Experimental Design
The experiment will be conducted in Bathinda, Sangrur, and Ferozepur districts of Punjab, India. Seed villages were selected using administrative data on subsidized crop residue machinery ownership. Eligible seed villages have at least one service provider and at least two individually owned Super seeders (the two machines need not belong to different owners). Selection was stratified by district. Selected seed villages were required to be at least 5 km from other selected seed villages. The planned sample is 130 seed villages, distributed across three districts as follow:
- Bathinda: 44
- Sangrur: 44
- Ferozepur: 42
Recruitment will use a common (neutral) explanation of the study objectives. For listing in the flyer, providers must be based (reside) in a treatment seed village, own an operational Super Seeder, and intend to offer services to other farmers during the upcoming sowing season (2026 winter sowing season). The field team will: (1) Obtain both consent for research participation and consent for publication of contact details, (2) Verify each provider’s preferred public name, base village, and telephone number, and (3) Maintain a log/record of recruitment attempts, eligibility check result, refusals (if any), and flyer listing consent. The recruitment script will explain that participation does not guarantee additional customers and that participation is completely voluntary. Recruitment team members will not arrange bookings, introduce customers, recommend prices, offer marketing advice, or describe service providers as government-endorsed or quality-certified.
Households in treatment seed villages will receive Punjabi-language flyers (in Gurmukhi script) with the following information, whose English translation is:
- Header line: “For wheat sowing services using a Super Seeder, contact any of the service providers listed below.”
- A list of verified service providers (Name and phone number), listed in no particular order
- Name of the district and the village
- Disclaimer: “These service providers are listed only to help farmers book Super Seeder wheat sowing services. The names are listed in no particular order. This list is for information only and does not imply endorsement of any service provider.”
One neighboring village per treatment village will also receive the same flyer as distributed in the corresponding treatment village. The criteria for selecting the neighboring village is as follows. First, we find a list of qualified candidate neighboring villages based on this criteria: (1) The candidate neighbor village’s polygon must be adjacent to the treatment village’s polygon and (2) The candidate neighbor village’s polygon must not be adjacent to any other seed village’s polygon (treatment or control) or any another selected neighbor. If there is only one neighbor candidate that meets these two criteria, that village is selected. If there are multiple neighbor candidates, the candidate neighbor with the least number of super seeder machines purchased via the government subsidy program is selected. If there remains a tie at this stage, the candidate village is selected with the closest centroid-to-centroid straight-line distance. In all cases, the selected village neighbor must be unique i.e. each picked neighbor village can only be assigned to one treatment village. At the end of this process, 4 treatment villages remained without a valid neighbor. The following manual procedure was utilized for these exceptions: (1) pick the neighbor village whose polygon minimally shares a border with another already picked neighbor village, (2) if selection still unsatisfied, pick the neighbor village whose polygon minimally shares a border with a treatment village, and (3) if selection remains unsatisfied, pick the neighbor village whose polygon minimally shares a border with another control village. For the exceptions that we resolved, we were able to entirely use (1) and minimally used (2). We never needed to assign a village neighbor based on (3).
Control seed villages will receive no study-provided information or flyers. No flyers will be distributed in any neighboring villages to control villages and service providers and farmers in control villages may continue using existing advertising, information sharing, and coordination channels.
Flyers will be distributed only within treatment seed villages (and their one corresponding selected neighbor village). Local people will be paid and recruited to distribute flyers in the treatment and one neighboring village. The goal will be to reach as many households as possible either by visiting door-to-door or distributing flyers at community gathering places, like the Gurudwara.
Treatment assignment will be stratified by district and will be at the seed village level (i.e. clustered RCT design). The experimental unit is the seed village and the intervention is intended to expand demand for services offered by Super Seeder service providers based in the randomized seed village. The study design additionally specifies a geographic allocation procedure to assess separation between treatment and control locations and covariate balancing (to limit contamination as much as possible). 5 km of separation was determined to be the proper distance as service provision in outside villages falls sharply after 5 km, on average.
Villages were randomized into treatment and control using district-stratified matched-pair randomization. Within each district, villages were ordered according to the number of Super Seeder service providers, a Mahalanobis index of baseline average super seeder service-market characteristics in each village (average number of super seeder units available per provider for providing services, 2025 acres served per machine, 2025 farmers served per machine, villages served per machine, and maximum service distance), and the total number of super seeder machines purchased in a machinery till date using the crop residue management subsidy. Adjacent villages in this ordering were grouped into pairs. Within each matched pair, one village was randomly assigned to treatment and the other to control with equal probability. Randomization was conducted using a pre-specified random-number seed. The number of villages assigned to treatment and control by district is as follows:
- Bathinda: 22 treatment villages and 22 control villages
- Sangrur: 22 treatment villages and 22 control villages
- Ferozepur: 21 treatment villages and 21 control villages
With the exception of the total number of super seeder machines, all variables are constructed from baseline survey data. Total number of super seeder machines was calculated using district-level government-supplied super seeder purchasing data based on government subsidy.
Baseline survey of service providers was conducted in August and September 2026 across all study villages prior to the rice harvesting season. All the service providers will be revisited after the wheat sowing season (tentatively scheduled to begin at the end of November) to collect endline service provider-level information. At this time, all respondents will receive one small gift in appreciation of their time. Each gift will be valued between INR 200 - 300.