Hitting potential hot spots: insurers are incorporating catastrophe modeling in underwriting individual risks.

Increased computer power together with a stream of scientific advances has raised catastrophe modeling to a level of sophistication that is light years away from the first generation of models introduced in the early 1980s, major modeling companies say. The catalyst for greatest change was Hurricane Andrew in 1992, the powerful storm that resulted in the largest insured loss up to that time.

As this process has unfolded, insurers have moved from using catastrophe models primarily for portfolio-based applications--estimating probable maximum loss in the event of a catastrophe by creating a large set of storms laid over a portfolio of business--to implementing these models in individual risk underwriting decisions.

"It's much more valuable now for companies to not just look for a PML analysis from a model to determine how much reinsurance to buy, but to use the catastrophe models in-house, that is, for pricing, developing more sophisticated underwriting guidelines, analyzing where they might want to grow in the future and of course, portfolio optimization," said Karen Clark, president and chief executive officer of AIR Worldwide Corp., Boston, one of the top three U.S. modeling firms.

"The biggest trend we see right now is an intense focus on underwriting, and the more advanced applications and deployment of modeling within the underwriting process," said Paul VanderMarck, executive vice president of products for modeler Risk Management Solutions, based in Newark, Calif. "Almost independent of a particular model is the notion of moving the use of catastrophe modeling out of the back office, of out of a centralized portfolio management function, and using modeling actively during underwriting."

This has come about, Clark noted, because insurers faced with rising reinsurance costs after Hurricane Andrew's losses tended to raise their retention levels to save some premium dollars. As a result, they are reinsuring only for extremely large losses now and must absorb the test. Over the years, however, catastrophe losses have continued to increase, not just from hurricanes and earthquakes, but from other types of events such as severe thunderstorms.

"These kinds of catastrophe losses eat away at a company's bottom line year after year," Clark said. For example, AIR estimates that insurance companies annually average nearly $6 billion for severe thunderstorm losses alone." And most of those are not huge events; therefore, they're not covered by reinsurance," she said. "That means that companies retain a lot of those losses, so they need a way to manage them better internally."

That's where catastrophe modelers, stressing the latest technology, are offering their assistance.

In AIR's case, its new Web-based technology allows insurers to monitor in real time how exposure concentrations change with the addition of each new policy to their book of business, Clark said. Also, underwriters can use this technology to look at an individual property at a specific location, determining its distance from the coast, the severe thunderstorm hazard and the proximity to a potential target of a terrorist attack.

"This helps companies get a lot of very detailed information on individual locations and individual properties before they even get added to their portfolio policies," Clark said. With AIR's recently launched CATStation system, insurers can add underwriting guidelines that will trigger an automatic alert to underwriters when a rule is violated. The system also monitors concentrations of workers' compensation, life, accident and disability--a legacy of the insured losses across multiple lines on Sept. 11, Clark said.

"This is a real focus of attention for companies, wanting to be able to aggregate their exposure from all lines of business--not just the property line," she said.

Circles of Information

Benfield, the reinsurance intermediary, has its own concentration management tool that takes data in an insurer's portfolio and uses it to identify concentrations of exposures. "We typically do it within circles. If you're interested in tornado/hail or wildfire, we'd use a circle with a five- to 10-mile radius," said Paul Budde, senior vice president. "We also perform these analyses for potential terrorist attacks where we may be looking at a circle of 1,000 feet or a quarter-mile. We are able to vary the radius."

The tool operates by envisioning each of the insurer's risks in the portfolio as the center of the circle, then projecting 1,000 feet from there, accumulating all the other risks that come into the circle, Budde said. Each risk is then assigned an amount of total insured value. Seeing this, insurers can determine if they are over concentrated in one area, he said. "And that could be something they're concerned about, for California, wildfire, or tornado in the Midwest, or any other localized event," Budde said.

Insurance companies are especially focused these days on finding ways to streamline their underwriting processes to save time and money. To meet that need, Eqecat Inc., Oakland, Calif., has introduced an Integrated Catastrophe Management System that integrates its catastrophe models into the insurer's internal underwriting and policy management systems. Using this approach, the insurer is able to perform analyses automatically and issue reports using its own internal interface formats with little of no human intervention, said Richard Clinton, president.

"If all the modeling companies offered this capability, it would result in a seamless integration of catastrophe modeling into the insurer's operation and enable them to take advantage of all the models with little additional work," he said.

Changing Models

Some insurers typically use multiple catastrophe models, which can produce varying estimates of exposures. Others rely heavily on one model only. This proved especially problematic recently when one of the vendors issued changes in its models, resulting in significant increases in exposure estimates for insurers, Clinton said. "We heard of several insurance companies that had to go out and purchase $500 million or more of additional catastrophe cover to protect their rating simply because of a modeling change," Clinton said. "A lot of companies are feeling this pain, which could have been mitigated through the ongoing use of multiple models. A key element in financial risk management is diversification, and the same principle should be applied to catastrophe risk management."

The use of two or three different models for earthquake, hurricane or other events helps a company to achieve a consensus on exposures, he note. "There's uncertainty in the science and there's uncertainty within the modeling process itself," Clinton said. "One of the ways you can quantify the modeling uncertainty is to use multiple models." Reinsurers, in particular, know that there are biases built into these models so they use several to obtain as great an understanding of their exposures as they can, he added.

And there's the problem of the varied ways in which corporations and brokers report their exposures to underwriters.

"There's no standard format for how that data is put together," VanderMarck said. "Our commercial insurance clients literally spend hours--sometimes days--cleaning up location schedules from Fortune 500 accounts." Location schedules, typically provided on an Excel spreadsheet, give detailed information on an insured's individual properties, he added.

"One important dynamic you see in the market these days is a lot of attention on improving that flow of data. Adoption of a standard format, like the ACORD format, will dramatically improve the consistency and quality of the data coming into underwriters," VanderMarck said. Last year, the Association for Cooperative Operations Research and Development published a standard data format for location information. Many think the adoption of such a standard is the solution to this problem.

Catastrophe modelers say they support this concept. "It makes sense," Clinton said. "Every modeling company has their little quirks in how they want to use data and what they think is important or not. But ultimately, it's going to benefit the insurance industry if we become consistent in our data requirements and data input formats."