📈 economics

Screening and Segmenting: A Consumer Surplus Perspective

This paper demonstrates that when a monopolist engages in simultaneous second- and third-degree price discrimination by adjusting both prices and qualities, consumer welfare is determined by the interplay of cost structures and demand elasticities, with the consumer-optimal segmentation ensuring identical quality for consumers of the same value across all segments.

Original authors: Dirk Bergemann, Tibor Heumann, Michael C. Wang

Published 2026-03-04

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Dirk Bergemann, Tibor Heumann, Michael C. Wang

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are the owner of a massive, magical bakery. You sell cakes of all different sizes and flavors. You know that some customers are willing to pay a fortune for a tiny, perfect slice, while others just want a huge, plain slab of cake for a few dollars.

In economics, this is called price discrimination. You want to charge everyone exactly what they are willing to pay to make the most profit. But there's a catch: you can't see inside your customers' heads to know their true budget.

To solve this, you use two tricks:

The Menu (Second-Degree): You offer a menu: "Small cake for \ $5, Medium for \$ 10, Large for $15." Customers choose their size. The small cake is actually a bit too small (you "distort" the quality) so that rich people don't accidentally buy it for cheap.
The Groups (Third-Degree): You put customers into different rooms. In the "VIP Room," you charge more. In the "Student Room," you charge less.

The Big Question:
What happens if you combine these tricks? You put customers into specific groups (segments) and then give each group its own custom menu of cakes and prices. Does this help the customers, or does it just help the baker make even more money?

This paper by Bergemann, Heumann, and Wang answers that question. Here is the breakdown in simple terms.

1. The "Magic Mirror" of Segmentation

The authors discovered something surprising about the best way to organize these groups to help customers.

Imagine you have a group of people who all love chocolate and are willing to pay $10.

The Intuition: You might think, "If I put all the chocolate lovers in one room, I can charge them a specific price."
The Reality: The paper proves that in the best possible scenario for customers, everyone with the same willingness to pay gets the exact same cake size in every single room.

If you are a \ $10 chocolate lover, you get a medium cake whether you are in the VIP room or the Student room. The only thing that changes is the **price tag**. In one room, you might pay \$ 8; in another, you might pay $12. But the cake itself is identical.

The Analogy: Think of it like a concert. The "quality" is the view of the stage. The paper says that in the best scenario, if you have the same budget, you get the same view of the stage no matter which ticket section you are in. The only difference is that some sections have a "cover charge" (a fixed fee) added to the ticket price, while others don't.

2. The "Goldilocks" Zone: When Segmentation Helps

The authors found that segmentation isn't always good for customers. It depends on a tug-of-war between how much people want the product (Demand) and how hard it is to make the product (Cost).

The "Too Easy" Scenario (Bad for Customers):
Imagine your bakery is in a place where making cakes is very cheap and easy (low cost), and people are very sensitive to price (high demand elasticity).
- What happens: The baker is already being forced to offer great deals to get people to buy. If you try to split people into groups, the baker just gets smarter at tricking them. The customers end up worse off.
- The Verdict: If the market is already "elastic" (people are price-sensitive), don't segment. Keep everyone in one big room.
The "Just Right" Scenario (Good for Customers):
Imagine making cakes is a bit tricky and expensive (high cost), and people are less sensitive to price.
- What happens: The baker is currently making terrible cakes for the poor customers to save money. By splitting the market, the baker is forced to make better cakes for the poor group because they are isolated from the rich group.
- The Verdict: Segmentation can force the baker to be nicer to the lower-value customers, increasing their happiness.

3. The "Concentration" Trick

How does the baker actually do this? The paper describes a process called Concentration.

Imagine you have a crowd of 100 people: 90 poor people and 10 rich people.

The Problem: The baker ignores the poor people because there are too many of them, and the rich people are too few to justify making a special cake for them.
The Fix: You take 80 of the poor people and put them in a separate room. Now, in that room, the poor people are the majority. The baker realizes, "Hey, I have a huge crowd of poor people here! I better make a decent cake for them to sell to them."
The Result: The poor people in that specific room get a better cake than they would have in the big mixed room. The rich people in the other room get what they want. Everyone wins (or at least, the poor people win).

4. The "Slippery Slope" Warning

The paper also warns us about a specific type of cost.

If the cost of making the product is very rigid (it's hard to change the quality), segmentation is more likely to help customers.
If the cost is very flexible (it's super easy to tweak the product), segmentation often hurts customers because the baker can just tweak the product to squeeze more money out of everyone without actually improving anything.

Summary: The Takeaway for Policymakers

If you are a government regulator watching a company like Netflix, Spotify, or an airline:

Don't ban segmentation automatically. Sometimes, splitting customers into groups actually forces companies to offer better deals to the "little guys."
Look at the "Elasticity." If customers are very sensitive to price (they will leave if you raise prices even a little), segmentation is likely bad for them. The company is already being forced to be fair.
Look at the Costs. If the company has high costs to change their product, segmentation might be a good thing because it forces them to serve different groups better.

In a nutshell: Market segmentation is like a double-edged sword. It can be a tool for companies to exploit us, but if the market conditions are just right, it can be a tool that forces companies to treat us better. The key is knowing exactly when the sword cuts the right way.

1. Problem Statement

The paper investigates the welfare implications of market segmentation when a monopolist engages in simultaneous second-degree (screening via quality differentiation) and third-degree (pricing across distinct groups) price discrimination.

The Setup: A monopolist sells vertically differentiated products (quality $q$ ) to a continuum of buyers with private valuations ( $v$ ). The seller knows the aggregate distribution of values but can partition the market into arbitrary segments (submarkets).
The Mechanism: Within each segment, the seller posts a profit-maximizing menu (nonlinear pricing). Across segments, the seller can charge different prices for the same quality.
The Objective: The authors seek to characterize the consumer-optimal segmentation—the specific partition of the market that maximizes aggregate consumer surplus, given that the seller will always respond optimally to any segmentation by maximizing profit.
The Core Question: Under what conditions does market segmentation benefit consumers, and what is the structure of the segmentation that maximizes their welfare?

2. Methodology

The paper's primary methodological contribution is a dramatic reduction of the optimization problem's dimensionality.

From Distributions to Functions: The original problem involves optimizing over the space of distributions of distributions (segmentations $\sigma \in \Delta(\Delta V)$ ), which is infinite-dimensional and unwieldy. The authors reduce this to maximizing a single functional over a single function: the inverse hazard rate ( $h(v)$ ).
Local Information Rent: They define a "local information rent" function, $u(v, h) = h \cdot Q(v-h)$ , where $Q$ is the supply function (inverse marginal cost). This function captures the contribution of a buyer of type $v$ to total consumer surplus, given the inverse hazard rate $h$ at that value.
Majorization Constraint: The authors establish that any feasible segmentation must satisfy a majorization constraint. If $h^*$ $h^{*}$ is the inverse hazard rate of the aggregate market and $h$ $h$ is the average inverse hazard rate induced by a segmentation, then $h^*$ $h^{*}$ must majorize $h$ $h$ ( $h^* \succ h$ $h^{*} ≻ h$ ).
- Concentration: Grouping low-value buyers together lowers their inverse hazard rate.
- Dilution: Spreading low-value buyers across segments where higher types are missing raises their inverse hazard rate (requires creating gaps in support).
- The majorization constraint formalizes the trade-off: one can lower $h$ at some values only if one raises it at others (or vice versa) to maintain the cumulative distribution.
Uniform Segmentations: The authors prove that the optimal segmentation is always uniform. In a uniform segmentation, every buyer of a given type $v$ faces the same inverse hazard rate across all segments where they appear. This implies that while prices may vary, the quality allocation is identical for a given type across all segments.

3. Key Contributions

Dimensionality Reduction: The paper transforms a complex optimization over distributions of distributions into a standard optimal control problem over a single function $h(v)$ subject to linear majorization constraints.
Characterization of Consumer-Optimal Segmentation: It provides a complete characterization of the segmentation that maximizes consumer surplus, showing that identical buyers receive identical quality in every segment, with price differences manifesting only as fixed fees.
Conditions for Zero Segmentation: It identifies precise, sharp conditions under which any segmentation harms consumers (i.e., zero segmentation is optimal). This contrasts with previous literature (e.g., Haghpanah & Siegel, 2023) which suggested segmentation always helps in generic markets with finite goods. The authors show that with a continuum of qualities, segmentation gains can vanish entirely.
Elasticity Nexus: The paper explicitly links the optimal segmentation structure to the interaction between demand elasticity and cost convexity (supply elasticity).

4. Key Results

A. Structure of the Optimal Segmentation (Theorem 1)

The maximum consumer surplus is achieved by solving:
$\max_{h \prec h^*} \int u(v, h(v)) dF^*(v)$
where $h \prec h^*$ denotes the majorization constraint.

Uniform Quality: In the consumer-optimal segmentation, buyers with the same value $v$ receive the same quality $q(v)$ in every segment.
Price Variation: Prices differ across segments only by a fixed fee (base price). The marginal price schedule (quality menu) is uniform across segments.

B. Convex Segmentations (Theorem 2)

Under log-concave demand (monotone inverse hazard rate) and log-concave supply (convex or not-too-concave costs), the optimal segmentation has a simple "convex" structure:

Nested Intervals: Segments are nested intervals $[v_m, \bar{v}]$ sharing the same upper bound but differing in their lower bounds.
Two-Regime Solution:
- For high values ( $v \geq \hat{v}$ ): The inverse hazard rate remains at the aggregate level $h^*(v)$ .
- For low values ( $v < \hat{v}$ ): The inverse hazard rate is lowered to the level that maximizes local information rents, $h(v) = \arg\max_h u(v, h)$ .
No Dilution: Under these regularity conditions, the optimal strategy involves only concentration (grouping low types) and never dilution (creating gaps).

C. The Elasticity Threshold (Corollary 3)

The paper derives a sharp condition for when segmentation is always harmful to consumers:

Condition: Zero segmentation is optimal if and only if the aggregate inverse hazard rate is everywhere below the rent-maximizing level: $h^*(v) \leq h(v)$ for all $v$ .
Economic Interpretation: If aggregate demand is sufficiently elastic relative to the cost structure (specifically, if the supply is inelastic enough), the monopolist's screening in the aggregate market already provides near-optimal rents. Any attempt to segment the market merely redistributes buyers inefficiently, reducing surplus.
Iso-elastic Costs: For power cost functions $c(q) = q^\gamma/\gamma$ , zero segmentation is optimal if the demand elasticity $\eta$ satisfies $\eta \leq \frac{\gamma}{1-\gamma}$ .

D. Comparison with Finite Goods

The paper clarifies the discrepancy with Haghpanah and Siegel (2023), who found segmentation always helps in markets with finite goods. The authors show that as the quality space approaches a continuum, the gains from segmentation can vanish entirely if the cost convexity and demand elasticity satisfy the threshold conditions derived above.

5. Significance and Implications

Regulatory Policy: The findings suggest that blanket bans on market segmentation may be counterproductive. Segmentation can be welfare-enhancing when demand is inelastic relative to supply costs. Conversely, in markets with highly elastic demand and rigid supply, segmentation is uniformly detrimental, and regulators should be wary of allowing it.
Firm Strategy: Firms can use these insights to design segmentation policies that balance profit and consumer welfare. The result that quality menus should remain uniform across segments while prices vary via fixed fees offers a concrete operational guideline.
Big Data and Personalization: The paper informs debates on personalized pricing. It suggests that the ability to segment markets does not automatically guarantee consumer benefits; the outcome depends critically on the underlying cost and demand elasticities.
Generalizability: The framework is extended to adverse selection (where costs depend on buyer type) and Pareto-efficient outcomes (balancing consumer and social surplus), demonstrating the robustness of the majorization approach.

In summary, the paper provides a rigorous, unified framework for understanding how market segmentation interacts with nonlinear pricing. It moves beyond the binary "segmentation is good/bad" debate by providing a precise, testable threshold based on market primitives (elasticities) that determines when segmentation benefits or harms consumers.