User interface effects in preference elicitation

The experiment employs a between-subjects, within-subjects design implemented in a controlled laboratory setting. The between-subjects intervention consists of an input format treatment (three levels): Participants are randomly assigned to one of three user interface formats for submitting their bids in a Becker–DeGroot–Marschak (BDM) mechanism: a) Button input: participants adjust their bid using increment/decrement buttons labeled –10, –5, –2, –1, +1, +2, +5, and +10. b) Slider input: participants indicate their bid using a continuous slider. c) Numeric box input: participants type their bid directly into an open text box.

Within each between-subjects treatment, participants complete two valuation rounds for four cookie products as follows:
- Round 1: participants view four alternative cookie packages and place bids for each based on visual information only.
- Round 2: participants taste the cookies and then place bids again for all four products.

At the end of the session, one product and one round are randomly selected as binding, and participants have their decisions realized with 20% chance according to the BDM mechanism procedure.

2025-11-18

2025-12-17

Bids or WTP

The study follows a three level between-subjects factorial design combined with two within-subject valuation rounds. Participants are randomly assigned to one of three input format conditions (buttons, slider, or numeric box). Randomization is conducted at the individual level using the Qualtrics randomizer to ensure equal allocation across conditions.

Within each treatment combination, participants complete two valuation rounds involving the same four cookie products. In Round 1, participants place bids based only on visual information about the packages, while in Round 2, they first taste the cookies before bidding again. This within-subject structure allows measurement of how experience (tasting) affects willingness to pay, holding interface constant.

At the end of the session, one product and one round are randomly selected as binding. Participants have their choices realized with 20% chance implemented following the BDM procedure, introducing real monetary incentives while maintaining feasibility for a lab setting. The design therefore enables identification of the causal effects of (i) input format on elicited willingness to pay, as well as (iii) within-subject changes in valuation before and after product experience.

Randomization is implemented at the individual level through the Qualtrics platform. Each participant is automatically and independently assigned to one of the three input format conditions (buttons, slider, or numeric box). The Qualtrics randomizer uses built-in random number generation to ensure equal allocation probabilities across all treatment. Assignment information is stored as embedded variables for use in subsequent data analysis.

The unit of randomization is the individual participant.

No

~64 subjects * 3 between-subjects treatments = 192 subjects

192 subjects * 4 products * 2 rounds = 1536 observations

1536/3 ~ 512 observations per treatment arm

With N=64 subjects/treatment we can detect a medium Cohen's effect size of Delta=0.4 in bids with assumed Intra-class Correlation Coefficient across the M=2 rounds of up to rho=0.3 (Kupper and Hafner, 1989): N = [2*[(Z_{1-a/2}+Z_{beta})^2]*(1+(M-1)*rho))]/[M*(Delta^2)] Where z represents the statistics and using the conventional values for α (type I error) of 0.05 and β (type II error) of 0.20 since we consider a power of 80%. Kupper, L. L., & Hafner, K. B. (1989). How Appropriate are Popular Sample Size Formulas? The American Statistician, 43(2), 101–105. https://doi.org/10.1080/00031305.1989.10475628