Will Banning Personalized Pricing Work?
As AI makes it easier for businesses to tailor prices to individual customers, Maryland recently became the first state to prohibit the use of personal data in setting prices. We asked Yale SOM economist Jidong Zhou whether such restrictions are likely to work as intended—and whether they will benefit consumers.
Setting prices differently for different customers has been a practice for a long time. Are AI tools making personalized pricing more common or more effective?
Yes. Firms have often tried to charge different customers different prices, but AI makes this practice easier, cheaper, and more precise. In the past, price discrimination often relied on fairly broad categories: business versus leisure travelers, coupon users versus non-coupon users, students versus non-students. Today, firms have access to much richer information. A consumer’s browsing behavior, purchase histories, app usage, location data, social media activity, and even items left in online shopping carts can all leave useful digital traces. AI tools make it much easier to analyze this kind of scattered data and generate predictions on consumer preferences. In economic terms, this moves firms closer to “first-degree price discrimination,” in which each consumer could be charged a different price based on their preferences. We are not fully there, of course, but the direction is clear: pricing is becoming more data-driven and more individualized.
Firms could raise their posted sticker prices and then offer individualized discounts to selected consumers through emails, apps, or loyalty programs. Economically, this achieves the exact same result as personalized pricing.
AI also makes it much cheaper to carry out personalized pricing at scale. Airlines, hotels, and ride-hailing platforms already adjust prices as demand changes, and algorithmic pricing makes these adjustments faster and more individualized. Similarly, online retailers and subscription services can test different discounts, send coupons to shoppers who seem likely to leave without buying, or offer special retention deals to customers who appear likely to cancel. Digital platforms can change targeted prices or offers almost instantly, so the practical cost of using personalized pricing has fallen dramatically.
The push to regulate personalized pricing reflects some discomfort with the practice. What does your research say about how this kind of pricing affects consumers overall?
A common concern is that personalized pricing, compared with uniform pricing, allows firms to extract more surplus from consumers. This concern goes back at least to Pigou’s classic analysis more than a century ago: in the textbook monopoly case, if a firm knows exactly how much each consumer is willing to pay, it can charge each person that amount and leave consumers with no surplus. This is the standard reason people worry about personalized pricing.
But that conclusion depends on two strong assumptions: that pricing is perfectly personalized and that the firm has monopoly power. Once we relax either assumption, the effect on consumers becomes more complicated. For example, in a competitive market, personalized pricing can actually help some consumers. In particular, consumers who view competing products as close substitutes may benefit because firms have strong incentives to undercut each other to win their business. By contrast, consumers with strong brand preferences, or those who do not shop around, may be worse off, because firms know they are less likely to switch and may charge them more.
My own research emphasizes that the overall effect of personalized pricing is therefore nuanced. It depends on market conditions, including production costs, the degree of competition, and of course also the distribution of consumer preferences in the population. In general, personalized pricing is more likely to benefit consumers in aggregate when costs are lower, when there are more competitors, or when more consumers have weak preferences across competing products.
Can a ban like the one proposed in Maryland be effective? Is there a better approach?
A ban like Maryland’s may address some consumer concerns, but it is unlikely to fully stop personalized pricing, even if we assume that personalized pricing is bad for consumers—which, as discussed above, is not always true. One potential loophole is that firms could raise their posted sticker prices and then offer individualized discounts to selected consumers through emails, apps, or loyalty programs. Economically, this achieves the exact same result as personalized pricing, but legally, it remains entirely permissible under the bill’s exceptions for promotions, loyalty programs, retention offers, or consent-based discounts. As a result, the ban may simply push firms to reframe personalized high prices as personalized discounts without changing the underlying practice.
Regulating personalized pricing is difficult in practice. To evaluate whether such a regulation is working, policymakers may want to monitor how affected firms adjust their posted prices after the law takes effect. If regular prices rise while individualized discounts become more common, the law may have changed the form of personalized pricing rather than its substance. More generally, policymakers may want to regulate the conditions under which personalized pricing is used, rather than trying to ban it outright. For example, they could require firms to disclose when personal data affect prices or discounts, and to keep records that regulators can audit. Of course, these measures also have practical limits: disclosures may be too vague to be useful, and audits of complex pricing algorithms may require substantial technical expertise. Still, this approach may be more workable than a simple ban, because it could curtail predatory “price gouging” while still preserving legitimate, targeted discounts that can benefit consumers in competitive markets.