Hierarchical constrained egalitarianism in TU-games. (Q1867813)

The Dutta-Ray [\textit{B. Dutta} and \textit{D. Ray}, Econometrica 57, No. 3, 615--635 (1989; Zbl 0703.90105)] constrained egalitarian solution for cooperative games, based on the Lorenz-ordering and the Lorenz-core, is extended here to the case where some exogenous asymmetry between players may arise from considering hierarchical systems. A hierarchical system \(\Sigma\) is the same as a Kalai-Samet [\textit{E. Kalai} and \textit{D. Samet}, Weighted Shapley values, in: Roth, A. E. (ed.), The Shapley Value: Essays in Honor of Lloyd S. Shapley, Cambridge University Press, New York, USA, 83--99 (1988; Zbl 0708.90098)] weight system, so that it is defined by giving an ordered partition of the player set into layers and a positive weight distribution among the players. To achieve this purpose, \textit{U. Ebert}'s [Using equivalent income of equivalent adults to rank income distributions, Social Choice Welfare 16, 233--258 (1999)] weighted Lorenz-ordering, that extends the classical Lorenz-ordering to the case where just a weight distribution is given -- i.e. with trivial layer partition reduced to \(N\) --, is also adapted to the possible existence of different layers, giving rise to a hierarchical Lorenz-ordering as inequality measure. The hierarchical Lorenz-ordering, that depends only on the relative weights, is defined by means of a curve composed of one weighted Lorenz-curve for each layer, and it gives a (partial) dominance notion for coalition payoff vectors. Then, the \(\Sigma\)-Lorenz core is introduced in a recursive way and enables the authors to define viable coalition whenever the set of its Lorenz-core undominated payoffs is nonempty, and hence the set of \(\Sigma\)-constrained egalitarian solutions of the game when the grand coalition is viable. It is shown that this hierarchical constrained egalitarian solution (HCES) consists of one allocation at most, or else it is empty. A first algorithm is then provided for computing the HCES in case of trivial hierarchical systems, and it is shown to be successful for convex games; the proof essentially follows from the fact that, in a convex game, the set of coalitions that maximize the weighted average value is closed under union. A second algorithm -- an adaptation of the former -- refers to nontrivial hierarchical systems. If it gives an allocation in the Lorenz-core of N, this is the HCES of the game. For convex games, this algorithm always gives the HCES. Finally, it is shown that each core element of a nonnegative convex game is the HCES of the game for some hierarchical system, and this recalls a similar result of \textit{D. Monderer}, \textit{D. Samet} and \textit{L. S. Shapley} [Int. J. Game Theory 21, No. 1, 27--39 (1992; Zbl 0778.90094)] for the weighted Shapley values and the core of a convex game.