International Game Theory Review

Nash equilibrium leaves the impression that each player foresees perfectly and responds optimally. Must human-like, rational agents really acquire both these faculties? This paper argues that in some instances neither is ever needed. For the argument repeated play is modelled here as a constrained, decentralized, second-order process driven by noncoordinated pursuit of better payoffs. Some friction feeds into — and stabilizes — a fairly myopic mode of behavior. Convergence to equilibrium therefore obtains under weak and natural conditions. An important condition is that accumulation of marginal payoffs, along the path of play, yields a sum which is bounded above.

The article presents multilevel game theory, as a generalization of conventional single-level game theory as it has developed since von Neumann and Morgenstern (1944). We define a multilevel game structure, multilevel games, payoffs and distribution rules, upward feasible strategies and the solution concept multilevel Nash equilibrium (MNE) in such games. A MNE must be, for each player, a best reply against itself with respect to alternative strategies that may have other players deviate as well, in contrast to the NE for conventional games where simultaneous deviations by more than one player are not considered. Although every pure or mixed MNE must give the same outcome as a NE of the extensive form representation, a NE is not necessarily a MNE. It is shown that a MNE need not exist in pure or mixed strategies and, if it does, it may not be unique. In the former case, the multilevel structure is considered unmaintainable.

In this paper we study the geometrical properties of the set of Lorenz allocations of a transferable utility game. We provide procedures to compute a single Lorenz allocation, and even the entire set of Lorenz allocations, that rely solely on linear optimization techniques. These procedures only require a finite number of elementary operations and are therefore easy to implement.

In this paper we generalize the studies of Bondareva–Shapley for a general core having weights in the definition. These weights were introduced by Billera but not used in the form as here. Moreover we study such a core with conditions of equalities for some coalitions analogously as those obtained a posteriori in the assignment games due to Shapley and Shubik.

The grim-trigger strategy introduced by Friedman is often used in economic models, mainly because of its simplicity, to show that collusion can be sustained by means of a subgame perfect equilibrium. In this work we show that, under simple conditions, it is possible to improve on the grim-trigger strategy while retaining subgame perfection and in some cases adding weak renegotiation proofness (in the sense of Farrell and Maskin). The basic idea is that, following a deviation, the cheater, instead of continuing in the Nash equilibrium forever, chooses an autopenalty which signals, in a very strong way, that he/she would not deviate any more if cooperation were reestablished. We check the nice working of this strategy in the Cournot's oligopoly model.

The paper is a kind of generalization of Rubinstein bargaining model. Rubinstein assumed that preferences of the players were constant in time, and he analyzed models in which preferences of each player were defined either by constant discount rate or by constant bargaining cost. In this paper, a bargaining model is presented, in which preferences of each player are expressed simultaneously by sequence of discount rates and sequence of bargaining costs varying in time. The results presented in the paper concern subgame perfect equilibria. There is a theorem concerning sufficient and necessary conditions for the existence of subgame perfect equilibrium of the game. Moreover, some theorems presenting forms of subgame perfect equilibria for various cases of the model analyzed have been proved here. A possibility of delay in reaching an agreement is also considered in the paper. If we analyze a class of strategies, that depend on the former history, a delay can appear for some models. The adequate examples are presented. In the paper, some applications of the bargaining model are also described.

This note develops a mechanism for the derivation of individual player's payoff functions in cooperative stochastic differential games with nontransferable payoffs. An illustrative example is provided.