LL2-19 — Ch19 Hungry for innovation: from GM crops to agroecology#

Late lessons from early warnings: science, precaution, innovation (EEA Report No 1/2013), Part C “Emerging issues”. Chapter title as printed: “Hungry for innovation: pathways from GM crops to agroecology”. Report pages 458–485 (PDF pages 460–487). Body text pp. 458–478; references pp. 478–485.

Reading notes: the full text extract was read page by page to the final marker (PDF 487 / report 485). The PDF was checked visually for all body pages except pp. 459 and 461, including both boxes, all three figures and the footnotes. The extraction was clean. Author affiliations are not on the chapter pages. They come from the report’s contributor annex (pp. 685–696) and acknowledgements (p. 5). A few passages elsewhere in the report that refer to this chapter (pp. 10, 644, 646, 660) are cited and labelled as outside the section. Anything marked [background] is my general knowledge rather than the report, and needs verifying before use. The EEA authorises reproduction with acknowledgement (copyright notice, report p. 2).

Authors and standpoint#

Authors (from the report’s annex, not the chapter): - David A. Quist: senior scientist at GenØk – Centre for Biosafety, Tromsø, since 2005. His PhD (UC Berkeley, 2004) studied transgenic DNA introgressed into traditional maize in Oaxaca. He is a founding member of the European Network of Scientists for Social and Environmental Responsibility (ENSSER) and was Norway’s representative on the Cartagena Protocol’s Ad Hoc Technical Expert Group (AHTEG) on risk assessment (p. 696). He was also on the LL2 editorial team (p. 5). - Jack A. Heinemann: professor of genetics and molecular biology at the University of Canterbury (NZ) and director of its Centre for Integrated Research in Biosafety (INBI). He served on the Cartagena AHTEG on risk assessment (p. 691). - Anne I. Myhr: senior scientist at GenØk since 2003. Her PhD was on precaution and ethics in GMO risk governance. She “has been a member” of Norway’s Scientific Committee for Food Safety (VKM) (the annex does not say whether this is current) and has worked on GMO risk-assessment capacity building in developing countries and on socio-economic impacts under Cartagena (p. 695). - Iulie Aslaksen: ecological economist at Statistics Norway since 1981. Her work covers biodiversity policy, sustainable-development indicators and precautionary perspectives (p. 685). - Silvio Funtowicz: professor at the Centre for the Study of the Sciences and the Humanities, University of Bergen. He was at the EC Joint Research Centre (early 1990s–2011) and has published on post-normal science, including the book Uncertainty and Quality in Science for Policy with Jerry Ravetz (p. 689). [background] He and Ravetz are generally credited with originating the post-normal science concept. He was also on the LL project advisory board (p. 5).

Mix. Three biosafety scientists, an ecological economist and a science-for-policy scholar. No agronomist, plant breeder, development or agricultural economist, farmer organisation, regulator or industry voice is among them.

Stance. Openly normative. The chapter is critical of GM crops as developed so far and of IP-driven innovation policy, and advocates agroecology, participatory research and public-good research priorities. Its intellectual frame comes from science and technology studies and post-normal science: the STEPS “3D agenda”, Stirling on lock-in, Rudner on value-laden science. It is structured less as a classic late-lessons case history than as an innovation-policy contrast between two “pathways”. Box 19.2 carries the classic early-warning argument under the question-marked title “GM crops: a late lesson case in the making?” (p. 468).

Panels. None. There is no industry, regulator, crop-science or development-economics commentary and no right of reply, on one of the most polarised topics in the volume. (For fairness: the report states that “Case study chapters have been peer reviewed from a broad range of perspectives” (p. 5, outside section), but reviewers and their comments are not identified.)

Self-citation and network ties. - Heinemann’s own works are cited for central framing claims: Hope not Hype (2009, Third World Network) on pp. 460, 463, 464, 467, 470 and 474, and an in-press UNCTAD chapter on pp. 459, 465 and 475. He co-authored Kiers et al. 2008 (pp. 461, 464, 475). Further Heinemann items: Heinemann & Kurenbach 2008/2008b (Third World Network; p. 462, the sole source for Box 19.1’s verdict that HT transfer “has not, and will not” benefit poor farmers), Heinemann 2008a (The Ecologist; p. 475), Heinemann & Goven 2006a/b (pp. 467, 469) and Heinemann & El-Kawy 2012 (p. 470). - Box 19.2’s methodological case draws on Aslaksen 2006, Myhr 2002 and Funtowicz 2003 (p. 470). Funtowicz co-authored Benessia 2012 (p. 477). Funtowicz & Strand 2003 appeared in a volume edited by Traavik; Traavik 1999 (p. 468) is a report of the Norwegian Institute of Gene Ecology (GenØk, per the reference list), the home institution of Quist and Myhr; Traavik also co-authored Bøhn 2008 and Myhr & Traavik 2002 (reference list). - The critique of EU “Innovation Union” policy (pp. 461, 463, 476, 477) rests on van den Hove et al. 2011 and 2012. Van den Hove was on the LL2 editorial team (p. 5), and the EEA’s Executive Director, Jacqueline McGlade (foreword p. 8; annex p. 694), co-authored both papers. An EU agency’s report is thus carrying a critique of EU innovation policy co-written by its own leadership. That is relevant to standpoint, though not evidence that the critique is wrong. - [background, verify] Heinemann is widely reported as an IAASTD lead author. The IAASTD is the chapter’s most-cited framing source, and was itself contested: several governments did not fully endorse it, and industry participants withdrew. - [background] Quist co-authored (with Chapela) the 2001 Nature paper reporting transgenes in Oaxacan landraces, a landmark contested-science episode. Nature later said the evidence did not justify publication, and follow-up studies conflicted. The chapter’s point that safety-minded scientists face “personal and professional attacks” (p. 468, citing Delborne 2008 and Waltz 2009) is close to the lead author’s own experience, which goes undisclosed. [background, verify] Delborne 2008 (“Transgenes and Transgressions: Scientific dissent as heterogeneous practice”, per the reference list) is, to my recollection, itself a study of the Chapela–Quist controversy. This is context, not a rebuttal.

How the report’s editors framed it (outside the section). - Part C covers cases where “there is often little science, and very little direct hindsight” (p. 10). - The synthesis summarises the chapter more softly than the chapter does itself, saying that “in addition to some ‘top down’ genetic engineering”, bottom-up approaches “are proving capable…” (p. 644). - The editors note “some opening up of research on GM seeds” after the 26-scientist complaint (p. 646), which the chapter does not mention. - They pair the chapter with Ch3 (leaded petrol’s “roads not taken”) as raising “responsible and socially relevant innovation” questions (p. 660). - On the same page (p. 660) the editors note recent improvements in public access to company data submitted to regulators (EFSA 2013), which “follows recent controversies over the food additive aspartame and GM maize (Seralini et al., 2012; Genewatch, 2012)”. They cite the controversy as a driver of transparency, not as evidence of harm.

Section-by-section notes#

Summary box (p. 458)#

• Two “examples of food and agricultural innovation”, GM crops and agroecological methods, show “how different innovation strategies affect future agricultural and social options”.
• GM crops suit “high-input monoculture agricultural systems that are highly productive but largely unsustainable”. IP rights “often close down, rather than open up further innovation potential”.
• Agroecology is “participatory”, “better suited” to sustainable food security, but needs “a broader range of incentives and supportive frameworks”.
• Introduces the “top-down transfer of technology” versus “bottom-up” dichotomy (Altieri). Closing warning: without governance of innovation “we will design agriculture to fail”.

19.1 Introduction (pp. 459–460)#

• Incentives shape innovation. Incentive systems “set at the highest levels of policymaking largely determine who is innovative and what innovative products will look like” and “favor those who will most benefit” (p. 459). Despite more than enough production, current policies “fail nearly a billion people” (FAO 2010, 2011a). Central question: “Has modern agriculture, despite good intentions, been unwittingly designed to fail?” (p. 459)
• Problems listed: trade, price volatility, energy inefficiency, losses and waste, and “decades” of degradation from high-input farming plus “centuries” of damage from inefficient traditional farming (IAASTD 2009a). Traditional farming is not romanticised (see also p. 471).
• Invention is not enough. Innovation must “meet real needs and be effectively accessed, supported and adopted by farmers” (p. 459). Science helps most where limits are “simple technical issues” rather than institutional or social constraints.
• Agriculture is “multifunctional” (IAASTD): food, fibre, fuel, income, nutrition, identity, skills, biodiversity, greenhouse gases. Yet “food production remains local” (p. 459).
• Three rival framings of food insecurity: too little production; surplus with local access failures (MEA 2005); farmers as “biomass” producers for competing food, feed and fuel markets (Pengue 2005a). Policy is shifting “from ‘how much’ through to ‘how long’ to just ‘how’” (p. 459).
• Energy. Industrial agricultural practices “on average require 10 calories of exogenous energy… for every 1 calorie of food produced” (Giampietro 1993; UNEP 2011) (p. 459). The chapter’s own parenthesis defines the energy broadly (“used for everything from petrochemical production, extraction, transport etc.”). [background] Ratios of this size usually describe the whole food system (processing, transport, retail, household), not farming alone; see the limitations section.
• IP and competitiveness (p. 460). Governments frame innovation “as a means for economic competitiveness by using the promise of returns on intellectual property (IP)” (Heinemann 2009). So “the problems identified for solution will tend to be those that can be packaged and sold — usually to the largest/wealthiest/most lucrative market and largely bypass the poor” (Spielman 2007).
• The Green Revolution was “perhaps the most evident early warning” (the chapter’s hedge). Seed and agrochemical packages raised cash-crop yields “in Asia but not Africa” and proved “incompatible with the cultural and social structures” in many places. It was a “useful stop-gap solution” but not sustainable for local food security, diet diversity or poverty exit (IAASTD 2009b) (p. 460).
• Micronutrients. An IAASTD quote: adequate protein and energy “for over 85 % of people”, but only “two-thirds” get sufficient micronutrients. Diet diversity has declined with staple monocultures (p. 460).
• Neglected innovations are locally adaptable ecological practices that “do not lend themselves to commodification”, e.g. ICIPE’s push-pull system in Kenya (p. 460).
• Case selection is explicit (p. 460). GM is “driven by production goals and short-term profit maximisation incentives”. Agroecology is “better suited” to systems in transition but “requires a broader range of incentives and shelters” (Tilman 2002). Agroecology may better answer “the call from affluent consumers” for quality, sustainability and equity, as well as small-scale livelihoods.
• Footnote 1 (p. 460), a limited concession: “While hypothetically not all GM crops would necessarily require high-input or monoculture farming methods, their development to date has focused on ‘technology traits’” suited to such systems. The word “hypothetically” keeps the concession narrow. The stated target is the innovation system and trait portfolio, not the technique per se (though p. 476 later indicts “‘biotech’ crops” as such). The same paragraph adds that “economies of scale allow the farmer to outweigh the higher costs of production” in high-input systems.

19.2 Innovation: what kinds and for whom? (pp. 461, 463)#

• EU “Innovation Union” (EC 2011) calls innovation “the key” to jobs, a greener society, quality of life and “maintaining our competitiveness on the global market”. The European Council (EU-Council 2011) links growth to ideas “turned into new marketable products and services”. Footnote 2 gives New Zealand’s product/process/marketing definition (p. 461).
• “How innovation is conceived shapes how it is promoted, and who benefits from the promotion.” (p. 461) The EC treats “expensive patenting, market fragmentation, slow standard-setting and skills shortages” as barriers because they “prevent ideas getting quickly to market”. The chapter calls this a “preoccupation with how efficiently technology products flow from knowledge holders to technology users”, Altieri’s “top-down transfer-of-technology approach” (p. 461). This is the proponents’ implied mental model: innovation as product flow, with market friction as the only barrier.
• Kiers et al. 2008: “[i]nnovation is more than invention… innovation demands sophisticated integration with local partners”. Altieri’s bottom-up approach builds on “local people, their knowledge and their autochthonous natural resources” through participation (p. 461).
• Concession on outcomes. “Either pathway could lead to policy decisions to drive efficiencies in food production, lower food costs through increased supply, and become a means out of poverty.” (p. 461)
• The critical innovator. Pathways differ “in who is considered the critical innovator and thus who should primarily benefit from innovation policies” (p. 461). Top-down relies on a specialist producer whose technologies “shape the agroecosystems in which they are to be applied”. The HT-crop-plus-herbicide system demands low agrobiodiversity, high capital (multi-row sprayers) and “a scale investment and specialised farmer”, which is “incompatible” with smallholders (p. 461).
• Research question (pp. 461, 463): does top-down innovation deliver the claimed benefits to small farmers and food-insecure countries? The answer: “the promise of this approach to deliver the expected benefits will continue to be elusive when the pace and scale of innovations are prioritised over considerations for the intertwined institutional, governance and societal issues” (p. 463). Such pathways “may condition innovation directions, diversity and distribution away from” locally adapted innovation (STEPS 2010). “The lure of short-term wealth production from predominantly productivist frameworks” must be “re-balanced” toward long-term sustainability goals, supporting innovations that address “the non-technological, social, institutional, organisational and behavioural aspects” (van den Hove 2012). “There is increasing evidence that the top-down approach to innovation will not achieve the expected stimulus to innovation (Baldwin, 2011)”, and private IP incentives “may actually have a negative effect on the progress in certain fields, including biotechnology” (Murray 2007) (p. 463).

Box 19.1 Herbicide tolerant GM crops: a technology for developing country agriculture? (pp. 462–463)#

• Promise acknowledged. From the 1990s, HT systems gave “near exclusive reliance on a single agrochemical product (‘Roundup’)”, promising “lower labour costs” and no-till compatibility that can reduce erosion (Duke 2008) (p. 462).
• Erosion. “These advantages are, however, disappearing” (Service 2007; Pengue 2005b; Benbrook 2012). Continuous glyphosate use (Powles 2008) drove “a rapid evolution of glyphosate-resistant weeds” (Binimelis 2009; Duke 2008; Heap 2012; Heinemann 2008; NRC 2010) (p. 462).
• Treadmill. “overuse of a single product leads to tolerance and tolerance is overcome with more product”, with more resistant species. Some farmers return “to tilling and using other (and possibly more toxic) herbicides” (p. 462).
• Health and environment. The chapter says “indications of harm” from intensive glyphosate use “has been documented in the scientific literature and remains a concern”, yet cites only Greenpeace 2009 (an NGO report, not peer-reviewed literature) and Séralini 2012 (p. 462). A thin, contested base for the claim as worded.
• Smallholder fit. The package is most economical with aerial or multi-row spraying, “not… possible in a mixed cropping landscape” (Binimelis 2009) (p. 462).
• Deskilling. Mortensen 2012: IWM approaches “are based on knowledge-intensive practices, not on saleable products, and lack a powerful market mechanism to push them along”. The system is “‘deskilling’” (Binimelis 2009). The package also encourages expansion into marginal land and deforestation (Morello 2007; the sentence is garbled in the original) (p. 462).
• Alternative: integrated weed management (IWM). It combines rotation, cover crops, competitive cultivars, “judicious” tillage and “targeted herbicide application” (Mortensen 2012). The chapter claims IWM is affordable, gives “the same high yields and profits” (Liebman 2008 and others), improves soil “even above no-till approaches based on herbicides” (Venterea 2006), and avoids resistance “of the magnitude” of simplified control (Davis 2007) (p. 462). IWM still uses herbicide: this is integration, not prohibition.
• Forecast on stacking (Mortensen 2012): extra HT genes “will encourage continued neglect of public research and extension in integrated weed management” (p. 462).
• Verdict. HT transfer to poor farmers, which the chapter says “has demonstrated not to be a sustainable approach for addressing the needs of developed country agriculture”, “appears to be another example of a top-down approach that has not, and will not produce the beneficial outcomes for the poor farmer” (Heinemann 2008b, i.e. Heinemann & Kurenbach, a Third World Network paper). The verb “appears” is a hedge, but the evidential support is a single self-citation. Viable bottom-up approaches exist: “all that is lacking is the political will and institutional capacity” (p. 462).
• Income (p. 463). Gains are not “uniform or sustainable” (Botta 2011, Argentina). The highest-yielding GM varieties owe this to “traditional breeding, rather than… genetically engineered traits” (Gurian-Sherman 2009). Even mature US-south cotton adopters had no “net economic benefit” compared with “other high yield varieties” (Jost 2008, which the reference list shows is a single study of Georgia cotton systems). Patent seed rent is an upfront cost “too high” for poor farmers (Delmer 2005).

19.3 GM crops as a top-down path out of poverty and hunger (pp. 463–465)#

• Commercial GM is restricted “so far and for the foreseeable future” to crop plants: cotton, maize, canola and soy. Footnote 3 adds papaya, sugar beet and possibly alfalfa in the US, and tomato and pepper at low levels in China (p. 463).
• “Despite more than 30 years of research and development and nearly 20 years of commercialisation… only two traits have been significant in the marketplace — herbicide tolerance and insecticide production.” (p. 463)
• Concentration (industry data, James 2011 / ISAAA):
• Five countries (USA, Brazil, Argentina, India, Canada) account for 91%, the next five for 8%, and “1 %” is spread among “just seven other countries”. Read literally, this implies only 17 countries grow GM crops, whereas Figure 19.1’s own title calls them “the 17 largest producing countries” (see the limitations section).
• The “17 so-called GM ‘mega-countries’” had 159 million ha in 2011, “seemingly a large figure, but in reality is just 3 % of the world’s agricultural land”.
• Near-total conversion of soy in Argentina and the US, US sugar beet and Indian cotton (p. 463).
• Figure 19.1 (p. 464).
• Left: a pie of GM hectares, dominated by the USA (69 million ha), then Brazil, Argentina, India, Canada and China; the smallest shown is Spain, at “an industry estimated 100 000 hectares”. The caption also says countries range down “to < 50 000 hectares”, which is inconsistent with Spain being the lowest shown.
• Right: countries grouped by GM share of agricultural land:
  • 12–17%: USA, Argentina, Paraguay, Canada, Brazil.
  • 5–9%: Pakistan, Uruguay, India, Philippines.
  • 1–3%: Myanmar, Burkina Faso, South Africa, Australia, Bolivia, China.
  • <1%: Mexico, Spain.
• Point made: rankings by proportion differ from rankings by hectares.
• Figure 19.2 (p. 465). GM share of agricultural land: USA 17%, Argentina 17%, Brazil 12%, Canada 14%, India 6%; the global share is 3%. [background] The denominator (FAOSTAT “agricultural land”) includes permanent meadows and pasture; see the limitations section.
• Why adoption is “patchy” (pp. 464–465): 1. Consumer rejection by high-income markets (Gaskell 2010), since the traits “provide no direct benefit to consumers and may be introducing unintended adverse effects”. The case for adoption rests on indirect financial and management benefits to “certain kinds of farmers”. 2. Subsidies undermining developing-country markets. Adopters are “usually large-scale commodity growers” in monocultures or two-crop rotations. OECD agroecosystems “rely on heavy taxpayer subsidies to remain viable” (Kiers 2008), and GM crops are “most commonly crops that benefit from subsidies” (Pechlaner 2010). The second reason is that these subsidies’ “use in developed countries undermines the market for these crops in developing countries”. Via Hoffman 2011, quoting OECD 2006: support was 30% of farm receipts in 2003–05, “almost USD 1 billion per day”, costing developing countries “about USD 17 billion per year”, about five times aid to agriculture. The GM-specific causal link is asserted; subsidies cover non-GM commodities too. 3. Cost. “The incentive brought by subsidies” gives the third reason: the “high rent of GM seeds and associated management inputs” confines them to countries that “redistribute wealth to farming for export, whether rich or poor” (Delmer 2005).
• Export treadmill. Poor countries following this lead risk “a loss leading treadmill where they produce agricultural goods at a net social loss” (Heinemann in press; Pengue 2005b). GM has not spread where “upfront costs are too high” (p. 465).

19.3.1 Top-down incentives homogenise tool building (pp. 465–466)#

• The feed-the-world debate is framed as genes versus environment (Marris 2008; Vanloqueren 2009). The chapter concedes the dichotomy “is in essence artificial”, but finds “an underlying truth”: the emphasis on seed-based tools “is an unavoidable outcome of how innovation in the top-down model works”, because capturing genotypes as IP is “far easier” than commodifying cover crops, rotations, composting, training or credit (pp. 465–466). “When a singular, centralised and highly specialised approach to agricultural development is followed, such as through genetic engineering, it can stifle other approaches that might produce even more desirable outcomes.” (p. 466)
• Market-size asymmetry (p. 466). Where inputs can homogenise the agroecosystem, “a small number of varieties based on a proprietary genotype can be sold to a large number of farmers”. Management techniques are “knowledge- rather than product-intensive”, site-customised, and “require more investment relative to the size of the market”. Their benefits are “better distributed because they are not concentrated back to a seed producer”. Hence “these asymmetries in investment incentives mean that management-based approaches do not receive the same levels of support and investment as do approaches that are easily recaptured in the marketplace.” This is the chapter’s core mechanism.
• Selling homogenisation. “The success of the green revolution was its ability to convert very different lands into similar agroecosystems using external fertilisers and other inputs”, at great long-term environmental, fossil-fuel and emissions cost. Therefore it “will not be the model for future agriculture” (p. 466).
• The Nature editor (Editor 2010) is quoted in support of both new varieties using fewer inputs (e.g. via root research, stress resistance) and “lower-tech research” into rotation, mixed farming, soils and waste. One-quarter to one-third of food is lost (p. 466).
• Promise gap. Top-down approaches have failed “to deliver on promises of a wide range of trait innovations needed by farmers”, such as salt and water-stress tolerance (p. 466).

19.3.2 Effects on the knowledge pipeline (pp. 466–467, 470–471)#

• Funding. Public agricultural R&D was “just under twice” private R&D at the start of the century (IAASTD 2009b): about USD 12 billion a year public in developing countries, about USD 10 billion in high-income countries. The CGIAR budget is “only 12 %” of the combined R&D of the six largest breeding and GE companies (Spielman 2007) (p. 466).
• Three deeper shifts (pp. 466–467): 1. The public-to-private shift in R&D responsibility “is unequivocal in high-income countries”. 2. Public R&D now carries “‘industry-driven’ priorities”, leveraging public money toward “often privately held” innovation and compromising pro-poor research (Spielman 2007). 3. Public funding has “direct or indirect ties to industry” (Knight 2003; Lotter 2009b; Seabrook 2011).
• IP instruments. Patents and patent-like plant variety protection (PVP) were “newly applied to agriculture only in the last decades of the 20th century”. Patents give more control because PVP carries a research exemption. The pro-IP case, that IP creates “net social benefits” (Pray 2007), is stated fairly, then criticised for ignoring “effects on the innovation pipeline” (p. 467).
• International institutions. The chapter says leading bodies “dismissed prevailing IP instruments as agents of constructive economic or food security change in developing countries at least at their stage of development” (WHO 2005; World Bank 2007); the final qualifier matters. Seed saving and exchange “have become incompatible with these more severe IP instruments as shown in the conversion of behaviour in the US”, and would, if adopted by developing countries, undermine “farmer by farmer breeding” (p. 467). [my assessment] “Dismissed” may still overstate those sources.
• Concentration. Seed control is concentrating in “a very small number of multinational corporations”. The four-firm concentration ratio (CR4) “breached a critical threshold” (World Bank 2007). UK POST 2011 is quoted: seed companies “may have reduced incentives to develop conventional varieties”; in US soy “conventional breeding now mainly left to universities and to small seed producers”; patents “may also limit public-sector research” (p. 467). This is an authoritative, non-advocacy source, though the chapter introduces it as “the UK Parliament now says”; [background] a POST note is a briefing by Parliament’s science office, not a statement of parliamentary position.
• Public IP is no fix. Public institutions’ “unprecedented accumulation of IP claims in biotechnology” (Graff 2003) behaves “consistent with top-down innovation models”. This “creates a feedback loop in which the best-funded researchers are those with top-down innovation interests, and they in turn out-compete other researchers… from future funding”. The loop stifles the “public-good knowledge commons” (p. 467). It is inferred, not documented.
• Anti-commons. Murray & Stern 2007 is quoted across pp. 467 and 470: IP is associated with “a statistically significant but modest decline in knowledge accumulation”, “at least for research of the type published in Nature Biotechnology”. The chapter keeps the source’s qualifiers.
• Interim verdict (p. 470). If this framework is followed, “the outcome is likely contrary to the stated objectives”. Top-down “will fail in the long run to produce food security because it does not have the necessary incentives to create resilient and sustainable production systems”, and is “the wrong pathways” for a broader remit. The hedge follows: “we do not argue that top-down innovation is irrelevant at all times and in all countries. Indeed, the right mix of innovation is essential. Likewise, seed-based versus environmental approaches both have value in all agroecosystems at all times.” The problem arises when providers “become out of balance in scale, power or access to information, then one can smother the other”. “When public institutions must act in a way that is consistent with how companies must act, then the imbalance between farmer and knowledge access grows.” (pp. 470–471)

Box 19.2 GM crops: a late lesson case in the making? (pp. 468–470)#

• Framing (p. 468). Benefits and harms “are still being verified” despite “science-based calls for greater scrutiny… from early on in the US FDA” (Drucker 2012) “and elsewhere (Traavik, 1999)”. Drucker’s book is openly polemical (its subtitle, per the reference list: “How the Venture to Genetically Engineer Our Food Has Subverted Science, Corrupted Government, and Systematically Deceived the Public”); [background, verify] The author (usually spelled Steven M. Druker) founded the Alliance for Bio-Integrity and led its 1998 lawsuit against the FDA, so the source is a litigant’s account; the book cited as 2012 “(Part One)” was published in full in 2015. Footnote 5: the FDA memos (1991, 1992a, 1992b) are “3 of 24 internal FDA documents” obtained by FOIA by the Alliance for Bio-Integrity. “The literature is accumulating indicators both of inflated benefit claims and of evidence of adverse effects” (Bøhn 2008; Botta 2011; Hilbeck 2012; Jost 2008; Mesnage 2012; Rosi-Marshall 2007; Service 2007). Possibly overstated benefits: lower pesticide use (Service 2007), “reduced use of more toxic pesticides” (Mesnage 2012; Séralini 2009), yields (Gurian-Sherman 2009), income (Jost 2008). Mesnage 2012, per the reference list, is an in vitro study of cytotoxicity to human cells of Bt toxins and a glyphosate-based herbicide: topically relevant to the less-toxic-pesticides claim, but weak evidence for real-world toxicity comparisons.
• Organising questions (p. 468). “At what point is there sufficient evidence to be concerned and take action about the effects of GM crops on human health and the environment? How strong is the evidence of safety vs. risks?”
• Global exposure. As commodities, GM crops occur at low levels in many products: “exposure is global even if production is mainly in a few countries” (p. 468).
• Absence of evidence. Many risk assessments equate “‘no evidence of harm’” with safety: “the safety of GM crops is presumed when there is a lack of evidence of harm, as if this were equivalent to evidence of lack of harm, when it clearly is not”. Such conclusions are “assumptions-based, rather than evidence-based” (Spök 2004). “Critically, when this lower standard of safety assurance is followed, as is the case with the mainstream risk assessment approaches today, important effects may be missed.” Hedge: “it is plausible that there simply are no effects to be found”. The real question is whether current methods could detect effects (p. 468).
• Obstacles to biosafety research (p. 468):
• Material access. Industry contracts restrict it. “26 scientists” said agreements made independent research “virtually impossible” (Pollack 2009).
• Funding. The USDA spent “USD 1.8 billion” on biotech research in 1992–2002, only “approximately 1 % (USD 18 million)” on risk (Mellon 2003).
• Career risk. Scientists publishing unfavourable findings faced “personal and professional attacks” (Delborne 2008; a 1999 Lancet editorial; Waltz 2009a, 2009b), “and in some cases” “threats or loss of research funding and dismissal” (Lotter 2009a, 2009b). No cases are named in the text.
• Barriers in risk assessment (p. 469). Context given first: release is “preceded in many countries by a pre-market risk assessment”, and “most countries use international guidance (e.g. OECD/Codex Alimentarius, Cartagena Protocol on Biosafety)”. Then: “Policies can undermine the effectiveness of risk assessment (Pavone, 2011) by allowing risk standards which increase the likelihood that adverse effects, if occurring, would not be identified”: 1. Developer-generated data, with regulators lacking independent testing capacity, bring “‘the funding effect’” (Krimsky 2004; Diels 2011). 2. Confidentiality prevents independent review (AHTEG 2012). 3. Low sample sizes “bias the outcome towards no observation of differences” (Marvier 2002), and hazards the regulator did not ask about go untested. 4. The regulator’s choice of what to test “may miss unintended changes to other gene products or metabolites or the effects of cooking and processing”. “Surrogate” transgenic proteins, usually made in bacteria, leave “the actual protein produced by the GM plant untested” (Freese 2004). 5. “no regulatory framework requires mandatory toxicity or allergenicity testing” of consumption (or inhalation, Kroghsbo 2008). “Commonly, only 90-day (usually rat) feeding trials are conducted”, with long-term conclusions drawn from them (Séralini 2009; Spiroux de Vendômois 2010), and “indications of adverse health impacts only manifested after 120 days (Séralini, 2012)”. [background] This is the chapter’s most contested citation. 6. “Reference lines” broaden background variation so that signals are “drowned in statistical noise” and judged “‘within the range of biological variation’” (Antoniou 2012; Dolezel 2009).
• Monitoring (p. 470). Post-release monitoring is generally lacking. Where it is mandated there is “very little information on its effectiveness and no uniformity” (Züghart 2008, 2011; AHTEG 2012).
• Precaution inside the science (p. 470). Precaution is legitimised in GMO law (EU, Cartagena) but treated “as a risk management tool and not part of the scientific risk assessment”. The chapter acknowledges that critics “consider it easily misused as a barrier to trade and the cause of more regulation”, and replies that “this misrepresents how precaution may be appropriately applied”. Two roles in the science: 1. “the need for precaution and the need for scientific rigor are not incompatible but complementary” (Groth 2000). 2. Value judgements within science (Funtowicz 2003; Rudner 1953) about “levels of evidence, directions of error” (Brosi 2009; Lemons 1997), and about “what we know, do not know, and cannot know” (Aslaksen 2006; Myhr 2002). Formally acknowledging uncertainties and the choice of error type, and communicating them to decision-makers, are called “key components of rigorous science-based risk assessment”.
• Conclusion (p. 470): “The critical late lesson that may be emerging from GM crops is not the evidence of harm — the early indications of harm are just emerging — but the persistence of the same institutional patterns that led to the old late lessons already learned from asbestos, benzene and BSE (Harremoes, 2001)” (i.e. LL1), where “weak risk assessment standards were implemented that prevented identifying the harm and taking precautionary action”. Remedy: “the appropriate application of the precautionary approach to risk standards”. The primary warning is institutional, not harm-based, but the authors also assert, in passing, that “early indications of harm are just emerging”.

19.4 The bottom-up path towards sustainable farming (p. 471)#

• Bottom-up approaches bring users into innovation so that technologies fit “the ecological, socio-cultural and technical dimensions” (STEPS 2010; Wagner 2007). They build sustainable communities and local markets (Altieri 2011a; UNEP-UNCTAD 2008).
• Agroecology “fashion[s] our technological solutions to fit nature (Schumacher, 1973)” rather than “engineering nature”. It minimises agrochemical and energy inputs by using “ecological synergisms” (Altieri 1995; UNEP 2011 “green agriculture”). It “also values… conventional breeding and genotype optimisation”, but tends to use management changes “that remove, rather than adapt to, the problem” (Lal 2009).
• Resilience. Biodiversity and soil organic matter buffer climate extremes (Hajjar 2008). After Hurricane Mitch, agroecological farms “incurred less damage than neighbouring conventional monoculture farms” (Altieri 2011a). [background] This is from Holt-Giménez’s survey; Mitch was 1998.
• Selection. Agroecology, “including compatible organic certification schemes”, was chosen because it “is delivering excellent results” (p. 471). Organic is folded into agroecology. “The real innovation potential does not stop with the farmer, but often starts there.”
• Not traditional farming. Traditional farming “can… be as destructive… as any high external input industrial ‘modern’ farm” (IAASTD). Agroecology is “far more sophisticated, knowledge intensive, and integrative” than “kitsets of seeds, fertilisers and agrochemicals”, and so needs extension services and broad participation (p. 471).

19.4.1 Bottom-up incentives homogenise productivity and resilience rather than tools (pp. 471–472)#

• Instead of converting soils into “near replicas” of those that elite varieties need, agroecology supports soil conservation and “farmer seed exchanges” for local varieties (Badstue 2007; Jarvis 2008). Agrobiodiversity creates resilience and reduces pest attraction. Emphasis “on the farmer rather than the breeder” (p. 472).
• Yield-gap argument. Quoting Molden 2007, attributed to “the UN FAO”: “75 % of the additional food” could come from raising low-yield farmers to “80 % of what high-yield farmers get”. Hence development “at landscape rather than global or even national levels”. The chapter endorses Pretty 2011 on “the best possible seeds and breeds and their management in local ecological contexts” (p. 472), a genotype role conceded again. [background] Molden 2007 is the IWMI-led Comprehensive Assessment of Water Management in Agriculture, so the FAO attribution is loose.

19.4.2 Bottom-up innovations are participatory (pp. 472–473)#

• Participatory and open-collaborative models (Baldwin 2011; Ceccarelli 2006; Toomey 1999; Witcombe 1996). “The choice of relevant instrument can encourage ongoing innovation” even under IP (p. 472).
• IP choice. “Patent-like PVP and patents” restrict farmer use of protected germplasm, and poor farmers cannot reach distant rights-holders (Howard 2009). “PVPs which recognise breeder’s rights” allow farmers to keep developing and exchanging varieties, so “a wave of innovation” extends from an initial variety (Gyawali 2007; Steinberg 2001). Centralised licensing “can bottleneck technology transfer”, which is associated with a GM “‘yield gap’” (Fernandez-Cornejo 2006) (p. 472). The distinction between the two kinds of PVP is not explained, and the terminology is inconsistent: p. 467 calls PVP generally “patent-like” (while noting its research exemption gives less control than patents), whereas pp. 472, 475 and 476 separate “patent-like PVP” from PVP that recognises breeder’s rights. [my interpretation] It probably means stronger UPOV 1991-style PVP versus regimes with breeders’ exemption and farmers’ privilege.
• Figure 19.3 “Participatory IP instruments” (p. 473). A conceptual diagram, not data. A novel-genotype seed is sold, local gene × environment interactions produce varied phenotypes (“in situ trait”), and desirable ones are captured as new seed by farmers or breeders, repeatedly. Caption: elite-variety benefits “are more quickly adapted to local conditions… and may inspire new breeder income”.
• Social infrastructure. Seed-saving knowledge and exchange infrastructure “are also social resources that if (or when) lost may be difficult to re-establish” (Howard 2009). Decentralised public breeding and in situ conservation may be “fundamental to the survival of billions” under climate change (p. 472). Farmers as ecosystem managers need “freedom to innovate” and government “confidence” in their innovation (Hoffman 2011).
• Expert blind spot (Toomey 1999): “Professional breeders, often working in relative isolation from farmers, have sometimes been unaware of the multitude of preferences — beyond yield, and resistance to diseases and pests — of their target farmers” (harvest, storage, taste, maturity, fodder) (p. 472). Farmer field schools cut insecticide use (Indonesia, Bangladesh, Vietnam) and raised yields (China, India, Pakistan) (Van den Berg 2007) (pp. 472–473).

19.4.3 Bottom-up approaches deliver the right kind of innovation to the right kind of users (pp. 473–474)#

• “Agroecological bottom-up innovations are relevant and work.” (p. 473) Environmental co-benefits (Pimentel 2005); restoring soil fertility (de Jager 2005).
• Badgley 2007, “The world’s largest meta analysis comparing science-lead industrial and agroecological (organic) farming systems”. Per the chapter, organic “could match” industrial yields, with “the same or superior yields” and fewer inputs; the study “provided evidence to suggest” (the chapter’s hedge) that mature conversions (more than five years) “consistently out produced” industrial operations; and “it takes about five years of intensive work to rehabilitate soils”, which past studies missed by lumping young and mature conversions together (pp. 473–474). This characterisation needs checking; see the limitations section.
• UNEP-UNCTAD 2008: “average crop yield increase of 116%” for “organic and near-organic projects” across 114 cases, more than 1.9 million African farmers and about 2 million ha. Report conclusion: organic “can increase agricultural productivity and can raise incomes with low-cost, locally available and appropriate technologies”. It can also be more profitable (Edwards 2008; Nemes 2009) (p. 474).
• Pretty 2008: 286 initiatives, 37 million ha, 57 countries, 12.6 million farms, “average crop increase of 79 %”. UK Foresight 2011 / Pretty 2011: 40 African projects in 20 countries; by 2010, 12.75 million ha and 10.39 million farmers, “a doubling of crop yields, on average (2.13-fold increase)” over 3–10 years. The chapter comments: “The results speak for themselves” (p. 474).
• Global North. The Wisconsin 12-year trial found diverse low-input systems “can be as productive per unit of land” (Posner 2009; the paper cited in the reference list is Part II, “Economic and Risk Analysis 1993–2006”, not a yield paper as such). Rodale’s 30-year trial reports better yields, economics, energy efficiency and “human health indexes” (Rodale 2011) (p. 474).
• Scaling. UNEP 2011 modelling of 40-year “green” investment against business as usual shows higher yield, soil quality, water efficiency, GDP and employment, and lower CO2. “Therefore the potential scale-up… appear to be immense.” (p. 474) A scenario model is being treated as evidence.

19.5 Case example: water stress (pp. 474–475)#

• The likely top-down product is drought-tolerant genotypes. Progress is coming “through classical breeding… especially augmented through marker-assisted selection” (CIMMYT 2012). Footnote 6 notes MAS involves no in vitro modified nucleic acids (pp. 474–475).
• Prediction (p. 475). GE “has not been as successful” because “drought tolerance depends on the action of multiple genes”, needing changes “all at once, rather than adding genes singularly”. Adapted varieties need “more responsive breeding” than the lengthy GE commercialisation process allows (Gurian-Sherman 2009; Heinemann 2008a, in press). The chapter also says “plants with ever more extreme adaptation of genotypes will likely continue to exacerbate the depletion of the water table”, which is unsourced. Genotype innovation gets emphasis “especially from industry” because seed products are IP-protectable.
• Management alternatives. Bottom-up management solutions, “some of which have been in practice for decades”, raise soil water retention “as well as improve the genetics of crop plants” (Heinemann 2008a; Lotter 2003; Pimentel 2005; Scialabba 2007). The listed package, which starts with “locally-adapted drought tolerant varieties” and continues with cover crops, polycropping, fallows, compost, agroforestry and small dams, is said to “all raise water levels” (Altieri 2002; Lal 2006). So the bottom-up option includes a genotype component. Feeding the world in 2050 may be impossible “unless soil quality and water retention capacity are raised” (Hoffman 2011). Lal 2008: “If soils are not restored, crops will fail even if rains do not” (p. 475). Such innovations “could not be easily described or protected by patents”, so they “are not innovations in the currently practiced top-down model”.
• Problem shifting. The private sector “will offer solutions to a problem that either possibly cannot be solved using technologies that are described under prevailing IP instruments or which will only shift the problem in time or space, addicting us to finding and producing even more extreme genotypes” (p. 475).

19.6 Conclusions (pp. 475–477)#

• Lock-in. Top-down innovation’s “largely productivist objectives… tend to shut down rather than open up innovation and options”, and science as a public good becomes “conditioned within certain notions of progress” (Callon 1994, printed as “Callo”). “This framework will only continue to create technological lock-ins and path dependence to specific research choices at the expense of others (Stirling, 2007).” (p. 475)
• Right mix. “both top-down and bottom-up approaches will have their roles to play, but getting them in the right mix, order and framing is critical to ensure their benefits and risks are more evenly distributed”. This means “rebalancing innovation towards the public good” and attending to “interlinked institutional, organisational and social changes” (p. 475).
• Recommendations (p. 475). Operationalise IAASTD (2009a) and SCAR (2012). SCAR: “Approaches that promise building blocks towards low-input high-output systems, integrate historical knowledge and agro ecological principles… should receive the highest priority for funding”. “A public sector free from political incentives for top-down innovation is an essential capacity”. Small businesses and farmers, “benefiting from proprietary knowledge”, remain essential. The chapter is not anti-proprietary per se.
• Scientists’ responsibility (p. 476). Scientists are “a powerful force from within”. “The modern techno-science culture took shape just after WWII”, with US nuclear power and Soviet space travel as “exemplars”. “The convergence of internal culture, economic and political power was and is an irresistible force. Scientists today cannot shirk from their role and their responsibility on how science is done and governed”.
• Convergence and distribution (p. 476). Top-down innovation is “most effective” when specialised science converges with bottom-up local optimisation. “Long term, there will be enough food if agriculture both intensifies and remains local.” Comparative claim: if the bottom-up approach is followed, “the transfer of knowledge and further innovation potential is augmented, and success far more likely” than under top-down, “where the innovation potential downstream is severely limited”. Bottom-up proceeds “disproportionately flow to adopters rather than the providers”, so bottom-up providers “are easily displaced by wealthier and more powerful champions of top-down policies”. With an expanded remit, “the bottom-up innovation models are the responsible innovation models” (De Schutter 2011).
• Strongest indictment (p. 476). As “black-box technology” protected by “particularly restrictive IP instruments (patents and patent-like PVPs)”, “so-called ‘biotech’ crops (GM and similar) and their co-technologies are expensive to buy, destroy local seed savings and exchange practices, and prevent further farmer tinkering”. “When working as advertised”, top-down innovation “(which ironically is maintained by extra-market subsidies), undermines the stated national and international goals of poverty reduction, sustainability, and increases food insecurity”. It contributes to a feedback loop that concentrates “wealth and power into a smaller number of companies and large farms” (Botta 2011; Spielman 2007; USDA 2009; World Bank 2007). “Top-down providers are invariably attracted to the largest markets (real or subsidised), the most uniform agroecosystems, and the highest volume farmer. They therefore will always serve last” smallholders on “< 2 hectare plots”. No data in the chapter support “increases food insecurity”.
• Cultural homogenisation (p. 476). Products needing “ecosystem, cultural and financial homogenisation” erode local knowledge and traditional foods. Policymakers must remove obstacles to bottom-up approaches (De Schutter 2009, 2011).
• Framing (p. 477). “For as long as the problems needing products are framed as technological rather than social, behavioural or political, then innovation will be directed toward technological products” (van den Hove 2012). Fixes treat “symptoms”. Lederberg 1970: “Our imperfect solutions aggravate every problem.” A commonly attributed Einstein line is also quoted.
• STEPS “3D Agenda” (p. 477):
• Directions: top-down paths (“highly specialised, centralised and capital intensive”) “tend to shape innovation towards technological lock-ins”. “Economic and political forces that promote these innovation trajectories then become hard to reverse, or crowd out alternative approaches, such as agroecology.”
• Distribution: “legal instruments of knowledge control” define access “based on who can afford to pay for it” and “marginalise those for whom innovations are more critically needed”.
• Diversity: “Diversity in innovation buffers against lock-ins” and supports adaptation to “future uncertainties”.

19.7 Lessons learned (pp. 477–478)#

• Eisenhower (1961) is quoted on both dangers: domination of scholars “by Federal employment, project allocations, and the power of money”, and “public policy could itself become the captive of a scientific technological elite” (p. 477).
• Core lesson (p. 477). Framed as “The early warning, or perhaps late lesson”: following top-down, “usually technologically oriented” approaches, “the desired outcomes for addressing food insecurity will not be achieved”. They “will most likely fail to deliver on the large promises of food security and alleviation of poverty, mainly because these approaches contribute to a feedback cycle that concentrates resources, knowledge, and influence as witnessed in the seed and agrichemicals sector”. “Through this power, top-down providers can artificially homogenise both the conception of the problem to be solved and the solutions”. Questioning “the rationality of the approach gets lost” in discussion “over the use of the approach” (Pavone 2011; italics in original). The chapter suggests, with a hedge (“Perhaps… is needed”), “greater reflection and social deliberation into why and for whom agricultural innovations should be produced”.
• Close (pp. 477–478). Bottom-up approaches “are proving capable” but “are incapable of flourishing where invention is limited to what can be easily described by prevailing IP instruments. Change the directions, distribution and diversity of innovation, and you change the world.”

References (pp. 478–485): the evidence base#

About 178 entries by my count, including two duplicates and one merged entry (my classification): - Intergovernmental and official: IAASTD (×4), FAO, UNEP, UNEP-UNCTAD, De Schutter (UN Special Rapporteur), WHO 2005, World Bank 2007, OECD, UNCTAD/Hoffman, MEA, AHTEG, EU-SCAR, UK POST, UK Foresight, USDA ERS, NRC 2010. - Mainstream peer-reviewed: e.g. Tilman 2002, Powles 2008, Marvier 2002 and 2007, Murray & Stern 2007, Graff 2003, Spielman 2007, Liebman 2008, Mortensen 2012, Diels 2011, Van den Berg 2007, Badgley 2007, Pretty 2008 and 2011. - Advocacy, NGO and journalistic: Greenpeace 2009; Earth Open Source (Antoniou 2012); Drucker 2012; Third World Network (Heinemann & Kurenbach 2008b, Heinemann 2009; Edwards 2008); Rodale; Vanity Fair (Barlett & Steele 2008, “Monsanto’s harvest of fear”); New Yorker; NYT. [background] Gurian-Sherman 2009 and Mellon & Rissler 2003 are Union of Concerned Scientists authors (the reference list does not say so). - Séralini group (2009, 2010, 2012; Mesnage 2012) for the health “indications”. - Sources sympathetic to the technology appear only for hectares (ISAAA, which the chapter itself labels an “industry source”) and to state the “general argument” for IP (Pray & Naseem 2007, an academic development-studies paper, not an industry source). Duke & Powles 2008 (“Glyphosate: a once-in-a-century herbicide”) is cited for HT advantages and for resistance. - Duplicate or erroneous entries (Heinemann & Goven 2006a = 2006b; Shorett 2003a = 2003b; Heinemann & Kurenbach 2008 = 2008b; NRC 2010 and OECD 2006 run together in one entry; “Callo” for Callon & Bowker 1994) suggest light copy-editing.

Case timeline (as far as the chapter supports one)#

The chapter gives no chronological case history and identifies no “effective action”. Its argument is that action has not come. The dated items it uses are below; bracketed items are background.

Implied lags. - Biosafety: from the early FDA concerns (1991–92) to 2012 with harms “still being verified” (p. 468), over 20 years. - Herbicide resistance: [my inference; the chapter gives no dates] about 10–15 years from HT adoption (“Starting in the 1990s”) to a recognised treadmill, with the anticipated response being stacked tolerance traits (p. 462). - Innovation policy: the Green Revolution warning is presented as unheeded after 40-plus years (pp. 460, 463).

The chapter discusses no actual EU GMO authorisation, coexistence or labelling practice, a notable gap given the publisher.

The authors’ own lessons and conclusions#

Lessons derived from their evidence#

1. Incentives shape content. Policy incentive systems “largely determine who is innovative and what innovative products will look like” (p. 459). IP incentives select problems “that can be packaged and sold” (p. 460), producing an “unavoidable” seed-based bias (p. 465).
2. Investment asymmetry. It starves management-based, knowledge-intensive innovation (p. 466; examples pp. 462, 475).
3. GM fits large, subsidised, homogenised systems. Adoption is concentrated and limited by consumer rejection, subsidies and cost (pp. 463–465; Figs 19.1–19.2).
4. The HT treadmill. HT systems generate a resistance treadmill and deskilling; stacked tolerance deepens neglect of IWM (p. 462).
5. Overstated benefits. GM benefits are overstated: yields come from conventional breeding, and farmer income gains are uneven (pp. 463, 468).
6. The knowledge pipeline. Privatisation and IP accumulation, including by public bodies, reshape it: a funding feedback loop, a modest anti-commons effect, reduced conventional breeding (pp. 466–467, 470).
7. Risk assessment. It is structurally weighted against detecting harm and treats “no evidence of harm” as safety (pp. 468–470).
8. The emerging late lesson is institutional (p. 470).
9. Bottom-up innovation works. Agroecological and participatory innovation delivers yield, income and environmental gains, and surfaces user needs experts miss (pp. 472–474).
10. Lock-in and diversity. Top-down pathways create lock-in and concentrate power; diversity buffers against lock-in (pp. 475, 477).

Predictions stated with certainty or near-certainty#

(Hedging words in the source are noted, since they are easy to drop.) - Top-down “will fail in the long run to produce food security” (p. 470; unhedged, though the preceding sentence says the outcome is “likely contrary” to stated objectives). - Top-down “will most likely fail to deliver on the large promises of food security and alleviation of poverty” (p. 477; “most likely”). - Providers “will always serve last” smallholders (p. 476; unhedged). - HT transfer “appears to be another example of a top-down approach that has not, and will not produce the beneficial outcomes for the poor farmer” (p. 462; “appears”). - Stacked HT traits “are likely to” further undermine sustainable agriculture by encouraging neglect of IWM (p. 462, quoting Mortensen 2012). - GE drought tolerance has “not been as successful” as breeding, and top-down solutions “will only shift the problem in time or space” (p. 475). - “current indications are that we will surely design agriculture to fail” (p. 460).

Recommendations and advocacy#

• Rebalance innovation “towards the public good”, including institutional, organisational and social innovation (p. 475).
• Operationalise IAASTD and SCAR 2012, giving low-input, high-output agroecological research “the highest priority for funding” (p. 475).
• Build “a public sector free from political incentives for top-down innovation” (p. 475).
• Prefer IP instruments allowing downstream farmer and breeder innovation (pp. 472–473).
• Put precaution inside risk assessment (error types, evidence levels, stated ignorance) and strengthen monitoring (p. 470).
• Scientists should take responsibility for how science is governed (p. 476).
• Adopt the STEPS 3D agenda (p. 477). Hold “social deliberation into why and for whom” (p. 477).
• Treat bottom-up models as “the responsible innovation models” (p. 476). Policymakers should remove obstacles to them (p. 476).

Mechanisms and dynamics#

1. Appropriability steers direction. The master mechanism: whether a developer can capture returns determines how much investment an innovation attracts. - Genotypes are capturable through patents or PVP. Rotations, cover crops, composting, training, push-pull, IWM, and soil and water management are not (pp. 460, 465–466, 475). - Management innovations also face fragmented, site-specific markets, and their benefits are “not concentrated back to a seed producer” (p. 466). They are under-invested even when socially superior. - The same logic explains which GM traits exist: those that sell to the largest, most uniform, often subsidised markets (pp. 463, 476). Stress tolerance, promised for years, went undelivered (p. 466). - This is standard public-goods economics, applied convincingly but not quantified.

2. Homogenisation as the business model. Top-down tools need uniform environments. Inputs “homogenise the environment to support proprietary genotypes”, which was the Green Revolution’s “success” (p. 466). The HT package demands scale, mechanisation and low agrobiodiversity (pp. 461–462). Technology and farming system co-produce each other. Bottom-up homogenises “productivity and resilience rather than tools” (p. 471). The chapter extends homogenisation to culture and local knowledge (p. 476).

3. Framing as power. - The EU defines innovation as marketable products for competitiveness and barriers as market frictions (p. 461). - Providers “can artificially homogenise both the conception of the problem to be solved and the solutions” (p. 477). Debate then shifts from an approach’s rationality to its use (p. 477). - Technological framing of social, behavioural or political problems yields symptom fixes (p. 477). - The chapter flags the genes-versus-environment framing as artificial (p. 465), yet keeps using it. - It names competing framings of food insecurity (production, access, biomass competition) (p. 459) and sides with “how we produce and consume” (p. 459).

4. Who counts as the innovator. A distributive choice built into definitions: specialist producer or farmer (p. 461). Bottom-up treats farmers as experts, with innovation starting with them (pp. 471–472).

5. Evidence production and the burden of proof (Box 19.2). Safety knowledge is structurally tilted toward null results: - developer-generated and funded data (the funding effect); - confidentiality blocking replication; - underpowered designs; - surrogate test proteins; - short feeding trials; - variance-inflating comparators; - weak monitoring; - restricted material access, scarce risk funding (1%) and career risk for dissenters (pp. 468–470).

The epistemic core is treating “no evidence of harm” as “evidence of lack of harm” (p. 468): an implicit burden of proof on those alleging harm and a low standard of proof for safety. The remedy is explicit error-type and uncertainty choices inside the science (p. 470). The chapter concedes there may be “no effects to be found” (p. 468).

6. Public research drifts toward private logics. Public institutions patent, adopt industry priorities and partner with industry (pp. 466–467). This creates a loop in which top-down researchers win funding and crowd out others (p. 467). “When public institutions must act in a way that is consistent with how companies must act”, farmers’ knowledge access worsens (p. 471). The empirical anchors are modest (Graff 2003; Murray & Stern’s “modest” effect). The loop itself is inferred.

7. Concentration narrows the pipeline. CR4 passes a “critical threshold” (p. 467). Firms have “reduced incentives to develop conventional varieties”, and in US soy conventional breeding retreats to universities and niche firms (POST 2011, p. 467). Licensing “can bottleneck technology transfer” (p. 472). Concentration feeds power over framing (p. 477): a “feedback cycle that concentrates resources, knowledge, and influence” (p. 477).

8. Treadmills from simplification. Single-tactic control of an adaptive system selects for resistance. More product, or stacked products, deepens dependence (p. 462). This is the best-evidenced dynamic, and it has a deskilling side-effect (p. 462).

9. Lock-in through knowledge loss. - Alternatives decay during use of the dominant path. Farmers lose weed-management know-how, and research and extension neglect IWM (p. 462). Seed-saving knowledge and networks may be “difficult to re-establish” (p. 472). - There is also economic and political lock-in: the “Economic and political forces that promote these innovation trajectories then become hard to reverse, or crowd out alternative approaches” (p. 477), with path dependence in research choices (p. 475). - [my assessment] This irreversibility is institutional and cognitive rather than physical, a less common emphasis among the Late Lessons cases.

10. Distribution of costs, benefits and risks. - Benefits. They flow to providers and large farms (p. 476). Bottom-up proceeds “flow to adopters rather than the providers” (p. 476). So bottom-up champions lack resources and are “easily displaced” (p. 476): the alternative’s virtue, distributed benefit, is also its political weakness. - Costs. Seed and input rents fall on farmers and are prohibitive for the poor (pp. 463, 465). Taxpayers fund the subsidies (p. 464). Developing countries lose about USD 17 billion a year (p. 464). Export-oriented poor countries risk a “loss leading treadmill” (p. 465). - Risks. Exposure is global through commodity chains, while production is concentrated (p. 468). Environmental costs: resistance, deforestation, energy and emissions (pp. 459, 462, 466).

11. Policy supports shape apparent competitiveness. GM crops cluster in subsidised commodities (p. 464). Top-down innovation “ironically is maintained by extra-market subsidies” (p. 476). The GM-specific causal link is asserted.

12. Promise versus delivery. Promises of feeding the poor, stress tolerance and fewer pesticides turned into two traits for the richest markets (pp. 463, 466), plus “inflated benefit claims” (p. 468). The chapter explains the gap by incentives, not bad faith.

13. The actors’ mental models (as portrayed). The chapter quotes almost no proponents, firms, regulators or adopting farmers directly. Their thinking is reconstructed from policy texts and outcomes, itself a limitation. - Innovation policymakers: innovation as product flow; IP as incentive; friction as barrier; innovation as competitiveness and jobs (p. 461). - IP proponents: IP yields “net social benefits” (Pray 2007), a view blind to pipeline effects (p. 467). - Risk assessors: “no evidence of harm” means safety; “within the range of biological variation” reassures; precaution is a management add-on (pp. 468–470). These are institutional habits rather than individual choices. - Professional breeders: isolated from and “unaware of” farmers’ multidimensional preferences (p. 472). - Scientists generally: shaped by post-WWII techno-science culture, an “irresistible” fusion of internal culture with economic and political power (p. 476). Separately, the chapter warns that “a singular, centralised and highly specialised approach to agricultural development… can stifle other approaches” (p. 466); that sentence describes a development pathway, not scientists’ mindset as such. - Adopting farmers: drawn by “lower labour costs through a simplified weed management strategy” (p. 462). Rational within subsidised, large-scale systems (pp. 464–465). - Rich-market consumers: sceptical of “claims of net benefit” given no direct benefit to them (p. 464).

14. Language and framing devices. - Terms the chapter critiques: “no evidence of harm” and “within the range of biological variation” (pp. 468–469); “Innovation Union” (p. 461); “so-called” used for “Green Revolution” (p. 460), “mega-countries” (p. 463) and “‘biotech’ crops” (p. 476). - The chapter’s own devices: - Metaphors: “treadmill” (pp. 462, 465), “deskilling” (p. 462), “addicting us” (p. 475), “black-box technology” (p. 476), “smother” (p. 470). - Slogans: “design agriculture to fail” (pp. 458, 460); “Change the directions, distribution and diversity of innovation, and you change the world” (p. 478). - Evaluatives: “The results speak for themselves” (p. 474), “immense” (p. 474), “stark realisations” (p. 470). - The chapter is a framing intervention as much as an analysis.

15. Complexity. The chapter works with multifunctionality and cross-scale coupling (p. 459), adaptation and resistance (p. 462), feedback loops (pp. 467, 477), resilience through diversity (pp. 471–472), gene × environment interaction (p. 473) and problem-shifting “in time or space” (p. 475). Fixes “aggravate every problem” (Lederberg, p. 477).

16. Innovation effects. The distinctive claim is that the precautionary critique is also pro-innovation, about diversity rather than speed. IP-led innovation closes down options (p. 458). Participatory IP allows “a wave of innovation” (p. 472). Diversity “buffers against lock-ins” (p. 477). Prioritising pace and scale over institutions makes benefits “elusive” (p. 463).

Transferable insights (technology-neutral)#

1. Ownership-tied incentives favour what can be packaged and sold over what is most needed. Knowledge-intensive, context-specific, non-excludable solutions are systematically under-supplied. - Evidence: pp. 460, 465–466, 475; examples pp. 460, 462, 475. - Strength: moderate. Standard appropriability economics and coherent examples; no quantified counterfactual.

2. Products that need uniform operating conditions serve large, homogeneous, well-resourced users first and heterogeneous, poor users last. - Evidence: pp. 461–462, 464–466, 476; Figs 19.1–19.2. - Strength: moderate. Adoption patterns fit; rival explanations untested.

3. How policy defines “innovation” is a distributive choice. Product- and competitiveness-centred definitions pre-select problems, solutions and beneficiaries. Powerful actors can shape the problem to fit their solution, moving debate from “whether/why” to “how to use”. - Evidence: pp. 459, 461, 475, 477. - Strength: suggestive to moderate. Well argued from the EU’s own language; capture asserted more than traced.

4. Promises narrow to what pays. Broad early promises narrow in delivery to applications with the largest paying markets. - Evidence: pp. 463, 466, 468. - Strength: moderate to strong for the narrow trait portfolio (documented by 2012); moderate for “inflated claims”, which partly rest on contested sources.

5. Single-tactic control of an adaptive system produces a treadmill. Resistance is met with more product or stacked products, displacing the problem and deepening dependence. - Evidence: p. 462. - Strength: strong. Well documented by 2012 (NRC 2010; Powles 2008; Heap) and follows from selection theory.

6. A simplifying technology can erode the skills, services and social infrastructure that alternatives need, creating lock-in through knowledge loss. - Evidence: pp. 462, 472. - Strength: moderate. Specific sources (Binimelis; Mortensen; Howard); little quantification.

7. Supplier concentration narrows the innovation pipeline. Dominant firms under strong property rights neglect lower-margin or non-proprietary alternatives. - Evidence: pp. 467, 472. - Strength: moderate. Authoritative source (UK POST), essentially one example (US soy).

8. When public research institutions adopt private incentives, public-good research is crowded out and knowledge diffusion slows. This may self-reinforce through funding. - Evidence: pp. 466–467, 470–471. - Strength: moderate for the diffusion effect (Murray & Stern: “modest”); suggestive for the funding loop (inferred).

9. Developer-controlled evidence is structurally biased toward “no evidence of harm”. Contributing features: confidentiality, restricted independent access, low power, surrogate test materials, short durations, broad comparators, weak post-market monitoring. - Evidence: pp. 468–470. - Strength: moderate. The individual mechanisms are well grounded (funding effect, Diels 2011; power, Marvier 2002; access, the 26-scientist statement). The claim that harms were missed here rests on contested studies.

10. “No evidence of harm” is not “evidence of no harm”. A null result is only informative if the search could have detected the effect. Decision-makers should be told what was not, and cannot be, tested.

    • Evidence: pp. 468, 470.
    • Strength: strong as an epistemic principle; moderate as a description of practice.

11. Safety research lags development research when funding follows the product pipeline.

    • Evidence: p. 468 (1% of USDA biotech research, 1992–2002); echoed in the report synthesis (p. 646, outside section).
    • Strength: moderate. A specific figure from a single source (Mellon & Rissler 2003, a conference paper; [background] by Union of Concerned Scientists staff). The report’s editors add a parallel EU figure, about 3% of EUR 28.5 billion for nano-, bio- and information technology spent on hazards (p. 646, outside section).

12. Researchers reporting unwelcome findings can face professional costs, chilling safety research.

    • Evidence: p. 468.
    • Strength: suggestive. Documented episodes exist in the cited literature but none is named, some involved contested studies, and the lead author’s history is relevant context.

13. Precaution can operate inside the science, through deliberate error-type choices, explicit evidence thresholds and stated limits of detection, not only in management after it.

    • Evidence: p. 470.
    • Strength: asserted, with a sound philosophical basis (Rudner 1953; Lemons 1997). No applied demonstration.

14. The recurrence of familiar institutional patterns can itself be an early warning before harm is shown. Examples: weak standards, developer-controlled evidence, low-power tests.

    • Evidence: p. 470.
    • Strength: suggestive. A useful reframing but at risk of unfalsifiability.

15. Evaluation windows bias comparisons. Alternatives that take years to mature look worse in short-run trials.

    • Evidence: p. 474.
    • Strength: suggestive. One contested meta-analysis, and the chapter’s reading needs checking.

16. Users hold knowledge that central developers miss. Participation surfaces valued attributes, and property regimes permitting downstream modification let improvements compound.

    • Evidence: pp. 472–473.
    • Strength: moderate. Participatory-breeding and field-school evidence; Fig. 19.3 is conceptual.

17. Where the provider captures little value, the alternative is politically weak. Its champions are “easily displaced” by wealthier promoters of proprietary paths.

    • Evidence: p. 476.
    • Strength: suggestive. A compelling collective-action logic, asserted without cases.

18. Concentrated production can create diffuse, global exposure through supply chains. [my extrapolation] Exposure may then fall outside governance confined to producing jurisdictions; the chapter does not make this point.

    • Evidence: p. 468 (“exposure is global even if production is mainly in a few countries”).
    • Strength: strong as a structural point, but asserted in the chapter without data on actual exposure levels.

19. A technology’s apparent competitiveness can depend on policy supports rather than intrinsic merit.

    • Evidence: pp. 464, 476.
    • Strength: suggestive. A correlation; the causal claim is not shown.

20. Diversity of innovation pathways hedges against lock-in and uncertainty. Governing direction, distribution and diversity, not just rate, is a coherent frame for technology choice.

    • Evidence: pp. 475, 477–478.
    • Strength: asserted. A borrowed normative framework (STEPS; Stirling), untested here.

Limitations, contestation and bias check#

Advocacy over analysis. - The chapter openly argues for one pathway. Its conclusions escalate beyond the evidence presented: “increases food insecurity” and “will always serve last” (p. 476), “will fail” (p. 470; “will most likely fail”, p. 477) and “has not, and will not” (p. 462, though introduced by “appears to be”). - The hedges are real but thin: “not… irrelevant” and “both have value” (p. 470), the “right mix” (p. 475), fn 1’s “hypothetically” (p. 460). - The editors summarise the chapter more cautiously than it speaks (p. 644, outside section).

No panel, no reply. There is no regulator, industry, crop-science or development-economics commentary on a polarised topic. All authors come from one intellectual community. The lead author was on the editorial team and one author on the advisory board (p. 5). The chapter’s own “funding effect” and conflict-of-interest logic (p. 469) applies to institutional and intellectual commitments too, but it is not turned inward.

Asymmetric evidentiary standards (the main fairness problem). - GM benefit claims get forensic scrutiny: power, independence, confidentiality (pp. 468–469). - Agroecology claims are accepted at face value (pp. 473–474), though several share the weaknesses criticised: - Project syntheses (UNEP-UNCTAD 2008; Pretty 2008, 2011) are mostly before-and-after comparisons of selected, often successful projects, with self-reported outcomes and no controls. - Rodale is an organic-advocacy institute. - UNEP 2011 is a scenario model presented as showing “immense” potential. - “The results speak for themselves” (p. 474) is the reasoning the chapter would reject from industry.

Selective citation. [background, verify] - Sources are used for narrow points while their headline findings, which cut the other way, go unmentioned: - NRC 2010 is cited only for weed resistance (p. 462); its overall assessment of US farm-level benefits was broadly favourable. - WHO 2005 is cited on IP (p. 467); it also judged marketed GM foods unlikely to present health risks. - Marvier et al. 2007 is cited only for “long lag time frames” (p. 469); it is widely read as finding non-target invertebrates more abundant in Bt than insecticide-sprayed fields (though less abundant than in unsprayed non-GM fields for some taxa, so it does not cut only one way). - World Bank 2007 is said to have “dismissed” IP instruments (p. 467); it also saw potential in transgenic crops for poor farmers. - Contrary peer-reviewed work available before publication is absent: - Bt cotton income and yield gains in India (Qaim & Zilberman 2003; Kathage & Qaim 2012); - Bt cotton biocontrol benefits in China (Lu et al. 2012); - organic yield-gap meta-analyses finding yields about 20–25% lower on average (Seufert et al. 2012; de Ponti et al. 2012).

Contested health evidence. - The health “indications” rest mainly on the Séralini group (2009, 2010, 2012; Mesnage 2012, in vitro), Greenpeace, Earth Open Source and Drucker. For environmental “evidence of adverse effects” Box 19.2 cites Rosi-Marshall 2007 (stream ecosystems), Bøhn 2008 (Daphnia; co-authored by Traavik of GenØk) and Hilbeck 2012 (ladybirds); [background] each was disputed in the literature. - [background] Séralini 2012 drew detailed criticism from EFSA and national agencies in late 2012, before publication of this report, and was retracted by the journal in November 2013 (later republished elsewhere). The chapter does not acknowledge the criticism. The report’s editors mention the “controversies over… GM maize (Seralini et al., 2012)” only as a spur to regulatory data transparency (p. 660, outside section). - The wording is hedged (“indications”, “may”). The source selection is not.

Weaker evidence chosen over stronger. [background] Documented regulatory and containment failures would have supported the institutional-pattern thesis far better than contested toxicology: an unapproved-for-food maize variety entering the US food supply (2000); unauthorised GM rice in US exports (2006). Neither is used.

Framing of numbers. - “Just 3 %” (p. 463; Fig 19.2) divides by FAOSTAT “agricultural land”, which includes permanent pasture (about 4.9 billion ha). [my approximate calculation] Against cropland (about 1.5 billion ha), 159 million ha is about 10%. Both are defensible; “in reality is just 3 %” is a minimising frame. - [background, verify] ISAAA reported GM cultivation in about 29 countries in 2011, not only 17; ISAAA’s “mega-countries” are those growing at least 50,000 ha. The chapter’s text (“leaving a total of 1 % … among just seven other countries”, p. 463) reads as if the 17 were the whole total, while its own Figure 19.1 title says “the 17 largest producing countries” (p. 464). - [background] The “10:1” energy ratio (p. 459) is usually a whole-food-system figure, though the chapter’s parenthesis does define the energy broadly. - The “75%” yield-gap quote is from Molden 2007 (IWMI-led), attributed to “the UN FAO” (p. 472).

Badgley et al. 2007 characterisation needs checking. [background, verify] - Badgley reported organic:conventional yield ratios of about 0.92 (developed countries) and about 1.8 (developing countries). - The developing-country ratio came largely from comparisons with low-input local practice, not “industrial” farming, contrary to the chapter’s framing (p. 473). - The study was criticised on data selection and nitrogen assumptions. - I cannot confirm from memory that it found mature conversions “consistently out produced industrial operations” or that it “exposed the reason” for contrary findings (p. 474). The chapter itself softens the first claim (“provided evidence to suggest”), but states the second without a hedge.

Conflation and dichotomy. - The chapter concedes the gene-versus-environment dichotomy “is in essence artificial” (p. 465), then builds on it. - It maps GM onto top-down and agroecology onto bottom-up, folds organic into agroecology (pp. 471, 473–474), and leaves proprietary conventional breeding and hybrids, also top-down, largely outside the frame. - Many attributed harms belong to industrial commodity agriculture generally: monoculture, subsidies, export orientation, deforestation. Footnote 1 (p. 460) concedes GM is not inherently high-input, but p. 476 indicts “‘biotech’ crops” as such. - The recommended remedies run through top-down institutions (IAASTD, SCAR, public funding).

Alternative explanations not engaged. - [background] For low adoption (pp. 464–465), a prominent rival account (e.g. Paarlberg 2008) blames restrictive biosafety regulation and fears of losing European export markets. If right, precautionary regulation itself contributed to non-adoption that the chapter reads as rejection on the merits. - The subsidy argument does not separate GM from non-GM commodities (p. 464).

Internal tensions. - Hedge versus conclusions: “not… irrelevant” (p. 470) against “increases food insecurity” (p. 476). - Toxicity testing: “no regulatory framework requires mandatory toxicity or allergenicity testing” sits beside “commonly, only 90-day… feeding trials are conducted” (p. 469), and beside the chapter’s own acknowledgement that most countries follow OECD/Codex and Cartagena guidance (p. 469). [background] That guidance did require assessing newly expressed proteins’ toxicity and allergenicity; the real point is about long-term whole-food testing. The EU made 90-day whole-food studies mandatory by implementing regulation in April 2013, around publication. - Surrogate proteins: [background] the critique (p. 469) omits that regulators generally require evidence that the surrogate and plant-produced proteins are equivalent. - PVP: described generally as “patent-like” (p. 467) yet PVP that “recognise[s] breeder’s rights” is contrasted with patents and “patent-like PVP” (pp. 472, 475, 476); inconsistent terminology rather than a contradiction.

Hindsight and foresight bias. As a Part C emerging issue, the risk is precautionary anticipation bias. Box 19.2 maps GM onto asbestos, benzene and BSE before harm is established. Making the lesson institutional rather than harm-based (p. 470) protects it from disconfirmation by an absence of harm.

Fair in the other direction. - Much of the structural critique drew on authoritative, non-activist sources: IAASTD, the UN Special Rapporteur, UK POST, Nature, NRC on resistance, Murray & Stern. - The HT treadmill, seed concentration, the narrow trait portfolio, the role of conventional breeding in yield gains, and the absence-of-evidence critique were well founded. [background] Several were largely confirmed later. - The innovation-policy analysis is thoughtful and anticipates later “responsible innovation” discourse, a term the chapter itself uses (p. 476). - It is explicit about case selection (pp. 460, 471), concedes traditional farming’s harms (pp. 459, 471), concedes that “Either pathway” could cut food costs and offer “a means out of poverty” (p. 461), states the pro-IP case fairly (p. 467), keeps sources’ qualifiers (“modest”), acknowledges critics of precaution (p. 470), allows there may be “no effects to be found” (p. 468), and says seed-based and environmental approaches “both have value in all agroecosystems at all times” (p. 470). - Overall. Strongest: the political economy of innovation incentives, the treadmill, pipeline and concentration effects, and risk-assessment epistemics. Weakest: health-harm insinuation and the uncritical agroecology yield record. In between: distributive claims, well reasoned but under-evidenced.

Pointers for the hindsight strand (background knowledge, NOT verified; check before use)#

• Resistance and stacking. Glyphosate resistance kept spreading. Dicamba- and 2,4-D-tolerant stacked traits were commercialised in the mid-2010s, followed by widespread dicamba drift damage (2017–18) and US court vacaturs of dicamba registrations (2020, 2024). Tests the p. 462 forecasts.
• Glyphosate health. IARC Group 2A (2015). EFSA and ECHA disagreed. EU renewals (2017; 2023, for 10 years). Large US Roundup litigation and settlements. Keep separate from GM-food safety.
• GM food safety. NASEM 2016: no substantiated evidence GE foods are less safe; no clear aggregate US yield-potential gain from GE traits; Bt reduced insecticide use and losses; HT resistance a major problem. EU-funded long-term rat studies (GRACE; G-TwYST, about 2018) found no adverse effects of NK603. Séralini 2012 retracted in 2013.
• Farmer outcomes. A 2014 meta-analysis (Klümper & Qaim) reported average yield, pesticide and profit benefits, mainly from insect resistance in developing countries; contested. Indian Bt cotton faced pink bollworm resistance from the mid-2010s. Burkina Faso phased out Bt cotton around 2016 over fibre quality.
• Traits and adoption. About 190 million ha in about 29 countries by 2019, still dominated by the same five countries and by HT and insect-resistance traits. GE drought-tolerant maize (2013) and wheat (Argentina, 2020 onward). Conventional drought-tolerant maize spread widely in Africa. Tests p. 475.
• Concentration and IP. Major seed and agrochemical mergers (2016–18) (tests p. 467). The US Supreme Court’s Bowman v. Monsanto (2013) upheld the ban on replanting patented seed (tests the p. 467 seed-saving claim).
• Transparency. The EU’s 2019 food-chain risk-assessment transparency regulation (study disclosure, notification of commissioned studies, verification studies) addresses several Box 19.2 concerns.
• Agroecology and organic. Meta-analyses (Seufert 2012; Ponisio et al. 2015) found organic yields about 19–25% lower on average, with the gap narrowed by diversification. Agroecology gained institutional standing: FAO symposia (2014, 2018), the Committee on World Food Security’s High Level Panel of Experts report (2019), the EU Farm to Fork organic target (2020), a CGIAR initiative. Sri Lanka’s abrupt 2021 synthetic-fertiliser ban (an imposed input ban, not agroecology as defined here) cut yields and was reversed, a cautionary case for transition design.
• Hunger. The SDGs replaced the MDGs (2015). FAO revised its undernourishment methodology (2012). After years of decline, FAO estimates showed hunger stalling and then rising from the mid-to-late 2010s, and sharply around 2020.
• New breeding techniques. These post-date the chapter; how its IP and top-down analysis applies to them, and the EU debate over whether they are GMOs, are open. Note only: outside the reports.

Notable quotes#

1. “If we fail to address the governance of innovation in food, fibre and fuel production now, then current indications are that we will design agriculture to fail.” (p. 458)
2. “the problems identified for solution will tend to be those that can be packaged and sold — usually to the largest/wealthiest/most lucrative market and largely bypass the poor” (p. 460)
3. “How innovation is conceived shapes how it is promoted, and who benefits from the promotion.” (p. 461)
4. “farmers have entered into a treadmill where overuse of a single product leads to tolerance and tolerance is overcome with more product” (p. 462)
5. “these asymmetries in investment incentives mean that management-based approaches do not receive the same levels of support and investment as do approaches that are easily recaptured in the marketplace.” (p. 466)
6. “the safety of GM crops is presumed when there is a lack of evidence of harm, as if this were equivalent to evidence of lack of harm, when it clearly is not.” (p. 468)
7. “The critical late lesson that may be emerging from GM crops is not the evidence of harm — the early indications of harm are just emerging — but the persistence of the same institutional patterns that led to the old late lessons already learned from asbestos, benzene and BSE” (p. 470)
8. “When public institutions must act in a way that is consistent with how companies must act, then the imbalance between farmer and knowledge access grows.” (p. 471)
9. “Through this power, top-down providers can artificially homogenise both the conception of the problem to be solved and the solutions” (p. 477)
10. “Change the directions, distribution and diversity of innovation, and you change the world.” (p. 478)

Open questions#

1. Evidence for agroecology. How well does its yield evidence (Badgley; UNEP-UNCTAD; Pretty) hold against controlled comparisons, and what did later meta-analyses find?
2. Why adoption stayed low. Was limited adoption in poorer countries driven by cost and unsuitability (the chapter), by regulatory and trade barriers (the rival view), or both? The answer changes the lesson about precaution.
3. Which forecasts held up after 2013? Did the HT treadmill and stacking forecasts (p. 462) play out? How did smallholder outcomes from insect-resistant crops compare with “will always serve last” (p. 476)?
4. Risk-assessment reform. Did the Box 19.2 patterns change, especially in the EU, and did changes reveal missed harms or confirm null findings?
5. Is the lesson falsifiable? Is “institutional pattern as early warning” (p. 470) falsifiable? What outcome would show the pattern was not a warning here?
6. Public research. How did the public-private R&D balance and public patenting evolve, and is there evidence for the funding loop (p. 467)?
7. Intellectual property. Which IP regimes, if any, enabled the downstream “waves of innovation” of Fig. 19.3 (p. 473)? What happened to farmers’ seed rights?
8. Scale. Can low-margin, knowledge-intensive innovation scale without concentrated provider revenues? If providers capture little (p. 476), who funds the extension services agroecology needs (p. 471)?
9. Who the authors were. Would disclosing the authors’ positions and histories, or an adversarial panel, have changed the conclusions?
10. Missed examples. Why rely on contested toxicology rather than documented regulatory and containment failures that fit the institutional thesis better?

Audit log#

Independent audit against the full text extract (PDF pp. 460–487), with PDF checks of Fig. 19.1 (p. 464) and the italics on p. 477, and spot checks of the outside-section pages cited (pp. 5, 8, 10, 644, 646, 660; annex pp. 685, 689, 691, 694, 695, 696). Web sources were not used.

• Authors: Myhr’s VKM membership changed from “former member” to the annex’s “has been a member”.
• Authors: Funtowicz “co-developed post-normal science” replaced with the annex’s wording; the attribution is now tagged [background].
• Panels: added the report’s statement that case-study chapters were peer reviewed (p. 5), for fairness.
• Self-citation: added p. 467 to the Hope not Hype page list; listed further Heinemann items, including Heinemann & Kurenbach 2008b as the sole source for Box 19.1’s verdict.
• Network ties: added that Traavik 1999 is a GenØk report and that Traavik co-authored Bøhn 2008 and Myhr & Traavik 2002 (all from the reference list).
• Background: added a flagged note that Delborne 2008 appears to be a study of the lead author’s own controversy.
• Editors’ framing: added p. 660, where the editors cite the Séralini controversy only as a spur to regulatory data transparency.
• Energy (p. 459): quote corrected to “on average require… food produced”; the chapter’s broad parenthesis added; the whole-food-system gloss tagged [background].
• Green Revolution (p. 460): restored the chapter’s hedge “perhaps the most evident early warning”.
• Footnote 1 (p. 460): restored “hypothetically”, changed “key concession” to “limited concession”, and added the economies-of-scale sentence.
• 19.2 (p. 461): added the concession that “Either pathway” could lower food costs and provide a way out of poverty.
• 19.2 (p. 463): added the productivist “re-balancing” passage (van den Hove 2012) and Baldwin 2011’s “increasing evidence” claim.
• Box 19.1: corrected the citations for “advantages… disappearing” (Service; Pengue; Benbrook) and for resistance.
• Box 19.1: noted that glyphosate harm is said to be “documented in the scientific literature” while Greenpeace is cited.
• Box 19.1 verdict: restored the “appears to be” hedge, the “developed country agriculture” clause and the single self-citation.
• Box 19.1 (Jost 2008): the Georgia scope is now sourced to the reference-list title, and the “other high yield varieties” comparator added.
• 19.3 (p. 463): noted that the text implies only 17 GM countries, whereas Fig. 19.1’s title says “the 17 largest”.
• Fig. 19.1: added the caption’s internal inconsistency (<50,000 ha against Spain at 100,000 ha as the lowest shown).
• Fig. 19.2: the pasture-in-denominator point is now tagged [background].
• Patchy adoption (p. 464): restated the second reason as the chapter gives it (developed-country subsidies undermine developing-country markets) and the third as subsidy-linked cost.
• 19.3.1 (p. 466): added the “singular, centralised and highly specialised approach… can stifle other approaches” sentence.
• 19.3.2 (p. 467): restored the qualifier “at least at their stage of development” on the WHO/World Bank “dismissed” claim; the overstatement comment is now tagged [my assessment].
• 19.3.2 (p. 467): noted that the chapter presents the POST briefing as “the UK Parliament now says”.
• Interim verdict (p. 470): added “likely contrary”, the stated reason (lack of incentives) and the hedge that seed-based and environmental approaches “both have value in all agroecosystems at all times”.
• Box 19.2 framing: Drucker is now characterised from the reference-list subtitle, with the litigant and spelling points tagged [background, verify]; footnote 5 detail (3 of 24 documents) added.
• Box 19.2 framing: full citation lists restored.
• Box 19.2 (Mesnage 2012): corrected. It is cited for the less-toxic-pesticides claim, where it is topically relevant but weak (in vitro), not for pesticide use.
• Box 19.2: added the box’s organising questions and the “important effects may be missed” sentence.
• Box 19.2: added the regulatory-context sentence (OECD/Codex/Cartagena guidance) and the unintended-changes/cooking-and-processing point.
• Box 19.2: corrected the 120-day quote to “indications of adverse health impacts”; added Séralini 2009 and Spiroux de Vendômois 2010; tagged “most contested” as [background].
• Box 19.2: added the named sources for career-risk claims (Delborne; Lancet 1999; Waltz; Lotter).
• Box 19.2 precaution: added the chapter’s acknowledgement of critics, Funtowicz 2003, Brosi 2009, Lemons 1997 and the error-type communication sentence.
• Box 19.2 conclusion: added Harremoës 2001 (LL1) and the remedy; the bold “institutional, not harm-based” softened to “primary warning”, noting the chapter’s assertion that “early indications of harm are just emerging”.
• 19.4.2 PVP: “conflicts with p. 467” softened to inconsistent terminology across pp. 467, 472, 475, 476.
• 19.4.3: restored Badgley’s “provided evidence to suggest” hedge; added UNEP-UNCTAD’s “organic and near-organic” scope; noted that the cited Posner 2009 is the economic and risk-analysis paper.
• 19.5: the management package now includes “locally-adapted drought tolerant varieties” and the claim that management “improve[s] the genetics of crop plants”, which the notes had omitted.
• 19.6: completed the “right mix” quote (benefits and risks “more evenly distributed”).
• 19.6: added the p. 476 comparative claim that bottom-up success is “far more likely”.
• 19.6: added the “black-box”/”(GM and similar)” framing and the sources for the concentration claim.
• 3D agenda: corrected the subject of “become hard to reverse” to the “Economic and political forces” behind lock-in.
• 19.7: added the “early warning, or perhaps late lesson” framing and the “Perhaps” hedge on social deliberation.
• References: count corrected from about 170 to about 178; added the Heinemann & Kurenbach 2008/2008b duplicate and the NRC/OECD merged entry.
• References: reclassified Pray & Naseem 2007 as academic, not “industry-side”; the UCS affiliations are now tagged [background].
• Timeline: the 26 scientists’ description now follows the chapter and editors (“academics”), with “corn-insect researchers” tagged [background]; added GenØk and Lancet details for 1999; the herbicide-resistance lag is now tagged [my inference].
• Predictions: the header now reads “certainty or near-certainty”, and each item notes the source’s own hedges (“appears”, “most likely”, “are likely to”, “current indications”).
• Mechanisms 9 and 13: the corpus-level comment is now tagged [my assessment]; corrected the misapplication of the p. 466 “singular, centralised” sentence to scientists’ mindset.
• Insight 11: the source is identified (Mellon & Rissler 2003, UCS affiliation tagged [background]); added the editors’ parallel EU 3% figure (p. 646).
• Insight 18: removed the unsourced “escaping governance” gloss from the insight itself (kept as [my extrapolation]); strength qualified as asserted without data.
• Limitations (selective citation): Marvier 2007 background balanced (fewer non-target invertebrates than in unsprayed fields for some taxa).
• Limitations (contested evidence): Rosi-Marshall, Bøhn and Hilbeck moved from “health” to environmental evidence, with contestation tagged [background]; added the editors’ p. 660 handling of Séralini.
• Limitations (numbers): added the ISAAA 50,000 ha mega-country threshold and the chapter’s 17-country ambiguity; the energy point is now tagged [background].
• Limitations (internal tensions): the toxicity-testing and surrogate-protein points are now tagged [background], with a link to the chapter’s own mention of OECD/Codex guidance; the PVP point softened.
• Advocacy and fairness: the escalation list now notes the source’s hedges; added further concessions the chapter makes (pp. 461, 470).
• Hindsight pointers: the hunger-trend wording made less date-specific (still unverified background).
• Notable quote 7: restored the omitted clause “the early indications of harm are just emerging” instead of an ellipsis.
• Digest: added the peer-review statement; restored “current indications are”; added a patchy-adoption bullet; added the Box 19.1 verdict with its “appears” hedge and its source.
• Digest: added the Box 19.2 “no effects to be found” concession and the drought-tolerant-varieties element of the management package.
• Digest: nuanced the “institutional, not harm-based” lesson and expanded the hedges (pp. 460, 461, 470).
• Digest: qualified insight 12’s strength and fixed the garbled “10% of cropland” phrasing; tagged the pasture and Séralini points [background].
• Digest: added the 17-country ambiguity and the self-citation and network caveat.