The bounded core for games with precedence constraints (Q1945083)

In this paper, the authors consider cooperative games in which there are restrictions on cooperation. These restrictions are modeled by a hierarchical structure on the set of players (cf. [\textit{U. Faigle} and \textit{W. Kern}, Int. J. Game Theory 21, No. 3, 249--266 (1992; Zbl 0779.90078)]), who are motivated by the existence of a partial order in the sense that if a coalition can be formed and a player belongs to a coalition, then, all players preceding him in the partial-order relation are also in the coalition. The classic model of cooperative games as a special case appears when all players are incomparable. It is a known fact that the core of cooperative games with transferable utility and with precedence constraints can be unbounded. Therefore, they appear in the literature various attempts to define a subset of the core that is bounded and having good properties. Although this core, which is a convex polyhedral set, can have unbounded faces, this paper considers the so-called bounded core that is defined as the union of all bounded faces. The bounded core can be interpreted as the set of core elements such that every player gets the most out of their subordinates. It coincides with the core in classic games. The paper shows that the bounded core is nonempty when the core is nonempty. To do this, they formulate a generalization of the Bondareva-Shapley theorem (cf. [\textit{O. N. Bondareva}, Probl. Kibern. 10, 119--139 (1963; Zbl 1013.91501)]) for games with precedence constraints given in [\textit{U. Faigle}, Z. Oper. Res. 33, No. 6, 405--422 (1989; Zbl 0685.90103)]. Furthermore, the bounded core is a continuous correspondence on games with coinciding precedence constraints. They generalize some well-known properties of solutions on a set of classic games: Pareto optimality, covariance under strategic equivalence, anonymity, boundedness, boundedness w.r.t. singletons, and the two-person zero-inessential game property. Furthermore, they generalize several properties of reduced games (called reduced-game property, weak reduced-game property, converse reduced-game property, reconfirmation property, and reconfirmation property w.r.t. classical games, respectively), for which they first define a generalization of the so-called Davis-Maschler reduced game. If the set of players has more than four elements, they axiomatically characterize the bounded core by the two-person zero-inessential game property, anonymity, covariance under strategic equivalence, reduced-game property, reconfirmation property w.r.t. classical games, converse reduced-game property, and boundedness. Moreover, the core is the maximum solution that satisfies the two-person zero-inessential game property, anonymity, covariance under strategic equivalence, reduced-game property, reconfirmation property w.r.t. classical games, converse reduced-game property, and boundedness w.r.t. singletons. Using variants of the reduced-game property, they obtain new characterizations of the bounded core and the core on the so-called games with connected hierarchies, so that these results are valid for any set of players. The proof of the axiomatic characterization of the bounded core is similar to the proof of the axiomatization of the prekernel (cf. [\textit{B. Peleg}, Int. J. Game Theory 15, 187--200 (1986; Zbl 0629.90099)]). Finally, they study the independence of the sets of axioms of the main results.