? Is current practice in Stress Testing & Scenario Analysis providing the robust input into business practice and top level executive decisions it should provide ? Copyright © 2008 Bancstreet Capital Partners Ltd. All Rights Reserved. www.bancstreet.com What happens in a true crisis? Market participants have to survive in a world where everything is upside down. Copyright © 2008 Bancstreet Capital Partners Ltd. All Rights Reserved. www.bancstreet.com What happens Happens in in aa true Truecrisis? Crisis ? What Events cause a sudden move of prices, volatilities and correlations to a spectrum well outside the normal range of observations. Correlations change, break down or even get reversed. Disappearing liquidity may make correlations irrelevant. Stable patterns may take a long time to re-establish themselves. Copyright © 2008 Bancstreet Capital Partners Ltd. All Rights Reserved. www.bancstreet.com What solutions are there today? ... And are they the on-coming train at the end of the tunnel; or a new dawn, after these dark days? Copyright © 2008 Bancstreet Capital Partners Ltd. All Rights Reserved. www.bancstreet.com Integrating Stress Testing at business level What tools do we need as a prerequisite to better stress testing and scenario modelling ? Copyright © 2008 Bancstreet Capital Partners Ltd. All Rights Reserved. www.bancstreet.com The Stress Testing Space Credit (counterparty) risk Funding liquidity Scenario analysis Traded portfolio risk Sensitivity analysis Risk Factors Methodology Operational risk Application Copyright © 2008 Bancstreet Capital Partners Ltd. All Rights Reserved. www.bancstreet.com Space Dimension 1: Risk Factors External factors the changes of which affect the firm’s balance sheet. Assets side: Traded portfolio risk: Interest rate, Equity, FX, Commodity and Credit Derivatives Credit (counterparty) risk Liabilities side: Funding liquidity stress scenarios: changes in client behaviour, own rating, funding costs and collateral requirements. Historical scenarios: LCTM & Asian crises. Operational Risk: not often tested due to data problems. Copyright © 2008 Bancstreet Capital Partners Ltd. All Rights Reserved. www.bancstreet.com Case 3:Counterparty Risk - Simulating the joint effects of market movements and credit events in Trading Book Alpha Testing as a Basle II short-cut Credit Risk Market Risk Liquidity Risk Alpha testing is a short-cut developed in the Basel II "cook book" to measure the interaction of market risk and credit events. A global bank has calculated the ‘alpha’ coefficient for a trading book with 5,000 counterparties 100,000 long-term derivative contracts Monthly time steps in simulation Copyright © 2008 Bancstreet Capital Partners Ltd. All Rights Reserved. www.bancstreet.com Space Dimension 1: Risk Factors External factors the changes of which affect the firm’s balance sheet. Assets side: Traded portfolio risk: Interest rate, Equity, FX, Commodity and Credit Derivatives Credit (counterparty) risk Liabilities side: Funding liquidity stress scenarios: changes in client behaviour, own rating, funding costs and collateral requirements. Historical scenarios: LTCM & Asian crises. Operational Risk: not often tested due to data problems. Copyright © 2008 Bancstreet Capital Partners Ltd. All Rights Reserved. www.bancstreet.com The Asian Contagion – A Stress Test Example Asian contagion in financial markets Trading Market Risk Liquidity Risk Trading Credit Risk Asian Currencies Declined Market Liquidity Dried Up Credit Spreads Widened Equities Fell Enterprise Liquidity Dried Up Credit Quality Declined Interest Rates Unstable Financial System Under Stress Defaults Increased Copyright © 2008 Bancstreet Capital Partners Ltd. All Rights Reserved. www.bancstreet.com Credit Risk – Stress Testing . . .To produce perturbations of combined Distributions of simulated future credit losses Credit risk factors May be shocked . . . Exposure Defaults PROBABILITY Shocks 0 Recoveries t0 tN Correlations to other factors Copyright © 2008 Bancstreet Capital Partners Ltd. All Rights Reserved. www.bancstreet.com Space Dimension 1: Risk Factors External factors the changes of which affect the firm’s balance sheet. Assets side: Traded portfolio risk: Interest rate, Equity, FX, Commodity and Credit Derivatives Credit (counterparty) risk Liabilities side: Funding liquidity stress scenarios: changes in client behaviour, own rating, funding costs and collateral requirements. Historical scenarios: LTCM & Asian crises. Operational Risk: not often tested due to data problems. Copyright © 2008 Bancstreet Capital Partners Ltd. All Rights Reserved. www.bancstreet.com Operational Risk: not often tested due to data problems - Frequency of events (should include “near-misses”) Copyright © 2008 Bancstreet Capital Partners Ltd. All Rights Reserved. www.bancstreet.com Operational Risk: not often tested due to data problems - Severity, measured in money terms Please compare the shape of the rel. frequency. distribution with the one on next slide Copyright © 2008 Bancstreet Capital Partners Ltd. All Rights Reserved. www.bancstreet.com Different types of operational risks Single loss < $1m Quality assurance errors Larger ‘blunders’ Traditional controlled self assessment tools Material risk concentrations Not recognised by managers as the problem has been corrected Structural operational risk breakdowns outside managers’ experience Measure using actual Use extreme value theory and Monte Carlo losses simulation Evaluate exposure to processes, estimate probability of main causes Within the control of business managers Absorbed by normal operating costs Possible candidates for insurance Value Sufficient ‘buffer’ capital at group level Material Risk Concentrations Relevance of scenarios will require senior executive level input and involvement. Proritisation requires an assessment of sources of risk at a strategic level, some not part of the experience of line management. Drill beneath risks that are observable in in-house and syndicated, industry-wide incident databases. Copyright © 2008 Bancstreet Capital Partners Ltd. All Rights Reserved. www.bancstreet.com Operational Risk: not often tested due to data problems - Exposure (any confidence level can freely be chosen) Kurtosis and skewness statistics would have told you about the fat tail even without looking at the graph Copyright © 2008 Bancstreet Capital Partners Ltd. All Rights Reserved. www.bancstreet.com Why does our headline still hold true… … that Operational Risk often is not tested due to data problems? Adapted from Janet Rogers, Federal Reserve Bank of NY Qualitative Approach - Judgemental Develop Operational Risk Function Self Assessment & Action Plans Develop Policies & Control Framework Define KRI’s & Escalation Triggers Quantitative Approach - Optimising Develop Feedback & Improve Performance Allocate Risk Capital Track & Measure Losses Copyright © 2008 Bancstreet Capital Partners Ltd. All Rights Reserved. Measurable Performance Develop OpRisk Measurement Model www.bancstreet.com Space Dimension 2: Application Risk factor sensitivity Evaluate effect of extreme market movements Counterparty risk assessment Verify economic capital allocation Business risk evaluation Communication tool Copyright © 2008 Bancstreet Capital Partners Ltd. All Rights Reserved. www.bancstreet.com Space Dimension 3: Methodology 3 (4) approaches, depending on whether one counts historical simulation (3.) and full Monte Carlo (4.) as one or two separate Sensitivity analysis (1.) Source of shock not identified, instantaneous risk parameter move. Scenario analysis (2.) Single scenario identification; then “deterministic” simulation used Parameter calibration Historical: maximum loss events “limited by history”, once-in-a-thousand years events occur every decade. Hypothetical: use statistical methods to work out Worst Case risk factor values that are still plausible (e.g. by principal component analysis) even when risk factors go “out-of-range”. Copyright © 2008 Bancstreet Capital Partners Ltd. All Rights Reserved. www.bancstreet.com Covering the entire business Scenarios that are generated with different assumptions and approaches cannot safely be aggregated to help make decisions for a whole business or portfolio. The set of scenarios should be carefully adapted to the business as a whole. The approaches and assumptions must be formulated in such a way that they can be made to apply to the business as a whole. Tools to automatically detect which risk factors are operative, and the “quality check” data on those risk factor inputs A suitable infrastructure and process is needed to make the combining of external data (risk factors) and internal data (exposures) cost effective and reliable and timely Copyright © 2008 Bancstreet Capital Partners Ltd. All Rights Reserved. www.bancstreet.com Integration at business level takes... A solution built around a path Monte Carlo simulation capability that is linearly scalable to handle any size of institution and any horizon for analysis, still using a time step that corresponds to market or macroeconomic adjustments paths. Has an extensive pricing library, with an in-built capability to handle structured products, too. Can handle thousands of risk factors so that both risk analyses and performance attribution analysis can be made granular enough to incorporate all important “decision levers”. And: has a standardised way of recording the financial characteristics of all instruments for quick and easy integration (from existing or newly introduced base systems). And: supports formulating different scenarios and incrementing them, with a capability to easily analyse and compare the results. Copyright © 2008 Bancstreet Capital Partners Ltd. All Rights Reserved. www.bancstreet.com Questions? for further information please contact: Wolfgang Schwerdt Email: ws@bancstreet.com Pekka Luoma Email: pekka.luoma@bancstreet.com www.bancstreet.com Bancstreet Capital Partners Ltd Bancstreet House 21 Albert Road Hounslow, Middlesex United Kingdom Copyright © 2008 Bancstreet Capital Partners Ltd. All Rights Reserved. www.bancstreet.com