The optimal design of rewards in contests

Using contests to generate innovation has been and is widely used. Such contests often involve offering a prize that depends upon the accomplishment (effort). Using an all-pay auction as a model of a contest, we determine the optimal reward for inducing innovation. In a symmetric environment, we find that the reward should be set to c(x)/(c' (x) - beta) where c is the cost of producing an innovation of level x and beta is the weight attached by the designer to the sum of efforts. In an asymmetric environment with two firms, we find that it is optimal to set different rewards for each firm. There are cases where this can be replicated by a single reward that depends upon accomplishments of both contestants.