An experimental study of sorting in group contests

We study experimentally the effects of sorting in contests between groups of heterogeneous players whose within-group efforts are perfect substitutes. The theory predicts that higher aggregate effort will be reached when variation in ability between groups is lower, i.e., by a more balanced sorting. In the experiment, we assign subjects to four types -- A, B, C, and D -- ranked by their cost of effort, with A having the lowest and D having the highest cost, and conduct contests between two groups of two players each. In the Balanced treatment, (A,D) groups (i.e., groups comprised of a type A and a type D player) compete with (B,C) groups, whereas in the Unbalanced treatment, (A,B) groups compete with (C,D) groups. We find substantial heterogeneity and overinvestment of efforts by all types in both treatments, including the "underdog" (C,D) group which surprisingly is not demoralized by the unbalanced matching. Despite strong overbidding, relative aggregate efforts are remarkably close to equilibrium predictions both between treatments and between groups within each treatment. The results confirm the prediction that balanced sorting leads to higher aggregate effort.