A Randomised Controlled Trial to Enhance Bamboo Supply and Quality in India Technical Production Information, Quality-Market Access, Behavioural Implementation Support, and Spillovers

This study will test practical ways to help bamboo-growing farmers increase the quantity and quality of bamboo they produce and sell. The interventions are designed to address three common barriers: limited technical knowledge about bamboo management, weak links to better-paying markets, and difficulty turning good intentions into action.

The first intervention is a production and action-plan training. Farmers will receive simple, practical advice on how to manage bamboo clumps better. This will include guidance on cleaning clumps, removing old or dead culms, thinning overcrowded clumps, identifying mature culms, avoiding premature harvesting, and planning when key bamboo-management activities should be carried out.

The second intervention is a quality and market-access intervention. Farmers will receive information on how bamboo quality affects price, how to sort and grade bamboo, how to compare buyers, and how to calculate the net price after transport, labour, commissions, and other deductions. Where possible, the intervention will also connect farmers with processors, industries, Farmer Producer Organizations, or other organized buyers who may be willing to pay more for better-quality bamboo.

The third intervention is a bundled intervention that combines the production training with the quality and market-access support. This will test whether farmers benefit most when technical advice is linked to a credible market opportunity.

In addition, some directly treated farmers will receive a behavioural support component. This will include a simple future-profit calculation, a discussion of avoidable losses from poor harvesting or selling decisions, a written bamboo action plan, a soft commitment to follow the plan, reminder messages or calls, and information on whom to contact for technical support. This component is designed to help farmers follow through on actions that may bring benefits later but require effort today.

Some households in treatment villages will not receive the intervention directly. These households will allow the study to understand whether information spreads naturally through neighbours, relatives, traders, or farmer groups. Control villages will not receive the intervention during the study period and will provide a comparison group for measuring impact.

2027-01-01

2028-01-01

The effectiveness of the interventions will be assessed through outcomes at four levels: behavioural mechanisms, adoption and management behaviour, market behaviour, and final production and income. This follows the structure of agricultural extension evaluations that first examine knowledge, then adoption, then production, market participation, and welfare outcomes. In this study, the main production outcomes are bamboo area, number of clumps, clumps managed, mature culms per clump, total harvestable mature culms, culms harvested, and culms sold. The main market outcomes are buyer search, sale channel, quality grade, price per culm, marketing costs, net price, and bamboo income.
To reflect multidimensional learning, the outcome framework will distinguish awareness, technical know-how, subtle know-how, adoption, and correct combinations of practices. Simple awareness will capture whether a farmer has heard about clump cleaning, quality grading, buyer premia, or delayed harvest. Technical know-how will capture whether the farmer knows how to perform specific tasks. Subtle know-how will capture whether the farmer understands complementarities, for example that quality premia require mature harvesting, sorting, and credible buyer access, or that clump cleaning must be timed with labour availability and future harvest planning. Mistakes in combinations, such as harvesting immature culms after receiving a quality-premium message or grading culms without comparing net prices, will be measured as a separate outcome category.
The universal behavioural mechanism index will combine future-profit salience, loss awareness, action-plan existence and quality, plan implementation, self-efficacy, buyer trust, and willingness to wait for a premium. This index will be measured for all households and used as a mechanism outcome. A separate treatment-fidelity index will be created only for psychology-overlay households to verify whether the behavioural intervention was delivered as intended. The treatment-fidelity index will not be used as a main impact outcome because it is mechanically related to treatment assignment.

The study uses a randomized controlled trial to evaluate interventions designed to improve bamboo production, quality upgrading, and market participation among bamboo-growing farm households in India. The experiment is implemented among listed bamboo-growing households across selected villages in Maharashtra and Karnataka.

The study follows a multi-stage randomized design. Villages are first stratified by bamboo cultivation system, distinguishing villages where bamboo is primarily grown as monocrop, mixed crop, or border crop. Within these strata, villages are assigned to different experimental arms. The main treatment arms are a production/action-plan intervention, a quality-market-access intervention, and a bundled intervention that combines both components. The comparison group includes control villages, with some control villages classified as geographically or economically closer to treatment areas and others as far pure-control villages.

Within treatment villages, only a randomly selected subset of sampled households receives the intervention directly. Other sampled households in the same treatment villages do not receive direct treatment and are included to measure possible spillovers within villages. This structure allows the study to distinguish direct effects of the intervention from indirect effects that may arise through neighbours, relatives, farmer groups, traders, or local market interactions.

The production/action-plan intervention provides practical information on bamboo clump management, thinning, cleaning, rejuvenation, mature culm identification, and harvesting practices. The quality-market-access intervention provides information on bamboo grading, quality requirements, buyer options, price comparison, and market channels, including possible links to processors, industries, farmer collectives, or organized buyers. The bundled intervention combines the production and market components.

A behavioural support component is cross-randomized among directly treated households. This component includes future-profit salience, loss framing, an individualized bamboo action plan, soft commitment, reminders, and technical-support cues. It is designed to test whether behavioural support improves follow-through on bamboo management and marketing decisions.

The primary comparisons will estimate the effect of each treatment arm relative to the far pure-control group. Additional comparisons will examine within-village spillovers, differences between near-control and far pure-control villages, and the added effect of the behavioural support component among directly treated households. Outcomes will be measured through household surveys, field verification, and, where available, buyer or transaction records. The main outcomes include bamboo management practices, number of clumps managed, harvestable culms, bamboo quality, buyer search, sale channel, price received, and bamboo income.

Not available

Randomization will be conducted in office by computer using a reproducible random-number generator. A fixed random seed will be used so that the village and household allocations can be replicated and audited.

The randomization will be implemented in stages. First, villages will be stratified by bamboo cultivation system and location. Within each stratum, villages will be randomly assigned by computer to one of the experimental village arms: production/action-plan, quality-market access, bundle, near-control, or far pure-control. Second, within treatment villages, sampled households will be randomly assigned by computer to direct treatment without behavioural support, direct treatment with behavioural support, or spillover status. Households in control villages will be sampled from villages assigned to near-control or far pure-control status.

The randomization will not be conducted by field enumerators. Field teams will receive the final assignment list after randomization is completed. The randomization file will record the random seed, strata, village IDs, household IDs, and assigned treatment status.

The study uses more than one level of randomization.

The first unit of randomization is the village. Villages will be randomly assigned to one of the main experimental environments: production/action-plan treatment, quality-market access treatment, bundled treatment, near-control, or far pure-control. Village-level randomization is used because information, learning, buyer contacts, and market behaviour may spill over across households within the same village.

The second unit of randomization is the household within treatment villages. Within villages assigned to one of the active treatment arms, sampled households will be randomly assigned to receive the intervention directly, to receive the intervention directly with the behavioural support component, or to remain untreated within the village as spillover households.

Thus, the design is a two-stage randomized design: village-level randomization determines the treatment environment, and household-level randomization within treatment villages determines direct exposure to the intervention and the behavioural support component.

Yes

Planned number of clusters: 149 villages. The study plans to randomize approximately 149 village clusters across the production/action-plan, quality-market access, bundled treatment, near-control, and far pure-control arms. Within treatment villages, sampled households are further randomized to direct treatment, behavioural-support treatment, or spillover status.

The study plans to sample approximately 3,300 bamboo-growing farm households at baseline across 149 village clusters. This sample is expected to yield around 3,000 completed endline observations after allowing for approximately 10% attrition.

Production/action-plan treatment: 825 households
This includes 275 direct-treatment households without behavioural support, 275 direct-treatment households with behavioural support, and 275 spillover households.

Quality-market access treatment: 825 households
This includes 275 direct-treatment households without behavioural support, 275 direct-treatment households with behavioural support, and 275 spillover households.

Bundled production plus quality-market access treatment: 825 households
This includes 275 direct-treatment households without behavioural support, 275 direct-treatment households with behavioural support, and 275 spillover households.

Near-control group: 275 households

Far pure-control group: 550 households

Total planned sample: 3,300 households across approximately 149 village clusters.

The power calculations are based on a planned baseline sample of 3,300 bamboo-growing farm households across approximately 149 village clusters, allowing for around 10% attrition, which gives an expected endline sample of about 3,000 households. The calculations assume 80% power, a 5% two-sided significance level, and adjustment for village-level clustering. For the main supply-side outcome, log(1 + bamboo area in acres), the standard deviation of the log outcome is 0.321. The minimum detectable effect is approximately 0.093 log points, equivalent to about 0.29 standard deviations, or a 9.7% increase in bamboo area. For the second main production-stock outcome, log(1 + number of bamboo clumps), the standard deviation of the log outcome is 1.477. The minimum detectable effect is approximately 0.278 log points, equivalent to about 0.19 standard deviations, or a 32% increase in the number of clumps. Summary: Accounting for village-level clustering, 80% power, and a 5% two-sided significance level, the study is powered to detect an effect of approximately 0.093 log points, or 9.7%, for log(1 + bamboo area in acres), where the standard deviation is 0.321. For log(1 + number of bamboo clumps), where the standard deviation is 1.477, the study is powered to detect an effect of approximately 0.278 log points, or 32%. The calculations are based on a planned baseline sample of 3,300 households and an expected endline sample of about 3,000 households after attrition.