How social identity affect network formation and public good provision in long run?

Clarification on Random Assignment and Control Condition:

In this trial, participants were randomly assigned to one of nine treatment groups, including five new experimental conditions and four re-implemented (previously studied) conditions. Each treatment consisted of multiple sessions with 12 participants per session, and each subject was randomly assigned to a session.

New Treatments (8 sessions × 12 subjects = 96 subjects per treatment):
1) I30-50Rounds;
2) O30-50Rounds;
3) I80-50Rounds;
4) O80-50Rounds;
5) Medium80.

Re-implemented Treatments (4 sessions × 12 subjects = 48 subjects per treatment):
1) I30;
2) O30;
3) I80;
4) O80.

Treatment Naming Conventions:
1) “I” indicates participants started each game with two in-group members;
2) “O” indicates participants started each game with two out-group members;
3) “Medium” indicates participants started with one in-group and one out-group member;
4) Treatments with “50Rounds” in the name had 50 rounds of gameplay; others had 25 rounds.

N/A

2025-05-14

2025-11-30

Participants' contribution level in playing the public good, and who they choose to play the public good game (the network).

The key outcome variables (endpoints) of interest in this experiment include:

1. Network Formation Metrics:
1.1 Average number of neighbors per subject per round.
1.2 Number of links initiated per subject per round.
1.3 Number of links deleted per subject per round​

2. Linking Decisions:
2.1 Probability of initiating links with in-group vs. out-group members.
2.2 Probability of deleting links with in-group vs. out-group members.

3. Contribution Behavior:
3.1 Choice of plans (Plan 1 to Plan 4) by participants in each round.
3.2 Modal plans of neighbours

4. Network Segrereation:
4.1 Composition of in-group members in network

We have completed Phase 1 of the project and found that the results are not statistically significant (see, Chen, Ying, Tom Lane and Stuart McDonald. 2025. "How social identity affect network formation and public good provision in long run?." AEA RCT Registry. April 04. https://doi.org/10.1257/rct.15651-1.0). In other words, there is no meaningful difference between using the mean and the median as the basis for contributions. Therefore, we are proceeding to Phase 2 of the implementation.

We would like to note that the implementation plan for Phase 2 differs slightly from the original design, due to changes in experimental resources.

Originally, the project relied on two laboratories (hereafter referred to as Lab A and Lab B). However, Lab A is currently inactive, and as a result, we plan to hire one or two new laboratories (Lab C and Lab D) to continue the experiments.

Given that this phase involves a comparison between 25-round and 50-round treatments, and the previously collected 25-round data from Lab A is no longer usable, we will conduct both 25-round and 50-round treatments in both Lab C. Specifically, in addition to implementing the following four new 50-round treatments in Lab C (I30-50Rounds, O30-50Rounds, I80-50Rounds and O80-50Rounds), we will also re-implement the following four original 25-round treatments (I30, O30, I80 and O80). This approach ensures internal consistency and comparability of data, minimizing potential bias due to differences across laboratory environments.

If we finally plan to hire Lab D, then we will split the sessions planned for Lab C across Lab C and Lab D.

All other design elements remain consistent with Chen, Ying, Tom Lane, and Stuart McDonald. 2025. "How social identity affect network formation and public good provision in long run?" AEA RCT Registry. April 04. https://doi.org/10.1257/rct.15651-1.0.

N/A

Public lottery: Subjects in the pool are randomly invited to participate in the experiment. In the invitation email they are allowed to choose to participate one of the sessions, while they do not know which session belongs to which treatment.

Individual.

Yes

672 undergraduate/postgraduate students.

There are 5 new treatments: 1. I30-50Rounds: 8 sessions * 12 subjects/session = 96 subjects 2. O30-50Rounds: 8 sessions * 12 subjects/session = 96 subjects 3. I80-50Rounds: 8 sessions * 12 subjects/session = 96 subjects 4. O80-50Rounds: 8 sessions * 12 subjects/session = 96 subjects 5. Medium80: 8 sessions * 12 subjects/session = 96 subjects And we re-implement 4 old treatments: 1. I30: 4 sessions * 12 subjects/session = 48 subjects 2. O30: 4 sessions * 12 subjects/session = 48 subjects 3. I80: 4 sessions * 12 subjects/session = 48 subjects 4. O80: 4 sessions * 12 subjects/session = 48 subjects

672 undergraduate/postgraduate students.