Learning in Speculative Bubbles: An Experiment

Résumé

Does traders' experience reduce their propensity to participate in speculate bubbles? This paper studies this issue from a theoretical and an experimental viewpoint. We focus on a game in which bubbles, if they arise, are irrational, as in the Smith, Suchanek, and Williams (1988)'s set up. Our theoretical results are based on Camerer and Ho (1999)'s Experience-Weighted Attraction learning model. Adaptive traders are assumed to adjust their behavior according to actions' past performance. In the long run, learning induces the market to converge to the unique no bubble equilibrium. However, learning initially increases traders' propensity to speculate. In the short run, more experienced traders thus create more bubbles. An experiment shows that bubbles are very pervasive despite the fact that subjects have become experienced. Our estimation of the EWA model also indicates that learning is at work.