Changing the answers: the gap between the risk we are taking and the risk we can afford to take may have grown. (Property/Casualty).

The fundamental questions of risk management are timeless and universal. The answers, though, have changed in ways that have significant implications.

I was reminded of the first point during dinner with close friends and a guest they had invited, a bright, successful businesswoman. The guest wasn't afraid to make bold statements: "Life, like business, is nothing more than the assessment of probabilities." She then explained, in convincing detail, how the success of her family's business depended on identifying and controlling the various factors that influenced their probability of success. "At every stage in our production process, we know exactly how much risk we are taking. If we can control it, we do, and if we can't, we try to hedge it."

Like many corporate executives, she and her family paid close attention to the three fundamental questions of risk management: How much risk are we taking? How much risk can we afford to take? What should we do about the difference?

Although these questions are timeless, what we've learned during the past two decades forces us to revise our answers to them. First, we've learned that rare events aren't as rare as we thought. The so-called normal distribution, the subject of countless lectures for MBA candidates, has turned out to be a misleading guide to risk. For example, until the mid-1980s, it was widely assumed that distribution of investment returns was roughly normal. The death blow to this view came Oct. 19,1987, when market indices fell by more than 20% in a day--an event previously considered so improbable that it might occur once in billions--yes, billions--of years.

But the rapid market recovery quickly overcame statistical anxiety, until the collapse of Long Term Capital Management in 1998 again challenged conventional wisdom. Since then, numerous studies have demonstrated that market returns are not normal, but instead exhibit leptokurtosis (try this term at a cocktail party!). This simply means that the market typically oscillates between days when returns are near zero and days when returns are quite large, positive or negative.

Potential for extreme loss is by no means confined to the stock market, as numerous hurricanes, earthquakes and the events of Sept. 11 have so forcibly reminded us. With respect to each type of event, history and science are imperfect guides, so that forecasting their frequency or their severity is itself fraught with uncertainty. As we cannot safely assume that a fair coin will produce heads exactly half the time, we cannot safely assume that an event that has occurred only once in the past 100 years will occur in the future only once in 100 years. There is a substantial likelihood that what we consider a 100-year event may, in fact, occur more frequently than what we have observed. At harvest time, when rain is unwelcome, I once heard a pessimistic farmer ask another, "What's the forecast?" "Weatherman says about a 10% chance of rain," was the reply. "Well," said the first farmer, "this time of year, that's about all that it takes."

A second lesson we've learned is that diversification is often absent when you need it most. In asset management and insurance, diversification is highly valued. Both aim to combine risks that are imperfectly correlated so as to maximize the return achieved for a given overall amount of risk. Despite this focus on diversification, the fact is that under the most extreme conditions diversification disappears. In the investment world, for example, circumstances that produce truly dismal returns for some assets create dismal returns for most other assets as well. Although diversification does occur when returns are modest, in the most extreme scenarios experienced in the past two decades, the returns exhibited by supposedly diverse asset classes have been almost perfectly correlated. In the property/casualty world, extreme events like Sept. 11 create a similar interdependence across classes of business--property, workers' compensation, business interruption--that normally would be independent of one another.

Taken together, these two points have enormous implications. If extreme events are more common than we thought, and if extreme events destroy the diversification of risk we thought would protect us, then our firms are far more vulnerable to extreme events than we thought. How much risk are we taking? More than we thought!

How much risk can we afford to take? If the answer hasn't changed much, then the gap between the risk we are taking and the risk we can afford to take has increased. And to close this gap, firms should be increasing, rather than reducing, their purchase of insurance, reinsurance, and other risk-reducing instruments.

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