Near term hurricane models overestimate losses, says Karen Clark & Co.

Karen Clark & Co., the catastrophe risk consultancy, has updated its report looking into near-term hurricane catastrophe models. The company have been investigating near-term risk modelling for a few years now and this is the third update to their report which looks at near-term hurricane risk models from the major risk modelling firms Risk Management Solutions (RMS), AIR Worldwide and EQECAT Inc.

The latest report suggests that these near-term models have severely overestimated hurricane losses over a cumulative five year period from 2006 to 2010. This follows the same trend of overestimation that Karen Clark & Co. reported in previous editions of the report.

The report found that each of the risk modelling companies near-term hurricane models initially projected insured losses of at least 35% above the long-term average for the period 2006-2010. AIR lowered this to around 16% in 2007. According to the report EQECAT has only made very small adjustments to its estimated loss increases of 35-37%. RMS has introduced some modifications to its hurricane risk model during 2009, but it still predicts losses 25% above long-term averages.

The report suggests that if you take the long-term average annual insured loss from hurricanes as $10 billion then the risk models percentage increases translate the cumulative loss for 2006 through 2010 as $60.4 billion, $68.2 billion and $67.2 billion respectively for AIR, EQECAT and RMS. Actual cumulative losses for those years were much lower at $15.2 billion according to the report.

“Now that we have completed the first five-year near term hurricane model projected period, it has become clear, as our previous reports have found, that a short time horizon is not sufficient for credibly estimating insured losses from hurricanes,” said Karen Clark, President and CEO, Karen Clark & Company. “Tropical cyclone activity changes markedly year to year, and even a near-record season of storm activity such as 2010 does not necessarily translate into large insured losses.”

“While most scientific research focuses on the number of storms that develop, insurance companies are most interested in landfalling hurricanes that cause significant losses, so the hurricane frequency paradox is an important issue for the industry,” said Ms. Clark. “While landfall activity fluctuates over time, since 1900, seven of 11 decades have been average or below, including every decade since 1950. Despite the active 2004 and 2005 seasons, and the enormous losses from Hurricane Katrina, the first decade of the 21st century was average in terms of landfalls and insured losses in the United States. There is simply no clear basis for concluding we are in a period when losses associated with hurricanes should be expected to be well above the long term average.”

You can read the full press release from Karen Clark & Co. here and download the report in PDF format here.

Obviously this could throw into doubt the projections made by risk modelling firms on the potential insured losses that could be experienced by U.S. hurricane catastrophe bonds. However, modelling the likely severity of a hurricane season is extremely technical and very difficult to predict accurately due to the many factors involved. How anyone could have predicted the benign nature of the 2010 hurricane season, considering the forecasts for an active season and the fact that it was extremely active but did not result in landfalling, loss causing storms seems almost mystical. So perhaps saying the models are wrong over a period as short as five years is a bit hasty?

We’re interested to know what our readers think, please let us know by commenting below. It would be particularly interesting to hear the counter argument from any risk modelling firms who use near-term hurricane risk models and to hear what re/insurers who utilise the models think.