Revealed Preference

A decisionmaker is said to weakly prefer Brie to Camembert if he or she likes the first cheese either more than or just as much as the second. Such a preference is, moreover, said to be revealed whenever the decisionmaker chooses Brie from a menu of options containing both of the cheeses in question. The logic behind this terminology is straightforward: An agent that views Camembert as superior to Brie will never choose the latter when the former is available, and therefore any observed choice of Brie in the presence of Camembert implies a weak preference in the opposite direction.

The concept of revealed preference grew out of the work of neoclassical economists. These neoclassical economists and their contemporaries showed that the standard practice of modeling economic agents as maximizers of numerical “utility” functions did not depend for its validity on any of a variety of questionable auxiliary assumptions about the form of the functions being maximized (such as additive separability in the variables upon which they depend). However, as this independence came to be understood, the logical strength and psychological content of the utility maximization hypothesis became increasingly unclear. On the one hand, had this hypothesis been reduced to a tautology, excluding no logical possibilities and thus capable of “explaining” any pattern of behavior? And on the other, did the new, neoclassical notion of utility continue to reflect the view of human [Page 838]decision making originally adopted by the founders of the utilitarian school?

Working in the context of consumer demand theory, Paul Samuelson sought to identify the “refutable implications” of utility maximization for behavior in market environments. In response to the first question previously listed, he showed that—far from being tautological—this hypothesis has definite implications that can be expressed as a prohibition against the preferences revealed by the decisionmaker's actions coming into conflict with each other. With regard to the second question, Samuelson's contribution made clear that in psychological terms, agents in economic models are typically endowed with both the wellintegrated personalities and the substantial cognitive resources needed to behave consistently across different choice problems. Indeed, it is this internal consistency of the decisionmaker's behavior, rather than any assumption about the tastes or values lying behind it, that has come to be seen as the essence of the utility maximization hypothesis.

While Samuelson's definition of a revealed preference was phrased in terms of choices among consumption bundles in a market setting, the idea that opinions or other mental states can be deduced from observed behavior applies much more generally. For example, in the theory of decision making under uncertainty, an agent's assessments of the relative likelihoods of different events (such as a particular company going bankrupt within the next year or a particular political party winning a majority in the next national election) are revealed by his or her choices among bets contingent on the unknown information.