VaR and the Meltdown

Posted by Peter Orr on Jan 05, 2009

RISK!If you didn't catch it, the NYT magazine this weekend had a cover story on risk which posed the question, was management or specifically risk management more responsible for the current financial mess in which we now sit? Not unexpectedly, Mr. Black Swan himself got a good dose of coverage railing against the utter folly of VaR and seemingly anyone who attempts to quantify anything about risk in finance. The other corner is represented by the leadership at RiskMetrics, Sunguard, etc weighing in with the "calculating risk has benefits" position, given it (VaR) provides relevant and useful information the majority of the time. Taleb's point is the "majority of the time" doesn't matter much after insolvency.

Given the topic, the article provides predictable variations of common platitudes: "Guns (quantifying risk) don't kill people, people (dumb risk management) kill people," and "Those who ignore the lessons of history are bound to repeat it (particularly if you only use 2 years of data as a basis for your VaR calc)."

I couldn't help but notice how much the article echoes the debate about the degree of risk versus uncertainty present in financial management, and what to do about it. In fact, one way to look at the position of people like Taleb/Mandelbrot is that the uncertainty about which Dr. Knight wrote in his 1916 dissertation is really the driving force behind socio-economic variable movement and as such, if you're going to do any modeling, fractals are your best only choice. They argue fractals and power laws are the only things that give you a prayer of appreciating the potential magnitude out in the tails of the uncertainty.

I think the article properly highlighted one VaR shortcoming that has been reasonably well-known but under-addressed, "[VaR] failed to distinguish between leverage that came from long-term, fixed-rate debt…and loans that can be called at any time and…blow you up in 2 minutes." Coincidentally this is highly germane to my disagreement with Dr. Black, and perhaps even the mindset behind a lot of quant training.

I wish there was a bit on behavioral finance. The author hints at that quoting a risk manager, "It has to do with the human condition. People like to have one number they can believe in." It might've been nice if he touched on concepts like availability bias, overconfidence, and herding as additional contributors to the problem. What ramifications this all has for future policy is anyone's guess. In the end, we are still doing this democratic/capitalist experiment and it doesn't appear we're in any danger of getting it *right* anytime soon.

Of course, I'm sure some risk managers out there reveled in one of the author's concluding and resigned ruminations, "Maybe it would have been better if the people in charge had a better understanding of risk." I suppose that would've been nice.