Strive to be first and avoid being last: An experiment on relative performance incentives

Managers often use tournament incentive schemes which motivate workers to compete for the top, compete to avoid the bottom, or both. In this paper we test the effectiveness and efficiency of these incentive schemes. To do so, we utilize optimal contracts in a principal-agent setting, using a Lazear-Rosen type model that predicts equal effort and efficiency levels for three tournament incentive schemes: reward tournaments, punishment tournaments, and tournaments combining reward and punishment. We test the model s predictions in a laboratory experiment and find that the combination of reward and punishment produces the highest effort from agents, especially in contests of a relatively larger size. Punishment is shown to be more effective and, in larger contests, more efficient than rewards, and it is also the mechanism with the lowest variance of effort. Finally, we show that behavior in all mechanisms is consistent with a model of basic directional and reinforcement learning.