The Anonymity Markup: Why Saying No to Tracking Can Cost You More

Based on the research ofWang, Jiang and Singh, "Impact of the Invisibles: Personalized Pricing on Platform with Privacy-Preserving Users," Information Systems Research, 2026

The Mechanism: Pooling as a Pricing Weapon

Everyone who declines a tracking prompt, rejects a cookie banner, or checks out as a guest assumes they are stepping outside the pricing game, not changing its rules. A new theoretical paper in *Information Systems Research* by Jieqiong (Julie) Wang, Zhaohui (Zoey) Jiang, and Param Vir Singh, all of Carnegie Mellon's Tepper School of Business, argues the opposite happens. As privacy rules expand what consumers can withhold, the authors build a two-stage model in which a platform first designs how to segment its users, then the platform and third-party sellers compete on price within whatever segments it created. The model's central device is what the authors call "fuzzy segmentation": rather than treating every anonymous, privacy-preserving user as an unpriceable blank, the platform strategically pools them with a selectively chosen subset of consumers who did share data, and prices that combined group as one.

The model's first result inverts a textbook intuition. Economists have long treated finer segmentation as pro-competitive: slice the market into narrower slivers, and price competition inside each sliver gets sharper, pushing prices down. Wang, Jiang, and Singh's model shows fuzzy segmentation does the opposite. By deliberately blurring the line between who is known and who is not, the platform softens the rivalry that fine-grained data would otherwise sharpen, and seller profits rise as a result. The second result is the one worth sitting with: the model finds that privacy-preserving consumers, the very people who opted out specifically to avoid being singled out for a worse deal, can still face a price premium at equilibrium. The third result spreads the damage further. Data-sharing consumers who would have preferred to buy from a third-party seller can also come out worse off, purely because the platform's segmentation design happened to lump them into the mixed group with privacy-preserving users. None of this requires the platform to behave badly in any conventional sense. It only requires the platform to notice that not-knowing-who-you-are is itself useful raw material for pricing, once it is allowed to choose who gets grouped with whom.

A Members-Only Discount Is an Everyone-Else Markup

The model is abstract, but a mechanism this close to home has already left fingerprints. In 2014, five Northeastern University researchers, Aniko Hannak, Gary Soeller, David Lazer, Alan Mislove, and Christo Wilson, ran identical, simultaneous queries against 16 popular e-commerce sites, comparing results from real browser profiles against clean, cookie-free synthetic accounts, to measure price discrimination and search-result steering. Among the sites where they found personalization, two travel retailers stood out for a specific pattern: "Cheaptickets and Orbitz implemented price discrimination by offering reduced prices on hotels to 'members.'" Flip that finding around and it says something sharper than a loyalty discount: the shopper who did not log in, who did not hand over an account's worth of identifying data, was the one left paying the un-discounted rate. Logging in was effectively a way of opting into a data-sharing segment; staying anonymous was the default that cost more.

That 2014 study predates the regulatory wave the Wang-Jiang-Singh model is built to explain, and it measured ordinary account-based personalization rather than the privacy-law-driven anonymity the paper models, so it is a precursor, not a replication. But it is exactly the kind of real-world groove the model's mechanism would carve, at a moment when identifying yourself to a seller was still mostly a matter of convenience rather than a privacy decision governed by law. What has changed since is scale: once regulation and platform policy made going unidentified a one-tap, default-on choice for hundreds of millions of people, the "member discount, stranger markup" pattern stopped being a quirk of two travel sites and became a structural feature available to any platform managing a large population of people it genuinely cannot identify.

The Opt-Out Majority Is Not a Fringe Case

The reason this mechanism matters now, rather than as a curiosity, is that "privacy-preserving users" are not a marginal sliver of the market. When Apple's App Tracking Transparency framework went live with iOS 14.5 in April 2021, analytics firm Flurry found that only 4% of U.S. iPhone users, in a sample of 2.5 million daily active users, actively chose to let apps track them; worldwide, in a 5.3-million-user sample, the opt-in rate was 12%. Reported the other way, roughly 96% of U.S. users and 88% of users worldwide left tracking off within the first two weeks. Four years on, that is still not a passing phase. Adjust's mobile-marketing data puts the industry-wide ATT opt-in rate, measured among users who were actually shown the prompt, at just 35% as of the second quarter of 2025, up only marginally from 34.5% in Q2 2024 and 34% in Q2 2023. In other words, roughly two in three prompted users have declined to be tracked, consistently, for three straight years, not as a one-time overreaction to a new prompt but as a stable behavioral default.

Anonymity does not make you invisible to pricing. It makes you indistinguishable from whoever else opted out, and you inherit whatever price that pooled group gets charged.

That stability is the condition the Wang-Jiang-Singh model assumes rather than treats as a curiosity: a large, durable population of consumers who cannot be individually priced because they have exercised a real legal and technical option to withhold data. The GDPR gives EU residents a right to object to having their personal data processed, including for direct marketing, under Article 21, and comparable mechanisms exist elsewhere; ATT is Apple's own platform policy rather than a government regulation, but it performs the identical economic function the paper models, converting a previously identifiable user into one the seller cannot price individually. The model's point is that regulation succeeding at its stated goal, letting people withhold their data, is precisely the condition under which fuzzy segmentation becomes available and profitable for the platform to deploy.

The Welfare Trap Hiding Inside Consumer Protection

This is where the paper's framing turns pointed rather than merely technical. Privacy regulation is built and sold on a simple premise: giving consumers control over their data protects them. Wang, Jiang, and Singh's model does not dispute that consumers gain control. It disputes that control and protection are the same thing. In their model, the platform's optimal response to a population of privacy-preserving users is not to price them generously as an olive branch, and not to ignore them as unpriceable, but to fold them into mixed segments precisely sized to soften the platform's and its sellers' competitive pricing pressure. The anonymous user pays a premium not despite opting out, but in some sense because of it, since anonymity is the raw material the platform uses to build the segment. And the collateral damage lands on people who did share their data: the model shows that data-sharing consumers who would have preferred a third-party seller can take a hit from being swept into the same mixed segment as privacy-preserving users, a negative spillover that has nothing to do with their own choices and everything to do with the platform's segmentation design.

The authors are explicit that this is a model result, not a field measurement: there is no claim here that any specific real platform runs fuzzy segmentation today, and no claim about the exact magnitude of any price premium in a live market. What the model offers instead is a mechanism, rigorously derived, for how a rational, profit-maximizing platform would respond to a privacy-regulated world with a large anonymous population, and a prediction, worth testing empirically, that the response would not be uniformly kinder prices for people who chose privacy. That prediction cuts directly against the policy narrative that privacy regulation straightforwardly helps consumers, and it does so through ordinary profit-seeking rather than through any violation of the rules.

What This Means Before the Next Privacy Rule Gets Written

For a platform operator, the model is close to a playbook: fuzzy segmentation is a real, legally available lever once your anonymous population is large enough to be worth pooling deliberately, and the Northeastern-era "member discount" pattern shows the underlying logic already has real-world precedent, just not yet at the scale or legal footing the model describes. For a regulator or privacy advocate, the uncomfortable implication is that giving consumers a bigger, cleaner off-switch for data sharing does not, by itself, guarantee that those consumers end up better priced; the welfare effect depends on exactly how platforms are permitted to re-segment the users who flip that switch, which is a market-design question, not merely a consent-design one. For the consumer checking the "do not track" box, the honest takeaway is less comforting than the marketing around privacy tools usually suggests: anonymity protects your identity. It does not, on its own, protect your price. Whether privacy regulation leaves you better off depends less on whether you were allowed to disappear, and more on which group the platform decided to make you disappear into.