Seminario CEDE - Rachid Laajaj

Farmer’s learning is a decisive driver of the process of technology adoption and dis-adoption, yet few studies take a close look at this process. To address this, we implemented a randomized control trial in Kenya where 10 farmers in each treated village participated in agronomic research trials, which gave them the opportunity to test new inputs in their own parcels during 3 rainy seasons. Drawing from detailed data collection on various forms of knowledge during 6 consecutive seasons, we take a deep dive into the learning process and differentiate it by the initial skill level of each farmer. We draw 10 lessons regarding how the treatment affected farmers’ beliefs, know-how, practices, input use, returns and the role of social connections in this process. The findings point to a costly and challenging learning given the multidimensionality of input and practices. Differences by skill level are large, in particular for complex things to learn such as the complementarity between different practices or the relative comparison between different inputs. The results are not always consistent with the standard learning models, hence we propose an alternative model which, although less tractable, better captures the complexity of learning about the potential returns and learning how to farm in a multi-dimensional decision environment.