[Paper Review] Fraction-like rates from preferential voting

Rosa Camps, Xavier Mora+1 more|arXiv (Cornell University)|Jan 22, 2010

Game Theory and Voting Systems17 references3 citations

TL;DR

This paper proposes a method to derive continuous, fraction-like mixing proportions from preferential voting data using a combination of Zermelo's strength method and a novel CLC (Condorcet-Llull-Campbell) projection. The approach ensures continuity, satisfies single-choice voting consistency and unanimous decomposition, and aims to meet the Condorcet-Smith majority principle, with empirical results showing improved compliance over standard methods like fair bets or Zermelo’s method alone.

ABSTRACT

A method is given for determining a mixed social choice out of a paired-comparison matrix. The method combines a projection procedure introduced in previous papers of the same authors and a classical method due to Zermelo. The resulting method is proved to have certain desirable properties, which include: compliance with a majority principle, clone consistency, and continuity of the mixing fractions with respect to the data.

Motivation & Objective

• To develop a continuous, well-defined method for deriving mixing fractions from preference matrices in collective decision-making scenarios such as budget or prize distribution.
• To ensure the method satisfies key axiomatic properties: continuity in preference scores, single-choice voting consistency, and unanimous decomposition.
• To investigate whether combining Zermelo’s method with a CLC projection can satisfy the Condorcet-Smith majority principle while preserving desirable behavioral properties.
• To address limitations of existing methods—such as fair bets and Zermelo’s method—regarding continuity near reducible matrices and compliance with majority principles.

Proposed method

• The method begins with Zermelo’s strength method to compute initial preference-based ratings from the Llull matrix of paired comparisons.
• It applies a CLC (Condorcet-Llull-Campbell) projection to transform the initial ratings into a form that respects Condorcet-Smith majority consistency.
• The resulting scores are normalized to produce mixing fractions that sum to one, representing proportional shares.
• The approach ensures continuity of the output fractions with respect to small changes in preference scores, even near reducible matrices.
• It leverages the fair bets method as a benchmark for continuity and behavioral properties, using it as a reference point for comparison.
• The method is designed to satisfy quantitative monotonicity, ensuring that raising an option’s rank increases its assigned fraction.

Experimental results

Research questions

• RQ1 Is the fair bets method continuous in the preference scores, even in neighborhoods of reducible matrices?
• RQ2 Does the combination of the CLC projection with the fair bets method satisfy the Condorcet-Smith principle?
• RQ3 Do the fair bets maintain the inversion property when preceded by the CLC projection?
• RQ4 Can the CLC projection restore or preserve quantitative monotonicity when applied to Zermelo’s method?
• RQ5 Does the combined method produce more consistent and fair mixing fractions than Zermelo’s method or fair bets alone?

Key findings

• For the example with 18 voters and full rankings, the proposed method yields mixing fractions of a: 0.325, b: 0.286, c: 0.214, d: 0.175, which better reflect Condorcet preferences than fair bets (a: 0.323, b: 0.378, c: 0.174, d: 0.124).
• The fair bets method is found to be continuous near reducible matrices, as demonstrated by the limit behavior of (1−ε, ε, ε)T as ε↓0.
• The method satisfies single-choice voting consistency: when voters choose only one option, the mixing fractions match the vote shares exactly.
• The CLC projection improves the behavior of Zermelo’s method by restoring continuity and potentially enabling compliance with the Condorcet-Smith principle.
• Numerical results suggest that the fair bets method may satisfy the inversion property after CLC projection, though this remains an open question.
• The method avoids the failure of quantitative monotonicity seen in standard Zermelo’s method when combined with CLC projection, suggesting improved behavioral consistency.