The Limits of Price Discrimination with a Bayesian Seller

Hanrui Zhang; Wei Tang; Yilin Li; Yuan Deng

With Bayesian private production costs, third-degree price discrimination achieves only the surplus pairs that form a proper projection of a polytope whose essential constraints are discounted flow conservation, and those pairs can be optim

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · grok-4.5

2026-07-15 04:42 UTC pith:YOOMQT5T

load-bearing objection Abstract-only: clean polytope/flow-network generalization of BBM 2015 to Bayesian private cost, with a poly-time LP claim that looks worth a referee if the extremal-market decomposition holds. the 2 major comments →

arxiv 2607.12615 v1 pith:YOOMQT5T submitted 2026-07-14 econ.TH cs.GT

The Limits of Price Discrimination with a Bayesian Seller

Yuan Deng , Yilin Li , Wei Tang , Hanrui Zhang This is my paper

classification econ.TH cs.GT

keywords third-degree price discriminationBayesian private costmarket segmentationseller surplusbuyer surpluspolytope projectiondiscounted flow networkextremal markets

verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper asks what welfare outcomes a monopolist can achieve by segmenting a market when its own production cost is random and private. Different ways of slicing the market into segments produce different pairs of seller surplus and buyer surplus; the set of attainable pairs is no longer the simple triangle of Bergemann, Brooks and Morris once cost is Bayesian. The authors prove that this set is exactly the projection of a polytope defined by a polynomial number of linear inequalities, the important ones being flow-conservation constraints in a discounted flow network. Consequently any linear combination of the two surpluses can be optimized by a polynomial-time algorithm that finds the corresponding market segmentation. Along the way they show every market decomposes into a convex combination of piecewise equal-surplus “extremal markets” that preserve both surpluses exactly.

Core claim

When production cost is Bayesian and private to the seller, the region of (seller surplus, buyer surplus) pairs attainable by third-degree price discrimination coincides with a proper projection of a polytope whose essential facets are the flow-conservation constraints of a discounted flow network; this yields a polynomial-time algorithm for optimal market segmentations under any linear welfare objective.

What carries the argument

A polytope of polynomial size whose projection recovers the attainable surplus region; its essential inequalities encode flow conservation in a discounted flow network whose nodes and arcs are indexed by the possible costs and by the equal-surplus pieces of extremal markets.

Load-bearing premise

Every market admits a convex decomposition into piecewise equal-surplus extremal markets that preserves both seller surplus and buyer surplus exactly.

What would settle it

Exhibit a market and Bayesian cost distribution whose attainable (seller surplus, buyer surplus) region is not equal to the claimed polytope projection, or show that optimizing a linear combination of the two surpluses is NP-hard.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

Any linear welfare objective (weighted sum of seller and buyer surplus) admits a polynomial-time optimal market segmentation.
The geometry of third-degree price discrimination with Bayesian costs is completely described by a polytope of polynomial description complexity.
Extremal markets that are piecewise equal-surplus with respect to each possible cost are the only vertices needed to span the attainable surplus set.
The classical fixed-cost triangle of Bergemann–Brooks–Morris is recovered as the special case of a single-point cost distribution.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

The discounted-flow representation may extend to multi-product or multi-period settings where costs are correlated across markets.
If the number of cost atoms grows, the polytope size remains polynomial only in the number of atoms, suggesting a natural complexity threshold for continuous cost distributions.
The same extremal-market decomposition could be used to characterize robust or worst-case surplus guarantees under cost misspecification.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Desk Editor's Note

Abstract-only: clean polytope/flow-network generalization of BBM 2015 to Bayesian private cost, with a poly-time LP claim that looks worth a referee if the extremal-market decomposition holds.

read the letter

Colleague —

We only have the abstract, so this is provisional. The punchline is a genuine within-subfield advance: they take Bergemann–Brooks–Morris (2015) and replace fixed production cost with Bayesian private cost, then claim the achievable (seller surplus, buyer surplus) region is a proper projection of a polytope with polynomially many linear constraints whose essential ones are flow-conservation equalities in a discounted flow network. That immediately gives a poly-time LP for any linear combination of the two surpluses. If the proofs check out, that is useful method territory for information design and monopoly segmentation.

What looks new and well-posed: the discounted flow network, the polytope-projection characterization, the poly-time algorithm, and the structural lemma that every market is a convex combination of piecewise equal-surplus extremal markets (now with respect to the different possible costs) that preserves both surpluses exactly. That is the natural generalization of BBM’s fixed-cost extremals, and they present it as the en-route result that underpins the projection. No free parameters, no data-fitting; pure structural + algorithmic claim. Self-citation pattern is just the expected BBM foundation, not circular.

The soft spot is exactly the one the stress-test flags, and it is load-bearing rather than cosmetic. Because the geometry is “complex” once cost is Bayesian, the whole polytope claim rests on that convex decomposition preserving both surpluses after the prior is integrated. If the weights cannot be chosen independently of realized cost, or if expected surpluses leak, the lifted polytope over- or under-approximates and the flow constraints are incomplete. We cannot verify the argument from the abstract alone, so soundness stays provisional. Everything else (poly-time LP, finite-type assumptions common in this literature) is plausible once that lemma is granted.

Who it is for: people who already work on BBM-style information design, third-degree price discrimination, or algorithmic mechanism design with surplus objectives. A serious referee in that circle should see the full paper; the claim is sharp enough and the method (flow network + projection) is concrete enough that desk rejection would be a mistake. I would not bring the abstract to reading group, but I would read the full text if it lands. I would not cite yet—no proofs in hand—but I would accept it for peer review.

Bottom line: send it out. The decomposition is the one thing the referees must pressure-test; if it holds, this is a clean, citable extension.

Referee Report

2 major / 2 minor

Summary. The paper generalizes Bergemann, Brooks and Morris (2015) on the limits of third-degree price discrimination to the case of a Bayesian, privately known production cost. It claims that the set of achievable (seller surplus, buyer surplus) pairs coincides with a proper projection of a polytope defined by a polynomial number of linear constraints, the essential ones being flow-conservation constraints in a discounted flow network. As a consequence it asserts a polynomial-time algorithm for optimal market segmentations under any linear objective in the two surpluses. En route it claims a structural decomposition: every market is a convex combination of piecewise equal-surplus extremal markets (with respect to the possible costs) that exactly preserves both seller and buyer surplus.

Significance. If the claimed polytope projection and surplus-preserving decomposition hold, the paper supplies a clean geometric and algorithmic extension of BBM 2015 to a setting whose welfare region is known to be more complex. The discounted-flow-network representation and the resulting poly-time LP for linear surplus objectives would be reusable tools in information design and monopoly pricing with private costs. The extremal-market decomposition itself is a natural generalization of the fixed-cost extremal markets of BBM and, if correct, is of independent structural interest.

major comments (2)

The abstract’s central identification of the achievable set with a proper projection of a flow-conservation polytope rests on the structural claim that every market admits a convex decomposition into piecewise equal-surplus extremal markets that preserves both seller and buyer surplus exactly under Bayesian costs. Only the abstract is available for this review, so the derivation of that decomposition (weights independent of realized cost, exact preservation after integrating the prior) cannot be checked. If the preservation step fails, the projected polytope may over- or under-approximate the true region and the algorithmic claim collapses. This is the load-bearing step that must be verified from the full proofs.
The abstract asserts a polynomial number of linear constraints and a poly-time algorithm, but does not state the precise type-space assumptions (finite support sizes for values and costs, etc.) that make the flow-network dimension polynomial. Without those assumptions made explicit and the projection argument written out, the complexity claim cannot be confirmed from the abstract alone.

minor comments (2)

The abstract is clear and well-structured, but a full manuscript would benefit from an early formal statement of the type-space finiteness assumptions that underwrite the polynomial bound.
A short comparison paragraph locating the discounted flow network relative to the ordinary flow arguments used in the fixed-cost BBM analysis would help readers see where the Bayesian generalization requires new ideas.

Circularity Check

0 steps flagged

No significant circularity: abstract claims a structural polytope characterization generalizing external BBM 2015 work, with no definitional or fitted reduction visible.

full rationale

The abstract-only text presents a pure theoretical characterization: the achievable (seller surplus, buyer surplus) region under Bayesian private cost equals a proper projection of a polytope whose essential constraints are discounted flow-conservation equalities, yielding a poly-time LP for any linear surplus objective. En route it asserts a convex decomposition of any market into piecewise equal-surplus extremal markets that exactly preserve both surpluses, generalizing the fixed-cost extremal markets of Bergemann–Brooks–Morris (2015). BBM 2015 is external prior literature (non-overlapping authors), not a self-citation that forces the new result by construction. There are no fitted parameters, no data-driven “predictions,” no uniqueness theorem imported from the present authors, and no renaming of a known empirical pattern. Because the full text is unavailable, no equation-level reduction can be exhibited; on the abstract’s face the derivation is self-contained structural reasoning rather than circular. Residual concerns about whether the decomposition preserves surpluses under the Bayesian prior are correctness risks, not circularity. Score 0 is therefore the honest finding.

Axiom & Free-Parameter Ledger

0 free parameters · 4 axioms · 2 invented entities

Abstract-only review: free parameters are not visible (theory paper, no data fits). Axioms are the standard finite-type / Bayesian-mechanism-design background plus the BBM segmentation framework and the private Bayesian cost model. Invented modeling entities are the discounted flow network and the generalized piecewise equal-surplus extremal markets; both are mathematical devices whose independent evidence is internal to the claimed proofs.

axioms (4)

domain assumption Seller production cost is drawn from a known Bayesian prior and is private to the seller.
Stated as the modeling premise that generalizes fixed-cost BBM; without it the polytope claim does not apply.
domain assumption Third-degree price discrimination via market segmentation as in Bergemann–Brooks–Morris (2015) is the welfare-generating mechanism.
The paper is explicitly a generalization of that framework; the achievable-set geometry inherits its segmentation language.
standard math Standard linear-programming / polyhedral duality and flow-conservation reasoning apply to the constructed network.
Needed for the claim that polynomially many linear constraints define a polytope whose projection is the welfare region and that linear objectives are poly-time solvable.
domain assumption Markets admit a finite (or otherwise poly-size) type/support representation so that the constraint system is polynomial.
Implicit in the polynomial-time and polynomial-constraint claims; not spelled out in the abstract but required for those complexity statements.

invented entities (2)

discounted flow network no independent evidence
purpose: Encode the essential linear constraints whose flow conservation characterizes achievable surplus pairs under Bayesian cost.
Introduced as the combinatorial object whose conservation laws give the polytope; no external physical referent, only mathematical utility inside the proof.
piecewise equal-surplus extremal markets (Bayesian-cost generalization) no independent evidence
purpose: Serve as the extreme points whose convex combinations recover any market while preserving seller and buyer surplus.
Generalizes BBM’s equal-surplus extremal markets to multiple possible costs; existence and surplus-preservation are load-bearing and proved only inside the paper.

pith-pipeline@v1.1.0-grok45 · 6152 in / 2749 out tokens · 32656 ms · 2026-07-15T04:42:23.499844+00:00 · methodology

0 comments

read the original abstract

We study the limits of third-degree price discrimination when the production cost is Bayesian and private to the seller, generalizing the seminal work of Bergemann, Brooks and Morris (2015). The rough setup is the following: A monopoly seller sets different prices for buyers in different "segments" of the market so as to maximize seller surplus. Different ways in which the aggregate market is decomposed into segments lead to different welfare outcomes, i.e., (seller surplus, buyer surplus) pairs. When the production cost is Bayesian, the region of achievable welfare outcomes can exhibit complex shapes beyond the clean characterization by Bergemann, Brooks and Morris for the case with a fixed cost. We show that with a Bayesian cost, this region coincides with a proper projection of a polytope defined by a polynomial number of linear constraints, the essential ones of which correspond to flow conservation in a "discounted" flow network. As a result, we give a polynomial-time algorithm that computes optimal market segmentations in terms of any linear combination of the seller surplus and the buyer surplus. En route, we establish the following structural property: Any market can be written as a convex combination of "extremal markets" in a way preserving the seller surplus and the buyer surplus. These extremal markets are piecewise equal-surplus with respect to different possible costs, generalizing a similar notion introduced by Bergemann, Brooks and Morris when the cost is fixed.

The Limits of Price Discrimination with a Bayesian Seller

Core claim

What carries the argument

Load-bearing premise

What would settle it

If this is right

Where Pith is reading between the lines

discussion (0)