|
|||
|
|
|
Blissfully Blinded by Models
Put yourself for a moment in the place of a turkey. For the past 1,000 days, your owner has fed you and treated you well. If you were an analytical bird, you might trend your risk history and determine that it looks very favorable. Projecting past experience to the future, you could reasonably conclude that life is safe and should continue to be so.
No matter how analytical a turkey you might be, you would have no way of knowing that tomorrow is Thanksgiving and that your string of peaceful days would soon end with you on the dinner table. The tragedy that suddenly strikes such a turkey is an unforeseen horror. This type of event, which arises in contradiction to our most carefully crafted models, is known as a Black Swan event. The term has recently been popularized by the premier genius of our time, Nassim Taleb, in his book "The Black Swan." The title refers to a problem first popularized by the British philosopher David Hume. Hume noted that most people, having seen only white swans, concluded that there were no black ones.
One can readily see the wisdom in this simple observation, and yet this is exactly the trap into which we repeatedly fall. Every day without a tremor leads the residents of San Francisco to feel increasingly safe. Every dry day in New Orleans leads to more rebuilding on the flood plain. What's worse, we are analytical birds and know that New Orleans has suffered a major flood on average every 11 years for the past 270 years. Perhaps our memories are not so good. But there could be another explanation. According to Taleb, our attempts to predict future disasters, especially very large ones, are severely hampered by our use of the wrong models. For example, the 1987 market crash, according to the risk models, was a 20 standard deviation event. In other words, such a market decline should only occur once every several billion lifetimes of the universe. The collapse of Long Term Capital Management should have occurred with slightly greater frequency--according to the models. We are good at managing small-impact events. With workers' compensation losses, we have vast amounts of data. Life insurance is similarly predictable. But a single workers' compensation or life insurance claim is not going to threaten the entire U.S. economy. Massive losses should be of paramount importance to risk managers. And yet, these are precisely the types of events that our standard actuarial models fail to predict. The first thing we have to do is admit that we are blind to huge Black Swan events. The evidence is clear on this point. The second thing we have to do is take a long hard look at the models upon which we rely to predict future events. Taleb does a great service to the risk management profession by exposing our greatest flaw. We take solace in the fact that our current methods make huge losses seem rare. This is comforting. It is probably easier for the turkey to get through each day comforting itself with the fact that the past 1,000 days have been uneventful. But that does not change the fact that Thanksgiving is tomorrow. We should learn from Mr. Taleb how not to be turkeys. BEAUMONT VANCE is the risk management columnist for Risk & Insurance®. He manages risk for Sun Microsystems Inc. November 1, 2007 Copyright 2007© LRP Publications |
|
|
|
Entire contents copyright © 2013 Risk and Insurance® All rights reserved.
May not be reproduced in any form without written permission.
|