Adoption and spillover of new cassava technologies in West Africa: Econometric models and heterogeneous agent programming

Innovations in cassava production are critical for food security and poverty alleviation among many of the most vulnerable populations in Africa, Asia, and Latin America. In the West African context, new varieties have been developed by the International Institute for Tropical Agriculture (HTA) in Nigeria, but their spread to neighboring countries has been limited. The purpose of this dissertation is to ask what determines the adoption and spillover of the HTA innovations, so as to predict more accurately what groups of farmers in what locations are likely to benefit from them. The dissertation's findings can help guide research and extension efforts immediately in West Africa, and point to improved modeling methods for the analysis of spillovers throughout the world. The analytical approach developed here focuses on the variation across farmers in resource endowments and market conditions. Standard methods use survey data to construct a single representative-farm model for specific farm types. The new method preserves the diversity among farmers found in survey data. The result is more efficient use of those data, to determine who benefits from new techniques in one area and who might adopt it elsewhere. The survey data used in the dissertation are from the Collaborative Study on Cassava in Africa (COSCA) conducted by IITA in Nigeria, Ghana and Ivory Coast. The COSCA data are used first in an econometric model to determine possible interaction effects between the various technologies currently available in Nigeria, to assess underlying constraints to adoption and predict spillover potential in neighboring countries. The interaction effects are between improved cassava varieties adopted at the household level and mechanized processing technologies adopted at the village level. The new varieties are labor-using, whereas mechanized processing is labor-saving. Results show significant complementarity between the two technologies in the Nigerian context, with no evidence of significance as to the order in which they are adopted. Low investments in credit and extension, and perhaps a weak physical infrastructure, were determined to be important factors limiting adoption in Ghana and Ivory Coast. Using insights from the econometric analysis, the same data are then used to develop a heterogeneous-agent linear programming (HALP) model to assess the profitability of cassava technologies and thus ex-ante spillover potential across the full sample of farm households in Ghana and Ivory Coast. For varietal improvement, HALP model results show that there is significant potential for spillovers outside of Nigeria, especially Ivory Coast's humid climatic zone, with adoption occurring at a rate above 70 percent, rates similar to those in Nigeria. Similar potential rates exist for Ghana assuming that current low productivity levels have been affected by a high incidence of plant disease and pest attack. The study therefore recommends targeting IITA's improved varieties in the humid climatic zones, including Ghana's semihot zone, while targeting mechanized processing to all regions.