This report describes the methods and findings of a randomized controlled trial evaluating the impact of coffee agronomy training and phone-based advisory services on farmer practices and observed coffee yield. In-person training was provided in randomly selected villages over the course of two years by Hanns R. Neuman Stiftung (HRNS) and TechnoServe in two separate regions of Western Uganda encompassing six districts. Messages reinforcing this training were sent to a subset of farmers in villages where training was offered by Precision Development (PxD), and standalone messages were sent to a subset of farmers in villages where no training was offered. The program period spanned the onset of the COVID-19 pandemic and associated lockdowns, which significantly affected how training could be delivered and likely reduced its impact. Coffee yield per tree was measured for the fourth main harvest after the start of training. The number of coffee cherries per tree was estimated for three trees per farm by removing and weighing green cherries by size category ahead of harvest. This number was multiplied by the weight per ripe cherry during harvest time to arrive at an estimate of the ripe yield weight. Coffee practices were observed at the same time as green cherry measurement. We find that coffee yield per tree was 4.1% higher in villages where HRNS offered training compared to in villages where no training was offered, and 10.5% higher in villages trained by TechnoServe among a sample of farmers who showed initial interest in training. As only about half of farmers in treatment villages were verified to have attended a significant amount of training, yield impacts on trained farmers could be as high as 8.2% and 18.1% for HRNS and TechnoServe, respectively. That the regions where the two implementers worked were quite different, so these differences may be due at least partially to context, as opposed to trainer programming. Farmers’ use of recommended coffee practices increased in line with the observed yield impact. Knowledge of these practices was also improved by training. However, practices about which knowledge improved the most were not always those for which the greatest physical changes were seen. This suggests that there are other constraints to adoption of recommended practices, and that training can serve as a reminder and encouragement to implement practices of which farmers are already aware. There is no evidence that the improvement in coffee yield was offset by a decline in households’ cash income from other sources or less food crop production, or that it reduced women’s income sources controlled by women. Nor did training affect women’s control over coffee income, which remained low. In contrast to impacts on the directly observed outcomes of yield, practices, and knowledge, we do not find statistically significant impacts on farmers’ reported coffee revenues, costs, or profits. We attribute this to the difficulty of accurately recalling expenditures and income related to coffee, both of which occur over several months, and to the discrepancy in time periods covered by recall data on coffee costs and sales, versus prospective yield estimates. Sending farmers recorded phone messages that reinforced training content resulted in a near-statistically significant impact on practices equivalent to a 3.1 percent increase in yield in HRNS training villages, but no impact in the TechnoServe sample. This implies a total combined yield increase of 5.5 percent for farmers who received both HRNS training and reinforcement messages, and a 2.7 percent increase for those who did not. There is suggestive evidence that a standalone mobile phone-based intervention may have had a similar impact on coffee practices, though this is measured with less statistical precision.

Yield gains of in-person training translate to an annual return on investment of between 15% and 44% depending on the region and implementer. The total ROI depends crucially on whether these yield impacts are sustained over time. A previous evaluation for which farmer practices were observed at multiple points post-training found persistent effects. Based on results from that study, we estimate a total 14- year net ROI of between 19% and 251% (by region and implementer) beyond recovery of the initial investment. Using the imprecisely estimated impacts of the phone-based interventions, and projecting yield effects based on farm practices, we calculate annual ROIs for these of between 225% and 335%, assuming they are deployed at a scale of one million farmers.