Bivariate scoring rules: unifying the characterizations of positional scoring rules and Kemeny's rule (Q6564045)

In this paper the author studies social preference functions (SPFs) which, given the voters' strict preferences \(R\) over some set of alternatives, compute a non-empty set of winning rankings \(\succ\) over a feasible subset of the alternatives. Two of the most prominent classes of SPFs are positional scoring rules and Kemeny's rule, both of which have repeatedly been characterized in the literature. In this paper the author unifies these two independent streams of research by characterizing the class of \textit{bivariate scoring rules}, which contains both Kemeny's rule and all positional scoring rules, in particular the Borda rule. Roughly, bivariate scoring rules can be seen as variants of Kemeny's rule that weight comparisons between alternatives differently, depending on their positions in the voters' preference relations. Intuitively, a bivariate scoring function quantifies how important it is for a voter that his \(l\)-th best alternative is ranked ahead of his \(k\)-th best alternative in the output ranking. The author characterizes this class of bivariate scoring rules by mainly relying on two axioms called reinforcement and local agenda consistency, which formalize consistency notions with respect to variable electorates and variable agendas. More precisely, the main theorem says that an SPF is a bivariate scoring rule if and only if it satisfies anonymity, neutrality, continuity, faithfulness, reinforcement and local agenda consistency.\N\NIntuitively, \textit{reinforcement} requires that if some rankings are chosen for two disjoint elections, then precisely the common winning rankings are chosen in a joint election.\N\N\textit{Agenda consistency} postulates of an SPF \(f\) that if \(X, Y\) are sets of alternatives such that \(X \subseteq Y\), then for every preference profile \(R\) for an electorate \(N\), \(F(R, X) = \{\succ\hspace{-1mm}|X ;\ \succ \in f(R, Y)\}\), where \(f(R, Y)\) is the set of winning rankings given preference profile \(R\) and the set of alternatives \(Y\). Positional scoring rules satisfy this condition, but Kemeny's rule does not.\N\N\textit{Local agenda consistency} is based on the idea that a winning ranking \(\succ\) for a large set \(Y\) should only inherit to a subset \(X\) if the alternatives in \(X\) appear consecutively in \(W\).\N\NBased on his main theorem, the author also infers characterizations of the class of positional scoring rules and of Kemeny's rule as corollaries. More precisely, Corollary 1 says that an SPF is a positional scoring rule if and only if it satisfies anonymity, neutrality, continuity, faithfulness, reinforcement and agenda consistency. And Corollary 2 says that an SPF is Kemeny's rule if and only if it satisfies anonymity, neutrality, continuity, faithfulness, reinforcement, local agenda consistency and independence of infeasible alternatives.\N\N\textit{Independence of infeasible alternatives} demands that the winning rankings should only depend on the available alternatives.\N\NThese two corollaries can be seen as variants of the characterizations by \N\textit{J. H. Smith} [Econometrica 41, 1027--1041 (1973; Zbl 0286.90008)]\N and \N\textit{H. P. Young} [Am. Polit. Sci. Rev. 82, No. 4, 1231--1244 (1988; \url{doi:10.2307/1961757})],\N which demonstrates that his main theorem unifies the axiomatic research on positional scoring rules and Kemeny's rule.\N\NIt should be noted that the proof of the main theorem requires ten lemmas which are grouped into four different subsections (Appendices A1, A2, A3 and A4) and together require about thirteen full pages. Fortunately, the author gives also a sketch of the proof, which is still one and a half page long.