Learning in dynamic games

Chong Huang, University of Pennsylvania


Many dynamic strategic environments feature players' learning about the fundamentals from both exogenous and endogenous information sources. While players resolve uncertainties about the fundamentals by receiving informative signals (exogenous learning), previous plays often provides counter evidence (endogenous learning). As a result, the comparison between these two learning processes largely determines players' equilibrium behaviors. In addition, different games have different features, which have significant interplays with the endogenous learning. In this dissertation, I study the effects of learning on players' behaviors in three dynamic environments. ^ Chapter 1 discusses the interaction between coordination and social learning in a dynamic global game. Short-lived players need to coordinate to succeed at some state of the fundamentals. They learn the state from their own private information and from previous players' actions. Coordination motives significantly change the way public history aggregates private information, and thus affect the comparison between the two learning processes. Such a comparison, in the meantime, influences the coordination outcome. ^ Chapter 2 studies effects of speculators' learning in a dynamic regime change game, with a policy maker trying to build her reputation to defend against speculative attacks. Long-lived speculators collect private signals about the policy maker's type in every period, while they also learn the policy maker's type from her previous defending decisions. The comparison between the speed of the potential reputation building (endogenous learning speed) and that of the individual learning (exogenous learning speed) determines whether attacks can happen and the dynamics of attacks in the equilibrium with attacks (if exists). ^ Chapter 3 analyzes the seller's learning about the current buyer's value and the existence of new high value short-lived buyers in a Coasian bargaining model. New buyers, if exist, arrive stochastically and will take any available offer. So if no transaction happens, the seller becomes more pessimistic about the existence of new buyers. At the same time, the seller uses prices to screen the current buyer's value. Therefore, the seller has to balance these two learning processes, which determines the strategic delay and price fluctuations in the bargaining.^

Subject Area

Economics, General|Economics, Theory

Recommended Citation

Chong Huang, "Learning in dynamic games" (January 1, 2012). Dissertations available from ProQuest. Paper AAI3509139.