Canada’s Black Swans

While reading the April 2008 issue of Reinsurance recently, I came across a brief article by the editor, Mark Geoghegan, on a book by Nassim Nicholas Taleb called The Black Swan (published by Random House). In fact, the cover of the magazine featured a picture of a flock of white swans with a pink one in the middle (I guess black swans are too hard to reproduce on the magazine cover).

If you are wondering what a black swan is in this context, I will quote directly from the inside dust cover of the book: “A Black Swan is a highly improbable event with three principal characteristics: It is unpredictable; it carries a massive impact; and, after the fact, we concoct an explanation that makes it appear less random, and more predictable, than it was.”

Taleb’s book focuses on both positive and negative forms of ‘black swans.’ Positive examples of black swans would be successes that were not seen in advance or anticipated — i. e., the Internet, Google, eBay, J. K. Rowling’s success with the Harry Potter book series, cell phones, etc. One negative example of a black swan would be the subprime meltdown in the United States.

The black swan analogy goes back to the Middle Ages, when it was common belief that all swans were white. With the discovery of Australia, we learned that black swans also exist. Taleb is trying to make us aware of the limitations of our knowledge. It is human nature to try to imagine a world that is orderly and in which everything can be explained. His thesis is that the world is more random than we wish to believe.

As I said earlier, black swans can be either negative or positive. Since insurers and reinsurers are in the risk-taking business, and our business is to handle the downside of risk, I would like to touch on a few examples of black swans within our own industry.

NOT PLAYING BY THE RULES

Hurricane Andrew in 1992 was a black swan, in that it opened our eyes to how costly hurricanes could be. The World Trade Center loss in 2001 and Hurricane Katrina in 2005 were also black swans. If we look at these two latter events, we see an illustration of the kind of rationalization that occurs after the event. Following the World Trade Center loss, we have companies selling anti-terrorism software to help predict the cost of potential future terrorist attacks. That is fine if you assume the terrorists will play by the rules you set up in your software.

After Katrina, the software modelling companies that all got the loss estimates wrong blamed it on poor data from the insurance companies. They also said they were modelling for wind and not for flood. But I think part of the problem is the modelling companies’ love with Pareto curves. Put very simply, a Pareto curve illustrates an equilibrium found in the distribution of many small variables to a few large ones.

Unfortunately, real life does not follow Pareto curves: the lines are more jagged. An excellent article in the October 2001 issue of Scientific American predicted that if any hurricane exceeding an F3 in force were to hit New Orleans, the dykes would not hold and the city would flood. As Scientific American is available in newsstands and not some obscure publication, why didn’t the modelling companies take this information into consideration? If they had, then their Pareto curve would be fine up to an F3 hurricane. But at F3, the line in their curve would have gone up to a vertical position. At this point, the city floods, electricity, water supplies, sewage and communications cease and the patient dies! However, modellers are not used to vertical lines or the fact that actual losses are jagged affairs and cannot be smoothed out on a curve.

RELYING ON MODELS

In a recent speech to the Langdon Hall Financial Services Forum on May 6, 2008, OSFI Superintendent Julie Dickson made the following comments about the risks associated with an overreliance on models. “Many people have suggested that ‘models’ played a big role in the turmoil,” she said. “Models are all about taking what you have experienced in the past and trying to make sense out of it, so that if history repeats itself, you do not make mistakes that you could have avoided if only you had properly considered your own data and experience. Thus, while models are important, they should not be blindly relied on because they are based on the past, as well as on confidence intervals (they are right 99% of the time or 95% of the time).

“As we know, it is the tails of distribution that pose the problem. And, as many have realized, increasingly we seem to be in the tails, not in the range of the expected, which requires that even more judgment be brought to bear.”

In Canada, Quebec’s Ice Storm in 1998 was our black swan. As a result of heavy ice accumulation over a short period of time, more than 2 million people woke up without power. Some people went close to a month before their power was restored.

Although the essence of a black swan is its unpredictability, I feel we are facing future black swans in Canada due to the way we carry on our business.

For example, in Canada we give “Guaranteed Replacement Cost” [GRC] on homeowner contracts. In other words, as long as our clients insure their homes for the same amount our replacement software comes up with, we will rebuild their houses with like material and size, no matter what the cost. This is great both for consumers, because it gives them peace of mind, and for insurance brokers, because they have no exposure to their E&O policies for insufficient coverage. This system works well for a single-risk fire loss of a house. But in the event of a catastrophic loss, this is not so good for the insurance company.

Take, for example, two Western provinces: B. C. and Alberta. Both provinces have hot economies. Currently in both provinces, given the current loss frequency, company officials tell me it is hard to find labour to repair cars or houses. If we had a severe storm hitting Calgary or Edmonton, or a major quake hitting Vancouver, from where would the labour force come to repair the homes? In B. C. especially, if we had to rebuild thousands of damaged homes there, contractors would quickly learn we have no limits on our policies, thanks to GRC, and would use the situation as a financial windfall. In the Kelowna fires of 2003, where only 250 homes were destroyed, we saw the average cost of rebuilding a Cdn $500,000 home was closer to Cdn $800,000 and higher. On a much larger scale, the danger is that uncontrollable costs would push insurance companies through the top of their reinsurance program and back into their net capital and surplus. I am not advocating abolishing GRC on homeowner business.

I would suggest we cap it at 50% above the replacement cost value. That is, a Cdn$500,000 homeowner policy would have a GRC endorsement, but with a ceiling recovery of Cdn$750,000. Insurers and reinsurers are all financial institutions. Corporate governance impels us to act conservatively and wisely to protect our abilities to respond to unexpected losses. We have seen the mess the banks have incurred in the subprime mortgage situation, in which losses cannot be quantified and estimates keep changing for the worse.

Since there will continually be unpredictable events or black swans, we should not think we can foresee everything with our current risk models. We should at least have a limit on our financial obligations as an ultimate safety when all else fails. In the meantime, I heartily recommend reading the book The Black Swan, by Nassim Nicholas Taleb.