Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Theorem 6, p. 31 — comonotonic inf-convolution of Wasserstein-robust VaRs, (36)–(37), optimal allocation (X, 0, …, 0)

Disproved
TailRiskSharing.RobustVaR.theorem6_robust_var_comonotonic

by mikedeng1 · Oct 9, 2026 · Mathlib 0df444a (Lean v4.33.1)

comonotonicityinf-convolutionoptimal-allocationp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1risk-sharingvalue-at-riskwasserstein-distance

Let (Ω,F,P)(\Omega,\mathcal F,\mathbb P)(Ω,F,P) be an atomless probability space, X∈L∞X\in L^\inftyX∈L∞ and Λ∈{L,R}\Lambda\in\{L,R\}Λ∈{L,R}. For a risk measure ρ\rhoρ and δ>0\delta>0δ>0, [ρ]δ1(X)=sup⁡{ρ(Y):Y∈L∞, W1(Y,X)≤δ}[\rho]^1_\delta(X)=\sup\{\rho(Y):Y\in L^\infty,\ W_1(Y,X)\le\delta\}[ρ]δ1​(X)=sup{ρ(Y):Y∈L∞, W1​(Y,X)≤δ} is its robust version over the order-111 Wasserstein ball, and ⊞\boxplus⊞ is the inf-convolution over comonotonic allocations An+(X)\mathbb A^+_n(X)An+​(X). Consider n≥1n\ge1n≥1 agents.

  1. If 0<δ1≤⋯≤δn0<\delta_1\le\dots\le\delta_n0<δ1​≤⋯≤δn​ and α∈(0,1)\alpha\in(0,1)α∈(0,1), then
⊞i=1n[VaRαΛ]δi1(X)=[VaRαΛ]δ11(X)+∑i=2nδiα.\mathop{\boxplus}_{i=1}^n[\mathrm{VaR}^\Lambda_\alpha]^1_{\delta_i}(X)=[\mathrm{VaR}^\Lambda_\alpha]^1_{\delta_1}(X)+\sum_{i=2}^n\frac{\delta_i}{\alpha}.⊞i=1n​[VaRαΛ​]δi​1​(X)=[VaRαΛ​]δ1​1​(X)+i=2∑n​αδi​​.
  1. If 1>α1≥⋯≥αn>01>\alpha_1\ge\dots\ge\alpha_n>01>α1​≥⋯≥αn​>0 and δ>0\delta>0δ>0, then
⊞i=1n[VaRαiΛ]δ1(X)=[VaRα1Λ]δ1(X)+∑i=2nδαi.\mathop{\boxplus}_{i=1}^n[\mathrm{VaR}^\Lambda_{\alpha_i}]^1_{\delta}(X)=[\mathrm{VaR}^\Lambda_{\alpha_1}]^1_{\delta}(X)+\sum_{i=2}^n\frac{\delta}{\alpha_i}.⊞i=1n​[VaRαi​Λ​]δ1​(X)=[VaRα1​Λ​]δ1​(X)+i=2∑n​αi​δ​.
  1. In either setting, the allocation (X,0,…,0)(X,0,\dots,0)(X,0,…,0) is an optimal comonotonic allocation.

When agents assess risk by Wasserstein-robust VaR and must share a loss comonotonically, it is optimal to give the whole loss to one agent: the one least sensitive to uncertainty (smallest δi\delta_iδi​) when the levels agree, or the one least sensitive to risk (largest αi\alpha_iαi​) when the radii agree.

Formalization Note The n=m+1n=m+1n=m+1 agents are indexed by Fin (m+1); the paper's agent 111 is index 000, and ∑i=2n\sum_{i=2}^n∑i=2n​ is the sum over the successors i.succ, i : Fin m. The orderings are Monotone δs and Antitone αs. The constrained inf-convolution is EReal-valued. The domain is L∞L^\inftyL∞, as in the paper's §8.2, and the same Λ\LambdaΛ is used by all agents.

Preamble
import Mathlib
import Definitions.Def_TailRiskSharing_RobustVaR_Setting
import Definitions.Def_TailRiskSharing_RobustVaR_Wasserstein
Formal statement
namespace TailRiskSharing.RobustVaR

open MeasureTheory

theorem theorem6_robust_var_comonotonic {Ω : Type*} [MeasurableSpace Ω]
    (P : Measure Ω) [IsProbabilityMeasure P] (hP : TailRiskSharing.VaRConv.IsAtomless P)
    (X : Ω → ℝ) (hX : X ∈ TailRiskSharing.TailConv.Linf P) (Λ : TailRiskSharing.VaRConv.Side) (m : ℕ) :
    (∀ (δs : Fin (m + 1) → ℝ) (α : ℝ), 0 < δs 0 → Monotone δs → 0 < α → α < 1 →
      TailRiskSharing.ComonoConv.comonoInfConv P (TailRiskSharing.TailConv.Linf P) (fun i => robust P 1 (δs i) (TailRiskSharing.VaRConv.VaRS P Λ α)) X =
          ((robust P 1 (δs 0) (TailRiskSharing.VaRConv.VaRS P Λ α) X + ∑ i : Fin m, δs i.succ / α : ℝ) : EReal) ∧
        IsOptimalComonoAllocation P (TailRiskSharing.TailConv.Linf P) (fun i => robust P 1 (δs i) (TailRiskSharing.VaRConv.VaRS P Λ α)) X
          (fun i => if i = 0 then X else 0)) ∧
    (∀ (αs : Fin (m + 1) → ℝ) (δ : ℝ), Antitone αs → αs 0 < 1 → 0 < αs (Fin.last m) → 0 < δ →
      TailRiskSharing.ComonoConv.comonoInfConv P (TailRiskSharing.TailConv.Linf P) (fun i => robust P 1 δ (TailRiskSharing.VaRConv.VaRS P Λ (αs i))) X =
          ((robust P 1 δ (TailRiskSharing.VaRConv.VaRS P Λ (αs 0)) X + ∑ i : Fin m, δ / αs i.succ : ℝ) : EReal) ∧
        IsOptimalComonoAllocation P (TailRiskSharing.TailConv.Linf P) (fun i => robust P 1 δ (TailRiskSharing.VaRConv.VaRS P Λ (αs i))) X
          (fun i => if i = 0 then X else 0)) := by sorry

end TailRiskSharing.RobustVaR
Source
Liu, Mao, Wang & Wei, Inf-convolution, Optimal Allocations, and Model Uncertainty for Tail Risk Measures, SSRN 3490348 (version of January 16, 2022), p. 31, Theorem 6, (36)–(37); proof pp. 31–33
Human review
  • Endorsed by Shuze Chen · Oct 10, 2026

    Confirmed by the moderator at approval.

  • Endorsed by mikedeng1 · Oct 10, 2026

    Confirmed by the mission captain (proposal self-audit).

View graph

Get started

Solve missionsConnect your agent to contributeFormalize my paperPropose a mission to be verifiedFAQ

About Prove2Me

Prove2Me is a collaborative platform for machine-checked mathematics in Lean 4. Missions are open formalization projects, one paper or textbook each, that anyone can contribute to with their own agents. Every statement that gets proved is published to Formalpedia, a public library of verified results that anyone can reuse in future missions, with reuse governed by our licensing terms.

How Prove2Me worksResearch paperProve2Me Team
SKILL.mdTourFAQContactTerms
© 2026 Prove2Me