Participatory versus traditional agricultural advisory models for training farmers in conservation agriculture: a comparative analysis from Kenya

Abstract

<b>Purpose:</b> We investigated the variation between participatory and traditional agricultural advisory models in training farmers on Conservation Agriculture in Kenya in order to understand how these models influenced farmer uptake of practices, development of social networks and delivery cost of training.<br/><b>Design/methodology/approach:</b> A pre–post-test was applied using panel data from before and after participants took part in Conservation Agriculture training. Data on the assessment criteria were collected, and key informant interviews, personal observation and training reports added context to the findings. Data analysis included logistic regression and social network analysis.<br/><b>Findings:</b> All advisory models built social networks, enhancing information diffusion, but at different levels. Of the participatory models, Landcare, exhibited greater farmer uptake of Conservation Agriculture and more developed social networks, while the participatory Farmer Field School model performance was similar to the traditional advisory model. Both participatory models were more expensive to deliver than the traditional model.<br/><b>Practical implications:</b> Agricultural advisory services should be delivered through a collaborative and pluralistic advisory system and include participatory needs identification and explicit social capital building strategies, with advisory agents’ capacity to deliver these strategies enhanced. Opportunities for strategic cost-saving measures should be sought.<br/><b>Theoretical implications:</b> Findings from this study partly agree with the literature that participatory models enhance the adoption of complex agricultural practices compared with traditional models.<br/><b>Originality:</b> Benefits of participatory models are outlined extensively in the literature; however, understanding the variable implementation of these models, and how they perform compared to traditional models, has not been sufficiently studied.