Theorem 1 (only if), p. 9 / Appendix A — equilibria in all games of 𝒢(N, f^𝕎, 𝕎) force one ω with f^W = GWSV[ω] of the distributed welfare (10)
ProvedCostSharingPNE.Char.theorem_1_only_ifThis is the "only if" (necessity) direction of Theorem 1 of Gopalakrishnan–Marden–Wierman, the part to which Appendix A is devoted ("we only need to prove one direction").
Let , , be the players, a nonempty set of local welfare functions, and the corresponding distribution rules (not necessarily budget-balanced). For each let be the welfare actually distributed by ,
If every game in possesses a pure Nash equilibrium, then there is a single weight system such that, for every ,
Together with the converse (generalized weighted Shapley values guarantee equilibria, Hart–Mas-Colell), this gives the full characterization of Theorem 1: equilibrium existence in all games forces every rule to be a generalized weighted Shapley value of the welfare it distributes, with one common weight system.
Formalization Note Same conventions as CostSharingPNE.Char.theorem_1: the game class quantifies over every number of resources, every assignment of welfare functions from to resources (repetitions allowed) and every family of nonempty action sets. The rules are indexed by the elements of , and two different may distribute the same welfare ; the statement still asks for one serving all of them.
import Mathlib import Definitions.Def_CostSharingPNE_Char_Setting import Definitions.Def_CostSharingPNE_Char_WeightSystem
namespace CostSharingPNE.Char
theorem theorem_1_only_if (n : ℕ) (hn : 1 < n) (𝕎 : Set (Welfare n)) (h𝕎 : 𝕎.Nonempty)
(f : Welfare n → Rule n) (hpne : GuaranteesPNE 𝕎 f) :
∃ ω : WeightSystem n, ∀ W ∈ 𝕎, ∀ S : Finset (Fin n), ∀ i ∈ S,
f W i S = gwsv ω (distributed (f W)) i S := by sorry
end CostSharingPNE.Char