Labor Constraints to Technology Adoption: Evidence from Burundi

Last registered on September 10, 2023


Trial Information

General Information

Labor Constraints to Technology Adoption: Evidence from Burundi
Initial registration date
March 07, 2022

Initial registration date is when the trial was registered.

It corresponds to when the registration was submitted to the Registry to be reviewed for publication.

First published
March 08, 2022, 1:53 PM EST

First published corresponds to when the trial was first made public on the Registry after being reviewed.

Last updated
September 10, 2023, 6:39 PM EDT

Last updated is the most recent time when changes to the trial's registration were published.



Primary Investigator

University of California, Berkeley

Other Primary Investigator(s)

PI Affiliation
University of California, Berkeley
PI Affiliation
University of Burundi, Université Clermont Auvergne
PI Affiliation
Henan University

Additional Trial Information

Start date
End date
Secondary IDs
Prior work
This trial does not extend or rely on any prior RCTs.
Agricultural productivity in Sub-Saharan countries is low despite the existence of technologies available to farmers that might profitably increase yields. In this project, we propose that labor market frictions constrain the diffusion of new agricultural technologies. Specifically, we argue that take-up of new agricultural technologies might hinge upon the availability of a rural labor force skilled in these technologies, and that farmers may underinvest in training others in these skills if they cannot fully appropriate the returns of this training. We design an RCT in partnership with the NGO “One Acre Fund” in Burundi (OAF) to test the hypotheses that 1) contracting frictions that prevent farmers from fully appropriating returns from training contribute to inefficiently low level of training on these technologies in equilibrium and 2) that encouraging training and the increasing the stock of skilled workers can increase the adoption of productivity-enhancing technologies.
External Link(s)

Registration Citation

Cefala, Luisa et al. 2023. "Labor Constraints to Technology Adoption: Evidence from Burundi." AEA RCT Registry. September 10.
Experimental Details


We use two interventions that are both designed to increase the return to farmers for undertaking the training of village laborers in improved agricultural planting practices.

• Treatment 1 – Financial Incentives: We offer One Acre Fund farmers financial incentives if they spend a minimum amount of time ¬training a laborer in the village in how to use the improved agricultural technology (planting practices)
• Treatment 2 – Labor Insurance: In a smaller sample of villages, we tell farmers that if they train the laborer, the team will ensure that the worker will work for the farmer for two days at the prevailing wage of a laborer that does unskilled labor in the village.
Intervention Start Date
Intervention End Date

Primary Outcomes

Primary Outcomes (end points)
Primary Outcomes

• First stage (at training):
o Does farmer train laborer (spends certain amount of time training laborer during training session)?

o Perfomance of laborer in conducting improved planting practices under time pressure (incentivized task). This mimics the real work conditions laborers face when hired by a farmer.

Post Training: We focus on three classes of outcomes after our intervention:
1. Whether the improved agricultural technology (planting practices) are adopted by farmers
2. Whether a laborer is hired to conduct the improved agricultural technology (planting practices)
3. Total earnings and days worked of workers selected to be trained as part of the project

We describe these in more detail below:

a. Adoption:
a. The training focuses on two modern planting practices: 1) correct spacing of plants, 2) correct dosage and application of fertilizer and compost. Our primary outcome will be spacing because this outcome we are able to conduct field audits to verify the self-reports. We focus on beans – the main crop in the current planting season and the one for which the farmers received the training. We also look at other crops for which spacing can also apply (maize and potatoes) as robustness.

b. Labor demand:
We measure:
a. Whether and how many days an individual hires a laborer to do the improved agricultural practices
b. Whether and how many days an individual hires someone invited to be trained
c. In T2 only – we measure whether the worker for the individual that trained them, and what work that individual did via self-reports and direct observations.

c. Labor supply
We measure days worked and earnings during the agricultural season

In addition, if we see effects at this stage we will also measure yields and profits at the end of the agricultural season
Primary Outcomes (explanation)
i. For the spacing technology – we include both field audits and self-reported data:
1. Field audits: we measure i) the number of randomly selected points on a field with correct spacing and ii) whether a field has any randomly selected points with correct spacing. Fields to audit are randomly selected among the plots planted by the respondent in the current season. We over-sample beans plots (our main outcome of interest). Correct spacing consists of planting lines apart from one another with a consistent and correct distance (within a margin of error), as well as also planting pockets consistently and with correct spacing (within a margin of error)
2. Self-reported data: The number of the farmer’s plots with i) any correct spacing (as described above) and ii) where the majority of the plot (≥50% of the surface) is spaced correctly.

Secondary Outcomes

Secondary Outcomes (end points)
Secondary outcomes:

a. Adoption:
a. We plan to also look at the number of plots with self-reported adoption of correct dosage practices. Since this measure is not verifiable, we consider it a secondary outcome
b. We will also look at area planted with the improved agricultural technology
b. Labor supply
a. We will explore days worked and earnings by the type of work done (modern planting techniques vs. traditional, other jobs)
Secondary Outcomes (explanation)
Correct dosage consists of utilizing the correct quantity, and applying correctly, fertilizer, compost and seed

Experimental Design

Experimental Design
In the experiment, we invite a sample of experienced farmers – who master our planting technology and are, on average, net buyers of labor – to select a potential laborer who does not know the improved agricultural planting practices.

We then randomize farmers (at the village (sous-colline) level) into one of three treatments:

• Treatment 1 – Financial Incentives : We offer One Acre Fund farmers financial incentives if they spend a minimum amount of time training a laborer in the village in how to use the improved agricultural technology (planting practices)
• Treatment 2 – Labor Insurance : In a smaller sample of villages, we tell farmers that if they train the laborer, the team will ensure that the worker will work for the farmer for two days at the prevailing wage of a laborer that does unskilled labor in the village.
• Control – Farmers are given an unconditional financial incentive to farmers equivalent to the conditional incentive given to farmers in Treatment 1. It is suggested to them that they could train the laborer they brought in the planting practices but no further incentives are given.

We then construct our sample as follows. We randomly sample farmers and workers who take part in the training. This sample is used to measure the returns to training for the trainers and trainees.

In addition, we sample other individuals from the same village (sous-colline) who do not participate in the training, which we call our “spillover sample”. In this sample, we oversample individuals who have previously hired laborers to work on their fields.
Experimental Design Details
The planting technology that we encourage farmers to teach during this training is made up of two components. The first is component is spacing between pockets and lines. The second is proper application of fertilizer, compost and seed. The techniques that we teach in our training are more time-intensive than traditional planting methods in Burundi.

The spillover sample that we use is used to measure whether the returns to training also accrue to those not involved in the training (that is, whether individuals who do not incur the cost of training still accrue its benefits).

In August/September 2023 - we conduct an additional test of labor insurance. In this treatment, we ask 150-200 workers to identify a casual laborer who doesn't know the improved agricultural technology, works as a laborer and is not a member of their household. We then randomly assign employers and laborers to two groups. One is a control group in which a worker receives an unconditional amount of money. The second group is a labor insurance group. In this group, the employer and worker are told that the worker is going to receive an additional subsidy payment paid by us if they return to work for the employer during the planting season - which aims to increase the probability that the employer believes the employee will return to work for him or her during planting season. This randomisation is completed at the individual level.
After treatment status is assigned, individuals are told that in the upcoming days we will hold a training event in the village for several days. This training event will involve us providing land where individuals could train laborers (or others) in tasks they think are valuable for them. Our core outcome measure is then who decides to attend the training event, and the amount of time they spend training any person at the event.


We plan to look at the following sources of heterogeneity:

• We plan to look at heterogeneity in our treatment effects contingent on whether the worker invited to training had ever previously been hired by the person training them – as measured through self-reports
• By previous usage of technology
• By previous demand for labor
Randomization Method
Office by computer
Randomization Unit
Village (sous-colline)
Was the treatment clustered?

Experiment Characteristics

Sample size: planned number of clusters
~90 villages
Sample size: planned number of observations
In each village we sample 40-55 individuals: 15 Trainers 15-30 Spillover 10 Trainee
Sample size (or number of clusters) by treatment arms
10% Treatment 2 and the remainder split between Treatment 1 and Control.
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)

Institutional Review Boards (IRBs)

IRB Name
UC Berkeley
IRB Approval Date
IRB Approval Number


Post Trial Information

Study Withdrawal

There is information in this trial unavailable to the public. Use the button below to request access.

Request Information


Is the intervention completed?
Data Collection Complete
Data Publication

Data Publication

Is public data available?

Program Files

Program Files
Reports, Papers & Other Materials

Relevant Paper(s)

Reports & Other Materials