Reconstructing preferences in a voting context: Some clues from experimental data on voting rules: Antoinette Baujard and Herrade Igersheim
Introduction How can we identify the “correct” individual preferences or metarankings – ranking on rankings of preferences1 – to be used at the collective level in “real life”? How can we implement a true liberal collective choice? We know that aggregating individual preferences into a collective choice – hence selecting a legitimate winner – is the aim of voting rules. Besides, voting rules are also used to observe individual preferences. But ballot papers reveal information on individual choices rather than on individual preferences. The voting rule properties play a significant role as for the passage from actual preferences to choices not only from a positive perspective, but from a normative one. Let us call actual individual preferences initial or first preferences. The latter are translated into different ballot choices according to the properties of the voting rule at stake. Therefore, to plead for or against a voting rule, one has first to take into account how this voting rule (re-)constructs and modifies individual first preferences, and second, to examine whether the reconstruction it leads to matches the normative values one wants to promote. In this chapter, we may consider the primary translation of individual preferences into these individual ballot choices as the reconstruction of individual (first) preferences into individual metarankings. The characteristics of such metarankings hence strongly depend on the voting rule properties. Thus, the examination of voting rules may help to answer whether there is any “better”, if not ideal or satisfying, mechanism to implement a liberal collective choice based on metarankings. Specific studies of voting rules on the basis of experimental data should allow proposing better clues for such quest. We shall resort to experimental data on alternative voting systems which have been obtained during a large-scale field experiment conducted on April 22, 2007, in parallel with the first round of the French presidential elections (see Baujard and Igersheim 2007, 2009). The aim of our chapter is to give new insights into the influence of a voting rule on the expression of individual preferences. The issue of preferences’ modification or (re-)construction is inherent in voting theory. For a given voting method, a collective decision has to be deduced from the ballots which derive from individual preferences. Yet, as stressed above, ballots correspond to a very
specific way to express one’s individual preferences. We will attempt to show here that voting methods have two kinds of influences. First, the mechanical influence which is linked to the requested information for a given voting rule shall be called the filtering problem; one can say that this problem deals with the difficulty of identifying individual (first) preferences. Second, the endogenous influence of voting rules which relates to the modification of individual preferences once translated into ballot papers shall be called the metaranking issue; it deals mainly with the distinction between individually and collectively oriented individual preferences. Further, we claim that, compared with the voting rules usually used in our democracies, some deal in a more convincing way with the problems induced by these two kinds of influences, i.e., to reduce the filtering problem and to increase individuals’ chances of voting according to their metaranking. Following more particularly Goodin (1986), we opt in this chapter for a specific definition of metaranking in terms of “public-oriented preferences” or “ethical preferences”. We will show that approval voting and evaluation voting – the two alternative voting rules we tested in 2007 – and more particularly approval voting, are more able to cope with these two constraints than plurality vote, even with the two-round system. In particular, they may give more insights of what actual preferences are, and give a better approximation of what the metarankings should be. Our chapter is organized as follows: section “Experimental design and main results” is a presentation of the experimental design and the main results. Sections “The filtering effect of voting rules (1): restriction of ballot information” and “The filtering effect of voting rules (2): strategic behaviors” are devoted to the filtering effect, and section “The endogenous influence of voting rules (3): voting behaviors” to the endogenous effect. Section “Conclusion: three arguments about metarankings”, which exposes the metaranking issue and gives arguments to resolve it, concludes the chapter.