Three wildfire catastrophe models cleared the California Department of Insurance's Pre-Application Required Information Determination in 2025: Verisk's Wildfire Model for the United States on July 24, Karen Clark and Company's US Wildfire Reference Model Version 3.0 on August 1, and Moody's Analytics' North America Wildfire Model Version 2.0 on August 4.
The CDI now accepts rate applications built on any of them. The first five carriers through the Sustainable Insurance Strategy filed essentially identical 6.9% average statewide increases, which is not what three independent models usually produce.
Key Takeaways
- Three models cleared PRID in 2025, with a fourth petition from Cotality filed in November 2025 and still in review, giving carriers a vendor choice California has never had.
- The first five SIS filers, Mercury, CSAA, USAA, Pacific Specialty, and California Casualty, all landed at 6.9% average statewide increases.
- Model use is conditional: SIS participants must write at least 85% of their statewide market share in CDI-designated wildfire-distressed ZIP codes.
- The FAIR Plan's April 2026 model-based filing moved territories from cuts of 78% in parts of the Central Valley to increases above 300% in the highest-hazard Sonoma and Sierra Nevada zones.
- California surplus lines homeowners policies grew from roughly 50,000 in 2023 to approximately 320,000 by end of 2025, a 540% increase.
What PRID Actually Certified
The approval is narrower than a model validation, and the constraint it lifted is specific.
California Code of Regulations Section 2644.4.5, effective January 2, 2025, established the legal basis for using catastrophe model output in a rate filing for the first time in the state's history. Before it, Proposition 103 ratemaking required homeowners rates set on the prior 20-year average of actual incurred losses, which anchored the indicated rate to whichever fires happened to fall inside that window. The 2017-2021 sequence pushed the average up sharply, but by then the FAIR Plan's exposure had already expanded 123% between September 2020 and September 2024, to 451,799 policies covering $458 billion of residential structure, without a single admitted carrier having filed a model-based rate.
PRID does not certify the stochastic event set or independently test vulnerability functions. It evaluates whether the methodology rests on accepted science, whether outputs are consistent and reproducible, whether climate conditioning and property and community mitigation are incorporated, and whether the vendor's documentation can support a prior-approval filing. Verisk had cleared Nevada before California opened. KCC arrived with approvals in 24 states. Cotality's v25.1 petition, filed November 2025, is still under review, which suggests the vendor community reads this regime as durable rather than as a one-cycle experiment.
The Translation From EAL to Rate Exhibit
The certified output is an average annual loss per policy year. Turning it into a filed rate is where the actuarial judgment concentrates, and California has no precedent for any of it.
The credibility step is the sharpest problem, because it can quietly undo the regulation. Standard practice blends model output against the carrier's own historical experience. In California the historical experience period is precisely the data source the regulation replaced. A carrier that mechanically credibility-weights a high-EAL model output against thin, low-loss history produces a lower indicated rate than the model alone, which returns it partway to the constraint it just escaped. CDI's conditions of use meet this by requiring model output to be filed alongside the EAL supporting documentation, so an examiner can judge whether the blend is actuarially reasoned rather than suppressing the indication.
Territory construction is the second. Prop 103 wildfire relativities came from geographic smoothing of actual loss ratios, producing wide zones with little internal differentiation in high-hazard areas. A probabilistic model produces parcel-level loss probabilities, and the actuary must aggregate them into the filed territory structure while keeping real separation between the WUI interface tier and adjacent geography. The FAIR Plan's April 2026 filing shows the magnitude at stake: territory-level changes running from cuts of 78% in parts of the Central Valley to increases above 300% in the highest-hazard Sonoma and Sierra Nevada zones.
None of it is free. SIS participants must write at least 85% of their statewide market share in CDI-designated wildfire-distressed ZIP codes, and the December 2024 net cost of reinsurance regulation lets them include California ceded reinsurance cost in rates only as SIS participants. Carriers outside the strategy get neither: they price off the 20-year average and absorb full reinsurance cost inside the indication. Against that backdrop, five carriers arriving at the same 6.9% reads as year-one conservatism rather than model convergence.
The Holdout Book and the Tower Underneath It
Two mismatches are running at once, and both get worse the longer SIS adoption stays narrow.
The first is selection. Over a five to ten year cycle, a carrier pricing at EAL will win risks where its indicated rate sits above what a historical-average carrier charges, which is the lower-hazard end of the book. The WUI properties where the model produces materially higher EAL than history are the ones model-based carriers price out of and holdouts keep writing below model. The holdout book concentrates the highest-hazard risks, its experience deteriorates faster than its indications reflect, and Prop 103 filings catch up at five to seven year lags.
That is already visible in where the business went. Surplus lines homeowners policies grew from roughly 50,000 in 2023 to approximately 320,000 by end of 2025, a 540% increase driven by admitted carriers non-renewing WUI exposure they could not price under the historical average. California homeowners direct written premium was $13.7 billion across admitted carriers in 2023, on a 2013 through 2022 underwriting result of negative 10.9%.
The second mismatch sits in the reinsurance tower. Property cat rate-on-line fell roughly 10% risk-adjusted at the June-July 2025 California renewals, in the immediate aftermath of the January 2025 Palisades and Eaton fires and their estimated $40 billion of insured losses, with KBW projecting 15-20% declines at January 2026. Reinsurers and ILS investors were pricing those layers off model output while their cedants priced primary business off history, so attachment points set as a percentage of premium were calibrated against a premium base that was itself below modeled expected cost.
SIS closes that gap only as primary premiums move toward EAL. Until they do, a cedant whose model output implies a different probable maximum loss at the 250-year return period than its current tower assumes is carrying that difference net. Five carriers and Travelers' stated intent is not yet enough adoption for model-based pricing to set the market rate, which means the holdout block is large enough that the selection clock is running now rather than five years out.
Further Reading
- Illinois's Rate-Review Law Turns Actuarial Credibility Into a Statutory Test – A newly enacted deemer-based prior-approval regime that cites California's Proposition 103 experience as a rate-adequacy cautionary case, with its own credibility-data statutory test for homeowners and auto filings.
- California Court Upholds FAIR Plan Assessment Recoupment Surcharge – the lighter-scrutiny recoupment fee pathway that runs alongside the PRID-certified rate filings this article covers.
- AI Wildfire Models in California Rate Filings: Credibility, Territory, and Certification Challenges in H2 2026 – The three actuarial methodology problems PRID approval created but did not resolve: credibility blending when AI model scores diverge 3x to 5x from historical data, parcel-to-territory aggregation, and ASOP 38 certification for proprietary AI models.
- How California's First Approved Cat Model Changes the Property Rate Indication Formula – The rate indication mechanics of replacing Prop 103 backward-looking data with AAL, the reinsurance cost pass-through allocation, and the 85% writing mandate's territorial cross-subsidy in detail.
- California FAIR Plan's 35.8% Wildfire Rate Hike: Cat Models, Reinsurance Costs, and a Territory Dispersion Story – The April 2026 FAIR Plan rate revision as the first prior-approval filing to combine cat model output and net cost of reinsurance, with detailed territory dispersion analysis.
- California's Public Wildfire Model Enters Its RFP Stage – SB 429's state-owned, fully open wildfire model now in its RFP stage, and the actuarial certification conflict a competing public model creates for appointed actuaries filing against these three PRID-approved vendor models.
- Verisk Synergy Studio Rewrites the Cat Modeling Playbook – How Verisk's cloud-native platform consolidates 110-plus cat models, and what the architecture change means for the pricing actuaries and pricing teams that consume model output.
- Supershear Earthquakes: The $13 Billion Blind Spot in Cat Models – A parallel case study in how certified models can systematically underestimate tail losses for a specific hazard mechanism, and the actuarial implications for relying on model output in rate exhibits without independent sensitivity analysis.
- June 2026 Catastrophe Renewals: 15% Pricing Decline Signals Structural Softening – The mid-year renewal context: rate-on-line softness and California's cat XL pricing dynamics as SIS filings begin shifting primary premium bases.
- Wildfire Losses Growing 12% Annually: Swiss Re Sigma's Structural Findings – The macro context behind California's wildfire exposure trajectory and why backward-looking historical averages persistently underestimate expected annual loss in WUI geographies.
- NAIC Catastrophe Risk Task Force Consolidation 2026 – How NAIC's cat risk framework is evolving in response to the California model approval precedent and what a federal or multi-state PRID analog might eventually look like.
- RiskScan 2026: Overlapping Perils and Actuarial Model Limits – Secondary peril interactions and the challenge of modeling wildfire in conjunction with drought, convective storm, and post-fire debris flow in the same cat model framework.
- Moody's RMS SCS HD Models and the Secondary Peril Cat Budget – How the HD architecture's physics-based hazard, 2,700-curve vulnerability library, and financial module represent the same vendor-model upgrade cycle now underway for SCS as happened for wildfire in California.
Sources
- Verisk Sets Precedent as First to Complete Wildfire Cat Model Review Process in California (Verisk press release, July 24, 2025)
- KCC Completes Review of California Wildfire Model (Insurance Journal, August 1, 2025)
- California Approves First Wildfire Risk Model for Insurance Pricing (Beinsure)
- CDI Completes Review of Verisk's Wildfire Model (Reinsurance News)
- California Department of Insurance Completes Final Review of Wildfire Model (Insurance Journal, July 24, 2025)
- California Certifies First Wildfire Catastrophe Model Tool from Verisk (Insurance Business Magazine)
- Regulatory Update: CAT Models and SIS Filings (PIFPAC)
- California Homeowners Insurance: Current State of the Market and Implications of the Los Angeles Wildfires (Milliman)
- California's New Take on Wildfire Loss Models, and What This Means for Property Insurers (Insurance Journal / Cotality, August 2025)
- CDI Public Cat Model Webinar (California Department of Insurance, December 2025)
- California Homeowners Insurance News: 2026 Market Roundup (Latent Insurance)
- California FAIR Plan Says Debut $750M Cat Bond "Augments" Robust Reinsurance Program (Artemis)