The economics of BT cotton production in India--a meta analysis.

Chakraborty, Kalyan

Abstract

India became the largest Bt cotton planter and the second largest cotton producer and exporter in the world in 2008. Since its introduction in 2002 the Bt adoption rate increased 152 fold and generated increased economic benefit of USD 3.2 billion to Indian cotton farmers. However, there is a continuous debate over the economic impact of Bt cotton on Indian farmers. This study undertook a vote counting based meta-analysis to 16 empirical studies on economics of Bt cotton in India and evaluated the inclusion and significance of three broad categories of factors determining the net return. The study shows that farm inputs and farm characteristics are the factors most often included in the studies. However, the economic outcome is most likely to be significantly determined by irrigation, use of pesticide, and insecticide. The study concludes that the empirical evidences suggest after its commercial release in 2002 Bt cotton has a significant positive impact on the economic conditions of the cotton growers in India.

Keywords: Transgenic, Bt-cotton, economic impact, biotechnology, bollworm

1. INTRODUCTION

The success story of transgenic (Bt cotton) cotton in India is spectacular. Within a span of six years the adoption rate of Bt cotton increased 150 fold, from 44,000 hectare in 2002 to 7.6 million hectare in 2008 (Table 1). It is estimated that the planted area for Bt cotton is expected to increase from 82 per cent in 2008-09 to 87 per cent in 2009-10 of total planted area for cotton in India (USDA, 2009). Bt cotton has generated US$ 3.2 million increased income for 5 million farmers, reduced the use of insecticide by half, nearly doubled the yield (568 kg/ha), and transformed India from an importer to a major exporter of cotton in the world (USDA, 2009).

According to an USDA estimate for MY 2009-10 India's cotton production is expected to exceed 25 million U.S. bales and India's exports are expected to be 7.8 million bales. The distribution of cotton in major growing states between 2002 and 2008 is reported in Table 2. For MY 2008, Maharastra with 3.13 million hectare represents 42 per cent share of all Bt cotton produced in India followed by Gujarat (18%), Andhra Pradesh (18%), Northern Zone (11%, Punjab, Haryana, and Rajasthan) and Madhya Pradesh (8%).

Although India is the second largest producer of cotton in the world and its planted area represents 25 per cent of global area of cotton India's cotton productivity is well below the world average (Table 3).

There are various factors contribution to the low yield for cotton such as, lack of irrigation, pest-pressure, and factors characterized by small-scale and resource-poor farming system. In India 65 per cent of cotton is cultivated under rain-fed conditions, thus, variability in yield is largely dependent on the arrival of monsoon (ISAAA, 2009). Another major factor is pest infestation, especially American bollworm causing annual loss of estimated US$ 300 million (APCoAB, 2006). This implies that every year farmers need to spend a large amount of money on pesticide. Although cotton constitutes 5 per cent of the planted area in India it consumes 50 per cent of expenditure (US$ 660 million) on insecticide (Manjunath, 2004). Due to a high dose and more frequent use of pesticides some insects such as, helicoverpa armigara has developed resistance against most of the commonly used chemicals. In an effort to reduce the use of high cost pesticide and provide protection against bollworm (helicoverpa armigara) the genetically modified Bt cotton (Bacillus thuringiensis) was approved by the government of India in 2002 for its commercial cultivation. Bt cotton contains a gene (cry1AC for the first event) of Bacillus thuringiensis (Bt) which is naturally occurring soil bacterium that produces a protein toxic to bollworm providing resistance to the plants and significantly reducing the need for chemical insecticide (APCoAB, 2008). Until 2008-09 government of India approved five events and over 280 hybrids for the commercial cultivation in different agro-climatic regions (Table 4).

Empirical studies around the world suggest that Bt technology is a major factor in boosting cotton productivity with additional positive effect on human health and environment due to reduced use of pesticide and increased farmers' net return (Zilberman et al. 2007; Huang et al. 2006; Shankar and Thirtle, 2005). Since the commercial introduction of Bt cotton in India over the last eight years several academic journals, independent researchers, and various news media published results from numerous studies assessing the economic impact of Bt cotton adoption on major cotton producing states.

The objective of the current study is to step back and take a stock of what we have learned from the past empirical studies on the agronomy of Bt technology adoption and the factors influencing the economic impact of the technology on India farmers. Empirical studies that use regression and/or some form of statistical analysis using farm level data are the major focus in this paper. Because many of these studies considered very few factors in the regression model, examined the technology in a limited geographic area, and/or used trial-farm data instead of market-data such studies cannot offer a general predictive understanding of the real economic impact of Bt technology. Generalization of results from any single study would be restrictive due to: (i) limited farmers/plots/villages sampled; (ii) time/crop-year considered; and (iii) variables included. (Pattanayaka, 2003)

The basic idea of this paper is to provide an answer to the question--whether a group of studies can collectively provide a better snapshot of the determinants of the economic impact of Bt cotton among Indian farmers instead of qualitative comparison of individual study results. Meta-analysis uses several quantitative methods for synthesizing results from several research methods including opinion surveys, correlation studies, experimental and quasi-experimental studies, and regression studies (Cook et al. 1992). The current study provides one of the simplest meta-analysis called the vote counting method--the analysis counts the number of studies that found a statistically significant result. For example, if 10 out of 16 empirical studies found a significant negative correlation between the use of insecticide and the yield from Bt cotton we can be fairly confident of the general correlation between a lower use of insecticide and Bt cotton yield. Vote counting provides a convenient starting point for a systematic assessment of multiple studies.

The organization of the current paper is as follows: next section provides a brief conventional review and background of the literature on the economics of Bt cotton in India followed by a section on the structure/organization of the meta-analysis for this study. The fourth section discusses the method and the vote counting results followed by a section on discussion and conclusion.

II. THE BACKGROUND

Following the formal release of Bt cotton in India in 2002 several empirical studies have reported higher yield and less labor and insecticide use for Bt cotton cultivation against non-Bt or other varieties (Barwale, et al. 2004; Bennett et al. 2004; Morse et al. 2005a, 2005b; Naik, 2001; Pemsl et al. 2004; Qaim and Zilberman, 2003; Ismael and Morse, 2005). However, there are studies that suggest no significant advantage in terms of yields and gross margin (Gala, 2005; Sahai and Rahman, 2004; Kumbamu, 2006). Qayam and Sakkhari (2005) found non-Bt farmers had a higher profit and lower cost than Bt farmers. Several studies in the literature provide a survey of studies on yield difference between Bt and non-Bt cotton among Indian farmers (Gruere, 2008; Crost et al. 2007; Smale et al. 2006; Ramaswami; 2005).

Most of these studies use different methods to compare the effect of Bt and non-Bt technology in terms of yields, pesticide use, seed costs, and net revenues. Smale et al. (2006) noted that the reported effects of Bt cotton studies mainly vary due to the extensive heterogeneity of growing environment, pest pressures, farmer practices, and social context. Further, Doss (2006) noted that it is difficult to compare the productivity gains between adopters and non-adopters of technologies because farmer's adoption decisions are correlated with other factors affecting productivity. For example, if a correlation between Bt adoption and high yields exists it could be due to a positive effect caused by the technology or it could be a self-selection effect (also called endogeneity) if only efficient farmers adopt the technology more frequently than their peers (Crost et al. 2007). As a result, the gains from Bt cotton cannot be generalized to all farmers, all states, and for all times. This inconsistency is the major reason behind the controversy over the use of Bt cotton and its benefits for Indian farmers during the early years of its introduction.

III. THE METHOD AND THE VOTE COUNTING

A systematic review of 25 empirical studies on yield/economics of Bt cotton adoption and the results from these studies is presented in the Appendix Table A, B. Table A summarizes the author(s), geographic location, methods, data, and the results from 17 peer-reviewed publications and Table B reports the same for 8 non-peer reviewed publications on the economic impact of Bt cotton adoption. The results vary widely across location, time, sample, methods, and studies however, a clear convergency is observed on the positive impact of the technology on Bt farmers. For each study the text and the tables are reviewed to identify the variables that fit the three general categories of determinants--(a) farm characteristics; (b) farm inputs; (c) farmer/household characteristics. However, the overview of the Bt adoption studies reported in Table 5 is restricted to a total of 16 peer-reviewed publications and manuscripts (in the review process) that have used either regression or some form of statistical methods analyzing economic impact of Bt technology.

Several variables within each broad category are identified and vote counting method is applied to each. For each study (under each category), first it is determined whether the variable was included in the study and then if it has a statistically significant positive or negative relationship with the yield or net revenue. If a study did not report the result for a particular variable the cell is marked as '0.' The second column of Table 5 shows the total number of studies that included each variable and the sixth column "included per cent" shows per cent of studies containing each variable. It is revealed from the table that the farm characteristics and farminputs are almost equally important determinants of Bt cotton adoption and hence its economic impact. In contrast, farmer/household characteristics are relatively less often included in those studies. Column 7 of Table 5, "significant per cent included" provides a better assessment of a variable in terms of its significance in those studies that included those variables. To this end farm characteristics (68%) and farm inputs (71%) are most likely to have a statistical significant effect. For individual variables such as, farm location, Bunny seeds, other costs, and family income are statistically significant in 100% of the models that include these variables.

However, we need to recognize two caveats regarding statistical significance as pointed out by McCloskey and Ziliak (1996). Statistical significance says nothing about the size of the influence of various factors in a production function estimates when Bt is used as a dummy dependent variable unless marginal probabilities are reported (for example, probit/togit models). Also, given the predisposition towards 'significance' in the literature researchers most often select the variables with significant coefficient in their models.

IV. WHAT FACTORS INFLUENCE THE ECONOMIC IMPACT OF BT COTTON?

Farm Characteristics

Farm characteristics include location, farm size, irrigation/rain-fed, and soil quality. The variable 'location' (used by one study) is associated with the farm's proximity to city/town with financial/credit availability. Among the two mostly used variables are 'farm size' (50%) and 'irrigation' (44%), the former is significant 50 per cent and the later is significant 71 per cent when included. Farm size measures the size of land holding for Bt cotton cultivation. Most of the studies found Bt technology is scale-neutral implying small and large farms have similar yield advantages. However, limited number of studies found opportunity cost of time saved from Bt cultivation (for off-farm activities) by the large farmers is higher than the small farmers. Soil quality was found a significant factor for half of the studies when included in the model. Rao and Dev (2007) found three types of soil such as, dark, medium, and light in Andhra Pradesh and 91 per cent of Bt cotton is cultivated in dark soil which has less salinity and is favorable to cotton cultivation. The variable 'irrigation' measures the proportion of land covered by irrigation which is devoted to Bt cotton cultivation including cost of providing the irrigation. Some studies found Bt cotton performs better under irrigation and generally use 20 per cent more water than conventional cotton (Sadeque, 2008). Most of the variables included in 'farm characteristics' are consistently positive implying positive influence on Bt yield when included in the study.

Farm Inputs

Cotton is both investment and labor intensive crop and some studies found plant protection is the most dominant cost (37%) of total variable cost. The list of variables included in 'farm inputs' is not exhaustive but representative of the farm-level control for costs. This list includes cost of spray for insecticide/pesticide (for sucking pest and bollworm), fertilizer/manure, integrated pest management (IPM), human labor, Bt, Bunny & Other seeds, and other costs. Pesticide, insecticide, and fertilizer/manure when included in the model are significantly correlated with Bt cotton yield for 89, 87, and 57 per cent, respectively. However, the direction of correlation is mixed depending upon the regional agro-climatic conditions and pest pressures. For example, in the production function estimates where yield from Bt cotton is considered as a dependent variable inputs such as, number of sprays for insecticide/pesticide or practice of IPM are expected to have a negative correlation. However when the growing conditions are such that the farmers face a high pest pressure and draught which significantly increases farmer's perceived risk of crop failure, results in considerable increase in the use of such inputs. It should be noted that Bt plots are in general higher yielding and have higher revenues than non-Bt plots although the costs are higher. However, overall the net revenues from Bt plots are significantly higher than non-Bt (Morse et al. 2007).

The variable 'human labor' is typically based on hours of labor used including spraying and harvest. Studies have used different form/type of labor such as, hired/ family, male/female, and landless/partial land owners. In most of the Indian provinces picking of cotton bolls during harvesting season is typically done by female hired workers. Depending upon the model used the direction of correlation between labor and Bt/nonBt yield varies. For example, labor requirements for bollworm insecticide-sprays are likely to be less for Bt plots but the labor requirements for harvesting are likely to be significantly higher due to higher yields for Bt cotton. The measure of labor endowment has been included 44 per cent of the studies and was found positive and significant for 57 per cent of the studies that included the variable. The three types of 'seed' variables included in Table 5 most of the studies measured 'Bt-seed' costs higher than conventional and other hybrid seed costs and when included in the model it is found to be positive and significant 71 per cent of the studies.

Farmer/Household

A large number of farmers in India grow both Bt and non-Bt cotton varieties hence, variables included in this sub-section provide some control for producer-related factors which might influence the preference of the Bt technology. The variables mostly included are--age, education, experience, family size, family income, and net revenue. Except for experience, all of these variables are significant in the range of 33 and 100 per cent when included. Farmer's education, family income, and net revenues have positive correlations with adoption of Bt technology. The variable 'family income' for farm household includes all income generated from agriculture other than cotton cultivation. Studies found non-Bt adopters had a higher family income from other agricultural crops while adopters had a higher income from cotton and livestock, as a result, adopters generally concentrated more on cotton than non adopters (Morse, et al. 2007). The studies that included 'net revenue' found higher yield and net revenue over-compensated the higher cost of Bt cultivation leading to a positive correlation between Bt yield and 'net revenue.' Family size measures the number of adult members in the family who could potentially join as farm labor. Studies that included the variable (19%) found to have mostly insignificant correlation with Bt yields. One possible reason might be that for most parts of India farmers generally hire seasonal and landless labors for cotton cultivation rather than using family labors who are involved some non-farm income activities.

V. SUMMARY AND CONCLUSIONS

This paper reviews 25 studies 16 of which uses some form of statistical analysis/model related to economic impact of Bt cotton in India. Three key determinants of Bt technology adoption such as, farm characteristics, farm inputs, and farmer/household characteristics are identified. These determinants provide a useful organizing framework for conceptual and empirical evaluations of the economics of Bt cotton in India. After reviewing 16 empirical studies a meta-dataset of specific variables within three broad categories of determinants of Bt adoption is developed. Using vote counting based on meta-analysis to this dataset this study provides a richer picture of economics of Bt cotton than can be developed from qualitative comparison of the individual study features and results. From this point of view the current study generates two distinct meta-statistics on the empirical literature related to economics of Bt adoption i.e., inclusion and influence of factors. The study found all three categories of determinants are most likely to be included in the analysis of Bt technology adoption--especially over 50 per cent of the studies include variables such as farm size, pesticide, and insecticide.

When the 'significance' criterion is used, most of the studies found that Bt yield effect is mostly statistically correlated to 'farm inputs' category. Variables such as, irrigation, pesticide, insecticide, and Bt-seed have the greatest statistical power implying they are statistically significant most of the time when included in the model. However, the author recognizes some of the limitations of this study. For example, the three broad categories of determinants and the specific variables within the set reported in column seven (Table 5) (significant per cent included) is not a perfect overlap. Further, the set of impact studies included in the meta-analysis may not represent a random sample of the true population hence, might be bias because the meta-dataset only considered the peer-reviewed published and unpublished empirical studies that used statistical analysis/model. This study did not include the studies that are based on informal surveys, in-depth interviews, and focus groups with participating farmers measuring cost and benefits from Bt cotton adoption.

The study found the direction of the correlations is mostly consistent for the 'farm characteristics' but it is not consistent for 'farm inputs' and 'farmer/household' categories. Similar conclusions can be drawn in regard to the statistical significance of for specific variables included in the meta-analyses. Considering the growing interest among the academic researchers, policy makers, and independent research agencies around the world on India's Bt cotton success story, it is time to overview the stock of the literature. Reviewing 25 Bt cotton adoption studies within a span of seven years (2002-09) and conducting a simple meta analysis of 16 empirical studies the present study provides a first step towards this direction. The general conclusion from this study is that the economic impact of Bt cotton is mostly determined by the factors such as, irrigation and the use of pesticides, insecticides, and Bt-seeds.

Appendix

Table A
Summary of Peer Reviewed Published Studies on the Economic Impact of
Bt Cotton on Farmem in India

Author/Study/Date     Location          Sample/Data

Herring (2009)        India             Essay

Subramanian and       Maharastra,       Farm statistical
Qaim (2009)           Karnataka,        survey, 341
                      Andhra Pradesh,   farmers: 133-Bt
                      Tamil Nadu        plots; 301-non-Bt
                                        plots 2002-03

Maertens (2009)       Andhra Pradesh    3 Villages, 246
                                        households, panel
                                        data 2001-2008

Herring (2008)        Andhra Pradesh    One village, 2006

Rao, Rao,             Andhra Pradesh,   180 Farmers
Naraiah,              Guntur            surveyed
Malathi, and                            practicing IPM
Reddy (2007)                            with and without
                                        Bt and non-
                                        IPM , 2004-05

Roy, Herring,         Gujarat, 4-       45 Farmers were
and Geisler           districts:        interviewed; 35-
(2007)                Junagadh,         Bt and 10-non-Bt
                      Bhavnagar,        farmers,
                      Sabarkantha,      Summer, 2004
                      Vadodara

Qaim, Subramanian,    Maharastra,       Farm statistical
Naik, and             Karnataka,        survey, 341
Zilberman (2006)      Andhra Pradesh,   farmers: 133-Bt
                      Tamil Nadu        plots; 301-non-
                                        Bt plots 2002-03

Morse, Bennett,       Maharastra,       Farm statistical
Ismael (2007)         Jalgaon           survey: 137-
                                        Bt plots; 95-
                                        non-Bt plots.
                                        2002-03

Crost, Shankar,       Maharastra,       Farm level survey
Bennett, and          Jalgaon           data from 6
Morse (2007)                            villages: 718-plots;
                                        338-farmers;
                                        84-Bt only;
                                        122-non-Bt only;
                                        134-both, 2002-03

Kambhampati,          Maharastra        Survey by Mahyco:
Morse, Bennett,                         Khandesh,
and Ismael (2006)                       Marathawada,
                                        Vidarbha
                                        7,751-plots 20021,
                                        580-plots 2003

Bennett,              Maharastra;       7,751-plots 2002
Kambhampati,          Gujarat,          1,580-plots 2003
Morse, and            Madhya Pradesh,
Ismael (2006)         Karnataka

Kambhampati, Morse,   Gujarat           Interview: 22 Up-
Bennett, and                            stream and down-
Ismael (2005)                           stream companies:
                                        Sabarkantha,
                                        Ahmedabad,
                                        Gandhinagar,
                                        Mahesana, 2004

Morse, Bennett,       Gujarat           Farm statistical
and Ismael                              survey: 622.
(2005a,b)                               farmers; 306-plots
                                        official Bt;
                                        169-plots
                                        unofficial Bt;
                                        151- non-Bt

Bennett, Ismael,      Maharastra        Survey by Mahyco:
Kambhampati,                            Khandesh,
and Morse (2004)                        Marathawada,
                                        Vidarbha 7,751-
                                        plots 20021,580
                                        -plots 2003

Pemsl, Waibel,        Karnataka         Survey data:
and Orphal (2004)                       Dharwad and
                                        Belgaum. 100
                                        Bt and Non-Bt
                                        farms; Irrigated
                                        -44, Non-irr-66;
                                        2002-03

Barwale, Gadwal,      Andhra Pradesh,   Survey by Mahyco:
Zehr, and Zehr        Gujarat,          1069-farms,
-2004                 Karnataka,        2002-03
                      Madhya Pradesh,
                      Maharastra,
                      Tamil Nadu

Qaim and Zilberman    Maharastra,       157-Trial Plots
(2003)                Madhya Pradesh    from 25 districts,
                      Tamil Nadu        2001

Author/Study/Date     Method Used

Herring (2009)        Narrative

Subramanian and       Social Accounting
Qaim (2009)           Matrix (SAM)
                      Multiplier model

Maertens (2009)       Probit model of
                      Quasi-Panel

Herring (2008)        Narrative

Rao, Rao,             Two-way contingency
Naraiah,              table testing
Malathi, and          correlation between
Reddy (2007)          Bt adoption and
                      practice of IPM.
                      Regression analysis.

Roy, Herring,         Survey data analysis
and Geisler
(2007)

Qaim, Subramanian,    Regression analysis:
Naik, and             Estimating translog
Zilberman (2006)      production function

Morse, Bennett,       One-way ANOVA
Ismael (2007)         table, comparison
                      between Bt adopters
                      and non-adopters

Crost, Shankar,       Estimated Cobb-
Bennett, and          Douglas production
Morse (2007)          function using panel
                      data fixed effect
                      model

Kambhampati,          Estimated frontier
Morse, Bennett,       production function
and Ismael (2006)     using panel data

Bennett,              Estimated Cobb-
Kambhampati,          Douglas production
Morse, and            function using
Ismael (2006)         2-year panel data

Kambhampati, Morse,   Survey data analysis
Bennett, and
Ismael (2005)

Morse, Bennett,       Farm survey
and Ismael            analysis
(2005a,b)             and Econometric
                      Stepwise
                      Regression

Bennett, Ismael,      Farm survey
Kambhampati,          analysis
and Morse (2004)

Pemsl, Waibel,        Stochastic partial
and Orphal (2004)     budgeting;
                      Multiple linear
                      regression

Barwale, Gadwal,      Farm survey
Zehr, and Zehr        Analysis
(2004)

Qaim and Zilberman    Regression analysis
(2003)                of pesticide use
                      function and
                      production function
                      with Pesticide

Author/Study/Date     Results

Herring (2009)        Concludes Bt cotton has proved a scale-neutral
                      partial solution to a pressing agronomic problem,
                      bollworm destruction. Evidences found in the
                      literature that the technology is pro-poor, which
                      implies larger harvest creates more work for
                      labourers.

Subramanian and       Found Bt cotton is associated a substantial
Qaim (2009)           overall generation of rural employment
                      especially among female workers. The returns to
                      saved management time in alternative activities
                      are higher for large farmers hence the large
                      farmers benefit more from Bt adoption in an
                      economy-wide framework.

Maertens (2009)       The magnitude of the coefficient measuring the
                      change in probability to adopt Bt cotton after
                      having heard of one additional institutional
                      source is about 4-40 times the magnitude of the
                      coefficient of learning from one additional
                      fellow farmer.

Herring (2008)        The study found that "innocent error in fording
                      Bt cotton failure, revolving around germplasm,
                      spurious seeds, and trait differentiations."
                      Economics and environmental integrity
                      coincided--environmental toxin declined with
                      Bt technology

Rao, Rao,             The study found plant protection expenditure is
Naraiah,              lowest when IPM is practiced with non-Bt
Malathi, and          varieties. Practice of IPM would reduce the cost
Reddy (2007)          of plant protection and increase the net return.
                      No significant reduction of plant protection cost
                      was observed when Bt varieties were adopted
                      without IPM.

Roy, Herring,         Found similar to conventional cotton farmers, Bt
and Geisler           cotton farmers are more likely to use saved seeds
(2007)                (34%) and both large and small farmers used
                      'loose' seeds, officially approved and illegal
                      transgenic seeds. Laborers are benefited in
                      terms of amount of wages earned due to higher
                      yields from Bt cotton.

Qaim, Subramanian,    Found Bt plots had less spray (2.6 times), higher
Naik, and             yields (34%), and higher profits (Rs. 2,161) more
Zilberman (2006)      than conventional cotton plots. Andhra Pradesh
                      suffered losses because of high pesticide use and
                      severe draught, and spurious Bt seeds.

Morse, Bennett,       On average Bt plots are higher yielding (27-42%)
Ismael (2007)         and have higher revenue (34-53%) than non-Bt
                      plots. Overall, the gross margins of Bt plots are
                      significantly higher (59-66%) than non-Bt.
                      Between adopters and non-adopters, about the
                      half of the increase in yield for Bt adopters is
                      due to 'farmer effect' and the rest is due to
                      Bt-trait.'

Crost, Shankar,       Found efficient farmers adopt Bt cotton at a
Bennett, and          higher rate than less efficient farmers. When
Morse (2007)          self selectivity bias is controlled, Bt cotton
                      farmers are found to have 31% yield advantage
                      over non-Bt farmers.

Kambhampati,          Average yield for Bt plot increased by 45% and
Morse, Bennett,       63% for 2002-03 compared to non-Bt plot.
and Ismael (2006)     Increased performance of Bt could be any of the
                      following factors: Bt gene, the base cotton
                      variety used performed well in local conditions
                      and the farmers might be more efficient.

Bennett,              For 2002, the production function estimates
Kambhampati,          indicate Bt technology has a 33% positive effect
Morse, and            on yield per acre after allowing for influence of
Ismael (2006)         insecticide sprays, soil type, and irrigation. For
                      both Bt and non-Bt varieties yield rise under
                      irrigation however, the impact of irrigation is
                      highly significant under Bt variety.

Kambhampati, Morse,   Companies sold less pesticide as a result of
Bennett, and          farmers used Bt cotton. Due higher quality Bt
Ismael (2005)         cotton need to be separated from other cotton for
                      higher price. Farmers are less reliant on credit
                      for purchasing pesticide and higher price for Bt
                      cotton would benefit the farmers most.

Morse, Bennett,       Average yield benefits 37% for MECH-12, 20%
and Ismael            for MECH-162, 14% for unofficial Fl, -5% for
(2005a,b)             unofficial F2, Gross margin also follow the same
                      order - +132%, +73%, +37%, +20%.

Bennett, Ismael,      For Bt cotton adopters the cost of bollworm spray
Kambhampati,          is lower by 72% in 2002 and 83% in 2003.
and Morse (2004)      Average yield for Bt cotton increased by 45%
                      (2002) and 63% (2003) over non-Bt. The average
                      gross margin for Bt was higher by 74% (2002)
                      and 49%(2003) than non-Bt.

Pemsl, Waibel,        The study found under non-irrigated conditions
and Orphal (2004)     there would be no reason for farmers to adopt Bt
                      cotton. Under irrigated conditions stochastic
                      dominance shows pesticide alone is the superior
                      strategy. The regression analysis found pest
                      pressure and cotton output price largely
                      determine the yield.

Barwale, Gadwal,      On average Bt cotton yields were 30% higher
Zehr, and Zehr        than non-Bt hybrids and Bt had a cleaner quality
-2004                 and color. Bt farms had 1.93 less spray for
                      bollworm when combined had an additional Rs.
                      18,000/ha compared to non-Bt.

Qaim and Zilberman    Average yields for Bt hybrids exceeded yields for
(2003)                non-Bt and popular checks by 80% and 87%
                      respectively. Bt hybrids were spread three times
                      less against bollworm than non-Bt and popular
                      checks.

Appendix

Table B
Summary of Non-peer Reviewed Studies/Reports and Working Papers

ISAAA (2009)      India          State level           Farm data
                                 published data,       analysis
                                 2002-2008

Krishna,          Maharastra,    Personal interview:   Cobb-Douglas
Zilberman,        Karnataka,     362-farmers-2004      production
and Qaim (2009)   Andhra         344-farmers-2006      function (plot);
                  Pradesh,                             Tobit damage
                  Tamil Nadu                           function (farm)

Subramanian and   Maharastra,    Farm statistical      SAM-Village
Qaim (2009)       Karnataka,     survey, 341           level
                  Andhra         farmers: 133-Bt       microdata
                  Pradesh,       plots; 301-non        simulation
                  Tamil Nadu     -Bt plots 2002/3
                                 Kanzara, 2004

Rao and Dev       Andhra         Stratified sample     Regression
(2008)            Pradesh:       survey, 437-Bt        analysis:
                  Warangal,      farmers; 186-         Estimated
                  Nalgonda,      Non-Bt; 2004/5        production
                  Guntur.        and 2006/7            function.
                  Kurnool                              Farm survey
                                                       analysis

Rao and Dev       Andhra         Stratified sample     Regression
(2007)            Pradesh:       survey, 437-Bt        analysis:
                  Warangal,      farmers; 186-         Estimated
                  Nalgonda,      Non-Bt; 2004/5        production
                  Guntur,                              function. Farm
                  Kurnool                              survey analysis

                  All major      IMRB-Survey of        Farm data
                  cotton         5,950 farmers,        analysis
                  growing        60% Bt adopters,
                  states in      2006/7
                  India

Gandhi and        Gujarat,       Farm statistical      Regression
Namboodiri        Maharastra,    survey, 694           analysis and
(2006)            Andhra         farms 2004            farm survey
                  Pradesh,                             analysis
                  Tamil Nadu

Morse, Bennett,   Maharastra     7,751-plots in        Kruskal-Wallis
and Kambhampati   (Khandesh,     2002; 1,580-plots     non-parametric
(2005)            Marathawada,   in 2003; 22-          tests; Frontier
                  Vidarbha);     companies (seed/      production
                  Gujarat, AP,   input/textile)-       function
                  Karnataka      2004

ISAAA (2009)      Between 2002 and 2008 Bt cotton acreage
                  increased by 42%, the consumption of pesticide
                  decreased by 22%, and the insecticide use
                  reduced by 20%. Since the introduction of Bt
                  technology in 2002, the yield increased from 308
                  kg/ha to 580 kg/ha in 2008.

Krishna,          Bt technology and varietal richness increased the
Zilberman,        mean yield and decreased yield variability.
and Qaim (2009)   Farmers' economic welfare can be greatly
                  enhanced by introducing diversity among GM
                  varieties.

Subramanian and   Bt cultivation increases the aggregate return to
Qaim (2009)       labor by 42% and the return to hired female
                  labourer is 55%. For poor landless labourers,
                  their household income increases by 134% under
                  Bt cultivation than under non-Bt.

Rao and Dev       Compared to 2004/5, yield for Bt cotton in 2006/7
(2008)            increased by 42% and the use of chemical
                  insecticide decreased by 56%. There was a net
                  gain to the cotton farmers for Rs. 7,122 crores in
                  2006/7. Benefits in rain fed farming were lower
                  and were not statistically significant.

Rao and Dev       Study finds that the overall Bt cotton decreased
(2007)            pesticide cost by 18%, increased total cost by
                  17%, and increased income by 83%. Regression
                  estimates find 36% advantage of Bt over non-Bt
                  and Bt has a positive effect on farm employment.
                  However, due to bad weather in 2004/5 in
                  Andhra Pradesh total net income was negative
                  for both Bt and Non-Bt.IMRB (2006)

                  For India as a whole, Bt increases yield by 50%,
                  reduces pesticide costs by 32%, and increases
                  profit by 162%. All states gain from Bt cotton, the
                  biggest relative profit increase is in Maharastra
                  (375%), Andhra Pradesh (217%), Gujarat (198%)
                  and Madhya Pradesh (156%).

Gandhi and        Regression analysis found Bt provides 31% yield
Namboodiri        gains but costs 7% higher for a higher profit of
(2006)            74% on average. However, the percentage
                  increase of yield, costs, and profits vary from
                  state to state. Survey analysis also found almost
                  all farmers in Andhra Pradesh and Tamil Nadu
                  want to plant Bt in the future.
Morse, Bennett,   The average difference in yield between Bt and
and Kambhampati   non-Bt was 45% in 2002 and 63% in 2003. Two
(2005)            most important benefits of growing Bt cotton are
                  increasing household income and reducing
                  insecticide use. Higher profit is not due to
                  reduced costs, but due to higher yield from Bt
                  variety.

References

APCoAB--Asia-Pacific Consortium on Agricultural Biotechnology- A Status Report. Dec. (2006), New Delhi, India.

Barwale R. B., Gadwal V. R., Zehr U., & Zehr B. (2004), Prospects for Bt Cotton Technology in India, AgBio Forum, 7(1&2): 23-26.

Bennett, R. M., Ismael, Y., Kambhampati, U., & Morse, S. (2004), Economic Impact of Genetically Modified Cotton in India, AgBio Forum, 7(3): 1-5.

Bennett, R., U. Kambhampati., S. Morse., and Y. Ismael (2005), Perceptions of the Impacts of Genetically Modified Cotton Varieties: A Case Study of Cotton Industries in Gujarat, India, AgBio Forum, 8(2&3): 161-171.

Cook T., Cooper H., Corday D., Hartman H., Hedgel L., Light R., et al. (1992), M eta-analysis for Exploration: A Casebook, Russell Sage Foundation. New York.

Crost, B., B. Shankar, R., R. Bennett., and S. Morse (2007), Bias from Farmer Self Selection in Genetically Modified Crop Productivity Estimates: Evidence from Indian Data, Journal of Agricultural Economics, 58(1): 24-36.

Doss, C. R. (2006), Analyzing Technology Adoption using Micro-studies: Limitations, Challenges, and Opportunities for Improvement, Agricultural Economics, 34: 207-219.

Dev, S. M., and N. C. Rao (2007), Socioeconomic Impact of Bt cotton. CRSS-Monograph. No. 3. Center for Economic and Social Studies, Hyderabad, India at www.cess.ac.in.

Gala, R. (2005), India's Bt cotton fraud (Press Release). London: Institute of Science in Society, http://www.i-sis.org.uk/IBTCF.php

Gandhi, V. P., and N. V. Namboodiri: The Adoption and Economics of Bt Cotton in India: Preliminary Results from a Study. Indian Institute of Management, Ahmedabad, India. Working Paper No. 2006-09-04.

Gruere, G. P.; P. M. Bhatt., and D. Sengupta (2008), Bt Cotton and Farmer Suicides in India: Reviewing the Evidences. International Food Policy Research Institute, USA.

Herring, R. J. (2009), Persistent Narratives: Why is the 'Failure' of Bt Cotton in India's Story Still with us? AgBio Forum, 12(1): 14-22.

Herring, R. J. (2008), Whose Numbers Count? Probing Discrepant Evidence on Transgenic Cotton in the Warangal District of India. International Journal of Multiple Research Approaches, 2:145-159.

Huang, J., H. Lin., R. Hu., S. Rozelle., and C. Pray (2006), Eight Years of Bt Cotton in Farmer Fields in China: Is the Reduction of Insecticide use Sustainable? Working Paper, Center for Chinese Agricultural Policy, Chinese Academy of Science.

IMRB--International Business and Industrial Research Division. Bt cotton cultivation in India--A Report to Understand the Awareness, Perceptions, and Acceptability of Bt Cotton Seeds among Cotton Growing Farmers. Prepared by AICBA--2006.

ISAAA--International Service for the Acquisition of Agri-biotech Applications--The Dawn of a New Era: Biotech crops in India. Brief No. 39. 2009 New Delhi, India.

Kambhampati, U., S. Morse., R. Bennett., and Y. Ismael (2006), Farm-level Analysis of Cotton Production in Maharastra, Outlook on Agriculture, 34(4): 291-297.

Krishna, V., D. Zilberman., and M. Qaim (2009), GM Technology Adoption, Production Risk and Onfarm Varietal Diversity. Paper Presented at the AAES and ACCI Annual Conference, Milwaukee, July 26-28, 2009.

Manjunath, T. M. (2006), Bt Cotton: Protein Expression in Leaves is Most Critical. Current Science, 90: 278-279.

Maertens, A. (2009), Understanding the Effects of Social Learning and Social Pressures in Bacillus Thuringiensis Cotton Technology Adoption Decisions in India. Working Paper, Cornell University, Department of Applied Economics and Management.

McCloksey D., and Ziliak, S. (1996), The Standard Error of Regression. Journal of Economic Literature, 35: 97-114.

Morse, S., Bennett, R. M., & Ismael, Y. (2005a), Comparing the Performance of Official and Unofficial Genetically Modified Cotton in India, AgBio Forum, 8(1): 1-6.

Morse, S., Bennett, R. M., & Ismael, Y. (2005b), Genetically Modified Insect Resistance in Cotton: Some Economic Impacts in India, Crop Protection, 24(5): 433-440.

Morse, S., R. Bennett and Y. Ismael (2007), Isolating 'Farmer' Effect as a Component of the Advantage of Growing Genetically Modified Varieties in Developing Countries: A Bt Cotton Case Study from Jalgaon, India. Journal of Agricultural Science, 145: 491-500.

Morse, S., R. Bennett., and U. Kambhampati (2005), Monitoring the Impact of GM Cotton in India. Working Paper/Project. University of Reading, UK.

Naik, G. (2001), An Analysis of Socioeconomic Impact of Bt Technology on Indian Cotton Farmers. Center for Management in Agriculture, Indian Institute of Management, Ahmedabad, India.

Pattanayak. S. K., D. E. Mercer, E. Sills, and J. C. Yang (2003), Taking Stock of Agroforestry Adoption Studies, Agroforestry, 57: 173-186.

Pemsl, D., Waibel, H., and Orphal, J. (3004), A Methodology to Assess the Profitability of Bt Cotton: Case Study Results from the State of Karnataka, India. Crop Protection, 23(12): 1249-57.

Qaim, M. and Zilberman, D. (2003), Yield Effect of Genetically Modified Crops in Developing Countries, Science, 299 (5608): 900-902.

Qaim, A., and K. Sakkhari (2005), Bt Cotton in Andhra Pradesh - A Three-year Assessment. AP, P 49.

Qaim, M., A. Subramanian., G. Naik., and D. Zilberman (2006), Adoption of Bt Cotton and Impact Variability: Insights from India, Review of Agricultural Economics, 28(1): 48-58.

Ramaswami, B. (2005), Indian Experience with GM Crops: Socioeconomic Impacts and Institutional Change. Paper Presented at the Workshop, Italy, May 31, 2005. Belfercenter. STPP, Kennedy School of Govt.

Rao, C. A. R., M. S. Rao., P. Naraiah., B. Malathi., and Y. V. R. Reddy (2007), Profitability of Cotton on a Pest Management Continuum in Guntur District of Andhra Pradesh, Agricultural Economics Research Review, 20, July, 273-282.

Rao, N.C., and S. M. Dev (2008), Biotechnology in Indian Agriculture - Evidences from Panel Studies on Bt Cotton. Paper Presented at The Institute of Economic Growth, New Delhi, India, Sept. 2324.

Roy, D., R. J. Herring., and C. C. Geisler (2007), Naturalising Transgenics: Official Seeds, Loose Seeds, and Risk in the Decision Matrix of Gujarati Cotton Farmers, Journal of Development Studies, 43(1): 158-176.

Sadeque, N. (2008), After Disastrous Track Record in 40 Countries Bt Cotton is 'Welcomed' in Pakistan. Financial Post Online at www.dailyf-post.com/btcotton.htm

Sahai, S. and S. Rehman (2004), Bt Cotton 2003-04 Fields Swamped with Illegal Variants, Economic and Political Weekly, June 26, 39(24): 2673-74.

Kumbamu, A. (2006), Ecological Modernization and the 'Gene Revolution': The Case Study of Bt Cotton in India, Capitalism Nature Socialism, 17(4): 7-31.

Shankar, B., and C. Thirtle (2005), Pesticide Productivity and Transgenic Cotton Technology: The South African Smallholder Case, Journal of Agricultural Economics, 56(1): 97-116.

Smale, M., P. Zambrano., and M. Cartel (2006), Bales and Balance: A Review of the Methods used to Assess the Economic Impact of Bt Cotton on Farmers in Developing Economies, AgBio Forum, 9(3): 195-212.

Subramanian, A., and M. Qaim (2009), Village-wide Effects of Agricultural Biotechnology: The Case of Bt Cotton in India, World Development, 37(1): 256-67.

USDA. Economic Research Services Briefing Room Online at www.ers.usda.gov/briefing/India/policy.htm

Zilberman, D., H. Ameden., and M. Qaim (2007), The Impact of Agricultural Biotechnology on Yields, Risks, and Biodiversity in Low Income Countries, Journal of Development Studies, Jan, 43(1): 6378

KALYAN CHAKRABORTY

Emporia State University, Emporia

* Paper presented at the International Conference on Business and Economic Issues Organized by the Indian Journal of Economics and Business, Dec 19-21, New Delhi, India

Table 1
Commercialization of Bt-Cotton in India

Crop Year   Area Bt-Cotton   Per cent of Total   Number of
                 (Hectare)         Cotton Area     Farmers

2002/03             44,500                0.58      54,000
2003/04            100,000                1.31          --
2004/05            500,000                5.57     300,000
2005/06          1,300,000               14.38   1,000,000
2006/07          3,800,000               41.27   2,300,000
2007/08          6,200,000               68.88   3,800,000
2008/09          7,600,000               82.00   5,000,000

Source: ISAAA--2008

Table 2
Adoption of Bt Cotton in India, by Major States, 2002 to 2008
(000 Hectares)

State             2002   2003   2004    2005    2006    2007    2008

Maharastra          25     30    200     607   1,840   2,800   3,130
Andhra Pradesh       8     10     75     280     830   1,000   1,320
Gujarat             10     36    122     150     470     980   1,360
Madhya Pradesh       2     13     80     146     310     500     620
Northern Zone *     --     --     --      60     215     682     840
Karnataka            3      4     18      30      85     145     240
Tamil Nadu           2      7      5      27      45      70      90
Others                                             5       5       5
Total               50    100    500   1,300   3,800   6,200   7,605

* Punjab, Haryana, Rajasthan Source: ISAAA, 2008

Table 3
Top-10 Cotton Producers in the World, 2008-09

Country        Haruested Area   Yield (lbs/   Production (480lbs/
                  (000 acres)         acre)    bales) (000 bales)

China                  14,703         1,169                35,800
India                  23,161           466                22,500
USA                     7,569           813                12,815
Pakistan                7,166           603                 9,000
Uzbekistan              3,509           629                 4,600
Brazil                  2,051         1,287                 5,500
Turkey                    853         1,182                 2,100
Australia                 405         1,719                 1,450
Turkmenistan            1,483           437                 1,350
Greece                    618           893                 1,150

Source: NCC of America (Cropdata)

Table 4
Approval of Bt Events and Hybrids for Commercial Cultivation

Year     Events                                        No of Hybrid
                                                          Varieties

2002/3   MMBL's Bollgard I                                        3
2003/4   MMBL's Bollgard I                                        3
2004/5   MMBL's Bollgard I                                        4
2005/6   MMBL's Bollgard I                                       20
2006/7   MMBL's Bollgard I & II; JK Seed's Event 1,
         Nath Seed's GFM Event                                   62
2007/8   MMBL's Bollgard I & II; JK Seed's Event 1,
         Nath Seed's GFM Event                                  162
2008/9   MMBL's Bollgard I & II; JK Seed's Event 1,
         Nath Seed's GFM Event; and CICR Event                  281

Source: ISAAA, 2008

Table 5
Votes on the Determinants of Bt Cotton Adoption--Samples with
Statistical Models (16 studies)

                       Numbers      Numbers
                                  Significant
Variables              Included   Pos.   Neg.   Insignificant

Farm Characteristics
  Location                    1      1      0               0
  Farm size                   8      3      1               4
  Irrigation                  7      5      0               2
  Soil quality                4      2      0               2
Farm Inputs
  Pesticide                   9      5      3               1
  Insecticide                 8      2      5               1
  Fertilizer/Manure           7      3      1               3
  IPM Technology              1      0      1               0
  Human labor                 7      4      0               3
  Bt-Seed                     7      4      1               2
  Bunny & other seed          3      1      0               2
  Others                      3      2      1               0
Farmer /Household
  Age                         6      0      2               4
  Experience                  3      0      0               3
  Education                   7      2      1               4
  Family size                 3      0      1               2
  Net revenue                 4      4      0               0
  Family income               2      2      0               0

                                  Significant
                        Percent       Percent
Variables              Included      Included

Farm Characteristics         31            68
  Location                    6           100
  Farm size                  50            50
  Irrigation                 44            71
  Soil quality               25            50
Farm Inputs                  35            71
  Pesticide                  56            89
  Insecticide                50            87
  Fertilizer/Manure          44            57
  IPM Technology              6             0
  Human labor                44            57
  Bt-Seed                    44            71
  Bunny & other seed         19            33
  Others                     19           100
Farmer /Household            26            62
  Age                        37            33
  Experience                 19             0
  Education                  44            43
  Family size                19            33
  Net revenue                25           100
  Family income              12           100