Three Models, One Green Light: CDI Wildfire Cat Model Certifications and What They Require of P&C Rate Filings

Three companies submitted their wildfire catastrophe models to the California Department of Insurance in January 2025 under the new Pre-Application Required Information Determination procedure. All three cleared the process before summer ended: 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. A fourth petition, from Cotality (the rebranded CoreLogic), arrived in November 2025 and remains in review. The CDI is now accepting rate applications that use any of the three approved models, and by the time Travelers announced its intent to participate in California's Sustainable Insurance Strategy in April 2026, the question had shifted from whether the state would permit cat modeling in rate filings to which model a carrier selects, what the regulatory conditions of use require, and how the choice plays out in competition with carriers that have not yet made the transition.

These are not hypothetical concerns for an actuarial department that will file its next California homeowners rate revision in 2026 or 2027. The first five carriers through the SIS process filed essentially identical 6.9% average rate increases, which suggests either tight model convergence or a deliberate conservative posture in year one. Neither outcome resolves the harder question of what happens as the cohort of model users grows, their cat loads start reflecting probabilistic wildfire risk rather than smoothed historical loss, and the carriers still pricing off historical data face a worsening adverse selection dynamic over five to ten development years.

Wildfire cat models cleared through CDI's PRID process: Verisk (July 24), KCC (August 1), Moody's Analytics (August 4, 2025)

85%

Market share in CDI-designated wildfire-distressed ZIP codes each carrier must write to qualify for cat model use in SIS rate filings

6.9%

Average statewide rate increase filed by the first five SIS participants: Mercury, CSAA, USAA, Pacific Specialty, and California Casualty

What the PRID Process Actually Evaluates

California Code of Regulations Section 2644.4.5, which took effect January 2, 2025, established the legal basis for using catastrophe model output in rate filings for the first time in the state's history. Before that regulation, Proposition 103's ratemaking framework required insurers to set homeowners rates using the prior 20-year average of actual incurred losses, a constraint that systematically underweighted tail risk by anchoring indicated rates to whatever fires happened to have occurred in a given 20-year window. The Northridge earthquake and the 1991 Oakland Hills fire burned into early editions of that window; the 2017-2021 catastrophe sequence drove the 20-year average up sharply, but the FAIR Plan's exposure had already expanded 123% from September 2020 to September 2024, reaching 451,799 policies covering $458 billion of California residential structure, before a single admitted carrier had filed a model-based rate.

The PRID process is not a model certification in the traditional actuarial sense. CDI does not validate the model's stochastic event set or independently assess its vulnerability functions. What the PRID evaluation covers is whether the model's methodology is grounded in accepted science, whether its outputs are consistent and reproducible, whether it incorporates climate conditioning and mitigation at the property and community level, and whether the vendor can provide a transparent documentation framework sufficient to support a prior-approval rate filing under Prop 103. Verisk's model had already completed review in Nevada before the California process opened. KCC's model held approvals in 24 states prior to the California evaluation. Moody's RMS North America Wildfire Model brought a global natural hazard portfolio background to the review. Each vendor passed the six-month evaluation on a different timeline, but all three are now available to admitted carriers preparing SIS filings.

Cotality, which rebranded from CoreLogic in 2024, submitted its PRID petition for the Cotality U.S. Wildfire Model v25.1 in November 2025. That review is ongoing as of June 2026, and its eventual approval would extend the vendor set to four, giving carriers a broader competitive basis for model selection. The existence of a fourth petition in the pipeline suggests the modeling vendor community views California's ratemaking regime as a durable market, not a one-cycle regulatory experiment.

From EAL to Rate Exhibit: The Actuarial Translation Problem

The core output that a CDI-approved model provides for rate filing purposes is the average annual loss, or EAL in some vendor nomenclature. This is the probability-weighted expected loss per policy year, derived from running tens of thousands of synthetic wildfire seasons and applying the carrier's geocoded exposure to each event's modeled intensity, spread, and vulnerability. The EAL replaces the historical loss exhibit's 20-year average as the indicated expected loss frequency-severity product for wildfire peril in the rate indication formula.

Translating EAL into an indicated rate change involves several steps that have no prior California precedent. The actuary must first reconcile the model's peril split: wildfire loss in a California homeowners form spans direct fire damage, smoke, ash contamination, and debris removal, and the model's EAL needs to map to the policy's coverage triggers rather than just structure replacement cost. Second, the credibility blending that typically offsets model output against historical experience raises a conceptual problem in California specifically, because the historical experience period is the very data source the regulation replaced. A carrier that straightforwardly credibility-weights a high-EAL model output against its own thin, low-loss historical data will get a lower indicated rate than the model alone would produce, which partly defeats the regulatory purpose. CDI's conditions of use address this by specifying that carriers must present model output alongside the supporting documentation for the EAL calculation, giving examiners the ability to evaluate whether the blending approach is actuarially sound rather than mechanically suppressing the model's loss indication.

Third, the EAL must be territory-ized. California homeowners territorial relativities for wildfire under Prop 103 ratemaking were historically derived from geographic smoothing of actual loss ratios, which produced wide zones with minimal internal differentiation in high-hazard areas. A probabilistic wildfire model produces parcel-level loss probabilities, and the actuary must aggregate those up to the territory structure the filing uses while preserving meaningful differentiation between the WUI interface tier and adjacent lower-hazard geographies. That aggregation methodology is a new actuarial exhibit with no prior California template, and its defensibility before CDI examiners will depend on the carrier's ability to explain how the modeled parcel-level variance translates into the filed territorial relativities. The FAIR Plan's April 2026 filing, which produced territory-level changes ranging from cuts of 78% in parts of the Central Valley to increases above 300% in the highest-hazard Sonoma and Sierra Nevada zones, illustrated the magnitude of territorial redistribution that model-based ratemaking can produce.

The 85% Writing Mandate and What It Costs

Using a CDI-approved cat model in a rate filing is not free. The Sustainable Insurance Strategy requires each participating carrier to write at least 85% of its statewide market share in CDI-designated wildfire-distressed ZIP codes. The intent is explicit: the state is permitting forward-looking risk pricing in exchange for a commitment that admitted capacity remains available in the very geographies where historical pricing proved inadequate. Without that commitment, the model approval would simply enable more precise underwriting triage rather than improving coverage availability.

For a carrier with a large California homeowners book concentrated in lower-hazard suburban geographies, the 85% mandate may require geographic expansion into WUI territories where its current market share is negligible. That expansion carries genuine underwriting risk: the first renewal cycle after a catastrophic WUI fire will test whether the carrier's EAL-based rates were adequate, whether its reinsurance attachment points and limits were calibrated correctly against the model's return-period output, and whether the policy form language covers the contamination and air quality claims that accompanied the January 2026 Palisades and Eaton fires. Carriers that sized their California treaty programs against historical loss patterns rather than model output entered the January 2026 wildfire season with a different risk profile than their reinsurance tower implied.

The net cost of reinsurance regulation, finalized in December 2024, pairs with the writing mandate by allowing carriers to include California-specific ceded reinsurance costs in their rate filings, provided they are SIS participants. This is the second major element of the regulatory exchange: the state both permits cat model use and permits reinsurance cost pass-through, but both permissions are conditional on the writing commitment. Carriers that stay outside the SIS cannot file reinsurance cost loads and cannot use cat models. They are left pricing off the 20-year historical average and absorbing their full reinsurance cost within the indicated rate, which becomes progressively harder to reconcile with Prop 103 hearing requirements as reinsurance costs rise.

Early Adopters and the Adverse Selection Clock

California homeowners generated $13.7 billion in direct written premium across admitted carriers in 2023. The 10-year underwriting result from 2013 through 2022 was negative 10.9% on a direct earned premium basis, reflecting the accumulated loss from catastrophe years that the Prop 103 backward-looking framework was too slow to price adequately. By the time the SIS filings arrive at renewal, a carrier using an approved cat model will be pricing wildfire exposure against a probabilistic EAL that accounts for forward-looking climate conditioning, fuel load, topography, and community-level mitigation. A carrier still filing off the 20-year historical average will be pricing against the actual loss record, which in current WUI geographies reflects significant underpricing relative to modeled expected cost.

That gap is the adverse selection mechanism. Over a five to ten year development cycle, model-based carriers writing at EAL-indicated rates will tend to attract the risks where their indicated rate exceeds what historical-average carriers are charging, which is precisely the lower-hazard end of the book. The WUI properties where the model produces materially higher EAL than the historical average would indicate are the properties that model-based carriers price out of reach and holdouts continue to write at below-model rates. The holdouts absorb a higher concentration of the highest-hazard risks. Their loss experience deteriorates faster than their rate indications reflect. Their Prop 103 filings eventually catch up, but at five to seven year development lags that leave them materially underpriced during a period of rising wildfire frequency and severity.

This is not a theoretical projection. California surplus lines homeowners policies grew from roughly 50,000 in 2023 to approximately 320,000 by end of 2025, a 540% increase driven almost entirely by admitted carriers non-renewing WUI exposure they could not price adequately under the 20-year historical average constraint. The SIS and its cat model approval are designed to reverse that dynamic, but the reversal depends on SIS adoption becoming broad enough that model-based pricing sets the market rate rather than competing against a large block of historical-average pricing from holdouts. As of June 2026, five admitted carriers have filed under SIS and Travelers has announced intent. The holdout population is still large enough that adverse selection is the pricing risk, not just a theoretical concern five years out.

Cat XL Pricing and the Modeled-Loss Transition

The downstream effect on California property cat reinsurance pricing is more complicated than the simple narrative of model-driven rate increases forcing cat XL repricing upward. Property cat rate-on-line declined roughly 10% on a risk-adjusted basis at the June-July 2025 California renewals, even in the immediate aftermath of the January 2025 Palisades and Eaton fires that produced an estimated $40 billion in insured losses, the costliest wildfire event on record. KBW was projecting 15-20% risk-adjusted rate-on-line declines at January 2026 renewals, driven by reinsurer capital levels that had recovered faster than loss experience from California might have justified.

That pricing softness reflected a structural disconnect: reinsurers and ILS investors were pricing California wildfire layers off model output, but their cedants were still pricing primary insurance off historical loss averages. The premium that admitted carriers collected from California homeowners insureds did not reflect modeled expected loss, so the cat XL attachment points and limits that primary carriers purchased were calibrated to a premium base that was itself underpriced relative to modeled risk. SIS adoption begins to close that gap. As admitted carriers file cat model-based rates and their primary premiums start moving toward EAL-indicated levels, the premium base supporting cat XL purchases grows, reinsurance attachment points calibrated as a percentage of premium rise in absolute dollar terms, and the reinsurance tower's relationship to modeled aggregate exposure becomes more coherent.

What this means for California property cat pricing at the January 2027 renewal cycle is not a rate-on-line surge. The capital supply environment is soft, and the reinsurance market has demonstrated repeatedly that it will not reprice aggressively after a single catastrophe season if capital levels are adequate. The more durable effect is on attachment point calibration and limit adequacy. Cedants whose cat model output implies a different probable maximum loss profile at the 250-year return period than their current tower structure assumes have an actuarial obligation to reconcile that discrepancy before they submit their next reinsurance treaty renewal. The PRID approval process created the model output. The rate filing process will eventually make that output public in rate exhibits. Reinsurers who are not yet seeing California primary carriers file cat model-based rates will see it in CDI public rate filings before they see it in renewal submissions, and the smarter cedants are updating their treaty structures to match the model before their reinsurers bring it up in negotiation.

Vendor Selection: Verisk, KCC, and Moody's RMS Differences

The existence of three approved models creates a choice that California actuaries have never faced before. Prior to PRID certification, the question was whether a vendor's model would survive regulatory review. Now the question is which of the three certified models to use, and whether to commission a secondary model for sensitivity analysis in the rate exhibit.

Each model brings different methodological antecedents. Verisk's Wildfire Model for the United States was built through Extreme Event Solutions, the former AIR Worldwide business, and reflects AIR's longstanding approach to probabilistic hazard modeling through stochastic event set construction and parcel-level vulnerability functions. KCC's model entered California with 24 existing state approvals, built on Karen Clark's academic background in catastrophe model architecture and a methodology that emphasizes stochastic simulation of fire weather conditions at fine resolution. Moody's Analytics' North America Wildfire Model carries the former RMS architecture, which approaches wildfire hazard through fire progression modeling tied to topography, fuel type mapping, and ignition probability, with a structural vulnerability library developed from post-fire survey data.

For a California rate filing, the practical model differences surface in two places: the EAL at the ZIP code or parcel level for specific geographic strata, and the coefficient of variation around that EAL, which affects the indicated risk load above the expected loss. Carriers that run all three models and find material EAL divergence for their specific geographic exposure mix have an actuarial judgment to make about which model's indicated loss is more defensible for their filing territory structure. CDI's PRID approval does not rank the models or establish a primary-model requirement; a carrier can select any approved model and is responsible for supporting its choice on actuarial grounds. That flexibility is appropriate but also means the carrier owns the model selection rationale, which will be subject to examiner scrutiny in a rate hearing if the filed EAL is substantially different from what CDI's in-house review of other carriers' filings produces as a reference range.

The secondary model question is related. PRID approval is specifically for rate filing support; carriers can and do use non-PRID models for underwriting triage, loss reserving, and reinsurance purchasing without triggering the certification requirement. A carrier might rationally use a fourth vendor model for reinsurance treaty optimization while using its certified PRID model for rate exhibits, as long as the two are internally consistent enough that the reinsurance tower the carrier buys actually covers the tail losses the rate filing model implies. Where they diverge materially, the actuary has a documentation problem.

The Other States Watching the PRID Template

California is the first state to complete a forward-looking wildfire catastrophe model review under a formal regulatory certification procedure, and the first to tie that certification to an affirmative coverage expansion commitment from participating carriers. Colorado, Oregon, and Washington have all experienced significant wildfire loss events in the past five years, all three face analogous insurance availability concerns in WUI communities, and all three have insurance commissioners who have observed the California PRID process closely.

Colorado adopted a wildfire home hardening statute in 2022 and has been building toward a model-based ratemaking framework, but has not yet established a formal PRID analog. Oregon's Division of Financial Regulation has engaged with the NAIC's catastrophe risk task force on the question of forward-looking cat model use in personal lines rate filings. Washington has the added complexity of participating in the NAIC's state-based regulatory architecture while dealing with a homeowners market that sustained significant wildfire-related non-renewals in the 2020-2024 period.

What these states are watching is not just the existence of the PRID process but its enforceability. California's Sustainable Insurance Strategy derives its leverage from Proposition 103's prior-approval requirement: carriers want cat model use and reinsurance cost inclusion in rates, and the only legal path to both runs through the SIS's writing commitment. States without comparable prior-approval authority have fewer tools to impose a coverage expansion condition in exchange for model-based pricing permission. The NAIC's Strengthen Homes Act model law, which provides a blueprint for state-run mitigation grant programs, addresses one piece of the availability problem but does not replicate California's exchange structure. Regulators in other western states looking to adapt the PRID approach will need to work within their own ratemaking authority frameworks rather than directly copying a structure built for Prop 103.

The June 2026 State of Play

The FAIR Plan's exposure stood at approximately $650 billion as of late 2025, up roughly 42% year over year, concentrated in WUI geographies where admitted market withdrawal has been most severe. The FAIR Plan's October 2026 rate revision will produce an average 29.8% statewide increase, the second significant FAIR Plan adjustment in seven months and one that reflects both the January wildfire losses and the Plan's shift to model-based ratemaking in its April 2026 filing. The Plan also issued its debut $750 million Golden Bear Re catastrophe bond and subsequently secured an additional $400 million from a second Golden Bear Re cat bond, shifting a meaningful share of its wildfire tail risk to the ILS market at spreads that the January 2026 losses narrowed but did not eliminate.

For admitted carriers, the SIS writing commitments filed by Mercury, CSAA, USAA, Pacific Specialty, California Casualty, and Travelers represent the first wave of capacity restoration in the admitted market. Whether that wave broadens through 2026 and 2027 depends largely on whether the cat model-based rate levels are confirmed as adequate over the next several renewal cycles and whether the reinsurance market continues to supply protection at attachment structures that make WUI writing economically viable for admitted carriers after the writing commitment kicks in.

From tracking California admitted carrier rate filings across the 2022-2025 non-renewal cycle, the single most persistent actuarial complaint was that Prop 103 ratemaking forced carriers to price California wildfire from a historical loss record that systematically excluded the tail event probability the exposure actually carried. The PRID approvals resolved that constraint. The resolution created new ones: model selection, exhibit construction, credibility blending in a low-precedent environment, writing commitment calibration, and reinsurance treaty alignment. P&C actuaries working on California property lines in 2026 are not operating in a simpler regulatory environment than they were three years ago. They are operating in one where the tools finally match the risk. Using them well is the work ahead.