Essays in the Empirical Theory of Preferences

Abstract

In this dissertation, I investigate various empirical aspects of the theory of preferences and decision. A classical result in revealed preference shows that when one observes a subject's choice on a rich enough collection of choice sets, the weak axiom of revealed preference is both necessary and sufficient for the choice data to be rationalized by a preference relation. In Chapter 1, I provide a complete characterization of how far these complete data assumptions may be relaxed, while still retaining a suitably powerful weak axiom. I then explore connections between these richness conditions and the classical literature on demand integrability. Relative to the existing literature which focused weakening the analytic regularity conditions under which the system of partial differential equations defining the integrability problem can be solved, in Chapter 2 prove a ``nothing assumed" integrability theorem, that not only imposes no regularity conditions on model primitives, but also applies to arbitrary data sets, finite or infinite, in contrast with the traditional literature which requires an infinite set of observations. Finally, in Chapter 3, I develop a least squares regression theory for a novel form of choice data. I show that for a wide range of decision theoretic models, across a variety of domains, constructing statistical tests of consistency for these models may be reduced to a standard problem of testing multiple linear moment inequalities. Applications to trade, welfare, and the eliciting of subjective beliefs, are provided.