Nonnegative product rewards of earlier stopped history and later fresh Esscher increment factor exactly
ProvedAvramDividend.Classical.esscher_ordered_bounded_stops_history_future_increment_joint_reward_factoravram-classicalbounded-stoppingesscherjoint-lawmarkovnonnegative-integralordered-stopsstopped-history
For τ≤σ≤H genuine bounded stopping times, an arbitrary adapted earlier-history observable Y to any measurable space β, and any nonnegative measurable rewards f(Y) and g(X_(σ+u)−Xσ), the genuine finite-terminal Esscher Q_T expected product equals the stopped Esscher likelihood expectation of f(Y) at τ times the deterministic Q_u expected reward of g(X_u). The equality is valid in extended nonnegative reals including infinite expectations; no global tilted path law or WH independence is assumed.
Preamble
import Mathlib.MeasureTheory.Measure.Prod import Mathlib.MeasureTheory.Integral.Lebesgue.Map import Mathlib.MeasureTheory.Measure.Typeclasses.Probability import Theorems.Thm_AvramDividend_Classical_esscher_ordered_bounded_stops_history_future_increment_joint_product import Theorems.Thm_AvramDividend_Classical_chat04_esscher_weighted_finite_horizon_probability import Theorems.Thm_AvramDividend_Classical_levy_boundedStopping_stoppedFuturePath_measurable open AvramDividend.Classical MeasureTheory ProbabilityTheory Set open scoped NNReal ENNReal
Formal statement
theorem AvramDividend.Classical.esscher_ordered_bounded_stops_history_future_increment_joint_reward_factor
{Ω : Type*} [mΩ : MeasurableSpace Ω] {P : Measure Ω}
[IsProbabilityMeasure P] {𝓕 : Filtration ℝ≥0 mΩ}
(X : SpectrallyNegativeLevy P 𝓕) (θ : ℝ) (hθ : 0 ≤ θ)
(τ σ : Ω → ℝ≥0)
(hstopτ : IsStoppingTime 𝓕 (fun ω => (τ ω : WithTop ℝ≥0)))
(hstopσ : IsStoppingTime 𝓕 (fun ω => (σ ω : WithTop ℝ≥0)))
(horder : ∀ ω, τ ω ≤ σ ω)
(H : ℝ≥0) (hboundσ : ∀ ω, σ ω ≤ H)
(u T : ℝ≥0) (hT : H + 1 + u ≤ T)
{β : Type*} [MeasurableSpace β] (Y : Ω → β)
(hY : Measurable Y)
(hYτ : ∀ (B : Set β), MeasurableSet B →
∀ s : ℝ≥0, MeasurableSet[𝓕 s]
((Y ⁻¹' B) ∩ {ω : Ω | τ ω ≤ s}))
(f : β → ℝ≥0∞) (hf : Measurable f)
(g : ℝ → ℝ≥0∞) (hg : Measurable g) :
let QT : Measure Ω := P.withDensity (fun ω => ENNReal.ofReal
(Real.exp (θ * X.X T ω - (T : ℝ) * X.ψ θ)))
let QU : Measure Ω := P.withDensity (fun ω => ENNReal.ofReal
(Real.exp (θ * X.X u ω - (u : ℝ) * X.ψ θ)))
let Qτ : Measure Ω := P.withDensity (fun ω => ENNReal.ofReal
(Real.exp (θ * X.X (τ ω) ω -
(τ ω : ℝ) * X.ψ θ)))
(∫⁻ ω, f (Y ω) *
g (X.X (σ ω + u) ω - X.X (σ ω) ω) ∂QT) =
(∫⁻ ω, f (Y ω) ∂Qτ) *
(∫⁻ ω, g (X.X u ω) ∂QU) := by sorrySource
Apply accepted full genuine ordered-stop joint product law a8cb4e02 to pushforward pair (Y,Δσ(u)). The pinned Mathlib lintegral_prod_mul theorem (private exact signature checked at revision 0df444a) factors any two nonnegative measurable reward tests. The second marginal map X_u Q_u is S-finite because accepted finite-terminal Esscher unit probability is preserved by the measurable increment map. Convert both marginals with lintegral_map. The proof requires no additional strong Markov assumption.