Is It Time to Break up With Your Deficient Risk Scoring Analytics?

Risk scoring is a fundamental part of the property and casualty underwriting process, allowing underwriters to sort and rank the quality of submissions. This process culminates in critical business decisions on quoting, declination, referral, and pricing, which taken together can make the difference between an insurer’s survival and its failure. The best insurers make these decisions in a manner that is disciplined, consistent, and data-driven. Those who fail to do this fall prey to adverse selection, pay high reinsurance costs, and suffer at the hands of disapproving rating agencies.

Given this high stakes game, why does the industry continue to rely on oversimplified, unproven, and outdated risk scores for natural catastrophe underwriting?

Many commercially available catastrophe risk scores appear similar on the surface. They use straightforward 1-to-10 or 1-to-30 scoring systems that enable a frontline underwriter to quickly separate good risks from bad. Unfortunately, the scores provided by these vendors reveal significant flaws, such as:

  • They fail to take into account structural fragility. In other words, a 1975 brick structure and a 1932 wood frame house receive the same score. This makes no sense because the two houses have vastly different risk profiles for virtually every peril: wind, hail, fire, earthquake, and even winter storm.
  • The exact definition of each score is elusive. What does a “4 out of 10” really mean? Is this a percentile ranking, a damage extent, or an arbitrary binning?
  • The scores themselves are incomplete. Earthquake risk scores that don’t consider liquefaction or landslide? Flood risk scores that fail to capture tropical cyclone induced precipitation? Yes, they exist. And you should avoid them.
  • The scoring is inconsistent with the metrics used for reinsurance purchasing. The models that determine a carrier’s reinsurance strategy often give entirely different answers than the scores that guide underwriting. This inconsistency results in poor decision making.
  • There is no “ground truth” to the scores. The scores are based on models which did not consider real-life detailed insurance claims. Without this ground truth, a score is as subjective as an abstract painting.

Recognizing this problem, RMS undertook a simple goal: provide superior risk and loss metrics for every cat-exposed country, for every peril, accessible via API … for integration into any underwriting workflow.

Today, RMS has more than 50 data layers across 30 countries and nine perils enabling this insight. The data is based on its industry-leading models, which have been calibrated on billions of dollars of real-world claims and are used in production by the world’s biggest and most sophisticated (re)insurers. And each data layer provides a near-instant view of expected damage that allows fast, accurate underwriting decisions. 

To bring more transparency to insurers’ scoring systems, RMS provides an exact definition for each score it provides. For example, an earthquake score of 6 out of 10 means expected damage of 15 to 20 percent at 100 years. These metrics are fully consistent with the upstream catastrophe models used for portfolio management and reinsurance purchasing. And most importantly, the scores take into account critical attributes such as building construction, height, age and occupancy, as well as peril-specific attributes such as basement and first floor height.

Figure One: RMS Flood Risk Scores differentiate between two houses in the same location based on their construction attributes. Almost all other scores fail to do this.

Given today’s razor-thin underwriting margins and the wealth of new data sources available to insurers, catastrophe underwriting must evolve. Insurers who survive the next decade will have underwriting guidelines and risk appetites that are more discerning, granular, and data-driven. Luckily, there are products on the market that enable this transformation to occur. But there are also legacy solutions that don’t incorporate the latest science, fail to capture all sources of risk, and provide scoring information of insufficient resolution and quality.

After US$150 billion of cat loss in the past two years coupled with anemic carrier profitability, it’s time to rethink cat underwriting. If your underwriting management accomplishes one thing this year, it should be to back-test and validate the scoring systems used for catastrophe risk selection and pricing. Does it align with your actual claims experience? Does it still work for you? Is it guiding your underwriting strategy adequately? These are questions every carrier must ask itself continuously, in order to survive in an industry that is increasingly competitive, sophisticated, and data-driven.

To learn more about RMS risk scoring assets, click here.

Chris Folkman is a senior director of product management at RMS, where he is responsible for specialty lines including terrorism, casualty, wildfire, marine cargo, industrial facilities, and builders' risk. He has extensive experience on both the broker and carrier sides of insurance, where he has led many aspects of property and casualty operations including underwriting, pricing, predictive analytics, regulatory affairs, and third-party commercial coverage and claims.

Prior to RMS, he was a managing director at CompWest Insurance Company, a workers’ compensation start-up that was acquired by Blue Cross Blue Shield of Michigan. Chris holds a bachelor's degree from Stanford University. He is a licensed insurance broker and a Chartered Property and Casualty Underwriter (CPCU).