Binary strategy-proof social choice functions with indifference (Q2143915)

The authors give a new characterization of binary (i.e., there are only two alternatives, say \(a\) and \(b\)) social choice functions, which are coalitionally strategy-proof (i.e., no group of voters is able to achieve a better outcome for the group by misrepresenting their preferences). The authors assume that the individual preferences allow for indifferences. The set of voters \(V\) may be finite or infinite. The social choice function is supposed to satisfy universal domain, i.e, it is defined for every possible profile, but other conditions such as anonymity are not required. The authors give an example that shows that coalitional strategy-proofness does not coincide with individual strategy-proofness in case the set of voters is infinite. In order to formulate Theorem 2.12 we need the following definitions, assuming that the two alternatives are \(a\) and \(b\). Given a profile \(p\), \(I(p)\) is the set of all voters \(i\) which are indifferent between \(a\) and \(b\), \(S(p) = V - I(p)\), \(D(a, p)\) is the set of all voters \(i\) in \(S(p)\) which strictly prefer \(a\) to \(b\) and \(D(b, p)\) is the set of all voters \(i\) in \(S(p)\) which strictly prefer \(b\) to \(a\). Given two profiles \(p\) and \(q\), let \(E(p,q)\) be the set of all voters \(i\) in \(V\) such that \(p(i) = q(i)\). We say that \(p\) dominates \(q\), denoted by \(p \sqsupset q\), if either \(V = E(p, q) \cup [I(p) \cap D(a, q)]\) -- in which case we write \(p \sqsupset_{a} q\) - or \(V = E(p,q) \cup [I(p) \cap D(b, q)]\) -- in which case we write \(p \sqsupset_{b} q\). A social choice function \(\phi\) is compatible with the dominance relation \(\sqsupset\), when \(p \sqsupset_{\phi(p)} q\) implies \(\phi(p) = \phi(q)\). The first characterization result of the authors is Theorem 2.12: A binary social choice function \(\phi\) is coalitionally strategy-proof if and only if \(\phi\) is compatible with \(\sqsupset\). The second characterization of coalitionally strategy-proof binary social choice functions is in terms of functions of type \(\psi\). A binary social choice function \(\phi\) is of type \(\psi\) if there is suitable indexed double collection \((\mathcal{I}, \mathcal{F})\) = \((I_{\lambda}, \mathcal{F}_{\lambda})_{\lambda \in \Lambda}\), where \(\Lambda\) is a well-ordered set and for every index \(\lambda \in \Lambda\), \(I_{\lambda}\) is a proper subset of the set \(V\) of voters and the family \(\mathcal{F}_{\lambda}\) is a superset closed family on \(V - I_{\lambda}\), and an alternative \(x\) such that \(\phi\) can be written as \(\psi_{(\mathcal{I},\mathcal{F}), x}.\) For the definition of the social welfare function \(\psi_{(\mathcal{I}, \mathcal{F}), x}\) we refer to the paper itself. In Theorem 4.2 the authors prove that every binary coalitionally strategy-proof social choice function \(\phi\) with range \(\{a, b\}\) (i.e., onto) can be written as a function of type \(\psi\).