Action Bias or Illusion of Control? - Complexity Controlled, Within Subject

How to cite

Psurek, Aleksander; Mollerstrom, Johanna; McCabe, Kevin. 2025. "Action Bias or Illusion of Control? - Complexity Controlled, Within Subject." AEA RCT Registry. AEARCTR-0015612. https://doi.org/10.1257/rct.15612-2.0

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

Aleksander Psurek* , Johanna Mollerstrom , Kevin McCabe

* Primary investigator

Registry ID
AEARCTR-0015612
DOI
https://doi.org/10.1257/rct.15612-2.0
First registered
Last updated
Status
In development
Trial dates
2025-07-01 – 2025-10-15
Countries
Private
JEL codes
C91
Registry record
http://www.socialscienceregistry.org/trials/15612

Abstract

In response to our previous laboratory experiment, pre-registered under "Action Bias or Illusion of Control?" (RCT ID: AEARCTR-0012294), we have identified the complexity of our willingness-to-pay / willingness-to-accept elicitation as a potential confounding factor. In this updated design, we again study the presence of an illusion of control and an action bias in two simple laboratory games. We designed these games to be as similar to each other as possible, with both involving the rolling of a pair of dice and relating to the values these dice come up on. The primary difference between these games is the presence of an element of luck (stochasticity, risk) in one game but not in the other, thus allowing us to separate the illusion of control from any present action bias.

In our previous value elicitation we used a third price auction stylized as an incrementing / decrementing price clock, though allowing asynchronous resolution. We similarly focused on a comparison across rather than within subject. In our new design, we are offering subjects a fixed price offer at two price points and have subjects play both dice games. In this way we simplify the value elicitation to decrease significantly the cognitive demands on subjects, while still testing for the presence of both the illusion of control and the action bias, and allow for tighter comparisons by using a within subject design. We do however give up the ability to measure the size of these effects at the individual level.

Experimental design

We conduct a lab experiment employing computer interfaces and physical dice. Subjects participate in both a stochastic and a deterministic game involving a pair of dice. Subjects either roll the pair of dice themselves or the experimenter rolls the pair of dice on their behalf. The outcome of the dice roll is non-instrumental in determining the payoff for the subject and subject are informed of this fact. The order of the dice games is randomized and counterbalanced.

The subject determines who will roll the pair of dice for them, for a price. This price may be $0.25 or $1.00, depending on the treatment. In one treatment, the subject starts with the ability to roll the dice and they can sell this ability to the experimenter for a fixed price. In the other treatment, the experimenter starts with the ability to roll the dice and the subject can buy this ability from the experimenter for a fixed price. Regardless of who rolls the pair of dice, the payoffs from the dice rolling game are identical for the subjects.

Design details

Randomization method
Randomization is done at the session level, with the flipping of two coins determining the order of session treatments. The treatments are "sell high" and "sell low", as well as "buy high" and "buy low." All treatments contain their own control groups, as the main analysis will pertain to within-subject variation. Within the subject, their behavior during the deterministic portion can be thought of as the control, with the stochastic treatments being the treatment. Sessions will be ordered based on the flipping of two coins, with each outcome HH, HT, TH, and TT mapping to the treatments as such: HH = "sell high", HT = "sell low", TH = "buy high", and TT = "buy low." Sessions are planned to have 16-20 participants with an expected number of 3 sessions per treatment. If there are less than 50 observations per treatment, additional sessions will be added with a random assignment to treatment, from those treatments with an insufficient observation count.
Randomization unit
Experimental session (though individuals do not interact)
Clusters
200
Observations
200
Arms
50 per each treatment
Primary outcomes
We are interested in the percentage of subjects who buy (or do not sell) the right to personally roll the pair of dice. Percentage of subjects who buy the right to roll dice personally in the stochastic game. Percentage of subjects who buy the right to roll dice personally in the deterministic game. Percentage of subjects who do not sell the right to roll dice personally in the stochastic game. Percentage of subjects who do not sell the right to roll dice personally in the deterministic game.

Cited by

No citation data available for this registration yet.

Keywords