Coordination with third-party negative externalities and social distance

How to cite

Ouvrard, Benjamin; Bonroy, Olivier; Garapin, Alexis. 2025. "Coordination with third-party negative externalities and social distance." AEA RCT Registry. AEARCTR-0015975. https://doi.org/10.1257/rct.15975-1.0

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

Benjamin Ouvrard* , Olivier Bonroy , Alexis Garapin

* Primary investigator

Registry ID
AEARCTR-0015975
DOI
https://doi.org/10.1257/rct.15975-1.0
First registered
Last updated
Status
In development
Trial dates
2025-05-12 – 2025-12-31
Countries
Private
JEL codes
C72; C92
Registry record
http://www.socialscienceregistry.org/trials/15975

Abstract

Building on the work of Bland and Nikiforakis (2015) and Cason et al. (2022), we investigate whether reducing the social distance between “active” participants (i.e., those making decisions) and “inactive” participants (i.e., those whose outcomes depend on the active participants' choices) in a coordination game with third-party negative externalities can promote coordination towards a “pro-social” equilibrium (i.e., such that the “inactive” participants do not incur any losses). Before playing the coordination game, subjects play a “paintings game” either individually, or in groups (Chen and Chen, 2011) to create a shared experience.
The experimental conditions vary based on the level of information available to the “active” subjects in the main game: (i) some receive no information about the other participants (baseline condition); (ii) others know whether they played with the “inactive” subject during the paintings game (incomplete information treatment), and (iii) some know whether they played with both the “inactive” and the other “active” subjects during the paintings game (complete information treatment). Within these treatments, we will also control for the composition of the groups, considering past interactions between participants during the paintings game.
We also account for allocation preferences and measure subjects' beliefs regarding the chosen actions by other subjects and their perceived social distance (Aron et al., 1991).
Our main hypotheses are that: (i) subjects' choice of the prosocial action depends on the level of information provided about a shared experience with the inactive player; (ii) subjects' choice of the prosocial action depends on the level of information provided about a shared experience with the active player; (iii) subjects' choice of the prosocial action depends on the group’s composition and on the level of information provided about a shared experience with the players.

Intervention

We consider different versions of the coordination game that differ regarding (i) whether, or not, subjects share a common experience in the paintings games; and (ii) the level of information communicated to “active” subjects in a between-subject design (i.e., they participate in one treatment only). The different conditions are:
i) Baseline: subjects play the paintings game individually and do not receive any information on the other players;
ii) Treatment 1: subjects play the paintings game in groups (green, blue, orange) and do not receive any information on the other players;
iii) Treatment 2: subjects play the paintings game in groups (green, blue, orange) and receive information on whether, or not, they played with the “inactive” subject during the paintings game;
iv) Treatment 3: subjects play the paintings game in groups (green, blue, orange) and receive information on whether, or not, they played with the “inactive” and the other “active” subjects during the paintings game.

The comparison between the Baseline and Treatment 1 identifies the impact of sharing a common experience in the painting games on participants’ decisions in the coordination game. Comparing Treatment 1 with Treatments 2 and 3 isolates the effect of making salient a shared experience (or its absence) in the painting games on those same decisions.

Experimental design

The structure of our experiment is as follows:
i) Subjects play a paintings game;
ii) Then, they play a coordination game with third-party negative externalities;
iii) Next, they play an allocation game;
iv) Finally, we measure their beliefs, their perceived social distance and collect some socio-demographic characteristics

1) Paintings game:

Subjects play the game either individually (Baseline) or in groups (Treatments 1 to 3).

If they play the game individually, they first have five minutes to individually analyze five pairs of paintings by Klee and Kandinsky, identified only as “Artist 1” and “Artist 2”. Then, they are given seven minutes to analyze a sixth pair of paintings and determine the artist behind each.

If they play the game in groups, subjects are randomly assigned to one of three groups: blue, orange, or green. First, they have five minutes to individually analyze five pairs of paintings by Klee and Kandinsky, identified only as “Artist 1” and “Artist 2”. Then, they are given seven minutes to analyze a sixth pair of paintings and determine the artist behind each. For the identification of the sixth pair, group members can communicate via an open chat, but not with members of other groups (e.g., green subjects can only communicate with green subjects, etc.). Each subject submits their choice individually, and correct answers are rewarded.

2) Coordination game:

A. Main structure of the game:

Subjects are randomly allocated in fixed groups of three (partner design): two C subjects and one Z subject. Roles remain fixed for the rest of the experiment and subjects do not know the exact identity of the other group members.

Subjects play a similar version of the coordination game proposed in Bland and Nikiforakis (2015) and Cason et al. (2022), where only C players make a decision. Precisely, at each round of the game, they can choose between action M and action J.

If both C subjects choose action J, then each C subject earns 5 Experimental Currency Units (ECU, with 1 ECU = €0.5) while subject Z earns 4 ECU. The (J;J) equilibrium is the pro-social one.

If both C subjects choose action M, then each C subject earns 7 ECU, while subject Z earns z ECU.

If C subjects do not coordinate on the same action, everyone earns 0 ECU.

Each round of the game corresponds to a specific value of z: -10; -8; -6; -4; -2; 0; 2; 4. Therefore, subjects play a total of eight rounds and are informed about the number of rounds, with the value of z that appears in random order between groups. To avoid any form of learning, no feedback will be provided at the end of each round. Only C subjects make decisions in this game. However, subjects are informed that, at the end of experiment, Z participants will be informed of all decisions made by the C participants in their group for all rounds.

We will control for group composition in the treatments and program the software to obtain:
- Groups such that all subjects already played together during the paintings game;
- Groups such that one of the C subjects and the Z subject played together during the paintings game;
- Groups such that only the C subjects played together during the paintings game.

B. Treatments:

We consider four types of experimental conditions:

- In the baseline and treatment 1 experimental conditions, C subjects will not receive any information on the other players when making their decisions.
- In treatment 2 (incomplete information condition), at each round C subjects will be informed on whether, or not, they played with the “inactive” subject during the paintings game.
- In treatment 3 (complete information condition), at each round C subjects will be informed on whether, or not, they played with the “inactive” and the other “active” subjects during the paintings game.

C. Payment:

Subjects will be informed that, at the end of the experiment, one of the rounds will be randomly selected and their payoff for that round will be added, or deduced, to a fixed dotation of 30 ECU (i.e., €15).

3) Allocation game:

After the main game, subjects, still in the same group, play an allocation game. They will be presented the same eight situations as in the previous game (in a random order), except that they will have to determine their preferred allocation of payoffs. Precisely, subjects will simply have to indicate their preferred allocation of payoffs resulting from the choice of action M or action C.

At the end of the experiment, one of the choices of the C subjects will be randomly selected in each group and the allocation decided by this subject will be implemented at the group level, and the payoffs added to the final payoff of each subject.

4) Additional variables:

Finally, we will ask subjects:
- Their beliefs (C subjects only) regarding the most frequently chosen action by other C subjects depending on their group in the paintings game (blue, orange, green). Correct beliefs will be remunerated;
- Their beliefs (Z subjects only) regarding whether other subjects in the session identified correctly the artists of the sixth pair in the paintings game. Correct beliefs will be remunerated. We ask this question to Z subjects since they do not have an active role in the experiment and we want to avoid widening the payoff gap between them and the C subjects;
- Their perceived social distance with respect to other subjects depending on their group in the paintings game (blue, orange, green), in the treatments only;
- Socio-demographic characteristics (age, gender, education).

Design details

Randomization method
Randomization made by computer (Hroot)
Randomization unit
Experimental sessions
Clusters
Non relevant.
Observations
We will recruit a minimum of 189 subjects (i.e., a session with 27 subjects will be conducted for the baseline, and a minimum of two sessions with 27 subjects will be conducted for treatments 1 to 3). Per experimental condition, this will result in a minimum of 9 independent observations.
Arms
A minimum of 54 subjects will be recruited for treatments 1 to 3, and 27 for the baseline.
Primary outcomes
Our primary analysis will focus on two key outcomes: (i) the actions chosen by active subjects, and (ii) whether coordination occurs, including the nature of the outcome if coordination is achieved. We will estimate a probit model to analyze subjects' choice of actions.
Secondary outcomes
After the main game, we will measure subjects’ allocation preferences following the methodology proposed by Bland and Nikiforakis (2015) and Cason et al. (2022) (see below). We will also assess subjects' beliefs regarding the most frequently chosen action by other players, based on their group during the paintings game, as well as their perceived social distance (between each type of subject within the groups formed during the paintings game). Additionally, we will gather basic socio-demographic information, including age, gender, and education level.

Cited by

No citation data available for this registration yet.

Keywords