T03 — Interests, power and the political economy of knowledge#

Cross-cutting thematic synthesis across Late lessons from early warnings: the precautionary principle 1896–2000 (EEA 2001, “LL1”) and Late lessons from early warnings: science, precaution, innovation (EEA 2013, “LL2”). Strand A working document, 26 September 2026.

Scope. How interests shaped what was known, when it was known, what was said about it and what was done: corporate conduct, manufactured doubt, research control and funding effects, secrecy, capture and dual mandates, lobbying and rule-changing, trade and jurisdiction, the treatment of early warners, and the LL2 Part D analysis of why business did not act. Throughout, the synthesis separates documented misconduct from ordinary incentive effects and from sincere but mistaken belief.

Conventions. - Section ids (LL1-05, LL2-07, …) follow working/late-lessons/. Page numbers are report pages. - “Hindsight” means the post-publication record in hindsight/<id>.md (to September 2026). Those files flag where web search was unavailable and where items rest on secondary sources; those caveats carry over. - Three voices are kept apart: - Reports: what the chapters say. - Evidence: what the chapters’ own sources and the later record show. - Analysis: my interpretation, marked as such. - Evidence-strength ratings: - Strong: several cases with contemporaneous documentary evidence or official or judicial findings, not overturned since. - Moderate: several cases, but resting partly on protagonist accounts or secondary histories, or a documented pattern whose causal effect is unmeasured. - Suggestive: one or two cases, or inference from timing and outcome. - Asserted: claimed without supporting evidence.

Summary#

• The best-evidenced lessons are about mechanisms, not motives or frequencies. Seven cases rest on contemporaneous internal documents or official findings: tobacco, vinyl chloride, beryllium, PCBs, asbestos, Minamata and lead. Together they document four things: producers knowing more than regulators; deliberate efforts to keep scientific questions open; control of research agendas and publication; and campaigns to change the rules of evidence. Later litigation and archives have strengthened every one of these (hindsight LL2-07, LL2-08, LL1-06, LL1-05, LL2-03, LL2-25).
• Most of the reports’ cases are not misconduct cases. Benzene, DES, PCE, CFCs, fisheries, SO2, MTBE (as written), antimicrobials, EE2 and several others show incentive effects: limits set at what was feasible, costs shifted to others, promote-and-protect mandates, indifference. Some show sincere error, and the BSE inquiry and the beryllium panel argue for that reading in cases the chapters present as bad faith.
• The state is often the interested party. Examples include MITI, the Health Ministry and Kumamoto Prefecture at Minamata; MAFF; the US DOE; the USDA; Canada’s DFO and fisheries ministers; the Japanese nuclear regulator; and the European Commission on hormones. Hindsight adds Flint and Camp Lejeune. This weakens LL2’s claim that harms were “for the most part” caused by “irresponsible corporations” (LL2-00, p. 11). The better generalisation concerns whoever controls exposure and information.
• Part D (LL2-25) is a strong diagnosis and a weak base rate. Its account holds up well: harms enter firms’ decisions only through leaky liability, regulatory and reputational channels, and the most useful concept is its distinction between “business actions” and “political actions”. Its “virtually all” is built in by case selection. Firms differed within sectors, and some later moved (hindsight LL2-25).
• The main limits are observability and symmetry. Concealment surfaces mainly through litigation, so litigated sectors look worse and unlitigated ones cleaner. The reports scrutinise industry-linked evidence far more than evidence from allied advocates, several of whom were expert witnesses or had undisclosed funding. Later cases show unfounded alarms can also be organised and sustained.

1. How the reports frame interests and power#

Reports (LL1). - LL1 treats interests as one cause among several. The Preface says lack of political will “seems to be an even more important factor” than the availability of trusted information (LL1-00, p. 4; asserted). - The synthesis finds warnings ignored “because of short-term economic and political interactions” in some cases: asbestos, PCBs, the Great Lakes and SO2 (LL1-16, p. 168). It locates the main failure in information not reaching decision-makers (“institutional ignorance”, p. 171). - Its independence lesson (lesson 10) rests on the “dependence of risk assessment on information produced and owned by the very actors whose products are being assessed” (LL1-16, p. 179; LL1-17, p. 194). The editors add: “Not all of these cases demonstrate the delaying or distorting effect of non-independent sources” (p. 179). - The chapters also make two points about how evidence and proof are judged. Whether demands for more analysis count as “paralysis by analysis” or “prudent and careful evaluation” depends on who bears the costs (LL1-16, p. 182). And choosing the level of proof is “a key political decision with profound ethical implications” (LL1-17, p. 193).

Reports (LL2). LL2 moves interests to the centre. - Framing. The Preface makes “the relationship between knowledge and power” the volume’s thesis and calls for “the power structures of knowledge” to change (LL2-00, pp. 7–8). - Part A’s summary. The nine health cases show “slow and sometimes obstructive behaviour by businesses” (p. 10), and harms were “for the most part” caused by “irresponsible corporations” (p. 11). - Dedicated chapters. There is a chapter on tobacco-industry manipulation of research (LL2-07) and a Part D chapter on why business did not act (LL2-25). The latter explicitly moves attention from regulators, LL1’s focus, to firms (p. 608). - The closing chapters. “Opposition from powerful corporations” heads Gee’s seven barriers to action (LL2-27, p. 645). The conclusion names decisions “made by a few people on behalf of many” (LL2-28, p. 671) and “the vested interests of some powerful” stakeholders (p. 672), yet declares unequal political power “well beyond the scope of this report” (p. 672).

Analysis. Three points follow. - The shift has a cost. LL1’s distinction between knowledge that fails to reach decision-makers and knowledge that is discounted because of interests is analytically sharper than LL2’s broader corporate framing, and it survives hindsight better (hindsight LL1-16, Claim 3). - Both volumes prescribe mainly information, independence and transparency. The diagnosis of weak political will (LL1-00, p. 4) implies remedies about power that neither develops (LL2-28, p. 672). - LL2 is not neutral ground on this theme. Case authors were chosen for “substantial involvement” (LL2-00, pp. 9–10). No Part A panel comes from a company whose conduct is at issue or from a regulator defending itself, and two of the more sceptical commentaries (Guidotti on beryllium, Castaño on fish) were left off the panel list (LL2-00, p. 16; LL2-06, p. 145; LL2-05, p. 130).

2. Three kinds of explanation: misconduct, incentives, sincere error#

2.1 Working criteria (analysis)#

• Documented misconduct. Contemporaneous internal records, or findings by a court or official inquiry, show at least one of the following:
• concealment of known findings;
• falsification or deliberate editing of evidence;
• covert sponsorship;
• a stated intent to create doubt or delay;
• suppression of researchers.
• Incentive effect. Conduct that follows predictably from who pays and who gains, without evidence of deception. Examples:
• limits set at what is “feasible”;
• catch limits above advice;
• costs left with third parties;
• studies not done;
• indifference.
• Sincere but mistaken belief. Decision-makers held a reassuring model, and the record suggests they believed it. Examples:
• threshold “tolerance” doses;
• the “placental barrier”;
• dilution by tall stacks;
• “remote” human risk from BSE.
• Asserted motive. A chapter attributes motive without documentary support. Motive attribution is always an inference, and the categories overlap: self-serving bias can make an incentive feel like sincere belief (LL2-25, p. 614).

2.2 Case classification#

2.3 What the classification shows (analysis)#

• The misconduct cases share structural features: high-value products, few producers with dominant process knowledge, long latency, and eventual litigation that forced documents out. That points to conditions worth screening for, not a universal corporate trait.
• Motive claims are strongest where internal intent was written down. Tobacco and vinyl chloride are the clearest cases; as the audited notes on LL2-07 observe, the chapter’s core evidence is contemporaneous internal intent rather than retrospective judgement. Motive claims are weakest where chapters infer them from outcome: DES, radiation, PCE, mobile phones.
• In several cases the reports read as bad faith what inquiries read as sincere error. BSE and beryllium are the clearest. The difference matters for remedies. Deliberate deception calls for disclosure duties and sanctions. Motivated belief calls for independent challenge, pre-set triggers and “exit or escape opportunities” (LL2-06, pp. 149–150; LL2-25, p. 616).

3. Recurring patterns#

P1. Producers know first, and private knowledge runs ahead of public positions#

Reports. Producers held the first and best information and chose what to reveal (LL2-25, p. 611; LL2-08, pp. 180–186; LL2-07, pp. 158, 161, 164).

Evidence. - Vinyl chloride. Companies learned privately in 1966 that 1% of PVC workers and 6% of vat cleaners had a bone disease, and agreed to “use discretion” (LL2-08, p. 183). US and European firms kept Maltoni’s 250 ppm cancer result from NIOSH in 1973 under a secrecy agreement (pp. 184–186). - PCBs. Monsanto publicly called toxicity claims “simply not true” while its 1969 plan accepted worldwide contamination (LL1-06, p. 65). - Beryllium. In 1974 a Japanese producer reported five cases below the limit to US executives (LL2-06, p. 134). In 1989 a co-author of the limit told the company he could no longer defend it (p. 136). - Other cases: - DBCP: 1958 rat data showed testicular damage (LL2-09, p. 204); - asbestos: industry held unpublished mouse data (LL1-05, p. 54); - benzene: the API’s 1948 statement that zero was “the only absolutely safe level”, alongside its recommendation of “50 ppm or less” (LL1-04, p. 39); - Minamata: Chisso’s own doctor induced the disease in a cat in 1959 (LL2-05, p. 101); - climate: a firm publicly denying “the existence of the problem that they had privately identified” (LL2-25, p. 610).

Hindsight. The pattern has been strengthened. - The PCB documents are now primary: the same manager signed the “simply not true” letter and sat on the committee whose plan called the evidence “beyond questioning” (hindsight LL1-06, Claim 2). - An internal memo of 1984–85 foresaw MTBE well contamination (hindsight LL1-11). - Fluorochemical makers knew of toxicity by 1970 (hindsight LL2-25), and ExxonMobil’s internal climate projections were accurate while its advertorials expressed doubt (hindsight LL2-25, Claim 3).

Cases. - For: 10 (tobacco, vinyl chloride, beryllium, PCBs, asbestos, Minamata, DBCP, benzene, climate, and MTBE on hindsight). - Against or complicating: - PCE: “the science has not been hidden” (LL2-04, pp. 84, 88); - DES: warnings were published and absorbed, with no suppression described (LL1-08, p. 88); - PCBs in the 1930s: the 1937 evidence was published, with public-health officials present (hindsight LL1-06, Claim 3); - CFCs before 1974: genuine ignorance (LL1-17, Table 17.1, p. 192); - BSE: the gap was between departments, not producer concealment (hindsight LL1-16, Claim 3).

Strength: strong. The evidence is contemporaneous internal records in at least seven cases, corroborated since. The generalisation is moderate: it has been tested mainly where litigation exposed records.

Analysis. The asymmetry is sharpest when the producer controls the process or the exposed workforce (Chisso, LL2-05, p. 95; vinyl chloride reactor cleaners, LL2-08, p. 186). Where knowledge sat in published but ignored literature (PCE, DES, 1930s PCBs), the problem is uptake, not concealment.

P2. Manufactured doubt: keeping questions open, and “more research” as delay#

Reports. - Where revenue depends on non-regulation, doubt becomes a product. Tobacco documents speak of research as the “antidote” (1978) and of teams “to keep the controversy alive” (1988) (LL2-07, p. 154). - Firms “can contribute effectively to a ‘paralysis by analysis’” (LL2-25, p. 610). - Asymmetric evidentiary standards mark manufactured doubt: universal or mechanistic proof is demanded of harm but not of safety (LL2-05, pp. 102, 112).

Evidence. - Vinyl chloride. An MCA committee rejected a report it had funded, then accepted it once it said the “etiologic agent of the disease is unknown”. Goodrich declined further research so industry could act as if the cause were unknown (LL2-08, p. 184). - Beryllium. Once the limit was plainly inadequate, the company “initiated more research and convened a conference that propagated the need for further research” (LL2-25, p. 610, from LL2-06, p. 138). - Minamata. Chisso demanded proof of the chemical mechanism while attacking researchers for “reductionism”. Allies promoted amine and wartime-explosives theories (LL2-05, pp. 102–103). - PCBs. Monsanto called for “extensive research, on a worldwide basis” (LL1-06, p. 65). - Lead. The industry-funded paradigm treated ubiquitous lead as “natural” (LL2-03, pp. 56–57). - Climate. The fossil-fuel-funded Global Climate Coalition campaigned against action (LL2-14, p. 330). The chapter calls doubt-making one cause, “not the only one”. - Weaker instances: - benzene consultants introducing “irresolvable” dose-response arguments (protagonist account, LL1-04, pp. 44–45); - Bayer’s “semantic slips” on Gaucho, which Ch 25 itself qualifies with “seems” (LL2-25, p. 610); - PCE, where “artificially and purposefully creating doubt” is uncited (LL2-04, p. 88).

Hindsight. Strengthened as an existence claim. - A US federal court found that tobacco manufacturers had conspired to deny health effects (hindsight LL2-07). - Cross-industry document studies catalogue 28 doubt tactics, five shared by all five industries or organisations examined (Goldberg and Vandenberg 2021; hindsight LL2-02, Claim 6). - The French Senate called the asbestos industry’s joint committee “un ‘modèle’ de lobbying, de communication et de manipulation” that exploited “pseudo incertitudes scientifiques” (hindsight LL1-05). - The explanatory claim, that doubt-making accounts for the scarcity of confirmed false positives (LL2-02, pp. 33–34), remains untested.

Cases. - For: 7 documented (tobacco, vinyl chloride, beryllium, Minamata, PCBs, lead, climate) and 4 weaker (benzene, Gaucho, PCE, Great Lakes: a 1967 Shell report, LL2-25, p. 610; LL1-12, p. 127). - Against or complicating: - Some criticism was valid. The EPA revised its second-hand smoke assessment “in response to valid criticisms” (LL2-07, p. 153), and exposure misclassification was a real question (LL2-07 notes). - Beryllium’s blood test was unvalidated in the 1980s, and scepticism was “not unreasonable for non-scientists” (LL2-06, p. 148). - The Wong brain-cancer retraction the chapter attributes to industry pressure (LL2-08, p. 188) is consistent with later evidence (hindsight LL2-08). - The “tobacco-style” charge in the mobile-phone case is unevidenced (hindsight LL2-21).

Strength. Strong that deliberate doubt-making exists and recurs, on documented intent in at least six cases. Moderate that it caused specific delays, which are rarely measured: the EPA’s release was “probably delayed” (p. 153) but “considerably delayed” on p. 166. Suggestive that it can be diagnosed in real time. Bero’s five diagnostic questions (LL2-07, p. 167) are untested and depend on knowing sponsorship, which is exactly what is hidden.

Analysis. The most useful marker is the asymmetry of demanded proof (LL2-05, p. 112), not the existence of dissent. Legitimate critics apply the same bar to evidence of safety. Delay-seekers demand certainty only of harm, and they shift ground as each objection is answered, as the beryllium defence did from overexposure to insufficient knowledge (LL2-06, pp. 137–138).

P3. Control of the research pipeline and the funding effect#

Reports. Firms need not falsify results. Choosing which studies are done, how they are designed and where they are published shifts the apparent weight of evidence (LL2-25, pp. 610–611; LL2-07, pp. 154–161; LL2-26, pp. 636–637).

Evidence. - Tobacco. Among 106 reviews, tobacco-industry affiliation was the only factor associated with concluding that second-hand smoke is harmless (OR 88.4, 95% CI 16.4–476.5). Quality score and peer review were not (LL2-07, Table 7.1, p. 161). Seventy per cent of one industry body’s “peer-reviewed” projects studied other pollutants (pp. 156–157). None of 11 symposia on passive smoking was peer reviewed (p. 159). - Lead. Health research was funded and run by Ethyl and GM “for over 40 years”, with GM controlling publication and constraining the regulator’s reporting (LL2-03, pp. 50, 56; LL2-25, p. 611). - Minamata. A committee “primarily sponsored by” Chisso “attempted to obscure” the organic-mercury theory (LL2-05, p. 103). The chapter says Kumamoto University took Chisso funding and banned clinical research (p. 104). - BPA. - Government-funded and industry-funded studies split 94 of 104 versus 0 of 11 in finding harm, in a protagonist-compiled sample (LL2-10, p. 228). - Nine of 21 EFSA panel members declared industry links (p. 225). - A consultancy “active for the tobacco industry” claimed credit for the classification outcome (p. 225). - Gaucho. Bayer relied “almost exclusively” on its own research (LL2-25, p. 611). The ministry told public researchers not to measure below Bayer’s 10 ppb detection limit, and Bayer sat on the protocol committee (LL2-16, p. 373). - Other cases: - antimicrobials: industry-supported studies outnumbered independent ones (LL1-09, p. 97); - GM crops: safety evidence that is developer-funded, confidential and underpowered, and about 1% of USDA biotech research spent on risk (LL2-19, pp. 468–470); - research spending: EU hazard research was put at 3% (LL2-27, p. 646) or 1% (LL2-28, p. 679). The figures are unsourced and inconsistent.

Hindsight. - A Cochrane review confirmed that industry sponsorship shifts drug-trial conclusions (RR 1.34) in ways “that cannot be explained by standard ‘Risk of bias’ assessments” (hindsight LL2-07). - A 2016 analysis found industry ties in 40% of Bt-crop efficacy papers, associated with 50% more favourable outcomes (hindsight LL2-19). - A widely cited 2000 herbicide safety review was retracted in December 2025 over undisclosed company authorship, a retraction some toxicologists contest (hindsight LL2-26, LL2-27). - But in chemicals the evidence remains mostly case-level. For systematic reviews, conflicts co-occur with lower quality (Cochrane 2019), so the claim that bias is independent of quality is only partly true (hindsight LL2-07). - Most telling, CLARITY-BPA was publicly funded in both arms and still reproduced the split between guideline and academic studies. That points to paradigm and evidence rules as well as funder (hindsight LL2-10).

Cases. - For: 9 (tobacco, lead, Minamata, BPA, Gaucho, antimicrobials, GM, beryllium’s sponsored textbook (LL2-06, pp. 135–136), and asbestos via the industry-run research council (hindsight LL1-05)). - Against or complicating: - Industry-funded work produced protective findings: low-dose benzene risk and an industry proposal below 1 ppm (hindsight LL1-04); the skin route of beryllium sensitisation, from joint studies with NIOSH (hindsight LL2-06); Schering’s key EE2 life-cycle study (LL2-13, pp. 282, 290). - Critics’ funding goes unexamined. The Hardell mobile-phone studies were part-funded by a telecom operator, disclosed in a footnote but never discussed, while Interphone’s industry money is scrutinised at length (LL2-21, fn 11, p. 517).

Strength. Strong for the association in pharmaceuticals and for documented agenda control in tobacco and lead. Moderate for environmental chemicals, where evidence is case-level and tallies are small and protagonist-compiled. Suggestive for the mechanism: the reasons are “not clear” (LL2-07, p. 167).

Analysis. The robust lesson concerns which studies exist, the agenda, more than how studies are done. That is why hindsight’s most promising remedy is structural: a duty to register commissioned studies before results are known, plus disclosure of raw data. The EU adopted both in Regulation 2019/1381 (hindsight LL1-16, Claim 5; LL2-25).

P4. Changing the rules: standards of proof, metrics, definitions and assessment procedure#

Reports. Interested parties move from contesting studies to reshaping the rules that decide what counts. Ch 25 calls these “political actions”, aimed at “influencing these political and regulatory contexts in the pursuit of profits” (LL2-25, p. 615).

Evidence. - Tobacco (strongest). - “Sound science” and “good epidemiological practice” campaigns promoted “unreasonably high standards of proof” (LL2-07, p. 162). - A proposed rule would have discarded relative risks below 2.0 (p. 164). - US data-access and data-quality laws applied only to publicly funded data (p. 163). - BAT believed legislated “structured risk assessment” would “remove the possibility” of smoking restrictions, and it sought to limit officials’ discretion (Panel 7.1, pp. 164–165). - Standard-setting bodies. - Benzene limits reflected what was “easily achievable”, and corporate scientists sat on the limit-setting committee (LL1-04, pp. 43, 46). - The vinyl chloride limit came from a voluntary body reflecting “what the industry felt was achievable” (LL2-08, p. 182). - Industry shaped ISO tar-yield and ASHRAE “comfort” ventilation standards (LL2-07, p. 162). - Metrics and definitions. - Brush planned to resist tightening the limit “as measured and calculated by Brush” (LL2-06 notes, p. 137). - Minamata’s 1977 criteria were tightened as claims surged, set by those who pay (LL2-05, pp. 108–110). - Gaucho’s assessment asked whether seed treatment was “solely responsible, at national level, for all” losses, a question the authors say “had never been asked” (LL2-16, p. 379). - DBCP’s smell was accepted as a warning, though it registered above the limit itself (LL2-09, p. 205). - Burden of proof through litigation. The US Supreme Court’s 1980 benzene decision required OSHA to make a threshold finding of “significant risk” (LL1-04, p. 40).

Hindsight. - The rule-changing strategy recurred. The US EPA’s 2018–21 “transparency” rule was vacated for determining “outcomes rather than process”. A 2025 executive order revived its logic and discourages “overly precautionary assumptions” (hindsight LL2-07, LL2-26). - The European Risk Forum, a successor to the tobacco-initiated Risk Forum, claims a role in the EU “innovation principle”, now in Horizon Europe (hindsight LL2-07; external/critiques). - The effects are mixed. The EU never adopted the RR < 2.0 rule (LL2-07, p. 164). The EU passed strong tobacco law, and the Court of Justice upheld it partly using the Commission’s impact assessment (hindsight LL2-07). The EU also chose symmetric transparency for food-chain risk assessment. - The rhetoric is no longer only corporate: “sound science” framing has entered US federal science policy (hindsight LL2-27, Claim 3).

Cases. - For: 9 (tobacco, benzene, vinyl chloride, beryllium, Minamata, Gaucho, DBCP, BPA classification (LL2-10, p. 225), and nanomaterials, where a “competitiveness” ground was offered for moving the definition threshold (LL2-22, p. 540)). - Against or complicating: the failed EU rule, and protective rule changes such as EU hazard classes and the 2016 US chemicals law (hindsight LL1-17).

Strength. Strong on intent (documents in tobacco and beryllium; payer-set criteria at Minamata). Mixed on realised effect.

Analysis. Panel 7.1 guards against a genetic fallacy: who promoted a procedure does not settle whether it is good governance (LL2-07 notes; p. 165). The transferable test is whether a proposed rule is symmetric, and whether it removes discretion to act on weight of evidence.

P5. Capture, dual mandates and the state as an interested party#

Reports. Agencies charged with both promoting an industry and protecting the public favour the industry. Examples include MAFF, the USDA, DFO, the French agriculture ministry and the US DOE (LL2-25, p. 612; LL1-15, pp. 157, 164–165). Independence is lesson 10 (LL1-17, p. 194).

Evidence. - BSE. Fear for export confidence and public spending dominated MAFF’s first 20 months. Notifiability was resisted as a danger signal (LL1-15, pp. 159–160). The feed ban was limited to protect renderers’ markets, and offal tissues were chosen by commercial value (p. 160). - Beryllium. The Energy Secretary said worker safety had been the “last priority” (LL2-06, p. 132). The authors attribute the death of OSHA’s 1975 proposal to industry, DoD and DOE lobbying and the 1980 election (p. 135). - Minamata. The trade ministry demanded that wastewater bans “should never be implemented”: “Never stop it!” (LL2-05, p. 99). A health-ministry committee was told not to name the factory and was then dissolved (p. 102). The prefecture co-financed Chisso’s liabilities (p. 109). - Nuclear. Japan’s Diet commission found the accident “profoundly manmade” and cited “regulatory capture” (LL2-18, pp. 441–443). - Fisheries. - A new governor replaced state experts (LL1-02, p. 20). - Harris recommended a catch limit above the level his own analysis required, to avoid “drastic” social consequences (LL1-02, pp. 21–22). - Ministers rejected a cut because “the scientists had been wrong before” (LL2-17, p. 413). - EU catch limits averaged 47% above advice after 2003 (LL2-17, p. 421). - Other cases: - the USDA “tended to align itself” with pesticide makers (LL1-12, p. 128); - French risk assessment was “performed jointly with the industry” by three officials handling 20,000 applications a year (LL2-16, pp. 382–383); - the NNI both promotes and oversees nanotechnology (LL2-22, pp. 546–548; co-authored by Andrew Maynard).

Hindsight. - Existence of the problem is confirmed. The IAEA’s 2015 report found an unchallenged assumption of safety “a major factor”, and Japan’s new regulator itself called it a “safety myth” (hindsight LL2-18). Japan’s Supreme Court held the state liable at Minamata (hindsight LL2-05). North Sea cod’s 2026 catch limit was set against zero-catch advice (LL1-02 summary). The NNI never separated promotion from oversight (hindsight LL2-22). - But causation is complicated by the BSE inquiry. It found MAFF’s policy decisions not producer-biased, and found the Department of Health, which had no producer interest, “as eager as MAFF” to avoid alarm (hindsight LL1-15, LL1-16). - Separated, public, “independent” assessors can also diverge by about 1,000-fold on the same BPA evidence (hindsight LL2-00, LL2-10). - Promote-and-protect conflicts resurface after separation as emergency derogations: 206 EU emergency authorisations for restricted neonicotinoids (hindsight LL2-16, LL2-25).

Cases. - For: 11 (BSE, beryllium, Minamata, nuclear, fisheries ×2, Great Lakes, Gaucho, nanomaterials, DBCP, where the USDA deferred to Shell (LL2-09, p. 205), and hormones, where the Commission’s interest in public confidence and harmonised rules led it to overrule and not publish its own committee (LL1-14, p. 150)). - Complicating: BSE per Phillips; BPA divergence among public bodies.

Strength. Strong for the existence of dual-mandate conflicts and state-side distortion, backed by official inquiries and court findings in several cases. Moderate for the claim that the dual mandate causes delay. The urge to reassure appears in bodies without it.

Analysis. Hindsight on LL2-00 is the key correction: the reliable generalisation concerns the actor controlling exposure and information, whether firm, utility, ministry or regulator. Separating assessor from promoter is, as LL1 said, “necessary, if not sufficient” (LL1-16, p. 179).

P6. Secrecy, legal privilege, and litigation as the main window#

Reports. Knowledge was routed through lawyers to keep it privileged or unrecorded. The Wells memo advised to “cease producing information … helpful to plaintiffs” (LL2-07, p. 156). Without litigation-released documents, most of the tobacco record “would have remained undiscovered” (p. 169). Compensation should be paired with ways to extract corporate information (LL2-08, p. 179).

Evidence. - Firms: - the vinyl chloride secrecy agreement (LL2-08, pp. 184–186); - INBIFO’s unpublished findings on sidestream smoke toxicity, and the lawyer-managed Japanese spousal study (LL2-07, p. 158); - Chisso’s suppression of cat 400 (LL2-05, p. 101); - company removal of records, which biased a benzene reanalysis (LL1-04, p. 42); - asbestos records refused to Selikoff (LL1-05, p. 55). - Public bodies: - the EU’s unpublished Lamming findings (LL1-14, p. 150); - a DoE–MAFF–Brunel contract that held back the national effluent-oestrogen survey from about 1990 to 1994 (LL2-13, p. 284); - the edited SEAC draft on BSE (LL1-15, p. 161); - Minamata’s dissolved committee (LL2-05, p. 102).

Hindsight. Litigation and archives remain the main route by which concealed knowledge surfaces, typically decades later, for example the vinyl chloride archive, the 1969 PCB plan and the MTBE memo (hindsight LL2-25, lesson 2). New non-litigation windows exist: - EU study pre-notification and proactive publication (Regulation 2019/1381); - the Court of Justice’s Blaise ruling against giving applicant studies “preponderant weight”; - EPA’s TSCA §8(d) order for unpublished vinyl chloride studies (hindsight LL1-16, LL2-08).

Cases. For: 12 (the five firm-side and four public-body episodes above, plus DBCP, where manufacturer knowledge did not reach users (LL2-09, p. 211), Gaucho and BSE).

Strength: strong that secrecy occurred and delayed recognition. The implied observability bias is also well supported, and is acknowledged by Bero (LL2-07, p. 152) and by later synthesisers (hindsight LL2-07).

Analysis. Secrecy is not only corporate: public bodies withheld reassuring findings (hormones) as well as alarming ones (EE2, BSE, Minamata). The common driver is protecting a decision already taken.

P7. Treatment of early warners and whistleblowers#

Reports. Warners were dismissed, marginalised or retaliated against, and protection should not depend on being proved right (LL2-24, pp. 582–585; LL2-03, pp. 70–71; LL2-16, lesson 8; LL1-16, p. 179, “shooting the messenger”).

Evidence. - Lead. Henderson’s contracts were terminated after he refused industry-controlled work (LL2-03, p. 51; the link rests on timing). Patterson faced contract losses (p. 58), sourced mainly to his own letter. Needleman’s “character assassination” is given no detail (p. 61). - Fisheries. The inshore fishers’ Keats report was dismissed as “biased pseudoscience” (LL1-02, p. 21). The chapter describes civil-service silence inside DFO (LL2-17, pp. 414–415). - Gaucho. Anonymous testimonies describe legal threats and cancelled programmes (LL2-16, p. 378). Bayer sued three beekeeper leaders for “discrediting” Gaucho and lost all three cases (p. 380). - DBCP. Dole sued a Swedish documentary; the suit was later ruled a SLAPP (LL2-09, Box 9.2, p. 208). - Minamata. The company doctor Hosokawa resigned without publishing, and a mother saw congenital disease first (LL2-05, pp. 101, 105–106). - Lay and worker warners were often right: the DBCP workers who compared notes at lunch (LL2-09, p. 204); Love Canal residents (LL1-12, p. 127); factory inspector Deane (LL1-05, p. 53).

Hindsight. - Flint regulators met independent testers and a paediatrician with “aggressive dismissal, belittlement, and attempts to discredit” (Flint Water Advisory Task Force; hindsight LL2-03). - The protections have gaps. The EU Whistleblower Directive covers breaches of EU law, not warnings about lawful hazards. France abolished its 2013 health-and-environment alert commission in 2026. The 2024 anti-SLAPP Directive names researchers (hindsight LL2-24, LL2-16). - A German whistleblower prize went in 2015 to the author of a retracted GM feeding study (hindsight LL2-24). This illustrates the false-alarm cost that Gee’s own panel acknowledged.

Cases. For: 8 or more. Complicating: the warners in these chapters were selected for vindication. Several accounts are the warners’ own. Some celebrated warners were wrong (mobile phones, LL2-21; the GM feeding study, LL2-19).

Strength: moderate. The existence of retaliation is documented in some cases (Gaucho suits, the SLAPP ruling, Flint), but often rests on timing or first-person accounts. Protection must operate before vindication, and the evidence base is selected on vindication.

P8. Cost-shifting, liability, and the incentive not to know or admit#

Reports. - Social costs reach firms only through liability, regulation or reputation, and each channel leaks (LL2-25, pp. 608–612). - Costs fell on families, health services, insurers and building owners (asbestos), and on public water budgets (MTBE) (LL2-25, p. 612; LL1-11, p. 119). - Manville’s 1982 bankruptcy was “a means of dealing with” asbestos claims (LL2-25, p. 612). - Proof rules decide who absorbs uncertainty (LL2-24, pp. 586–593). - Research that might confirm harm is itself a cost (LL2-25, p. 609).

Evidence. Liability also deters admission: - Monsanto’s plan rejected stopping production partly because “we would be admitting guilt by our actions” (LL1-06, p. 65, as corrected by hindsight); - the beryllium limit was “fundamental to our product liability defense” (LL2-06, p. 137); - Guidotti argues that as uncertainty fell “the stakes increased”, making defence of earlier positions more attractive (LL2-06, pp. 148–149).

Other routes: - DBCP liability was shifted by an indemnity contract, not removed (LL2-09, p. 209); - nuclear liability caps socialise tail risk (LL2-18, pp. 445–446); - the Minamata state co-financed the polluter while deciding who counted as a victim (LL2-05, p. 109).

Hindsight. Strengthened. - The Manville Trust pays asbestos claimants 5.6% of scheduled value. A US appeals court dismissed a strategic asbestos bankruptcy in 2023 (hindsight LL2-25). - About a dozen US states have settled PCB claims with Monsanto without admission of liability (hindsight LL1-06). - Fukushima costs run to about 100 times the revised European liability cap (hindsight LL2-18). - The 2024 EU Product Liability Directive eases proof but keeps the development-risks defence, which protects exactly the harms unknowable at launch (hindsight LL2-25). - No assurance bonding for uncertain hazards was found (hindsight LL2-25). - A French Senate inquiry found that pooled asbestos compensation diluted employer accountability (hindsight LL1-05).

Cases. For: 10 (asbestos, MTBE, PCBs, beryllium, DBCP, nuclear, Minamata, fisheries, EE2 and SO2 (LL1-10, pp. 103–104)).

Strength. Strong for cost-shifting and liability gaps. Moderate for liability deterring admission (two documented statements plus an organisational analysis). Suggestive for Guidotti’s case for “exit or escape opportunities”: the sequence fits, but interest-alignment explains it equally well (hindsight LL2-06).

P9. Trade, export and jurisdictional arbitrage#

Reports. Restriction in one place displaces hazards to less-regulated places, and exporters obstruct information-sharing (LL2-A3, pp. 724–726). Trade law shapes precautionary choices (LL1-05, Box 5.1, p. 57; LL1-14, p. 153).

Evidence. - Hazards moved abroad: - DBCP exports continued after the US ban, with English-only labels (LL2-09, pp. 207–209); - DES stayed in use abroad for years (LL1-08, p. 89); - Turkish shoemakers switched to benzene because it was cheaper (LL1-04, p. 39); - older vinyl chloride technologies persisted in lower-resource settings (LL2-08, p. 181); - lead exports and e-waste (LL2-03, pp. 66–69). - Information blocked. Canada led opposition to listing chrysotile under the Rotterdam Convention from 2004 and blocked it in 2011; only in 2012 did Ottawa say it would stop opposing (LL2-A3, pp. 725–726). Russian pressure held up Thailand’s move to a ban (p. 724). - Trade disputes and competition: - The WTO upheld the French asbestos ban on qualitative evidence (LL1-05, p. 57). - The US retaliated against, then won against, the EU hormone ban (LL1-14, p. 153). - The US claimed its unilateral aerosol ban cost it market share (LL1-07, p. 80), a first-mover penalty. - Governments invoked free movement to justify national inaction on invasive species (LL2-20, Box 20.4, p. 501).

Hindsight. - The hormone dispute was settled by beef quotas, not science, at other exporters’ expense (hindsight LL1-14). - Asbestos use continued at about 930,000 t in 2025, mainly in Asia (hindsight LL1-05). - DBCP was never listed under the Rotterdam Convention (hindsight LL2-09). - Pfizer’s 1999 argument that imports undermined the EU growth-promoter ban was answered only in 2026. The EU extended the ban to imports, which immediately suspended some Brazilian products (hindsight LL1-09). - Canada, the WTO complainant, banned asbestos itself in 2018 (hindsight LL1-05).

Cases. For: 10.

Strength. Strong for displacement and exporter obstruction (asbestos, DBCP). Moderate for the first-mover penalty (one claimed instance, corroborated externally). Moderate that trade adjudication decides cases on procedure rather than science (hormones).

P10. Countervailing interests: when precaution arrives because another interest aligns with it#

Reports. Industries harmed by a hazard become early warners and allies of precaution, though they “seem to be a minority” (LL2-25, p. 609).

Evidence. - Harmed sectors pushed for action: - Arcachon oyster farmers on TBT (LL1-13, p. 136; LL2-25, p. 609); - Newfoundland inshore fishers (LL1-02, p. 21); - French beekeepers (LL2-16, pp. 372–373, 380); - US and Canadian insurers, who refused asbestos cover in 1918 (LL1-05, p. 54). - Users moved before producers. Swedish farmers asked for a growth-promoter ban in 1984 to protect consumer confidence, and Danish farmers and feed makers halted avoparcin voluntarily (LL1-09, pp. 95–96). Retailers dropped BPA (LL2-10, p. 225; LL2-27, p. 647). Gee notes responsible behaviour came mostly from “companies selling hazardous products rather than by their manufacturers” (LL2-27, p. 647). - Commercial alignment: - GM wanted lead out to protect catalytic converters, and the phase-out followed (LL2-03, p. 60); - refinery profits aligned with SO2 abatement (LL1-10, p. 104); - Du Pont’s turn on CFCs was partly commercial, per sources in the notes (LL1-07 notes). - Counter-pressure against precaution. Water companies and pharmaceutical firms lobbied against the cost of EE2 controls (LL2-13, p. 294). Some horticulture firms favoured legislation so that competitors could not undercut them (LL2-20, pp. 498–499).

Hindsight. - Harmed third parties are most effective when they have legal standing, as the beekeepers did through the courts (hindsight LL2-25). - Investor pressure is volatile: majority votes in 2017–21 had faded to 21% or less by 2024 (hindsight LL2-25). - A rival refiner declined MTBE (hindsight LL2-25), and Materion co-drafted a stricter beryllium limit with the steelworkers’ union (hindsight LL2-06).

Cases. For: 12 or more.

Strength: moderate. Instances are many and documented, but the causal weight of interest-alignment against other drivers is rarely separated.

Analysis. One of the most transferable and least emphasised lessons: action often waited less for proof than for an organised interest that bore the harm, held standing, or profited from the alternative.

4. The Part D analysis: why business did not act on early warnings (LL2-25)#

What the chapter says. Le Menestrel and Rode (business ethics and behavioural economics) re-read cases from both volumes. Their headline: “In virtually all reviewed cases it was perceived to be profitable for industries to continue using potentially harmful products or operations” (p. 607). They give three lessons (pp. 615–616): 1. Value conflict under economic rationality. Ethical values count “if and only if” they add to expected profit (p. 608). Harms to others enter decisions only through liability, regulation or reputation (p. 608), and each channel leaks: - publics are unaware or feel unaffected, and firms can shape opinion “when this appears cheaper” than cutting the risk (p. 609); - regulators depend on firm data and hold dual mandates (pp. 611–613); - liability places the burden of proof on victims and can be escaped through insolvency (p. 612). 2. “Political actions.” These are defined as efforts “not aimed at maximising profits within the political and regulatory contexts but rather … at influencing these political and regulatory contexts in the pursuit of profits” (p. 615). They can erode society’s warning capacity itself, as tobacco’s discrediting of institutions did (p. 615). 3. Psychology and culture. Five mechanisms are identified: - the neglect of rare, unexperienced harms; - discounting of distant harms; - self-serving bias that makes uncertainty “a welcome ‘excuse’” (p. 614); - an ideology that profit-seeking itself serves society (p. 616); - cultures of denial (“the blunder was denial”, Sells, Manville, p. 614).

Recommendations (advocacy, untested): internalise costs and consider assurance bonding (pp. 611–612, 616); have public institutions analyse and publicly disclose sector “dilemmas and temptations” (p. 616); make corporate political actions transparent (p. 617). The authors warn that blaming business “with hindsight” may not be constructive, and that failure to act is “not necessarily” wilful (p. 616). Secrecy about political actions “can be seen as a signal” of bad faith (p. 617).

Assessment. - Strengths. - The three-channel model explains why precaution is unlikely where harm does not feed back to the decider. It is illustrated by several cases, though the reputation leak is mostly argued (moderate). - The distinction between business and political actions is the chapter’s most useful contribution (moderate as analysis). It lets an analyst be charitable about decisions within the rules and strict about efforts to change the rules or the evidence. It also defeats the “we complied with the law” defence where the firm shaped the law. - The claim that firms can bias research without having to “manipulate results or engage in other forms of misconduct” (p. 610) is well corroborated since (hindsight LL2-25, Claim 4). - Weaknesses. - Selection on the dependent variable. The cases were chosen as failures, so “virtually all” is close to built in. There is no comparison group of firms that acted (hindsight LL2-25). - Two causal stories. A calculative story (deliberate doubt-making) and an unconscious one (bias, culture) sit side by side, unreconciled. - Second-hand and thin evidence. Some evidence is thin, including an unpublished mimeo cited for consumer reactions to disclosure (p. 609). There are also errors: Canada’s DFO is called “US”, and the Friedman quotation is truncated before “without deception or fraud”. - A double-edged success story. The one success story, Manville’s 1986 voluntary cancer label on fiberglass (p. 615), is double-edged. The warning was later judged largely unwarranted, and the chapter’s “nearly five years” timeline is wrong: glass wool stayed in IARC Group 2B until 2001 (hindsight LL2-25, Claim 7). - A partly contradictory example. The German post-Fukushima phase-out, offered as a case where a vivid event sped up regulation (p. 613), restored a 2002 plan; one study puts its social cost at about USD 12 billion a year. That fits the chapter’s own caution about over-reaction (hindsight LL2-25). - Hindsight verdicts (hindsight LL2-25): - Strengthened: claims 3–5, on private recognition versus public denial, bias without misconduct, and cost-shifting and dual mandates. - Held up as description, but not a base rate: claim 1. - Adopted only partially: the remedies, and never for uncertain hazards. The EU introduced study pre-notification (2019/2021), a mandatory lobbying register (2021) and lobbying disclosure in reporting standards (2023), later narrowed (2026). - Evidence of heterogeneity: firms differed within sectors. A rival refiner declined MTBE, and Materion co-drafted a tenfold-tighter limit.

Overall strength. Moderate as an explanation of why failures happened; not usable as a base rate for how firms respond. Pair it with Gee’s observation that responsible behaviour came mostly from users rather than manufacturers (LL2-27, p. 647), and with the counter-examples above.

5. Counter-evidence, complications and critiques#

1. Selection and observability. - All LL1 cases and nearly all LL2 historical cases were chosen because harm materialised (LL1-00, pp. 11–13; LL2-00, pp. 9–10). - Misconduct surfaces mainly through litigation discovery (LL2-07, pp. 152, 169). Litigated sectors are therefore over-represented, and the eight-strategy template has been checked mainly against them. - Critics’ strongest point is against frequency readings: the reports cannot say how often warnings of comparable strength were wrong. Marchant accepts the mechanism claims, including producer control of research and slow response once evidence existed (external/critiques §4).
2. Protagonist authorship and asymmetric disclosure. Many authors had stakes in their stories. Some interests are undisclosed or undiscussed in the chapters: - Expert-witness roles: Michaels was a former beryllium expert witness and later head of OSHA (LL2-00). Cranor was the plaintiffs’ methodology expert in Milward, the case he praises (hindsight LL2-24). Ozonoff had a litigation role in PCE/TCE cases (hindsight LL2-04). Grandjean was a disclosed PFAS litigation expert (hindsight LL2-26). The IJOEH industry-document studies came from authors serving as expert witnesses (LL2-07, p. 169). - Advisory or institutional roles: Jim Bridges advised the EU at the WTO (LL1-14 digest). The 2013 hormones “update” is a Commission litigator’s summary (LL2-A3, pp. 736–737). The growth-promoter “vindication” rests largely on the authors’ own institutions (hindsight LL1-09). - Funding: the Hardell group had telecom-operator funding (LL2-21, fn 11). - Advocacy-commissioned figures: the Greenpeace-commissioned Dutch asbestos figure (hindsight LL1-05); the Yablokov Chernobyl estimate; a renewables-body liability estimate (LL2-18 digest).

The reports rarely apply to allies the scrutiny they apply to industry. Bero concedes that both sides “construct” evidence but examines only industry (LL2-07, pp. 152, 166). 3. Motive attributions exceed the documents in several chapters. - “Covertly subordinated” health (LL1-15, p. 164) conflicts with the BSE Inquiry’s findings. - “Economic interests predominated” for DES rests on no internal documents (LL1-08, p. 90). - “Purposefully creating doubt” for PCE is uncited (LL2-04, p. 88). - “Politics entered the scene” for radiation is unsourced (LL1-03, p. 34). - The mobile-phone “spin” charge is unevidenced (hindsight LL2-21). 4. Genuine uncertainty was sometimes real. Some episodes read as manufactured doubt contained valid science: - the EPA’s second-hand smoke revisions (LL2-07, p. 153); - the time-lag bias in publication, which the industry overstated but did not invent (LL2-07 notes); - the validity of the beryllium blood test (LL2-06, p. 148); - the Wong retraction (hindsight LL2-08); - the vinyl chloride multi-site cancer list, weakened by IARC 2012 (hindsight LL2-08). 5. Public authorities as primary causes. At Minamata the state is judicially liable. Flint and Camp Lejeune (hindsight LL2-00) show public bodies causing and concealing harm. DDT use is driven by public-health programmes (LL2-11). The “irresponsible corporations” generalisation (LL2-00, p. 11) is weakened. 6. Firms changed course, sometimes early. - Materion co-drafted the 2017 beryllium standard (hindsight LL2-06). - Industry-linked benzene work proposed a sub-1 ppm limit (hindsight LL1-04). - The Danish feed industry, Swedish farmers, Du Pont (1986–88) and S.C. Johnson (1975, misattributed in LL2-27, p. 647) all moved.

Obstruction by business is recurring, not constant (hindsight LL2-00, Claim 2). 7. Organised alarms exist too. - The MMR–autism alarm caused lasting harm and entered US official communication after 2025 (hindsight LL2-02). - LL2’s concluding chapter repeats a GM health claim resting on a study later retracted (LL2-28, p. 674; hindsight LL2-19). - A lens that tests only for manufactured doubt will miss manufactured alarm. 8. Evidence tools cut both ways. - Impact assessment helped delay EU tobacco law and also helped the Court uphold it (hindsight LL2-07). - “Transparency” can target public data (US data laws) or all data (EU Regulation 2019/1381). - Rhetoric about “sound science” is now used by governments as well as firms (hindsight LL2-27). 9. Forced precaution has costs that interested parties can legitimately raise. - Sugar-beet yields fell after the French neonicotinoid ban (hindsight LL2-16). - Hormone-ban trade retaliation was real (LL1-14, p. 153). - The UK Food Standards Agency judged in 2005 that keeping one BSE measure (the Over Thirty Months rule) would cost about £2bn per death prevented (hindsight LL1-15).

Cost arguments are not all doubt-making. But industry forecasts were often inflated: vinyl chloride compliance costs were overestimated about fourfold (hindsight LL2-08), and lead and hormone cost claims were overstated (LL2-03, p. 60; LL1-14, p. 150). 10. Critics’ reading of the reports. Brussels critics read LL2 as agency advocacy, and the European Risk Forum reframed the debate procedurally through the “innovation principle” (external/critiques §§6–7). Neither touches the documented mechanisms; both bear on the reports’ generalisations about corporate blame and false-alarm rarity.

6. Technology-neutral diagnostic questions#

Each question is tied to the patterns and evidence above. They are meant to be asked of any emerging technology and of those developing, promoting and regulating it.

1. What does the developer know internally that is not in the public or regulatory record, and what would reveal a gap? Examples: registration of commissioned studies before results are known; access to raw data; comparison of technical and public-facing documents from the same period. (P1, P6; LL2-08, pp. 183–186; LL1-06, p. 65; hindsight LL2-25, lessons 1–3.)
2. Who funds, designs and controls the studies behind the safety claims, and which questions are not being studied at all? (P3; LL2-07, pp. 155–161; LL2-03, p. 56; LL2-16, p. 373; LL2-26, pp. 626–629.)
3. Are demands for “more research” or “sound science” coming from parties who benefit from delay, and do they apply the same evidentiary bar to evidence of safety as to evidence of harm? (P2; LL2-05, p. 112; LL2-06, p. 138; LL2-07, p. 162.)
4. Is anyone trying to change the rules rather than contest the evidence? Rules here include standards of proof, the metric of compliance, detection limits, case definitions and the question put to assessors. If so, would the proposed rule apply symmetrically? (P4; LL2-07, pp. 162–165; LL2-06, p. 137; LL2-05, pp. 108–110; LL2-16, pp. 373, 379.)
5. Does any body that assesses or oversees the risk also have a mandate, budget or political incentive to promote the activity? Who, beyond the promoter, has reasons to reassure? (P5; LL1-15, pp. 157–162; LL2-06, p. 132; LL2-22, pp. 546–548; LL2-18, pp. 441–443; hindsight LL1-15 on the Department of Health.)
6. Have public assurances of safety gone beyond what the evidence and advisers support? Such assurances make later precaution look like an admission of error. (LL1-15, pp. 161–162; LL2-06, p. 137; LL1-06, p. 65.)
7. Is the conduct in question within the rules (“business action”) or aimed at changing rules, evidence or opinion (“political action”), and is the latter disclosed? (LL2-25, pp. 615, 617; hindsight LL2-25, lesson 11.)
8. If harm emerges decades later, who will pay, and does the liability design reward finding out or not finding out? Design features include the burden of proof, caps, state-of-knowledge defences, insolvency and financial assurance. (P8; LL2-25, pp. 609, 612; LL2-24, pp. 588–603; LL2-18, pp. 445–446; hindsight LL2-25, lessons 4–6.)
9. How would a person inside or outside the organisation raise a concern about a lawful but possibly hazardous activity, and what protects them before they are proved right? (P7; LL2-24, pp. 582–585; LL2-16, pp. 378, 380; LL2-09, p. 208; hindsight LL2-24 on the scope of the EU directive.)
10. Which parties bear the harm, and do they have standing, data and voice to act as a counterweight? Which harmed parties have none? (P10; LL2-25, p. 609; LL1-13, p. 136; LL2-16, p. 380; LL1-02, p. 21.)
11. Where restriction happens in one jurisdiction, where does the activity go, and who is positioned to block information-sharing or trade measures? (P9; LL2-09, pp. 207–209; LL2-A3, pp. 724–726; LL1-04, p. 39; hindsight LL1-09 on import rules.)
12. Are claims about compliance costs, infeasibility or economic ruin being checked against outcomes in comparable past cases? (LL2-08, pp. 187, 192 and hindsight, about fourfold overstatement; LL2-03, p. 60; LL1-14, p. 150; LL2-06, p. 135.)
13. Once a provisional limit, definition or classification is adopted, what forces its review, and whose interests attach to keeping it? (LL2-06, pp. 133, 137, 150; LL2-05, pp. 109–110; LL1-04, pp. 41, 43; hindsight LL2-06, lesson 1.)
14. Is the evidence of interested distortion documented, or inferred from timing and outcome? Are the funding, legal roles and institutional stakes of critics and advocates disclosed to the same standard as the developer’s? (Section 5, items 2–3; LL2-21, fn 11; LL2-04, p. 88; LL1-08, p. 90.)
15. Would the same scrutiny catch an unfounded alarm promoted by an interested advocate, and is there an agreed test for lifting a precaution? (Section 5, item 7; hindsight LL2-02 on MMR; hindsight LL2-19; hindsight LL1-15 on exit criteria.)

Method note#

• Read: all 47 digests; full notes for LL2-07 and LL2-25 and targeted sections of the notes for LL1-06, LL2-03, LL2-05, LL2-06, LL2-10, LL2-13, LL2-21 and LL2-27, plus searches across all notes; hindsight overviews and relevant claims for 25 sections; the external critiques file.
• Checked against source text (resumed pass, 26 September 2026): every phrase quoted from the reports was matched against working/text/chunks/, and key figures and page numbers were spot-checked. Six misquotations or paraphrases shown as quotations, and two page references, were corrected. Other page references follow the audited notes and digests. Hindsight items that rest on secondary sources (later PCB settlements, some Minamata rulings, the Los Angeles DBCP fraud dismissals) keep that status.