AEA RC T Registr y

The American Economic Association's registry for randomized controlled trials

The experiment consist of four treatments: Baseline, Performance, Color and Feedback. In Baseline, participants repeatedly solve real effort tasks. In Performance, some participants are excluded from a team based on their performance. In Color, the exclusion is due to a random assignment of color. In the Feedback treatment, participants learn the same information as in Performance with respect to their relative ranking but there is no exclusion possibility.

2019-07-16

2019-10-31

Within a treatment: performance difference within different types of participants (excluded vs. included in Performance and Color, and message receivers vs others in Feedback) in a given round, and performance difference of the same type of player across different
rounds.
Across treatments: performance in a given round; performance difference between different types of participants in the final round.

Performance: the number of correct answers given by a player in a given round of a real effort task.

Within a treatment: difference in the number of attempts, error rate, and beliefs about their own performance within different types of participants across different rounds. Across treatments: In a given round, the number of attempts, error rate, and beliefs about own performance. In the final round, the number of attempts, error rate, and beliefs about own performance across different types of participants. The questionnaire at the end of the experiment includes the following: (1) Whether participants think the outcome of Task 3 is fair, (2) An open question on what they think the experiment is about, (3) Their estimate of how performance of the excluded or message receiving players changes after they learn that they have been excluded or received a message, (4) Gender, major, ethnicity, and personality traits, (5) whether participants have ever experienced social exclusion in real life and how they coped with it.

The experiment consists of 8 rounds of real-effort task and two rounds of group discussion. The experiment is organised in blocks; in Block 1, the real effort task is repeated thrice. In Block 2, players are allocated into groups of four and engage in a group discussion. Then the real effort task is repeated thrice. In Block 3, players are first informed of the possibility of social exclusion if exclusion treatments (Color or Performance) or of the possibility of learning whether they are the worst performer (Feedback). Then, they first perform the real effort task, then engage in a group activity (excluded persons cannot), and engage in the real effort task again.

computer randomization

Groups are formed according to subjects' initial performance (non-random), player roles are randomly assigned in each group. Treatments are distributed across the sessions, also non-randomly to have a balanced set.

Yes

In Baseline, there are no clusters. In other treatments, data is clustered at the group level even though performance remains strictly independent of others. In baseline, the aim is to get at least 50 observations, and in other treatments, about 30. This results in a total of 170 clusters.

410

50-30-30-30 for Baseline-Color-Performance-Feedback treatments

When comparing the performance difference between included and excluded players in last round of real-effort task in the Color treatment to that of the Performance treatment, we will use a two-tailed multiple linear regression (t-test, single regression coefficient). With α = 0.05, power = 0.80, total sample size = 28 (the Color treatment) + 28 (the Performance treatment), and the number of predictors = 3, the minimum effect size Cohen's f^2 is 0.1455200. A similar analysis holds when comparing the Feedback treatment to Performance. Finally, we will pool all the data and compare Baseline performance to the treatments when controlling for social exclusion. Then, in a multiple regression with 3 predictors (treatment, exclusion, previous round performance), and 140 observations (50 from Baseline, and 30 each from the other three treatments), minimum detectable effect is 0.0568657.