Quantitative Halmos-Savage theorems and robust large financial markets

Christa Cuchiero; Georg K\"ostenberger; Irene Klein; Thorsten Schmidt

arxiv: 2504.06686 · v3 · pith:H3TGF4SLnew · submitted 2025-04-09 · 🧮 math.PR

Quantitative Halmos-Savage theorems and robust large financial markets

Christa Cuchiero , Irene Klein , Georg K\"ostenberger , Thorsten Schmidt This is my paper

Pith reviewed 2026-05-22 21:07 UTC · model grok-4.3

classification 🧮 math.PR

keywords Halmos-Savage theoremrobust FTAPKnightian uncertaintylarge financial marketsno asymptotic arbitrageNAA1NAA2one-period model

0 comments

The pith

A quantitative Halmos-Savage theorem for convex sets of probability measures enables robust no-arbitrage characterizations in large financial markets.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper establishes a quantitative version of the Halmos-Savage theorem and its dual for convex sets of probability measures that may not be dominated by a single measure. This generalizes earlier quantitative versions and is applied to derive robust versions of the fundamental theorem of asset pricing in large financial markets. The focus is on a one-period setting where the absence of asymptotic arbitrage of the first and second kind is characterized under Knightian uncertainty. Readers care because these results provide tools to handle model uncertainty in financial modeling of large markets.

Core claim

We establish a quantitative version of the classical Halmos-Savage Theorem for convex, potentially non-dominated sets of probability measures and its dual counterpart, generalizing previous quantitative versions. These results are then used to derive robust versions of the fundamental theorem of asset pricing (FTAP) in large financial markets in a one-period setting, characterizing the absence of arbitrage under Knightian uncertainty. To this end, we consider robust formulations of no asymptotic arbitrage of first kind (NAA1), which is the large market analogue of No unbounded profit with bounded risk (NUPBR), as well as no asymptotic arbitrage of second kind (NAA2). Finally, we characterize

What carries the argument

The quantitative Halmos-Savage theorem for convex sets of probability measures, which supplies a rate-controlled approximation result that transfers to robust separation conditions for asset pricing.

Load-bearing premise

The sets of probability measures under consideration are convex and the market model is restricted to a one-period setting.

What would settle it

A concrete counterexample consisting of a non-convex family of measures in which the quantitative bound on the Halmos-Savage distance fails to hold would refute the claimed generalization.

read the original abstract

We establish a quantitative version of the classical Halmos-Savage Theorem for convex, potentially non-dominated sets of probability measures and its dual counterpart, generalizing previous quantitative versions. These results are then used to derive robust versions of the fundamental theorem of asset pricing (FTAP) in large financial markets in a one-period setting, characterizing the absence of arbitrage under Knightian uncertainty. To this end, we consider robust formulations of no asymptotic arbitrage of first kind (NAA1), which is the large market analogue of ``No unbounded profit with bounded risk'' (NUPBR), as well as no asymptotic arbitrage of second kind (NAA2). Finally, we characterize asymptotic arbitrage of first and second kind in the robust one-period binomial model in terms of the model parameters.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

Extends quantitative Halmos-Savage to non-dominated convex sets and applies it to robust NAA1/NAA2 in one-period large markets under Knightian uncertainty.

read the letter

The main thing here is a quantitative Halmos-Savage result that works for convex families of measures without requiring a dominating probability, plus its use to characterize robust no-arbitrage in large one-period markets. They also spell out the conditions explicitly for the robust binomial model in terms of parameters. This generalizes earlier quantitative versions and moves them into the setting of Knightian uncertainty for asymptotic arbitrage of first and second kind. The dual statements are included as well. The paper does a clean job of stating the convexity assumption up front and showing how the classical theorem lifts to this case, then feeding the bounds directly into the FTAP-type results for NAA1 and NAA2. That linkage is the useful step for people working on robust large-market models. The one-period restriction keeps the technical load manageable while still addressing the large-market asymptotics. On the softer side, the one-period frame is a real limit for anyone who needs dynamic trading or continuous-time limits, even though the authors flag it clearly. Convexity is the natural setting for their duals, but it does leave out some non-convex ambiguity specifications that appear in other robust-finance papers. Without the full proofs in front of me I cannot check the error controls or measurability details in the non-dominated case, but nothing in the setup looks circular or self-referential. This is aimed at mathematical finance researchers who already know the classical Halmos-Savage and FTAP literature and want quantitative robust versions for large markets. It is solid enough on its own terms to deserve referee time rather than a desk rejection.

Referee Report

2 major / 2 minor

Summary. The paper establishes a quantitative version of the classical Halmos-Savage theorem for convex (possibly non-dominated) sets of probability measures, together with its dual counterpart. These results are applied to derive robust characterizations of no asymptotic arbitrage of the first kind (NAA1, the large-market analogue of NUPBR) and no asymptotic arbitrage of the second kind (NAA2) in one-period large financial markets under Knightian uncertainty, thereby obtaining robust versions of the fundamental theorem of asset pricing. The paper concludes with an explicit parameter-based characterization of asymptotic arbitrage in the robust one-period binomial model.

Significance. If the quantitative bounds and dual results hold with the stated error controls, the work supplies useful tools for robust no-arbitrage analysis in large markets. The extension to non-dominated convex families addresses a setting that arises naturally under Knightian uncertainty, and the explicit binomial-model characterization provides a concrete, falsifiable illustration of the general theory.

major comments (2)

[§3, Theorem 3.2] §3, Theorem 3.2: the quantitative bound on the Halmos-Savage constant appears to rely on a uniform integrability argument that is only sketched; a complete proof controlling the constant under non-domination would strengthen the central claim.
[§5.1, Definition 5.3 and Theorem 5.4] §5.1, Definition 5.3 and Theorem 5.4: the robust NAA1 condition is formulated via a supremum over the convex set; it is not immediately clear whether the quantitative Halmos-Savage result directly yields the stated equivalence without an additional uniform integrability hypothesis on the admissible strategies.

minor comments (2)

[§2 and §4] Notation for the convex set of measures is introduced in §2 but reused with slight variations in §4; a single consistent symbol would improve readability.
[§6] The binomial-model section would benefit from an explicit statement of the parameter ranges under which NAA1 fails.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the careful reading and constructive comments on our manuscript. We address each major comment below.

read point-by-point responses

Referee: [§3, Theorem 3.2] §3, Theorem 3.2: the quantitative bound on the Halmos-Savage constant appears to rely on a uniform integrability argument that is only sketched; a complete proof controlling the constant under non-domination would strengthen the central claim.

Authors: We appreciate the referee's observation. The argument in the proof of Theorem 3.2 does rely on uniform integrability to control the constant, and while the key steps are present, we agree that a more explicit and complete write-up would improve clarity, particularly in the non-dominated case. We will expand the proof in the revised version to include all details of the uniform integrability argument and the resulting bound. revision: yes
Referee: [§5.1, Definition 5.3 and Theorem 5.4] §5.1, Definition 5.3 and Theorem 5.4: the robust NAA1 condition is formulated via a supremum over the convex set; it is not immediately clear whether the quantitative Halmos-Savage result directly yields the stated equivalence without an additional uniform integrability hypothesis on the admissible strategies.

Authors: The robust NAA1 condition is defined via the supremum precisely to allow direct application of the quantitative Halmos-Savage theorem to the convex set of measures. The equivalence in Theorem 5.4 holds without an additional uniform integrability hypothesis on strategies because the NAA1 definition already encodes the relevant boundedness in the large-market setting. To improve clarity, we will insert a short remark after Definition 5.3 explaining this direct applicability. revision: yes

Circularity Check

0 steps flagged

No significant circularity

full rationale

The paper establishes a quantitative generalization of the classical Halmos-Savage theorem for convex (possibly non-dominated) families of measures, together with its dual, and applies the result to obtain robust characterizations of NAA1 and NAA2 in a one-period large-market setting under Knightian uncertainty. The derivation proceeds from standard measure-theoretic and duality arguments with explicitly stated hypotheses; no step reduces a claimed prediction or theorem to a fitted parameter, a self-referential definition, or a load-bearing self-citation whose content is itself unverified. The one-period restriction and convexity assumption are declared at the outset and serve as the natural setting for the robust FTAP extensions rather than being smuggled in. The work is therefore self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

Results rest on standard measure-theoretic probability and the explicit convexity assumption for sets of measures; no free parameters or new entities are indicated.

axioms (2)

standard math Standard axioms of probability and measure theory
Foundation for Halmos-Savage theorem and its quantitative extension.
domain assumption Convexity of the set of probability measures
Stated as the setting for the quantitative theorem in the abstract.

pith-pipeline@v0.9.0 · 5663 in / 1157 out tokens · 49100 ms · 2026-05-22T21:07:56.528300+00:00 · methodology

discussion (0)

Forward citations

Cited by 2 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

Can the $L^1$-$L^\infty$ duality be restored for non-dominated families of probability measures?
math.PR 2026-05 unverdicted novelty 7.0

A minimal P-complete extension of the sigma-algebra makes L^∞(P) the dual of the space of signed measures absolutely continuous w.r.t. at least one member of P.
Can the $L^1$-$L^\infty$ duality be restored for non-dominated families of probability measures?
math.PR 2026-05 unverdicted novelty 6.0

A canonical smallest P-complete extension of the sigma-algebra restores the isometric isomorphism between L∞(P) and the dual of signed measures absolutely continuous w.r.t. at least one measure in the non-dominated family P.