Knowledge that Transforms

The authors study repeated games in which players observe a public outcome that imperfectly signals the actions played. They provide conditions guaranteeing that any feasible, individually rational payoff vector of the stage game can arise as a perfect equilibrium of the repeated game with sufficiently little discounting. The central condition requires that there exist action profiles with the property that, for any two players, no two deviations--one by either player--give rise to the same probability distribution over public outcomes. The results apply to principal-agent, partnership, oligopoly, and mechanism-design models, and to one-shot games with transferable utilities.

The Generalized Extreme Value Model was developed by McFadden for the case with discrete choice sets. The present paper extends this model to cases with both discrete and continuous choice sets and choice sets that are unobservable relative to the analyst. We also propose behavioral assumptions that justify random utility functions (processes) that have a max stable structure i.e., utility processes where the finite dimensional distributions are of the multivariate extreme value type. Finally we derive non-parametrically testable implications for the choice probabilities in the continuous case.

This paper presents conditions under which a person's beliefs about the occurrence of uncertain events are quantified by a capacity measure, i.e., a nonadditive probability. Additivity of probability is violated in a large number of applications where probabilities are vague or ambiguous due to lack of information.The key feature of the theory presented in this paper is a separation of the derivation of capacities for events from a specific choice model. This is akin to eliciting a probability distribution for a random variable without committing to a specific decision model. Conditions are given under which Choquet expected utility, the Machina-Schmeidler probabilistically sophisticated model, and subjective expected utility can be derived as special cases of our general model.