Theorem 5.1, p. 446 — a distribution-invariant ρ is consistent with ⪰icx iff it satisfies (A2) and is risk averse
ProvedConvexRiskFn.Order.theorem_5_1Let be a probability space on which a uniform random variable exists, let and . Let be distribution invariant: whenever for all . Then
Here consistency means: for all , if for every increasing convex for which the expectations exist, then . (A2) means a.s. implies . Risk aversion means for every -algebra , every and every version of the conditional expectation.
The theorem characterizes, among law-invariant risk functions, exactly those that respect the increasing convex order, the counterpart of second-order stochastic dominance for costs; mean–risk models such as mean–semideviation and CVaR-based ones are special cases. No convexity of is assumed.
Formalization Note The existence of a uniform random variable and , , are the standing assumptions of the rest of §5.2 (p. 445). Elements of are functions with MemLp X p P; the order in (A2) is the almost-sure order of ; the increasing convex order is the published StochasticOrders_MonotoneConvex_IcxOrder.
import Mathlib import Definitions.Def_StochasticOrders_MonotoneConvex_IcxOrder import Definitions.Def_ConvexRiskFn_Order_Setting open MeasureTheory
namespace ConvexRiskFn.Order
/-- Theorem 5.1, p. 446: under the standing assumptions of §5.2 (`𝒳 = ℒ_p(Ω, ℱ, P)`, `p ∈ [1, ∞)`,
a uniform random variable exists), a distribution-invariant risk function is consistent with the
increasing convex order iff it satisfies (A2) and is risk averse. -/
theorem theorem_5_1 {Ω : Type*} [MeasurableSpace Ω] (P : Measure Ω) [IsProbabilityMeasure P]
(p : ENNReal) (hp1 : 1 ≤ p) (hpt : p ≠ ⊤) (hU : ∃ U : Ω → ℝ, IsUniformRV P U)
(ρ : (Ω → ℝ) → ℝ) (hρ : DistInvariant P p ρ) :
ConsistentIcx P p ρ ↔ A2 P p ρ ∧ RiskAverse P p ρ := by sorry
end ConvexRiskFn.Order
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.