Cheap talk and honesty

How to cite

Fries, Tilman; Parra, Daniel. 2025. "Cheap talk and honesty." AEA RCT Registry. AEARCTR-0015630. https://doi.org/10.1257/rct.15630-1.0

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

Tilman Fries* , Daniel Parra

* Primary investigator

Registry ID
AEARCTR-0015630
DOI
https://doi.org/10.1257/rct.15630-1.0
First registered
Last updated
Status
In development
Trial dates
2025-03-26 – 2025-12-31
Countries
Colombia (Bogotá)
JEL codes
C91, D83
Registry record
http://www.socialscienceregistry.org/trials/15630

Abstract

We propose an experiment on cheap talk. A receiver guesses a secret number, and receives advice from a sender with misaligned incentives. Across three treatment conditions, we vary whether the sender can provide informed advice and how cognitively demanding the receiver finds it to respond to the advice in a sophisticated way. This allows us to identify how receivers reason about the sender's advice and whether honesty-related concerns can explain overcommunication and systematic belief bias.

Intervention

We propose an experiment on cheap talk. A receiver guesses a secret number, and receives advice from a sender with misaligned incentives. Across three treatment conditions, we vary whether the sender can provide informed advice and how cognitively demanding the receiver finds it to respond to the advice in a sophisticated way. This allows us to identify how receivers reason about the sender's advice, and whether honesty-related concerns can explain overcommunication and systematic belief bias, two key empirical patterns identified by the experimental literature (see, e.g.,Cai and Wang, 2006; Blume, Lai, and Lim, 2020; Groseclose, 2021).

Experimental design

Senders and receivers are told that the computer randomly generated a secret number j by drawing once from {1,2,3,4,5,6,7}. The sender then chooses a number X in {1,2,3,4,5,6,7} to send to the receiver. The receiver observes the message and guesses a value for j. The sender is incentivized to induce a high guess, while the receiver has incentives to be accurate. In three treatments, we vary whether the sender can send the true secret number as a message and how coginitively demanding it is for a (subset) of receivers to make a sophisticated guess.

Design details

Randomization method
Randomization into treatments by random assignment of participants to sessions, further randomization (drawing secret numbers) implemented by the experimental software (otree).
Randomization unit
Treatments are randomized on the session level. Secret numbers are drawn on the sender-receiver-round level.
Clusters
We will have either matching groups of size 8 or 10. The exact size of the matching groups will depend on how many participants attend each session, something we have no direct control over.
Observations
500-544
Arms
We aim to run sessions until we have observations of 70 receivers in Baseline and Expert and 66 non-decode receivers in Decode. This implies a sample size of 140 participants (70 senders and 70 receivers) in Babbling and Expert and a sample size of ca. 220 - 264 participants in Decode, depending on the average realized matching group size.
Primary outcomes
Message sent by senders, guess made by receivers.
Secondary outcomes
See the attached pdf.
Sponsors
Name: Pontificia Universidad Javeriana Location: Bogotá Url: https://www.javeriana.edu.co/inicio

Cited by

No citation data available for this registration yet.

Keywords