Contest Schemes and Dynamic Incentive Provision

Simple contest schemes, which tend to use extreme contemporaneous peer benchmarking, are generally thought to be less efficient at motivating productive and cooperative effort than explicit bonus contracts that optimally weight observed own and peer outputs. We show, however, that contest schemes can outperform explicit bonus contracts in a two-period dynamic setting if periodic output is correlated over time, and contract terms can be renegotiated. In our model, all parties update their beliefs about ex ante unknown time-invariant noise upon observing first-period outputs, and second-period wage offers are updated accordingly. This creates first-period shirking incentives for the agents. We show that relative to bonus contracts, the extreme contemporaneous peer benchmarking in contest schemes can weaken or even eliminate the adverse incentives created by prior-period benchmarking, even though the available information and the learning process are the same under both schemes. This dynamic advantage allows contest schemes to outperform repeated bonus contracts for a wide range of parameter values, despite imposing higher static efficiency losses in each period. If competing agents are assigned to sufficiently similar tasks, a contest scheme’s ability to sever the statistical link between periods can yield not only higher early productive effort compared with bonus contracts, but also less uncooperative behavior, despite putting the agents in a more contest-like situation.