September 1, 2003 by David Lalonde & Dr. Tim Doggett
U.S. insurers were reminded of the destruction of severe thunderstorms earlier this year when close to 400 tornadoes, 1,000 hailstorms and 1,500 severe straight-line windstorms wreaked havoc across 18 states over a 10-day period between May 2-11. From a meteorological perspective, it was one of the largest severe thunderstorm outbreaks in U.S. history. From an insurance perspective, the storms caused the largest insured loss ever as a result of a severe thunderstorm system. The most recent loss estimate from the U.S.-based Insurance Services Office (ISO) places these losses at more than US$3.1 billion.
Canada has had its share of significant insured losses due to severe thunderstorms. The 1987 Edmonton tornado, a 1985 tornado in Barrie, Ontario and the 1991 hailstorm in Calgary stand out in the minds of many insurers. According to the IBC, these events caused insured losses of $215 million, $132 million and $411 million respectively. These and many other examples illustrate that Canadian insurers have experienced significant losses in the past and can be assured of experiencing similar, and even more significant losses in the future.
According to the catastrophe models developed by AIR Worldwide Corp., estimated losses from severe thunderstorms are very similar to estimated earthquake losses on an average annual basis. It is not until you analyze the less frequent events, on the tail of the distribution, that earthquake loss potential becomes significantly higher.
AIR’s model shows the expected average annual industry loss from severe thunderstorms in Canada is roughly $300 million, slightly less than the same loss from earthquake. In analyzing the potential occurrence loss – the loss from a single event in any given year – Canadian insurers could see severe thunderstorm losses of more than $600 million in one year out of every 20 years, which is more than twice the occurrence loss from earthquake over the same return period. At the tail of the loss distribution – events that happen once every 100, 250, 500 or 1,000 years – earthquake risk is more severe, but Canadian insurers can still expect to see a single occurrence loss from severe thunderstorm approaching $2 billion at least once in a 100 years.
How can insurers appropriately assess and manage a risk that has the potential to cause catastrophic losses, but only on rare occasion? Because of the relative infrequency of these events and changing exposures, actuarial techniques based on past loss experience may not fully account for the true potential for large losses. Rather, estimating future losses from these storm systems requires an understanding of their meteorological characteristics and their impact on property. Advances in catastrophe modeling software give insurers the ability to manage severe thunderstorms risk at an extremely detailed level.
The same conditions that favor the development of severe thunderstorms in the U.S. are also found in Canada. The Canadian severe thunderstorm season is at its peak in late June and July. When the situation is favorable, it is common for a large number, or “families”, of storm events to occur over the course of several days, further increasing the risk associated with this extreme weather.
In Canada, the severe weather threat is most pronounced in two regions. The first stretches from Lake Erie up along the St. Lawrence Seaway, including the cities of Windsor, Toronto, Ottawa, and Montreal. In this case, warm, humid air that builds up over the Gulf of Mexico is carried north and east by the prevailing wind, which can interact with fronts. When these factors couple with the jet stream, which is commonly located in southern Canada during mid-to-late summer, conditions are ripe for severe weather in this area.
The second region is found in the southern plains of central Canada, stretching from the Rocky Mountains across to southeastern Manitoba, including the cities of Edmonton, Calgary, Regina, and Winnipeg. In this area, storms are initiated along the Rockies when air is lifted as it flows over the mountains. As these storms then migrate eastward, they can become severe as they interact with the often-present jet stream.
While these areas are most at risk from severe thunderstorm perils, it is important to realize that most of southern Canada experiences some level of risk from these events. In addition, since severe thunderstorm systems in sparsely populated areas often go unreported, the level of risk from hail, tornadoes, and straight-line winds is even greater than observed data would reveal.
Since the late 1980s, catastrophe models have been used by insurers and reinsurers to manage risk from hurricanes and earthquakes. More recently, high-resolution models have been introduced that effectively capture the risk from severe thunderstorms at an extremely localized level. Insurers can use these models to make better risk management decisions and increase the profitability of their portfolios. Specifically, modeling helps insurers minimize accumulations of risk, strategically price their policies, and determine various reinsurance and risk transfer strategies.
Catastrophe modeling provides insurers with the estimated loss potential for individual risks, books of business or an entire portfolio. Insured losses are calculated by applying individual company policy conditions to the total damage estimates derived from the model. Policy conditions may include deductibles by coverage, site-specific or blanket deductibles, coverage limits and sub-limits, loss triggers, coinsurance, attachment points and limits for single or multiple location policies, and risk specific reinsurance terms. After all of the insured loss estimations have been generated, the data can be analyzed for pricing decisions and risk transfer strategies.
For example, the model produces complete probability distributions of losses, also known as “exceedance probability curves”. Output includes probability distributions of gross and net losses for both annual aggregate and annual occurrence losses. These “probabilities” can also be expressed as return periods. For instance, the loss associated with a return period of 10 years is likely to be exceeded only 10% of the time or, on average, in one year out of 10 years.
Model output may be used to run sensitivity tests, develop underwriting guidelines, analyze policy conditions, make sound decisions regarding the purchase of reinsurance, estimate consistent loss costs for catastrophe-prone areas, and for overall catastrophe risk management. Output can also be customized to any desired degree of geographical resolution down to the street address.