Peer Effects under Different Relative Performance Feedback and Grouping Procedures

We conduct a laboratory experiment to test theoretical predictions about subjects' performance in an effort task conditional on their peer group's composition and relative performance feedback. Subjects are grouped either randomly or according to their ability, with the feedback being the best or average performance of their group. While theory-derived hypotheses on aggregate treatment differences cannot be confirmed, we find evidence when gender differences are taken into account. Male subjects perform significantly better when they compare themselves with the best peer instead of the average, while the opposite is true for females. With respect to the grouping treatment, we find that random grouping is beneficial for male subjects, and ability grouping for female subjects. These differences are explained by gender differences in (non-linear) reactions to the reference point and an aversion of females to competitive environments.