Pesticide Usage by Cotton Farmers in India Changes over a Decade

With India emerging as a leading cotton producer in the world, and considering the large-scale adoption of Bt cotton cultivation, there is a need to understand the patterns of pesticide use by cotton farmers, especially as environmental, ecological, and health concerns surrounding pesticide use continue to be debated.

Globally, synthetic pesticides have become the predominant method for controlling pests.1Consumption of pesticides is particularly high in cotton cultivation as the crop yield is seriously affected by pest attacks. The potential production losses due to pests in the absence of pest control mechanisms worldwide have been estimated at around 82% for cotton (Oerke 2006). In India, the loss of cotton yield due to bollworm was estimated to be around 50%–60% (Narayanan and Ramaswami 2007). Given this, it is no surprise that cotton farmers use a large amount of pesticides as compared to other farmers. India is no exception to this, and farmers use a wide range of pesticides for damage control. The use of pesticides in India’s cotton production is critical because it has the largest area under cotton cultivation and, as such, it is emerging as the leading producer of cotton. India is also the second largest exporter of cotton in the world. Cotton forms roughly 5% of the gross cropped area in the country while consuming 36%–50% of the total pesticides in the country (Devi 2010; FICCI 2013; Bhardwaj and Sharma 2013).

Pesticide consumption for plant protection in India is estimated to be around 600 grams per hectare, while countries such as Taiwan, China and Japan consume 17 kilogrammes (kg), 13 kg and 12 kg per hectare respectively (Singh et al 2014). However, with respect to overall consumption, India ranks 10th in the world. From the perspective of farmers’ utilisation of pesticide, several concerns demand attention. First, there have been concerns regarding the prevalence of some hazardous and banned pesticides, which makes current pesticide levels highly risky to health, environment, and ecology (Shetty et al 2008; Devi 2010; Kouser and Qaim 2011; Taneja 2017;BBC 2017; Kranthi 2017). Second, the question of irrational usage of pesticides has received attention (Nagaraju et al 2002; Shetty 2004; Ranga Rao et al 2009; Colvin et al 2012). Third, there have been concerns regarding the counterfeit nature of the pesticides available in the Indian market (Ramaswami 2002; Pray and Nagarajan 2014) as well as the “deskilling” of farmers owing to the rapid introduction of new variants (Stone 2011).

The issue of pesticide usage in cotton becomes particularly relevant in the context of the commercialisation of genetically modified cotton hybrid seeds, Bacillus thuringiensis, or Bt cotton varieties since 2002. However, there is an ongoing debate on whether cultivation of Bt cotton has resulted in lower pesticide usage (Krishna and Qaim 2012; Narayanan and Viswanathan 2015a, 2015b). Studies show that although total pesticide usage has significantly reduced (Carpenter 2010; Kathage and Qaim 2012; Krishna and Qaim 2012), there has been an increase in pesticide sprays against the outbreak of secondary pests such as jassids and aphids (Stone 2011; Nagrare et al 2009). In spite of the association between the growing cultivation of Bt cotton and pesticide use among cotton farmers, not much is known in terms of the trends and determinants of pesticide usage since the commercialisation of Bt cotton.

In this paper, we use unit-level National Sample Survey (NSS) data to explore pesticide usage and changes from 2002–03 to 2012–13. The rest of the paper is structured as follows. We begin with an explanation of the background of the data and methodology, followed by presentation of the results of the study. We then conclude with a discussion of the findings and policy implications.

Data and Methodology

We use unit-level data from the situation assessment surveys (SAS) of agricultural households conducted by the National Sample Survey Office (NSSO) in its 59th and 70th rounds. This is an important source for data on the agricultural situation of farmers and farm households in India. Some recent studies (Birthal et al 2015; Agarwal and Agrawal 2017) have used the data from these surveys to understand the important aspects of farming among Indian farm households. These surveys were conducted in 2002–03 and, again, a decade later in 2012–13, to capture the condition of agricultural households in India. The households were visited twice to collect information related to farming and other socio-economic characteristics of agricultural households.2 The survey collected extensive information on consumption expenditure, income and productive assets, indebtedness, farming practices and preferences, resource availability, awareness of technological developments and access to modern technology in the field of agriculture, as well as information on crop loss, crop insurance, etc.

Although farm households were surveyed during both the years, the definition of a farm household differed in the two rounds of the surveys. In 2002–03, a “farmer” was defined as a person who operated some land (owned, leased, or otherwise possessed) and engaged in agricultural activities—including cultivation, poultry, fishery, piggery, beekeeping, vermiculture, sericulture, etc—on that land for a period of 365 days preceding the survey. A farmer household was defined as one in which at least one member was a “farmer” thus defined (NSSO 2014). In the 2012–13 survey, the land possession criterion was dropped. Instead, the survey replaced the concept of agricultural households with that of an “agricultural production unit,” that is, one which produces field crops, horticultural crops, livestock and products of any other specified agricultural activity (NSSO 2014: 2). Furthermore, agricultural households were defined as households receiving value of more than ₹3,000 as annual agricultural produce for agricultural activities—including cultivation, poultry, fishery, piggery, beekeeping, vermiculture, sericulture, etc—and in which at least one member was self-employed in agriculture during the 365 days prior to the survey.3 Given this background, we study pesticide expenditure of cotton farmers during the kharif season with data from the 59th and 70th NSS rounds.4

Selection Bias in Pesticide Usage

Since not all farmers use pesticides, there may be inherent differences between those who use pesticides and those who do not. This exclusion of non-users in the subsample of pesticide users is likely to bias our estimates of pesticide usage (expenditure per hectare). This is particularly problematic if observations are selected in a process that is not independent of the outcome variables of interest. Therefore, we address the selection bias using Heckman’s (1976, 1979) two-stage correction. In the first stage (selection model), we estimate a probit model of whether or not a farmer uses pesticides by controlling for farm and farmer characteristics. This provides us with a correction factor, termed the inverse Mills ratio, that is used in the second stage (outcome model) which is estimated using ordinary least squares. In the second stage estimation, we take into account an exclusion restriction for identification purpose, that is, there is at least one explanatory variable in the first stage which is excluded from the second stage equation. This implies that the process of selection into the sample of pesticide users and the extent or intensity of pesticide usage does not follow from the same process. Before we proceed with our analysis, we present the descriptive statistics related to pesticide expenditure by farmers in both the years.

Size-class of Cotton Holdings

For our analysis, we estimate the number of cotton farmers and cultivated area under cotton in the kharif season for both the survey years. We restrict the analysis to the kharif season as cotton is largely a kharif crop. Around 38.5 lakh and 68.1 lakh households cultivated cotton in 2002–03 and 2012–13, respectively.5 Average land under cotton cultivation was 1.21 hectares (ha) in 2002–03 compared to 1.17 ha in 2012–13. This meant that, as per the NSS data, the area under cotton cultivation during the kharif season of 2002–03 and 2012–13 was 46.5 lakh ha and 75.2 lakh ha, respectively. These numbers are lower than the official estimates of 76.7 lakh ha and 119.8 lakh ha for 2002–03 and 2012–13, respectively.6 Such a discrepancy could arise due to two reasons. First, cotton is cultivated during the rabi season in some states like Tamil Nadu, which has not been considered in our analysis. Second, the way NSS has defined farm households and the manner in which it samples households may result in such discrepancies. It is worth noting that the total cultivable land estimated by the NSSO survey is far lower than the total cultivable land as estimated in the agricultural land census data.7

From the NSS estimates, we find that the number of households cultivating cotton has increased by two-thirds while the area under cotton cultivation increased by around three-fifths in the decade under consideration. However, both 2002–03 and 2012–13 being drought years (the former being more severe), the trends should be interpreted with caution. Table 1 (p 44) shows the changes in the percentage of households cultivating cotton and the average cotton area cultivated by farmers across cotton growing states.

From Table 1, we find that around 4% of farm households in 2002–03, and 6% of farm households in 2012–13 cultivated cotton along with some other crop. We also find that, in India, the percentage of farmers who cultivated only cotton (monocropped cotton) increased threefold between 2002–03 and 2012–13. The average land cultivated by cotton farmers is higher than that of farmers cultivating other crops. In 2012–13, the average cropped area during the kharif season was 0.94 ha for all cultivating households, while average area cultivated under cotton was 1.18 ha (Ranganathan 2015). This indicates that cotton is primarily cultivated by larger farmers. Also, during the decade, the average land size of monocropped cotton increased from 1.12 ha to 1.25 ha whilst that of intercropped cotton fell from 1.22 ha to 1.12 ha.

It should be borne in mind that there are three agroecological regions—north, central and south—for cotton production and there are wide variations in the extent of irrigation of cotton crop both between and within the regions. The northern zone has a distinctly higher percentage of net sown area under cotton being irrigated, as compared to the southern and central zones. Across states, we find that a higher proportion of farm households cultivated cotton in Maharashtra, Punjab and Haryana. Furthermore, there is a considerable increase in the proportion of farmers growing monocropped cotton in these two states. Gujarat, a major cotton producing state saw a large increase in the proportion of farm households cultivating cotton and a decline in the average area under cotton cultivation. In all the four southern cotton growing states, there was an increase in the proportion of farm households cultivating cotton and the average land sizes on which cotton was cultivated.

Table 2 shows that about three out of 10 farmers growing cotton with other crops in 2002–03 and 2012–13 had semi-medium landholdings, that is, land sizes of 2 ha–4 ha. Compared to 2002–03, a higher proportion of farmers in 2012–13 were small and marginal landholders while a lower proportion were in other size-classes of land cultivated. This means that even though the average land under cotton increased in each of the size-classes, the average area cultivated under cotton by these households declined from 1.22 ha to 1.15 ha. Among the households that cultivated cotton as a monocrop, a smaller proportion of households were small and marginal landholders, while a higher proportion belonged to other land class categories. This meant that the average area allotted by households to monocropped cotton increased from 1.12 ha in 2002–03 to 1.25 ha in 2012–13.

Cost of Cultivation

Table 3 reports the cost composition of households that cultivated crops other than cotton, households that cultivated cotton with other crops, and households that cultivated only cotton. Costs reported are paid-up costs and include expenditure on purchased items only.8

While there has been a marked decrease in the share of seed costs in the total cost (per ha) for non-cotton growers, there has been an increase in the seed cost share for cotton growers; the increment being larger for monocropped cotton growers vis-á-vis farmers growing cotton with other crops. Although pesticide expenditure is a major component of total costs for cotton growers (apart from seed, fertilisers, and human labour), the share of pesticide costs have fallen considerably. In the absence of information on whether a farmer used Bt or non-Bt technology, it can be surmised that reduction in bollworm pressure by use of Bt technology may be associated with lower pressure of bollworms regionally. This may, in turn, induce non-Bt growers to reduce their pesticide applications (Krishna and Qaim 2012: 47). However, the emergence of secondary pests such as jassids and aphids could be a negative effect of the reduction in pesticide use following the widespread use of Bt technology by 2012–13 (Stone 2012; Herring 2013). This period has also witnessed the regulatory ceiling placed on the price of Bt cotton in Andhra Pradesh, followed by other states. This is likely to have reduced the seed costs to some extent. Some of the primary survey-based studies indicate that in their study regions, the proportion of expenditure on pesticides reduced from 41% in 1999 to 21% in 2010 (Narayanan and Viswanathan 2015b).

Pesticide Expenditure and Farmer Characteristics

Table 4 presents statewide variations in pesticide expenditure per ha and as percentage of total costs over the two survey years for cotton growing states and overall for India.

Pesticide expenditure per ha increased by about 1.5 times for crops other than cotton over the decade. However, pesticide expenditure per ha for farmers cultivating cotton with other crops remained almost the same and for farmers cultivating only cotton, the pesticide expenditure per ha decreased by a factor of 0.65. As a share of the total costs, pesticide expenditure remained constant for crops other than cotton, while it reduced for cotton with other crops and cotton monocrop. Such patterns are evident in all states except Karnataka and Tamil Nadu, where the pesticide expenditure per ha and share of it in total costs increased for those farmers monocropping cotton.

In terms of pesticide expenditures per ha by size-class of landholding (Table 5), there is an inverse relationship between average pesticide expenditure per hectare and size-class of landholding for all three categories of farm households. For both the years and all categories of farmers, the differences were statistically significant at 5%.

Since around two-thirds of Indian cotton is grown in rainfed conditions (Gaurav 2014), it is worth examining the differences in pesticide expenditure by irrigation status. As shown in Table 6, there are considerable differences in pesticide expenditure between irrigated and unirrigated farmers. In both the years, pesticide expenses per unit area were considerably higher among cotton monocroppers than the other two groups. While in 2002–03, unirrigated monocropped cotton registered higher pesticide expenditure per hectare than irrigated ones, in 2012–13, irrigated growers spent more on pesticides per hectare than the other two groups. The difference between pesticide expenses across irrigated and unirrigated farmers was statistically significant at 1% level for farmers cultivating cotton with other crops.

Another line of investigation that may show variations in pesticide expenditures by farmers is the tenancy pattern. Table 7 presents the variations in pesticide expenses per hectare across tenancy categories. In our analysis, households that cultivated their own land were categorised as owner cultivators; the households that had their own land and leased-in some more land were categorised as tenants; and those who cultivated on leased-in land were categorised as landless tenants for the purpose of our analysis.

Among farmers cultivating crops other than cotton, landless tenants spent more on pesticides per hectare than tenants or owner cultivators in both 2002–03 and 2012–13. The difference though, was statistically significant at 5% level in 2002–03 but not in 2012–13. For farmers cultivating cotton with other crops, tenants spent more on pesticides than owner cultivators and landless tenants in both the years. In 2012–13, the difference was high and statistically significant at 1% level. For farmers who monocropped cotton, there has been a considerable decline in average per hectare expenditure on pesticides irrespective of tenancy patterns. It is also evident that landless tenants growing monocropped cotton spent a lot more on pesticides per hectare than the tenants and owner cultivator in both the years. However, as a percentage of the total costs, the pesticide expenditure among monocropped cotton farmers was almost the same across three categories of farm households in 2002–03. In 2012–13, the share of pesticide expenditure in total costs was lowest for landless tenants, while it was highest for tenants. The changes over the decade could be attributed to changes in the proportion of farmers across different tenancy categories, cultivating cotton during these two study years, as well as heterogeneity in pest pressure and changing pest ecology. However, the patterns indicate that farmers leasing-in land spend more (as a proportion of total expenditure) on pesticides than landowners.

Extending our analysis of the variation in pesticide expenditure to farmers who are insured as opposed to those who are not, Table 8 reports the pesticide expenditure for households, categorised according to their status of insuring cotton. In 2002–03, the survey collected information from households on whether they insured any of the crops but did not collect information related to each of the specific crops cultivated by the households. So, for farmers cultivating cotton and other crops, we do not have information on whether specifically cotton was insured. This information was, however, collected in 2012–13 and hence highlighting the differences between insured and non-insured cotton growers is possible for 2012–13. Also, in 2002–03, the survey collected information only on whether the household insured any crop. In 2012–13, the survey collected information on whether the household insured a particular crop along with the loan, as well as whether the household insured the crop additionally, by purchasing insurance apart from that accompanying the loan.

In 2002–03, pesticide expenditure for the insured was significantly lower than that of the uninsured at the 5% level. However, in 2012–13, the differences in pesticide expenses between the two groups were not statistically significant. Among farmers cultivating cotton with other crops in 2012–13, farmers insuring cotton additionally spent the least on pesticides while those whose cotton was insured with the loan spent the most for pesticides. The same was observed among monocropping cotton farmers in 2012–13, but as a percentage of total costs. Monocropping farmers who did not insure their cotton spent the least on pesticides, while those who insured it along with the loan product spent the most. This is weakly suggestive of possible adverse selection and moral hazard issues. Note that 2002–03 was a severe drought year while 2012–13 was not particularly good either. The findings related to insurance have to be read with the caveat that the proportion of insured farmers in the data across both the years is quite small.

In order to understand whether training in agriculture may influence pesticide expenditure patterns, we present variations by households in which the household head has received agricultural training, as against those who have not received training (Table 9, p 48).

In 2002–03, for farmers cultivating cotton with other crops, those who attended training spent less on pesticides compared to those who did not. However, among monocroppers, those who attended training spent more on pesticides than those who did not. In 2012–13, for farmers not cultivating cotton, the amount spent on pesticides was similar, irrespective of training status. Pesticide expenses as percentages of total costs, however, were higher for farmers attending agricultural training than for those who did not. This difference is not statistically significant. On the other hand, for farmers cultivating cotton with other crops, and as a monocrop, those who attended training spent far less per hectare than those who did not, and the differences were statistically significant at 10% level.

A related factor that could influence the expenditure on pesticides is access to extension services and information about agricultural practices pertaining to pesticides. We have considered cases where farmers had direct access to private extension services such as input dealers and non-governmental organisations (NGOs). Table 10 shows the variations in pesticide expenses for those farmers accessing private extension, and those who did not access these directly.

Those who reported having accessed private extension services spent more on pesticides as compared to those who did not. The differences were statistically significant at 1% level for non-cotton farmers for both 2002–03 and 2012–13. For farmers growing cotton with other crops, the difference was significant at 1% level in 2012–13, but not in 2002–03. For monocropping cotton farmers, although those who reported accessing extension services had considerably higher pesticide expenditures per hectare than those who did not, the difference was not statistically significant at the conventional levels.

Therefore, we find that characteristics such as size of landholding, irrigation, tenancy, agricultural insurance, training, and access to private extension, seem to influence the amount and share spent on pesticides by cotton farmers.

Determinants of Pesticide Expenditure

The previous section analysed the variations in pesticide expenditure per hectare across different categories of farmers and across different parameters of importance. However, these were not analysed in a multivariate context, where different variables of interest were considered for the analysis. Next, we present the results of the Heckman selection model, which corrects for selection bias. Table 11 (p 48) presents the results of the analysis for farmers who cultivated cotton (as a monocrop or with other crops) during the kharif seasons of 2002–03 and 2012–13. The exclusion criteria were imposed in the form of two variables: whether the household head attended agricultural training; and whether the household head accessed any private extension service such as input dealers, NGOs, etc. These two controls were included in the selection model, but not in the outcome model.

From Table 11, we find that the age of the household head is associated with the likelihood of pesticide usage, but not with the amount of money spent on pesticides. The probability of pesticide usage increased initially with education, and then decreased with higher education. The area under cotton cultivation significantly influences the likelihood as well as the extent of pesticide expenditure. The likelihood of pesticide usage increased with land size initially and then decreased after a particular point. The selection model indicated a non-linear relationship with a decline in pesticide expenditure per hectare: decreasing up to around six hectares and then increasing thereafter. The finding that an additional hectare of cotton cultivation is associated with significantly lower pesticide expenditure per hectare, suggests that smallholders spend disproportionately higher on pesticides per unit area of cotton production, as compared to farmers with larger holdings.

The role of credit in pesticide usage is evident. The outstanding loan amount of the household is not associated with the likelihood of pesticide usage, but is associated with pesticide expenditure. With a percentage increase in the loan amount, the pesticide expenses per hectare increased by around 2% in 2002–03, and by 4% in 2012–13. The result is statistically significant at the 5% level. The share of cultivated land under irrigation and leased-in is positively associated with pesticide expenditure.

Attending agricultural training has a negative association with the choice of pesticide, while access to private extension services have a negative association with the probability of pesticide use.


The patterns of pesticide usage by cotton farmers in India have been analysed using two rounds of data, collected in the SAS by the NSSO in 2002–03 and 2012–13. This is an important decade, during which cotton cultivation in India underwent significant changes. Since the commercialisation of transgenic Bt cotton seeds in 2002, Bt cotton has almost entirely replaced conventional seeds across the country. With India emerging as a leading producer and exporter of cotton globally, narratives of concern about health, environmental, and ecological risks associated with pesticide use in cotton—the crop that uses pesticides most intensively—have emerged. This decade is also important from the perspective of the agrarian crises that have been considerably severe in India’s cotton belt. Our findings indicate that there has been a considerable increase in the area under cotton cultivation, as well as the number of households cultivating cotton during the period of analysis. We also find evidence of increasing monocropping by cotton farmers. While cotton cultivation is mostly taken up by farmers with large landholdings, we find some evidence of a greater number of small and marginal farmers cultivating cotton in 2012–13 than in 2002–03. Upon analysing the cost of cultivation, we find that there has been a significant change in the composition of costs for cotton farmers. Pesticide costs per hectare in real terms (at 2002–03 prices) and as a share of total costs were lower in 2012–13 as compared to 2002–03. However, the share of seed costs in the cost of cultivation has increased over the same period. This is primarily on account of the near universal adoption of Bt cotton with considerably higher seed costs per hectare of production. However, issues of growing resistance to Bt toxins and incidence of secondary pests such as whiteflies and jassids in cotton are serious concerns and further investigation into the relationship between proliferation of Bt hybrids and use of specific pesticides needs to be undertaken (Herring 2013; Gaurav 2014).

On correlates of pesticide expenditure per hectare, we find that smaller farmers tend to spend more, although larger farmers were more likely to use pesticides. This may be attributed to the tendency of small farmers to use purchased inputs more intensively, given their lower size of landholdings. Furthermore, availability of credit was positively associated with pesticide expenditure during both the study years, suggesting that credit constraints are an important barrier to pesticide usage. Concerns about indebtedness of Bt cotton farmers in particular has been hotly debated (Gruère and Sengupta 2011). Our findings suggest that there is need for a further investigation of this linkage.

Our finding that irrigation and tenancy increased the spending on pesticides deserves attention. While irrigation and spraying of pesticides often takes place together, those leasing in land may exhibit risk aversion towards pest losses, thereby spending more on pesticides per hectare of land. The average pesticide expenditure in the district where a farmer resides was positively associated with farmer’s pesticide expenditure. This is indicative of not only access to pesticides, but also of potential social learning effects (Munshi 2004; Krishnan and Patnam 2013). Training in agriculture and access to private extension services were linked to the likelihood of pesticide usage, but not with the money spent on pesticides. The association of pesticide use with training may not arise only because of the training content, but attending training could be an indicator of the access and interest of the farmer in knowledge and extension.

Our findings have the following policy implications: First, since more marginal and small farmers are cultivating cotton, particularly by monocropping, there is a strong case for cost reduction and/or appropriate insurance. Since seeds are contributing significantly to the cost of cotton cultivation, there is a need to regulate the seed quality, as well as costs for the farmer. Also, facilities to test seed quality, like testing soil health, need to be set up so that such significant investments by farmers are made without risk. Second, given the regulatory issues surrounding tenancy and reduction in average agricultural landholdings in the country, higher pesticide expenditure per hectare by tenants could be due to anxieties involved in losses caused by pest attacks and also lack of access to formal financial services. Last but not the least, our findings could be relevant in identifying heterogeneity in response to public and private extension services that influence farm practices regarding pest management. Lessons from public programmes such as insecticide resistance management (IRM) under the Technology Mission on Cotton (TMC), could throw light on localised variations in sustainable pest management. Success stories from the use of biopesticides and non-pesticidal management (NPM) as well as integrated pest management (IPM) practices should be widely disseminated among farmers. Since the spread and success of these programmes have been low, understanding the barriers to pesticide usage as well as determinants of the extensive and intensive margin may offer useful insights to influence the programme performance. This is also particularly relevant given the higher pesticide expenditures for those depending on input advisories from agricultural input dealers andNGOs, as public extension services have weakened.

A limitation of our study is that the information on pesticide expenditure is useful, but is limited, in the sense that it does not provide us with information on the quantity and types of pesticide used. It could so happen that farmers are using a lesser amount of costly pesticides (required to tackle bollworms), but more of cheaper pesticides (targeting secondary pests, for instance), thereby creating difficulties in assessing quantitative and qualitative changes in pesticide usage. There could also be some price effects with pesticide prices decreasing in 2012–13, which could have reduced the expenditure, although the quantity of pesticide used may be the same or may have increased. We also do not have information on the usage of biopesticides andNPM practices by farmers in the sample, and the expenditure data does not distinguish between spending on chemical pesticides and biopesticides. These are the limitations of using only expenditure data. A primary survey, collecting information related to the types of pesticide and quantity used, could address this information shortfall.


1 Animal pests (insects, mites, nematodes, rodents, slugs and snails, birds), plant pathogens (viruses, bacteria, fungi, chromista) and weeds are collectively called as pests, and are among the biotic factors that have a high potential to damage crop production.

2 Different households were visited in the two survey years. In 2002–03, the information collected during two visits to the households was related to the kharif and rabi seasons, while in 2012–13, the information collected was related to the July to December 2012 and January to June 2013 periods.

3 In the 2002–03 and in 2012–13 rounds, households that were entirely agricultural labour/fishing/artisan/agricultural service households were excluded from the sampling frame.

4 Kharif crops are also known as monsoon crops. Sowing of cotton usually takes place in the first week of June, coinciding with the onset of the southwest monsoon. Harvesting takes place after six to eight months. However, there are state-specific variations in the sowing and harvest months. The availability of reliable
irrigation also influences farmers’ choice of sowing and harvest.

5 The average cotton area among those who cultivated cotton was calculated using analytical weights. The sum of weights was multiplied with the average land sizes to estimate the total area cultivated.

6 Cotton Corporation of India, Government of India (

7 For a more detailed description of issues involved in sampling and underestimation of cultivable land, see Kumar (2016).

8 All the costs reported are in 2002–03 constant prices. The 2012–13 prices were multiplied by the ratio of average of monthly consumer price index for agricultural labourers (CPI–AL) from July to December 2002 and the average of monthly CPI–AL from July to December 2012.


Agarwal, Bina and Ankush Agrawal (2017): “Do Farmers Really Like Farming? Indian Farmers in Transition,” Oxford Development Studies, Vol 45, No 4, pp 1–19.

BBC (2017): “The Indian Farmers Falling Prey to Pesticide,” British Broadcasting Corporation,

Bhardwaj, Tulsi and J P Sharma (2013): “Impact of Pesticides Application in Agricultural Industry: An Indian Scenario,” International Journal of Agriculture and Food Science Technology, Vol 4, No 8, pp 817–22.

Birthal, Pratap, Devesh Roy, M Tajuddin Khan and Digvijay Singh Negi (2015): “Farmers’ Preference for Farming: Evidence from a Nationally Representative Farm Survey in India,” Developing Economies, Vol 53, No 2, pp 122–34.

Carpenter, Janet (2010): “Peer-reviewed Surveys Indicate Positive Impact of Commercialised GM Crops,” Nature Biotechnology, Vol 28, No 4, pp 319–21.

Colvin, John et al (2012): “Socio-economic and Scientific Impact Created by Whitefly-transmitted, Plant-virus Disease Resistant Tomato Varieties in Southern India,” Journal of Integrative Agriculture, Vol 11, No 2, pp 337–45.

Devi, Indira (2010): “Pesticides in Agriculture—A Boon or a Curse? A Case Study of Kerala,” Economic & Political Weekly, Vol 45, Nos 26–27, pp 199–207.

FICCI (2013): “Indian Agrochemicals Industry: Imperatives of Growth,” Knowledge and Strategy Paper released at 3rd National Agrochemicals Conclave 2013, Federation of Indian Chambers of Commerce and Industry, viewed on 16 October 2016,

Gaurav, Sarthak (2014): “Risk and Vulnerability of Agricultural Households in India,” unpublished PhD thesis, Indira Gandhi Institute of Development Research (IGIDR), Mumbai.

Gruère, Guillaume and D Sengupta (2011): “Bt Cotton and Farmer Suicides in India: An Evidence-based Assessment,” Journal of Development Studies, Vol 47, No 2, pp 316–37.

Heckman, James (1976): “The Common Structure of Statistical Models of Truncation, Sample Selection and Limited Dependent Variables and a Simple Estimator for Such Models,” Annals of Economic and Social Measurement, Vol 5, No 4, pp 475–92.

— (1979): “Sample Selection Bias as a Specification Error,” Econometrica, Vol 47, No 1, pp 153–61.

Herring, Ronald J (2013): “Reconstructing Facts in Bt Cotton: Why Scepticism Fails,” Economic & Political Weekly, Vol 48, No 33, pp 63–66.

Kathage, Jonas and Martin Qaim (2012): “Economic Impacts and Impact Dynamics of Bt (Bacillus thuringiensis) Cotton in India,” Proceedings of the National Academy of Sciences, Vol 109, No 29, pp 11652–56.

Kouser, Shahzad and Martin Qaim (2011): “Impact of Bt Cotton on Pesticide Poisoning in Smallholder Agriculture: A Panel Data Analysis,” Ecological Economics, Vol 70, No 11, pp 2105–13.

Kranthi, Kesavraj (2012): “Pesticide Use in Bt Cotton–reply,” online post, 28 June 2012, Development Dialogue: Voices of Unheard, Agrarian Crisis,….

Krishna, Vijesh V and Matin Qaim (2012): “Bt Cotton and Sustainability of Pesticide Reductions in India,” Agricultural Systems, Vol 107, pp 47–55.

Krishnan, Pramila and Manasa Patnam (2013): “Neighbours and Extension Agents in Ethiopia: Who Matters More for Technology Adoption?” American Journal of Agricultural Economics, Vol 96, No 1, pp 308–27.

Kumar, Deepak (2016): “Discrepancies in Data on Landholdings in Rural India: Aggregate and Distributional Implications,” Review of Agrarian Studies, Vol 6, No 1.

Munshi, Kaivan (2004): “Social Learning in a Heterogeneous Population: Technology Diffusion in the Indian Green Revolution,” Journal of Development Economics, Vol 73, No 1, pp 185–215.

Nagaraju, N, H M Venkatesh, H Warburton, V Muniyappa, T C B Chancellor and John Colvin (2002): “Farmers’ Perceptions and Practices for Managing Tomato Leaf Curl Virus Disease in Southern India,” International Journal of Pest Management, Vol 48, No 4, pp 333–38.

Nagrare, Vishlesh Shankar, Sandhya Kranthi, V K Biradar, N N Zade, V Sangode, G Kakde, R M Shukla, D Shivare, B M Khadi and Keshav Raj Kranthi (2009): “Widespread Infestation of the Exotic Mealybug Species, Phanacoccus Solenopsis (Tinsley) (Hemiptera: Pseudococcidae), on Cotton in India,” Bulletin of Entomological Research, Vol 99, No 5, pp 537–41.

Narayanan, Lalitha and Bharat Ramaswami (2007): “Pesticides Use Pattern among Cotton Cultivators in Gujarat,” Frontiers of Agricultural Development in Gujarat, Ravindra Dholakia (ed), Centre for Management in Agriculture, Indian Institute of Management: Ahmedabad. pp 77–98.

Narayanan, Lalitha and P K Viswanathan (2015a): India’s Tryst with Bt Cotton: Learning from the First Decade, New Delhi: Concept Publishing.

— (2015b): “Technology Diffusion and Adoption in Cotton Cultivation: Emerging Scenario in Gujarat,” AgBioforum, Vol 18, No 2, pp 209–20.

NSSO (2014): “Key Indicators of Situation of Agricultural Households in India,” National Sample Survey Office, Government of India, available at, last accessed on 16 October 2016.

Oerke, Eric-Christian (2006): “Crop Losses to Pests,” Journal of Agricultural Science, Vol 144, No 1, pp 31–43.

Pray, Carl and Latha Nagarajan (2014): “The Transformation of the Indian Agricultural Input Industry: Has It Increased Agricultural R&D?” Agricultural Economics, Vol 45, No S1, pp 145–56.

Ramaswami, Bharat (2002): “Understanding the Seed Industry: Contemporary Trends and Analytical Issues,” Indian Journal of Agricultural Economics, Vol 57, No 3, pp 417–29.

Ranga Rao, G V, V Rameshwar Rao, V P Prasanth, N P Khannal, N K Yadav and C L L Gowda (2009): “Farmers’ Perception on Plant Protection in India and Nepal: A Case Study,” International Journal of Tropical Insect Science, Vol 29, No 3, pp 158–68.

Ranganathan, Thiagu (2015): “Farmers’ Income in India: Evidence from Secondary Data,” Agricultural Situation in India, Vol 72, No 3, pp 30–70.

Shetty, Paddu Krishnappa (2004): “Socio-ecological Implications of Pesticide Use in India,” Economic & Political Weekly, Vol 39, No 49, pp 5261–67.

Shetty, Paddu Krishnappa, M Murugan and K G Sreeja (2008): “Crop Protection Stewardship in India: Wanted or Unwanted,” Current Science, Vol 95, No 4, pp 457–64.

Singh, Surendra Pal, Kavita Gupta and Sandeep Kumar (2014): “Judicious Use of Pesticides in Sustainable Crop Production and PGR Management,” National Bureau of Plant Genetic Resources E-Publication No NBP–14–02, viewed on 16 October 2016,

Stone, Glenn Davis (2011): “Field v Farm in Warangal: Bt Cotton, Higher Yields, and Larger Questions,” World Development, Vol 39, No 3, pp 387–98.

— (2012): “Constructing Facts: Bt Cotton Narratives in India,” Economic & Political Weekly, Vol 47, No 38, pp 62–70.

Taneja, Sonam (2017): “Why India Continues to Use Lethal Pesticides,” Down to Earth, 59173.

Be the first to comment

Leave a Reply

Your email address will not be published.


This site uses Akismet to reduce spam. Learn how your comment data is processed.