Speaker: Matthew Ryan
Affiliation: The University of Auckland
Title: Inference with Ambiguous Priors and an Economic Application
Date: Tuesday, 13 Sep 2011
Time: 4:00 pm
Location: Room 6115, Owen Glenn Building
This paper considers statistical inference when the prior takes the form of a belief function (Dempster, 1967; Shafer, 1976) rather than a probability. We review some approaches to this non-standard inference problem and discuss their properties. The paper also develops an economic application, in which entrepreneurs learn about a new market or technology over time. We demonstrate that these learning dynamics, when embedded in an equilibrium model of price determination, can produce an “investment bubble”: a boom in investment despite unfavourable market data – a frequentist evaluation would lead one to reject the new technology – followed by the inevitable crash. The investment boom and bust in tech stocks of the late 1990’s is a recent example of the phenomenon. Curiously, the initial boom is driven not by the increasing exuberance of over-optimistic entrepreneurs, but by the diminishing resistance of their more conservative employees and financiers. This is a Joint work with Luca Rigotti (Pittsburgh) and Rhema Vaithianathan.