The importance of continuing to enhance your organization's Stress Testing program Robert G. Kula, EVP KeyCorp October 5, 2011 R Kula Moodys confer Final 10 5 11 1
Disclaimer • The views expressed here are my own, and not necessarily those of KeyCorp. • I helped draft Key’s Presentation and Publishing Disclosure Guidelines. R Kula Moodys confer Final 10 5 11 2
Background KeyCorp: • Has assets of approximately $90 billion. • Is headquartered in Cleveland, Ohio. • Provides a wide range of retail and commercial banking, commercial leasing, investment management, consumer finance, and investment banking products and services to individual, corporate and institutional clients. • Has 1,040 full service retail banking branches in 14 states, and 1,547 ATMs. R Kula Moodys confer Final 10 5 11 3
My Group is part of Risk Management Quantitative Risk Analysis (QRA) is focused on: • Economic Capital Attribution and Reporting. • Commercial Risk Rating models. • Stress Testing. • Model Risk Control. A number of Moody’s tools are leveraged in the bank: • RiskCalc, RiskFrontier, RiskAnalyst, Economy.com, etc. • And we are a participant in the credit research database. R Kula Moodys confer Final 10 5 11 4
The connection between Stress Testing, ERM, and Economic Capital – and some high level “wisdom” • Economic Capital attribution was ERM before ERM. • Economic Capital attribution was stress testing before stress testing. • All implausible stress testing scenarios are somewhat plausible. • No matter how severe the stress is that you define, an even more extreme stress can be developed. • You need to document everything you do and everything you think about. (As well as everything you didn’t think about.) R Kula Moodys confer Final 10 5 11 5
“wisdom” continued • Bad times are a function of good times, i.e., bad = f(good). The trick is knowing when this will happen. What are the early warning signs? • Some loans that were good for a real long time will default during bad times. • Unfortunately, despite everything being done in the industry over the last two years (risk appetites, risk tolerances, stress testing, etc.), future downturns cannot be prevented. • When in doubt, you can always blame the model or modeler. • Despite what anyone says, no risk rating model is entirely “through-the-cycle” or purely “point-in-time.” All are a mix of the two. R Kula Moodys confer Final 10 5 11 6
Economic Capital and Stress Testing go hand in hand. Economic Capital attribution is a form of stress testing – looking out over the next year. Probability of Loss Equates to Confidence Level Capital Base Stress #1 Stress #2 | | | Reserves Reserves X X X Potential “Unexpected” Potential “Unexpected” Zero Expected Losses for Which Losses Against Which It Losses Loss Capital Should be Held Would be too Expensive to Hold Capital Loss 0% 100% Rate R Kula Moodys confer Final 10 5 11 7
Learnings around Stress Testing • Expanded Stress Testing activities are good for the business (i.e., good risk management), and a necessary component of an Internal Capital Adequacy Process (ICAP). • Coordinated bank-wide stress tests need to have a consistent starting point (or kickoff) – the agreed upon scenario (e.g., the forecasted macro economic indicators). • Stress Testing is a repeatable process, not a one-time project – there needs to be a dedicated team with clear roles & responsibilities (cutting across Treasury/Finance, Risk Management, the LOBs, etc). • Stress Testing needs to connect directly into Capital Planning, Profit Planning, and Strategic Planning. • There needs to be a focus on continuous improvement… R Kula Moodys confer Final 10 5 11 8
…For example, for Credit Risk, there are a number of models/methodologies that can be leveraged: • Risk Rating models – mapping macro economic variables to the drivers of capital for each risk (e.g., in the case of Credit Risk, the drivers are PD and LGD). • Historical transition matrices. • Economic Capital models (run at various confidence levels). And the more granular the better (i.e., segmenting by product, industry, geography, etc.). R Kula Moodys confer Final 10 5 11 9
…But remember that Stress Scenarios are developed to be applied across all risk types. Macro Scenario – Variables • GDP. • Unemployment. • Home Prices. • Credit spreads. • Consumer Price Index. Risks • Credit, Market, Operational, Liquidity, etc. R Kula Moodys confer Final 10 5 11 10
Learnings around Stress Testing (cont’d) • You need to continually strive for the right balance of quantitative analysis (e.g., modeling) and qualitative analysis (e.g., expert/fundamental credit analysis). • A risk management culture is enhanced with the development of a well-defined risk appetite and a set of risk tolerances. This brings even more meaning to activities such as stress testing (e.g., how far out of tolerance will my stress take me?). • Strong governance (e.g., an ERM Committee & the Board) is essential. • Third party perspectives and benchmarking can help enhance the overall process. • You should be thinking about stress testing all the time… R Kula Moodys confer Final 10 5 11 11
…and connecting your Stress Testing scenarios to the real world. • Something (e.g., Greece, price of oil, Egypt, Japan, etc.) actually causes the stress scenario – and this is reflected as a worsening GDP, higher unemployment, etc. R Kula Moodys confer Final 10 5 11 12
Making a list of Macro risks is easy. Translating each risk to an actionable scenario is more difficult. (And sometimes it’s hard to differentiate the causes from the effects.) Financial System - Lack of product demand - Failure to execute (creating credit risk, market risk, and operational risk) - Unintended consequences of regulation - Responses to the pressures to grow U.S. Economy - Poor leadership - Lack of Confidence/Fear - Lack of agreement on methodologies to avoid a prolonged slump - Drag from Europe World - Global warming - Oil - Weather events - Nuclear event(s) - Pandemic R Kula Moodys confer Final 10 5 11 13
Common challenges in the industry in developing enterprise-wide stress tests • Maintaining deep historical data, with good data quality. • Drilling down to the “right” level of granularity of the models. • Having well-defined Stress Testing documentation, which revolves around the: – Data used. – Assumptions made. – Models chosen. – Calculations. – Scenario(s) defined (e.g., economic conditions). • Understanding inter-risk correlations (is there diversification or a magnification effect?). • Strong Model Governance. R Kula Moodys confer Final 10 5 11 14
With more and more models being developed and utilized for Stress Testing: • Model risk control and model governance becomes even more essential. – Model Risk is a material risk. – Promoting good model building helps reduce model risk. R Kula Moodys confer Final 10 5 11 15
Things are still being defined and there are many things to think about • Proposed Guidance on Stress Testing. • Section 165(i) of the Dodd-Frank Act. • Reverse stress testing, living wills, etc. What can you do??? • Closely follow all regulatory and industry developments, and participate in industry trade groups to review and comment on proposed rules as they are published. • Stay closely engaged with bank regulators – what are best practices? And what are the range of practices? R Kula Moodys confer Final 10 5 11 16
Many Choices for Next License Plate • SCAP19 ??? CCAR??? CCAR II??? SIFI??? ICAP??? R Kula Moodys confer Final 10 5 11 17
Recommend
More recommend
