Fortunately, the robust technology many insurers use to estimate occurrence losses can also be used to estimate losses resulting from multiple event seasonsfor a single peril or even across multiple perils.

Should insurers expect more hurricane seasons like 2004?

With five landfalling hurricanes last year (Gaston was recently reclassified as a hurricane) and four significant losses, many considered the 2004 season to be unusual. Total insured losses from the hurricanes are estimated to exceed $22.5 billion. However, neither the number of hurricanes nor insured losses were record-breakers.

The most active year in the last 100 years was 1985, when six hurricanes made landfall in the United States. Last year is one of only four since 1900 in which five or more hurricanes made landfall. Looking further back in the historical record, seven hurricanes made landfall in 1886four in Texas and three along Floridas Gulf coast.

With regard to insured losses, 2004 was certainly above average, but not highly unusual. By simulating all historical hurricanes of the past 104 years using current property numbers and values, AIR Worldwide has concluded that 2004 is one of eight years in which aggregate losses would have exceeded $20 billion.

In fact, the return period of both the number of storms we witnessed in 2004 and their resulting insured losses are well within the range to which most companies manage their risk. It is therefore clear that companies need to actively manage the risk of multiple events in a single year.

What is the frequency of multiple-event seasons?

While it is generally accepted that a well-designed probabilistic catastrophe model must properly account for the annual frequency of extreme events, differences in approach become magnified when the focus is on analyzing multiple events within a single season.

In addition to standard meteorological parametersincluding central pressure, landfall location, radius of maximum winds, forward speed, and track angle at landfallmodels that accurately simulate multiple-event seasons will identify a specific year and date for each simulated storm.

This approach makes it extremely easy for companies to accurately determine the probability of multiple-event seasons, either for the industry as a whole or for individual company portfolios.

Models using this approach produce both robust occurrence and aggregate loss estimates across multiple perils and for individual events within a multiple event season.

The following are a few examples of techniques that will enable insurers to more effectively analyze and prepare for future multiple event seasons:

Estimate the frequency of multiple hurricanes in any given year.

Estimating the probability that five or more hurricanes, for example, will make landfall is straightforward using catastrophe modeling systems based on the previously described approach.

By specifying five hurricanes, the system will filter the catalog for years with five or more hurricanes. An exceedance probability curve is generated, which will identify the likelihood that five or more hurricanes will make landfall.

Estimate the probability of multiple insured losses exceeding a determined amount.

Catastrophe modeling software can be used to determine the probability that a company will experience multiple losses of at least “X” amount of dollars in a given year.

For example, if a particular company wants to assess the probability of three or more losses of $50 million each, the user simply specifies the amount of loss the company would retain for each eventor $50 millionand the minimum number of such events in any year (or three). This allows companies to identify their need for multiple reinstatements or aggregate covers.

Make real-time decisions for short-term covers.

Undoubtedly, the rapid succession of Hurricanes Charley, Frances, Ivan and Jeanne exhausted some companies catastrophe cover with two months left in the hurricane season. In this situation, it is important that any short-term cover a company secures is priced with consideration that the season is already half over.

For catastrophe models in which each simulated event is identified by the day of the year on which it occurs, the user can specify the dates of coverage inception and expirationfor example, from Oct. 1 through Dec. 31and the software will only consider the likelihood of hurricanes occurring between those two dates.

Manage the risk of multiple losses from multiple perils.

While most insurers need to manage more than just hurricane risk, some catastrophe modeling systems only allow the user to estimate losses from one peril at a time.

However, an accurate understanding of the loss potential requires a catastrophe modeling system that can perform a combined-peril analysis, since simply adding results from individual peril analyses will not produce a correct combined distribution.

Catastrophe modeling is now used for a wide range of insurance decision-making, including ratemaking, underwriting decisions, reinsurance purchasing, new business opportunities, portfolio optimization and exposure concentration analysis.

In the aftermath of the 2004 hurricane season, more effective management of catastrophe risk is now a high priority for many insurers. Those that better understand the models underlying their catastrophe management systems will be better positioned to handle the financial impact of future catastrophic events.

Uday Virkud is senior vice president at Boston-based AIR Worldwide Corp.

Reproduced from National Underwriter Edition, February 25, 2005. Copyright 2005 by The National Underwriter Company in the serial publication. All rights reserved.Copyright in this article as an independent work may be held by the author.