Collective Risk Social Dilemma and Equilibrium Reasoning

Last registered on June 23, 2026

Pre-Trial

Trial Information

General Information

Title
Collective Risk Social Dilemma and Equilibrium Reasoning
RCT ID
AEARCTR-0018962
Initial registration date
June 22, 2026

Initial registration date is when the trial was registered.

It corresponds to when the registration was submitted to the Registry to be reviewed for publication.

First published
June 23, 2026, 8:50 AM EDT

First published corresponds to when the trial was first made public on the Registry after being reviewed.

Locations

Primary Investigator

Affiliation
university of southampton

Other Primary Investigator(s)

Additional Trial Information

Status
In development
Start date
2026-06-22
End date
2026-07-18
Secondary IDs
Prior work
This trial does not extend or rely on any prior RCTs.
Abstract
This study uses a lab experiment to examine whether failures of equilibrium reasoning affect support for carbon taxes in a Collective Risk Social Dilemma (CRSD) setting. Pairs of participants play a consumption-based CRSD game with a tipping point. Consumption produces a Byproduct, and exceeding the byproduct production threshold triggers a 70% probability of losing all tokens. The game is played in three variants: a baseline without tax, a game with tax that affects players equally, and a game with tax that creates inequality through redistribution.
The tax alters the game's equilibrium, making both the private and social equilibria superior. If a failure of contingent thinking (Niederle and Vespa, 2023) is present, participants may still vote against it if they fail to anticipate how opponents will adjust their behaviour under the new rules.
After experiencing both a taxed and untaxed game, participants vote for their preferred version to play in a final round, with their stated beliefs about opponent behaviour elicited after the vote. At this stage, some players undergo an intervention to correct the FCT.
The study aims to shed light on why carbon taxes may lack public support even when they improve outcomes, and whether inequality created by tax design compounds this effect.
External Link(s)

Registration Citation

Citation
Bell, Alice. 2026. "Collective Risk Social Dilemma and Equilibrium Reasoning ." AEA RCT Registry. June 23. https://doi.org/10.1257/rct.18962-1.0
Experimental Details

Interventions

Intervention(s)
Intervention (Hidden)
Intervention Start Date
2026-06-22
Intervention End Date
2026-07-18

Primary Outcomes

Primary Outcomes (end points)
Presence of failure of contingent thinking (FCT), effect of FCT on vote choice, inequality effect on FCT, effect of the intervention on FCT, redistribution between players
Primary Outcomes (explanation)

Secondary Outcomes

Secondary Outcomes (end points)
Payoff of the collective risk social dilemma (CRSD) game; success in avoiding the tipping point; assessment of opponents' behaviour; political, ecological, and risk preferences' effects on CRSD.
Secondary Outcomes (explanation)

Experimental Design

Experimental Design
This experiment uses a variant of the Collective Risk Social Dilemma (CRSD) game, adapted for a consumption context following Bachler et al. (2024). Players make repeated contribution decisions over multiple rounds, facing the risk of a catastrophic loss if a threshold is exceeded. The experiment uses three variations of the game, differing in whether a tax is applied and whether that tax creates inequality between players. After experiencing multiple game variants, participants vote on which game they would like to play again. A randomly selected subset of participants receive an intervention during the voting stage designed to support reasoning about how their opponent is likely to behave. Participants then complete a questionnaire eliciting attitudes relevant to the experiment.
Experimental Design Details
•The CRSD game involves multiple rounds during which players contribute to a public good; if the threshold is not met, all players incur negative consequences. CRSD has several features which make it a good fit for modelling climate change: There are repeated decisions before the outcome is evident, the remaining private good is at risk if the threshold is not met, and the value of the public good is unknown
Following Bachler et al. (2024), the CRSD game was adapted for a consumption context in this experiment. In each round, players are given four tokens. They must choose in each round how many tokens to invest in a good. When a token is contributed to the good at the start of a round, four times as many tokens are returned to the player at the end of the round. Tokens not contributed to the good are kept.
All tokens held at the end of the round are equivalent to one unit of byproduct produced in the round. If across all rounds a threshold of 140 units of byproduct is passed, a catastrophic event, in which all players lose all their tokens, occurs with 70% probability.
The game is played in Groups of 2, which remain consistent throughout the experiment. These groups are randomly assigned and anonymous. There is no communication between players. There are 10 rounds per section of the game, regardless of whether the threshold is passed at any point. If the threshold is crossed, whether the catastrophic event occurs will be randomly generated.
This experiment uses three variations of the CRSD game. A baseline game, A game with a carbon tax added, equivalent to 50% of the token returns from investing in the good. A game with a carbon tax as before, which also creates inequality between the two players by redistributing some of the earnings from investing in the good. All pairs play the untaxed game and either the equal- or unequal-taxed game.
Success in the game occurs when the pair completes all 10 rounds without passing the threshold; a pair is not treated as successful if they pass the threshold, but the catastrophe is not realised.
The experiment will consist of 5 stages. Section 1: all players play an initial CRSD game with or without a welfare-improving tax. Section 2: players then play the CRSD game they did not play in Section 1, so they have experienced both types of game.
In stage 3, players will be asked to vote on the game they have played so far that they would like to play again. Before the vote results are revealed, players will be asked how they expect their opponent to behave in each game. The winning game will be the one with the majority of votes, with ties broken at random by a computer.
This voting stage is important for evaluating the presence of failure in equilibrium reasoning because it allows us to see whether players still demand the game with lower payoffs, even after experiencing both games. It also shows whether they can correctly assess their opponent's strategy.
In a selection of randomly selected groups, an intervention will be implemented during the voting phase to address failures in equilibrium reasoning. A feature of FCTs is that when the problem is presented to players in a way which helps them focus on all contingencies, they can optimise the problem. The intervention aims to assess whether this feature is present.
Finally, in Section 4, they play the winning game they voted for,, and Section 5 then presents a post-experimental questionnaire to elicit risk aversion, inequality aversion, environmental attitudes, and opinions on taxes.
It will use the Revised New Environmental Paradigm (NEP) scale (Dunlap et al., 2000) to elicit views on Environmentalism. To elicit risk aversion, we will use the method described in Eckel and Grossman (2008). For inequality aversion, we will use a variation of the dictator game. where players can choose how much to split with their opponent. Finally, we will use the Resistance to Change Beliefs Scale (White et al., 2020) to establish conservatism.
The 10-round game, belief elicitation task, and risk and inequality aversion tasks are incentivised, with tokens earned in these games converted into a bonus payment.
Players will be excluded if they fail any one of an attention check question, a question using a visual allusion designed to catch the use of AI, or if they cannot answer a set of understanding questions after the instructions after 4 attempts.

In the game without tax, there are two equilibria. In the cooperative one, players coordinate so their combined contributions stay just below the threshold, and each contributes at least 4 tokens. For example, Player 1 puts in 10 tokens, and Player 2 puts in 8, totalling 18 and yielding payoffs of (70, 64), maximising joint welfare while staying below the threshold.
In the non-cooperative equilibrium, both players put all 40 tokens into the good, crossing the threshold and leaving each with an expected 48 tokens, accounting for the probability of the catastrophic event. A key problem is that one player can cross the threshold alone; for example, Player 1 putting in all 40 tokens is enough to do so without any contribution from Player 2. This means that cooperation is fragile

In the game under equal tax, due to the tax-effective threshold for the number of tokens that can be put into the good without exceeding it, the threshold rises to 58 combined tokens, meaning no single player can cross it with just their own contribution. Even if Player 1 puts in all 40 tokens, gaining a payoff of 80, Player 2 can respond with 18, with a payoff of 58, to stay just below 58 total tokens put into the good. This removes the non-cooperative equilibrium entirely, meaning coordination is the only equilibrium.

This is welfare-improving if the groups were not coordinating under no tax, as both now earn more than 48 tokens, and equal if the groups cooperate under the game with tax.

In the unequal tax, total joint welfare is unchanged relative to the equal-tax game, but it is redistributed. Both players still fare better than in the no-tax non-cooperative outcome of 48 tokens each, but the more heavily taxed player may end up worse than under no-tax cooperation, potentially giving them reason to prefer the untaxed game.
Randomization Method
Treatment assignment to game variant (equal or unequal tax) is determined by systematic sequential assignment, cycling through conditions in order as participants arrive. FCT intervention assignment is determined independently for each individual by a computer-generated random draw with 50% probability.
Randomization Unit
There are two levels of randomisation. Game variant is assigned at the individual level on arrival, with participants subsequently paired within treatment by instruction completion time, making the pair the effective unit of analysis. The FCT intervention is assigned at the individual level independently, with each participant having an equal probability of assignment regardless of their partner's assignment.
Was the treatment clustered?
Yes

Experiment Characteristics

Sample size: planned number of clusters
104 pairs
Sample size: planned number of observations
208 individuals
Sample size (or number of clusters) by treatment arms
26 pairs, untaxed game followed by equal tax
26 pairs, untaxed game follwed by unequal tax
26 pairs, equal tax game follower by untaxed game
26 pairs unequal tax game followed by untaxed game

104 individuals with intervention to correct failure of contingent thinking,
104 without intervention
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
IRB

Institutional Review Boards (IRBs)

IRB Name
University of Southampton Faculty of Social Sciences Ethics Committee
IRB Approval Date
2025-12-16
IRB Approval Number
108896
Analysis Plan

There is information in this trial unavailable to the public. Use the button below to request access.

Request Information

Post-Trial

Post Trial Information

Study Withdrawal

There is information in this trial unavailable to the public. Use the button below to request access.

Request Information

Intervention

Is the intervention completed?
No
Data Collection Complete
Data Publication

Data Publication

Is public data available?
No

Program Files

Program Files
Reports, Papers & Other Materials

Relevant Paper(s)

Reports & Other Materials