Cooperation and Spillovers under Group-Level Regulatory Allocation

How to cite

zhang, chenxi; Lao, Yao. 2026. "Cooperation and Spillovers under Group-Level Regulatory Allocation." AEA RCT Registry. AEARCTR-0018023. https://doi.org/10.1257/rct.18023-1.0

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

chenxi zhang* , Yao Lao

* Primary investigator

Registry ID
AEARCTR-0018023
DOI
https://doi.org/10.1257/rct.18023-1.0
First registered
Last updated
Status
Completed
Trial dates
2025-11-21 – 2025-11-28
Countries
China (Shanghai)
JEL codes
JEL Classification: C91, D91, D64
Registry record
http://www.socialscienceregistry.org/trials/18023

Abstract

This study examines how regulatory interventions affect cooperation in repeated social dilemmas, with a particular focus on spillover effects that persist after external enforcement is removed. We employ a laboratory experiment in which 252 participants play a two-person finitely repeated Prisoner's Dilemma (T=90, R=60, P=55, S=10) over 24 rounds divided into two phases.In Phase 1 (Rounds 1–12), participants are randomly and fixedly matched and subject to one of five between-subjects treatments that vary the scope and information structure of a forced-compliance regulation mechanism. Unlike standard punishment mechanisms, the regulation imposes no financial penalties; instead, defection choices are automatically corrected to cooperation, allowing us to isolate the behavioral effects of regulation from deterrence. The five treatments are: No Monitor (baseline), Two Monitor (both partners monitored), One Monitor (one partner monitored with immediate disclosure), Random Monitor (one partner monitored with delayed disclosure), and Half Monitor (half of pairs fully monitored with delayed disclosure). In Phase 2 (Rounds 13–24), all regulation is removed and participants are re-matched with new partners, enabling identification of post-regulatory spillover effects.

Intervention

The experiment implements five between-subjects treatments that vary the scope and information structure of a forced-compliance regulatory mechanism in Phase 1 (Rounds 1–12). In all treatments, if a monitored participant chooses to defect, the system automatically corrects the choice to cooperation; no financial penalty is imposed.
No Monitor (control): No regulation in either phase. Participants make unrestricted choices throughout.
Two Monitor: Both partners in each pair are monitored and subject to forced compliance in Phase 1. Monitoring status is disclosed to both parties before Phase 1 begins.
One Monitor: One participant per pair is randomly selected for monitoring in Phase 1. Both partners are informed of who is monitored before Phase 1 begins (immediate disclosure). At the start of Phase 2, participants are additionally informed of their new partner's Phase 1 monitoring status.
Random Monitor: One participant per pair is randomly monitored in Phase 1, but monitoring status is disclosed to participants only after Phase 1 ends (delayed disclosure). During Phase 1, participants are unaware of whether they or their partner are monitored.
Half Monitor: Half of all pairs are randomly selected for full monitoring (both partners) in Phase 1; the remaining pairs are unmonitored. Monitoring status is disclosed only after Phase 1 ends (delayed disclosure).
In Phase 2 (Rounds 13–24), all regulation is removed across all treatments. Participants are re-matched with new partners and play the Prisoner's Dilemma without any enforcement.

Experimental design

No Monitor (control): No regulation in either phase. Participants make unrestricted choices throughout.
Two Monitor: Both partners in each pair are monitored and subject to forced compliance in Phase 1. Monitoring status is disclosed to both parties before Phase 1 begins.
One Monitor: One participant per pair is randomly selected for monitoring in Phase 1. Both partners are informed of who is monitored before Phase 1 begins (immediate disclosure). At the start of Phase 2, participants are additionally informed of their new partner's Phase 1 monitoring status.
Random Monitor: One participant per pair is randomly monitored in Phase 1, but monitoring status is disclosed to participants only after Phase 1 ends (delayed disclosure). During Phase 1, participants are unaware of whether they or their partner are monitored.

Half Monitor: Half of all pairs are randomly selected for full monitoring (both partners) in Phase 1; the remaining pairs are unmonitored. Monitoring status is disclosed only after Phase 1 ends (delayed disclosure).

Design details

Randomization method
Treatment assignment was implemented at the session level via computer. Participants registered for experimental sessions without advance knowledge of treatment conditions. Each session was pre-assigned to a single treatment arm by the experimenter using a computer-generated randomization schedule prior to recruitment. Within each session, participant-to-partner matching was performed by the oTree using random assignment. For treatments involving individual-level monitoring (One Monitor, Random Monitor, Half Monitor), the software randomly determined which participant within each pair — or which pairs within the session — would be subject to the forced-compliance mechanism, with equal probability and without experimenter discretion.
Randomization unit
The individual participant is the unit of randomization. Treatment assignment (No Monitor, Two Monitor, One Monitor, Random Monitor, Half Monitor) was conducted at the session level: each experimental session was assigned to a single treatment condition, and participants self-selected into sessions without knowledge of treatment assignment. Within-session randomization determined pair formation and, for applicable treatments, which member of each pair was monitored.
Clusters
Not applicable. The experiment does not use a clustered design in the conventional sense. Randomization occurs at the session level; sessions range from 16 to 24 participants. A total of 11 sessions were conducted across the five treatment arms.
Observations
252 participants × 24 rounds = 6,048 participant-round observations in total.
Arms
No Monitor (control): 60 participants Two Monitor: 32 participants (all monitored) One Monitor: 72 participants (36 monitored + 36 unmonitored in Phase 1) Random Monitor: 44 participants (22 monitored + 22 unmonitored in Phase 1) Half Monitor: 44 participants (24 monitored + 20 unmonitored in Phase 1) Total: 252 participants
Primary outcomes
The binary cooperation decision (1 = cooperate, 0 = defect) in each round of Phase 1 (Rounds 1–12), aggregated as the individual-level average cooperation rate. This captures the direct effect of regulatory treatments on compliance-era behavior. The binary cooperation decision in each round of Phase 2 (Rounds 13–24), aggregated as the individual-level average cooperation rate. This is the primary outcome of interest, capturing the spillover effect of Phase 1 regulatory experience on voluntary cooperation after enforcement is removed.

Cited by

No citation data available for this registration yet.

Keywords