In previous articles, we have discussed the increases in catastrophe losses that insurers have seen over the past several years. One peril causing significant pain is severe thunderstorm, which includes tornadoes, hail, and straight-line winds — all of which can cause devastating property damage.
Of particular concern to insurers is hail, which has been a major driver of recent catastrophe losses. In fact, according to a Verisk Underwriting analysis of A-PLUS™ data, between 2008 and 2012, hail caused more than 4.5 million claims resulting in $32.1 billion of insured losses — an average of $6.4 billion each year. That’s almost three times higher than the preceding five-year period between 2002 and 2007, when average annual hail losses were $2.3 billion.
From a property damage perspective, the most common casualty from hail is roofing. Analysis of claims estimates by Xactware determined that 36 percent of all property claims nationally involved roof repairs, which accounted for almost 24 percent of all property claim costs between 2008 and 2012.
Cost of doing business
Traditionally, most insurers viewed losses from severe thunderstorms as a cost of doing business and a risk they couldn’t manage effectively. However, if losses from the peril continue at the current rate, severe thunderstorm losses will soon become unsustainable.
Fortunately, there are actions insurers can take to stem losses caused by hail and other severe thunderstorm components. A comprehensive approach to severe thunderstorm risk management includes a macro-level methodology that analyzes risk from the perspective of the peril and portfolio and a micro-level methodology that looks at risk from the individual property perspective.
By assessing risk at the macro and micro levels, carriers can reduce future losses by establishing general underwriting guidelines and rates based specifically on the risk that severe thunderstorms pose to their portfolios. It’s also critical to stem the tide of losses for preexisting damage by taking actions to reduce claims for damage occurring before policy inception.
The macro approach: The case for modeling to support underwriting
Effectively managing severe thunderstorm risk requires an understanding of how the peril affects your organization at the portfolio level. Currently, many insurers rely on historical loss experience data. However, the highly localized nature of the severe thunderstorm peril almost guarantees historical data isn’t sufficient for developing an underwriting strategy.
Catastrophe models overcome the inadequacies of historical data. By simulating tens of millions of individual hailstorms, tornadoes, and straight-line windstorms that severe thunderstorms generate, models effectively augment the historical record to produce the critical amount of information needed for actuarial analysis and underwriting, even in highly localized areas.
Using model results, various rates and underwriting practices for adjacent territories or different construction types may be more easily justified to producers and regulators. Models may also operate on hypothetical data sets, such as samples of typical homes of various construction types in a given ZIP code, to ensure the models capture granular differences in expected losses. The insurer thereby achieves a robust rating plan and a greater degree of confidence in the precision of regional rules and parameters.
Catastrophe models also make it is easy to test the sensitivity of financial outcomes to various underwriting scenarios. For example:
- What would be the effect on annual expected catastrophe losses if business retracted by 10 percent in one area and grew by 10 percent in another area of the same state?
- When underwriting new business, how do expected losses change by targeting specific combinations of property characteristics more resistant to hail damage?
- Are the savings insurers expect by writing actual cash value policies for the roof worth the loss of policyholders who choose a different carrier offering full replacement value?
Model-driven analysis leads to better decisions in many aspects of underwriting: the choice of risk data requested on the insurance application, policy terms, and growth strategy. Catastrophe models extend the organization’s ability to execute the classic strategies of risk management — risk avoidance, mitigation, and sharing — even for seemingly frequent perils such as hailstorms.
The micro approach: Assessing risk at the property level
In conjunction with developing underwriting rules to stem future losses related to hail and severe thunderstorm, insurers must effectively manage the risk of insuring homes with preexisting hail damage.
On-site home inspections are a common approach to mitigate risk on individual properties. However, since inspections are costly, many carriers don’t inspect all properties. Furthermore, very few carriers feel they have good processes in place to maximize inspection selection and spend, especially when it comes to identifying existing roof damage.
While on-site property inspections are still necessary to confirm the presence of hail damage, new solutions combining historical weather and claims data will provide carriers with a scientific approach for determining which properties should be inspected. This approach will help carriers reduce inspection spend and avoid claims for damage that occurred before policy inception.
High-resolution historical weather data is essential for effectively identifying policies with a high probability of damage from hailstorms. If historical weather data isn’t of adequate resolution, results won’t have sufficient detail. Some systems identify every property within a 6x6-mile (10x10-km) grid as having the same likelihood of preexisting damage. That’s not very useful, considering the small damage footprint from hailstorms.
Identifying properties with a high likelihood of preexisting damage is enough to determine the need for a property inspection. However, combining that intelligence with historical claims data will provide a more refined assessment.
By analyzing the claim activity of neighboring properties, insurers can identify properties with past roof-related claims and homes with the potential for unreported damage. For example, if a new or existing policy has no history of roof-related claims, analysis may show a number of hail claims filed within a quarter mile of that property in the past three months. That’s an indication a damage-producing hailstorm affected the property (which can be reinforced by historical weather data), making it a good candidate for a roof inspection.
When using weather and claims information, carriers today are able to make smarter, more informed decisions on which properties should get inspections.
Eye in the sky
Today, inspecting the roofs of homes with a low probability of preexisting hail damage is likely not worth the expense. However, considering the high frequency of hail events and claims losses, it’s important to know as much as possible about the age, size, condition, and cost to replace the roof. For homes with recent claim activity, insurers can extract that information from the claim estimate.
In the future, ultra-high-resolution aerial images combined with remote sensing technology will provide roof information for many properties without the expense of an on-site roof inspection. Combining the technology with the currently available macro and micro approaches to roof risk management will provide an efficient way to help control claims costs for the increasingly costly severe thunderstorm peril.
Jimmy Engström is manager, strategic innovation and marketing, for Verisk Insurance Solutions – Underwriting. He’s responsible for developing new products and go-to-market strategy for the underwriting segment. Mr. Engström has held numerous product and marketing roles within Verisk Analytics and holds the Associate in Risk Management (ARM) designation.
Scott Stransky is a senior scientist in AIR Worldwide’s research and modeling group. He’s responsible for leading the development of AIR’s U.S. and Canada severe thunderstorm models. Before joining AIR, he completed his bachelor’s and master’s degrees at MIT.