How Information Structure Impacts Aggregation of Information in Markets

We study information aggregation in financial markets. We introduce two new stability concepts: absolute stability (sensitivity of rational expectations equilibrium prices to signals (bits) that are lost or garbled), and relative stability (the difference between prices in the rational expectations and the corresponding private-information equilibria). The concepts are applied to two canonical settings of information aggregation, namely: (i) winner-take-all contracts, where a binary payment depends on the signals held by the majority of the agents; (ii) assets whose payoffs depend on the average of the signals across the agents. The former is unstable under both concepts, and hence, would lead rational agents to doubt the reliability of prices. The latter is robust, and hence, prices aggregate information more reliably. Experiments, designed to avoid any effect of risk attitudes, confirm the predictions. Through second-price auctions, we observe how participants trust prices less in the winner-take-all setting. Our findings have important implications for the design of prediction markets, for historical evidence on the Efficient Markets Hypothesis, and for recent experimental work on information aggregation