The Ex Ante Likelihood of Bubbles

The limits of arbitrage explain how a speculative bubble is sustained; they do not explain how likely one is to occur. To do that, you need a theory about the thing that sporadically causes arbitrageur constraints to bind. I propose a first such theory, which is based on social interactions between speculators. The theory says that bubbles should be more likely in assets where increases in past returns make excited-speculators relatively more persuasive to their peers. I empirically verify this ex ante prediction about bubble likelihoods and show that it is robust to some ex post disagreement about bubble definitions.