In early 2011, catastrophe modeling company Risk Management Solutions (RMS) released an update of its U.S. Atlantic hurricane model. Dubbed RMS version 11, the model would prove to be highly influential and controversial. It shook up the industry with insured loss projections spiking 20% to 100%—or more—in some locations. Some loss estimates in Texas and the mid-Atlantic states doubled, and substantial increases were seen in parts of Florida.
This update factored in lessons learned from 2008’s Hurricane Ike and several other major wind events that occurred after RMS v. 10, an earlier catastrophe model update whose changes had been driven by earthquake, not wind. Version 11 also projected higher inland wind speeds and increased losses due to storm surge.
Scott Clark, risk and benefits officer for the Miami-Dade County Public School District and former RIMS president, remembers when RMS v. 11 was being rolled out. “I was in the London market at the time, and the property syndicates [at Lloyd’s] were being told the model would have a significant impact on Texas and the Gulf Coast, but not in coastal parts of Florida.” Clark, the risk manager of the largest employer in South Florida, was in for a rude awakening.
RMS v. 11 more than doubled the school district’s probable maximum losses (PML), from nearly $900 million to almost $1.9 billion. However, his story had a happy ending. With the help of his broker at Arthur J. Gallagher & Co., Clark turned to California-based engineering and modeling company GRC Miyamoto. The firm helped the school district recode its property schedule using specific RMS codes instead of the ISO codes the school district had originally submitted. As a result, its 250-year PML was reduced by close to a billion dollars. Clark was resourceful but also fortunate. Many others saw their PMLs and insurance premiums spike after RMS v. 11. Now, the question becomes: How will this next update affect premiums?
Enter RMS Version 13
This summer, the company is planning to release another updated hurricane model. RMS version 13 (models are named for the years in which they are released) is expected to lower loss expectations in wind-prone areas, a suggestion that has some insurance purchasers in catastrophe-prone regions hoping they will get a small cost break on premiums, especially on policies where large limits are sought.
Claire Souch, senior vice president of model solutions in RMS’ London office, said that RMS version 13 will “soften” loss expectations in most states. But compared to previous forecasts, the biggest difference will be in coastal Southeast states, such as Georgia, the Carolinas and Florida, she said.
The changes are driven by RMS research conducted since 2005, showing, among other things, that fewer hurricanes in the North Atlantic basin have made landfall. The model update, which was expected to be released by the San Francisco-based firm in July, reflects trends that RMS has observed over the last several years.
Loss assumptions are also expected to improve with regard to the phenomenon known as “leakage,” or the payment of flood losses on policies that actually do not cover the peril. RMS says that insurers are doing better with this problem due to better claims handling and policy language, and because more homeowners have purchased government-issued flood coverage.
In fact, whereas loss expectation conditions will be lower along the Southeastern coast and Florida, in particular, all of RMS’ hurricane property damage forecasts for the Atlantic region over the next five years will be lower than those of RMS version 11.
And yet, despite encouraging updates from RMS earlier this year, insurance buyers and brokers remain skeptical as to whether changes brought about by RMS v. 13 will have more than a minor impact on catastrophe insurance rates and industry capacity. “I don’t want to get my hopes up,” said Clark.
Modeling’s Rising Importance
Over the past couple of decades, models made by the industry’s “big three” companies—RMS, AIR Worldwide and Eqecat—have become a staple of the property-catastrophe risk insurance process. The process of using models to gauge the likelihood and severity of natural disasters began to take shape about 20 years ago, when Hurricane Andrew hit South Florida, causing $15.5 billion in insured losses in 1992. Their rise to prominence has been steady, especially since rating agencies and global reinsurers began to take the projections of these companies very seriously when making their rating and underwriting decisions. Today, all commercial property insurers are required to have some relationship with one of the three major modelers to meet the requirements of reinsurers and rating agencies.
With so much at stake, insurance buyers are holding their breath to see what kind of changes the new RMS model might bring. In the lead-up to the release, most have been watching and waiting before making any predictions. Lori Gray, risk management division chief for Prince William County, Virginia, located some 20 miles from Washington, said she got “no feedback from the markets” regarding the impact of the new hurricane model prior to her July 1 renewal.
Even though Hurricane Sandy damaged the county’s Occoquan Bay National Wildlife Refuge enough to warrant nearly $2 million in relief from the U.S. Department of the Interior, it was mostly considered to be a storm-surge-related event. Since taking over as risk manager for the county in 1999, Gray said “we really haven’t had any significant [wind-related] damage that’s impacted our insurance.”
Though Gray’s employer has been mostly spared, other insurance buyers with significant wind exposures are naturally curious about how the changes will impact premiums and industry capacity. And for good reason.
In the spring, forecasters from Colorado State University predicted 18 named storms for the 2013 hurricane season, with nine of those forecasted to become hurricanes and four expected to be major hurricanes. The National Oceanic and Atmospheric Administration’s Climate Prediction Center warns there could be even more storms to hit—up to 20, compared to the average of 12.
In the lead-up to the new model’s release, Dave Marcus, South Florida-area chairman for Arthur J. Gallagher, and also Scott Clark’s broker, said he didn’t expect RMS v. 13 to be as influential on catastrophe insurance rates as v. 11 was. Early on, one of his associates heard from a cat property underwriter at Lloyd’s of London that some Florida rates may be reduced by an average 15% as a result of RMS v. 13. But he took that estimate with a grain of salt.
Underwriters at Lloyd’s might be anticipating a reduction in PMLs for Florida due to the model revision, but given his experience with RMS v. 11, Marcus said he won’t tell all of his clients to expect that much of a change.
“When we cut Scott [Clark’s] PML by 55% back in 2011, his premiums didn’t come down that much,” said Marcus. “My gut tells me that if what you buy is still within the RMS model’s PML for the risk in question, it most likely won’t have an impact on an account like Scott’s,” said Marcus. Instead, he expects that “there will be more of an effect on pricing in upper layers versus primary limits purchased.”
This means that, for instance, a risk manager with a $100 million PML on a particular hurricane risk before release of the new model might have as many as five carriers contributing to a layered program. One may provide a $25 million primary layer, while another company offers limits of $25 million excess of $25 million. And so on.
Given such a situation—with a PML decline of 15% that drops the total to $85 million—Marcus believes the uppermost layer of coverage reaching beyond the new PML should end up costing less. “The top layers will most likely be most impacted,” said Marcus. On the other hand, he predicts there will be “no change” on the first few layers of coverage contained within the original loss expectations.
Duncan Ellis, U.S. property practice leader at Marsh, points out that the model changes will not impact insurance rates directly, but more so the annual average losses (AALs), which “are a component of the development of the premium number and the PMLs” that dictate how much risk an insurance market can accept. Still, the indications for risk purchasers are positive. “A decrease in the PML would allow an insurance company to participate to a larger degree or at a more beneficial level to the insured,” said Ellis.
Projections Just Part of the Picture
A May bulletin from Willis explained that the key drivers to the updates in RMS’ medium-term hurricane rates reflect the academic community’s ongoing research into hurricane activity, landfall proportion of Atlantic basin hurricanes as well as the addition of new active baseline forecast models.
In addition, Willis noted that there is an updated criteria for classifying hurricanes at landfall as well as the inclusion of the 2012 hurricane season’s data extracted from sources like historical North Atlantic hurricane database (HURDAT) and sea-surface temperature data. Since there have been new findings applied to changes in the geographic distribution of hurricanes with varying sea-surface temperatures—and improved methods to implement these findings—this has led to change. Overall, RMS’ latest medium-term forecast appears to decrease significantly between 2011 and 2013 for all classes of hurricanes in Florida and the Southeast, according to Willis.
Still, as recently as May, Willis suggested that “things are still very preliminary,” with regard to the update, but “given the numerous changes to the model, the trend is for possible decreases in the modeled losses.” All the same, the broker stressed that “we need to remember that a portfolio’s make up can greatly affect the results. “Around the same time, Clark said that he had “heard bits and pieces about how version 13 is supposed to fix some of issues version 11 had created, but for me the proof is in the pudding. I’ve seen too many versions come out with flaws, where until you run it and see how it models you [individually] you really don’t know if it will be to your benefit.”
As was Clark’s experience in 2011, the lesson with this update seems to be that it will affect various portfolios differently. Similarly, insurance companies using catastrophe modeling would be wise to recognize that it is only one of several components in their risk management toolbox, said Rich Attanasio, vice president of property casualty ratings at A.M. Best. “We view catastrophe models as an important tool but not the only tool,” he said. Attanasio thinks insurers must realize both the benefits and limitations of models, and craft their own comprehensive view of risk.
Before the active, destructive hurricane seasons of 2004 and 2005, says Attanasio, there may have been an over-reliance on what the models were saying. But today, more insurers have come to realize that a PML “isn’t everything.” Ultimately, insurers need to understand “how those nuances play,” and adjust their underwriting and capital standards appropriately. Until the market reacts, it will remain difficult for policybuyers to know if this update will lead to premium breaks.