Bucklin Voting is Broadly Resistant to Control

Electoral control models ways of changing the outcome of an election via such actions as adding/deleting/partitioning either candidates or voters. These actions modify an election’s participation structure and aim at either making a favorite candidate win (“constructive control”) or prevent a despised candidate from winning (“destructive control”), which yields a total of 22 standard control scenarios. To protect elections from such control attempts, computational complexity has been used to show that electoral control, though not impossible, is computationally prohibitive. Among natural voting systems with a polynomial-time winner problem, the two systems with the highest number of proven resistances to control types (namely 19 out of 22) are “sincere-strategy preference-based approval voting” (SP-AV, a modification of a system proposed by Brams and Sanver) and fallback voting. Both are hybrid systems; e.g., fallback voting combines approval with Bucklin voting. In this paper, we study the control complexity of Bucklin voting itself and show that it behaves equally well in terms of control resistance for the 20 cases investigated so far. As Bucklin voting is a special case of fallback voting, all resistances shown for Bucklin voting in this paper strengthen the corresponding resistance for fallback voting.

💡 Research Summary

The paper investigates the computational resistance of Bucklin voting to the full spectrum of electoral control actions, a topic that has become central in the study of election security. Electoral control refers to a set of manipulative actions—adding, deleting, or partitioning candidates or voters—that aim either to make a distinguished candidate win (constructive control) or to prevent an undesirable candidate from winning (destructive control). There are 22 standard control scenarios when the four basic actions are combined with the two goals. The prevailing approach to protecting elections from such attacks is to show that the decision problem associated with each control scenario is computationally intractable (NP‑hard), thereby rendering the attack impractical for large instances.

Prior work identified two hybrid voting systems—sincere‑strategy preference‑based approval voting (SP‑AV) and fallback voting—as the most robust among systems with polynomial‑time winner determination, each resisting 19 out of the 22 control types. Fallback voting, for instance, merges approval voting with Bucklin’s round‑by‑round majority check. Bucklin voting itself, however, had not been examined in isolation. The authors fill this gap by providing a systematic complexity analysis of Bucklin voting across all control scenarios that have been studied to date.

The Bucklin rule works as follows: each voter submits a strict ranking of all candidates. The election proceeds in rounds; in round k every candidate receives one point from each voter who ranks the candidate among his or her top k choices. The first round in which a candidate’s cumulative score exceeds half of the total number of voters determines the winner. This “cumulative majority” mechanism blends features of plurality, runoff, and rank‑order voting, and it is this structure that the authors exploit in their reductions.

For each of the 20 control cases examined (the remaining two cases are still open or known to be polynomial), the authors construct a polynomial‑time many‑one reduction from a classic NP‑complete problem—such as SAT, 3‑SAT, Exact‑Cover, or Subset‑Sum—to the corresponding Bucklin control decision problem. The reductions are carefully designed to respect the peculiarities of Bucklin’s scoring: candidate addition reductions encode a subset‑sum instance by arranging voter rankings so that a target candidate reaches the majority threshold only if a specific subset of “added” candidates is present. Deletion reductions embed Exact‑Cover by ensuring that removing a particular set of candidates forces the remaining ones to satisfy the cover constraints. Voter addition and deletion reductions manipulate the number of voters who rank the distinguished candidate within the critical round, mirroring the satisfaction of clauses in a SAT formula. Partition reductions are the most intricate; they split the election into two sub‑elections and require that the distinguished candidate wins in at least one sub‑election while the overall winner of the combined election is controlled. This forces the construction to simulate a two‑stage logical formula, again yielding NP‑completeness.

The paper’s main theorem states that Bucklin voting is resistant (i.e., the control problem is NP‑hard) to 20 of the 22 standard control actions. The two exceptions—constructive control by partition of voters with runoff and destructive control by partition of voters with runoff—remain either polynomially solvable or open, mirroring the status in the literature for other systems. Because Bucklin voting is a special case of fallback voting (fallback reduces to Bucklin when all voters approve every candidate), the resistance results automatically strengthen the known resistances for fallback voting.

In the discussion, the authors emphasize the practical implications: a voting system that is easy to compute winners for (Bucklin’s winner can be found in polynomial time) yet hard to manipulate via control offers a compelling balance between efficiency and security. They also note that the remaining open cases present fertile ground for future research, both in terms of pinpointing exact complexity and in exploring algorithmic heuristics that might succeed on realistic election sizes. Finally, the paper suggests that the methodology—systematic reductions tailored to the cumulative‑majority nature of Bucklin—could be adapted to analyze other rank‑based systems, potentially expanding the catalog of elections that are provably resistant to control.