Can GM-Technologies Help the Poor? The Impact of Bt Cotton in Makhathini Flats, KwaZulu-NatalIntroductionProgress with GM cropsGM crops in South AfricaCotton in Makhathini FlatsThe Makhathini Flats surveyCotton productionCharacteristics of the smallholders and adoptionAgricultural inputs and nonagronomic constraints to cotton productionAgronomic constraints to cotton productionReasons for adoption and future intentionsAdoption, yield, cost, profitability and efficiency of adopters and nonadoptersAdoption modelFarm accounting measures of the impact of Bt cottonProduction efficiency of adopters and nonadopters: stochastic production frontiersSocial benefits: environment and healthConclusionsReferencesCan GM-Technologies Help the Poor? The Impactof Bt Cotton in Makhathini Flats, KwaZulu-NatalCOLIN THIRTLE, LINDIE BEYERSImperial College London, UKUniversity of Pretoria, Pretoria, South AfricaYOUSOUF ISMAELUniversity of Reading, UKandJENIFER PIESSE*KingÕs College London, UKUniversity of Pretoria, Pretoria, South AfricaSummary. — The results of a two-year survey of smallholders in Makhathini Flats, KwaZulu-Natalshow that farmers who adopted Bt cotton in 1999–2000 benefited according to all the measuresused. Higher yields and lower chemical costs outweighed higher seed costs, giving higher grossmargins. These measures showed negative benefits in 1998–99, which conflicts with continuedadoption, but stochastic efficiency frontier estimation, which takes account of the labor saved,showed that adopters averaged 88% efficiency, as compared with 66% for the nonadopters. In1999–2000, when late rains lowered yields, the gap widened to 74% for adopters and 48% fornonadopters.Ó 2003 Elsevier Science Ltd. All rights reserved.Key words — KwaZulu-Natal, South Africa, Bt cotton, stochastic frontiers, efficiency1. INTRODUCTIONAlthough there has been an adverse reactionto genetically modified (GM) plants in Europeand some developing countries, others, parti-cularly China and South Africa, have followedthe United States in rapidly adopting GMtechnologies. For Bt cotton, there is now evi-dence from China (Pray, Ma, Huang, & Qiao,2001) and Mexico (Traxler, Godoy-Avila,Falck-Zepeda, & Espinoza-Arellano, 2002) ofincreased profits and positive impacts on theenvironment and health, due to reduced pesti-cide use. This study, based on a sample surveyand farm records, investigates the farm-levelimpacts of the first use of Bt cotton by Africansmallholders.Section 2 of this paper provides some back-ground on GM crops, especially in SouthAfrica and on Bt cotton, particularly inMakhathini Flats. Section 3 analyzes the sam-ple of 100 farmers, in order to give a broadWorld Development Vol. 31, No. 4, pp. 717–732, 2003Ó 2003 Elsevier Science Ltd. All rights reservedPrinted in Great Britain0305-750X/03/$ - see front matterdoi:10.1016/S0305-750X(03)00004-4www.elsevier.com/locate/worlddev*Earlier versions of this paper were presented at the 5thInternational Conference on Biotechnology, Science andModern Agriculture: a New Industry at the Dawn of theCentury, Ravello, Italy, June 2001; at a workshop onBiotechnology and Rural Livelihood-Enhancing theBenefits, ISNAR, the Hague, Netherlands, June 2001;and at Future Opportunities in Rural Africa, a work-shop organized by IFPRI for USIAD, Washington DC,November 2001. We thank Marnus Gouse and Lwandleand Knassi Mqadi for help with the survey, VictoriaMcKenzie for research assistance, Monsanto, VUNISA,Cotton SA and Innovation Biotechnology for theirgenerous help, the Department for International De-velopment, London and the Rockefeller Foundation forfinancial support and three anonymous referees for theircomments. Final revision accepted: 24 November 2002.717overview of the characteristics of the small-holders, but specifically concentrating on iden-tifying the innate differences between adoptersand nonadopters. Section 4 begins by reportingan adoption model, which is followed by farmaccounting results, comparing yields, inputlevels, costs and gross margins of the adoptersand nonadopters. A first major contribution ofthis paper is that with data for two seasons, theeffects of the technology can be separated fromthe innate differences between adopters andnonadopters, so that the comparisons are notbiased. But, both yields and gross margins arepartial measures of efficiency, which fail to takeaccount of major inputs such as labor. Thesecond main contribution is efficiency mea-surement using stochastic efficiency frontiers.These model the efficiency with which all inputsare converted into outputs, using only the morereliable input and output quantity data andavoiding prices since there are none for land orfamily labor. The final part of this section notesthat there are also social benefits, which areenvironmental and health improvements, andthe final section offers concluding comments.2. PROGRESS WITH GM CROPSIn 2001, the leading countries, by GM areaplanted, were the United States, Argentina,Canada, China, South Africa and Australia,with the first three accounting for 96% of thearea (James, 2002). More than 85% of GMcrops were bred for tolerance of specific herbi-cides and insecticides, but almost all the restare insect-resistant corn (maize) and cotton.These crops contain the genes controlling theproduction of a natural insecticide, Bacillusthuringiensis (Bt), which acts specifically onLepidoptera, including bollworm in cotton andstem borers in maize, and is harmless to allother insects.A survey by Marra (2001) shows that the vastmajority of GM crop studies have been con-ducted in the United States, typically using trialplot data from the biotechnology industry.Most studies of the performance of GM cropsshow increased yields and lower levels of laborand pesticide use (Fernandez-Cornejo & Klotz-Ingram, 1998; Gianessi & Carpenter, 1999).Fernandez-Cornejo, Klotz-Ingram, Jans, andMcBride (1999) report that for cotton specifi-cally, better quality and hence higher producerprices are a further benefit. The more analyticalstudies, such as Falck-Zepeda, Traxler, andNelson (2000) apply economic surplus mea-sures to determine the beneficiaries of GMcrops, but again most are for the United States.The only application of an efficiency frontier isa data envelopment analysis of cotton inGeorgia, United States, by Ward, Flanders,Isengildina, and White (2002). Much of thedeveloping country literature is ex-ante studies,such as Qaim (2000), since data are