Learning in Speculative Bubbles: Theory and Experiment 

Abstract

Does learning reduce or fuel speculative bubbles? We study this issue in the context of the Bubble Game proposed by Moinas and Pouget (2013). Our theoretical analysis based on adaptive learning shows that i) in the long run, learning induces convergence to the unique no-bubble equilibrium, ii) in the short run, more experienced traders create more bubbles, and iii) learning is more difficult when more steps of reasoning are necessary to reach equilibrium. These predictions are consistent with our experimental observations. We find that reinforcement learning rather than belief-based learning is driving behavior in our experiment.

Keywords: financial markets, adaptive learning, speculation, bubbles