Learning in Dynamic Games

点击翻译按钮获取中文摘要

This project develops an axiomatic intuitive learning model that can be used as an alternative way of selecting an equilibrium from a class of dynamic games, where the conventional approach like strategic stability is often ineffective. The conventional rational expectations equilibrium is often ineffective because it is founded on a strong "rationality" assumption. Each economic agent must perfectly foresee all others' strategies and beliefs without prior communication or coordination. One alternative to the conventional approach is called "rationalizability". The rationalizability literature develops solution concepts that are intuitively more plausible than the conventional approach, but admit too many outcomes to be an effective solution concept. Another alternative is to develop axiomatic learning models with very specific assumptions about the behavior of individual economic agents. The contribution of this project comes from developing axiomatic learning models that capture the intuition of rationalizability and make precise predictions about the outcome of economic problems. The project develops an axiomatic intuitive learning theory and uses it to generalize the Coase theorem for durable good monopolists, an important economic theorem about monopoly behavior. The theory is also used to determine the conditions under which learning will lead to cooperation. Finally, the results obtained in the first two parts of the project will be applied to alternating offer bargaining models, bargaining where each party can stop the learning process once and for all by accepting the offer.