Help

BI’s Article search uses Boolean search capabilities. If you are not familiar with these principles, here are some quick tips.

To search specifically for more than one word, put the search term in quotation marks. For example, “workers compensation”. This will limit your search to that combination of words.

To search for a combination of terms, use quotations and the & symbol. For example, “hurricane” & “loss”.

Login Register Subscribe

New hurricane model shakes up insurance market

Treat projections as guidance only, experts advise

Reprints

This year's revamp of Risk Management Solutions Inc.'s U.S. catastrophe model has thrust the industry into discussions on whether insurers are overly reliant on modeling.

While catastrophe models are an important piece of how risk is evaluated within the industry, it is crucial to use them as tools rather than as absolute views of risk, industry experts say.

Newark, Calif.-based RMS' updated hurricane model—RiskLink V11 U.S. Hurricane Model—is causing a stir in the insurance market because it has significantly changed its view of U.S. hurricane exposures, experts say.

Some of the RMS model changes include higher inland wind speeds; increases in building vulnerability; updates of secondary modifiers, such as roof type and construction; and increases in modeled storm surge losses.

“Were companies so tied up in what we refer to as delusional exactitude in the models that they left themselves susceptible to this sort of development? Absolutely the case,” said John DeMartini, head of the catastrophe risk management practice for Towers Watson & Co. in Stamford, Conn.

A primary insurance company writing catastrophe-exposed property coverages in the mid-Atlantic region, which was careful not to underwrite coastal exposures because the prior version of the RMS model provided high loss estimates in those areas, is seeing probable maximum loss increases of 150% to 200% under the updated model, Mr. DeMartini said.

Those increases are based on new inland exposures and reduced coastal exposures in the updated version of the model.

“Those same companies are crushed by this model change,” Mr. DeMartini said. “They were overly reliant on models, and many of them were overly reliant on one model.”

RMS 11.0, released in February, is the fourth generation of hurricane modeling since 2003, said Robert Muir-Wood, chief research officer at RMS in London.

RMS makes major upgrades to models about every five to six years, leaving them fairly stable in the interim, Mr. Muir-Wood said.

“When we produce a new release, there clearly is the potential that there can be, in certain areas and certain classes of business, a significant change in the losses from the previous model,” Mr. Muir-Wood said.

The upgrade includes new knowledge and science in modeling that is accumulated over a number of years, he said.

“We try—and we probably could do a better job of it—but we try and communicate to our clients about the knowledge which we are learning as we go along,” Mr. Muir-Wood said.

Bill Kennedy, CEO of global analytics and advisory for Guy Carpenter & Co. L.L.C. in New York, advises clients to use a blended approach whenever possible to help address some of the problems with variability around the models.

“Our view is that the models themselves are probably far more influential than they should be in the marketplace,” Mr. Kennedy said. “The models, when they were first developed, were established to provide guidelines for risk management. I think the industry has been lulled to sleep by the lack of credible, objective, transparent assessment of the models themselves.”

While the elements that create hurricane models are very complex, the models are just tools that by definition provide estimates, said Jayanta Guin, senior vp of research and modeling for AIR Worldwide Corp. in Boston.

“The users of the models, which are insurance companies and reinsurance companies, they also have to do their own due diligence with which tool they're using and what's inside that tool,” he said. “Therefore, there is a certain degree of ownership that they need to have of these models.”

Tom Larsen, product architect for EQECAT Inc. in Oakland, Calif., said that models do not produce an absolute answer to risk.

“We are continually working with our clients to remind them that computer models are a complement to informed decision-making and not a replacement for informed decisions,” he said.

Reinsurance companies and large insurance companies agree about the importance of storm models, but many of them use a blended approach by developing their own internal models that augment vendors' models, experts say.

“At Swiss Re, we have our own internal models, and we've always believed that some of the vendor tools were not accurate in the way they depicted their losses,” said Andrew Castaldi, head of catastrophe perils for the Americas for Swiss Reinsurance Co. Ltd. in Armonk, N.Y.

“And I'd have to say that, at least in the latest RMS version, they moved more toward the way that we were thinking, that storm surge was a bigger component of loss than it was in the past.”

Swiss Re has developed its own catastrophe model and has 50 individuals worldwide dedicated to natural disasters, such as scientists, structural engineers, mathematicians and statisticians, Mr. Castaldi said.

“We feel better judging or evaluating this risk on our own tools and technology than the tools and technology of someone else,” he said.

The Zurich-based reinsurer licenses one vendor model as a view of what the rest of the industry is examining, Mr. Castaldi said.

Mark Verheyen, senior vp and chief risk officer for CNA Financial Corp. in Chicago, said hurricane models are “fundamentally too important from a risk management perspective to not use them and to not rely on them.”

While the models are the most sophisticated tool available for an insurer to understand its exposures, “they are just a tool,” Mr. Verheyen said.

CNA licenses one model from a provider. But when looking to make a reinsurance placement, the Chicago-based insurer evaluates the outputs from multiple models, he said.

“There needs to be a certain amount of humility around the limitations of human knowledge,” Mr. Verheyen said.

While the models consider vast amounts of information and variables, there are instances that the model cannot consider elements the underwriter has access to.

“Any individual company may be overreliant on the model, depending on how much discretion they're giving their underwriters in terms of deviating from model indications,” he said.

Users of the storm models need to understand how it is working and what exactly it's doing, said Carl Hedde, senior vp and head of risk accumulation for Munich Reinsurance America Inc. based in Princeton, N.J.

“It's important, though, that models not be used in a vacuum,” Mr. Hedde said. “Test it for reasonableness as well as the scientific component. Application of judgment is still an important piece of model use within our industry.”

The Princeton, N.J.-based U.S. operations of Munich Reinsurance Co. evaluates its models and makes adjustments that take into account its own view of risk, with more than 30 scientists evaluating the models.