Geometric construction of voting methods that protect voters first choices

We consider the possibility of designing an election method that eliminates the incentives for a voter to rank any other candidate equal to or ahead of his or her sincere favorite. We refer to these methods as satisfying the Strong Favorite Betrayal Criterion" (SFBC). Methods satisfying our strategic criteria can be classified into four categories, according to their geometrical properties. We prove that two categories of methods are highly restricted and closely related to positional methods (point systems) that give equal points to a voter's first and second choices. The third category is tightly restricted, but if criteria are relaxed slightly a variety of interesting methods can be identified. Finally, we show that methods in the fourth category are largely irrelevant to public elections. Interestingly, most of these methods for satisfying the SFBC do so only weakly," in that these methods make no meaningful distinction between the first and second place on the ballot. However, when we relax our conditions and allow (but do not require) equal rankings for first place, a wider range of voting methods are possible, and these methods do indeed make meaningful distinctions between first and second place.

💡 Research Summary

The paper investigates whether an election method can be designed that removes any incentive for a voter to rank any other candidate equal to or ahead of his or her sincere favorite. This stronger version of the Favorite Betrayal Criterion is called the Strong Favorite Betrayal Criterion (SFBC). While the ordinary FBC only forbids a voter from gaining a more preferred outcome by placing another candidate ahead of his favorite, it still allows a voter to gain an advantage by ranking another candidate equal to the favorite. SFBC eliminates even that possibility: no situation may exist where a voter can obtain a more preferred result by ranking another candidate equal to or ahead of his true favorite.

To analyze SFBC the author adopts a geometric framework closely related to Saari’s simplex representation of electorates. Each possible ballot type (preference order) is a coordinate in a d‑dimensional space; the vector p whose components p_k give the fraction of voters casting each ballot lies in the unit simplex S(d). Under the linearity assumption, the condition that a particular candidate i wins can be expressed as a set of linear inequalities ∑k u{ijk} p_k > 0. Each inequality defines a hyperplane (a boundary) separating the simplex into regions that correspond to different winners. The normal vector of a boundary is precisely the coefficient vector u_{ijk}.

SFBC imposes a geometric restriction on these boundaries. If a voter swaps his favorite from first to second place (or makes it equal), the sign of any inequality that determines the winner must not change. In other words, the normal vectors must be invariant under the permutation that exchanges the first‑ and second‑rank positions for any candidate. This requirement dramatically limits the admissible normal vectors and therefore the admissible voting rules.

The analysis yields four distinct categories of SFBC‑compliant methods:

1. Type 1 (single‑condition boundaries) – Each boundary is defined by a single linear inequality. This class coincides with positional (point‑system) methods in which each candidate receives a fixed number of points for each rank. To satisfy SFBC the point values for first and second place must be identical; antiplurality (one point for every non‑last rank) is the canonical example. The normal vectors are simple differences w_i – w_j, guaranteeing invariance under a first‑second swap.

2. Type 1b (mixed boundaries) – Some boundaries satisfy multiple linear conditions simultaneously. These can be interpreted as hybrids of positional scoring and runoff stages, similar to the Bucklin method. As long as the combined normal vectors retain the required symmetry, the method remains SFBC‑compliant.

3. Type 2 (multi‑condition boundaries for all candidates) – Every boundary satisfies more than one linear condition. This class is heavily restricted; when the SFBC restriction is relaxed to the ordinary FBC, it encompasses a variety of recently proposed systems such as range voting and majority‑defeat disqualification approval. Nevertheless, preserving a strict distinction between first and second place generally demands additional non‑linear rules.

4. Type 3 (universal multi‑condition boundaries) – Every possible linear condition is satisfied simultaneously. Such methods make no meaningful distinction among candidates and are therefore irrelevant for public elections.

The paper shows that the first two categories essentially force the election rule to treat the first and second ranks identically; they satisfy SFBC only “weakly” because the ballot’s first‑place designation has no effect on the outcome. The third category can produce more interesting methods, but only if the SFBC requirement is softened. The fourth category yields degenerate systems with no practical use.

An illustrative example is antiplurality voting, which gives each candidate one point for every ballot that does not rank the candidate last. This method meets SFBC because a voter never gains by moving a rival ahead of or equal to his favorite; however, the first‑place slot is purely ceremonial. The author argues that most realistic SFBC‑compliant systems will share this weakness. To obtain a voting rule that meaningfully distinguishes first from second place while still protecting the favorite, one must accept the weaker FBC or introduce additional procedural layers (runoffs, thresholds, etc.).

In conclusion, the geometric analysis demonstrates that strict adherence to SFBC forces voting methods into a narrow set of highly constrained forms, most of which collapse the distinction between first and second preferences. Consequently, for practical public elections, designers are likely to settle for “weak” protection of the favorite or combine SFBC‑compatible scoring with auxiliary mechanisms to mitigate strategic incentives.