Information frictions and sustainable purchase behavior

Participating consumers are randomly assigned to one of four groups: a control group and three treatment groups designed to relax different information frictions related to the sustainability attribute of food products. The sustainability information takes the form of an Eco-Score, an aggregate sustainability label ranging from A+ (most sustainable) to E- (least sustainable). The three treatment arms differ in the components of the information friction they target:

T1 (Salience): a short video that increases salience of the sustainability attribute.
T2 (Salience + digital access): the salience video plus information about a digital tool that lowers the cost of searching for Eco-Score information at the point of purchase.
T3 (Salience + digital access + explanation): the T2 content plus an explanation of how the Eco-Score is calculated and how to interpret it, lowering the processing cost of the information.

Treatments are delivered through a labJS-based online survey distributed by email by the cooperation partner (a major Swiss retailer) to members of its loyalty program. Post-treatment behavior is measured via transaction-level scanner data from the retailer.

2026-05-27

2027-05-26

Consumer-level purchase behavior measured from transaction-level scanner data over a 60-day pre / 60-day post window:

Average Eco-Score of purchased products (restricted to products with an assigned Eco-Score).
Purchase count by Eco-Score bracket.
Purchase value (CHF) by Eco-Score bracket.
Percentage share of purchases (counts) by bracket.
Percentage share of value by bracket.

Products are sorted into three brackets: (i) high Eco-Score (B- or higher), (ii) low Eco-Score (C+ or lower), (iii) no Eco-Score assigned.

For consumer i in period t and bracket b:

- Average Eco-Score: y_it = S̄_it, the mean Eco-Score across all Eco-Scored products purchased by consumer i in period t.

- Purchase count by bracket: y_itb = Σ_{k∈b} N_ikt

- Purchase value by bracket: y_itb = Σ_{k∈b} V_ikt

- Share of purchases by bracket: y_itb = Σ_{k∈b} N_ikt / Σ_{b'∈B} Σ_{l∈b'} N_ilt

- Share of value by bracket: y_itb = Σ_{k∈b} V_ikt / Σ_{b'∈B} Σ_{l∈b'} V_ilt

A "period" aggregates 30 days of transactions, giving two pre- and two post-treatment observations per consumer. Consumers with zero purchases remain in the sample and contribute zero-valued outcomes for count and share outcomes; the average Eco-Score is coded missing for consumer-periods in which no Eco-Scored product was purchased.

Consumer–product-category-level outcomes in three categories with high baseline Eco-Score variation: raw eggs, refrigerated orange juice, and UHT milk. Within each category, the same five outcomes as the primary set.
Product–canton–period outcomes: purchase counts and purchase values by product, canton, period, and treatment arm, for the most-sold products covering at least 80% of purchases in each of the three categories.
Medium-term effects measured over a 6–12 month pre/post window.

Consumer–product-category outcomes mirror the primary outcome definitions but are restricted to purchases within each focus category.
Product–canton–period outcomes define markets at the cantonal level (26 Swiss cantons) and aggregate transactions to the product–canton–period–treatment cell.

The study is a four-arm randomized controlled trial (one control group, three treatment arms) delivered via a labJS online survey distributed by email to members of a major Swiss retailer's loyalty program. Each participant is randomly assigned to one of the four arms with equal probability, watches the assigned video, and completes a short post-video survey. Only customers who complete the survey are included in the analysis sample.
Treatment effects are estimated from transaction-level scanner data linked to participants via their loyalty-program ID, over a 60-day pre- and 60-day post-treatment window. The treatment date common to all treated customers is defined as the day of the first survey response.
A follow-up survey is planned after the close of data collection.

Not available

Computer-based randomization implemented in labJS, using its built-in randomizer to assign each participant to one of the four arms with equal probability upon entering the survey.

Individual consumer (loyalty-program member).

No

Not clustered

15,000 consumers. If the target enrollment is not reached within one month of survey rollout, the analysis will proceed with the available sample.

~3,750 consumers per arm: 3,750 control / 3,750 T1 (Salience) / 3,750 T2 (Salience + digital access) / 3,750 T3 (Salience + digital access + explanation).

The MDE was derived by simulation from the retailer's transaction database. A 15,000-customer sample was randomly drawn from March–June 2025, customers were randomly assigned to the four arms (~3,750 per arm), and a placebo regression equivalent to the primary consumer-level count specification was run for each of 100 simulated assignments. The standard error was taken as the average across simulations. At α = 0.05 and power = 0.80, MDE = 2.80 × SE = 0.065, corresponding to a minimum-detectable treatment effect of approximately 6.7% in purchase counts. This is in the range of effects documented in comparable settings: Dubois et al. (2021) ≈14% for in-store nutrition labels, Cawley et al. (2015) ≈8.3% for supermarket nutrition ratings, and Katare & Zhao (2024) ≈25% for online behavioral interventions. Cawley, J., Sweeney, M. J., Sobal, J., Just, D. R., Kaiser, H. M., Schulze, W. D., Wansink, B. (2015). The impact of a supermarket nutrition rating system on purchases of nutritious and less nutritious foods. Public Health Nutrition, 18(1), 8–14. Dubois, P., Albuquerque, P., Allais, O., Bonnet, C., Bertail, P., Combris, P., Lahlou, S., Rigal, N., Ruffieux, B., Chandon, P. (2021). Effects of front-of-pack labels on the nutritional quality of supermarket food purchases: evidence from a large-scale randomized controlled trial. Journal of the Academy of Marketing Science, 49(1), 119–138. Katare, B., Zhao, S. (2024). Behavioral interventions to motivate plant-based food selection in an online shopping environment. Proceedings of the National Academy of Sciences, 121(50), e2319018121.