Harsanyi power solutions for graph-restricted games (Q532510)

The paper considers cooperative transferable utility games with limited communication structure [\textit{R. B. Myerson}, Math. Oper. Res. 2, 225--229 (1977; Zbl 0402.90106)]. In such games players can cooperate if and only if they are connected in the communication graph and a solution can be obtained by applying standard solutions to a restricted game that takes account of the cooperation restrictions. The authors introduce Harsanyi power solutions for graph restricted games which are based on Harsanyi solutions for cooperative transferable utility games (cf. [\textit{V. A. Vasil'ev}, Optimizatsiya 28(45), 102--111 (1982; Zbl 0486.90091); Diskretn. Anal. Issled. Oper., Ser. 1 10, No. 2, 17--55 (2003; Zbl 1060.91020)]). A Harsanyi power solution distributes Harsanyi dividends (cf. [\textit{J. C. Harsanyi}, Ann. Math. Stud. 40, 325--355 (1959; Zbl 0084.36501)]) such that the dividend shares of players in a coalition are based on power measures for nodes in corresponding communication graphs. The payoff to each player \(i\) is equal to the sum of its shares in the dividends of all coalitions of which he or she is a member. The power measure is a function that assigns a nonnegative real number to every node in the graph. A special power measure is the so-called degree measure that assigns to every player in a communication graph the number of players with whom he is directly connected. On the class of cycle-free graph games, the corresponding Harsanyi power solution is equal to the position value (cf. [\textit{P. Borm, G. Owen} and \textit{S. Tijs}, SIAM J. Discrete Math. 5, No. 3, 305--320 (1992; Zbl 0788.90087)]), while it equals the Shapley value (cf. [\textit{L. Shapley}, Ann. Math. Stud. 28, 307--317 (1953; Zbl 0050.14404)]) on the class of complete graph games. Applying the equal power measure, which assigns equal power to all players, they obtain the Myerson value as the corresponding Harsanyi power solution. After weakening some of the axioms used in Borm et al. [loc.\,cit.] to axiomatically characterize the position and Myerson values on the class of cycle-free graph games, in particular, the degree measure property and the communication ability property, they generalize these axioms to obtain two axiomatic characterizations of all Harsanyi power solutions joint with the properties of component efficiency, additivity and superfluous link property. Both give characterizations for the position value and the Myerson value as special cases. So, the difference between the position value and the Myerson value (on cycle-free graph games) is not about using the degree measure property or the communication ability property, but about which power measure to use. It turns out that one of the axioms is not satisfied on the class of all graph games. By replacing this axiom by two invariance axioms, in particular invariance with respect to the so-called inessential links in unanimity graph games and to the worths of non-connected coalitions, they also obtain an axiomatic characterization on the class of all graph games. Various applications are discussed. In assignment games (cf. [\textit{L. Shapley} and \textit{M. Shubik}, Int. J. Game Theory 1, 111--130 (1971; Zbl 0236.90078)]) they compare the total payoff to the sellers and the total payoff to the buyers and consider the case that buyers and sellers cannot trade directly with each other, but need intermediaries. ATM-games (cf. [\textit{E. Bjorndal, H. Hamers} and \textit{M. Koster}, Math. Methods Oper. Res. 59, No. 3, 405--418 (2004; Zbl 1148.91324)]) models a situation of \(n\) banks on a single location, where some banks have an automated teller machine (money dispenser) and others not. The banks may agree to cooperate in the sense that customers of banks not having an ATM are allowed to make use of the ATMs of the other banks, resulting in cost savings because using ATMs is a relatively cheap way of cash withdrawals. They consider situations when there is only one single bank having an ATM or when there are multiple banks having an ATM and compare the payoffs proposed by the Harsanyi power solution with the ones proposed by the Shapley, position, Myerson and \(\tau\)-values (for the latter, cf. [\textit{S. H. Tijs}, in: Game theory and mathematical economics, Proc. Semin., Bonn/Hagen 1980, 123--132 (1981; Zbl 0467.90087)]), the nucleolus (cf. [\textit{D. Schmeidler}, SIAM J. Appl. Math. 17, 1163--1170 (1969; Zbl 0191.49502)]) and the so-called aggregate allocation solution [Bjorndal et al., loc.\,cit.]. Finally, they consider the cooperative game associated with a second-price sealed bid auction with \(n\) bidders when all bidders collude and reveal their private valuations [\textit{D. A. Graham, R. C. Marshall} and \textit{J. F. Richard}, Am. Econ. Rev. 80, 493--510 (1990)] and compare the Shapley value and the Harsanyi degree solution.