No Crystal Ball For Predicting Risk
"Taking the angst out of taking a gamble" (Finance, July 21) was timely. But the value-at-risk (VAR) technology is less reliable than some academic proponents would like us to believe. The actual probability of a specified loss can vary by as much as one order of magnitude (10 times) from the probability estimated by value-at-risk methods.
The article suggests that a bank might notify the Federal Reserve that "we estimate that there is a 5% chance our portfolio will decline $20 million or more in the next two days." In fact, given the properties of the value-at-risk historical estimation method, that reported 5% may be as little as 0% or as large as 20%. The seeming precision of VAR estimates is deceptive. Having worked both sides of the street, I find that communication between financial rocket scientists and working risk managers is sometimes dangerously incomplete. The unreliability of value-at-risk estimates has not to my knowledge been adequately communicated to regulators or risk managers.
Richard B. Hoppe