Presenting a property portfolio—especially a large and complex one—to the insurance markets is a challenge. Loss experience, stability of operations, and a solid understanding of the operational hazards and external exposures to the portfolio all come into play. Today, more than ever, insurers rely on computer-based modeling to gauge their risk exposure to catastrophic natural hazard (CAT) events such as earthquakes, hurricanes, and floods. Depending on these loss projections and the uncertainty surrounding the risk, insurers determine rates and how much reinsurance to purchase, among other things.
Gathering the best data possible, presenting it to insurers, and understanding the models are critical pieces of an organization’s property insurance program. Although a model’s loss projections can never be 100 percent accurate, one thing is certain: The quality of the data that goes into the CAT model has a direct correlation to the quality of the results that are generated. And the greater certainty underwriters have about the risk they are underwriting, the more appropriately they can price the premium.
Introduction of Probabilistic Modeling
The insurance industry didn’t begin computer modeling of natural hazard exposures until the mid- 1980s, when the first sophisticated CAT loss-simulation technology was developed. As claims data accumulated and computing power increased, improvements were realized in the ability to assess the structural integrity of a wide range of properties when exposed to natural hazards.
Today’s CAT models are faster, capable of managing numerous variables, and can accommodate a greater amount of and more refined data. As a result, we now better understand the range of structural responses that can occur. For example, a precast, reinforced concrete, tilt-up, walled structure with a wood roof will respond much differently than an un-reinforced masonry structure with a steel roof, when subjected to a 6.5 magnitude earthquake event. It is in an organization’s best interest to be able to make such a distinction to underwriters.
Using the Models Effectively
Every CAT model available today is built upon sound scientific knowledge, solid engineering principles, and a bank of historical claims data. However, gaps do exist in the available data and assumptions must be made. For example, the promulgation of tropical cyclone storm sets or earthquakes in sufficient numbers to fill out a 10,000-year event set requires some assumptions.
This leads to the issue of modeled uncertainty. Quality of construction, adherence to building codes, and integrity of building materials are three of many variables that can never be fully known. There are no two hotels, office buildings, or casinos that are identical, but for expediency, they are each coded and put into the model as if they were similar.
Modeled uncertainty, whether actual or perceived, will manifest itself as insurance premium. The key to reducing uncertainty is to use complete data. It is the best and the only weapon at your disposal.
There are four pivotal property characteristics that most often drive the damage functions of a structure and are used to model loss projections: type of construction, how the building is being occupied, the year the property was built, and its number of stories. CAT models now allow for more than 100 construction types. Steel construction alone can have up to 12 sub-categories, depending on the CAT model used. Secondary building characteristics, including the shape of the roof, the soil it was built on, and architectural elements, also can be used to further differentiate properties, but not without the primary data first being provided.
The model will handle any omission in these data sets by way of default algorithms, which can increase the uncertainty of results. For example, vague, incorrect, or incomplete addresses given for a property’s location will often result in the property being located at the center of the applicable postal zone, which can have a dramatic impact on windstorm modeling if it alters the property’s distance from the coast. When possible, the best approach is to provide the latitude and longitude for each and every property.
The data that you develop, the models that you run, and the way your broker presents your risk to insurers play a significant role in how underwriters price the risk and, ultimately, the premium you pay.
Depending on how your insurance program is presently positioned, significant premium savings are not uncommon when insured properties are properly assessed, described, and coded for modeling purposes. Moreover, all savings realized now are likely to continue every year unless there is a significant change to your property portfolio.
Poor data quality leads to insurer uncertainty, which leads to higher premiums.
This is the new normal for CAT placements.
Four Tips for Improving Data Quality
- Provide latitude and longitude coordinates for each location rather than street address.
- Individually identify and describe each structure in a multi-building complex by its own set of primary risk characteristics.
- If a primary risk characteristic is not known find the answer—do not guess. All entries made in the Primary Risk Data Set should be validated.
- As a general rule, if multiple occupancy types. For example, a multi-story building with a retail shop on the first floor, offices on the second floor, and residential apartments on the third and fourth floors, should be modeled as a residential property.