Quantifying the degree of risk aversion of spectral risk measures
Pith reviewed 2026-05-23 22:27 UTC · model grok-4.3
The pith
A functional on spectral risk measures quantifies their degree of risk aversion via two axioms.
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
Core claim
The author constructs a functional on the space of spectral risk measures that quantifies their degree of risk aversion. The functional is obtained by imposing normalization on the subspace of CVaRs together with linearity. This setup formalizes direct comparisons of risk aversion across different spectral risk measures and yields two computable formulas along with associated properties.
What carries the argument
The functional on spectral risk measures defined by the normalization axiom on CVaRs and the linearity axiom.
If this is right
- Spectral risk measures receive a single comparable number for their risk aversion.
- Linearity implies that convex combinations of risk measures receive the corresponding weighted average aversion value.
- Normalization fixes the value of the functional exactly on all CVaR measures.
- The two formulas allow direct computation of the aversion degree for any given spectral risk measure.
- Properties such as monotonicity with respect to the weighting function follow from the axioms.
Where Pith is reading between the lines
- Portfolio optimization routines could incorporate a target value of the functional to control overall conservatism.
- The same axiomatic approach might extend to define aversion degrees for non-spectral risk measures.
- Empirical studies could check whether the ordering induced by the functional aligns with observed market risk premia.
Load-bearing premise
The normalization on CVaRs and the linearity axiom together correctly capture the intuitive notion of degree of risk aversion.
What would settle it
Finding two spectral risk measures such that the functional assigns a lower aversion value to the one that places greater weight on tail losses than the other.
read the original abstract
I propose a functional on the space of spectral risk measures that quantifies their ``degree of risk aversion''. This quantification formalizes the idea that some risk measures are ``more risk-averse'' than others. I construct the functional using two axioms: a normalization on the space of CVaRs and a linearity axiom. I present two formulas for the functional and discuss several properties and interpretations.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The manuscript proposes a functional on the space of spectral risk measures that quantifies their degree of risk aversion. The functional is constructed via two axioms—a normalization condition on the family of CVaR measures and a linearity axiom on the vector space spanned by spectral risk measures—and two explicit formulas for the functional are derived, along with a discussion of properties and interpretations.
Significance. If the two axioms are accepted as correctly capturing the intended ordering, the construction supplies a parameter-free, axiomatically grounded ranking of spectral risk measures by risk aversion. This could be useful for comparing tail-sensitive risk measures in portfolio selection and regulatory applications, and the axiomatic approach itself is a strength that makes the proposal reproducible and falsifiable in principle.
major comments (2)
- [Axiomatic construction (linearity axiom)] The linearity axiom is used to extend the normalization from CVaRs to the full space of spectral risk measures, but the manuscript provides no verification that the resulting functional respects the natural partial order on spectra: if ϕ₁(t) ≥ ϕ₂(t) for all t then the assigned degree satisfies f(ρ₁) ≥ f(ρ₂). Without this monotonicity property the functional may fail to align with the intuitive notion that greater tail weight corresponds to strictly higher risk aversion.
- [Properties and interpretations] The central claim that the functional 'quantifies the degree of risk aversion' rests entirely on the appropriateness of the two axioms; the manuscript does not supply an independent check (e.g., comparison against known orderings of specific spectral measures beyond CVaR or against empirical risk-aversion rankings) that would confirm the axioms produce the intended comparisons.
minor comments (2)
- [Abstract] The abstract states that two formulas are presented but does not indicate their form; a one-sentence description of each formula would improve readability for readers who do not reach the main text.
- [Notation and definitions] Notation for the spectrum function ϕ and the functional f should be introduced consistently in the first section where they appear, rather than relying on the reader to infer the mapping from the CVaR normalization.
Simulated Author's Rebuttal
We thank the referee for the constructive comments. We address the two major comments point by point below and indicate the revisions we will make.
read point-by-point responses
-
Referee: [Axiomatic construction (linearity axiom)] The linearity axiom is used to extend the normalization from CVaRs to the full space of spectral risk measures, but the manuscript provides no verification that the resulting functional respects the natural partial order on spectra: if ϕ₁(t) ≥ ϕ₂(t) for all t then the assigned degree satisfies f(ρ₁) ≥ f(ρ₂). Without this monotonicity property the functional may fail to align with the intuitive notion that greater tail weight corresponds to strictly higher risk aversion.
Authors: We agree that monotonicity with respect to the pointwise order on spectra is a natural requirement. The functional is constructed to be linear on the vector space spanned by spectral risk measures and normalized on the CVaR family; because the normalization assigns higher values to CVaRs with higher confidence levels (which correspond to spectra that are larger in the pointwise order) and linearity preserves inequalities, the resulting functional is monotone. We will add an explicit proposition and short proof of this monotonicity property in the revised manuscript. revision: yes
-
Referee: [Properties and interpretations] The central claim that the functional 'quantifies the degree of risk aversion' rests entirely on the appropriateness of the two axioms; the manuscript does not supply an independent check (e.g., comparison against known orderings of specific spectral measures beyond CVaR or against empirical risk-aversion rankings) that would confirm the axioms produce the intended comparisons.
Authors: The primary justification remains the two axioms, which are chosen to capture the intended ordering. Nevertheless, we accept that concrete illustrations strengthen the interpretation. In the revision we will add a short subsection comparing the functional values on standard families (e.g., the Wang transform and the proportional-hazards transform) and on convex combinations of CVaRs, confirming that the ordering aligns with the intuitive notion of increasing tail sensitivity. revision: yes
Circularity Check
Axiomatic definition with no reduction to inputs or self-citations
full rationale
The paper explicitly constructs the functional via two stated axioms (normalization on the CVaR family and linearity on the vector space of spectral risk measures) rather than deriving it from data, fitted parameters, or prior results. No equations or claims reduce by construction to their own inputs, no self-citations are invoked as load-bearing uniqueness theorems, and the presentation is a definition rather than a prediction or renaming of an empirical pattern. The derivation chain is therefore self-contained against external benchmarks.
Axiom & Free-Parameter Ledger
axioms (2)
- domain assumption Normalization on the space of CVaRs
- domain assumption Linearity axiom
Reference graph
Works this paper leans on
-
[1]
Spectral measures of risk: A coherent representation of subje c- tive risk aversion
Acerbi, C. Spectral measures of risk: A coherent representation of subje c- tive risk aversion. Journal of Banking & Finance 26, 7 (2002), 1505–1518
work page 2002
-
[2]
Why the 1-wasserstein distance is the area between the two marginal cdfs
De Angelis, M., and Gray, A. Why the 1-wasserstein distance is the area between the two marginal cdfs. arXiv preprint arXiv:2111.03570 (2021)
-
[3]
On law invariant coherent risk measures
Kusuoka, S. On law invariant coherent risk measures. In Advances in Mathematical Economics, S. Kusuoka and T. Maruyama, Eds., vol. 3. Springer, Tokyo, 2001, pp. 83–95
work page 2001
-
[4]
Rockafellar, R. T., and Uryasev, S. Conditional value-at-risk for general loss distributions. Journal of banking & finance 26, 7 (2002), 1443– 1471
work page 2002
-
[5]
Advances in risk-averse optimization
Ruszczy´nski, A. Advances in risk-averse optimization. In Theory Driven by Influential Applications. INFORMS, 2013, pp. 168–190
work page 2013
-
[6]
On Kusuoka representation of law invariant risk measures
Shapiro, A. On Kusuoka representation of law invariant risk measures. Mathematics of Operations Research 38, 1 (2013), 142–152
work page 2013
-
[7]
Yaari, M. E. The dual theory of choice under risk. Econometrica: Journal of the Econometric Society (1987), 95–115. 9
work page 1987
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.