Issue-by-issue voting: an experimental evaluation

Multiple referenda consists in making a common decision about each of a set of binary issues, given the preferences of a set of voters. Asking voters for their preferences on all combinations of values is not feasible in practice, because of the exponentially large number of such combinations; on the other hand, voting issue-byissue on each of the issues can lead to strongly paradoxical outcomes. This paper proposes to measure experimentally to which extent it is suboptimal to vote issue-by-issue voting, in function of the voting rule to be implemented, and the nature of the voters’ preferences (arbitrary, separable or additive). For this we use randomly generated separable profiles, which turns out to be a difficult problem. keywords. Computational social choice; preferences; voting; combinatorial domains; random generation.