As part of the AAIS Webinar Series, AAIS hosted a virtual presentation on June 13, 2023, featuring AAIS Partner, Davies. Moderated by AAIS Personal Lines Product Manager, Linda Jancik, the session explored how wind models are created, used, and regulated. Featured guest speakers, Greg Fanoe, Director & Consulting Actuary at Davies, Sandra Darby, Property & Casualty Division Actuary at the Maine Bureau of Insurance, and Shaveta Gupta, Catastrophe Risk & Modeling Actuary at the NAIC, discussed how this data is gathered from inside the storm, why it’s collected, and how it is used by insurance carriers to price policies. The panel also analyzed hurricane models from the regulation side, explaining how regulators use this data to develop legislation to further protect consumers and ensure a healthy market.

Creation of Hurricane Models

According to Fanoe, hurricane models are generally a merge of meteorological expertise, engineering expertise, and insurance expertise. "Most models will start with the [meteorological] portion of the model, which is projecting out potential hurricanes, determining the frequency with which they'll occur, and the path they'll take if they do occur,” said Fanoe. “Then, we project the path [a hurricane] will take, including where it will make landfall, what will happen once it makes landfall, and how much the wind speed will slow down.” Then, there is the wind field assessment, which determines what areas are being affected and by how much wind. “This allows you to map out all of the homes and buildings that are impacted by wind and how much wind there is,” said Fanoe. “Wind fields feed directly into the [engineering] component of the model.” The combination of the meteorological and engineering components produces an estimate of how much damage is done to every home within the wind field of the model, which leads to insurance expertise. “This component looks at what homes are written by an insurer, what the limits are on those homes, and what the deductibles are on those homes,” Fanoe explained. “What the overall limits are is how reinsurance comes into play. That is used to project the loss that a particular insurance company is going to take related to that storm." While this is a very simplified take on how hurricane models work, it is important to understand that since they utilize meteorological, engineering, and insurance expertise, it takes multiple different experts to create them. There also can be a lot of different versions of these models. That being said, almost every model will have a long-term and short-term version of that model. “The only difference between the long and short-term versions of the model is in the frequency assessment of the hazard model,” Fanoe shared. “A long-term model will use usually 100-year or more averages of the actual landfall rates of hurricanes in the U.S. to project that frequency. A short-term model will be based on shorter-term averages.”

How Insurance Companies Use Hurricane Models

Since hurricane models are very complicated, Darby revealed that the Maine Bureau of Insurance requests information on each CAT model in the filings as well as the version number that they're using for their CAT load. Then, they look down to the Florida Commission of Hurricanes to see if they've approved that model. If Florida has approved it for use in Florida, then Maine allows it for use in their rate filings. “We follow this process because I do not have the expertise to review the CAT model as it is,” Darby explained. “That being said, we do have 3,500 miles of coastline here in Maine, and even though our hurricanes are usually category one or two, we're still concerned about having a large coastline in the future.” With this in mind, it is important that insurance companies are building in enough load in their rates so that they maintain solvency.

Education & Regulation of Hurricane Models

While some states, like Maine, do not have a coastline that is much impacted by severe hurricanes, Gupta believes it has become increasingly important for regulators to build knowledge and understanding of CAT models. “[Regulators] are key stakeholders when it comes to managing the insurance marketplace within their states,” she said. “They need to ensure that their markets are healthy and solvent with increased catastrophic and climate risk and that there is availability and affordability of insurance.” Historically, the knowledge and understanding of CAT models have been limited in the regulatory community for many reasons. “One is simply the lack of access to model documentation due to the proprietary nature of these commercial CAT models,” Gupta shared. “So, the documentation or the knowledge doesn't exist in a public state that is readily available for regulators.” The other, she explained, is the fact that these are complex models. “There are not enough standard resources that exist within individual state DOIs and within the NAIC around these CAT models,” Gupta stated. She is currently trying to change this with her work at the NAIC. Gupta’s role focuses on bridging the educational gap and building knowledge within the regulatory community. “We have actually developed a foundational course around these CAT models that is peril agnostic,” she revealed. “The course focuses on different components of the CAT model framework, the input-output, and the application of CAT models.” The NAIC plans to expand this training program to specific perils, which will go more in-depth depending upon individual states' needs.

If you would like to view the presentation again in its entirety, please click the video above.

Questions? Please don't hesitate to reach out to any of the featured speakers through the contact information below.

Linda Jancik
Product Manager – Personal Lines (AAIS)
lindaj@aaisonline.com

Greg Fanoe, FCAS, MAAA
Director & Consulting Actuary (Davies)
gfanoe@merlinosinc.com

Sandra Darby
Property & Casualty Division Actuary (Maine Bureau of Insurance)
sandra.c.darby@maine.gov

Shaveta Gupta, CPCU, ARM, ARe, CCM, CCRMP
Catastrophe Risk & Modeling Advisor (NAIC)
sgupta3@naic.org