IMPACT OF GENETICALLY MODIFIED MAIZE ON SMALLHOLDER RISK IN SOUTH AFRICA 16 th ICABR Conference June 25-27, 2012 Ravello, Italy Greg Regier*, Timothy Dalton, Jeffery Williams Kansas State University Source: Google images
Objective Is genetically modified (GM) maize a beneficial technology for smallholders in low-income countries? H 1 : GM maize reduces net returns risk H 2 : GM maize has higher output H 3 : GM maize leads to lower cost
Literature Review Bt Maize, Philippines Higher yields and net returns; Yorobe and Quicoy 2006 ; same results when controlling for selection bias; Mutuc and Yorobe 2007 Yield advantage is smaller controlling for censoring; Mutuc, et al Bt Maize, South Africa Higher output, declining as pest pressure decreases, net returns depends; Gouse, Piesse and Thirtle 2006, Gouse et. al 2006 RR Maize, South Africa Higher output, lower labor use; Gouse, Piesse, and Thirtle 2006 Seed cost cancels gain in yield efficiency; Gouse, Piesse, Thirtle and Poulton 2009
Location: KwaZulu-Natal, South Africa
Background Information Hlabisa and Simdlangetsha Annual rainfall of 980 mm (38 inches) Marginal land - 13% arable Average maize yield is 1500 kg/ha (24 bu/acre) 39% land ownership by smallholders Labor supply characteristics Urban migration 26% working age population HIV-positive
Data 212 maize plots (184 households) Plot size 0.49 hectares, farm size1.85 hectares One season, Farmer Characteristics: Head of household average age of 55 years Pension is top income source for 53% ($168/month) A majority of maize consumed at home High access to credit
Maize Types Conventional Hybrids Pannar Carnia GM Hybrids Bt – insect resistant Roundup Ready © (RR) – herbicide tolerant BR “stacked”
Maize Yield, Cost, and Net Returns * Indicates significantly higher at 5% using a one-sided t-test Seed TypeNYield (kg/ha) Maize Revenue ($/ha) Input Cost ($/ha) Labor Cost ($/ha) Net Returns ($/ha) HlabisaBR Pannar RR GM * Non-GM *234 SimdlangetshaBR Bt Carnia Pannar RR GM Non-GM *-338
H 1a : GM Maize Reduces Risk - Stochastic Dominance
H 1b : GM Maize Reduces Risk - Stochastic Efficiency with Respect to a Function (SERF) RRAC = 2 (moderately risk averse) RRAC = 4 (extremely risk averse) RRAC = 0.37 (slightly risk averse)
H 1b : GM Maize Reduces Risk – SERF RRAC= 2RRAC = 4
H 2 : GM Maize has Higher Output Maize output = f(labor, fertilizer, herbicide, seed, land, land prep cost, Hlabisa, RR, Bt, assets, experience with herbicide, education)
Production Function Results OLS: LinearOLS: QuadWLS: Quad2SLS: Herbicide2SLS: Labor Coef. Intercept *** *5.3 Labor3.26***2.73* ***-0.4 Fertlizer1.35* ** Herbicide **-26.3 Seed Land993.56*** * * *** Total Cost Land Prep *-15.28** Hlabisa Dummy308.83*** **215.9* RR Dummy217.27***137.45**131.61*332.0***38.9 Bt Dummy N212 Adjusted R- squared ***,**,* indicates significantly different than zero at 1%, 5% and 10% respectively
H 3a : GM Maize has Lower Cost Total Cost = f(maize output, labor price, fertilizer price, herbicide price, seed price, land, land prep price, Hlabisa, RR, Bt, assets, experience with herbicide, education)
Cost Function Results OLS - LinearOLS - Quadratic WLS - Quadratic Treatment Effects- Quadratic Coef. Intercept * ** ** ** P(labor)134.59*** P(fertilizer)237.79** ** ** ** P(herbicide)2.91** * P(seed)14.64***79.49*75.83*70.78* Land389.67*** *** *** *** P(land prep)-0.75***9.28*9.68**8.75** Maize Output0.05***0.69***0.61***0.62*** Hlabisa Dummy *** *** *** *** RR Dummy-63.83***-77.67***-69.60*** *** Bt Dummy Inverse Mills Ratio λ 49.77** Adjusted R-squared ***,**,* indicates significantly different than zero at 1%, 5% and 10% respectively
H 3b : GM Maize has Lower Cost - Kernel Density Estimator Total Cost
H 3b : GM Maize has Lower Cost - Kernel Density Estimator Average Cost
Conclusion H 1 : GM Maize Reduces Risk SERF RR maize producers must be compensated between $18 and $221 per hectare to switch varieties H 2 : GM Maize has Higher Output Production function 8-13% RR maize advantage; N.S.-20% controlling for endogeneity bias H 3 : GM Maize leads to Lower Cost Cost function 18-23% lower costs for RR maize; 33% controlling for selection bias Nonparametric regression At least 17% lower costs for RR maize
THANK YOU! Acknowledgements: Bill and Melinda Gates Foundation by the provision of data under the Global Development Grant OPP 53076, “Measuring the Ex-Ante Impact of Water Efficient Maize for Africa.” Assistance from Marnus Gouse in the understanding of data. Source: Google images
Future Research More advanced techniques to control for selection bias Control for censoring Tradeoff between no-till and intercropping Labor supply Constrained or not Effect on GM maize adoption by country Impact over several years in multiple regions
Weighted Risk Premiums Relative to BR, Simdlangetsha
Weighted Risk Premiums Relative to BR, Hlabisa
Two-Stage Least Squares (2SLS) Regression = predicted value of endogenous variable = parameter of all exogenous variables = parameter of instrumental variables
Treatment Effects Model Step 1: Probit Model Step 2: Include inverse Mills ratio in least squares regression
Histogram of Total Cost
Kernel Density Estimator: Average Cost RR = 0.5, non-RR = % lower costs estimate
Family and Hired Labor (Hours/Hectare) SiteSeed TypeChildMaleFemaleHiredWorkgroup Total HlabisaBR Pannar RR GM **192 Non-GM 18** 153** 177** 68**20 437** SimdlangetshaBR Bt Carnia Pannar RR GM Non-GM 65** 94** 118* ** **,* Indicates significantly higher at 1% and 5% respectively using a one-sided t-test.