Decentralized Trading with Private Information

Joint with: MIchael Golosov and Guido Lorenzoni

We contribute to the recently developed theory of asset pricing in decentralized markets. We extend this literature to characterize an environment in which some agents have superior private information. In our model, agents have an additional incentive to trade assets to learn information that other agents have. First, we show that uninformed agents can learn all the useful information the long run, and that the long-run allocations are Pareto efficient. In the long run, therefore, the allocations coincide with those of the standard centralized market equilibrium such as in Grossman-Stiglitz. Second, we show that agents

with private information receive rents, and the value of information is positive. This is in contrast with the centralized markets in which prices fully reveal information and the value of information is zero. Finally, we provide characterization of the dynamics of the trades.

For more information, contact Harold Cole.