Learning Competitive Equilibrium in Laboratory Exchange Economies

Crockett, Spear and Sunder [2005b] propose an algorithm whereby boundedly rational agents with standard neoclassical preferences learn competitive equilibrium in a repeated static exchange economy. In this paper a laboratory market is instituted to examine the hypothesis that people are at least as sophisticated as these agents. The adopted market institution strongly restricts the space of agent actions, facilitating the identification of decision rules. Evidence for learning competitive equilibrium is mixed due to strong heterogeneity in decision-making. Some subjects clearly demonstrate the ability to learn across periods. However, a majority exhibit little evidence of learning, and many are, in fact, simply content to satisfice, though the opportunity to do better was fairly straightforward. The presence of satisficers permits noncompetitive outcomes within this particular market institution, but I conjecture learners may stimulate convergence to competitive equilibrium in others.