Hoping for normal: a register of the assumptions beneath national policy

Britain keeps a careful account of what could go wrong. The National Risk Register, published by the Cabinet Office and last updated in Jul 2026, catalogues pandemics, cyber attacks, floods, and dozens of other threats, each with a reasonable worst case.^[1] The Cabinet Office describes it as the external version of the classified National Security Risk Assessment beneath it.^[2] The register proposed here would complement that apparatus.

The gap is in what the apparatus stands on. Every risk in the register is assessed against a background of things the state believes without examining them. It believes its alliances hold, its markets supply, its infrastructure suits its weather, and the next emergency will resemble the last one. Those beliefs are assumptions, and the invalidation of any one of them produces the same thing a hazard produces.

Assumptions of this kind are documented, where they are documented at all, inside individual programmes and departments. Business cases carry assumption logs, defence planning has its own planning assumptions, and the Office for Budget Responsibility publishes the economic determinants behind its forecasts. Each of those sits inside a delivery silo and is tested against a project. At the level where the state's cross-cutting foundations sit, nothing is written down and nobody is responsible.

What the absence costs is time

An assumption nobody has written down is invisible, and invisible assumptions are the frame through which everything else gets read, so evidence that the world has changed arrives looking like turbulence inside the existing model. A tariff is a trade dispute. A buckled rail is a bad summer. A regulator's fine is a badly run company. Each event is absorbed as an exception to a rule still believed, the response runs late, and when the exceptions accumulate past the point of comfort the argument turns to denial and then to waiting for conditions to return to what they were.

The people doing this are not being stupid, and the diagnosis does not depend on anybody's failure of nerve. An official who declares a national assumption dead early carries the career risk alone, while an official who is wrong alongside everyone else carries none. Nobody owns the decision that an assumption has failed, so nobody makes it, and hoping is what the arrangement leaves.

Writing an assumption down changes what the same evidence looks like. A belief on paper, with the conditions of its own failure written beside it, stops being the frame and becomes a proposition that events can contradict. That is the whole mechanism. The death of a national assumption can often be seen coming years or decades ahead.

Six entries the register should have carried

Each entry below is written the way the register would write it: the assumption in one line, its failure mode, what rests on it, tests at two levels, and an owner. The two levels do different work. An assumption is at risk when named indicators show it under stress, and the response at that level is preparation, sized to what being wrong would cost. An assumption no longer holds when its terminal tests fire, and the response at that level is execution of what was prepared.

The United States anchors British security and the trading framework

Eighty years of British policy have rested on this line, and the failure mode is rupture. Defence planning assumes American reinforcement through the North Atlantic Treaty Organisation, intelligence arrangements assume American partnership, and the trading system Britain operates under is the one the United States designed and anchored. A register entry would have recorded that breadth and separated the limbs, because they can fail independently.

The trading limb went first, and the tests for it were behavioural. What matters about an anchor state is how it treats the rules it built. In May 2016 Washington blocked the reappointment of a sitting member of the World Trade Organization's Appellate Body on the substance of his rulings, the first time any member had done that to a judge who was not its own national.^[3] From 2017 the blocking became systematic and has been sustained by every administration since.^[4]

The terminal test fired on 2 Apr 2025. The United States imposed a universal 10 per cent tariff with country rates up to 50 per cent, allies included, under the International Emergency Economic Powers Act, a sanctions statute never used for tariffs before.^[5] The choice of instrument is the evidence. A state constrained by rules establishes that an action is lawful before taking it. The tariffs reversed that ordering, and the reversal was visible at once. The first suit arguing the statute authorises no tariffs at all was filed the following day, and a second within 12 days.^[6] Three courts agreed, ending with the Supreme Court on 20 Feb 2026.^[7] Later the same day the administration imposed a replacement surcharge under a different statute, and when the surcharge was struck down in turn in May 2026 the administration appealed.^[8]

The security limb has not failed. Treating the whole assumption as dead because its trading limb died would be the false declaration the register's critics predict, and separating the two limbs is what prevents it.

• At risk when. The anchor state acts against the system's rules in ways it would once have used the system to contest.
• No longer holds when. It acts against the framework itself, in a form no procedural reading explains.
• Status. Trading limb: no longer holds, from 2 Apr 2025. Security limb: at risk, since 2025.
• Owner. The National Security Adviser, with the trading limb delegated to the Permanent Secretary at the Department for Business and Trade.

The at-risk indicator was visible for nine years before the terminal test fired. Britain spent them waiting to see whether normal service would resume.

The rules-based trading system functions as a backstop

This assumption was built on while it was dying. From 2017 the case for leaving the European Union without an agreement rested on a fallback. Britain could trade on World Trade Organization terms, the argument ran, under rules with a court behind them. The Lancaster House position of Jan 2017, that no deal was better than a bad deal, worked only if those terms were a floor a country could stand on, and campaign economists made the same case at report length.^[9]

The floor was giving way while the argument was being made. The United States had been blocking appointments to the Appellate Body since 2016, and without interruption from 2017. On 11 Dec 2019 the body fell below quorum and stopped hearing appeals, seven weeks before Britain formally left the European Union and a year before the transition period ended.^[10] The backstop's enforcement mechanism ceased to exist before the country stepped onto it. A workaround now operates, the Multi-Party Interim Appeal Arbitration Arrangement, which Britain joined in Jun 2025 and of which the United States is not a member.^[11]

• At risk when. Appointments to the appellate mechanism are blocked, or its membership falls toward quorum.
• No longer holds when. The mechanism cannot hear an appeal.
• Status. No longer holds, from 11 Dec 2019.
• Owner. The Permanent Secretary at the Department for Business and Trade.

Nothing in this entry needed foresight. The evidence was public, contemporaneous, and reported at the time, and national policy was still being built on the assumption while its failure ran in plain view.

The next pandemic will look like influenza

This assumption was the wrong shape. The state held a pandemic assumption. The UK Influenza Pandemic Preparedness Strategy of 2011 was the operative national plan going into 2020, and Exercise Cygnus tested it in 2016 against a flu scenario, finding severe gaps in surge capacity and equipment.^[12] Non-influenza outbreaks were on the risk register of the period, in an emerging infectious diseases category rated at up to 100 deaths, beside a pandemic influenza scenario running to as many as 750,000.^[13] The risk was written down, and the weighting given to it was the assumption that shaped the preparation.

The COVID-19 Inquiry's Module 1 report of 18 Jul 2024 found that the United Kingdom prepared for the wrong pandemic and had placed too much weight on a single scenario.^[14] The Inquiry traced the persistence of that reliance partly to groupthink and the absence of structured challenge, and recommended moving away from single worst-case scenarios and institutionalising external red teams, which is an assumptions discipline arriving under another name.^[15] Countries whose planning had been reshaped by severe acute respiratory syndrome and Middle East respiratory syndrome, South Korea and Taiwan among them, carried a broader threat model into 2020 and moved faster in the early weeks.^[16]

• At risk when. A novel pathogen of a different transmission class causes a contained international outbreak.
• No longer holds when. Such a pathogen achieves sustained human transmission across borders.
• Status. At risk from 2003 and again from 2012. No longer holds, from Jan 2020.
• Owner. The Permanent Secretary at the Department of Health and Social Care, with the Government Chief Scientific Adviser.

The outbreaks of 2003 and 2012 were the at-risk indicator firing twice. Each firing would have obliged a documented answer to the question of what changes in British preparation if the next pandemic is not influenza. No part of the apparatus owned that question, and no answer was ever recorded.

Private ownership under regulation delivers investment in essential utilities

This assumption decayed from within, and the promise of 1989 decomposes into three parts a register would list separately, because each carries its own tests and they failed in sequence.

The first was that private equity capital would fund the investment. The water companies were floated essentially debt-free, with around £5bn of historic debt written off and a cash dowry of roughly £1.5bn, and Thames Water's gearing was in single figures at the start of the 1990s.^[17] Ownership models that secured debt against the asset ended equity financing in substance long before any visible service failure.

The second was that distributions would be paid from performance and not from borrowing. At sector level the arithmetic is the test. Water companies in England and Wales paid out roughly £78bn in dividends by 2023 while accumulating £60bn to £64bn of net debt from a zero-debt start.^[18] The industry answers with roughly £190bn of infrastructure spending over the same period, itself largely funded from bills and borrowing.^[19] Thames Water is the vivid instance, with £7.2bn paid in dividends between 1989 and 2022, against net debt of £19.77bn at the Jul 2026 results.^[20]

The third was that the regulator would protect the consumer interest. The Water Services Regulation Authority, Ofwat, acquired the power to block a dividend only in May 2023.^[21] Its finding that Thames Water had declared dividends without meaningful regard to performance for customers and the environment came provisionally in Dec 2024 and finally in May 2025, on payments made in Oct 2023 and Mar 2024, under a licence condition introduced months before.^[22]

• At risk when. Gearing rises above the level assumed in the price review, or distributions exceed free cash flow after capital investment.
• No longer holds when. Debt taken on across the sector approaches the total distributed, on a sustained basis.
• Status. No longer holds. The sector crossed that threshold during the 2000s, and no declaration followed.
• Owner. The Permanent Secretary at the Department for Environment, Food and Rural Affairs, on numbers published by the regulator.

Every figure above came from published company accounts, and the accounts were read. The National Audit Office, the Public Accounts Committee, and the Environment Committee returned to the water companies repeatedly for three decades. What they asked was whether the companies were well run and whether the regulator was effective, which are questions inside the model. Whether the model's founding belief had stopped being true was a different question, and it belonged to nobody. The consequences arrived together in the end, in a record £122.7m fine, a rescue bid abandoned by its bidder^[23], a creditor takeover awaiting a vote as this paper is published, and a government committed to abolishing the regulator following the Independent Water Commission.^[24][25]

British infrastructure operates within the climate it was designed for

This assumption erodes from outside, has no single day of death, and carries the most objective tests on the register because they are physical. Britain's railway rails are stressed to a neutral temperature of 27°C, chosen for the country's historical temperature range.^[26] On 18 and 19 Jul 2022 rail temperatures were measured at 62°C, air temperature reached 40.3°C for the first time on record, track buckled, and the East Coast Main Line closed.^[27] Summer 2025 was the warmest on record, and a summer of that kind is now roughly 70 times more likely than it would be in a climate without human-caused warming, a return period of one year in five against one in 340.^[28] Jun 2026 broke the all-time record for the month at 37.7°C under the country's first three-day red heat warning, with blanket speed restrictions and services suspended across multiple operators.^[29]

Water tells the same story. Sewers are conventionally designed for a one-in-30-year storm, a standard the 2025 national sustainable drainage standards flag as mismatched against the one-in-100-year benchmark used elsewhere in planning, while Met Office research finds extreme rainfall intensity rising 5 to 15 per cent with each degree of warming.^[30] On flood defences, the National Audit Office found in Nov 2023 that the Environment Agency had missed its asset condition target for five consecutive years, leaving more than 200,000 properties at avoidably increased risk, with the protection programme cut from 336,000 properties to 200,000.^[31]

• At risk when. Design thresholds are exceeded in any year, or attribution moves an event's return period materially inside the design life of the assets built to it.
• No longer holds when. No terminal test can be written for this entry. Erosion has no single event, so the at-risk tier carries the whole burden here, and the entry's discipline is a standing obligation to rebaseline design standards.
• Status. At risk, continuously, since 2022 at the latest.
• Owner. The Permanent Secretary at the Department for Transport for the rail envelope, with the Environment Agency's Chief Executive for the flood and drainage envelope.

Climate adaptation is the hardest case for this paper, because the machinery already exists. The Climate Change Act 2008 mandates a risk assessment and an adaptation programme, the Department for Environment, Food and Rural Affairs is the named department, and the Climate Change Committee is an independent statutory scrutineer reporting to Parliament. That machinery produced an assessment in May 2026 finding credible planning for three of 45 adaptation outcomes.^[32] The scrutiny asks whether adaptation is progressing at the necessary rate, which is again a question inside the model. No moment exists at which the design envelope itself is declared void and every standard built on it becomes due for rebaselining, because that declaration is nobody's to make.

Global energy markets will continue to supply Britain at tolerable prices

This assumption has now failed twice in four years, which makes recurrence its failure mode. Gas sets the marginal wholesale electricity price in Great Britain 98 per cent of hours, on the 2021 figure, while generating just over 40 per cent of the electricity, so a movement in gas prices reaches every household and employer in the country within one billing cycle.^[33] Government holds no lever over those prices and its only instrument is subsidy. That position was tolerable for as long as the assumption held. With the assumption dead, ministers carry the political consequence of affordability and inflation for something they do not control and cannot quickly fix. Storage, renewable capacity, and domestic production take years to build.

A register entry would have named the dependencies. Britain's direct imports of Russian gas were under 4 per cent of supply in 2021, and the exposure ran instead through the European market, which the British wholesale price tracks and which ran on Russian pipeline gas.^[34] In normal conditions a fifth of the world's liquefied natural gas and a quarter of its oil transit the Strait of Hormuz.^[35]

• At risk when. A supplier or transit route carrying a material share of the European or global market comes under military or political threat.
• No longer holds when. Such a route or supplier is interrupted, and prices move beyond the range the subsidy of last resort can absorb.
• Status. No longer holds, twice: Feb 2022 and Feb 2026.
• Owner. The Permanent Secretary at the Department for Energy Security and Net Zero, on storage and price data published by National Gas and the regulator.

Russia's invasion of Ukraine in Feb 2022 fired the terminal test the first time. The response was £44bn of subsidy, with the exposure left in place.^[36] The second firing was identifiable in real time, as rhetoric escalated between Israel, the United States, and Iran, and it arrived on 28 Feb 2026 when Iran closed the Strait of Hormuz to normal commercial traffic. In Mar 2026 the Organisation for Economic Co-operation and Development cut the United Kingdom's growth forecast by more than that of any other Group of Twenty economy.^[37] Six months after the closure, transits were running at a few a day against a pre-crisis figure of about 85,^[38] and gas storage stood around 30 per cent full, against Britain's roughly 12 days of demand and Germany's 89.^[39] Nothing had been prepared, because the dependency had never been written down as one.

The proposal

Britain should maintain a national assumptions register alongside the National Risk Register, held by the Cabinet Office and published on the same model, as a public document resting on a classified annex. The precedent is the government's own arrangement, since the risk register is already described as the external version of the classified assessment.

Each entry carries seven things:

1. The assumption in one sentence, with sub-assumptions where it decomposes, because a headline assumption can hold in name while its limbs fail in sequence.
2. What rests on it, so the blast radius is recorded before anyone needs it.
3. At-risk tests, with a documented response and a named owner. At least one test in every entry must be a published number the owner does not produce, so that a fired test is visible to anyone who can read the source.
4. No-longer-holds tests, with a documented response and a named owner, or a statement of why no terminal test can be written.
5. A review trigger and cadence, so entries cannot rot in silence.
6. Publication, with the classified annex reserved for content that genuinely cannot be public.
7. Whether Britain can act on the assumption or only watch it. Some assumptions sit inside the state's own arrangements, and a failing test can be answered by changing them. That private ownership under regulation delivers investment in essential utilities is one, because Parliament wrote that regime and can rewrite it. Others sit outside entirely. That the United States remains a stable and dependable ally is the largest of them, and no British decision secures it. The distinction sets what a response can be. Where the assumption is Britain's to hold, the response protects it. Where it is not, the response is to see the failure early and to have the alternative already built, because an assumption of that kind gives no notice when it goes.

Four of the six entries below sit outside British control. That is the ordinary case rather than the exception, and it is the reason the register is worth keeping. An assumption nobody in Britain can secure is one Britain can still watch, and watching it is the difference between preparing and hoping.

Beyond the entry format, the register needs the requirements that stop it becoming a filing cabinet.

New policy should state which register entries it depends on, and the current status of each, on the face of its business case. The trading-system entry is the argument for this. Business cases and Cabinet committee clearance already ask what a policy requires in order to work.

Coverage must not be controlled solely by the departments with most to lose from an entry. An assumption that is never listed has no owner, no tests, and no obligation, which makes omission a larger risk than any fiddling of thresholds. The scrutiny committee should be able to compel an entry; the National Audit Office, the Office for Budget Responsibility, the Climate Change Committee, and departmental chief scientific advisers should be able to nominate within their domains; and the register should publish the nominations it has rejected, with reasons.

Ownership is personal, in the way Britain already makes financial accountability personal through accounting officers. The line to draw is between finding and response. The named official owns the finding, recording that a test has fired, on the evidence, on a date, as a matter of record. The minister owns the response, because deciding what Britain does about a failed national assumption is a political judgement and belongs with a politician. That split keeps the constitutional lines intact. A fired test with no response written beside it then sits on the public record.

Responses at the at-risk level are sized to what being wrong would cost. Some firings will turn out to be false alarms, and a register that never produces one has its tests set too high. Most responses are cheap, being contract clauses, second suppliers, retained optionality, and a written contingency, costed and agreed before the pressure arrives.

The Joint Committee on the National Security Strategy is the natural scrutiny body, spanning both Houses and sitting adjacent to the risk assessment. Its access to classified material has been contested since at least 2012, when it complained that it could give Parliament no assurance about the assessment's adequacy. That problem has to be solved at the outset, because the register's most consequential entries would otherwise sit in an annex the committee cannot read.^[40] Two things follow. Access terms should be a founding condition, on the ring-fenced model the Intelligence and Security Committee already operates. The existence of every entry, its owner, and the date of its last review should be public even where its content is not.

None of the machinery is new. Assumption-based planning has been documented since the RAND Corporation published James Dewar's work in 2002, with the same anatomy of assumptions, signposts, and hedging actions.^[41] The Office for Budget Responsibility already publishes the determinants behind its forecasts, some of them audited independently, which is the closest British precedent for a published and owned set of assumptions. The Bank of England tests the solvency of the banking system on a regular cycle with a named owner and consequences attached, from a position of statutory independence that no departmental owner will have. The COVID-19 Inquiry has already recommended retiring single-scenario planning and institutionalising red-team challenge. The machinery closest to this proposal, the National Resilience Planning Assumptions, runs in the opposite direction, translating assessed risks into consequence benchmarks for planners.

Dewar's method has existed for 24 years without being adopted at national scale anywhere, which is itself evidence about feasibility. Its requirement is an owner with the standing to be publicly wrong, and no existing structure supplies one, which is what the ownership split set out above is for.

Objections

The first objection is hindsight, that six entries are easy to write in 2026 and would not have been written in advance. The trading-system entry answers it. That failure needed no foresight, only the reading of public evidence while policy was being built on the assumption. The energy entry answers it a second time, since the second firing was identifiable as rhetoric escalated, months before the strait closed.

The second is selection. Six failures were chosen from an unstated population, and a register that fires on everything is a permanent amber wall rather than a signal. A register written in 2005 would have carried assumptions that held: sterling's floating exchange rate as a shock absorber, the City's access to global capital, and the supply of health workers through international recruitment. Entries that hold are the demonstration that the tests discriminate, and a register whose entries all fail has been drawn to flatter itself.

Government already runs horizon scanning, which is the third objection. The functions differ. The Government Office for Science's futures work looks outward for weak signals of things that do not yet exist, while the register watches beliefs already held. Nobody is currently tasked with declaring that something long believed has stopped being true.

The fourth objection is diplomatic cost. Recording doubts about an ally's reliability has a price once the ally reads the entry, and the risk register's practice of naming hostile states does not transfer, because naming an adversary costs nothing diplomatically and naming an ally's unreliability costs a great deal. The classified annex exists for that case, on the model the state already runs.

The fifth objection is that the register relocates the problem it describes. The official who must declare a national assumption dead faces the same career asymmetry as the official who currently says nothing, and pre-agreed tests only remove discretion where the tests are mechanical. Two things reduce the exposure, and neither removes it. Every entry must carry at least one test that is a number somebody else publishes, so the fired-and-undeclared state is computable from outside. The finding and response split does the rest, leaving the official an arithmetical statement to make and the minister the political one. What remains is a reduction from raising an unprompted alarm about something in nobody's job description to confirming a published number.

Politicisation is the objection an owner would make. A standing list of fired tests with named officials attached is a gift to any opposition, and the predictable consequences are entries drafted vaguely enough to survive, resistance to the naming convention through the ordinary channels, and ministers contesting a declaration instead of executing a response. The answer is that the register makes the argument specific. An opposition attacking a fired-and-actioned test has nothing to work with, and an opposition attacking a fired-and-ignored test is performing the register's intended function. On the related legal exposure, responses should be recorded as planning commitments and not as duties owed to individuals, which is a drafting matter with precedent in the risk register's own status.

Gaming is the seventh objection, that owners will write tests which cannot fire. Publication answers most of that objection, and the requirement for an externally produced test answers the rest, because a threshold on a number the owner does not control cannot be quietly moved.

The last is that the register becomes another shelf document, a fourth artefact beside the risk register, the planning assumptions, and the capabilities assessment. That is the likeliest failure, and the answer to it is to name the minimum viable version now. Publication, the nomination route, and the review cadence can each be dropped and the register still works. Without a named owner and a response agreed in advance there is nothing worth building, and that is the component to defend when implementation starts trading things away.

Where this lands

The Cabinet Office already publishes a risk register on this model, resting on a classified assessment beneath it. The machinery, the publication route and the precedent for the split all exist. What is missing is the second register and a decision to keep it.

The smallest useful version is the six entries in this paper, written up properly, with owners named and tests attached. A single team could produce that in a quarter. It would not have prevented any of the six failures. It would have meant that when each test fired, somebody's name was against it, a date was on it, and a response was written before the pressure arrived rather than assembled afterwards by an inquiry.

None of the six entries needed prescience. Each needed somebody obliged to notice, early enough for the noticing to be worth something, and in every case that obligation belonged to nobody at all.

---

1. Cabinet Office, National Risk Register 2026, published 14 Jul 2026. ↩︎

2. Cabinet Office, National Risk Register 2026: "The National Risk Register (NRR) is the external version of the National Security Risk Assessment (NSRA)." ↩︎

3. World Trade Organization, Dispute Settlement Body meeting summary, 23 May 2016. The member was Seung Wha Chang. ↩︎

4. Chatham House, Reforming the World Trade Organization, chapter 4: Dispute settlement in crisis, Sep 2020. ↩︎

5. Center for Strategic and International Studies, Liberation Day tariffs explained; Tax Foundation, Liberation Day tariffs. The 10 per cent baseline took effect 5 Apr 2025 and country rates on 9 Apr 2025. ↩︎

6. New Civil Liberties Alliance filed on 3 Apr 2025, reported by Reason; the Liberty Justice Center filed V.O.S. Selections v. Trump on 14 Apr 2025, reported by UPI. ↩︎

7. Learning Resources, Inc. v. Trump, Supreme Court of the United States, 20 Feb 2026, decided 6–3. Congressional Research Service, Supreme Court Rules Against Tariffs Imposed Under IEEPA. ↩︎

8. Peterson Institute for International Economics, How will Trump's new 15 percent tariff fare in court?; Skadden, US Trade Court Strikes Down Section 122 Tariffs, May 2026. ↩︎

9. Theresa May, The government's negotiating objectives for exiting the EU, 17 Jan 2017. Economists for Free Trade published No Deal is the Best Deal for Britain. ↩︎

10. Congressional Research Service, The WTO's Appellate Body Loses Its Quorum. House of Commons Library, Brexit timeline. ↩︎

11. European Commission, Multilateral trading order strengthened as UK joins interim appeals system, 26 Jun 2025. ↩︎

12. Department of Health and Cabinet Office, UK Influenza Pandemic Preparedness Strategy 2011. Exercise Cygnus ran in Oct 2016 against an H2N2 influenza scenario. ↩︎

13. Written evidence to the Joint Committee on the National Security Strategy, The National Security Risk Assessment process is not fit for purpose, citing the 2017 National Risk Register rating for emerging infectious diseases. ↩︎

14. UK Covid-19 Inquiry, Module 1: the resilience and preparedness of the United Kingdom, HC 18, ordered to be printed 18 Jul 2024. The report finds that “The UK prepared for the wrong pandemic”, and at Flaw 1, Reliance on a single scenario, that “too much weight was placed on this single scenario”. ↩︎

15. Module 1 recommendations 3 and 9. See UK Government response to the Covid-19 Inquiry Module 1 report. ↩︎

16. On Taiwan, whose National Health Command Center was established after the 2003 outbreak and which began screening in Dec 2019, see Lessons learned from Taiwan's response to COVID-19, European Journal of Public Health. South Korea's statutory reform followed its 2015 outbreak of Middle East respiratory syndrome. The comparison is illustrative rather than controlled, since geography, population density, and prior public trust also shaped early outcomes. ↩︎

17. House of Lords Industry and Regulators Committee, Debt, dividends, and financial resilience. The sector was floated with historic debt written off and a cash dowry. ↩︎

18. Sixteen water monopolies have paid out a total of £78bn in dividends, Apr 2024, analysing company accounts; House of Lords Industry and Regulators Committee, as above. ↩︎

19. The £190bn is the water industry's own figure for infrastructure spending since privatisation, advanced as the answer to the dividends and debt criticism. The House of Lords Industry and Regulators Committee examined that defence and recorded the objection that the investment was funded largely through customer bills and company borrowing rather than new shareholder equity. ↩︎

20. Thames Water, Annual results 2025-26; Guardian analysis of company accounts, dividend payments 1989 to 2022. ↩︎

21. Defra, Government backs Ofwat's announcement on water company dividends, 20 Mar 2023. The licence conditions took effect on 17 May 2023 under powers in the Environment Act 2021. ↩︎

22. Ofwat, Thames Water's dividend payments and Condition P30: final decision, 28 May 2025, following a provisional finding on 19 Dec 2024. ↩︎

23. KKR drops bid to buy troubled British utility Thames Water, Jun 2025. KKR abandoned a recapitalisation of around £4bn and its preferred-bidder status lapsed. ↩︎

24. Water Magazine, Thames Water seeks creditor approval for further funding drawdowns, 25 Aug 2026. Senior creditors named a proposed new board on 24 Aug 2026, contingent on approval of the turnaround plan. ↩︎

25. Ofwat, Ofwat fines Thames Water nearly £123m, 28 May 2025. House of Commons Library, Water reform: A new vision for water, on the Independent Water Commission and the government's response. ↩︎

26. Network Rail, Why rails buckle in Britain. ↩︎

27. Met Office, A milestone in UK climate history, on the 40.3°C reading at Coningsby on 19 Jul 2022. The 62°C rail temperature was measured in Suffolk on 18 Jul 2022. ↩︎

28. Met Office, Summer 2025 is the warmest on record for the UK. The attribution compares the present climate with one without human-caused warming. ↩︎

29. Met Office, June 2026 heatwave: a recap of the temperature records. The reading was 37.7°C at Lingwood, Strumpshaw Hill, on 26 Jun 2026, against a previous record for the month of 35.6°C set in 1976. ↩︎

30. National Standards for Sustainable Drainage Systems, 2025, on the disparity between the 1-in-30-year sewer design standard and the 1-in-100-year design flood. Kendon et al., Update to the UKCP Local projections, Nature Communications, 2023. ↩︎

31. National Audit Office, Resilience to flooding, HC 189, 9 Nov 2023. ↩︎

32. Climate Change Committee, A Well-Adapted UK, the Fourth Independent Assessment of UK Climate Risk, 20 May 2026. ↩︎

33. University College London Institute for Sustainable Resources, Electricity prices dictated by gas producers who provide less than half of UK electricity, Sep 2022, on 2021 data. ↩︎

34. Department for Business, Energy and Industrial Strategy, Energy imports from Russia, 30 Jun 2022. Russian gas was 3.7 per cent of UK supply in 2021. ↩︎

35. International Energy Agency, Strait of Hormuz. ↩︎

36. National Audit Office, Energy bills support: an update, HC 232, 14 Nov 2024, reconciling the 2022 to 2024 schemes at £44bn, of which the Energy Price Guarantee was £24.0bn. ↩︎

37. Organisation for Economic Co-operation and Development, Economic Outlook, Interim Report March 2026. The UK forecast was cut 0.5 percentage points, from 1.2 to 0.7 per cent. ↩︎

38. Al Jazeera, Are ships passing Hormuz more willing to defy Iran or the US? What data shows, 20 Aug 2026. ↩︎

39. National Audit Office, Energy bills support: an update, HC 232, 14 Nov 2024, reconciling the schemes at £44bn. Storage days from Centrica doubles Rough gas storage capacity, Jul 2023. For the Aug 2026 level see British Gas boss warns UK risks winter fuel shortage, Aug 2026, reporting Centrica market data and the remarks of its chief executive. ↩︎

40. Joint Committee on the National Security Strategy, First Review of the National Security Strategy, HL Paper 265 / HC 1384, session 2010-12, on the Committee's access to the classified assessment. ↩︎

41. James A. Dewar, Assumption-Based Planning: A Tool for Reducing Avoidable Surprises, RAND Corporation, 2002. ↩︎