Just one number? Then you're wrong! Efforts to find one perfect number to measure risk are just plain silly.

In Douglas Adams' witty science fiction series, A Hitchhiker's Guide to the Galaxy, a gigantic supercomputer named Deep Thought is asked to find "The Answer to the Great Question of Life, the Universe and Everything." After laboring for seven-and-a-half million years, it finally reports its answer to an eager audience:

"'Forty-two,' said Deep Thought, with infinite majesty and calm.

'Forty-two!' yelled Loonquawl. 'Is that all you've got to show for seven-and-a-half million years' work?'

'I checked it very thoroughly,' said the computer, 'and that quite definitely is the answer. I think the problem, to be quite honest with you, is that you've never actually known what the question is.'"

Most readers find it hilarious to imagine that the answer to such an enormous and consequential question could be just one single number! But firms that adopt Enterprise Risk Management--ERM for short--often engage in a similar quest to answer a big question--what is our firmwide risk?--with a single number.

It is time to correct the pervasive but mistaken assumption that there is a single perfect number for measuring risk. There isn't.

Early attempts at ERM measured firmwide risk by a single number called Value at Risk, or VaR. This number answers the question, "How bad could our firm's losses be during a relatively bad time period?" The time period varies with the nature of the institution: for insurers it is typically a year. The answer was a percentile: 95% of the time we will lose less than, say, $X million. This amount, then, was the 95% Value at Risk. Since the percentile is arbitrary, any percentile could be used.

Some brilliant academics subsequently identified a technical characteristic of VaR that made it problematic--"incoherent" in their terminology. Their examples of this problem were highly unusual and would virtually never be encountered in the real world. Nonetheless, many ERM practitioners, wary of having their risk measures described as "incoherent," abandoned VaK and adopted alternative risk measures that were "coherent," such as Tail Value at Risk (TVaR).

Overlooked was the fundamental but mistaken assumption that there is some single risk measure that is perfect. In fact, firmwide risk is accurately described by a distribution of potential outcomes that cannot be adequately summarized by a single number. So it becomes important to use multiple numbers to describe a firm's overall risk. And these numbers need to be made meaningful--the use of percentiles is indeed rather arbitrary.

One useful approach is to identify undesirable consequences of increasingly severe losses, in order to calculate the probability of these consequences and track them over time. How large a loss would trigger, say, a one-notch downgrade? How about two notches? What are the probabilities of such losses? Are these probabilities acceptable, and are they larger or smaller than last year? These are the questions anal numbers that make risk measurement understandable and manageable to senior managers and directors.

William H. Panning, a Best's Review columnist, is executive vice president at Willis Re Inc. He can be reached at [e-mail].