Methodology

Ex-post – conceptual framework

NERICAs are relatively new and are not yet widely adopted and the diffusion (exposure) and adoption of the NERICAs by farmers is non-random. This makes it difficult to estimate the causal effect of NERICA adoption due to the bias resulting from the non-randomness of the exposure to the technology. Further, since the diffusion of the technology is incomplete, it means the impact on a population can not be estimated and thus only the potential impact can be estimated. This necessitated the search for an appropriate methodology for impact assessment. Thus impact assessment is conducted using counterfactual outcomes framework proposed by Rubin (1974). The framework addresses problems of selection bias, overt and hidden bias as well as the problem of the endogenous treatment variable or non-compliance. For further explanation on the framework refer to Diagne (2006).

Data collected

Data is collected at three levels; country, community and household levels. There are two major surveys through which data is collected; the light survey and the deep survey. The objective of collecting light survey data is to get a quick overview of farming systems and crops grown in the study areas. The deep survey captures details on rice farming including input use patters, productivity, consumption patterns and income.

AfricaRice employs impact assessment methodology grounded within the ‘counterfactual’ outcomes or Average Treatment Effect (ATE) framework underlying modern evaluation theory and practice. The main components of this methodology, which is also at the center of the impact assessment training courses conducted since 2002 for NARS economists, comprise data collection at several levels and analysis as follows:  

1. Community and household surveys on knowledge and adoption of varieties and on seed acquisition

2. Household and plot-level surveys to collect data on areas and yield by variety, input use, income, food intake, children’s schooling etc

3. Countrywide census or survey data on rice areas and farm populations

4. Estimation of dynamic models of adoption based on the Average Treatment Effect (ATE) methodology

5. Estimation of impact on various household-level outcomes based on ATE methodology

6. Estimation of ex-ante and ex-post impact on economic and environmental outcomes at the national and continent-wide levels.

Apart from in Côte d’Ivoire where AfricaRice is fully responsible, data collection and documentation work is conducted by NARS economists in the ROCARIZ network using AfricaRice funding.