Abstract. Experimentalists frequently claim that human subjects in the laboratory violate game-theoretic predictions. It is here argued that this claim is usually premature. The paper elaborates on this theme by way of raising some conceptual and methodological issues in connection with the very definition of a game and of players ’ preferences, in particular with respect to potential context dependence, interpersonal preference dependence, backward induction and incomplete information.

Abstract not found

Abstract not found

An observer makes a number of observations of an industry producing a homogeneous good. Each observation consists of the market price, the output of individual firms and perhaps information on each firm’s production cost. We provide various tests (typically, linear programs) with which the observer can determine if the data set is consistent with the hypothesis that firms in this industry are playing a Cournot game at each observation. When cost information is wholly or partially unavailable, these tests could potentially be used to derive cost information on the firms. This paper is a contribution to the literature that aims to characterize (in various contexts) the restrictions that a data set must satisfy for it to be consistent with Nash outcomes in a game. It is also inspired by the seminal result of Afriat (and the subsequent literature) which addresses similar issues in the context of consumer demand, though one important technical difference from most of these results is that the objective functions of firms in a Cournot game are not necessarily quasiconcave.

We provide general conditions under which an economic theory has a universal axiomatization: one that leads to testable implications. Roughly speaking, if we obtain a universal axiomatization when we assume that unobservable parameters (such as preferences) are observable, then we can obtain a universal axiomatization purely on observables. The result "explains" classical revealed preference theory, as applied to individual rational choice. We obtain new applications to Nash equilibrium theory and Pareto optimal choice.

A universal type space of interdependent expected utility preference types is constructed from higher-order preference hierarchies describing (i) an agent’s (unconditional) preferences over a lottery space; (ii) the agent’s preference over Anscombe-Aumann acts conditional on the unconditional preferences; and so on. Two types are said to be strategically indistinguishable if they have an equilibrium action in common in any mechanism that they play. We show that two types are strategically indistinguishable if and only if they have the same preference hierarchy. We examine how this result extends to alternative solution concepts and strategic relations between types.

on earlier drafts of this paper. Anat Bracha, Rustam Ibragimov and Dmitry Shapiro gratefully acknowledge the

In this paper we develop tests for whether play in a game is consistent with equilibrium behavior when preferences are unobserved. We provide necessary and sufficient conditions for observed outcomes in extensive game forms to be rationalized first, partially, as a Nash equilibrium and then, fully, as the unique subgame-perfect equilibrium. Thus one could use these conditions to find that play is (a) consistent with subgame-perfect equilibrium, or (b) not consistent with subgame-perfect behavior but is consistent with Nash equilibrium, or (c) consistent with neither. Further, we discuss the relevance of the test outcomes for rationalization of data by multiple preference profiles.

Centre de recherche et développement en économique (C.R.D.E.) / Centre interuniversitaire de recherche en économie quantitative (CIREQ) and Département de sciences économiques, Université de Montréal

Abstract not found