Parallelism in Information Production: Moral Hazard and Relative Performance Evaluations

We analyse how to design incentive schemes for agents producing information. The agents may, for example, be divisional managers, market researchers, OR consultants, financial analysts, corporate auditors, research scientists or political pollsters. We show that an agent should be compensated more for making a correct rather than an incorrect prediction, and for agreeing rather than disagreeing with colleagues. Depending on the environment, correctness should be compensated more or less than agreement. Also, compensation should increase with the difficulties of making correct or conforming predictions. One consequence of the result is that it may sometimes reduce a principal's total incentive costs to engage with more information producers performing parallel investigations, because the possibility to make relative performance evaluations of the agents may substantially reduce the costs of motivating the agents in performing their duties as desired.