Some regrettable grading scale effects under different versions of evaluative voting (Q2235100)

This paper concerns evaluative voting, a single-winner voting method in which voters are asked to evaluate each candidate on a pre-determined numerical scale; the candidate receiving the largest total is elected. Alternative voting is the particular case with the binary scale \(\{0,1\}\). The analysis of the paper is empirical, drawing on a field experiment conducted in parallel with the first round of the 2017 French presidential election. The willing voters in four polling stations were offered to vote experimentally twice: with approval voting and with another modality of evaluative voting. The aim is to study the effects induced by negative grades and scale length under evaluative voting. Among the 11 candidates, five of them are considered major (i.e. viable) candidates and six of them minor. The former are selected here according to their real results (Fillon, Le Pen, Macron and Mélenchon) or the political context (Hamon). The major candidates are divided into two groups, popular and polarizing, depending on whether or not they attract a significant proportion of intermediate grades among non-approving voters. Theoretically, for a given scale length, voters should not be sensitive to the specification of grades with or without negative ones. The experiment suggests that the presence of negative grades favours non-viable candidates. As for major candidates, there is no difference in the scores of polarizing candidates and the increase in the scores of the popular candidates is weaker than that of the non-viable candidates and sometimes too weak to be taken into account. Concerning the length of the scale, it follows from the data that the normalized scores of the candidates are unchanged under various scales, except for some non-viable candidates. On the other hand, it is found that the length of the scale changes the use of the extreme grades. The lengthening of the scale leads to a significant decrease of the proportion of voters who use the entire extent of the grading scale. The authors interpret this result as suggesting that a major part of voters favour expressive voting over strategic voting.