Consistency and revenue sharing in hierarchies, with Mikel Álvarez-Mozos and René van den Brink.
Abstract: We study the effects of the reorganization of productivity in situations where agents are organized in a hierarchical structure. Specifically, we analyze the consequences of an agent at the bottom of the hierarchy being removed from the hierarchy, or promoted within it. We introduce new axioms related to the consistency principle in this context, which preclude agents not involved in these changes have their payoffs change as a consequence. We use these axioms, along with some standard axioms in the literature, to characterize four values for games with a hierarchical structure. One of the new axioms is satisfied by all four solutions, while the remaining four come naturally in two pairs: each of the four values satisfies one axiom from each pair. Out of the characterized values, three are known in the literature and are obtained by applying the Shapley levels value to a game with a levels structure derived from the hierarchical structure. The fourth value follows from the remaining combination of axioms; we define it through a recursive algorithm. Finally, we compare the four values to other values in the literature with respect to the proposed axioms.
Jané-Ballarín, M. (2025): "The essential coalitions index in games with restricted cooperation". International Journal of Game Theory 54, 34.
Abstract: We propose a new power index, which we call the essential coalitions index. The index is defined on the class of simple games with cooperation restricted by a cooperation index, which encompasses other restriction models, including Myerson’s communication games. In general, the games we study are not simple themselves. We introduce the essential coalitions of restricted games as an analogue to the minimal winning coalitions of a simple game; in particular, we show they provide a representation of the game. Their namesake index is defined to distribute payments based on the agents’ presence in essential coalitions. In terms of its characterization, the new index differs from the Shapley value of the restricted game in only one property. On the other hand, under a suitable normalization, the payoff distribution of the essential coalitions index generalizes that given by the Deegan-Packel index. Furthermore, this normalization admits an intuitive probabilistic interpretation, which we apply to the analysis of power distribution in a legislative chamber.
Non-obvious Manipulability with Groups in Shapley–Scarf Housing Market, with Louise Demoor, Pierre Nunn, Subhajit Pramanik, Antoine Prévotat and Makoto Yokoo.
Abstract: In Shapley-Scarf housing markets, top trading cycles (TTC) is the unique mechanism satisfying individual rationality, Pareto efficiency, and strategy-proofness. We weaken strategy-proofness to non-obvious manipulability (NOM). Ordinary NOM is too weak: individual rationality and Pareto efficiency already imply it. We therefore introduce NOM with groups, where an agent knows her group-mates' preferences but is uncertain about outsiders. We propose TTC with super-groups, a class of staged integration mechanism satisfying group rationality, Pareto efficiency, and NOM with groups. All such mechanisms have the same best- and worst-case envelope, and every mechanism satisfying these three properties has exactly this envelope.
Abstract: We study the complexity of the Banzhaf and Shapley-Shubik indices in weighted voting games with cooperation restricted by an incompatibility graph. In this model, a coalition can only cooperate if it forms an independent set of this graph. We show that the Banzhaf index is strongly #P-complete when the complement of the incompatibility graph is planar and perfect. As a consequence, the problem is strongly #P-complete even if the incompatibility graph has O(n) maximal independent sets. We also show that in some classes of graphs, the complexity is the same as in the unrestricted case: we show that deciding whether the index is non-zero is weakly NP-complete if the incompatibility graph has O (p(n)) maximal independent sets, where p is a polynomial, and that computing the index is weakly #P-complete if the incompatibility graph has O(1) maximal independent sets. These results also apply to the Shapley-Shubik index on incompatibility games.
Abstract: We study the effects of collusion in the polluted river problem. In this model, agents along a river network have to share the costs of cleaning the entire river. We define a collusion neutrality property which requires that, if all agents upstream of a given point agree to collude and act as one, then the cost to be paid by this new entity equals the sum of the costs originally allocated to the colluding agents. This property is related to the Unlimited Territorial Integrity doctrine used to settle international disputes over water streams flowing through several territories. We introduce a new cost sharing method, called the Upstream Recursive Equal Sharing (URES) method, and axiomatically characterize it using our new collusion neutrality axiom. We also provide a new characterization for the well-known Local Responsibility Sharing (LRS) method that uses this axiom and which differs from the URES method only in one property that explains how costs are shared at the source of the river. Moreover, we derive and characterize a family of collusion neutral methods where the different methods are distinguished by a parametrized clean source property that balances between Unlimited Territorial Integrity and Absolute Territorial Sovereignty and that contains the LRS method as an extreme and the URES method as an intermediate method. Finally, we introduce and characterize the other extreme method in this class, which we call the limit recursive sharing method. All members of the family can be computed by means of a recursive algorithm.