The Impact of Behavioral Framing on Farmer Engagement: A Randomized SMS Extension Experiment in Côte d’Ivoire

2026-11-01

2026-12-27

Primary outcomes
1. Farmer Engagement Index (FEI). The first primary outcome is a composite measure of participants’ engagement with the SMS-based agricultural advisory service during the intervention. The index will capture five dimensions of engagement: (i) recall of agricultural recommendations received through the SMS service; (ii) comprehension of the recommendations; (iii) active interaction with the advisory service, including responses, requests for clarification, or contact with designated project or extension personnel where applicable; (iv) information-seeking behavior related to the recommendations; and (v) discussion and continued use of the information received. The five domain scores will be standardized and equally weighted to construct the Farmer Engagement Index. Higher values will indicate greater engagement with the advisory service.
2. Adoption of at least one promoted agricultural practice. The second primary outcome is a binary indicator equal to one if a participant adopts at least one eligible agricultural practice promoted through the SMS intervention during the prespecified outcome measurement period, and zero otherwise. A practice will count toward this outcome only if it was included in the finalized experimental message library, was relevant to the participant’s farming context, had not already been fully adopted by the participant at baseline, and was implemented during the study period. Adoption will be measured primarily through the endline survey. Where operationally feasible and prespecified before random assignment, reported adoption will also be independently verified for the full sample or a randomly selected validation subsample.

The Farmer Engagement Index will be constructed from five prespecified domains: (1) message recall, (2) message comprehension, (3) active interaction with the advisory service, (4) information-seeking behavior, and (5) message discussion and continued use.
Message recall will be measured as the proportion of a prespecified subset of relevant agricultural recommendations that the participant correctly recalls at endline. Message comprehension will be measured as the proportion of prespecified questions correctly answered regarding the recommended practice, the agricultural problem addressed, and the expected consequences of implementing or not implementing the recommendation.
Active interaction with the advisory service will be constructed from applicable binary indicators capturing whether the participant responded to an SMS prompt, contacted a designated project or extension representative in response to a study message, or requested clarification or additional agricultural information through an approved study channel. Information-seeking behavior will be constructed from indicators capturing whether the participant consulted an extension agent or project representative, requested additional technical information, consulted another knowledgeable farmer, or attended a relevant agricultural training or field demonstration after receiving related SMS advice.
Message discussion and continued use will be constructed from indicators capturing whether the participant discussed at least one study message with another farmer or household member, reported reading most or all study messages, saved or revisited at least one message, and correctly identified the approximate timing or agricultural context in which at least one promoted recommendation should be implemented.
All component variables will be coded so that higher values indicate greater engagement. Applicable components within each domain will first be averaged to obtain a domain score. Each of the five domain scores will then be standardized using the mean and standard deviation of that domain among participants assigned to the loss-framed condition. The five standardized domain scores will receive equal weight and will be averaged to construct the Farmer Engagement Index. Higher values of the resulting index indicate greater engagement with the SMS-based agricultural advisory service.
A domain score will be calculated when valid data are available for at least half of the component items assigned to that domain. The overall Farmer Engagement Index will be calculated when valid scores are available for at least four of the five domains. If four domains are available, the index will equal the average of those four standardized domain scores. Participants with fewer than four valid domain scores will be treated as missing for the complete-case confirmatory analysis.

Secondary outcomes will include: (i) message recall and comprehension; (ii) intention to adopt recommended agricultural practices; (iii) perceived relevance and usefulness of the SMS advisories; (iv) agricultural self-efficacy; (v) agricultural risk perception, including perceived probability and severity of adverse agricultural outcomes and perceived ability to prevent or control them; (vi) information-seeking behavior; and (vii) willingness to continue receiving the SMS advisory service and willingness to pay for continued access.

This study is an individually randomized, two-arm field experiment involving smallholder farmers in Côte d’Ivoire. The experiment is designed to test whether the behavioral framing of agricultural advisory messages affects farmer engagement with an SMS-based advisory service and the adoption of recommended agricultural practices.
Eligible participants will complete a baseline survey before random assignment. Participants will then be randomly assigned in equal proportions to one of two experimental conditions: a gain-framed SMS condition or a loss-framed SMS condition. Participants assigned to the gain-framed condition will receive agricultural advisory messages that emphasize the benefits or positive outcomes associated with implementing recommended practices. Participants assigned to the loss-framed condition will receive substantively equivalent messages that emphasize the risks, costs, losses, or forgone benefits associated with not implementing the same practices.
All participants will receive agricultural advisory messages; there is no no-message control group. For each agricultural recommendation, the underlying technical advice will be the same across the two experimental conditions. The messages will differ only in their behavioral framing. Other features of message delivery, including language, timing, frequency, sender identity, and available response options, will be kept equivalent across the two conditions to the extent operationally possible.
The intervention will run for eight weeks, with participants expected to receive approximately three SMS advisory messages per week. Message timing will consider the agricultural calendar and the relevance of individual recommendations to participants’ crops and farming conditions.
Randomization will occur at the individual-participant level using reproducible statistical code, with a 1:1 allocation ratio between the two framing conditions. Randomization will be stratified by geographic study area and gender. Principal crop may be included as an additional stratification variable if the final recruited sample contains sufficiently large numbers within crop categories to avoid sparsely populated strata. All stratification variables and their definitions will be finalized before random assignment.
Random assignment will take place only after recruitment, eligibility verification, completion of the baseline survey, resolution of duplicate or household records, and finalization of the randomization dataset. Recruitment personnel and baseline enumerators will not know participants’ framing assignments during recruitment and baseline data collection.
Outcomes will be measured using endline survey data and records of participant engagement with the advisory service. Where prespecified and operationally feasible, reported adoption of promoted agricultural practices will also be independently verified for the full sample or a prespecified randomly selected validation subsample. The primary analysis will estimate intention-to-treat effects by comparing participants according to their original randomized framing assignment.

Not available

Randomization will be conducted by computer using reproducible statistical code in Stata. Eligible participants will be randomly assigned in a 1:1 ratio to either the gain-framed or loss-framed SMS messaging condition. Randomization will be stratified by geographic study area and gender. Principal crop will be included as an additional stratification variable only if the final sample contains sufficiently large numbers of participants within each crop category. A fixed random-number seed will be selected and archived before randomization, together with the randomization script and treatment-assignment file.

The unit of randomization is the individual participant. Eligible farmers will be individually randomized in a 1:1 ratio to either the gain-framed or loss-framed SMS messaging condition. There is no additional level of randomization.

No

600 individual participants

600 farmers

300 farmers assigned to the gain-framed SMS messaging condition; 300 farmers assigned to the loss-framed SMS messaging condition.