The Sunset That Will Not Set

In the spring of 2016, something remarkable happened to the price of a wet house.

For years, a home that had flooded carried its history around like a limp. The average insurance quote for a property with a flood claim stood near £4,400 — where a quote could be had at all — and the housing market, which is not sentimental, priced accordingly: the Bank of England later measured the discount on flooded properties at 1.6 per cent against dry neighbours on the same street. The discount was not a cruelty. It was a sentence, in both senses — a judgment, and a piece of information. It said: water has been here, water knows the way back, and somebody should think about that before exchanging contracts.

Then Flood Re switched on, the average quote fell to £1,316, and the Bank of England — in one of the most quietly devastating staff papers of the decade, 1.75 million property transactions matched against Environment Agency flood maps — recorded what happened next. The 1.6 per cent discount did not merely shrink. It reversed, to a premium of 1.8 per cent. Transaction volumes for previously flooded homes rose. Buyers, relieved of insurance anxiety, paid an average of £4,083 more per flooded property than the pre-scheme market would have charged them. The paper’s authors, constrained by the good manners of central banking, put it like this: the scheme “completely offsets the negative pricing effect of flood risk, irrespective of the risk measure we employ.”

Read that conclusion the way an actuary reads it. Britain did not merely subsidise flood insurance in 2016. Britain instructed its housing market to stop telling the truth about water — and the market, ever obedient to a subsidy, complied. Over the scheme’s lifetime the wealth transferred into flood-zone property values is estimated at £4.4 billion. That is not the cost of the scheme. That is the size of the lie — the gap between what these houses are worth with the truth suppressed and what they would be worth with it spoken.

A machine for shredding messages

It is worth pausing on what an insurance premium actually is, because the people who design schemes like this appear never to have done so. A premium is a message. Into it flows everything the modern world knows about a property’s relationship with water — claims histories, hydrological models, river gauge data, climate projections, the surveyor’s note about the airbricks — compressed by competition into a single number and delivered, once a year, to the one person who can act on it: the owner. Hayek made his whole reputation on the observation that prices are how dispersed knowledge reaches the people who need it. An actuarially priced flood premium is dispersed knowledge with a direct debit attached.

Flood Re is a machine for intercepting that message and replacing it with a different one. Under the scheme, an insurer may cede the flood-risk portion of a household policy to a state-established reinsurer at a fixed tariff — and the tariff is set not by flood risk but by council tax band. A Band D house that floods one year in ten and a Band D house that floods one year in a thousand pay the identical £263. The banding rests, for connoisseurs of British administrative comedy, on property valuations conducted in 1991 and never updated since — so the risk signal has been replaced by a wealth signal, and the wealth signal is thirty-five years out of date. The pool is topped up by a levy on every household insurer in the country, currently £160 million a year, which lands on every home policy in Britain at roughly £10.50 apiece. If you insure a flat on a hill in Halifax, you are paying it. The scheme’s chief executive, Perry Thomas, described the resulting flow with a candour his board must occasionally regret: subsidies run “from poorer people in the north to some of Britain’s richest boroughs in the country’s south.”

He was not exaggerating. Band H properties — the top of the 1991 tree, the rectories and riverside villas — make up 0.6 per cent of Britain’s housing stock and roughly 12 per cent of Flood Re’s book: a twentyfold overrepresentation. These are households paying around £1,000 a year in premiums and, in Thomas’s own words, making “claims in the millions of pounds.” One claim in eight is for a property that has flooded before. The scheme has lately raised the top-band tariffs to their regulatory ceiling, which is to its credit; even at the ceiling, £1,613 for a repeatedly inundated manor on the Thames is not a price. It is a tip.

Now, honesty compels the acknowledgment that Flood Re was not conjured by fools, and the problem it answered was real. Before 2016, thousands of households faced quotes they could not pay or no quotes at all — homes that could not be insured could not be mortgaged, could not be sold, could not be left. The Joseph Rowntree Foundation’s analysis warned, plausibly, of “extensive social blight.” There is even a respectable formal model finding that a substantial flood-insurance subsidy — around 46 per cent — can be justified on ordinary welfare grounds. The case for helping was and is sound. But observe what was actually chosen. A society that wishes to help people in wet houses can send them money — visibly, on a budget line, where Parliament can see it and voters can weigh it. What Britain chose instead was to help them by falsifying the price — to bury the assistance inside the insurance system, where it distorts every decision downstream: what buyers pay, what lenders advance, what developers build, and what nobody bothers to defend against. The first method is a transfer. The second is a transfer wearing a blindfold, handed to the market as fact.

The safeguard that was supposed to hold

The designers, to be fair, saw some of this coming, and built three safeguards no comparable scheme has ever had: a statutory sunset in 2039, written into the Water Act 2014; an exclusion for homes built after 1 January 2009, so that the subsidy could not summon new houses onto the floodplain; and, latterly, a published transition plan. These are real. They deserve to be weighed. So let us weigh the second one against what the planning system has been doing while Flood Re held its umbrella over the existing stock.

Between 2022 and 2024, 11 per cent of the roughly 397,000 new homes built in England — about 44,000 of them — went up in areas of medium or high flood risk, a share rising from the prior decade’s 8 per cent. On current trends, some 115,000 of the government’s promised 1.5 million new homes will stand in the highest-risk zones. The Environment Agency, the body that actually knows where the water goes, is a statutory consultee with no veto: since 2016, nearly three thousand homes have been approved over its explicit objections. The planning framework’s “sequential test,” which directs development away from the wettest land, turns out to be guidance rather than law — and following the Mead Realisations judgment, failing it is merely a factor to be “weighed.” At Yatton the inspector acknowledged the test was failed and approved the scheme anyway; at Faversham the test was not conducted at all. In December 2024 the framework’s refusal threshold was softened from “clear reason” to “strong reason,” a one-word amendment that tells you everything about the direction of travel. And a Defra review found that more than half of planning authorities “rarely or never” inspect completed developments for compliance with the flood conditions attached to their consents. The safeguard, in short, was a turnstile bolted to an open field — and Aviva’s chief executive said as much this February: the post-2009 exclusion as a deterrent to floodplain building “simply hasn’t happened.”

Here the scheme’s architecture produces its cruellest joke, and it is worth savouring the mechanism. Those 44,000 new homes are excluded from Flood Re. Their buyers — young families, mostly, buying new-builds off plans that did not mention the river — will meet the full actuarial price of their postcode at precisely the moment the scheme’s existence has taught the entire market that flood risk is not something house prices need to reflect. The subsidy anaesthetised the market’s risk perception; the exclusion ensures the anaesthetic will wear off on the operating table, one completion at a time. A developer, meanwhile, sells the house long before the water arrives and bears none of this — the cleanest principal-agent problem you will ever see with planning permission.

The American tutorial

If you wish to know how this story ends, you do not need a model. You need a passport. The United States has been running the experiment since 1968, at full scale, with the results published annually by its own government auditors.

The National Flood Insurance Program was designed — the word matters — as a transition. The 1966 task force that conceived it recommended subsidised premiums for existing homes and actuarial rates for new ones, with the subsidies expected to phase out by attrition within roughly twenty-five years. Every element of that sentence should sound familiar. Fifty-eight years later, the programme has borrowed $36.5 billion from the US Treasury, still owes $22.5 billion, pays over $280 million a year in interest alone, and is described by its own administrator, FEMA, as never having generated a profit. The academic record confirms it did exactly what a suppressed price signal must do: peer-reviewed studies find the programme measurably drove migration toward higher-risk land, and a county-level analysis found flood subsidies had summoned 8 per cent more houses into harm’s way. Today, 2.5 per cent of its policies generate 48 per cent of its claims by value — a quarter of a million “repetitive loss” properties, rebuilt and reflooded on the public dime, America’s answer to Band H.

And when America tried to stop? This is the part of the tutorial to read twice. In 2012, Congress passed the Biggert-Waters Act with near-unanimous bipartisan support, phasing out the subsidies at last. Then the phase-out produced actual numbers on actual renewal notices — premiums in some cases rising from $1,000 to $31,500 — and in 2014, with equally impressive bipartisan speed, Congress reversed itself. The cycle ran again a decade later: an actuarial repricing called Risk Rating 2.0 was implemented in 2021, and by 2025 senators were queueing to demand its abolition. Economists have a name for this, coined by James Buchanan: the Samaritan’s dilemma. A state cannot credibly promise to withdraw help from concentrated, visible, sympathetic losers, because the losers vote and the diffuse payers do not notice. The commitment to end the subsidy is announced in the era of abstraction and abandoned in the era of renewal notices — every time, in every legislature, because the incentive structure is the legislature.

Against this, Flood Re’s defenders hold up the Water Act’s statutory sunset: 2039 is the law, and repealing it needs primary legislation. So it does. But consider what has been quietly assembled on the other side of the ledger. Ceding to Flood Re grew 20 per cent in 2024/25 — dependence deepening, thirteen years from a transition that requires it to be withering. The properties at risk in England were re-counted in 2024 at 6.3 million — one in five homes — heading for 8 million by mid-century as the climate does what it has been advertising. Uptake of the scheme’s own resilience programme runs at one household in three offered. The scheme’s chair wrote, in the current annual report, that if Flood Re left the market today “many households currently supported by the Scheme would face significant increases to their insurance premiums — if they were able to obtain cover at all.” And the scheme’s public framing of 2039 has already migrated from a date to a condition: exit “hinges on faster risk reduction.” Note the manoeuvre, because you will hear it again every year until 2039. A deadline that hinges is not a deadline. It is an opening position. The rhetorical scaffolding for the extension is not being built; it stands complete, painted, and awaiting only the ribbon — and the ribbon will be cut by whichever housing minister is in office when the first re-pricing letters land in marginal constituencies. Parliament in 2038, staring at several hundred thousand households facing the restoration of true prices in a single year, will be invited to choose between an act of arithmetic and an act of mercy. Consult the American precedent — twice around the same circuit — for how legislatures choose.

Thirteen years of compound interest on a falsehood

What makes the 2039 question graver than a mere budget fight is what has been accumulating underneath the administered price. Every year of suppression adds buyers who paid the inflated price, lenders who advanced against it, and owners who skipped the flood door because the premium never asked them to fit one. The Bank of England’s arithmetic implies a housing stock carrying billions of pounds of value that exists only while the subsidy exists — assets alive in the way a patient on a ventilator is alive, mortgaged at valuations the first honest premium would embarrass. The scheme’s own chief executive has warned that lenders’ working assumption about all this is that “they don’t need to do anything.” The banks, in other words, have looked at a statutory sunset backed by the full faith and credit of the British legislature, and priced it at approximately nothing. On the evidence assembled above, one struggles to call them irrational.

So the options in 2039 are three, and it is not too early to say them plainly. Let the prices speak, and accept a sharp, concentrated, politically combustible correction that lands on whoever happens to be holding the deeds — many of whom bought in good faith at prices the state’s own scheme inflated. Extend, and become the NFIP with better manners: the temporary made permanent, the liability compounding, the floodplain filling, the honest houses on the hill in Halifax paying their £10.50 forever. Or the likeliest course — a fudge, a “reformed” scheme, a taper that hinges, a sunset moved politely over the horizon in the last mile, as sunsets in this genre invariably are.

There remains one party to these negotiations who has not been introduced, because she requires no introduction and accepts no terms. The river was there before the Water Act and will be there after it. She has read neither the transition plan nor the council tax banding of 1991; she is unmoved by the sequential test, indifferent to the planning balance, and has never once responded to consultation. Britain has spent a decade adjusting every number in this story except hers — the premium, the house price, the refusal threshold, the sunset — and hers is the only number that was ever real. The properties at risk stand at 6.3 million and rising. She is patient, the river. She can wait until 2039.

She may look in before then.

Principal sources: Garbarino, Guin & Lee, Bank of England Staff Working Paper No. 995 (2022; Journal of Risk and Insurance, 2024); Flood Re Annual Report 2025/26 and Transition Plan (2023); Aviva new-build flood analysis (Feb 2026); Environment Agency NaFRA2 (2024); Climate Change Committee, Progress in Adapting to Climate Change (2025); TCPA on the sequential test (2025); GAO-23-105977 and GAO testimony (2026) on the NFIP; Cato Policy Analysis No. 923 on the NFIP’s history and the Samaritan’s dilemma; Kousky & Kunreuther on subsidy-induced development; Kornai on soft budget constraints; Hayek, “The Use of Knowledge in Society” (1945).