A carrier that non-renews a wildfire-exposed book is not being cowardly. It is being solvent.
Reserves have to hold, the treaty has to renew, and rate adequacy in a wildland-urban interface ZIP code is not a question of nerve. Walking away from a book you cannot price is defensible underwriting, and anyone who has sat in a reserving meeting knows it.
The problem is what the exit does and does not do. It ends your obligation. It does not end the loss. The loss stays exactly where it was, on the same ridge, in the same channel, and it reappears on a line item nobody underwrote.
the exit is priced. the aftermath is not.
Non-renewal is the most precisely priced decision in property insurance. Everything downstream of it is not priced at all.
The pattern is not in dispute. The Senate Budget Committee's 2024 investigation collected county-level non-renewal data from 23 carriers representing roughly 65% of the U.S. homeowners market across 249 million policies, and found that in 2023, among counties with at least 10,000 policies in force, 48 of the 50 with the highest non-renewal rates were coastal, low-lying delta, or high wildfire risk. Non-renewal rates rose fastest in the counties most exposed to wildfire and hurricane risk between 2018 and 2023. Carriers are reading hazard correctly and acting on it.
Now follow the loss. Start with the residual market, because it is the clearest arithmetic in the industry. The California FAIR Plan — the state's insurer of last resort — reported $768 billion of total exposure as of June 2026, a 250% increase since September 2022, across 696,562 policies in force, up 157% over the same period. That growth is not new risk being created. It is voluntary-market risk changing hands.
Then the part underwriters tend to skip past: the FAIR Plan is funded by assessments on its member insurers, which is to say every admitted carrier writing property in California. In 2025 the California Department of Insurance authorized the plan to collect $1 billion from those insurers to shore up reserves, and allowed carriers to recoup up to half of the levy from their own policyholders statewide. Carriers that had already shrunk their exposure in the canyons still wrote that check.
| who holds the loss after the policy ends | what they hold |
|---|---|
| the owner | rebuild cost, plus an asset that is harder to sell and harder to finance |
| the lender | collateral whose insurability is a loan covenant |
| the residual market | the growth — $768 billion of FAIR Plan exposure by June 2026 |
| every admitted carrier in the state | assessment exposure to that residual market |
| the county | a tax base and bonding capacity tied to property values |
| the state and federal taxpayer | disaster aid, buyouts, and the suppression bill |
| the landscape | the same fuel load, the same channel, the same missing marsh |
The last row is the one that matters, because it is the row that determines all the others next season.
There is a second exposure that survives the exit, and it sits on the other side of the balance sheet. Insurers are among the largest institutional asset owners in the world. The mortgage paper, municipal debt, and real assets in the general account are written on the same coastlines and the same fire-prone counties the underwriting side just left. You can non-renew a homeowner. You cannot non-renew a watershed you also hold bonds against.
climate risk reduction is the only line that shrinks the loss
Climate risk reduction is spending that changes the physical outcome of the next event — fuel load, channel capacity, wetland storage, dune volume, canopy cover — rather than changing who pays for it. That distinction is the whole argument. Most of what the underwriting side currently calls climate action is reassignment.
| the move | what it changes | what it does to the hazard |
|---|---|---|
| reprice | premium adequacy on your book | nothing |
| tighten terms | your share of a given loss | nothing |
| non-renew | who holds the exposure | nothing — it relocates it |
| reduce the hazard | the physical outcome of the next event | the only one of the four that lowers expected loss |
The evidence base for the fourth row is unglamorous and reasonably good. Swiss Re Institute's 2026 sigma report put 2025 insured natural catastrophe losses at $107 billion against $220 billion of economic losses — a record 49% insured share, which still means more than half the loss was uninsured — and estimated the global natural catastrophe protection gap at $424 billion, up from $395 billion a year earlier. In the same work, Swiss Re found catastrophe adaptation projects carry a median projected benefit-cost ratio of 1.86. The National Institute of Building Sciences, in its 2019 Natural Hazard Mitigation Saves study, found $11 saved per $1 invested in adopting current model building codes, $6 per $1 for federal mitigation grants, and $4 per $1 for above-code design.
Read those numbers honestly. They are ratios for hazard mitigation at building, parcel, and program scale — not a promised return on any financial instrument, ours included. What they establish is narrow and useful: reduction is the only category of spend where the money comes back through the loss column rather than through someone else's premium.
Repricing changes the invoice. Reduction changes the event. Only one of those shows up in next year's loss ratio as anything other than a smaller book.
the ridge belongs to nobody's premium
Here is why reduction is hard, and it is not a modeling problem.
Parcel-scale mitigation has an owner and a payer. Defensible space, a Class A roof, a raised first floor, a floodproofed loading dock — one property, one decision-maker, one premium credit. The industry knows how to price that, and codes and rating plans increasingly do.
The landscape-scale determinants have neither. The ridge above the subdivision, the forty miles of channel upstream of the distribution center, the marsh in front of the terminal, the wet meadow that decides whether a fire arrives as a ground burn or a crown run — these set the loss for hundreds of policies at once, and no single policyholder can buy them. No premium prices them. No claim reaches them. They are boundary conditions in your catastrophe model, not variables, which is a polite way of saying they are somebody else's problem until they are everybody's loss.
That is a coordination failure, not an actuarial one. Our existing insurer writing takes it apart from three angles, and this post does not repeat them: there is no risk transfer on what the word "transfer" actually moves, insurability is the first domino on why the exit signal arrives before the price signal, and the insurer investment that actually reduces losses on the investment case itself.
what it takes to fund a ridge
Ensurance funds the present condition of a named place — before a trigger, without a claim. That is the entire structural difference from the products already on your desk.
Insurance is ex post: it pays after a loss, and the business model requires the event. Parametric covers are faster but still ex post: a defined trigger, then liquidity — worth having, and worth being clear-eyed about, which is the argument in pre-arranged reactionary finance. Carbon and biodiversity credits are ex ante: they price a counterfactual about a future that has not happened. Ensurance is ex nunc — from now. It funds a system that is presently producing the services you are exposed to, and reads its condition as the thing measured.
In plain instrument terms: a certificate funds one named place and its present condition directly. A coin funds protection across the protocol, indirectly, through market activity. Neither is a policy. Neither carries a limit, a trigger, an indemnity, or a claims process, and calling them insurance would be both wrong and unhelpful. They belong on the loss-prevention line, beside fuel treatment contracts and floodplain easements — with the difference that the funding is continuous rather than grant-cycled, and the unit of account is the condition of a place rather than the occurrence of an event.
For the fiduciary version of the point: capital does not have to care about the ridge. The architecture does the caring. An instrument built so its return depends on a functioning watershed routes money to the watershed whether or not the allocator has an opinion about wet meadows.
And our own stage, stated plainly: the agents, coins, and certificates are live onchain with real but small volumes. We are not a market you can allocate a reinsurance program into. We are a mechanism you can fund one geography through, and the honest version of that assessment is in is nature an asset class yet.
if you want the operational versions
These already exist. Start where your book hurts.
- wildfire insurance non-renewal: what to do — the policyholder-side playbook your non-renewed insureds are searching for
- flood insurance is disappearing — here's how to become insurable — the same problem on water
- if you could pay for the storm to lose its teeth, would you? — the reduction question posed to the people who pay for the loss
- from liability to asset: how insurers are navigating the CWRC — what happens when a state writes a resilience code
- nature finance has a timing problem — the timing argument in protocol language
one decision
Not a new product. Not a brochure. One question, and it is answerable in a single meeting:
Name one geography where you are still writing and intend to keep writing. Then look at what the landscape-scale hazard in that geography actually costs to change, who else is exposed to it, and what a funded condition looks like over a three-to-five-year horizon rather than a policy term.
The insurer view is written in loss ratios, exposure concentration, and TNFD reporting language, not in ecology. If the geography is real, start the conversation and bring the map.
The book can walk away. It will still get the bill. The only question left is whether it pays for a smaller event or a larger assessment.