ABSTRACT
As a first attempt to formally model this tension, we analyze a variant of a model proposed by Morris and Shin (2002). The model features a set of agents endowed with both private and common information who simultaneously submit predictions. With some probability, agents are rewarded on the basis of the accuracy of their individual prediction relative to the final outcome. With complementary probability, agents are rewarded on the distance between their prediction and the average prediction across agents (the consensus forecast). This probabilistic interpretation of the rewards captures the notion that the market designer may only observe the value of the idea with some probability, which we refer to as the prediction market intensity. The mixed nature of rewards means that agents care not only about their own assessment of the final outcome, but also – because of the incentives to coordinate with other agents – about their assessments of other agents’ assessments, and about the assessments of other agents about the assessments of other agents, and so on. Higher order beliefs play a key role in these markets, as in Keynes’ (1936) celebrated metaphor of financial markets as beauty contests.6
