Adoption of quality- and market-improving practices and revenues earned by smallholder coffee farmers in Eastern Uganda

Coffee is a major cash crop and source of income for thousands of smallholder farmers in Uganda. High quality Arabica coffee is highly demanded in the international market and, thus, has high income generation potential across the value chain including for primary producers. However, the quality of Arabica coffee grown in Uganda is persistently low. A key cause identified is sub-optimal harvesting techniques among smallholder coffee producers. World Agroforestry Centre (ICRAF) implements a training program in collaboration with the largest coffee exporter in the area to encourage adoption of optimal harvesting. A module which focuses on selective and frequent picking and is developed by coffee experts and researchers is incorporated into ongoing extension services in the area. This project analyzes the role of information dissemination and estimates the impact of the training program in encouraging optimal harvesting of coffee to improve coffee quality, quantity sold, and revenues for smallholder farmers. There are 3 treatment variations. This study aims to identify which strategy is most effective to encourage selective harvesting of coffee.

1.Information treatment. In this set of farmer groups, a training module is given using the standard extension approach. Extension workers are trained by and with the materials prepared by coffee experts and researchers. Topics to be covered include: what is coffee quality and how to achieve ideal quality; what is selective harvesting; cost and benefits of selective picking; how to pick more effectively. Pictures of coffee cherries picked at different stages are included in the training programme to show how ideally picked cherries (and others) should look like. Cost-benefit calculations are provided to help farmers make informed decisions.
A picking demonstration is held to show how to effectively pick cherries.

2.Information treatment + peer farmer meetings. In this set of randomly selected groups, coffee farmers are exposed to the same information treatment as above. In addition, a peer farmer who selectively picks and sells coffee without sorting talks about their experience during the meeting. Farmers’ questions are answered by the peer farmer during the meeting. Hearing the positive experience of peer farmers about selective harvesting is expected to encourage selective picking among farmers (through peer-to-peer social learning).

3.Information treatment + buyer visits and calls. In this set of randomly selected groups, farmers are exposed to the information treatment and, additionally, a senior staff (e.g. project manager) from the buyer attends the meetings. This variation aims at improving the credibility of information provided during the training and restoring the producers’ confidence in the exporter. In addition, an automatic voice message (phone call) from the buyer is sent to farmers during the harvest season starts that the exporter hopes to receive hand-picked high-quality cherries from them. The goal is to make each individual farmer feel that his/her behaviour matters to the buyer and create loyalty thereby making farmers more motivated to meet buyer’s expectations to produce high quality coffee.

2018-07-30

2018-08-18

adoption of selective picking, coffee quality (subjective), quantity of coffee harvested and sold, coffee revenues

trust in and loyalty to buyer, agriculture network density

The study takes advantage of a World Agroforestry Centre (ICRAF) program implemented in collaboration with the largest coffee exporter in Eastern Uganda. The program aims to train about 4000 coffee farmers on an optimal harvesting method to increase coffee quality and quantity harvested. Our sample consists of 88 farmer groups and 1600 coffee farming households which are randomly selected. These 1600 farmers are randomly assigned to the three treatment variations and the control group (see abstract). The study uses an encouragement design.

Strafied randomization (on region, coffee harvest level, education of the household head) at the cluster level was done using statistical software Stata in office. We used the command randtreat which deals with misfits.

Randomization was done at the farmer group level

Yes

88 farmer groups in total

1600 households

20, 19, and 19 farmer groups in treatment variation 1, 2, and 3, respectively, and 30 in the control group.