Computer predictions of catastrophes aren't providing the right answers, says Anthony Hilton
Katrina has been a sobering influence on the insurance industry, not because of the scale of the losses, though these are indeed scary, but the brutal way it has exposed the frailties of the underwriters' computer modelling, disaster planning and risk control.
Before Katrina there was a growing confidence in the industry, reflected in the speeches of its leaders that advances in technology and mathematics now being applied in insurance had resulted in a step change for the better in how to price risk and how to quantify potential losses.
Katrina has shown that what has been achieved so far is a lot less than it was cracked up to be. In the words of Nick Prettejohn, chief executive of Lloyd's, the industry has been reminded that a model is only as good as its inputs and the assumptions which are made when it is written.
If these assumptions turn out to be wrong then the model gives the wrong answer. Being new to modelling the industry probably put a bit too much faith in it, he says.
In fairness, it is the case that all models need adjusting in the face of real life events - the investment banking and hedge fund industry finds that out on a regular basis when markets behave in ways that, according to the computer, they shouldn't.
Each problem once encountered is eliminated for the future, but there is no escape from the losses incurred when it first emerges. So with the hurricanes.
The four storms which hit Florida last year and Katrina and Rita this year were the first real tests of the computers' ability to predict. The fact that they got it wrong does not mean they are useless - but it does mean they need to be improved.
Interestingly, it was an American who spilled the beans about how fundamental the problems are. GE Solutions chairman and chief executive Ron Pressman told an audience at the Insurance Institute that his firm was completely rethinking the scale of the losses it faced and revisiting all its other business.
Its model had failed to pick up on the flooding - and so the environmental clean-up and the mould. The model would be okay if these disasters happened only once every 200 years, but the reality of every 20 years or less left the firm massively exposed.
The danger now though is that the industry might draw inadequate lessons. Clearly it is worried about hurricanes and the growing incidence and cost of natural disasters - and those problems will no doubt be revisited.
But what it also needs to understand is that its computer modelling in other areas is probably similarly flawed and could lead to similar financial pain when the wind blows the wrong way.
In particular, the industry has in recent years picked up a huge amount of risk from the banks by buying collateral debt obligations and similar instruments used to hedge credit risk.
The volumes changing hands are huge, and while many profess that they have the measure of these derivatives, and that the models continue to give comforting answers, the fact is that these have not been tested by a financial Katrina, a truly extraordinary financial event.
Until there has been such a test, no one can be sure what will happen. But given that the banks are more sophisticated than the insurance industry and have been much better at modelling volatility, it is probably safe to say that the losses, if they come, will fall on the insurance industry's side of the fence. IT
' Anthony Hilton is City editor of the London Evening Standard