Quant opinion: The three myths of modern risk management part 1
Financial institutions in the post-crisis era have found themselves in two camps: those who have already spent an exorbitant amount of cash on state-of-the-art risk systems and those that are going to spend on the systems some time soon. And yet, investors and investment risk professionals have little confidence in the outcome of these efforts. The reason is simple: professionals with over a decade of experience have an intuitive and acute awareness of the three myths of modern risk management.
Myth 1: volatility is a measure of risk. There is plenty of evidence to suggest that our treatment of volatility as a measure of risk was in part responsible for the excessive risk taking in the period of 2005-2007. While the markets continued to appreciate in value, investors kept pouring more money into the same limited set of investments. In practice this meant that buying S&P 500 at 1100 was riskier than buying the same index at 1400. In fact, many institutions did not follow blindly their risk systems in that period, preferring instead to rely on common sense.
Myth 2: correlation is a measure of diversification. Most investors by now have acknowledged that low correlations in calm market conditions exaggerate the benefits of diversification. The reason for this lies in the economics of systemic crises.
Consider a supply chain of automotive companies. Steel makers supply steel to car makers and in a normal economic environment the returns of steel stocks and automotive stocks may diverge based on relative negotiating power between the two groups of companies. Lower steel prices would be good news for the car makers. An investor holding both set of stocks would observe low return correlations. In a crisis environment, however, the whole supply chain struggles with lower volumes and all companies perform equally poorly. So just when the investor needs the benefits of diversification, the portfolio is faced with a single risky exposure.
Myth 3: the underlying asset does not matter. By buying a stock, an investor becomes a shareholder of a company. The investor runs the risk of mismanagement, adverse market for the company’s product, cost overruns, failed product introductions and a multitude of other quantifiable risks. The resulting volatility of the stock price is a mere reflection of true risks after the event. The risk of a major draw down, above all, depends on the price paid for an investment in the first place. All of the above, although intuitive from an experienced investor’s point of view, is ignored by the modern risk management practice.
It is quite understandable then that despite all ongoing efforts to beef up the risk management function, experienced professionals are skeptical. All mainstream risk management systems are based on the three myths. They are useful in evaluating investment risks if everything goes well. And if not, then the experienced professionals will yet again be blamed for inadequate monitoring.
We have been successfully applying an alternative risk management framework which does not use volatilities and correlations as an input. Graham Risk is named after Benjamin Graham, who first suggested a measure of risk based on the difference between the price and the intrinsic value of an asset. The full article is available on http://waronvar.com/?p=156. More about Graham Risk is available on www.waronvar.com
In general, people want to simplify things, it make life easier if a few (one will even better) number can summarize the current situation for us. So we have
Point taken – one big issue is oversimplification. But whereas stock
indices are in general a reasonable gauge of what the daily moves of the
stock markets are, volatility indices are not even close in terms of
describing the underlying level of ex ante risk.
• I am fully on this side of the war. The general myth is that modern quantitative finanacial analysis, which in the end is 99% about volatilities and correlations, is an effective tool for risk management because it “looks scientific”. The over-quantification of finance has made everyone blind to the “common sense” risks of asset bubbles and improper asset values (in the G-D sense). After all, the academics of Modern Portfolio Theory “deny” that an investor like Graham and Dodd coul even exist!
Volatility and VaR are stictly “backward looking”, so they work as long as the past is what is going to happen in the future. But during bubbles this is a naive assumption. Note that volatility and VaR are typically low during periods of over-optimism just before a bubble bursts. Totally useless.
The one thing to remember is that even market aggregates can become overvalued, so the G-D value principle applies to the market in aggregate as well as for individual companies, a fact missed by some value investors.
I have written a white paper on the subject here (note also my comments on risk and volatility):
I am fully supportive of the war against VaR and other nonsensical measures.
Simple is good. Simplistic is not.
A good example of volatility not equalling risk is a strategy which involves systematically selling low delta out-of-the-money options. The equity curve will show a smooth move higher month after month, thus volatility is “low”. But clearly this low vol (or high Sharpe) in no way reflects the risk being taken.
I’m with all of you on the limits of modern portfolio and G-D methods. In my experience they work in “normal” markets (1 sigma). In heavy trending markets or bubbles they can cost dearly. If you do not add technicals and sentiment indicators you might be sitting in a one or two legged stool. Remember markets can be irrational longer than you can remain liquid. As examples look at Silver or the Dollar today. Even with this four legged stool you have to be very careful in those heavy trending or bubble markets (lessons of 2008 crash and 2009 rally)
The CEO of Deutsche Bank, Josef Ackermann, once made the astute comment that financial analysts overused linear regression at their own risk. A statistics text proved the same point by using linear regression on Galileo’s data from dropping different weights from different heights, getting precise but meaningless results because gravity takes the form of constant acceleration, not constant velocity. Volatility and correlation methods (more precisely, covariance analysis) are overused in finance and fail because i) market probabilities are time dependent and ii) the relationship between standard deviations and percentiles quickly becomes impossible to measure “down the tail” or in the presence of sufficient leverage. The saying “market risk will hurt you but liquidity risk will kill you” has come to supplant the old Keynes: “markets will remain irrational longer than you can remain solvent.”
“Overused” for sure and I would add “abused” and “over-hyped”. As soon as I entered the financial industry I spent months and months trying to make sense of returns based risk analysis. In the end, I quickly convinced myself that it was highly useless and potentially misleading. Unfortunately, it is also ubiquitous and pervasive because it provides the illusiong of scientific-looking risk management.
That being said, data can be very helpful to understand what is driving the markets at any given time. Personally, I have found it very useful to combine elements of fundamental as well as technical analysis depending on the situation and especially with a great deal of common sense.
Market prices are good indicators of rationally evaluated economic value
Securitised credit has improved allocative efficiency and financial stability
Mathematical analysis can deliver robust quantitative measures of trading risk
Market discipline can be used as an effective tool in constraining harmful risk taking
Financial innovation can be assumed to be beneficial
“Market prices are good indicators of rationally evaluated economic value”
After spending over a decade of hands-on asset valuation experience I tend
to avoid a concept such as rationally evaluated economic value.
“Securitised credit has improved allocative efficiency and financial
Trying to be unbiased, not sure if there is evidence either way
“Mathematical analysis can deliver robust quantitative measures of trading
Agree, with a caveat that mathematics is a fantastic tool, but it requires
correct application and the right set of data.
“Market discipline can be used as an effective tool in constraining harmful
“Financial innovation can be assumed to be beneficial”
Agree. Contrary to the general media bias, I do not hold financial
innovation as such responsible for the crisis. The problem is not that
there is too much innovation. The problem is there is not sufficient
innovation to fully understand and implement the new tools, financial
instruments and systems.
Most basic myth of risk management is that they (financial firm most important owners, their C-level representatives and insiders, installed regulators and politicians/representatives) give a damn about it.—
• I agree with almost all of that. Correlation as a measure of dependence isn’t that bad, but models that suppose correlation is constant under all circumstances are very obvious culprits. For a long time I thought more general modeling of dependence structures (copulas etc.) would be a promising line of research but sooner or later you run into the same problem: you make an assumption about a parameter that governs dependence being constant. Enter massive risk underestimation from stage left.
Linear regression…don’t get me started. I am an econometrician, and a good econometrician spends most of his or her time being extremely paranoid about linear regressions. Sadly, modern software allows everyone to “estimate” linear regression models, without having the slightest understanding of what the results mean or how much they can be trusted. The Galileo example is a classic but there are many other. I vividly remember an article (in the refereed (!) journal “Economics Letters”) which regressed a movie’s box office total on a set of dummies indicating which oscars that movie had won. Surpisingly, winning an oscar for best actress had a NEGATIVE effect on box office total….if you believed the regression result was not fatally flawed because the regressor had failed to do any diagnostic testing.
In the end, linear regression models are little idiot machines that can be very good at extracting the maximum from simple datasets..if the practitioner knows what he or she is doing.
The following are the replies ::-
Myth 1: Volatility is a measure of risk – According to me Volatility is not the measure of Risk infact Volatility is the byproduct of Risk . Whenever we are having Risk in the Financial markets then Volatility originates but the same Financial world have created few models like VAR to compute the impact of the volatility but the real truth is ” Volatility is byproduct of Risk ”
Myth 2 : Correlation is a measure of diversification :- I agree with the fact that low the level of correlation the better the diversification you are having . Correlation is something that matches your returns with an Index but at the same time if you are having 100% correlated than there is low level of Diversification you are having in your portfolio.
Myth 3 : The underlying asset does not matter :- I may agree because Underlying matter and at the same time you have to have backup plan as well…., like in the case of Options underlying matters but we are having backup as well …., My answer can change in case i am having more scenarios.
I hope we can have more discussions on the same topic … ,
FACEBOOK ACCOUNT and TWITTER. Don't worry as I don't post stupid cat videos or what I eat!