The neoclassical theory of consumer behavior makes strong assumptions about the informational
and computational bases of consumer behavior. The core assumption is that
consumer behavior is reasonably characterized as the maximization of expected lifetime
utility subject to a budget constraint and conditional on the available information. In short,
consumer behavior can be characterized as the solution to a discounted dynamic programming
problem. For economists, this approach has many attractions: it meshes well with
traditional notions of economic rationality, it is theoretically tractable (at least in its standard
formulations), and it generates predictions that are readily testable (at least apparently).
Unfortunately, the available empirical evidence often conflicts with these predictions.
Using a unique, survey-based data set, we uncover some additional difficulties for neoclassical
theory. For example, we consider the response of consumers to a very simple
optimization subproblem, the solution of which is independent of other choices for a neoclassical
consumer. We find that the behavior of even highly educated consumers deviates
radically from the neoclassical predictions: they postpone the receipt of income. Furthermore,
it appears that many consumers believe that a smooth income stream aids them to