Article

Asiatic cotton can generate similar economic benefits to Bt cotton under rainfed conditions

  • Nature Plants 1, Article number: 15072 (2015)
  • doi:10.1038/nplants.2015.72
  • Download Citation
Received:
Accepted:
Published online:

Abstract

American cotton (Gossypium hirsutum L.), transformed with Bacillus thuringiensis Cry genes (Bt G. hirsutum) that confer resistance to lepidopteran pests, is extensively cultivated worldwide. In India, transgenic Bt G. hirsutum was commercially released in 2002 and by 2014 95% of farmers had adopted Bt G. hirsutum1. The economic benefits of Bt G. hirsutum over non-Bt G. hirsutum are well documented and include increase in yields, increase in farmers' net revenue and reduction in pesticide application against lepidopteran pests2,​3,​4,​5,​6,​7,​8,​9. However, it is unclear to what extent irrigation influences the performance of Bt G. hirsutum on smallholder farming in India, and if, in the absence of irrigation, growing Bt G. hirsutum provides greater economic benefits for Indian smallholder farmers compared with growing the Asiatic cotton Gossypium arboreum L. Here, we compare the economic impact of growing Bt G. hirsutum with growing G. arboreum under rainfed conditions in the Indian state of Maharashtra, and show that G. arboreum can generate similar net revenue, and thus similar economic benefits for smallholder farmers compared with growing Bt G. hirsutum. We also compare the economic impact of growing Bt G. hirsutum under rainfed conditions with growing Bt G. hirsutum under irrigated conditions and show that even though Bt G. hirsutum yields increase with irrigation, the net revenue does not significantly increase because farmers using irrigation spend significantly more than farmers growing Bt G. hirsutum without irrigation. We conclude that our data provide a broader insight into how socio-economic data needs to be incorporated into agro-ecological data when planning strategies to improve cotton farming in India.

The Asiatic cotton G. arboreum L. historically occupied the majority of the cotton-growing area in India, but its cultivation was drastically reduced first with the introduction of the American cotton G. hirsutum L. in the early 1980s, and then reduced further with the introduction of the transgenic Bt G. hirsutum in 200210. Both G. hirsutum and the transgenic Bt G. hirsutum tend to have higher yields than G. arboreum10,11, but G. arboreum harbours a suite of valuable agronomic traits that could potentially benefit farmers, such as tolerance to a variety of biotic and abiotic stresses, including drought11,​12,​13. Although many studies have compared the performance of Bt G. hirsutum to non-Bt G. hirsutum2,​3,​4,​5,​6,​7,​8,​9, it remains to be quantified the extent to which growing the widespread Bt G. hirsutum impacts the economy of farmers compared with growing G. arboreum. Given that the majority of cotton is grown under rainfed conditions in Maharashtra14, and predictions indicate that climate change and groundwater depletion will limit access to irrigation in the coming decades15,​16,​17, here we compare the economic impact of growing Bt G. hirsutum with growing G. arboreum under rainfed conditions in the Indian state of Maharashtra. In the analysis, we also include farmers growing Bt G. hirsutum under irrigated conditions to determine the influence of irrigation on the field performance of Bt G. hirsutum. Using a stratified sampling design, 51 smallholder farmers (possessing less than 12 acres of landholding) were comprehensively interviewed and their farms investigated (see Methods section for detailed information on the study design). With this sample size we are able to detect large differences between the mean values of the variables considered (further information on the power of our test can be found in the Supplementary Information).

To compare the economic impact of growing Bt G. hirsutum with growing G. arboreum under rainfed conditions, we first compared how much money smallholder farmers spent on cultivating each variety. Farmers cultivating Bt G. hirsutum spent more money on seeds, fertilizers and harvesting than farmers cultivating G. arboreum (Table 1). Farmers growing Bt G. hirsutum also spent more money on pesticides (Wilcoxon rank-sum test; P < 0.001). This may be because while Bt G. hirsutum is largely resistant to lepidopteran pests, it is still sensitive to non-lepidopteran pests for which chemical pesticides are needed18,​19,​20. In contrast, G. arboreum is resistant to several lepidopteran and non-lepidopteran pests including bollworms, spider mites, leaf curl virus, aphids and leafhoppers21,​22,​23,​24,​25.

Table 1: Expenditure, yield, revenue and net revenue for each cotton practice.

We found that farmers cultivating Bt G. hirsutum experienced marginally higher yields than farmers cultivating G. arboreum (Student's t-test; P = 0.06), which did not translate into significantly higher revenue (Student's t-test; P = 0.14). These findings demonstrate that, although Bt G. hirsutum may often deliver higher yields than G. arboreum10,11, farmers growing G. arboreum in Maharashtra can achieve revenues comparable to those of Bt G. hirsutum under rainfed conditions.

Farmers growing G. arboreum benefitted from a higher market price for their product than farmers growing Bt G. hirsutum (Wilcoxon rank-sum test; P < 0.001). This suggests that although G. arboreum fibres are less amenable to textile production than G. hirsutum fibres11,12, mills are willing to pay more for G. arboreum, because its fibre is scarce (G. arboreum is estimated to represent less than 3% of the cotton area in India), while still valued26,27.

The final aspect to consider before drawing conclusions about the economic impact of growing Bt G. hirsutum compared with growing G. arboreum is the net revenue obtained by the farmers; that is, how much is left to impact farmers' income once expenditure has been subtracted from the revenue gained by selling the crop. Table 1 shows that the net revenue was not significantly different between farmers growing Bt G. hirsutum and farmers growing G. arboreum. However, the coefficient of variation (CV = standard deviation/mean) was much higher among G. arboreum farms (CV = 83) than among Bt G. hirsutum farms (CV = 58), indicating that net revenue among G. arboreum farms was more variable than the net revenue among Bt G. hirsutum farms. Taken together our findings show that, although net revenue is more variable among G. arboreum farms, farmers growing G. arboreum under rainfed conditions can obtain similar net revenue to farmers cultivating Bt G. hirsutum under rainfed conditions.

The next analyses compared the economic impact of growing Bt G. hirsutum under rainfed conditions with the cultivation of Bt G. hirsutum under irrigated conditions. It is established that irrigation increases Bt G. hirsutum yields4,28, but little is known about how irrigation affects both expenditure and net revenue of smallholder farmers growing Bt G. hirsutum. We found that even when the costs of irrigation are excluded, farmers growing Bt G. hirsutum using irrigation spent more money on cotton production than farmers growing Bt G. hirsutum without irrigation (Student's t-test; P < 0.001; Table 1). Yield and revenue were also higher if Bt G. hirsutum was irrigated (Student's t-test; P < 0.01), however, not high enough to compensate for the increase in expenditure, and there were no significant differences in net revenue between irrigated and rainfed Bt G. hirsutum farms (Student's t-test; P = 0.23). These data demonstrate that farmers cultivating Bt G. hirsutum achieve higher yields if they use irrigation, but do not accrue significantly higher net revenue because they spend more to cultivate the crop.

To identify expenditure that could be reduced without impacting yield, we compared the influence of several inputs—fertilizer, pesticide and manure expenditure, and frequency of irrigation—on yield (Table 2). Fertilizer expenditure was the main driver of yield among rainfed G. arboreum farms, while the most important factors determining yield on rainfed Bt G. hirsutum were fertilizer and manure expenditure (Table 2a,b, respectively). Among irrigated Bt G. hirsutum farms, the most important factor determining yield was the frequency of irrigation (Table 2c). Further analyses revealed that the relationship between irrigation frequency and yield was positive and non-linear (Supplementary Fig. 1). Taken together these data suggest that farmers cultivating Bt G. hirsutum under irrigated conditions could consider reducing expenditure on fertilizers and pesticides without major negative impacts on yield while increasing irrigation frequency could significantly increase yield.

Table 2: Influence of inputs on yield.

Analyses were performed to assess how yields and expenditure influenced net revenue. The resulting regression models showed that yields were positively associated with net revenue across all cotton practices (Fig. 1d-f). In contrast, the relationship between expenditure and net revenue was not significant across all cotton practices. The more farmers spent on G. arboreum, the more net revenue they obtained (Fig. 1a), but there was no discernible relationship between expenditure and net revenue on either irrigated or rainfed Bt G. hirsutum farms (Fig. 1b,c). This suggests that farmers growing Bt G. hirsutum use their expenditure less effectively to increase their net revenue than farmers growing G. arboreum, which may be the result of a lack of adequate skills in the cultivation of Bt G. hirsutum29.

Figure 1: Relationship between total expenditure and net revenue (left column), and yield and net revenue (right column) in rainfed G. arboreum (a,d), rainfed Bt G. hirsutum (b,e) and irrigated Bt G. hirsutum (c,f).
Figure 1

F-statistics and the resulting statistical significance of each relationship are shown in the top left of each graph.

We hypothesized that the socio-economic background of the farmers could vary between farmers adopting different cotton practices. To assess the socio-economic status of the farmers interviewed, we integrated several socio-economic variables into one composite index that we called the ‘wealth index’. The following socio-economic variables were included in this index: irrigated area owned by the farmer, non-farm income, household size and number of months experiencing food insecurity (a full description of the wealth index can be found in the Methods section). Our index indicated that farmers growing Bt G. hirsutum under irrigated conditions were wealthier than farmers growing Bt G. hirsutum under rainfed conditions (Supplementary Fig. 2). In contrast, no significant differences in wealth were found between farmers cultivating G. arboreum and farmers cultivating Bt G. hirsutum, either under rainfed or irrigated conditions. This indicates that farmers' wealth is not significantly different between farmers that cultivate different cotton varieties (G. arboreum and Bt G. hirsutum), but farmers cultivating Bt G. hirsutum with irrigation are wealthier than farmers cultivating Bt G. hirsutum without irrigation.

The wealth index was also used to assess the impact of wealth on expenditure on cotton farming. We used multiple regression to study the impact of wealth on pesticide, fertilizer and manure expenditure (Fig. 2). We found that farmers cultivating Bt G. hirsutum (rainfed or irrigated) spent more money on pesticides and fertilizers than farmers cultivating G. arboreum (Fig. 2a,b, respectively), and this expenditure was independent of farmers' wealth (Supplementary Table 1 shows B estimates and P values of these analyses). In contrast, farmers' expenditure on manure did not vary between cotton practices, but varied depending on the wealth of the farmer: wealthier farmers spent more money on manure (Fig. 2c). These findings indicate that farmers' expenditure on pesticides and fertilizers is dependent on their cotton practice, while farmers' expenditure on manure is independent of the cotton practice, but dependent on farmers' wealth.

Figure 2: Relationship between farmers' wealth and expenditure (how much money farmers spent on a particular input).
Figure 2

a, Pesticide expenditure. b, Fertilizer expenditure. c, Manure expenditure. Each colour represents data from a particular cotton practice: rainfed Bt G. hirsutum in red, irrigated Bt G. hirsutum in green and rainfed G. arboreum in blue. The shade area is the 95% confidence around the line.

Once we established that wealth did not vary between farmers growing Bt G. hirsutum and farmers growing G. arboreum, we wanted to know the main reasons why farmers adopted either Bt G. hirsutum or G. arboreum. Results from the interviews indicated that farmers growing G. arboreum under rainfed conditions preferred to grow G. arboreum instead of Bt G. hirsutum because G. arboreum cultivation tends to require less expenditure on inputs and seeds (58% of farmers growing G. arboreum gave this reason), and to be better suited for rainfed conditions (42%). Farmers growing Bt G. hirsutum gave the following two reasons for growing Bt G. hirsutum instead of G. arboreum: they expected higher yields with Bt G. hirsutum (71 and 73% of farmers growing Bt G. hirsutum under rainfed and irrigated conditions gave this reason, respectively), and more resistance to bollworms (35 and 27% of the farmers, respectively). Remarkably, farmers growing Bt G. hirsutum both under rainfed and irrigated conditions claimed that G. arboreum was not an option for them because of the low availability of G. arboreum seed (12 and 40% of the farmers in each respective group claimed that it was difficult to obtain G. arboreum seed). Taken together, these data suggest that the main reasons why farmers adopt Bt G. hirsutum instead of G. arboreum is the expectation of higher yields and more resistance to bollworms, but also because G. arboreum seed is hardly available. In contrast, farmers choose to grow G. arboreum instead of Bt G. hirsutum because with G. arboreum cultivation they expect lower expenditure and better performance under rainfed conditions.

We found evidence demonstrating that under rainfed conditions, G. arboreum cultivation can generate similar economic benefits for farmers as Bt G. hirsutum cultivation in Maharashtra. Although farmers growing Bt G. hirsutum have more stable net revenue, they also need to spend more to obtain similar net revenue than farmers growing G. arboreum. This indicates that under rainfed conditions, the economic benefits associated with Bt G. hirsutum cultivation are not necessarily realized. In these conditions, other cotton varieties such as the Asiatic cotton G. arboreum may offer an alternative for cotton farmers in Maharashtra and perhaps in other cotton-cultivating areas. When farmers growing Bt G. hirsutum use irrigation, they obtain higher yields than under rainfed conditions, but our data demonstrate that without effective management of the expenditure, higher yields do not translate into higher net revenue. We conclude that our study provides insights into how the potential of Bt G. hirsutum cultivation is constrained under rainfed conditions in India, and how even though Bt G. hirsutum yields increase with irrigation, this does not necessary translate into an increase in the economic benefits received by Indian smallholder farmers.

Methods

Survey

An independent survey of Indian farmers growing cotton was carried out in Amravati district of Maharashtra state. In total, 51 smallholder farmers were interviewed twice; the first survey was carried in 2013 during the main cotton season (August to September) and the second one after harvest (March 2014). Farms were assigned to one of the three following cotton practices: (a) Bt G. hirsutum under rainfed conditions (n = 17), (b) Bt G. hirsutum under irrigated conditions (n = 15) or (c) G. arboreum under rainfed conditions (n = 19). Ideally, we would have also included G. arboreum under irrigated conditions, but this cotton practice was hardly found in Amravati. Farmers were chosen using a stratified sampling design that accounted for regional variability, that is, at least two of the three types of cotton practices were sampled in all villages except for one. Regional maps were also consulted to ensure that the 18 villages sampled shared precipitation and temperature conditions. Moreover, we homogenized our sample by ensuring that farmers and their farms met the following criteria: (1) maximum 12 acres of landholding area; (2) farms placed on black soils (the substrate most commonly used to grow cotton in Maharashtra); and (3) farmers had sown that particular variety of cotton at least during the last 3 years. The NGO ‘Community Action for Rural Development’ (CARD), based in Anjangaon, gave us the local support needed to find the farmers that met our criteria.

Each farm was visited to verify the information gathered through the survey, and to test leaves for the presence or absence of Bt endotoxins Cry1Ac and Cry2A, using the QuickStix Combo Kit (AS 012LS, EnviroLogix, Maine, USA).

When farmers were asked for the main reasons to adopt either Bt G. hirsutum or G. arboreum in their fields, we considered as main reasons the ones that were given by more than one-quarter (25%) of the farmers interviewed in each particular group.

Wealth index

We considered eight socio-economic variables to create the wealth index: household size, years of schooling, household assets (TV and vehicles), non-farm income, months experiencing food insecurity, and total and irrigated landholding area. With these variables, we created a matrix of the Pearson's correlation coefficients and the corresponding P values to reveal correlations between variables (Supplementary Table 2). Once such correlations were revealed, we selected the variables to be included in the wealth index. As one variable (‘number of months experiencing food insecurity’) was significantly correlated to five variables, it was included in the farmers' wealth index to account for the variability of the other variables. The two other variables not related to this one (area of irrigated land and household size) were also included in the index. Finally, non-farm income (Rs month−1) was also included as, even though it was correlated to the ‘number of months experiencing food insecurity’, it was considered an important contributor to wealth. To obtain then the wealth index per farmer, the value of each of these four variables was divided by the maximum value of that variable in the sample, and the resulting value was either added if it contributed to wealth (area of irrigated land and non-farm income) or subtracted if it negatively affected wealth (household size and food insecurity).Wealthindex=IrrigatedlandholdingareaMaxvalueinthesample+Non-farm incomeMaxvalueinthesampleHouseholdsizeMaxvalueinthesampleMonthsfoodinsecurityMaxvalueinthesample

Statistical analyses

To compare the mean values of the economic variables considered, the independent Student's t-test was used whenever the data were normally distributed or could be normalized, and the variances of the two groups were equal. When these assumptions were not met, the assumption-free Wilcoxon rank-sum test was used. When the means of a variable were compared between the three cotton practices (that is, for the wealth index), one-way ANOVA was used as the distributions within groups were normally distributed or could be normalized, and the variances were fairly equal.

General linear models were used to study the influence of several independent variables (fertilizer, pesticide and manure expenditure, and irrigation frequency) on yield, and to obtain the Akaike's information criterion (AIC). Delta AIC (Δi) and Akaike weight (wi) were subsequently calculated. The model(s) with the highest wi and with Δi < 2 were considered the best among the whole set of candidate models30.

We used multiple regression to test the influence of wealth on expenditure, including the cotton practice as a dummy variable. The Durbin–Watson statistic was used to test the assumption of independent errors, and the variance inflation factor was produced to check that no multicollinearity occurred between the two explanatory variables (cotton practice and farmers' wealth). If these assumptions were not met (that is, for manure expenditure), generalized linear models were used specifying the appropriated error distribution (that is, quasipoisson for manure expenditure).

Finally, Box–Cox transformations and ANOVA tests were used to study the relationship between two variables (for example, total expenditure versus net revenue). All the analyses were conducted using R, version 2.15.2 (ref. 31).

References

  1. 1.

    Global Status of Commercialized Biotech/GM Crops: 2014. Brief No. 49 (ISAAA, 2014).

  2. 2.

    & Yield effects of genetically modified crops in developing countries. Science 299, 900–902 (2003).

  3. 3.

    & Genetically modified insect resistance in cotton: some farm level economic impacts in India. Crop Prot. 24, 433–440 (2005).

  4. 4.

    & Farm-level economic performance of genetically modified cotton in Maharashtra, India. Rev. Agr. Econ. 28, 59–71 (2006).

  5. 5.

    & Adoption of Bt cotton and impact variability: insights from India. Rev. Agr. Econ. 28, 48–58 (2006).

  6. 6.

    Peer-reviewed surveys indicate positive impact of commercialized GM crops. Nature Biotechnol. 28, 319–321 (2010).

  7. 7.

    & Impact of Bt cotton on pesticide poisoning in smallholder agriculture: A panel data analysis. Ecol. Econ. 70, 2105–2113 (2011).

  8. 8.

    Field versus farm in Warangal: Bt cotton, higher yields, and larger questions. World Dev. 39, 387–398 (2011).

  9. 9.

    & Economic impacts and impact dynamics of Bt (Bacillus thuringiensis) cotton in India. Proc. Natl Acad. Sci. USA 109, 11652–11656 (2012).

  10. 10.

    & in Economic and Environmental Sustainability in the Asian Region (eds Gill, S. S., Singh, L. & Marwah, R.) 419–442 (Routledge, 2010).

  11. 11.

    & in Genetics and Genomics of Cotton, Plant Genetics and Genomics: Crops and Models (ed. Paterson, A. H.) 69–97 (Springer, 2009).

  12. 12.

    & in Cotton: Biotechnological Advances (ed. Zehr, U. B.) 15–40 (Springer, 2010).

  13. 13.

    et al. Identification and expression of six drought-responsive transcripts through differential display in Desi cotton (Gossypium arboreum). Mol. Biol. 42, 492–498 (2008).

  14. 14.

    in Agricultural Growth and Productivity in Maharashtra. Trends and Determinants Ch. 4 (Allied Publishers, 2011).

  15. 15.

    & Water resources in India. Curr. Sci. India 89, 794–811 (2005).

  16. 16.

    & Satellite-based estimates of groundwater depletion in India. Nature 460, 999–1003 (2009).

  17. 17.

    et al. Constraints and potentials of future irrigation water availability on agricultural production under climate change. Proc. Natl Acad. Sci. USA 111, 3239–3244 (2014).

  18. 18.

    Poor crop management plagues Bt cotton experiment in India. Nature Biotechnol. 20, 1069 (2002).

  19. 19.

    et al. Farm-scale evaluation of the impacts of transgenic cotton on biodiversity, pesticide use, and yield. Proc. Natl Acad. Sci. USA 103, 7571–7576 (2006).

  20. 20.

    & in Economic and Environmental Sustainability in the Asian Region (eds & ) 391–418 (Routledge, 2010).

  21. 21.

    & Effect of some antibiotic compounds in cotton on post-embryonic development of spotted bollworm (Earias vittella F.) and the mechanism of resistance in Gossypium arboreum. Proc. Indian Acad. Sci. (Anim. Sci.) 91, 67–77 (1982).

  22. 22.

    & Assessment of the resistance of cotton germplasm (Gossypium spp.) to aphids (Homoptera, Aphididae) and leafhoppers (Homoptera: Cicadellidae, Typhlocybinae): methodology and genetic variability. Plant Breed. 127, 376–382 (2008).

  23. 23.

    et al. Evaluation of Gossypium species for resistance to cotton leaf curl Burewala virus. Ann. Appl. Biol. 157, 135–147 (2010).

  24. 24.

    & Novel cotton germplasm with host plant resistance to twospotted spider mite. Field Crop Res. 134, 114–121 (2012).

  25. 25.

    et al. Development of Gossypium hirsutum lines with cytoplasmic bollworm tolerance from Gossypium arboreum and Gossypium tomentosum. Plant Breed. 128, 712–715 (2009).

  26. 26.

    & Studies on improved cotton: Part I – Fibre quality parameters. Ind. J. Fibre Textile Res. 36, 24–34 (2011).

  27. 27.

    Long Live Swadesi Cotton. Cotton Statistics and News (Cotton Association of India, 2013).

  28. 28.

    et al. Productivity and quality of Bt cotton (Gossypium hirsutum) as influenced by planting geometry and nitrogen levels under irrigated and rainfed conditions. Indian J. Agr. Sci. 84, 1069–1072 (2014).

  29. 29.

    Agricultural deskilling and the spread of genetically modified cotton in Warangal. Curr. Anthropol. 48, 67–103 (2007).

  30. 30.

    & Model Selection and Multimodel Inference: a Practical Information-Theoretic Approach (Springer-Verlag, 2002).

  31. 31.

    R Development Core Team R: A Language and Environment for Statistical Computing (R Foundation for Statistical Computing, 2013).

Download references

Acknowledgements

Sincere thanks go to V. Ladole and all his team from the Indian NGO ‘Community Action for Rural Development’, whose collaboration was key to the success of this study. We especially thank S. Ladole, R. Ramrao, H-H. Tao, D. Jorda and M. Manoj for helping us with the fieldwork. This research was funded by the John Templeton Foundation.

Author information

Affiliations

  1. Department of Zoology, University of Oxford, Oxford OX1 3PS, UK

    • Carla Romeu-Dalmau
    • , Michael B. Bonsall
    •  & Katherine J. Willis
  2. Department of Plant Sciences, University of Oxford, South Parks Road, Oxford OX1 3RB, UK

    • Carla Romeu-Dalmau
    •  & Liam Dolan
  3. Oxford Martin School, 34 Broad Street, Oxford OX1 3BD, UK

    • Carla Romeu-Dalmau
    • , Michael B. Bonsall
    • , Katherine J. Willis
    •  & Liam Dolan
  4. St. Peter's College, Oxford, OX1 2DL, UK

    • Michael B. Bonsall
  5. Department of Biology, University of Bergen, PO Box 7803, Bergen 5020, Norway

    • Katherine J. Willis
  6. Royal Botanic Gardens, Kew, Richmond TW9 3AE, UK

    • Katherine J. Willis

Authors

  1. Search for Carla Romeu-Dalmau in:

  2. Search for Michael B. Bonsall in:

  3. Search for Katherine J. Willis in:

  4. Search for Liam Dolan in:

Contributions

All authors contributed to the study design and are responsible for the integrity of the work as a whole. L.D. and K.W. secured funding for the project. C.R.D. did the fieldwork and analysed and interpreted the data along with M.B.B. C.R.D. wrote the manuscript with the input of all co-authors, especially L.D.

Competing interests

The authors declare no competing financial interests.

Corresponding authors

Correspondence to Carla Romeu-Dalmau or Liam Dolan.

Supplementary information