Essays in equilibrium selection and adaptive behavior
Loading...
Date
1993
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
This dissertation consists of two essays on experiments on the problem of equilibrium selection. The literature includes a variety of models in which multiple equilibria arise. One approach to the selection problem is to view the equilibria as the fixed points to some specific dynamical system; depending on initial conditions, a selection dynamic picks one of the equilibria. The accuracy of selection dynamics is an empirical question. In addition to the accuracy of selection dynamics it is also important to consider their precision. Using both of these characteristics a measure of predictive success can be constructed to compare the selection dynamics. The first experiment examines people's behavior in a generic game with multiple Pareto ranked equilibria. The experiment was designed to discover whether behavior diverges at the separatrix predicted by two specific selection dynamics. The myopic selection dynamic uses only information from the previous period. The percipient dynamic uses information from all previous periods. Both predict the same two sets of initial conditions as the initial conditions which indicate selection of a particular equilibrium. As predicted, the equilibrium selected is sensitive to small changes in initial conditions. Those sessions with initial conditions that would lead both dynamics to select the Pareto dominated equilibrium converge to that equilibrium. The experiment provides some striking examples of coordination failure growing out of small historical accidents. The second experiment examines people's behavior in a full information environment in which selection dynamics characterize the equilibria differently. The data is also used to test a convergence result for adaptive behavior in general. The results support the predictions made by the percipient dynamic as well as general convergence result. The point is made, however, that the general convergence result lacks the precision of a specific formalization such as the two specific forms examined. This lack of precision is demonstrated by a measure of predictive success like the one mentioned above. This measure is used to compare the precision and accuracy of both the general convergence result and the specific forms of the two selection dynamics.
Description
Vita.
Keywords
Major economics