Reconstructing facts in Bt cotton: Why scepticism fails

In disputes around Bt cotton, a “triumph narrative” is alleged to have emerged from researchers - mainly economists - catering to vested interests of the biotech industry, its funding and allied journals promoting biotechnology. This column explains why the ‘conspiracy theory’ fails, and then illustrates why the main claims of the peer-reviewed literature demonstrating agro-economic success of Bt cotton are consistent with the near universal adoption of the technology by farmers in India.

Bt cotton has assumed a surprisingly large place in both global debates around genetic engineering and in Indian politics. Controversies have raised important questions in the sociology of knowledge and politics of science – how are facts relevant to public policy settled? More radically, are ‘facts’ socially constructed in instrumental ways such that they themselves are a modernist fiction? Extreme scepticism characterises a postmodernist and constructivist unsettling of the very notion of facticity (Latour 2003).

What is at stake?

Politics that follow from epistemic scepticism are often powerful. Naomi Oreskes and Erik Conway (2010) demonstrated in ‘Merchants of Doubt’ how small numbers of scientists tied to industry were able to create doubt about the science behind global warming, the link between smoking and cancer, ozone depletion and other phenomena in which settled science would hurt corporate interests if not strategically unsettled. Doubt has its uses. The critical move is to flatten the terrain intellectually – all studies are weighted equally, there are always dissenting ‘studies’ and there is no way to judge among them. More powerfully, opponents of the climate science consensus charge that scientists have a vested interest in finding dangers to get funding to study them. Attaching an interest to researchers is a strategy for undermining conclusions without accepting any burden of proof. Convergence around these important problematics appeared in the pages of the Economic and Political Weekly (EPW) via Glenn Stone’s “Constructing Facts: Bt Cotton Narratives in India” (Stone 2012). Chandrasekhara Rao and I are implicated via our own EPW article (Herring and Rao 2012), on the “Failure of Bt Cotton”.

Stone’s scepticism seems at first blush sweeping and radical – “we simply cannot say how Bt seed has affected cotton production in India” (Stone 2012: p 63). Why? ‘Facts’ as presented in the numerous peer-reviewed articles on Bt cotton are ‘constructed’ within narratives and contaminated by interests. There is then no way to adjudicate between a nongovernmental organisation (NGO) – constructed narrative of Bt failure, and a “triumph narrative” of peer-reviewed journal articles. The important question for social scientists and policymakers would then be to find the interests and mechanisms of distortion, since permanent agnosticism is not a viable strategy for dealing with the world.

‘Constructed’ pro-GM facts?

What does Stone discover? He claims that the “triumph narrative” of Bt cotton in India “flows mainly from economists and the biotech industry (and its academic allies)” in “industry-journal authentication systems” (peer-reviewed journals) which “serve the interests of their constituent parties” (Stone 2012). This system “creates pro-GM facts through the interaction of a different set of interested parties”. The convergence of multiple studies using different methods on the success of a single trait – insect resistance – improving agro-economic results on farms is produced by this conspiracy. “The triumph narrative has been generated, authenticated, and disseminated by a particular system of interacting parties with overlapping interests”. The mechanism is a “cosy alliance between GM manufacturers and ostensibly independent researchers” (Stone 2012).

Stone makes strong claims, even by the standards of conspiracy theories. What is the evidence? How do we know that researchers in dozens of studies showing positive effects of insect resistance in Bt cotton are part of some “cosy alliance” and only “ostensibly” independent? What exactly makes peer-reviewed journals normatively predisposed to “pro-GM facts”? Are journals, like researchers, only ostensibly independent?

This is an especially awkward case for Stone to sustain. One of the journals that publish “pro-GM facts” is World Development – cited twice in his reference list. But Stone (2011) himself publishes in World Development. Should we discount his good empirical work in Warangal district because it is published in the “industry-journal authentication system”? Do authors published in World Development really get cheques from Monsanto¹? The breadth of journals indicted for having published “pro-GM facts” strains credulity – from the EPW in India to the Proceedings of the National Academy of Sciences in the US.

Caveats about pro-Bt narrative

What is valuable in Stone’s EPW article are caveats about triumph narratives of silver bullets from genetic engineering. Bt produces one trait; it affects only biotic stress² from one class of insects. Yields are driven by numerous traits, characteristics of germplasm³, and biotic and abiotic stresses that vary continually. There will be variation, field-to-field, season-to-season. Hence, variation across studies simply reflects the nature of agriculture. Finding the yield effect of a single trait – insect resistance induced by the Bt transgene – presents special difficulties. Pest pressure varies greatly by season, influenced by many factors, including weather. Bt gives farmers one great advantage when bollworm pressure is high; when pressure is low, the trait is less valuable in protecting harvestable yield. In that case, the extra expense of Bt seeds is comparable to an insurance policy payment. Compare irrigation water. Yield effects of irrigation water in years of scanty rain are large, even decisive, but add less to yield in wet years. Yet no one doubts the contribution of irrigation to agriculture. It is true that isolating the effect of one trait is difficult even conceptually. And cross-sectional studies do suffer the cultivator bias problem,⁴ as Stone notes. This is the reason for longitudinal analysis, which controls for cultivator bias by using data from the same farmers in the same fields before Bt and after Bt - nothing changes but the Bt trait over seasons (Herring and Rao 2012).

Agro-economic success of Bt cotton

Despite the inevitable variation in field studies, there is a very strong centre of gravity around the success of Bt cotton agro-economically – lower cost of production per unit and thus higher net returns to the farmer. To draw an analogy to ‘Merchants of Doubt’, it is difficult to ascertain how much cancer is caused by tobacco smoking; the variables are multiple and interactive. Sample sizes are small in clinical trials, genetics play a role, and other behaviours matter, as well as environmental and other factors. Studies are always limited and flawed, and anecdotal evidence will produce exceptions to generalisation: for example, people smoking for 60 years without cancer. But problems in assessing the precise contribution of smoking to cancer would not justify concluding that we do not know if there is a relationship. Denial would be contrary to public health, as most nations now recognise, but consonant with the interests of tobacco companies who for many years sought to unsettle medical consensus. Stone wants to unsettle the dominant findings on Bt cotton by levelling the epistemic field. ‘Facts’ from networks claiming that the failure of Bt cotton is total and ‘genocidal’ are equated with ‘facts’ from extensive rigorous fieldwork. Facts are all ‘constructed’. Ironically, this epistemic relativism does not prevent Stone from constructing a refutation of the Bt success narrative.

Are effects of Bt on cotton yields declining?

Stone’s rebuttal of the Bt “triumph narrative” has two legs. First, he argues that Bt could not have been a major driver of dramatically higher yields because Bt adoption was low in the period of increasing yields. Second, an unexamined Figure 1 shows yields falling in recent years despite virtually universal adoption of Bt technology by farmers (Stone 2012: 68). Let us examine these two claims, beginning with Stone’s adoption data. The data are systematically inaccurate in purporting to give us the national area under Bt cotton yearly, and thus the extent of diffusion of the technology. Measuring diffusion of cultivars⁵ is easy if there is a single source or effective control. Bt cotton violated these conditions egregiously. It was in fields illicitly for three years before the Government of India discovered it in 2001, even though none could be grown legally without government approval (Scoones 2006, Jayaraman 2001). Once the ‘desi Bt’ of Navbharat Seeds was banned in 2001 by the Genetic Engineering Approval Committee in Delhi, ‘Bt variants’ diffused underground from growers engaged by Navbharat to produce hybrids containing the Bt transgene (in this case, the cry1Ac, the same gene used in Bt brinjal). As farmers found the Bt seeds successful, underground diffusion accelerated – best documented in Gujarat (Roy 2006). How do we then know the area under Bt seeds? Stone confidently presents International Service for the Acquisition of Agri-biotech Applications (ISAAA) data to show area under Bt cotton nationally in his Figure 1. ISAAA quite explicitly recognises the ubiquitous illegal diffusion of biotech seeds in various countries, including India, but their figures exclude uncountable stealth seeds

Figure 1. Cotton yields and percentage of fields planted with Bt seed

Source: Yield data are from the Cotton Corporation of India; Bt adoption data are from ISAAA, Reproduced from Stone (2012).

Stealth seeds seriously complicate analysis of India, Brazil, China, Argentina, Ukraine, Thailand, Vietnam, Pakistan and other countries – the area under illegal biotech seeds is unrecorded (Herring 2007, Herring and Kandlikar 2009). This was especially true in India before the big price drop in Bt cotton hybrids in 2006, when the price advantage of illicit/desi seeds was reduced significantly⁶. Three Bt hybrids became legal in March 2002. Jayaraman (2004) cited “industry sources” as estimating that more than half the transgenic cotton in India came from illegal hybrids in 2004. Gujarati seed producers and farmers suggested to me a much higher figure for that state. Data from the Gujarat State Seeds Producers’ Association indicated that about 34% of the cotton-seed packets sold nationally were transgenic in the 2004-05 growing season. Of this 34%, less than a third were legal, more than two-thirds illegal⁷. The Press Trust of India reported on 10 February 2004 that “an illegal variety of Bt cotton covers nearly 90% of the cotton area in Gujarat”. Questions in Parliament revealed widespread knowledge of – and concern about – diffusion of illegal Bt but no conclusive data on extent. One especially knowledgeable observer of the cotton sector estimated 2.5 million hectares under stealth Bt seeds in 2005, that is, almost twice the area in official statistics. All estimates point in the same direction – rapid expansion of Bt acreage below the radar of official statistics and uncounted in ISAAA data (Herring 2005).⁸

Adding stealth Bt seed acreage to official acreage would increase the area under Bt cotton in Stone’s Figure 1 and thus move the curve up and address part of his puzzle. Use of Bt cotton was much more extensive, and spreading much faster, than Stone’s figure suggests. Data on illicit seeds are inherently problematic, but are we on firmer ground with yield data? Stone rejects the Ministry of Agriculture data used by Herring and Rao (2012: 51) in favour of data from the Cotton Corporation of India. They give quite different views – opposite yield trends in recent years. Which facts are factual? It may not matter too much for two reasons. First, no national agricultural statistics for India are rock solid in any event; short time series on crop production are problematic generally. Second, yields themselves are not the only or best indicator of technology success, especially for a technology that brings only one trait.

Farmers’ income important, not yields alone

The more important question for farmers is income, not yield. Here the Bt effect in reducing the cost of production of a marketed unit of cotton is important. An insect-resistance trait may also prevent total loss of crop – as happened in Gujarat in 20019 – but prove less necessary in normal years. Kathage and Qaim (2012) find in their longitudinal study that yield increases were 24% but income increases were 50% for Bt cotton farmers. Reduction of pesticide costs is a big part of the difference. The Bt trait will vary in effect on yields as pest pressure varies, from very large effects to low effects. That effects vary tells us nothing about the success of the technology once we understand the purpose of the trait and the inherent variability of crops in fields.

Could Bt cause falling yields in cotton?

There is much media attention to the ‘failure of Bt cotton’ and associated farmer suicides. If there is any local or short-term decline in yields, is there any way Bt could be responsible? It is possible that cotton’s very profitability is bringing more marginal lands into production, lowering average yields. Between 2009 and 2012, an additional 3 million hectares came under cotton cultivation, most of it Bt. It is hard to imagine that this land stayed out of cotton if it were prime soil with water; at least some expansion of cotton must be on less productive land. Moreover, with the boom in Bt cotton, counterfeit seeds multiplied; we do not know how much Bt land is actually planted with authentic Bt seeds (Herring and Kandlikar 2009). More broadly, yields will be affected by changes in pest pressure, drought, and perhaps, regionally, as yet unknown effects of climate change.

Farmers’ behaviour indicates economic value of Bt cotton

Whatever the discrepancies between ministry and corporation data on national yields, or yearly variations, farmers’ behaviour reinforces the conclusion that Bt provides cotton with a trait of economic value. The bulk of empirical work accords with farmers’ adoption logic. Some critics claim that adoption data mean nothing because – there is no choice: non-Bt cotton seeds have disappeared from the market. This is not true. Non-Bt cotton seeds are provided with Bt cotton packets for planting refugia,10 though farmers often throw them away. We also know from seed producers such as Nuziveedu that fewer non-Bt hybrid seeds were produced because they lost favour with farmers and eventually became unsaleable (Herring 2008). Moreover, since 2006 farmers have had the choice to buy even more expensive second generation stacked-gene Bt seeds – but could stay with the first generation one-gene Bt, or buy even cheaper stealth seeds. What have they done? Over 80% of the Bt cotton seed sold in 2011 was the more expensive two-gene Bt seeds, even though the less expensive original version is still available.

Concluding facts

Stone’s initial objective was to destabilise the broad consensus in peer-reviewed literature that he characterises as a “triumph narrative”. Yet his mechanisms are unproven and implausible – the conspiracy is too grand, the actors too diverse. He himself contradicts this initial claim with a conclusion consonant with the peer-reviewed literature he attacks. In assessing the “isolated impact” of Bt’s effects on cotton yields, he concludes, “Kathage and Qaim’s (2012) multivillage fixed-effects study gives us the number 24%; Stone’s (2011) multi-village before-and-after analysis gives us 18%; and Gruère and Sun’s (2012) trend analysis gives us 19%” (2012: 68). That is a tight range and impressive level of measured contributions to yield increases, especially for an agricultural innovation with a single trait; it is also consistent with the behaviour of farmers.

The radical scepticism of postmodern constructivist facticity is irrelevant to the Bt cotton question. Once we have these facts straight, it is clear that Bt cotton represents neither suicide seeds nor silver bullets, but a remarkably valuable technology. Stone’s article does not destabilise the broad consensus on its usefulness to farmers. Nor are doubts about the facts on Bt cotton sufficiently compelling or grounded to undermine further research and development in agricultural biotechnology, despite their political deployment for just this purpose.

A version of this column has appeared in Economic and Political Weekly.

Notes:

Image Source: ‘By Claude Renault (Crop applied by uploader) (http://www.flickr.com/photos/clodreno/128382985/) [CC-BY-2.0 (http://creativecommons.org/licenses/by/2.0)], via Wikimedia Commons’.

Monsanto is an American multinational chemical and agricultural biotechnology corporation.
Biotic stress is the negative impact of living organisms such as bacteria, parasites etc. on living organisms such as plants. Abiotic stress refers to the negative impact of non-living factors such as intense heat or drought that may cause harm to living organisms such as plants.
Germplasm is living tissue from which new plants can be grown. Germplasm contains the information for a species’ genetic makeup - a valuable natural resource of plant diversity.
Higher yields on Bt farms could reflect the fact that farmers who adopt new technology are advantaged in ways independent of the seeds - better educated, with better resources, credit, social standing, connections and so on.
Cultivar refers to a variety of a plant that was produced for and is maintained by cultivation.
The reason for the more rapid adoption of illegal hybrids was primarily price – as little as 30% of the cost of legal hybrids. The prices of official seeds dropped dramatically in 2006, after which the stealth market contracted but did not disappear (Roy 2006, Roy et al. 2007, Ramaswami, Pray and Lalitha 2011.
Mimeo tables and personal communication, October 2005; and conversations with D B Desai, Navbharat Seeds, June 2005.
Estimates of unofficial Bt diffusion are imprecise, but conservative in that they apply only to packaged and branded stealth seeds; farmers also used ‘loose seeds’ straight from the gin, produced by Bt hybrids, at very low cost.
It was the ‘bollworm rampage’ of 2001 that allowed discovery of illegal Bt seeds – all the regular cotton hybrids were devastated, but Bt fields stood undamaged. See Jayarama 2001, Herring 2005, Scoones 2006.
Best practice in use of any insecticide is to avoid insects developing resistance, as has happened with many synthetic chemical pesticides. With insecticidal proteins such as Bt, the idea is to plant non-Bt plants in every field so as to lower selection pressure for resistant.