Impacts of sustainable intensification training on adoption, yield and profits: Evidence from a Randomized Controlled Trial

The intervention will provide selected farmers with structured learning opportunities through training workshops, farmer field schools, and demonstration plots. The training will focus on a bundled package of sustainable intensification practices, including the use of improved maize varieties (drought-tolerant maize varieties, early maturing maize varieties, and pest and disease resistant maize varieties), crop rotation, minimum tillage, organic manure, climate-smart planting basin, cover cropping, intercropping, mulching, and crop residue retention. The training will involve building farmers capacity and knowledge on the use of tailored bundled sustainable intensification practices in their maize plots. We shall prioritize training farmers in different villages with tailored bundled sustainable intensification practices that addresses climate and production challenges specific to each village.

2026-04-30

2026-10-28

Adoption rates, Yield, Profits

Profit from maize production will be calculated as the difference between total revenue and total production costs. Total revenue will be computed as the quantity of maize harvested multiplied by the prevailing farm-gate price at the time of harvest. Total production costs will include expenditures on seeds, fertilizers, agrochemicals, hired labor, family labor,and other production-related expenses. The value of family labor will be imputed using the prevailing local wage rate.

Knowledge scores, Production cost, Income

Knowledge of Sustainable Intensification Practices will be measured using a standardized score based on farmers’ responses to knowledge questions related to the promoted practices, collected at baseline, after training sessions and at endline

This study employs a randomized controlled trial (RCT) to evaluate the impact of a farmer training intervention on knowledge, adoption of sustainable intensification practices, maize yield, and profit among maize farmers. The intervention consists of training workshops and demonstration activities designed to promote a bundled set of sustainable intensification practices tailored to meet farmers needs.

The units of analysis are smallholder maize farmers residing in five selected villages in Buea, Cameroon. The study focuses on individual farming households engaged in maize production as their primary or important agricultural activity. In the absence of existing data on maize farming population, a complete household listing exercise was first conducted in each village by enumerators with the help of lead farmers and extension agents to identify all maize-producing households. From this listing, a sample of eligible farmers was randomly selected for participation in the study, yielding a total sample of 320 farmers (approximately 64 farmers per village). These farmers constitute the panel that is followed over time for baseline and endline data collection.

A stratified random sampling approach will be used, where each village constitutes a stratum. Within each village, eligible maize farmers identified through a listing exercise are randomly selected to participate in the study. Equal sample sizes are maintained across villages to ensure comparability and to avoid overrepresentation of any single village in the analysis. This stratified sampling design ensures that each village is equally represented in the sample, thereby improving external validity and facilitating balanced comparisons across geographic contexts.

To further improve baseline comparability between groups, a covariate-constraint randomization (re-randomization) procedure will be implemented. Multiple random assignments will be simulated, and each allocation is evaluated based on balance across key baseline covariates. Standardized mean differences and the Mahalanobis distance will be used to assess balance. Only allocations that satisfy pre-specified balance thresholds will be retained, and the final assignment is randomly selected from the set of acceptable allocations.

The treatment consists of a structured agricultural training program focused on sustainable intensification practices in maize production. The intervention will be delivered only to farmers assigned to the treatment group. The program will include training workshops, demonstration plots, and farmer field schools.

The training will focus on a bundled set of practices, including the use of improved maize varieties (drought-tolerant maize varieties, early maturing maize varieties, and pest and disease resistant maize varieties), crop rotation, minimum tillage, organic manure, climate-smart planting basin, cover cropping, intercropping, mulching, and crop residue retention. The intervention will be implemented after baseline data collection and randomization are completed. Training activities will be delivered during the maize production cycle to allow farmers to apply learned practices in real time. Farmers in the control group will not receive any training during the study period and continue with their usual farming practices.

To support implementation fidelity and promote adherence to the training content, field extension agents will conduct regular monitoring visits to farmers assigned to the treatment group throughout the main maize production period (April to October). These visits will provide technical guidance, reinforce key messages from the training sessions, and assist farmers in applying the promoted sustainable intensification practices. Monitoring activities will be structured and standardized to ensure consistency across villages. To preserve the integrity of the experimental design and minimize contamination, extension agents will restrict advisory support to treatment farmers and will not deliver training-related information to control group farmers. In addition, monitoring protocols will emphasize limited interaction between treatment and control farmers regarding intervention-specific content. These measures aim to enhance compliance among treated farmers while reducing the risk of spillovers into the control group.

The study will collect both baseline and endline data, with primary outcomes measured at endline approximately 9 months after the intervention.

Primary Outcomes
-Adoption of sustainable intensification practices: will be measured using an index capturing uptake of improved maize varieties, crop rotation, intercropping, mulching, and residue retention
-Maize yield: will be measured as kilograms of maize harvested per hectare
-Farm profit: will be measured as revenue minus total production costs per hectare

Secondary Outcomes
-Knowledge of sustainable intensification practices: will be measured using a standardized index based on farmers’ responses to structured knowledge questions
-Maize revenue (income): will be total value of maize output
-Production costs: will including inputs (seeds, fertilizer, agrochemicals) and labor (including imputed family labor)

The evaluation will be conducted across five villages. Randomization will be implemented at the individual farmer level, with farmers assigned to either a treatment group (receiving the training intervention) or a control group (not receiving training during the study period). The study will follow a panel design, with the same farmers surveyed at baseline and endline.

The study population will consist of smallholder maize farmers residing in five selected villages where maize production is a primary agricultural activity. Within each village, a sample of 64 farmers will be selected, yielding a total sample size of 320 farmers. The same farmers will be tracked and re-interviewed at the endline to construct a balanced panel dataset.

To minimize spillover effects and contamination between treatment and control farmers, care will be taken to avoid assigning neighboring farmers to different experimental groups where possible. During the sampling and randomization process, spatial information collected during the household listing exercise will be used to identify geographically proximate households. Where feasible, farmers located within close proximity (e.g., within the same neighborhood or immediate spatial cluster) will be assigned to the same treatment status. This will reduce the likelihood of information diffusion from treated to control farmers and helps preserve the integrity of the experimental contrast between groups.

The study design will follow a panel design, tracking the same farmers at baseline and endline. We might not be able to reach all the respondents initially sampled. Attrition will be minimized through careful tracking procedures, and balance tests will be conducted to assess whether attrition is systematic.

The primary estimation approach will be an intention-to-treat (ITT) analysis comparing outcomes between treatment and control groups. Baseline covariates and stratification fixed effects will be included to improve precision. Standard errors will be clustered at the village level to account for intra-village correlation.

To further improve statistical power and control for unobserved heterogeneity over time, a difference-in-differences (DiD) specification will also be estimated using the panel structure of the data. This model will exploit variation across baseline and endline observations for the same farmers and estimate treatment effects by comparing changes in outcomes over time between treatment and control groups. The DiD specification will also act as a robustness check.

The study will examine heterogeneity in treatment effects by gender to assess whether the intervention has differential impacts on male and female farmers. Gender will be defined based on the gender of the household head and/or the primary decision-maker in maize production. To estimate differential effects, interaction terms between treatment assignment and gender indicators will be included in the regression specifications. This will allow for comparison of treatment effects across gender groups for key outcomes, including adoption of sustainable intensification practices, maize yield, and farm profit. Given potential limitations in statistical power for subgroup analysis, these results will be interpreted as exploratory but will provide important insights into gender-specific constraints and responses to agricultural training interventions.

Further robustness checks will be conducted to assess sensitivity to attrition and alternative specifications.

Not available

Randomization will be implemented using statistical software (stata) by the research team. Farmers will be randomly assigned to treatment and control groups within village-level strata using a reproducible random number generator. A covariate-constrained re-randomization procedure will be applied, whereby multiple random assignments are generated and only those satisfying pre-specified balance criteria on baseline covariates (including education, farming experience, farm size, and baseline yield) are retained. The final assignment will be randomly selected from the acceptable set.

The unit of randomization in this study is the individual maize farming household. No additional cluster-level treatment assignment is implemented. Randomization will be conducted at the individual level within villages, with villages serving as strata to ensure geographic balance across study sites. However, care will be taken during the randomization process to minimize potential spillovers by avoiding, where feasible, the assignment of geographically proximate farmers within the same village to different treatment arms. This helps reduce contamination and diffusion of training effects between treatment and control farmers while preserving the integrity of individual-level randomization. In addition to individual-level randomization, village-level stratification is used to control for potential heterogeneity in agro-ecological conditions and other location-specific factors.

No

320

320

160 farmers control, 160 farmers exposed to trainings

We will sample 320 farmers from five villages, with about 160 assigned to receive treatment and 160 assigned to the control group. We shall rely on the convention 0.05 significance level and a power of 0.8, allowing for the possibility to detect a minimum effect size of approximately 0.20 - 0.30 standard deviations in primary outcomes (yield, adoption, and profit). At this stage we do not take into account the potential gains in precisions from including baseline covariates and adjusting for clustering. Using baseline standard deviations, these effect sizes will be translated into meaningful units. For example, for maize yield, the minimum detectable effect corresponds to approximately X–Y kg/ha, representing a Z% change relative to baseline mean yield. For adoption outcomes, the minimum detectable effect corresponds to approximately A–B percentage point changes. These estimates will be updated using observed baseline variation prior to final analysis. Although randomization is conducted at the individual farmer level, outcomes may be correlated within villages due to shared environmental, institutional, and market conditions. To account for this intra-cluster correlation, all standard errors will be clustered at the village level. This adjustment ensures valid statistical inference while preserving the unbiasedness of treatment effect estimates.