Directions Magazine (DM): CoreLogic offers RiskMeter Online as an SaaS solution to insurance underwriters, who can, for example, use your Storm Surge reports to estimate water depth during a hurricane. What is the measurable value to underwriters of the new data about the location of lightning strikes?
Howard Botts (HB): Most insurance companies are moving to a by-peril rating system for homeowner policies in which each potential hazard, such as wildfire, coastal storm surge, hail, damaging winds, lightning, etc., is modeled against their claims data warehouse to determine the relationship of each variable to loss (claims) potential. This information is then used to guide underwriting and pricing decisions.
Lightning claims represent over $1 billion in annual losses for property and casualty insurers with the average claim exceeding $6,000. By knowing which homes have significantly higher lightning risk, insurers can work with their policy holders to help mitigate losses, price policies to reflect risk and potentially offer policy discounts for homes with surge protection.
Intensity is much less relevant than frequency. Identifying properties that are located where cloud-to-ground lightning strikes occur most frequently is the most valuable tool. This is primarily due to the fact that all lightning strikes contain significant damaging force, unlike other hazards such as hail and wind. The damage associated with hail and wind can vary greatly depending upon the intensity of the wind and size of the hail.
DM: What is the spatial granularity of the lightning strike data?
HB: Lightning cloud-to-ground strike (point data) was collected using over 100 sensors and aggregated into a series of 250 meter grid cells covering the coterminous 48 states.
DM: In your opinion, is it likely that the insurance underwriter will use this higher level of granularity, or would it be more likely that a lighting potential index (e.g. “heat” map) would be created showing variability over a given area?
HB: Insurers want data that is granular enough to make property-level underwriting and pricing decisions. CoreLogic maintains a rich set of highly granular hazard risk data, as well as parcel data on 136+ million properties and building characteristics that, when used with other customer specific information, provides a complete view of the risk potential. Heat maps may be useful for visualizing risk, but 99 percent of the geospatial data used by insurance companies is never presented in map form. Instead, it flows straight into application pre-fill for agents and into real-time underwriting pricing and decision engines.
DM: You state: “To ensure precision in the analysis, the data accounts for a variety of small-scale variances that can easily impact the level of lightning risk from neighborhood to neighborhood, including mountains, valleys, lakes, rivers and other defining geographic features.” Is the risk score computed before these geomorphological influences are factored?
HB: The risk model is generated by aggregating lightning strikes in the 250 meter grids cells without regard to the underlying geography. The lightning strike data enables us to evaluate ground features and terrain characteristics that occur in areas with strike frequency. Until now, these ground level factors would be lost to generalization when evaluating lightning strike data with less spatial granularity. For example, when viewing Panama City, Florida on the Gulf of Mexico, the beach front areas have a lower risk of lightning strikes than areas of the city further inland, thus reflecting the impact of sea breezes. In contrast, the more mountainous areas outside of Denver, Colorado will experience a higher lightning strike score than areas with less topographic relief. However, that does not mean that all areas with lower relief or all coastal areas will have the same correlation with lightning risk as these two areas. We believe that the most accurate method of evaluating cloud-to-ground lightning risk is based on utilizing the lightning strike density.
DM: How are these risk factors modeled with respect to lightning strikes? That is, can you quantify on a percentage basis the effect of each factor that influences lightning strikes?
HB: The goal of the lightning risk database is to reflect the actual distribution of lightning strikes over a long historical period rather than to model the underlying geography and then predict the number of strikes. Given that the dataset reflects actual cloud-to-ground strikes, it would be possible to correlate these with on-the-ground features and terrain, but since the goal was to create actionable information for insurers and enterprise risk managers, we have not quantified the underlying factors driving lightning strike activity.
DM: Is there now an element of predictive analysis that you can provide on risk?
HB: The lightning risk score is highly predictive of actual P&C claims. The extent of how predictive lighting risk is for overall property loss is part of the proprietary models insurance companies build and the information is not shared publically.