Stable voting procedures for committees in economic environments (Q5955034)

The paper seeks to extend the investigation of the social choice problem of finding satisfactory voting rules, involving a one-dimensional set of alternatives, to multidimensional sets. The approach may be encapsulated as follows. Let \(G\) denote a committee (simple game) described by a set of players \(N = \{1,2,\dots,n\}\) and a set \(W\subseteq 2^N\), where \(2^N\) denotes the collection of non-empty subsets of \(N\) of winning coalitions; \(G\) is the pair \((N,W)\). Let \(A\) denote the set of alternatives and \(\succsim^i\)-denote the preferences of the player \(i\) in \(N\). The question being investigated is whether there exists a game fomr (abb, GF) \(\Gamma\), described by the sets of strategies, \(\sum^i\), for each player \(i\) in \(N\), and the outcome function \(\pi :\sum^1\times \dots\sum^n=\sum^N\to A\), (so \(\Gamma = (\sum^N,\pi,A))\) which implements, in strong Nash equilibria (abb. SNE), the core allocations, when \(A\) and \(\succsim^i\) are suitably restricted, and, furthermore, the power structure induced by the game form on the set of players \(N\) is the same as (that of the given committee) \(G\). Some formal definitions and results follow. Let \(\mathcal P\) be a set of (admissible preference) orderings over \(A\). Let \(\succsim^N\) be an \(n\)-tuple of preference profiles, i.e., \(\succsim^N\in{\mathcal P}^N\). \(\sigma\in\Sigma^N\) is a strong Nash equilibrium (SNE) of the game in strategic form, \((\Gamma,\succsim^N)\), if for all \(S\in 2^N\) and \(\mu^S\in\Sigma^S\) there exists \(i\in S\) such that \(\pi(\sigma)\succsim^i\pi(\mu^S,\sigma^{N\setminus S})\); \(SNE(\Gamma,\succsim^N)\) denotes the set of SNE's of \((\Gamma,\succsim^N)\). An effectivity function (abb. EF) is a function \(E : P(N) \to P(P(A))\) satisfying \(E(N) = 2^A\), \(E(\emptyset) = \emptyset\), \(A\in E(S)\) for all \(S\in 2^N\) and \(\emptyset\notin E(S)\) for all \(S\in P(N)\). If \(B\in 2^A\) and \(S\in 2^N\) then \(S\) is \(\alpha\)-effective for \(B\) if there exists \(\sigma^S\in \Sigma^S\) such that \(\pi(\sigma^S,\mu^{N\setminus S})\in B\) for all \(\mu^{N\setminus S}\in \Sigma^{N\setminus S}\). The \(\alpha\)-EF of \(\Gamma, E^\Gamma_\alpha\), is defined by: \(E^\Gamma_\alpha(\emptyset) = \emptyset\) and \(E^\Gamma_\alpha(S) = \{B\in 2^A : S\) is \(\alpha\)-effective for \(B\}\), for \(S\in 2^N\). The core of \(E\) with respect to \(\succsim^N\), \(C(E,\succsim^N)\) is the set of undominated alternatives, i.e, an alternative \(x\) such that there is no \(B\in 2^A\) and \(S\in 2^N\) satisfying \(\{B\in E(S)\) and \(y\succ^i x\) for all \(y\in B\}\). \(\Gamma = (\Sigma^N,\pi,A)\) (partially) implements the core over \({\mathcal P}^{*N}\subset{\mathcal P}^N\) if for every \(\succsim^N\in {\mathcal P}^{*N}\), \((\pi(SNE(\Gamma,\succsim^N))\subset C(E,\succsim^N))\) \(\pi(SNE(\Gamma, \succsim^N)) = C(E,\succsim^N)\). The simple game \(G=(N,W)\) is assumed to be monotonic, i.e., \([S\in W\) and \(S\subset T\subset N]\Rightarrow T\in W\), and proper, i.e., \(S\in W\Rightarrow N\setminus S\notin W\) for all \(S\in 2^N\). If \(E\) is an EF, denote by \(W_E\) the set \(\{S \in 2^N : E(S) = 2^A\}\). \(\Gamma\) is a representation of \(G\) if \(G = (N,W_{E^\Gamma_\alpha})\), the simple game associated with \(\Gamma\). \(\Gamma\) is a strong representation of \(G\) if it is a representation of \(G\) and \((\Gamma,\succsim^N)\) has a SNE for all convex and continuous preference profiles \(\succsim^N\). \(V = \cap\{S : S\in W\}\) is the set of vetoers. If there are no vetoers, i.e., \(V = \emptyset\), then the Nakamura number of \(G\), \(v(G)\), is given by \(v(G) = \min\{\text{cardinality of }U : U\subset W\) and \(\cap\{S : S\in U\} = \emptyset\}\). The authors show that (1) if the admissible preferences are also bounded.i.e., for each \(a\in A\), the set \(\{x\in A: x\succsim a\}\) is bounded and \(A = R^m\) then any proper and monotonic game \(G\) has a game form representation \(\Gamma\) which implements the core (and, therefore, if the core is non-empty, which is the case if \(\nu(G)\geq m +2\), then \(\Gamma\) has a SNE for all admissible preference profiles); (2) if \(A\) is a compact convex subset of \(R^m\) with full dimension, i.e., affine hull \(A = R^m\), and if there are vetoers then \(G\) has a strong representation \(\Gamma\). In general, however, strong representations do not exist and examples are provided; in particular, if the Nakamura number of \(G\) is small enough in relation to the dimension of \(A\), strong representations do not exist. However, it is also shown that there are examples of games without vetoers where exists a strong representation. In the special case when \(A\) is a compact interval of the real line, there is always a game form representation which partially implements the core.