While it may take a certain type of re/insurance audience to appreciate these jokes (and yes, everyone laughed), his point is valid about forecasting for hurricanes. This is supposed to be an average hurricane year, with Colorado State University (CSU) having the prediction of 12 named storms, with six hurricanes are likely to make landfall.
But the flaws in cat modelling run much deeper. Hurricane Ike last year once again threw into stark relief how difficult it is to predict what a catastrophic event will actually do in reality. What had been seen as a relatively mild storm and initially seen as a fairly minor event, grew in size and cost as the claims grew and grew and the magnitude of the destruction became clear as time marched on.
The flaws in cat modelling are well documented. Just think of the terrorism attacks at the World Trade Center, London, Madrid and Mumbai, as well as Hurricane Katrina’s devastation to flood defences.
Modellers have learnt from these and the many other hurricanes that have provided additional information to modelling vendors as well as insurers.
Vendor models have increased resolution of their simulations to try and fill gaps in knowledge and the quality of information form insurers has improved in quality. But what needs to change is that modelling needs to be seen as an integrated tool – something to be used just to aid underwriting, but is and always will be fundamentally flawed. For instance, how long will it take for the lessons learned in Ike to filter down into the models – and how much business will have been underwritten using them by the time the updated models hit the streets?
So as we continue to look to the weather channels to see what might be forming over the Atlantic’s unusually warm seas (and by the way there are no tropical cyclones at this time), we may have to heed the venerable Mr Kramer’s advice and get ourselves a dart board to play with.