Here we review some economic studies on the benefits of Bt cotton adoption, the biological and ecological underpinnings of the Indian cotton system, and the question of whether the Bt cotton technology was needed, under what conditions might it give economic benefits, is its adoption linked to farmer suicide, who might profit from Bt cotton adoption, and lastly how climate change will impact rainfed cotton production.
Review of the economics of Bt cotton adoption
Numerous economic studies based on field experiments and survey data have found economic benefits for Bt cotton adoption in India [38], and yet, controversy persists (see the cogent expośe by Stone [39]). Econometric analyses ignore the underpinning ecology of the system and disregard underlying agro-ecological principles of yield formation (see Additional file 1) [40]. In particular, prior economic studies of Indian cotton compared the failed insecticide technology to the Bt technology option, ignoring the question of whether either technology was needed in the first place. Econometric analyses tell little about the root causes of the problem being evaluated or alternatives to the current production system and, most important, provide little insight into what is foremost an ecological problem with economics superimposed [16, 41]. Specifically, studies in ecologically disturbed environments limited to isolated small plots typical of published results from India, rather than in larger landscape and historical frameworks, are known to bias results against untreated checks (see [34, 42–44]), inputs such as fertilizer and water are often not experimentally controlled [45], in south-central India, ground water is being used unregulated and unpriced [26], industry data have been used to predict unrealistic estimates of potential yield [20], and generally, important agronomic aspects of the systems (e.g., irrigated vs. rainfed, planting density, varieties, pest dynamics, etc.) and the crucial effects of weather are ignored. For example, higher yields were found for Bt compared to conventional cotton [46], but in critique of this and related studies, a high degree of variation in productivity and profits with other social and economic explanations was found for yield differences including a “placement bias” of irrigation and “good growing conditions” [22]. Insecticides continue to be used in Bt cotton in India, and as in China productivity effects of Bt cotton and pesticide use likely depend on the action of natural control agents with the profitability of damage control measures increasing with the severity of ecosystem disruption due to insecticides [47].
Biological and ecological underpinnings
Economic analyses commonly assume that cotton pests must be controlled to prevent economic losses. However, in an 11-year study of rainfed cotton at Nagpur, MH, India (near Yavatmal) using the non-Bt G. hirsutum hybrid NHH-44 under organic and conventional practices, Blaise [5] found higher yields and lower pest damage in organic cotton. Fundamental to understanding this result is that annual emergence of the key pest PBW in spring is poorly timed to attack rainfed cotton and absent high inoculum from irrigated cotton (Fig. 5), large PBW populations fail to develop in non-Bt rainfed cotton. This biology reduces and usually obviates the need for Bt cotton and disruptive insecticides (see Figs. 3 and 5), thus avoiding ecological disruption and outbreaks of bollworm and other secondary pests (sensu [11]) that may be far more damaging and difficult to control than PBW [8, 9]. Prior to 1970, bollworm was not an important pest in Indian cotton [48].
That insecticide can induce pest outbreaks in Indian cotton has abundant parallels worldwide [11, 12, 34, 47, 49–51]. A well-documented example is irrigated industrial cotton in the Central Valley of California during the 1960s and 1970s. While PBW is limited there by winter temperatures [52], insecticide was used to control a presumed pest, the plant bug Lygus hesperus Knight. The insecticide induced severe outbreaks of bollworm, defoliators, and the resurgence of Lygus [34, 51, 53–55] (see Additional file 1), and the studies over several years showed that compared to the insecticide treatments, higher yields of the same quality accrued in the very large untreated check areas [34, 51]. Central Valley farmers had been spending money on insecticide to lose money on lower yields causing economist U. Regev [56] to call this the first documented case of market failure in pest control [6]. Bt cotton has not made inroads in the Central Valley of California though herbicide tolerant cotton is grown [52].
More germane to India is irrigated cotton in the desert valleys of southern California and Arizona where the invasive PBW caused heavy losses during the mid-1970s to the mid-1980s, and insecticide use for its control caused severe outbreaks of bollworms, budworms, defoliators, and whiteflies [11, 27, 57]. The problem was initially solved using high-density short-season cotton that produced high yields, but the technology requires early crop termination and plowing of the stubble to destroy residual dormant PBW populations (see Additional file 1) [32]. Short-season cotton was replaced by fertile Bt cotton despite no increases in yield because implementation requirements were less stringent, it gives excellent control of PBW [58], and in an industrial setting, the costs of the Bt technology are an acceptable cost-effective alternative to short-season cotton (see Additional file 1). Bt cotton is “softer” on natural enemies than insecticides [59, 60] enabling reductions in insecticide use [61] that allows secondary pests (e.g., bollworms, budworms, whiteflies) to recede to prior low pest status [33]. Though not significantly different, mean natural enemy densities in Bt cotton are consistently lower than those in unsprayed non-Bt cotton [62], and the efficacy of some natural enemies is reduced when feeding on Bt-intoxicated prey [63, 64]. For pests such as bollworms and defoliators having high reproductive capacities (500–1000 eggs/female/week) or pests with high tolerance to Bt toxins (e.g., plant bugs, whiteflies, and mealybugs), a small reduction in natural enemy density and efficacy may increase pest density and trigger insecticide use and ecological disruption [33]. Plant bugs have increased in Bt cotton in China and the USA, but this has been dismissed as due to reduced pesticide use [58, 65–67], despite strong evidence that insecticide use increases plant bug resurgence [33, 55] (see Additional file 1). In India and Pakistan, sucking insect pests had been of minor concern in cotton but are now increasing in Bt cotton and contribute to yield losses and increased insecticide use [68, 69].
In sharp contrast to industrial cotton farms globally, most Indian cotton farms in south-central India are <1 ha and are rainfed. Furthermore, F1 Bt hybrids are sold as a value-capture mechanism to discourage seed saving by millions of small farmers who cannot be controlled by threats of lawsuits as occurs in industrial agriculture in more developed areas. Single toxin hybrids may produce 25 % non-Bt seeds and 6.25 % in two toxin hybrids with the seasonal expression of Cry2Ab having a wide range with the levels being tenfold higher than for Cry1Ac [7]. Quality control of Bt seed in India is lax [2]; resistance to Bt occurs in PBW [70–72] and in bollworm [2, 73, 74]; and as elsewhere susceptibility to Bt toxins varies greatly among pests (Additional file 1) with insecticide resistance in Bt tolerant pests further complicating pest control [9, 14]. Insecticide use, while initially lower in Bt cotton, has greatly increased against sucking pests [69, 75] (Additional file 1: Table S1)
The need for Bt cotton in India must be reevaluated on biological and economic grounds [69] using properly unbiased field experiment unfettered by onerous corporate intellectual property constraints. These constraints were outlined in a 2009 letter to the US Environmental Protection Agency from 26 leading university entomologists from the US corn-belt protesting the restricted access to GMO seeds for experimental purposes: “…No truly independent research can be legally conducted on many critical questions involving these [GMO] crops” [76]. Corporate intellectual property constraints on GMO crops are an impediment worldwide to system-level analysis.
Does Bt cotton provide economic advantage?
Average profits per ha in rainfed cotton are computed as revenues from the sale of seed cotton minus average costs of seed, insecticide, and other production costs. On the revenue side, during 2002 to 2013 cotton prices (e.g., Bengal Desi) ranged from Rs 1875 to 4583/100 kg of seed cotton (i.e., ~$0.31–$0.75/kg assuming Rs 61 per $) [37]. A midrange value of $0.51/kg is used in our calculations. Prior to the advent of hybrid varieties, seed costs were nil to low (Rs 8–9/kg), but as fertile local varieties became unavailable, farmers increasingly bought F1 hybrid seed that for Bt varieties cost ~Rs 2111 per kg. Two kg of seed are required at traditional planting densities of two plants/m2 (~$69 ha−1), but optimal planting densities may be threefold or much higher, depending on the variety (see Additional file 1). At the traditional density, the cost of seed plus insecticide was Rs 5500 ha−1 (~$90 ha−1) for both Bt adopters and non-adopters [46] with an additional Rs 5386 for other production costs [43]. These prices vary in time and region, and hence, they must be viewed as reference values.
Observed average yields for rainfed cotton in the four target states ranged from 300 kg ha−1 to 1200 kg ha−1 lint cotton with yields in a few districts of AP and KA exceeding 1200 kg ha−1 [35]. Note that lint cotton is roughly 30–35 % by weight of seed cotton. In the absence of pest damage, average simulated rainfed yields for AP, GJ, KA, and MH accord with these values (Fig. 8a) with roughly half of the area producing less than 500-kg lint cotton ha−1 (Fig. 8b). (Yields in irrigated cotton would be at or above the upper end of this range.) Total costs of production for the small proportion of rainfed farms with yields of 1320-kg cotton ha−1 are ~8 % of the revenues. In contrast, production costs for farms with yields of 500 kg ha−1 are ~21.1 % of the total revenues resulting in a net income <$2 day−1 ha−1 farmed. At 250 kg ha−1, costs consume ~42.2 % of the revenues resulting in a net income of <$1 day−1 ha−1 (see Fig. 8a). Costs as a proportion of revenues decrease exponentially with increasing yields (9200yield− 0.965) becoming 100 % at ~78.5-kg lint cotton ha−1.
In rainfed areas, low yields and high variability are substantial sources of risk [22, 77] (see Figs. 4 and 7), with the high costs of Bt cotton seed and continued use of insecticide being added destabilizing factors. Debt has long been a dominating factor in Indian agriculture, and recently, official sources of credit have greatly decreased, and usury costs (5–10 % per month) to money lenders and others sources have become an added burden for poor farmers seeking to fund new technology adoption [23].
Did Bt cotton increase yield?
Bt cotton is not a yield enhancing technology, rather, it is designed to protect the yield potential of the variety that carries the trait from damage from some but not all lepidopteran pests (see above and Additional file 1). Average yields for India during 1975–2007 are illustrated in Fig. 9a with the apparent increase in the national average after 2004 attributed to Bt cotton adoption [19], despite the fact that the adoption rate was only 8 % in 2005 and 42 % in 2006 [78], that government subsidies for fertilizer (primarily urea) during 2003–2011 increased approximately fivefold (i.e., 110 to 600 billion Rs or about US$2 to 10 billion) [79], that data from irrigated and rainfed cotton were conflated in the average, and that agronomic practices and varieties were improving (see below). Variety improvements, fertilizer, rainfall amount and time, reduced pesticide use, and changes in planting densities can have large effects independent of the Bt technology. The large effects of genotype × spacing in rainfed cotton were demonstrated using fertile non-Bt G. hirsutum and Desi (G. arboreum) varieties with one variety yielding 1967 kg ha−1 of lint cotton at 16.6 plants m−2 that was >60 % higher than that at 5.5 plants m−2 [80].
The post-2004 yield data appear to be on the same increasing trend (dashed line in Fig. 9a) as before the introduction of Bt cotton when improved hybrid varieties began entering the market. We posit that the stippled area below the dashed line is a rough estimate of yield loss commonly observed in ecologically disrupted cotton systems with increasing insecticide resistance (e.g., California). Cotton in MH is predominantly rainfed, and yields before 2002 were increasing, but the level was well below the national average (Fig. 9b), while yield gains in AP were flat (not shown). Yield stagnation occurred nationally during the period 2005 to 2013 at about ~510-kg lint cotton ha−1 [2, 75] (see Additional file 1: Table S1). Yields in GJ and AP peaked before 2007 and then declined sharply, while yields in MH and KA continued to increase (Fig. 9c).
Are farmer suicides linked to Bt cotton adoption?
The reason for individual suicides is varied, and in India, it must be viewed against the webbed nuance of social (e.g., caste, religious and cultural), ecological, and economic factors of Indian agricultural society (see [23]).
A tenfold greater age-standardized suicide death rate occurs in the southern states of India compared to northern states [81]. Economists examined the national suicide data for 1997–2007 and concluded that there was no link of farmer suicides to Bt cotton adoption [19]. Indeed, plots of annual suicides in GJ and KA show no trend with time (Fig. 10a) or on the national total of suicides. Suicides in GJ were about 500 per year (<3 % of the national total), while cotton production was 24–39 % of the national total [37], and average yields were mostly >500-kg lint cotton ha−1 with CV <50 % (Fig. 7b). Farmer suicide rates in KA were high at about 2000 annually, but cotton production was only 2–4 % of the national total with predicted yields <500 kg and CV >50 % being very common (Fig. 7c). In KA, these are indicators of risk not only in cotton but also in other parts of the agriculture sector due to low and highly variable rainfall. In AP and MH, suicides are strongly increasing with time (Fig. 10b) and on the national total (Fig. 10c) with the increase beginning before the introduction of Bt cotton in 2002. The contributions to national cotton production in MH (15–26 %) and AP (13–19 %) are similar, and high-risk areas (i.e., yield <500 kg, CV >50 %) are common in both states (see Fig. 7a, d). In MH, 90 % of the farmers grow some rainfed cotton, and the state is a hot spot for suicides [23]. Regressing combined annual suicides in AP, GJ, KA, and MH (y) on the national total (x) yields an increasing relationship (y = 0.90x − 7461.3, R
2 = 0.823).
Revisiting the raw annual suicide data for AP, GJ, KA, and MH during the period 2001–2010 [82], 86,607 of 549,414 suicides were by farmers, and 87 % were males with the numbers peaking in the 30–44 age class (Additional file 1). Details concerning cotton cultivation, farm size, and other important factors were not reported for each suicide, and hence, as in [19], we were forced to aggregate the data by state. Total suicides per year per state were regressed singly on states averages of proportion of area seeded to rainfed cotton (t = 15.64), average farm size (t = −8.90), cotton growing area (×103 ha, t = 1.90), area of Bt cotton (×103 ha, t = 1.69), proportion of area with Bt cotton (t = 1.13), and simulated average yield/ha (t = −0.47) that includes the effects of weather (see “Methods”; Additional file 1).
Excluding the proportion of area seeded to rainfed cotton, linear multiple regression without transformation of the data (Eq. 3) shows suicides decrease with increasing farm size and yield but increase with the area under Bt cotton cultivation (i.e., Btarea is ha ×103). Farm size and yield are measures of poverty and risk, while the increase in Bt area is a surrogate for high costs of Bt technology adoption and continued use of insecticide.
$$ \begin{array}{l}\mathrm{suicides}=14,056\kern0.5em -6933.0{\mathrm{farm}}_{\mathrm{size}}-2.067\mathrm{yield}+0.3029\kern0.5em {\mathrm{Bt}}_{\mathrm{area}}\\ {}\\ {}\kern2.25em {R}^2=0.76,\ df=36,\kern0.75em F=38.66,\end{array} $$
(3)
The means of the independent variables and t values for the regression coefficients are farmsize(1.568 ha, t = − 9.93), yield(603.8 kg/ha, t = − 3.25), Bt
area
(743.4 ha, t = 2.84).
We note that the high variability associated with low yield is a further dimension of risk (Fig. 7).
Factors affecting suicide rates
Economic distress can be a proximal cause of suicide, and at least seven factors appear to have influenced this in rainfed areas where cotton is a cash crop: (1) weather-related intrinsic low average yields and high variability [23] (Figs. 7 and 8); (2) increasing insecticide use before 2002 that increased costs and yield losses due to ecological disruption by induced pests (see text; Fig. 9); (3) high costs of Bt cotton seed, fertilizers, insecticide, and ecological disruption and crop loss after the introduction of Bt cotton (e.g., Fig. 8); (4) crop losses due to ill adapted and possibly ineffective Bt varieties [7, 19]; (5) increased usury costs to fund the new technologies; (6) suboptimal planting densities [80]; and (7) the uncertain effects of weather (e.g., drought or excessive rain as occurred in 2013) on pest and yield. Factors 2–5 are industry driven, and increase bankruptcy rates as farmers assume the gamble in the monsoon [23]. As in China, risk-averse subsistence Indian farmers likely use greater quantities of pesticides that do not increase yield potential but may increase ecological disruption and risk of crop failure [83]. Sound information to enable cotton farmers to make informed decisions about plant densities, varieties, pest control practices, and whether to plant Bt cotton is largely unavailable [84]. This information gap was exploited by the insecticide industry before the introduction of Bt cotton and by the seed and chemical industry since. Technology promoters appear more interested in instrumentalizing the technologies for profit than in explaining the underlying causes of yield variation or to find satisfactory cheaper alternative solutions (e.g., high-density short-season cotton). Promoters willingly offer apparent solution with little regard for consequences. Instrumentalization of insecticides and Bt technology and their adoption by small stakeholders in India has hindered progress on ecologically based cotton production [85] and has contributed to human suffering. Economic analyses decoupled from the ecological and historical roots of the production problem confound the situation, and making the claim that the market determines the usefulness of Bt cotton and insecticide in India (and elsewhere) shameful (sensu [56]).
Does Bt cotton have a role in Indian cotton?
Absent resistance to Bt toxin [71], quality Bt cotton provides excellent control of PBW [86] and good control of bollworm and would appear to have a role in high-yielding irrigated long-season cotton in India where PBW is a chronic problem and where the marginal cost of the technology is low. Proponents of Bt cotton claim that it reduces the input of disruptive insecticides and greatly reduces ecological disruption, but ultimately, this has not proven to be the case as year 2013 insecticide use is at year 2000 levels [69] (Additional file 1: Table S1). Most cotton in India is rainfed, and Bt cotton provides only prophylactic protection against late season spillover of pink bollworm from irrigated fields (see Figs. 1, 5) and against outbreaks of bollworm and Bt susceptible pests caused by area-wide insecticide disruption. Bt cotton is only a partial solution because it is ineffective against many cotton pests and may help induce outbreaks of others (e.g., sucking pests such as aphids, mealy bugs, jassids, and plant bugs) and result in increased insecticide use [72]. More importantly, the cost of Bt cotton as a proportion of the total revenues to small farmers in rainfed areas can be very high (Fig. 8), and effective implementation of the technology requires among other things such as quality control of varieties, optimal planting density, resistance management to preserve susceptibility to Bt [87], and avoidance of insecticide use. These factors are major constraints in the predominant rainfed small farm cotton culture of south-central India.
The propensity of Indian cotton farmers to extend the season to harvest late-season bolls is counter-productive as PBW and bollworm infestations can then develop and increase intra-season carryover of pests. Short-season cottons can be high yielding and are designed to set and mature bolls quickly, lessening the need for insecticide for late-season pink bollworm (and bollworm) infestations. Customizing the technology developed with high-density short-season non-Bt cotton for irrigated cotton in southern California [32] (Additional file 1) appears to be a viable solution for irrigated cotton in India [80]. Variations of this technology would also be applicable to rainfed cotton in central and south India where rainfall is low and variable and in areas of south India where a portion of the PBW population does not enter diapause [88]. Furthermore, physiologically based demographic models (see “Methods”) as used here could provide a rapid way of evaluating short-season cotton varieties and control strategies for cotton pests in the different regions of India and elsewhere (e.g., [89]).
Climate change effects on rainfed cotton
Climate change will affect cropping systems globally, and in India, the generalized physiologically based plant simulator INFOCROP was used to simulate the effects of climate change increases in [CO2] and temperature on cotton [90]. Global climate model (GCM) temperature scenarios A2, B2, and A1B that respectively project 3.95, 3.20, and 1.85 °C increases in mean temperatures with marginal increases in rainfall were used in the study. They found that productivity in northern India may decline marginally while in rainfed cotton in central and southern India productivity may either remain the same or increase. Our results for rainfed cotton in central and southern India, assuming a 2 °C rise in temperature and no change in rainfall agree with their results. Our model predicts yield increases in most areas of less than 8 % (Fig. 11a, b) with yields decreasing in very low rainfall areas (Fig. 11c). Cotton tolerates high temperatures, and predicted yield increases are due to increases in temperatures during the monsoon season that effectively increase season length (Fig. 3a, b). The results do not change the conclusions about the relative risks of growing rainfed cotton in south and central India, but other high-temperature intolerant crops could be adversely impacted.