The Fear Register

Y2K

National governments · UN · World Bank · Gartner · the IT industry · the press

What was forecast Two-digit year fields rolling from 99 to 00 would cause systems to fail simultaneously worldwide. The institutional forecasts were sober; the ambient forecast was not. Predictions in wide circulation included power grid failure, aircraft falling from the sky, banking collapse, hospital equipment stopping, and lifts halting between floors. The US Senate had a dedicated Special Committee. The UN convened national coordinators. Governments issued preparedness advice; some advised households to hold cash, water and food.

What happened Effectively nothing. A scattering of minor faults — some card terminals, some display errors, a handful of monitoring systems at nuclear plants that did not affect safety. No grid failed. No aircraft fell.

The two readings Averted: roughly $300–600bn was spent globally (US estimates alone run near $100bn), millions of lines of code were genuinely remediated, and real bugs were found and fixed during that work — which is documented, not asserted. Overstated: the scale of consequence was never supported, and the household-preparedness register in particular went far past the engineering claim.

Counterfactual test This is the one case in the pillar with a genuine natural experiment, and it is the reason the record sits under unresolved rather than averted. Countries widely reported as spending very little — Italy, South Korea, Russia — experienced no obvious excess of failures. The frequently cited contrasts are the Russian military's reported $4m against the US military's $4bn, and British Telecom's spending against a South Korean operator running comparable 1980s hardware for far less. But the comparison is weaker than it is usually presented, including by me in the first draft of this record (verified 8 Sep). The £1.6m figure quoted for Italy was the running cost of its national coordination office, not national expenditure — Gartner's assessment was that it excludes what Italian government departments and businesses actually spent on testing, upgrading and replacing systems. The low-spend countries were not necessarily low-spend. Until somebody assembles actual sector-level remediation spend against observed failure rates, this is suggestive and uncontrolled, not decisive.

What the defenders answer That the low-spending countries ran less date-dependent legacy infrastructure, that multinationals fixed systems inside those countries anyway, and that the critical fixes were concentrated in a small number of systems that were in fact done. These are plausible and they are not measured. Nobody has done the work that would settle it — a sector-by-sector comparison of remediation spend against observed failure rates. Until someone does, the honest verdict is that the engineering problem was real, the remediation was partly justified, and the magnitude of consequence sold to the public was never established.

The harm Money, at a scale that dwarfs most entries in either pillar. Fear — surveys found substantial minorities expecting serious disruption, and people withdrew savings and stockpiled. And the durable cost: Y2K taught a generation that official warnings are inflated. That lesson was applied to every subsequent warning, correct ones included. A credibility drawdown is a real harm and it is almost never counted.

Who gained The IT consultancy and contracting industry, unambiguously and enormously — an entire remediation market created by a deadline that could not be renegotiated, with COBOL contractors recalled from retirement at exceptional rates. Analyst firms, whose forecasts drove procurement and whose business was selling those forecasts. Software and hardware vendors, for whom "not Y2K-compliant" was the most effective upgrade argument ever written. The press, for a story with a fixed date and a countdown. And governments, which gained a visible competence exercise with no losing outcome: if nothing happened, they had prevented it.

Why it anchors this pillar Note the incentive structure, because it recurs in every record below. Every actor with authority to assess the risk was also a beneficiary of the response to it. Not one of them had to lie. Nobody in the chain was rewarded for saying the number was too high, and everybody in it was insured against being wrong in the cautious direction.

BSE and variant CJD

UK government · SEAC · academic modellers · the press

What was forecast After the March 1996 admission of a probable link between BSE and vCJD, published modelling carried enormous upper bounds, and they escalated downwards over time as data arrived (verified 8 Sep): Imperial College modelling reported in 1997 put the upper bound as high as 10 million — the first of three appearances by the same modelling group in this study, alongside Report 9 in 2020 and the 20-million-lives-saved estimate, in 1998 at around 500,000, and by 2000 at up to 136,000. Each was published as a scientific projection and reported as a forecast. The long, unknown incubation period made the upper bound genuinely unbounded.

What happened 178 UK cases in total as of 2023. The epidemic peaked in 2000 and effectively ended. Against a 1997 upper bound of ten million, that is an overshoot of roughly five orders of magnitude.

The real failure was earlier and the opposite The 1990 assurance that British beef was entirely safe, and a minister feeding a beefburger to his four-year-old daughter on camera, is the part that did the lasting damage. The government first understated with theatrical confidence, then the modelling overstated, and the public was asked to trust the same institution through both. The credibility loss came from the reversal, not from either number.

The harm Around 4.4 to 4.5 million cattle destroyed (verified 8 Sep). A worldwide ban on British beef. Costs to the UK in the billions. Farming suicides. And a permanent blood-donation exclusion for people who lived in Britain in those years, still in force in several countries decades later — a precaution whose cost is borne by blood supplies elsewhere.

Who gained Competing beef producers, immediately and substantially. The rendering and feed regulations that followed benefited larger integrated producers over small farms. And the modelling community gained standing and funding — the BSE episode is a founding moment of UK infectious-disease modelling as a policy input, which matters enormously in 2020.

H5N1 avian influenza

WHO · national health departments

What was forecast WHO's senior pandemic coordinator David Nabarro said publicly in 2005 that a human H5N1 pandemic could kill "5 to 150 million". WHO's own working figure was 2 to 7.4 million, which it called a more reasoned forecast, and it moved publicly to distance itself after being deluged with inquiries — while notably declining to say the 150 million figure was wrong or implausible (verified 8 Sep). The larger number was the one that ran worldwide.

What happened No human pandemic. Cumulative confirmed human H5N1 deaths worldwide over the following two decades number in the hundreds. The virus has never acquired efficient human transmission.

The harm Mass culling of poultry, devastating smallholders across Asia who were rarely compensated. National stockpiling of antivirals — the direct cause of the Tamiflu purchases recorded in Pillar I. And a range so wide (5 to 150 million is a factor of thirty) that it conveyed no usable information while carrying the full authority of the institution issuing it.

Counterfactual test Weak, and it matters. Unlike Y2K, almost nothing was done that would have prevented human transmission — the culling addressed poultry spread, not the mutation that never occurred. The non-event cannot be credited to the response. This is as close as the pillar gets to a clean case of a forecast that was simply far too high.

Who gained Roche and antiviral manufacturers, directly. Vaccine manufacturers, who secured pandemic advance-purchase agreements with liability indemnities. And WHO itself, which gained the pandemic-preparedness mandate, the funding and the standing that carried into 2009 and 2020.

Swine flu, H1N1

WHO · UK Department of Health · SAGE

What was forecast The UK's planning assumption was a reasonable worst case of up to 65,000 deaths, widely reported as the expectation. WHO declared a Phase 6 pandemic in June 2009.

What happened Around 457 UK deaths. The strain was less severe than seasonal influenza for most groups.

Correction to my own first draft (verified 8 Sep) I wrote that WHO had changed its pandemic definition shortly beforehand, removing a requirement for enormous numbers of deaths. That claim is contested and I stated it too firmly. WHO's position — and the position of several independent reviewers — is that it has never had a formal definition of pandemic influenza to change; what altered was descriptive wording on a website. The allegation is widely repeated and should be presented as a dispute, not a fact.

What does stand The conflict-of-interest finding. WHO collected declarations of interest from its pandemic advisers and never published them, later calling this an oversight (verified 8 Sep). The BMJ and the Council of Europe established that scientists who helped draft pandemic influenza plans had financial ties to Roche and GlaxoSmithKline — the manufacturers whose advance-purchase contracts a Phase 6 declaration triggered — and that WHO did not disclose those links.

The harm Billions in vaccine and antiviral purchases across Europe, much of it destroyed unused. The Pandemrix narcolepsy injuries in Pillar I are the direct consequence of vaccinating fast under a pandemic authorisation. And a substantial further drawdown of institutional credibility, eleven years before it was most needed.

Who gained GSK, Baxter, Novartis and Sanofi, under contracts that in many cases triggered automatically on the WHO declaration — an arrangement in which a global body's classification decision was itself the financial event. Governments gained a demonstration of preparedness. This record and Y2K share the same structure exactly.

West African Ebola

US CDC

What was forecast The CDC projected in September 2014 that cases could reach 1.4 million in Liberia and Sierra Leone by January 2015 in the absence of additional intervention.

What happened Around 28,600 cases and 11,300 deaths across the whole outbreak — a real catastrophe for the countries involved, and roughly fifty times below the projection.

Counterfactual test — and a correction against my own framing (verified 8 Sep) The figure was 555,000 cases, or 1.4 million corrected for underreporting, by 20 January 2015, explicitly conditional on around 70% of cases not being effectively isolated. Actual reported cases across the three countries by late January 2015 were about 22,000 — an overshoot of roughly 65-fold on the headline number. But the same model contained a second branch that I left out, and it was right. It predicted that if roughly 70% of patients were placed in treatment units and safe burials conducted, transmission would fall away sharply — and that is what happened, on approximately the predicted timeline. So the honest reading is not "the model was wrong"; it is that the worst-case branch travelled worldwide and the conditional branch did not. Models that hold human response constant do overshoot, and that finding stands. But my first draft omitted the CDC's successful prediction, which made the institution look worse than the record supports.

Who gained The number secured an international response that was, by common consent, overdue and insufficient — so the beneficiaries here include the affected populations. Record that honestly: this is the pillar's clearest case of an inflated figure doing good, and a study that cannot say so is not measuring, it is arguing.

"Don't Die of Ignorance" — the predicted UK heterosexual AIDS epidemic

UK government · Chief Medical Officer · the tombstone campaign

What was forecast A leaflet to every household, tombstone and iceberg television advertising, and public projections of a general heterosexual epidemic in Britain. Forecasts circulating in 1986–88 ran to tens of thousands of infections a year by the early 1990s.

What happened HIV in Britain remained concentrated in specific groups. The generalised heterosexual epidemic of the projections did not occur.

Counterfactual test — and I overstated the "averted" reading (verified 8 Sep) I called this the strongest averted case in the pillar. It is weaker than that. The campaign did coincide with changed sexual behaviour and Britain's rates stayed low. But the general research literature on mass-media fear campaigns of exactly this type — television and radio advertising that raises awareness of risk — finds no reliable effect on condom use, which is awkward for attributing the outcome to the campaign. Meanwhile the epidemiology of heterosexual transmission in Western populations was poorly understood in 1986, and by 1987 UK cases were already concentrated among haemophiliacs, people who injected drugs, and gay men. Figures such as a million infected by 1990 assumed a generalised spread the virus did not achieve here. Verdict: genuinely unresolved, leaning further toward "the projection was wrong" than my first draft allowed.

Why it belongs here anyway Because the honest position on a case you may want to use as evidence is we do not know. Also because a real cost was paid that is rarely counted: the campaign's framing of universal risk contributed to a period of intense stigma and fear directed at people who were already dying, and to blood-donor and immigration exclusions that outlived the science.

Who gained The public health establishment — in standing, which it may well have earned. It gained standing, and deservedly if the averted reading holds. Include it because it is the counterexample to the beneficiary pattern that Y2K and H1N1 establish.

Covid has its own pillar

Where it went The covid records — the AstraZeneca litigation and the Vaccine Damage Payment Scheme, the origins question and its policing, the Inquiry's unpublished Rule 9 statements, the data released only under court order, and the twenty-eight-day counting asymmetry — are now Pillar V: Inside the Window.

Why they were moved out of here Keeping them in this register implied they were the same kind of object as the swine flu forecast or the September dossier. Those are closed — outcome known, inquiry reported, archive open. Covid is none of those things, and a register that mixes settled and unsettled cases lends the authority of the settled ones to conclusions that have not earned it. Pillar V documents only conduct that can be evidenced now — court orders, publication decisions, definitions, confidence levels — and states in advance what would close each open question. It is deliberately not scored.

Iraqi weapons of mass destruction

UK Government · US Administration · CIA · SIS · presented to the UN Security Council

What was claimed That Iraq possessed chemical and biological weapons and was reconstituting a nuclear programme. The UK's September 2002 dossier stated the intelligence was beyond doubt and carried the claim that weapons could be deployed within 45 minutes of an order. Colin Powell presented the case to the UN Security Council in February 2003 with vials and diagrams. Supporting claims included the Niger yellowcake documents and mobile biological laboratories.

What happened No weapons of mass destruction existed. The Iraq Survey Group concluded in 2004 that Iraq had ended its programmes in 1991. The Niger documents were forgeries. The mobile-laboratory claim rested largely on a single source, codenamed Curveball, who later admitted fabricating it. The 45-minute claim referred, at most, to battlefield munitions and not to strategic weapons.

Established by The Butler Review (2004) found that government judgements went to (although not beyond) the outer limits of the intelligence available — and, a finding I omitted from my first draft and which cuts against the harsher reading, that there had been no systematic attempt to distort the evidence (verified 8 Sep). What it did establish is caveat-stripping: Foreign Office advice of December 2001 was couched in more definitive terms than the language used by the JIC and omitted the JIC's caveats, and the 45-minute claim was given undue prominence though uncorroborated. German intelligence had warned that Curveball was unreliable; the Niger documents were crude forgeries identified as such by IAEA inspectors within hours. The Chilcot Inquiry (2016) concluded that the judgements were presented with a certainty that was not justified, that the legal basis was far from satisfactory, and that military action was not a last resort. In the US, the Senate Intelligence Committee found key judgements unsupported by the underlying reporting.

The harm 179 British service personnel dead, over 4,400 American. Iraqi deaths are contested and enormous — Iraq Body Count records over 200,000 documented violent civilian deaths, and survey-based estimates run far higher. Regional destabilisation and the conditions for the emergence of ISIS. And a collapse in public trust in government statements that is arguably the single largest credibility event in modern British political life.

Why it is the pillar's most important record Because the mechanism is the one every other record shares, operating at maximum stakes. Caveats stripped from an assessment. A range presented as a point estimate. A single unverified source promoted to a certainty because it was the source that said what the policy needed. Institutional endorsement — the UN podium — used to convert a contested claim into a settled one. Change the subject matter and this is the Vioxx paper, the September dossier and the reasonable worst case, structurally identical.

Who gained Defence contractors and the reconstruction industry, with Halliburton and KBR the most documented. Oil majors, subsequently. But the honest answer is more uncomfortable and more useful: the primary gain was political and institutional — an administration that had chosen a course and required a justification, and intelligence services whose product became decision-shaping rather than decision-supporting, which is a large increase in institutional importance. Nobody in the chain was rewarded for saying the case was thin.

The Gulf of Tonkin incident

US Administration · Department of Defense · NSA

What was claimed Two attacks by North Vietnamese torpedo boats on US destroyers, on 2 and 4 August 1964. The second was used to secure the Gulf of Tonkin Resolution, which passed the House unanimously and the Senate with two votes against, and became the legal basis for the Vietnam War.

What happened The first engagement was real. The second did not occur. An NSA historical study by Robert Hanyok — Skunks, Bogies, Silent Hounds and the Flying Fish — concluded that no attack took place on 4 August (verified 8 Sep). The agency had intercepted North Vietnamese communications indicating no attack; cables relating to the 2 August engagement and to the recovery of damaged boats were re-translated, mistranslated and selectively presented to fit an attack narrative. Note the timeline: Hanyok's study was written for the NSA's own classified journal in 2001 and not declassified until 2005, alongside 140 other documents — four years of internal suppression on top of thirty-seven years of external.

The harm Approximately 58,000 US deaths and Vietnamese deaths in the millions.

Why it belongs beside Iraq Because it establishes that the Iraq case is not an aberration of one administration, and because the declassification took forty-one years. Any judgement about a current claim of this kind is being made without the evidence that will eventually settle it — which is the study's central epistemic problem stated in its sharpest form.

Who gained An administration that obtained a war authorisation it had drafted in advance of the incident, and the defence industry of the subsequent decade. But the belief that made it feel obligatory rather than opportunistic was the doctrine of credibility — that American resolve had to be visibly demonstrated or allies would defect and adversaries advance. Held sincerely and across the administration, it converted a disputed radar contact into a duty. Note the effect on the reporting chain: a second attack that did not happen was reported because reporting it was consistent with what everyone expected, which is how conviction contaminates intelligence without anyone lying.

The Kuwaiti incubator babies

Citizens for a Free Kuwait · Hill & Knowlton · US Congressional Human Rights Caucus

What was claimed A 15-year-old Kuwaiti girl identified only as "Nayirah" testified in October 1990 that she had watched Iraqi soldiers remove babies from incubators and leave them to die. The account was cited repeatedly by the US President and by senators in the debate authorising the Gulf War, which passed the Senate by five votes.

What happened Nayirah was revealed in 1992 to be Nayirah al-Sabah, daughter of the Kuwaiti ambassador to the United States — a fact concealed from the hearing. Her appearance was arranged by Hill & Knowlton, then headed by a former chief of staff to Vice-President Bush, and paid for out of a roughly $10m campaign by the Kuwaiti government-funded Citizens for a Free Kuwait (verified 8 Sep).

A precision the record needs "The testimony was fabricated" is the popular version and is slightly too strong. Later interviews indicate she had seen one baby outside an incubator, for no more than a moment, and had not been a hospital volunteer as presented. The accurate description is a fragment of observation wildly distorted and staged — which is more instructive than invention, because a pure fabrication is easier to detect and easier to dismiss.

Why it matters for the study's method The single cleanest demonstration that an atrocity claim can be manufactured, laundered through a human rights forum and a respected NGO, and used to move a legislature — with no state intelligence service involved at all. It should make you slower to accept a vivid, emotionally decisive detail from any direction, including one that supports a conclusion you already hold.

Who gained The Kuwaiti government-in-exile, which paid for the campaign. Hill & Knowlton, on the largest foreign-government contract of its kind to that date. And the war's proponents, who obtained a narrow vote.

The Benghazi massacre justification for intervention in Libya

UK and French governments · NATO · UN Security Council Resolution 1973

What was claimed That Gaddafi's forces were about to carry out a massacre of civilians in Benghazi, justifying intervention under the responsibility to protect.

What happened The House of Commons Foreign Affairs Committee, in its Third Report of Session 2016–17 (HC 119, 6 September 2016), concluded that the government failed to identify that the threat to civilians was overstated and that the rebels included a significant Islamist element, and that the policy was not informed by accurate intelligence (verified 8 Sep). It found the Prime Minister ultimately responsible for the failure to develop a coherent strategy, and listed the consequences: political and economic collapse, inter-militia and inter-tribal warfare, humanitarian and migrant crises, widespread human rights violations, Gaddafi's weapons dispersing across the region, and the growth of ISIL in North Africa.

Kept as unresolved Gaddafi's public threats were real and his prior conduct was not reassuring; the committee found the specific massacre claim overstated, not invented. That distinction is the whole record and it should not be flattened in either direction.

Who gained Politically, governments that obtained a rapid multilateral authorisation. Structurally, the responsibility to protect doctrine gained a precedent — and then lost most of its credibility for future use, which is a cost borne by populations in later crises who did not get an intervention.

Satanic ritual abuse

Social services · police · therapists · evangelical campaigners · broadcasters

What was claimed Organised networks conducting ritual sexual abuse and murder of children, recovered through therapeutic techniques and suggestive interviewing of children. In Britain the Cleveland (1987), Rochdale (1990) and Orkney (1991) cases saw children removed from families in dawn raids.

What happened No corroborating evidence was ever found. Jean La Fontaine's three-year study for the Department of Health, The Extent and Nature of Organised and Ritual Abuse (1994), examined the records of 84 British cases going back to 1988 and found no evidence substantiating any of them (verified 8 Sep). Children had been subjected to lengthy repeated interviews and held in children's homes for periods ranging from three months to ten years; most were released in early 1991 after a court found the social services investigations flawed and the allegations untrue. The Clyde Report on Orkney was severely critical. Convictions in comparable US cases, including McMartin, were overturned; some people served years in prison.

The harm Children taken from parents and, in several cases, subjected to repeated interviewing that generated the accounts. Families destroyed. Innocent people imprisoned. And a genuine harm to actual abused children, whose accounts were subsequently harder to credit.

Who gained A therapeutic industry built on recovered memory; expert witnesses; campaigning organisations; and broadcasters. Note the structure: social workers and police were acting in good faith on expert guidance that had no evidence base — the same shape as the prone-sleeping record in Pillar I.

The "crack baby" epidemic

US media · medical commentary · legislatures

What was claimed That prenatal cocaine exposure produced a permanently damaged generation — a "biological underclass" incapable of normal development. The claim traced substantially to one small 1985 study with no adequate controls.

What happened Hallam Hurt's Philadelphia cohort followed 224 full-term babies born between 1989 and 1992, half exposed to cocaine in utero, all from low-income families, for close to twenty-five years (verified 8 Sep). It found no statistically significant differences in long-term health or life outcomes between the exposed and unexposed children. Hurt's own summary: poverty is a more powerful influence on the outcome of inner-city children than gestational exposure to cocaine.

The harm Children stigmatised and placed in special education on the label alone. Mothers prosecuted and imprisoned, and infants removed, overwhelmingly poor and Black. The panic supplied moral cover for sentencing law whose disparities lasted decades.

Who gained Politically, an argument for punitive drug policy. Institutionally, the framing removed the need to address poverty, which was the actual variable.

Video nasties

UK press campaign · National Viewers' and Listeners' Association · Parliament

What was claimed That uncertificated home video was causing measurable harm to children. The Parliamentary Video Enquiry of 1983 produced a survey claiming that 37% of children under seven had seen a "video nasty" (verified 8 Sep). Brian Brown, credited as an associate director of the enquiry, disclosed that his own research unit at Oxford Polytechnic had repudiated the framework, context and conclusions of the entire document over its methodology. Critics then replicated the survey using invented film titles and obtained comparable rates — children will say yes to an exciting-sounding name whether or not it exists.

What happened The Video Recordings Act 1984. Seventy-two titles prosecuted; some now carry an ordinary certificate. No robust evidence of the claimed harm was ever produced.

The pattern to note A fabricated or worthless survey, cited in Parliament, becoming law within two years. The same shape recurs with mobile phones, video games and social media — each new medium arrives with a harm claim that outruns its evidence, and the fact that the last one was overstated never adjusts the confidence of the next.

Who gained Newspapers running the campaign, campaigning organisations, and legislators with a low-cost demonstration of protecting children. And the proposition underwriting all of it: that depicted violence produces real violence. It has poor evidential support and has never lost its constituency, because it explains a frightening thing simply and costs nothing to hold. Follow it through the century and the medium changes while the claim does not — penny dreadfuls, cinema, comics, video, then video games, then social media. The belief is the durable asset; each new medium merely renews its lease.

Alar on apples

Natural Resources Defense Council · 60 Minutes · a public relations campaign

What was claimed That daminozide residue on apples posed an intolerable cancer risk to children. Broadcast to a mass American audience with a rotting-apple-and-skull graphic, and promoted through a professional PR campaign timed with the report's release.

What happened Broadcast 26 February 1989 as "A" is for Apple, based on the NRDC report Intolerable Risk, with the exclusive negotiated for the NRDC by the public relations firm Fenton Communications (verified 8 Sep). Apple purchases fell by as much as 60% and growers reported losses of around $100m; some families lost farms. The EPA proposed a ban on food uses in May 1989, the manufacturer withdrew the product voluntarily, and the EPA terminated its food uses that November.

The qualifier I omitted, and it matters Washington apple growers sued CBS and lost. They could not establish that the facts in the broadcast were false, and on appeal the court noted the absence of evidence that Alar was safe for children (verified 8 Sep). The risk calculation did rest on very high animal doses and unrealistic consumption assumptions, and that criticism stands — but "a baseless scare" is not what a court found, and my first draft implied a cleaner verdict than exists.

Why it is in the register Because the actor here is an environmental NGO with a media strategy, not a corporation or a government. The mechanism — a real hazard, a risk estimate built on worst-case assumptions, and a communications campaign that presented the worst case as the expectation — is identical to the mechanism in the epidemic-forecast group. It should be scored by the same rule.

Who gained The NGO, in profile and funding. The PR firm. Organic and alternative producers. And broadcasters. Underneath all four is the belief that does the actual work in every food scare in this register: natural is safe, synthetic is suspect. Alar was a growth regulator on an otherwise identical apple. That premise is not a commercial position — it is a widely and sincerely held one, which is precisely what makes it worth monetising. The companies did not create it. They rented it.

MSG and "Chinese restaurant syndrome"

A 1968 letter to the New England Journal of Medicine · the food press

What was claimed That monosodium glutamate causes headaches, palpitations and numbness. The origin was a single anecdotal letter to the New England Journal of Medicine in April 1968, signed Robert Ho Man Kwok, describing weakness, numbness and palpitations after eating in northern Chinese restaurants and naming salt, cooking wine or perhaps MSG as possible causes. A month later the journal printed ten responses from other doctors reporting similar discomfort — no two of which listed the same symptoms, appearing all over the body and at widely varying times after eating.

A twist that is itself disputed (verified 8 Sep) In 2019 an American orthopaedic surgeon, Howard Steele, claimed he had written the letter himself for a $10 bet, invented the name, and been refused permission to retract it. Treat that as contested rather than established: Dr Robert Ho Man Kwok was a real researcher who died in 2014, and his family and colleagues maintain he wrote it. Either way the letter was an anecdote, not a study.

What happened Double-blind studies have consistently failed to reproduce the effect at dietary levels. Regulatory bodies classify MSG as safe. The belief persists, and "No MSG" remains a marketing claim.

The harm Fifty years of damage to Chinese restaurants and to Chinese communities specifically — a food scare that attached to an ethnicity and stayed there. Small per person; very large in aggregate; and a demonstration that a single unreviewed letter in a respected journal can outlive fifty years of contrary evidence. Compare the Porter and Jick letter in Pillar I, which ran the opposite way and helped start an opioid epidemic. Same format, same journal, opposite direction, both catastrophic.

Who gained Food manufacturers selling on a negative claim; the journal, marginally, in citations. And a prejudice that acquired a scientific-sounding name. Glutamate occurs in parmesan, tomatoes, mushrooms and human milk, and has never generated a scare in any of them. The syndrome was called Chinese restaurant syndrome, and the suspicion attached to a cuisine rather than to a molecule — which is why the same compound under a European label caused no alarm whatever. What gained was an existing discomfort, given a clinical vocabulary and a journal citation to rest on. That is not the absence of a beneficiary; it is a beneficiary nobody wanted to name.

The Population Bomb

Paul Ehrlich · Sierra Club

What was forecast Opening line: The battle to feed all of humanity is over. Hundreds of millions would starve in the 1970s; England was given a serious chance of not existing by the year 2000. Compulsory population control was argued for, and food aid to countries judged beyond saving was to be withheld under a triage doctrine.

What happened The Green Revolution. Global food production outpaced population growth; the proportion of humanity in extreme poverty and hunger fell dramatically over exactly the period forecast. Ehrlich lost the 1980–1990 wager with Julian Simon on resource prices outright.

The harm Not merely a wrong book. Population-control programmes drew legitimacy from this framing. During India's Emergency of 1975–77, more than 8 million men were sterilised — around 6 million in 1976 alone — under state quotas, with proof of sterilisation required to draw a salary or obtain medical treatment, and roughly 2,000 deaths from botched operations (verified 8 Sep). Coercive one-child enforcement in China drew on the same intellectual climate. This is a forecast whose failure has an identifiable body count and a coercion count.

The qualifier the record needs Famine did occur in the 1970s — in India, Bangladesh, Cambodia and parts of Africa — and starvation claimed some four to five million lives that decade. What did not happen was Ehrlich's mechanism: the deaths were overwhelmingly the product of war andpolicy failure rather than of global food supply being outrun by population, and the world death rate did not rise as he said nothing could prevent.

Who gained The author, in sales and a lifelong platform — note that he has never retracted and remains professionally honoured, which is itself a finding about how forecasting error is penalised. Foundations and agencies whose population programmes were funded on the premise. And governments for whom it licensed coercive domestic policy.

The Limits to Growth

Club of Rome · MIT

What was forecast System-dynamics modelling of population, industrial output, food, pollution and resources, projecting overshoot and collapse in the twenty-first century. Widely reported as predicting specific resource exhaustion dates — gold, mercury, oil — within decades.

What happened, honestly The exhaustion figures reported in the press were not the book's forecasts (verified 8 Sep): they came from a table of static reserve calculations sourced to a US Bureau of Mines publication, which the text itself qualified. The book did not predict collapse by 2000, and the criticism that it did — the most widely repeated criticism of it — is answering something it never said. Graham Turner's comparisons of the model against observed data, in 2008 and again at forty years in 2014, found population, food, industrial output, pollution and resource decline tracking the "standard run" reasonably closely; collapse in that run does not begin until mid-century.

Why it is here as unresolved Because it cuts both ways and the study needs at least one record that punishes careless use by either side. The popular version of the forecast failed; the model's own timeline has not yet been tested. Citing it as a debunked prediction is as sloppy as citing it as vindicated.

Who gained The Club of Rome in influence, and an entire consultancy and conference economy around global modelling. Also, and fairly, the field of systems modelling gained methods that are genuinely useful.

Peak oil

Petroleum geologists · think tanks · the energy press

The part that was right, which I left out Hubbert predicted in 1956 that US crude production would peak between 1965 and 1970. It did — at about 9.6 million barrels a day in 1970, declining thereafter for nearly four decades (verified 8 Sep). The method worked on the case it was built for, and a register that opens with "peak oil failed" is starting from a false premise.

What was forecast, and failed The later global extension of it. Campbell and Laherrère revisited Hubbert in the 1990s and 2000s and placed successive dated peaks, including non-OPEC production peaking by 2010, followed by supply collapse and economic crisis.

What happened US production bottomed at around 5 million barrels a day in 2008 and then hydraulic fracturing and horizontal drilling took it to roughly 9 million by 2014 and past 12 million by 2019, making the United States the world's largest crude producer (verified 8 Sep). The forecasts consistently failed because they modelled a fixed resource base against extraction rates and did not model technological response to price, which is the same omission as the Ebola model's fixed human behaviour.

The transferable finding The commonest structural error in this entire pillar: a model that holds the response constant. Populations change their behaviour, engineers change their methods, markets change their prices. Forecasts that assume otherwise overshoot in a predictable direction and by predictable amounts, and that is a testable regularity your study could actually measure.

Who gained Investment products and newsletters sold on scarcity; think tanks; and, awkwardly, the case for renewable investment, which used the argument and did not need it.

The coming ice age

Popular press · Newsweek (1975) · Time

What was claimed That global cooling was under way and a new ice age was approaching. Newsweek's 1975 piece is still the most-cited artefact of the episode.

What the literature actually said Handle this one precisely, because it is used carelessly by everyone. Peterson, Connolley and Fleck surveyed peer-reviewed papers from 1965 to 1979 and found 7 papers predicting cooling, 20 neutral and 44 predicting warming (verified 8 Sep). There was no scientific consensus on cooling. The Newsweek author later published a note acknowledging the article's errors.

Correct classification Not a failed scientific forecast. A media amplification of a minority position — the same category as the eggs, Alar and video-nasty records. It is genuine evidence about how science reaches the public, and no evidence at all about whether the science was wrong. Using it as the latter is the single most common error made with this case, and if your study makes it, everything else in it will be discounted.

Who gained Magazines, at the time. Since the 1990s, the episode's main use has been rhetorical, and the beneficiaries have been the arguments it is deployed in.

The ozone layer

Molina & Rowland (1974) · UNEP · the Montreal Protocol

What was forecast That CFCs would deplete stratospheric ozone, raising ultraviolet exposure and skin cancer rates. The Antarctic hole was observed in 1985 and was larger than the models had predicted.

What happened The Montreal Protocol (1987) phased out CFCs, was ratified universally, and was strengthened repeatedly. Atmospheric equivalent chlorine peaked in 1993 and has fallen since — down about 11.5% from that peak by 2020 (verified 8 Sep). The Antarctic hole is measurably shrinking: the 2024 hole was below the 1990–2020 average and among the smallest since 1992, and its later seasonal onset is described by the WMO as a robust indication of recovery. The 2022 WMO assessment projects a return to 1980 values around 2066 over the Antarctic, 2045 over the Arctic and 2040 elsewhere.

Counterfactual test — and it passes Unlike Y2K, this one has the evidence Y2K lacks. Atmospheric chlorine concentration is directly measured, it tracks the phase-out schedule, and ozone recovery follows it. There is a measured dose-response between the action and the outcome. That is what an averted disaster looks like when you can actually demonstrate it, and it is precisely the evidence the Y2K record has to say is missing.

Where the criticism does land The industry ran a doubt campaign — DuPont argued for years that the science was unproven — and some popular coverage was apocalyptic beyond the science. But the forecast was correct and the response worked. Filing this as a scare that came to nothing inverts the case completely.

Who gained Worth stating plainly, because it complicates the beneficiary pattern in the opposite direction: DuPont held the patents on the replacement refrigerants and shifted position once substitutes were available. A commercial interest can be aligned with a correct outcome. The presence of a beneficiary tells you nothing on its own about whether a claim is true — which is why the beneficiary column in these registers is a lead to follow, never a verdict.

Acid rain

Scandinavian scientists · US EPA · NAPAP · UNECE

What was forecast That sulphur dioxide and nitrogen oxides were acidifying lakes and damaging forests across Scandinavia, Canada and the eastern United States.

What happened The US Clean Air Act Amendments of 1990 created a sulphur dioxide cap-and-trade programme. Emissions fell by around 90%, at a small fraction of the cost industry had forecast. Lake pH recovered measurably. The European convention achieved comparable reductions.

Where you have a genuine case — and it is stronger than I first wrote (verified 8 Sep) The NAPAP assessment reporting in 1990 found considerably less than the rhetoric had claimed. Damage to Appalachian red spruce is attributed in that assessment to the conifer swift moth rather than to acid deposition. It reported that only about 4.2% of lakes in the eastern United States were acidic, and that for many of those the cause was natural acidity or surface-mining runoff rather than acid rain. It concluded that 1990 levels of pollution-caused acid deposition were not harmful to agriculture or human health. The German Waldsterben narrative — most of the country's woodland dead by the 1990s — did not happen.

Correction to my own first draft I originally wrote that "the lakes were real, the dying forests largely were not". On NAPAP's findings the lake claim was also far weaker than the public case for action implied, and I understated your position. One caveat I will not skip: the specific figures above circulate heavily on partisan sites, and I have not read them out of the NAPAP report itself — check them against the primary document before using them, because this is precisely the kind of number this register exists to be sceptical about, whichever direction it points.

Why that version is stronger for your study Because it is true, it is defensible, and it identifies a mechanism that generalises to every record in both pillars: the gap between what the science supported and what was said in public. "It was all a lie" is refutable in one sentence with an emissions chart. "The public claim ran ahead of the evidence even where the policy was right" cannot be refuted, and it is the thesis your register actually supports.

Who gained Low-sulphur coal producers over high-sulphur, and the scrubber industry. Environmental agencies gained a durable success. And cap-and-trade gained the precedent later used for carbon.

AIDS in the United States

US Administration · CDC · the press

What was warned Clinicians and the CDC identified a fatal transmissible immune disorder from 1981 and pressed for funding and a public response.

What happened There was no presidential mention of AIDS until September 1985, despite internal memos from 1982 estimating thousands of infections, and officials were briefed to tell Congress funding was more than adequate while researchers were asking for resources (verified 8 Sep). The clearest single measure of the disparity: in 1982 the NIH spent about $8,991 per AIDS death against $34,841 per Legionnaires' disease death — nearly four times as much on a condition affecting a more politically sympathetic population. Blood bank physicians resisted screening on grounds of cost while minimising the risk, which is the same failure, on the same timeline, as the UK contaminated blood record in Pillar I.

Why it belongs opposite the UK campaign Britain over-forecast a general epidemic and acted hard; America under-responded to a real one. The same institutions can err in both directions within the same decade on the same disease, which is fatal to any simple thesis about institutional bias running one way. It is the single most useful record in this pillar for keeping the study honest.

Who gained Everyone whose position was made easier by saying nothing. Politically, inaction was cheaper than action for a constituency without political weight — the beneficiary of a silence is whoever avoids the cost of speaking.

Pandemic preparedness warnings

WHO · national risk registers · Exercise Cygnus (2016)

What was warned Pandemic influenza sat at or near the top of the UK National Risk Register for years. Exercise Cygnus in 2016 tested the response and concluded the UK's preparedness was currently not sufficient to cope with the extreme demands of a severe pandemic (verified 8 Sep): inadequate critical-care surge capacity requiring thousands more beds, large parts of the health service needing to be switched off, clinicians adopting a battlefield triage mentality, a shortage of personal protective equipment, and no joint tactical-level plans.

And what followed The report was withheld, on stated grounds of biosecurity and of not alarming the public, and was published only in 2020 after a freedom-of-information campaign by an NHS doctor. The government's strategic pandemic roadmap was never revised in response; its last update was 2014.

What happened The identified gaps were substantially unremedied in 2020.

The uncomfortable connection The same institutions that over-forecast H5N1 in 2005 and H1N1 in 2009 then under-prepared for the pandemic that came. Those are not contradictory; they are the same failure. The reward is for the visible announcement, not for the unglamorous stockpile. Announcing a threat is cheap and legible; maintaining capacity for fifteen years is expensive and invisible. Every incentive in the Y2K record points the same way.

Who gained Every treasury and every minister who deferred the spending. I previously recorded this as a saving rather than a gain, which is a distinction without a difference: the saving is realised now, in the current term, against a cost that falls on a successor. That is not the absence of an interest but one of the most ordinary interests there is — the discount rate applied to somebody else's problem. Under-preparation has a beneficiary in every case, and it is always whoever holds the office before the event.

The financial crisis warnings

Raghuram Rajan (2005) · a handful of economists and regulators

What was warned Rajan, then chief economist at the IMF, presented Has Financial Development Made the World Riskier? at Jackson Hole in August 2005 (verified 8 Sep). His argument was specific, not vague unease: financial-sector pay structures, combined with complex products, gave bankers large cash incentives to take risks with other people's money and imposed no penalty on them for the losses. Larry Summers rose from the audience to call the paper's premise slightly Luddite and largely misguided, warning that regulation would cut the sector's productivity. Others warning of housing and securitisation risk were dismissed in similar terms.

What happened 2008.

Why it is in this pillar rather than Pillar IV Because the mechanism is a fear-event mechanism in reverse: an institutional consensus dismissed a correct warning as alarmism, using precisely the vocabulary this study is examining. Note carefully — "that's alarmist" is itself a claim that can be wrong, and it is made with exactly the confidence, and by exactly the bodies, that this register catalogues.

Who gained Everyone paid on transaction volume: banks, ratings agencies paid by issuers, and the political consensus that had claimed to have ended boom and bust.