A ‘disconnect’ between perceived, modeled and actual catastrophe damages

There is a disconnect between insurance industry’s perceptions of potential catastrophe damages, modeled predictions of potential cat losses and actual catastrophe losses, according to David Lightfoot, managing director at Guy Carpenter & Company.
Lightfoot was speaking at an event marking the opening of Lloyd’s Toronto office on Nov. 30. The event featured the launch of a new report from the Institute for Catastrophic Loss Reduction and sponsored by Lloyd’s entitled Reducing the risk of earthquake damage in Canada: Lessons from Haiti and Chile.
Lightfoot talked about the potential financial implications of an earthquake on property and casualty insurers in Canada.
In the past, he noted, there have often been significant discrepancies between what the insurance industry believed the potential loss damages of a catastrophe might be, what catastrophe models predicted they might be and what the damages actually were.
During his presentation, Lightfoot showed a bar graph showing perceived, modeled and actual damages for four major catastrophic events: Typhoon Mireille in Japan (1991), Hurricane Andrew in Florida (1992), the 1994 Northbridge earthquake and Hurricane Katrina in New Orleans in 2005.
For each of the above events, the industry estimates of potential damages arising from such an event were half or less than the actual insured damages. And in all four instances, the modeled damages either exceeded or doubled the actual damages.
“There is a model uncertainty and a real cognitive disconnect between perception and what the models are saying,” Lightfoot said, noting the current industry expectation that a Vancouver earthquake could cause approximately $20 billion.
“That type of event, fortunately, has not happened,” Lightfoot said. “So is there a disconnect? I question that.”
The flip side to that question, added Lightfoot, is what should insurers be doing to prepare financially for such events. “One of the big things you can do is look at the quality of the data that [goes into] the model.”
Lightfoot said a fair amount of work has already gone into ensuring proper insurance to value figures are being input into the cat models. Also, he said, insurers will want to do things such as making sure multiple location policies are included in the data more accurately.