Self-allocation in contests

We consider contestants who must choose exactly one contest, out of several, to participate in. We show that when the contest technology is of a certain type, or when the number of contestants is large, a self-allocation equilibrium, i.e., one where no contestant would wish to change his choice of contest, results in the allocation of players to contests that maximizes aggregate equilibrium effort. For a class of oligopoly models that are equivalent to contests, this implies output maximization.