Fair Cooperation

In the experiment, participants are randomly assigned to one of four treatment conditions. These conditions are defined by unique combinations of two levels of redistribution rules—low tax and high tax—and two levels of task meritocracy—low and high. To be specific, the conditions are as follows: (1) Low Tax, Low Meritocracy; (2) High Tax, Low Meritocracy; (3) Low Tax, High Meritocracy; and (4) High Tax, High Meritocracy. In each condition, participants engage in a task that follows the same format but varies in tax and meritocracy levels. After completing this task, participants of the same treatment condition are randomly matched in groups of four members. A predetermined percentage of each member's earnings, referred to as the tax, is collected based on the assigned treatment. This tax is pooled and evenly distributed among all group members. Participants' earnings in the tasks are determined by their performance and redistribution outcomes.

In the Low Tax treatments, participants are subjected to a tax rate of 5%, while a tax rate of 50% applies in the High Tax treatments. We construct tasks with distinct levels of meritocracy while ensuring a consistent format. Specifically, we employ matrix reasoning and three-dimensional rotation problems, both of which are typical measures of cognitive ability. Drawing upon data from prior research in the field of intelligence, we select sets of problems with both low and high accuracy rates to represent tasks at different meritocracy levels. Tasks at a low meritocracy level are designed such that, given sufficient effort, participants are anticipated to answer most questions correctly. In contrast, tasks at a high meritocracy level are structured such that individual performance is primarily determined by cognitive ability.

2024-06-17

2024-09-30

Our primary outcomes include both individual and group-level investments in a public goods game. These investments are examined in both a one-shot scenario and across repeated interactions, under different treatment conditions.

Our secondary outcomes include individual perceived fairness regarding the redistribution of their earnings in the meritocracy tasks.

Participants are asked to rate their perceptions of fairness regarding their redistribution outcomes on a scale from 0 to 10.

The experiment is designed to be conducted on the platform Qualtrics, with participants recruited through Prolific. It consists of three parts, where participants engage in incentivized problem-solving tasks and economic decision-making scenarios. In the first part, participants undertake an incentivized task characterized by distinct redistribution rules and meritocracy levels based on their assigned treatment conditions. In the second part, they participate in a one-shot public good game to assess immediate cooperative behavior, without receiving any feedback from this game. In the third part, participants engage in an eight-round repeated public good game with feedback to assess cooperation dynamics across repeated interactions. Finally, participants provide information regarding their perception of fairness regarding their redistribution outcomes in the first part, as well as their personality traits and demographic characteristics.

The experiment is structured so that participants progress sequentially, with clear instructions provided at the beginning of each part, and experimental tokens used as rewards (at a conversion rate of 50:1 to British pounds).

In the first part, participants solve ten problems, comprising five on matrix reasoning and five on three-dimensional rotation. Each correct answer earns a participant ten experimental tokens. These problems, adapted from the intelligence literature (Condon and Revelle, 2014), maintain a consistent format with two levels of meritocracy: low and high, corresponding to the Low Meritocracy and High Meritocracy treatments, respectively. After completing this task, participants of the same treatment condition are randomly matched in groups of four members. In each group, a tax, either 5% or 50% of each member's earnings (for the Low Tax and High Tax groups respectively), is collected, pooled, and distributed evenly among all members of the group. Within a group, members are assigned roles (labeled A, B, C, and D), which remain unchanged throughout the experiment, indicating individual performance along with outcomes. Upon successful matching, each participant is presented with a table detailing their performance, redistribution, and earnings, as well as the performance of their group members.

The group that is formed in the first part is kept for both the second and third part. That is, participants engage in the games in the second and the third part within the same group as in the first part and this is clearly known by the participants.

In the second part, participants engage in a one-shot public good game, starting with an endowment of 100 experimental tokens. In this game, they choose an amount between 0 and 100 to invest in a group project, which yields a marginal per-capita return of 0.4 (i.e., each experimental token invested rewards 0.4 tokens to everyone in the group). The outcome of this part is only revealed at the end of the experiment.

In the third section, participants engage in the same game as in the second part, with this game repeating across eight rounds. The outcome of each round and its previous rounds are presented to the participants in a table format, including investments of each group member, the project returns, and the details of earnings.

After the third part, participants are asked to rate their perceived level of fairness, on a scale of 0 to 10, regarding the redistribution in the first part. Finally, participants complete a survey on their personality traits (openness, conscientiousness, extraversion, agreeableness, and neuroticism) and demographic characteristics (age, gender, and education levels). Comprehension checks are included in the first two parts, as well as attention checks in the third part and the personality traits survey.

The randomization is implemented by Qualtrics.

Individual.

No

160 groups.

640 individuals (4 per group).

40 groups (160 individuals) in each treatment condition.

With the planned sample size and a significance level of 0.05 and power of 0.80, our study is designed to detect effect sizes as small as 0.31 for participants' investments in the group project when comparing any two treatment arms at the individual level, and as small as 0.63 at the group level.