In 2024, the Global Federation of Insurance Associations told the International Association of Insurance Supervisors that "insurers have proven they are capable and up to the task of managing this climate risk." That same year, State Farm, the largest home insurer in California, had been paying $1.26 for every premium dollar collected over the previous nine years. It had already stopped writing new policies. It was about to nonrenew 72,000 more.
Both statements are documented in the industry's own filings. They were made to different audiences through different institutional channels for different purposes, and neither is false. The assertion went to an international regulatory body shaping supervisory guidance. The withdrawal was communicated to a state insurance department reviewing rate adequacy. The homeowner opening a nonrenewal letter saw neither document. The distance between those two communications contains the entire structure of how climate risk knowledge moves through the economy, and how it stalls.
The Wholesale Price of Catastrophe
The clearest price signal for climate risk anywhere in the global economy is not a consumer product. It lives in the reinsurance market, where insurance companies buy their own coverage against catastrophic losses.
Munich Re and Swiss Re began using their loss databases to warn about climate change in the early 1990s. Those databases now show insured losses from natural catastrophes exceeding $100 billion for six consecutive years. In 2024, only 43% of $318 billion in global economic losses were insured at all. Reinsurance pricing responds to these signals with a speed and transparency that retail markets do not. At the January 2023 renewals, the Guy Carpenter U.S. Property Catastrophe Rate-On-Line Index rose 35%, the sharpest increase in seventeen years. Loss-free accounts saw increases of 20% to 50%. In Florida, reinsurance costs tripled between 2021 and early 2024.
These are wholesale prices set between sophisticated institutional actors sharing the same catastrophe models, loss histories, and climate projections. The pricing is blunt because the information is shared. No homeowner in Florida or California has ever seen the Guy Carpenter index. There is no mechanism by which it reaches a renewal notice.
After the 2023 reset, reinsurers didn't just raise prices. They raised attachment points, meaning primary insurers had to retain more lower-layer losses before reinsurance kicked in. The reinsurers' share of natural catastrophe losses fell from 20% in 2022 to 12% in 2024, even as total losses climbed. Reinsurers posted returns on equity above 16% for three straight years. The wholesale market was working, and working well, for the institutions operating within it.
For primary insurers, losses that once would have been passed upward now sat on their own books. Those retained losses became the pressure driving nonrenewals, market exits, and rate filing requests. State Farm's March 2024 letter to the California Department of Insurance cited "inflation, catastrophe exposure, reinsurance costs and the limitations of working within decades-old insurance regulations." The reinsurance signal was embedded in that list, aggregated alongside other factors, stripped of the underlying pricing data, and framed as justification for a rate increase that would take months or years to approve.
When reinsurance rates subsequently softened, falling 12% at the January 2026 renewals, the savings did not propagate back to retail consumers at the same speed. Florida retail premiums remained 51% above 2021 levels, according to a broker-compiled analysis of state filings. The signal travels down the chain with friction and delay. It travels up with friction and delay. The homeowner experiences the net result of both lags simultaneously, without seeing either one.
From Science to Price to Silence
Between climate science and the reinsurance pricing it informs sits another layer of translation: the catastrophe modeling firms. Companies like Moody's RMS, Verisk, and CoreLogic build proprietary models that convert climate projections, building inventories, and historical loss patterns into probabilistic estimates of insurable loss. These models are the instruments through which physical risk becomes financial risk. Reinsurers and primary insurers negotiate using their outputs. An insurer's rate filing may reference a model's conclusions without the regulator, let alone the consumer, being able to interrogate the model's underlying assumptions about fire behavior, storm frequency, or building vulnerability. The models have existed for over two decades. California prohibited their use in rate filings until 2024. The knowledge was there, translated into a form the financial system could act on, and the regulatory system chose not to look at it.
Between the insurer's actuarial analysis and the policyholder's renewal notice sits the rate filing process itself. It functions as consumer protection and information filter simultaneously, and those two functions are in tension.
An insurer submits a rate filing to the state insurance department with actuarial justification. The department reviews it. In California, median approval time was 305 days over the 2020–2024 period. In Colorado, 331 days. During that interval, the risk the filing was designed to price continues to evolve. Insurance pricing uses past experience to predict future costs, and the filing process means those predictions must be made for a more distant and therefore more uncertain future.
California's framework illustrates how specific policy choices shape the information environment over decades. Under Proposition 103, rate increases above 7% for personal lines trigger a mandatory public hearing if requested. From 2010 to 2023, many insurers repeatedly filed increases just below the 7% threshold. The regulatory limit constrained filing requests even as underlying risk intensified. Until 2024, California was the only state that prohibited insurers from using forward-looking catastrophe models in rate filings, requiring them to base rates on historical wildfire losses. The state also excluded reinsurance costs from rate calculations.
The California DOI's own description of the reform is revealing:
"Wildfire catastrophe models have existed for more than 20 years, and every other U.S. state allows insurance companies to set their rates using this modeling."
For two decades, the most sophisticated tools for pricing wildfire risk existed and were used everywhere except in the state with the most wildfire exposure.
The industry's trade association, APCIA, has argued that when regulators suppress or delay rate adjustments, it "masks socially beneficial climate change risk signals, forcing other policyholders and taxpayers to subsidize those living in high-climate-risk regions." A notable formulation. The industry acknowledges a gap between actuarial knowledge and consumer pricing, and locates responsibility for it in the regulatory process. The formulation omits the industry's own role in the information chain: the portions of rate filings designated as trade secrets, the aggregation of risk factors in public communications, the catastrophe model outputs that inform reinsurance negotiations but never appear in any document a homeowner might encounter.
Publicly available and publicly understood are separated by a distance that is itself structural. A finalized rate filing sitting in a state insurance department database, written in actuarial notation, accessible to anyone who knows where to look and how to read it, is technically public. Functionally, for the homeowner trying to understand their own risk, it might as well not exist.
Where the Gap Becomes a Bill
The FAIR plan is where the distance between private knowledge and public pricing collapses into a number someone has to pay.
California's FAIR Plan grew from roughly 240,000 policies in September 2021 to 573,739 by March 2025, a 139% increase. Exposure reached nearly $600 billion. When the Palisades and Eaton fires struck in January 2025, the plan's retained earnings of $510 million were depleted. Total estimated losses reached $4.1 billion. The plan triggered its reinsurance tower, recovering $1.45 billion. It was not enough.
In February 2025, Insurance Commissioner Ricardo Lara approved a $1 billion assessment on the FAIR Plan's member insurers, the first in over thirty years. Each member insurer's share is calculated based on its market share of premiums written two years prior. For assessments up to $1 billion, insurers may recoup 50% from policyholders through surcharges, subject to the Commissioner's approval. Above $1 billion, they may recoup 100%.
A homeowner in Sacramento who has never held a FAIR Plan policy, who lives nowhere near a wildfire zone, will pay a surcharge on their private insurance premium to cover losses on policies written in Pacific Palisades.
The surcharge will appear as a line item on their bill. It will not explain the catastrophe models that priced the risk, the regulatory framework that kept premiums below actuarial adequacy for years, or the reinsurance dynamics that shifted losses back onto primary insurers.
The pattern extends across states. Louisiana's Citizens Property Insurance Corporation levied assessments after Hurricanes Katrina and Rita that ran for nearly twenty years, ending only in April 2025. Those assessments applied to virtually every property insurance policyholder in the state. In March 2025, the Louisiana Department of Insurance discovered that Liberty Mutual had been overcharging roughly 138,000 policyholders for Citizens assessments for four years, failing to update the percentage as it decreased. Even the surcharge meant to make the cost of the gap legible can itself be wrong, and no one notices for years.
In Florida, where Citizens' policy count peaked at 1.42 million in October 2023, the potential for a catastrophic hurricane to trigger assessments on all policyholders drove the state to aggressively depopulate the plan, pushing it down to 385,000 policies by the end of 2025. The underlying risk remained, redistributed to private carriers where it became less visible. Florida's average homeowner premium remains $3,815 in 2026, according to a broker-compiled analysis of state filings, still 51% above 2021 levels.
When the gap between what actuaries know and what consumers pay grows wide enough, it closes all at once, through catastrophe, and the cost arrives as an assessment on people who had no role in creating it and no access to the information that would have let them see it coming.
The Record That Appeared and Disappeared
Even when the system attempts to produce comprehensive information, the information itself proves unstable.
In January 2025, the Treasury Department's Federal Insurance Office released what it called "the most comprehensive data on homeowners insurance in history," covering more than 246 million policies across 330 insurers from 2018 to 2022. Consumers in the highest-risk ZIP codes paid 82% more in premiums and faced nonrenewal rates 80% higher than those in the lowest-risk areas. Premiums increased 8.7% faster than inflation.
APCIA called the report an "incomplete explanation about the affordability and availability of insurance." NAMIC called it "failed and flawed." Florida, the largest catastrophe-exposed state in the country, had declined to participate in the underlying data call. Seven states' data was missing entirely. By fall 2025, the Trump administration had deleted the report from FIO's website, part of a broader pattern of climate data removal following the January 2025 repeal of the Executive Order on Climate-Related Financial Risk.
The NAIC launched a new homeowners data call in March 2026, requiring insurers to submit detailed policy-level data by June 15. A public report is expected in early 2027. The data will cover 2018 through 2025, overlapping almost entirely with the deleted FIO report but collected through a different institutional channel, one year later, with results arriving one year after that. The knowledge will exist again. The gap between its existence and its availability to the people who need it will have widened by two years.
What the Architecture Produces
Each actor in this chain is behaving rationally within its own institutional logic.
Reinsurers price risk aggressively because their survival depends on it. Catastrophe modelers build proprietary tools because their business model depends on it. Primary insurers file for rate increases because their actuaries tell them the current rates are inadequate. When those filings are delayed or reduced, they nonrenew policies or exit markets, because writing insurance below cost is not viable. Regulators delay or reduce rate increases because their statutory mandate includes consumer protection and affordability. Industry trade groups tell international standard-setters that insurers can manage climate risk, because admitting otherwise would invite regulatory intervention their members oppose. The federal government collects comprehensive data, publishes it, then deletes it, because the political valence of climate information shifted between administrations. In 2023, insurers lost money on homeowners' coverage in eighteen states, up from eight in 2013. In some of those states, rates were regulated. In others, including Iowa, where at least four insurers stopped writing all insurance in 2023, they were not. The industry's argument that deregulation alone would close the gap is contradicted by its own behavior in deregulated markets.
No single actor is lying, exactly. Each communicates the portion of reality that serves its institutional function. The asymmetry is also sustained from the receiving end. A homeowner has rational incentives not to seek out information that would devalue their largest asset. A real estate market has rational incentives not to surface risk data that would suppress transactions. The gap persists in part because actors on the receiving end also benefit from not completing the circuit. The McKinsey estimate that restoring stability to California's homeowners insurance market may require $8 to $10 billion in additional annual premiums, a 50 to 65% increase, carries direct implications for property values that no participant in a home sale is eager to price in. (McKinsey is a management consulting firm; the estimate is an analytical projection, not a regulatory finding or peer-reviewed figure.)
The aggregate effect is an information environment in which the people with the most sophisticated understanding of climate risk are the people farthest from its physical consequences, and the people closest to those consequences receive the least usable information about them. A homeowner in a wildfire zone gets a nonrenewal letter. It is typically a single page. It states that the policy will not be renewed, provides an effective date, and may cite general underwriting criteria. It contains no catastrophe model output. It offers no explanation of reinsurance attachment points. It makes no reference to the Guy Carpenter index or the twenty-year prohibition on forward-looking models in the state where they were most needed. It provides the phone number for the FAIR Plan.
That number connects the homeowner to a plan whose premiums are proposed to rise 36%, whose exposure-to-premium ratio of 0.28% is a fraction of the 1.5 to 2.0% seen in comparable high-risk states, and whose continued operation required a billion-dollar assessment on every private policyholder in California. All of which reflects what the models have known for years. None of which has been communicated to the people who will pay for it, through any channel, in any form they can act on.
-
NAIC data call deadline: Insurers writing at least $50,000 in homeowners premium must submit detailed policy-level data to the NAIC by June 15, 2026, with a public report expected in early 2027 that will cover the same years as the deleted FIO report.
-
California's 85% coverage mandate: New regulations require insurers to increase their policy writing in fire-prone, underserved areas by 5% every two years until reaching 85% of their statewide market share, a phased test of whether regulatory mandates can reverse the withdrawal pattern documented here.
-
Race and place dimensions: A Brookings Metro analysis from April 2026 identified race- and place-based factors influencing homeowners insurance in the climate change era as a distinct research area requiring its own treatment, a dimension this essay's structural focus does not address.
-
Colorado River governance cliff: The seven states sharing the Colorado River have not agreed on rules for after 2026, when current guidelines expire, and the Interior Department has said it may set new rules unilaterally if no agreement is reached this summer.