Payoffs in nondifferentiable perfectly competitive TU economies (Q1851227)

The author studies a value of finite-type market games defined in the following way: Let \((T,C)\) be a standard measurable space with a set \(T\) of agents and a \(\sigma\)-algebra \(C\) of coalitions in \(T\). An economy \(E\) is a triple \((u,a,\eta)\), where \(u: T\times \mathbb{R}^k_{+} \rightarrow \mathbb{R}\) and \(a: T\rightarrow \mathbb{R}^k_{+}\) for some \(k\), and \(\eta\) is an element of the set \(NA\) of nonatomic probability measures on \((T,C)\). Here \(a_t=a(t)\) is agent \(t\)'s initial endowment of commodities, and \(u_t=u(t,\cdot)\) is his utility function on the space of commodity bundles \(\mathbb{R}^k_{+}\), and \(\eta\) is the population measure. Let \(\mathcal{G}_F = \bigcup_{\eta\in NA} \mathcal{G}_F(\eta)\) where each \(\mathcal{G}_F(\eta)\) is the set of so-called \textit{finite type} economies \(E=(u,a,\eta)\) associated with \(\eta\), defined by the following conditions: (1) \(a\) and \(u\) are \(H\)-measurable and \(H\times B\)-measurable, respectively, for a finite subfield \(H\) of \(C\) (\(B\) is the Borel \(\sigma\)-field of \(R^k_{+}\)); (2) for any \(t\in T\), \(u_t\) is a continuous and nondecreasing function, satisfying \(u_t(0)=0\), \(u_t(x) = o(\| x\| )\) as \(\| x\| \rightarrow \infty\), and \(\int_Ta_t\,d\eta (t)>0\). Now, for any finite-type economy \(E\in \mathcal{G}_F\), a game is defined in the form of its characteristic function \(v_E: C\rightarrow \mathbb{R}\), by \(v_E(S)=\max \{\int_T u_t(x_t)\,d\eta (t)\}\) for any coalition \(S\), where ``max'' is taken over the set of all commodities \(x:T\rightarrow \mathbb{R}^k_{+}\) satisfying \(\int_S x_(t)\,d\eta (t) = \int_S a_t\,d\eta (t)\). Let \(MG\) be the set of all such games \(v_E\) and let \(FA_{+}\) be the set of bounded and finitely additive measures. The author studies a value \(\Psi: MG\rightarrow FA_{+}\) satisfying the three standard axioms (Efficiency, Anonymity and Additivity) and the Contraction axiom (\(\| \Psi(v)-\Psi(v')\| \leq \| v-v'\| \) for any \(v\) and \(v'\)). The main result of the paper (with a nontrivial proof) is that \(\Psi\) coincides with the Mertens value restricted to the set \(MG\).