Underwriters 5, models 0

Dear friend,

The events of the last two months have proven once again that there is nothing new under the sun - especially if it's anything to do with reinsurance. Back in the old days (and I hasten to avoid saying, "the good old days"), before underwriters collected proper aggregate information and relied upon a combination of guile, cunning and seat-of-the-pants intuition to set rates, big losses used to meander their way around the market in long-lasting slow-motion swirls of indelible red ink.

After a large event, no one was ever 100% clear about their exact financial position, and consequently, rates would react a lot more slowly to events.

So we had a paradoxical situation in that the market's own in-built inefficiencies and abject lack of accurate underwriting information helped to smooth the underwriting cycle!

Of course the downside was that because information was forthcoming so slowly, it was also very difficult to be agile on the capital-raising front. Many London market figures recount with frustration their inability to make the most of the great opportunities (as they saw them) presented by big losses such as Hurricane Betsy in the 1960s, because they only realised that they had a hard-market opportunity after it was too late.

The post-Betsy market is legendary in London for written lines being reduced because syndicate and company stamp capacity had been exceeded right across the market.

Of course, I have argued many times that if they had been able to raise the cash quickly enough, underwriters might have seen rates stabilise fast, and the window of hard-market opportunity slam shut as quickly as it had opened. But surely, that is what we are all striving to achieve - not hard markets or soft markets, but profitable markets, year in year out.

Anyway, fast forward to today, and everyone's got the storm-tracker website up on their desktop, and if the sales patter from some of the more ambitious modelling firms were to be believed, all anybody has to do is bash in a few co-ordinates and press the magic button, and hey, presto! - instant, wholesome, accurate, loss estimates.

So how come almost all the upper-range Katrina damage estimates produced by reinsurers in that flurry of announcements that came around Monte Carlo have been revised upwards, and many almost doubled? Models have failed their latest test in spectacular fashion. From a company's point of view, the one thing that software models exist for is to be able to give the chief executive officer a quick answer to the one burning question he needs to be able to answer fast: "What is the worst-case scenario?" At the beginning of the 21st century - and a good decade into the so-called information age - it is almost comforting to know that despite the enormous strides taken and huge investments made, when a major loss event hits even the wealthiest and most sophisticated economy on earth, there is still such a huge margin of error that modelling algorithms fail.

It is within that margin of error and uncertainty that the best underwriters make their money - good luck everyone.

You'll see what your peers are thinking and keep one step ahead of everyone else - and you might win a bottle of champagne!