The Perils of Personalized Pricing

Fixed prices for goods and services is a relatively recent invention. In the 19th century, Quaker merchants argued that haggling was morally wrong—it was unfair for two customers buying the same good to walk away having paid different amounts. John Wanamaker, a Quaker and the founder of Philadelphia’s Wanamaker’s department store, took it one step further, inventing the price tag by writing a price on a rectangle of cardboard and attaching it to each item in his store.

We’ve lived with fixed prices since then, but the legacy now appears to be fading, as online shopping combined with vast troves of consumer data ushers in an era of personalized prices.

“Online retailers can record your transaction history, buy information from data brokers, track social media posts to understand brands you like, and from this learn a whole lot about your preferences,” says Jidong Zhou, professor of economics at Yale SOM. “Based on that, they can come up with models that are nearly specific enough to describe individual consumers and how much each of those consumers is willing to pay.”

But who benefits as fixed prices fade: companies or consumers? A new study by Zhou and Andrew Rhodes of the Toulouse School of Economics finds that, in aggregate, personalized prices benefit consumers in markets for goods that most consumers buy. Consumer welfare is hurt by personalized pricing, however, for niche products that are purchased by a smaller number of consumers, or when certain companies have a stranglehold on the consumer data that’s available.

These results construct a bridge between two longstanding economic theories. One, from the 1920s, argues that when a firm knows what each consumer is willing to pay in a monopolistic market, that firm will charge each consumer as much as it possibly can, extracting all the value. Another paper published a few decades later argues that these results reverse once a second competing firm is introduced: companies bid themselves down to the lowest possible price to woo consumers away from their competitor.

“Our paper is saying, wait a minute, let’s think about this problem a little more carefully,” Zhou says. First of all, consumers may decide not to buy a product in a given market. This is referred to as coverage: markets can have high coverage, in which most consumers buy something, or low coverage, in which few consumers do. Second, consumer preferences are often more complicated than a simple inclination for product A over product B, especially when there are more than two competing products in the market.

Zhou and Rhodes introduce several of these more realistic complexities into a model of transactions in which companies have full knowledge of consumer preferences. They find that personalized pricing tends to create winners and losers: for those consumers who have a similar preference for all products, firms will compete for them fiercely; while for those who strongly favor one product over others, the favored firm can charge them a high price. In particular, contrary to what one might expect, rich consumers are not always losers as they may have a similar preference for all products. They also find that personalized pricing benefits consumers, on average, in markets with high coverage—that is, markets in which most people buy the product—especially for products that are relatively inexpensive to produce. In contrast, for products that are expensive to produce and so purchased by relatively few consumers, personalized pricing harms consumers on average.

Zhou gives the following hypothetical to illustrate why this is the case: Assume a certain widget in a high-coverage market costs $4 to produce. A consumer is willing to pay $10 for a version of the widget from Firm A and $6 for a widget from Firm B—a difference of $4. If Firm B knows this, it will lower its price nearly to the cost of production in order to try to get the consumer’s business. Because the difference in the consumer’s willingness to pay is $4, Firm A will then lower its price to $8 in order to remain competitive. Regardless of which company they choose, the consumer can be better off in this case than when there is no personalized pricing.

This outcome starts to shift, however, as the cost of production creeps up. If, for instance, the cost of manufacturing widgets is $7, Firm B can’t lower its prices far enough to attract the consumer described above, as they aren’t willing to pay more than $6 for their widgets. So Firm A can charge the full $10.

“As the cost of production becomes higher in a given market we start to see more of this type of consumer,” Zhou says. “As that happens, one company may gain power over consumers and what was once a competitive market moves closer to a monopoly.” With this shift toward a monopolistic market, companies that know a lot about consumer preferences are able to extract a great deal of value.

These results are exacerbated in cases where there is an imbalance in data ownership. A company like Amazon knows far more about its consumers than the local hardware store, which makes it capable of targeting consumers with prices in a way the local hardware store cannot. In these cases of information asymmetry, Zhou says, we should be “more suspicious” about the effect of personalized pricing on consumer welfare.

For policymakers who think about the growing place of personalized prices in the marketplace, Zhou suggests a more attentive consideration of a given market’s contours. If coverage is high, personalized pricing will likely benefit the average consumer. But if coverage is low (e.g., because the cost of producing the good is high), or consumer data is held unevenly, “then personalized pricing can be bad for consumers.”