Understanding Overconfidence

How to cite

Dutta, Amrita; Serra-Garcia, Marta; Gneezy, Uri. 2026. "Understanding Overconfidence." AEA RCT Registry. AEARCTR-0018159. https://doi.org/10.1257/rct.18159-1.1

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

Amrita Dutta* , Marta Serra-Garcia , Uri Gneezy

* Primary investigator

Registry ID
AEARCTR-0018159
DOI
https://doi.org/10.1257/rct.18159-1.1
First registered
Last updated
Status
In development
Trial dates
2026-03-03 – 2026-12-31
Countries
Private
JEL codes
D91, C91
Registry record
http://www.socialscienceregistry.org/trials/18159

Abstract

We aim to understand what drives individual overplacement when participants are asked questions about their ability or intelligence. Our main hypothesis is that overplacement is driven by a desire to not perceive oneself in the very bottom of the distribution (lowest quartile), rather than a desire to place oneself at the very top of the distribution (highest quartile). To test this hypothesis we will compare participants’ beliefs when asked to place themselves in two cases: below or above median (Median condition), and in one of four quartiles of the distribution (Q1-Q4, Quartiles condition). We also hypothesize that there will be a difference by gender, according to which men will display a stronger desire to avoid the bottom quartile than women. The study will consider two domains, an objective, verifiable task (IQ test) and a subjective, unverifiable task (IQ prediction without any test).

Intervention

This multi-wave study investigates overplacement in self-assessment. Wave 1 (N=600) utilized a domain-specific design to contrast an objective, verifiable task (IQ test) with a subjective, unverifiable task (driving ability). In Wave 2 (N=400), participants are randomized into one of two task types (Subjective IQ Prediction vs. Easy IQ Test) and one of two response scales (Quartile vs. Median).

Experimental design

The overall study employs a multi-wave between-subjects design to examine the interaction between task subjectivity, difficulty, and response scale granularity.

Wave 1: 600 participants were randomized into experimental cells crossing two domains, IQ (cognitive test) and driving ability, with two scale types (quartile vs. median), along with the presence or absence of monetary incentives for accuracy.

Wave 2: 400 additional participants are randomized into a 2x2 design (100 participants per cell) within the IQ domain. This wave crosses two new task types, Subjective Prediction (no test) and Easy IQ Test (simplified assessment), with the same two response scales (quartile vs. median).

Design details

Randomization method
Randomization is done using the built-in randomizer tool in the Qualtrics survey flow.
Randomization unit
Individual
Clusters
1000 individuals (not clustered).
Observations
1000 individuals.
Arms
The study comprises a total of 1,000 participants across two waves of data collection. Wave 1 (N=600): Participants were randomized into six blocks crossing domain and scale type: Unincentivized IQ (Quartile): 100 individuals Unincentivized IQ (Median): 100 individuals Incentivized IQ (Quartile): 100 individuals Incentivized IQ (Median): 100 individuals Driving (Quartile): 100 individuals Driving (Median): 100 individuals Wave 2 (N=400): Participants are randomized into four blocks: Subjective IQ (Quartile Scale): 100 individuals. Subjective IQ (Median Scale): 100 individuals. Easy IQ Test (Quartile Scale): 100 individuals. Easy IQ Test (Median Scale): 100 individuals.
Primary outcomes
Categorical self-assessed placement of IQ performance and driving skills relative to a reference group of 99 other participants.

Cited by

No citation data available for this registration yet.

Keywords