What happens when risk stops being something you can buy your way out of
Methodology note: this piece was written collaboratively between a human and an AI — the human providing the instincts, provocations, editorial judgement and voice; the AI providing research synthesis, intellectual scaffolding and drafting. Full method at [athomehefeelslikeatourist.blog].
Section One: Five Families
Most public argument about climate change never reaches the insurance question at all. It settles instead into two familiar postures. On one side, doomsterism — the sense that nothing short of civilisational transformation matters, which tends to produce paralysis rather than action. On the other, a quieter denial that isn’t about the science any more — it’s about the arithmetic, the claim that serious mitigation costs more than the country, the household, the government can bear. Both postures have the same effect: neither one asks the more specific and more answerable question of what happens to risk itself as the climate changes, and whether that risk can still be shared out, priced, and paid for the way risk has always been managed.
That’s the question this essay stays with. Not whether climate change is happening, not how frightened anyone should be about it. How the risk gets mitigated, and specifically, how much of it can still be insured — pooled, priced, and paid out on — and how much is starting to fall outside that mechanism altogether.
Here are five families. None of them is real. Each stands in for a much larger group of households in a broadly similar position — ranked not by wealth but by exposure. The Williams family, median household income £42,000, lives in Worcester, England. The Johnson family, median household income $80,000, lives in the foothills east of Sacramento, California. The Chen family, median urban household income around $15,000, lives in Guangzhou, on the Pearl River Delta. The Akhter family, median household income around $3,500, lives in Khulna district, Bangladesh. The Paeniu family, median household income around $5,000, lives in Funafuti, the capital of Tuvalu.
Each has a nominal climate exposure and a nominal ability to insure against it, and the gap between the two widens as you move down the list. The Williamses can still buy a buildings and contents policy, though flood exclusions are creeping into the small print in the postcodes that need cover most. The Johnsons can too, though California’s wildfire-exposed foothills are already finding that cover isn’t guaranteed at any price on the open market. The Chens sit inside a state-directed system where the question of private insurance barely arises in the same form. The Akhters have almost no access to formal insurance markets at all — the risk is real, the mechanism to price and pool it isn’t. The Paenius face a number that breaks the exercise entirely: a nation whose insurable value approaches zero against what a warming ocean puts at stake.
That gradient — from “insurance is a routine household expense” to “insurance is not a coherent concept here” — runs through what follows.
Section Two: Where We Stand Now
Where do we stand now? 2026 is on course to average around 1.51°C above pre-industrial levels — the first calendar year to sit above the 1.5°C figure that anchored a decade of climate diplomacy. Western Europe has just had its hottest June–July on record, 2.79°C above the historical average, beating the previous record set in 2022. July was the hottest month ever logged for the world’s non-polar oceans.
The mechanism is simple. Dry soil stops cooling the air above it, so heat and drought reinforce each other, and this year that loop was compounded by a developing El Niño on top of the underlying warming trend.
Fire follows heat. The fire weather index across Europe has risen by around 50% over the past 40 years, and the continent now sees two to three times as many extreme-heat days as before. Fire suppression resources are still deployed at the same level — they’re being asked to do a job they were no longer built for.
That’s the facts. What follows is about risk, pricing, and who pays for it.
Section Three: How Insurance Actually Works
Insurance works by turning an uncertain loss for one household into a certain, small cost for many households. A given house either burns down or it doesn’t, and for that one household the outcome is binary and expensive. Spread across ten thousand houses that face the same small annual chance of fire, the total loss to the pool becomes predictable even though each individual loss stays unpredictable. That predictability is what gets priced and sold as a premium.
This only works if the losses in the pool are independent of each other. A textbook example: two people each face a 20% chance of a $2,500 loss. Alone, each carries a standard deviation of risk around $1,000. Pooled together, that volatility drops to $707 — a meaningful reduction, achieved simply by combining two risks that have nothing to do with each other. Add thousands more independent policyholders and the effect compounds: the law of large numbers lets an insurer predict, with real confidence, what the whole pool will cost to cover. It sounds simple, and in principle it is. Though there is devil in the detail. Which is why actuaries make such a good living.
A hurricane, a flood, or a heatwave doesn’t behave like that. It doesn’t damage one house at random — it damages thousands of houses in the same region on the same day. The more the pool is drawn from a single flood plain, coastline, or heat corridor, the less diversification the pool actually delivers, because the losses inside it are correlated rather than independent. Adding more policyholders in that pool doesn’t reduce the insurer’s uncertainty — it increases the total amount that could all come due at once.
Insurers manage that correlated risk by passing it upward, to reinsurers. Reinsurance widens the pool again — spreading correlated regional risk across a global book of business, which restores some of the stabilising effect the local pool lost. This is why a flood in Norfolk can end up, several contractual layers removed, priced into a reinsurance treaty written in Zurich or Bermuda. It’s also the point at which climate change stops being a local underwriting problem and becomes a global capital-markets one: if enough regions are becoming correlated-risk regions at the same time, reinsurers are absorbing the same concentration problem the local insurer was trying to escape.
Climate change attacks this system at both ends of the chain. It reduces geographic diversification — the central limit theorem effectively stops working when losses are correlated across whole regions rather than independent within them — and it raises genuine uncertainty about future frequency and severity, on top of the correlation problem. In the worst-affected cases, the calculated pure premium approaches infinity, and no price exists at which both sides of the transaction would agree to trade. That is the mechanical description of an uninsurable risk. Not “expensive.” Not “correlated.” A calculation that doesn’t converge on a number.
| Loss type | Independent or correlated | Typically insurable? |
|---|---|---|
| House fire | Independent | Yes — priced routinely |
| Car accident | Independent | Yes — priced routinely |
| Regional flood | Correlated (geographic) | Increasingly restricted or excluded |
| Coastal storm surge | Correlated (geographic) | Withdrawing in exposed markets |
| Small-island inundation | Fully correlated (whole territory) | No functioning market |
Section Four: What Has Already Happened
The theoretical break described in Section Three is already showing up in real underwriting decisions, not modelled projections. Several major US insurers have stopped renewing policies in the states hit hardest by climate risk — Allstate and State Farm both pulling back coverage in California, AAA declining to renew some Florida policies. When that happens, homeowners fall to state-run “insurer of last resort” programmes, which typically charge several times the premium of the lost policy while providing less cover.
Florida shows the pattern in its clearest form. The state’s home insurance nonrenewal rate reached roughly 3% in both 2023 and 2024 — triple the national average — and in three counties (Hendry, Okeechobee, Glades) the 2022 nonrenewal rate ran between 24% and 33%. Nonrenewal rates there have more than tripled since 2018.
California’s version is wildfire rather than flood, and it’s already required a state bailout. Losses from the Palisades fires pushed California’s own state-run insurer of last resort — itself designed to be the backstop — to need a $1 billion bailout.
The withdrawal isn’t confined to the states that make headlines. A Senate Budget Committee analysis found Oklahoma, driven by increasingly severe convective storms, ranks among the highest nonrenewal states in the country, alongside Nebraska, Ohio, South Dakota and Montana — places with no coastline and no wildfire season, dropped for hail and wind instead. Homeowners in the highest-risk 20% of US ZIP codes are already paying 82% more in premiums than those in the lowest-risk areas.
Britain’s version of the same mechanism runs through Flood Re, the industry-funded reinsurance pool set up in 2016 to keep flood cover affordable for high-risk homes. More than half a million UK homes have been covered through it since launch. But it was built as a bridge to a “risk-reflective” market, not a permanent solution, and it’s due to wind down in 2039, currently running on £3.2 billion of reinsurance capacity and increasingly reliant on capital markets to replenish itself. It seems risks are rising quicker than the industry’s capacity to reflect them.
The scheme also has a hard edge that produces a UK version of the American withdrawal pattern — and the pool of homes on the wrong side of that edge is still growing. It doesn’t cover homes built since 2009 and excludes blocks of more than three flats, which means new development in known flood-risk areas — and there’s a lot of it — enters the market already outside the safety net. One in nine new homes built in England between 2022 and 2024 went up in an area of medium or high flood risk, and planning permission has been granted for more than 7,000 homes in England’s highest flood-risk areas alone, with the Environment Agency recording 267 cases in a single year where permission was granted directly against its own advice. Nothing in the planning system stops this pool from growing. Every year adds more homes to the excluded side of the scheme, at the same time as the scheme itself is running down its own reserves.
Bank of England staff modelling published in May 2026 estimated that under a severe climate scenario, the share of UK mortgage holders without flood cover could rise from around 5% today to 7–10% by 2050 — as many as 910,000 households — and to around 16%, or 1.39 million households, in the year after a severe flood pushes premiums up sharply.
The mechanism that turns an uninsured home into a genuine household crisis is the mortgage, not the flood. A house that can’t be insured can’t be mortgaged, and a house that can’t be mortgaged can’t be sold at anything like its expected price. Bayes Business School research already finds homes at flood risk selling at an 8% discount, rising to 32% for the highest-risk properties. That is the mechanism working exactly as Section Three described it, priced directly into a family’s largest asset before the water has even arrived.
None of the numbers in this section look catastrophic taken individually, and that is precisely the trap. A 3% nonrenewal rate, a 5% mortgage protection gap, one in nine new homes — a reasonable reader can look at any of these and conclude it isn’t them, so there’s nothing urgent to think about. But these figures describe risk in the technical sense used in Section Three — a known probability distribution, the kind that can be priced. What actually sits underneath them is uncertainty in the older, stricter sense: nobody knows the shape of the distribution a decade out, only its direction. And on every measure in this section, the direction has only ever moved one way. A gradient that is currently mild and moving in a single direction is not a reason for calm. It’s a description of where the mild part ends.
| Market | What’s happened | Source of pressure |
|---|---|---|
| Florida | Nonrenewal rate ~3% statewide, up to 33% in worst counties | Sea level rise, hurricanes, flooding |
| California | Major insurers withdrawing; state backstop needed $1bn bailout | Wildfire |
| Oklahoma, Nebraska, Montana, others | Rising nonrenewals despite no coastal/wildfire exposure | Severe convective storms, hail |
| UK (Flood Re) | Half a million homes covered, but scheme time-limited to 2039, excludes post-2009 builds, and the excluded pool keeps growing | Flood, with mortgage-market spillover already modelled by the Bank of England |
Section Five: What Might Happen Next
One trajectory is already well underway: the risk is leaving insurers’ own balance sheets and moving to the capital markets. The global insurance-linked securities market reached $120 billion at year-end 2025, up from $107 billion the year before, and the catastrophe bond segment alone hit $63.9 billion outstanding by the end of Q1 2026, a new record. A cat bond is, in substance, a bet that a named disaster won’t happen — hedge funds and pension funds take on several years of catastrophe risk in exchange for high returns, and if the disaster does happen, some or all of their investment covers the claim instead of theirs. This is the reinsurance mechanism from Section Three, taken one step further: instead of spreading correlated regional risk across a global insurance book, it’s spread across the entire global capital markets, where a Californian wildfire and a Bermudan hurricane are just two more uncorrelated line items in someone’s diversified portfolio.
This solves a genuine problem, and it solves it for the wrong tier of the system. Insured losses from natural disasters are on track to exceed $100 billion globally in 2026 — the seventh consecutive year that’s happened — and cat bonds are demonstrably good at keeping that scale of loss from sinking individual reinsurers. What they don’t do, and aren’t built to do, is put a policy back in the hands of the Florida homeowner who’s just been dropped, or the Somerset household outside Flood Re’s 2009 cut-off. The capital is flowing toward the instrument that professional investors can price and trade, not toward the household risk that started the whole chain. Even inside the industry, the question of whether cat bond growth will ever translate into affordability relief for ordinary policyholders is openly unresolved — the benefits are real, but they may stop at the reinsurance layer and never reach the primary market.
A second, smaller-scale mechanism is closer to reaching the household directly: parametric insurance. Rather than paying out against an assessed loss — the traditional, and slow, claims process — a parametric policy pays a fixed amount the moment a defined trigger occurs: a windspeed above a threshold, rainfall above a threshold, a temperature above a threshold. It sidesteps the correlated-loss problem in one way (payout doesn’t depend on individually assessing thousands of claims after the same event) without solving it in another (the total exposure across a region is still the same total exposure). It’s growing fastest in agriculture and in developing-economy contexts precisely because it doesn’t require the dense claims infrastructure that indemnity insurance does — which makes it a plausible route into markets, like Bangladesh’s, that formal insurance has never really reached.
A third possibility is pricing adaptation. Nothing in the current model rewards a household for having done the things that reduce its actual exposure — solar generation that keeps the lights on through a grid failure, insulation that blunts a heatwave, drainage and flood defences installed at the household’s own cost. The data to do this now exists in a way it didn’t a decade ago: smart meters, connected batteries, and flood-sensor networks all generate exactly the kind of granular, verifiable evidence an actuary would need to write a lower premium for a genuinely lower-risk property. Energy providers are already circling the adjacent version of this problem — treating home batteries and EV chargers as assets the grid can call on at moments of peak demand, in exchange for payment. There’s no equivalent yet on the insurance side. The technical building blocks for one are already sitting in most of the smart homes being built today.
None of these three futures resolves the underlying trade. Capital markets are absorbing the correlated risk that indemnity insurers can no longer hold alone, but at the wrong tier. Parametric products are reaching new markets, but without changing the total amount of risk in them. Adaptation pricing is technically possible, but nobody in the chain — insurer, energy provider, or state — has yet built the model that pays a household back for reducing its own exposure. Each is a plausible part of an answer. None of them, on its own or together, is a plan.
Section Six: The Timing Gap
And any plan has a fundamental problem. Every actor in this system is working from a different clock, and none of them is the clock the physical risk is actually running on.
The household’s clock runs short, for reasons that have nothing to do with information. A family buying a home weighs a mortgage against a monthly premium that looks, this year, entirely reasonable — because it is entirely reasonable, priced against this year’s risk. The 25- or 30-year horizon of the mortgage itself is rarely the horizon anyone actually reasons about at the point of purchase. This isn’t a failure of intelligence or diligence. It’s how humans weigh a certain, immediate cost against an uncertain, distant one — the same discounting instinct that makes a cigarette today feel cheaper than a decade of lung capacity, applied to a house instead of a habit.
The insurer’s clock runs backward, and in some markets that isn’t a habit but a legal requirement. California’s Proposition 103 constrained insurers to price against historical, retrospective loss data rather than forward-looking models, until reforms in 2026 finally allowed catastrophe models that price for expected future loss rather than only what’s already happened. An actuary is, by training and often by law, the last person in the chain permitted to act on where the risk is going rather than where it’s been. Section Three’s law of large numbers was built for a stable world, where the past is a reasonable guide to the future. Nothing about how the model is built adapts automatically to a world where that stops being true.
The state’s clock runs on an electoral cycle, and the costs and the decisions sit in different terms. A council granting planning permission for a flood-zone development today isn’t in office in 2050, when the actuarial and mortality data catches up with the decision. 267 homes were granted planning permission in a single year directly against the Environment Agency’s own advice — not because nobody knew, but because the person authorising the permission and the person living with the consequence are never the same person, or even close in time.
Each of these clocks is individually defensible, and together they guarantee the same outcome: repricing arrives after the fact, not before it. The household discounts a future it can’t fully price. The insurer is constrained, by habit or by regulation, to price the past. The state authorises today what only becomes a liability decades out. Nobody in the sequence is lying, and nobody is obviously behaving irrationally, given the incentives they’re actually facing. But the physical risk itself doesn’t wait for any of these clocks to synchronise. It moves on its own schedule, and every actor in the chain finds out how far behind they were only once the gap has already opened into a real, priced, uninsurable loss.
This is the specific point at which a market-failure story stops being adequate. It isn’t that the market failed to notice climate risk — the cat bond growth in Section Five proves the opposite, that sophisticated capital notices and prices this risk quickly and well. It’s that the market prices it in a different place, and on a different timescale, than the one where a household actually lives inside the consequence. The information exists. The mechanism to act on it, at the point where it would still matter to the family signing the mortgage, doesn’t.
Section Seven: Who Fills the Gap
The honest answer is: nobody, on a durable basis. Only stopgaps, and every stopgap already described here is under strain or built with an expiry date.
Private capital won’t do it, because it’s already shown what it does instead. It withdraws from the household risk it can no longer price with confidence, and it redirects toward the reinsurance-layer instruments it can price intelligibly. That’s not a moral failure. It’s the system working exactly as a capital market is supposed to work — capital goes where the risk-adjusted return makes sense, and household flood insurance in a repeatedly-flooded postcode increasingly doesn’t.
The state is doing it, currently, almost everywhere — but by default, not by design, and always as a bridge rather than a destination. Flood Re winds down in 2039 on reserves that are already under pressure. California’s FAIR Plan — funded by the state’s admitted insurers, not the taxpayer directly, but a backstop all the same — now covers over 668,000 homes, its exposure up more than 52% in a single year, and it has already needed a $1 billion bailout after the Palisades and Eaton fires. Florida runs the same structure under a different name. None of these was designed as a permanent home insurance system. Each was designed as the place risk lands while the market “transitions” to something more sustainable. None of them has a transition plan. They’re simply where the risk goes once the private market declines it.
Further down the exposure gradient, even that stopgap barely exists. International climate finance — the “loss and damage” funding agreed in principle at COP28 in Dubai — is the only candidate for a family like the Akhters or the Paenius, and it operates at the scale of a national fund, not a household policy. It is not, and was never designed to be, the equivalent of a FAIR Plan quote landing in someone’s inbox.
So the actual formulation isn’t “the state fills the gap.” It’s “the state currently absorbs the gap, without closing it.” The cost doesn’t disappear when a market withdraws. It moves — from a household’s premium to a taxpayer’s levy, from an insurer’s balance sheet to a government’s contingent liability — and it sits there, waiting for the moment the levy or the balance sheet turns out to be smaller than the risk it was quietly asked to hold.
Section Eight: Back to the Families
It’s tempting to let all of this stay abstract, if none of it looks like it touches you yet. Every government reading this essay’s home country — the US, the UK, the rest of Europe, everywhere — is already carrying a long list of expensive headaches, and it’s a reasonable question whether insurance markets deserve to be near the top of it. Whether to keep relying on short-term fixes that themselves add emissions. How fast, and at what cost, to actually get to a genuinely lower-carbon economy. Whether some moonshot technology solves more of this than policy ever could. What any of it means for employment, in the places where it changes fastest. And behind all of that, the larger questions this essay hasn’t gone near — food supply, water, migration — that can crowd out a conversation about insurance premiums entirely, reasonably, because they are in most tellings the bigger story.
None of this is a story about sudden collapse. It’s a slower thing than that — a long, uneven negotiation between a rising threat and a set of institutions trying, with mixed success, to adjust to it. Insurance sits well down most people’s list of things to think about in that negotiation, most of the time. But whether or not it’s top of mind, it is quietly billing someone. And for some of the five families it introduced, it has already stopped being a live question at all — not because the risk went away, but because the point at which it could still have been managed has already passed.
Here’s what that costs, in the terms each family actually experiences it.
| Family | What insurance now costs | What’s happening to it | Global mitigation premium (2% of income) |
|---|---|---|---|
| Williams | Flood Re–subsidised premium today, roughly halved vs. unassisted flood-risk pricing | Scheme winds down 2039; new-build homes since 2009 already excluded | £840/year |
| Johnson | ~$3,000/year on California’s FAIR Plan (vs. ~$1,571 on the open market where still available) — up to $32,000/year in the highest-risk ZIP codes | FAIR Plan exposure up 52% in a year; 29% rate rise approved for late 2026 | $1,600/year |
| Chen | Minimal formal private market; risk absorbed through state-directed infrastructure spending rather than household premiums | Model depends on continued state capacity, not on a household purchasing decision | $300/year |
| Akhter | No meaningful formal insurance market at household level | Risk uncovered by any mechanism except (limited, contested) international climate finance | $70/year |
| Paeniu | No functioning market; the concept doesn’t resolve to a number | The question isn’t the price of cover — it’s whether “elsewhere” is a place that exists yet | $100/year |
Section Nine: The Alternative
There’s a second way to think about the number in that last column, and it isn’t insurance in the sense this essay has used the word so far. Everything in Sections Three through Eight describes a market mechanism — individual households, individual properties, individual premiums, priced one policy at a time by an institution trying to decide whether it can still afford to take the risk on. The 2% figure works on a completely different logic: not a household buying cover against its own loss, but every household paying a fixed, modest share of its income into a single global pool, sized to fund mitigation and adaptation before the losses happen rather than compensation after they do.
The distinction that matters is where the money goes. A market premium has to cover the insurer’s expected payout, its capital reserve, its cost of doing business, and a return for the capital sitting behind it — Section Three’s mechanics, plus a margin. A pooling arrangement built for mitigation doesn’t need to fund a return for anyone. There’s no shareholder on the other side of the transaction. Every pound or dollar collected goes toward the thing the money is actually for — sea walls, flood defences, grid resilience, agricultural adaptation — rather than a fraction of it being retained as the price of transferring the risk to someone willing to hold it.
At 2% of household income, the annual cost for each of the five families looks like this: £840 for the Williamses, $1,600 for the Johnsons, around $300 for the Chens, around $70 for the Akhters, around $100 for the Paenius. The same percentage, applied to five very different incomes, produces five very different sums — which is itself the point. The burden scales with what a household actually has. Nobody’s asked to pay a Sacramento premium on a Khulna income.
Set against what Sections Four through Eight have already shown, this is not a large number. The Johnsons are already paying close to that figure for FAIR Plan cover that only protects against fire, not flood, theft, or liability. The Williamses are already inside a scheme, funded by an industry-wide levy, that does something structurally similar to this at a national scale — it just does it for flood risk alone, and only until 2039.
And the alternative to paying it is not a lower cost. It’s the same cost, deferred, and very likely larger by the time it arrives. For the Johnsons, the inaction path this essay’s own figures already describe: a FAIR Plan whose exposure grew 52% in a single year, rate increases running well ahead of inflation, and a market that keeps contracting rather than stabilising. For the Williamses, a scheme with a fixed end date and a growing pool of homes it was never built to cover. For the Chens, water stress across a plain that fills a very large number of dinner tables, arriving on a timeline nobody in Section Six’s chain is actually pricing against. For the Akhters, displacement on a scale — potentially one in seven people in Bangladesh by 2050 — that no premium, at 2% or any other percentage, converts back into a place to live once the delta itself has gone. For the Paenius, the same fact this essay opened with: a nation whose insurable value approaches zero, because there may be very little nation left to insure.
The 2% figure was never really a comparison between two prices. It’s a comparison between paying something now, while the thing being paid for still has a chance of working, and paying considerably more later, for a repair that in at least one case on this list may not be technically possible at all.
Coda: Climate isn’t local
Of course, you might say none of this is really about you. You live in Harrogate, or somewhere similarly inland and unremarkable on a flood map. Or you live in Vermont, a state whose winters people still complain are too cold. The current government talks about green energy, so perhaps that’s already handled. There’s enough to think about already — a mortgage, a job, a family — without adding a risk pool on the far side of the world to the list. And you couldn’t point to Tuvalu on a map, if you’re honest about it.
All of that is reasonable. Most of it is true, as far as it goes. People are entitled to a limited amount of attention, and climate risk has to compete for it against everything else a life actually contains.
But it doesn’t hold for everyone right now, and one day it may not hold for you. Some of the people this essay has been about are not choosing not to worry. They’re past the point where worrying would have changed anything.
There are solutions. None of them is free. One way or another, the cost gets paid — by a premium, by a levy, by a government balance sheet, or by a family with nowhere left to live. The only real choice is which of those it turns out to be.
Climate isn’t local. And you can’t afford to be.
At Home He Feels Like A Tourist 
Leave a comment