Open Access
Karavaev, Andrei
Graduate Program:
Doctor of Philosophy
Document Type:
Date of Defense:
June 16, 2008
Committee Members:
  • Kalyan Chatterjee, Committee Chair
  • James Schuyler Jordan, Committee Member
  • Sophie Bade, Committee Member
  • Leonid N Vaserstein, Committee Member
  • game theory
  • Microeconomics
This dissertation consists of three essays in economics. The first essay studies information trading in fixed networks of economic agents who can only observe and trade with other agents with whom they are directly connected. We study the nature of price competition for information in this environment. The linear network, when the agents are located at the integer points of the real line, is a specific example I completely characterize. For the linear network there always exists a stationary equilibrium, where the strategies do not depend on time. I show that there is an equilibrium where any agent has a nonzero probability of staying uninformed forever. Under certain initial conditions this equilibrium is a limit of equilibria of finite-horizon games. The role of a transversality condition is emphasized, namely that the price in the transaction should not exceed the expected utility of all the agents who get the information due to the transaction. I show that the price offered does not converge to zero with time. The second essay focuses on the problem of using idiosyncratic shocks and random matchings. Many economic models use a continuum of negligible agents to avoid considering one person's effect on aggregate characteristics of the economy. Along with a continuum of agents, these models often incorporate a sequence of independent shocks and random matchings. Despite frequent use of such models, there are still unsolved questions about their mathematical justification. In this paper we construct a discrete time framework, in which major desirable properties of idiosyncratic shocks and random matchings hold. In this framework the agent space constitutes a probability space, and the probability distribution for each agent is replaced by the population distribution. Unlike previous authors, we question the assumption of known identity - the location on the agent space. We assume that the agents only know their previous history - what had happened to them before, - but not their identity. The construction justifies the use of numerous dynamic models of idiosyncratic shocks and random matchings. The third essay examines inefficient equilibria in Games Played Through Agents. Games Played through Agents is a class of complete information games with multiple principles and multiple agents. Previous studies found conditions for the existence of the efficient equilibria, in which the sum of all the payoffs is maximized. In this paper, we prove that for a game with two principals and one agent only, any inefficient equilibrium is Pareto-dominated by an efficient one. A counterexample shows that this result does not hold for more than two principals. We also demonstrate that any efficient equilibrium is weakly dominated by the Truthful Nash Equilibrium.