Scoring run-off paradoxes for variable electorates (Q5927694)

A finite set of voters strictly orders a finite set of candidates or alternatives. A scoring rule assigns points to the candidates according to the individual rankings and a run-off rule selects a subset of the candidates for a subsequent round of voting. It is well-known that this situation suffers from a `no-show' paradox where a voter or group of voters may obtain a more favorable outcome by abstaining than by voting. As voters may choose not to vote, the size of the electorate may vary. In this paper, the authors investigate the likelihood of such a situation for the case of three candidates and different sizes of electorate. The authors subdivide the paradox according to whether voters are added to or subtracted from the electorate and whether the change is advantageous to a loser or disadvantageous to a winner. The authors first determine necessary and sufficient conditions for each of these four cases to arise. The authors determine probabilities under the assumption of the Impartial Culture condition where each voter chooses from a uniform distribution and the Impartial Anonymous Culture condition where each voting situation is equally likely. The authors compute the probabilities of the paradoxes under various scoring rules for both small electorates and large electorates, giving limiting values in the latter case. As the size of the electorate increases, the probability of a given paradox decreases, but not always monotonically, even after distinguishing even and odd electorates. The authors also show that abstention is not necessarily the optimal strategy for a group of voters when it is available.