Understanding how Bt Cotton ‘deskilled’ farmers in India
The adoption of Bt cotton in India has led to agricultural deskilling, and there is evidence of over-reliance on social learning rather than careful trial and study of new seeds and practices. This is the central message of a startling new study carried out in the state of Andhra Pradesh, in the Warangal district. The study was done by Glenn Davis Stone, an anthropologist at Washington University in St Louis, USA, and published by the journal World Development.
Field-level studies of Bt cotton in India now number in the dozens. The clear majority of studies by economists do reveal advantages in cotton yield, and often in pesticide usage, for Bt cotton, but there are several reasons for agreeing that the results to date are inconclusive. One issue is that measures of central tendency obscure the enormous variability across time and space. Consider the major cotton-producing states: yields in Gujarat have surged from below the national average before Bt cotton to leading the country by 2005, while yields in Madhya Pradesh have decreased since Bt arrived.
Within sub-state units such as the district or mandal, villages vary greatly in prosperity, access to information, and other factors affecting use of new technologies, which may help explain cases like Maharashtra where studies show a “complex, confusing picture of farmers’ spraying behaviour and a startling degree of variability in their cotton output”, according to one earlier study. It is doubtful that there is any such thing as a typical cotton growing village in India, said another. [SciDev.net has a report on the study and its findings.]
Stone has said that another persistent problem has been selection bias. Early adopters are known to be a sample biased towards successful farmers. Bt-adopters have been found on average to own 58% more land and 75% more non-land assets; to own up to 36% more land; to be not only richer in land, but better educated and more diversified. Bt-adopters have also been found to be more effective farmers by comparing the non-Bt yields of adopters (i.e. farmers who planted both types) with the yields of non-adopters; the adopters’ conventional yields were found to have produced 29–43% more than the other conventional yields.
Research to date has very rarely controlled for this bias, and many studies fail to even specify how their samples were drawn. The problem is key because almost all studies have focused on the years immediately following the introduction of Bt cotton, when yield differences mainly reject the agricultural prowess of a biased group of early adopters (and also reject how this group happened to fare their first time trying a new technology).
A related problem is bias in cultivation practices: prior to the institution of price caps in some states in 2006, Bt seeds cost four times as much as conventional seeds, and would have been planted in the yields with best irrigation and then benefited from unusual care and expense. This accords with the fact that adopters spent more on bollworm sprays for their Bt plots than for their conventional plots. “In Warangal I have seen many cases of farmers lavishing extra resources and attention on their Bt yields,” wrote Stone in his paper.
“The 2007 season marked the first time virtually all farms in the sample planted exclusively Bt cotton. In 2007, most input shops stocked little if any non-Bt cotton seed, and no farmers in the sample reported with confidence that they had planted any non-Bt seed in 2007. In some cases farmers said they were not sure if they had bought Bt seed or not; farmers often buy seeds that others are buying without knowing much about them. Therefore it is impossible to specify how many packs of non-Bt seed were bought, but we can be certain that the number is vanishingly small. By 2008, I believe the number to be zero: all of the eight input shops I interviewed in Warangal City and four villages had only Bt cotton, and no vendors or farmers knew where one could find a box of non-Bt seed. Most people had stopped even identifying Bt cotton as such.”
[The formal citation: Stone, G. D. Field versus Farm in Warangal: Bt Cotton, Higher Yields, and Larger Questions, World Development (2010). Paper available here.]
From a farm-level perspective there appears to have been a general management failure of which the bollworm damage was merely a symptom. Such management failure has been theorised as “agricultural deskilling” which may be synopsised as follows:
* Farm management skill (in non-industrial contexts) is based not on static “indigenous technical knowledge” but on the ability to “perform”. It is not static, but rather an ability that must be continually updated and refined, especially when there are changes in market conditions, input technologies, pests and diseases, government policies, and even new ideas. This ongoing process of learning to perform with given technologies under changing conditions is agricultural skilling.
* How skilling actually occurs is complex. Drawing on work by behavioral ecologists, it is helpful to distinguish between environmental learning, which is based on evaluations of payoffs from various practices, and social learning, in which adoption decisions are based on imitation.
* Social learning is an indispensable part of human adaptation but it has intrinsic biases. One is prestige bias, in which a farmer chooses which farmer to emulate on the basis of prestige, regardless of the other farmer’s actual success with the trait being copied. Another is conformist bias, in which a farmer adopts a practice when (and because) it has been adopted by many others. Reliance on “pure social learning” should be high when environmental learning is costly and/or inaccurate. Social learning may lead to the spread of maladaptive beliefs, especially when the environment changes very rapidly.
* Failure of the ongoing process of learning to perform under changing conditions is agricultural deskilling, a condition differing in some key respects from the better-known industrial deskilling.
Specific causes of deskilling in Warangal cotton farming were identified as inconsistency, unrecognisability, and an excessively rapid rate of change in cotton seed. Patterns of seed choice gave conspicuous evidence for deskilling. Although choice of seed is one of the most serious decisions the farmer makes each year, farmers in all study villages relied heavily on “pure social learning,” producing a surprising pattern of highly localised seed fads, driven not by local agroecology but by marketing and happenstance. In counterpoint to the classic model of farmers adopting new seed only after careful evaluation of test plots, Warangal farmers showed a keen desire for new and untested seeds, which encouraged the churning of the seed market with new releases (including releasing seeds under multiple names).
In his discussion, Stone has said: “We have, on one hand, a global constituency that contests the spread of agricultural biotechnology on mostly political-economic grounds including effects on intellectual property regimes, funding priorities, and other articulations between the industrialised and developing worlds. On the other hand, we can recognise nexuses of corporate biotechnology, academic science, and state trade interests with a keen interest in developing-world success stories. There is much at stake, and the claim that transgenic technologies are ‘just another tool for the farmer’ is true only in the studiously myopic sense that the textile mills in England’s Industrial Revolution were ‘just another tool’ for making cloth. But the debate has followed a trajectory with enormous emphasis on empirical field-level measurements, and given the pervasive vested interests and strong antipathies, claims of resounding field-level ‘success’ or ‘failure’ have found ready audiences.”