An impact assessment of EAMDA's banana initiative to increase technology adoption by smallholder farmers in Kenya

Last registered on June 18, 2018


Trial Information

General Information

An impact assessment of EAMDA's banana initiative to increase technology adoption by smallholder farmers in Kenya
Initial registration date
November 21, 2017

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
November 21, 2017, 9:09 PM EST

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

Last updated
June 18, 2018, 7:33 AM EDT

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


Primary Investigator

University of Sydney

Other Primary Investigator(s)

PI Affiliation
University of Nairobi
PI Affiliation
Save the Children
PI Affiliation
World Bank

Additional Trial Information

On going
Start date
End date
Secondary IDs
Approximately 67% of the Kenyan population and approximately 80% of the poor live in rural areas, deriving their livelihoods from agriculture. However, agricultural productivity is low and diminishing; production declined by 5% and 2.3% in 2008 and 2009, respectively (World Bank 2009). While productivity enhancement and poverty reduction is a multifaceted challenge, adoption of modern technology especially in agriculture can play important role in poverty reduction. Successful technology adoption, however, appears to be constrained by a lack of access to information, and know-how, among others. Providing information and training on modern inputs and cropping practices and encouraging farmers to set goals can go a long way to adopt modern agricultural technology.

EA-MDA, in partnership with the Banana Growers Association of Kenya (BGAK), is taking concrete steps to solve the problems faced by small holder banana farmers. In Kenya, banana is consumed both as a fruit as well as cooked food, and it is an importance source of carbohydrates, essential vitamins and minerals. There are approximately 270,000 smallholder banana farmers in Kenya, and about half of them are women. To enhance the capacity of the farmers and FOs, EA-MDA will train farmers and FOs on aspects such as land preparation, planting and variety selection, weed control, harvesting, grading and post-harvest handling and record keeping.
The proposed study attempts to evaluate the effectiveness of EA-MDA’s interventions using a randomized control trial. The study will find the extent to which the interventions increase the rate of adoption of new technologies, productivity, and ultimately, farmers’ income. There is also growing evidence on various behavioural constraints in technology adoption and fostering income. Drawing on lessons from behavioural literature, we will also test the extent to which a nudge for goal setting may affect technology adoption will also be tested. This particular aspect is designed not only for assessing the underlying behavioural constraints in technology adoption but also for improving the program of our implementation partner.

In addition, the study will measure spill-over effects and social learning, which are often important policy parameters in themselves.
The proposed study will inform policy makers on how farmers and farmer organizations can learn to manage their operations and adopt modern technologies. The results of this study will fill an important gap in terms of information to feed into agricultural development and poverty reduction policy in Kenya and other developing countries.
External Link(s)

Registration Citation

Chowdhury, Shyamal et al. 2018. "An impact assessment of EAMDA's banana initiative to increase technology adoption by smallholder farmers in Kenya." AEA RCT Registry. June 18.
Former Citation
Chowdhury, Shyamal et al. 2018. "An impact assessment of EAMDA's banana initiative to increase technology adoption by smallholder farmers in Kenya." AEA RCT Registry. June 18.
Sponsors & Partners

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

Request Information
Experimental Details


The program “Building a competitive banana industry in Kenya” that we evaluate is developed by East Africa Market Development Associates (EA-MDA) with financial support from the Alliance for a Green Revolution in Africa (AGRA). The program is attempting to tackle four interrelated problems:
Problem 1, Production: Farmers and Farmer Organizations (FOs) do not know how to produce modern variety and high quality banana for high-end retail stores such as Nakumatt and Naivas.
Problem 2, Post-harvest handling: Smallholder farmers lack knowledge on post-harvest handling of banana thereby causing significant waste and spoilage when bananas are being moved from farms to markets.
Problem 3, Marketing: Smallholder farmers currently lack access to high-value markets and cannot sell their bananas to intermediaries (consolidators) who procure for such markets.
Problem 4, Behavioural constraints in translating goals into actions: This is an add-on component that we intend to pilot to strengthen the core model of EA-MDA. Training on new technology may fail to increase adoptions due to various obstacles (such as failing to get started, becoming distracted) that cause friction from goals to actions. There are strong evidences in psychology literature on potential strong effects of simple plans. We intend to develop an intervention on goal setting that will involve working with the farmers to set individualized implementation plans, setting a course of actions and a simple reminder.
To address these problems, EA-MDA is offering information, training and marketing links. The training will provide farmers with necessary agronomic practices and expertise on the handling of planting materials from nursery to the fields and the care required in the initial phase of production to get the right quality. It will also provide training on post-harvest technologies so that farmers acquire the technical expertise to ensure that products are properly sorted and graded, and there are no losses when banana is harvested. In addition, EA-MDA will work with intermediaries who procure for large supermarket chains so that farmers can sell their high-quality banana to them and receive a price premium. Finally, it will also work with FOs to build their technical and institutional capacity for sustainability and scaling up.
In order to evaluate their impacts, EA-MDA will work with some FOs/villages while others will act as comparison groups. In FOs/villages that will have interventions, two sets of interventions are considered: all members of the farmer groups will be provide with information, training and marketing links, while half of the training participants will be trained on goal setting. Intended beneficiaries of EA-MDA’s interventions are smallholder farmers, their FOs, as well intermediaries involved in agricultural value chain. Special considerations will be provided to female smallholder farmers who historically have had lower access to information, training and other services than their male counterparts. EA-MDA expects that due to its interventions, farmers will adopt modern technologies and farming practices; input suppliers will supply modern inputs and technologies; and procurers will purchase output from them.
Intervention Start Date
Intervention End Date

Primary Outcomes

Primary Outcomes (end points)
In this evaluation, we will use groups of outcome indicators related to technology adoption, yield and income. The specific outcome indicators to be used for the study are listed below
• Banana cultivation (outcome class 1)
- Dummy variable if a farmer has used any land for banana cultivation
- Number of plots used for banana cultivation
- Total amount of land used for banana cultivation

• Technology adoption (outcome class 2)
- Dummy variable equal to one if the farmer has cultivated any of the modern varieties (FIA, Giant, Cavendish, Uganda Green, Williams, Grand Nine or Valery); and zero if not cultivated (including those who did not do any banana cultivation).
- Dummy variable if used banana plantlet that is purchased (as a validation of whether actually used TC)
- Dummy variables if the farmer used Tissue Culture (TC) plantlets
- Number of plots used for modern varieties (specified earlier in this outcome class) of banana cultivation
- Total amount of land used for modern varieties of banana cultivation (by aggregating the size of the plots that are used mainly for banana cultivation, and banana varieties include modern ones)

• Farming practices for banana cultivation (outcome class 3)
- Dummy variable if the farmers have done weeding
- Dummy variable if the farmers have hired labour for weeding
- Dummy variable if the farmers have used any fertilizer or manure
- Dummy variable if the farmers have used any chemical fertilizer
- Dummy variable if the farmers have used pesticides
- An index of improve farming practices from the above indicators
- Amount of money spent for plantlets, hiring labour, fertilizer and pesticides for banana cultivation.

• Banana production (outcome class 4)
- Total amount of banana produced in the last season. The local measures of production will be converted to kg.
- Banana yield (in kg) per acre
- Value of total banana production. (Unit price of sales will be used for the farmers who sold any banana. For the farmers who did not sell any banana, median price for the village will be used for imputing value)

• Sales (outcome class 5)
- Total amount (in kg) of banana sold
- Total revenue from banana sales
- Dummy variable if sold through EAMDA supported marketing agents
- Dummy variable if have a written contract for the banana sales

• Household income (outcome class 6)
- Net income from banana production - by deducting the direct expenses for banana cultivation from the value of production. This variable will not include opportunity cost of own labour.
- Total household income (including banana, other crops, livestock, non-farm business and wage employment). Data from section 12 will be used for this total income calculation, and remittance and pension income will be included in this total income.
- Cash savings, household assets, outstanding debt
Primary Outcomes (explanation)

Secondary Outcomes

Secondary Outcomes (end points)
In addition to these 6 outcome classes, we will look into several set of indicators to explore possible mechanisms of change and for spillover effects
- Assess the outcome class 3 for other major crops as spillover effects on general farming practices.
- General farming practices on using hybrid seeds, fertilizer and irrigation in the last one year. These are variables from Section 7 of the questionnaire, and are not asked for specific crops.
- Total outstanding loans and use of borrowed money for covering cost of banana cultivation

For the endline, the study is also collecting data on three additional modules as secondary outcomes. These are
- Locus of control (10 item LOC module)
- Time preference (preference from hypothetical payment options with different interest rates at present vs. 3 months/4 months/15 months in future)
- Risk preference (preference from 6 options of increasing risk and expected pay-offs)
Secondary Outcomes (explanation)

Experimental Design

Experimental Design
The study adopts a randomized control design that involve both village and household level randomization. The study targeted 90 villages as study sites that were identified by the EAMDA for possible intervention. We aimed to survey 50 households per village at baseline with a total sample of 4,500 farmer households. At the first stage of randomization, villages were randomly assigned to five groups to vary the intensity of interventions. The 30 villages that are assigned to ‘pure control’ group will not have any intervention from EAMDA during the evaluation period.

In the other four types of villages, 15 villages in each type, the number of intervention households are 10, 20, 30 and 40 respectively. In these villages, the treatment households are randomly selected from the baseline survey, and the list is provided to the implementation teams to invite at the training sessions. The control households in these intervention villages are our “spillover sample”. Finally, half of the households from the treatment group were randomly selected for receiving the goal setting intervention. Therefore, we have five types of villages and four types of households in the study. While the village types vary in the number of treatment households, the household types are
Group 1: Control group (E1 in the figure above)
Group 2: Spillover group (A2, B2, C2 and D2)
Group 3: Treatment households (A1, B1, C1 and D1)
Group 4: Goal setting households (A2G, B2G, C2G and D2G, which is a sub-sample of Group 3)
Experimental Design Details
Randomization Method
Randomization done in office by a computer
Randomization Unit
Both village level (whether to provide training) and individual level (for spillover and goal setting exercise) randomization
Was the treatment clustered?

Experiment Characteristics

Sample size: planned number of clusters
90 villages
Sample size: planned number of observations
We expected to interview 4,500 households from 90 villages. On average, 53 households were interviewed per village with a total sample of 4,719 households during baseline.
Sample size (or number of clusters) by treatment arms
60 villages in treatment groups, 30 villages in control group. The 60 villages are further randomized into four groups (15 villages in each) for intensity of interventions (10, 20, 30 and 40 households). These treatment households are randomly selected from the eligible households (53 on average per village)
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
There are two levels of power calculation. Firstly, using village level randomization, we estimate the direct effects (on treatment households in treatment villages) and spillover effects (on control households in treatment village) against the control villages. Using 90 clusters, of which 2/3 treatment, with 18 households per cluster, the standardized MDE is 0.2 (at 80% power, 5% level of significance and 0.1 ICC). This 0.2 is a conservative estimate as it does not account for the larger number of households in control villages and baseline correlation of the outcomes. Secondly, at the individual level randomization (i.e. comparing treatment and control households in treatment villages), the MDE is much lower. For example, with 2700 households from the treatment villages (assuming 10% attrition from the intended 50 HH per cluster) divided into a treatment and control, the MDE for binary outcomes is 0.03 if the control group proportion is 0.1 (e.g. TCB use) to 0.054 if the control group proportion is 0.5 (e.g. fertilizer use for banana). For the goal setting component, which is randomized at household level, these MDEs are 0.046 and 0.076 respectively (with 1350 HHs in treatment and control groups).

Institutional Review Boards (IRBs)

IRB Name
Human Research Ethics Committee, University of Sydney
IRB Approval Date
IRB Approval Number
Analysis Plan

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

Request Information


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