Adaptive Stress Testing

834 views

Published on

Presentation with Alan Laubsch at Enterprise Risk Management Symposium in Chicago on 24 April 2013. See http://www.ermsymposium.org/2013/

0 Comments
0 Likes
Statistics
Notes
  • Be the first to comment

  • Be the first to like this

No Downloads
Views
Total views
834
On SlideShare
0
From Embeds
0
Number of Embeds
4
Actions
Shares
0
Downloads
15
Comments
0
Likes
0
Embeds 0
No embeds

No notes for slide
  • Here’s the agenda for the next hour:First I’ll introduce a framework for Adaptive Stress Testing. The idea around Adaptive Stress testing is maximize learning feedback, very much like the trial and error evolutionary process of adaptation to new environments.Attention to Early Warning Signals is a crucial component to making Adaptive Stress Testing work in practice. Due to dramatic phase transition properties in complex systems, early warning and early response is essential for adapting successfully to changes in the environment. We take an interdisciplinary perspective, and look at lessons ranging from earthquake monitoring to epidemiology, as well as looking at some of the significant early warning signals we detected prior to the GFC and the current European sovereign crisis.Secondly, I’ll introduce StressGrades, which is a methodology we have been developing at Riskcommons over the last 18 months. StressGrades integrates market intelligence into the stress testing process to highlight escalating risks. I’ll show some backtesting results, focusing on StressGrade performance leading up to the GFC.StressGrades can also be effectively used in Reverse Stress Testing portfolios to explore and monitor key vulnerabilities.I will also introduce some of our ideas around visualizing risks. I believe effective visualization techniques will grow in great importance for risk managers, especially in focusing on unusual activity and getting a better understanding of interrelationships.I’ll conclude with some thoughts about key Adaptive Stress Testing practices, and ideas for sparking network intelligence around stress testing. And we’ll leave 15 minutes for Q&A at the end.
  • Here’s an interesting example of real time data visualization – cloud network graphs. This one was yesterday’s world news. Economist has some interesting clouds about finance and economics. Increasingly, people are using real time news and especially twitter to assess how scenarios are evolving. One thing is clear – we increasingly need to integrate interdisciplinary perspectives to better understand scenarios and how they could potentially evolve.
  • We live in an increasingly complex and fast moving world. Predict and control no longer works. A more sensible approach is Sense and Respond. A mountain biking race would be a good analogy. Of course we do all the homework and map out the course, get GPS and weather forecasts. But what makes the real difference is executing on the course, sensing and responding to the changing conditions of the course. This is a paradigm we will take to stress testing.
  • In addition to sensing tremors, we need to be aware of underlying fault lines, and how our portfolios are positioned.Here’s a chart of earthquake fault lines and nuclear plants. It should come as no surprise that Japan has suffered the first earthquake induced nuclear reactor meltdown. Japan has 10% of the world’s earthquake activity. Obviously not a good place to build nuclear plants at all. But prior to Fukishima, it was considered safe and unthinkable. Classic disaster myopia.
  • Early WarningYou can take a range of perspectives on early warning, ranging from imminently short term (e.g., jumps in equity implied volatility before corporate events) to very long term (e.g., macro-economic imbalances). It makes sense for us to address the whole time spectrum.We can approach this from a long term and short term perspective:Long Term: top down analysis based on diagnosing structural risks (especially bubbles) and scenario analysisBottom Up: based on specific portfolio vulnerabilities, driven by short term market factorsIn both cases, we identify key variables to monitor (the stakeout) and focus on any unusual movements.
  • FinancialsDaily upside excessions1.89%Daily downside excessions1.32%Weekly upside excessions1.32%Weekly downside excessions1.79%
  • This is an interesting graph from a reputation risk consultancy. It shows the rapid phase transition of reputation risk, and how the only time to potentially exert control is before the exponential viral spread. However, most firms typically only respond at the inflection point, after a tipping point has been crossed and there is little chance to exert control.
  • Here’s the agenda for the next hour:First I’ll introduce a framework for Adaptive Stress Testing. The idea around Adaptive Stress testing is maximize learning feedback, very much like the trial and error evolutionary process of adaptation to new environments.Attention to Early Warning Signals is a crucial component to making Adaptive Stress Testing work in practice. Due to dramatic phase transition properties in complex systems, early warning and early response is essential for adapting successfully to changes in the environment. We take an interdisciplinary perspective, and look at lessons ranging from earthquake monitoring to epidemiology, as well as looking at some of the significant early warning signals we detected prior to the GFC and the current European sovereign crisis.Secondly, I’ll introduce StressGrades, which is a methodology we have been developing at Riskcommons over the last 18 months. StressGrades integrates market intelligence into the stress testing process to highlight escalating risks. I’ll show some backtesting results, focusing on StressGrade performance leading up to the GFC.StressGrades can also be effectively used in Reverse Stress Testing portfolios to explore and monitor key vulnerabilities.I will also introduce some of our ideas around visualizing risks. I believe effective visualization techniques will grow in great importance for risk managers, especially in focusing on unusual activity and getting a better understanding of interrelationships.I’ll conclude with some thoughts about key Adaptive Stress Testing practices, and ideas for sparking network intelligence around stress testing. And we’ll leave 15 minutes for Q&A at the end.
  • Here’s the agenda for the next hour:First I’ll introduce a framework for Adaptive Stress Testing. The idea around Adaptive Stress testing is maximize learning feedback, very much like the trial and error evolutionary process of adaptation to new environments.Attention to Early Warning Signals is a crucial component to making Adaptive Stress Testing work in practice. Due to dramatic phase transition properties in complex systems, early warning and early response is essential for adapting successfully to changes in the environment. We take an interdisciplinary perspective, and look at lessons ranging from earthquake monitoring to epidemiology, as well as looking at some of the significant early warning signals we detected prior to the GFC and the current European sovereign crisis.Secondly, I’ll introduce StressGrades, which is a methodology we have been developing at Riskcommons over the last 18 months. StressGrades integrates market intelligence into the stress testing process to highlight escalating risks. I’ll show some backtesting results, focusing on StressGrade performance leading up to the GFC.StressGrades can also be effectively used in Reverse Stress Testing portfolios to explore and monitor key vulnerabilities.I will also introduce some of our ideas around visualizing risks. I believe effective visualization techniques will grow in great importance for risk managers, especially in focusing on unusual activity and getting a better understanding of interrelationships.I’ll conclude with some thoughts about key Adaptive Stress Testing practices, and ideas for sparking network intelligence around stress testing. And we’ll leave 15 minutes for Q&A at the end.
  • Here’s the agenda for the next hour:First I’ll introduce a framework for Adaptive Stress Testing. The idea around Adaptive Stress testing is maximize learning feedback, very much like the trial and error evolutionary process of adaptation to new environments.Attention to Early Warning Signals is a crucial component to making Adaptive Stress Testing work in practice. Due to dramatic phase transition properties in complex systems, early warning and early response is essential for adapting successfully to changes in the environment. We take an interdisciplinary perspective, and look at lessons ranging from earthquake monitoring to epidemiology, as well as looking at some of the significant early warning signals we detected prior to the GFC and the current European sovereign crisis.Secondly, I’ll introduce StressGrades, which is a methodology we have been developing at Riskcommons over the last 18 months. StressGrades integrates market intelligence into the stress testing process to highlight escalating risks. I’ll show some backtesting results, focusing on StressGrade performance leading up to the GFC.StressGrades can also be effectively used in Reverse Stress Testing portfolios to explore and monitor key vulnerabilities.I will also introduce some of our ideas around visualizing risks. I believe effective visualization techniques will grow in great importance for risk managers, especially in focusing on unusual activity and getting a better understanding of interrelationships.I’ll conclude with some thoughts about key Adaptive Stress Testing practices, and ideas for sparking network intelligence around stress testing. And we’ll leave 15 minutes for Q&A at the end.
  • Here’s the agenda for the next hour:First I’ll introduce a framework for Adaptive Stress Testing. The idea around Adaptive Stress testing is maximize learning feedback, very much like the trial and error evolutionary process of adaptation to new environments.Attention to Early Warning Signals is a crucial component to making Adaptive Stress Testing work in practice. Due to dramatic phase transition properties in complex systems, early warning and early response is essential for adapting successfully to changes in the environment. We take an interdisciplinary perspective, and look at lessons ranging from earthquake monitoring to epidemiology, as well as looking at some of the significant early warning signals we detected prior to the GFC and the current European sovereign crisis.Secondly, I’ll introduce StressGrades, which is a methodology we have been developing at Riskcommons over the last 18 months. StressGrades integrates market intelligence into the stress testing process to highlight escalating risks. I’ll show some backtesting results, focusing on StressGrade performance leading up to the GFC.StressGrades can also be effectively used in Reverse Stress Testing portfolios to explore and monitor key vulnerabilities.I will also introduce some of our ideas around visualizing risks. I believe effective visualization techniques will grow in great importance for risk managers, especially in focusing on unusual activity and getting a better understanding of interrelationships.I’ll conclude with some thoughts about key Adaptive Stress Testing practices, and ideas for sparking network intelligence around stress testing. And we’ll leave 15 minutes for Q&A at the end.
  • Adaptive Stress Testing

    1. 1. www.riskcommons.org | www.fna.fi Alan Laubsch, alaubsch@gmail.com | Kimmo Soramaki, kimmo@fna.fiAdaptive Stress TestingHarnessing Network Intelligence in Stress Testingand Reverse Stress TestingERM Symposium, Chicago April 24 2013
    2. 2. 2www.riskcommons.orgwww.riskcommons.org | www.fna.fi Alan Laubsch, alaubsch@gmail.com | Kimmo Soramaki , kimmo@fna.fi 2www.riskcommons.org | www.fna.fi Alan Laubsch, alaubsch@gmail.com | Kimmo Soramaki , kimmo@fna.fi 2Agenda1. A Framework for Adaptive Stress Testing2. Signal or Noise?3. Introducing StressGrades™4. Network Approaches to Stress Testing5. Summary and Conclusions
    3. 3. 3www.riskcommons.orgwww.riskcommons.org | www.fna.fi Alan Laubsch, alaubsch@gmail.com | Kimmo Soramaki , kimmo@fna.fi 3Scenarios are continually emerging and evolving• Integrate interdisciplinary perspectivesSource: infomous.com/
    4. 4. 4www.riskcommons.orgwww.riskcommons.org | www.fna.fi Alan Laubsch, alaubsch@gmail.com | Kimmo Soramaki , kimmo@fna.fi 4Implication of Complexity: “Sense and Respond”Dynamic Steering: continual feedback
    5. 5. 5www.riskcommons.orgwww.riskcommons.org | www.fna.fi Alan Laubsch, alaubsch@gmail.com | Kimmo Soramaki , kimmo@fna.fi 5Seek to understand systemic fault lines…• …and how is your portfolio is positioned relative to fault lines.• Major challenge: disaster myopia (see “Why Banks Failed the Stress Tests”by A. Haldane, 2009)Earthquake activity vs Nuclear power plantsSource: http://googlemapsmania.blogspot.com/2011/03/nuclear-power-plants-earthquake.html
    6. 6. 6www.riskcommons.orgwww.riskcommons.org | www.fna.fi Alan Laubsch, alaubsch@gmail.com | Kimmo Soramaki , kimmo@fna.fi 6www.riskcommons.org | www.fna.fi Alan Laubsch, alaubsch@gmail.com | Kimmo Soramaki , kimmo@fna.fi 6I. Macro: identify structural risks (potential risks)• Stress Library based on Thought Leaders (Innovators)• Awareness of systemic cycles, in particular credit and asset bubbles• Financial or economic imbalances (e.g., capital flows, consumption vs. saving)• Examples: Shiller – (a) tech bubble (2000) and (b) housing bubble (2005)II. Micro: monitor potential precipitating events (visible risks)• Focus on short term market movements, especially outliers and regime shifts• Early Warning: identify amplification mechanisms and critical (tipping) points• Examples: vol spike in (a) tech stocks and (b) US mortgage securities & financialsAdaptive Stress Testing Framework
    7. 7. 7www.riskcommons.orgwww.riskcommons.org | www.fna.fi Alan Laubsch, alaubsch@gmail.com | Kimmo Soramaki , kimmo@fna.fi 7www.riskcommons.org | www.fna.fi Alan Laubsch, alaubsch@gmail.com | Kimmo Soramaki , kimmo@fna.fi 7Designing an Adaptive Stress LibrarySource: Wikipedia; see Geoffrey Moore’s “Crossing the Chasm” (1999)• Diffusion of ideas and innovation follow a predictable course after atipping point is crossedTwo key perspectives for stress testing1. Stress Library: Innovators2. Early Warning: Early AdoptersInnovators:Roubini, Rosenberg, Shiller,Rogoff, Reinhart, Ferguson, …Key early adoption signals:- Outliers clustering, vol spikes,super-exponential trends- Adoption by hedge funds andbroker dealers.
    8. 8. 8www.riskcommons.orgwww.riskcommons.org | www.fna.fi Alan Laubsch, alaubsch@gmail.com | Kimmo Soramaki , kimmo@fna.fi 8US Financials Case StudyFinancial Meltdown (“Roubini”) scenario escalates from ’07 and peaks March ’09 and thendeclines… inverse Financial Recovery scenario emerges-20.0%-15.0%-10.0%-5.0%0.0%5.0%10.0%15.0%20.0%Daily 99% VaR Backtest (.94 decay, Student t)Feb 27 „07 outlierSource: Alan Laubsch, “Equities as Collateral In U.S. Securities Lending Transactions”,The RMA Executive Committee on Securities Lending & RiskMetrics, March 2011March 6 Market bottomJune 1 Market peaksChart: U.S. Financials “death star pulse”
    9. 9. 9www.riskcommons.orgwww.riskcommons.org | www.fna.fi Alan Laubsch, alaubsch@gmail.com | Kimmo Soramaki , kimmo@fna.fi 9Tipping Point Dynamics require early detection and action• Limited window of opportunity for exerting control• What are early warning signals of a phase transition?Source: “Building A Reputation Risk Management Capability”, Diermeier & Loeb, 2011Invisible/Potential Visible & amplifying
    10. 10. 10www.riskcommons.orgwww.riskcommons.org | www.fna.fi Alan Laubsch, alaubsch@gmail.com | Kimmo Soramaki , kimmo@fna.fi 10www.riskcommons.org | www.fna.fi Alan Laubsch, alaubsch@gmail.com | Kimmo Soramaki , kimmo@fna.fi 10Agenda1. A Framework for Adaptive Stress Testing2. Signal or Noise?3. Introducing StressGrades™4. Network Approaches to Stress Testing5. Summary and Conclusions
    11. 11. 11www.riskcommons.orgwww.riskcommons.org | www.fna.fi Alan Laubsch, alaubsch@gmail.com | Kimmo Soramaki , kimmo@fna.fi 11www.riskcommons.org | www.fna.fi Alan Laubsch, alaubsch@gmail.com | Kimmo Soramaki , kimmo@fna.fi 11Exogenous vs Endogenous Crises• Nassim Taleb’s “Black Swan” claims that crises arise from unknowableevents that cannot quantified or predicted• Historical examples: Eisenhower heart attack; Lincoln & Kennedy assassinations;asteroid impact or flood basalt eruptions resulting in mass extinctions; 911;• Didier Sornette’s “Dragon King” thesis holds that most financial crises areendogenous in nature and can be diagnosed in advance, can bequantified, and have some predictability• Examples of endogenous crises in history: rise of Fascism; rise of dictators(Hitler, Mao);‟29 Great Depression, ‟87 Black Monday, ‟89 Japan Bubble; ‟01 TechBubble; GFC; current ecological crisis• Endogenous structural risk combined with exogenous precipitating eventis common (e.g., forest fire)Source: Alan Laubsch “Integrated Risk Management - Early Overview”, RiskMetrics
    12. 12. 12www.riskcommons.orgwww.riskcommons.org | www.fna.fi Alan Laubsch, alaubsch@gmail.com | Kimmo Soramaki , kimmo@fna.fi 12www.riskcommons.org | www.fna.fi Alan Laubsch, alaubsch@gmail.com | Kimmo Soramaki , kimmo@fna.fi 12Phase transitions can result from amplifying feedback• Super-exponential instability and change characterizes phase transitionsSee: http://www.er.ethz.ch/presentations/Endo_Exo_Oxford_17Jan08.pdfSource: Sornette et al., Endogenous versus Exogenous Origins of Crises (2008)
    13. 13. 13www.riskcommons.orgwww.riskcommons.org | www.fna.fi Alan Laubsch, alaubsch@gmail.com | Kimmo Soramaki , kimmo@fna.fi 13www.riskcommons.org | www.fna.fi Alan Laubsch, alaubsch@gmail.com | Kimmo Soramaki , kimmo@fna.fi 13Subprime CDO volatilities spiked 7 & 4 months before the meltdown-100.0%-80.0%-60.0%-40.0%-20.0%0.0%20.0%40.0%60.0%80.0%100.0%120.0%7/19/20068/19/20069/19/200610/19/200611/19/200612/19/20061/19/20072/19/20073/19/20074/19/20075/19/20076/19/20077/19/20078/19/20079/19/200710/19/200711/19/200712/19/20071/19/20082/19/20083/19/20084/19/20085/19/20086/19/2008DateSpreadChange300%+ increase in vol fromDec 12 to 21 06357% vol spike onFeb 23 07RM 2006 99% VaR bands vs 2006-1 AAA spread changesOne major outlier, a 12 sd move on Feb 23 07, the dayafter the $10.5bn HSBC loss announcementBacktesting summary:2.4% upside excessions0.81% downside excessionsMajor ratings agencies initiate reviews and/ordowngrades week of July 9 07Source: Alan Laubsch “Subprime Risk Management Lessons”, RiskMetricsGS exitssubprime
    14. 14. 14www.riskcommons.orgwww.riskcommons.org | www.fna.fi Alan Laubsch, alaubsch@gmail.com | Kimmo Soramaki , kimmo@fna.fi 14www.riskcommons.org | www.fna.fi Alan Laubsch, alaubsch@gmail.com | Kimmo Soramaki , kimmo@fna.fi 14The Dec ‟06 and Feb ‟07 spikes in volatility can be seen as tremors(foreshocks) that cascaded into a major earthquake2006-1 AAA0501001502002503003504004501/19/20062/19/20063/19/20064/19/20065/19/20066/19/20067/19/20068/19/20069/19/200610/19/200611/19/200612/19/20061/19/20072/19/20073/19/20074/19/20075/19/20076/19/20077/19/20078/19/20079/19/200710/19/200711/19/200712/19/20071/19/20082/19/20083/19/20084/19/20085/19/20086/19/2008The first tremor(vol up 300% Dec 12-21)Feb 23 07, first majoroutlier, 350% vol increasein 1 day, 12sd moveJune 20 07, ML triesto liquidate BearSubprime CDOsAbsolute Spread LevelsMajor ratings agenciesinitiate reviews and/ordowngrades week ofJuly 9 07bps• Absolute spread moves were small, but rate of change was super-exponential. Parallels to failure andrupture process in material science (pressure to break point)
    15. 15. 15www.riskcommons.orgwww.riskcommons.org | www.fna.fi Alan Laubsch, alaubsch@gmail.com | Kimmo Soramaki , kimmo@fna.fi 15www.riskcommons.org | www.fna.fi Alan Laubsch, alaubsch@gmail.com | Kimmo Soramaki , kimmo@fna.fi 15Feb 27 „07 DJIA outlier marks the beginning of a phase transition withincreasing waves of volatility• Increasing amplitude of volatility is a telltale sign of endogenous crises-5.00%-4.00%-3.00%-2.00%-1.00%0.00%1.00%2.00%3.00%4.00%5.00%1/23/20063/23/20065/23/20067/23/20069/23/200611/23/20061/23/20073/23/20075/23/20077/23/20079/23/200711/23/20071/23/20083/23/20085/23/2008DJIA daily returns vs 99% VaR bands (.94 decay, t dist)Feb 27, 6th biggest outlier in DJIAhistory 4 days after largest spike insubprime spreadsSource: Alan Laubsch “Integrated Risk Management - Early Overview”, RiskMetrics
    16. 16. 16www.riskcommons.orgwww.riskcommons.org | www.fna.fi Alan Laubsch, alaubsch@gmail.com | Kimmo Soramaki , kimmo@fna.fi 16Gold Early Warning
    17. 17. 17www.riskcommons.orgwww.riskcommons.org | www.fna.fi Alan Laubsch, alaubsch@gmail.com | Kimmo Soramaki , kimmo@fna.fi 17Let‟s look more closely at Outliers
    18. 18. 18www.riskcommons.orgwww.riskcommons.org | www.fna.fi Alan Laubsch, alaubsch@gmail.com | Kimmo Soramaki , kimmo@fna.fi 18Downside outlier Clustering Escalates from Oct 2012
    19. 19. 19www.riskcommons.orgwww.riskcommons.org | www.fna.fi Alan Laubsch, alaubsch@gmail.com | Kimmo Soramaki , kimmo@fna.fi 19www.riskcommons.org | www.fna.fi Alan Laubsch, alaubsch@gmail.com | Kimmo Soramaki , kimmo@fna.fi 19Agenda1. A Framework for Adaptive Stress Testing2. Signal or Noise?3. Introducing StressGrades™4. Network Approaches to Stress Testing5. Summary and Conclusions
    20. 20. 20www.riskcommons.orgwww.riskcommons.org | www.fna.fi Alan Laubsch, alaubsch@gmail.com | Kimmo Soramaki , kimmo@fna.fi 20www.riskcommons.org | www.fna.fi Alan Laubsch, alaubsch@gmail.com | Kimmo Soramaki , kimmo@fna.fi 20Typical stress testing processes generate much data,but not necessarily intelligence“We run over 180 stress scenarios against each of our counterparties on a dailybasis. But we don‟t know what to do with the information” – Risk manager atglobal bankKey questions:• With overwhelming amount of data, which scenarios to focus on?• …given market conditions (systemic)• …given our portfolio exposures (specific)
    21. 21. 21www.riskcommons.orgwww.riskcommons.org | www.fna.fi Alan Laubsch, alaubsch@gmail.com | Kimmo Soramaki , kimmo@fna.fi 21DStressPStressFigure: Student t distributions and Z-scoresWe estimate DStress w/market implied vols (and correlations for multi factor scenarios) andPStress using a distributional assumption (e.g., Normal or Student t).StressGrades™ harness market intelligence highlightemerging risksWe define three components of StressGrades™:1. PStress = Market Implied Probability of a Stress Scenario2. DStress = Distance to stress scenario in standard deviations (z-score)3. StressQ = Quantile (percentile) historical rank of stress scenario (e.g., StressQ = .82implies stress levels have exceeded current levels 18% of the time)
    22. 22. 22www.riskcommons.orgwww.riskcommons.org | www.fna.fi Alan Laubsch, alaubsch@gmail.com | Kimmo Soramaki , kimmo@fna.fi 22www.riskcommons.org | www.fna.fi Alan Laubsch, alaubsch@gmail.com | Kimmo Soramaki , kimmo@fna.fi 22Stress Grades™ provide early warning and can bebacktestedS&P 500 Case Study:• Since 1987, the biggest one day drop in the S&P 500 was a 9.6% fall on Dec 1 ‟08,which we use to calibrate and backtest our StressGrade scenario.• DStress escalates from -24sd to -2sd before Dec „08 drop. Regime shift warningFeb„07
    23. 23. 23www.riskcommons.orgwww.riskcommons.org | www.fna.fi Alan Laubsch, alaubsch@gmail.com | Kimmo Soramaki , kimmo@fna.fi 23www.riskcommons.org | www.fna.fi Alan Laubsch, alaubsch@gmail.com | Kimmo Soramaki , kimmo@fna.fi 23S&P500 cont’d: Super-exponential increase in PStress:170x on Feb 27 then another 1300x before Dec 1 ’08• Note log scale on the PStress Chart below (right scale)
    24. 24. 24www.riskcommons.orgwww.riskcommons.org | www.fna.fi Alan Laubsch, alaubsch@gmail.com | Kimmo Soramaki , kimmo@fna.fi 24www.riskcommons.org | www.fna.fi Alan Laubsch, alaubsch@gmail.com | Kimmo Soramaki , kimmo@fna.fi 24ETF analysis reveals systemic risk early warning signals• Implied Probability of Stress Event (PStress) for major ETFs shows super-exponential escalation during GFCLogScaleSource: Alan Laubsch, “Introduction to StressGrades©”, riskcommons.org, 2011
    25. 25. 25www.riskcommons.orgwww.riskcommons.org | www.fna.fi Alan Laubsch, alaubsch@gmail.com | Kimmo Soramaki , kimmo@fna.fi 25www.riskcommons.org | www.fna.fi Alan Laubsch, alaubsch@gmail.com | Kimmo Soramaki , kimmo@fna.fi 25Super-exponential increase in PStress preceded marketcrash…• … and broad declines in PStress from peak levels signaled market recoverySource: Alan Laubsch, “Introduction to StressGrades©”, riskcommons.org, 2011
    26. 26. 26www.riskcommons.orgwww.riskcommons.org | www.fna.fi Alan Laubsch, alaubsch@gmail.com | Kimmo Soramaki , kimmo@fna.fi 26www.riskcommons.org | www.fna.fi Alan Laubsch, alaubsch@gmail.com | Kimmo Soramaki , kimmo@fna.fi 26Highlight escalating and large risks in StressGrades™ HeatMapMouse over to getmore informationabout each scenarioStressGrade™ (PStress)%changeinStresGrade0 100 200 300 40010%-10%100%1000%2. PIGS Inflationary Bust- StressGrade 385 up 550%0%Highest Priority:Escalating and HighPStressEarly Warning:Moderate PStressbut escalating
    27. 27. 27www.riskcommons.orgwww.riskcommons.org | www.fna.fi Alan Laubsch, alaubsch@gmail.com | Kimmo Soramaki , kimmo@fna.fi 27www.riskcommons.org | www.fna.fi Alan Laubsch, alaubsch@gmail.com | Kimmo Soramaki , kimmo@fna.fi 27Outlier Analysis can identify regime shiftsRank % Move in PStress Scenario [each hyperlinked]1 780% to 389 PStress Sovereign default2. 690% to 355 PStress Deflationary bust….23. 55% to 80 PStress Gold spike• Sample Outlier Analysis: 5% threshhold StressGradiealyss• 23 of 80 scenarios were outliers• 6 Outliers Average over 12 months
    28. 28. 28www.riskcommons.orgwww.riskcommons.org | www.fna.fi Alan Laubsch, alaubsch@gmail.com | Kimmo Soramaki , kimmo@fna.fi 28www.riskcommons.org | www.fna.fi Alan Laubsch, alaubsch@gmail.com | Kimmo Soramaki , kimmo@fna.fi 28Network Graphs allow visualization of interrelationships• Potential to integrate stress themes into interactive network graphs andplay movie of changing correlation and volatility dynamics over time
    29. 29. 29www.riskcommons.orgwww.riskcommons.org | www.fna.fi Alan Laubsch, alaubsch@gmail.com | Kimmo Soramaki , kimmo@fna.fi 29www.riskcommons.org | www.fna.fi Alan Laubsch, alaubsch@gmail.com | Kimmo Soramaki , kimmo@fna.fi 29Agenda1. A Framework for Adaptive Stress Testing2. Signal or Noise?3. Introducing StressGrades™4. Network approaches to Stress Testing5. Summary and Conclusions
    30. 30. 30www.riskcommons.orgwww.riskcommons.org | www.fna.fi Alan Laubsch, alaubsch@gmail.com | Kimmo Soramaki , kimmo@fna.fi 30www.riskcommons.org | www.fna.fi Alan Laubsch, alaubsch@gmail.com | Kimmo Soramaki , kimmo@fna.fi 30DStress NetworkHow are asset stresses coordinated?We calculate the Euclidean distance betweenpairwise series of daily DStress values.Keep only most important links that form thebackbone dependencies, i.e. present a datareduction.Size of node scales with risk as defined byaverage DStress during the period: Large node,high risk. Small node, low risk.The network shows us the coordination ofstress among the assets in a portfolio.Jan 20 - April 19 2013http://www.fna.fi/demos/erm/dstress-tree.html
    31. 31. 31www.riskcommons.orgwww.riskcommons.org | www.fna.fi Alan Laubsch, alaubsch@gmail.com | Kimmo Soramaki , kimmo@fna.fi 31www.riskcommons.org | www.fna.fi Alan Laubsch, alaubsch@gmail.com | Kimmo Soramaki , kimmo@fna.fi 31Stress Testing a Portfolio - Opening up the Black BoxPartial correlation measures the degree ofassociation between two random variables,controlling for other variablesNetwork of statistically significant partialcorrelations of dailyt returns for a wide setETFs during 2009-2013• link = partial correlation• green node = positive return• red node = negative return• node size scales with absolute returnWe can use the partial correlations toundestand linkages within a standardportfolio stress test model
    32. 32. 32www.riskcommons.orgwww.riskcommons.org | www.fna.fi Alan Laubsch, alaubsch@gmail.com | Kimmo Soramaki , kimmo@fna.fi 32www.riskcommons.org | www.fna.fi Alan Laubsch, alaubsch@gmail.com | Kimmo Soramaki , kimmo@fna.fi 32Calculating Partial Correlation• We build regression models for daily returns of e.g. Oil and Gold based on all otherassets of interest and look at the correlation of their model residuals (i.e. what isleft unexplained by the other factors) -> Partial correlationModel 1: Regress Gold on all other assets except OilModel 2: Regress Oil on all other assets except Gold• Gold residuals = vector of differences between observed Gold values and valuespredicted by Model 1• Oil residuals = vector of differences between observed Oil values and valuespredicted by Model 2• Partial correlation between Oil and Gold is the correlation between Oil residuals andGold residuals32
    33. 33. 33www.riskcommons.orgwww.riskcommons.org | www.fna.fi Alan Laubsch, alaubsch@gmail.com | Kimmo Soramaki , kimmo@fna.fi 33www.riskcommons.org | www.fna.fi Alan Laubsch, alaubsch@gmail.com | Kimmo Soramaki , kimmo@fna.fi 33The Network for an Oil shockhttp://www.fna.fi/demos/erm/cascade-oil-01.html
    34. 34. 34www.riskcommons.orgwww.riskcommons.org | www.fna.fi Alan Laubsch, alaubsch@gmail.com | Kimmo Soramaki , kimmo@fna.fi 34www.riskcommons.org | www.fna.fi Alan Laubsch, alaubsch@gmail.com | Kimmo Soramaki , kimmo@fna.fi 34Shocking multiple nodes• We use multivariate percentiles (based on the multivariate normaldistribution) to simultaneously shock Financials, German Stocks and Gold• First we estimate the mean and covariance matrix of these three assetreturns from theobserved data.• Then, for the first percentile, we find the schocks x, y, and z such that thejoint probability P(XLF < x AND EWG < y AND GLD < z) = 0.01 and themarginal probabilities are equal, i.e., P(XLF < x) = P(EWG < y) = P(GLD < z)• A similar calculation finds the 99th percentile.
    35. 35. 35www.riskcommons.orgwww.riskcommons.org | www.fna.fi Alan Laubsch, alaubsch@gmail.com | Kimmo Soramaki , kimmo@fna.fi 35www.riskcommons.org | www.fna.fi Alan Laubsch, alaubsch@gmail.com | Kimmo Soramaki , kimmo@fna.fi 35The Network for Multiple Shockshttp://www.fna.fi/demos/erm/cascade-three-01.html
    36. 36. 36www.riskcommons.orgwww.riskcommons.org | www.fna.fi Alan Laubsch, alaubsch@gmail.com | Kimmo Soramaki , kimmo@fna.fi 36www.riskcommons.org | www.fna.fi Alan Laubsch, alaubsch@gmail.com | Kimmo Soramaki , kimmo@fna.fi 36Is it correct?The test:• We develop a model where we use the network structure to estimate manysmall models (some of which are based on estimates)• We see how well cascading predictions works by predicting values for a outof sample data set whose values are known.• We compare results to a normal linear model• Result: Predictions based on partial correlation network are as good forsingle asset shock, and just slightly worse for multiple asset shock-> The partial correlations do open up the modeland provide more insights into asset dynamics
    37. 37. 37www.riskcommons.orgwww.riskcommons.org | www.fna.fi Alan Laubsch, alaubsch@gmail.com | Kimmo Soramaki , kimmo@fna.fi 37www.riskcommons.org | www.fna.fi Alan Laubsch, alaubsch@gmail.com | Kimmo Soramaki , kimmo@fna.fi 37Agenda1. A Framework for Adaptive Stress Testing2. Signal or Noise?3. Introducing StressGrades™4. Network approaches to Stress Testing5. Summary and Conclusions
    38. 38. 38www.riskcommons.orgwww.riskcommons.org | www.fna.fi Alan Laubsch, alaubsch@gmail.com | Kimmo Soramaki , kimmo@fna.fi 38www.riskcommons.org | www.fna.fi Alan Laubsch, alaubsch@gmail.com | Kimmo Soramaki , kimmo@fna.fi 38Summary: sense and respond to emerging risks• Use algorithms and visualization techniques to detect signals amidstnoise (e.g., super-exponential rates of change)• Prioritize attention to relevant macro fault lines and specificportfolio vulnerabilitiesAnticipateMost of the focus at most companies is on what’s directly ahead. The leaders lack“peripheral vision.” This can leave your company vulnerable to rivals who detectand act on ambiguous signals.- 6 Habits of True Strategic Thinkers, Paul Schoemaker, Mar 20, 2012
    39. 39. 39www.riskcommons.orgwww.riskcommons.org | www.fna.fi Alan Laubsch, alaubsch@gmail.com | Kimmo Soramaki , kimmo@fna.fi 39Summary: Architect stress tests to adapt to marketintelligence• As early warning signals are observed:1. Focus on affected systemic fault lines (and related nodes)2. Assign higher probability of stress3. Apply more severe stress scenarios• Proactive response is essential1. War game scenarios to better understand potential impacts andconsequences over time, and practice playing out various permutationsof scenarios across the enterprise2. Take advantage of calm periods to reduce concentration risks, increasecapital and liquidity buffers. Get prepared to weather more severestorms ahead.
    40. 40. 40www.riskcommons.orgwww.riskcommons.org | www.fna.fi Alan Laubsch, alaubsch@gmail.com | Kimmo Soramaki , kimmo@fna.fi 40www.riskcommons.org | www.fna.fi Alan Laubsch, alaubsch@gmail.com | Kimmo Soramaki , kimmo@fna.fi 40Conclusions• Adaptive stress testing practices• Experiment: explore emerging vulnerabilities and seek touncover risk concentrations• Learn: intelligent feedback loops: market signals andsubjective perspectives (scenarios)• Practice: play through various scenario permutations• Early detection and adaptation is crucial for systemic risks• Harness market intelligence to prioritize attention“The future is already here — its just not very evenly distributed.”William Gibson

    ×