We study how rework routing together with wage and piece rate compensation can strengthen incentives for quality. Traditionally, rework is assigned back to the agent who generates the defect (in a self routing scheme) or to another agent dedicated to rework (in a dedicated routing scheme). In contrast, a novel cross routing scheme allocates rework to a parallel agent performing both new jobs and rework. The agent who passes quality inspection or completes rework receives the piece rate paid per job. We compare the incentives of these rework allocation schemes in a principal-agent model with embedded quality control and routing in a multi-class queueing network. We show that conventional self routing of rework cannot induce first-best effort. Dedicated routing and cross routing, however, strengthen incentives for quality by imposing an implicit punishment for quality failure. In addition, cross routing leads to workload allocation externalities and a prisoner’s dilemma, thereby creating the greatest incentives for quality. Firm profitability depends on demand levels, revenues, and quality costs. When the number of agents increases, the incentive effect of cross routing reduces monotonically and approaches that of dedicated routing.

We present a general methodology to automate the search for equilibrium strategies in games derived from computational experimentation. Our approach interleaves empirical game-theoretic analysis with reinforcement learning. We apply this methodology to the classic Continuous Double Auction game, conducting the most comprehensive CDA strategic study published to date. Empirical game analysis confirms prior findings about the relative performance of known strategies. Reinforcement learning derives new bidding strategies from the empirical equilibrium environment. Iterative application of this approach yields strategies stronger than any other published CDA bidding policy, culminating in a new Nash equilibrium supported exclusively by our learned strategies.

We examine a proposed auction algorithm using quantum states to represent bids and distributed adiabatic search to find the winner 1. When the auctioneer follows the protocol, the final measurement giving the outcome of the auction also destroys the bid states, thereby preserving privacy of losing bidders. We describe how a dishonest auctioneer could alter the protocol to violate this privacy guarantee, and present methods by which bidders can counter such attacks. We also suggest possible quantum circuit implementations of the auction protocol, and quantum circuits to perpetrate and to counter attacks by a dishonest auctioneer.

We present a synthesis of the various folk theorems for repeated games using a model that accommodates both finitely and infinitely repeated games with discounting. We derive a central result for this model and show that the various folk theorems follow as a consequence. Our result encompasses theorems involving epsilon equilibria and incomplete information.

This paper proposes a theory of cooperation over finite horizons, focusing on public good contribution games, that implies the broadly documented feature of decreasing cooperation over time. The central assumption is that there are two types of players: those who only care about their own material payoffs, and those who reciprocate others ’ contributions. The main result is that if reciprocity functions satisfy some regularity conditions, then generically there is a unique perfect equilibrium, in which contributions are decreasing. In this equilibrium, selfish players contribute to induce subsequent contributions by reciprocal players, and this incentive diminishes as the end of the play approaches. The model explains the puzzling restart effect and is consistent with various other empirical findings. In one-shot games, the model predicts no contributions. We also report the results of a series of experiments, using a probabilistic continuation design in which after each round, the game is restarted with low probability. The results support the implications of our model that the restart effect is present even with experienced players, whereas, in one-shot games, contributions disappear with experience. We show that experienced players correctly foresee the pattern of contributions, suggesting that the declining pattern comes from equilibrium play. We also identify the presence of conditional reciprocity among experienced players, and document that selfish players (identified exogenously) stop contributing earlier than reciprocal players, as implied by the model.

Abstract: The link between economic theory and experimental data is much tighter than is commonly supposed. Many presumed paradoxes arise because the theory is incorrectly applied. We go through several examples, emphasizing the theory as seen by a theorist. The main problem with the theory is that in some instances it lacks predictive power – We highlight where this is the case and current theoretical work designed to remedy the problem. 1 We are grateful to NSF grant SES-03-14713 for financial support, to Drew Fudenberg and Tom Palfrey for many conversations on this topic, to Colin Camerer for helpful comments and to Guillaume Freche for encouraging us to do this.

The computability of Nash equilibrium points of finite games is examined. When payoffs are computable there always exists an equilibrium in which all players use computable strategies. However, there is a computable sequence of games for which the equilibrium points do not constitute a computable sequence. For this reason, there can be no algorithm that, given arbitrary payoffs, computes a Nash equilibrium point for the game. Even for games with computable equilibrium points, best responses of the players may not be computable. In contrast, approximate equilibria, and error-prone responses are computable.

We conducted a two-stage game experiment with a non-excludable public good. In the first stage, two subjects choose simultaneously whether or not they commit to contributing nothing to provide a pure public good. In the second stage, knowing the other subject's commitment decision, subjects who selected not to commit in the first stage choose contributions to the public good. We found no support for the evolutionary stable strategy equilibrium, and the ratio of subjects who did not commit to contributing nothing increased as rounds advanced; that is, the free-riding rate declined over time. Furthermore, this behavior did not arise due to altruism or kindness among subjects, but from spiteful behavior of subjects. Correspondent: Until July 31, 2003 Tatsuyoshi Saijo 228-77 Humanities and Social Sciences California Institute of Technology Pasadena, CA 91125 USA Phone 1-626-395-4216 (phone and fax) E-mail: saijo@hss.caltech.edu From August 1, 2003 Tatsuyoshi Saijo Institute of Social and Economic Research Osaka University Ibaraki, Osaka 567-0047 Japan Phone 81-6 (country & area codes) 879-8582 (office)/878-2766 (fax) E-mail: saijo@iser.osaka-u.ac.jp 1.

The concept of bounded rationality is not well-defined. Several aspects of bounded rationality are discussed. Two different kinds of bounded rationality are distinguished. First, rationality may be bounded in the sense that the player cannot perform all the necessary calculations within the time frame of the game. This applies not only to numerical calculations but also to any kind of information processing, for example, figuring out logical consequences, or self-examination of prefer- ences. Recent studies on the effect of bounded rationality concentrate on this type, studying the effect of limited computational capability on the set of Nash-equilibria of the game. Some objections to the approach taken in these papers in dealing with the question of bounded rationality are presented.

CIRJE Discussion Papers can be downloaded without charge from: