Twisting in the Wind.

Tornadoes present one of the biggest challenges to modeling for homeowners coverage.

The United States, which ranks first in the world for tornado frequency, is in the very thick of tornado season, that stretch of often turbulent weather that generally lasts from March through June in what's come to be known as Tornado Alley -- Oklahoma, Texas, Nebraska and Kansas--and in many southern states. In the northern part of this nation, twisters usually become more prevalent beginning this month and extending through mid-July.

These fearsome funnel clouds, usually spawned along a squall line, can wreak widespread destruction with little, if any, warning. Last year, insured losses from tornadoes that struck parts of four states were estimated at $75 million. And two years ago, in one spectacular example on May 2, 1999, witnesses reported that 76 tornadoes swept through 18 states, leaving 49 people dead, more than 800 injured and entire neighborhoods destroyed. The width of the largest funnel, which touched down in Oklahoma, is thought to have exceeded 1 mile at times, said Swiss Reinsurance Co. in its recent report, "Twister! The Professional Reinsurer's Perspective."

This type of occurrence, known as a macro-event in catastrophe-modeling parlance, presents one of the biggest challenges to the experts who create the commercial or in-house models on which insurance companies build their rating structures for homeowners insurance. That's because these macro-events include numerous small micro-events, such as the scores of tornadoes sighted during the ay 2, 1999, superstorm.

Karen M. Clark, president and chief executive officer of Applied Insurance Research Inc., Boston, said her company had developed a severe thunderstorm model that includes hailstorms and other straight-line windstorms that typicially are not caused by one tornado or one hailstorm, but by an outbreak--the superstorm that triggers multiple micro-events.

Andrew Aftermath

As insurers began clamoring for new tools to assist them in managing their catastrophe risks in the wake of 1992's Hurricane Andrew--the single-most costly natural disaster in U.S. history--catastrophe models began multiplying. The market is now dominated by AIR and two other private firms: Risk Management Solutions Inc. of Menlo Park, Calif.; and Eqecat Inc. of New York.

In this fiercely competitive business, catastrophe modelers try to stay ahead of the pack by pouring funds into research and development of new models and products, as well as improving their analytical approach.

But arriving at that final model can be a complicated process.

For its modeling of tornadoes, AIR has drawn on historical clam from the Severe Storms Forecast Center, which provides a lot of detailed information on micro-events, Clark said. "It talks about each tornado, where it touched down, what the F-scale was and the maximum size for hail in the hailstorms," she said. "Then what we have to do is group them into these macro-events." The F-scale, or Fujita wind-damage scale, is used to rate the intensity of a tornado by examining the damage it has caused.

The company's statistical analysis considers each twister's location and when it has occurred relative to other tornadoes, Clark said. AIR then tries to cross-check these events with articles in publications and other reference materials that discuss individual tornadoes.

"We also use the Property Claims Services data to see what the insurance industry has typically defined as an event--what the dates were and where the losses were by state," she said. "We use that to validate our model as well."

Although AIR works with a data sample of more than 50,000 tornado touchdowns, that wealth of information still doesn't fill in all the pieces of the puzzle. "If you just look at locations where tornadoes have occurred, you might get an erroneous picture of where they are likely to occur in the future because tornadoes can happen in many places, but not be reported," Clark said.

Reporting Increases

As Andrew Castaldi, senior member of Swiss Re's catastrophe/peril group, Armonk, N.Y., observed: "It's the old saying: 'If a tree falls in the woods and no one's there, do you hear it? 'The tornado statistics go back 50-odd years, but in many places there are holes. That's because there may not have been enough population there to even make notice of a tornado since there wasn't much damage."

The Swiss Re report notes that the number of tornadoes recorded each year in the United States has more than tripled since 1953. "Although this figure sounds alarming, the growth more likely reflects increased reporting than increased frequency," the report said. "Climatologists maintain that the frequency of tornadoes has remained constant over time and that 'people-factors,' rather than meteorological ones, account for the increasing numbers of reported tornadoes. This is simply because there are greater densities of industry and population, and more people to see and report the events, coupled with a greater public awareness of severe weather."

Therefore, creating a model based solely on the historical data can deliver the wrong picture of where the tornado risk is highest, Clark said.

What AIR has done is develop techniques for what it calls smoothing the data and augmenting it to try to fill in any gaps--a matter of devising where tornadoes probably occurred, but were not included in the historical data because they were not reported. "That requires input from our team of meteorologists as well as our statisticians," Clark said.

Castaldi, a contributing author to the Swiss Re report on tornadoes, said his company has been developing its own tornado model to be used internally for a rating approach rather than a large loss. In building its own model, the company realized the challenges that vendors have faced, "because there are only certain places that you can look for the initial data," he said.

But during the last four years, he noted, better technology has yielded better data on twisters. "We're starting to get a lot more information, and we're getting a better idea of where the tornadoes will occur, where the Tornado Alley is and what we can expect as fur as damage losses," he said.

Indicator Light

The primary concern of tornado researchers and storm chasers is to save lives by providing better warning systems in advance of twisters. In that vein, news publications last year carried the preliminary results of an eight-week experiment called the Severe Thunderstorm Electrification and Precipitation Study, or STEPS, which ended in Kansas in July 2000. One of the surprising findings: The study team observed that lightning suddenly ceased in some updraft regions in storm clouds. Then, moments later, tornadoes formed in those areas, leading the researchers to speculate that these lightning-free zones usually precede tornadoes. This observation could help forecasters make better short-term predictions for severe weather, researchers said.

While the insurance industry strongly endorses efforts to prevent deaths and injuries during tornadoes, its focus is strictly on mitigating property loss, Castaldi said. "For property insurance liability the difference in warning time makes no difference--people can't get up and move their house or their business" if a tornado is bearing down on them, he said.

Last year, estimated insured losses from tornadoes that struck parts of Louisiana, Texas, Arkansas and Mississippi over Easter weekend reached $75 million in perils of flooding, hail, tornado and wind, said the Insurance Services Office Inc.

State Farm, the leading writer of homeowners multiple peril in Louisiana in 1999, with 35% of the market share, said it received 11,532 homeowners insurance claims, resulting in some $27.7 million in insured losses. It also received another 9,815 automobile insurance claims with about $16.7 million in insured losses. The company leads the state in private-passenger auto insurance, with 34.7% market share. In Texas, State Farm said it paid $14.1 million in insured losses from 2,788 homeowners insurance claims and some 2,943 auto insurance claims.

ISO designated the tornadoes a catastrophe, defined as any event that creates insured losses of more than $25 million. It affected companies that write about 80% of the property insurance in the market.

Risk Management

One thing insurers can do to mitigate property losses from tornadoes is to support efforts to improve building codes and construction materials. In its report, for example, Swiss Re includes detailed drawings and text explaining how homeowners can check the four most vulnerable areas of their houses--roofs, windows, doors and garage doors--for potential weakness.

Swiss Re also supports groups such as the Institute for Business and Home Safety, which has helped pool the resources of a number of insurers and reinsurers to gather information and begin testing ways to make homes safer during tornadoes. In this program, the institute is working with model communities to encourage homeowners to incorporate certain safety measures into their homes' construction in hopes of saving lives and reaping better insurance rates.

But constructing so-called safe rooms, which can be built to withstand a tornado's force even if the rest of the house blows away, can cost homeowners from $3,000 to $5,000, Castaldi said. Therein lies an obstacle. "The average person says that's a pretty big amount of money to swallow and asks if the insurance company can give him some of the money back," Castaldi said. Insurers are responsive, he noted, but the problem is that if the building is still going to fall clown except for that one room, and the average insurance policy premium might be $300 a year, insurers wonder how much they can afford to give back. "So unless people have extra money, they don't bother doing that type of construction," Castaldi said.

The Insurance Information Institute reported that insured losses from the Oklahoma-based macro-event in 1999 were $1.5 billion, making it the costliest disaster of its type in U.S. history. Insurers say they expect to see even greater losses from future tornado activity because the U.S. population is expanding into undeveloped areas of the country and, Clark said, "property values are increasing anyway, even if people don't move."

That's why Swiss Re thinks it's in the best interests of insurers, reinsurers and the public alike to gain a better understanding of tornadoes and improve the technologies for tracking them so that the number of casualties and amount of property damage can be markedly reduced.

"We think that's an area that's tended to be overlooked," Castaldi said. "We do see a void in the insurance area as far as knowledge and technology--and that's something we're trying to fill."

Big Tornado Losses Loom

Although a single twister can cause considerable damage, its impact is usually moderate--at least when compared to hurricane and earthquake damage, says a report by Swiss Reinsurance Co.

But as the reinsurer points out, even moderate tornado losses surpass those expected in common property business and, as a result, many small and midsize insurers can be confronted with a threat that is not covered sufficiently in their reinsurance treaties.

"The concentration of wealth, insurable equity and, in some instances, the population situated in tornado-prone areas all suggest that tornado losses of a magnitude similar to those of the May 1999 Oklahoma event are likely to occur more often" says the report, "Twister! The Professional Reinsurer's Perspective."

In the past, industry losses exceeding $1.5 billion were associated with more severe catastrophe-peril events, such as hurricanes and earthquakes. Now, however, losses of that magnitude and more could follow severe tornado events, the report says.

"As such, meteorologists, federal and local authorities, joined by the general public and the insurance industry, are all becoming increasingly concerned with the potential catastrophic severity of tornadoes and hailstorms." Swiss Re says.

"Historically, tornado and hail losses have been more of a frequency problem than one of individual severity," states the report, which was produced in 2000 and reviewed tornado statistics up to and including 1999.

"According to NOAA's [National Oceanic and Atmospheric Administration] Storm Prediction Center, 1,351 tornadoes occurred across the United States in 1999," the report says. "This annual total is eclipsed only by the 1,424 recorded in 1998. The total estimated insurable losses due to tornado and hail events over the last two years averaged well over $4 billion annually. These past two years are no exception. Over the past 25 years, the total property catastrophe losses due to tornadoes and hailstorms ($42 billion) amounted to more than the catastrophe losses associated with either hurricanes ($34 billion) or earthquakes ($17 billion). Comparing the annual loss due to tornadoes, hurricanes, and earthquakes in the United States over the last 25 years illustrates that tornado losses, while more prevalent, are less volatile than hurricanes or earthquakes.

"Individually, only a few of tile more dramatic tornado events cause enough devastation to capture the media's attention. Consequently, the relatively high loss frequency and individual event severity potential of $2 billion are a cause for concern to many insurers who are financially unprepared to meet the increased loss potential," the report says.