Wildfire catastrophe models and their use in California for ratemaking

Catastrophe models are typically used to develop insurance rates for low-frequency, high-severity catastrophic events. This paper explores the benefits of using catastrophe models in ratemaking compared to other actuarial techniques and provides an overview of current practices for wildfire ratemaking and the use of catastrophe models in California.

The use of catastrophe models for the earthquake and hurricane perils is also explored to demonstrate how catastrophe models can be used in ratemaking in a manner that allows for transparency to consumers and regulators while also protecting the intellectual property of companies that build catastrophe models. The use of catastrophe models in ratemaking has promoted the expansion of insurance offerings for earthquake and hurricane coverage. Similar approaches can be considered to increase the availability of coverage in the California admitted insurance market, which is increasingly prone to wildfire risk.

Over the last several years, California has experienced unprecedented wildfire activity. The deadliest recorded wildfire season in California’s history was 2018 and, in 2020, 4% of the state’s total land was affected by wildfires.¹ Insurers in the state responded by not renewing 235,520 homeowners insurance policies in 2019, which was a 31% increase from the prior year.² Large insurance companies in the state, such as State Farm and Allstate, have recently stopped writing new homeowners insurance policies.³ The California Department of Insurance (CDI) has responded by limiting nonrenewals by insurers⁴ and adopting a regulation for insurers to provide discounts on insurance premiums that encourage wildfire mitigation efforts. The regulation also provides greater transparency for consumers to better understand the wildfire risk score of their property, which is used in determining the insurance premium, as well as actions that can be taken to lower that score and the amount of premium reduction realized if additional wildfire mitigation measures are implemented.⁵

Adjustments for catastrophe losses in ratemaking

In ratemaking, it is common to remove catastrophe losses from the experience period data that is used to project future losses. Including catastrophe loss data without adjustments directly in a ratemaking analysis can exacerbate the uncertainty of future loss cost projections due to the relatively low frequency of these events compared to non-catastrophic loss events. This is typically addressed in the ratemaking process by removing the actual historical catastrophe experience and replacing it with an expected annual catastrophe loss amount for the prospective period. This can be accomplished through either a non-modeled approach or through the use of a catastrophe model.⁶ This is also outlined in Actuarial Standards of Practice 39, Section 3.4, Using a Provision for Catastrophe Losses, which states the following:

“In ratemaking, actuaries generally use historical data or modeled losses to form the basis for determining future cost estimates. The presence or absence of catastrophes in any historical data used to form future cost estimates can create biases that diminish the appropriateness of using that data as the basis for future cost estimates. The actuary should address such biases by adjusting the historical data used to form future cost estimates and determining a provision for catastrophe losses.”⁷

An example of a non-modeled approach is used in the CDI Prior Approval Rate Template (Rate Template). According to California Code of Regulations (CCR), Title 10, Section 2644.5 – Catastrophe Adjustment, “the catastrophic losses for any one accident year in the recorded period are replaced by a loading based on a multi-year, long term average of catastrophe claims. The number of years over which the average shall be calculated shall be at least 20 years for homeowners multiple peril fire.”⁸ The Rate Template then requires the input of a projected ratio of catastrophe loss and defense containment cost expenses (DCCE) to non-catastrophe loss and DCCE⁹ to calculate the catastrophe loading. Using 20 years of an insurer’s actual historical losses without adjustments can create bias in the future wildfire loss projections, and a method for adjusting the historical data is not provided in the Rate Template. Using the historical wildfire experience without adjustments fails to account for the increasing frequency of wildfires due to climate change. The Rate Template also assumes a relationship between non-catastrophe and catastrophe losses and this relationship is used to project future catastrophe losses. However, it can be difficult for those completing the Rate Template to determine causal relationships between catastrophe losses and non-catastrophe losses such as water and theft.

Other potential concerns with using historical data are that housing locations, building codes, and vegetation growth will change over time, making historical data less representative of the current or prospective period risk and, in turn, less effective at estimating prospective insurance catastrophe losses. In California, housing units in the Wildland-Urban Interface (WUI), which is the “line, area, or zone where structures and other human development meet or intermingle with undeveloped wildland or vegetative fuels,”¹⁰ increased 39% from 1990 to 2020.¹¹ Building designs and building codes may also change, which may increase or decrease the amount of insurance losses sustained, even with catastrophic events that may be similar to those in historical data.¹²

Use of catastrophe models and the California Earthquake Authority

Grossi and Kunreuther provide the following with respect to catastrophe models:

“With respect to natural disasters, there are limited data available to determine the probabilities of events occurring and their likely outcomes. In the absence of past data, there is a need for insurers to model the risk. Catastrophe models serve this purpose by maximizing the use of available information on the risk (hazard and inventory) to estimate the potential losses from natural hazards.”¹³

One example of the use of catastrophe models in California is with the California Earthquake Authority (CEA), which is a “not-for-profit, publicly managed, privately funded entity” established in 1996 by the California Legislature. The CEA provides earthquake insurance policies through participating insurance providers.¹⁴ The California Insurance Code (CIC), Section 10089.40, requires the CEA to not only set actuarially sound rates but to establish rates “based on the best available scientific information for assessing the risk of earthquake frequency, severity, and loss.”¹⁵ To meet these requirements, the CEA uses catastrophe models to develop rates and mitigation discounts and to determine the claim paying capacity for their portfolio.¹⁶ All other California earthquake insurers can utilize catastrophe models for setting rates as promulgated by CCR, Title 10, Section 2644.4 – Projected Losses.¹⁷

Wildfire catastrophe models and their use in California

Despite catastrophe models being considered among the best available scientific information for the earthquake peril, insurance companies are not currently permitted to use them when determining a catastrophe provision for their overall statewide wildfire rate.¹⁸ Insurance companies are permitted to segment risks or create a rate differential using a wildfire catastrophe model. However, this is complicated by the fact that CIC, Section 1861.07, states that information provided to the Commissioner “shall be made available for public inspection.”¹⁹ Information used to design and run catastrophe models may be considered trade secret or proprietary by the modeling vendors, which can limit an insurance company’s ability to utilize catastrophe models in a rate filing.

Catastrophe models and regulatory oversight: The Florida Commission on Hurricane Loss Projection Methodology

Florida provides an example for how catastrophe models can be utilized for ratemaking in a manner that allows regulatory oversight and consumer transparency. In 1992, Hurricane Andrew made landfall near Homestead, Florida, which resulted in nearly $16 billion (in 1992 dollars) of insurance losses for the industry and the insolvency of 11 insurance companies. Traditional actuarial models, most of which at the time relied on historical data, similar to the CDI Rate Template methodology, predicted that losses from the event would not exceed $6 billion. However, a catastrophe model from a newly formed catastrophe modeling company estimated that insured losses would be around $13 billion.²⁰ After Hurricane Andrew, the Florida Legislature began to support and encourage the use of catastrophe models for hurricane ratemaking,²¹ stating that “it is the public policy of this state to encourage the use of the most sophisticated actuarial methods to ensure that consumers are charged lawful rates for residential property insurance coverage.”²² In 1995, the Florida Legislature enacted Section 627.0628 of the Florida statutes, creating the Florida Commission on Hurricane Loss Projection Methodology (FCHLPM).²³

The hurricane catastrophe model approval process first starts with the FCHLPM adopting and revising hurricane standards, which occurs every two years. Modeling vendors that desire to have models approved must provide a notification to the FCHLPM that a model is ready for a review. Information such as statements of compliance with each standard, statements from various experts (professional engineer, statistician, actuary, meteorologist, and computer scientist), and the designation of trade secret information is then provided to the FCHLPM for review.²⁴ The Florida Legislature has allowed exemptions from public records law requirements as it has “found that it is a public necessity to protect trade secrets used in designing and constructing a hurricane or flood loss model.”²⁵ An extensive review process, which includes a public meeting, is performed to determine model acceptability. Based on this review, certain modeling revision submissions may be permitted. Once a model is determined acceptable, insurance companies can use the approved model for ratemaking purposes in the state.²⁶

Model vendors can use the approval from the FCHLPM as further validation of their models for use in insurance ratemaking. When the Version 21.0 North Atlantic Hurricane Model from Risk Management Solutions, Inc. (RMS) was approved by the FCHLPM, RMS stated that this was a “key milestone” for the model and that the approval “underscores the continued quality and reliability of our North Atlantic Hurricane Models.”²⁷ CoreLogic, Inc. has also stated that “insurers and reinsurers deserve to have confidence that the model they choose has undergone a rigorous certification process.”²⁸ Many other states also look to the FCHLPM approval of a model as an initial step in the model’s review for use in their own state.²⁹

Conclusion

Given the current state of California’s residential and commercial property insurance markets, expanding the use of catastrophe models in wildfire ratemaking in California would give insurance companies more actuarially sound and industry-accepted tools to understand, underwrite, and rate the risk, and more comfort with offering insurance in California. Both insurance companies and regulators can look to the actions around the use of catastrophe models for earthquake insurance in California over the last 30 years to increase insurance availability and determine a path forward for wildfire risks. The CEA currently has over 1 million policyholders, and private companies also provide their own earthquake insurance policies to consumers in the state.³⁰ An approval process similar to Florida’s by the FCHLPM can also be implemented to provide a more streamlined model review process for regulators as well as more transparency for consumers. Model vendors can use this rigorous review process when marketing their products to clients or in discussions with regulators in other states. As demonstrated by the CEA and with hurricane models in the state of Florida, the use of catastrophe models in all aspects of ratemaking should allow insurance writers to charge rates that are not excessive, inadequate, or unfairly discriminatory while engaging regulators and consumers in a comprehensive review process.

---

¹ Cal FIRE. Our Impact: Defending People, Wildland, and California’s Way of Life. Retrieved July 18, 2023, from https://www.fire.ca.gov/our-impact.

² Chiglinsky, K. & Chen, E, (December 4, 2020). Many Californians Being Left Without Homeowners Insurance Due to Wildfire Risk. Insurance Journal. Retrieved July 18, 2023, from https://www.insurancejournal.com/news/west/2020/12/04/592788.htm.

³ Kamisher, E., Reyes, M., & Carson, B, (June 2, 2023). It’s Not Just State Farm. Allstate No Longer Sells New Home Insurance Policies in California, Los Angeles Times. Retrieved July 18, 2023, from https://www.latimes.com/business/story/2023-06-02/allstate-state-farm-stop-selling-new-home-insurance-in-california.

⁴ California Department of Insurance (November 5, 2020). Insurance Commissioner Lara Protects More Than 2 Million Policyholders Affected by Wildfires from Policy Non-Renewal for One Year. Press release. Retrieved July 18, 2023, from https://www.insurance.ca.gov/0400-news/0100-press-releases/2020/release113-2020.cfm.

⁵ Cal. Code Regs. tit. 10, § 2644.9.

⁶ Werner, G. & Modlin, C, (2016). Basic Ratemaking, Fifth Edition (Casualty Actuarial Society), p. 98.

⁷ Actuarial Standards Board (2000). Actuarial Standard of Practice No. 39, Treatment of Catastrophe Losses in Property/Casualty Insurance Ratemaking.

⁸ Cal. Code Regs. tit. 10, § 2644.5.

⁹ California Department of Insurance (June 5, 2023). Prior Approval Rate Filing Instructions. Retrieved July 18, 2023, from https://www.insurance.ca.gov/0250-insurers/0800-rate-filings/0200-prior-approval-factors/upload/PriorAppRateFilingInstr_Ed06-05-2023.pdf.

¹⁰ U.S. Fire Administration. What Is the WUI? Retrieved July 18, 2023, from https://www.usfa.fema.gov/wui/what-is-the-wui.html.

¹¹ Jergler, D. (June 12, 2023). Report: Wildfire Fears at ”All-Time High” and Driving Insurer Decisions in Property Market. Insurance Journal. Retrieved July 18, 2023, from https://www.insurancejournal.com/news/west/2023/06/12/724932.htm.

¹² Grossi, P. & Kunreuther, H. (2004). Catastrophe Modeling: A New Approach to Managing Risk (Springer), p. 45.

¹³ Ibid., p. 35.

¹⁴ CEA. History of the California Earthquake Authority (CEA), Retrieved July 18, 2023, from https://www.earthquakeauthority.com/About-CEA/CEA-History.

¹⁵ Cal. Ins. Code § 10089.40.

¹⁶ Pomeroy, G. Use of Catastrophe Models by California Earthquake Authority. Press release. Retrieved July 18, 2023, from https://ains.assembly.ca.gov/sites/ains.assembly.ca.gov/files/CEA%20Use%20of%20Catastrophe%20Models%20-%20GP%20Statement.pdf.

¹⁷ Cal. Code Regs. tit. 10, § 2644.4.

¹⁸ Cal. Code Regs. tit. 10, § 2644.5.

¹⁹ Cal. Ins. Code § 1861.07.

²⁰ Hirji, Z. (September 27, 2012). Risky Business: Modeling Catastrophes. Earth. Retrieved July 18, 2023, from https://www.earthmagazine.org/article/risky-business-modeling-catastrophes.

²¹ Florida Commission on Hurricane Loss Projection Methodology. Hurricane Standards Report of Activities as of November 1, 2021, Retrieved July 18, 2023, from https://fchlpm.sbafla.com/media/fkwlhql1/2021_hurricaneroa.pdf.

²² Fla. Stat. § 627.0628.

²³ Florida Commission on Hurricane Loss Projection Methodology, “Hurricane Standards Report,” op cit., p. 11.

²⁴ Ibid., pp. 52-66.

²⁵ Ibid., p. 12.

²⁶ Ibid., pp. 52-66.

²⁷ Moody’s RMS. (June 8, 2021). Version 21.0 of RMS North Atlantic Hurricane Models Certified by Florida Hurricane Commission on Loss Projection Methodology,” Press release. Retrieved July 18, 2023, from https://www.rms.com/newsroom/press-releases/press-detail/2021-06-08/version-210-of-rms-north-atlantic-hurricane-models-certified-by-florida-hurricane-commission-on-loss-projection-methodology.

²⁸ Hazard HQ Team (June 5, 2023). FCHLPM Certifies CoreLogic Catastrophe Modeling Software for 2023 Atlantic Hurricane Season, CoreLogic. Retrieved July 18, 2023, from https://www.corelogic.com/intelligence/fchlpm-certifies-corelogic-catastrophe-modeling-software-for-2023-atlantic-hurricane-season/.

²⁹ Hemenway, C. (June 3, 2011). New RMS Hurricane Model Approved In Florida. PropertyCasualty360. Retrieved July 18, 2023, from https://www.propertycasualty360.com/2011/06/03/new-rms-hurricane-model-approved-in-florida/?slreturn=20230529205346.

³⁰ CEA, History of the California Earthquake Authority (CEA), op cit.