Learning and Beliefs in Non-Centralized Markets.

Abstract

This dissertation uses microeconomic theory to examine learning and beliefs in markets that are not centralized, in order to refine our knowledge of how information asymmetries affect market outcomes. I approach problems in this space from a game theoretic perspective, first building an appropriate market game to study the question of interest, then characterizing the equilibria of the game, and finally applying those characterizations to the initial problem.

The dissertation consists of three distinct chapters. The first of these examines the conditions that affect learning and efficiency in a decentralized market. Past research addressing this issue has shown in several models that learning is difficult and asymmetric information is costly. In a steady state environment, the first chapter shows that both learning and efficiency are possible if and only if the good being traded is divisible. This chapter identifies the divisibility of trade as a significant friction in decentralized markets.

The second chapter studies how beliefs regarding a market front-runner affect surplus in a fragmented market. A front-runner is an agent who has advance knowledge of market orders, gained from either an illegal or a technological advantage. So far, research involving front-runners has placed restrictions on the behavior of other agents, assuming that agents cannot change their strategies based on the presence of a front-runner. This work provides a framework to relax those restrictions in order to allow agents to best respond to a front-runner. The research then suggests that the impact of a front-runner on traders' strategies generally leads to a decrease in total surplus.

The third chapter again contributes to the literature on learning and efficiency in a decentralized market. This chapter studies the conditions that are necessary for learning and long-run efficiency in a non-stationary environment. Past research has obtained long-run efficiency if trade is divisible and bargaining is flexible; this research shows that only divisible trade is necessary. Furthermore, while past research requires traders use complex strategies in order to learn, our research demonstrates that traders may use simple strategies, and will learn by participating in the market and observing the actions of their match partners.