In this article, we discuss a study to identify an effective agricultural knowledge transfer channel for smallholder farmers in Ethiopia, using a randomised training experiment together with focus group discussions, key informant interviews and a survey. We also examine the factors that determine learning among smallholder farmers. Our results revealed that involving extension agents and model farmers leads to above-average knowledge transfer. However, learning from extension agents is significantly more effective than learning from model farmers. Additionally, we found that trust, effort, and locus of control are important determinants of learning. On the other hand, we found no evidence that farmers exert more effort when they are trained by extension agents, hence this rules out effort as a mechanism for higher learning from the extension agents. Based on these results, we conclude that, on average, the extension agent system is more effective at conveying agricultural knowledge than model farmers are and that policy-makers can use the two channels as complements rather than substitutes.
