Discrete Choice Experiment on Ecological Attributes in the Luxury Sector

Participants complete an online discrete choice experiment involving leather bags. Each participant first selects the bag they like most from nine options. This selected bag is then used throughout the choice experiment, so the alternatives differ only in price, material, and the presence of a third-party environmental certification.

Participants are randomly assigned to one of four experimental groups defined by product segment and treatment condition. They evaluate the bags either as luxury goods or as standard consumer goods and are assigned either to the information treatment or to the control group. All participants first complete a set of eight choice tasks, choosing each time between two product alternatives and an opt-out option.

Participants assigned to the information treatment then receive additional factual information about the firms' segments that use the certification standard in their leather supply chains. Participants assigned to the control group receive a text of comparable length that repeats previously provided information about the certification. After this intervention, all participants complete the same set of eight choice tasks again.

2026-08-28

2026-08-31

The primary outcome is the participant's choice among product A, product B, and an opt-out option. The primary treatment effect is the coefficient on the Eco-label × Treatment × Post interaction, estimated separately for the luxury and standard segments.

Each participant contributes sixteen choice occasions to the main analysis, with eight choice tasks completed before and eight after the information stage. The same eight tasks are used in both sequences, with task order and the left-right position of the product alternatives randomized.

The secondary outcomes include the valuation of the eco-label, recycled leather, and alternative leather across the luxury and standard segments. We also estimate willingness to pay for these attributes and examine how their valuation varies with green consumption values and conspicuous consumption orientation.

Green consumption values are measured using the six-item GREEN scale (Haws, Winterich and Naylor, 2014), and conspicuous consumption orientation using the eleven-item scale of Chaudhuri, Mazumdar and Ghoshal (2011). Both scales are administered in French translations obtained by back-translation, scored as the mean of their items, and standardized.

Willingness to pay for the eco-label and material attributes is estimated using a companion specification in which price is entered as a continuous percentage deviation from the participant's stated price.

Each choice task presents two product alternatives and an opt-out option. The alternatives differ in price, material, and the presence of an environmental certification.

We constructed a six-task efficient design for the pilot using priors derived from the literature review and the focus group. To respect realism we constraint the appereance of the ecolabel with the attribute genuine leather. The pilot design was selected based on D-efficiency and the minimum expected variance of the Eco-label × Treatment × Post interaction coefficient.

Based on the pilot results, two additional choice tasks were added to improve the identification of trade-offs that were insufficiently represented in the initial design, resulting in the final eight-task design.

Not available

Randomization is performed by computer using oTree. At session creation, participants are assigned evenly across the four experimental cells, and the assignment order is randomized using a session-specific seed.

Individual participant

No

The design is not clustered

1100 participants

275 participants per experimental cell (luxury-treatment, luxury-control, standard-treatment, standard-control), for a total sample of 1100 participants.

The study is powered to detect a 10 percentage point change in the probability of choosing the eco-labelled alternative between the first and second choice sequences in the treatment group relative to the control group. The expected effect is negative in the luxury segment and positive in the standard segment. Power calculations use one-sided tests at the 5% significance level and 80% power. With 550 participants per segment, the minimum detectable effect is approximately 7 percentage points under independence across choice occasions and approximately 8-10 percentage points when allowing for correlation between repeated choices made by the same participant.