Bargaining and Conflict with Private Information

How to cite

Cox, Caleb; Baranski, Andrzej. 2025. "Bargaining and Conflict with Private Information." AEA RCT Registry. AEARCTR-0015715. https://doi.org/10.1257/rct.15715-1.0

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

Caleb Cox* , Andrzej Baranski

* Primary investigator

Registry ID
AEARCTR-0015715
DOI
https://doi.org/10.1257/rct.15715-1.0
First registered
Last updated
Status
In development
Trial dates
2025-04-16 – 2026-08-31
Countries
Private
Registry record
http://www.socialscienceregistry.org/trials/15715

Abstract

This pilot study examines the bargaining decisions of human participants to avoid conflict in a game with private information. Each of two players will privately observe a "secret number," which are iid random draws from a uniform distribution between 0 and 100. The players then bargain over how to divide 120 experimental currency units or "tokens." If no agreement is reached, conflict occurs. In conflict, the player with the higher secret number gets 100 tokens, while the other player gets 0 tokens.

Between subjects, we compare two bargaining protocols. In the SPLITS condition, the two player simultaneous indicate whether they would agree to each of 13 possible divisions of 120 tokens, from 0 for self and 120 for other to the opposite extreme by increments of 10 tokens. If any split is mutually agreed, payoffs are determined according to the agreement. If multiple splits are mutually agreed, one is randomly selected by the computer. If no split is mutually agreed, conflict occurs.

In the ULTIMATUM condition, one player is randomly assigned at the proposer, and the other as responder. The proposer chooses an offer from the same menu used in the SPLITS condition. The responder then observes the offer and chooses whether to accept or reject. If the proposal is accepted, then payoffs are determined according to the agreement. Otherwise, conflict occurs.

Within subjects, we also compare treatments with and without free-form chat before bargaining. In Part 1, participants first play 8 rounds without chat, with random matching into pairs in each round. Next, in Part 2, participants play 8 additional rounds with chat. Chat occurs after the players observe their own secret numbers, but before bargaining.

We also include an additional 8 rounds played in teams in Part 3. Participants are randomly assigned into teams of 2, which remain fixed. In each round, teams are randomly matched. Teammates can chat privately with one another and they can chat with the other team, as in Part 2. Teammates always receive identical payoffs.

Experimental design

This pilot study examines the bargaining decisions of human participants to avoid conflict in a game with private information. Each of two players will privately observe a "secret number," which are iid random draws from a uniform distribution between 0 and 100. The players then bargain over how to divide 120 experimental currency units or "tokens." If no agreement is reached, conflict occurs. In conflict, the player with the higher secret number gets 100 tokens, while the other player gets 0 tokens.

Between subjects, we compare two bargaining protocols. In the SPLITS condition, the two player simultaneous indicate whether they would agree to each of 13 possible divisions of 120 tokens, from 0 for self and 120 for other to the opposite extreme by increments of 10 tokens. If any split is mutually agreed, payoffs are determined according to the agreement. If multiple splits are mutually agreed, one is randomly selected by the computer. If no split is mutually agreed, conflict occurs.

In the ULTIMATUM condition, one player is randomly assigned at the proposer, and the other as responder. The proposer chooses an offer from the same menu used in the SPLITS condition. The responder then observes the offer and chooses whether to accept or reject. If the proposal is accepted, then payoffs are determined according to the agreement. Otherwise, conflict occurs.

Within subjects, we also compare conditions with and without free-form chat before bargaining. In SPLITS-CHAT and ULTIMATUM-CHAT conditions, players can chat with each other freely for up to 3 minutes after observing their secret numbers, but before bargaining.

Also within subject, we examine conditions played in teams. In the SPLITS-TEAMS and ULTIMATUM-TEAMS conditions, players are randomly matched into fixed teams of 2. Teammates can chat with each other for up to 1 minute before entering a phase of chat with the other team of up to 3 minutes. During this phase, teammates can also chat privately in a separate chat window. In ULTIMATUM-TEAMS, the team in the responder role has an additional opportunity to chat for up to 1 minute after seeing the proposer's offer.

Each session begins with 8 rounds of the SPLITS or ULTIMATUM conditions in Part 1, matched randomly into pairs of 2 within a matching group of 8 in each round. Next, there are 8 additional rounds of the corresponding CHAT condition in Part 2, again matched randomly into pairs of 2 within the same matching groups of 8 in each round. In Part 3, there are 8 additional rounds of the corresponding TEAMS condition, with teams fixed but randomly matched into groups of 2 teams in each round. In each of the three blocks of 8, one round is randomly selected by the computer for payment.

Finally, all participants will complete several secondary tasks, including social value orientation, cognitive reflection test, multiple price list risk preference elicitation, and a demographic questionnaire.

Hypotheses:
Due to adverse selection, we hypothesize that no split will be offered or accepted that does not give one of the two players at least 100 tokens.

Our null hypothesis regarding chat is that the incentive to report and inflated secret number will lead to no credible communication of private information.

We have no a priori hypothesis about the difference between individual and team behavior. Instead, we use team chats primarily to gain insight into the decision process through chat content analysis.

Design details

Randomization method
Randomization done in office by a computer.
Randomization unit
Groups of participants (sessions) are randomized to experimental conditions.
Clusters
This is a pilot study. For the pilot, we plan to run 8 matching groups (4 sessions).
Observations
64 participants
Arms
32 participants (4 matching groups, 2 sessions) in each of the SPLITS and ULTIMATUM treatments.
Primary outcomes
Decisions to accept or reject potential splits, Proposed splits
Secondary outcomes
Chat contents (particularly truthful or untruthful communication of private information), Social Value Orientation, Risk Preferences, Cognitive Reflection Test scores (3 item)

Cited by

No citation data available for this registration yet.

Keywords