EXPERIENCE-WEIGHTED ATTRACTION LEARNING IN COORDINATION GAMES - PROBABILITY RULES, HETEROGENEITY, AND TIME-VARIATION

In earlier research we proposed an ''experience-weighted attraction (E WA) learning'' model for predicting dynamic behavior in economic exper iments on multiperson noncooperative normal-form games. We showed that EWA learning model fits significantly better than existing learning m odels (choice reinforcement and belief-based models) in several differ ent classes of games. The econometric estimation in that research adop ted a representative agent approach and assumed that learning paramete rs are stationary across periods of an experiment. In addition, we use d the legit (exponential) probability response function to transform a ttraction of strategies into choice probability. This paper allows for nonstationary learning parameters, permits two ''segments'' of player s with different parameter values in order to allow for some heterogen eity, and compares the power and legit probability response functions. These specifications are estimated using experimental data from weak- link and median-action coordination games. Results show that players a re heterogeneous and that they adjust their learning parameters over t ime very slightly. Legit probability response functions never fit wors e than power functions, and generally fit better. (C) 1998 Academic Pr ess.