Proof of Proposition 1, (15) — without full cross-targeting ADX1 earns at most 2wφ(1 − α + η₁) < (13)
ProvedAdDataSharing.Base.revenue_bound_openad-exchange-revenueonline-advertisingp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1
Assume the standing assumptions and suppose allows cross-targeting. Then:
- for every profile in which not both advertisers share data with , its revenue is at most
- this bound is attained when one advertiser shares data with and the other, on , cross-targets it on ;
- the bound is strictly less than the closed revenue of (13).
This gives the "only if" direction of Proposition 1: without full cross-targeting, opening is strictly worse for .
Formalization Note. The bound holds for every such profile, not only equilibria, so no equilibrium hypothesis is needed.
Preamble
import Mathlib import Definitions.Def_AdDataSharing_Base_Model
Formal statement
namespace AdDataSharing.Base
/-- Proof of Proposition 1, p. 28, display (15): if ADX1 is open and not both advertisers
contract with it, its revenue is at most `2wφ(1 − α + η₁)`; this bound is attained when one
advertiser contracts with ADX1 and the other cross-targets on ADX1; and it is strictly less than
(13). -/
theorem revenue_bound_open (P : Params) (hP : P.Standing) :
(∀ σ : Plan × Plan, ¬ (σ.1.ex = .adx1 ∧ σ.2.ex = .adx1) →
revenue P true σ ≤ 2 * P.w * P.φ * (1 - P.α + P.η₁)) ∧
(∀ s : Bool, revenue P true (⟨.adx1, s⟩, ⟨.adx2, true⟩) = 2 * P.w * P.φ * (1 - P.α + P.η₁)) ∧
2 * P.w * P.φ * (1 - P.α + P.η₁) < 2 * P.w * P.φ * (1 - P.α + P.η₂) := by sorry
end AdDataSharing.Base
Source
Johnson, Jungbauer, Preuss, Online Advertising, Data Sharing, and Consumer Control, SSRN 3898798 (Nov. 2, 2022), p. 28, proof of Proposition 1, display (15)
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.