Credit Stress Loss Alexandre Kurth WM&SB Risk Control 22 October, 2009
Alphabet of Credit Risk Measures frequency of losses mean: Expected Loss Standard deviation Risk Measure: Unexpected Loss Confidencelevel of economic capital Statistical Loss, e.g. Expected Shortfall, CoC amount of loss Expected Loss expected credit risk costs direct costs Economic Capital cost of economic capital Stress Losses unacceptable risks absorbed by revenues absorbed by capital 1
What is Stress Testing About? Principles for sound stress testing practices and supervision; Basel Committee on Banking Supervision Consultative Document, issued for comment March 13, 2009 Quote: Stress testing is an important risk management tool that is used by banks as part of their internal risk management and, through the Basel II capital adequacy framework, is promoted by supervisors. Stress testing alerts bank management to adverse unexpected outcomes related to a variety of risks and provides an indication of how much capital might be needed to absorb losses should large shocks occur. Moreover, stress testing is a tool that supplements other risk management approaches and measures. It plays a particularly important role in: providing forward-looking assessments of risk; overcoming limitations of models and historical data; supporting internal and external communication; feeding into capital and liquidity planning procedures; informing the setting of a banks risk tolerance; and facilitating the development of risk mitigation or contingency plans across a range of stressed conditions. 2
Overview of Stress Loss Three mutually reinforcing Pillars Reports, Limits, Planning Measure, monitor and NEW control BASEL risk under CAPITAL stress ACCORD conditions MINIMUM CAPITAL REQUIREMENTS Combined bankwide stress test SUPERVISORY REVIEW OF CAPITAL ADEQUACY Portfolio specific stress tests MARKET DISCIPLINE Reverse stress test 3
Credit Stress Testing Landscape Real Estate Credit Portfolio WM&SB Lombard, Securities Financing Corporate Clients sticky portfolio which is managed in longterm view liquid portfolio which can be managed on shortterm basis majority of portfolio cannot be managed on short-term basis Swiss Real Estate Scenario (Replication 90ies-crisis) Scenario Analyses Scenarios: Global macro-economic scenarios Factor model to forecast PDs and LGDs Concentration Analyses Reverse Stress Testing Market risk scenarios (e.g. liquidity and epicenter scenarios: Collateral Concentration: (Issuer Group, Hedge Funds, Mutual Funds, Country) Industry concentration Top Client List: Expected Tail Loss, limit, exposure, expected Loss Industry concentration Define hypothetical bank-breaking scenarios, e.g. leading to a predefined target loss, not based on PDs, EADs, LGDs, not using a history based model [cf. Principles for sound stress testing practices and supervision, Par. 9] 4
Applications of EL, Credit VaR, Stress Loss, Loss Forecast Influence from risk measures on risk relevant applications Risk Measures Notional Expected Loss Credit VaR Stress Losses Actual Loss Forecast Applications * RWA Basel II Capital Pricing Performance Measurement Portfolio Thresholds / Triggers Strategy Risk Control Business Plan Basel II ICAAP (Pillar II) General Provisioning Pool CLLP there is no one-size-fits-all measure *) Applications are not exclusively driven by these measures; but they are influenced; there may be others as well. 5
Uncertainty of Stress Losses distribution of credit losses conditional on the stressed economic situation current EL stress loss The stress loss measure does not suggest the worst possible outcome Even if the economy turns out to be as assumed in the stress scenario, the loss outcome may significantly vary around the expected value (stress loss) alternative measure for stress loss could be a quantile or ETL of the distribution or confidence interval 6
ABSCHNITT 1 Lombard Stress
Collateral Concentration Even without a general market downturn, idiosyncratic, event-driven shocks proved disastrous in the past. Portfolio concentrations of exposures collateralized by marketable assets (e.g. Lombard) are measured by 'single' shocks 10-day price drop for selected historic events Accounting fraud / insolvency, Dec 01 Enron Accounting fraud, June 02 Accounting fraud / insolvency, Dec 03 Grounding, Oct 01 Share price manipulation, Jan 06 Worldcom Parmalat Swissair Livedoor Accounting fraud / CEO resignation, Feb 03 Failed strategy / asbestos risk, Oct 02 Failed clinical trial, Sep 06 Ahold ABB SGX Pharmaceuticals Accounting and civil fraud, June 02 Legal issues, July 06 Tyco Bet and win Dot.com bubble, Oct 00 General company crisis, Aug 98 Dot.com bubble, May 00 Accounting problems, Jan 04 Yahoo! Netscape Qualcomm Adecco Proposed shocks Vioxx, Oct 04 Merck 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% 8
Example of Collateral Concentration All figures in CHF Mio Title MV LV exposure X X X Title A 15 6 Title B 2 1.2 Title C 3 1.8 Total 20 9 6 X Total 5 3 Loss: CHF 1 Mio 9
Collateral Concentration Analysis Results Collateral concentration of issuer groups CHF mio Volatility Adtv Mcap Stress exposure Eff [actual] Issuer Group Issuer Rating May09 Feb09 # stress exp. 1 (1) Issuer A 100 105 86 [1664] Equity 25% 315 86'716 26 Debt A1 8 Fiduciary 67 Structured products 0 95% Shock 2 (2) Issuer B 65 90 1 [314] Equity 28% 384 112'354 65 Debt Aa2 0 Fiduciary Structured products 0 3 (4) Issuer C 27 28 1 [99] Equity 46% 753 37'946 27 Debt A2 0 Fiduciary Structured products 0 illustrative figures 10
Market Scenario Stress Testing in Credit Client relationship Equities Fixed Income Precious Metals Cash Hedge Funds Mutual Funds Struct. Products other Traded Products Banking Products Collateral Exposure market value Before stress event Scenarios / market stress events 10-day risk factor shocks for Equity indices FX rates Interest rates Country spreads Credit spreads Shock transmission Risk factor shocks are translated into product-specific shocks Exposure side only receives FX shock to account for currency mismatches Stress exposure Additional unsecured exposure following a market stress event Stress ex. Equities Fixed Inc. P. Metals Cash H. Funds M. Funds St. Prod. other Traded Products Banking Products Collateral Exposure market value After stress event 11
Stress Test Scenario Description for Securities Financing The scenario suite consists of selected Group market stress as well as portfolio-specific scenarios Scenario 1 All markets down Scenario 2 EM crisis (pegs break) Scenario 3 Credit crunch Scenario 4 FX shocks only Scenario 5 specific scenario Scenario 6 specific scenario Libor rate shocks in bp Currency shocks vs. CHF Equity index shocks Topix Hang Seng CAC DAX FTSE 100 S&P500 SMI Topix Hang Seng CAC DAX FTSE 100 S&P500 SMI Topix Hang Seng CAC DAX FTSE 100 S&P500 SMI Topix Hang Seng CAC DAX FTSE 100 S&P500 SMI Topix Hang Seng CAC DAX FTSE 100 S&P500 SMI Topix Hang Seng CAC DAX FTSE 100 S&P500 SMI BRL RUB SGD HKD JPY GBP EUR USD BRL RUB SGD HKD JPY GBP EUR USD BRL RUB SGD HKD JPY GBP EUR USD BRL RUB SGD HKD JPY GBP EUR USD BRL RUB SGD HKD JPY GBP EUR USD USD BRL RUB SGD HKD JPY GBP EUR SGD HKD JPY GBP USD EUR CHF SGD HKD JPY GBP USD EUR CHF SGD HKD JPY GBP USD EUR CHF SGD HKD JPY GBP USD EUR CHF SGD HKD JPY GBP USD EUR CHF SGD HKD JPY GBP USD EUR CHF Market Risk Stress Test Scenario Portfolio specific Scenarios Note that the above charts only contain the most relevant risk factors. 12
Stress Scenario Results Stress Exposure in CHF million (indexed Mar 06) 6 5 4 3 2 1 0 1.00 1.78 0.87 0.82 1.27 1.33 1.71 2.54 Mar 06 May 06 Aug 06 Nov 06 Feb 07 May 07 Aug 07 Total Stress Exposure Nov 07 4.81 3.31 3.02 3.24 3.41 Feb 08 May 08 Aug 08 Nov 08 Feb 09 2.37 2.18 1.45 1.39 1.26 Mar 09 Apr 09 May 09 Jun 09 Jul 09 illustrative figures Stress results show portfolio exposition against a scenario Stress results are not a forecast Significant increase of Stress Exposure during the subprime crisis until November 2008 Continuous decrease of Stress Exposure since November 2008 13
ABSCHNITT 2 Macro-economic Stress Testing
Sensitivities Drivers Sensitivities Forecast Identify macroeconomic factors for each segment which impact the default rate, i.e. Interest Rate IR, GDP growth, RE prices, FX rate EUR/CHF Analyse historic time series to estimate how a change in a driver impacts the default rate resp. LGD, e.g. IR up 1% => PD up by x%, i.e. default rate for services (1.00%) changes by 1.00%*39% to 1.39%, if IR goes up by 1% Use sensitivities to forecast future PDs and LGDs for stress scenarios defined in terms of drivers, e.g. PD=f(IR, GDP growth, RE prices, ) Sensitivities for various economic drivers are estimated on portfolio relevant segmentation: Real Estate Lombard, Securities Financing Corporate Clients private mortgages IPRE private clients market stress e.g. Energy intensive Financial Services Manufacturing Services Real Estate Construction Restaurant Hotels 15
Macro-economic Stress Testing Methodology Explained Sensitivities to macroeconomic variables estimated based on internal history of impairment data Main drivers: GDP growth, interest rates, FX rate EUR/CHF, House Price changes PD = β + β GDPgrowth + β InterestRates + + β Volatility 0 1 1 2 2... K K + ε Application to given macroeconomic scenario % 0 2 4 6 8 1995q1 1996q1 1997q1 1998q1 1999q1 2000q1 Illustrative example 2001q1 2002q1 2003q1 2004q1 2005q1 2006q1 2007q1 2008q1 2009q1 2010q1 Upper confidence bound Fitted Values / Forecasts under stress scenario Lower confidence bound Application Current Average PD in Portfolio segment Stressed Default Rate Levels time Past time-series of default rates, used to derive sensitivities to macro-variables like GDP, interest rates Future stress scenario, specifying GDP, interest rates, FX-rates, etc. 2009q1 2008q1 2010q1 2009q1 2011q1 2010q1 16
ABSCHNITT 3 Real Estate Stress
A New Swiss Real Estate Crisis Replication of the Swiss real estate bubble from the 90ies Starting point: favorable economic environment in Swiss real estate market Inflation picks up throughout the country Property prices go up, too, resulting in a general bubble SNB is forced to react and raises interest rates sharply Illustration sample figures The economy markedly slows down, unemployment starts to grow, the property markets show first signs of correction 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 A widespread fall of property prices follows Loss Rate Bankruptcy Rate Booked Losses 18
Credit Risk Drivers in Real Estate Portfolio Portfolio Slice along LtV and Affordability for private mortgages Loan-to-Value LGD Equity Red Area Losses expected due to - limited equity buffer - stressed affordability Orange Area No losses expected because of - very good affordability and - limited or missing loss potential Affordability Default Affordability = costs (mortgage, maintenance) income (corrected by 2 nd homes) 19
Impact of Scenarios Equity Real Estate Price Decline Shift to upper areas Affordability Interest Rate Increase Shift to right area Unemployment Rate Increase 0.1% 0.35% Increase of Default Rates 20
Market Activity Map for Swiss Owner-occupied Properties Market activity map for Swiss Owner-occupied Properties based on MS segmentation (mobilité spaciale) Market activity map is based on Swiss real estate market variables such as prices, construction activity; and macro-economic variables such as unemployment rate, tax rates and population (change, forecast) 21
Real Estate Scenario Cockpit Real Estate Scenario Cockpit Price forecast of price index per MS region (mobilité spaciale) auf Basis based on macroeconomic parameters and scenarios Forecasts per MS-Region for SFH Condominiums MFH Office Buildings The Real Estate Scenario Cockpit supports the visualization of the (long-term) sensitivity of the Swiss real estate market based on normal scenarios stress loss calcuation based on various stress scenarios 22
ABSCHNITT 4 Reverse Stress Testing
The "Reverse" part of Stress Testing Current scenarios ask "what happens if" whereas reverse stress testing asks "what needs to happen that we lose x" Conventional stress test Define scenario mainly based on history in terms of macroeconomic variables, e.g. GDP, IR, etc. Apply model based on internal risk measures parameters and historic time series Obtain stress loss figures, usually in a severe but not (necessarily) bank breaking dimension Reverse stress test Find out which (sub-) portfolio would be affected to which extent, NOT on the basis of historic time series Find hypothetic trigger events which could cause such an effect Define outcomes which threaten the viability of the whole firm, e.g. target loss breaking regulatory capital ratios Principles for sound stress testing practices and supervision; Basel Committee on Banking Supervision Consultative Document, issued for comment March 13, 2009 (cf. Principle 9, p.18) 24