REVIEW 5 minor 15 references
Positional Voting and Doubly Stochastic Matrices
T0 review · 0 major / 5 minor · reviewed 2026-08-14 · deepseek-v4-flash
Pith's one-line read For any fixed profile, the rankings achievable by changing the positional voting weights are exactly those whose corresponding faces intersect the convex hull of the cumulative column sums of the profile matrix.
desk verdict A modest but correct paper that re-proves known voting bounds cleanly and adds a genuinely useful convex-hull characterization; deserves a serious referee. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
The load-bearing device is the identity $T_w p = Q_p w$: the result of a positional rule with weights $w$ on profile $p$ is just the product of the profile matrix $Q_p$ (whose rows and columns sum to the number of ballots) with $w$. Since $Q_p$ has constant row and column sums, shifting and rescaling makes it doubly stochastic, and the classical decomposition theorem for doubly stochastic matrices expresses it as a convex combination of permutation matrices—i.e., as a genuine (possibly fractional) ballot profile. For the geometric characterization, the cone $W$ of strict weighting vectors is generated by $n-1$ explicit vectors $v_k$, so the achievable results are exactly the conical combinations of $s_k = Q_p v_k$; discarding the constant direction leaves the convex hull of the cumulative column sums $t_1,\dots,t_{n-1}$. Faces of the braid arrangement encode rankings (with ties as lower-dimensional faces), so 'which face intersects the hull' is the complete answer to which rankings are possible.
What would settle it
Take a specific small profile (say n=4 with a handful of integer ballots), compute the vectors $t_1,t_2,t_3$, and exhaustively enumerate all faces of the braid arrangement that intersect their convex hull; then separately enumerate all rankings obtained by sampling or rationally searching every weighting vector in $W$. If any ranking appears in one list but not the other, Theorem 4.2 is false. Because both lists are finite and computable for small n, this is a direct computational check.
Extended reading notes
Core claim
For a fixed profile $p$, with profile matrix $Q_p = [q_1 \cdots q_n]$ whose $(i,j)$ entry counts voters ranking candidate $i$ in position $j$, the paper's Theorem 4.2 shows that a ranking (or partial ranking with ties) is achievable by some positional voting weighting vector exactly when the corresponding face of the braid arrangement intersects the convex hull of the vectors $t_k = q_1 + \cdots + q_k$ for $k=1,\dots,n-1$. Because adding a constant to all candidates' totals does not change the ranking, this hull is studied in the sum-zero hyperplane, and every point in it can be written as a conical combination of the $t_k$. The coefficients of that combination translate, through the cone generators of the weighting space, into explicit weights that realize the desired outcome. The paper also proves that arbitrary prescribed results vectors can be realized by infinitely many profiles for up to $n-1$ linearly independent weighting rules (Theorem 3.3), that at most $n! - (n-1)!$ strict rankings are possible from any profile (Theorem 3.9), and that there exist profiles achieving the maximum (Theorem 3.6).
Load-bearing premise
The load-bearing premise is that a constructed 'profile' may have negative or fractional ballot counts; the paper's conversion to genuine nonnegative-integer ballots is proved only when the goal is an ordinal ranking, so cardinal score outcomes may not be realizable with real voters.
Editorial extensions
If this is right
- For any actual ballot profile, no ranking can be forced by any positional weighting vector unless its face intersects the convex hull of the cumulative column sums $t_1,\dots,t_{n-1}$.
- A desired ranking can be reverse-engineered: pick a point in that hull belonging to the ranking's face, write it as a conical combination of the $t_k$, and the coefficients directly give a weighting vector that realizes it.
- At most $n! - (n-1)!$ strict rankings are possible from a single profile, and profiles attaining this maximum exist; both statements now follow from a half-space argument on the braid arrangement.
- Paradoxical profiles—in which different positional rules yield very different winners—can be constructed explicitly by choosing a matrix with constant row and column sums and expanding it into ballots via the doubly stochastic decomposition theorem.
- When only ordinal rankings matter, any profile can be replaced by a nonnegative integer ballot profile without changing the set of achievable rankings, so the geometric characterization applies to real elections.
Reading between the lines
- The convex-hull criterion gives an audit tool: from a published profile one can precompute every ranking an election official could induce by choosing weights, so attempted manipulation becomes detectable before the election.
- Because the conical coefficients in the decomposition are generally not unique, the same outcome can usually be produced by multiple weighting vectors; this raises a robustness question the paper leaves open—which of those weights is least sensitive to small changes in the profile.
- The same linear-algebra mechanism may extend beyond single-winner positional rules to other aggregation schemes whose tally is a linear map on a ballot-count matrix, such as multiwinner scoring or committee elections.
- The half-space proof of the upper bound only needs the weighting space to be a convex cone inside the sum-zero hyperplane, so the counting method could, in principle, be adapted to restricted weight families (e.g., integer weights or weights with prescribed ties) to give sharper bounds.
Signed reviews
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper revisits positional voting systems through the lens of doubly stochastic matrices. The main results are: Theorem 3.3, a linear-algebra proof of the Daugherty–Eustis–Minton–Orrison theorem that any n−1 linearly independent weighting vectors and any desired result vectors can be realized by some (possibly fractional or negative) profile; Theorem 3.6, a construction of profiles achieving the maximal n!−(n−1)! strict societal rankings; Theorem 3.9, an elementary proof of Saari's upper bound on the number of achievable strict rankings; and Theorem 4.2, a geometric characterization of the possible outcomes from a given profile in terms of the convex hull of the partial column sums of Q_p, together with an algorithm for choosing weights to realize a desired outcome. The arguments are largely self-contained and rely on the Birkhoff–von Neumann theorem, basic convex geometry, and the braid arrangement.
Significance. If the results hold, the paper is a valuable expository and methodological contribution: it gives transparent, constructive re-proofs of known results in algebraic voting theory and offers a new geometric description of achievable outcomes. The paper is honest about its limitations: the constructed profiles in Theorem 3.3 may have negative entries, and nonnegative integer profiles are guaranteed only for ordinal outcomes, not for cardinal score outcomes. The proofs are detailed and checkable, and Theorem 4.2 provides a concrete, worked algorithm for weight selection. These strengths make the paper suitable for publication after a minor revision.
minor comments (5)
- [Section 4, Theorem 4.2] The statement should clarify whether the zero weighting vector is admitted. If the closure of W is meant to include the all-zero weight, then the all-tie outcome 0 is possible for every profile, but 0 does not necessarily lie in the convex hull of t_1,...,t_{n-1}; for example, with n=3 and a single voter ranking 1≻2≻3, the convex hull is the segment from (1,0,0) to (1,1,0) and does not contain 0. Please either explicitly exclude the zero weighting vector or add {0} to the convex hull in the statement.
- [Sections 2 and 4] The notation W is used inconsistently: Section 2 defines W as the open cone of strictly decreasing sum-zero vectors, while Proposition 4.1 describes the closed cone of nonincreasing vectors. The closure should be denoted consistently (e.g., \overline{W}) throughout, especially in Proposition 4.1 and Theorem 4.2.
- [Abstract and Section 3, Theorem 3.3] The abstract says the paper gives a 'simple and explicit construction of paradoxical profiles' without noting that the constructed profiles may contain negative and fractional entries. The body is transparent about this (see the remark after Theorem 3.3 and Proposition 3.8), but the abstract should carry the same qualifier, since the cardinal constructions are not literal ballot counts.
- [Section 3, Theorem 3.9] There is a typo in the paragraph before the proof: 'oberve' should be 'observe'.
- [Section 4, Example 4.3] The randomized sampling procedure is described as likely to produce a 'nearly exhaustive' list of possible rankings; since the theorem gives exact geometric conditions, the authors could note that a deterministic enumeration of the chambers intersecting the convex hull is also possible, though the sampling heuristic is fine for practical illustration.
Circularity Check
No significant circularity: the paper's derivations are self-contained and reproduce prior results rather than assuming them.
full rationale
The paper's central claims—Theorem 3.3 (constructing profiles realizing arbitrary results vectors for independent weights), Theorems 3.6 and 3.9 (Saari's bound on possible strict rankings), and Theorem 4.2 (convex-hull characterization of possible outcomes)—are proved from the definitions of positional voting, the Birkhoff–von Neumann theorem, and elementary linear algebra. The only external results invoked are standard mathematical tools (Birkhoff–von Neumann, supporting hyperplane theorem), and they are used as tools rather than as the conclusions being derived. The proof of Theorem 4.2 is a direct computation: Proposition 4.1 expresses any nonstrict weighting vector as a conical combination of the vectors v_1,...,v_{n-1}, so Q_p applied to the closure of W is the conical hull of the vectors s_k = Q_p v_k; since s_{n-k} = t_k - (Nk/n)1, every nonzero conical combination differs from a convex combination of the t_k by a multiple of the all-ones vector, which does not change the braid face, and every convex combination of t_k arises from a nonzero weight vector w = sum b_i v_{n-i}. There is no fitted parameter renamed as a prediction, and the paper does not rely on self-citations for load-bearing premises: citations to Saari and to Daugherty et al. are for results that the paper re-proves or extends rather than assumes. The stated limitation that Theorem 3.3 constructs profiles with possibly negative entries, with Proposition 3.8 restoring nonnegative integer profiles only for ordinal outcomes, is a scope caveat rather than a circularity. Accordingly, no circular step is present.
Assumptions & free parameters
assumptions (3)
- standard math Birkhoff-von Neumann theorem: every doubly stochastic matrix is a convex combination of permutation matrices
- standard math Supporting hyperplane theorem for convex sets in finite-dimensional vector spaces
- standard math Dimension of the space of n x n matrices with constant row and column sums is (n-1)^2+1
Cite this review
Pith. "Pith review of Positional Voting and Doubly Stochastic Matrices." pith.science (2026). https://pith.science/paper/R5YJLA6C
@misc{pith2026190806506,
author = {Pith},
title = {Pith review of: Positional Voting and Doubly Stochastic Matrices},
year = {2026},
howpublished = {\url{https://pith.science/paper/R5YJLA6C}},
note = {Machine review of arXiv:1908.06506}
}
read the original abstract
We provide elementary proofs of several results concerning the possible outcomes arising from a fixed profile within the class of positional voting systems. Our arguments enable a simple and explicit construction of paradoxical profiles, and we also demonstrate how to choose weights that realize desirable results from a given profile. The analysis ultimately boils down to thinking about positional voting systems in terms of doubly stochastic matrices.
Reference graph
Works this paper leans on
-
[1]
Topics in Hyperplane Arrangements, volume 226 of Mathemat- ical Surveys and Monographs
Marcelo Aguiar and Swapneel Mahajan. Topics in Hyperplane Arrangements, volume 226 of Mathemat- ical Surveys and Monographs . American Mathematical Society, Providence, RI, 2017
work page 2017
-
[2]
A new way to aggregate preferences: Application to Eurovision song contests
J´ er´ emy Besson and Celine Robardet. A new way to aggregate preferences: Application to Eurovision song contests. In Advances in Intelligent Data Analysis VII, 7th International Symposium on Intelligent Data Analysis, volume 4723, pages 152–162, 2007
work page 2007
-
[3]
Three observations on linear algebra
Garrett Birkhoff. Three observations on linear algebra. Univ. Nac. Tucum´ an. Revista A., 5:147–151, 1946
work page 1946
-
[4]
Richard A. Brualdi. Notes on the Birkhoff algorithm for doubly stochastic matrices. Canad. Math. Bull., 25(2):191–199, 1982
work page 1982
-
[5]
Karl-Dieter Crisman and Michael E. Orrison. Representation theory of the symmetric group in voting theory and game theory. In Algebraic and Geometric Methods in Discrete Mathematics , volume 685 of Contemp. Math., pages 97–115. Amer. Math. Soc., Providence, RI, 2017
work page 2017
-
[6]
Eustis, Gregory Minton, and Michael E
Zajj Daugherty, Alexander K. Eustis, Gregory Minton, and Michael E. Orrison. Voting, the symmetric group, and representation theory. Amer. Math. Monthly , 116(8):667–687, 2009
work page 2009
-
[7]
Notes on Birkhoff–von Neumann decomposition of doubly stochastic matrices
Fanny Dufoss´ e and Bora U¸ car. Notes on Birkhoff–von Neumann decomposition of doubly stochastic matrices. Linear Algebra Appl., 497:108–115, 2016
work page 2016
-
[8]
The Borda count and its real-world alternatives: Comparing scoring rules in Nauru and Slovenia
Jon Fraenkel and Bernard Grofman. The Borda count and its real-world alternatives: Comparing scoring rules in Nauru and Slovenia. Aust. J. Political Sci. , 49, 2014
work page 2014
Show all 15 references
-
[9]
Hodge and Richard E
Jonathan K. Hodge and Richard E. Klima. The Mathematics of Voting and Elections: A Hands-On Approach, volume 30 of Mathematical World. American Mathematical Society, Providence, RI, 2018
2018
-
[10]
Linear algebra
Serge Lang. Linear algebra. Undergraduate Texts in Mathematics. Springer-Verlag, New York, third edition, 1989
1989
-
[11]
Donald G. Saari. Millions of election outcomes from a single profile. Soc. Choice Welf. , 9(4):277–306, 1992
1992
-
[12]
Donald G. Saari. Basic Geometry of Voting . Springer-Verlag, Berlin, 1995
1995
-
[13]
W. A. Stein et al. Sage Mathematics Software (Version 8.8) . The Sage Development Team, 2019. www.sagemath.org
2019
-
[14]
Chambers of arrangements of hyperplanes and Arrow’s impossibility theorem
Hiroaki Terao. Chambers of arrangements of hyperplanes and Arrow’s impossibility theorem. Adv. Math., 214(1):366–378, 2007
2007
-
[15]
A certain zero-sum two-person game equivalent to the optimal assignment problem
John von Neumann. A certain zero-sum two-person game equivalent to the optimal assignment problem. In Contributions to the Theory of Games, Vol. 2 , Annals of Mathematics Studies, no. 28, pages 5–12. Princeton University Press, Princeton, N. J., 1953. Jacqueline Anderson, Depa...
1953
Reviewed August 14, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.