Aon says many models still fall short of actual losses

Aon chose the Monte Carlo Renez Vous to issue a paper on catastrophe models, warning users not to rely on them without significant adjustments.

Following hurricane Katrina in 2005 there was a widespread backlash from reinsurers and insurers regardingthe accuracy of commercially available models. It was presumed that this would lead to a reduction in reliance on modelling; however,usage has not only increased, but has spread more widely among rating agencies, regulators, and investors.

According to Aon, studies which reassess the hurricanes which impacted the US in 2004 and 2005 show that models still underestimate the actual recorded losses by an average of 25% for personal lines insurers and 50% for complex commercial insurers.

“With this kind of potential for model miss, how can the output from today’s catastrophe models be relied upon?” the report asks.

Factors for non-modelled perils, data quality issues, model issues and even variations in damageability should all be accounted for and quantified as best they can.

The report adds: “Adjustments should be made to event frequencies in order to best reflect management’s belief, not just the belief of a given modelling firm. Finally, every event will allow for new lessons to be learned, more claims data to be analysed, and adjustments to the underlying components of the models. The models are much better than they were 15 years ago, let alone three years go and they will only further improve in time.”