An investment bubble is a period of excessive, and predictably unpro table, investment (DeMarzo, Kaniel and Kremer, 2007, p.737). Such bubbles most often accompany the arrival of some new technology, such as the tech stock boom and bust of the late 1990 s and early 2000 s. We provide a rational explanation for investment bubbles based on the dynamics of learning in highly uncertain environments. Objective information about the earnings potential of a new technology gives rise to a set of priors, or a belief function. A generalised form of Bayes Rule is used to update this set of priors using earnings data from the new economy. In each period, agents who are heterogeneous in their tolerance for ambiguity make optimal occupational choices, with wages in the new economy set to clear the labour market. A preponderance of bad news about the new technology may nevertheless give rise to increasing rm formation around this technology, at least initially. To a frequentist outside observer, the pattern of adoption appears as an investment bubble.
Ambiguity Belief Function Investment Bubble Inference