Discrimination as default

How to cite

Siddique, Abu; Goerg, Sebastian; Vlassopoulos, Michael; Zenou, Yves. 2026. "Discrimination as default." AEA RCT Registry. AEARCTR-0019349. https://doi.org/10.1257/rct.19349-1.0

Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.

Abu Siddique* , Sebastian Goerg , Michael Vlassopoulos , Yves Zenou

* Primary investigator

Registry ID
AEARCTR-0019349
DOI
https://doi.org/10.1257/rct.19349-1.0
First registered
Last updated
Status
In development
Trial dates
2026-08-24 – 2026-11-30
Countries
Private
JEL codes
C91, J71
Registry record
http://www.socialscienceregistry.org/trials/19349

Abstract

This randomized online experiment, implemented in Qualtrics with participants recruited through Prolific, examines whether discriminatory purchasing patterns persist after seller identity ceases to predict product quality. Participants first learn a quality difference between two seller types and are subsequently assigned to information or procedural interventions under routine or deliberative decision environments. We test whether seller-based purchasing gaps persist after quality differences and measured beliefs have converged, and whether different interventions reduce this persistence.

Intervention

Participants are assigned to Group X or Group Y. In Phase 1, one seller is more likely to provide good products in Group X, conditional on the product rating. In Group Y, both sellers are equally likely to provide good products. In Phases 2-4, seller identity does not predict quality in either group.

In Phase 3, Group X participants are assigned to one of five conditions:
- Routine Control: no intervention and a decision timer.
- Routine Information: participants are told that seller identity no longer predicts quality, with a decision timer.
- Routine Procedure: participants confirm the product rating before deciding, with a decision timer.
- Deliberative Information: participants receive the information intervention without a decision timer.
- Deliberative Procedure: participants complete the rating-confirmation procedure without a decision timer.

Group Y remains in the routine control condition. In Phase 4, all participants return to routine decisions without an intervention. Which seller is initially designated as the better seller is independently randomized.

Experimental design

The experiment has four phases. Phase 1 contains 50 decisions, Phase 2 contains 10 decisions, Phase 3 contains 15 decisions, and Phase 4 contains 10 decisions.

Five arms belong to Group X and experience a Phase 1 quality difference between sellers. These arms are Routine Control, Routine Information, Routine Procedure, Deliberative Information, and Deliberative Procedure. The sixth arm is Group Y, in which both sellers have equal quality throughout the experiment, and serves as the pure control group for comparison.

In Phases 2-4, both sellers follow the same quality schedule. In Phase 3, Group X participants receive their assigned intervention and decision environment. In Phase 4, all participants return to routine decisions without an intervention.

The primary Phase 3 analysis compares each Group X intervention arm with Routine Control. The main interaction test compares the relative effectiveness of Information and Procedure across routine and deliberative environments. We also examine whether seller-based purchasing persists when Phase 2 beliefs about the two sellers are equal or nearly equal.

The target sample is 1,500 participants, with 250 participants in each arm.

Design details

Randomization method
Randomization is performed by computer within Qualtrics. Participants are assigned with equal probability to one of six arms. The seller designated as initially better is independently randomized between Seller A and Seller B with equal probability.
Randomization unit
The unit of randomization is the individual participant. Treatment arm and the identity of the initially better seller are both randomized at the participant level.
Clusters
N/A
Observations
1500
Arms
250
Primary outcomes
Participant-level rating-adjusted seller gap in Phase 2. Participant-level rating-adjusted seller gap in Phase 3. Participant-level rating-adjusted seller gap in Phase 4. Phase 3 differences in seller gaps across treatment arms. Phase 3 seller gap when Phase 2 beliefs about the sellers are equal or nearly equal. Phase 3 crossover effect between intervention type and decision environment.

Cited by

No citation data available for this registration yet.

Keywords