Sam Friedman’s blog is a frequent stop of mine when I surf the internet for insurance issues. He is the editor of the National Underwriter, a trade publication serving the property and casualty segment of the insurance industry. I like Mr Friedman, he is a journalist first and foremost, one who examines the issues with great attention to balance in his reporting. He has opinions on insurance issues though, for instance he is against Gene Taylor’s multi peril bill HR 3121. However our offline conversations on that issue tell me his opinions do not stand in the way of doing the topic justice as he is open to reporting all sides of the issue in his blog and trade publication.
It was his post on weather modeling and it’s use in the industry for setting rates and determining the amount of risk to take on that got me interesting in the topic. In fact my first blog post dealt with that subject and featured one of Sam’s earlier posts on the topic. Featured was Karen Clark, who literally helped write the book on weather modeling:
One big problem is that catastrophe models are not reliable predictors of when or where a monster hurricane is going to strike, according to Karen Clark, vice chair of AIR Worldwide, one of the leading modeling firms.
“A model is just that—a model. It is still based on many simplifying assumptions, with a high degree of uncertainty,” she said. “It can tell you, given where you have your insureds, what the maximum probable loss would be should a storm of a certain strength hit a certain area, but the question of frequency is more vexing.”
Mr Friedman has been talking with Ms Clark again and did another post on the subject. How interesting that one of the founders of Weather Modeling now criticizes how they are used by property and casualty insurers:
Have property insurers become too dependent on computerized catastrophe models, and if so, what is the alternative, given the potentially huge exposures they face? Those were some of the provocative questions raised yesterday by one of the pioneers of cat modeling, Karen Clark.
Insurers have “stopped thinking about risks independently,” according to Ms. Clark, president and chief executive officer of Karen Clark & Company in Boston. You may be more familiar with her as founder of the first cat modeling company, Applied Insurance Research–later known as AIR Worldwide Corp. after its acquisition by the Insurance Services Office in 2002.
Ms. Clark is now a cat guru of sorts. Her consulting firm, according to her Web site, “helps senior executives and boards of directors make sure their companies have in place effective risk management processes that conform to best practices” when it comes to disaster exposures.
It’s ironic that now that she’s out of the business of cat modeling per se, she feels free to challenge insurer use of such technology to better assess disaster exposures.
Her message seems to be that cat models, while useful, are a blunt instrument, while emphasizing that underwriting should remain as much an art as a science. In other words, don’t blindly follow the models, but instead treat each risk individually. Go with your gut, and use common sense.
Speaking yesterday in New York during an Association of Professional Insurance Women luncheon, Ms. Clark emphasized that models are not designed to replace underwriters, but instead are merely best estimates. They certainly shouldn’t be the final word on which risks are acceptable and which are not, she added. (For Phil Gusman’s complete coverage of the speech, click here.)
“We’ve become a modeling society,” Ms. Clark said, noting that some insurers had confessed to her they followed whatever the models told them, even if the numbers don’t look quite right!
That jives with what I heard in London last October at an ACORD forum, when underwriters admitted that many employed two or even three competing models–which rarely agreed on exposure–hoping to get a range and a comfort zone within which to safely operate.
Ms. Clark’s candid speech and my own experience in London makes me wonder why insurers should even bother with computer models if they don’t really tell them anything definitive about the risks they face.
The comments to the entry were telling (mine excepted which were of the rah rah variety as I’m thankful to have a place to learn about the subject). Those on the inside seemed to agree with Ms Clark’s premise on the need for some human intelligence and judgment in the underwriting process.