📈 economics

A Note on Market Segmentation and Bertrand Competition

This paper demonstrates that in Bertrand competition with a finite number of firms and consumers having bounded willingness to pay, equilibrium profits are zero regardless of the market segmentation profile.

Original authors: Zhang Xu, Mingsheng Zhang, Wei Zhao

Published 2026-08-11

📖 6 min read🧠 Deep dive

CC0 1.0

Original authors: Zhang Xu, Mingsheng Zhang, Wei Zhao

Original paper dedicated to the public domain under CC0 1.0 (http://creativecommons.org/publicdomain/zero/1.0/). ✨ 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 a bustling marketplace where everyone is trying to sell the exact same lemonade. In the world of economics, this is called "Bertrand competition." It's a high-stakes game where sellers constantly undercut each other's prices to steal customers. If you sell a cup for $1, I'll sell mine for $0.99, and you'll drop to $0.98, until the price hits rock bottom and nobody makes any profit. Now, imagine a new twist: what if the sellers could peek at each customer's wallet before setting a price? This is "market segmentation" or "personalized pricing." With modern technology, companies can collect data to guess exactly how much you are willing to pay. The big question economists have been asking is: Does this ability to peek at wallets help sellers stop the price war? Can they use data to group customers and charge them different prices, effectively calming the competition and keeping their profits high?

This paper, written by Zhang Xu, Mingsheng Zhang, and Wei Zhao, dives straight into that question. They set up a mathematical model to see if splitting the market into different groups (segmentation) can save firms from the zero-profit trap of fierce price competition. Their answer is a definitive, mathematical "no," but with a crucial caveat: under specific conditions. They prove that if there are a finite number of firms and customers have a maximum limit on how much they are willing to pay (a bounded willingness to pay), then even with perfect data and segmentation, the competition will still drive profits down to zero. It doesn't matter how clever the segmentation strategy is; as long as these specific conditions hold, the pressure to undercut the competition remains too strong.

The Great Lemonade War: Why Data Can't Save the Sellers (When Budgets are Bounded)

Let's picture a town with a few lemonade stands (let's say $n$ stands) and a crowd of thirsty people. Each person has a secret "Willingness to Pay" (WTP)—the most they are willing to spend for a cold drink. Maybe some are rich and willing to pay $5, while others are on a budget and will only pay $0.50. The paper assumes this maximum amount is bounded, meaning no one is willing to pay an infinite amount of money.

In a normal, chaotic market, these stands compete by lowering prices. If one stand charges $2, another will charge $1.90 to steal the customers. This race to the bottom is the classic "Bertrand Competition." The paper asks: What if the stands could use a super-advanced computer to sort the crowd? They could create a "market segmentation profile," essentially putting customers into different buckets based on their data. Maybe Bucket A gets a price of $4, and Bucket B gets a price of $2. The hope is that by tailoring prices, the stands can avoid fighting over the same customers and keep their profits up.

The authors, however, show that this strategy fails completely within the specific context of their model. They prove a theorem that says: If there are a finite number of firms and a maximum price limit (bounded WTP), the profit for every firm in a stable equilibrium will be exactly zero, regardless of the market segmentation profile.

How does this happen? The logic is a bit like a game of "hot potato" with prices. The paper uses a clever mathematical argument to show that if the price is even slightly above zero, there is always a temptation for a firm to adjust its strategy. Imagine the firms are all agreeing (implicitly) to charge a price of $1. The paper shows that if the price is $1, a smart firm can slightly adjust its strategy to grab a huge chunk of the market for a tiny bit less, or in this specific proof, by shifting its pricing strategy to capture the "safe" range of customers without triggering a total price war.

The proof gets a little technical, but the core idea is this: The authors look at the highest possible price that actually happens in the market (let's call it the "ceiling"). They show that if this ceiling is above zero, there is always a "gap" or a safe spot where a firm can change its price just a tiny bit to make more money than it would by sticking to the current plan. It's like finding a loophole in the rules. Because every firm is rational and wants to make money, they will all eventually exploit this loophole. This constant jockeying for position forces the price down until the profit disappears entirely.

The paper explicitly rules out the idea that market segmentation can "alleviate" (reduce) the intensity of competition under these specific bounded conditions. It's not just that segmentation might fail; the authors prove that under the assumption of bounded willingness to pay, the profit must be zero. They don't just suggest it; they provide a rigorous mathematical proof that leaves no room for doubt within the scope of their assumptions.

So, in this story, the "super-technology" of data collection doesn't give the lemonade stands a superpower if budgets are capped. Instead, it turns out that the sheer pressure of having a few competitors fighting over a limited pool of customers is too strong to be tamed by clever grouping. The market segmentation profile, no matter how complex or "measurable" it is, cannot stop the race to the bottom in this specific scenario. The only way to escape this zero-profit trap, according to the paper, would be to change the fundamental rules of the game—perhaps by having an infinite number of customers with no upper limit on what they are willing to pay, which is a scenario the authors mention as a topic for future work, but not one that applies to the real world of bounded budgets.

In short, if you are a seller in a market with a few competitors and customers who have a cap on their spending, don't expect your data analytics team to save your profits. The math says the competition will win, and the profit will vanish, no matter how you try to segment the crowd under these specific conditions.

Technical Summary: A Note on Market Segmentation and Bertrand Competition

Problem Statement
The paper addresses a central question in industrial organization: how does market segmentation (a mechanism for price discrimination) affect competition among firms? While the effects of price discrimination have been extensively analyzed in monopolistic settings and in markets with differentiated goods, the benchmark scenario of Bertrand competition with homogeneous goods remains a critical area of inquiry. Specifically, the authors investigate whether market segmentation can alleviate the intensity of competition and sustain positive equilibrium profits when the number of firms is finite and consumers' willingness to pay (WTP) is bounded.

Methodology
The authors construct a formal game-theoretic model with the following components:

Agents: A finite set of firms ( $N = \{1, \dots, n\}$ ) and a unit mass of consumers.
Consumer Heterogeneity: Consumers possess a willingness to pay (WTP), denoted by $\omega$ , drawn from a bounded interval $\Omega := [0, \bar{\omega}]$ with a measure $\mu$ .
Market Segmentation: A segmentation profile is defined by a measurable function $\pi: \Omega \to \Delta(M)$ , where $M$ represents the set of market segments. This function maps a consumer's WTP to a probability distribution over segments, effectively determining how information about consumer types is revealed to firms.
Strategies: Firms employ mixed strategies, represented as measurable functions $\sigma_i: M_i \to \Delta(\mathbb{R}_+)$ , mapping observed segments to price distributions.
Profit Function: Firm $i$ 's profit depends on the price profile $p$ and consumer WTP $\omega$ . The demand allocation follows standard Bertrand rules: if a firm's price exceeds the minimum price in the market or the consumer's WTP, demand is zero; otherwise, the total demand at the minimum price is allocated among firms offering that price.

The authors analyze the Nash Equilibrium of this game, where no firm can increase its expected profit by unilaterally deviating from its strategy profile.

Key Contributions and Results
The paper's primary contribution is Theorem 1, which establishes a definitive result regarding equilibrium profits under the specified conditions:

Zero-Profit Theorem: In any Nash Equilibrium of Bertrand competition with a finite number of firms and bounded consumer WTP, the equilibrium profit for every firm is identically zero ( $Q_i \equiv 0$ for all $i \in N$ ).
Robustness to Segmentation: This result holds under any market segmentation profile. Regardless of how the market is segmented or how information is disclosed to firms, the competitive pressure in the homogeneous goods setting drives profits to zero.

Proof Sketch
The proof proceeds by contradiction:

Assume there exists a Nash Equilibrium where the essential supremum of the transaction price distribution, denoted as $a$ , is strictly positive ( $a > 0$ ).
The authors identify a price point $\hat{a} \in (a/n, a]$ where the transaction price distribution is atomless (continuous).
They demonstrate that if the transaction price is in the range $[\hat{a}, a]$ , the total profit is bounded by $a \cdot G([\hat{a}, a])$ . Consequently, at least one firm must earn a profit share of at most $\frac{1}{n} a \cdot G([\hat{a}, a])$ .
The authors construct a profitable deviation for this firm: shifting its strategy to concentrate all probability mass on the price $\hat{a}$ (specifically, moving mass from prices $\ge \hat{a}$ to exactly $\hat{a}$ ).
By calculating the profit difference, they show that this deviation yields a strictly higher profit because $\hat{a} > a/n$ and the mass of the transaction price distribution in the relevant interval is positive.
This contradiction implies that the assumption $a > 0$ is false; therefore, the equilibrium transaction price must be zero, leading to zero profits.

Significance and Claims
The paper offers an "affirmative negative answer" to the question of whether market segmentation can soften Bertrand competition. The authors claim that segmentation does not provide a mechanism for firms to escape the zero-profit outcome in homogeneous goods markets with finite firms and bounded valuations.

The authors position their work as a simplification and generalization of existing literature, specifically Jann and Schottmüller (2015). While Jann and Schottmüller considered correlated equilibria, the authors note that their results can be interpreted within the framework of Bayesian Correlated Equilibrium (referencing Bergemann and Morris, 2016).

Future Directions
The authors modestly outline a specific avenue for future research: investigating whether the zero-profit result holds if the consumers' WTP range is unbounded ( $\Omega \in [0, \infty)$ ). They pose the question of whether a market segmentation profile could be designed to induce a "Folk Theorem" outcome (sustaining positive profits) in such an unbounded setting, referencing Baye and Morgan (1999).

The Great Lemonade War: Why Data Can't Save the Sellers (When Budgets are Bounded)

Technical Summary: A Note on Market Segmentation and Bertrand Competition

More like this