Belief Disagreement and Instability under Stochastic Recall

How to cite

Tang, Li; Kuang, Pei; Weber, Michael. 2026. "Belief Disagreement and Instability under Stochastic Recall." AEA RCT Registry. AEARCTR-0017671. https://doi.org/10.1257/rct.17671-1.0

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

Li Tang* , Pei Kuang , Michael Weber

* Primary investigator

Registry ID
AEARCTR-0017671
DOI
https://doi.org/10.1257/rct.17671-1.0
First registered
Last updated
Status
In development
Trial dates
2026-01-19 – 2027-01-18
Countries
Private
JEL codes
E31, D84, C83
Registry record
http://www.socialscienceregistry.org/trials/17671

Abstract

We study how imperfect human memory shapes cross-sectional belief disagreement and within-individual belief instability. In Part 1 of the experiment, decisions makers need to form beliefs about the unobserved value of a hypothetical company. They receive a sequence of 20 binary signals. We examine how disagreement is affected by the degree of information ambivalence and recalling efforts. In Part 2, we follow up with respondents one day after the Part 1 to re-elicit their beliefs and recalls.

Intervention

In Part 1 of the experiment, participants are told about a hypothetical company that can be of good or bad quality and receive a sequence of 20 signals or news about it sequentially. Each news item is either positive or negative, with the mix of good and bad news varying across conditions. In some conditions, participants are given small incentives to improve recall accuracy. Some participants receive “stop-and-think” instructions to increase recall effort and the probability of correct recall. Participants then state their belief about the company’s type based on what they remember. In Part 2, we follow up with these respondents one day after Part 1 and re-elicit their beliefs about the company’s type based on what they remember from Part 1.

Experimental design

In Part 1 of the experiment, participants are presented with a sequence of 20 news items about a hypothetical company that can be of good or bad quality. Each news item is either positive or negative, and the mix of good and bad news varies across conditions. Participants form a belief about the company’s quality after seeing the sequence. They are also asked to recall the number of good and bad news items they received. Some participants also receive “stop-and-think” instructions to increase recall effort and the probability of correct recall. One day after Part 1, we re-invite these respondents to participate in Part 2, where we re-elicit their beliefs and recall.

Design details

Randomization method
Done within qualtrics using javascript.
Randomization unit
Individual-level.
Clusters
We plan to collect data from around 2000 participants in Part 1 and follow-up with them one day later in Part 2.
Observations
2000 participants in Part 1.
Arms
500 in each groups.
Primary outcomes
Our primary variable of interest is participants’ belief about whether the company is of good or bad quality, as well as the associated probability, after observing the news sequence. The main outcome is the level of cross-sectional belief disagreement and within-individual belief instability for participants exposed to the same information in both Parts 1 and 2. We also collect measures of participants’ recall of the number of positive and negative news items they received in both Parts 1 and 2.

Cited by

No citation data available for this registration yet.

Keywords