Characterizations of voting rules based on majority margins

In the context of voting with ranked ballots, an important class of voting rules is the class of margin-based rules (also called pairwise rules). A voting rule is margin-based if whenever two elections generate the same head-to-head margins of victory or loss between candidates, the voting rule yields the same outcome in both elections. Although this is a mathematically natural invariance property to consider, whether it should be regarded as a normative axiom on voting rules is less clear. In this paper, we address this question for voting rules with any kind of output, whether a set of candidates, a ranking, a probability distribution, etc. We prove that a voting rule is margin-based if and only if it satisfies some axioms with clearer normative content. A key axiom is what we call Preferential Equality, stating that if two voters both rank a candidate $x$ immediately above a candidate $y$, then either voter switching to rank $y$ immediately above $x$ will have the same effect on the election outcome as if the other voter made the switch, so each voter’s preference for $y$ over $x$ is treated equally.

💡 Research Summary

This paper investigates the normative status of “margin‑based” voting rules—rules whose outcomes depend solely on the pairwise margins of victory or loss between candidates. The authors provide a series of characterizations that show a voting rule is margin‑based if and only if it satisfies a small set of intuitively appealing axioms, thereby giving a clear normative justification for the margin‑based property.

The core of the analysis begins with a formal definition: for any profile P and any two candidates x, y, let #_P(x, y) be the number of voters who rank x above y, and define the margin M_P(x, y) = #_P(x, y) – #_P(y, x). A rule F is margin‑based if whenever two profiles P and Q have identical margin matrices (M_P = M_Q), then F(P) = F(Q). While mathematically natural, the authors ask whether this invariance should be adopted as a fairness principle.

Two fundamental axioms are introduced for the domain of linear (complete) rankings:

1. Preferential Equality – If two voters both rank candidate x immediately above y, then swapping x and y in the ballot of either voter has exactly the same effect on the election outcome. In other words, each voter’s marginal preference for y over x is treated equally. This axiom captures a strong form of voter‑symmetry at the level of adjacent swaps.

2. Neutral Reversal – Adding a pair of completely reversed linear orders to any profile does not change the outcome. The intuition is that two voters with opposite preferences cancel each other out, leaving the collective result unchanged.

The authors prove (Theorem 2.9) that on the domain of linear profiles a voting rule satisfies Preferential Equality and Neutral Reversal if and only if it is margin‑based. This result gives a clear normative reading: any rule that respects equal treatment of adjacent swaps and the cancellation of opposite ballots must ignore any information beyond the pairwise margins.

The paper then shows that the stronger axiom of Homogeneity (Smith 1973) allows Neutral Reversal to be replaced by the weaker Block Invariance (Holliday & Pacuit 2025): the outcome must remain unchanged when exactly one copy of every possible linear order is added to the profile. This substitution makes the characterization more palatable to scholars who find Neutral Reversal controversial.

Extending beyond linear orders, the authors consider profiles that may contain ties. They introduce two additional axioms:

• Tiebreaking Compensation – If two profiles have the same margin matrix and the same total number of voters, then the rule must produce the same outcome. This captures the idea that the absolute size of the electorate should not affect a margin‑based decision.

• Neutral Indifference – When a candidate is tied with another in a voter’s ranking, swapping the tied positions does not affect the outcome. This ensures that the rule is indifferent to the precise placement of tied candidates.

With Preferential Equality strengthened to handle ties, together with Tiebreaking Compensation and Neutral Indifference, the authors obtain a full characterization of margin‑based rules on the unrestricted domain (Theorem 3.8).

Beyond margin‑based rules, the paper systematically maps the relationships among several related classes:

• Head‑to‑Head rules: depend on the full pairwise count function #_P.
• C2 rules: depend only on #_P, ignoring the total number of voters.
• C2w rules: depend only on the weak preference count #_wP (the number of voters not preferring y over x).
• Quasi‑margin‑based rules: depend on both the margin matrix and the total electorate size.

The inclusion diagram forms a diamond: Margin‑based ⊂ C2 ⊂ Head‑to‑Head and Margin‑based ⊂ C2w ⊂ Head‑to‑Head, with C2 and C2w incomparable. The authors give concrete examples: the weak Pareto rule is C2 but not margin‑based; the “winning‑votes” version of the Minimax rule is C2 but not margin‑based; the strict Pareto rule is C2w but not C2.

To illustrate the practical relevance, the paper presents empirical comparisons between the standard Minimax rule (margin‑based) and its winning‑votes variant (C2). Using four real‑world election databases, they report the frequency with which the two versions disagree when there is no Condorcet winner (e.g., 2521 cases with 1645 disagreements). Similar divergences are shown for other elections, highlighting that the choice of margin‑based versus non‑margin‑based specifications can materially affect outcomes.

A particularly striking case study involves Instant‑Runoff Voting (IRV), which violates Preferential Equality. The authors construct a profile where a 3 % group of Democratic voters swapping an adjacent pair does not change the winner, while an equally sized 3 % group of Republican voters swapping the same pair does change the winner. This demonstrates how IRV can treat identical marginal preferences asymmetrically, contrary to the fairness intuition encoded in Preferential Equality.

In conclusion, the paper provides a robust normative foundation for margin‑based voting rules by linking them to a small set of transparent axioms. It also clarifies the landscape of related rule families, offers concrete examples of violations, and supplies empirical evidence of the impact of these theoretical distinctions. The results are valuable for scholars designing voting systems, for political scientists interpreting election outcomes, and for policymakers seeking rules that respect voter symmetry and pairwise fairness.