On elections with robust winners

Abstract

We study the sensitivity of election outcomes to small changes in voters' preferences. We assume that a voter may err by swapping two adjacent candidates in his vote; we would like to check whether the election outcome would remain the same given up to delta errors. We show that this problem can be viewed as the destructive version of the unit-cost swap bribery problem, and demonstrate that it is polynomial-time solvable for all scoring rules as well as for the Condorcet rule. We are also interested in identifying elections that are maximally robust with respect to a given voting rule. We define the robustness radius of an election with respect to a given voting rule as the maximum number of errors that can be made without changing the election outcome; the robustness of a voting rule is defined as the robustness radius of the election that is maximally robust with respect to this rule. We derive bounds on the robustness of various voting rules, including Plurality, Borda, and Condorcet.