Determinants of Adoption of Sustainable Rice Technologies: Experimental Evidence from India.

All farmers receive basic information about AWD before the experiment begins. Farmers in the treatment group then participate in a practical, hands-on AWD training session, while farmers in the control group do not receive any training. After the training session, all farmers complete the same survey.

Within the survey, all farmers first receive the same basic information about microbiome-based soil treatments: The microbiome is a biological treatment (e.g. bacteria, fungi, etc.) made from beneficial microbes that live around the rice plant roots and in the soil. It is applied as seedling dip or as a spray. Farmers who use microbiomes may see healthier root growth, better nutrient uptake, and sometimes lower greenhouse gas emissions such as methane. Results can vary depending on soil type and farming conditions.

And are then asked:

How likely are you to try microbiome next season? (0 = Not at all likely, 10 = Very likely)

After this, each farmer is randomly assigned to one of three short vignettes that provide additional information. These vignettes are shown individually in the survey, and each farmer only sees one version. Immediately after the vignette, farmers report how willing they are to try microbiomes next season.

The low adoption norm frame tells farmers that only 1/10 farmers in their area used microbiomes last season, which allows us to measure how a low descriptive norm affects willingness to adopt.

The high adoption norm frame instead presents that 7/10 farmers used microbiomes, allowing us to measure the effect of high perceived adoption rates on adoption intentions.

The safety reassurance frame informs farmers that the microbiome product does not contain GMOs and is not an antibiotic. It is described as consisting of beneficial soil microbes similar to those already found naturally in paddy soils. This frame is designed to reduce uncertainty about product safety and test whether reassurance increases farmers’ willingness to try microbiomes.

Full detail on the experimental interventions in the attached document.

2025-09-01

2026-01-15

Intention to adopt AWD (binary) (outcome 1)

Intention to try microbiomes after randomized information frames (0–10) (outcome 2).

Outcome 1 measures farmers’ forward-looking willingness to adopt AWD. It allows us to capture learning and confidence effects that may not yet translate into current-season adoption. Survey question: Are you willing to practice improved water management such as AWD in the future? (Yes/No).

Outcome 2 measures farmers’ willingness to try microbiome-based soil treatments after receiving the randomized information frames. It captures belief updating and shows whether social norms or safety reassurance shift intentions for a technology that farmers are largely unfamiliar with. Survey question: How likely are you to try microbiomes next season? (0 = Not at all likely, 10 = Very likely)

Full detail on the outcomes in the attached document.

Our project consists of two experiments:

Farmers in selected padasekharams receive hands-on AWD training, while control clusters receive only basic information. This experiment isolates whether technical guidance increases AWD adoption intentions.

All farmers receive basic microbiome information and are then individually randomized into one of three information frames. This experiment identifies how social norms and safety reassurance shape adoption intentions.

Each experiment targets a different mechanism; technical knowledge for AWD and belief updating for microbiomes.

Full detail on the experimental design and hypotheses in the attached document.

Randomization has been computed by a computer.

The design includes two separate randomization steps.
For the AWD part, we randomize at the padasheram level (village clusters), assigning each cluster to either receive the AWD training or serve as control. We use cluster-level randomization because AWD is a shared water-management practice, and treating individual farmers would create a high risk of spillovers within the same irrigation system.

For the microbiome section, the randomization happens at the individual level. Each farmer is randomly assigned to one of the three information versions within the survey. This makes sense since microbiome-based soil treatments can be adopted independently, and the information is given privately to each respondent.

Full information on randomization in the attached document.

Yes

For the AWD part, we randomize at the padasheram level (village clusters), assigning each cluster to either receive the AWD training or serve as control. We use cluster-level randomization because AWD is a shared water-management practice, and treating individual farmers would create a high risk of spillovers within the same irrigation system. Around 15 padasherams are included.

For the microbiome section, the randomization happens at the individual level. Each farmer is randomly assigned to one of the three information versions within the survey. This makes sense since microbiome-based soil treatments can be adopted independently, and the information is given privately to each respondent. It includes around 880 individuals.

Full information on randomization in the attached document.

Around 880 observations.

Sample size includes 7 treated clusters and 8 control clusters in the AWD experiment.

The sample size in the microbiome experiment consist of around 880 farmers fairly equally distributed between treated and control group.

Around 880 observations and equal assignment, the experiment is powered to detect effects of 7–10 percentage points (binary) and 0.30–0.35 SD (continuous) at alpha=0.05, power = 0.80