LL2-27 digest — Ch27 More or less precaution? (David Gee)#

EEA Report No 1/2013, pp. 643–669 (text pp. 643–662). There are no panels. Box 27.5 consists of “edited extracts” from René von Schomberg (European Commission research directorate, writing in a personal capacity). Gee originated the Late Lessons project, was EEA editor for both volumes and co-wrote three of the cases he summarises (Chs 3, 8, 21). This is the project lead’s own case for wider use of the precautionary principle (PP), not an independent review.

Core argument#

The PP has two roles: - a justification for acting earlier on early warnings; - a trigger for debate about which innovation pathways a society wants (p. 644).

Seven barriers “explain much about the decades‑long delay between warnings and action” (p. 645): 1. corporate opposition; 2. confusion over definitions; 3. complex, multicausal systems; 4. the clash between the proof needed for scientific causation and the lower bar needed for timely policy; 5. skewed cost analysis; 6. short-termism; 7. little public engagement.

The chapter supplies a set of tools: - An EEA working definition of the PP (p. 649): act under “complexity, uncertainty and ignorance” against serious or irreversible threats, “using an appropriate strength of scientific evidence” and weighing pros and cons “and their distribution”. - A typology of knowledge states (Table 27.1, p. 656). - A strength-of-evidence scale (Table 27.2, p. 658). - Twelve “criteria for action” (Box 27.4, p. 653). - A participatory risk-analysis cycle (Fig. 27.2, p. 660).

Key evidence#

• Product defence. Industries controlled research, used reassuring language (“natural”, “sound science”) and manufactured doubt (pp. 645–646). The leaded-petrol industry held a research “virtual monopoly” from 1925 to the 1970s (p. 646).
• Skewed research spending. Only about 3 % of EUR 28.5 billion of EU nano, bio and IT research went on hazards (unsourced here), and 1 % of USDA biotech research (p. 646).
• Profit and externalised costs. Le Menestrel and Rode (Ch 25) found that continuing was perceived as profitable in “virtually all” reviewed cases. Gee attributes the dominance of short-term corporate interests “mainly” to externalised costs (p. 647).
• Speed depends on the signal. Once compelling human evidence appeared, just 4–7 cases of very rare cancers or of sperm reduction triggered prompt action on DES, VCM and DBCP (p. 645). Tobacco took more than 40 years (p. 647).
• Causation criteria misused. Under multicausality, missing Bradford Hill features are weak grounds for dismissal (pp. 652–654). Methods lean toward false negatives, and results can track funders (p. 654).
• Ignorance is not uncertainty. Mesothelioma and the ozone hole were “surprises” from ignorance (p. 654). Precaution matters most where the knowledge-to-ignorance ratio is low (p. 655).
• “Differential standards”. Bradford Hill argued for them (pp. 656–657). Sweden’s 1973 law demands strong evidence of safety from manufacturers but only a “scientific suspicion of risk” from regulators (p. 658).
• Hidden choices. “Not established” judgements seldom say who bears the consequences of error (“risk takers or risk makers”) or what purpose the evidence is being judged for. And “no evidence of harm” is not “evidence of no harm” (pp. 657–658).
• Uncounted costs and lock-in. Lost trust after false reassurance “can sometimes be as significant as the economic costs” (BSE, Chernobyl, Fukushima; p. 658). Late action “consolidated technological monopolies … at unrealistically low prices” (p. 659).

Main mechanisms#

• Disputes about the standard of proof are presented as disputes about facts. Failing to acknowledge different strengths of evidence “has led to several ill‑founded debates” (p. 657).
• Expert habits lean toward false negatives: reductionism, “context stripping”, misuse of statistical significance, hubris.
• Who owns and funds research shapes what becomes known.
• Firms become locked into defending their product.
• Costs of action are concentrated while benefits are diffuse.
• Late action produces technological lock-in.
• Short institutional horizons meet long, irreversible lags.
• Harms fall unevenly. The definition requires attention to distribution across groups, regions and generations (p. 649). The differing consequences for “risk takers or risk makers” are “seldom discussed” (p. 658). Those exposed before birth get no offsetting benefit (p. 650).

Transferable insights (technology-neutral)#

1. Set evidence thresholds by purpose and consequence, not by the standard of scientific proof (pp. 649, 656–658). Strong.
2. The burden and standard of proof can be split between the parties (p. 658). Strong as a description of legal designs.
3. “No evidence” often means “never studied”, and “not established” judgements rarely say who bears the error (pp. 657–658). Strong as a logical and communication point. How often this happened across cases is asserted here and evidenced elsewhere (Ch 26).
4. Missing causal indicators in complex systems are weak grounds for dismissal (pp. 652–654). Moderate–strong.
5. Control and funding of research shape what is known (pp. 646, 654). Moderate. The funding effect is strong in pharmaceuticals and tobacco.
6. The defence repertoire is to control research, use reassuring language and manufacture doubt (pp. 645–646). Strong historically, suggestive as prediction.
7. Concentrated costs versus diffuse benefits bias decisions toward inaction (pp. 647, 659). Strong as a political-economy argument. The chapter illustrates it by cross-reference to Ch 23 but does not measure it.
8. Delay entrenches incumbents and suppresses substitutes (p. 659). Moderate/suggestive.
9. Ignorance calls for proxies, analogy, monitoring and adaptable options (pp. 654–656). Strong as a concept, moderate as guidance.
10. Specific, prompt harms trigger action, while diffuse, delayed, multicausal harms are disadvantaged (pp. 645, 650). Moderate. This is an inference from the chapter’s juxtapositions, not its stated thesis.
11. False reassurance costs trust (p. 658). Moderate.
12. Value judgements run from framing onward, so participation should include framing (pp. 659–660). Moderate.
13. Institutional time horizons are mismatched with hazard lags (p. 659). Moderate–strong as diagnosis. It is plausible given the documented lags, but Section 27.7 is two paragraphs with no case analysis.

Main caveats#

• One side argued. The title asks “more or less?” but only “more” is argued. The main critics (Sunstein, Wiener, Graham and Hsia, the European Risk Forum) appear only in the bibliography. Many supportive works are listed but uncited too.
• Some balance shown. The chapter concedes that Carson “overstates” the case. It reports Le Menestrel and Rode’s warning against hindsight blame. It calls the antibiotics ban only “arguably” precautionary. It notes that balancing warnings against false alarms is “very challenging” (p. 660). It sets a floor of “some plausible evidence” (p. 655). And it concedes “Mistakes will be made” (p. 662).
• Missing counterweights. There is little on risk–risk trade-offs, regrettable substitution or when to lift measures. The definition’s “pros and cons of action” leaves room for them, but the chapter gives no case where precaution itself did harm.
• Contested examples. Two of the five cases the chapter says “describe precautionary actions” are contested as successes: the hormones ban and EU GMO restrictions.
• Mixed scale. Table 27.2 puts legal, statistical and IPCC-derived standards on one probability scale. “Balance of probabilities” sits in a 33–65 % band.
• Selective Interphone quote. It omits the study’s lead finding, “Overall, no increase in risk”, which was checked against the abstract. The point about binary misreading survives, but the example is framed to favour the chapter.
• Hill. The chapter calls his dose-response feature a “linear” criterion; Hill did not require linearity.
• Probable errors. “Johnson & Johnson … 1977” is likely S.C. Johnson, 1975. “Baby toys” should be bottles (Directive 2011/8/EU). France’s TBT ban is dated “1984”, but is usually dated 1982. The “financial collapse of 2009” is conventionally dated 2007–08.

Later evidence#

• Supportive: BPA (EFSA 2023 tolerable daily intake; EU food-contact ban 2024), neonicotinoid bans, reform of statistical-significance practice, IPCC confidence.
• Less supportive: mobile phones (a 2024 WHO-commissioned review found no brain-cancer link) and GM food.

The chapter’s tools have held up better than some of its example “emerging hazards”.