State Farm General Insurance Company paused new homeowners business in California on May 27, 2023, citing "a challenging reinsurance market" as one of three drivers in its public statement.2 Ten months later, in March 2024, the same carrier announced the non-renewal of approximately 30,000 in-force California homeowners and rental policies, with cancellations beginning July 3, 2024.3 The parent State Farm group's nationwide homeowners book ran a 99.4% loss ratio in Q3 2023, up roughly twenty-six points from 73.4% in the same quarter a year earlier — the highest single-quarter result since Q4 2017, and one S&P Global attributed in significant part to Hawaii's Maui wildfire losses.1 The carrier's California-only loss experience is reported separately on California Department of Insurance stat-page filings rather than in the consolidated public summaries, but the parent-level signal alone was the kind of number that lands on a treaty desk before it lands anywhere else.
Most of the press coverage framed the California events as a story about state regulation, the wildfire problem, and the limitations of Proposition 103. All of that is real. None of it explains why a carrier the size of State Farm moved as fast as it did in California specifically.
The number that explains it is not a State Farm California loss ratio that any outside party can compute exactly. It is the general magnitude of swing — seventeen points, twenty points, twenty-six points — at which a treaty underwriter stops asking questions and starts repricing. State Farm cited the reinsurance market in its own filing. The leading indicator behind the May 2023 announcement was visible to the carrier's reinsurance panel before it was visible to regulators or to the broader public.
What a 17-Point Swing Reads Like at a Reinsurer's Desk
A loss ratio that moves seventeen points in a year is roughly $170M of unexpected loss on a $1B premium book. On a $400M book it is $68M. The dollar figure varies. The diagnostic question for a treaty underwriter does not.
Year-over-year loss ratio variance inside a stable underwriting model and a stable peril distribution should sit in single digits. A two or three point move is normal noise. A five point move asks for a phone call. Seventeen points is a different category of finding. It says one of three things happened, and a treaty underwriter will work through them in order:
- The peril distribution shifted faster than the rating model could absorb, which is a climate diagnosis and shows up across the cedent's peer group at similar magnitudes.
- The model itself was misspecified at the prior rate filing, which is a competence diagnosis and shows up uniquely at this cedent.
- The book composition changed underneath the rate, which is a discipline diagnosis, and is the one that lands hardest in a treaty meeting.
The first answer can be defended with industry data. The other two cannot. A cedent producing a seventeen-point swing has to explain to the reinsurer which of the three is the dominant cause, and the reinsurer has already done its own decomposition before the meeting starts.
The Mechanism Following the Swing
State Farm cited reinsurance affordability and availability when it announced the May 2023 new-business pause.2 The May filing referenced the inability to obtain catastrophe reinsurance at terms the carrier could absorb. The March 2024 non-renewal of in-force policies followed the same logic, applied to the renewal book rather than the new-business pipeline.3 The market mechanism produced the entire sequence without an enforcement order or consent decree appearing anywhere in the chain.
A reinsurer reading a seventeen-point swing has a small set of moves available at the next placement: re-rating the layer, tightening warranty language around disclosure of underwriting changes (often with a specific carve-out for territories where the loss development was worst), cutting the limit they are willing to write at the prior price, and in some cases declining to participate at all. The cedent walks into the next June 1 with a meaningfully different placement than it walked into the prior one with, and the rate the carrier needs to file with the regulator to make the new program economic is larger than the rate the regulator is likely to approve in the timeframe the carrier needs.
The State Farm sequence demonstrates that the gap between the loss ratio showing up and the carrier's strategic options collapsing can be measured in months rather than years. The carrier survived by withdrawing in stages from a market it had been the largest writer in, executing the move in the window between the loss ratio becoming visible and the next reinsurance renewal closing the door on a different option.
Where AI-Driven Rating Sharpens the Risk
ML rating models that retrain on recent loss data have a property that has not yet been priced into most treaty conversations. The retraining cycle is short relative to the climate cycle the book is exposed to. A model fit on a 2018-2024 loss window will read recent fire seasons as the new baseline, and indicate rates accordingly. If the next season delivers losses outside that window, the indicated rate was wrong, the binds at the indicated rate compound, and the loss ratio shows up in the third or fourth quarter as a signal the carrier did not see coming.
By the time the carrier files for the corrective rate increase, the data the regulator will use to evaluate the filing already contains the bad year. In California specifically, Proposition 103 has historically been interpreted to prohibit the use of forward-looking catastrophe models in rate filings, and the prior-approval process can stretch well beyond the statutory review period when intervenors participate.4 The carrier is asking for a rate against a loss baseline that the regulator's review will treat as the new normal, while the underlying peril distribution may already be shifting again in the same direction. The rate-making lag is partly codified.
Aon's data showed that secondary perils, the wildfires and severe convective storms and atmospheric rivers that don't carry hurricane names, accounted for roughly 86% of global insured losses in 2023.5 Those are exactly the perils where the historical training window is most likely to be stale relative to the current distribution, and exactly the perils where a seventeen-point swing in a single year becomes plausible without anything visibly wrong with the carrier's underwriting discipline.
For carriers using AI-driven rating, the chain runs faster than it does for traditional actuarial pricing. The model sees the prior loss data, retrains, and indicates a rate. The book grows or shrinks against that rate within a quarter or two. The next loss season either confirms the indication or contradicts it. If it contradicts, the loss ratio prints, the reinsurer reads it, and the conversation at June 1 is the conversation State Farm walked into in early 2023, sized to whatever the cedent's actual book looks like.
We covered the treaty mechanics of that conversation in a companion piece on why the reinsurance treaty is where AI risk becomes existential, and the underlying concentration dynamic in our look at Merced County and what 113 years of underwriting actually buys you. The seventeen-point swing is the leading indicator that ties both threads together. It is the metric a reinsurer can compute from public filings without ever opening the cedent's model documentation, and the one a board can compute on the same source data.
What the State Farm Sequence Establishes
The useful claim from the State Farm episode is narrow. Reinsurance pressure, not regulatory enforcement, drove the largest single carrier withdrawal in a major homeowners market in recent memory. The pressure was triggered by loss ratios that moved outside the band a treaty underwriter is willing to absorb without repricing. The carrier had a rating, a brand, and a parent of unusual scale, and still chose withdrawal over recapitalizing into a market the reinsurance program no longer made economic sense in.
A regional or specialty carrier facing the same loss ratio dynamic does not have State Farm's optionality. The replacement reinsurance market is smaller for it, the rating sensitivity is higher, and the timeline between the loss ratio printing and the surplus consequence becoming binding is shorter. The State Farm sequence is therefore a warm-up rather than the worst case. The mechanism that produced the withdrawal is the same one that produces insolvency at carriers without the resources to choose withdrawal.
The CEO Question
The question to ask the chief actuary, the head of underwriting, and whoever owns the rating models, in that order, is short.
What is our largest loss ratio swing in any single line over the last three years, and could we explain it to a treaty underwriter without referring to weather?
The first half of the question is a number. Most carriers can produce it from the same data they file with the NAIC. The second half is a paragraph. It requires the carrier to have a written explanation of the swing that decomposes the move into peril shift, model misspecification, and book composition change, with evidence for the share of the swing attributed to each. The decomposition is the document a treaty underwriter is going to ask for at the next renewal whether or not the carrier has prepared it. A cedent walking into the conversation without the decomposition concedes the framing to the reinsurer.
The work to produce the decomposition is not large in absolute terms. It is a model change log written for a treaty underwriter, an aggregation report broken out by AI-rated and legacy actuarial portions of the book, and a paragraph from the chief actuary on which of the three diagnoses fits best. The work is small. The cost of not having it the first time the loss ratio prints outside the band is the conversation State Farm had with its reinsurance program in early 2023, sized to whatever a regional or specialty carrier's surplus can absorb, in a placement window that closes faster than the rate filing it would take to fix it.
A loss ratio swing of seventeen points in a year is the leading indicator. The reinsurer sees it before the regulator does, acts on it before the regulator does, and the action is binding before the rate filing the carrier needs has been approved. CEOs who can produce the decomposition in advance are buying themselves a different conversation at the next June 1. The ones who cannot are letting the reinsurer write the explanation for them.
Related
- The Reinsurance Treaty Is Where AI Risk Becomes Existential
- Merced County, Camp Fire, and What 113 Years of Underwriting Buys You
- The One Treaty Clause Your AI Strategy Lives or Dies On
- Hippo's 273% Loss Ratio: What Happens When AI Can't See Climate
Footnotes
-
"State Farm had the highest loss ratio in the analysis at 99.4%, an increase of 26 percentage points from 73.4% a year earlier. Losses in Hawaii from the Maui wildfires helped push the insurer's loss ratio to its highest level in any three-month period since the fourth quarter of 2017." — S&P Global Market Intelligence: State Farm keeps homeowners market share lead in Q3 despite rise in loss ratio (December 2023) ↩
-
"State Farm General Insurance Company®, State Farm's provider of homeowners insurance in California, will cease accepting new applications including all business and personal lines property and casualty insurance, effective May 27, 2023… This decision was made due to historic increases in construction costs outpacing inflation, rapidly growing catastrophe exposure, and a challenging reinsurance market." — State Farm Newsroom: State Farm General Insurance Company®: California New Business Update (May 26, 2023) ↩
-
"State Farm General Insurance Co. said it will non-renew 30,000 California homeowners, rental dwelling, and other property insurance policies… These actions are California-specific and will occur on a rolling basis over the next year, beginning on July 3, 2024, for homeowners, rental dwelling, residential community association and business owners policies and on August 20, 2024, for commercial apartment policies." — Insurance Journal: State Farm Nonrenewing 30K California Homeowners, Renters (March 20, 2024) ↩
-
"Until recently, and for the past three decades since the passage of Proposition 103, insurers had only been allowed to use historical modeling, looking at the past 20 years' worth of climate data when justifying increases to customers' premiums… The proposed regulation will allow insurance companies to use forward-looking, so-called 'catastrophe modeling' when asking the California Department of Insurance (CDI) to approve requested rate hikes." — Coverage Cat: California's Proposition 103 ↩
-
"Aon pegged insured natural catastrophe losses at $118bn for 2023, with perils traditionally considered as primary or peak – tropical cyclone, earthquake and European windstorm – accounting for just 14 percent of the total." — Reinsurance News: Global insured losses from natural disasters hit $118bn in 2023: Aon (January 2024) ↩
