Stress Testing the Banking System: Comparing the EU and US Experience

Transcription

1 Stress Testing the Banking System: Comparing the EU and US Experience Tagung "Reform der Finanzmarktregulierung" 23 rd September, 2011 Markus Schwaiger Deputy Head Financial Markets Analysis and Surveillance Division

2 Agenda Macroeconomic Stress Testing: Methodologies and Approaches Methodological Approaches: Top-down and Bottom-up Comparison between US SCAP and EU-wide Stress Test Macroeconomic Stress Testing and Systemic Risk - 1 -

3 Macroeconomic Stress Testing A stress test is a what if scenario that takes the world as given but assumes a major change in one or more variables in order to see what effect this would have on various indicators. (IMF, 2003) Haldane (2009) - Executive Director for Financial Stability at Bank of England: 2008 will be remembered as the year stress testing failed

4 Methodologies: Top-down vs. Bottom-up Stress Testing Top-down Bottom-up Supervisory Actions Bank Actions + Pros Cons Scenario definition Calculation of expected losses (EL) or loan loss provisions, most often based on reporting data Aggregation of results Very limited Consistency in assumptions, methods, and risk parameters allows for best comparability of results (more objective perspective) Banks incapable of doing stress testing are provided with quality results Supervisory reporting data usually of lower granularity; more precise information from banks internal portfolio data Banks not (or limitedly) involved in assessment Scenario definition Provision of benchmarks Provision of bottom-up ST template Validation, discussions with banks Aggregation of Results Examine changes in risk parameters Calculation of ELs, or NPLs Detailed database (quality and granularity); internal models; own forecasts Increasing banks risk awareness Encouraging banks to stress test their portfolios Different models and parameters across individual banks considerable challenges in aggregating individual results Consistency checks more difficult Much more time-consuming - 3 -

5 Top-down Solvency Stress Testing Scenarios designed by supervisory authority Regulatory reporting data Publicly available (market) data Authority performs stress test(s) for each bank individually and on a the system level Credit risk Domestic exposure Non-domestic exposure FX loans Market risk Interest rate risk Equity risk FX risk Other issues Sovereign risk Securitization Liquidity / funding Contagion risk Feedback to the real economy Publication of results (aggregates or individual results) Input for banking supervision Off-side analysis On-site analysis Feedback to banks (single bank- and sectoral results) - 4 -

6 Bottom-up Solvency Stress Testing Provision of methodology and macroeconomic scenarios, along with benchmark parameters Participating banks run stress tests using either their own tools and risk management infrastructure or apply the benchmark parameters Credit Risk Domestic exposures Foreign exposures Market Risk Interest rate risk Equity, FX risk Volatility risk Other (occasional) issues Sovereign risk Securitization Liquidity risk / funding risk Review of bottom-up results Result and process discussions with banks Result publication - 5 -

7 US SCAP and EU-wide Stress Test Comparison US SCAP EU-wide Stress Test Institution Fed EBA (plus EC, ECB and NSAs) Organizational complexity Medium Very high Approach Bottom-up and top-down Bottom-up / Top-down in Review Process Number of banks 19 in 2009 (consolidated) 91 in 2011 (consolidated) Coverage 2/3 of total assets; and >50% of US loans 60% of total assets Quality assurance Fed internal Multilateral peer review (EBA, NSAs, etc.) Criteria to pass Capital ratios above predefined threshold Capital ratios above predefined threshold Consequences Recapitalization (governmental support package); banks need to design action plan Recapitalization; other measures (restructuring, etc.) Main objective Communication Communication Transparency: Methodology Moderate Full disclosure a priori Transparency: Result Individual results Individual results - 6 -

8 US SCAP and EU-wide Stress Test Comparison US SCAP EU-wide Stress Test GDP growth scenario (1) 2009 Test 2011 Test Baseline forecast [%] 0.1% 3.7% GDP shock, stress scenario [pp] -2.8pp -4.0pp Losses in the stress scenario [EUR bn] 448 (2) 163 Recapitalization costs [EUR bn] [in% of GDP] (3) [0.4%] [0.2%] Capital (4) Tier 1 Common Equity EBA Core Tier 1 Capital threshold (minimum to pass) [%] 4% 5% Capital level, pre stress [%] 5.3% 8.9% Capital level, post stress [%] 3.0% 7.4% Reduction in capital ratio [pp] -2.3pp -1.5pp Number of banks below capital threshold 10/19 20/91 Notes: All figures consider the situation as of the starting point of the stress tests. Potential capital raisings or any other bank actions during the stress testing time-horizons are disregarded in these figures. (1) GDP growth is cumulated over the two-year time-horizon. (2) Losses of 448 EUR bn equal 599 USD bn (FX rate as of 7 May 2009: ). (3) Referring to EU and US 2008 GDP. (4) Esssentially, Tier 1 Common Equity is defined as Tier 1 capital, less preferred stock, less qualifying trust preferred securities, and less minority interests in subsidiaries. EBA Core Tier 1 was designed especially for the purpose of the 2011 EUwide Stress Test (and should not been confused with with the Basel III Common Equity Tier 1 capital). This measure takes the existing EU definition of Tier 1 net of deductions of participations in financial institutions and it excludes hybrid instruments including existing preference shares, but recognises existing government support measures

9 US SCAP and EU-wide Stress Tests Applaus and Critique Transparency a priori: disclosure of methodology/scenarios ex post: full coverage of (un)stressed exposures (incl. sovereign), positions, capital components most important part of exercise Risk coverage Including funding (costs), sovereign exposures, securitization Backstop measures For banks <5% rel. clear measures communicated Unclear how banks just above 5% are handled Severity and ignorance criticized Stress level deemed too low Test not focused on current risks (sovereign debt in particular) EU-wide Stress Test Stress test done too late; expired data base Information and disclosure Unlike any supervisory disclosure before Well timed (first half of 2009) Strategy Clear communication of purpose ( restore confidence in lareg U.S. banks ) Outcome as desired: Clear differentiation among institutions Backstop measures Very clear communication of purpose: stress test assessing capital needs; Banks not passing the test have one month to provide action plan Severity, ignorance, timeline criticized US SCAP Stress level deemed too low Sovereign debt disregarded at all Fed did not stick on publication timeline

10 US SCAP and EU-wide Stress Tests Media Statements EU-wide Stress Test The banking component can no longer be separated from sovereign and institutional developments. This is why Friday s publication of stress tests results, while useful, is unlikely to be the gamechanger it could have been two years ago (FT, 14 July 2010) [The EBA] is going to repeat the main error of its predecessor organisation last year, and fail to factor in a potential sovereign default. (FT, 17 July 2011) ( ) since the tests did not require banks to factor in a sovereign default, even though such an event is increasingly likely, a narrow pass is no guarantor of capital strength. (FT USA, 18 July 2011) Nevertheless they were an important exercise in disclosure. Analysts can now apply their own stress tests to bank data ( ). (The Economist, 18 July 2011) US SCAP A very well done public relations exercise ( ) have not stressed the balance sheets of the banks in a meaningful way ( ). (S. Johnson, CNN Interview, 7 May 2009) (R)egulators and bank executives are concerned about how disclosure is handled because weaker institutions could suffer a collapse in stock prices. (Bloomberg, 1 May 2009) ( ), putting regulators in the awkward position of picking winners and losers ( ). (NYT, 24 April 2009) Initially, ( ) planned to give each bank a numerical grade ( ) But ( ) that might unfairly stigmatize banks ( ). At all banks suggestion, we ( ) went to a PASS/PASS system. Congratulations, banks! (Saturday Night Live, 9 May 2011) - 9 -

11 US SCAP and EU-wide Stress Tests Success or Failure? US SCAP Purpose / Goal - Ensure adequate system capital to promote lending and restore investor confidence (Fed, Staff Report 409, 2009) Outcome - Individual capital requirements evaluated - Differentiation of banks (see next slide) - Restored confidence in US banks - Large media coverage EU-wide Stress Test - The objective ( ) is to assess the resilience of the EU banking system (EBA, Press Release, 13 January 2011) - Restore confidence in EU banks and EU banking system as a whole - Transparency was more appreciated than the results - Higher risk coverage was welcomed (funding, securitization, etc.) - Sovereign debt problem was not fully assessed

12 US SCAP and EU-wide Stress Tests Success or Failure? Major Bank Stock Price Correlation with Financial Stock Price Index 1 0,95 0,9 Correlation Coefficient 0,85 0,8 0,75 0,7 0,65 Key Events 0,6 0,55 Big Four Banks Big Eight Banks Regional Banks All Banks Geithner announcing SCAP (Feb 2009) Result Publication (7 May 2009) 0,5 Figure based on Lehnert, A. (2011): U.S. Stress Tests 2009, ATC Workshop on Stress Testing

13 US SCAP and EU-wide Stress Tests Institutional Setting ESRB EBA Board of Supervisors Management Board Sub-Groups (for specific issues) Quality Assurance Task Group EBA Committees EBA Stress Test Team EBA Stress Test Expert Group (EBA, EC, ECB, ESRB, NCB, NSA) ECB EC Euro Area (e.g. FSC) EU Committees (ATC, EFC, etc.) External Stakes (e.g. IMF, governments) NCBs and NSAs Banks Ministries of Finance Working Group on Backstop Measures Public (analysts, broader public, etc.) EU-wide Stress Test

14 Agenda Macroeconomic Stress Testing: Methodologies and Approaches Methodological Approaches: Top-down and Bottom-up Comparison between US SCAP and EU-wide Stress Test Macroeconomic Stress Testing and Market Failures

15 Stress Testing and Market Failures Haldane (2009) - Executive Director for Financial Stability, BoE: 2008 will be remembered as the year stress testing failed. Risk management failed Keynes test : It is better to be roughly right than to be precisely wrong. Haldane (2009) diagnoses three market failures Disaster myopia Network externalities Misaligned incentives Additionally, I d like to highlight the issue of Measure myopia 14

16 Stress Testing and Market Failures Measure Myopia What is wrong with this picture? 16% 14% 12% 10% 8% 6% 4% 2% Source: 0% OeNB 2005/ / / / /09 Source: OeNB Historical Capital Ratios for the Austrian Banking System during 2005/ /09 Tier 1 Ratio (Core plus Non-Core Tier 1) Capital Adequacy Ratio 15

17 Stress Testing and Market Failures Measure Myopia Ratios are static and retrospective Numerator: usually includes entities (e.g. hybrid capital, minority interest), which are not necessarily loss-absorbing in the case of an event Denominator: Risk-weighted Assets (RWA) - Difficulties in capturing off-balance entities - Risk models may give misleading results Ratios behave procyclical: capital reduces, and risk (i.e. RWA) rises during times of financial distress; both lowers the capital ratios. Sufficient solvency at the banking system s level does not necessarily imply a stable system crisis was induced by systemic risk, which capital ratios are unable to measure Interconnectedness Shock transmission, contagion 16

18 Stress Testing and Market Failures Disaster Myopia We are seeing things that were 25-standard deviation moves, several days in a row. David Viniar, Goldman Sachs CFO, in August Cited in Haldane (2009). What does that mean? If we assume a bell-shaped curve (i.e. elliptic distribution), a 7-standarddeviation-move is expected to occur once every 14 bn years which is roughly the age of the universe. Disaster myopia refers to agents propensity to underestimate the probability of adverse outcomes, in particular if dealing with very small probabilities (Haldane, 2009). Behavioral Finance/Economics: Kahnemann & Tversky s ( ) choice experiments: subjective probabilities Heuristics in human decision making (instead of Neoclassical optimization) - Availability, representativeness, overconfidence are central concepts - Irrelevance of history: overconfidence discourages taking lessons from past statistics In the end, all these systematic human patterns lead to risk pricing that were set much too low! 17

19 Stress Testing and Market Failures Disaster Myopia Daily VaR (1-day; 99,9 percentile) for S&P 500, 10 yrs Source: Federal Reserve Bank,

20 Stress Testing and Market Failures Disaster Myopia Daily VaR (1-day; 99,9 percentile) for S&P 500, 70 yrs Source: Federal Reserve Bank,

21 Stress Testing and Market Failures Disaster Myopia Frequency Distribution of Selected Austrian Macro Variables 45 Annual GDP growth rates 80 Unemployment Rates % -10% -5% 0% 5% 10% 15% (exkl , ) % 4% 8% 12% 16% 20% 24% (exkl , )

22 Stress Testing and Market Failures Misaligned Incentives Principal-agent problems in stress testing - at the internal level: risk managers risk-takers within firms risk return - at the external level: financial firms authorities/public take return bear cost Internal level - If ROEs are high, power moves from back to front offices, risk managers pessimistic opinion is ignored. - Hence, communication of adverse scenarios and stress test results to management becomes difficult. External level - Downside safety net: privatizing profits and socializing losses 21

23 Stress Testing and Market Failures Network Externalities Why is the banking industry different from (say) the retail industry? Consider the insolvency of a major competitor Network model for the Austrian interbank market In the end, the influence of systemic risk depends on - the collective behavior of the financial network s agents, - the intensity of connectivity within the financial system, and - the interactions between financial markets and the macro economy. Note: The figure shows for each bank its largest loan exposure to other banks. The size of the marker reflects the size of the bank. 22

24 Macroeconomic Stress Testing and Systemic Risk Systemic risk is the major building block of the financial crisis! Systemic risk is concerned with the joint distribution of losses of all market participants and requires modelling how losses are transmitted through the financial system. (Cont et al, 2010) Systemic risk is disregarded in almost all famous stress testing exercises! Top-down stress tests could in principle consider interdependencies during stress, i.e. a common loss distribution and shock transmission / contagion All bottom-up approaches, incl. famous ones, neglect this important aspect. Considerable problems due to non-transparency of counterparty exposure (complexity) 23

25 Macroeconomic Stress Testing Conclusion Stress testing will not cut the Gordian knot! It is an important risk management tool for financial supervision as well as banks themselves, but not a panacea for all policy questions. The potential of stress testing to detect future crisis will depend on learning from the past (disaster myopia ) 24

26 Thank you for your attention.

27 References Cont, R., et al. (2010): Too Interconnected to Fail: Contagion and Systemic Risk in Financial Networks. URL: proba.math.univ-parisdiderot.fr/mathfipronum/confrisk/slides/cont.pdf ECB (2010): Recent Advances in Modelling Systemic Risk using Network Analysis. European Central Bank, January Haldane, A. (2009): Why banks failed the stress test. Speech at the Marcus Evans Conference on Stress Testing, February IMF (2009): Assessing the Systemic Implications of Financial Linkages. In: Global Financial Stability Report, Ch. 2. Kahnemann, D. and Tversky, A. (1973): On the psychology of prediction. Psychological Review, 80, Kahneman, D. and Tversky, A., Eds. (2000): Choices, values and frames. New York: Cambridge University Press. Lehnert, A. (2011): U.S. Stress Tests 2009, ATC Workshop on Stress Testing