Optimal Timing of Selection Contests

Abstract

A principal must choose an agent to perform a task and faces a tradeoff in deciding when to hold a selection contest. Agents’ types evolve and thus a later contest is more accurate. However, an agent’s effort in the contest diminishes her task performance and the less time until the task, the costlier the effort. We explore the optimal timing decision and show that later contests are better whenever types are more disperse and whenever ability at the task gives agents an edge in the contest. We find that the expected performance of the selected agent is invariant to the average ability and can either increase or decrease with the number of participants. We also recast the timing decision as a more general screening problem in which the principal suffers the signaling costs of the chosen agent. We argue that the optimal screening mechanism is equivalent to that in Chakravarty and Kaplan (2006), is stochastic, and favors agents with low hazard ratios.