The inverse problem for power distributions in committees (Q2629512)

Given a certain power index, the author studies the question of how to design a voting procedure such that the resulting distribution of power (according to the power index in question) among the committee members meets or almost meets a fixed distribution. \textit{N. Alon} and \textit{P. H. Edelman} [Soc. Choice Welfare 34, No. 3, 371--377 (2010; Zbl 1201.91045)] gave explicit bounds stating that a Banzhaf vector whose weight is concentrated on \(k<n\) players, has to be near to the Banzhaf vector of a game with \(n - k\) null players, which is essentially a game with just \(k\) players. The author shows that such Alon-Edelman type results exist for other power indices like e.g. the Public Good index or the Coleman index to prevent actions, while such results are principally impossible to derive for e.g. the Johnston index. The author studies the Shapley-Shubik index, the absolute Banzhaf index, the Coleman index to prevent action, the Coleman index to initiate action, the König-Bräuninger index (KB), the Public Help index (PHI), the absolute Johnston index, the absolute Public Good index, the absolute Deegan-Packel index, the absolute Shift index and the absolute Shift-Deegan-Packel index. It turns out that for the power-indices KB and PHI it makes no sense to ask for Alon-Edelman type results, since the respective power indices do not admit power distributions where most of the power is concentrated on a small number of players. For the other considered power indices such a concentration is in principle possible. Nevertheless, there can be no such result for the Johnston index, which has been shown by analytical power index calculations for a certain class of parameterized weighted games.