Equilibrium Transitions in Sampling Dynamics: An Experimental Investigation *

Abstract

We experimentally investigate sampling best response dynamics (sBRD) (Oyama et al., 2015) that predict sampling-based information (i.e., a random sample of opponents' play) facilitate equilibrium transitions than full information. In the experiment, subjects played a coordination game in one of three treatments differing in sampling sizes. We observe more efficiency-improving transitions with a larger sampling size. We identify two channels for equilibrium transitions by combining experiments, structural estimations and simulations. Strategic teaching emerges as the main driving factor, and the sBRD does facilitate efficiency-improving equilibrium transitions. However, the sampling effect is weakened by subjects' lagged responses, as they place larger weights on earlier signals, especially under smaller sampling sizes. Overall, sampling promotes equilibrium transition given a small number of teachers, while revealing full information works better when sufficiently many teachers are present.

Keywords: Equilibrium Transition, Sampling Dynamics, Strategic Teaching, Adaptive Learning JEL Codes: C73, C91, C92, D83