Default Bias and Coin-flipping

How to cite

Fu, Jingcheng; Zhang, Xing; Zhong, Songfa. 2026. "Default Bias and Coin-flipping." AEA RCT Registry. AEARCTR-0018697. https://doi.org/10.1257/rct.18697-1.1

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

Jingcheng Fu* , Xing Zhang , Songfa Zhong

* Primary investigator

Registry ID
AEARCTR-0018697
DOI
https://doi.org/10.1257/rct.18697-1.1
First registered
Last updated
Status
In development
Trial dates
2026-06-10 – 2026-08-10
Countries
Private
Registry record
http://www.socialscienceregistry.org/trials/18697

Abstract

This study examines how alternative decision procedures affect choices between risky lotteries. In an online, incentivized, within-subject experiment, participants repeatedly choose between pairs of lotteries under six different procedures: direct choice (benchmark), random default, delegation to a random device, randomized advice, and two procedures combining the random default with delegation to a random device or randomized advice. We study whether and how these procedures shift choice behavior relative to direct choice and random default, and whether procedural effects correlate with choice-level and individual-level measures of decision complexity.

Intervention

Participants complete binary lottery choice tasks over seven sessions, each on a separate day.

In Sessions 1–6 (Treatment Sessions), participants complete tasks under each of the six conditions, one condition per session, in a randomized order. In the benchmark condition (C0), participants directly choose between two lotteries. In the default condition (C1), one lottery is randomly assigned as the default, and the participant decides whether to keep it or switch. In the delegation condition (C2), participants may choose a lottery directly or delegate the decision to a fair coin flip. In the advice condition (C3), participants may choose directly or request a random assignment that they can then keep or switch. In the default–delegation condition (C4), one lottery is randomly assigned as the default, and the participant decides whether to keep it, switch to the other lottery, or delegate the decision to a fair coin flip. In the default–advice condition (C5), one lottery is randomly assigned as the default, and the participant decides whether to keep it, switch to the other lottery, or request a randomized recommendation that can then be accepted or rejected.

In Session 7 (Elicitation Session), participants make direct choices only, with each choice presented three times in randomized order.

Experimental design

This is a within-subject experiment with six treatment conditions. Participants complete Sessions 1–6 under each of the six conditions (64 choices per session), with condition order randomized across participants. Session 7 is the Elicitation Session, consisting of 192 binary lottery choices under the benchmark (direct choice) condition, with each choice set presented three times in randomized order to allow measurement of choice stochasticity. Participants complete each session on a separate day.

Each lottery pair consists of two lotteries with four equiprobable monetary outcomes. Lottery pairs are drawn from 16 parameterized choice sets grouped into three categories: (1) "hard" choices where lotteries have equal expected values but differ in dispersion; (2) epsilon-perturbed repetitions of hard choices with small expected value differences; and (3) "easy" choices where one lottery has a clearly higher expected value. Choice set parameters are randomly generated within pre-specified ranges.

Participants share some choice sets across conditions (for cross-condition comparisons) and face some choice sets unique to each condition (to reduce perceived repetition). At the end of the experiment, one decision is randomly selected for payment.

Design details

Randomization method
The order of conditions in Sessions 1–6 is randomized at the individual participant level. Within each session, the order of choice tasks is also randomized at the participant level. The presentation order of the two lotteries is randomized at the participant–choice level. All within-task random assignments (defaults and coin flips) use fair randomization with equal probability. Randomization is implemented by the experimental software.
Randomization unit
Individual participant. All conditions are within-subject; randomization of condition order and within-task assignments occurs at the individual level.
Clusters
The target analysis sample is 500 participants who complete all experimental sessions. Because participation requires completing multiple sessions, some attrition is expected due to participants failing to complete all sessions. We anticipate an attrition rate of approximately 5–10%. To achieve the target analysis sample of 500 participants, approximately 550 participants will be recruited initially. Participants who do not complete all sessions will be excluded from the main analysis sample.
Observations
Around 500 * 576 = 288,000 (number of participants × number of choice tasks per participant). All participants complete all six conditions (within-subject design). Each participant makes 64 decisions per treatment condition in Sessions 1–6 and 192 decisions in the Elicitation Session (Session 7), for a total of 576 decisions per participant.
Arms
Around 500 * 64 = 32,000 (number of participants × number of choice tasks per participant). Each participant makes 64 decisions per treatment condition.
Primary outcomes
The primary outcomes are (i) whether the participant accepts or switches away from an assigned default (in conditions with defaults: C1, C4, C5); (ii) whether the participant delegates the decision to the random device (in delegation conditions: C2, C4); (iii) whether the participant seeks randomized advice and accepts the advice after receiving a randomly assigned lottery (in advice conditions: C3, C5).
Secondary outcomes
Secondary outcomes include measures of (i) choice variability; (ii) choice heterogeneity; (iii) violations of transitivity; and (iv) response times, measured at the choice level and/or the participant level using Elicitation Session data.

Cited by

No citation data available for this registration yet.

Keywords