(A.21), p. 65 — α_i = ρσ_α X_i + e_i with e_i ~ N(0, σ_α²(1−ρ²)) and X_1,…,X_N, e_1,…,e_N mutually independent
ProvedImpactInvesting.Portfolio.regression_decompositiongaussianimpact-investingindependencep2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1regression
Assume (A1) with the normalisations , , and let , . Define the regression residuals
so that , which is display (A.21) under the normalisations. Then
- each is normal with mean and variance ;
- the random variables are mutually independent.
This is the first step of the paper's proof of Proposition 2: it separates the part of the excess return explained by the impact factor from a residual independent of all impact factors.
Formalization Note. The law of is stated as P.map (e i) = gaussianReal 0 (σα^2 (1-ρ^2)). Mutual independence of the variables is iIndepFun over the index type Fin N ⊕ Fin N (left copy: , right copy: ). Indices are 0-based.
Preamble
import Mathlib import Definitions.Def_ImpactInvesting_Portfolio_OrderStat import Definitions.Def_ImpactInvesting_Portfolio_Setting open MeasureTheory ProbabilityTheory
Formal statement
namespace ImpactInvesting.Portfolio
theorem regression_decomposition {Ω : Type*} [MeasurableSpace Ω] {N : ℕ} (P : Measure Ω) [IsProbabilityMeasure P]
(X α : Fin N → Ω → ℝ) (σα ρ : ℝ) (hA1 : IsA1 P X α σα ρ) (hσα : 0 < σα)
(hρ₁ : -1 < ρ) (hρ₂ : ρ < 1) :
(∀ i, P.map (fun ω => α i ω - ρ * σα * X i ω) =
gaussianReal 0 (σα ^ 2 * (1 - ρ ^ 2)).toNNReal) ∧
iIndepFun (fun k : Fin N ⊕ Fin N =>
Sum.elim (fun i => X i) (fun i ω => α i ω - ρ * σα * X i ω) k) P := by sorry
end ImpactInvesting.Portfolio
Source
Lo & Zhang, Quantifying the Impact of Impact Investing, SSRN 3944367 (draft of 29 June 2023), p. 65, Appendix B.2, proof of Proposition 2, (A.21)
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.