Abstract
The contribution of technological change to agricultural productivity in developing countries has long been documented. It is believed that the adoption of new agricultural technologies, such as high-yielding varieties, could lead to significant increases in agricultural productivity and stimulate the transition from low-productivity, subsistence agriculture to a high-productivity agro-industrial economy. The article uses the local average treatment effect (LATE) to estimate the impact of adoption of improved rice varieties on rice farmers' productivity in the three major rice ecologies of Nigeria. A stratified random sampling was adopted by the study to select a sample of 500 rice farmers across ecologies. Findings of the analysis indicated that adoption of improved varieties helped raise farmers' area harvested and yield per hectare, respectively, by 0.39 hectare and 217.9 kg/ha for NERICA and 0.51 hectare and 210.4 kg/ha for other improved varieties, thereby increasing their productivity. In addition, NERICA varieties performed better than any other upland improved variety and the impact of its adoption on both area harvested and yield was greater among female rice farmers than among their male counterparts. Intervention programs to increase the dissemination of high-yielding rice varieties to areas with low productivity are, therefore, a reasonable policy instrument.
Acknowledgments
This study was jointly conducted by the University of Ibadan, Nigeria, and the National Cereals Research Institute, with scientific support from AfricaRice. The authors acknowledge the financial support of the Bill and Melinda Gates Foundation, the European Union with technical support by IFAD, and the African Development Bank. They also thank lecturers in the Department of Agricultural Economics at the University of Ibadan for their useful and constructive comments on the manuscript.
Notes
1. Others improved varieties released at the same time with NERICA were FARO 53 and FARO 54.
2. The usual third requirement that the instrument be “uncorrelated with the unobserved error term” made in classical IV can be weakened by the CitationAbadie (2003) generalization of the LATE identification estimation through the local average response function (LARF).
3. (i) Independence of the instrument: Conditional on X, the random vector (Y00; Y01; Y10; Y11; D0; D1) is independent of Z; (ii) exclusion of the instrument: P(Y1d = Y0d|x) = 1 for
; (iii) first stage: 0 < P(Z = 1|x) < 1 and P(d1 = 1|x) > P(d0 = 1|X); (iv) monotonicity: P(d1
d0|x) = 1.
4. The first survey was carried out from December 2008 to February 2009, while the second survey, which collected data on household expenditure, was carried out during the second half of 2009.
5. ICT was measure by access to new technologies for communication (radio, television, and mobile telephone).
6. This is introduced in the model to test for the linearity of age with adoption and access to NV seed.