Taleb Testimony
Taleb Testimony
Breakdown
Investments L.P.
INTRODUCTION
Mr. Chairman, Ranking Member, Members of the Committee, thank you for giving me the
opportunity to testify on the risk measurement methods used by banks, particularly those
concerned with blowup risk, estimates of probabilities of losses from extreme events ("tail risks"),
What is the VaR? It is simply a model that is supposed to project the expected extreme loss in an
institution’s portfolio that can occur over a specific time frame at a specified level of confidence.
Take an example. A standard daily VaR of $1 million at a 1% probability tells you that you have
less than a 1% chance of losing $1 million or more on a given day2. There are many modifications
around VaR, "conditional VaR" 3 , so my discussion concerns all quantitative (and probabilistic)
1 The author thanks Daniel Kahneman, Pablo Triana, and Eric Weinstein for helpful discussions.
2 Although such definition of VaR is often presented as a "maximum" loss, it is technically not so in an open-ended
exposure: since, conditional on losing more than $1 million, you may lose a lot more, say $5 million.
3 Data shows that methods meant to improve the standard VaR, like "expected shortfall" or "conditional VaR" are
equally defective with economic variables --past losses do not predict future losses. Stress testing is also suspicious
because of the subjective nature of "reasonable stress" number --we tend to underestimate the magnitude of outliers.
methods concerned with losses associated with rare events. Simply, there are limitations to our
Thirteen years ago, I warned that "VaR encourages misdirected people to take risks with
shareholders’, and ultimately taxpayers’ money." I have since been begging for the suspension of
these measurements of tail risks. But this came a bit late. For the banking system has lost so far,
according to the International Monetary Fund, in excess of 4 trillion dollars directly as a result of
faulty risk management. Most of the losses were in the U.S. and will be directly borne by
taxpayers. These losses do not include the other costs of the economic crisis.
Data shows that banks routinely lose everything earned in their past history in single blowups --
this happened in 1982, 1991, and, of course now. Every time society bails them out -- while bank
risk-takers retain their past bonuses and start the game afresh. This is an aberrant case of
1. These problems have been obvious all along. My first encounter with the VaR was as a
derivatives trader in the early 1990s when it was first introduced. I saw its underestimation of the
risks of a portfolio by a factor of 100 --you set up your book to lose no more than $100,000 and
you take a $10,000,000 hit. Worse, there was no way to get a handle on how much its
Using VaR after the crash of 1987 proved strangely gullible. But the fact that its use was not
suspended after the many subsequent major events, such as the Long-Term Capital Management
"Jumps" are not predictable from past jumps. See Taleb, N. N. (in press) "Errors, robustness, and the fourth quadrant",
International Journal of Forecasting (2009).
2
blowup in 1998, requires some explanation. Furthermore, regulators started promoting VaR (Basel
- Alas, we cannot "measure" the risk of future rare events like we measure the temperature.
By robustness, I mean that the measure does not change much if you change the model,
technique, or theory. Indeed risk estimation has nothing to do with the notion of measure. And
the rarer the event, the harder it is to compute its probability --yet the rarer the event, the larger
the consequences5.
- Furthermore, the type of randomness we have with economic variables does not have a well-
tractable, well-known structure, and can deliver vastly large events --and we are unable to get a
handle on "how large". Conventional statistics, derived on a different class of variables, fail us
here6 7 8
.
3. VaR encourages "low volatility, high blowup" risk taking which can be gamed by the
4 My recollection is that the VaR was not initially taken seriously by traders and managers. It took a long time for the
practice to spread --and it was only after regulators got involved that it became widespread.
5 See Taleb N.N. and Pilpel, A. (2007) Epistemology and Risk Management, Risk and Regulation, 13
6 We are in the worst type of complex system characterized by high interdependence, low predictability, and
vulnerability to extreme events. See N.N. Taleb, The Black Swan, Random House, 2007.
7 There are other problems. 1) VaR does not replicate out of sample -- the past almost never predicts subsequent
blowups. (see data in the Fourth Quadrant). 2) A decrease in VaR does not mean decrease in risks; often quite the
opposite holds, which allows the measure to be gamed.
8 The roots of VaR come from modern financial theory (Markowitz, Sharpe, Miller, Merton, Scholes) which, in spite
of its patent lack of scientific validity, continues to be taught in business schools. See Taleb, N.N., (2000), The Black
Swan: The Impact of the Highly Improbable, Random House.
3
Figure 1- A typical "blow-up" strategy with hidden risks: appearance of low volatility, with a
high risk of blowup. The trader makes 11 bonuses, with no subsequent "clawback" as losses
are borne by shareholders, then taxpayers. This is the profile for banks (losses in 1982,1991,
and 2008) and many hedge funds. VaR encourages such types of risk taking.
I have shown that operators like to engage in a "blow-up" strategy, (switching risks from visible to
hidden), which consists in producing steady profits for a long time, collecting bonuses, then losing
everything in a single blowup.9 Such trades pay extremely well for the trader --but not for society.
For instance, a member of Citicorp's executive committee (and former government official)
collected $120 million of bonuses over the years of hidden risks before the blowup; regular
Blowup risks kept increasing over the past few years, while the appearance of stability has
increased.10
Many people favor the adjunct application of VaR on grounds that it is "not harmful", using
arguments like "we are aware of its defects". VaR has side effects of increasing risk-taking, even
9 Taleb, N. N. (2004) “Bleed or Blowup: What Does Empirical Psychology Tell Us About the Preference For Negative
4
by those who know that it is not reliable. We have ample evidence of so called "anchoring"11 in
overconfidence.
5. VaR-style quantitative risk measurement is the engine behind leverage, the main
Leverage12 is a direct result of underestimation of the risks of extreme events --and the illusion
that these risks are measurable. Someone more careful (or realistic) would issue equity.
April 28, 2004 was a very sad day, when the SEC, at the instigation of the investment banks,
initiated the abandonment of hard (i.e. robust) risk measures like leverage, in favor of more
Regulators should understand that financial markets are a complex system and work on increasing
the robustness in it, by preventing "too big to fail" situations, favoring diversity in risk taking,
allowing entities to absorb large shocks, and reducing the effect of model error (see "Ten Points
for a Black Swan Robust Society", in Appendix II). This implies reliance on "hard", non-probabilistic
11 Numerous experiments provide evidence that professionals are significantly influenced by numbers that they know
to be irrelevant to their decision, like writing down the last 4 digits of one's social security number before making a
numerical estimate of potential market moves. German judges rolling dice before sentencing showed an increase of 50%
in the length of the sentence when the dice show a high number, without being conscious of it. See Birte Englich and
Thomas Mussweiler, “Sentencing under Uncertainty: Anchoring Effects in the Courtroom,” Journal of Applied Social
Psychology, vol. 31, no. 7 92001), pp. 1535-1551; Birte Englich, Thomas Mussweiler, and Fritz Strack, “Playing Dice with
Criminal Sentences: the Influence of Irrelevant Anchors on Experts’ Judicial Decision Making,” Personality and Social
Psychology Bulletin, vol. 32, no 2 (Feb. 2006), pp. 188-200.
12 There is a large difference between equity and credit bubbles. Equity bubbles are benign. We went through an
5
measures rather than more error-prone ones. For instance "leverage" is a robust measures (like
the temperature, it does not change with your model), while VaR is not.
Furthermore, we need to examine the toxicity of models; financial regulators should have the
same test as the Food and Drug Administration does. The promoter of the probability model must
be able to show that no one will be harmed even if the event is rare. Alas, the history of medicine
shows translational gaps, the lag between the discovery of harm and suspension of harmful
pre-modern medicine14. But we cannot afford to wait 200 years to find out that the medicine is far
1996-1997
estimate, namely the risk of rare events. It gives people a misleading sense of precision.
13 "When William Harvey demonstrated the mechanism of blood circulation in the 1620s, humoral theory and its
related practices should have disappeared, because the anatomy and physiology on which it relied was incompatible with
this picture of the organism. In fact, people continued to refer to spirits and humors, and doctors continued to prescribe
phlebotomies, enemas, and cataplasms, for centuries more --even when it was established in the mid-1800, most notably
by Louis Pasteur, that germs were the cause of disease." Noga Arikha "Just Life in a Nutshell: Humours as common
sense", in The Philosophical Forum Quarterly, XXXIX, 3
14 Most of the use of probabilistic methods lacking both mathematical and empirical justification can be attributed to
the prestige given to modern finance by the various Nobel memorial prizes in economics. See P. Triana, 2009, Lecturing
Birds on Flying: Can Mathematical Theories Destroy the Markets?, J. Wiley.
6
VaR encourages misdirected people to take risks with shareholders,’ and ultimately taxpayers’
2003
Fannie Mae's models (for calibrating to the risks of rare events) are pseudoscience. (New York
"What happened to LTCM will look like a picnic compared to what should happen to you". (Lecture,
2007
Fannie Mae, when I look at its risks, seems to be sitting on a barrel of dynamite, vulnerable to the
slightest hiccup. But not to worry: their large staff of scientists deems these events "unlikely". (The
Black Swan)
Banks are now more vulnerable to the Black Swan than ever before with “scientists” among their
staff taking care of exposures. The giant firm J. P. Morgan put the entire world at risk by
introducing in the nineties RiskMetrics, a phony method aiming at managing people’s risks. A
related method called “Value-at-Risk,” which relies on the quantitative measurement of risk, has
7
APPENDIX II: TEN PRINCIPLES FOR A BLACK SWAN ROBUST WORLD (FINANCIAL TIMES, APRIL
8, 2009)
1. What is fragile should break early while it is still small. Nothing should ever become too
big to fail. Evolution in economic life helps those with the maximum amount of hidden risks - and
out should be nationalised; whatever does not need a bail-out should be free, small and risk-
bearing. We have managed to combine the worst of capitalism and socialism. In France in the
1980s, the socialists took over the banks. In the US in the 2000s, the banks took over the
3. People who were driving a school bus blindfolded (and crashed it) should never be
given a new bus. The economics establishment (universities, regulators, central bankers,
government officials, various organizations staffed with economists) lost its legitimacy with the
failure of the system. It is irresponsible and foolish to put our trust in the ability of such experts to
get us out of this mess. Instead, find the smart people whose hands are clean.
4. Do not let someone making an "incentive" bonus manage a nuclear plant - or your
financial risks. Odds are he would cut every corner on safety to show "profits" while claiming to
be "conservative". Bonuses do not accommodate the hidden risks of blow-ups. It is the asymmetry
of the bonus system that got us here. No incentives without disincentives: capitalism is about
systems survive thanks to slack and redundancy; adding debt produces wild and dangerous
gyrations and leaves no room for error. Capitalism cannot avoid fads and bubbles: equity bubbles
6. Do not give children sticks of dynamite, even if they come with a warning. Complex
derivatives need to be banned because nobody understands them and few are rational enough to
know it. Citizens must be protected from themselves, from bankers selling them "hedging"
7. Only Ponzi schemes should depend on confidence. Governments should never need
cannot stop the rumours. Simply, we need to be in a position to shrug off rumours, be robust in
8. Do not give an addict more drugs if he has withdrawal pains. Using leverage to cure the
problems of too much leverage is not homeopathy, it is denial. The debt crisis is not a temporary
9. Citizens should not depend on financial assets or fallible "expert" advice for their
retirement. Economic life should be definancialised. We should learn not to use markets as
storehouses of value: they do not harbour the certainties that normal citizens require. Citizens
should experience anxiety about their own businesses (which they control), not their investments
10. Make an omelet with the broken eggs. Finally, this crisis cannot be fixed with makeshift
repairs, no more than a boat with a rotten hull can be fixed with ad hoc patches. We need to
rebuild the hull with new (stronger) materials; we will have to remake the system before it does so
9
itself. Let us move voluntarily into Capitalism 2.0 by helping what needs to be broken break on its
own, converting debt into equity, marginalising the economics and business school establishments,
shutting down the "Nobel" in economics, banning leveraged buy-outs, putting bankers where they
belong, clawing back the bonuses of those who got us here, and teaching people to navigate a
Then we will see an economic life closer to our biological environment: smaller companies, richer
ecology, no leverage. A world in which entrepreneurs, not bankers, take the risks, and companies
are born and die every day without making the news.
10