VAR LECTURE - VALUE AT RISK

"The same false assumptions that underestimates stock-market risk, mis-price options, builds bad portfolios, and generally misconstrue the financial world are also built into the standard risk software used by many of the world's banks. The method is called Value at Risk." Benoit Mandelbrot

"Organised common sense - very basic knowledge - is an enormously powerful tool. There are huge dangers with computers. People calculate too much and think too little." Charlie Munger

“I am extremely sceptical of automated, algorithmic, Value at Risk, and other business school sanctioned approaches to risk management. None of these approaches saved Lehman, Bear Stearns, Fannie, Freddie, AIG, WaMu, Wachovia or any of the other institutions that used these and other ostensibly more sophisticated risk management strategies.” Bill Ackman

“We will never put our faith in computer “risk models”; instead, we apply sound judgement and common sense to our assessment of risk.” Seth Klarman

“The use of the value-at-risk models which made no sense from my judgment because they cut off the tail.” David Einhorn

“If the financial crisis taught nothing else, it showed how elegant financial models that calculate risk to decimal point precision act like a sedative towards critical thinking and even common sense.” Allan Mecham

“VAR tells you how volatile your current portfolio was in the past. That is all. VAR is entirely backward looking. You have to recognise that the future will be different. VAR doesn’t blow up portfolios, people do.” Colm O’Shea

“We have never had and would never use any form of quantitative risk control because all quantitative risk control models use historical volatility. It is like driving by looking in the rear-view mirror.” Martin Taylor

“A 99% Value-at-Risk calculation does not evaluate what happens in the last one percent… This is like an airbag that works all the time, except when you have a car accident.” David Einhorn

"Once a bank adopts VAR the managers manage to it because they .. have no choice but to keep the VAR numbers shiny for their clients as well as for their own internal risk managers. This is extremely dangerous, for VAR is a complex formal system. The essence of all formal systems is to carve away some information so as to cast what remains into bold relief, to make choices about what is important and what is not. And the truth about all models that exceed a certain point of complexity is that things left out eventually become invisible even if they never become insignificant." Andrew Redleaf

“The rocket scientists (at financial institutions) managed to create a missile that landed on themselves.” Anthony Sanders, Professor of Finance, Arizona State University

"I find it preposterous that a single number reflecting past price fluctuations could be thought to completely describe the risk in a security." Seth Klarman

"Things like Gaussian curves and Value at Risk (VAR) were some of the dumbest ideas ever put forward." Charlie Munger

“The only problem is that [Gaussian] curve is not applicable to behavior in markets, and people find that out periodically.” Warren Buffett

"The idea that you have a bell-shaped curve is false. You have outlying phenomena that you can't anticipate on the basis of previous experience." George Soros

"The precision that goes into saying that this is a two standard deviation event or this a three standard deviation event, and therefore we can afford to take this much risk and all that, it's totally crazy." Warren Buffett

"The key hazard of quantative risk management is the illusion of control the models and their results impart to us. No model has an R-sqare of 1.00. Even if you have a so-called statistically significant outcome, which is 95% certain - 95% still leaves 5% you know nothing about. The devil is in the residuals, as all of us have discovered to our sorrow." Howard Marks

"Most market participants reconcile the existential Black Swan risk by using the coping mechanism we call “disaster myopia”: If an event is too uncertain and unpredictable as to its timing or magnitude—if it can never be part of the quantitative model you use to evaluate risk—you simply put it out of your mind to avoid cognitive dissonance. That’s also what ostriches do." Frank Martin

"Those who trust the models do so at their own peril. They will be hurt as the models are wrong. The world may be different than those used in the model's history." Paul Singer 

"If everybody that operated in financial markets had never had any concept of standard errors and so on, they would be a lot better off." Warren Buffett

"Using backward-looking models - a common coping tool - computing the probability of something disastrous that has never happened before usually produces a number close to zero. With a probability that slim, most become disaster myopic." Frank Martin

"Over time, markets will do extraordinary, even bizarre things. A single, big mistake could wipe out a long string of success. We therefore need someone genetically programmed to recognize and avoid serious risks, including those never before encountered. Certain perils that lurk in investment strategies cannot be spotted by use of models commonly employed today by financial institutions." Warren Buffett

"The real danger of VAR was [is] the same as that of all statistical systems that become substitutes for human judgement. By encouraging bankers to surrender their judgement, VAR radically increased [increases] the possibility of disaster." Andrew Redleaf

"Just as Justice Stewart found it impossible to formulate a test for obscenity but nevertheless asserted, "I know it when I see it," so also can investors - in an inexact but useful way - "see" the risks inherent in certain investments without reference to complex equations or price histories." Warren Buffett

"Do not trust financial market risk models. Reality is always too complex to be accurately modelled. Attention to risk must be a 24/7/365 obsession, with people - not computers - assessing and reassessing the risk environment in real time." Seth Klarman

"A good deal of common sense has to be utilized in conjunction with risk model output. We generally assume risk models and stress tests are significantly underappreciating true risk." John Burbank

"One of the things I have observed during my career that I think is interesting is that the basics of investing do not change only the terminology or lexicon seems to. VAR, sharpe ratio, Market Neutral - whatever that means, tail risk, black swans, Sortino ratio. To me, there just seems to be some perverse human characteristic that likes to make easy things difficult." John Phelan

"I discovered along the way that the economists and social scientists were almost always applying the wrong maths to the problems, what became later the theme of the Black Swan. Their statistical tools were not just wrong, they were outrageously wrong - they still are. Their methods underestimated "tail events", those rare but consequential jumps. They were too arrogant to accept it. This discovery allowed me to achieve financial independence in my twenties, after the crash of 1987." Nassim Nicholas Taleb

“The thing that really destroys people are what the academics would call six-sigma, or five-sigma, or seven-sigma events, which are things that are never supposed to happen, basically. Sigma is a method of describing the probabilities that they will happen in any given period — the number of sigmas. Financial markets don’t lend themselves well to modeling based on that. You know, they do most of the time, until it doesn’t work. And when it doesn’t work, you know, chaos reigns. And there are more six-sigma events that happen in financial markets — or theoretical six- sigma events — than any study of probability curves would ever come up with.” Warren Buffett

"The modern financial system seems almost designed for systemic trouble because it continues to rely on VaR (value at risk), carrying the antiquated intellectual baggage of efficient markets and normal distributions into the world of risk management." Frank Martin

"Life in financial markets has got no relation to standard deviation." Warren Buffett

"We have seen many instances when probabilistic models turned out not to have made sufficient allowance for an ‘improbable disaster.’" Howard Marks

"The defect of VaR alone is that it doesn't full account for the worst 5% of expected cases. But these extreme events are where ruin is found. It's also true that these extreme changes in securities prices may be much greater than you would expect from the Gaussian or normal statistics commonly used." Ed Thorp

"Investors should be skeptical of history-based models. Constructed by a nerdy-sounding priesthood using esoteric terms such as beta, gamma, sigma and the like, these models tend to look impressive. Too often, though, investors forget to examine the assumptions behind the symbols. Our advice: Beware of geeks bearing formulas." Warren Buffett

"A lot of the modern risk-management techniques created a totally false illusion of safety. The idea that by quantifying risk using a tool like VaR [Value at Risk] that you can therefore control it is one of the slightly more ridiculous things to have come along in years." James Montier

“Traditional parametric VAR [Value at risk] stress tests assume that returns and volatility follow a normal distribution. They are useful in normalised environments. But in stress tested environments? Not so much. They need to be supplemented with empirical stress tests which use historical data as far bask as it goes.” Paul Marshall

“I’ve never used VAR. We had to use it at Soros to get banking lines from great companies like GS. But basically, [I’m] very unsophisticated - I watched my PnL every day and if it started acting in a strange manner in what I would expect out of a matrix, my antennae would go up. I’ve always used my PnL because I’ve found all those risks models are great until complete chaos happens and then all the correlations break down and they can suck you into a false sense of security. If you’re watching you’re PnL with your antennae up, and you’ve been doing it for 30 or 40 years - I’ve found it a much better warning system than some of these mathematical models which are useful, they’re just not useful when you need them.” Stanley Druckenmiller

“I really don't [think that there's value in using quantitative tools to measure and manage risk]. We don't do it at all. First of all, what is risk? It's the probability of a negative event in the future. What do we know about that? What does the past tell us about that? The past has relevance, but it's not absolute…. I don't think risk can be measured. I don't think the past is absolutely applicable. In fact, the big money is lost at the juncture when the past stops being applicable, which happens eventually. And I don't think that the possibilities in the future can be synthesized into inputs that are sufficient to let a computer decide what's risky or not.” Howard Marks