Evaluating Risk In Current Markets

Value at Risk: Evaluating Risk in the Current Market

Crisis

Value at Risk: Evaluating Risk in the Current Market

CrisisBarry Schachter

for presentation November 12, 2008

Risk Management CrisisRisk Management Crisis

Tower Group, NEEDHAM, Mass., Dec. 3, 2007 /PRNewswire/ -- The recent crisis in the subprime lending market exposed structural deficiencies in the financial system globally, particularly around the area of risk management…poor integration of risk management across an organization masks the interdependencies of risk and financial indicators -- potentially exposing financial institutions to severe losses.

2

Valuation FailureValuation Failure

IMF Survey Magazine, Jan. 15, 2008As trading volumes [in structured securities] declined, many market participants were forced to use pricing models that relied on historical data. However, the recent performance of some of these subprime loans has been much worse than the record would have suggested. This has caused valuation models to break down.

3

Risk Measurement or Mysticism?


Financial Times, “The pseudo-science hurting markets,” by Nassim Nicholas Taleb, Oct. 23, 2007MPT [Modern Portfolio Theory] produces measures such as “sigmas”, “betas”, “Sharpe ratios”, “correlation”, “value at risk”, “optimal portfolios” and “capital asset pricing model” that are incompatible with the possibility of those consequential rare events I call “black swans”…

4



…one expert cheerfully declared that "no one in the industry really has much faith in VAR any more". Or as another official at a big bank noted: "VAR has a role as a tool but it is just one tool. There is no point in using it unless you also engage your brain.”…VAR has become so central to the financial world this decade - and so closely watched by shareholders and regulators - that it is difficult to ignore.Financial Times, “Volatility wrecks financial world's value at risk models,” Oct. 12, 2007

5

Financial Risk Measurement 101 (or so)Value at Risk (the love/hate relationship)Who’s sorry now?Stress Testing (promises unfulfilled)To the caves (or worse) for risk managers?

6

Overview and SummaryOverview and Summary

Origins and rationale – to combine diverse risks into a single number for management information

Current status – employed in financial and non-financial firms; enshrined by financial regulators and required for measuring risk-based capital.

VaR looks at “tail risk”

7

Risk Measurement 101Value at Risk (“VaR”)

Risk Measurement 101Value at Risk (“VaR”)

Parametric – Variance/Covariance – Normal (everyone’s favorite straw man)

Monte CarloHistorical Simulation …Most banks' models performed poorly which is not surprising given the popular

use of historical simulation. While historical simulation provides stable VaR numbers, it has weak forecasting power and is entirely inappropriate during regime shifts… RiskMetrics Group, “Did VaR Forecast the U.S. Subprime Crisis?” Aug. 22, 2008

8

Risk Measurement 101 VaR VaRieties

Risk Measurement 101 VaR VaRieties

Origins and Rationale – address VaR deficiencies in “plausible but abnormal” market environments

Current Status – about the same as VaR (Stress star is waxing while VaR’s star is, perhaps, waning)

Stress Scenarios are economic reconstructions or constructions

9

Risk Measurement 101 – Scenario Stress TestingRisk Measurement 101 – Scenario Stress Testing

A Pause to Reflect Before Proceeding

A Pause to Reflect Before Proceeding

10

This talk has a built-in biasRemember, “Guns don’t kill people…”Remember, too, hindsight is 20-20

The Love/Hate Relationship with VaR –

What’s Wrong With VaR?



11

If it isn’t in the data, it isn’t in the VaR“…a few years of past calm is taken to indicate many

more years of future market tranquility. In effect, when the real risk of a market collapse is high, the ratio of VAR to the value of the underlying assets is low. Finance directors look at the low ratio of VAR to assets and say: “We’re not using our balance sheet efficiently. Leverage it up.”BreakingViews.com, “Learning lessons from subprime crisis,” Dec. 10, 2007.





12

Assumptions relating the past to the futureHistorical simulation is the most agnostic method but

pays a price for this (maybe especially now)How bad is bad – VaR doesn’t know as a percentile

measureVaR can be “gamed” – e.g., John Rusnak (Allfirst)





13

Many nuanced refinements (filtered HS, GARCH, mixtures of distributions, Extreme Value Theory)

The irresolvable is question is dependence on model quality at two levels

pricing modelsforecast model

The Love/Hate Relationship

What Have We Learned Recently?



14

Credit ratings (models) provided flawed inputsThe behavior of CDO tranche ratings can differ markedly

from that of corporate ratings. In addition, tranching is found to have an important impact on the probability of large losses. Investors who narrowly focus on ratings and draw direct parallels with corporate exposures can seriously misjudge the value-at-risk of CDOs.BIS Quarterly Review, “Credit fundamentals, ratings and value-at-risk”, March 2008





15

Complement VaR with Stress TestingInstitutions that have performed well had adaptive risk

measurement processes and systems that could rapidly alter underlying assumptions in risk measures to reflect current circumstances; in particular…well designed stress tests allow firms to estimate the economic benefits of diversification and the impact of correlation risk in stressed markets Senior Supervisors Group, “Observations on Risk Management practices during the Recent Market Turbulence”, Mar. 2008





16

Take pricing models with a grain of salt“Emanuel Derman …writes, “To confuse the model with

the world is to embrace a future disaster driven by the belief that humans obey mathematical rules.” “The models were more a tool of enthusiasm than a cause of the crisis.” New York Times, “In modeling risk, the human factor was left out”, Nov. 5, 2008.





17

The GIGO principle still applies…investors failed to weigh data that indicated some

[CDOs] were particularly risky. "It was the risk management software and systems that hadn't been properly adjusted and set up to look for the specific issues that caused these problems," says Silitschanu. "It's not as if these metrics were raising red flags and risk managers were saying don't worry about it.“Wall Street & Technology, “Is the Risk Manager to Blame…”, Feb. 11, 2008.





18

Far Tails don’t look like near tailsCounterparty risk affects risk of loss (we knew it but

didn’t believe it)Liquidity affects risk of loss (we knew it but didn’t know

how to model it (still don’t))

Who’s Sorry Now? – The Blame Game

Who’s Sorry Now? – The Blame Game

19

Michael Alix former Bear CRO, hired by NY Fed on 11/3/08 – “in his new job at the central bank, Alix will help oversee the financial safety and soundness of banks, which are inspected by Federal Reserve examiners. "That's incredible," said James Cox, a Duke University law professor and securities law expert. "This is not reassuring. ... What is there in this person's experience and skill package" that qualifies him for the Fed position? Daily News, Nov. 4, 2008

Who’s Sorry Now? – The Blame Game continuedWho’s Sorry Now? – The Blame Game continued

20

I am about to prove the "nonneutrality of representation". It means that risk measures CAUSE a disproportionate increase in risk taking on the part of the unsuspecting user…Yes, many of you (providers of risk measures) will be sued by the innocent people who lost their savings --like tobacco companies who were sued by the innocent smokers.You will be held accountable. I will make sure it happens. NNT’s Blog, “Many of You Will Be Sued”, Oct. 1, 2008.

Who’s Sorry Now? – The Blame Game continuedWho’s Sorry Now? – The Blame Game continued

21

In the CRO job 99% of the days there is nothing going wrong. The only test you get of how well you are doing – short of pouring out risk reports and looking ponderous and prudent in meetings – is what happens to the firm during times of market crisis…if the firm gets blown away, that suggests you did not do a very good job…The job of the risk manager …should involve more than simply running a value at risk calculation on the computer. Rick Bookstaber’s Blog, “Risk Management and Shake-up Time at the Investment Banks”, Oct. 12, 2007

Will Scenarios Save Us?Will Scenarios Save Us?

22

Not necessarily…Bear Stearns used scenarios extensively

What’s relevant? How long can we use the 1987 equity crash?

What do you do once you have calculated one?How many points define a high-dimensional space?Where do probabilities come in?

Problems of risk measurement are real, but they are not the main story here (e.g., agency problems, transparency impairment and margin impacts).

Regulators’ prescriptive approaches to risk measurement stifle innovation, underestimate complexity, create compliance complaisance

Common sense, good judgment , healthy anxiety are essential complements to risk measurement

Complexity happens, perfect foresight doesn’t exist

23

ConclusionConclusion

Evaluating Risk In Current Markets

Business

risk var

risk models

financial risk measurement

tail risk risk measurement

risk measures

interdependencies of

real risk

risk managers