Attention and Salience in Preference Reversals

Last registered on August 18, 2020

Pre-Trial

Trial Information

General Information

Title
Attention and Salience in Preference Reversals
RCT ID
AEARCTR-0005985
Initial registration date
July 06, 2020

Initial registration date is when the trial was registered.

It corresponds to when the registration was submitted to the Registry to be reviewed for publication.

First published
July 08, 2020, 5:06 PM EDT

First published corresponds to when the trial was first made public on the Registry after being reviewed.

Last updated
August 18, 2020, 11:06 AM EDT

Last updated is the most recent time when changes to the trial's registration were published.

Locations

Region

Primary Investigator

Affiliation
University of Zurich

Other Primary Investigator(s)

PI Affiliation
University of Zurich

Additional Trial Information

Status
In development
Start date
2020-08-19
End date
2020-09-15
Secondary IDs
Abstract
The preference reversal phenomenon refers to a pattern of decisions under risk where decision makers explicitly value long-shot lotteries above more moderate ones but then choose the latter, in contradiction with Expected Utility Theory. The phenomenon is typically demonstrated in paradigms involving pairs of lotteries consisting of a relatively safe lottery, called the P-bet (for "probability"), and a riskier lottery offering a larger prize (a long shot), called the $-bet. Decision makers often choose the P-bet in the direct choice task, but explicitly value the $-bet above the P-bet, in contradiction with the most basic tenets of decision theories under risk, and specifically with the indifference between a lottery and its certainty equivalent. This phenomenon reveals an inconsistency between elicitation methods which should be equivalent.

Salience theory postulates that reversals can occur because certain states are overweighted due to their salience. From a process perspective, this has a direct implication for visual attention. To test this hypothesis we conduct an eye-tracking experiment analyzing decisions under risk while implementing two treatments which manipulate salience. See the hypotheses in the attached pdf.
External Link(s)

Registration Citation

Citation
Alós-Ferrer, Carlos and Alexander Ritschel. 2020. "Attention and Salience in Preference Reversals." AEA RCT Registry. August 18. https://doi.org/10.1257/rct.5985-2.0
Experimental Details

Interventions

Intervention(s)
Intervention Start Date
2020-08-19
Intervention End Date
2020-09-15

Primary Outcomes

Primary Outcomes (end points)
Saccades, fixations, lottery evaluations, and proportion of preference reversals
Primary Outcomes (explanation)
Lottery evaluations: Lowest-acceptable selling price of the lottery.

Proportion of predicted reversals: Proportion of the $-bets being evaluated higher than the P-bets, when the P-bet was chosen.

Fixations: Number of fixations in pre-specified Areas of Interest (AOIs), averaged across the relevant lotteries.

Transitions: Number of saccades and transitions from outcome-to-outcome AOIs, averaged across the relevant lotteries. Specific transitions represent the states according to Salience Theory (Bordalo et al., 2012).

Secondary Outcomes

Secondary Outcomes (end points)
Secondary Outcomes (explanation)

Experimental Design

Experimental Design
Classical preference reversal experiment conducted in individual lab sessions where subjects choose and evaluate lottery pairs. There will be two within-subject treatments for evaluating lotteries and eye movements will be measured with an eye tracker.
Experimental Design Details
Each subject faces 96 decisions constructed using 32 lottery pairs.
Each of the 32 lottery pairs consists of a $-bet and a P-bet. Each bet will have two possible outcomes, expressed in ECU (experimental currency units). Payments will be exchanged to CHF at the end of the experiment (exchange rate 1 ECU = 2.5 CHF). The $-bet consists of a high outcome (>11 ECU) with a low probability (<45%) and a second, low outcome, while the P-bet consists of a moderate outcome (<9 ECU) with high probability (>60%) and a second, low outcome. Each subject faces three decisions for each fixed lottery pair: (i) choose between the two lotteries (Choice), (ii) evaluate the $-bet, or (iii) evaluate the P-bet. Choices and evaluations are interspersed according to a pseudo-randomized sequence.

Subjects face two (within-subject) treatments for the evaluation of lotteries: Isolation and Joint evaluation. In the Isolation Treatment, the $-bet and P-bet are evaluated without the other lottery (the one not being evaluated) being present. Black circles (as placeholders) are shown where the probabilities and outcomes of the alternative lottery would have been presented.
In the Joint Treatment, the $-bet will be evaluated while a perturbed version of the P-bet is present, and vice versa (see below for the perturbed versions).
Half the lottery pairs will be evaluated jointly and the remaining half in isolation. We constructed sequences which ensure that each subject evaluates half the lotteries jointly and half the lotteries in isolation. These sequences also balance that each lottery will be evaluated jointly by half the subjects and in isolation by the other half of the subjects. Each sequence is unique, meaning that no subject evaluates the same lotteries jointly and in isolation as any other subject.

For payment, one decision (a choice or an evaluation) will be chosen randomly for each subject and either the chosen lottery will be played out (if a choice decision is sampled) or the Becker-DeGroot-Marshak mechanism will be applied (if an evaluation decision is sampled), that is, either the lottery will be played out or a monetary value will be paid.

Perturbed lotteries are designed in such a way that the higher outcome of each lottery is changed by less than 3% and the lower outcome is changed accordingly to keep a similar expected value of the lottery. The perturbation does not change the salience ranking of the states. Salience states were calculated using the same salience function as in Bordalo et al. (2012). See the attached pdf for the equation.

The Lotteries and the different outcomes are separated by two axes. Lotteries are presented on the left and right side of the screen visually, separated by a vertical line. The two possible states of a lottery (outcome with probability) are visually separated by a horizontal line. For each subject, the position of the $-bets and P-bets and the position of the higher outcome (HO) for each lottery is counterbalanced:
{$ left/P right; $ right/P left}×{HO($) top; HO($) bottom}×{HO(P) top; HO(P) bottom}.
Randomization Method
We constructed sequences that determine which lotteries will be evaluated jointly or in isolation. The computer assigns sequences to subjects.
Randomization Unit
Lotteries within subjects. (Each subject evaluates half the lotteries in isolation and half jointly. Randomization determines which lotteries will be evaluated jointly and isolated for a subject.)
Was the treatment clustered?
No

Experiment Characteristics

Sample size: planned number of clusters
64 individuals
Sample size: planned number of observations
Transitions and fixations in 6144 trials (2048 choices, 2048 joint evaluations, 2048 isolated evaluations). 128 proportions of preference reversals (2 per individual). 64 individual average number of fixations in each evaluation treatment. 64 individual average number of saccades on the most salient state and 64 individual average number of saccades on the least salient state during choice phase.
Sample size (or number of clusters) by treatment arms
It is a within treatment, 64 individuals.
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
We aim to collect data that could detect a small-to-medium-sized effect (Cohen's d=.35), which requires a sample size of N=52 for a one-tailed Wilcoxon-Signed-Rank test with alpha=.05 and a power of .80. We increase this number to 64 participants to be a multiple of 16 (=2 treatments x 8 counterbalancing measures). A total of 64 subjects yields an actual power of .87 (with d=.35 and alpha=.05). A sample of N=64 is able to detect a minimal effect of d=.31 (with power of .80 and alpha=.05). We will, therefore, collect data from 64 participants.
Supporting Documents and Materials

There is information in this trial unavailable to the public. Use the button below to request access.

Request Information
IRB

Institutional Review Boards (IRBs)

IRB Name
Human Subjects Committee of the Faculty of Economics, Business Administration, and Information Technology, University of Zurich
IRB Approval Date
2020-05-18
IRB Approval Number
OEC IRB # 2020-026

Post-Trial

Post Trial Information

Study Withdrawal

There is information in this trial unavailable to the public. Use the button below to request access.

Request Information

Intervention

Is the intervention completed?
No
Data Collection Complete
Data Publication

Data Publication

Is public data available?
No

Program Files

Program Files
Reports, Papers & Other Materials

Relevant Paper(s)

Reports & Other Materials