scientific article; zbMATH DE number 1071776

This paper is an account of the author's recent work on voting and aggregation paradoxes, and social choice procedures. It pulls together salient features of several earlier works by the author, seeking to shed light on long standing debates on suitable ways of ranking candidates which are based on expressed rankings of candidates by individual voters. One interesting aspect of the paper is the light it throws on comparisons of positional voting methods. Let \(n\) denote the number of candidates in an election; \(P_n\) denote the set of strict, complete and transitive rankings over the \(n\) candidates. A preference profile of the voters is denoted by \(p\), a point in the \(m\)-fold Cartesian product \(P^n\times\cdots\times P^n\), where \(m\) is the number of voters. A positional voting method is described by a voting vector \(w^n= (w_1,\dots, w_n)\), where \(w_j\geq w_{j+1}\), \(j= 1,\dots, n-1\), \(w_1> 0\), \(w_n=0\); \(w_j\) is the point assigned to a voter's \(j\)th ranked candidate, the candidates being ranked, socially, by tallying the total points scored and comparing these totals. The Borda Count method (\(\text{BC}^n\)) is defined by the voting vector satisfying \(w_j= n-j\), for each \(j\), and the plurality method (\(\text{PM}^n\)) is defined by the voting vector \((1,0,\dots,0)\). Let \(f(p, w^n)\) denote the \(n\)-candidate election ranking for a given preference profile \(p\) when the voting vector is \(w^n\). Let \(C_1,C_2,\dots,C_K\) denote all the possible subsets of two or more candidates and \(| C_j|\) denote the cardinality of \(C_j\). Let \(w^{| C_j|}\) denote the voting vector used if the election were to take place out of \(C_j\) and call \(W^n= (w^{| C_1|},\dots, w^{| C_K|})\) the system voting vector. The election outcomes over all subsets \(C_1,\dots, C_K\) for a given profile \(p\) and voting system \(W^n\) is \((f(p, w^{| C_1|}),\dots, f(p, w^{| C_K|}))\). Denote this by \(F(p, W^n)\). Let \(U^n\) be the set of all possible rankings over the sets \(C_1,\dots, C_K\). A ranking in \(U^n\) need not be the result of an election according to a voting system, but the subset of \(U^n\), \(D^n(W^n)\equiv \{F(p, W^n)/p\) is a profile; there are no restrictions on the number of voters\} corresponds to election outcomes according to \(W^n\), for some profile. Voting paradoxes are interpreted to mean conflicts or reversals of rankings, over subsets of the set of candidates, produced by a given voting system, for the same profile. An arbitrary ranking in \(U^n\) would typically involve such inconsistencies, only a subset of points in \(U^n\) would be free of paradoxes. From this point of view, the smaller the set \(D^n(W^n)\) the less the paradoxes admitted by the election procedure \(W^n\). The following theorem provides an idea of the kind of comparisons of positional voting methods achieved. Theorem: For \(n\geq 3\), there is a lower-dimensional algebraic set \(\alpha^n\) so that if \(W^n\not\in\alpha^n\) then \(D^n(W^n)= U^n\). In particular, \(\text{PM}^n\not\in \alpha^n\). Further, if \(w^n\neq \text{BC}^n\) then \(D^n(\text{BC}^n)\neq D^n(w^n)\). The theorem shows, in particular, that the plurality method admits all possible paradoxes while the Borda Count admits the least and, in this sense, is the ``best positional voting method. It is an interesting feature of the works that combinatorial problems are reduced to geometric equivalents. A second interesting aspect of the studies is the examination and interpretation of Condorcet's paradox and Arrow's impossibility theorem, in which symmetry consideration of profiles and loss of information involved in pairwise voting procedures, in particular of transitivity, are key features, and lead to the conclusion that social choice impossibility results have a very benign interpretation, once they are properly viewed. The arguments here, however, seem less compelling. The paper is engagingly written, providing a good sense of the overall direction and program of research, but provides no proofs, for which readers are referred to published articles and books. It is a valuable piece for anyone interested in social choice theory.

Recommendations

Cited in

(12)

This page was built for publication:

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4359091)