An exact frontier for rank comparisons
A mathematical preprint reports an exact ceiling on Spearman’s rho once Spearman’s footrule is fixed. The question is a trade-off between rank agreement and average absolute separation: for every admissible footrule value, how high can rho go? The paper gives a closed-form upper boundary and explicitly constructs a unique optimizer, or coupling, that reaches it.
That result describes more than the best possible score. The paper says the full attainable region consists of all rho values between sharp lower and upper boundaries at each allowable footrule. In other words, it identifies both the ceiling and the values that can actually occur below it within the stated mathematical class.
The proof turns ranking geometry into transport
The population here is not a survey or an observed ranking database. It is the class of couplings of U and V, each uniformly distributed on the interval from 0 to 1, with a prescribed mean absolute difference. A coupling is the joint pairing of the two variables; the fixed mean difference plays the role of the footrule in the optimization.
The optimization is recast as an optimal-transport problem with a linear moment constraint. In practical terms, the proof examines all permitted pairings while holding their mean separation fixed, then uses transport duality to certify that no permitted pairing can produce a higher value.
That certificate has two pieces: a feasible coupling and a feasible dual potential. The coupling’s support lies inside the potential’s contact set, establishing primal–dual equality. This turns an explicit candidate arrangement into an optimality result rather than merely a proposed curve.
Because the construction is explicit and unique, the theorem describes not only the maximum value but also the pairing that achieves it for each admissible footrule. Its certainty is conditional on the mathematical assumptions, rather than an estimate with a statistical error range.
A sharper account of distance and spread
Once the boundary is established, the paper derives a sharper form of the ordinary squared-mean lower bound. For couplings with uniform marginals, the squared displacement moment is at least the square of the mean plus a sharp correction, Vmin(m). The correction is nonnegative and equals zero exactly on the exceptional set C.
The study also gives an exact mean–variance picture for the absolute displacement |U−V|. At a prescribed mean absolute difference, its variance lies between Vmin(m) and Vmax(m); the lower envelope is attained by C_x and the upper envelope by B_x. These explicit copula families show that the two bounds are attainable extremes, not just loose inequalities.
The correction matters because it captures a restriction imposed by uniform marginals. The paper’s result says that the usual squared-mean bound can be improved precisely for this class of couplings, with equality confined to the exceptional cases identified by C.
From continuous pairings to finite rankings
The same results are carried into finite rankings by embedding permutations into uniform-marginal couplings. That yields refined lower and upper inequalities for normalized absolute rank displacements and normalized squared rank displacements. The continuous transport problem therefore supplies a common framework for finite permutation bounds.
Those finite bounds have continuum and asymptotic sharpness, but the supplied analysis cautions that they need not be optimal for every fixed ranking size. Limit-level exactness therefore does not settle the best possible inequality for each particular finite size.
A test for constant aggregation
Another application concerns generalized mixability, using a centered uniform construction and the aggregation rule Ψ(u,v)=|u+v|. In this setting, the paper characterizes the exact centers C at which constant aggregation is possible. At a prescribed mean m, the minimum variance is Vmin(m).
Put less formally, the result quantifies how close these paired values can come to producing a constant total. It is a precise statement for the centered-uniform construction and this particular aggregation rule, not a universal guarantee for every distribution or every way of combining values.
The remaining gap: two dependence measures
The paper then turns to Chatterjee’s rank correlation, ξ, and the copula correlation ratio, η. For continuous X and Y, it gives the outer bound max{0, ϱ_lower(ξ)} ≤ η ≤ min{ϱ_upper(ξ), 2ξ}. An outer bound is a fence: allowable pairs must stay within it, but being inside does not by itself show that every pair is attainable.
To show what is attainable, the authors construct an inner ξ–η enclosure with explicit models. It covers approximately 81.4% of the area of the outer enclosure, according to the paper’s reported comparison. The exact ξ–η region remains unresolved, and the sharpness of the constructive inner curves ηℓ and ηu has not been established.
That gap matters to the interpretation. The upper ξ–η bound is explicitly non-sharp under the conditional-i.i.d. restriction, and sharpness over all copulas does not automatically carry over to that narrower class. The paper therefore offers a useful outer fence and an attainable inner region, but not a final map.
The approximately 81.4% area figure should also be read in context. The supplied analysis notes that the paper mentions numerical validation and area evaluation but does not provide a reproducible dataset or protocol, so the comparison is part of the reported mathematical analysis rather than an independently replicated measurement.
What the result can—and cannot—say
Everything here is a mathematical conclusion within stated assumptions. The work analyzes couplings, finite permutations and constructed conditional-i.i.d. models; it does not estimate effects from human or other empirical observations, report confidence intervals or statistical tests, or establish causation. The central certainty is deductive: for the model class, the boundaries and optimizers are exact.
The supplied document is arXiv version 1, dated 20 August 2026. Its unresolved questions are concrete: the exact ξ–η region, whether either inner curve is sharp, whether these mathematical bounds can become reliable finite-sample estimators, and how finite-ranking inequalities can be made optimal for each fixed size.
For readers interested in rank correlations or dependence measures, the contribution is a set of exact theoretical frontiers with several linked applications. Its reach is correspondingly clear: it strengthens what can be said about idealized pairings and finite-ranking inequalities while stopping short of empirical performance claims.
Paper data and sources
Original title: The exact Spearman rho-footrule region via optimal transport with applications to finite rankings, mixability, and Chatterjee's rank correlation
Authors: Jonathan Ansari, Marcus Rockel
Journal/Repository: arXiv
Status: Preprint, not yet peer-reviewed
First online: 2026-08-20
DOI: Not available
Original paper · Full text