Today, I submitted a comment on the Federal Trade Commission's Proposed Enforcement Policy Statement regarding personalized pricing. I argue that the Commission rightfully acknowledges personalized pricing's potential for consumer deception and offer suggestions for how to best enforce the policy.
I thank the FTC for welcoming comments on its recently proposed Enforcement Policy Statement Regarding Personalized Pricing. I am a research fellow for the Foundation for American Innovation, a think tank focused on ensuring that innovation aligns with American values and protects citizens’ constitutional rights.
Personalized pricing, also known as first-degree price discrimination, uses data to sell identical or nearly identical products at different prices under identical or nearly identical cost conditions. Practically, this translates to charging customers differently based on their attributes rather than market conditions. General outrage over unexpected personalized pricing has forced companies such as Instacart, Target, and Amazon to reverse course, demonstrating that this can be a materially misleading omission for at least a significant minority of consumers.
I recommend the following adjustments and clarifications in the final statement:
- Emphasize that the risk of deception is greatest in markets with few sellers and where products cannot be resold, since this is where omission is most likely to mislead consumers to their detriment.
- Encourage a simple transparency statement with pre-drafted language. Do not set disclosure of data types as the standard for compliance, since total pricing transparency can facilitate tacit collusion in at least some cases. Emphasize that dynamic pricing does not require disclosure.
- As other comments have recommended, the Commission should exempt the need to disclose introductory rates, negotiated prices, and loyalty or rewards programs whose only data use is purchasing frequency with the seller. Consumers do not expect introductory rates to last indefinitely and understand negotiated prices are personal, while loyalty members reasonably expect rewards to vary based on purchase quantity and frequency (but only on that basis).
- Base allegations of unfairness on use of assessed vulnerability or protected class rather than on higher prices. Perfect price discrimination usually raises prices for some customers, even where the price for all but one consumer is marketed as a discount. A measure of unfairness based on higher prices inappropriately treats all personalized pricing as suspect, even when it is clearly disclosed.
- Clarify that consent to a company's use of data for pricing also permits the firm to sell or share access to that data and its insights for the same purpose. If disclosure has a chilling effect on firms selling data to third parties, it will allow dominant firms with rich data pools to insulate themselves from competition.
Personalized Pricing Has Ambiguous Economic Effects
As the policy statement points out, the economic effects of employing algorithms that consider personal attributes to set prices is understudied. Personalized pricing expands the market by compensating companies for offering discounts to some customers with higher payoffs from other customers. Definitionally, first-degree price discrimination reduces consumer surplus, which is the difference between a consumer's willingness to pay and the price.
In many cases, personalized pricing increases overall welfare by helping firms recover high fixed costs and expanding access to more price-sensitive customers. It may also intensify competition as firms target discounts toward rivals' loyal customers. However, under a monopoly, personalized pricing can allow a firm to extract the whole consumer surplus without attracting competition. This is because the incumbent's ability to target discounts to disloyal customers while maintaining high prices for loyal ones deters competition, even if barriers to entry are low. A monopolist's data advantage can also enable it to better personalize product bundling and fidelity rebates in a way that excludes rivals.
In oligopolies and markets with differentiated products, whether first-degree price discrimination encourages fiercer competition or tacit collusion is fact-specific and remains an open question in the literature. For this reason, it is more accurate to say that the more complicated the personalized price scheme, the more difficult to assess whether it hurts consumers rather than to say, as the policy statement asserts, the more complex the price setting, the less likely consumers benefit. Within this context, the FTC's mandate requires it to inform Congress of the economic effects of personalized pricing, assess whether pending mergers facilitate more perfect price discrimination to the detriment of competition, and, relevantly, prosecute where personalized prices are unfair and deceptive.
Focus on Markets with Few Sellers, Goods Tethered to Customers
The policy statement asserts that consumers expect the price on the shelves of retail stores to be the same for any other consumer simultaneously shopping at the same store. This is true, and much state legislation focuses on the price-setting for physical goods and services. Large retailers do often use the reams of data collected on their customers to vary prices through targeted coupons, but to avoid the kind of consumer harm that animates the policy statement, the Commission should focus on markets with few sellers where products and services are tied to specific customers. Often this translates to digitally mediated markets rather than traditional retailers.
To sustain first-degree personalized pricing requires three market conditions. First, it requires the ability to discriminate among customers. As the Commission notes, this capacity is ever-increasing. The average American would be surprised to learn that Home Depot operates a thriving marketing platform, that last year Kroger netted $1.5 billion from its "alternative profit" business (of which selling customer data is the key driver), or that ads constitute the majority of Amazon's e-commerce profits.
Second, personalized pricing requires sufficient market power to set prices. Marginal cost is independent of the buyer, thus in a competitive market, sellers continue to benefit from undercutting the personalized price of a competitor until the price reaches marginal cost for all buyers. In practice, geography limits the available number of sellers, and consumers gravitate toward familiar sites, such that even in markets with relatively undifferentiated products, sellers still have some price setting power.
Third, personalized pricing requires limited arbitrage. Significant price variance among customers only makes sense when there is a weak or nonexistent resale market. Physical goods can be resold. Personalized pricing expands the market to buyers who would otherwise be priced out; these are exactly the consumers willing to go through the hassle of reselling the goods sold to them at a discount. Perishable, inferior, and Giffen goods (low-income necessities whose demand increases with price) are all exemptions from this rule, as are goods that require significant storage. Nonetheless, the resale capacity of physical goods usually offers consumers a bulwark against first-degree price discrimination.
These conditions—which, when combined, facilitate the perfect price discrimination that eliminates consumer surplus—are much more prevalent in digitally mediated markets and markets where products are tied to their purchasers. Digital goods, like a streaming subscription or software access, and physical services linked to a customer's account have no resale capacity and sometimes operate in markets with only a handful of sellers. Hotel accommodations, car rentals, and apartment leases are all examples of markets that, although not always purchased digitally, offer limited arbitrage and therefore should be prioritized. The final statement should emphasize the risk of deception in these types of markets specifically, since this is where omission is most likely to misleads consumers to their detriment. Where FTC authority stops short, such as in regulating airlines, it can nevertheless highlight price discrimination abuses to the relevant bodies, such as the Department of Transportation.
Simple Disclosures Better than Exhaustive Transparency
Disclosures can easily become a box-checking exercise for companies and a nuisance to customers. Simple transparency statements, like the one required by New York's Algorithmic Pricing Transparency Act, which requires sellers to use specific language to disclose whether personal data was used to set a price, more meaningfully inform customers than lengthy privacy notices. While simple disclosures are appropriate, listing the basis for price-setting and the data types that contributed to it is impractical in some markets. Rideshare platforms, for example, consider countless personal data points whose disclosure would lead consumers to miss "battery percentage" in the laundry list of other legitimate factors. Total pricing transparency could also facilitate algorithms in colluding to set supracompetitive prices without intent or evidence. The more difficult it is to discern the basis of the pricing decisions of a rival company's algorithm, the harder it is for another algorithm to converge on a shared pricing strategy.
In some markets, disclosures that exclude the types of data used to set prices may meaningfully mislead the consumer as to the breadth of personal information employed. However, given both the possibility of aggressive compliance overwhelming the customer and of complete price transparency encouraging tacit collusion, the final statement should not set the listing of data types and the basis of personalization as an industry-agnostic expectation.
Insofar as a simple disclosure raises the hackles of consumers, it encourages companies with straightforward data usage to volunteer the data types they use as a way to reassure consumers. The Commission's 2012 definition of personal data as data that "can be reasonably linked to a specific consumer, computer, or other device" remains a good benchmark, as evidenced by the New York law's alignment with it. Crucially, this does not require disclosure when "over 65" is considered alongside weather conditions to set prices for a rainy day special with a senior discount, since the overlap of those two demographics still cannot reasonably be indirectly linked to an individual. The Commission should emphasize that disclosure is not necessary for dynamic pricing based on locality, weather, time of day, and other features that are specific but not personal. A narrow definition of personal data ensures that it translates to a meaningful warning to consumers rather than a fact-of-life acknowledgement.
Use of Purchasing Frequency Not Always Deceptive
Merely raising prices for continued customers based on their purchase history is not deceptive unless the program states or obviously implies it rewards loyalty. In some industries, offering an introductory rate that raises over time is common and encourages frequent switches among competitors despite switching costs. The final statement should specify that introductory rates are not deceptive because consumers do not expect the price used to overcome switching cost and hassle will remain the price indefinitely.
Loyalty programs are not unfair or deceptive, and an aggressive pursuit of the FTC mandate should not fault companies for a long-accepted marketing tactic. However, customers understand loyalty programs to reward purchasing frequency and quantity, not to include loyalty among a host of other unlisted considerations. Loyalty schemes that vary prices on a basis other than that expectation are deceiving customers if they do not clearly disclose that other variables contribute to personal prices. Conversely, the final statement should specify that customers enrolled in a rewards program expect price differences based on purchasing frequency with the seller and, if this is the only personal data used, a disclosure is not necessary to avoid deception.
Negotiated Prices and Price Matching Do Not Violate Section 5
The policy statement should clarify that the FTC does not consider negotiated prices a potential violation of Section 5. Consumers understand negotiated prices are personal by their nature. Negotiation empowers consumers to address prices that creep higher after repeat purchases, like internet provision. Negotiation also allows consumers who are highly committed to a particular seller to nonetheless lower costs through their purchasing power. When consumers strongly prefer one product over another supplier's very similar one, suppliers set a high uniform price to maximize profit from the captive market segment. Negotiated prices allow large but loyal buyers to talk prices down.
Unfairness Best Measured by Inputs, Not Outputs
The policy statement asserts that higher prices based on personal data may be a substantial injury. If unfairness is based on whether a consumer pays a premium instead of receiving a discount, firms will simply raise the sticker price and offer individualized discounts. The problem with linking any subset's higher prices to substantial injury is that it offers no metric or test to weigh the market expansion of personalized pricing against the injury to consumer welfare it causes.
Firm use of protected classes or vulnerability indices is a better starting place for establishing when personalized prices cause unavoidable harm that outweighs whatever countervailing benefit it causes. A person's immutable or protected characteristics and his vulnerability given circumstances unrelated to market considerations cannot be reasonably avoided and cause substantial injury. The Commission's examples of hotels raising prices for funeral attendees or retailers charging home invasion victims more for security cameras both fall under this type of regulatory focus.
Confirming that firms do not use these criteria is more difficult than simply assessing whether their algorithms raise prices for some customers (they do). However, if the FTC were to set the former as an expectation for avoiding unfairness, it would encourage firms to adopt more straightforward pricing mechanisms whose inputs can be verified.
Privacy Can Sometimes Hinder Competition
The policy statement argues that businesses may violate Section 5 of the FTC Act if they do not sufficiently verify consumers consented to their data being used for the purpose of personalized pricing. If a company discloses its own use of data for pricing, then sells or shares the data to third parties without disclosure, one may consider this an omission. However, in order to violate Section 5, the omission must be to the detriment of the consumer. Counterintuitively, minimizing the entities who can access customer data may preserve a modicum of privacy, but it can often harm competition. Data exclusivity can fortify market dominance and deter new entrants, especially in markets where data is integral to the product. To the extent larger data pools are a competitive advantage, the FTC should be less concerned with whether firms disclose that they are selling data volunteered by customers and more concerned by restrictions on data resale. Given the competition concerns arising from data moats, the Commission should clarify that consenting to a company's use of data for pricing also permits the firm to sell or share data access and its insights for the same purpose.
Personalized Pricing Defies Conventional Antitrust Thinking
In The Antitrust Paradox, Robert Bork defined the "whole task of antitrust" as the effort to improve allocative efficiency without damaging productive efficiency. Oddly, perfect price discrimination flips the conventional monopolistic balance of those ideals, such that a monopoly or collusive set of sellers can improve allocative efficiency by supplying exactly the amount of goods demanded, while stunting or harming productive efficiency by diverting resources to assess willingness to pay. Personalized pricing offers new meaning to Senator John Sherman's comment, “It is sometimes said of these combinations that they reduce prices … but all experience shows that this saving of cost goes to the pockets of the producer.” Yet better-informed sellers can more effectively meet market demand while wasting less. The task before the Commission in this new algorithmic era is to prevent unfairness, deception, and unreasonable restraints of trade while still encouraging sellers to compete by targeting rivals' loyal customers and expanding market access. The policy statement succeeds in acknowledging the faults of personalized pricing, but appropriately addressing those faults will prove more difficult.
