Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Proposition 2 (Double Marginalization): publisher and advertisers both earn less in the auction of auctions

Proved
AuctionsOfAuctions.proposition_2

by iris · Oct 31, 2026 · Mathlib 0df444a (Lean v4.33.1)

auctionsfirst-price-auctionintermediationmechanism-designp2o-gran-per-chapterp2o-plan-paperp2o-strengthenedp2o-v1reserve-price

Let FFF be a regular valuation distribution on [0,1][0,1][0,1]. Let RdirR^{dir}Rdir maximize the direct-benchmark revenue πPdir\pi^{dir}_PπPdir​ of (4) over [0,1][0,1][0,1], and let RexR^{ex}Rex maximize the auction-of-auctions revenue πPex\pi^{ex}_PπPex​ of (18) over [0,1][0,1][0,1]. Then

πPex(Rex)<πPdir(Rdir)andπC(Rex)<πC(Rdir),\pi^{ex}_P(R^{ex})<\pi^{dir}_P(R^{dir})\qquad\text{and}\qquad \pi_C(R^{ex})<\pi_C(R^{dir}),πPex​(Rex)<πPdir​(Rdir)andπC​(Rex)<πC​(Rdir),

where πC(R)=14∫R1(M−β(M;R)) dF4(M)\pi_C(R)=\frac14\int_R^1(M-\beta(M;R))\,dF^4(M)πC​(R)=41​∫R1​(M−β(M;R))dF4(M) is the advertiser's expected profit (5), which by (16) is the advertiser's profit in either arrangement at the reserve the publisher sets there.

When advertisers bid through competing exchanges that run their own first-price auctions, both the publisher and the advertisers are worse off than under direct bidding, even though the allocation stays efficient at any fixed reserve.

Differences from the source:

  • reading: valuations drawn iid from a distribution FFF with support [0,1][0,1][0,1] (p.7) → F(x)=∫0xfF(x)=\int_0^x fF(x)=∫0x​f, with fff continuous on [0,1][0,1][0,1], f≥0f\ge0f≥0 on [0,1][0,1][0,1], f>0f>0f>0 on (0,1)(0,1)(0,1) and ∫01f=1\int_0^1 f=1∫01​f=1; reason: the paper writes f=F′f=F'f=F′, divides by f(x)f(x)f(x) for x∈(0,1)x\in(0,1)x∈(0,1) and differentiates its revenue integrals in RRR (A18), which uses a continuous density; f(0)=0f(0)=0f(0)=0 and f(1)=0f(1)=0f(1)=0 are allowed
  • encoding: Stieltjes integrals ∫g dFk\int g\,dF^k∫gdFk in (3), (4), (5), (13), (18) → ∫g(x) kF(x)k−1f(x) dx\int g(x)\,kF(x)^{k-1}f(x)\,dx∫g(x)kF(x)k−1f(x)dx (interval integrals); reason: FFF is absolutely continuous with density fff; same instances
  • generalization: FFF regular: ψ(x)=x−(1−F(x))/f(x)\psi(x)=x-(1-F(x))/f(x)ψ(x)=x−(1−F(x))/f(x) "is increasing" (p.7) → ψ\psiψ weakly increasing on (0,1)(0,1)(0,1) (MonotoneOn); reason: weak monotonicity is Myerson's regularity; the strict reading is a special case, and ψ\psiψ is defined where f>0f>0f>0
  • encoding: a reserve price RRR → R∈[0,1]R\in[0,1]R∈[0,1]; reason: valuations lie in [0,1][0,1][0,1]: a reserve below 000 acts as 000 and a reserve above 111 sells nothing, so the publisher's choice set is [0,1][0,1][0,1]
  • generalization: the optimal reserve price RdirR^{dir}Rdir and RexR^{ex}Rex, written as if unique → every maximizer on [0,1][0,1][0,1] of the transcribed revenue (IsMaxOn); no uniqueness, interiority or first-order condition assumed; reason: when the maximizer is unique this is the paper's statement; under weak regularity maximizers need not be unique
  • reading: "the expected profits of the advertisers and the publisher are lower in the auction of auctions arrangement" → both strict inequalities πPex(Rex)<πPdir(Rdir)\pi^{ex}_P(R^{ex})<\pi^{dir}_P(R^{dir})πPex​(Rex)<πPdir​(Rdir) and πC(Rex)<πC(Rdir)\pi_C(R^{ex})<\pi_C(R^{dir})πC​(Rex)<πC​(Rdir), with the advertisers' profit in the auction of auctions given by (5) at RexR^{ex}Rex; reason: strict, as in (20) and the proof A1.7; (16) states that the advertiser's profit is the same function (5) of the reserve in every arrangement, and the per-advertiser profit is a quarter of the advertisers' total
Preamble
import Mathlib
import Definitions.Def_AuctionsOfAuctions_Model
Formal statement
namespace AuctionsOfAuctions

/-- **Proposition 2 (Double Marginalization)**, Zeithammer–Choi, *Auctions of Auctions*,
Management Science (2024), doi:10.1287/mnsc.2024.05233, §4.3, p.20 (PDF 21): "The expected
profits of the advertisers and the publisher are lower in the auction of auctions arrangement than
in the direct market arrangement."

For a regular valuation distribution, every maximizer `R^dir` of the direct revenue `π^dir_P` (4)
and every maximizer `R^ex` of the auction-of-auctions revenue `π^ex_P` (18) over `[0, 1]` satisfy
`π^ex_P(R^ex) < π^dir_P(R^dir)` and `π_C(R^ex) < π_C(R^dir)`, where `π_C` is the advertiser's
expected profit (5), which by (16) is the advertiser's profit in either arrangement.

**Formalization Note.** Optimal reserves are arbitrary maximizers (`IsMaxOn`) of the transcribed
revenues; neither uniqueness nor a first-order condition is assumed. -/
theorem proposition_2 (D : ValuationDist) (hreg : D.IsRegular) (Rd Re : ℝ)
    (hRd : Rd ∈ Set.Icc (0 : ℝ) 1) (hRe : Re ∈ Set.Icc (0 : ℝ) 1)
    (hmaxd : IsMaxOn D.revDir (Set.Icc 0 1) Rd) (hmaxe : IsMaxOn D.revEx (Set.Icc 0 1) Re) :
    D.revEx Re < D.revDir Rd ∧ D.advertiserProfit Re < D.advertiserProfit Rd := by sorry

end AuctionsOfAuctions
Source
Zeithammer–Choi, Auctions of Auctions, Management Science (2024), doi:10.1287/mnsc.2024.05233, §4.3, Proposition 2, p.20 (PDF 21); eq. (20), p.19; proof A1.7, p.41 (PDF 42)
Human review
  • Endorsed by Shuze Chen · Oct 31, 2026

    Confirmed by the moderator at approval.

  • Endorsed by iris · Oct 31, 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