Kenya LR2020 cropping series randomization
KEN -20 -1502Last modified on October 1st, 2025 at 10:28 pm
-
Abstract
PAD operates the MoA-INFO platform in collaboration with the Kenya Ministry of Agriculture to provide free agricultural recommendations to farmers via SMS. Farmers can access information whenever they like by sending the word MENU or ORODHA to access a complete list of topics. In addition, farmers have an opportunity to opt into weekly “cropping series” (CS) which offer advice on crop management practices - from land preparation to harvesting and storage - throughout the season. While all the content in the CS is available on the menu for farmers to access at any time, most engagement with the platform comes from farmers who opt into the weekly cropping series advisory.
This gave us an opportunity to evaluate the effects of the MoA-INFO service by randomizing CS invitation messages. While we previously found that the MoA-INFO platform could increase farmer knowledge about topics such as Fall Armyworm (FAW), we had not measured the effects on behavior changes. During the long rainy season 2020, we selected seven CS that pertain to seed and fertilizer choice and post-harvest storage, and we randomized invitations to those seven CS messages. In September 2020, we collected information on farmer adoption of recommended practices and crop yield via a phone survey. Complementary information on platform engagement came from administrative data from the MoA-INFO platform.
We found that adopting more recommended practices, measured using an aggregate index of practices, was correlated with higher yield. Yet, there was no evidence suggesting that receiving text message advice led to statistically significant changes in the adoption of recommended practices or farm outputs (yield and harvest). -
Status
Completed
-
Start date
Q1 Jan 2020
-
Experiment Location
Kenya
-
Partner Organization
Kenya Ministry of Agriculture
-
Agricultural season
Long Rains
-
Experiment type
Impact Evaluation
-
Sample frame / target population
MoA-INFO Platform Users (farmers)
-
Sample size
2,939
-
Outcome type
Farming practices, Agricultural production / yield, Input adoption, Service engagement
-
Mode of data collection
Phone survey, PxD administrative data
-
Research question(s)/hypotheses
What are the effects of MoA-INFO cropping series (CS) messages on farmer practices and yields?
-
Research theme
Agricultural management advice, Communication technology
-
Research design notes
The trial sample consisted of all 11,336 MoA-INFO farmers from early planting constituencies who opted into both maize and bean CS, did not opt out as of the randomization, self-reported ward location, and were in a ward with at least one other farmer who met these criteria (since otherwise the ward would be dropped due to ward level fixed effects).
The sample was stratified at the ward level and randomized at the individual level because farming recommendations vary by the predominant agricultural ecological zone (AEZ) at the ward level. Out of the 11,336 farmers in the sample, 1,417 (12.5%) were randomized to be part of the control group that did not receive the selected CS messages: maize seed, maize fertilizer, maize topdressing, maize post-harvest storage, bean seed and fertilizers, bean planting, bean post-harvest storage. The remaining 9,919 treatment farmers received the complete cropping series messages (unless they opted out).
We conducted a phone survey on a sub-sample of 2,939 farmers to gather information on their farming practices, yield, and demographics.
Given that farmers had signed up to receive CS recommendations, we did not want to hold off sending CS invitations to many farmers. Therefore, we chose to have a small control group and measure a limited set of outcomes on behavior change.
-
Results
We found that adopting more recommended practices, measured using an aggregate index of practices, was correlated with higher yield, yet there was no evidence suggesting that receiving text message advice led to statistically significant changes in the adoption of recommended practices or farm outputs (yield and harvest).
We analyzed results by sub-groups for maize mono-cropping (N=1,260), bean mono-cropping (N=381), and maize and bean intercropping (N=1,635). We did not detect a significant impact by these three sub-groups because the analysis was underpowered given the observed effect levels and standard errors, yet differences among these groups provide some valuable insights. In the case of beans, treatment impacts on yield and on individual practices are always in the expected (positive) direction. For maize, treatment impacts on practices are sometimes in the expected (positive) direction and sometimes in the opposite (negative) direction, while the impacts on maize yield are consistently (insignificantly) negative. This difference seems to suggest that bean results are somehow more promising than maize results.
We also explored heterogeneous effects across different subgroups (previous CS opt-ins, superusers, gender, smartphone ownership, location) and found little evidence of such effects.
These results hint at the robustness of our agronomic content and approach (i.e., advising on a set of good agricultural practices), and highlight the importance of increasing the effectiveness of our information provision.