Axiomatizing the Harsanyi solution, the symmetric egalitarian solution and the consistent solution for NTU-games (Q1764289)

The authors study three classical solutions for NTU-games: the Harsanyi solution (H-solution), the symmetric egalitarian solution of Kalay/Samet (KS-solution) and the consistent solution of Maschler/Owen (MO-solution). The starting point is Hart's axiomatization of the H-solution as the unique solution defined on the class of partially positively games (\({\mathcal G}_{PSm}'\)) and saisfying six axioms: Efficiency (EFF), Scale Covariance (SC), Independence of Irrelevant Alternatives (IIA), Conditional Super-Additivity (CSA), Unanimity Games (UG) and Zero-Inessential Games (ZIG). The authors review this axiomatization and discuss Hart's result by focusing on the importance of the regularity conditions imposed on games. Observing that Hart's axiomatization is incompatible for solutions defined on the set of all games (\({\mathcal G}\)), they showed that after removing the axiom SC, his axiomatization uniquelly determines the KS-solution on \({\mathcal G}\). The second main result of the paper says that the MO-solution is the only solution defined on the class of positively games (smaller than \({\mathcal G}_{PSm}'\)) and satisfying four axioms, EEF, IIA, CSA and some new axiom called Recursive Conditional Random Dictatorship (RCRD). Next the authors consider four other axioms, where the first is some modification of the axiom IIA and the remaining three are different versions of the axiom RCRD. With the help of these new axioms, they reformulate Hart's axiomatization for H-solution and their two previous results for HS-solution and MO-solution. At the end, a comprehensive analysis of indepedence of the considered axioms is given.