Spotlight

I. Introduction and Summary of Recommendations

The International Center for Law & Economics (ICLE) is a nonprofit, nonpartisan research center based in Portland, Oregon. In comments to the Federal Trade Commission (FTC) in 2014,[1] 2018–19,[2] and 2022,[3] and to the Canadian Competition Bureau in 2025[4] and 2026,[5] we have explained why price differences that expand output and lower prices for price-sensitive consumers should not be deemed presumptively harmful. We have made the same point in work on the Kroger/Albertsons merger,[6] state grocery-pricing laws,[7] food-delivery fees,[8] and the FTC’s proposed enforcement policy statement on personalized pricing (Proposed Statement).[9]

The Proposed Statement says that, where consumers reasonably expect prices not to vary with their personal data, a seller that personalizes prices should disclose that fact, the basis for the personalization, and the types of data used.[10] Omitting any of the three is “likely to constitute an unfair or deceptive act or practice in violation of Section 5.”[11] Yet the same document reports that “the effects of personalized pricing on consumers are unclear.”[12]

The Commission thus proposes to presume unlawful a practice whose prevalence it does not know and whose effects, by its own account, could benefit or harm consumers. Section 5(n) requires the Commission to weigh the costs of enforcement mistakes in unfairness matters,[13] and a presumption can substitute for evidence only where theory predicts a high probability of harm.[14] Where theory predicts mixed effects, a presumption turns the agency’s uncertainty into the regulated firm’s liability.

The two possible errors also impose different costs. A seller that drops targeted discounts to avoid a disclosure dispute takes those discounts from the consumers who would have received them, and market forces may not restore them.[15] A seller that personalizes prices without disclosure, in a market where consumers can compare offers, faces rivals that can discount to win its customers. Disclosure mandates produce the first kind of loss nearly as readily as prohibitions do: A mandate raises the cost of the conduct it covers relative to conduct it does not, and firms may respond by changing what they do rather than by becoming transparent about it.[16] The Proposed Statement moves from “unclear” effects to “likely” illegality without weighing these errors.

The FTC’s deception policy recognizes that omissions can mislead consumers. But a consumer’s belief alone cannot establish that a seller conveyed a misleading message. The Proposed Statement’s second deception theory finds deception when a consumer “reasonably believes” a price is static or widely offered and the seller “fails to disclose” that it is personalized.[17] This assumption is unsubstantiated. A displayed price tells a buyer what she may pay; standing alone, it does not say what everyone else pays. The Commission’s International Harvester decision declines to treat such “pure omissions” as deceptive.[18]

Unfairness requires substantial injury that consumers cannot reasonably avoid and that countervailing benefits do not outweigh. The 11th U.S. Circuit Court of Appeals called those conditions “mandatory” in FTC v. Corpay Inc.[19] The Proposed Statement applies none of them to its seven hypothetical examples. Having called the practice’s effects “unclear,” the Commission cannot simply conclude that it is nevertheless “injurious in its net effects.”[20]

The Proposed Statement cites six economic works for the claim that “the more sophisticated personalized pricing practices become, the less likely consumers are to benefit.”[21] None tests disclosure, and none supports the claim as stated. Where the papers find consumer losses, they trace them to market structure, a seller’s pricing power, or unequal access to data among rivals. In each of the three empirical studies, most consumers pay less; in two, the market grows.[22]

The Proposed Statement also assumes that the benefits of personalization may “be realized without concealing” it.[23] But a notice that lets buyers avoid a higher price changes the prices a seller can profitably offer. A seller forced to discount either for everyone or for no one will discount less. The Commission has not examined how its proposed notice would change prices.

The notice’s second and third elements compound the problem. They require a seller to explain how it set a price—more information than consumers generally care about or use, and more than International Harvester contemplates for mandatory disclosure.[24]

The Proposed Statement improves on the 2022 advance notice of proposed rulemaking that preceded it.[25] It acknowledges that Congress has not authorized the Commission to prohibit personalized pricing,[26] drops the term “surveillance pricing,”[27] and proceeds through guidance, as we urged.[28] But useful guidance would explain how the Commission will assess materiality, injury, countervailing benefits, and reasonable avoidability in the cases it describes. The Proposed Statement announces a conclusion instead.

A document that tells every seller what to disclose also operates as a disclosure rule, whatever its disclaimer of legal effect.[29] Firms will plan as though it binds them, without the protections of Section 18 rulemaking.[30]

The Commission should withdraw the Proposed Statement and proceed case by case, as Corpay shows it can. If the Commission nonetheless issues a statement, it should preserve the concessions the proposal already makes and adopt the six revisions laid out in Part VII:

1. Withdraw or qualify the third economic claim and expand the evidence. Withdraw the claim that more sophisticated personalization is less likely to benefit consumers, or identify the conditions under which more detailed consumer information can cause harm.
2. Clarify what a price display represents. A displayed price, standing alone, does not represent that every consumer sees the same price. Whether a particular presentation conveys that claim depends on evidence about the market in which it appears.
3. Confine the deception theory to representations. Strike the “reasonably believes” theory, retain the first theory, and use extrinsic evidence to establish what consumers understood an alleged implied claim to mean.
4. Replace the three-element prescription with a safe harbor. A clear and conspicuous statement that prices may vary with customer data should defeat a claim of implied uniformity.
5. Identify practices the Commission is not targeting. Name individualized discounts, coupons, loyalty pricing, retention offers, and negotiated prices as practices the Commission does not regard as presumptively unlawful.
6. Analyze the examples or relabel them. For each of the seven hypothetical examples, identify the price that would have prevailed without personalization, the substantial injury and whether consumers could reasonably avoid it, and the gains and losses for consumers and competition.

Only after resolving these legal and evidentiary questions should the Commission test which of the avoidance steps it describes that a notice would enable consumers to take, just as it tested its mortgage disclosures.[31] It should separately assess how sellers would change their prices in response.

Finally, any prescription of a disclosure’s wording, placement, or format constitutes a rule, which requires Section 18 rulemaking procedures.

II. The Effects of Personalized Pricing Depend on Market Conditions

The Proposed Statement makes four concessions that any final statement should retain. First, Congress has not authorized the Commission to prohibit personalized pricing outright. Second, personalized pricing “is a long-established norm in some markets,” including car sales, residential real estate, and markets where buyers and sellers negotiate.[32] Third, prices vary for reasons unrelated to personalization, including local supply and demand, regional taxes, and the individualized risk assessments used in insurance and credit. Fourth, real-world personalization is imperfect. An analysis that takes these concessions seriously must account for competition and the limits of consumer data and pricing tools.

Ample evidence that personalized pricing generally harms consumers might make those circumstances less consequential. The studies the Commission cites provide no such evidence. They report mixed effects on consumer welfare and show that the circumstances often determine the outcome.

The Commission begins its economic account with a monopoly benchmark, then concedes that the benchmark does not describe most markets. Perfect personalized pricing “without business competition would allow a monopolist to capture the entire consumer surplus.”[33] But the Commission immediately acknowledges that “real-world personalized pricing would be less than perfect in its ability to identify consumers and determine their willingness to pay and, in the vast majority of circumstances, would have to contend with competitive pressures that limit pricing power.”[34] That is the Statement’s most significant concession: In actual markets, imperfect information and competition may limit or eliminate the harm predicted by the monopoly benchmark. Without empirical findings about particular markets, the Commission’s case rests heavily on what six economic studies say about consumer harm under realistic conditions. As discussed below, those studies do not support many of the Statement’s claims or proposals.

A. Targeted Pricing Can Intensify Competition

Personalized prices can intensify competition by letting a seller compete for a rival’s customers without cutting prices for all its existing buyers. Under uniform pricing, a seller must extend a price cut meant to attract new customers to buyers who would have paid the old price. That makes price cuts more costly and less likely. Targeted offers avoid that cost, and rivals can respond with targeted offers of their own, multiplying the potential benefits.

The literature on price discrimination—charging different buyers different prices—supports this dynamic, although its effects depend on the market. Thisse and Vives show that it can lower prices and profits in a duopoly market with two differentiated sellers.[35] Corts shows why the result depends on how firms rank customer groups and the groups’ relative importance.[36] In his model, price discrimination can lower prices in every segment or raise them in every segment. A calculation based on a monopolist cannot, however, predict what happens when competing sellers can respond. As the antitrust literature puts it, price discrimination “implies neither that a market is not performing ‘competitively’ nor that it would perform more ‘competitively’ without price discrimination.”[37]

B. Imperfect Data and Competition Limit Personalized Pricing

Imperfect data and competing offers limit a seller’s ability to personalize prices and the effects it can plausibly achieve. Data are noisy and often dated, and sellers estimate pricing models from samples rather than individual buyers. A consumer who can buy elsewhere—or decline to buy—will not pay more than a purchase is worth compared with her best alternative. Better data can improve a seller’s estimate and change the prices it can profitably charge, but the result remains an estimate.

In some markets, prices must depend on personal information. As then-Commissioner Ferguson observed:

This practice is ancient and commonplace for many products and services. The prices of insurance policies or of credit, for example, necessarily turn on a consumer’s personal details because the price must reflect the risk of providing the good or service to the consumer.[38]

The Commission also concedes that it does not know how widely sellers use personalized pricing.[39] The available evidence suggests limited use in the markets studied. A 2018 European Commission mystery-shopping study of 160 e-commerce sites in eight member states found personalized price differences in only 6% of tests. The median difference was less than 1.6%.[40]

Alberto Cavallo found no evidence that large U.S. multichannel retailers varied online prices with a shopper’s location or browsing history. He attributed that restraint to retailers’ fear of “antagonizing their customers.”[41] Stefano DellaVigna and Matthew Gentzkow found that most large U.S. food, drug, and mass-merchandise chains charged nearly uniform prices across stores despite substantial differences in local demographics and competition. The median chain forwent about $16 million a year in profit by doing so.[42]

Customers’ reactions already constrain a practice that firms forgo despite its potential returns. That restraint does not make enforcement unnecessary in every case, but it raises questions about what a general presumption would accomplish and what it would cost to apply one to the discounts that make up much of the practice.

C. The Cited Studies Do Not Support a General Presumption of Harm

The Proposed Statement cites six works in support of its three economic claims: Personalized pricing is likely to increase business profits; it benefits some customers while harming others; and more “sophisticated” practices are less likely to benefit consumers.[43] The Commission does not claim that personalized pricing has no benefits, causes net harm, or is prevalent. Its first two claims are accurate as far as they go, but leave out consequential findings. The cited literature does not support the third.

The six studies differ in what they can establish. Three are empirical studies but compare the observed setting with modeled pricing alternatives. Jean-Pierre Dubé and Sanjog Misra conduct field experiments on a business-to-business hiring platform and model the seller as a monopolist. Benjamin Reed Shiller estimates what personalization might have done for Netflix’s DVD-by-mail service in 2006, when the company had some pricing power but did not personalize prices. Nicholas Buchholz and coauthors model a Prague-based taxi-hailing platform whose drivers bid in an auction that does not rely on personalized data.

Two other papers are theoretical models without empirical data: Andrew Rhodes and Jidong Zhou’s model of competing firms in an oligopoly setting, and Rosa-Branca Esteves and Francisco Carballo-Cruz’s unpublished model of entry over two periods. The sixth work, by Ginger Zhe Jin, Liad Wagman, and Mengyi Zhong, summarizes the literature without adding a model or data of its own.

None empirically studies personalized pricing in a consumer retail market like those in the Proposed Statement’s examples. None tests the proposed disclosures, and none of the three empirical studies measures what happens when competitors respond.[44]

The Appendix to these comments includes a table that summarizes each paper’s setting and findings, how the Commission uses it, and how well it supports the Commission’s claims. Briefly, with respect to the claims, the studies show the following:

• Profits: The three empirical studies find profit gains in the settings they examine. But in Dubé and Misra and in Buchholz et al., most of the gains come from correcting or centralizing a uniform price rather than personalizing it. One theoretical paper shows that personalization under competition can reduce industry profits.
• Winners and Losers: The conclusion that some consumers gain while others lose is true, but also trivial. Because consumers are heterogeneous along indefinitely many dimensions—both observable and unobservable—every pricing practice, including uniform pricing, benefits some consumers while harming others. That differential alone says nothing about the size or direction of the net effect. Nor does it show whether the distribution of gains and losses is undesirable along dimensions policymakers may care about, such as income.
• Sophistication: The cited studies do not establish the third claim that more “sophisticated” practices are less likely to benefit consumers. The Proposed Statement does not define the term. Some forms of personalization can reduce consumer surplus, depending on competition, access to data, and how sellers use the data. But several cited works find that broader access to more data can benefit consumers. Where those works identify harm, they often trace it to market power or unequal access to data, rather than to the use of more data itself.

Assessing welfare also requires separating new sales from the redistribution of gains on sales that would have occurred anyway. A higher price on a purchase that still occurs transfers surplus from buyer to seller. A discount that induces an additional purchase that would not otherwise have occurred can create gains from trade. At the same time, price discrimination can also reduce welfare if it allocates a fixed supply to buyers who value it less. The traditional monopoly analysis developed by Robinson, Schmalensee, and Varian and surveyed by Stole weighs the effect on output against this allocation effect. Under its assumptions, output must increase for total welfare to rise. But neither higher output nor higher total welfare establishes that consumers collectively benefit.[45]

1. Seller Effects: Profit Estimates Omit Rivals’ Responses

The three empirical studies the Commission cites estimate profit gains from personalization, but none evaluates how rival sellers would respond. Rivals could offer deeper discounts, retain customers who would otherwise switch, or both, reducing the estimated gains.

1. Netflix. Shiller estimates that Netflix could have increased profits by 12.99% in 2006 by setting markups based on subscribers’ web-browsing histories, compared with 0.27% using demographics alone.[46] Of course, a tiered menu that used no consumer data would have increased profits by 22.48%.[47] Shiller also finds that the top 0.5% of consumers account for 42% of the estimated gain.[48] Notably, Netflix did not personalize its prices in 2006; these results come from a simulation.
2. ZipRecruiter. Dubé and Misra estimate that personalized pricing would raise ZipRecruiter’s profit by 19% relative to its optimal uniform price and by 86% relative to the price it actually charged.[49] But correcting the uniform price accounts for 55 of the 86 percentage points. That gain does not come from personalization.[50]
3. Liftago. Buchholz and coauthors model Liftago, a Prague app on which licensed taxi drivers bid for each ride. They simulate pricing counterfactuals for one route during evening hours.[51] Compared with actual pricing through driver competition, a single platform-set price for all riders would cut driver profit by 76% and increase the share of riders who walk away from 37% to 53%. Adding personalization would raise platform profit by a further 7.82% and driver profit by 13.44% relative to platform-set uniform pricing. Drivers would recover only a small fraction of what they lost when the platform took control of pricing. Most of the estimated effect comes from that change in control, rather than the use of consumer data.[52]

The studies model some customer responses, but none measures how competing sellers would respond. Dubé and Misra expressly exclude a competitive response from other platforms.[53] In the abstract, that dynamic should uniformly reduce the estimated effects. The distinction matters for Liftago. Buchholz and coauthors note that Prague’s regulatory setting is “different from most U.S. cities in which there is nearly free driver entry into the ride-hail market with firms such as Uber and Lyft.”[54] Where drivers can switch easily between competing platforms, a rival’s better offer would limit how much profit a platform could shift from drivers to itself.

The size of that competitive effect is uncertain, but its direction is clear: It would reduce the platform’s estimated profit gain. The papers therefore show that personalization can increase profits in the settings they model, while leaving the effect of competitive responses unanswered. In the limit case, competition could reduce any expected profit increase to zero. Indeed, although the Commission does not cite it for this point, Andrew Rhodes and Jidong Zhou model a market in which, when the market is fully covered, personalized pricing by all firms reduces industry profit.[55]

2. Consumer Effects: Outcomes Depend on the Market and the Benchmark

As the Commission acknowledges, the cited papers find that some consumers gain from personalized pricing while others lose. None of the three empirical studies, however, examines an ordinary retail buyer looking at a posted price in a U.S. store—the kind of buyer the Proposed Statement would affect.

The Commission also omits a finding common to all three studies: Most consumers gain surplus—the value they receive beyond what they pay—or receive lower price offers. In Buchholz et al., 62.5% of consumers gain surplus, while losses fall chiefly on riders least sensitive to price and waiting time.[56] Dubé and Misra estimate that roughly three-fifths of potential customers would receive a lower offer under personalization than under the profit-maximizing uniform price.[57] Shiller finds that “most consumers receive lower prices when prices are personalized, and hence the majority are slightly better off.”[58]

Rhodes and Zhou’s theoretical model shows why the broader result depends on the market and on which competitors have access to data.[59] When the market is fully covered— all potential consumers buy at the prevailing uniform price—and all firms personalize, industry profits fall and consumers collectively gain relative to uniform pricing. That result cuts against the Proposed Statement’s implicit presumption of consumer harm.

The result can reverse when the uniform price excludes some consumers. Personalization then brings more buyers into the market, but firms may collect enough from existing buyers to raise industry profits and reduce aggregate consumer surplus. In the authors’ numerical examples, industry profits begin to rise when uniform pricing excludes roughly 10%–20% of consumers. Aggregate consumer surplus begins to fall when it excludes roughly 20%–30%. Those thresholds depend on costs and the number of firms, while the model holds firms’ information fixed at a perfect level.[60] The remedies the authors consider—data sharing or restrictions on firms with more data—address situations in which only some firms can personalize. Neither is a consumer-disclosure remedy.

The benchmark also matters when interpreting the empirical studies. In Buchholz et al., replacing Liftago’s actual driver auctions with a single price set by the platform would cut consumer surplus by 33% and increase the share of riders who walk away from 37% to 53%.[61] But loss comes from replacing driver competition with centralized pricing. Moving from centralized uniform pricing to centralized personalized pricing reduces consumer surplus by a further 2.25%. Thus, the vast majority of the consumer-surplus loss comes from the platform taking control of pricing, rather than using consumer data.[62]

Dubé and Misra find a find a similar distinction. Moving ZipRecruiter from its actual uniform price of $99 to a profit-maximizing uniform price of $327 would raise profits by 55% and halve the share of prospective customers who buy, from 25% to 12%. Again, this entails no personalization. The paper measures the consumer-surplus effects of personalization against that $327 price rather than against $99, because they expect a maturing firm to charge the optimal uniform price even without personalization. Against that benchmark, theoretically optimal personalized pricing reduces total consumer surplus by 25%. The $499-capped personalization ZipRecruiter implemented reduces it by 2.04% while increasing profits by more than 8%.[63]

Personalization matters in both studies, but their headline changes cannot all be attributed to it. In Buchholz et al., personalization accounts for a small fraction of the consumer-surplus loss. In Dubé and Misra, the implemented policy reduces consumer surplus by about 2% against a uniform-price benchmark that itself halves the share of prospective customers who buy.

3. Distributional Effects: Consumer Gains, Losses, and Market Expansion

The studies’ most important finding is that personalized pricing affects consumers differently. In Rhodes and Zhou’s model, when the prevailing uniform price excludes some buyers, personalization both lowers aggregate consumer surplus and brings previously priced-out consumers into the market.[64] The studies find that the net effect of personalized pricing can sometimes be a reduction in aggregate consumer surplus.[65] They also find that personalization can expand access, that most consumers in the settings studied gain or receive lower offers, and that the beneficiaries tend to be more sensitive to price.[66] The Commission acknowledges that personalized pricing produces winners and losers, but it does not explain what follows from that dichotomy.

The FTC’s Unfairness Policy Statement treats a practice as unfair only if it is “injurious in its net effects.”[67] But the Proposed Statement does not claim that personalized pricing lowers aggregate consumer welfare. Instead, it points to particular consumers who pay more. That raises a question about distribution—who gains and who loses—that the cited studies help answer.

In Buchholz et al., losses fall chiefly on riders least sensitive to both price and waiting time, a profile unlike that of a passenger in an emergency. In Dubé and Misra’s study, larger employers hiring full-time workers with benefits tend to pay more, while smaller firms hiring part-time staff are most likely to gain. In Shiller’s study, subscribers whose browsing histories predict the highest willingness to pay face higher prices.[68] These findings do not support the impression left by the Statement’s examples that those paying more are the consumers least able to choose otherwise.

Section 5(n) also requires the Commission to weigh any “substantial injury to consumers” against “countervailing benefits to consumers or to competition.”[69] In the studies cited, however, the measured losses are small. Against a uniform-price benchmark, the estimated consumer-surplus losses are 0.5% in Shiller and 2.25% in Buchholz et al. For the capped personalization ZipRecruiter implemented, Dubé and Misra estimate a 2.04% loss against the optimal uniform price, although their estimate for uncapped personalization is much larger: 25% against that benchmark.[70]

Whether those numbers suggest substantial injury or not, they are static, partial-equilibrium sums with equal dollar weights. They do not show who gains or loses, and they value new participation only by the estimated surplus of each additional buyer. A measure that also considers distribution may reach a different result and the sign can even reverse. Dubé and Misra find that under inequality-averse welfare measures, personalization is preferred to uniform pricing “because the allocative benefits outweigh the decline in total surplus,” and that those benefits accrue mainly to smaller firms.[71]

Similarly, the assessment of benefits “to competition” permitted by Section 5(n) contemplates a separate calculation from the static consumer-effect balancing, one that includes the dynamic consequences of a larger market (e.g., entry, scale, investment, spillovers), not just consumer surplus.[72] Market size and access may, in fact, be more important than the consumer-surplus sign. Output expansion is antitrust law’s own measure of competitive benefit,[73] and traditional monopoly analysis treats output growth as necessary, though not sufficient, for price discrimination to increase total welfare.[74] Two of the three empirical studies find that personalization expands participation: About 7% more requests result in a ride in Buchholz et al., and about 3% more households subscribe in Shiller.[75]

The Proposed Statement instead treats a measured aggregate loss as presumptive injury without accounting for the majority who gain, the consumers newly able to buy, or the benefits of market expansion. That applies a welfare criterion the Statement has not explained or defended and does not complete the balancing Section 5(n) requires.

4. The Studies Do Not Support the Sophistication Claim

The Proposed Statement’s cited authorities do not establish that “the more sophisticated personalized pricing practices become, the less likely consumers are to benefit.”[76]

The Commission cites Dubé and Misra to suggest that the effects on consumers depend on the data available to a seller—and, implicitly, that more data means less consumer welfare.[77] The study finds a more complicated relationship. The pages the Commission cites give an example in which more detailed data raise consumer surplus and conclude that “increasing the granularity of the consumer data available to a firm can increase consumer surplus and even reduce deadweight loss.”[78] The authors report “a nonmonotonic relationship between the granularity of data and consumer surplus under personalization.”[79] In other words, more detailed data do not consistently help or harm consumers.

Across 62 potential personalization schemes, the authors find that “allowing the firm to target on more granular data can be good for consumer surplus,” that “granularity per se does not harm consumers,” and that “granting the firm access to more information is not per se worse for the consumer, as it can lead to segmentation schemes that allocate more surplus to the consumer.”[80] They warn that “overregulation of the types of data that firms can use for personalized pricing purposes could exacerbate rather than offset some of the harm to consumers.”[81] All 62 schemes do lower total consumer surplus relative to the optimal uniform price, but the size of that loss does not move consistently with the granularity of the data.[82] That finding undercuts the Statement’s general claim about sophistication.

The Commission cites Rhodes and Zhou for the finding that “when only some companies in a market can personalize prices, consumers can be worse off than when either all or no companies do so.”[83] That summary is accurate, but it does not support the sophistication claim. The harm in the model comes from unequal access to personalized data across firms, not from increasingly detailed data.[84] Indeed, the finding cannot depend on the amount or sophistication of the data, because the model gives every personalizing firm perfect information about each consumer. The authors leave the effects of differing amounts of information for future work.[85]

Buchholz and coauthors provide the Commission’s best evidence for its sophistication claim, and it is thin. The paper finds that, relative to a uniform price set by the platform, personalizing prices based only on riders’ sensitivity to waiting time leaves aggregate consumer surplus unchanged. Adding their sensitivity to price lowers it by 2.25%.[86] Yet nearly all of the consumer loss comes from the platform taking control of pricing. Replacing driver auctions with a uniform platform price costs riders a third of their surplus. The platform’s other pricing policies, from a menu using no rider data to full personalization, differ by less than four percentage points.[87] As the authors put it, “the losses in consumer surplus from personalized pricing are mostly due to the platform’s ability to centralize pricing and not its knowledge of consumer preferences.”[88]

Adding information thus has either no effect or a small negative effect on aggregate consumer surplus in the settings Buchholz and coauthors model. It also brings important benefits: 62.5% of riders are better off, the average fare falls, and about 7% more requests result in a ride. Indeed, the authors reach the opposite conclusion from the Commission’s about total welfare: “Conditional on the platform exercising this pricing power, the more information the platform incorporates into its pricing policy, the larger the overall welfare gains.”[89] Meanwhile, higher fares fall on the riders least sensitive to price.[90] The paper primarily demonstrates that a dominant platform with unconstrained pricing power can take surplus from its least elastic customers. That is much stronger support for the Commission’s market-power admission than for its sophistication claim.

Jin, Wagman, and Zhong’s literature review likewise does not establish the Commission’s general claim.[91] First, it is a literature review with no original model or data; it summarizes and assesses other economists’ work and is not independent evidence. Second, the authors warn that “as a firm with market power collects extremely fine-grained personal data, it can approach this limit case and potentially hurt all consumers.”[92] The claim is confined to the market-power setting, and the paper’s next sentence adds that “the concern about first-degree price discrimination can be mitigated by market competition.”[93]

Even where the paper claims that “access to finer-grained consumer information” allows a firm to “selectively undercut its rivals’ customer base,”[94] its support for the Statement’s sophistication claim is limited. As the paper notes, the result is a function of “raising switching costs for targeted consumers and creating data asymmetry among firms and consumers.”[95] Asymmetry can arise at any level of sophistication (although presumably more sophisticated data in some firms’ hands makes asymmetry more likely).

The review’s conclusion emphasizes the conditions the Proposed Statement overlooks. It says that “in competitive markets where multiple firms have access to consumer data, personalized pricing can intensify competition and benefit consumers,” and that greater pricing sophistication holds “the potential for both significant consumer benefits and serious market failures.”[96] Even where a firm has market power, “the consumer benefits can exceed the costs in net, depending on the market and product specifics.”[97] The authors’ first policy recommendation addresses the Statement’s central defect: “Regulatory frameworks should focus on market structure and competitive dynamics rather than treating all forms of personalized pricing equally.”[98]

The remaining paper the Commission cites for its sophistication claim, Esteves and Carballo-Cruz’s theoretical duopoly model, attributes harm to the asymmetry of access to data, not the sophistication of the data used. Consumer harm arises when an incumbent’s exclusive data prevent a rival from entering, leaving the incumbent as a monopolist.[99] When entry does occur, aggregate consumer surplus is highest if both firms personalize, lower if only the incumbent does, and lowest if neither does.[100] In other words, consumers collectively benefit as more firms personalize.[101] The authors conclude that, when entry costs are low, “price discrimination intensifies competition and overall consumer surplus is above its non-discrimination counterpart.”[102] In those markets, they find “there are no reasons to block price discrimination.”[103]

Other research confirms that more information can raise or lower consumer surplus, under both competition and monopoly. Mark Armstrong and John Vickers show that data identifying captive buyers, who consider only one seller, let firms charge those buyers more while competing harder for buyers who compare offers. Under the paper’s demand conditions, discrimination harms consumers collectively when the firms are sufficiently similar and can benefit them when the firms are sufficiently asymmetric.[104] In Brian C. Albrecht’s model, complete information available to all firms can maximize aggregate consumer surplus even though sellers charge loyal buyers the full value of the product: Competition drives prices for contested buyers down to the cost of serving them.[105] Dirk Bergemann and coauthors show that perfect price discrimination leaves consumers with zero surplus, their worst feasible outcome. But they also identify partial market segmentations that leave consumers better off than uniform monopoly pricing would.[106]

Different information and different market conditions determine whether greater sophistication harms or helps consumers. Better estimates of willingness to pay, the ability to identify a rival’s customers, and wider access to the same data can affect prices and welfare differently. Buchholz and coauthors find one refinement that slightly lowers consumer surplus and another that leaves it unchanged. The competitive models identify conditions under which more informative or more widely available data benefit consumers.

Evidence that personalization can sometimes benefit consumers is not, by itself, an objection to disclosure; nor is disclosure obviously an answer to firms’ variable ability to harm consumers with personalized pricing. The studies compare pricing policies and access to information; the Proposed Statement concerns what sellers tell buyers. The Commission has yet to ask how its proposed notices would affect consumer choices, firms’ access to data, their incentives to change prices, and the resulting gains and losses.

There is, however, no basis for the Proposed Statement’s broad, unqualified claim about sophistication. A final statement should withdraw the claim or identify when more detailed information can harm consumers. The relevant conditions would include the extent of competition, the amount and distribution of data, differences among firms’ ability to personalize, and the ease of entry.[107] The Commission must also specify the pricing benchmark against which it measures harm and separately assess how disclosure would change consumer choices and seller prices.

III. Consumer Expectations Alone Cannot Establish Deception

The Proposed Statement advances two theories of deception. Under the first, retailers “may deceive consumers in violation of Section 5 when they represent, expressly or by implication, that a price is static or widely offered when in fact it is personalized.”[108] That theory follows settled deception law by asking what the seller’s express or implied representations conveyed.

Under the second, retailers “may similarly deceive consumers when a consumer reasonably believes that a price for a good or service is static or widely offered, and the merchant fails to disclose that the price is in fact personalized.”[109] This theory rests on a consumer’s expectation without identifying what, if anything, in the seller’s presentation conveyed that prices were uniform.

A. A Displayed Price Does Not Promise Uniformity

A price tag makes a narrow, useful offer: A buyer may purchase an item at the displayed price. Standing alone, it does not say that every other buyer sees the same price at the same moment. Whether a seller’s broader presentation conveys that message depends on how consumers understand it, and that understanding can differ across markets.

U.S. shoppers already encounter many forms of price variation: clipped, downloaded, and register-generated coupons; loyalty-card prices posted beside shelf prices; student, senior, and military discounts; airline fares that change between search and booking; retention offers available only when a subscriber tries to cancel; and negotiated prices for cars, homes, and professional services. Prices for insurance and credit products, as the Commission itself says, “necessarily vary from purchaser to purchaser.”[110]

The Commission acknowledges that personalized pricing is established in some markets. In others, it infers that consumers reasonably expect “that the price they see for a product or service is the same price that any other consumer at the same place and time would see.”[111] But the absence of established personalization does not show that consumers read a displayed price as a promise of uniform treatment. In a store with a loyalty program, that expectation is already false, and loyalty-card holders know it. The Commission might call a loyalty price “widely offered,” but that answer concedes the point: Consumers may expect to see the terms of a price without expecting everyone to pay the same amount.

The Proposed Statement also leaves the expectation undefined. How many consumers must hold it, and how consistently?[112] A new customer may expect something different from a customer placing his 100th order. Expectations can change as pricing practices change. In a 2015 report on differential pricing, the Council of Economic Advisers said disclosure rules “should be based on a comparison of the compliance costs and the expected benefits from increased transparency.” It predicted that buyers “may come to expect personalized pricing in specific online settings, just as they do offline.”[113] A trigger based on current expectations risks freezing a norm that, as the Commission notes, varies by market. It also gives sellers little guidance when they set prices.

The Commission can still bring a deception case when a seller’s presentation conveys a uniform-price claim and the evidence establishes that claim. The Deception Policy Statement says the Commission “will often be able to determine meaning through an examination of the representation itself.” In other situations, it “will require extrinsic evidence that reasonable consumers reach the implied claims,” such as copy tests, surveys, or consumer testimony.[114] An alleged implied promise of uniform pricing calls for that evidence. The policy’s presumptions of materiality—for express claims, intended implied claims, and information about central product characteristics such as cost—apply only after the Commission identifies a claim or omission likely to mislead. They do not establish that a price display conveys a promise of uniformity.[115]

B. Deception Requires More Than an Assumed Expectation

The FTC’s deception test has three elements: A representation, omission, or practice must be likely to mislead; the Commission assesses it from the perspective of a consumer acting reasonably in the circumstances; and it must be material, meaning likely to affect the consumer’s conduct or decision.[116]

1. A Pure Omission Does Not Convey a Claim

An omission can satisfy the first element only when the context gives the silence a misleading meaning. In International Harvester, the Commission distinguished a half-truth—“where a seller fails to disclose qualifying information necessary to prevent one of his affirmative statements from creating a misleading impression”—from a “pure omission,” or “a subject upon which the seller has simply said nothing, in circumstances that do not give any particular meaning to his silence.”[117]

The Commission explained that it “does not treat pure omissions as deceptive.” Otherwise, deception would expand “virtually beyond limits” because “[i]ndividual consumers may have erroneous preconceptions about issues as diverse as the entire range of human error.” Moreover, deception law provides no cost-benefit analysis of the disclosure that a finding would compel.[118] The Commission therefore held that a manufacturer’s failure to disclose a fuel-system hazard was unfair, but not deceptive, allowing “a cost-benefit analysis” before ordering disclosure.[119] The Proposed Statement’s second theory treats exactly such a preconception as a representation.

International Harvester recognized an exception when silence occurs “under circumstances that constitute an implied but false representation.” That can include “ordinary consumer expectations as to the irreducible minimum performance standards of a particular class of good,” such as the expectation that a car can reach highway speed.[120] The Proposed Statement’s second theory appears to invoke that category. But the category is defined by an expectation that virtually every buyer of the class of goods holds and that the seller knows its product does not fulfill. The Commission has acknowledged that personalized pricing is well established in some markets. Expectations of uniform pricing therefore vary by market and consumer. The Statement turns an expectation it has not identified into a representation it need not prove.

If a consumer’s mistaken preconception is enough to create a duty to speak, a duty to speak, sellers must anticipate every consequential misunderstanding a buyer might hold. Sellers cannot survey that set of assumptions, and enforcers cannot define it in advance.[121]

The Proposed Statement’s citation to P. Lorillard Co. v. FTC, introduced with a “cf.” signal, does not establish a general duty to disclose the fact, basis, or data sources behind personalized pricing.[122] Lorillard promoted Old Gold cigarettes as lowest in nicotine, tar, and resin while selectively invoking a Reader’s Digest test. Read as a whole, the test found the differences among brands practically insignificant and did not support a claim that any brand was less harmful. The court therefore found that Lorillard’s own partial affirmative statements created a misleading net impression. The problem “was not that it did not print all that the Reader’s Digest article said, but that it printed a small part thereof in such a way as to create an entirely false and misleading impression.”[123]

Lorillard did not hold that silence alone is deceptive or that firms must disclose everything consumers might consider relevant. The other case cited in the same footnote, FTC v. Standard Education Society, concerned salespeople who falsely told prospects that a set of books was free and that its regular price exceeded the offered price—affirmative statements the Commission found false.[124]

2. The Materiality Rationale Remains Untested

The Proposed Statement grounds materiality in steps an informed consumer might take to avoid a personalized price. Consumers unaware of personalization “cannot take steps to avoid the higher prices that may result from it, such as using a virtual private network or private browsing functionality, choosing a different retailer whose prices are static or widely offered, rather than personalized, or simply declining to complete the transaction.”[125]

A virtual private network masks the address a server uses to infer location, but does nothing about personalization keyed to a logged-in account, a loyalty number, or a stored payment card—the personalization the Commission’s model disclosure describes as a price derived “from data about that consumer’s previous purchases from the same retailer through the same login account.”[126] Private browsing can clear cookies after a session, but it leaves purchase history held by the seller intact. A consumer would need to know which other retailers use uniform prices before switching to one; a notice from the current seller does not supply that information. And a consumer can decline any purchase at a price she considers too high, personalized or otherwise.

Materiality then depends on something narrower than the Commission claims: Would learning that a price was personalized cause consumers to shop elsewhere or delay buying? The Commission can test that proposition but has not done so. Nor does a presumption that information about cost is material resolve the issue. Materiality is the third element of deception; a presumption about it cannot establish the first—an act or practice likely to mislead.

C. Colgate-Palmolive Concerns False Claims, Not Silent Prices

FTC v. Colgate-Palmolive Co. does not establish a general duty to disclose personalized pricing.[127] The passage the Commission cites says a seller deceives consumers by falsely claiming that a product usually sells at an inflated price and is now available at a special discount.[128] The Court offered that example by analogy. The case itself concerned a television commercial that presented an undisclosed mock-up as an actual product demonstration.

Even so, the discount example turns on an affirmative false claim. The seller represents that the higher reference price is genuine and the lower price is a bargain. The Court used it to show that a false claim can be material even when buyers receive a product worth what they paid. It did not hold that a seller must disclose every fact that influenced its price.

A retailer that advertises “$40, marked down from $80” makes a false claim if the item did not ordinarily sell for $80. A retailer that simply lists an item for $40 makes no comparable claim about a discount, price uniformity, or its pricing method. The Proposed Statement initially respects that distinction by tying possible liability to an express or implied representation that a personalized price is static or widely offered. Its second theory drops the requirement by treating a consumer’s belief as sufficient without a seller-created representation. Colgate-Palmolive does not support turning a rule against false price claims into a general duty to disclose personalized pricing.

IV. The Proposed Statement Does Not Establish Unfairness

Section 5(n) limits the Commission’s authority rather than listing factors for it to balance. It provides that the Commission “shall have no authority” to declare a practice unfair “unless the act or practice causes or is likely to cause substantial injury to consumers which is not reasonably avoidable by consumers themselves and not outweighed by countervailing benefits to consumers or to competition.”[129] As the 11th U.S. Circuit Court of Appeals explained in Corpay, “[a]s ‘unless’ conveys, the three factors that follow it are mandatory for a finding that a practice is ‘unfair.’”[130]

The Proposed Statement, however, casts each requirement as a possibility: Injury “may be” substantial, consumers “may not” be able to avoid it, and benefits “may also” be realized another way. It then presents seven hypothetical examples without applying the statutory requirements to any of them.

A. The Examples Do Not Apply the Unfairness Test

The Proposed Statement closes with seven hypothetical examples of personalized pricing that, “without adequate disclosure,” would raise Section 5 concerns. A food-delivery service quotes a higher price to a customer it believes cannot leave home. A grocery chain charges more for milk to a household with several children. A hotel charges more to a traveler it believes is attending a funeral. A rideshare service charges more to a rider with no competing apps installed, and more again to one headed to a medical facility during an apparent emergency. A retailer charges a recent crime victim more for a security camera and raises its website price for a shopper standing in its parking lot.[131]

The examples differ sharply on whether a consumer can avoid the higher price. The shopper in the parking lot can check the shelf price at little cost. A rider without a competing app may be able to download one, though that takes time and effort. Indeed, the example’s surcharge depends on the seller’s belief that the rider is less likely to switch. By contrast, a passenger headed to an emergency room may have no practical chance to compare fares. The Proposed Statement groups these situations together despite their differences on the very question Section 5(n)’s avoidability requirement asks.

A finding of substantial injury also needs a comparison price—the price the seller would have charged but for personalization. The Proposed Statement supplies neither that price nor a method for finding it. The comparison is not necessarily the lowest price offered to anyone, but the price the seller would have set had it not personalized. It might be a uniform price or a price based on different customer information. Data that could support a surcharge for a household could, for example, also support a family discount. The grocery example does not say which outcome would occur.

A defined comparison price would also distinguish a loss in total consumer surplus from a redistribution of gains among consumers. Higher prices injure the buyers who pay them. If personalization also lowers prices for others, those buyers gain. Section 5(n) then requires the Commission to weigh the injury against benefits to consumers or competition. It cannot do so without estimating who gains, who loses, and by how much.

The statute requires a finding that injury is “not outweighed by countervailing benefits to consumers or to competition.”[132] Yet the Commission says the effects of personalized pricing on consumers are unclear.[133] Instead of estimating those effects, the Proposed Statement assumes that “[a]ny benefits to consumers or competition from personalized pricing may also be realized without concealing the fact and nature of the personalized pricing.”[134] Whether sellers would keep personalizing under a disclosure requirement, and which benefits would survive, is a question about how firms respond to compliance costs. The Commission should test those responses using its Section 6(b) record.

The FTC’s Unfairness Policy Statement adds two constraints absent from the proposal. The Commission “is not concerned with trivial or merely speculative harms,” and “[e]motional impact and other more subjective types of harm . . . will not ordinarily make a practice unfair.”[135] It also “will not find that a practice unfairly injures consumers unless it is injurious in its net effects.”[136] The Proposed Statement cannot establish net injury while describing the practice’s effects as “unclear.” To the extent it relies on consumers’ reported sense that personalized pricing is unfair, it relies on the kind of subjective harm that the Unfairness Policy Statement and the legislative history of Section 5(n) exclude.[137]

The Commission also identifies three ways consumers might avoid a higher price. It says they may lack the tools to “modify their behavior to avoid triggering higher prices, dispute or correct inaccurate information collected about them that is leading to higher prices, or avoid the collection of that data in the first place.”[138] A notice might help consumers compare offers or limit data collection without explaining every feature of a seller’s pricing model. Disputing an error, though, requires knowing which information the seller used. Limiting collection requires knowing where the information came from. Categories such as “purchase history” or “location” answer neither question.

This does not mean the Commission should demand a more detailed notice. It means the avoidability analysis cannot depend on steps the prescribed notice would not enable. A final statement should identify which steps the notice would make possible and confine its analysis to them.

The proposal’s unfairness theory turns on concealment: Consumers may be unable to avoid a higher price “if the fact or nature of personalization of the price has been concealed by the retailer.”[139] Consider the rider headed to a medical facility during an apparent emergency. Telling him that the fare reflects data suggesting an emergency removes the concealment but leaves both his circumstances and the surcharge intact. The same problem applies to a hotel guest believed to be attending a funeral. Both examples depend on the possibility that some personalized prices are unfair even when fully disclosed—a question the Commission expressly sets aside.[140]

FTC v. Corpay Inc. shows what an unfairness case supported by evidence can look like. The 11th U.S. Circuit Court of Appeals upheld summary judgment for the Commission on a claim that Corpay charged fleet-card customers fees it had not disclosed.^[141] The record showed what customers were told and what the company knew about their confusion. A telephone survey by the Commission’s expert found that Corpay told only 7.02% of affected customers about all the fees before charging them. The survey also measured advance notice for each fee.

Internal emails and complaints showed that sales staff omitted the fees. Corpay kept its terms and conditions offline for three years and instructed employees not to send electronic copies. Eight internal surveys from 2016 through 2020 showed that its fees confused and surprised customers. Corpay’s expert testified that customers bought no less fuel after a fee was disclosed. The court found that testimony did not answer the evidence that Corpay had failed to disclose the fees in the first place.

None of the Proposed Statement’s seven examples supplies comparable evidence about what consumers were told, what sellers knew, or how many consumers could have avoided the charge.

B. Credit and Insurance Laws Do Not Create a General Pricing-Disclosure Duty

The Commission invokes the Fair Credit Reporting Act (FCRA) to support a duty under Section 5 to disclose the reasons for a personalized price. It says the FCRA requires notice “not only when an adverse action is based on a consumer report (for example, the charging of a higher price), but also of the specific basis for that adverse action.”[142] Section 1681m(a) does not require that explanation. The Proposed Statement’s own footnote describes the provision more accurately: It requires an adverse-action notice and tells consumers how to obtain a free copy of their report “so they can identify and dispute any inaccuracies.”[143]

The notice must identify the adverse action; provide any credit score used, along with the key factors that adversely affected the score; identify the consumer reporting agency; and explain the right to a free report and to dispute its contents. The key factors explain the score, not the creditor’s decision. FCRA also requires “a statement that the consumer reporting agency did not make the decision to take the adverse action and is unable to provide the consumer the specific reasons why the adverse action was taken.”[144]

The reason-giving duty the Commission describes comes from the Equal Credit Opportunity Act (ECOA), which it does not cite. ECOA entitles an applicant subject to adverse action to a statement of reasons, which suffices “only if it contains the specific reasons for the adverse action taken.”[145] Even there, a creditor may discharge the obligation by notifying the applicant of the right to request those reasons, rather than volunteering them.[146] The duty applies to specified credit decisions, including a denial or revocation, a change to the terms of an existing arrangement, or a refusal to grant credit on substantially the terms requested.[147]

FCRA’s other notice provisions are equally specific. If a lender uses a consumer report to offer credit “on material terms that are materially less favorable than the most favorable terms available to a substantial proportion of consumers,” Section 1681m(h) requires a risk-based pricing notice. That notice identifies the report’s role in setting the terms, the reporting agency, and the right to a free copy and contact information for obtaining it. If the lender used a credit score, the notice must also provide the score and specified accompanying information, including the factors that adversely affected it.[148]

FCRA does require disclosure of “the nature of the information” behind certain decisions based on information from someone other than a consumer reporting agency, including a denial of credit or an increase in its charge. But the duty arises only after a consumer makes a written request within 60 days.[149] Disclosure to a consumer who asks differs from a notice on every displayed price.

Congress enacted these targeted requirements. That suggests Section 5 did not already compel them across markets. The same point applies to the state insurance provisions the Commission cites: Each reflects a legislative or regulatory decision to require notice in a particular market.[150]

Section 5(n) allows the Commission to consider established public policies as evidence, but says “[s]uch public policy considerations may not serve as a primary basis for such determination.”[151] The Proposed Statement quotes that limit in footnote 21, then relies on FCRA and state insurance law as its only statutory support for a disclosure duty.[152] The FTC’s Unfairness Policy Statement says public policy most often supplies “additional evidence on the degree of consumer injury.” When the Commission relies on it heavily, the policy should be “widely shared, and not the isolated decision of a single state or a single court.”[153] Two state insurance statutes, one state insurance regulation, and a credit statute with narrower requirements do not establish the general duty the Proposed Statement proposes.

V. Disclosure Requirements Need Clear Limits and Consumer Testing

The Proposed Statement leaves unclear how much detail a personalized-pricing notice must provide. It calls for disclosure of personalization, its basis, and the types of data used.[154] Its example of an adequate notice describes a price “based on a consumer’s estimated willingness to pay derived from data about that consumer’s previous purchases from the same retailer through the same login account.”[155] That explanation identifies the purpose and source of the data without revealing the full pricing model.

If the Commission retains these three elements instead of adopting a safe harbor, it should clarify what satisfies them when prices depend on several data sources. That clarification should, however, stop at what a notice must convey. If the Commission goes on to mandate the format of the disclosure—its wording, placement, or presentation—the document ceases to be a statement of enforcement policy and becomes a disclosure rule, one that would require the procedures of Section 18 rulemaking.

Disclosing what a consumer will pay differs from explaining how the seller arrived at the price. Clear final prices, honest reference prices, and understandable loyalty terms help consumers compare offers.[156] The Proposed Statement’s second and third elements instead ask sellers to describe their pricing methods and the data they use.

In International Harvester, the Commission advised that “[i]n most cases it is appropriate to limit mandatory disclosure to those core aspects of a transaction that virtually all consumers would consider essential to an informed decision.”[157] Price clearly meets that threshold. Whether the model behind the price does is far less clear. Even a 2014 White House report drew this line, calling it “implausible for consumers to be presented with the full parameters of the data and algorithms shaping their online and offline experience.”[158] The Council of Economic Advisers’ 2015 follow-up described a narrower option: a rule “requiring disclosure that prices or search results may vary across users.” It said such a rule “should be based on a comparison of the compliance costs and the expected benefits from increased transparency.”[159]

A. More Disclosure Can Mean Less Useful Information

Petra Persson models how a firm can satisfy a disclosure mandate while defeating its purpose. The firm complies, then sends “an increasing swarm of irrelevant cues” that obscures the required information and leaves readers worse off.[160] Her model formalizes a prediction Frank H. Easterbrook and Daniel R. Fischel made in 1984: Firms “may cease disclosing some category of useful information and switch to some obfuscatory (but complying) information.”[161] Persson also shows how a firm told to disclose information in an easy-to-understand format can complicate its product until that requirement no longer helps.[162] She concludes that “simple disclosure rules can be completely impotent and have no impact on social welfare.”[163]

At the same time, the consumer who must read what the firm produces bears a cost of her own. Mandated disclosures “provide diminishing returns and increasing costs,” while “[o]ptimal disclosure is not maximal disclosure.” Yet commentators often “take for granted that increased disclosure is beneficial—and cheap.”[164] Additional information can “corrode[] the decision-making value of information” by forcing consumers to separate what matters from what does not.[165] The consumers least equipped to do that may be the very people a mandate aims to help.[166]

The Proposed Statement’s second and third disclosure elements ask a shopper at checkout to do exactly that. A notice saying a price reflects “estimated willingness to pay derived from data about that consumer’s previous purchases” describes a process.[167] It does not tell the shopper what she would pay elsewhere or what action to take. Nor would a better-drafted paragraph solve the broader problem. A single disclosure prescribed for every seller and market replaces judgments tailored to particular transactions with “a blunt ex ante standard of disclosure.” As Geoffrey A. Manne observes, “[t]here is no reason to believe that the costs imposed by the blanket standard, although different in nature, are any smaller in magnitude” than the costs it displaces.[168] The Commission should instead ask whether each added element improves consumers’ decisions sufficiently to justify its costs in attention and changes to sellers’ behavior. The Proposed Statement does not ask that question.[169]

The Organization for Economic Co-operation and Development (OECD) accordingly advises governments to “limit the prescriptiveness and extensiveness” of mandated disclosures to what is necessary. It also warns of an accumulation problem: Disclosures that seem defensible one by one can overwhelm consumers in combination. Quoting Omri Ben-Shahar and Carl Schneider—“[o]ne disclosure by itself may seem trivial, but en masse disclosures are overwhelming”—the OECD warns that “even a well-designed personalised pricing disclosure may be easily overlooked in the presence of a relatively more salient health warning.”[170]

Accumulation is a problem even at this Commission. The Rule on Unfair or Deceptive Fees governs how sellers display total prices for live-event ticketing and short-term lodging.[171] In March and April 2026, the Commission sought comment on rules for rental-housing fees[172] and for fees in online food and grocery delivery.[173] While each of those disclosures is confined to its market, a personalization notice would not be. It would apply to every seller that personalizes, joining the fee disclosure the Commission already requires, along with the privacy notices, cookie banners, and terms that already compete for the same checkout screen. The consumer’s attention to that screen does not grow with the number of notices on it.

Research by Lisa Robinson, Kip Viscusi, and Richard Zeckhauser suggests that generic warnings can weaken the force of warnings about serious risks. They found that 69% of surveyed consumers thought a generic Proposition 65 cancer warning on breakfast cereal conveyed a risk as great as a cigarette warning from 1966. “Taking cigarettes as the benchmark for wolves,” they write, “the Proposition 65 warnings are largely about puppies.”[174] A notice on every personalized price would likewise attach the same signal to a loyalty discount and to the emergency rideshare surcharge in the Commission’s example. The OECD’s review of online disclosures adds that consumers “might mistake, for example, neutral information for a warning, focus their attention unduly on disclosed information, rather than other product characteristics,” or react to it emotionally.[175]

Howard Beales and Todd J. Zywicki similarly warn that “disclosures may misdirect attention, if they imply that consumers should pay attention to information that has no real significance for them.”[176] A personalization notice could do just that for a shopper receiving a discount.

Anna Priester and coauthors, whom the Commission cites in support of disclosure, asked participants to imagine buying a digital movie and rate the fairness of the price they were told they received. Individually personalized prices drew lower fairness ratings than prices set for customer groups, though the difference was marginal in both studies.[177] Every participant was told the pricing rule, so the researchers did not test the effect of disclosing it. Their study therefore measures reactions to a described practice, not whether a notice helps consumers understand a price or act on it.

Finally, the academic case for algorithmic-pricing transparency is narrower than the Proposed Statement suggests. Oren Bar-Gill, Cass R. Sunstein, and Inbal Talgam-Cohen tie harm to algorithms that exploit consumers’ misperceptions. They describe pricing based on preferences and wealth as causing “limited harm,” and a disclosure that a seller charges wealthier consumers more as revealing a “(not particularly alarming) fact.”[178]

B. Consumer Testing Does Not Support the Proposed Notice

The OECD two personalized-pricing notices. Working for its Committee on Consumer Policy, Ireland’s Economic and Social Research Institute recruited 210 participants in Dublin and 210 in Santiago, Chile. Each completed eight purchase tasks on simulated retail sites. Some saw no notice. Others saw a statement that “[w]e used your personal data in calculating this/these price(s) for you.” A stronger version added that the price “may be higher or lower than for other customers.” The stronger notice increased stated awareness among Irish participants after they received a definition of personalized pricing, but awareness remained low. The study “failed to provide evidence of a significant impact of the disclosures on participants’ purchasing behaviour.”[179] Between 62% and 94% of Irish participants and between 85% and 96% of Chilean participants did not recall seeing a notice.[180] The authors concluded: “Neither disclosure was effective in protecting participants from paying more than they needed to.”[181]

While interesting, the experiment has two important limits. First, prices varied randomly rather than according to personal data, a limitation the authors identify themselves.[182] At the same time, 66% of Dublin participants and 51% of Santiago participants said personalized pricing should not be allowed.[183] The absence of a significant effect on purchases therefore does not show that participants approved of the practice. Second, results from simulated sites in two cities need testing before they can support a U.S. retail policy. The notices were prominent and repeated—likely more so than a notice at checkout[184]—yet most participants did not recall seeing them.[185]

Willem H. van Boom and coauthors tested how notice wording affected stated purchase intentions. They first studied 426 Dutch university students choosing among hotel rooms, then repeated the experiment with 186 U.S. participants recruited online.[186] Notices suggesting that personalization served the buyer’s interests—because the price might be lower than others paid or was “appropriate”—raised stated purchase intentions. In the U.S. study, the effect operated entirely through participants’ belief that the practice served their self-interest.[187] Notices also tended to raise stated purchase intentions compared with no notice.[188] It should be noted in this regard that, under the Proposed Statement, sellers would draft their own notices and choose the wording. The van Boom study suggests that wording, rather than disclosure alone, can shape consumers’ responses.

Allison Nicoletti and Christina Zhu studied roughly 22 million loan applications after the Truth in Lending Act–Real Estate Settlement Procedures Act (TILA-RESPA) Integrated Disclosure rule took effect. They compared covered mortgages with exempt open-end loans, which let borrowers draw credit repeatedly. Borrowers became more likely to withdraw covered applications or leave them incomplete, a result the authors interpret as consistent with lower information-processing costs. Approval rates also fell by about 5 percentage points. Compliance costs and frictions in selling loans to investors reduced credit availability.[189] The authors draw no conclusion about net welfare and call their findings “primarily descriptive.”[190]

Calorie labels can change purchasing behavior, although their average effect is small. A 2025 Cochrane review by Natasha Clarke and coauthors finds strong evidence of a small reduction in calories selected or purchased. The evidence on calories actually consumed is less certain.[191] Pasquale E. Rummo and coauthors found 24.7 fewer calories purchased per transaction across 2,329 Taco Bell restaurants. Most restaurants subject to labeling were in California. For restaurants outside California, the pooled estimate showed no statistically significant change over the three- to 24-month follow-up.[192]

These studies offer a narrower lesson than the Proposed Statement draws. Calorie labels are brief. Mortgage forms span several pages, but they use standardized fields and quantities that borrowers can compare when choosing a loan. A notice that prices may vary with personal data could be brief and standardized. An explanation of how a seller set a price may be longer and harder to standardize. It also gives consumers no alternative price to compare at the moment of purchase.

None of these studies establishes how consumers would use the Commission’s proposed notice at a U.S. retail checkout. The Commission should first resolve the legal and evidentiary questions discussed above and define the scope of any disclosure duty. If it proceeds, it should conduct or commission consumer testing comparable to its mortgage-disclosure research.[193] The tests should ask whether consumers understand the notice, can use it to compare offers, and make different purchasing decisions. Testing could inform a safe harbor or an eventual Section 18 rule. The Commission must separately assess how disclosure would change sellers’ prices and consumer welfare.

VI. Disclosure Would Change Sellers’ Pricing Decisions

The Commission cannot assume that personalized pricing’s benefits would survive its proposed disclosure requirement.[194] If a notice helps consumers avoid higher prices, sellers may lower those prices, raise discounted prices, change how they make offers, or narrow eligibility for discounts. The Commission must assess the resulting gains and losses across consumers before concluding that disclosure would preserve the benefits.

The Proposed Statement assumes that “[a]ny benefits to consumers or competition from personalized pricing may also be realized without concealing the fact and nature of the personalized pricing.”[195] That assumption treats sellers’ conduct as fixed and the notice as a costless addition. But it is a mistake to “treat regulated behavior as static” when regulation makes some conduct more expensive.[196] As Geoffrey A. Manne explains, “[d]isclosure regulation makes some forms of behavior more expensive relative to others. Rational actors will respond by shifting some conduct into the comparatively cheaper outlets. And these alternative behaviors may actually be less beneficial than the regulated, deterred behavior.”[197]

The size of that shift depends on how closely other practices can substitute for personalized pricing.[198] Sellers have many alternatives to an individualized discount that requires a three-part notice and may still invite enforcement scrutiny.[199] They can charge everyone one higher price, vary prices by location or time of day, offer coupons through intermediaries, or use bundles and loyalty tiers. They can also let consumers choose among offers. Buchholz and his coauthors describe one example: pricing based on a rider’s expected wait time rather than purchase history.[200] These alternatives may preserve some gains from personalization, but they may miss consumers who would have received a targeted discount. The losses will be hard to observe. The Commission may see fewer personalized prices without seeing the discounts that sellers stopped offering.[201] And the benefits cannot be assumed: “It is insufficient either to ignore unanticipated consequences or to presume that the only relevant effects will be the salutary, intended ones.”[202]

Price discrimination also depends on limits to buyers’ ability to obtain another buyer’s lower price. If a notice lets a buyer qualify for a discount at a cost below the price difference, the seller must account for that possibility when setting both prices. It may lower the higher price or raise the discounted price until seeking the discount is no longer worthwhile. Consumers need not actually change their behavior for this constraint to affect prices. An assessment that holds prices fixed would miss these effects.[203]

Where disclosure helps consumers in these models, it does so because sellers change their behavior. Xi Li and Krista J. Li find that making it cheaper for consumers to manipulate the data used in pricing can reduce consumer surplus.[204] In their model, sellers raise prices for buyers with low willingness to pay, while other buyers incur costs to obtain discounts. Disclosing the scope of data collection, by contrast, weakly increases consumer surplus in the same model. Sellers collect less data, consumers spend less effort manipulating it, and prices fall for buyers with low willingness to pay.

Consumers’ information choices can also change the offers competing firms make. S. Nageeb Ali, Greg Lewis, and Shoshana Vasserman study buyers who choose which verifiable information to reveal to each seller.[205] A buyer can obtain a lower offer from a rival by revealing information to that rival while withholding it from a preferred seller. Under the model’s demand assumptions, the authors construct an equilibrium in which every consumer type benefits relative to a market without personalization. Competition for the buyer drives those gains. A notice alone may give consumers no comparable control over their information. To apply this result, the Commission must identify what its notice would enable consumers to do and how competing sellers would respond.

Consumers who currently receive lower personalized prices could lose out if disclosure prompts sellers to withdraw those offers. Dubé and Misra estimate that roughly three-fifths of potential business customers would receive lower offers under personalization than under a profit-maximizing uniform price.[206] Those customers’ potential losses belong in the analysis. Whether the discounts would disappear depends on whether they would remain profitable after consumers respond to the notice.

Evidence on geographic pricing shows that a move toward common prices can affect consumers unequally. DellaVigna and Gentzkow find that food-store prices near the 10th percentile of demand elasticity within a chain are only 0.4% higher than prices near the 90th percentile. A model with common marginal costs predicts a difference of about 17%.[207] Their calibration also suggests that uniform pricing raises prices at stores serving the lowest-income decile by roughly 10% relative to stores serving the highest-income decile, compared with flexible pricing. The authors caution that competitive responses, including entry and exit, could change those results.[208] That caution reinforces the need to account for consumers who benefit from existing price differences.

Rules that push sellers toward uniform prices offer another warning. When sellers must extend a discount to everyone or no one, they may discount less. The Antitrust Modernization Commission reached that conclusion about the Robinson-Patman Act, which has exposed sellers to litigation over price differences and pushed them toward rigid uniform pricing.[209] The FTC has likewise alleged that Amazon’s “anti-discounting” tactics, “[b]y taming price cutters into price followers,” suppress price competition and deprive shoppers of lower prices.[210] Competition can produce individualized discounts, and a disclosure regime that raises the cost or legal risk of offering them weakens sellers’ incentives to do so. Research on pay transparency points to a similar risk: when individualized terms become visible, firms may narrow the differences and offer fewer favorable terms.[211]

The effect of personalization on consumers depends on how firms compete for different buyers. If a notice changes which customers a seller can attract with a discount, rivals may change their offers in response. The welfare comparison must therefore account for prices across competing sellers, rather than treating the avoidance of one seller’s higher price as the entire consumer benefit.

The Commission needs evidence on these pricing responses before concluding that disclosure would preserve the benefits of personalization. Its January 2025 interim staff perspective reports initial findings from the Section 6(b) study and acknowledges limits on their representativeness and on conclusions about effects.[212] Further aggregated evidence should show which consumers receive higher and lower prices, what the proposed notice would enable them to do, and how sellers would adjust their offers. The Commission should compare gains and losses across consumers under the resulting prices.

VII. Recommendations: Withdraw or Substantially Revise the Proposed Statement

The Proposed Statement concedes that personalized pricing’s effects on consumers are unclear. The literature and evidence do not support its broader inferences, and the Commission has not done the analysis needed to justify the proposed policy. The Commission should withdraw it. If it proceeds, we urge it to adopt the following recommendations:

Qualify the third economic claim and develop the evidence. Withdraw the claim that more sophisticated personalization is less likely to benefit consumers, or identify the conditions under which more detailed consumer information can cause harm. Build on the Section 6(b) release with aggregated evidence on pricing, consumer welfare, and disclosure’s likely effects. Once the Commission has addressed the evidentiary and legal concerns and clarified what it proposes to require, it should test whether consumers can use the notice to compare offers, correct information, or avoid higher prices. It should separately assess how sellers would change their prices and which consumers would gain or lose.

Clarify what a displayed price display conveys. State affirmatively that a displayed price, standing alone, does not represent that every consumer sees the same price at the same time. Whether a seller’s broader presentation conveys that claim depends on evidence from the market in which it appears.

Ground deception claims in established law. Strike the Proposed Statement’s second theory of deception, which finds that a seller deceives consumers whenever they “reasonably believe” a price is static or widely offered and the seller does not disclose personalization. Retain the first theory, which requires a seller’s express or implied representation. A false claim that a price is uniform or generally available can violate Section 5 if the Commission proves each element of the Deception Policy Statement’s three-part test. Where the claim is implied, the Commission should use extrinsic evidence to establish what consumers understood.

Provide a disclosure safe harbor. Replace the requirement to disclose the basis of personalization and the types of data used with a safe harbor: A clear and conspicuous statement that prices may vary based on customer data defeats a claim of implied uniformity. This would tell sellers how to avoid implying that prices are uniform without requiring an explanation of how they set each price.

Identify practices the Commission does not presume unlawful. Name individualized discounts, coupons, loyalty pricing, retention offers, and negotiated prices as practices the Commission does not regard as presumptively unlawful. That clarification would reduce the risk that uncertainty deters beneficial discounts. The protection should exclude cases where a seller inflates a nominal public price merely to present every individualized price as a discount. The public price must be offered in good faith to a meaningful number of consumers for a meaningful period. Within that limit, a personalized price no higher than the price generally offered at the same time should be safe.[213]

Analyze the examples or present them as questions. For each of the seven hypotheticals, the Commission should identify the relevant alternative price, explain the substantial injury and whether consumers can reasonably avoid it, and assess gains and losses across consumers and any countervailing benefits to competition. Otherwise, it should identify the examples as questions for further study. Corpay illustrates the evidence an unfairness claim requires: a customer survey, internal surveys and emails, complaint data, and evidence of what customers were told about each fee. The Commission also supported its deception claims with the advertisements and internal documents.[214]

Appendix: How the Proposed Statement Interprets the Economic Works It Cites

The Proposed Statement relies on six works to summarize the economics of personalized pricing.[215] This table identifies each paper’s methodology, the claim the Commission cites it to support, its relevant findings—including those the Commission omits—and whether it supports the Commission’s claim.

Page references are to the cited works; note numbers (n.8–n.13) are the Proposed Statement’s. Quotations from the Statement are at 3–4.

[1] Berin Szoka, Tom Struble, Geoffrey A. Manne & Ben Sperry, Comments of TechFreedom & the International Center for Law & Economics in Response to Big Data: A Tool for Inclusion or Exclusion? 11–12 (FTC Project No. P145406, 2014), https://www.laweconcenter.org/resources/comments-response-ftc-workshop-big-data-tool-inclusion-exclusion (“tailoring prices more closely to an individual’s willingness to pay means that more transactions will take place than otherwise would have, with more consumers having access to the product and more consumers’ desires being satisfied. And price differentiation generally means that the poor pay less, not more, for the same goods.”).

[2] Dirk Auer, Justin (Gus) Hurwitz, Geoffrey A. Manne, Julian Morris & Kristian Stout, Privacy, Antitrust, and the Economic Approach to the Regulation of Consumer Data 5–7 (ICLE Comment 4, FTC Hearings on Competition & Consumer Protection in the 21st Century, Project No. P181201, 2018), https://laweconcenter.org/resources/icle-comments-the-regulation-of-consumer-data; Int’l Ctr. for Law & Econ., The Weakness of Interventionist Claims: Summary Comments on the FTC Hearings on Competition & Consumer Protection in the 21st Century 10, 139–40 (2019), https://laweconcenter.org/resources/icle-comments-ftc-hearings-on-competition-consumer-protection-in-the-21st-century-summary (“there are many cases where personalized pricing reduces deadweight loss and increases access to a product or service for those with a lower willingness-to-pay”).

[3] Geoffrey A. Manne, Daniel Gilman & Kristian Stout, Comments of the International Center for Law & Economics: FTC Advance Notice of Proposed Rulemaking, Trade Regulation Rule on Commercial Surveillance and Data Security (Docket No. FTC-2022-0053, R111004, Nov. 21, 2022), https://laweconcenter.org/wp-content/uploads/2022/11/ICLE-Commercial-Surveilance-ANPR-Comments-v4.pdf.

[4] Geoffrey A. Manne et al., Comments on Canadian Competition Bureau Discussion Paper on Algorithmic Pricing and Competition (Int’l Ctr. for Law & Econ., Aug. 4, 2025), https://laweconcenter.org/wp-content/uploads/2025/08/Canada-Algorithmic-pricing-and-competition-1.pdf.

[5] Ian Adams, Eric Fruits & Brian Albrecht, ICLE Comments to Competition Bureau Canada on Examining the Food Supply Chain 9–12, 27 (Int’l Ctr. for Law & Econ. 2026), https://laweconcenter.org/resources/icle-comments-to-competition-bureau-canada-on-examining-the-food-supply-chain.

[6] Brian C. Albrecht et al., Food-Retail Competition, Antitrust Law, and the Kroger/Albertsons Merger (Int’l Ctr. for Law & Econ., White Paper No. 2023-10-17, 2023), https://laweconcenter.org/wp-content/uploads/2023/10/Kroger-Albertsons-Merger-Full-Paper.pdf.

[7] Eric Fruits, The Waterbed Effect and the Price of ‘Pricing Fairness’ (Int’l Ctr. for Law & Econ., Issue Brief No. 2026-03-17, 2026), https://laweconcenter.org/wp-content/uploads/2026/03/Grocery-Pricing-Legislation-2026.pdf.

[8] Eric Fruits, Too Much on the Menu: Disclosure Overload and Crowding Out in Food-Delivery Regulation (Int’l Ctr. for Law & Econ., Issue Brief No. 2026-08-21, Aug. 21, 2026), https://laweconcenter.org/wp-content/uploads/2026/08/Delivery-Pricing-Literature-Review-v3-2026.pdf; Brian C. Albrecht et al., Comments of the International Center for Law & Economics: Rule on Unfair or Deceptive Fees in Online Food Delivery Services (FTC Project No. P267101, May 18, 2026) [hereinafter Online Food Delivery Comments], https://laweconcenter.org/wp-content/uploads/2026/05/FTC-Delivery-Fees-2026.pdf.

[9] See Daniel J. Gilman, Fine Print for Every Price: The FTC’s One-Size-Fits-All Guidance, Truth on the Mkt. (Aug. 28, 2026),https://truthonthemarket.com/2026/08/28/fine-print-for-every-price-the-ftcs-one-size-fits-all-guidance; Alden Abbott, Your Price May Vary: The FTC’s Personalized-Pricing Puzzle, Truth on the Mkt. (Sept. 2, 2026), https://truthonthemarket.com/2026/09/02/your-price-may-vary-the-ftcs-personalized-pricing-puzzle.

[10] Fed. Trade Comm’n, Proposed Enforcement Policy Statement Regarding Personalized Pricing, Docket No. FTC-2026-1057, Doc. No. FTC-2026-1057-0001 (Aug. 19, 2026), https://downloads.regulations.gov/FTC-2026-1057-0001/content.pdf [hereinafter Proposed Statement].

[11] Id. at 1.

[12] Id. at 3.

[13] See Geoffrey A. Manne, Bringing Antitrust’s Economic and Institutional Limits to the FTC’s Consumer Protection Authority 12–13 (Int’l Ctr. for Law & Econ. 2014), https://www.laweconcenter.org/resources/bringing-antitrusts-economic-institutional-limits-ftcs-consumer-protection-authority (explaining that Section 5(n)’s requirements “serve to impose an error cost approach on unfairness questions, limiting both the likelihood and harm of erroneous over-enforcement”).

[14] See Geoffrey A. Manne, Error Costs in Digital Markets, in The GAI Report on the Digital Economy 33, 48 (2020), https://laweconcenter.org/resources/error-costs-in-digital-markets (“Rebuttable presumptions are sometimes employed as a cost-saving substitute for direct evidence when economic theory predicts a relatively high probability of competitive harm.”).

[15] See Adams, Fruits & Albrecht, supra note 5, at 2 (“If enforcement condemns lawful discounting, loyalty pricing, or algorithm-driven responses to market conditions, firms may abandon those practices across their operations to avoid liability. Market forces cannot readily restore the lost benefits.”).

[16] See Geoffrey A. Manne, The Hydraulic Theory of Disclosure Regulation and Other Costs of Disclosure, 58 Ala. L. Rev. 473, 477, 485–86 (2007) [hereinafter Manne, Hydraulic Theory] (“Disclosure regulation makes some forms of behavior more expensive relative to others. Rational actors will respond by shifting some conduct into the comparatively cheaper outlets. And these alternative behaviors may actually be less beneficial than the regulated, deterred behavior.”).

[17] Proposed Statement, supra note 10, at 5.

[18] Int’l Harvester Co., 104 F.T.C. 949, 1059 (1984).

[19] 15 U.S.C. § 45(n) (2024); FTC v. Corpay, Inc., 164 F.4th 807, 840 (11th Cir. 2026) (“As ‘unless’ conveys, the three factors that follow it are mandatory for a finding that a practice is ‘unfair.’”).

[20] FTC Policy Statement on Unfairness, 104 F.T.C. 1070, 1073 (1984) (appended to Int’l Harvester Co., 104 F.T.C. 949 (1984)).

[21] Proposed Statement, supra note 10, at 4 & n.13.

[22] See infra Part II.C and the Appendix.

[23] Proposed Statement, supra note 10, at 6.

[24] Int’l Harvester, 104 F.T.C. at 1062.

[25] Trade Regulation Rule on Commercial Surveillance and Data Security, 87 Fed. Reg. 51,273, 51,293 (Aug. 22, 2022) (advance notice of proposed rulemaking) (dissenting statement of Commissioner Noah Joshua Phillips); id. at 51,298 (dissenting statement of Commissioner Christine S. Wilson).

[26] Proposed Statement, supra note 10, at 4.

[27] See Gilman, supra note 9.

[28] Manne, Gilman & Stout, supra note 3, at 15, 47 (urging the Commission to explain how its Section 5 authority applies to data practices and noting that “[s]uch guidance would, of necessity, account for the assessment of countervailing benefits too”); Berin Szóka & Geoffrey A. Manne, The Federal Trade Commission: Restoring Congressional Oversight of the Second National Legislature 67 (Int’l Ctr. for Law & Econ. 2016), https://laweconcenter.org/resources/wstate-manneg-20160524-sd004 (arguing that guidance should explain how the Commission will measure “concepts like materiality, weighing injury with benefits, and measuring reasonable avoidability”).

[29] Proposed Statement, supra note 10, at 8.

[30] See Brief of Amici Curiae International Center for Law & Economics and TechFreedom in Support of Petitioner at 26–27, LabMD, Inc. v. FTC, No. 16-16270 (11th Cir. 2017), https://laweconcenter.org/resources/amicus-brief-icle-techfreedom-labmd-inc-v-federal-trade (arguing that treating informal guidance as the source of “legally enforceable duties” creates a “de facto rulemaking[]” that bypasses Section 18’s safeguards).

[31] See James M. Lacko & Janis K. Pappalardo, Fed. Trade Comm’n Bureau of Econ., Improving Consumer Mortgage Disclosures: An Empirical Assessment of Current and Prototype Disclosure Forms ES-1 (2007), https://www.ftc.gov/sites/default/files/documents/reports/improving-consumer-mortgage-disclosures-empirical-assessment-current-and-prototype-disclosure-forms/p025505mortgagedisclosurereport.pdf.

[32] Proposed Statement, supra note 10, at 1; see also id. at 3–4.

[33] Id. at 3 (emphasis added).

[34] Id. at 4.

[35] See Lars A. Stole, Price Discrimination and Competition, in 3 Handbook of Industrial Organization 2221, 2239–43 (Mark Armstrong & Robert Porter eds., 2007) (reviewing the models of Thisse & Vives and of Corts).

[36] Id.

[37] James C. Cooper, Luke Froeb, Daniel P. O’Brien & Steven Tschantz, Does Price Discrimination Intensify Competition? Implications for Antitrust, 72 Antitrust L.J. 327, 342 (2005).

[38] Andrew N. Ferguson, Comm’r, Fed. Trade Comm’n, Concurring Statement, Surveillance Pricing Intermediaries 1 (Matter No. P246202, July 23, 2024), https://www.ftc.gov/system/files/ftc_gov/pdf/surveillance-pricing-6b-ferguson-concurrence.pdf.

[39] Proposed Statement, supra note 10, at 3. In January 2025, Commission staff described its summary of the Section 6(b) responses as preliminary and disclaimed any assessment of illegal conduct. Fed. Trade Comm’n, FTC Surveillance Pricing 6(b) Study: Research Summaries—A Staff Perspective 2 (Jan. 2025), https://www.ftc.gov/system/files/ftc_gov/pdf/p246202_surveillancepricing6bstudy_researchsummaries_redacted.pdf [hereinafter Research Summaries] (“The format of this work aims to highlight initial observations . . . . This document does not reflect or suggest any assessment as to whether anyone has engaged in illegal conduct.”).

[40] Competition & Mkts. Auth., Pricing Algorithms: Economic Working Paper on the Use of Algorithms to Facilitate Collusion and Personalised Pricing ¶ 7.14, at 38 (CMA94, 2018) (describing the European Commission’s 2018 mystery shopping study of 160 e-commerce websites in eight member states: “Price differences were only observed in 6% of tests, and the median price difference observed was less than 1.6%.”); see also Competition & Mkts. Auth., The Commercial Use of Consumer Data ¶¶ 3.69–3.70, at 92 (CMA38, 2015) (“we did not receive a clear indication of examples where price discrimination using consumer data was being used to the likely detriment of consumers”; “Firms may be wary of damaging their reputation or brand value by being seen to do this.”).

[41] Alberto Cavallo, Are Online and Offline Prices Similar? Evidence from Large Multi-Channel Retailers, 107 Am. Econ. Rev. 283, 295 (2017).

[42] Stefano DellaVigna & Matthew Gentzkow, Uniform Pricing in U.S. Retail Chains, 134 Q.J. Econ. 2011, 2011 (2019).

[43] Proposed Statement, supra note 10, at 3–4 nn.8, 11–13.

[44] The paper the Commission relies on most, citing it for each of its economic claims, models the seller as a monopolist and states that it “do[es] not consider the impact of personalized pricing in a competitive market.” Jean-Pierre Dubé & Sanjog Misra, Personalized Pricing and Consumer Welfare, 131 J. Pol. Econ. 131, 180 (2023).

[45] See Stole, supra note 35, at 2231–32 & n.19.

[46] Benjamin Reed Shiller, Approximating Purchase Propensities and Reservation Prices from Broad Consumer Tracking, 61 Int’l Econ. Rev. 847, 861 (2020).

[47] Id. at 862 tbl.4.

[48] Id. at 867.

[49] Dubé & Misra, supra note 44, at 157–58.

[50] Id. at 134–35.

[51] Nicholas Buchholz et al., Personalized Pricing and the Value of Time: Evidence from Auctioned Cab Rides, 93 Econometrica 929, 930, 934 (2025).

[52] Id. at 952–53 & tbl.VII.

[53] Dubé & Misra, supra note 44, at 159 (“we also do not expect any competitive response from other platforms”).

[54] Buchholz et al., supra note 51, at 934.

[55] Andrew Rhodes & Jidong Zhou, Personalized Pricing and Competition, 114 Am. Econ. Rev. 2141, 2152 (2024) (stating in Proposition 1 that, where the market is fully covered, “relative to uniform pricing, personalized pricing harms firms and benefits consumers in aggregate”).

[56] Buchholz et al., supra note 51, at 954.

[57] Dubé & Misra, supra note 44, at 167.

[58] Shiller, supra note 46, at 862.

[59] Rhodes & Zhou, supra note 55, at 2152–53, 2162–63; see also Proposed Statement, supra note 10, at 4 n.13.

[60] Rhodes & Zhou, supra note 55, at 2157 & n.23 (thresholds); id. at 2163, 2165 (remedies).

[61] Buchholz et al., supra note 51, at 952.

[62] Id. at 952–53 & tbl.VII.

[63] Dubé & Misra, supra note 44, at 157, 165–67 & tbl.8. Every personalized price exceeds the uninformative $99 status quo, ranging from $126 to $6,292. A comparison with $99 thus shows every buyer paying more. The authors use the optimal uniform price “as our base case, not $99, since the former reflects the textbook inverse-elasticity-rule pricing that would be predicted for a maturing company.” Id. at 158, 165.

[64] Rhodes & Zhou, supra note 55, at 2144 n.7, 2155 (“With partial market coverage, personalized pricing increases welfare by expanding demand, but it boosts profit so much that in aggregate consumers suffer from it.”).

[65] Against a uniform price, consumer surplus falls 0.50% in Shiller and 2.25% in Buchholz et al. Shiller, supra note 46, at 861 tbl.3; Buchholz et al., supra note 51, at 953 tbl.VII. Dubé and Misra estimate a 25% decline against the optimal uniform price, even though most potential customers would receive lower price offers. Dubé & Misra, supra note 44, at 166–67. In Buchholz et al., adding price sensitivity to personalization based on wait sensitivity alone also lowers consumer surplus, but only slightly. Buchholz et al., supra note 51, at 953 tbl.VII.

[66] Buchholz et al., supra note 51, at 953 tbl.VII, 954–55 & fig.5 (the inside-option share rises 7.28% relative to uniform pricing; 62.5% of consumers gain; the most price- and wait-sensitive riders gain most); Dubé & Misra, supra note 44, at 167, 171–72 (63% of personalized prices fall below the optimal uniform price; smaller firms with part-time staff are most likely to benefit); Shiller, supra note 46, at 861 tbl.3, 862 (subscribers rise 3.12%; “most consumers receive lower prices when prices are personalized, and hence the majority are slightly better off”); see also Rhodes & Zhou, supra note 55, at 2153, 2155 (under partial coverage, personalized pricing “expands total demand and strictly increases total welfare”).

[67] FTC Policy Statement on Unfairness, supra note 20, at 1073.

[68] See Buchholz et al., supra note 51, at 954–55; Dubé & Misra, supra note 44, at 172; Shiller, supra note 46, at 862, 867.

[69] 15 U.S.C. § 45(n) (2024).

[70] Shiller, supra note 46, at 861 tbl.3; Buchholz et al., supra note 51, at 953 tbl.VII; Dubé & Misra, supra note 44, at 166–67 & tbl.8. Dubé and Misra compare personalized prices capped at $499 with the optimal uniform price of $327. Uncapped personalization reduces total consumer surplus by 25% against that benchmark. Capped personalization reduces it by 11.45% against the $249 uniform price ZipRecruiter later adopted. Id. at 166 tbl.8, 167.

[71] Dubé & Misra, supra note 44, at 167, 179.

[72] The Unfairness Policy Statement lists “reduced incentives to innovation and capital formation” among the costs of a remedy that must be considered. FTC Policy Statement on Unfairness, supra note 20, at 1073–74.

[73] See, e.g., Ohio v. American Express Co., 585 U.S. 529 (2018).

[74] See Stole, supra note 35, at 2231–32 & n.19.

[75] Buchholz et al., supra note 51, at 953 tbl.VII; Shiller, supra note 46, at 861 tbl.3. Buchholz et al. note that increased platform use need not mean an equal increase in total travel because their model’s outside option includes taking “another form of transportation.” Buchholz et al., supra note 51, at 939. The relevant question is how pricing affects users’ decisions to book trips on the platform; the distance traveled matters less to that question.

[76] Proposed Statement, supra note 10, at 4 & n.13.

[77] Id.

[78] Dubé & Misra, supra note 44, at 142–43.

[79] Id. at 131 (emphasis added).

[80] Id. at 171–72.

[81] Id. at 180.

[82] Id. at 136, 172.

[83] Proposed Statement, supra note 10, at 4 n.13.

[84] Rhodes & Zhou, supra note 55, at 2163 (“[A]symmetry in information and timing allows the firms to better segment the market, increasing industry profit. Consequently, consumer surplus ends up being lowest in the mixed regime.”).

[85] Id. at 2142 (“We focus on the limit case of personalized pricing, i.e., first-degree price discrimination, where firms can perfectly infer individual consumers’ preferences”); id. at 2162 n.30 (“it would also be interesting to consider other information structures and investigate how the impact of price discrimination changes with the amount of information that firms have access to”).

[86] Buchholz et al., supra note 51, at 953 tbl.VII.

[87] Id. at 952–53 & tbl.VII.

[88] Id. at 953.

[89] Id. at 932.

[90] Id. at 953 tbl.VII, 954–55 & fig.5.

[91] Ginger Zhe Jin, Liad Wagman & Mengyi Zhong, Surveillance Pricing: A Cautionary Summary of Potential Harms and Solutions, CPI Antitrust Chron., July 2025, at 4.

[92] Id. (emphasis added).

[93] Id.

[94] Id. at 5.

[95] Id. (citing Rhodes & Zhou, supra note 55).

[96] Id. at 9 (emphasis added).

[97] Id.

[98] Id.

[99] Rosa-Branca Esteves & Francisco Carballo-Cruz, Access to Data for Personalized Pricing: Can It Raise Entry Barriers and Abuse of Dominance Concerns? 10, 15, 18 (NIPE Working Paper No. 05/2021, Mar. 2021). The Proposed Statement misdates the paper by four years. See Proposed Statement, supra note 10, at 4 n.13 (citing the paper as “NIPE Working Papers, NIPE, Universidade do Minho (May 2025)”).

[100] Esteves & Carballo-Cruz, supra note 99, at 14 tbl.1.

[101] Id. at 15 (“conditional on entry, consumer surplus is in fact maximized when both firms are able to use data for personalized prices”).

[102] Id. at 1.

[103] Id.

[104] Mark Armstrong & John Vickers, Discriminating Against Captive Customers, 1 Am. Econ. Rev.: Insights 257, 257–58, 263 (2019).

[105] Brian C. Albrecht, Price Competition and the Use of Consumer Data 1–2 (working paper, June 12, 2026), https://briancalbrecht.com/albrecht_price_competition_consumer_data.pdf.

[106] Dirk Bergemann, Benjamin Brooks & Stephen Morris, The Limits of Price Discrimination, 105 Am. Econ. Rev. 921, 921–23 (2015).

[107] See Rhodes & Zhou, supra note 55, at 2164; Esteves & Carballo-Cruz, supra note 99, at 1, 17–18.

[108] Proposed Statement, supra note 10, at 5.

[109] Id.

[110] Proposed Statement, supra note 10, at 3; see also id. at 1 (personalized pricing is “a long-established norm in some markets”).

[111] Id. at 1.

[112] Gilman, supra note 9 (“How many consumers must hold that expectation? How consistently? Must a clear majority share it?”); see also Albrecht et al., Online Food Delivery Comments, supra note 8, at 7 (“A first-time user’s expectations differ materially from those of a customer placing a hundredth order.”).

[113] Council of Econ. Advisers, Big Data and Differential Pricing 17 (2015).

[114] FTC Policy Statement on Deception, 103 F.T.C. 174, 176 & n.8 (1984) (appended to Cliffdale Assocs., Inc., 103 F.T.C. 110 (1984)).

[115] FTC Policy Statement on Deception, supra note 114, at 182; Int’l Harvester, 104 F.T.C. at 1057; see also Geoffrey A. Manne, R. Ben Sperry & Berin Szoka, In the Matter of Nomi Technologies, Inc.: The Dark Side of the FTC’s Latest Feel-Good Case 5, 10–11 (ICLE White Paper 2015-1, 2015), https://laweconcenter.org/resources/nomi-white-paper (objecting to the Commission’s treatment of the absence of an in-store notice as an implied representation and observing that “for an implied promise to be deemed material (and thus deceptive) under the DPS, the FTC must adduce some proof: either that it was, in fact, intended, or that it was, in fact, material”).

[116] FTC Policy Statement on Deception, supra note 114, at 175–76.

[117] Int’l Harvester, 104 F.T.C. at 1057, 1059.

[118] Id. at 1059–60.

[119] Id. at 1063–64, 1070.

[120] Id. at 1058 & n.29; see also id. at 1062 (limiting mandatory disclosure “to those core aspects of a transaction that virtually all consumers would consider essential to an informed decision”).

[121] See Int’l Harvester, 104 F.T.C. at 1060 (“Since the seller will have no way of knowing in advance which disclosure is important to any particular consumer, he will have to make complete disclosures to all.”).

[122] Proposed Statement, supra note 10, at 6 n.24 (quoting P. Lorillard Co. v. FTC, 186 F.2d 52, 58 (4th Cir. 1950)) (“to tell less than the whole truth is a well-known method of deception”).

[123] P. Lorillard Co. v. FTC, 186 F.2d 52, 58 (4th Cir. 1950) (“The fault with this advertising was not that it did not print all that the Reader’s Digest article said, but that it printed a small part thereof in such a way as to create an entirely false and misleading impression, not only as to what was said in the article, but also as to the quality of the company’s cigarettes.”). The Commission itself cites Lorillard as a half-truth case, not a silence case. Int’l Harvester Co., 104 F.T.C. at 1057 & n.25.

[124] FTC v. Standard Educ. Soc’y, 302 U.S. 112, 114–16 (1937) (salespeople represented that a set of books was free, that the prospect had been specially selected, and that the regular price exceeded the offered price; the Commission found each representation false).

[125] Proposed Statement, supra note 10, at 5.

[126] Proposed Statement, supra note 10, at 7.

[127] FTC v. Colgate-Palmolive Co., 380 U.S. 374, 387 (1965).

[128] Proposed Statement, supra note 10, at 5 n.19 (quoting Colgate-Palmolive, 380 U.S. at 387).

[129] 15 U.S.C. § 45(n) (2024) (emphasis added).

[130] Corpay, 164 F.4th at 840 (“As ‘unless’ conveys, the three factors that follow it are mandatory for a finding that a practice is ‘unfair.’”); see also Proposed Statement, supra note 10, at 5 n.17 (citing Corpay, 164 F.4th at 839).

[131] Proposed Statement, supra note 10, at 7–8.

[132] 15 U.S.C. § 45(n) (2024).

[133] Proposed Statement, supra note 10, at 3.

[134] Id. at 6.

[135] FTC Policy Statement on Unfairness, supra note 20, at 1073.

[136] Id. at 1073.

[137] See Proposed Statement, supra note 10, at 6 n.24 (citing Priester et al. and Xia et al. on consumers’ perceptions of fairness); Geoffrey A. Manne & Kristian Stout, Comments of the International Center for Law & Economics, Informational Injury Workshop 40–41 (FTC Project No. P175413, 2017), https://laweconcenter.org/resources/comments-international-center-law-economics-matter-informational-injury-workshop-ftc (quoting the legislative history of Section 5(n): “Emotional impact and more subjective types of harm alone are not intended to make an injury unfair.”).

[138] Proposed Statement, supra note 10, at 6.

[139] Id. at 6.

[140] Id. at 6 n.20.

[141] Corpay, 164 F.4th at 825–27, 838–42.

[142] Proposed Statement, supra note 10, at 6.

[143] Proposed Statement, supra note 10, at 6 n.22.

[144] 15 U.S.C. § 1681m(a)(1)–(4) (2024); see id. § 1681g(f)(1)(C) (key factors). The quoted statement appears in § 1681m(a)(3)(B).

[145] 15 U.S.C. § 1691(d)(2)–(3) (2024).

[146] 15 U.S.C. § 1691(d)(2)(B) (2024).

[147] 15 U.S.C. § 1691(d)(6) (2024).

[148] 15 U.S.C. §§ 1681m(h)(1), (h)(5), 1681g(f)(1)(B)–(E) (2024).

[149] 15 U.S.C. § 1681m(b)(1) (2024).

[150] Proposed Statement, supra note 10, at 6 n.23 (citing Cal. Ins. Code §§ 791.02, 791.10; Va. Code Ann. § 38.2-602; Wash. Admin. Code ch. 284-30A).

[151] 15 U.S.C. § 45(n) (2024).

[152] Proposed Statement, supra note 10, at 6 n.21.

[153] FTC Policy Statement on Unfairness, supra note 20, at 1075–76.

[154] Proposed Statement, supra note 10, at 6.

[155] Proposed Statement, supra note 10, at 7.

[156] Adams, Fruits & Albrecht, supra note 5, at 27 (“retailers should clearly disclose final prices and truthfully present discounts and reference prices”); Albrecht et al., Online Food Delivery Comments, supra note 8, at 7–8, 12.

[157] Int’l Harvester, 104 F.T.C. at 1062 (adding that “virtually any piece of information may be useful to some consumers,” and that a disclosure duty grounded in unfairness requires “a full cost-benefit analysis, in which we weigh the consumer benefits of disclosure against their likely costs”).

[158] Exec. Office of the President, Big Data: Seizing Opportunities, Preserving Values 65 (2014).

[159] Council of Econ. Advisers, supra note 113, at 17.

[160] Petra Persson, Attention Manipulation and Information Overload, 2 Behav. Pub. Pol’y 78, 96–97 (2018).

[161] Frank H. Easterbrook & Daniel R. Fischel, Mandatory Disclosure and the Protection of Investors, 70 Va. L. Rev. 669, 709 (1984).

[162] Persson, supra note 160, at 101.

[163] Id. at 97. Both results come from a formal model, illustrated with examples rather than tested against market data. They show that the seller responses Persson describes are available to any firm subject to a disclosure mandate, but they do not show how often firms use them.

[164] Manne, Hydraulic Theory, supra note 16, at 480–81 (“As this Article suggests, however, mandatory disclosure has important and ill-considered consequences; it is not merely regulatory ‘chicken soup.’”).

[165] Id. at 506.

[166] Id. at 506–07 (“It is precisely individual shareholders who are not in a position to evaluate complex information regarding, for example, executive pay packages. But it is also precisely these individuals at whom mandatory disclosures are directed.”).

[167] Proposed Statement, supra note 10, at 7.

[168] Manne, Hydraulic Theory, supra note 16, at 510–11.

[169] Like other agencies before it. See id. at 497 (“Unfortunately, the SEC does not appear to consider the relevant, marginal question. . . . The broad-brushed nature of federal regulation effectively precludes a firm-by-firm marginal analysis.”); see also id. at 491 (“Regulators are limited in their ability to predict the effects of their regulations . . . . Even well-intended regulation is bound to have unintended consequences . . . .”).

[170] OECD, Enhancing Online Disclosure Effectiveness 14, 31 (OECD Digital Econ. Papers No. 335, 2022) (quoting Omri Ben-Shahar & Carl E. Schneider, The Failure of Mandated Disclosure, 159 U. Pa. L. Rev. 647, 686 (2011)).

[171] See Trade Regulation Rule on Unfair or Deceptive Fees, 90 Fed. Reg. 2066, 2066 (Jan. 10, 2025) (codified at 16 C.F.R. pt. 464).

[172] See Rule on Unfair or Deceptive Rental Housing Fee Practices, 91 Fed. Reg. 12,325 (Mar. 13, 2026) (advance notice of proposed rulemaking).

[173] See Proposed Statement, supra note 10, at 1 n.1.

[174] Lisa Robinson, Kip Viscusi & Richard Zeckhauser, Efficient Warnings, Not “Wolf or Puppy” Warnings, in The Future of Risk Management 227, 241 (Howard Kunreuther et al. eds., 2019).

[175] OECD, Enhancing Online Disclosure Effectiveness, supra note 170, at 6.

[176] Howard Beales & Todd J. Zywicki, Junkyard Dogs: The Law and Economics of “Junk” Fees, CPI Antitrust Chron., Apr. 2023, at 6, https://laweconcenter.org/resources/junkyard-dogs-the-law-and-economics-of-junk-fees/.

[177] Anna Priester et al., A Special Price Just for You: Effects of Personalized Dynamic Pricing on Consumer Fairness Perceptions, 19 J. Revenue & Pricing Mgmt. 99, 106–08 (2020).

[178] Oren Bar-Gill, Cass R. Sunstein & Inbal Talgam-Cohen, Algorithmic Harm in Consumer Markets, 15 J. Legal Analysis 1, 10, 25 (2023).

[179] OECD, The Effects of Online Disclosure About Personalised Pricing on Consumers 5, 9–13 (OECD Digital Econ. Papers No. 303, 2021).

[180] Id. at 5.

[181] Id. at 27.

[182] Id. at 27.

[183] Id. at 26.

[184] The OECD’s later report emphasizes the prominence and repetition of the tested notices: the disclosures were “succinct, substantial in size, seen repeatedly, and accompanied every single personalised price. OECD, Enhancing Online Disclosure Effectiveness, supra note 170, at 14.

[185] Id.

[186] Willem H. van Boom et al., Consumers Beware: Online Personalized Pricing in Action! How the Framing of a Mandated Discriminatory Pricing Disclosure Influences Intention to Purchase, 33 Soc. Just. Rsch. 331 (2020).

[187] Id. at 335–37.

[188] Id. at 343–45.

[189] Allison Nicoletti & Christina Zhu, Economic Consequences of Transparency Regulation: Evidence from Bank Mortgage Lending, 61 J. Acct. Rsch. 1827, 1846–51 (2023).

[190] Id. at 1832.

[191] Natasha Clarke et al., Calorie (Energy) Labelling for Changing Selection and Consumption of Food or Alcohol, 2025 Cochrane Database of Systematic Reviews, Art. CD014845, at 2, https://doi.org/10.1002/14651858.CD014845.pub2.

[192] Pasquale E. Rummo et al., Menu Labeling and Calories Purchased in Restaurants in a US National Fast Food Chain, 6 JAMA Network Open e2346851, tbl. 3 (2023).

[193] See Lacko & Pappalardo, supra note 31, at ES-1.

[194] Proposed Statement, supra note 10, at 6.

[195] Proposed Statement, supra note 10, at 6.

[196] Manne, Hydraulic Theory, supra note 16, at 476. The article discusses securities regulation, but its account of how regulated actors respond to disclosure mandates applies beyond that setting.

[197] Id. at 477.

[198] Id. at 485 (“[W]here behavior can be altered at a lower cost than the cost of disclosure, disclosure rules will induce behavioral changes rather than increased information flow.”); id. at 486 (“The magnitude of the hydraulic effect will turn primarily on the degree of cross-elasticity (substitutability) between the impeded (more costly) and the freer (less costly) sources of behavior.”).

[199] Proposed Statement, supra note 10, at 6 n.20 (declining “to take any position on whether some personalized pricing practices are unfair even when fully disclosed to consumers”).

[200] Buchholz et al., supra note 51, at 932, 952 (describing ETA-based pricing as “another non-personalized pricing policy” that platforms may adopt “due to consumer backlash,” under which “[w]ait-time pricing may then give consumers incentives to reveal their high wait-time sensitivity by picking a driver with shorter wait time and higher price”).

[201] Manne, Hydraulic Theory, supra note 16, at 477; see also id. at 487 (observing that, under seat-belt laws, a shift in behavior “does not impose the same risk on, say, pedestrians, who now suffer the ill effects of more reckless driving but earn none of the benefit of ‘safer’ accidents”).

[202] Id. at 477–78.

[203] Stole, supra note 35, at 2226 (conditions for discrimination). The price-gap example in the text is a conditional incentive argument, not an estimate of the proposed notice’s effect.

[204] Xi Li & Krista J. Li, Beating the Algorithm: Consumer Manipulation, Personalized Pricing, and Big Data Management, 25 Mfg. & Serv. Operations Mgmt. 36, 36 (2023).

[205] S. Nageeb Ali, Greg Lewis & Shoshana Vasserman, Voluntary Disclosure and Personalized Pricing, 90 Rev. Econ. Stud. 538, 558–60 (2023).

[206] Dubé & Misra, supra note 44, at 167, 171–72.

[207] DellaVigna & Gentzkow, supra note 42, at 2014, 2033.

[208] Id. at 2055–56 & fig.XI. See also Manne, Hydraulic Theory, supra note 16.

[209] See Fruits, Waterbed Effect, supra note 7, at 14–16 (the Robinson-Patman Act’s effect on discounting, the Antitrust Modernization Commission’s 2007 conclusion, and the pay-transparency evidence of Cullen and Pakzad-Hurson); Geoffrey A. Manne, Written Testimony, Reviving Competition, Part 5: Addressing the Effects of Economic Concentration on America’s Food Supply 19–20 (Int’l Ctr. for Law & Econ. 2022), https://laweconcenter.org/resources/testimony-of-geoffrey-a-manne-reviving-competition-part-5-addressing-the-effects-of-economic-concentration-on-americas-food-supply/ (“The RPA effectively imposes wholesale price uniformity on sellers which deters innovation (or, as noted, price reductions) and new entry.”).

[210] Complaint ¶¶ 7, 21–22, 262, FTC v. Amazon.com, Inc., No. 2:23-cv-01495 (W.D. Wash. Sept. 26, 2023).

[211] See, e.g., Zoë B. Cullen & Bobak Pakzad-Hurson, Equilibrium Effects of Pay Transparency, 91 Econometrica 765, 765 (2023) (finding that state laws protecting workers’ right to inquire about pay were followed by wage declines of about 2%, because employers “credibly refuse to pay high wages to any one worker to avoid costly renegotiations with others”).

[212] Fed. Trade Comm’n, Research Summaries, supra note 39, at 1–2, 10.

[213] See Abbott, supra note 9.

[214] Corpay, 164 F.4th at 838–42 (unfairness count proved through a customer survey, eight internal surveys, internal emails, complaint data, and evidence of what customers were told about the fees); see also id. at 825–27 (record evidence on what customers were told).

[215] Proposed Statement, supra note 10, at 3–4 nn.8, 11–13.

Interest of Amicus Curiae^[1]

The International Center for Law & Economics (ICLE) is a nonprofit research organization that applies economic analysis to legal questions, with particular expertise in competition, intellectual property, and digital markets. ICLE has participated as amicus in earlier stages of this litigation, including the original appeal, the ensuing rehearing and certiorari proceedings, and the appeal from the district court’s contempt order. Its scholars have written extensively about platform competition, investment incentives, and the consequences of judicial error in evaluating business conduct.

Summary of Argument

The 2021 injunction required Apple to allow developers to communicate with customers about alternative purchasing mechanisms, including through links within their apps. It established no commission rate or method of calculating one. Pet. App. 166a–167a. The Ninth Circuit nonetheless found a limit on “prohibitive” commissions in the injunction’s spirit, even though “the text of the Injunction does not address commissions at all.” Pet. App. 34a n.9.

The injunction rested on a finding of informational harm to consumers—that Apple’s anti-steering rules prevented consumers from making informed choices because they were left ignorant of non-Apple purchasing mechanisms. Apple’s revised guidelines alleviated that harm by allowing developers to inform users within the App Store of external payment alternatives. Apple also imposed a commission on those external payments.

The contempt order then treated the commercial attractiveness to developers of external purchasing as the measure of whether the informational harm had been remedied. A commission can affect developers’ willingness to offer links and thus the information consumers receive. But its effect on developers’ incentives does not, by itself, establish whether a lower commission would improve competition and consumer welfare. The injunction left Apple discretion to determine compensation for the services it continued to provide, and a commission did not violate the order merely because a lower rate would have been preferred by some developers.

A business required to change its conduct must still decide how to run its business. When an injunction leaves several permissible ways to comply, choosing a compliance mechanism that also preserves the most revenue is still compliance. The firm need not prefer an alternative because it is less profitable, and its decision to compare the financial consequences of available choices is entirely ordinary. The contrary approach would make the sacrifice of revenue a test of obedience even where an injunction imposed no such requirement.

Apple faced a practical question in determining the exact terms of compliance: how to collect compensation for services that developers continued to use when a user followed a link-out and paid elsewhere. Preserving that compensation is not inconsistent with the obligation to enable developer communications, and, indeed, doing so furthers the competition the obligation was intended to promote.

The underlying California Unfair Competition Law (UCL) judgment makes the competitive effects of commissions particularly relevant. The UCL reaches conduct beyond completed antitrust violations, but the standards applied in this case remain concerned with harm to competition and the benefits of the challenged conduct. A court invoking the competitive purpose of an injunction must account for the procompetitive effects of the defendant’s conduct, including its role in sustaining valuable services. Because those services require financial backing, it does not further competition simply to equate a larger financial advantage for developers with a better competitive outcome.

An external payment changes who processes a transaction. It does not eliminate the software development, distribution, and access to users that helped make the transaction possible. Commissions provide a way to finance the development, provision, and ongoing improvement of those services while allowing developers to enter with low initial charges and pay more as they succeed. They can also prevent developers from free riding by avoiding payment for valuable platform services simply by moving the final purchasing step elsewhere. Consideration of these functions is essential to evaluating a remedy intended to promote competition.

In finding Apple’s commission prohibitive, the courts below relied heavily on the conclusion that developers’ outside transaction costs exceeded the alleged three-percentage-point difference between Apple’s in-app and link-out commissions. Pet. App. 20a, 136a. That comparison may be relevant to assessing the competitive effects of a chosen commission, but it does not establish the value of Apple’s contribution. Nor does it capture the full range of reasons other than short-term cost that a developer might choose another purchasing channel. For these reasons, increasing use of that alternate channel will not necessarily promote competition. A rule that encourages migration by limiting compensation for services still supplied can reduce the resources and incentives to provide those services, ultimately harming competition and consumer welfare.

The resulting rate-setting problem is not trivial, and it necessitates exactly the sort of internal commercial deliberations that the district court treated as evidence of contempt. The Ninth Circuit rejected a permanent prohibition on commissions but contemplated a court-approved fee based on selected costs and uses of intellectual property. Pet. App. 34a–36a. Determining that fee requires decisions about common costs, continuing investment, and the allocation of charges across different users. A general instruction to promote competition answers none of those questions.

This Court has cautioned against imposing duties that require judges to administer complex commercial terms and that discourage beneficial conduct. See, e.g., Verizon Commc’ns Inc. v. Law Offices of Curtis V. Trinko, LLP, 540 U.S. 398, 414–15 (2004); Brooke Group Ltd. v. Brown & Williamson Tobacco Corp., 509 U.S. 209, 223 (1993). Those concerns have particular force when contempt supplies the means of imposing a pricing obligation absent from the injunction. The questions the injunction left unanswered create at least a fair ground of doubt about whether it barred Apple’s commission, which rules out civil contempt. Taggart v. Lorenzen, 587 U.S. 554, 561 (2019).

Argument

I. The Injunction Sought to Remedy Informational Harm. It Said Nothing About Commissions

The 2021 injunction rests on a violation of California’s Unfair Competition Law predicated on informational harm to consumers. Apple’s anti-steering rules prevented users from learning that purchasing options existed outside in-app purchase. See Pet. App. 34a n.9. Apple’s revised guidelines removed the source of that harm by permitting developers to tell users—through buttons, links, and calls to action—that other options exist. Once that information reaches consumers, the injunction has accomplished what it set out to do.

A. Apple’s compliance measures must be assessed as a whole

Apple did not simply replace a 30% commission with a 27% commission. Rather, Apple adopted a multi-pronged approach in response to the district court’s initial order and injunction which makes the repeated assertion of a 27% rate factually inaccurate.

First, Apple dropped the anti-steering provisions that the court had found to violate the UCL, so that links to external (non-Apple) purchasing mechanisms would be permitted. Apple also provided a mechanism by which approved developers could “include within their apps buttons or links with calls to action directing users to out-of-app purchasing mechanisms other than IAP.” JA506–507. That, in itself, lowers information costs for developers who wish to alert users to external payment options, for the users themselves, and, by extension, for external payment sites.

Second, Apple adopted a varied schedule of commissions for off-app purchases facilitated by that mechanism. For linked off-app purchases made more than “seven days after a user taps on an External Purchase Link and continues . . . to an external website,” JA519, Apple would charge no commission. For linked off-app purchases made within seven days, Apple would charge either 12% (for developers participating in the App Store Small Business Program) or 27% (for larger participating developers). JA518.[2] The program also specified a 12% rate for qualifying transactions otherwise subject to a 15% in-app commission. JA518.

Third, Apple reiterated recently adopted provisions under a separate settlement that permitted developers to send out-of-app communications to their users about alternative purchasing methods. JA521. These provisions permit developers to request basic contact information from their app users and to “send communications outside of the app to their user base about purchasing methods other than in app purchase.” JA504. Apple’s notice further explained that developers could use those communications to direct users to their websites without incurring a link-out commission at all. JA520–21.

Apple also provided a procompetitive rationale for its multi-pronged approach—one that neither the district court nor the court of appeals refutes—arguing that its approach balances “(1) Apple’s entitlement to a commission for purchases facilitated through the App Store, even if the user waits a small amount of time after leaving the App Store to actually finalize the purchase, with (2) the fact that the longer the time between the link out and the purchase, the more attenuated the connection between the purchase and Apple’s facilitation of the purchase.” JA519–520.

Further, while Apple did adopt a 27% commission on a subset of linked purchases made within one week of linking via apps on the App Store, it is not correct to characterize its compliance as imposing “a 27 percent commission . . . on off-app purchases, where it had previously charged nothing.” Pet. App. 51a. Rather, the 27% commission applied to only a subset of the previously barred purchases, and the effective commission rate was considerably lower.[3]

Moreover, Apple permitted previously barred in-app links to external purchasing mechanisms, with commissions of 0%, 12%, or 27% commissions depending on the timing and type of purchase and the developer’s eligibility for reduced rates. JA518–20. The prior bar on those links was a clear and categorical prohibition tantamount to an infinite price for using that route—it was not a zero-percent commission.

Thus Apple did not implement a 27% charge on transactions for which it previously charged nothing; rather, it opened a previously prohibited purchasing channel, with a maximum commission of 27% and an effective rate it expected to be lower.

B. Developer savings alone do not establish consumer benefit

Apple’s Notice of Compliance should be assessed as a whole, considering both its stated rationale and the informational harm the injunction was designed to remedy. JA500–522. The Ninth Circuit’s question—”whether charging a 27% commission has a prohibitive effect, in violation of the Injunction,” Pet. App. 20a—cannot be answered by isolating that rate from the other terms Apple adopted.

The expanded opportunities to communicate and the varied commission rates cannot simply be presumed to undermine the injunction’s informational objective. The 27 percent figure was the highest rate in a schedule that also included lower rates and commission-free purchases. Apple could reasonably expect those features to produce a lower effective rate. JA518–20. Treating the headline rate as controlling fails to address the Ninth Circuit’s own admonition that “[t]he district court must be mindful of the threshold upon which commissions become ‘prohibitive.’” Pet. App. 34a n.9.

Whether Apple’s response “thwarted the Injunction’s goals, and continued its anticompetitive conduct,” Pet. App. 51a, must also be assessed against the injunction’s express terms and the details of Apple’s approach. But the contempt order treated the commercial attractiveness of external purchasing to developers as the full measure of whether the informational harm had been remedied. It reasoned that Apple’s commission left external purchasing so financially unattractive that developers would decline to offer it, leaving consumers with a right to information but nothing to be informed about. Pet. App. 69a–87a.

A developer’s reluctance to offer an in-app link because of the commission can affect how readily consumers learn about an alternative. But a link both communicates an alternative and provides a direct route to a purchase; it is not the only way to inform users that purchasing is available through the developer’s website. Apple also permitted out-of-app communications directing users to developers’ websites without triggering a link-out commission. JA520–21.

The Ninth Circuit properly distinguished the mere existence of a commission from its effect on developers’ willingness to offer links. Pet. App. 34a n.9. In the courts’ analysis, however, developer savings became a proxy for consumer benefit. A lower commission may encourage more linking, but that alone does not establish that it benefits consumers when the commission also finances services they continue to use.

Recasting the injunction’s informational premise as a guarantee of a commercially attractive alternative does not enforce it. It sets a new commercial term the injunction never fixed, essentially modifying the injunction without proper process.

II. Choosing a profitable way to comply is still compliance

An injunction that does not enjoin a company from doing business entirely implicitly contemplates that it will undertake activities necessary to ensure its continued operation so long as they do not conflict with the terms of the injunction. This is especially true for the injunction at issue here, which is premised on increasing competition by permitting developers to direct users to outside purchasing options while still maintaining Apple’s platform, including its In-App Purchasing (IAP) as a viable alternative.

Incorporating a new operating procedure—here, enabling developers to direct customers to alternate payment mechanisms, Pet. App. 166a–167a—may entail any number of adjustments to other activities not addressed by the injunction and necessary to ensure effective continued operations. Managers implementing compliance with an injunction have two related responsibilities: identify the changes the order requires and decide how best to make those changes while preserving the quality and profitability of the business as a whole. Both are constituent parts of compliance.

This is especially true for managers of firms, who owe a fiduciary duty to exercise their considered business judgment to act in the corporation’s best interests. See Commodity Futures Trading Comm’n v. Weintraub, 471 U.S. 343, 348–49 (1985). And it is no less true when exercising their business judgment in choosing among lawful means of complying with a court order. See Schenker v. E. I. du Pont de Nemours & Co., 329 F.2d 77, 78–79 (2d Cir. 1964) (upholding directors’ chosen method of implementing a court-ordered divestiture as “a reasonable exercise of business judgment”). The business must continue to assess costs, respond to customers, and earn revenue. Financial analysis does not become evidence of evasion simply because, while determining how to comply with an injunction, the firm is also attempting to avoid a loss.

An injunction must identify required and forbidden conduct with specificity and reasonable detail, which necessarily leaves discretion over implementation on various dimensions. See Fed. R. Civ. P. 65(d)(1); Schmidt v. Lessard, 414 U.S. 473, 476 (1974) (“[Rule 65(d)] was designed to prevent uncertainty and confusion on the part of those faced with injunctive orders, and to avoid the possible founding of a contempt citation on a decree too vague to be understood.”). The firm knows more than the court about its operations, and it can use that knowledge to identify less costly ways to comply. Preserving that discretion allows the injunction to achieve its stated objective without unnecessary losses to the business and the customers it serves.

Where several alternatives may satisfy the same obligation, managers must assess the full consequences of each form of compliance. The underlying court-ordered obligation may affect how rivals compete and how users behave, and the commercial and operational consequences of those changes may require numerous adjustments throughout the defendant’s business that go well beyond the text of the injunction.

It was thus predictable that compliance would require internal deliberations over the precise mechanism to employ and the follow-on adjustments that mechanism might entail. These included, e.g., how to collect commissions on purchases made outside Apple’s payment system, Pet. App. 73a; whether to change the amount or basis of its charges, Pet. App. 78a; how to adapt its commission discounts and partner programs to the new payment arrangements, 3-ER-433; how to allocate responsibility for subscription management and refunds, JA509, JA511; how to revise its communications with users, Pet. App. 145a; and how to manage developer accounting, audits, interest charges, and payment offsets, JA520. Its implementation discussions also addressed the software changes, developer materials, and review procedures needed to meet the anticipated compliance deadline. 3-ER-443.

Indeed, Apple’s planning documents illustrate the work of translating the injunction into practice. One presentation treated persistent, discoverable links and in-app price comparisons as requirements, while identifying placement, frequency, design, and pricing language as matters still to be resolved. 3-ER-385, 387. The injunction did not specify those details, Pet. App. 166a–167a; rather, Apple had to determine how to give effect to the order on those dimensions and incorporate the resulting changes into its existing services.

The necessity of such considerations was, in fact, even contemplated by the district court. In its order denying Apple’s request for a stay of the injunction, the court acknowledged that “the injunction may require additional engineering or guidelines,” Order Denying Apple’s Motion to Stay Injunction Pending Appeal at 3, Epic Games, Inc. v. Apple Inc., No. 4:20-cv-05640-YGR (N.D. Cal. Nov. 9, 2021) (“Stay Order”), and that it could “envision numerous avenues for Apple to comply with the injunction,” specifically noting that Apple could “take steps to protect users, to the extent that Apple genuinely believes that external links would create issues.” Id. Tellingly, the court added that it was not “here to micromanage.” Id.

Consider Apple’s specification of lower commissions for small developers and qualifying subscription renewals as part of its compliance plan. JA518. Having elected to collect commissions on linked-out purchases, Apple also had to decide how its existing discounts would apply to those transactions. The injunction did not prescribe that treatment. For a profit-making enterprise obligated to provide a return to its shareholders, that decision inevitably included an evaluation of the revenue consequences of the various possible arrangements. Considering those consequences was thus an inherent function of compliance, not an indication of noncompliance.

Permitting in-app links to outside purchases also necessitated a reconsideration and revision of how to convey Apple’s resulting responsibilities and users’ risks when users opt to leave the App Store to make external payments. Thus, Tim Cook asked the compliance team to revise the customer notice to reflect that “Apple’s privacy and security standards do not apply to purchases made on the web.” Pet. App. 145a. Enabling those links required Apple to decide how to explain the consequences of outside payments to users, even though the injunction did not specifically address that issue.

In considering different commission rates, Apple also assessed how its pricing rationale would compare with “existing discounting approaches in other jurisdictions” and whether it could explain those rates “without implicitly devaluing our IP / proprietary technology.” Pet. App. 81a. Implementing the injunction had consequences both for Apple’s day-to-day operations as well as for its pricing and commercial relationships elsewhere, and incorporation of those effects was a function of compliance with the order, not of its evasion.

Because the injunction left compensation available, Apple’s decision to preserve it cannot itself show that the firm frustrated the order. Nor does the existence of a cheaper option for developers establish that the more remunerative choice was impermissible. That inference would assume the very pricing obligation in dispute: that Apple had to forgo otherwise available compensation to make the external alternative more attractive. As long as Apple was still providing services, it had to decide whether and how to collect compensation for them through other means. The injunction specified the communications Apple must permit; it did not establish a price for the platform services that remained in use. Pet. App. 166a–167a. Evaluating the revenue consequences of alternative collection arrangements was a necessary and predictable part of implementing the order.

The Ninth Circuit expressly stated that the district court had not found contempt merely because Apple selected the most advantageous option for its business. Pet. App. 16a. That distinction must have practical consequences for how a court evaluates commercial deliberations and their resulting decisions. Discussions about protecting revenue, calculating the cost of different options, and retaining a return on services supplied are consistent with ordinary efforts to comply. Managers can plan around an identified obligation and choose among ways of satisfying it. The possibility that another arrangement would leave more money with developers supplies no additional standard for determining what compliance requires.

This Court has recognized that “[e]conomic hardship can influence choice only as among two or more effective remedies.” United States v. E. I. du Pont de Nemours & Co., 366 U.S. 316, 327 (1961). That holding arose in the context of the Court’s rejection of an ineffective remedy, chosen by a court to avoid severe economic consequences for the defendant. What the Court rejects is the ineffectiveness, not the economic consideration, and the same distinction is relevant to managers implementing a presumptively effective remedy: Economic considerations may inform the choice among compliant alternatives. Once the firm has identified permissible alternatives, it has every reason to select the one that works best commercially.

Attention to commercial self-interest does not conflict with the competitive purpose of the court’s remedy. Indeed, it furthers it. The prospect of earning revenue encourages firms to compete by improving their products, attracting customers, and sustaining services that others value. See Trinko, 540 U.S. at 407–08. The injunction left Apple free to compete for users’ business alongside outside purchasing options. Requiring it to disregard costs and returns when choosing among compliant arrangements would weaken its incentive to improve the services on which that competition depends.

III. Apple’s Commissions can advance the UCL’s purpose of protecting competition

The competitive purpose of this injunction must be understood in light of the UCL standards that produced it. The judgment addressed restrictions on developers’ ability to inform users about alternative purchasing options. The district court found that those restrictions impaired informed choice and competition in the relevant market, and it ordered Apple to permit the specified communications. JA373–79, JA403–05. Apple’s response to the injunction complied with its express terms and addressed the underlying competitive harm identified in the decision.

The UCL reaches conduct that threatens an incipient antitrust violation or otherwise significantly harms competition. JA487 (citing the Supreme Court of California in Cel-Tech Commc’ns, Inc. v. L.A. Cellular Tel. Co., 20 Cal. 4th 163, 186–87 (1999)). It also reaches “unfairness to consumers . . . us[ing] the balancing test, which ‘weigh[s] the utility of the defendant’s conduct against the gravity of the harm to the alleged victim.’” JA487 (quoting Progressive W. Ins. Co. v. Super. Ct., 135 Cal. App. 4th 263, 285 (2005)).

The Ninth Circuit applied both standards here, explaining that they “are not mutually exclusive.” JA487. Under either approach, a conduct’s benefits to competition and consumers are just as much a part of its competitive effects as are its costs. As the Supreme Court of California noted, when fashioning equitable remedies under the UCL, the court must consider “the equities on both sides of the dispute.” Cortez v. Purolator Air Filtration Prods. Co., 23 Cal. 4th 163, 180 (2000). To ask only whether Apple could do more to favor developers would truncate the competitive inquiry that justified the original judgment and injunction.

A. Commissions finance valuable services and encourage continued investment in them

Apple’s IAP processes payments and collects compensation for the broader services Apple supplies. The Ninth Circuit described Apple’s developer agreement as providing tools and “access to Apple’s vast consumer base” in exchange for “a $99 fee and an ongoing 30% commission on developers’ iOS revenue.” JA418. Payment processing is one of those tools. But moving payment processing to another provider does not eliminate developers’ access to consumers and use of other tools, nor does it obviate the need for an annual fee and continuing commission.

In the absence of a commission, a developer could, for example, make an app available for download without charge and earn revenue later through purchases of content used in the app, effectively cutting off Apple’s mechanism for compensation for distribution, software, security, and access. The district court recognized this distinction and explained that Apple could collect a commission even without requiring use of IAP, although another collection method could be more costly to administer. JA347–48 & n.617. The distinction between payment processing and other attributes of Apple’s system does not disappear when a customer follows a link to complete a purchase outside Apple’s App Store. The app still uses the platform, and the developer may still have acquired the customer through Apple’s distribution services. Ensuring Apple receives compensation for those contributions does not become economically baseless when a different firm happens to handle the final payment.

In addressing Apple’s app-distribution and in-app-payment-processing restrictions under the UCL, the district court noted that “Apple has proffered mostly valid and non-pretextual procompetitive justifications. To a large extent that makes the conduct more than ‘not anticompetitive’ but potentially beneficial to consumers.” JA372–73. Specifically, the court found that Apple’s app distribution restrictions “increase security in the ‘broad’ sense by allowing Apple to filter fraud, objectionable content, and piracy during app review while imposing heightened requirements for privacy.” JA270. The court also recognized that “security and privacy have remained a competitive differentiator for Apple,” JA271, and noted that “[t]he witnesses are unanimous that user security and privacy are valid procompetitive justifications.” JA272.

The court also agreed “with the general proposition that Apple is entitled to be paid for its intellectual property,” and that Apple “invests enormous sums into developing new tools and features for iOS.” JA277. It also recognized that “[b]oth Apple and third-party developers like Epic Games have symbiotically benefited from the ever-increasing innovation and growth in the iOS ecosystem.” JA87.

Having observed that the evidence at trial did not resolve the question of the extent to which Apple’s investments in its ecosystem were specific to the App Store “with precision,” the court strained to qualify that lack of precision, finding that “the record is devoid of evidence that Apple set its 30% commission rate as a calculation related to the value of its intellectual property rights.” JA278. More curious still, the court therefore found “that with respect to the 30% commission rate specifically, Apple’s arguments are pretextual, but not to the exclusion of some measure of compensation.” JA278. That was, at least, a recognition that the UCL does not require a commission rate of zero.

The court nonetheless discounted the connection in its initial judgment, reasoning that “[a] UCL action is equitable in nature,” and that “[c]ourts have ‘broad discretion’ to fashion equitable remedies to serve the needs of justice.” JA373 (internal citations omitted).

Whatever mechanism the injunction leaves available, the underlying economic problem does not disappear once anti-steering provisions are removed. This Court has recognized that vertical price restraints can properly be used to prevent free riding. Leegin Creative Leather Prods., Inc. v. PSKS, Inc., 551 U.S. 877, 890–91 (2007) (explaining that absent vertical price restraints, discounting retailers can “free ride on retailers who furnish services” such as showrooms and demonstrations, and finding that it is proper to impose restraints that alleviate the problem by encouraging retailers to “invest in tangible or intangible services or promotional efforts that aid the manufacturer’s position as against rival manufacturers”).

Left unaddressed, free riding weakens the incentive to supply the services that made the underlying transaction possible in the first place and to keep investing in them. Id. Courts have recognized that vertical restraints—including anti-steering rules that restrict point-of-sale merchant steering—can curb negative externalities by preventing retailers from free-riding on network investments. See Ohio v. Am. Express Co., 585 U.S. 529, 547 (2018) (quoting Leegin, 551 U.S. at 890–91, for the proposition that “vertical restraints can prevent retailers from free riding and thus increase the availability of ‘tangible or intangible services or promotional efforts’ that enhance competition and consumer welfare”). The injunction here forecloses that particular tool. It does not, however, eliminate the economic problem a restriction on steering might otherwise have addressed.

A developer that uses Apple’s distribution, discovery, curation, and security-review services to acquire and retain users and then completes the resulting sale through a channel that contributes nothing toward the cost of those services captures the value of Apple’s investment without paying for it. This is the same dynamic that concerned this Court in Leegin. A commission on such purchases addresses that problem by collecting compensation when the developer earns revenue from using Apple’s services. Its effect on developers’ willingness to offer links must be assessed together with its role in financing those services. If anything, that approach is less restrictive of developer and consumer conduct than the vertical restraint this Court sustained in Leegin, because it regulates only compensation for services actually used, not communication about, or access to, an alternative purchasing channel.

The allocation of those charges also matters to competition. As this Court explained in Am. Express, prices on a two-sided platform reflect the interdependence of its users; higher merchant fees alone do not establish anticompetitive effects. 585 U.S. at 535–37. Commissions on successful transactions can help finance services available to developers who generate little or no revenue, supporting low initial charges that encourage entry and experimentation. Reducing those commissions may benefit the developers who pay them, but it can also shift charges to other participants or weaken incentives to maintain and improve the shared services.

The competitive inquiry must include Apple’s ability to continue competing. Developers benefit from improvements in the tools, distribution services, and customer experience that make a platform useful. Rules that reduce the return from supplying those services can weaken the incentive to improve them and shift their costs to other participants. A remedy aimed at competition must consider those consequences alongside the immediate benefit a developer receives from paying a lower fee.

B. Competition does not require every alternative to offer the same financial advantage

That inquiry also distinguishes permission to communicate from a guarantee about the commercial results of communication. Better information can help customers compare available offers and encourage sellers to improve them. The resulting competition may change prices, quality, and the division of revenue among businesses. Its value does not depend on ensuring that every purchasing channel attracts transactions or that the platform providing access to customers receives less compensation.

As noted, the Ninth Circuit’s commission analysis concentrated on the difference between Apple’s 30 percent in-app commission and its maximum 27 percent external-purchase commission, together with developers’ additional costs of completing a purchase elsewhere. It reasoned that those outside costs exceeded the difference and make the external alternative economically nonviable. Pet. App. 19a–21a.

That comparison focuses on only one source of commercial attraction (lower transaction charges), but developers may prefer a particular purchasing mechanism for other reasons. As the Ninth Circuit recognized in the initial appeal, it is incorrect to conclude that a developer would have no reason to choose another in-app payment processor if Apple charged a commission. Different refund practices, a closer customer relationship, or changes affecting fraud, for example, could make another arrangement attractive. JA478. Those features illustrate why alternatives can compete without offering an identical service at a lower price.

In fact, the district court also recognized that competition is multidimensional, and that part of the purpose of the injunction was to preserve competition through product differentiation—including by Apple. In its order rejecting Apple’s request for a stay of the injunction, the court made clear that it “did not enjoin [the availability of IAP,] but rather enjoined the prohibition to communicate [alternatives.].” Stay Order at 4. The intent was that Apple would have the incentive to compete by offering a better alternative. Thus the court acknowledged Apple’s assertion that IAP was more convenient for users and specifically noted that “if it can compete on pricing, developers may opt to capitalize on that convenience, including any reassur[ance] that Apple provides to consumers that it may provide a safer or better choice.” Id.

For similar reasons, it is a mistake to infer competitive effects from one developers’ idiosyncratic preferences. A channel unattractive to one developer may be useful to another. Some developers may value greater control over ongoing subscriptions, for example, while others—particularly less-established and smaller developers—may value the opportunity to establish and maintain a direct commercial relationship with users. Developers also differ in the costs of operating their own purchasing services and in the number of sales over which they can spread those costs. Requiring every alternative to produce a particular margin of savings would make one price comparison control the inquiry despite those differences.

Because of these differing developer characteristics and preferences (as well as the free-rider problem discussed above), whether a commission frustrates the conduct the injunction protects requires more than a superficial price comparison and recognition of one developer’s self-interested preference for a cheaper alternative. The commission’s effect cannot be understood simply by assuming that the competitive purpose of permitting information was to guarantee a cheaper transaction after accounting for outside payment costs. Rather, as the district court itself noted in its Stay Order, “[c]onsumer information, transparency, and consumer choice is in the interest of the public.” Stay Order at 4. Such considerations plainly implicate the full range of competitive attributes upon which a consumer might decide, not just price.

The original injunction removed specified restrictions on communications. Whether further limits on compensation would improve the competitive outcome requires an economic inquiry of its own. The breadth of the UCL supplies no answer to that inquiry.

IV. The spirit of an injunction cannot set commission rates

Nor can an appeal to the injunction’s “spirit” replace the requisite economic inquiry. In its 2025 decision, the Ninth Circuit recognized this Court’s instruction that “a decree will not be expanded by implication or intendment beyond the meaning of its terms,” and that those terms must be “‘read in the light of the issues and the purpose for which the suit was brought.’” Pet. App. 11a (quoting Terminal R.R. Ass’n of St. Louis v. United States, 266 U.S. 17, 29 (1924)). As noted above, those terms do not address commission rates at all. Nonetheless, citing one of its own earlier decisions, the Ninth Circuit argues that courts can “find a breach of the decree in a violation of the spirit of the injunction, even though its strict letter may not have been disregarded.” Pet. App. 11a-12a (citing Institute of Cetacean Research v. Sea Shepherd Conservation Society, 774 F.3d 935, 949 (9th Cir. 2014) (internal quotations omitted)).

A. Valuing Apple’s services and ensuring developers a margin of savings are different tasks

The injunction’s goal of promoting competition does not, by itself, tell Apple how much cheaper external purchasing must be. The contempt decision faulted Apple both for failing to establish that its commission reflected the value of its intellectual property and for choosing a rate that left insufficient savings after developers paid outside transaction costs. Pet. App. 135a–138a. Those criticisms ask different economic questions. One concerns what Apple contributes; the other concerns the financial attraction of a purchasing arrangement that depends on services supplied by other firms. The problem is that the two need not be the same.

Suppose a commission reasonably reflected the value of the software and distribution services a developer continued to use. That would not establish how much the developer must pay an outside processor, spend handling refunds, or devote to operating a website. Those costs could leave the developer with little reason to use the external channel even though the value of Apple’s contribution had been properly assessed. Conversely, a developer’s ability to operate an inexpensive purchasing service would not tell a court what Apple’s contribution was worth.

A court could require a commission to satisfy both conditions, but doing so would require it to decide which developers’ costs matter, what margin is sufficient, and which services Apple may charge for. That would be a substantive choice about the terms of trade. Neither a general objective of promoting competition nor an instruction to permit external purchasing communications determines those terms.

Similarly, a firm ordinarily prices by considering what customers will pay and how its proposed charge compares with alternatives. It need not begin every pricing decision by assigning a separate value to every component of its offering. If a particular valuation method is required, that is an additional constraint on the decision.

But the injunction prescribes neither that method nor a required relationship between the combined costs of the two purchasing channels. Pet. App. 166a–167a. Indeed, the injunction says nothing about commissions at all, let alone about how to calculate one. The Ninth Circuit recognized this and remanded the case to the district court with instructions to “amend the April 30 Order’s commission prohibition,” along with recommendations on how to do it. Pet.App.35a–36a. A later preference for one particular approach to pricing cannot establish that choosing a different approach was incompatible with the competitive spirit of the injunction.

B. The Ninth Circuit’s opinion would require continuing judicial decisions about prices and investment

This case illustrates how quickly a commission dispute becomes a rate-setting exercise “beyond the practical ability of a judicial tribunal to control.” Brooke Group, 509 U. S. at 223. The Ninth Circuit rejected the permanent ban on commissions and contemplated a prohibition lasting until the district court approved an appropriate fee. Its proposed approach would allow compensation for costs necessary to coordinate external links and for intellectual property directly used in external purchases, with reductions reflecting the non-use of intellectual property reading solely on Apple’s IAP. It also proposed accommodations for security and privacy features and possible assistance from experts or a technical committee. Pet. App. 34a–36a.

Those instructions require much more than merely identifying the expense of processing an external payment. The court would have to decide which parts of the platform contribute to an external purchase, what compensation they warrant, and how to divide costs among uses. Software can support an app whose customers buy through several channels. Distribution can bring a customer to the app long before a purchase. An expense can support both activities without being uniquely attributable to either transaction.

Describing an allocation as proportionate does not resolve the commercial question. The court would still need to choose a measure of use and explain why that measure should determine the allocation of costs or compensation. The number of transactions, the revenue they generate, and the value different developers receive can point toward different charges. A division based on current use can also change as developers respond to the resulting prices. The decision thus requires judgments about behavior and value, as well as accounting.

The problem becomes only more difficult over time. A new service may support several kinds of purchases; a change in outside processing costs may alter the relative attractiveness of different channels; an improvement may require expenditure before it generates revenue. If the permissible commission depends on maintaining a particular commercial advantage for external purchasing, those changes can require further judicial decisions. The court must then administer the continuing relationship among businesses whose products and costs keep changing.

California’s courts have recognized this difficulty in applying the UCL. In California Grocers Ass’n v. Bank of Am., a trial court limited a bank service fee through an injunction based on its assessment of costs and an allowable profit. 22 Cal. App. 4th 205, 211–12 (Ct. App. 1994). The appellate court held the ongoing limitations inappropriate, explaining the institutional problems with judicial supervision of a particular service fee. Id. at 218–19.

Further, the existence of broad remedial authority does not itself equip a court to choose and maintain commercial prices. See Willard v. AT&T Commc’ns of Cal., Inc., 138 Cal. Rptr. 3d 636, 641 (Ct. App. 2012) (collecting cases on the proposition that where a UCL action would drag a court of equity into a complex economic policy question like setting prices, abstention is appropriate because policy is “primarily a legislative and not judicial function”).

This Court has made the same economic concern central to its competition decisions. Trinko cautioned against requiring courts to administer terms of dealing, including prices, in ways that can impair investment and competition. 540 U.S. at 407–08, 414–15. And National Collegiate Athletic Ass’n v. Alston emphasized the difficulties of continuing judicial supervision and the importance of leaving room for business decisions in implementing relief. 594 U.S. 69, 102–06 (2021). These concerns apply with particular force when a court derives detailed compensation terms from an injunction that supplied no pricing standard.

The practical effect is the district court is put in the position of managing Apple’s business relationships with developers, right down to the cent of what is charged as a commission. This type of “micromanaging is beyond the scope of court oversight contemplated by injunction law.” LabMD, Inc. v. FTC, 894 F. 3d 1221, 1237 (11th Cir. 2018). See also Stay Order at 3 (noting the court is not “here to micromanage”).

C. Expanding an injunction by reference to its spirit can discourage beneficial conduct

A business subject to an injunction must decide which activities it can continue, how it should change its arrangements, and which investments remain worthwhile. If complying with the order’s stated terms leaves the firm exposed to contempt because a court later prefers a more expansive understanding of its purpose, the firm may decide to abandon commercially valuable choices the order left available.

And that pressure operates in only one direction. An appeal to an injunction’s spirit can be used to demand restrictions beyond those stated in the order; the firm cannot invoke that spirit to disregard an express command.

Adding such an enforcement rule therefore encourages firms to do more than the order requires whenever the court might later demand an additional restriction. In a competition case, that can mean foregoing prices, products, or investments that serve customers but expose the firm to a court’s disagreement about how best to promote competition.

The resulting costs can be difficult for courts to observe. Cf. Frederic Bastiat, That Which Is Seen, and That Which Is Not Seen (1850).[4] A complaint about continuing restrictions creates a dispute that can be investigated and reviewed. A service improvement abandoned because its permissible compensation is uncertain may produce no comparable record. Neither will a pricing arrangement discarded to avoid another round of litigation. An enforcement approach aimed at improving the competitive process must account for both these “unseen” lost opportunities as well as visible conduct that fails to comport with an injunction. That is the practical importance of the error-cost framework in competition law. See Frank H. Easterbrook, The Limits of Antitrust, 63 Tex. L. Rev. 1, 2–5 (1984); Geoffrey A. Manne & Joshua D. Wright, Innovation and the Limits of Antitrust, 6 J. Competition L. & Econ. 153, 157–59 (2010).

Apple supplies services that retain value regardless of where payment occurs. Commissions can support their continued provision, and the terms of compensation involve choices that the injunction does not resolve. Treating those choices as settled by an appeal to the injunction’s spirit of facilitating competition risks suppressing the conduct that competition law seeks to encourage. Those economic consequences leave a “fair ground of doubt as to the wrongfulness of the defendant’s conduct” under the injunction. Taggart, 587 U.S. at 561 (quoting California Artificial Stone Paving Co. v. Molitor, 113 U.S. 609, 618 (1885)).

Conclusion

Applying this Court’s contempt standard requires reversing the judgment below. The commission-related contempt determination and the remedial directions dependent on it should be set aside.

^[1] No counsel for a party authored this brief in whole or in part, and no party or counsel for a party made a monetary contribution intended to fund its preparation or submission. No person other than Amicus and their counsel made a monetary contribution to the preparation or submission of this brief.

[2] Apple’s link-out commission generally applied only to purchases made within seven days after a user followed an in-app External Purchase Link, but renewals of subscriptions initially purchased within that period could remain subject to commissions after the seven days expired. JA518–19. Apple estimated that this seven-day limitation brought the effective rate closer to 18%, rather than the 27% figure the decisions below treated as controlling. Pet. App. 77a n. 25.

[3] Epic challenged some of the assumptions underlying Apple’s estimate, and the district court questioned the extent of its evidentiary support. 6-ER-1140–42; Pet. App. 77a n.25. But even without accepting Apple’s specific estimate of an 18% effective commission, the record supplies a firm basis to expect a rate below 27%. Apple’s customer-spending data and experience with direct purchasing channels supported the expectation that some purchases would occur on that basis. JA518–20; Feb. 25, 2025 Tr. 1588:14–19, 1590:15–1591:18 (Oliver), Dkt. 1307. The presence of any such purchases would reduce the average commission below the headline rate.

[4] Available at https://mises.org/articles-interest/which-seen-and-which-not-seen.

ICLE Director of Competition Policy Dirk Auer joined The Center Edge to discuss AI safety, whether leading AI labs need an antitrust exemption to coordinate development slowdowns, how existing law applies to AI, and whether new regulation could shield established firms from competition.

New York City has found an unlikely culprit for unsafe deliveries: the name on a driver’s paycheck. Its proposed cure would force Amazon to bring thousands of drivers in-house. The evidence that this would improve safety is thin. The implications for union organizing are anything but.

Last month, New York City Mayor Zohran Mamdani formally endorsed the Delivery Protection Act (DPA), a City Council bill sponsored by Council Member Tiffany Cabán. Though framed as a warehouse-licensing and safety measure, the bill would also require Amazon to employ its delivery drivers directly rather than contract with independent delivery businesses.

Just six days before Mamdani’s announcement, New Jersey sued Amazon over substantially the same delivery arrangements. The state alleges that Amazon uses its control over Delivery Service Partners (DSPs) to suppress compensation and worsen working conditions, including by pressuring drivers to meet unsafe delivery expectations. The geographic markets alleged in the lawsuit include New York City.

Both governments portray Amazon as directing the work while DSPs bear formal responsibility. Yet New Jersey treats limits on DSP independence as part of the problem, while New York would virtually eliminate that independence. Different statutes explain the different legal claims, of course. But they don’t explain how substantially the same conditions and harms can result from too little DSP independence in New Jersey, yet be cured by eliminating DSPs as employers in New York.

That contradiction frames the DPA’s central problem. Its supporters haven’t shown that DSP employment causes the harms they cite, or that direct employment would reduce them. Amazon’s extensive safety programs further complicate the case for abolishing the DSP model in the name of safety. What the mandate would unquestionably change is how New York City delivery drivers unionize and who they bargain with.

Read the full piece here.

Executive Summary

The Digital Markets Act (DMA) directs its regulatory machinery toward “fairness” and “contestability,” yet defines neither goal clearly. Critics argue that this ambiguity gives the European Commission broad discretion, while defenders note that terse competition laws have gained meaning through enforcement and adjudication. This white paper asks whether two years of DMA enforcement have begun to fill those gaps. It reconstructs the Act’s implicit theory from its text, then tests that theory against the Commission’s impact assessment, designation and infringement decisions, specification proceedings, statutory review, and DMA–GDPR guidance.

The record reveals that the DMA’s goals are coherent but redistributive. Fairness means correcting a structural imbalance the Act conclusively presumes between gatekeepers and those who depend on their core platform services. Contestability supplies the remedy by transferring competitive opportunity, resources, and market share toward rivals and business users. User choice serves both goals, but the Act values it primarily when users move away from gatekeepers.

Four concepts explain this logic. The Structural Unfairness Axiom presumes that gatekeeper markets are inherently unfair. The Bidirectional Unfairness Inference allows structural asymmetries and disappointing outcomes to confirm each other as evidence of unfairness. The Scale–Unfairness Paradox treats the same growth that improves a platform as evidence of deeper entrenchment. Dialectical Fairness seeks to level up business users while leveling down gatekeepers. The enforcement record reflects these ideas by treating gatekeeper growth as weak contestability, defining fairness through prohibited conduct, elevating parity without effects analysis, and measuring success through third-party access and uptake rather than price, quality, or innovation.

What the record still lacks is calibration. The DMA identifies who should gain and who should lose without specifying how much, for how long, or against what benchmark. Because its logic assumes that covered markets were never fair, it supplies no baseline or endpoint. Compliance may therefore remain unverifiable and intervention potentially endless. That problem matters for gatekeepers, reviewing courts, consumers, and jurisdictions importing DMA-style rules without fully understanding the program they embody.

I. Introduction

The Digital Markets Act (DMA or Act)[1] has drawn persistent criticism because its twin goals—fairness and contestability—remain ambiguous.[2] Critics raise a familiar objection: Broad objectives allow the European Commission to shift the goalposts and exercise considerable discretion over enforcement and interpretation, weakening the system’s predictability and legal certainty.[3]

Yet the DMA is hardly unique in this respect. Landmark competition laws—the Sherman Act in the United States, most famously, and Articles 101 and 102 of the Treaty on the Functioning of the European Union (TFEU) to a lesser extent—are famously terse. Their underlying logic took shape gradually through litigation and enforcement.[4] Such openness may reflect a deliberate legislative choice to let enforcers, regulated parties, and courts fill the gaps.[5] It may instead, however, reveal poor drafting or a legislature that lacked a clear vision of what it sought to achieve.

Against this backdrop, this white paper asks a simple question: Have two years of enforcement clarified the DMA’s goals and, if so, how?

The question matters to both scholarship and policy. Scholars remain divided over two related issues: whether the DMA’s goals can be meaningfully defined and what those goals require. The literature has debated whether fairness and contestability are coherent, compatible, or even legally cognizable objectives. Few studies, though, have examined how concrete enforcement decisions, interpretive communications, and early institutional practice have given those concepts meaning.

Greater clarity would also provide practical guidance. It would help gatekeepers and firms seeking access to understand what compliance requires and why. It would also help jurisdictions considering DMA-style regulation—including the United Kingdom, Japan, and Brazil[6] —understand the rules they may be importing, including their logic, principles, underlying aims, likely results, and shortcomings.[7]

This white paper argues that the DMA does contain a coherent conception of fairness and contestability, but the Act never states it clearly. That conception emerges only by reading the law alongside the European Commission’s enforcement decisions. Properly reconstructed, fairness and contestability are not free-standing values in tension, as much of the literature assumes. They are, instead, two sides of a single redistributive program.

Fairness supplies a conclusive diagnosis: Gatekeepers and the businesses that depend on them suffer from a structural imbalance that designation conclusively presumes. Contestability supplies the remedy: a continuing transfer of competitive opportunity away from gatekeepers and toward actual or potential rivals. User choice serves both goals, but the DMA only values that choice when (and because) users exercise it away from gatekeepers. This structure explains why the Act can treat “fair and contestable” as a hendiadys—two terms expressing a single idea.

That coherence comes at a price. The DMA’s goals are directional but uncalibrated. The Act identifies who should lose and who should gain, but not how much, for how long, or against what benchmark. Because it presumes that markets for core platform services were never fair, it supplies no earlier equilibrium to restore and no baseline against which to measure progress. Any improvement may therefore prove insufficient, making compliance impossible to verify.

Two years of enforcement have brought us closer to understanding what fairness and contestability mean under the DMA. They have brought us no closer to knowing when, if ever, the Act’s goals will have been achieved.

The remainder of this white paper proceeds as follows. Section II situates the inquiry within the scholarly debate. It explains why the DMA’s goals require an abstract definition, why the Act fails to provide one, and how commentators have divided over what those goals entail. Section III reconstructs the DMA’s implicit theory of fairness and contestability from its recitals and operative provisions. It introduces four analytical constructs that organize the analysis: the Structural Unfairness Axiom, the Bidirectional Unfairness Inference, the Scale–Unfairness Paradox, and Dialectical Fairness.

Section IV tests that reconstruction against the enforcement record, including the impact assessment, designation and infringement decisions, specification proceedings, the Commission’s first statutory review, and joint guidelines governing the interplay between the DMA and the General Data Protection Regulation (GDPR). Section V distills the findings into an explicit reformulation of the DMA’s twin goals. Section VI examines the implications of what remains unknown. Section VII concludes.

II. The Debate Over the DMA’s Undefined Goals

Scholars agree that fairness and contestability must guide the DMA’s interpretation, but the Act defines neither goal nor clearly connects them to its obligations. This ambiguity has produced competing accounts of the DMA’s legal character, aims, and criteria for success.

A. Why the DMA’s Goals Require Definition

The early literature on the DMA’s goals converges on three related conclusions. First, the Act requires an abstract account of fairness and contestability. Their meaning cannot simply be inferred “by doing”—that is, from the prohibitions and obligations in Articles 5 through 7.

Giorgio Monti argues that conceptual clarity would guide courts and help stakeholders understand the Act’s policy objectives.[8] Richard Feasey and Monti likewise treat a clear statement of the DMA’s goals as essential to assessing both compliance and the Regulation’s success. In their account, the DMA creates an “obligation to provide a result on the market,” not merely a set of formal duties.[9] Jacques Crémer et al. similarly argue that courts and regulators need a working definition of fairness and contestability to interpret Articles 5, 6, and 7. The goals express policy aspirations, but they also guide the European Commission’s compliance assessments and any resulting judicial review.[10]

The DMA’s institutional design reinforces this point. The Act allows the Commission to specify what certain obligations require, reopen proceedings when compliance measures prove ineffective, and adapt the rules to new forms of unfairness or weak contestability. Each power presupposes a standard for deciding whether a compliance measure advances the Act’s goals. A list of practices labeled unfair cannot, by itself, supply that overarching standard.[11]

Second, the DMA does not provide one. Friso Bostoen observes that the Act seeks to ensure that markets in which gatekeepers operate “are and remain contestable and fair,” yet defines neither term.[12] Its closest approximation to a general definition of fairness appears in recital 33, which states that unfairness “should relate to an imbalance between the rights and obligations of business users where the gatekeeper obtains a disproportionate advantage.”

As Pinar Akman notes, that formulation leaves basic questions unanswered, including how fairness and contestability apply in markets that have already tipped toward a dominant platform.[13] Nicolas Petit argues that the vague terminology provides little reason to believe the legislature had a clear theory of competition or coherent economic policy in mind. The Act’s structure obscures both its primary objectives and the reasons the legislature selected some obligations over others, raising broader questions about its legitimacy.[14] Giuseppe Colangelo goes further, portraying fairness as a mantra that grants enforcers broad discretion rather than a meaningful legal standard. As he writes, “the endemic uncertainty surrounding the notion of fairness has traditionally made it unsuitable to operate as a stand-alone legal standard.”[15]

Third, the relationship between the DMA’s goals and obligations remains unclear, perhaps because the boundary between fairness and contestability is itself blurred. Monti argues that the Act does not identify which provisions advance which objective. Grouping obligations by policy objective, he suggests, would clarify the Act’s approach and aid judicial interpretation.[16]

Petit similarly finds that “the structure of the DMA makes it very hard to unpack the primary objective of each obligation.” Fairness and contestability do not necessarily reinforce one another. Conduct deemed unfair could, in some circumstances, encourage entry and thereby improve contestability.[17] Akman draws a related distinction: Contestability largely concerns market structure, while fairness primarily concerns conduct. She doubts that this “two-headed” design can deliver greater contestability in markets where gatekeepers already hold entrenched positions.[18]

B. Competing Accounts of the DMA’s Goals

A third conclusion follows from the first two: If the DMA does not adequately define the objectives that guide its interpretation, competing accounts will fill the gap. The literature offers several.

One dispute concerns whether the DMA extends competition law or creates a separate regulatory regime. This may appear to be a formal distinction, but it determines how far established competition-law concepts, benchmarks, and safeguards should constrain enforcement. Some scholars find the DMA’s effort to distance itself from competition law unconvincing because the Act’s recitals connect its intended improvements to traditional competition parameters, including innovation and consumer benefits.[19] Others view the DMA as more than streamlined enforcement of Article 102 TFEU. On this account, the Act looks forward and seeks to shape—and perhaps constrain—the conditions under which gatekeepers exercise power.[20] Still others cast it as a democratic reassertion of rulemaking authority over platforms that competition-law standards should not govern.[21]

Geoffrey A. Manne, Lazar Radic, and Dirk Auer classify the DMA as “digital competition regulation.” In their account, the Act departs from antitrust’s consumer-welfare standard and revives structural concerns and presumptions associated with neo-Brandeisianism and ordoliberalism—traditions the antitrust mainstream has largely rejected. The DMA places conduct that might otherwise fall under Article 102 within a separate framework whose goals, evidentiary standards, and presumptions more closely reflect the regulator’s policy preferences.[22]

Even sympathetic observers acknowledge uncertainty about the Act’s character. Pierre Larouche and Alexandre de Streel describe the DMA as occupying a “difficult and perhaps ominous epistemological position.”[23] Comparative analysis deepens the ambiguity because other jurisdictions have incorporated provisions identical to the DMA’s directly into their national competition laws. The result resembles a “Schrödinger’s regulation” in which the same substantive rules simultaneously are and are not competition law.[24]

A second dispute concerns the relationship between fairness and contestability. Crémer and his coauthors treat them as mutually reinforcing. They describe a “dual-action dial” under which a practice cannot qualify as fair if it undermines contestability.[25] By contrast, de Streel and his coauthors treat the goals as distinct but connected: Contestability lowers entry barriers and enables competitive pressure, while fairness balances rights and obligations within the platform ecosystem.[26] Petit rejects the premise that the two goals necessarily move together.[27] Akman likewise questions whether the same tools can achieve objectives that may pull in different directions.[28]

This disagreement extends to whether the DMA addresses the causes or merely the symptoms of weak competition. A regime focused on preserving contestability would logically target markets before they tip toward a dominant platform. The DMA instead regulates markets in which gatekeepers already exist. That choice leaves the Act’s underlying ambition unclear: Does it accept gatekeepers while regulating their conduct, or does it seek to reshape market structure and ultimately erode their positions?[29]

Because the DMA supplies no clear definitions, scholars have also reconstructed its broader purposes. Manne, Radic, and Auer identify three objectives shared by the DMA and similar regimes: leveling down gatekeepers, assisting rivals, and redistributing economic rents—returns above competitive levels—from gatekeepers to business users. They connect these objectives to the European Union’s broader pursuit of digital sovereignty and strategic autonomy.[30]

After reviewing gatekeepers’ compliance reports, Colangelo and Alba Ribera Martínez connect fairness and contestability to long-term goals of market modeling, ecosystem openness, neutralizing competitive advantages, and transparency. They characterize the fairness-related goals as expressly redistributive.[31] Oles Andriychuk offers a more direct assessment, characterizing the Act as “punitive.”[32]

C. Why Articles 5–7 Are Not Enough

One response is that the DMA needs no overarching definition because Articles 5 through 7 give fairness and contestability their full content. On this view, the two goals begin and end with the obligations themselves. But that account fails both functionally and on the Act’s own terms.

Recital 79 directs the European Commission to “apply a predefined standard” when identifying practices that are unfair or limit contestability. Such a standard would promote predictability and enable the Commission to exercise its other statutory powers. Article 12, for example, allows the Commission to extend the DMA through delegated acts to cover practices that the Act does not enumerate but that nonetheless qualify as unfair or contestability-limiting.

Other provisions likewise require the Commission to look beyond the obligations’ text. Article 18 authorizes remedies when gatekeepers have failed to achieve “fairness and contestability.” Article 19 permits market investigations to identify additional practices that are unfair or restrict contestability. Article 53 requires the Commission to assess every three years whether the DMA has achieved its objective of contestable and fair markets.

Article 8(1) makes the need for an overarching standard clearest. It requires a gatekeeper’s compliance measures to be “effective in achieving the objectives of this Regulation and of the relevant obligation.” This distinction implies that compliance depends on both the specific purpose of an individual obligation and the Act’s broader goals.[33]

Taken together, these provisions presuppose three things: a coherent meaning of fairness and contestability, a benchmark for measuring them, and an endpoint at which the Commission can declare them achieved. The Act supplies none. The question, then, is whether the DMA’s text and enforcement record allow those missing elements to be reconstructed.

III. Reconstructing the DMA’s Goals

Although the DMA never defines fairness or contestability, its recitals and operative provisions reveal a coherent direction. The Act treats gatekeeper scale and user dependence as sources of structural unfairness, seeks to improve business users’ position, and promotes rival entry and expansion by reducing gatekeepers’ competitive advantages.

Four analytical constructs organize this reconstruction. Together, they show that the DMA treats user choice as a tool for shifting activity toward rivals and contestability as market restructuring. The Act’s direction is clear, but it supplies no benchmark for measuring success or determining when intervention should end.

A. The DMA’s Implicit Theory of Fairness

The DMA defines fairness only indirectly, through its opposite. Recital 33 states that “[u]nfairness . . . should relate to an imbalance between the rights and obligations of business users where the gatekeeper obtains a disproportionate advantage.”[34] Recital 62 expands on that definition, holding that pricing or other general access conditions:

. . . should be considered unfair if they lead to an imbalance of rights and obligations imposed on business users or confer an advantage on the gatekeeper which is disproportionate to the service provided by the gatekeeper to the business users or lead to a disadvantage for business users in providing the same or similar services as the gatekeeper.[35]

The Act therefore treats digital markets dominated by gatekeepers as unfair because they distribute rights and obligations unequally across the supply chain. It attributes that inequality to persistent differences in bargaining power that allow gatekeepers to capture a disproportionate share of the value created on their platforms while business users receive inadequate returns on their contributions.[36]

This theory rests on the assumption that users, especially business users, depend on gatekeepers’ core platform services (CPSs). The Act does not distinguish between business users that depend on a gatekeeper and those that do not. Nor does it distinguish firms whose dependence results from their own choices from those that could not reasonably avoid it.[37] This omission implies that the Act treats dependence as an inherent structural feature of CPSs. Bargaining asymmetries increase dependence, which further entrenches those asymmetries and weakens business users’ position.

A closer reading reveals four related themes. First, the DMA treats gatekeeper scale and other structural market features as sources of unfairness that both produce and confirm unfair outcomes. Second, it conceives fairness as a relative redistribution that improves business users’ position while curtailing gatekeepers’ advantages. Third, it values user choice primarily as a means of encouraging switching away from gatekeepers. Fourth, it invokes fair prices, innovation, quality, and other desirable outcomes without providing benchmarks for measuring them. Together, these themes supply the DMA’s implicit theory of fairness.

1. The Structural Unfairness Axiom

The DMA begins with the premise that markets for core platform services (CPSs)—and gatekeepers in particular—are structurally unfair. This white paper calls that conclusive presumption the DMA’s Structural Unfairness Axiom (DSUA). The Act attributes the unequal distribution of rights and obligations between gatekeepers and their business and end users to market features such as network effects, data-driven economies of scale, tipping, and lock-in.[38] On this reading, those features are inherently unfair and create imbalances that only regulation can correct. Unfair inputs necessarily produce unfair outcomes.

The logic is deliberately circular. If an outcome appears unfair, the market characteristics that produced it are presumed unfair. If the market is structurally unfair, its outcomes are necessarily tainted. Competition law disciplines such reasoning through external benchmarks such as consumer welfare and efficiency. The DMA offers no comparable yardstick.

The Act instead asks whether allegedly unfair structural features have been neutralized and whether the resulting distribution is fair. Each inquiry depends on the other. This white paper calls that feedback loop the Bidirectional Unfairness Inference (BUI): Structural features are deemed unfair and therefore presumed to produce unfair outcomes, while outcomes perceived as unfair confirm the structural features’ inherent unfairness.

The DSUA also assumes that the potential for unfairness scales with the size of a CPS. Within the Act’s logic, a larger user base increases the risk that the platform’s structural features will produce unfair outcomes. Network effects and data accumulation can improve a platform’s quality while deepening users’ dependence on it. As the platform becomes more attractive, users become less likely to switch. The same features that make a CPS work well thus make it, in the DMA’s view, inequitable and prone to unfairness.

This white paper calls that dynamic the Scale–Unfairness Paradox (SUP): Greater scale can simultaneously improve the quality and attractiveness of a CPS and heighten the risk of structural unfairness. The result is a partially inverse relationship between DMA-style fairness and service quality. The dynamics that improve the service are the same dynamics that, under the DMA’s theory, undermine fairness.

The DMA’s designation criteria make this logic especially visible. The Act’s use of market capitalization, turnover, and user numbers[39] treat a company’s scale as the ultimate source of unfairness. High turnover might ordinarily suggest that a firm creates substantial value for consumers. The DMA instead reads it as evidence of leveraging potential, market tipping, and latent unfairness.

Once a firm crosses the Article 3 thresholds and becomes a gatekeeper, the Act prohibits specified conduct that would otherwise remain lawful and commonplace. A company does not become a gatekeeper because it acted unfairly. The Act treats its conduct as unfair because the company is a gatekeeper. The recitals extend this logic beyond the quantitative thresholds. They treat the ability to monetize a service, “superior access” to financial markets, and high relative growth as indicators of gatekeeper status and entrenchment—and therefore of unfairness or its potential.[40]

If this structural account is correct, the DMA does not restore fairness so much as construct it. Markets for CPSs are unfair under the Act because of their inherent characteristics, which means they were never fair in the sense the DMA now uses that term. The Act has no prior equilibrium to restore and no historical benchmark against which to measure future success.

This account also yields an implicit operational test. Conduct that increases user numbers, turnover, market capitalization, monetization potential, scale, or access to capital strengthens the structural conditions associated with gatekeeper status. By reinforcing the conditions the Act treats as unfair, such conduct becomes evidence of unfairness itself.

2. Dialectical Fairness: Leveling Up and Down

If the current distribution of rights and obligations is unfair, fairness requires a different allocation. Because the DMA identifies business users as the disadvantaged group, the new allocation must leave them better off relative to gatekeepers. Recital 62 illustrates this logic: Unfairness arises from both a “disadvantage for business users” and an advantage “disproportionate” to the gatekeeper.

Fairness under the DMA is therefore relative and dialectical. It measures each group’s position against the other and seeks movement in both directions: leveling up business users while leveling down gatekeepers. The goal is a new allocation that narrows the gap between them. Fairness must leave business users better off or gatekeepers worse off—and preferably do both.

The Act’s operative provisions reveal how this redistribution works. Article 5(2) restricts a gatekeeper’s ability to process, combine, or cross-use personal data without consent. The provision might appear to address privacy. Recital 36 frames it differently: Cross-use is problematic because it gives gatekeepers an advantage over competitors, allowing them to combine data sources, improve their services, and undermine contestability.[41]

The provision targets gatekeepers’ competitive data advantages rather than data collection itself. By contrast, other provisions—including Article 6(9)’s data-portability mandate—encourage data sharing and movement. The DMA instead treats gatekeepers’ data advantages as a source of unfairness precisely because those advantages have competitive value. On this view, gatekeepers should not compete using resources their rivals lack.

Recital 72 applies the same logic to consumer profiling. It states that greater transparency should discourage gatekeepers from making intensive profiling the industry standard, “given that potential entrants or start-ups cannot access data to the same extent and depth, and at a similar scale.”[42] The DMA thus de facto favors privacy-intensive models over data-driven ones. It both advantages firms without comparable data access and defines the terms on which the Act expects them to challenge incumbents.

Recital 46 and Article 6(2) follow the same pattern. Accumulating business-user data is not unfair per se. Unfairness arises when a gatekeeper uses that data in competition with the business users that generated it.[43] The provision prohibits such use in any market, even when the resulting competition might benefit end users.

Here, “unfair” means that business users could receive more. All else equal, they would benefit if gatekeepers exerted less competitive pressure against them. Fairness therefore operates as a mandate to relax the competition business users face by preventing gatekeepers from deploying competitively valuable resources against them.

Article 6(5)’s self-preferencing prohibition offers the clearest example. Gatekeepers may not favor their own products, but the Act does not prevent them from favoring third-party offerings. Article 6(7)’s interoperability mandate and Article 6(10)’s real-time data-access requirement likewise compel gatekeepers to share the products of their investments with third parties and competitors. These obligations reduce gatekeepers’ advantages and place the two sides on more equal footing.[44]

The underlying premise is that a gatekeeper should share its competitive advantages because its exclusive possession of them is inherently unfair. More generous access terms and lower prices for business users qualify as fair because they move the parties’ positions in both directions. They improve business users’ position, curtail the gatekeeper’s “disproportionate advantages,” and narrow the imbalance the DMA seeks to correct.

3. Choice as Directed Switching

Several DMA recitals suggest that practices become unfair when they restrict user choice. Recital 70, for example, instructs gatekeepers not to “subvert or impair user autonomy, decision-making, or choice.”[45] That language might suggest that choice is an independent value under the Act. A closer reading reveals a directional preference: The DMA encourages users to choose away from gatekeepers’ products and core platform services and toward rival offerings.

The Act’s asymmetries are systematic. Article 6(3) requires gatekeepers to make changing default settings easy, but only when those defaults direct users toward the gatekeeper’s own services. The Act does not prohibit defaults that steer users toward third-party products, even though such steering could also “subvert” user autonomy within the meaning of recital 70.[46]

Article 6(4) likewise requires gatekeepers to permit third-party apps and app stores and allows those third parties to prompt users to make their products the default. EU law imposes no corresponding obligation on other firms to open their systems to gatekeepers or anyone else. Defaults that might otherwise lock in users become unobjectionable—and perhaps desirable—when they benefit a firm other than a gatekeeper.

Article 5(3) prohibits parity clauses, which prevent business users from offering better terms through other channels, because they “unfairly” restrict business users’ freedom. The gatekeeper’s freedom to adopt such clauses receives no comparable weight, even when the clauses prevent businesses from free-riding on the platform.[47] Articles 5(4), 5(7), and 5(8) follow the same pattern by reducing reliance on gatekeepers and easing users’ path to alternative products and services. In each case, the Act promotes choices that move users away from gatekeepers.

If user choice were an overarching objective, the Act would express it as a general rule. All firms would have to make software easy to uninstall, and no firm could steer users toward its preferred channels. The DMA instead values choice only when it advances the Act’s other goals. It seeks to level down gatekeepers by directing users toward rivals and creating room for those rivals to challenge the gatekeeper’s position. The Act may even subordinate convenience to that goal, most visibly through mandatory choice screens that require users to take additional steps.

The DMA therefore values switching more than choice itself. Switching demonstrates contestability because it suggests that users have overcome entry barriers. It advances fairness by redirecting users and traffic away from gatekeepers. Under this logic, regulators may measure the effectiveness of choice mechanisms through observed switching rates and treat low switching as evidence of inadequate implementation. Choice becomes “meaningful” only when it leads away from the gatekeeper.

Taken to its logical conclusion, the Structural Unfairness Axiom implies that choices made before the DMA took effect were not genuinely meaningful. Structural unfairness tainted those choices, making them normatively suspect and unworthy of acceptance at face value.

4. Outcomes Without Benchmarks

Some commentators argue that the DMA seeks to create rights and opportunities rather than particular market outcomes.[48] Yet the Act does contain several outcome-oriented goals. Recital 5 refers to “ensuring fair economic outcomes” when market processes cannot do so. Recital 107 connects the Act’s objectives to “promoting innovation, high quality of digital products and services, fair and competitive prices, as well as a high quality and choice for end users.” Recital 45 asserts that advertising costs are “likely to be higher than they would be in a fairer, more transparent and contestable platform environment.”[49]

The Act provides no benchmark for these outcomes. It does not explain what constitutes a fair price, adequate innovation, or fair product quality. Competition law, the closest neighboring framework, relies on market power and the competitive counterfactual—a comparison with conditions absent the challenged conduct—to identify suboptimal outcomes.

The DMA expressly separates its analysis from those tools. Recital 23 makes market definition and efficiencies irrelevant to designation. Recital 11 states that the DMA pursues goals distinct from those of competition law and operates “without prejudice” to anticompetitive effects, objective justifications, and efficiency considerations.[50]

By excluding those considerations, the Act discards competition law’s principal method for determining when prices, output, or innovation are inadequate. Competition law asks whether market power distorts competition and harms consumers. The DMA supplies no equivalent test. Competition-law principles therefore cannot define “fair” outcomes under the Act, except perhaps at the margins.

B. The DMA’s Implicit Theory of Contestability

The DMA treats contestability as rivals’ ability to enter, scale, and take market share from gatekeepers. Although some provisions recognize competition on the merits, the Act primarily seeks to restructure markets by reducing gatekeeper advantages and promoting third-party growth.

1. Contestability as Market Restructuring

The DMA also leaves contestability undefined. Mirroring its negative account of fairness, recital 3 suggests that contestability is lacking when a gatekeeper’s position remains difficult to challenge “even by more innovative or efficient market operators.” It attributes that difficulty to high barriers to entry or exit and limited access to key inputs such as data.[51] Recital 5 adds that certain platform markets are “prone to tipping” and that gatekeepers hold entrenched and durable positions.

Two features of this account stand out. First, contestability extends beyond the formal freedom to enter a market. It asks whether rivals can attract users, achieve scale, and exert meaningful pressure on gatekeepers. The emphasis on effectiveness matters. Without observable entry, expansion, or shifts in market structure toward challengers, it becomes difficult to say that rivals can contest the market effectively. The DMA thus appears to infer inadequate contestability from rivals’ failure to gain a meaningful foothold. As with the Bidirectional Unfairness Inference, the absence of the desired outcome becomes evidence of a structural defect.

Second, contestability entails reducing entry barriers to induce actual entry and expansion by third parties, regardless of the barriers’ source. This includes advantages produced by otherwise legitimate business conduct, such as data accumulation. Recital 44, for example, objects to tying because it gives gatekeepers “a potential advantage in terms of accumulation of data” that “is liable to raise barriers to entry.”[52]

The treatment of online search follows the same logic. The DMA links contestability to gatekeepers’ data-driven quality advantages, then requires them to share query data with rivals. The premise is that access to those data will help rivals improve their services, attract users, and gain market share.[53]

Under this account, contestability requires reducing a critical advantage on one side so that the other can compete more effectively and take market share from incumbents. The Act treats access to data—and, by extension, scale and a large user base—as an illegitimate advantage when concentrated in gatekeepers, yet affirmatively provides that access to their rivals. That asymmetry reflects a broader objective of restructuring markets by dispersing market share among more firms.

2. Gatekeeper Growth as Evidence of Weak Contestability

If contestability requires entry and expansion by third parties, continued gatekeeper growth points in the opposite direction. The DMA operationalizes this implication of the Scale–Unfairness Paradox through Article 18.

In deciding whether to impose additional behavioral or structural remedies, Article 18 asks whether a gatekeeper has “maintained, extended or further strengthened its impact in the internal market.”[54] The inquiry does not consider whether that increased impact resulted from superior performance or a better product, even though growth in turnover and user numbers may reflect precisely those advantages.

Recital 75 confirms this approach. A gatekeeper “should be deemed to have maintained, extended or strengthened its gatekeeper position where, despite the enforcement actions taken by the Commission, that gatekeeper still holds or has further consolidated or entrenched its importance as a gateway for business users to reach end users.”[55]

The Act likewise presumes an “entrenched and durable position” when a core platform service reaches a very high number of users across several Member States for three years. By the same logic, contestability improves when a gatekeeper’s user numbers decline. If network effects, scale, and data advantages restrict contestability, reducing those advantages should enhance it.

Gatekeeper growth therefore cuts against the Act’s objectives even when it results from success on the merits. Taken to its logical conclusion, the DMA could treat a gatekeeper’s loss of market share as evidence that its compliance measures are working.

3. Limits of the Restructuring Account

Not every DMA provision fits the market-restructuring account. Article 6(5) permits a gatekeeper’s products to rank first when that result follows from transparent, fair, and nondiscriminatory criteria. This provision therefore supports an efficiency-based reading: A gatekeeper may prevail on the merits but may not leverage its position to distort competition in adjacent markets.

Other passages point in the same direction. Recital 53 states that offering better services or lower prices does not create a prohibited barrier to switching. Recitals 55 and 57 frame interoperability as a way to ensure that rival hardware competes on its merits.[56] These provisions support a consumer-oriented conception of contestability.

Other parts of the Act are difficult to reconcile with that conception. Article 5(2) restricts data combination without asking whether integration improves service quality. Article 6(11) requires search gatekeepers to share query and click data with rivals because the competitive advantage generated by legitimate scale and user engagement creates a structural obstacle. Article 6(2) bars gatekeepers from using business-user data even when doing so would produce a genuinely superior product. Recital 64 similarly treats the integration of communication services into an ecosystem as an entry barrier regardless of whether users actually multihome.[57]

This broad conception makes many consumer-facing integrations suspect. It also casts doubt on claims that the Act primarily seeks to benefit end users or increase innovation overall, rather than promote innovation by firms other than gatekeepers. If consumer welfare and innovation were controlling objectives, the DMA would distinguish efficient from inefficient entry and efficient from inefficient exclusion, as competition law does. It makes neither distinction.

The Act retains a residue of consumer-oriented contestability, but its broader structure subordinates consumer welfare to effective third-party entry and market restructuring.

C. The DMA’s Known Unknowns

A close reading reveals a more coherent account of fairness and contestability than the DMA expressly provides. Fairness rests on the Structural Unfairness Axiom (DSUA): Markets for core platform services are presumed structurally unfair because of their inherent characteristics. That presumption eliminates the need to prove anticompetitive conduct or harm.

The Bidirectional Unfairness Inference (BUI) reinforces the premise. Market structure is presumed to produce unfair outcomes, while outcomes deemed inadequate confirm that structural defects persist. The Scale–Unfairness Paradox (SUP) follows because growth and scale heighten the risk of unfairness even when gatekeepers earn them through competition on the merits. Dialectical Fairness supplies the remedy by leveling up business users while leveling down gatekeepers. The Act treats user choice as an instrument of that redistribution and directs it away from gatekeeper ecosystems.

The DMA’s conception of contestability likewise departs from the standard economic concept. Conventional contestability relies on the credible threat of entry to discipline incumbents. The DMA seeks to induce actual entry and expansion by transferring gatekeepers’ competitive advantages—including access to data, distribution, and interoperability—to rivals. It also treats continued gatekeeper growth as evidence of inadequate contestability regardless of the growth’s cause.

Taken together, these elements define fair and contestable markets as markets in which competitive advantages and market share move away from gatekeepers and toward rivals and business users. Consumer welfare remains an expected downstream benefit rather than the Act’s primary objective.

The Act nevertheless leaves five questions crucial to interpretation and enforcement unresolved. These are its “known unknowns.”

First, the DMA provides no benchmark for deciding when fairness or contestability has been sufficiently achieved. Article 8(8) presupposes an equilibrium without defining it. Article 18 authorizes remedies when the goals “have not been achieved” but supplies no standard of achievement. Articles 12, 19, and 53 assume that the goals can be measured without providing a yardstick. The “predefined standard” contemplated by recital 79 appears nowhere in the Act.

This measurement problem compounds the missing-baseline problem. If markets for core platform services were never fair, the DMA has no prior state against which to measure progress. Without a baseline, the Commission can always deem any improvement insufficient.

Second, the operative terms governing enforcement remain undefined. The Act does not explain what constitutes “equally effective interoperability,” “effective provision,” “effective use,” or “effective implementation.” Regulators cannot apply those terms consistently without first clarifying the goals that give them meaning.

Third, the Act does not identify which business or end users its obligations should benefit or how the Commission should manage tradeoffs among them. Neither group is monolithic. Small and large business users may have different interests, as may developers that use
“freemium” and paid models. A compliance measure may benefit one group at another’s expense, but the DMA supplies no principle for choosing among them.

Fourth, the DMA does not establish whether compliance turns on structural inputs, such as access, switching mechanisms, and data sharing, or distributional outputs, such as prices, market shares, quality, and switching rates. Under the BUI, disappointing outcomes after full structural compliance support an inference that structural defects remain. Structural compliance can therefore never guarantee protection from further intervention.

Fifth, the Act leaves consumers in an ambiguous position. Article 9(4) distinguishes the interests of third parties, including consumers, from “the objectives of this Regulation.” That distinction suggests that consumer interests do not themselves constitute an objective. Consumers instead appear to be indirect beneficiaries whose welfare the DMA expects to improve as a byproduct of market restructuring.[58]

The text therefore establishes a clear direction but no method of calibration. Section IV asks whether the enforcement record supplies what the text lacks.

IV. Testing the DMA’s Goals Against the Record

This section tests the preceding reconstruction against the DMA’s institutional and enforcement record, including the impact assessment, designation and infringement decisions, specification proceedings, first statutory review, and joint DMA–GDPR guidance. Across these sources, the Commission consistently treats gatekeeper advantages as structural problems and rival access, growth, and uptake as evidence of success, while supplying no benchmark for measuring achievement or resolving tradeoffs. Pinpoint references to institutional documents appear parenthetically in the text.

A. The Impact Assessment’s Built-In Conclusions

The Commission’s impact assessment is among the most revealing sources on the DMA’s goals.[59] It exposes the Regulation’s underlying chain of reasoning: assumptions about how digital markets work, harms inferred from those assumptions, and rules designed to remedy them.

That chain begins with a selective reading of the literature. The impact assessment cited almost no studies counseling against prescriptive platform regulation. It instead prioritized work supporting intervention, creating the appearance of consensus about both the harms of the digital economy and the urgency of addressing them.[60]

The Commission’s Regulatory Scrutiny Board (RSB) identified this weakness even after the Commission had “substantially redrafted” the assessment in response to an earlier rejection. The RSB concluded that the assessment:

. . . does not appropriately describe the shortcomings the initiative intends to address and does not provide a proper evidence base for them.

It also asked the Commission to:

. . . consider the negative consequences of curtailing the size advantages following from network economies and economies of scale for consumers [and] . . . more convincingly demonstrate for each of the selected core platform services that the identified weak contestability has negative effects in terms of higher mark-ups, lower quality of service, or reduced innovation (pp. 3–4).[61]

The Commission’s own quality-control body thus identified two central gaps before the DMA’s adoption: The assessment assumed harms rather than demonstrating them, and it discounted tradeoffs involving innovation.

The analysis also reflects what economist Harold Demsetz called the Nirvana fallacy—the mistake of comparing real institutions with idealized alternatives while ignoring the alternative’s imperfections and costs.[62] Echoing the mission letter’s promise to make markets “work better for consumers, business and society” (para. 8), the impact assessment seeks to make digital markets “better” without identifying either a baseline or an endpoint. Any market could conceivably perform better.

The assessment “fully recognise[s] the benefits that online platforms bring to the economy and society” (para. 21). It nevertheless argues that digital markets would perform even better if policymakers addressed “the most salient incidences of unfair practices and weak contestability” (para. 6). The assessment does not base its conclusion that markets function poorly on an observed decline in output, innovation, or economic performance. It instead infers poor performance from its starting assumptions. Because some ecosystems are “essentially impossible to contest . . . irrespective of how innovative and efficient [competitors] may be” (para. 27), “the likelihood increases that these markets do not function well—or may soon fail to function well” (para. 27).

Theoretical premises thus establish the imperfection of digital markets, while contrary views receive little consideration. The command to make markets work “better” supplies no stopping point. Identifying harm becomes both impossible, because no counterfactual exists against which to measure it, and unnecessary, because the theory presumes it. The DMA carries that method forward by dispensing with effects analysis and an efficiency defense.

The assessment draws an equally direct link between unfairness and size. It identifies gatekeepers’ “incomparable economic strength” (para. 88) through revenue, markups “amongst the highest in the world” (paras. 56, 129), profits that outperform the market (para. 96), and pandemic-era stock valuations (para. 95). Paragraph 88’s principal evidence for unfairness consists of app-store and social-network revenue figures.

The assessment then treats economic strength as de facto conclusive evidence of unequal bargaining power, which transforms ordinary business conduct into a potentially unfair practice (para. 41). Yet it never measures bargaining power or compares gatekeepers’ size with that of supposedly dependent business users. Some of those users rank among the world’s largest companies and include other gatekeepers. The assessment also fails to ask whether platforms depend on their business users or to identify a comparative benchmark.

A survey supplies the principal corroboration. It found that 88% of businesses reported “unfavourable trading conditions on large platforms,” which respondents attributed mostly to “a perceived imbalance of bargaining power” (para. 61). The assessment substantiates that perception only through proxies.

Its logic is straightforward. Gatekeepers’ size and economic success produce power. That power creates a bargaining imbalance, and the imbalance produces unfair outcomes. Curbing the growth, profitability, and revenue that provide gatekeepers with “economic strength” therefore becomes one of the Act’s principal objectives, if not the principal one. This reasoning supplies the foundation for the Structural Unfairness Axiom and Dialectical Fairness.

The impact assessment also portrays fairness and contestability as projects for redistributing wealth and restructuring digital markets. It measures anticipated improvements through a projected decline in the Herfindahl-Hirschman Index (HHI), a standard measure of market concentration (RSB Opinion, Annex I, Table 1.1). The stated aim is to make it easier for third parties to dislodge gatekeepers (para. 74). The assessment would achieve that aim by reducing competitive advantages it recasts as entry barriers, including data access, ecosystems, vertical integration, and zero prices (paras. 75–76).

Those measures would enable rivals to “[steal] market shares from” gatekeepers (para. 77) and reduce their “supra-normal profits” (para. 302). According to the assessment, more effective regulation produces a “more optimal . . . redistribution of profits” (para. 355).

Its treatment of scale is expressly asymmetric. The assessment acknowledges that the rules “would lead to curtailing the size of network effects and economies of scale thus reducing associated advantages for consumers” (para. 318). It does not deny that cost. Instead, it seeks to extend those advantages to “also non gatekeeper platforms benefitting from such advantages” (para. 318). Scale is desirable when challengers obtain it (paras. 29, 69, 308) and problematic when gatekeepers retain it.

The same logic governs innovation. The assessment treats the concentration of research and development (R&D) among a few firms as a “gap” for regulation to close rather than evidence of those firms’ contribution (para. 282). It accepts the possibility of reduced gatekeeper innovation on the premise that greater aggregate innovation, especially “amongst smaller businesses” (para. 104), will more than offset the loss (paras. 283–285). Yet the observable evidence available at the time showed higher R&D investment and greater innovation—the opposite of the predicted harm.

Finally, the impact assessment presumes the effects of the practices it regulates and writes those presumptions into law. It states that gatekeepers “monopolise data” (para. 38) without addressing data’s nonrivalrous character, meaning one firm’s use does not prevent another’s. It treats high markups as uniform evidence of inadequate competition across all core platform services (para. 56). It infers business-user “dependence” from revenue-share figures without explaining what would demonstrate independence (para. 58). It also assumes that gatekeepers will not pass on compliance costs (paras. 301–302) and excludes consumer-side costs from its calculations (para. 350).

The exercise was also quantitative. The assessment projected €13 billion in consumer surplus, €12 billion to €23 billion in income gains from increased R&D, and a 0.25-point decline in the HHI (Part 2, pp. 59–61). The DMA is therefore effects-based in a distinctive sense. The Commission calculated its effects once, across the entire regulatory scheme, and converted those estimates into statutory premises exempt from case-by-case verification. The benefits later invoked in the preamble and enforcement decisions were assumed before enforcement began, and subsequent decisions need not establish them again.

B. Designation Decisions: Contestability as Status

The DMA’s first enforcement acts designated gatekeepers under Article 3.[63] On their face, the seven decisions reveal little about the Act’s goals. Designation turns almost entirely on Article 3(2)’s quantitative presumptions, while fairness and contestability appear mostly in boilerplate. The decisions’ omissions and exclusions nevertheless reveal how the Commission understands those goals.

The principal omission concerns fairness. None of the seven decisions defines or applies the term. Four mention it only in titles and citations. The Apple decision mentions neither fairness nor contestability.

Fairness makes its sole substantive appearance in the Meta decision, paired with its statutory twin. The Commission states that multihoming across WhatsApp and Messenger “reinforce[s], rather than mitigate[s], the contestability and fairness issues” that led the legislature to include messaging services in the Act (Meta designation, para. 230). Fairness therefore does not function as an independent standard at the designation stage. Designation establishes whatever unfairness gatekeepers embody through classification alone.

The Commission gives contestability more content, but only by identifying what does not demonstrate a contestable market. Multihoming does not. As the ByteDance decision explains, “the fact that there is (some degree of) multi-homing does not preclude the possibility of weak contestability and unfair practices” (ByteDance, para. 136).

The presence of several large competitors also does not suffice. Contestability “can also be limited if there is more than one gatekeeper for a core platform service” (Meta designation, fn. 287, citing recital 32). The Commission also excludes conventional economic evidence. Rebuttal arguments “should relate directly to the quantitative criteria,” while “any justification on economic grounds such as those related to market definition or to efficiencies should be discarded, because it is not relevant to the designation as a gatekeeper” (Apple designation, para. 23, applying recital 23).

The ByteDance decision goes further by converting classic evidence of contestability into evidence of its absence. ByteDance argued that TikTok was “a recent entrant and challenger of several established ecosystems” and therefore could not hold an entrenched and durable position (para. 111). The Commission responded that Article 3(1)(c) “does not refer to an ‘unassailable’ position, but to an ‘entrenched and durable position’” (para. 156). It then presumed limited contestability from three years of high user numbers (para. 157) and concluded that “TikTok’s significant scale and growth . . . with an upward trajectory further supports the finding” of entrenchment (para. 158).

TikTok is the only successful large-scale entrant into a core platform market in the past decade. Its entry and growth would ordinarily provide strong evidence that the market remains contestable. The Commission treated them as evidence of the opposite based on the DMA’s formal, irrebuttable assumptions about core platform-service markets. The decision illustrates the Scale–Unfairness Paradox—the same entry and growth that gave TikTok scale also made it a gatekeeper.

Under the DMA, contestability operates as a status attached to firms rather than as a property of markets. The Commission infers that status from scale and duration, and firms cannot rebut it by pointing to entry, growth, switching, or multihoming—the very conduct that demonstrates contestability in its ordinary economic sense.

This approach reverses William J. Baumol’s conception of contestability, under which the threat of potential entry disciplines incumbents because incumbency offers no protection from competition.[64] Under the DMA, successful entry leads to incumbency, and incumbency triggers regulation.

C. Infringement Decisions: Prohibitions Without Endpoints

The Commission issued its first DMA infringement decisions against Apple’s anti-steering rules and Meta’s “consent or pay” model on April 23, 2025. These decisions provide the first applications of fairness and contestability to specific conduct. Both follow the same structure: They define the goals negatively through the elimination of practices labeled unfair or contestability-limiting, without identifying when a market becomes fair or contestable.

In the Apple decision, the Commission found that the App Store’s terms unlawfully restricted developers’ rights under Article 5(4) to communicate with users, promote offers, and conclude contracts outside the App Store.[65] Two findings stand out.

First, the Commission found that a disclosure sheet shown when users attempted to leave the App Store was “not neutral and objective” and could “deter end users from exercising their right under Article 5(4)” (para. 103). The sheet warned that Apple “is not responsible for the privacy or security of purchases made on the web.” According to the Commission, that warning implied that transactions outside Apple’s ecosystem were unsafe and therefore “unfairly reinforce[d] Apple as a gatekeeper” (para. 103).

Second, the decision prevents Apple from charging for “gatekeeper value,” defined as “the value of the core platform service provided by a gatekeeper, on top of its market value, because of its position as an important gateway” (para. 197). Apple may charge for facilitating a developer’s initial acquisition of a new user. It may not monetize the positional advantage associated with its gatekeeper status. The decision offers no criteria for determining when a fee crosses the line between legitimate compensation and payment for gatekeeper value.

The Commission adds that fairness may justify structural changes to a gatekeeper’s business and charging models when those restrictions are lawful, proportionate, and necessary to achieve “objectives of general interest recognised by the Union” (para. 212). Yet it does not specify how far those changes must go before the market becomes fair. Fairness remains an open-ended standard. A gatekeeper can eliminate conduct the Commission has labeled unfair without knowing whether it has reached the required endpoint.

The decision treats contestability in much the same way. Allowing Apple to charge fees on contracts concluded after steering would “reinforce developers’ dependence” and reduce contestability (para. 160). Contestability therefore means reducing dependence and increasing openness. The decision provides no criteria for determining when dependence has fallen far enough. Its enforcement analysis verifies that Apple removed the restrictions (para. 312). It does not ask whether developers can reach users in practice or whether alternative distribution channels have grown.

The Meta decision follows the same architecture.[66] Meta’s “consent or pay” model required users either to permit the combination of personal data across Meta services or to pay a subscription for an alternative. The Commission found that model inconsistent with Article 5(2).

The decision approaches fairness through imbalance and coercion. Designation, it states, “is a strong indication of the data controller’s power to impinge on the freedom of choice of the data subject.” Recital 4 links that power to “a serious imbalance in bargaining power and, consequently, unfair practices and conditions” (para. 183). The binary choice prevented users from withholding consent “without suffering a detriment” (para. 189), particularly given Meta’s “strong lock-in and network effects” (para. 190).

Contestability again turns on data as an input. The restriction on combining data seeks to prevent gatekeepers from exploiting “potential advantages in terms of accumulation of data” that “raise barriers to entry and hinder contestability” (para. 32). Article 5(2)’s choice mechanism therefore serves as a means of limiting gatekeeper data accumulation rather than as an independent end.

The decision does not explain what equivalence between paying with money and paying with data would eliminate coercion. Nor does it specify how much user control or data separation would make the market sufficiently open. A gatekeeper can identify what the Commission has prohibited yet cannot determine when its practices have become fair or contestable enough.

The Commission added a third data point on July 23, 2026, when it fined Google €890 million in two noncompliance decisions. It imposed a €460 million fine for favoring Google’s services in Google Search in violation of Article 6(5) and a €430 million fine for restricting steering from Google Play in violation of Article 5(4). The Commission gave Google 60 days to comply and backed that deadline with potential periodic penalty payments.[67]

The Commission had not published the full decisions at the time of writing. Based on its announcement, they appear to follow the Apple and Meta template: conduct-specific prohibitions, reduced dependence as the operative theory, and no articulated endpoint.

D. Specification Proceedings: Parity Without Effects Analysis

If the infringement decisions show what fairness and contestability prohibit, the specification proceedings show what they require. The DMA’s redistributive logic is clearest here.

In its preliminary findings on Alphabet’s search-data-sharing obligation under Article 6(11), the Commission introduced a “principle of parity.”[68] Under that principle, Alphabet may retain no advantage from data it collects unless eligible third-party search providers receive the same data. The Commission defines eligible providers to include chatbots with search functions (paras. 2–3).

Alphabet must therefore share all query, view, click, and ranking data that it collects to optimize its services. It must provide those data as frequently as it accesses them itself (para. 15). This parity principle goes beyond Article 6(11)’s text, which requires access only on “fair, reasonable and non-discriminatory” (FRAND) terms.

The broader standard follows from Section III’s reconstruction. If collecting and “hoarding” data is inherently unfair, forcing a gatekeeper to share those data on equivalent terms corrects the imbalance. A narrower interpretation could have required meaningful access without equalizing inputs. That approach might have helped rivals challenge Google Search while preserving some of Alphabet’s advantage. The Commission’s conception of fairness instead demands parity.

The preliminary findings extend that parity in three ways. First, Alphabet must anonymize the shared data, monitor third parties’ compliance with privacy restrictions, and enforce retention and encryption requirements (paras. 35–50). The gatekeeper thus bears the cost and risk of policing the rivals the obligation seeks to empower.

Second, Alphabet may charge only the incremental costs of preparing, storing, and distributing the dataset, plus a return capped at its weighted average cost of capital. It cannot recover the costs of collecting the data. Small and medium-sized access seekers must receive the data below cost even when Alphabet demonstrates that it cannot otherwise cover its costs efficiently (paras. 71–78).

Third, the post-FRAND standard requires prices that allow beneficiaries to optimize their services “without Alphabet maintaining or obtaining an unfair advantage.” That test does not explain whether a rival’s inefficiency may account for a continuing performance gap or whether any remaining Alphabet advantage is presumptively unfair. The latter interpretation would make the obligation perpetual and impossible to satisfy conclusively.

The Apple interoperability decision under Article 6(7) separates contestability even more clearly from competitive effects.[69] The Commission repeatedly states:

[It] is not required to demonstrate that each of the measures are necessary to enable contestability, i.e. that (some) third parties are able to provide a “competitive offering” or an “alternative solution” which is enough for achieving contestability.

Such an inquiry, the Commission concludes, “would re-import the effects analysis, which the legislator explicitly rejected” (paras. 77, 330).

The Commission neither measures contestability before intervening nor tests it afterward. Contestability instead operates as a legal category that follows from the Act’s assumptions. The decision then translates that category into strict parity. Article 6(7) requires access to the “same,” rather than merely “similar,” features Apple uses (para. 61). Parity extends across the user experience, including ease of use, setup, data-transmission speed, and energy consumption.

Rivals’ mere ability to compete does not suffice. They must compete using identical inputs. The reason a third party cannot offer a competitive product is irrelevant, whether the obstacle arises from connectivity limitations or the product’s inferior quality.

Interoperability is a “results-based concept” for which the gatekeeper bears sole responsibility (para. 92). “Allowing” interoperability means proactively enabling it. Apple may have to develop application programming interfaces (APIs), which allow software systems to communicate, test for bugs, and provide third parties with “adequate and timely assistance” (para. 80). Its solution may impose no “undue costs—including development costs” on beneficiaries (para. 81). Apple must provide all of this without charge.

Apple must also build interoperability into new and updated features from inception—that is, by design—and release it to third parties at launch. This obligation applies even when third parties do not know that Apple is developing the feature (paras. 71, 146, fn. 208).

Two aspects of the decision are especially revealing. First, although the Commission rejects effects analysis, it repeatedly asserts beneficial effects. Equal access “fosters contestability and fairness,” which “in turn, improve[s] the innovation potential of the wider online platform economy.” The Commission also asserts that “more contestability for complementary or supporting services or hardware will bring more contestability for operating systems” and that mandated sharing “will also indirectly increase Apple’s incentives to innovate within iOS itself” (paras. 95–98).

The decision offers no evidence for those propositions. The Act’s preamble already assumes the benefits, so the Commission sees no need to demonstrate them. Those assumed effects then exclude contrary evidence. Examining whether forced sharing might weaken Apple’s incentives to innovate “would introduce the requirement to investigate on a case-by-case basis the effects on competition of gatekeepers’ given conduct, which the legislator explicitly rejected” (para. 66).

The decision simultaneously declares that “[a] main goal of Article 6(7) . . . is to enable innovation by third parties” (para. 95). The source of innovation therefore matters more than its substance.

Second, the Commission concedes that full compliance cannot eliminate Apple’s advantage. Apple will retain an “intrinsic advantage” because “only Apple decides which iOS features are being prioritized, planned and developed” (para. 74). A fully level playing field is therefore unattainable. Apple will always manage its own operating system, and that unavoidable advantage can justify continuing intervention.

The gatekeeper ultimately becomes both a competitor and the custodian of its rivals’ competitiveness. The DMA preserves the platform’s “dual role”—owner and participant, platform operator and commercial actor—but repurposes it. Recast as a quasi-public utility, the gatekeeper must exercise control over its platform for third parties’ benefit and at its own expense.

E. The First DMA Review: Redistribution as Success

The Commission’s first statutory report offers its own assessment of what two years of DMA enforcement have achieved.[70] It also provides a revealing account of how the Commission applies fairness and contestability. Three themes stand out.

First, the report defines “competition on the merits” in a way that discounts gatekeepers’ competitive advantages. It states that Article 6(5) will “curb gatekeepers’ power from vertical integration, promoting merit-based competition and innovation” (p. 24). The implication is that vertical integration represents a source of exclusion rather than competition on the merits.

The report applies the same logic to data. Article 6(2) allows market participants “to compete on the merits” by preventing gatekeepers from using business-user data to compete against those users (p. 11). The provision targets that use precisely because the data can improve gatekeepers’ offerings and increase competitive pressure. Competition by gatekeepers becomes unmeritorious “exploitation.”

This reasoning extends to artificial intelligence (AI). The report recounts trade associations’ requests that the Commission apply Article 6(5) to Google’s AI Overviews because their integration into search may entrench Alphabet’s position and reduce traffic to third parties (p. 26). It also notes that Article 6(2) “may apply to the provision of AI services if gatekeepers rely on non-public data of business users to compete with them” (p. 12).

The report’s conception of competition and innovation is therefore source-specific. It favors competitive pressure from third parties while discounting pressure from gatekeepers. This approach echoes the Apple specification proceeding, which described a principal goal of Article 6(7) as enabling innovation by third parties.

Contestability likewise means “strengthen[ing] the ability of alternative search engines to compete” (p. 15) by improving their position relative to gatekeepers. That relative improvement counts as progress even if it worsens gatekeepers’ products or reduces the overall quality of available offerings. The report accordingly treats greater uptake of services offered by firms other than gatekeepers as evidence of success (pp. 6, 14). Greater uptake of gatekeeper services implies failure.

The report identifies products such as Firefox, Aloha, Opera, and Vivaldi as beneficiaries because they are “smaller,” without defining smallness beyond their size relative to a gatekeeper (p. 14). It even counts Alphabet’s decision to permit account registration without an email address as progress because it “further reduc[es] user dependency on Gmail” (p. 23). That change does not create competition with Gmail as an email service. It simply reduces the number of Gmail users.

Third-party access, uptake, and presence also substitute for evidence of market outcomes. The availability of third-party alarm-clock apps on iOS is presumed to increase innovation without examining their quality or consumer demand (p. 17). The report similarly treats the launch of interoperable messaging services as confirmation of demand for interoperability without supplying usage figures. At the time of writing, one of the two cited services had not launched in the European Union. The other had few downloads and an average App Store rating of 3 out of 5 stars.

The report thus resolves Section III’s input-output ambiguity in favor of inputs. It assumes that favorable outcomes will follow from access, interoperability, and third-party presence because the Act defines those inputs ex lege as fair and contestability-enhancing.

Second, the report exposes a tension among user choice, contestability, and consumer welfare. It describes Article 5(2) as “giving users back control over their data” through a “real and unbiased choice” (pp. 6, 8). It also presents the provision as a means of “limiting gatekeepers’ accumulation of personal data” (p. 7). The first objective can be satisfied regardless of what users choose. The second can be satisfied only if users choose against data combination.

The report correctly states that “Art 5(2) of the DMA does not require any target threshold of how users exercise their choice” (p. 8). That assurance sits uneasily with the report’s framing and the Act’s broader effort to reduce gatekeepers’ accumulation of a crucial competitive resource. The asymmetry remains: Data accumulation is desirable when Articles 6(9) and 6(10) facilitate it for third parties and undesirable when gatekeepers undertake it.

Third, the report casts gatekeepers as financiers of the DMA’s redistributive project. Article 6(10) requires them to provide commercially valuable data free of charge, even though Amazon charges for portability solutions outside the European Union (p. 10). Steering beyond initial customer acquisition must be free (p. 15). Gatekeepers must also fund alternative dispute-resolution mechanisms (p. 29).

The report applies the same approach to consumer profiling. Gatekeepers must disclose their profiling techniques to facilitate contestability by allowing rival platforms to distinguish themselves through stronger privacy guarantees (p. 39). The obligation requires gatekeepers to reveal information resembling business secrets without imposing a comparable duty on third parties.

The Commission goes further when compliance creates conflicts among business users. Suppliers—including hotels, airlines, and merchants—complained that Alphabet’s Article 6(5) compliance measures reduced their visibility in Google Search. The Commission responded that any solution must “consider the interests of all participants in the market, including merchants, train operators, hotels, airlines, as well as vertical search services and content creators” (pp. 25–26).

The DMA thus makes gatekeepers responsible for managing tradeoffs among the groups the Act seeks to benefit. Yet neither the Act nor the report provides a principle for resolving those conflicts.[71]

F. Joint DMA–GDPR Guidance: Choice as Data Redistribution

The final source is the draft joint guidance from the Commission and the European Data Protection Board (EDPB) on the relationship between the DMA and the General Data Protection Regulation (GDPR).[72] It offers the first joint institutional account of how the DMA’s goals interact with a neighboring and potentially competing regulatory regime.[73]

The Guidelines present the two regimes as harmonious. They describe them as “complementary in terms of goals and in terms of the protections provided.” They also posit a reinforcing cycle in which “[g]reater fairness in and contestability of digital markets lead to more choice for individuals, which in turn should increase incentives for gatekeepers and their business users to develop and implement data protection and privacy features.”

The Guidelines offer no evidence for that relationship or explain why greater contestability should improve privacy. More firms might instead have stronger incentives to monetize data to compensate for their lack of scale. The Guidelines then devote substantial attention to managing frictions that their harmony narrative obscures, including consent requirements for business-user data access under Article 6(10), anonymization under Article 6(11), and security exceptions under Article 7.

The Guidelines make the instrumental role of choice explicit. Article 5(2)’s specific-choice requirement “seeks to address the enhanced access to personal data that provides gatekeepers with potential advantages in terms of data accumulation, which in turn raises entry barriers and hinders contestability.” Offering that choice “is a means to address the accumulation of personal data by gatekeepers and the correlated erosion of market contestability.” Choice is therefore a mechanism for reducing gatekeepers’ data holdings.

The treatment of data portability reveals the same asymmetry. The Guidelines criticize data accumulation by gatekeepers while praising continuous portability under Article 6(9) for “addressing data accumulation by gatekeepers and providing a stable source of data for the improvement and creation of new services.” Data accumulation creates a structural harm when data remain with a gatekeeper and an innovation input when third parties receive them. The gatekeeper becomes, in the Guidelines’ words, a “stable source of data” for its rivals. This is Dialectical Fairness in practice.

The Guidelines also make an important concession. Under Article 5(2), gatekeepers must offer users who withhold consent a “less personalised but equivalent alternative” that is “not . . . different or of degraded quality.” The Guidelines do, however, permit degraded quality when “a degradation of quality is a direct consequence of the gatekeeper not being able to process such personal data.”

That exception recognizes that personal data can improve product quality and that restricting their use can make a service worse. This tension runs throughout the enforcement record. The DMA seeks to reduce gatekeepers’ data-driven advantages even though those same advantages may support product improvements.

The Guidelines complicate the analysis in one respect. They frame choice as protecting end users’ informational self-determination, meaning their control over personal data. That framing gives consumers’ interests independent weight that the First DMA Report rarely affords them. The redistributive flows are also conditional. Article 6(10) requires user consent before business users receive personal data, while Article 6(11) requires anonymization before search data may be shared.

Yet if end-user self-determination controlled the analysis, the direction of users’ choices would not matter. The Guidelines’ own account of Article 5(2) shows that it does. The choice mechanism aims specifically to reduce gatekeepers’ data accumulation and the competitive advantages it creates.

The Guidelines therefore reinforce the pattern developed across the enforcement record. Fairness operates as a structural presumption against gatekeepers, while contestability calls for redistributing advantages toward their rivals. The official rhetoric assumes harmony among privacy, competition, and innovation even as the detailed rules must manage tradeoffs among them. Theoretical premises, rather than demonstrated effects, continue to determine the expected outcomes.

V. Reformulating Fairness and Contestability

Reading the Act alongside its enforcement record permits a clearer account of its twin goals. Fairness emerges as gatekeeper-financed structural redistribution, while contestability directs competitive opportunity toward rivals and business users. Both rest on unrebuttable premises, assume broad benefits despite unresolved tradeoffs, and use consumer choice to advance a single redistributive program.

A. Unrebuttable Premises and Assumed Benefits

The DMA rests heavily on theory, but the law insulates that theory from contradictory evidence in two ways.

First, the Act converts its assumptions into statutory premises that enforcement cannot revisit. As the Commission repeatedly explains, contestability is a legal category rather than an empirical condition open to dispute. It follows from the definition of a gatekeeper, the list of core platform services, and the recitals. If market developments disprove those assumptions, a gatekeeper cannot rebut them during enforcement. Only legislative change can correct them.

Second, the DMA assumes that achieving fairness and contestability benefits everyone except gatekeepers. It does not account for tradeoffs among different business users or between business users and consumers. Nor does it assign weight to gatekeepers’ compliance costs, which the Act treats as effectively limitless.

B. Fairness as Structural Redistribution

Fairness under the DMA means correcting a presumed structural imbalance between gatekeepers and those who depend on their core platform services. Properly reformulated, it has four features:

It is axiomatic. The Structural Unfairness Axiom presumes unfairness from gatekeeper status and establishes it conclusively at designation. The Commission does not infer unfairness from specific conduct, and gatekeepers cannot rebut it with evidence of superior products, consumer benefits, or economic efficiency. The designation decisions expressly exclude such evidence.

It is dialectical. Fairness operates by leveling down gatekeepers while leveling up business users and rivals. It does not seek to restore a neutral baseline or prevent discrete harms.

It is distributive rather than commutative. It redistributes value and opportunity among groups rather than policing the fairness of particular exchanges. Advantages rooted in scale, integration, or data accumulation may therefore be neutralized or reassigned even when they improve products and services.

It is gatekeeper-financed. Gatekeepers bear the costs of correction, including free access, free portability, compliance systems, and reconciling data utility with privacy. The Act implicitly assumes that this redistribution leaves everyone else no worse off.

C. Contestability as Competitive Redistribution

Contestability under the DMA means transferring competitive opportunity from gatekeepers to actual or potential rivals. It also has four features:

It is status-based. The Commission infers limited contestability from a gatekeeper’s scale and duration. Contestability attaches to firms rather than markets, and evidence of entry, growth, switching, or multihoming cannot rebut it. Those same facts would demonstrate contestability in its ordinary economic sense.

It is source-dependent. Competition, innovation, and data accumulation support contestability when they come from firms other than gatekeepers and undermine it when they come from gatekeepers. Under the Scale–Unfairness Paradox, gatekeeper innovation and growth reinforce rather than rebut concerns about weak contestability.

It is produced rather than diagnosed. Legal obligations seek to create contestability rather than identify an existing market condition. This reverses Baumol’s conception, under which the credible threat of entry makes a market contestable by disciplining incumbents.

It is input-measured. The Commission measures success through third-party access, uptake, and presence. It assumes rather than demonstrates the resulting effects on price, quality, and innovation.

D. One Redistributive Program

Under this reformulation, fairness and contestability form a single program. Fairness supplies the diagnosis—a structural imbalance presumed at designation. Contestability supplies the remedy—an ongoing redistribution of competitive opportunity toward rivals and business users. User choice serves both goals and gains value when it moves activity away from gatekeepers.

This relationship explains why the DMA can treat “fair and contestable” as a hendiadys, or two terms expressing one idea. Fairness and contestability represent distinct stages of the same redistributive logic.

VI. Implications of the DMA’s Unresolved Goals

Three implications follow from the gaps that remain. The DMA supplies no benchmark or endpoint for compliance, lacks an internal mechanism for detecting and correcting regulatory error, and risks exporting these features to jurisdictions that adopt its language without fully embracing its redistributive aims.

A. No Benchmark or Endpoint

The most fundamental gap is the absence of any benchmark for determining when fairness or contestability has gone far enough. The Structural Unfairness Axiom treats core platform-service markets as structurally distinct and presumes they were never fair. The DMA therefore has no prior equilibrium to restore and no historical baseline against which to measure progress.

The enforcement record supplies no substitute. Infringement decisions define both goals negatively through the elimination of prohibited practices, without identifying an endpoint. Specification proceedings elevate parity into the governing standard while conceding that full parity is unattainable because only the gatekeeper decides which features of its platform to develop. The First DMA Report measures success through third-party presence, which can always increase, rather than through any outcome-based metric.

The Bidirectional Unfairness Inference compounds the problem. Unsatisfactory outcomes after full structural compliance support an automatic inference that structural defects remain. Compliance therefore cannot protect a gatekeeper from further intervention, and the gatekeeper can never demonstrate conclusively that it has done enough.

That indeterminacy undercuts defenders’ claim that the DMA offers greater predictability than case-by-case antitrust enforcement.[74]

B. No Mechanism for Detecting or Correcting Error

The DMA’s architecture forecloses the usual mechanisms for detecting and correcting regulatory error. It excludes effects analysis and efficiency defenses by design. The legislature quantified the Act’s projected benefits once and converted them into statutory premises that enforcement need not verify.

The enforcement record shows the result. Claims about innovation move from the impact assessment to the preamble and then into individual decisions without gaining evidentiary support. The Act excludes by law the possibility that compliance could reduce product quality or weaken innovation incentives. A regime that treats incumbent growth as failure and discounts quality improvements by gatekeepers cannot distinguish successful regulation from harmful regulation. Both appear as a reduction in gatekeeper advantage.[75]

The framework can distinguish them only if reducing gatekeeper advantage is itself the objective, regardless of cost. Yet the tradeoffs the Act refuses to consider do not disappear. It resolves them silently and in advance against gatekeepers and, potentially, against consumers who value the integrated products and data-driven quality the Act targets.

C. Implications for DMA Importers

The reconstruction also matters beyond the European Union. The United Kingdom, Japan, Thailand, Vietnam, Türkiye, and Brazil have adopted or are considering DMA-style regimes, often copying the EU law’s operative language and twin goals.[76]

These jurisdictions would import an internally coherent redistributive program rather than a neutral toolkit for creating fair and contestable digital markets. That program includes a conclusive presumption of structural unfairness, a remedial strategy that levels down incumbents and levels up rivals, an instrumental conception of user choice, and a source-dependent, input-measured account of contestability with no defined endpoint.

Legislatures that want this program should adopt it knowingly. Those seeking consumer-oriented contestability or greater innovation regardless of its source should not assume the DMA will deliver those outcomes. The Act was not designed to do so.

VII. Conclusion: A Direction Without a Destination

This white paper argues that the DMA contains a coherent but unstated conception of fairness and contestability. Fairness seeks to correct a presumed structural imbalance. It is axiomatic, dialectical, distributive, and gatekeeper-financed. Contestability transfers competitive opportunity toward rivals. It is status-based, source-dependent, legally produced, and measured through inputs rather than market outcomes. User choice serves both goals by directing activity away from gatekeepers.

The enforcement record confirms this redistributive logic. The impact assessment embedded its predicted harms and benefits in the Act’s premises. Designation decisions treat contestability as a status inferred from scale. Infringement decisions identify prohibited conduct without defining an endpoint. Specification proceedings elevate parity while excluding effects analysis. The First DMA Report and joint DMA–GDPR guidance treat rival access, uptake, and data accumulation as progress, largely without testing effects on price, quality, innovation, or consumer welfare.

Two years of enforcement have therefore clarified the DMA’s direction but not its calibration. The Act identifies who should gain and who should lose without saying how much, for how long, or against what benchmark. It assumes away tradeoffs among business users, gives consumer welfare no controlling weight, and offers no internal mechanism for distinguishing beneficial intervention from harmful intervention. Full compliance cannot guarantee finality because disappointing outcomes can always support an inference that structural defects remain.

These omissions matter within the European Union and wherever lawmakers borrow the DMA’s language. Jurisdictions seeking consumer-oriented contestability or greater innovation from any source should recognize that the Act embodies a more specific redistributive program.

We now understand more clearly what fairness and contestability mean under the DMA. The Act tells markets which direction to move, but it still cannot say when they have arrived.

VIII. Glossary of Terms

This white paper uses several original terms to describe recurring patterns in the DMA’s text and enforcement. These terms are analytical tools rather than statutory categories. The following glossary defines each and explains its role in reconstructing the Act’s goals.

A. BUI (Bidirectional Unfairness Inference)

The Bidirectional Unfairness Inference (BUI) describes reasoning used without objective or agreed benchmarks for market performance. Under the BUI, outcomes that depart from a regulator’s preferred state support an inference that the market is structurally unfair. Conversely, asymmetries in rights, obligations, or constraints imply that the resulting prices, economic rents, market structure, or distribution must also be unfair. Market structure and outcomes thus serve as mutual evidence of unfairness.

The BUI is autopoietic, meaning self-generating, and it is self-referential. Neither unfairness nor the relevant counterfactual—the outcome expected under alternative conditions—has an independent definition. Structural asymmetries and deviations from regulatory expectations therefore validate each other. Structure explains the outcomes, while the outcomes confirm the diagnosis of an unfair structure.

Consider Apple’s 30% commission on in-app purchases. The fee is presumed excessive because Apple controls the in-app payment system and restricts steering, practices characterized as unfair. Its perceived excessiveness then serves as evidence that the market itself is structurally unfair. The diagnosis rests on reciprocal inferences from market design and observed outcomes rather than an external benchmark.

B. Dialectical Fairness

Dialectical Fairness describes fairness as a relational and directional process that simultaneously reduces gatekeepers’ advantages and improves the position of rivals or business users. Gatekeepers are made worse off so that others may be made better off. Fairness lies in the relative shift between those positions rather than in reaching a fixed equilibrium.

Constraining gatekeepers and empowering rivals are therefore linked. Any persistent gatekeeper advantage remains a potential target for correction.

Dialectical Fairness is dynamic and nonterminal, with no built-in endpoint. It supports continuing, asymmetric intervention to limit the renewal or expansion of gatekeeper advantages and rebalance competitive positions, rather than restoring a neutral baseline or remedying discrete harms.

C. DSUA (DMA Structural Unfairness Axiom)

The DMA Structural Unfairness Axiom (DSUA) is the Act’s foundational assumption that digital markets suffer from endemic unfairness and weak contestability. The premise is axiomatic because evidence cannot rebut it during enforcement. Changing it would require legislative amendment or repeal. The DMA’s obligations and interventions flow from—and find justification in—this axiom.

D. Source-Dependent Competition and Innovation

Source-Dependent Competition and Innovation describes a view of competitive activity whose value depends on its source. Conventional competition law protects the competitive process without favoring which firm succeeds. The DMA instead credits competition and innovation from rivals and entrants while discounting the same activity from gatekeepers, regardless of their respective effects on consumers.

Gatekeeper competition and innovation may therefore count against fairness and contestability. Under the Scale–Unfairness Paradox, no gatekeeper can innovate or grow its way out of the presumption of unfairness.

E. SUP (Scale–Unfairness Paradox)

The Scale–Unfairness Paradox (SUP) arises because greater scale can simultaneously improve a core platform service’s quality and attractiveness while increasing the risk of structural unfairness. The DMA treats the same features that make gatekeepers successful—network effects, data, and user growth—as sources of bottleneck power and weak contestability.

The SUP produces a second paradox. By constraining gatekeeper scale while helping rivals grow, the DMA may eventually turn successful rivals into gatekeepers subject to the same restrictions.

Paragraph 84 of the Impact Assessment illustrates this logic:

The bigger the platform, the stronger the indirect network effects, the larger the amount of data and the higher its quality. This leads to increased insight into user profiles and preferences, allowing gatekeepers to offer them more personalized services and advertisements, thus attracting even more users and reinforcing consumer lock-in, single-homing and rendering switching to alternative platforms more difficult.

[1] Regulation (EU) 2022/1925 of the European Parliament and of the Council of 14 September 2022 on Contestable and Fair Markets in the Digital Sector (Digital Markets Act), 2022 O.J. (L 265) 1 [hereinafter DMA].

[2] See, e.g., Giuseppe Colangelo, In Fairness We (Should Not) Trust: The Duplicity of the EU Competition Policy Mantra in Digital Markets, 68 Antitrust Bull. 618 (2023).

[3] See Nicolas Petit, The Proposed Digital Markets Act (DMA): A Legal and Policy Review, 12 J. Eur. Competition L. & Prac. 529 (2021).

[4] On the Sherman Act as a “common-law statute” whose content courts developed through adjudication, see Leegin Creative Leather Products, Inc. v. PSKS, Inc., 551 U.S. 877, 899 (2007). On the incremental development of Articles 101 and 102 of the Treaty on the Functioning of the European Union, see Pablo Ibáñez Colomo, The Shaping of EU Competition Law (Cambridge Univ. Press 2018).

[5] See, e.g., Belle Beems, The DMA in the Broader Regulatory Landscape of the EU: An Institutional Perspective, 19 Eur. Competition J. 1, 27 (2023); Geoffrey A. Manne, Lazar Radic & Dirk Auer, Regulate for What? A Closer Look at the Rationale and Goals of Digital Competition Regulations, 22 Berkeley Bus. L.J. 201 (2025).

[6] See Digital Markets, Competition and Consumers Act 2024, c. 13 (UK); Act on Promotion of Competition for Specified Smartphone Software, Act No. 58 of 2024 (Japan); Projeto de Lei No. 4675/2025, Câmara dos Deputados (Braz.).

[7] See Manne, Radic & Auer, supra note 5 (arguing that the stated goals of DMA-style regulations diverge from their actual objectives).

[8] Giorgio Monti, The Digital Markets Act: Improving Its Institutional Design, 5 Eur. Competition & Regul. L. Rev. 90, 91–92 (2021).

[9] Richard Feasey & Giorgio Monti, DMA Process and Compliance, in Implementing the DMA: Substantive and Procedural Principles 94, 120 (Alexandre de Streel coord., Ctr. on Regul. in Eur. 2024) [hereinafter CERRE].

[10] Jacques Crémer et al., Fairness and Contestability in the Digital Markets Act, 40 Yale J. on Regul. 973, 975–77 (2023).

[11] Id. at 975–76; see also DMA, supra note 1, arts. 12, 18–19, recital 65.

[12] Friso Bostoen, Understanding the Digital Markets Act, 68 Antitrust Bull. 263 (2023).

[13] Pinar Akman, Regulating Competition in Digital Platform Markets: A Critical Assessment of the Framework and Approach of the EU Digital Markets Act, 47 Eur. L. Rev. 85 (2022).

[14] Petit, supra note 3.

[15] Colangelo, supra note 2, at 620.

[16] Monti, supra note 8, at 91–92.

[17] Petit, supra note 3.

[18] Akman, supra note 13.

[19] See Bostoen, supra note 12 (noting that the DMA’s recitals connect its anticipated improvements to innovation and consumer benefits, reinforcing the regime’s proximity to competition-policy benchmarks despite its ex ante structure).

[20] Monti, supra note 8.

[21] Rupprecht Podszun, Philipp Bongartz & Sarah Langenstein, Proposals on How to Improve the Digital Markets Act (working paper, Feb. 18, 2021), https://ssrn.com/abstract=3788571.

[22] Manne, Radic & Auer, supra note 5.

[23] Pierre Larouche & Alexandre de Streel, The European Digital Markets Act: A Revolution Grounded on Traditions, 12 J. Eur. Competition L. & Prac. 542 (2021).

[24] Manne, Radic & Auer, supra note 5, at 217 (describing a “Schrödinger’s [digital competition regulation]” scenario in which “the same substantive rules simultaneously are and are not competition law”).

[25] Crémer et al., supra note 10, at 1006–08.

[26] CERRE, supra note 9, at 14.

[27] Petit, supra note 3.

[28] Akman, supra note 13.

[29] Id.

[30] Manne, Radic & Auer, supra note 5.

[31] Giuseppe Colangelo & Alba Ribera Martínez, The Metrics of the DMA’s Success, 16 Eur. J. Risk Regul. 1017 (2025).

[32] Oles Andriychuk, Do DMA Obligations for Gatekeepers Create Entitlements for Business Users?, 11 J. Antitrust Enf’t 123, 123–32 (2023) (characterizing the DMA as “punitive”).

[33] DMA, supra note 1, recital 79 & arts. 8(1), 12, 18–19 & 53.

[34] DMA, supra note 1, recital 33.

[35] Id. recital 62.

[36] See id. recitals 2, 13, 20 & 32–33.

[37] The Act likewise does not distinguish companies whose dependence on a CPS resulted from a lack of foresight or diversification from those whose dependence was not self-inflicted.

[38] See DMA, supra note 1, recitals 2–5, 13 & 32–33 (attributing imbalances to features such as extreme economies of scale, strong network effects, data-driven advantages, lock-in, and a lack of multihoming).

[39] Id. art. 3(2).

[40] See id. recitals 15 & 17 (identifying monetization potential and access to capital markets as indicia of gatekeeper status); id. recital 25 (identifying high relative growth as an indicator of entrenchment).

[41] Id. recital 36.

[42] Id. recital 72 (emphasis added).

[43] Id. recital 46 (asserting further that it is unfair for a gatekeeper to benefit from its “dual role” as both provider and user of a core platform service).

[44] Id. arts. 6(5), (7) & (10).

[45] Id. recital 70; see also id. recital 37.

[46] Id. art. 6(3) & recitals 49 & 63.

[47] Id. art. 5(3) & recital 39.

[48] See, e.g., Christophe Carugati, The Digital Markets Act Is About Enabling Rights, Not Obliging Changes in Market Conditions, Bruegel (Sept. 6, 2023), https://www.bruegel.org/analysis/digital-markets-act-about-enabling-rights-not-obliging-changes-market-conditions.

[49] DMA, supra note 1, recitals 5, 45 & 107; see also id. recitals 6–7 (linking unfair practices to “negative societal and economic implications” and the “proper functioning of the internal market”).

[50] Id. recitals 11 & 23; see also id. recital 5 (stating that the DMA does not address the forms of market power covered by Articles 101 and 102 TFEU).

[51] Id. recital 3.

[52] Id. recital 44.

[53] Id. recital 61 & art. 6(11).

[54] Id. art. 18(3).

[55] Id. recital 75.

[56] Id. recitals 53, 55 & 57; see also id. art. 6(5) (permitting rankings based on transparent, fair, and nondiscriminatory conditions).

[57] Id. recital 64.

[58] See id. recital 79 & arts. 8(8), 9(4), 12(2)(b), 18–19 & 53.

[59] Eur. Comm’n, Commission Staff Working Document: Impact Assessment Report Accompanying the Proposal for a Regulation of the European Parliament and of the Council on Contestable and Fair Markets in the Digital Sector (Digital Markets Act), SWD(2020) 363 final (Dec. 15, 2020). Parenthetical paragraph references in this subsection refer to this document unless otherwise indicated.

[60] See Manne, Radic & Auer, supra note 5; Lazar Radic & Dirk Auer, A Europe Fit for the Age of Startups: Rhetoric and Reality in the EU’s Digital Package (Int’l Ctr. for L. & Econ. 2025), https://laweconcenter.org/wp-content/uploads/2025/07/Is-a-Europe-fit-For-the-Digital-Age-also-Fit-for-the-Start-up-Age.pdf.

[61] Eur. Comm’n Regul. Scrutiny Bd., Opinion on the Proposal for a Regulation of the European Parliament and of the Council on Contestable and Fair Markets in the Digital Sector (Digital Markets Act), SEC(2020) 437 (Dec. 10, 2020).

[62] Harold Demsetz, Information and Efficiency: Another Viewpoint, 12 J.L. & Econ. 1 (1969).

[63] Eur. Comm’n Decisions of Sept. 5, 2023, designating Alphabet, Amazon, Apple, ByteDance, Meta, and Microsoft as gatekeepers under Article 3 of Regulation (EU) 2022/1925; Eur. Comm’n Decision of May 13, 2024, designating Booking.com as a gatekeeper under Article 3 of Regulation (EU) 2022/1925. Parenthetical references in this subsection refer to the respective designation decisions.

[64] See William J. Baumol, Contestable Markets: An Uprising in the Theory of Industry Structure, 72 Am. Econ. Rev. 1 (1982); William J. Baumol, John C. Panzar & Robert D. Willig, Contestable Markets and the Theory of Industry Structure (Harcourt Brace Jovanovich 1982).

[65] Eur. Comm’n Decision of Apr. 23, 2025, Case DMA.100109—Apple (noncompliance with Article 5(4) of Regulation (EU) 2022/1925) [hereinafter Apple Anti-Steering Decision].

[66] Eur. Comm’n Decision of Apr. 23, 2025, Case DMA.100055—Meta (noncompliance with Article 5(2) of Regulation (EU) 2022/1925); see also Regulation (EU) 2016/679 of the European Parliament and of the Council of Apr. 27, 2016 (General Data Protection Regulation), 2016 O.J. (L 119) 1 [hereinafter GDPR] (providing the consent standard incorporated into Article 5(2) of the DMA).

[67] Press Release, Eur. Comm’n, Commission Fines Google €890 Million for Breaches of the Digital Markets Act (July 23, 2026), https://digital-markets-act.ec.europa.eu/commission-fines-google-eur890-million-breaches-digital-markets-act-2026-07-23_en.

[68] Eur. Comm’n Preliminary Findings of Apr. 16, 2026, Case DMA.100209—Alphabet (specification proceedings under Article 6(11) of Regulation (EU) 2022/1925), https://ec.europa.eu/competition/digital_markets_act/cases/202616/DMA_100209_1986.pdf.

[69] Eur. Comm’n Decision of Mar. 19, 2025, Specifying the Measures to Be Adopted by Apple Pursuant to Article 6(7) of Regulation (EU) 2022/1925 (interoperability with connected physical devices).

[70] Eur. Comm’n, Report on the Review of Regulation (EU) 2022/1925 (Digital Markets Act), COM(2026) 178 final (Apr. 28, 2026). Parenthetical page references in this subsection refer to this document.

[71] The report does not address whether a compliance solution must leave every affected business user no worse off—a Pareto improvement—or may trade some users’ interests against others’ and, if so, under what principle. See also id. at 48 (acknowledging “the concerns raised by the media industry, including in relation to traffic losses”).

[72] Eur. Comm’n & Eur. Data Prot. Bd., Joint Guidelines on the Interplay Between the Digital Markets Act and the General Data Protection Regulation (version for public consultation, Oct. 9, 2025), https://www.edpb.europa.eu/public-consultations/joint-guidelines-on-the-interplay-between-the-digital-markets-act-and-the_en. Quotations in this subsection are from this document.

[73] See GDPR, supra note 66.

[74] See DMA, supra note 1, recital 79 (invoking predictability as the rationale for a “predefined standard”).

[75] On error-cost analysis in this setting, see Manne, Radic & Auer, supra note 5; Colangelo & Ribera Martínez, supra note 31 (arguing that identifying the DMA’s goals is a prerequisite to measuring its success).

[76] See supra note 6. On the incorporation of DMA-style provisions directly into national competition laws, see Manne, Radic & Auer, supra note 5.

Introduction

The International Center for Law & Economics (ICLE) submits these comments on the Korea Fair Trade Commission’s (KFTC) proposed amendment to the Notification Guidelines for Business Combinations (Proposed Amendment), published for public comment on Sept. 9.[1] ICLE is a nonprofit, nonpartisan global research and policy center that applies law & economics methods to public-policy debates. Its scholars have written extensively on acquihires—transactions centered on hiring a firm’s team—and artificial intelligence (AI) partnerships.[2] ICLE has also commented on the KFTC’s merger review guidelines.[3]

The Proposed Amendment would treat transactions in which a firm hires a target’s key personnel, often alongside a license to its technology, as business transfers under Article 9(1)(iv) of the Monopoly Regulation and Fair Trade Act (MRFTA).[4] The KFTC describes the amendment as a clarification. In its view, acquihires already constitute business transfers under Article 9(1)(iv), and the amendment would make that clear so parties do not fail to notify. We welcome greater certainty for parties, but the proposed text goes beyond restating existing law.

The Proposed Amendment would expand the definition of a “business” to include organized personnel whose technology or knowledge enables them to perform a core business function. It would treat a “material part” of a business as transferred whenever the acquirer becomes able to carry on the same activities as the transferor. It would count any consideration paid to the transferor toward the transfer price, whatever its label, including payments to release rights over departing personnel and license fees. It would also treat the transferor’s cessation of business as implementation of the transfer.[5]

We share the principle behind the Proposed Amendment. Competition law should look to a transaction’s substance, not its form. A transaction that strips a firm of its competitive significance should not escape scrutiny merely because it takes the form of a hiring arrangement and no shares change hands. As drafted, however, the amendment would bring ordinary licensing and hiring within merger control. We recommend that the KFTC withdraw it or narrow it substantially. Our comments make three points.

First, the case for extending notification has not been made (Section I). No competition authority has found an acquihire anticompetitive. The United Kingdom’s Competition and Markets Authority (CMA) cleared the Microsoft/Inflection transaction in Phase 1. Brazil’s Administrative Council for Economic Defense (CADE) also cleared it through an expedited review, finding that “[t]he transaction is not likely to result in significant changes to the competitive landscape in Brazil and therefore does not require a more in-depth analysis.”[6] Structuring a transaction to stay below notification thresholds is not misconduct. It is consistent with shareholders’ interest in minimizing transaction costs. Nor does the concern that an acquihire might neutralize a nascent competitor warrant amending or clarifying competition law. Error-cost considerations counsel restraint: Wrongly deterring these transactions is costly and difficult to reverse, while errors from wrongly permitting them tend to correct themselves.

Second, most acquihires are not mergers in the competition-law sense (Section II). The Notification Guidelines define a business as a bundle of property rights. Employment contracts are not property rights in employees, who have a constitutional right to choose where they work.[7] An acquirer instead obtains a coordinated but revocable set of employment decisions, often alongside a nonexclusive license that leaves rivals free to use the same technology. Against this backdrop, the proposed “material part” test asks what the acquirer can do rather than what the transferor has lost. The proposed implementation rule also ties the parties’ obligations to decisions they do not control.

Third, the proposed transfer-price calculation would not track competitive significance and would create perverse incentives (Section III). It would count license fees paid for access without control. And because it would count only payments to the transferor—payments that return capital to a startup’s investors—it would burden compensated, cooperative transactions while leaving uncompensated raids on a startup’s workforce untouched.

I. The Case for Extending Notification Has Not Been Made

The KFTC explains that the Proposed Amendment seeks to prevent parties from evading notification obligations, noting that large global technology firms have used acquihires to recruit key personnel from AI startups.[8] This concern echoes statements by U.S. enforcers and legislators.[9]

A transaction’s exclusion from a notification regime says little about its competitive effects. Every threshold invites parties to structure transactions around it. Notification regimes deliberately leave many transactions unreviewed to focus scarce resources on those most likely to alter competitive conditions and give parties a predictable basis for planning.[10]

Acquihires also arise partly because conventional acquisitions face heightened scrutiny. When firms respond with licensing and hiring arrangements that leave the target independent, extending merger control to those arrangements tells them they are “damned if they buy and damned if they don’t.”[11]  The relevant question is whether the transactions the Proposed Amendment would capture are likely to restrain competition. The published materials offer no evidence that they are.

Before extending notification obligations to a new class of transactions, the KFTC should establish that the class poses a significant risk to competition, that existing law cannot adequately address that risk, and that the benefits of review outweigh the costs. The Proposed Amendment meets none of these conditions.

Published reviews and evidence from evolving AI markets favor restraint rather than a presumption of harm (Subsection A). Existing law reaches harmful hiring practices and allows authorities to weigh the evidence, reasonable justifications, and efficiencies (Subsection B). Acquihires also support team mobility, workers’ bargaining power, and startup investment. Wrongly deterring them would impose broad, often unobservable costs that could exceed those of wrongly permitting them, across sectors well beyond AI (Subsection C).

A. Published Reviews and Market Evidence Favor Restraint

The KFTC states that it developed the Proposed Amendment through exchanges with authorities in the European Union, Germany, and the United Kingdom.[12] But if anything, those authorities’ experience points toward restraint. The CMA and CADE have both reviewed Microsoft’s arrangements with Inflection AI and cleared the transaction. These published reviews provide a more useful starting point than a presumption that acquihires harm competition.

Microsoft reportedly paid about $620 million for a nonexclusive license to Inflection’s models and $30 million for a waiver of claims arising from its hiring of Inflection staff, most of whom joined Microsoft.[13] The CMA asserted jurisdiction but cleared the transaction at Phase 1. It found that Inflection had a very small share of chatbot use, had struggled to grow or sustain its user base, and exerted no material competitive constraint on rival services.[14]

CADE likewise cleared the transaction through an expedited review, concluding that it was “not likely to result in significant changes to the competitive landscape in Brazil and therefore does not require a more in-depth analysis.”^[15] CADE also observed:

Microsoft would not exceed a 30% market share in any of the segments related to the development and application of artificial intelligence tools, which would mitigate major vertical competition concerns.

It is worth noting that the artificial intelligence sector is rapidly evolving, with the boundaries of relevant markets not yet clearly defined; it is reasonable to assume that there remains scope for market contestability and significant gains before a state of stabilization is reached. Furthermore, there are no strong indications of a decline in rivalry that would warrant intervention.[16]

Germany’s Bundeskartellamt concluded that the transaction could, in principle, constitute a concentration but declined jurisdiction because Inflection had no substantial operations in Germany.[17] The European Commission took no decision after the Court of Justice curtailed referrals of transactions below notification thresholds in Illumina.[18] U.S. agencies have opened investigations but, to our knowledge, have brought no enforcement action against an acquihire.

The markets in which these transactions occur also show little sign of the entrenchment that critics predict. The cost of AI inference—running a trained model to produce an output—has fallen sharply. Epoch AI reports that “the price to achieve GPT-4’s performance on a set of PhD-level science questions fell by 40x per year. The rate of decline varies dramatically depending on the performance milestone, ranging from 9x to 900x per year.”[19]

New firms continue to enter at every layer of the AI supply chain, from computing resources to consumer services. Startups, rather than incumbents with the largest data holdings, produced the most successful early generative-AI services.[20] The empirical record on “killer acquisitions” also offers little basis for presuming harm. The study that popularized the term and prompted several investigations estimated that 5.3% to 7.4% of pharmaceutical acquisitions preempted future competition.[21] Evidence of comparable effects in digital markets remains limited.

One of that study’s authors, Florian Ederer, has since examined roughly 1,200 startup acquisitions by eight leading technology firms and found little evidence that they suppress innovation. Patenting in the relevant technology areas rose before acquisition and continued afterward where further acquisitions followed. Acquired patents received more citations, including from firms that hired the targets’ employees, and target inventors who stayed with the acquirer went on to patent more.[22]

We encourage the KFTC to take a wait-and-see approach while the evidence develops and generative AI remains in a formative stage. A notification obligation covering an entire class of transactions should rest on evidence that the class poses a significant risk to competition.

B. Existing Law Reaches Harmful Hiring Practices

The harms critics attribute to acquihires—hollowing out a rival or hoarding talent to deny it to competitors—are forms of conduct the MRFTA already reaches. Article 45(1)(viii) prohibits unfair interference with another firm’s business activities. The Enforcement Decree identifies unjustly inducing or hiring another firm’s personnel as one form of interference when it significantly impedes that firm’s business.[23] The KFTC’s guidelines give examples that closely match the concerns raised about acquihires: recruiting a large share of a firm’s core personnel with excessive benefits, or recruiting key personnel the hiring firm does not need to obstruct a competitor. The guidelines also allow the KFTC to consider reasonable justifications and efficiency gains.[24]

The talent-hoarding theory also rests on thin evidence. The most prominent study of this theory infers that incumbents keep researchers idle using data that shows only that certain software firms retained more skilled employees after their patent protection weakened. Yet ordinary management decisions could explain the same pattern.[25] Where hoarding does occur and impedes a rival, the KFTC can assess it under the existing rules on the evidence.

Other provisions of Korean competition law address related conduct. Article 5 reaches abuse of a dominant position, including unfair interference with rivals’ business activities. Article 40 reaches agreements among competitors, including restraints on hiring. Korea’s trade-secret law addresses the misappropriation of confidential information by departing employees.[26] These laws allow authorities to assess conduct on an actual record without burdening the many transactions that raise no concern.

C.  Acquihires Support Labor Mobility and Startup Investment

Acquihires serve important economic functions. They give founders and investors an opportunity for exit when a startup cannot raise further capital, preserve teams whose joint value exceeds the sum of individual hires, and reward entrepreneurial risk-taking. Sought-after engineers often prefer to move through an acquihire rather than leave individually, which could damage their standing with investors who might finance future ventures.[27]

Competition for AI talent has also boosted workers’ compensation.[28] Treating coordinated team moves as reportable business transfers would weaken those workers’ bargaining position. A team that can move together has a better alternative to its current employer—and therefore more bargaining power—than individuals who must move one at a time. Burdening employers who bid for teams would reduce competition for their services, contrary to competition law’s aims in labor markets.

For many target firms, failure is the realistic alternative to an acquihire. Practitioners and scholars describe acquihires as a soft landing for startups that cannot secure further funding, although the recent AI transactions that prompted regulatory scrutiny involved considerably better-capitalized targets.[29]

Making these exits costlier or less certain lowers the expected return on venture investment and, with it, the funding available to the next generation of startups.[30] Japan offers a useful contrast. Acquisitions now account for 88% of Japanese startup exits, and the Japanese government has made raising that share an explicit policy objective.[31]

The costs of enforcement errors are asymmetric. If the KFTC wrongly permits an acquihire, the error tends to correct itself because employees remain free to leave, join rivals, or found new firms. As Section II.B explains, acquired employees leave at higher rates than ordinary hires. If the KFTC wrongly deters acquihires, the costs fall on labor mobility and venture investment across the economy. Those costs are largely unobservable because the deterred transactions never occur.[32]

These costs would extend beyond AI and global technology firms. The Proposed Amendment applies to any acquirer that meets the MRFTA’s thresholds, and Chairman Joo Byung-ki has identified semiconductors and pharmaceuticals among the sectors of concern.[33] Their importance raises the costs of both mistaken intervention and inaction. Well-intentioned rules that chill procompetitive deals can do lasting harm to the competition they aim to protect.

II. Acquihires Do Not Fit the Business-Transfer Framework

The Proposed Amendment would bring acquihires within merger control by redefining a “business” under Article 9(1)(iv) of the MRFTA rather than amending the statute. This approach stretches provisions designed for transfers of assets, contracts, and an organized business to cover voluntary employee moves, often accompanied by a nonexclusive license.

The amendment departs from the existing definition of a business transfer and raises questions about the KFTC’s authority to expand statutory obligations through an administrative notice (Subsection A). Employees retain control of their human capital and remain free to leave, limiting the acquirer’s ability to retain a team (Subsection B). The proposed “material part” test overlooks what the transferor retains and whether rivals can still access its technology (Subsection C). Finally, the implementation rule ties the acquirer’s obligations to decisions it may not control, while conventional merger remedies cannot reliably restore a team to its former employer (Subsection D).

These problems call for a narrower approach tied to transferred property rights, the target’s loss of competitive significance, and acts within the parties’ control.

A. Personnel Moves Are Not Transfers of Property Rights

Article 9(1)(iv) of the MRFTA reaches the acquisition or lease of all or a material part of another company’s business, and the acquisition of all or a material part of its fixed operating assets.[34] The Notification Guidelines define a business as “a bundle of property rights organized for the company’s business purposes and functioning as an integrated whole,” including sales rights, intellectual-property rights, and permits of economic value.[35] A part of a business is “material” if it can operate as an independent business unit or its transfer significantly reduces the transferor’s sales.[36]

Both definitions describe something that changes hands. A business transfer moves assets, contracts, and the organization that uses them from one owner to another. The associated employment relationships generally pass to the transferee, subject to each employee’s right to object. In an acquihire, by contrast, employees resign and accept new offers, while the transferor retains its property.

The existing definition mentions personnel only as a component of sales rights, which include “sales-related organization, personnel, and agency relationships.”[37] Personnel therefore count when they form part of an organized business that changes hands together with its contracts and assets. The Proposed Amendment would make the movement of personnel and their know-how sufficient, even when the transferor transfers nothing else and continues to operate.

Adding organized personnel to a category defined by bundles of property rights makes the definition internally incoherent. Employees remain free to change employers after a deal. Their movement does not itself transfer property rights.

The amendment also raises a question of legal authority. The KFTC presents it as a clarification, but it would substantially expand statutory notification obligations. An administrative notice should not bring transactions within the statute’s scope when its existing provisions do not cover them. The National Assembly should decide whether to expand Article 9(1)(iv).

B. Employees Retain Control of Their Human Capital

Article 15 of the Korean Constitution guarantees every citizen’s freedom to choose an occupation.[38] Other legal systems recognize similar principles. Section 6 of the U.S. Clayton Act provides that “the labor of a human being is not a commodity or article of commerce.” The House of Lords held in Nokes v. Doncaster Amalgamated Collieries that, absent a contrary statutory provision, an employment contract cannot transfer without the employee’s consent.[39] An employer offers terms that individuals remain free to accept or reject.

An acquirer therefore gains the prospect that a team will move together, rather than ownership of its human capital. Coordinated offers, bonuses tied to the team’s joining, and agreements not to sue align the incentives of individuals who could otherwise decide separately. Each employee remains free to decline, leave later, or join a rival.[40] Economic theory supports the same distinction. Under the property-rights theory of the firm, a firm owns the complementary assets around which employees organize their work, but cannot own their human capital.[41]

The evidence confirms that employee retention is uncertain. A study of roughly 230,000 acquired startup workers finds that “acquired workers show significantly higher turnover rates than regular hires,” with attrition concentrated among the highest earners.[42] Other research finds the same pattern among senior and technical staff, and finds that acquisitions increase acquired employees’ later entrepreneurship.[43] Acquihire theories of harm depend on the acquirer retaining a team. That assumption weakens fastest for the people whose departure from the target supposedly causes the harm.

Some commentators identify the relevant asset as the know-how the team carries.[44] This argument conflates two kinds of knowledge. Codified know-how—source code, model weights, and documentation—moves through licenses or assignments, to which existing legal analysis already applies. Tacit know-how—knowledge embedded in employees’ skills and experience—is inseparable from the people who hold it. Those employees remain free to leave. Where departing employees take information they have no right to use, trade-secret law supplies the remedy rather than merger control.

C. ‘Material Part’ Test Should Assess What Is Transferred

The Proposed Amendment adds a third way to satisfy the first component of the “material part” test. It would suffice that the acquirer becomes able, after the transaction, to carry on the same business activities as the transferor.[45] This test focuses on the acquirer’s capabilities rather than the transferor’s loss, treating hiring arrangements coupled with exclusive and nonexclusive licenses alike. An incumbent already active in the field will often satisfy it by definition, as will almost any hire that adds capability. As an alternative to the existing criteria, it also bypasses the requirement that the transferred part be capable of operating as an independent business unit—the requirement that gives “material part” its meaning.

The other jurisdictional requirements do little to narrow the test. The price threshold requires consideration equal to only 10% of the transferor’s total assets or 10 billion South Korean won (KRW).[46] The thresholds for the parties’ size reach any acquirer group with KRW 300 billion in assets or turnover and any transferor with KRW 30 billion.[47] Because total assets count toward these thresholds, a well-capitalized startup holding venture funding may qualify despite having little revenue.

Consider a hypothetical transaction. A Korean electronics group hires 15 engineers from a chip-design startup holding KRW 40 billion in venture funding and pays it KRW 12 billion for a nonexclusive license to its design tools. The startup retains its remaining staff, continues to serve other customers, and remains free to license the same tools to the group’s rivals. Under the Proposed Amendment, the transaction could qualify as a business transfer. The engineers form an organized team whose knowledge enables them to perform a core function, the group can now carry on the startup’s design activities, and the license fee exceeds KRW 10 billion. Yet nothing in this sequence suggests that the transaction has transferred or eliminated the startup’s competitive significance.

A more accurate test would focus on what the transaction removes from the market, including whether the acquirer’s rivals retain commercially meaningful access to the target’s technology.[48] When the target remains free to license its technology to others and continue operating, the acquirer gains a workforce it cannot own and a license its rivals can also obtain. By contrast, a formally nonexclusive license that operates as exclusive in practice, or prevents the target from using or licensing its technology, may transfer the target’s competitive position and properly qualify as a business transfer.

The transactions that prompted the Proposed Amendment—including those involving Inflection, Character.AI, Windsurf, and Groq—rested on nonexclusive licenses under which the target retained the right to use and license its technology.[49] The guidelines should distinguish these arrangements from transfers that remove a business’s competitive significance.

The KFTC has publicly indicated that ordinary recruitment and the hiring of a handful of employees will not count as business transfers.[50] The Proposed Amendment contains no such exclusion. The final guidelines should state it expressly.

D. Implementation and Remedies Must Be Workable

The new implementation rule illustrates the mismatch. For organized personnel, the Proposed Amendment treats a business transfer as implemented when the transferor ceases its existing business.[51] Implementation determines when parties must submit a post-closing notification and what parties subject to pre-closing notification may do before clearance.[52]

The transferor’s board and investors decide when to cease operations, and the Proposed Amendment does not define cessation. A full wind-down would presumably qualify. A product pivot, a reduction in operations, or the sale of the remaining business to a third party is less clear. An acquirer could face liability for implementing a transaction before clearance because of a decision it neither made nor knew about.

The rule would also pressure startups to keep hollowed-out operations running during review, consuming investors’ capital to avoid triggering the acquirer’s obligations. The transitional provision compounds this uncertainty. It applies the amendment to business combinations whose notification obligation arises after its effective date, so a transaction signed years earlier could become notifiable if the transferor winds down afterward.[53]

Remedies present a further difficulty. Corrective measures for unlawful business combinations, such as orders to dispose of shares or transfer a business, presuppose assets the acquirer can relinquish.[54] The KFTC cannot order employees to return to a former employer without conflicting with their constitutional freedom to choose an occupation. Requiring the acquirer to dismiss the team would not restore competition, because nothing would ensure that the employees return to the target. Rescinding the license would deprive the target and its investors of the consideration that often keeps the target alive. A pre-closing hold would, in practice, prevent individuals from starting jobs they have already accepted.[55] A notification obligation that cannot lead to a workable remedy imposes costs without corresponding benefits.

Recommendation: Delete subparagraph (2) of the definition of a business, or limit it to personnel who move together with transferred property rights, such as exclusive intellectual-property rights, ownership of source code or model weights, or customer contracts. Replace the new “material part” criterion with a test focused on what the transferor loses and whether rivals retain access to the technology. Tie implementation to acts within the parties’ control.

III. The Rules Would Misidentify Reportable Transactions

Under the current Notification Guidelines, the “transfer price” includes consideration paid for the business or fixed operating assets, liabilities assumed, and annual fees for leases and management entrustment.[56] Each element reflects something the transferor gives up, such as ownership of assets or the right to operate a business for a specified period.

For organized personnel, the Proposed Amendment would count any economic consideration paid to the transferor for the transaction, “regardless of its designation.” Examples include payments to release rights relating to departing personnel and license fees for intellectual property needed for the business.[57] The KFTC presents this as a technical change to account for consideration spread across several contracts.[58] Its consequences are substantive. The transfer price determines whether a transaction meets the price requirement of the “material part” test. A personnel-based transfer that falls below that threshold is not notifiable, even if it satisfies the test’s other component.[59]

These rules would poorly distinguish transactions that warrant review. License fees and payments supporting lawful hiring do not establish that the acquirer gains control over the target’s competitive position, and their size does not establish competitive harm (Subsection A). Counting only payments to the transferor would burden cooperative deals, encourage uncompensated recruitment, and shift value toward selected employees at the expense of investors and those who remain. Unclear aggregation rules would compound the uncertainty and encourage parties to restructure payments to avoid notification (Subsection B).

A. Transaction Prices Do Not Establish Competitive Harm

A nonexclusive license fee buys access to technology without control over it. The licensor remains free to use the technology and license it to others, including the licensee’s rivals. For this reason, the U.S. Federal Trade Commission’s Premerger Notification Office has long treated exclusive licenses as potentially reportable asset acquisitions and nonexclusive licenses as outside the notification requirement.[60]

The transactions that prompted the Proposed Amendment illustrate the point. After Microsoft’s arrangements with Inflection, Inflection retained “the unfettered ability to commercialise” its models, including by licensing them to third parties. It also announced that it would offer its latest model through licensing partnerships and application programming interfaces (APIs) that let other software access its models.[61] Counting a nonexclusive license fee as the price of a business conflates access with control. The test proposed in Section II.C would capture a formally nonexclusive license that operates as exclusive in practice. The transfer-price calculation need not substitute for that inquiry.

The same problem affects the transaction-value threshold. Korea introduced it in 2021 on the premise that a high price for a target with little revenue may signal competitive significance that revenue-based thresholds miss. Authorities have rarely invoked it since.[62] Whatever the merit of that premise for acquisitions, it does not carry over to license-and-hire arrangements. Those payments buy a license the licensor can grant again and a set of employment decisions it cannot guarantee.

Payments to release rights over departing personnel serve a similar function. In the Microsoft/Inflection arrangements, the reported $30 million payment secured a waiver of claims arising from Microsoft’s hiring of Inflection staff. Such payments remove a litigation threat that would otherwise deter lawful mobility. They support the hiring arrangement without purchasing the target’s business.[63]

More generally, transaction prices reflect scarce talent, the opportunities new technology may create, the benefits of combining the target’s capabilities with the acquirer’s assets, and competition among bidders. None of these establishes that the transaction eliminates a competitive constraint imposed by the target.[64]

The price requirement compounds the problem. Measured against the transferor’s total assets, the 10% threshold captures the transaction’s size relative to the target rather than its effect on competition. For a small startup, almost any meaningful payment will meet it.

B. The Price Rule Distorts Deal Structures

The Proposed Amendment counts only consideration paid to the transferor. These payments allow a startup’s investors, remaining employees, and option holders to recover value. An acquirer that recruits a startup’s team without paying the startup would not meet the price requirement. The rule would therefore burden cooperative transactions while leaving uncompensated raids on a startup’s workforce untouched. This result runs counter to the Proposed Amendment’s stated purpose of protecting venture companies.[65] It would encourage structures that leave investors with less and substitute litigation for negotiated settlements.

The rule would also encourage parties to shift value from payments to the transferor toward compensation paid directly to individual employees. This would arbitrarily favor founders and senior researchers over investors and employees who stay behind. The disputes following the Windsurf transaction illustrate these conflicts over how to distribute a deal’s proceeds.[66]

The press release explains that “regardless of its designation” aims to capture consideration spread across several contracts, but the Proposed Amendment provides no aggregation rules.[67] It does not identify which agreements form part of a transaction or the period over which payments count. Nor does it explain whether accompanying agreements for cloud computing, chips, or distribution count toward the price, or whether payments to the transferor’s shareholders count at all. These gaps prevent parties from determining whether they must notify and encourage them to stagger payments or route them through affiliates.

Recommendation: Exclude nonexclusive license fees and payments ancillary to hiring, such as releases of noncompete covenants and agreements not to sue, from the transfer price. Specify which agreements count, the aggregation period, and whether payments to the transferor’s shareholders count. If the KFTC retains a price rule for personnel-based transfers, set a separate, substantially higher threshold to reduce the risk of capturing harmless transactions, rather than treating price as a measure of competitive significance.

IV. Conclusion and Recommendations

The concern that a firm might acquire a competitor’s business in substance without acquiring it in form supports a narrow rule. The Proposed Amendment instead would reach ordinary licensing and hiring, measure transactions by prices that do not establish competitive harm, and tie notification obligations to events outside the acquirer’s control. Employees remain free to leave, and conventional merger remedies cannot reliably restore a team to its former employer.

Published reviews and the evidence from AI markets instead favor restraint. Existing law already addresses harmful hiring practices, while acquihires can preserve teams, strengthen workers’ bargaining power, and provide exits that support startup investment. Wrongly deterring these transactions would impose broad, often unobservable costs extending beyond AI and global technology firms. Acquirers could shift toward uncompensated recruitment, staggered payments, and litigation, while costlier, less certain exits would reduce the returns that finance venture investment in Korea.

Firms would also face obligations they cannot reliably assess in advance under a combination of tests no peer jurisdiction has adopted. The United Kingdom, which the KFTC consulted, is moving toward narrower jurisdiction. Its government has proposed to narrow the open-ended share-of-supply test that gives the CMA broad discretion to assert jurisdiction, and the CMA has said it is less likely to investigate global transactions that other authorities are reviewing.

Guidelines serve their purpose when firms can predict how the KFTC will treat a transaction. Predictable merger control also helps make Korea an attractive place to found, finance, and acquire innovative companies. We recommend that the KFTC withdraw the Proposed Amendment and address harmful conduct through Articles 5, 40, and 45 of the MRFTA. Any expansion of statutory notification obligations should come from the National Assembly.

If the KFTC nonetheless proceeds, we recommend that it:

1. Delete subparagraph (2) of the definition of a business, or limit it to personnel who move with transferred property rights that together constitute an operating business.
2. Replace the new “material part” criterion with a test asking whether the transferor loses its ability to continue the business and rivals lose commercially meaningful access to the relevant technology.
3. Expressly exclude ordinary recruitment and provide safe harbors for nonexclusive licenses that leave the licensor free to license others and for transactions after which the transferor continues operating.
4. Exclude nonexclusive license fees and payments ancillary to hiring from the transfer price. Specify the agreements and periods covered by aggregation, clarify whether shareholder payments count, and set a separate, substantially higher threshold for personnel-based transfers.
5. Tie implementation to acts within the parties’ control and confirm that an individual’s start of employment never constitutes implementation.
6. Exclude transactions signed before the effective date, regardless of when the transferor ceases its business.
7. Publish worked examples addressing transactions like those involving Inflection, Character.AI, Windsurf, and Groq, and provide a channel for informal pre-notification consultation.
8. Review the amended guidelines after a defined period and publish data on notifications and their outcomes.

[1] Press Release, Korea Fair Trade Comm’n, Sinyuhyeong Gieop Gyeolhap-ui Sagakjidae Haeso-reul Wihan Gieop Gyeolhap-ui Singo Yoryeong Gaejeongan Haengjeong Yego [Advance Notice of Proposed Amendment to the Notification Guidelines for Business Combinations to Eliminate Blind Spots for New Types of Business Combinations] (Sept. 9, 2026) [hereinafter Press Release]; id. attach. (comparison of current and proposed provisions) [hereinafter Proposed Amendment]. This comment cites lettered subdivisions of the Notification Guidelines by their English equivalents (가 = a, 나 = b, and so on). Unless otherwise noted, translations of Korean-language materials are our own.

[2] See, e.g., Dirk Auer & Onyeka Aralu, ‘Reverse Acquihires’ and the Market for Talent: Antitrust Theory in Search of a Case, Concurrences No. 9-2026 (forthcoming) [hereinafter Auer & Aralu, Reverse Acquihires]; Dirk Auer & Mario A. Zúñiga, AI Partnerships and Competition: Damned if You Buy, Damned if You Don’t (Int’l Ctr. for L. & Econ., Working Paper, May 2026).

[3] Geoffrey A. Manne, Dirk Auer & Lazar Radic, Comment of the International Center for Law & Economics Concerning the Proposed Amendments to Korea’s Merger Review Guidelines (Int’l Ctr. for L. & Econ., Dec. 5, 2023).

[4] Dokjeom Gyuje mit Gongjeong Georae-e Gwanhan Beomnyul [Monopoly Regulation and Fair Trade Act], Act No. 3320, Dec. 31, 1980, amended by Act No. 17799, Dec. 29, 2020, art. 9(1)(iv) (S. Kor.) [hereinafter MRFTA].

[5] Proposed Amendment, supra note 1, §§ 4(b)(2), 4(c)(1), 4(d)(3), 4(f).

[6] Microsoft/Inflection, Merger Case No. 08700.005792/2026-71, Decision No. 475/2026/CGAA5/SGA1/SG, ¶ 19 (Braz. Admin. Council for Econ. Def. 2026), https://sei.cade.gov.br/sei/modulos/pesquisa/md_pesq_documento_consulta_externa.php?HJ7F4wnIPj2Y8B7Bj80h1lskjh7ohC8yMfhLoDBLddbNMaDk2AO6MZR6aBnSzp6KvSE3SnFkJaYX5WBvy-aRe0rVTcGut04mNHPDwDq9y_ebH8rs7cqfZFpufAf0t1Iw (“Neste contexto, entende-se que a Operação não possui o condão de acarretar alterações significativas ao ambiente concorrencial no Brasil, não demandando, deste modo, um maior aprofundamento da análise.”).

[7] Daehanminguk Heonbeop [Constitution] art. 15 (S. Kor.) (“All citizens shall enjoy freedom of occupation.”).

[8] Press Release, supra note 1.

[9] See Bloomberg Podcasts, FTC Chair Andrew Ferguson Talks Acquihires | Bloomberg Talks, YouTube (Jan. 16, 2026), https://www.youtube.com/watch?v=8u_hazxCplM; Letter from Sens. Elizabeth Warren, Ron Wyden & Richard Blumenthal to Gail Slater, Assistant Att’y Gen., Antitrust Div., U.S. Dep’t of Just., and Andrew N. Ferguson, Chairman, Fed. Trade Comm’n (Feb. 4, 2026), https://www.warren.senate.gov/wp-content/uploads/media/doc/final_-_warren_wyden_blumenthal_letter_to_the_department_of_justice_and_the_federal_trade_commission_on_big_tech_reverse_acqui-hires.pdf.

[10] See Auer & Zúñiga, supra note 2, at 6–7.

[11] Id. at 23–24; Brian J. Broughman, Matthew T. Wansley & Samuel N. Weinstein, No Exit, 100 N.Y.U. L. Rev. 1481, 1484–86 (2025) (situating license-and-hire transactions among responses to tougher merger enforcement).

[12] Press Release, supra note 1.

[13] Krystal Hu & Harshita Mary Varghese, Microsoft Pays Inflection $650 Mln in Licensing Deal While Poaching Top Talents, Source Says, Reuters (Mar. 21, 2024), https://www.reuters.com/technology/microsoft-agreed-pay-inflection-650-mln-while-hiring-its-staff-information-2024-03-21.

[14] Competition & Mkts. Auth., Microsoft Corporation’s Hiring of Certain Former Employees of Inflection AI and Its Entry into Associated Arrangements, Case ME/7103/24, Phase 1 Decision, ¶ 22 (Sept. 4, 2024), https://assets.publishing.service.gov.uk/media/6719ff5f549f63039436b3c8/__Full_text_decision__.pdf.

[15] Microsoft/Inflection, Decision No. 475/2026/CGAA5/SGA1/SG, ¶ 19.

[16] Id. ¶¶ 14–15.

[17] Press Release, Bundeskartellamt, Taking Over Employees May Be Subject to Merger Control in Germany—Bundeskartellamt Not Competent to Review Microsoft/Inflection Transaction as Inflection Has No Substantial Operations in Germany (Nov. 29, 2024), https://www.bundeskartellamt.de/SharedDocs/Meldung/EN/Pressemitteilungen/2024/29_11_2024_Microsoft.htm.

[18] See Joined Cases C-611/22 P & C-625/22 P, Illumina, Inc. v. Comm’n, ECLI:EU:C:2024:677 (Sept. 3, 2024), https://eur-lex.europa.eu/legal-content/EN/TXT/HTML/?uri=CELEX:62022CJ0611; Bundeskartellamt, supra note 17 (noting the Commission’s Sept. 18, 2024, announcement that it would not issue a decision).

[19] Ben Cottier, Ben Snodin, David Owen & Tom Adamczewski, LLM Inference Prices Have Fallen Rapidly but Unequally Across Tasks, Epoch AI (Mar. 12, 2025), https://epoch.ai/data-insights/llm-inference-price-trends.

[20] Auer & Zúñiga, supra note 2; Geoffrey A. Manne & Dirk Auer, From Data Myths to Data Reality: What Generative AI Can Tell Us About Competition Policy (and Vice Versa), CPI Antitrust Chron., Feb. 2024, at 2, 7–8.

[21] Colleen Cunningham, Florian Ederer & Song Ma, Killer Acquisitions, 129 J. Pol. Econ. 649 (2021).

[22] Florian Ederer, Regina Seibel & Timothy Simcoe, Digital (Killer?) Acquisitions (Nat’l Bureau of Econ. Rsch., Working Paper No. 35762, Sept. 2026), https://www.nber.org/papers/w35762.

[23] MRFTA, supra note 4, art. 45(1)(viii); Dokjeom Gyuje mit Gongjeong Georae-e Gwanhan Beomnyul Sihaengnyeong [Enforcement Decree of the Monopoly Regulation and Fair Trade Act], annex 2, item 8(b) (S. Kor.). This provision treats the conduct as an unfair trade practice rather than an abuse of dominance. The burden of proof is therefore lower: Plaintiffs need not show that the defendant holds a dominant position.

[24] Korea Fair Trade Comm’n, Bulgongjeong Georae Haengwi Simsa Jichim [Guidelines for Review of Unfair Trade Practices] pt. V, item 8(b) (S. Kor.).

[25] Geoffrey A. Manne, Gotta Catch ’Em All? Antitrust and the AI Talent Wars, Truth on the Mkt. (June 18, 2026), https://laweconcenter.org/resources/gotta-catch-em-all-antitrust-and-the-ai-talent-wars (discussing Shaolong Wu & Zefan Qian, Talent Hoarding and Upstream Innovation: Labor Market Distortions by Large Incumbents (working paper)).

[26] MRFTA, supra note 4, arts. 5, 40; Bujeonggyeongjaeng Bangji mit Yeongeopbimil Boho-e Gwanhan Beomnyul [Unfair Competition Prevention and Trade Secret Protection Act], Act No. 911, Dec. 30, 1961 (as amended) (S. Kor.).

[27] John F. Coyle & Gregg D. Polsky, Acqui-Hiring, 63 Duke L.J. 281 (2013).

[28] Auer & Zúñiga, supra note 2, at 14–15.

[29] See Elizabeth Pollman, Startup Failure, 73 Duke L.J. 327 (2023) (discussing soft-landing acquisitions and acquihires as alternatives to formal insolvency); Louis Lehot, What an Acqui-Hire Looks Like for an AI Startup Today, Medium (July 16, 2026) (noting that most acquihires involve companies that raised less than $5 million and could not raise another round).

[30] Geoffrey A. Manne, Samuel Bowman & Dirk Auer, Technology Mergers and the Market for Corporate Control, 86 Mo. L. Rev. 1047 (2022).

[31] Geoffrey A. Manne, Dirk Auer, Lazar Radic & Mario A. Zúñiga, Comments of the International Center for Law & Economics to the Japan Fair Trade Commission Re: Opinions on the Draft of the Business Combination Guidelines 3–4 (Int’l Ctr. for L. & Econ., Aug. 27, 2026).

[32] For an explanation of the error-cost framework, see Frank H. Easterbrook, The Limits of Antitrust, 63 Tex. L. Rev. 1, 15 (1984).

[33] Kim Nam-myung, Korea to Screen “Acqui-Hire” Deals as Corporate Mergers, Seoul Econ. Daily (Sept. 10, 2026), https://en.sedaily.com/finance/2026/09/10/korea-to-screen-acqui-hire-deals-as-corporate-mergers.

[34] MRFTA, supra note 4, art. 9(1)(iv).

[35] Korea Fair Trade Comm’n, Gieop Gyeolhap-ui Singo Yoryeong [Notification Guidelines for Business Combinations] § 4(b) (S. Kor.) [hereinafter Notification Guidelines].

[36] Id. § 4(c).

[37] Notification Guidelines, supra note 35, § 4(b).

[38] S. Kor. Const. art. 15.

[39] 15 U.S.C. § 17; Nokes v. Doncaster Amalgamated Collieries Ltd. [1940] AC 1014 (HL).

[40] Auer & Aralu, Reverse Acquihires, supra note 2; Onyeka Aralu & Dirk Auer, Acquihires and Antitrust: When Buying the Team Isn’t Buying the Company, Truth on the Mkt. (Apr. 9, 2026), https://truthonthemarket.com/2026/04/09/acquihires-and-antitrust-when-buying-the-team-isnt-buying-the-company [hereinafter Aralu & Auer].

[41] Oliver Hart & John Moore, Property Rights and the Nature of the Firm, 98 J. Pol. Econ. 1119 (1990).

[42] J. Daniel Kim, Startup Acquisitions as a Hiring Strategy: Turnover Differences Between Acquired and Regular Hires, 9 Strategy Sci. 118 (2024).

[43] Weiyi Ng & Toby E. Stuart, Acquired Employees Versus Hired Employees: Retained or Turned Over?, 43 Strategic Mgmt. J. 1025 (2022); J. Daniel Kim, Startup Acquisitions, Relocation, and Employee Entrepreneurship, 43 Strategic Mgmt. J. 2189 (2022).

[44] See Nina Fridman, When a Mass Resignation Becomes a Merger: Rethinking Asset Acquisitions for the AI Era, 93 U. Chi. L. Rev. 999 (2026); David T. Wong, Comment, An Acquisition by Another Name: Reverse Acquihires Under the Clayton Act, 135 Yale L.J. 3175 (2026) (treating human capital as an asset while distinguishing it from employees themselves).

[45] Proposed Amendment, supra note 1, § 4(c)(1).

[46] Notification Guidelines, supra note 35, § 4(c); Proposed Amendment, supra note 1, § 4(c)(2).

[47] MRFTA, supra note 4, art. 11(1); Shin & Kim LLC, Merger Control in Korea Vol. 13—KFTC to Bring “Acquihires” Within the Scope of Merger Review (May 29, 2026), https://shinkim.com/eng/media/newsletter/3312 [hereinafter Shin & Kim].

[48] Aralu & Auer, supra note 40.

[49] Auer & Aralu, Reverse Acquihires, supra note 2; Press Release, Groq, Groq and Nvidia Enter Non-Exclusive Inference Technology Licensing Agreement to Accelerate AI Inference at Global Scale (Dec. 24, 2025).

[50] Kim Nam-myung, supra note 33.

[51] Proposed Amendment, supra note 1, § 4(f).

[52] MRFTA, supra note 4, art. 11(8); Notification Guidelines, supra note 35, § 4(e).

[53] Proposed Amendment, supra note 1, addendum ¶ 2.

[54] MRFTA, supra note 4, art. 14.

[55] Auer & Aralu, Reverse Acquihires, supra note 2.

[56] Notification Guidelines, supra note 35, § 4(d).

[57] Proposed Amendment, supra note 1, § 4(d)(3).

[58] Press Release, supra note 1 at 2.

[59] Proposed Amendment, supra note 1, § 4(c).

[60] Fed. Trade Comm’n, Premerger Notification Off., Informal Interpretation 9307006 (July 1993), https://www.ftc.gov/legal-library/browse/hsr-informal-interpretations/9307006; Eric Fruits, Daniel J. Gilman, Onyeka Aralu & Brian Albrecht, ICLE Comments to FTC and DOJ on Premerger Notifications pt. IV.B (Int’l Ctr. for L. & Econ., May 22, 2026).

[61] Competition & Mkts. Auth., supra note 14, at 21.

[62] Shin & Kim, supra note 47, § 3.

[63] Broughman, Wansley & Weinstein, supra note 11; Auer & Aralu, Reverse Acquihires, supra note 2, pt. III (describing such covenants as ancillary to lawful hiring).

[64] Manne, Bowman & Auer, supra note 30; Geoffrey A. Manne et al., Comments of the International Center for Law & Economics: EU Draft Merger Guidelines—Public Consultation pt. II.A.3 (Int’l Ctr. for L. & Econ., June 22, 2026) (explaining that acquisition prices reflect complementarity, option value, and talent).

[65] See Kim Nam-myung, supra note 33.

[66] Auer & Aralu, Reverse Acquihires, supra note 2, pt. IV.

[67] Press Release, supra note 1, at 2.

I appreciate the opportunity to comment on the FTC’s Proposed Enforcement Policy Statement Regarding Personalized Pricing (“Proposed Policy”).[1] This comment offers my own views as a long-time student of and practitioner in the FTC’s consumer protection mission.[2]

The comment begins with a brief overview of personalized pricing and the issues it raises as a form of price discrimination in Section I. Section II then argues that when there are competing sellers, consumers can easily avoid any injury from personalized pricing by the simple expedient of searching for another price. Section III argues that the Commission’s premise that consumers generally expect one price for all customers is unsubstantiated. Consumers are well aware that different consumers pay different prices in ordinary purchase situations. Section IV considers discounts based on personal information. It argues that such practices benefit consumers, and that the Commission should create a safe harbor to avoid requiring disclosures for personalized discounts. Section V argues that disclosure is an inappropriate remedy for the problems the Commission seeks to address and should, in any event, be tested before it is required.

I. An Overview of Personalized Pricing

Some critics of the information economy have become concerned about personalized pricing, in which businesses use consumer-specific information to set different prices for different consumers. By definition, charging different prices for different consumers (assuming costs are the same) is price discrimination. The economics literature has long been clear that price discrimination can be either welfare increasing or welfare reducing, depending on the circumstances.[3] Whether a given instance of price discrimination is good or bad for consumers is thus an empirical question. Nevertheless, many scholars have argued that common forms of retail price discrimination lead to increased output and enhanced consumer welfare.[4]

As the Proposed Policy notes, “the extent to which businesses currently use personalized pricing is not well understood, and the effects of personalized pricing on consumers are unclear.”[5] Instead, the discussion is driven by concern about a wide variety of hypothetical business practices which are not known to have actually occurred, and which may never occur.

The hypothetical scenario that has attracted the most attention is the possibility that a monopolist could charge each individual consumer the maximum amount that consumer would be willing to pay, thus extracting all consumer surplus. Although a theoretically interesting polar case, such personalized pricing is not possible in practice: Statistical uncertainty remains. Even the best models cannot predict with certainty, and data, even in the physical sciences and especially in economics, are subject to measurement errors. Dubé and Misra show by example that, even in a monopoly model that incorporates statistical uncertainty, a database that allows different prices for different consumers may increase both profits and consumer surplus.[6] In their careful experimental study, personalized pricing never extracted more than 30% of consumer surplus that would result from uniform pricing, whereas “true perfect price discrimination … would extract all consumer surplus.”[7] Moreover, in their experiment, more than 60% of consumers received a lower price with personalized pricing. Why such a pricing outcome might warrant consumer protection enforcement is hardly obvious.

More important than the limitations of personalized pricing in the monopoly context is the fact that most industries are not monopolies. When there are competitors, the same tools and data that allow a monopolist to charge more to consumers who are more willing to pay also enable competitors to identify the consumers who are most overcharged and compete for their business through lower prices. Firms can also offer discounts to consumers who prefer their rival’s product, again pressuring the rival to lower price.[8] These uses of information-based pricing offer clear consumer benefits.

Empirical questions that underlie a public policy deserve empirical answers rather than adopting a policy that assumes an answer. The impact of price discrimination on consumer welfare is just such a question, as economists have recognized for decades. Empirical answers about the effects of personalized pricing as a method of price discrimination are, at present, lacking. Whether a particular implementation of personalized pricing benefits consumers or harms them should be central to the Commission’s enforcement decisions. Under the Proposed Policy, the Commission does not appear to ask that key factual question in deciding to bring a case.

The specific “personalized pricing” scenarios the Proposed Policy identifies as requiring disclosure highlight the important role of competition as a constraint. Marketing delivery services to those who are unable to leave home, milk to people with several children, hotels to people who need to travel, or home security to crime victims are examples that appear to be competitive markets. Information identifying these consumers makes competition, including offering a lower price, more likely, not less. Similarly, in the Commission’s example of a consumer standing in the store who is shown a higher price online, the consumer can buy the item off the shelf. Moreover, a competitor using the same information to offer online discounts to consumers who are in its rival’s stores would seem to produce clear consumer benefits. As discussed in the next section, consumers can easily avoid injury in such circumstances.

The most troubling of the Commission’s scenarios is the rideshare company charging a premium price for a consumer in dire need of transportation, particularly if the consumer has no reasonable alternatives. Here, however, the problem is more akin to duress, rather than a failure to disclose personalized pricing. The common law of contract has long policed closely related situations. For example, a seller may refuse to continue performance under a contract without a price higher than previously agreed. In another example, a seller who finds a prospective buyer in a sensitive position with few if any alternatives, raises the price to extraordinary levels. In either circumstance, common law courts scrutinize the contract with extra care and may refuse to enforce it.[9] Most importantly, although this exploitation may constitute a Section 5 violation,[10] the problem is not the failure to disclose. No amount of disclosure about how the price was set will either generate new alternatives or relieve the consumer of the dilemma.

II. When Alternatives Exist, Consumers Can Easily Avoid Any Injury from Personalized Pricing

Consumers cannot buy anything without first locating a seller. Although some may purchase from the first seller they find, consumers frequently gather information about the offerings of other sellers to choose the best option. That process of consumer search means that consumers do not come to a particular offering in ignorance. They have some information about prices at other sellers, or perhaps this seller at an earlier time, that provides context for the offered price. Even when a new seller or new product is first encountered, consumers can easily generate a context by searching for other offerings. In the online context that seems to be a primary concern of the proposed personalized pricing policy, the simple expedient of clicking the next link will generate additional information about whether the price is high or low.

Rather than focusing on the injury, the draft statement’s analysis of materiality and avoidability focuses on the process that allegedly generates the injury.[11] It is true that if consumers are unaware of personalized pricing they would not know to take steps like using a VPN or a private browser.[12] But the injury is not the process of setting the price, it is the high price that may result. Consumers can avoid the actual injury with the same simple process they routinely employ to avoid any price they think is high: check another seller. Consumers can therefore reasonably avoid any injury that might result from personalized pricing.

Of course, consumers may decide to purchase without additional search. That decision, however, is based on the perceived benefits and costs of further information gathering. The manner in which the seller determined the price does not affect the search decision. It is likely that consumers who choose not to search will pay a higher price on average, but not searching may nonetheless be a rational choice. In addition to avoiding the costs of search, some consumers prefer familiarity with one seller and will pay for that benefit. Choosing not to search also captures the Commission’s scenario of a rideshare company charging more for a consumer who has only one installed rideshare app. Moreover, installing another app would seem an easy way for a consumer to avoid injury. In any event, disclosure at the point of sale will not convert a non-searcher into a searcher and does nothing to reduce the costs of search.

The presence or absence of personalized pricing has no effect on search behavior in two of the most studied strategies in the economics literature. In a fixed sample size strategy, consumers choose a number of sellers to sample based on the costs of checking another price and the expected benefit of finding a lower price.[13] Personalized pricing does not affect either side of this equation. Another much-studied search strategy is an optimal stopping rule,[14] in which consumers continue to search until they find a price below some specified value. Again, nothing about the presence or absence of personalized pricing would affect this determination or the transaction price that results.

III. The Commission’s Premise that Consumers Expect One Price for All Is Unsubstantiated

The information that consumers need to make optimal choices is information about the price itself, not how it was determined. As Nobel Laureate Friedrich Hayek observed in his seminal paper The Use of Knowledge in Society, the beauty of the price system is that the price aggregates and summarizes all information about value and cost into a single statistic, one that provides the basis for decisions without exploration of the underlying details.[15] Whatever consumers may expect about how it was determined, that summary remains the same for the particular consumer and the particular proposed transaction.  In most circumstances consumers can remain blissfully ignorant of the process by which prices are set and the range of information used to set them.

Charging different prices to different consumers is as old as retailing. Consumers who are more willing to pay for a product or service, or less sensitive to price, are often charged higher prices. Consumer characteristics may serve as a proxy, as is the case with discounts for students or for the elderly. But many traditional price discrimination techniques rely on consumers who care more about price to self-identify, usually by their willingness to jump over some “hurdle” to qualify for a lower price. Rebate offers or discount coupons are common examples: Consumers who are more price sensitive are more willing to clip coupons or jump the hurdles necessary to claim the rebate. As one economics textbook noted, “[i]t is a rare product whose seller does not use the hurdle model of differential pricing.”[16]

Indeed, nothing is more common in the experience of every consumer than the simple fact that prices vary. Different sellers offer the same goods at different prices, in different locations, at different times, and to different consumers. Similar goods that are close substitutes are available at a similar array of prices. The only way for any consumer to know whether a particular price is a good one is to compare it to other prices—to search for a price they are willing to pay.

Even in circumstances most favorable to the Commission’s assertion that consumers have “reasonable expectations” of one price for all, it is highly likely that consumers understand that the transaction price will differ for different consumers. Consider a posted price for an item in a grocery store, arguably an express representation of one price for all. Consumers understand that at checkout some will pay the posted price, but others will pay less because they use a coupon on the purchase (and perhaps different coupons offering different discounts). Still others may receive a different loyalty program discount or take advantage of a volume discount. The range of prices expands even further when we consider competing brands, including store brands that may be functionally identical. Each of these arrangements means that different consumers pay different prices, based on their personal choices and behavior. Such pricing arrangements are ubiquitous in retailing, obvious to consumers, and make clear that the posted price is not “one price for all.”

The Proposed Policy cites evidence that consumers do not understand the nature and extent of data collection and use.[17] It does not follow that they therefore expect “one price for all,” despite their everyday experience to the contrary. Moreover, the Commission recognizes that consumers do expect individualized pricing in some circumstances, but it offers no principled distinction between those instances and others where consumers allegedly expect a single price, and no way for a company to assess what consumers “reasonably expect” in any particular instance.

In any event, consumer expectations about prices and how they are set are a foundation of quicksand for enforcement that actually benefits consumers. The Commission has a sad history of using claims of “deception” to attack discounters and other low-price sellers,[18] spurred on by competitors seeking to suppress competition that benefits consumers. Presumably most consumers who choose to buy “expect” that the price they are paying is a good one. By the logic of the Proposed Policy, however, the failure to disclose that the price is in fact higher than most other offerings in the market would seem to be a deceptive practice. That approach, however, would make the Commission a price regulator, something it clearly is not under its statute.

IV. The Proposed Policy Ignores Personalized Discounts that Benefit Consumers

Although the possible downsides of personalized pricing are at this point largely hypothetical, concrete benefits of information-based pricing abound. Using the same data and techniques as personalized pricing, loyalty programs and discount email marketing are widespread. Industry estimates indicate that millions of consumers are members of loyalty programs, often multiple programs, that offer discounts and other promotional benefits based on the consumer’s history with the seller. Capital One Shopping reports that discounts are the strongest drivers of loyalty program participation.[19] Zoho, a provider of online platforms for loyalty programs, reports that 80% of Americans are members of at least one loyalty program.[20] EMARKETER reports that the average consumer is a member of a surprisingly high 17 loyalty programs,[21] and notes that the best programs tailor offers and content based on the individual member’s preferences and behavior. Email marketing also makes extensive use of consumer information. Capital One reports that 77% of consumers want to receive emails offering discounts, and that personalized promotional emails with discounts are opened 29% more often and clicked 41% more often than those without.[22]

These programs offer clear consumer benefits, reducing the costs of living for American families. They use the same kinds of data, and the same or similar analytical techniques, to determine how large a discount (if any) to offer any individual consumer. Unless the Commission creates a safe harbor for such discounts, the analysis of the proposed policy statement would seem to require exactly the same personalized pricing disclosure that the policy statement contemplates when personalization leads to higher prices. In the Commission’s example of a price for a security system based on knowledge that the consumer was a recent crime victim, the seller would have to disclose that “this price is personalized based on data showing you are a recent crime victim” whether the price was higher or lower than the usual price. When the price is a discount, however, the disclosure offers no benefits whatsoever. Defining “discounts” with enforceable precision is a difficult challenge, but the Commission can clearly state that it does not intend to require disclosure when information is used to offer discounts, and consider whether a given price is in fact a “discount” on a case-by-case basis.

Although the Proposed Policy is based on disclosure rather than a prohibition of personalized pricing, given the well-known fact that any communication can confuse or mislead a significant number of consumers,[23] required disclosure when prices are discounted risks harming consumers. The legal risk, the cost of the disclosure itself, and the risk of consumer misinterpretation of the disclosure will discourage sellers from offering these discounts. The Commission’s past attacks on discounting argue strongly for a safe harbor, rather than reliance on the Commission’s enforcement discretion.

V. Disclosure Is an Inappropriate Remedy

The proposed policy statement would require that firms using personalized pricing “disclose not just that the price is personalized, but also the basis for that personalization and the types of data on which the personalization is based.”[24] In its evaluation of disclosures, the Taskforce on Federal Consumer Financial Law Report noted that “when disclosures are required, it should be with a clear, and articulated, goal in mind.”[25]

The contemplated disclosure fails this test. Is it simply intended to discourage personalized pricing? It will likely have that effect, but, as noted above, that will raise prices to the many beneficiaries of personalized discounts. Is it intended to motivate additional consumer search for a better price? If so, it would seem equally valuable to require any seller to disclose that its price is relatively high, whether personalized or not. Is the goal to encourage consumer use of behaviors and technologies that seek to conceal their information? If so, the disclosure should be at the point at which information is collected, when consumers can act on its message. Disclosure when a price is offered is in the wrong place and at the wrong time.

Rather than facilitating consumer choices, which should be the goal of all required disclosures, the proposed disclosure appears to be a “normative” disclosure – information provided not because consumers are interested, but because policymakers think consumers should take it into account.[26]

Like any disclosure, this one may be misunderstood. The Commission’s own research has shown that additional information can lead to worse consumer decisions, either because it is misunderstood or because consumers assign it more significance than it should have.[27] To ensure that consumers actually benefit from disclosure, the Commission should first test it, and only then consider requiring it.

Conclusion

In most circumstances, personalized pricing is unlikely to harm consumers. There may be harm, and a Section 5 violation, if there are instances in which personalized pricing borders on duress. Disclosure, however, does nothing to address the problem, and leaves a possibly better-informed consumer with the same unpalatable choice.

It is generally easy for consumers to avoid any injury from personalized pricing by the simple expedient of searching for a better price. The Commission’s premise that consumers generally expect one price for all is unsubstantiated. It seems highly likely that most consumers are aware that prices differ from one consumer to another because of coupon use, volume discounts, and loyalty programs. Because personalized discounts offer substantial consumer benefits, the Commission should create a safe harbor making clear that disclosures are not necessary when information is used to offer a discount. Disclosure, particularly untested disclosure, is an inappropriate remedy.

[1] Fed. Trade Comm’n, Proposed Enforcement Policy Statement Regarding Personalized Pricing, Matter No. P034101 (Aug. 19, 2026) [hereinafter “Proposed Policy”], https://www.ftc.gov/system/files/ftc_gov/pdf/p034101-ftc-enforcement-policy-statement-re-personalized-pricing-proposed-for-public-comment.pdf.

[2] I am currently Emeritus Professor of Strategic Management and Public Policy at the George Washington School of Business and a Senior Scholar at the GW Regulatory Studies Center. I was on the staff of the FTC in the Bureau of Economics and in the Bureau of Consumer Protection as an Assistant to the Director, Associate Director, and Acting Deputy Director from 1977 to 1986, and Director of the Bureau of Consumer Protection from 2001-2004. I am grateful to the International Center for Law and Economics for financial support.

[3] See Jean Tirole, The Theory of Industrial Organization 138 (1988) (“The welfare effects of third-degree price discrimination are ambiguous.”).

[4] See 14 Phillip E. Areeda & Herbert Hovenkamp, Antitrust Law: An Analysis of Antitrust Principles and Their Application ¶ 2340c (2024) (“most price discrimination is socially beneficial in that it produces higher output and thus yields greater consumer benefits than forced nondiscriminatory pricing”); see also Benjamin Klein, Price Discrimination and Market Power, in 2 Issues in Competition Law and Policy 977, *20 (Wayne D. Collins ed., 2008), available at https://papers.ssrn.com/sol3/papers.cfm?abstract_id=1657202 (“[P]rice discrimination can be used and in some circumstances accepted as a legitimate competitive business justification. . . .”).

[5] Proposed Policy, supra note 1, at 3.

[6] See Jean-Pierre Dubé & Sanjog Misra, Personalized Pricing and Consumer Welfare, 131 J. Pol. Econ. 131, 142 tbl. 1 (2023).

[7] Dubé & Misra, supra note 7, at 170. They note that “[e]ven with expanded data collection, it is unlikely that a firm could truly perfectly price discriminate using consumer data.” Id.

[8] See Jacques-François Thisse & Xavier Vives, On the Strategic Choice of Spatial Price Policy, 78 Am. Econ. Rev. 122 (1988).

[9] See, e.g., Restatement (Second) of Contracts §§ 175–76 (Am. L. Inst. 1981). See also Timothy J. Muris, Opportunistic Behavior and the Law of Contracts, 65 Minn. L. Rev. 521 (1981).

[10] The most promising theory is unfairness. There is substantial injury. By hypothesis, there is no alternative, and there are no offsetting benefits of the practice to consumers or competition. See, e.g., Holland Furnace Co. v. FTC, 295 F.2d 302 (7th Cir. 1961) (seller’s servicemen dismantled home furnaces and then refused to reassemble them until the consumers had agreed to buy services or replacement parts).

[11] Materiality is the injury inquiry in the context of deception: Would the consumer have done something differently to avoid the problem? Avoidability is the same question in the context of unfairness: Is there an injury-free alternative the consumer could reasonably have chosen?

[12] The proposed policy also points to correcting information and avoiding information collection in the first place as possible responses. But disclosure of personalization does not create an opportunity for correction, and existing disclosure requirements apply where correction rights exist. As discussed in Section V below, the point of sale is the wrong place for a disclosure aimed at information collection practices.

[13] Stigler examines a fixed sample size strategy in his seminal article on the economics of information. See George J. Stigler, The Economics of Information, 69 J. Pol. Econ. 213 (1961).

[14] See J. McCall, Economics of Information and Job Search, 84 Q.J. Econ. 113 (1970); Martin L. Weitzman, Optimal Search for the Best Alternative, 47 Econometrica 641 (1979).

[15] See Friedrich A. Hayek, The Use of Knowledge in Society, 35 Am. Econ. Rev. 519 (1945). Price conveys information even when the information is not observed directly, even when the information is not disclosed, and even when it may not be “consciously articulable” by any market participant. See Randal K. Quarles, Vice Chair for Supervision, Bd. of Governors of the Fed. Rsrv. Sys., Friedrich Hayek and the Price System Remarks at the Ninth Annual Conference of the William F. Buckley, Jr., Program at Yale 5 (Nov. 1, 2019), https://www.federalreserve.gov/newsevents/speech/files/quarles20191101a.pdf. For a formal model in a slightly different context, see Sanford Grossman, On the Efficiency of Competitive Stock Markets Where Trades Have Diverse Information, 31 J. Fin. 573 (1976).

[16] Robert H. Frank, Microeconomics and Behavior 434 (4th ed. 2006).

[17] See Proposed Policy, supra note 1, at 2–3 & nn. 4, 7.

[18] See Robert Pitofsky, Beyond Nader: Consumer Protection and the Regulation of Advertising, 90 Harv. L. Rev. 661 (1977); Robert Pitofsky, Randal Shaheen & Amy Mudge, Pricing Laws Are No Bargain for Consumers, 18 Antitrust 62 (2004); Timothy J. Muris & Paloma Zepeda, The Benefits, and Potential Costs, of FTC-Style Regulation in Protecting Consumers, 8 Competition L. Int’l 11 (2012).

[19] See Discount Marketing Statistics, Capital One Shopping (last updated June 15, 2026), https://capitaloneshopping.com/research/discount-statistics/.

[20] See 30+ Loyalty Program Statistics Businesses Should Know in 2026, Zoho Thrive (last visited Sept. 22, 2026), https://www.zoho.com/thrive/loyalty-program-software/statistics.html.

[21] See Arielle Feger, Why Loyalty Programs Are Hitting a Growth Ceiling, EMARKETER (July 31, 2026), https://www.emarketer.com/content/why-loyalty-programs-hitting-growth-ceiling.

[22] See Discount Marketing Statistics, supra note 20.

[23] As Karl Popper said, “Always remember that it is impossible to speak in such a way that you cannot be misunderstood: there will always be some who misunderstand you.” Karl Popper, Unended Quest: An Intellectual Autobiography 29 (1976) (emphasis in original). Academic studies of brief communications find that 20 to 30 percent of the audience misunderstand some aspect of the message. Regarding televised messages, see Jacob Jacoby, Wayne D. Hoyer & David A. Sheluga, Miscomprehension of Televised Communications 64 (1980). Regarding print communications, see Jacob Jacoby & Wayne D. Hoyer, The Comprehension and Miscomprehension of Print Communications: An Investigation of Mass Media Magazines (1987). Both studies compare advertisements with excerpts of editorial content designed to be roughly equal in length and find no significant differences in the extent of miscomprehension.

[24] Proposed Policy, supra note 1, at 1.

[25] Consumer Fin. Prot. Bureau, Taskforce on Federal Consumer Financial Law Report vol. I, at 326 (Jan. 2021, amended Jan. 2022) [hereinafter “Taskforce Report”], https://files.consumerfinance.gov/f/documents/cfpb_taskforce-federal-consumer-financial-law_report-volume-1_2022-01_amended.pdf. The CFPB’s Taskforce included four members with vast FTC experience, including me.

[26] See the discussion in the Taskforce Report, id. at 327–28 (noting that disclosures that reduce information costs help consumers achieve their goals but “normative disclosures seek to change their goals.”).

[27] See James M. Lacko & Janis K. Pappalardo, Fed. Trade Comm’n, Bureau of Econ., The Effect of Mortgage Broker Compensation Disclosures on Consumers and Competition: A Controlled Experiment 28–29 (Feb. 2004), https://www.ftc.gov/sites/default/files/documents/reports/effect-mortgage-broker-compensation-disclosures-consumers-and-competition-controlled-experiment/030123mortgagefullrpt.pdf. The experiment tested alternative disclosures on the closing forms required under the Real Estate Settlement Procedures Act, and found that when the broker’s compensation was disclosed, consumers were less likely to select the low-cost mortgage.

I. Introduction and Summary

The International Center for Law & Economics (‘ICLE’) welcomes the opportunity to respond to the Department for Business and Trade’s (‘DBT’) consultation on Swifter and Simpler Competition Redress, Regulatory Appeals and Competition Enforcement.[1] ICLE is a nonprofit, nonpartisan research centre that promotes rigorous economic analysis in public-policy debates. Its scholars have written extensively on competition enforcement, private litigation, and regulatory design. ICLE has also responded to the government’s Refining Our Competition Regime consultation, the Competition and Markets Authority’s (‘CMA’) merger-efficiencies review, and its proposed first conduct requirements under the Digital Markets, Competition and Consumers Act 2024 (‘DMCC Act’).[2]

ICLE submits these comments by email as a single document rather than through the online survey. Headings identify any specific consultation questions addressed in a section. Paragraph references are to the Consultation unless otherwise indicated.

The Consultation’s three chapters share an ambition to make redress, appeals, and enforcement ‘swifter and simpler’. ICLE supports that aim. Speed and simplicity help when they lower the cost of reaching sound outcomes. They can, however, do harm when they remove safeguards needed to reach those outcomes. We therefore assess each proposal through the error-cost framework that has long guided law & economics analysis of enforcement institutions: Does the reform reduce the expected cost of error, including false positives, false negatives, and enforcement costs?[3]

Our principal conclusions follow.

A. Opt-Out Collective Actions (Chapter 1)

The government’s diagnosis is broadly right, and most of its remedies point in the right direction. The regime has two linked problems. First, certification has become too easy, leaving funders’ appetite for returns to filter claims that the Competition Appeal Tribunal (‘CAT’) should assess for merit and proportionality. Seconds, even successful claims have returned very little to the consumers in whose name they are brought

ICLE supports codifying an absolute-suitability test with a weighted cost-benefit assessment (Q4–Q6), clarifying class representatives’ obligations through the CAT Rules (Q1–Q3), permitting damages-based agreements subject to disclosure (Q7–Q9), requiring budgets from both sides (Q13–Q14), and introducing settlement offers with automatic costs consequences (Q18–Q20). We oppose two proposals. Requiring the CAT to indicate at certification what funder return would be ‘reasonable’ (Q10) would import rate-of-return regulation into litigation finance. This risks blending into rate-of-return regulation of litigation finance with the associated cost inflation. It would be neither workable nor wise. Allowing the CMA to order redress based on a presumed overcharge (Q22–Q26) would turn an enforcement agency into an administrative damages tribunal without the experience, evidence, or standard of review that role requires.

B. Regulatory Appeals (Chapter 2)

ICLE supports transferring the CMA’s appellate functions to the CAT (Q44–Q45). Adjudication belongs in a judicial body, and the CMA’s resources are better spent on its primary functions.

That transfer should, however, preserve parties’ right to a meaningful appeal. Applying judicial-review principles ‘taking due account of the merits’ (Q47) would lower the standard that now governs energy appeals and replace water redeterminations with a materially more deferential process. Decisions that set the return on tens of billions of pounds of regulated investment warrant a genuine merits appeal.

C. Streamline Competition Act 1998 Enforcement (Chapter 3)

Several proposals would sensibly reorganise procedures and institutions. Others would remove statutory safeguards and replace them with CMA guidance. Most notably, omitting Rule 3 (Q65) would allow the official who issued a statement of objections to sit in judgment on it.

The proposals would also withdraw the right to make confidentiality representations (Q67), narrow access to the file (Q75, Q77, Q78), and codify appeal waivers in settlement (Q68). Together, these would weaken checks on an agency whose infringement decisions bind courts in follow-on damages actions. ICLE supports faster competition-law procedures, but notes that procedural fairness protects the parties, lends credibility to enforcement decisions, and makes the UK a more attractive place to do business.

II. The Opt-Out Collective Actions Regime

The opt-out regime began in 2015 with three objectives: better redress for victims of anticompetitive conduct, deterrence, and protection for defendants against unmeritorious claims.[4] A decade later, the evidence assembled in the Consultation suggests that the regime is falling short on redress and protection for defendants. It also pursues deterrence in a way that generates large private returns without commensurate public benefit.

The regime has grown rapidly since 2020. Claimants have filed cases with a cumulative estimated value exceeding £130 billion on behalf of some 655 million class members.[5] Claimant budgets routinely exceed £20 million, and defence budgets can be expected to match or exceed them.[6] Funders will not back a claim worth less than roughly £500 million.[7]

Yet in the only case to reach distribution, less than one per cent of the class made a claim. Approximately £216,000 reached consumers, while more than £10 million went to lawyers, funders, and other stakeholders.[8] In the only case to reach a judgment in the claimant’s favour, the Tribunal had to determine what a ‘fair’ commission for access to a digital platform would have been. As ICLE has explained elsewhere, this exercise in administered pricing is unlikely to produce stable or welfare-enhancing results. The judgment is now the subject of an application for permission to appeal.[9]

These outcomes reflect a well-understood divergence between the private and social incentives to litigate.[10] The lawyer or funder who selects and controls a claim captures much of its private return, while society bears the costs of adjudication and any over-deterrence. Research on ‘entrepreneurial litigation’ shows that when those who select claims are paid from the proceeds, they are likely to favour claims that promise returns to themselves over those that benefit the people they represent.[11] Opt-out proceedings sharpen that divergence because class members are dispersed, passive, and often unaware of the claim.

Two features of the UK regime make it especially susceptible to this dynamic. First, since Merricks, certification has been understood as a ‘relative suitability’ test that ‘does not involve a merits test’.[12] Second, because the regime covers only competition law, claimants have pleaded claims ‘not obviously rooted in competition’ as abuses of dominance. In some cases, the Tribunal has observed that the alleged conduct ‘seem[s] not to be competition law infringements at all’.[13]

Together, these features have made the regime increasingly attractive as an asset class for litigation finance rather than a route to redress: a portfolio of large standalone claims certified at a low threshold, whose value to investors depends on a handful settling or succeeding.

We agree with the government that the regime should be neither expanded nor further restricted at this stage.[14] The answer to the ‘dressing up’ problem is to enforce its existing boundaries. At certification, the Tribunal should require a pleaded competition claim with a real prospect of success, rather than accept a competition-law label alone. Cost-benefit analysis should also be undertaken; at the moment, only cost-effectiveness analysis is. This necessarily implies considering the market impact of the claim, including at the certification stage; this omission is the driver of claims which simply incur costs without improving markets (and sometimes harming them). The proposals discussed below would do much to achieve this.

A. Certification (Q4–Q6)

ICLE supports all three proposals in paragraph 29. Together, they would restore the ‘strong process of judicial certification, including a preliminary merits test’ that the government promised when it created the regime and that subsequent case law has eroded.[15]

Intrinsic (absolute) suitability. A relative-suitability test asks whether collective proceedings are preferable to individual proceedings. Because individual claims for small losses are rarely viable, the answer is usually clear, and certification does little to filter claims. An intrinsic test asks whether the claims are fit to be grouped and determined collectively under the Act. It also asks whether collective proceedings offer a reasonable prospect of a just outcome at proportionate cost.[16] Where claimants seek aggregate damages, they must offer a methodology with a realistic prospect of establishing class-wide loss and evidence that the necessary data exist.[17]

Some mature class-action jurisdictions take a similar approach. In the United States, Rule 23 requires a ‘rigorous analysis’ that will ‘frequently … entail some overlap with the merits’.[18] The class plaintiff must also show that damages can be measured on a class-wide basis consistent with its theory of liability.[19]

Greater weight for costs and benefits, including expected take-up. This is, in our view, the Consultation’s most important reform. The error-cost logic is straightforward. Certification imposes immediate and certain costs: defence spending that can match a £20 million claimant budget, years of management distraction, and, critically, settlement pressure largely independent of a claim’s merit.[20] The benefit is uncertain and, on the evidence so far, small: only a fraction of a damages award reaches a fraction of the class.

A test that gives this comparison real weight and counts likely take-up among the benefits would screen out claims most likely to enrich stakeholders while compensating few class members. It would also encourage funders and law firms to design claims and distribution plans around the class rather than headline quantum.

To make the test operational, the CAT Rules should require the proposed class representative (‘PCR’) to file a costed distribution plan and a reasoned estimate of take-up at certification. A projected return to the class disproportionate to projected costs should be a presumptive ground for refusal. The Tribunal’s recent refusal to certify Waterside v Mowi because litigation costs far exceeded the likely benefit and distribution return to consumers shows that the existing Rules permit this reasoning. Codification would make it the norm rather than the exception.[21] Importantly, this should be cost-benefit analysis and not just cost-effectiveness analysis: there should be an assessment of whether the lawsuit will improve market performance, and not just its costs, or this analysis is unduly truncated.

Sufficient evidence for aggregate damages. Aggregate awards make opt-out proceedings viable, but they also make them risky. The Tribunal should distinguish injured from uninjured class members and tie awards to a sound theory of loss so that they reflect the harm class members actually suffered.[22] Rule 79(2)(f) already requires the Tribunal to consider whether claims are suitable for an aggregate award. Under the current approach, though, suitability is only one factor in a broader assessment of ‘all matters it thinks fit’.[23] Suitability for aggregate damages should instead be a condition of certification, supported by credible evidence grounded in the facts of the case.

The government should also address the ‘multiple bites of the cherry’ problem noted in paragraph 61. Repeatedly filing reformulated claims after a refusal imposes costs on defendants and the Tribunal. The Rules should require permission to refile and allow the Tribunal to make payment of the earlier application’s costs a condition of that permission.

The concern that a statutory merits element would generate satellite litigation (¶ 27) is overstated. Parties already contest certification at length. In Evans v Barclays, the Supreme Court held that the weakness of a claim was properly treated as a powerful factor against opt-out certification under Rule 79(3)(a), without requiring a separate merits trial.[24]

B. Class Representatives (Q1–Q3)

ICLE agrees that greater certainty would help in all three areas identified in paragraph 19: funding arrangements, costs specialists, and consultative panels.

The class representative is the only participant in the proceedings whose incentives are meant to align with the class. Yet a funder or law firm typically recruits the representative, who has no financial stake in the outcome and relies on the funder to meet an adverse costs order.[25]

Gutmann v First MTR South Western Trains illustrates the problem. The class representative, a private individual, could not have met costs beyond the £15 million indemnity his funder provided, so the defendants pursued the funder directly.[26] The funder had advanced about £6.58 million of the representative’s own costs—less than half the adverse costs cover it had provided. The solicitors absorbed later budget increases through deferred and contingent fees. When the defendants sought costs above the cap, the funder argued that the lead solicitors should bear the excess. The solicitors, in turn, wrote to the Tribunal to rebut the funder’s account. The representative sat between two parties financing the litigation, each with a contingent interest, without an independent adviser on the dispute. The agency problem identified in the class-action literature is acute.[27]

The Tribunal’s decisions in Riefa and Hammond ask the right questions: Does PCR understand and genuinely negotiate the funding terms? Has the PCR obtained independent costs advice? Is there a way to make informed decisions on settlement?[28]

The CAT Rules should answer those questions, with guidance on what is proportionate for a given claim. Rule 78 should require the PCR to (i) disclose the litigation funding agreement (‘LFA’) to the Tribunal, including provisions on funder control, termination, and payment priority; (ii) confirm that the PCR obtained independent advice on the funding terms; and (iii) explain how the representative will decide questions of settlement and distribution, including through any consultative panel.[29] This would address the concern in paragraph 17 that PCRs might overspend to cover every base. The requirement is informed independence, not a prescribed governance structure.

C. Financing the Claim (Q7–Q14)

Damages-based agreements (Q7–Q9). ICLE cautiously supports lifting the prohibition on damages-based agreements (‘DBAs’) in opt-out proceedings. The economic case rests on competition. When third-party funders paid a multiple of their investment are the only source of finance, they gain substantial pricing power over the claims the regime exists to enable.[30] Allowing lawyers to take a share of recoveries would introduce a competing source of capital and—as the experience in Victoria, Australia suggests—may lower funding costs and improve returns to the class.[31] It would also align lawyers’ incentives more closely with the size of the recovery than hourly billing does.

The government identifies credible risks, but certification safeguards can address them. We recommend three: full disclosure of the DBA to the CAT through the Rule 78 process; approval of the percentage at settlement or judgment based on the sum actually distributed, rather than the headline award (see Section II.D); and CAT approval of ‘stacked’ arrangements in which a funder and law firm each take a return from the same recovery, so that the CAT can assess their combined cost to the class.

Funder returns: rate-of-return analysis of damages (Q10–Q12). The proposal in paragraph 52 responds to a genuine problem. Funders do not know what return the CAT will approve at the end of a claim, and that uncertainty raises the cost of capital.[32] But asking the CAT at certification to indicate a reasonable return and payment order would, in substance, regulate the rate of return on litigation finance, albeit reading on damages. It would bring the well-documented problems of that model into the CAT.

Rate-of-return regulation asks a decision-maker to determine in advance what return on invested capital is ‘reasonable’. Decades of utility regulation reveal three recurring problems.[33] First, the decision-maker cannot obtain all the information needed to set the rate correctly. The appropriate return on a funded claim depends on the funder’s cost of capital, how the claim’s risk relates to the rest of its portfolio, and the likelihood of success. The CAT cannot observe those factors at certification.[34]

Second, fixing a return as a multiple of cost, or damages, gives the regulated party an incentive to inflate its cost base or the damages estimate (e.g., by not settling). In litigation, an ‘approved’ multiple on invested capital or damages would reward spending more rather than litigating better. The Consultation’s evidence that claimant budgets have doubled within months of certification (¶ 60) suggests that this dynamic is already present. Third, an early indication would become an anchor for later decisions. The promised discretion to ‘revisit’ the assessment at the end of the claim (¶ 52) could prove illusory in practice.

A better response has three parts. First, funders should obtain certainty from their agreements and efficient operations so that they are competitive providers of finance, rather than advance CAT approval. If the PCR has disclosed the LFA, negotiated it with independent advice, and shown at certification that it preserves the PCR’s independence, the Rules should provide that the CAT will enforce the agreed return absent fraud, undue influence, or a material change in the claim. Second, DBAs and the transparency measures proposed above would discipline returns through competition. Third, the Rules should address returns that rise with the passage of time and thus reward delay past the next step-up date. They could bar recovery from the class of any increase triggered after a reasonable settlement offer of the kind discussed in Section II.D. That would target the incentive to delay without regulating the rate. Indeed, a regulated return on damages would likely discourage settlement and slow down lawsuits (e.g. by going to trial to extract more damages than settlement would give).

ICLE sees the force of the time-value argument for paying funders when the CAT makes an award or approves a settlement (¶ 53; Q11–Q12), but the proposed presumption raises a concern. If the funder receives its entire return from the gross sum before distribution, no participant retains a financial interest in whether the class receives anything. Gutmann shows how little may otherwise reach class members.[35]

We would support a presumption that the funder recovers its capital and costs on award or approval, while the CAT ties payment of the profit element to distribution. The funder could draw any remaining profit from undistributed funds only after the CAT finds that reasonable distribution efforts were made. This would preserve an incentive to design and carry out distribution effectively. The CAT’s residual discretion (Q12) is welcome but should not be the class’s only protection. Indeed, the major problem has been precisely that the claimant class is already too far back in the queue; further demoting their claims is not the answer here.

Budgets and costs conduct (Q13–Q14). ICLE supports requiring both parties to submit budgets after certification and to notify the CAT of material changes. Transparency on both sides reduces surprises when costs are assessed and helps the CAT exercise its powers under Rule 104.[36]

The Rules should not presume that vigorous defence of a claim worth hundreds of millions of pounds is a ‘scorched earth’ tactic (¶ 61). Costs sanctions should apply equally to both sides, turn on objective conduct—such as applications that fail comprehensively, repeated failure to follow directions, or refiling a refused certification application—and follow an assessment of the record. The cost-benefit test at certification offers the strongest discipline on claimant budgets. A realistic prospect that only meritorious claims will be certified offers the strongest discipline on defendant budgets.

D. ADR, Settlement, and Distribution (Q15–Q21, Q34–Q37)

Mediation (Q15–Q17). ICLE supports allowing the CAT to order mediation after the parties exchange expert reports and to account for an unreasonable refusal when awarding costs. This would align with the approach now taken under the Civil Procedure Rules.[37] At that point, both sides have enough information to value the claim. Mandatory ADR before filing would add costs without resolving disputes, as the government recognises (¶ 71).

Settlement offers with automatic costs consequences (Q18–Q20). ICLE supports extending a Rule 45-type mechanism to collective proceedings. The current asymmetry is striking. A funded PCR faces little downside from refusing a reasonable defendant offer. A defendant that refuses a Calderbank offer faces, at most, a discretionary costs sanction that, as Kent v Apple shows, may never be imposed.[38] Automatic consequences for both sides would create stronger incentives to settle.

Existing CAT approval under Rule 94 addresses the concern that such a mechanism might put undue pressure on PCRs (Q19). The CAT must find a settlement just and reasonable before approving it, so an offer cannot compel acceptance of a bad settlement.[39] The mechanism should otherwise follow Rule 45, with enough time for acceptance to accommodate CAT approval. The approval process itself (Q21) should develop through experience, with a review once the CAT has considered more settlements.

Distribution and undistributed sums (Q34–Q37). Listing certified claims on the CAT website is sensible and inexpensive. But when fees and returns depend on the gross award, no participant has a financial interest in getting money to the class. Evidence from the mature US regime shows that claims rates in consumer class actions are typically in single digits.[40]

The reforms proposed above would create that interest: counting expected take-up in the certification cost-benefit test, requiring a costed distribution plan, and tying payment of funder profit to distribution. Parties could also agree that undistributed sums revert to the defendant, an option the government rightly leaves to them (¶ 129).

Destination of undistributed damages (Q37). Large undistributed sums may show that the parties overstated the claim’s benefit to the class at certification. The CAT should be able to consider that record in later certification decisions involving the same representatives, funders, or firms.

ICLE also strongly advises against designating the Consumers’ Association (Which?), an active class representative and litigant before the CAT.[41] Directing undistributed damages to an organisation that brings collective proceedings creates a perception—and a real risk—of circularity: more certified claims and less distribution would mean more money for that organisation. US experience with cy pres awards to organisations connected to litigants has drawn scrutiny from the Supreme Court.[42] By contrast, the Access to Justice Foundation may have little connection to competition law, but it has no stake in the regime.

E. CMA-Ordered Redress and Presumed Overcharge (Q22–Q27)

ICLE supports simplifying the Redress Scheme Regulations (Q23),[43] allowing the CMA to accept binding redress undertakings when a business volunteers them in a settlement or commitments decision, and making clear that payments under a redress scheme reduce any subsequent damages claim and count at certification (Q24, Q27). These measures would make voluntary redress easier without displacing the adjudication of loss.

We oppose the two proposals that go further: a statutory presumption of overcharge and a power for the CMA to direct a business to establish a redress scheme (Q22, Q25–Q26).

Presumed overcharge. Damages compensate loss actually suffered. The Organisation for Economic Co-operation and Development (OECD) impact-assessment methodology and meta-studies of historical cartels discussed in paragraph 92 can help estimate the aggregate benefits of enforcement across many cases. The issue is to identify the true overcharge, because correcting the overcharge is how the economic harm of principled cases is addressed. The underlying studies report overcharges ranging from zero to well above 50 per cent, so any single default figure will systematically overcompensate some purchasers and undercompensate others.[44] A presumed overcharge will therefore not achieve optimal deterrence.

The Consultation would also apply the presumption to abuse-of-dominance cases, where ‘overcharge’ is often a poor fit because the alleged harm concerns exclusion rather than price. And because the same agency would find the infringement, impose a penalty, and set the presumed sum, the presumption could operate in practice as a second penalty under another name. It would increase deterrence without improving compensation.[45]

Mandatory redress. The government rejected this power in 2013 for reasons that remain relevant: assessing loss lies outside the CMA’s expertise and would divert resources from investigation and enforcement.[46] The CMA’s role has since expanded considerably, while Chapter 3 of the Consultation proposes to reduce safeguards around its infringement decisions (see Section IV below).

Giving the CMA power to order compensation, subject to a standard of review the Consultation does not specify, would place investigation, prosecution, adjudication, penalties, and damages assessment in one administrative body. ICLE warned against that concentration of functions in its response to the January consultation.[47] It would also sit uneasily with the principle that redress should follow from the merits of a claim rather than regulatory direction. If the government wants more redress in follow-on cases, it should make the collective regime fit for purpose through the certification reforms discussed above

F. Leniency and Public Enforcement (Q28–Q33)

Leniency is the most effective cartel-detection tool available to competition authorities. The decline in Type A applications from an average of 21 a year in 2014–16 to eight a year in 2022–24 is a serious warning sign.[48] The economics are well understood: a firm will report its own conduct only if the expected benefit of immunity exceeds the expected cost of exposure. As private damages exposure has grown, the calculation has shifted against reporting.[49] Immunity from penalties offers little to a firm whose main exposure is civil liability.

ICLE therefore supports extending immunity from civil liability to Type A applicants in principle (Q28), subject to two safeguards (Q29–Q30). First, the government should retain the set-aside for unrecoverable damages proposed in paragraph 108 and define it using the Damages Directive’s model. Under that model, the immunity recipient remains liable when victims cannot obtain full compensation from the other infringers.[50]

Second, immunity should depend on continued cooperation with claimants as well as the CMA. The US model, for example, conditions relief from treble and joint-and-several liability on ‘satisfactory cooperation’ with civil plaintiffs.[51] That would make the immunity recipient a source of evidence for claims against the remaining cartel members and help address the concern about compensation. The objection that the recipient ‘escapes’ liability has less force when its report enables the cartel’s detection and makes redress possible (¶ 103).

III. Simplifying the Regulatory Appeals Framework

ICLE supports transferring the CMA’s appellate functions to the CAT and bringing telecoms price-control disputes within its jurisdiction. Consolidation would give these disputes a judicial forum and reduce duplication. The CAT will need the capacity and sectoral expertise to take on the work. The transfer should also preserve meaningful scrutiny of decisions that shape billions of pounds in infrastructure investment. A genuine merits standard, with power to substitute a decision where the correct answer is clear, would allow the CAT to correct errors without unnecessary remittals.

A.  Transfer to the Competition Appeal Tribunal (Q44–Q46)

ICLE supports transferring the CMA’s appellate functions to the CAT, converting the remaining reference functions into appeals, and extending the CAT’s jurisdiction to price-control matters in telecoms appeals.

The current allocation assumes that sectoral regulation is best reviewed by a body with deep experience assessing pricing decisions across sectors. The Competition Commission, and later the CMA, built that expertise through decades of price-control appeals and references. But deciding a dispute between a regulator and a regulated firm is a judicial function. The Consultation itself accepts that a judicial body with the right technical expertise ‘would be equally capable of dealing with these issues’ (¶ 166).[52]

The CAT can acquire that sectoral depth by selecting and, where necessary, supplementing its Ordinary Members (Q46). It would thereby gain independence without sacrificing technical competence. Transfer would also free CMA resources: a single large regulatory appeal can absorb the equivalent of two or three market studies (¶ 167). The CAT already hears appeals from the Office of Communications (Ofcom), the Office of Gas and Electricity Markets (Ofgem), the Payment Systems Regulator (PSR), and the Civil Aviation Authority (CAA). Consolidation would improve consistency and help precedent develop across sectors, as the Penrose report recommended five years ago.[53]

Extending the CAT’s jurisdiction to price-control matters in telecoms appeals (Q45) follows the same logic. The CAT currently hears a telecoms appeal but must refer any price-control element to the CMA, then incorporate the CMA’s conclusions into its judgment (¶ 174).[54] Splitting one appeal between a judicial body and an administrative agency adds procedure, lengthens timetables, and divides the reasoning between institutions with different processes. If the CAT can determine price-control appeals in water and energy, as the Consultation proposes, it can do so in telecoms.

The proposed transfer also bears on the CMA’s independent panel. In March, ICLE argued that the panel should not be abolished without introducing merits-based appellate review at the CAT. The panel provided an independent, quasi-judicial layer of decision-making that helped justify deferential judicial review.[55] The Consultation now cites the panel’s impending abolition as a reason to move regulatory appeals out of the CMA (¶ 169). While that is sound, abolishing the panel does not eliminate the need for the functions it performs. The CAT is the right place for regulatory appeals. For merger and market decisions, the same logic points to merits review at the CAT, which the government has yet to propose.

The government should assess the CAT’s staffing needs before the transfer takes effect. Comparing the expected caseload with the pool of people who have the necessary sectoral and financial expertise, and who are eligible and willing to serve as Ordinary Members, would let the CAT build capacity in advance. Without that assessment, the reform could improve the forum while lengthening the queue—undermining the timeliness gains it seeks.

B. The Standard of Review (Q47–Q52)

The Consultation proposes that the CAT decide all regulatory appeals by applying ‘the same principles as would be applied by a court on an application for judicial review, while also taking due account of the merits of the case’. This follows the telecoms model in section 194A of the Communications Act 2003 (¶ 204(a)).[56] For most appeals within scope, it would reduce the intensity of review. Energy licence-modification appeals currently succeed when a decision is wrong on specified grounds, including errors of fact and failure to give appropriate weight to relevant matters. The CMA may substitute its own decision in price-control appeals, while water redeterminations involve a fresh decision on the merits.[57] The proposal would replace those approaches with a review in which the CAT ‘takes [the regulator’s] decision as a starting point’, accords it ‘an appropriate degree of respect’, and asks whether it is ‘materially wrong’ (¶ 202).

The government may argue that the telecoms formula also asks whether a decision is ‘wrong’ (¶ 201), but two differences matter. First, under the existing energy, water, and air regimes, an error of fact, a failure to give appropriate weight to relevant matters, or—in the air regimes—an error in the exercise of discretion can independently justify allowing an appeal. Under section 194A, the CAT considers the merits through the judicial-review grounds of illegality, irrationality, and procedural unfairness. A decision can therefore be wrong without meeting the threshold for intervention.

Second, treating the regulator’s decision as a starting point gives it a degree of deference. An appeal on specified grounds instead requires the reviewing body to form its own view of whether the decision was wrong, giving the regulator’s judgement only the weight the evidence warrants.

Research on regulatory disputes supports that distinction. Despoina Mantzari argues that deference to a regulator’s economic assessment should depend on the relative institutional competence of the regulator and the reviewing body, rather than on the regulator’s expertise alone.[58] The CAT’s own economic and technical expertise supports meaningful scrutiny, even when the regulator’s specialised judgement deserves weight.

ICLE’s concern is with the error costs of a deferential standard. Price controls set the allowed return on tens of billions of pounds of infrastructure investment, and regulatory error runs in both directions. If a regulator sets the cost of capital too low and an appeal cannot correct it, consumers may bear the cost of underinvestment for decades. If it sets the allowed return too high, consumers pay excess returns for a full control period. The Consultation itself acknowledges that the current arrangements ‘can … deter investments in UK markets’ (¶ 199).

From a law & economics perspective, the question is how the review standard allocates the risk of error among regulators, regulated firms, investors, and consumers. A merits standard, combined with the proposed standing for consumer bodies that ICLE supports, would allow errors in either direction to be corrected.

The Consultation’s principal argument for deference is that many elements of a price control, ‘e.g. the cost of capital’, are ‘complex matters requiring the exercise of regulatory judgement with no single right answer’ (¶ 199). That may be true, but a merits appeal can still give the regulator appropriate latitude within a range of defensible values. In the RIIO-2 appeals, the CMA upheld the networks’ challenges to the ‘outperformance wedge’ and, for five of the eight appellants, an innovation uplift unsupported by evidence. It nonetheless dismissed every challenge to the cost of equity.[59]

Fresh redetermination has not simply favoured regulated firms, either. In the PR24 redeterminations, the CMA raised the allowed return from 4.03 to 4.20 per cent but granted the five disputing companies £463 million in additional revenue—17 per cent of the £2.7 billion they sought.[60] A judicial-review standard would do more than respect regulatory judgement: it would leave wrong decisions standing when they fall short of irrationality. The lack of a single right answer strengthens the case for external review because a regulator may struggle to identify its own errors.

The standard also matters to the government’s growth agenda. Water and energy networks must raise substantial private capital over the coming decade, and investors account for the available appellate review when pricing risk. If an erroneous determination of the allowed return can be corrected only when it is ‘materially wrong’ under judicial-review principles, investors may demand a higher risk premium, with consequences for allowed returns and investment.

A merits appeal need not entail unrestricted de novo review. It asks whether the regulator’s decision has evidential support and accords with its statutory duties. It does not require the CAT to take over the regulator’s continuing administrative role. The weight due to a technical assessment will depend on the issue under review.

ICLE therefore recommends a genuine merits appeal set out in statute. A single list of grounds drawn from the existing energy, water, and air regimes should cover error of fact; error of law; failure to consider, or give appropriate weight to, relevant matters; failure to follow the applicable procedure; failure of a modification to achieve its stated effect; and error in the exercise of discretion. The CAT should decide whether the regulator was wrong on those grounds, giving its assessment the weight the evidence warrants.[61] This would largely extend the existing energy standard across sectors. The Consultation offers no evidence that the standard has produced excessive appeals or unmanageable proceedings.

Requiring a notice of appeal to identify whether each ground alleges an error of fact, law, or discretion (Q48) would help focus proceedings, provided a mistaken label does not bar a sound argument. ICLE also agrees with the approach to new evidence (Q49) and cross-examination (Q50) in paragraph 204(c): the CAT should retain discretion under its existing Rules to allow both. Restricting appeals to material available to the regulator could reward opaque reasoning, since an appellant may not know what evidence informed the decision until its publication. The principal barriers the proposals would create (Q52) are thus the review standard and the proposed rule on remedies.

On remedies (Q51), mandatory remittal should not be the default. When the CAT finds an error and the correct answer follows from that finding—for example, a discrete input to a price control—remittal would delay the outcome and invite another appeal. The CAT should be able to substitute its own decision on that point, as the CMA can in energy price-control appeals and as the appeal bodies can in the air and postal regimes (¶ 195).[62] It should remit cases when the regulator needs to exercise fresh judgement in light of the CAT’s findings. Confined in this way, substitution corrects an error without beginning a second round of regulation.

IV. Competition Act Enforcement and Procedural Fairness

ICLE supports faster enforcement, but the CMA’s decisions must remain reliable when they bind courts in follow-on damages claims. The Rules should keep final decision-makers separate from the team that investigated a case, preserve a meaningful chance to protect confidential information and inspect the evidence, and maintain safeguards when the CMA imposes interim measures. Settlement procedures can be simpler, provided they give parties adequate access to the case against them and clearly define the scope of their admissions.

A. The CMA’s Decision-Making Model (Q65–Q66)

Rule 3 of the CMA Rules embodies a simple principle: the officials who decide to issue a statement of objections (‘SO’) should not decide whether those objections are well founded.[63] The Consultation proposes removing Rule 3 and letting the CMA set its decision-making model in guidance. One option is an ‘SRO + 2’ model, in which the senior responsible officer (‘SRO’) who issued the SO would make the final decision alongside two colleagues who had not worked on the case (¶ 229).

Basic procedural fairness weighs against this change. Separating investigative and decision-making functions within an enforcement agency is an important safeguard against confirmation bias: the tendency to interpret new evidence as supporting a theory one has already formed.[64] The concern is not misconduct by CMA officials. Any decision-maker can be influenced by a view formed earlier in a case, and institutional rules should account for that risk.

Critics have long challenged the European Commission’s combination of these functions. The European Court of Human Rights has accepted such an arrangement in an administrative competition authority where a court can fully review its decisions on the facts and the law.[65] The UK’s case-decision-group model was a deliberate improvement on the Commission’s approach.

The consumer-protection regime offers no reason to remove Rule 3. It also separates functions, but the CMA sets that safeguard itself and can revise it. The difference concerns where the rule sits—in CMA guidance or in a statutory instrument. It does not diminish the need for separation. Nor does the SRO’s familiarity with the case outweigh the risk of confirmation bias. Two uninvolved colleagues may give undue weight to the views of the officer who led it.

Three considerations make separation especially important. First, infringement decisions under the Competition Act 1998 bind courts in follow-on damages actions,[66] which now include opt-out collective claims worth hundreds of millions of pounds. The reliability of the underlying decision matters more than it did in 2014. Second, although the CAT hears appeals on the merits, [67] an appeal is costly and slow, and a party that settles waives it. For a meaningful share of decisions, internal separation is the only check. Third, moving the safeguard from the Rules to guidance would let the CMA change it after a consultation it controls, without parliamentary scrutiny.

If the government considers Rule 3 too prescriptive, it should retain the core requirement in the Rules: people who took part in the investigation or decided to issue the SO should not make the final infringement or penalty decision. The CMA could set the decision-makers’ composition and procedures in guidance. The same principle should govern settlements (Q66). A party’s admission does not relieve the CMA of its duty to establish an infringement, and someone outside the case team that negotiated the settlement should make that judgement.

B. Confidentiality, Access to the File, and Interim Measures (Q67, Q71–Q78)

Rule 7 (Q67). Rule 7 should be amended, not omitted. The Consultation may be right that the CMA must process too many blanket confidentiality claims that offer no reasons. It should therefore require parties to identify and explain each claim, and allow the CMA to disclose information without further representations when a claim is unreasoned or manifestly unfounded. Parties should retain the right to make representations on substantiated claims. Competition Act 1998 investigations can uncover trade secrets whose disclosure is irreversible. Even well-resourced authorities have failed to protect such information.[68] A chance to be heard before disclosure is a minimal safeguard.

Confidentiality rings (Q71–Q74). ICLE supports a standardised framework that gives economic as well as legal advisers access to confidentiality rings, so parties need not negotiate the arrangements afresh in every case. This is already the case following Practice Direction 2/2026 and the framework there, which amounts to a rebuttable presumption, seems apt. Predictability would benefit both the CMA and the parties it investigates. Civil sanctions for breaches (Q72) are acceptable in principle if they require fault, are proportionate to the breach, and provide a safe harbour for prompt self-reporting. Employer liability should apply to systemic failures, not automatically to every breach (Q73). Extending the framework to mergers and markets (Q74) is sensible if the same safeguards apply.

Access to the file (Q75–Q77). Aligning the Rule 6(2) exceptions with the consumer-enforcement rules would let the CMA withhold ‘routine correspondence’ and documents ‘to the extent that they identify an individual’ (¶ 253).[69] Neither category defines its own limits. In an abuse-of-dominance investigation, correspondence the CMA regards as routine may matter to the defence. Any new exception should be narrow and require a list of withheld documents, so a party can request specific items.

Disapplying Rule 6 altogether in settlement cases (Q77) is more troubling. A settling party admits an infringement that binds courts in follow-on claims and gives up its right of appeal. It should understand the full case file before doing so, rather than see only the documents the CMA chooses to cite. The CMA could offer streamlined access if the party agrees to it. The Rules should not impose it as the default.

Interim measures (Q78). ICLE opposes removing Rules 13(1) and 13(2). Interim measures are the CMA’s most intrusive tool short of a final decision. They impose binding obligations on the basis of a prima facie case, before any finding of infringement.[70] Removing the right to inspect the file could require a business to change its conduct without seeing the evidence said to justify the direction. A targeted rule could address the efficiency concern: give the business access to documents on which the CMA relies, through a confidentiality ring where needed, within a short, fixed period. The safeguard should remain in the Rules rather than move to guidance.

C. Settlement Procedure (Q68–Q70)

Settlement benefits both sides. It saves CMA resources and gives the settling party a penalty reduction and an earlier end to the investigation. But settlement decisions also affect third parties: they establish binding infringement findings on which follow-on collective claims may rest.

Together, the proposals would allow an infringement decision without an SO (Q69–Q70), full access to the file (Q77), or a right of appeal (Q68). Each proposal reflects existing practice in isolation.[71] In combination, they could produce a decision that no one outside the case team has tested against the full evidence, yet that carries the weight of a binding infringement finding in litigation worth many times the penalty.

ICLE does not object to codifying the agreement not to appeal. The Rules should, however, also make clear that the settling party admits only the facts and infringement set out in the settlement decision, and that follow-on claimants cannot extend that admission. The decision should state that it resulted from a settlement.

When all parties settle before an SO, a summary statement of facts may serve as the section 31 notice if it sets out the CMA’s case with the specificity that section 31 requires and the parties have seen the documents on which it relies. The consumer-enforcement rules should not set the benchmark for this process. The procedural rules should instead reflect the difference between conduct whose illegality is usually clear and conduct whose illegality turns on contested economic analysis.

V. Conclusion

The Consultation is right to seek swifter, simpler competition redress, regulatory appeals, and enforcement. Several proposals would also improve accuracy: certification that tests claims’ merit and proportionality, settlement offers with meaningful consequences, a judicial forum for regulatory appeals, and standardised confidentiality arrangements. ICLE supports these proposals.

Other proposals would sacrifice accuracy for speed. Regulating the rate of damages, allowing the CMA to order redress based on a presumed overcharge, weakening review of regulatory decisions worth tens of billions of pounds, and removing the separation between investigation and decision-making in Competition Act 1998 cases would concentrate discretion while weakening checks on it. These proposals draw on the DMCC Act’s consumer-enforcement provisions, which address a narrower set of issues and remain largely untested. Their effects should be understood before they become a model for more complex and contested cases.

The reforms most likely to meet the government’s aims will align private incentives with public purposes: claims brought on their merits, funders rewarded when consumers receive compensation, and binding enforcement decisions reached through fair procedures.

[1] Dep’t for Bus. & Trade, Swifter and Simpler Competition Redress, Regulatory Appeals and Competition Enforcement (17 July 2026) [hereinafter Consultation].

[2] Geoffrey A. Manne et al., Refining Our Competition Regime: ICLE Response to the Department for Business and Trade Consultation Published 20 January 2026 (Int’l Ctr. for L. & Econ., 31 March 2026) [hereinafter ICLE Refining Comments]; Dirk Auer, Selcukhan Ünekbas & Mario A. Zúñiga, Comments of the International Center for Law & Economics: UK Competition and Markets Authority Call for Evidence for Merger Efficiencies Review (Int’l Ctr. for L. & Econ., 25 February 2026); Geoffrey A. Manne et al., Comments of the International Center for Law & Economics: The CMA’s Proposed Steering Conduct Requirements for Apple and Google (Int’l Ctr. for L. & Econ., 28 July 2026) [hereinafter ICLE Steering Comments].

[3] See Frank H. Easterbrook, The Limits of Antitrust, 63 Tex. L. Rev. 1 (1984); Geoffrey A. Manne & Joshua D. Wright, Innovation and the Limits of Antitrust, 6 J. Competition L. & Econ. 153 (2010).

[4] Consultation, supra note 1, ¶ 1.

[5] Id. ¶ 38 (citing CMS, European Class Action Report 2025, at 14, 36 (2025)).

[6] Consultation, supra note 1, ¶ 34.

[7] Id. ¶ 38.

[8] Id. ¶¶ 29, 116 (discussing Gutmann v First MTR South Western Trains Ltd [2025] CAT 72).

[9] Dr Rachael Kent v Apple Inc [2025] CAT 67 (application for permission to appeal pending); see ICLE Steering Comments, supra note 2, at 11 (discussing the Tribunal’s difficulty in determining a fee justified by the value provided).

[10] Steven Shavell, The Fundamental Divergence Between the Private and the Social Motive to Use the Legal System, 26 J. Legal Stud. 575 (1997); see also William M. Landes & Richard A. Posner, The Private Enforcement of Law, 4 J. Legal Stud. 1 (1975).

[11] John C. Coffee, Jr., Understanding the Plaintiff’s Attorney: The Implications of Economic Theory for Private Enforcement of Law Through Class and Derivative Actions, 86 Colum. L. Rev. 669 (1986); Jonathan R. Macey & Geoffrey P. Miller, The Plaintiffs’ Attorney’s Role in Class Action and Derivative Litigation: Economic Analysis and Recommendations for Reform, 58 U. Chi. L. Rev. 1 (1991); John C. Coffee, Jr., Entrepreneurial Litigation: Its Rise, Fall, and Future (2015).

[12] Merricks v Mastercard Inc [2020] UKSC 51, ¶¶ 59, 70, 74.

[13] Consultation, supra note 1, ¶ 7 (quoting Gormsen v Meta Platforms, Inc [2023] CAT 10, ¶ 30(1)); see also Gutmann v First MTR South Western Trains Ltd [2025] CAT 79 (holding that ‘[c]ompetition law is not a general law of consumer protection’).

[14] Consultation, supra note 1, ¶ 10.

[15] Dep’t for Bus., Innovation & Skills, Private Actions in Competition Law: A Consultation on Options for Reform—Government Response ¶ 5.55 (January 2013) [hereinafter 2013 Government Response]; Consultation, supra note 1, ¶ 21.

[16] Merricks v Mastercard Inc [2020] UKSC 51, ¶¶ 117–119.

[17] Id. ¶¶ 120–123.

[18] Wal-Mart Stores, Inc. v. Dukes, 564 U.S. 338, 350–51 (2011).

[19] Comcast Corp. v. Behrend, 569 U.S. 27, 34–35 (2013); see also In re Hydrogen Peroxide Antitrust Litig., 552 F.3d 305, 307 (3d Cir. 2008) (requiring courts to resolve factual disputes relevant to certification, even when they overlap with the merits).

[20] See In re Rhone-Poulenc Rorer Inc., 51 F.3d 1293, 1298–1300 (7th Cir. 1995) (Posner, J.) (describing the ‘intense pressure to settle’ that certification imposes even on defendants with strong defences); Merricks v Mastercard Inc [2020] UKSC 51, ¶¶ 118, 154, https://supremecourt.uk/uploads/uksc_2019_0118_judgment_6dd114686b.pdf (Lord Sales & Lord Leggatt, separate judgment) (noting that collective actions ‘confer substantial legal advantages on claimants and burdens on defendants which are capable of being exploited opportunistically’).

[21] Waterside Class Ltd v Mowi ASA [2026] CAT 32; Consultation, supra note 1, ¶ 29 & n.15; see also Competition Appeal Tribunal Rules 2015, SI 2015/1648, r. 79(1)–(2) (UK), https://www.legislation.gov.uk/uksi/2015/1648/article/79.

[22] See Comcast Corp. v. Behrend, 569 U.S. 27, 35–38 (2013).

[23] See Merricks v Mastercard Inc [2020] UKSC 51, ¶ 61 (‘The listed factors are not separate suitability hurdles, each of which the applicant for a CPO must surmount.’), https://supremecourt.uk/uploads/uksc_2019_0118_judgment_6dd114686b.pdf.

[24] Evans v Barclays [2025] UKSC 48.

[25] See Gutmann v First MTR South Western Trains Ltd [2025] CAT 21.

[26] Id.

[27] See Coffee, Understanding the Plaintiff’s Attorney, supra note 11; Macey & Miller, The Plaintiffs’ Attorney’s Role, supra note 11.

[28] Riefa v Apple Inc [2025] CAT 5, ¶¶ 89(5), 95, 114; Hammond v Amazon.com, Inc. [2025] CAT 42, ¶¶ 56, 67(1), 156(1).

[29] Competition Appeal Tribunal Rules 2015, SI 2015/1648, r. 78 (UK).

[30] See Jonathan T. Molot, Litigation Finance: A Market Solution to a Procedural Problem, 99 Geo. L.J. 65 (2010).

[31] Consultation, supra note 1, ¶ 42 (citing evidence from Victoria, Australia, after the introduction of group costs orders under s. 33ZDA of the Supreme Court Act 1986 (Vic.)).

[32] Consultation, supra note 1, ¶¶ 46–51 (discussing Merricks v Mastercard Inc [2025] CAT 28, ¶ 188; Gutmann v Apple Inc [2024] CAT 18, ¶ 38; and Gutmann v Apple Inc [2025] EWCA Civ 459, ¶ 78).

[33] See 1 Alfred E. Kahn, The Economics of Regulation: Principles and Institutions ch. 2 (1988); Harvey Averch & Leland L. Johnson, Behavior of the Firm Under Regulatory Constraint, 52 Am. Econ. Rev. 1052 (1962); Jean-Jacques Laffont & Jean Tirole, A Theory of Incentives in Procurement and Regulation (1993).

[34] See Friedrich A. Hayek, The Use of Knowledge in Society, 35 Am. Econ. Rev. 519 (1945).

[35] See Gutmann v First MTR South Western Trains Ltd [2025] CAT 21.

[36] Competition Appeal Tribunal Rules 2015, SI 2015/1648, r. 104 (UK).

[37] See Churchill v Merthyr Tydfil County Borough Council [2023] EWCA Civ 1416; Consultation, supra note 1, ¶¶ 74–76.

[38] Consultation, supra note 1, ¶¶ 78–80.

[39] Competition Appeal Tribunal Rules 2015, SI 2015/1648, r. 94 (UK).

[40] Fed. Trade Comm’n, Consumers and Class Actions: A Retrospective and Analysis of Settlement Campaigns 11–13 (September 2019) (finding a median claims rate of 9 per cent and a weighted mean of 4 per cent across the settlements studied); see also Consultation, supra note 1, ¶ 122 (noting that take-up in North America ‘will not often exceed ~10 per cent’).

[41] Consultation, supra note 1, ¶ 127.

[42] Marek v. Lane, 571 U.S. 1003 (2013) (Roberts, C.J., statement respecting denial of certiorari); Frank v. Gaos, 139 S. Ct. 1041 (2019) (per curiam); Martin H. Redish, Peter Julian & Samantha Zyontz, Cy Pres Relief and the Pathologies of the Modern Class Action: A Normative and Empirical Analysis, 62 Fla. L. Rev. 617 (2010).

[43] Competition Act 1998 (Redress Scheme) Regulations 2015, SI 2015/1587 (UK).

[44] See Oxera, Quantifying Antitrust Damages: Towards Non-Binding Guidance for Courts (December 2009) (study for the European Commission finding wide variation in estimated cartel overcharges); John M. Connor & Robert H. Lande, Cartel Overcharges and Optimal Cartel Fines, in 3 Issues in Competition Law and Policy 2203 (ABA Section of Antitrust Law 2008).

[45] See Wouter P.J. Wils, The Relationship Between Public Antitrust Enforcement and Private Actions for Damages, 32 World Competition 3 (2009).

[46] 2013 Government Response, supra note 15, ¶ 6.36; Consultation, supra note 1, ¶ 89.

[47] ICLE Refining Comments, supra note 2, at 5–8.

[48] Consultation, supra note 1, ¶¶ 103–104; see also Organisation for Economic Co-operation and Development (OECD), The Future of Effective Leniency Programmes: Advancing Detection and Deterrence of Cartels (2023).

[49] See Massimo Motta & Michele Polo, Leniency Programs and Cartel Prosecution, 21 Int’l J. Indus. Org. 347 (2003); Giancarlo Spagnolo, Leniency and Whistleblowers in Antitrust, in Handbook of Antitrust Economics (Paolo Buccirossi ed., 2008).

[50] Directive 2014/104/EU, art. 11(4), 2014 O.J. (L 349) 1.

[51] Antitrust Criminal Penalty Enhancement and Reform Act of 2004, Pub. L. No. 108-237, § 213, 118 Stat. 661, 666–67 (2004).

[52] See OECD, Recommendation of the Council on Transparency and Procedural Fairness in Competition Law Enforcement, OECD/LEGAL/0465 (2021) (calling for review by an impartial, independent adjudicative body).

[53] John Penrose MP, Power to the People: Independent Report on Competition Policy 18 (February 2021); Consultation, supra note 1, ¶¶ 152 n.62, 171.

[54] Communications Act 2003, c. 21, § 193 (UK).

[55] ICLE Refining Comments, supra note 2, at 5–8.

[56] Communications Act 2003, c. 21, § 194A(2) (UK); TalkTalk Telecom Group PLC v Office of Communications [2020] CAT 8, ¶¶ 137–140 (5 March 2020).

[57] Consultation, supra note 1, ¶¶ 192, 195 (table); see Gas Act 1986, c. 44, §§ 23D–23E (UK); Electricity Act 1989, c. 29, §§ 11E–11F (UK).

[58] Despoina Mantzari, Economic Evidence in Regulatory Disputes: Revisiting the Court-Regulatory Agency Relationship in the US and the UK, 36 Oxford J. Legal Stud. 565 (2016).

[59] Competition & Mkts. Auth., Energy Licence Modification Appeals 2021: Summary of Final Determination (28 October 2021) (upholding challenges concerning the outperformance wedge and the innovation-related ongoing-efficiency uplift, while rejecting challenges concerning the cost of equity).

[60] Competition & Mkts. Auth., Water PR24 References: Final Determinations—Summary (10 March 2026) (following an 18 March 2025 reference, raising the allowed return from 4.03 to 4.20 per cent and allowing £463 million in additional revenue, 17 per cent of the £2.7 billion sought).

[61] See Consultation, supra note 1, ¶ 195 (table); Gas Act 1986, c. 44, § 23D(4) (UK); Electricity Act 1989, c. 29, § 11E(4) (UK); Water Industry Act 1991, c. 56, § 12F(4) (UK); Transport Act 2000, c. 38, § 19B (UK); Civil Aviation Act 2012, c. 19, § 26 (UK).

[62] Consultation, supra note 1, ¶ 195 (table) (identifying substitution powers in energy price-control, air traffic and airport licence, and postal price-control appeals); Gas Act 1986, c. 44, § 23E (UK); Postal Services Act 2011, c. 5, § 59 (UK).

[63] Competition Act 1998 (Competition and Markets Authority’s Rules) Order 2014, SI 2014/458, sch., r. 3 (UK), https://www.legislation.gov.uk/uksi/2014/458/pdfs/uksi_20140458_en.pdf.

[64] See Ian S. Forrester, Due Process in EC Competition Cases: A Distinguished Institution with Flawed Procedures, 34 Eur. L. Rev. 817 (2009).

[65] A. Menarini Diagnostics S.r.l. v. Italy, App. No. 43509/08, ¶¶ 57–67 (Eur. Ct. H.R. 27 September 2011).

[66] Competition Act 1998, c. 41, § 58A (UK).

[67] Competition Act 1998, c. 41, sch. 8, ¶ 3 (UK).

[68] See ICLE Refining Comments, supra note 2, at 13–14 (discussing the inadvertent public disclosure of confidential Microsoft materials in the Microsoft/Activision proceedings).

[69] Competition & Mkts. Auth., Direct Consumer Enforcement Rules r. 5 (CMA201, April 2025).

[70] Competition Act 1998, c. 41, § 35 (UK).

[71] Competition & Mkts. Auth., Guidance on the CMA’s Investigation Procedures in Competition Act 1998 Cases ch. 14 (CMA8, December 2024).

I. Introduction

The International Center for Law & Economics (ICLE) supports Kuiper Systems LLC’s (Amazon Leo) application to provide direct-to-device (D2D) service in the 1.6/2.4 GHz mobile-satellite service (MSS) bands (ICFS File No. SAT-LOA-20260601-00224; Call Sign S00917).

Satellite broadband and D2D are emerging, capital-intensive markets. High fixed costs and long deployment cycles make it difficult to enter the consumer low-Earth-orbit (LEO) broadband market.[1] Amazon Leo’s deployment would help a competitor capable of operating at scale to overcome those hurdles, thereby strengthening competition and benefiting consumers.[2]

Amazon Leo’s proposed entry also shows how existing spectrum transactions can support new service. Amazon has agreed to acquire Globalstar, including its 1.6/2.4 GHz MSS authorizations, subject to the transaction’s closing.[3] The deal would pair Globalstar’s operating MSS network and spectrum authorizations with Amazon Leo’s planned D2D constellation. It would put scarce spectrum to more intensive use while giving companies a predictable basis for investment, consistent with the Federal Communications Commission’s (FCC) longstanding support for secondary-market spectrum transactions.

The FCC should grant the application and appropriate limited waivers so that another operator can expand D2D capacity using licensed spectrum that sees limited use. It should also address legitimate coexistence and interference concerns raised by terrestrial licensees and other satellite operators.

II. Competition, Connectivity, and Spectrum Use

LEO satellite broadband remains a developing market with steep entry costs. Before serving customers, a provider must invest heavily in satellites, launches, ground infrastructure, and regulatory approvals. Building a constellation—and earning a return on it—takes years.[4]

Those costs, along with spectrum constraints, likely limit how many providers can operate at scale. There is no fixed number of competitors that every broadband market needs,[5] but entry by a credible additional provider can put pressure on incumbents and give consumers more choice.[6] Timely application review and workable spectrum-sharing rules can reduce the uncertainty and delay that compound the costs of entry.[7]

Starlink dominates commercial LEO broadband service in the United States and nothing in the record suggests that it gained that position through anticompetitive conduct. Still, another provider capable of operating at scale would give consumers and terrestrial partners an alternative.

Amazon Leo proposes a D2D constellation of 5,105 satellites in five orbital shells, at altitudes ranging from 510 to 580 kilometers.[8] It plans to use the 1.6/2.4 GHz band and standards developed by the 3rd Generation Partnership Project (3GPP) to connect with compatible commercial handsets, without specialized customer equipment.[9] That design could extend messaging, location sharing, roadside assistance, and emergency communications to people beyond the reach of terrestrial networks, including those in rural and remote areas.[10] D2D could also provide backup connectivity when disasters disrupt terrestrial service.[11]

The application and Amazon’s proposed acquisition of Globalstar could also make more intensive use of existing MSS authorizations. MSS bands support satellite links to consumer devices, but use varies considerably by band.[12] The 2 GHz MSS band sees relatively limited use,[13] while Globalstar and Iridium actively serve customers in the Big LEO band. Globalstar uses 1.6 GHz frequencies for uplinks and 2.4 GHz frequencies for downlinks; Iridium uses 1.6 GHz frequencies in both directions.[14] Satellite operators are also developing D2D services through agreements involving MSS rights and terrestrial mobile spectrum.[15]

If the transaction closes, Amazon Leo could pair Globalstar’s operating network and MSS authorizations with its planned constellation. That combination offers a practical way to expand the capacity and reach of spectrum that already supports D2D service.

III. Waivers and Technical Review

Amazon Leo seeks limited waivers of several rules in Parts 2 and 25.[16] Some requests address filing procedures. Others concern frequency allocations, handset licensing, limits on signal strength, the treatment of links to users and ground stations, and rules for reviewing competing satellite applications.

The FCC may waive a rule for good cause when special circumstances warrant a departure and a waiver would serve the public interest. That includes cases in which strict application would frustrate the rule’s purpose.[17] The question for each request is whether the FCC can grant relief while retaining meaningful technical review and protection against interference.

Amazon Leo’s Schedule S request illustrates the distinction. Schedule S collects technical data for satellite applications. Amazon Leo entered a representative orbital plane in the form and supplied a separate spreadsheet with the full dataset the rules require.[18] If those materials let FCC staff assess individual satellites and potential interference, requiring Amazon Leo to enter the same data again in Schedule S would add work without strengthening review.

The FCC took a similarly tailored approach in its June 2026 Amazon Leo milestone order. It granted a limited waiver with conditions designed to preserve incentives for rapid deployment and efficient spectrum use.[19] The FCC should grant the requests here where Amazon Leo shows that substantive review and interference protections will remain intact, imposing conditions where needed to protect coexistence.[20]

A. Address Interference Through Targeted Conditions

Expedited review must still address credible technical concerns from terrestrial licensees and other satellite operators. The FCC should evaluate evidence of potential interference and, where needed, impose conditions tied to actual performance or require coordination. That approach can protect existing services while allowing more intensive non-geostationary orbit (NGSO) operations and more efficient spectrum use.[21]

Good-faith frequency coordination is essential to sharing spectrum. The FCC can address documented objections through conditions and coordination requirements, leaving operators room to negotiate the details of coexistence. Speculative disputes should not delay a decision on the application.

IV. Conclusion

The FCC should grant Amazon Leo’s application and the limited waivers that preserve substantive technical review and safeguards against interference. High entry costs and long deployment timelines make a credible additional LEO provider especially valuable. Amazon Leo’s planned constellation could expand D2D service through compatible handsets, give consumers and network partners another option, and provide connectivity where terrestrial networks are unavailable or disrupted.

If Amazon’s acquisition of Globalstar closes, the combined system could put Globalstar’s operating network and MSS authorizations to more intensive use. The FCC should address documented engineering concerns through tailored conditions and coordination requirements. Timely approval on those terms would support competitive entry, protect existing services, and make more efficient use of scarce spectrum.

[1] Kristian Stout & Michael Calabrese, Competition in the Low-Earth-Orbit Satellite Industry, Truth on the Mkt. (Oct. 10, 2023), https://truthonthemarket.com/2023/10/10/competition-in-the-low-earth-orbit-satellite-industry [hereinafter Stout & Calabrese, LEO Competition]. See also Kuiper Systems LLC, Application for Extension or Waiver of the Milestone Deadline, Order, ICFS File No. SAT-MOD-20260129-00065, DA 26-553, ¶ 1 (Space Bur. June 5, 2026) (calling Amazon Leo a “new competitor-at-scale” in low-Earth-orbit broadband).

[2] Michael Calabrese et al., Low Earth Orbit Satellites: Policies to Promote Spectrum Sharing, Foster Competition, and Close Digital Divides (Int’l Ctr. for L. & Econ. & New America Oct. 30, 2025), https://laweconcenter.org/wp-content/uploads/2025/10/Low_Earth_Orbit_Satellites_FINAL.pdf [hereinafter LEO Policy Working Group].

[3] Approval of the Proposed Assignment and Transfer of Control of Licenses and Authorizations Held by Globalstar to Amazon, Public Notice, GN Docket No. 26-134, DA 26-550, at 1 (June 4, 2026) (accepting the transfer applications for filing).

[4] Stout & Calabrese, LEO Competition, supra note 1.

[5] Eric Fruits et al., Dynamic Competition in Broadband Markets: A 2024 Update (Int’l Ctr. for L. & Econ. June 4, 2024), https://laweconcenter.org/wp-content/uploads/2024/06/Broadband-Competition-2024-Update.pdf [hereinafter Fruits et al., Dynamic Competition]; Eric Fruits et al., A Review of the Empirical Evidence on the Effects of Market Concentration and Mergers in the Wireless Telecommunications Industry (Int’l Ctr. for L. & Econ. Sept. 17, 2019), https://laweconcenter.org/wp-content/uploads/2019/09/ICLE-Telco_Merger_Lit_Review_Jud_Rpt_FINAL.pdf; Thomas W. Hazlett & Robert W. Crandall, Competitive Effects of T-Mobile/Sprint: Analysis of a “4-to-3” Merger, in Proceedings of the TPRC2024: The Research Conference on Communications, Information and Internet Policy (2024), https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4736059.

[6] Fruits et al., Dynamic Competition, supra note 5. See also Kuiper Systems LLC, DA 26-553, ¶ 1.

[7] LEO Policy Working Group, supra note 2.

[8] Application of Kuiper Systems LLC for Authority to Launch and Operate a Non-Geostationary Satellite Orbit System, Legal Narrative § II, ICFS File No. SAT-LOA-20260601-00224 (July 24, 2026) [hereinafter Kuiper D2D Application] (proposing 5,105 satellites in five orbital shells at altitudes of 510–580 km); see also Space Bureau Accepts for Filing Application of Amazon Leo for NGSO D2D System, Public Notice, ICFS File No. SAT-LOA-20260601-00224, DA 26-866, at 1 (Space Bur. Aug. 18, 2026) (accepting the application for filing).

[9] Kuiper D2D Application, supra note 8, at 7, 10.

[10] Id. at 10 (explaining that D2D would improve coverage for rural and remote users); see also id. at 2 (describing service for people beyond existing networks).

[11] Id. at 11 (explaining that D2D could keep rescue and emergency communications running when terrestrial networks fail); see also id. at ii (describing D2D as a fallback during terrestrial outages).

[12] LEO Policy Working Group, supra note 2.

[13] Timothy M. Farrar & J. Armand Musey, Spectrum for Emerging Direct-to-Device Satellite Operators (TMF Assocs. & Summit Ridge Grp. Jan. 27, 2025), https://summitridgegroup.com/wp-content/uploads/D2D-White-Paper-SRG-TMF-Final.pdf [hereinafter Farrar & Musey, SRG/TMF].

[14] Id.

[15] GSMA Intell., What Does the Early Data on D2D Usage Tell Us? (May 29, 2026), https://www.gsmaintelligence.com/blogs/what-does-the-early-data-on-d2d-usage-tell-us. See also Fed. Commc’ns Comm’n, Order, DA 26-398, ¶¶ 1–2 (Space Bur. Apr. 23, 2026), https://docs.fcc.gov/public/attachments/DA-26-398A1.pdf (reaffirming licensees’ exclusive mobile-satellite-service spectrum rights and encouraging secondary-market transactions).

[16] See Kuiper D2D Application, supra note 8, Ex. B (Waiver Requests), at 1, 6–9; 47 C.F.R. § 1.3 (2026); Ne. Cellular Tel. Co. v. FCC, 897 F.2d 1164, 1166 (D.C. Cir. 1990); WAIT Radio v. FCC, 418 F.2d 1153, 1159 (D.C. Cir. 1969).

[17] 47 C.F.R. § 1.3 (2026); Ne. Cellular Tel. Co. v. FCC, 897 F.2d 1164, 1166 (D.C. Cir. 1990); WAIT Radio v. FCC, 418 F.2d 1153, 1157, 1159 (D.C. Cir. 1969).

[18] Kuiper D2D Application, supra note 8, Ex. B (Waiver Requests), at 7.

[19] Kuiper Systems LLC, DA 26-553, ¶¶ 1–2.

[20] Kuiper Systems LLC, DA 26-553, ¶¶ 1–2; Gerald Adams, Out of Orbit: How Outdated Spectrum Rules Hold Back Satellite Broadband (Int’l Ctr. for L. & Econ. Aug. 12, 2026), https://laweconcenter.org/resources/out-of-orbit-how-outdated-spectrum-rules-hold-back-satellite-broadband [hereinafter Adams, Out of Orbit].

[21] See Adams, Out of Orbit, supra note 20; Eric Fruits & Kristian Stout, Comments of the International Center for Law & Economics Regarding Modernizing Spectrum Sharing for Satellite Broadband, SB Docket No. 25-157 (July 28, 2025), https://laweconcenter.org/wp-content/uploads/2025/07/2025-EPFD-Comments.pdf.

I. Introduction and Summary

The International Center for Law & Economics (ICLE) submits these comments in response to the Federal Communications Commission’s (FCC) Notice of Proposed Rulemaking (NPRM) on reducing local barriers to broadband deployment.[1] ICLE is a nonprofit, nonpartisan research center that applies law & economics to public policy. Its work promotes sound economic analysis and consumer welfare, particularly in dynamic, technology-driven markets such as telecommunications.

The NPRM proposes reforms to streamline the state and local approvals that providers must obtain before building wireline infrastructure. Deployment often requires numerous permits, rights-of-way authorizations, and access agreements. Slow, discretionary, or unpredictably priced approval processes increase the cost and risk of investment and can make otherwise viable projects uneconomical.

These barriers carry especially high costs in broadband markets. Broadband networks require substantial upfront investments that providers cannot recover if a project fails. They also generate broad economic and social benefits that providers cannot fully capture. Permitting delays and excessive fees therefore shape where and how quickly providers build next-generation networks—and whether they build them at all. The resulting costs fall hardest on rural and lower-income communities, the very communities that federal connectivity policy seeks to reach.

The FCC should adopt its proposals to require cost-based fees and establish firm deadlines, known as shot clocks, for local review of wireline-infrastructure applications. It should pair those deadlines with a strong deemed-granted remedy that treats an application as approved when a locality fails to act on time. The FCC should also prevent localities from evading fee caps by demanding noncash, or in-kind, contributions. Together, these reforms would reduce avoidable deployment costs and help broadband providers build the infrastructure needed to support next-generation applications, including artificial intelligence.

II. Local Barriers Deter Broadband Deployment

As the NPRM explains, providers must secure numerous permits, rights-of-way authorizations, and access agreements before deploying infrastructure.[2] Slow, inconsistent, or unpredictably priced approvals increase both the cost and risk of deployment. These barriers can deter investment even when a provider has capital available. They therefore shape the cost of individual projects, where and how quickly providers build next-generation infrastructure, and whether they build it at all.

A. Broadband Economics Magnify Regulatory Barriers

Two features of broadband deployment make the FCC’s proposed reforms especially important. First, broadband networks require high fixed costs, and providers cannot recover much of that investment if a project fails.[3] A provider must incur most network costs before serving a single customer. If the network attracts too few customers, much of that spending is lost.[4]

Providers spread these fixed costs across their customer base. Projects serving large or affluent populations are therefore more likely to produce an adequate return, while projects in rural and lower-income communities often remain marginal.[5] Every avoidable regulatory cost or delay pushes more of those marginal projects below the threshold for investment.

Second, broadband creates broad social benefits that providers cannot capture through subscription revenue.[6] Reliable connectivity can raise property values, improve educational outcomes, expand access to telemedicine in ways that make rural living more viable, and allow remote workers to bring income into lower-cost communities.[7] Providers cannot charge for a farmhouse’s increased resale value or a student’s improved grades.

These spillover benefits create an underinvestment problem. A broadband project may produce substantial value for society while generating too little revenue to justify private investment. AI, augmented reality, and virtual reality will widen this gap because their benefits will extend far beyond what any single broadband provider can monetize. Policies that reduce deployment costs and risks therefore create value beyond the revenue that providers earn.

B. Fragmented Approvals Create Holdout Risk

State and local requirements impose transaction costs and create holdout risks that can exceed the cost of the equipment itself. A single linear route may require agreements with several municipalities, county and state authorities, private landowners, and owners of thousands of utility poles.[8] Each party controls a segment of the project and has an incentive to seek the highest payment its approval can command. Yet no single party bears or fully accounts for the cumulative burden imposed on the network.

Many projects thus require consent from a long chain of independent parties, any one of which can block the entire build.[9] This structure creates a holdout problem. After a provider has invested in a route and obtained nearly every required approval, the final gatekeeper can demand far more than the underlying access is worth. The provider must then overpay, redesign the route, or abandon the investment already sunk into the rest of the project.

C. Delay and Uncertainty Deter Marginal Projects

Delay and unpredictability often impose greater costs than the fees themselves. A project’s economics depend heavily on when it begins generating revenue relative to when its costs accrue. Protracted or discretionary review postpones revenue while financing and development costs continue to mount. Each month of delay also consumes part of an asset’s useful economic life and increases the risk of technological obsolescence.[10] A 10-year asset that spends three years awaiting approval must repay its debt from seven years of revenue instead of 10.

Open-ended standards compound these costs. A provider can budget for a fixed fee. It cannot price a discretionary decision with the same confidence, so it must demand a higher expected return to compensate for the uncertainty. Firm deadlines with meaningful consequences for inaction, paired with fees tied to a government’s actual and direct costs, reduce the risk premium built into deployment decisions. This benefit extends even to projects completed on schedule, because providers must evaluate them ex ante based on the possibility of delay.

These burdens fall most heavily on marginal projects. Excessive delay and denial risk raise the minimum return needed to justify deployment, eliminating projects in rural and lower-income communities first.

The FCC can directly address these problems by establishing firm deadlines backed by a presumption of approval, limiting fees to reasonable government costs, and counting in-kind demands against those limits. These safeguards target the delay, discretion, and rent extraction that impose the largest avoidable costs and risks on broadband deployment.

III. Shot Clocks Need Effective Enforcement

As explained above, broadband providers need regulatory certainty when deciding whether to invest. ICLE supports the proposed 120-day shot clock, but the remedy for missed deadlines matters more than the precise timeframe.[11] If the record supports a shorter period to encourage deployment or a longer one to let localities manage their public rights-of-way, the FCC should adjust the deadline accordingly.

Whatever period the FCC chooses, it must provide swift and meaningful enforcement. A deadline offers little relief if a locality can ignore it, stall a deployment, and create a de facto moratorium on new builds. Although the NPRM declined to propose a deemed-granted remedy, it left the issue open, stating that the FCC “may revisit this decision if evidence submitted to the Commission suggests that a ‘deemed granted’ remedy is needed.”[12]

The economics of broadband deployment provide sufficient evidence of that need. Without an automatic consequence for inaction, providers remain exposed to the same uncertainty that the shot clock should eliminate. The FCC could wait for a more extensive record, but broadband deployment would suffer in the meantime. It should instead adopt a deemed-granted remedy under which an application receives approval when a locality fails to act before the deadline.

The FCC should also apply a single deadline to all authorizations required for access to a public right-of-way. Allowing localities to stack separate shot clocks for rights-of-way agreements, construction permits, road-closure permits, and other approvals would preserve the delay and uncertainty the rule seeks to eliminate.

A. Start the Clock at the First Procedural Step

The FCC should clarify that the shot clock begins to run when a provider initiates the first procedural step in the application process. A firm deadline does little to constrain delay if a jurisdiction can postpone its start by withholding acknowledgment of an application, imposing open-ended preapplication or completeness requirements, or excluding informal preliminary steps from the clock.[13]

These open-ended requirements introduce the same costly discretion as an unenforceable deadline. In the aggregate, they can recreate the de facto moratoria that the shot clock should prevent. Predictability requires an objective, provider-triggered starting point that localities cannot manipulate. The clock should therefore begin when the provider takes the first required procedural step, rather than when the locality later decides that the application is complete.

B. Allow Batch Applications

The FCC should also allow providers to submit batch applications. It has previously approved batching for small-cell deployments.[14] Although wireline projects differ in scale, the same logic applies when a provider proposes substantially uniform work, such as repeated attachments using the same design along a fiber route or materially identical installations using common construction methods.

Batch review would benefit providers and localities alike. Providers could avoid duplicative filings, fees, and staggered timelines. Localities could conduct one engineering and legal review of the shared elements instead of repeatedly examining identical specifications. That approach would accelerate deployment while easing the burden on resource-constrained local review staff.

IV. Fee Limits Must Track Actual Costs

The NPRM cites several examples of localities imposing fees that exceed the actual costs of processing permits and maintaining public rights-of-way.[15] These examples provide ample support for cost-based fee rules. Two additional disputes further illustrate what happens when gatekeepers can impose charges untethered from their costs.

A. Maple Grove Treated Broadband as a Revenue Source

Recent events in Maple Grove, Minnesota, show how directly some localities treat broadband deployment as a revenue opportunity. In March 2026, the city amended its code to require every “Broadband Service Provider” to obtain a cable-television franchise from the Northwest Suburbs Cable Communications Commission (NSCCC) before the city would act on any rights-of-way application.[16]

Gateway Fiber is a certificated telecommunications carrier that provides broadband internet access, rather than cable service. When it sought permits to continue construction, the city deferred its applications. A provider accustomed to receiving permits within four to six weeks suddenly could not obtain them at all.[17]

The local commission stated its financial motive plainly. It projected that cable-franchise-fee revenue would fall by about 10% annually and identified “new broadband fees” as the “most impactful” source of replacement revenue.[18] The city’s Request for Council Action likewise tied the ordinance to “declining cable franchise revenue.” It said nothing about coordinating construction, protecting pavement, or preventing conflicts among users of the rights-of-way—the interests that might justify such a requirement.[19] The ordinance sought to turn a broadband entrant into a replacement source of franchise revenue.

B. Pole-Replacement Charges Require Clear Standards

A recent pole-replacement dispute illustrates a related problem. Owners of essential infrastructure can use their gatekeeping position to impose charges untethered from cost. In February 2026, the FCC rejected Appalachian Power Company’s (APCo) “cost causer” policy. The policy required a company seeking to attach equipment to an APCo pole to pay upfront 100% of the cost of replacing a pole that already violated applicable standards, subject to a possible 50% refund.[20]

The FCC held that a new attacher may pay only the incremental cost of the taller or stronger pole its equipment requires. APCo could not charge the attacher for correcting a preexisting violation.[21] The dispute nonetheless continued after the ruling. Rather than apply the incremental-cost standard, APCo allegedly billed Comcast a flat share of roughly 20% of the full replacement cost, regardless of the costs Comcast’s attachment caused. APCo also reportedly treated 20% as a floor rather than a ceiling, prompting Comcast to file another complaint.[22]

Investor-owned utilities may have especially strong incentives to extract rents, but similar incentives operate when local governments and permitting offices control access needed for broadband deployment. A rule limiting fees to “reasonable” or “actual” costs will constrain those incentives only if the FCC defines the standard concretely and provides an effective remedy. Providers should be able to challenge systemic overcharges without relitigating the same issue one gatekeeper and one pole at a time.[23]

V. Fee Caps Must Cover In-Kind Contributions

A fee cap that excludes in-kind contributions constrains the form of compensation while leaving the total burden unchecked. When a locality conditions access on free service to municipal buildings, dedicated fiber or conduit, capacity on an institutional network, or a comparable obligation, the provider bears the added cost of supplying that service or facility.[24]

An in-kind obligation requiring $1,000 in additional spending imposes the same direct burden as a $1,000 cash payment. Its total cost may be even greater if the obligation displaces commercial uses or forces the provider to expand capacity sooner than planned. The precise cost depends on the provider’s available capacity and the alternative uses it could reasonably expect.

A cap limited to cash payments therefore allows localities to collect compensation above the nominal ceiling. These demands can increase subscriber prices, reduce service quality, and discourage investment. Although the effects depend on market conditions and the nature of the obligation, they fall hardest on marginal projects that broadband policy should encourage.

Cable franchising offers a cautionary precedent. Congress defined “franchise fee” broadly to prevent local authorities from evading the 5% statutory ceiling by demanding services or facilities instead of cash.[25] Localities cannot circumvent Congress’ limit simply by changing the form of compensation.[26] The FCC recognized this principle in its 2019 Section 621 order, which held that nonincidental in-kind obligations count toward the 5% cap.[27]

The same principle applies to the fee limits in this proceeding. The FCC should adopt its proposal to count the value of state and local in-kind demands toward any safe-harbor fee level.[28] Covered demands should include dedicated capacity, free service to public facilities, donated equipment, and similar assessments that function as payment.

To make the limit durable, the FCC should require the locality to justify the value assigned to each in-kind demand. That valuation should reflect the provider’s marginal cost of meeting the demand and its opportunity cost from displaced commercial uses.

VI. Section 253 Supports the Proposed Rules

A threshold objection is that Section 253 protects the ability to provide “telecommunications service,” while the FCC currently classifies broadband internet access as a Title I information service. That objection misconstrues both the statute and the deployments at issue. The FCC can apply firm deadlines and cost-based fee limits to the facilities covered by this proceeding without changing the classification of retail broadband. Section 253 protects the deployment of infrastructure capable of carrying telecommunications, regardless of the regulatory classification assigned to a particular retail service.

Section 253(a) reaches any state or local requirement that prohibits or has the effect of prohibiting “the ability of any entity to provide any interstate or intrastate telecommunications service.”[29] This language extends beyond firms that sell telecommunications directly to consumers. It also covers entities that deploy infrastructure capable of providing telecommunications services.[30]

The same fiber that delivers mass-market broadband often carries—or can carry—interconnected voice over internet protocol (VoIP), business-data and enterprise-transport services, dark-fiber leases, and point-to-point connections. Providers building these networks also commonly hold state telecommunications certificates and file tariffs for regulated services.[31] The relevant considerations are the network’s telecommunications capabilities and the burden imposed on them. A provider’s use of the same fiber to offer an information service does not place the infrastructure beyond Section 253’s reach.

The FCC should therefore adopt its proposal to prohibit state and local governments from imposing additional requirements on wireline telecommunications infrastructure because the facilities “may be used to provide other services.”[32] This commingled-facilities rule would prevent localities from accomplishing indirectly what Section 253 forbids them to do directly.

The proposal also follows the approach the FCC has taken in the cable context. Its mixed-use rule prevents franchising authorities from regulating noncable services carried over a cable operator’s mixed-use system.[33] Applying the same principle here places the shot-clock and fee-cap rules on settled statutory ground, even when providers also use the infrastructure to offer an information service.

VII. Conclusion

The proposed reforms reflect the basic economics of infrastructure investment. Broadband deployment requires large, largely sunk investments under substantial uncertainty. It also creates broad social benefits that no provider can fully capture through subscription revenue. State and local delay, discretion, and rent extraction raise the minimum return needed to justify deployment, eliminating marginal projects first. Rural and lower-income communities bear the greatest burden, even though federal broadband policy seeks to reach them most urgently.

Firm deadlines backed by a deemed-granted remedy, fees tied to a government’s actual, direct costs, and in-kind demands counted toward those limits would reduce the risk built into every deployment decision. The FCC can adopt these measures on settled statutory grounds. Section 253 reaches requirements that prohibit or effectively prohibit an entity’s ability to provide telecommunications service. That protection extends to infrastructure capable of carrying telecommunications, regardless of how the provider’s retail services are classified.

ICLE therefore urges the FCC to adopt a firm shot clock backed by a deemed-granted remedy, cost-based fee limits that include in-kind contributions, and a commingled-facilities rule that prevents localities from indirectly imposing requirements that Section 253 directly prohibits. These reforms would give providers the predictability needed to invest and help ensure that next-generation networks reach the communities that need them most.

[1] See Build America: Eliminating Barriers to Wireline Deployment, Notice of Proposed Rulemaking, WC Docket No. 25-253, FCC 26-40 (rel. June 25, 2026), https://docs.fcc.gov/public/attachments/FCC-26-40A1.pdf.

[2] See also Jeffrey Westling & Kristian Stout, The Permission Layer: Clearing the Way for AI-Era Broadband 13 (Int’l Ctr. for L. & Econ. Sept. 9, 2026), https://laweconcenter.org/resources/the-permission-layer-clearing-the-way-for-ai-era-broadband.

[3] Geoffrey A. Manne, Kristian Stout & Ben Sperry, A Dynamic Analysis of Broadband Competition: What Concentration Numbers Fail to Capture 20 (Int’l Ctr. for L. & Econ. 2021), https://laweconcenter.org/resources/a-dynamic-analysis-of-broadband-competition-what-concentration-numbers-fail-to-capture; see also Geoffrey A. Manne et al., Comments on Refreshing the Record in Restoring Internet Freedom and Lifeline Proceedings, WC Docket Nos. 17-108, 17-287 & 11-42, at 2 (Int’l Ctr. for L. & Econ. 2020), https://laweconcenter.org/resources/comments-on-refreshing-the-record-in-restoring-internet-freedom-and-lifeline-proceedings-in-light-of-the-d-c-circuits-mozilla-decision.

[4] See Comments of the International Center for Law & Economics, The State of Competition in the Communications Marketplace, GN Docket No. 26-78, at 8 (May 21, 2026), https://laweconcenter.org/resources/icle-comments-to-the-fcc-on-the-state-of-competition-in-the-communications-marketplace [hereinafter ICLE Competition Comments].

[5] Westling & Stout, supra note 2, at 4.

[6] Jeffrey Westling, The Last Mile Is a Paper Trail: Why Broadband Gets Stuck, Truth on the Mkt. (Apr. 22, 2026), https://truthonthemarket.com/2026/04/22/the-last-mile-is-a-paper-trail-why-broadband-gets-stuck.

[7] See Paraoma Sanyal et al., Fiber Deployment Has Significant Incremental Economic Benefits 32 (The Brattle Grp. 2024), https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5370089.

[8] Kristian Stout, Infrastructure Is Destiny: The Geography of the Next Technology Race 18 (Int’l Ctr. for L. & Econ. Issue Brief 2026), https://laweconcenter.org/resources/infrastructure-is-destiny-the-geography-of-the-next-technology-race.

[9] Robert Cooter & Aaron Edlin, The Falcon’s Gyre: Legal Foundations of Economic Innovation and Growth 6.21 (2014), https://www.law.berkeley.edu/library/resources/cooter.pdf.

[10] ICLE Competition Comments, supra note 4, at 11.

[11] Ajit Pai, Comm’r, Fed. Commc’ns Comm’n, Remarks at PCIA’s 2014 Wireless Infrastructure Show 3 (May 20, 2014), https://www.fcc.gov/document/commissioner-pai-remarks-pcia-2014; see also Comments of the International Center for Law & Economics and TechFreedom, In re Applications of Cellco P’ship & SpectrumCo LLC, WT Docket No. 12-4, at 9 (Mar. 16, 2012), https://laweconcenter.org/resources/in-the-matter-of-application-of-cellco-partnership-spectrumco-llc-for-consent-to-assign-licenses-wt-docket-no-12-4 (noting that the FCC’s wireless shot clock “has had little effect in practice”).

[12] Build America, supra note 1, ¶ 32.

[13] See Comments of the Free State Foundation, Eliminating Barriers to Wireline Deployments, WC Docket No. 25-253, at 5 (Nov. 18, 2025), https://freestatefoundation.org/wp-content/uploads/2025/11/FSF-Comments-Eliminating-Barriers-to-Wireline-Deployments-Final-111825.pdf.

[14] Accelerating Wireless Broadband Deployment by Removing Barriers to Infrastructure Investment et al., Declaratory Ruling & Third Report and Order, 33 FCC Rcd. 9088, ¶ 113 (2018), https://docs.fcc.gov/public/attachments/FCC-18-133A1.pdf.

[15] Build America, supra note 1, ¶ 35.

[16] Gateway Fiber Petition for Preemption and Declaratory Ruling Pursuant to Section 253(d) of the Communications Act, WC Docket No. 26-166, at 3 (June 17, 2026), https://www.fcc.gov/ecfs/document/26109834869/1.

[17] Id. at 2.

[18] Id. at 6.

[19] Id.

[20] Comcast Cable Commc’ns, LLC v. Appalachian Power Co., Memorandum Opinion and Order, Proceeding No. 25-330, Bureau ID No. EB-25-MD-002, FCC 26-6, ¶¶ 26–27 (rel. Feb. 5, 2026), https://docs.fcc.gov/public/attachments/FCC-26-6A1.pdf.

[21] Id. ¶ 2.

[22] Complaint of Comcast Cable Communications, LLC at 18, Comcast Cable Commc’ns, LLC v. Appalachian Power Co., Proceeding No. 25-330 (July 7, 2026), https://www.fcc.gov/ecfs/document/26109998224/1.

[23] Jeffrey Westling, The Cost of Holding Up Broadband, Truth on the Mkt. (July 14, 2026), https://truthonthemarket.com/2026/07/14/the-cost-of-holding-up-broadband.

[24] Comments of the International Center for Law & Economics, Implementation of Section 621(a)(1) of the Cable Communications Policy Act of 1984, as Amended by the Cable Television Consumer Protection and Competition Act of 1992, MB Docket No. 05-311 (June 18, 2019), https://laweconcenter.org/resources/icle-comments-on-implementation-of-section-621a1-of-the-cable-communications-policy-act-of-1984.

[25] See 47 U.S.C. § 542(b) (capping franchise fees at 5% of gross revenue), (g)(1) (defining “franchise fee” to include “any tax, fee, or assessment of any kind”); Implementation of Section 621(a)(1) of the Cable Communications Policy Act of 1984, Third Report and Order, 34 FCC Rcd. 6844 (2019).

[26] See Comments of the International Center for Law & Economics, supra note 24.

[27] See id.

[28] Implementation of Section 621(a)(1) of the Cable Communications Policy Act of 1984, supra note 25.

[29] 47 U.S.C. § 253(a).

[30] See Crown Castle Fiber, L.L.C. v. City of Pasadena, 76 F.4th 425, 433 (5th Cir. 2023).

[31] See Comments of Crown Castle Fiber LLC, Build America: Eliminating Barriers to Wireline Deployments, WC Docket No. 25-253, at 15–16 (Nov. 18, 2025).

[32] Build America, supra note 1, ¶ 68.

[33] See 47 C.F.R. § 76.43 (“A franchising authority may not regulate the provision of any services other than cable services offered over the cable system of a cable operator, with the exception of channel capacity on institutional networks.”).

I. Introduction and Overview

The International Center for Law & Economics (ICLE) is a nonprofit, nonpartisan research centre that applies economic analysis to legal and regulatory questions. We appreciate the opportunity to respond to the Competition Bureau’s examination, Behind the Price Tag: Examining the Food Supply Chain.[1] These comments address Question of Interest No. 3, which covers retail pricing and six related sub-questions.

ICLE uses law & economics analysis to determine whether challenged conduct harms consumers by raising prices, reducing output, lowering quality, or diminishing innovation. Our analysis builds on submissions to the Bureau concerning algorithmic pricing,[2] anti-competitive conduct and agreements,[3] merger enforcement guidelines,[4] and the future of Canadian competition policy.[5] We have also studied grocery competition in work on the proposed Kroger/Albertsons merger,[6] state grocery-pricing laws,[7] and online food-delivery fees.[8]

Canadian grocery prices rose sharply after 2021,[9] and Canadians reasonably want to understand why. That concern has focused attention on loyalty programs, pricing algorithms, and less visible changes such as “shrinkflation,” which reduces package size without changing the nominal price, and “skimpflation,” which lowers quality or reformulates a product to reduce costs.

The Bureau should evaluate each practice by its demonstrated effects on competition and consumers, not by whether it is new, automated, or unpopular. That inquiry must account for error costs—the harms caused by mistaken intervention or nonintervention. Wrongly condemning beneficial pricing practices may suppress them across the market, while failing to address harmful conduct may allow it to persist where entry barriers are high or switching is costly. Ambiguous cases therefore require market-specific evidence rather than presumptions based on the form of the conduct.

That caution applies in particular to Section 78(1)(k) of the Competition Act, which identifies “excessive and unfair selling prices” as a potential anti-competitive act without defining either term. High prices alone do not establish exclusion or harm to competition. Treating them as sufficient could turn abuse-of-dominance enforcement into price regulation. The Bureau has recognized that the Competition Act does not authorize it to regulate grocery prices and that price controls are “blunt tools” because authorities cannot reliably identify or continuously adjust the right price.[10]

These comments instead examine the conditions under which consumers can find, compare, and act on prices, and those under which firms can coordinate or exclude rivals. Across Item 3’s six sub-questions, we focus on consumer information, switching costs, entry barriers, deception, coordination, and exclusion. Grocery-price levels reflect broader macroeconomic forces and supply-chain conditions that competition enforcement cannot directly manage. The relevant questions are what conduct occurred, how it changed the choices available to firms and consumers, and what the evidence shows about its competitive effects.

A. Error Costs Counsel Evidence-Based Restraint

Competition enforcement operates under uncertainty. The error-cost framework, developed by Frank Easterbrook and extended in the modern literature, assesses legal rules by the consequences of mistaken decisions. A false positive, or Type I error, wrongly condemns conduct that benefits consumers or does not harm competition. A false negative, or Type II error, fails to stop conduct that does.[11]

These errors are not symmetric in a low-margin, high-volume industry such as grocery retail. False positives can prove both costly and durable. If enforcement condemns lawful discounting, loyalty pricing, or algorithm-driven responses to market conditions, firms may abandon those practices across their operations to avoid liability. Market forces cannot readily restore the lost benefits. False negatives are more likely to invite a market response because elevated margins attract entry, expansion, and substitution. As Easterbrook observed, a monopolist’s higher prices attract rivalry.

This asymmetry supports restraint in ambiguous cases, not inaction when the evidence clearly establishes harm. False negatives may persist where artificial barriers impede entry or consumers face high switching costs. Enforcers should require market-specific evidence of those conditions rather than presume that they exist.

B. Section 78(1)(k) Requires Cautious Enforcement

Bill C-56 (2023) amended the Competition Act’s abuse-of-dominance provisions to identify the imposition of “excessive and unfair selling prices” as an anti-competitive act.[12] Yet the Act defines neither “excessive” nor “unfair,” creating significant interpretive uncertainty. High prices are not inherently exclusionary. They allocate scarce goods and signal opportunities for entry. Deciding that a price is excessive requires judgments about common-cost allocation, the value of intangible assets such as brands and intellectual property, and the appropriate return for risk. Competition authorities cannot make those judgments reliably, particularly for retailers that sell many products and platforms that serve several interdependent groups of users.

The European Union’s experience enforcing Article 102 of the Treaty on the Functioning of the European Union illustrates the problem.[13] In United Brands v. Commission, the European Court of Justice established a two-part test that asks whether the margin between price and cost is excessive and, if so, whether the price is unfair in itself or compared with competing products.[14] Despite that framework, the European Commission rarely brings excessive-pricing cases because measurement remains difficult and the risk of error remains high.[15]

Comparative studies likewise counsel intervention only in exceptional cases. Surveys of EU and national case law, along with Organisation for Economic Co-operation and Development (OECD) roundtables, show that these cases typically involve regulated or quasi-regulated sectors, legal privileges, or earlier exclusionary conduct. Authorities rarely challenge prices as exploitative without some related restraint on competition.[16]

David S. Evans and A. Jorge Padilla,[17] Massimo Motta and Alexandre de Streel,[18] and Lars-Hendrik Röller[19] similarly argue that excessive-pricing claims are most defensible when linked to exclusionary conduct or durable structural barriers that competition law can identify and address. Policing high prices in markets where entry remains feasible creates a serious risk of measurement error and turns competition authorities into price regulators—a role for which they lack both the tools and the institutional mandate.

The Bureau should therefore presume that pricing conduct benefits competition unless exclusionary conduct prevents rivals from entering or expanding. High prices alone should not create liability without evidence that they result from, or reinforce, conduct that substantially lessens competition. The 2025 expansion of private access to the Competition Tribunal heightens the risk by allowing more private parties to bring cases.[20] Canadian law also lacks a screen comparable to the U.S. antitrust-injury doctrine, which requires a private plaintiff to show that its injury flows from harm to competition rather than from competition itself.[21]

The practices covered by the Bureau’s six sub-questions do not have a common competitive effect and require individual assessment. Our response to Sub-Question 1 illustrates the point. Pricing algorithms may automate lawful market monitoring, or they may facilitate coordination by pooling confidential data. Loyalty discounts may expand output, or they may exclude rivals when combined with market power and substantial foreclosure. Sub-Questions 4 through 6 present similar trade-offs. Disclosure may reduce consumers’ search costs, but it may also help competitors detect departures from a coordinated price.[22]

The Bureau should ask what conduct occurred, how it changed the choices available to firms and consumers, and what the evidence shows about prices, output, quality, innovation, entry, and switching. The analysis should turn on the conduct’s mechanism and demonstrated effects.

II. Sub-Question 1: Which Pricing Practices Help, and Which Hurt, Competition

The Bureau asks which pricing practices, including loyalty programs and pricing algorithms, help or harm competition. Most strengthen competition by improving price discovery, helping supply respond to demand, reducing waste and shortages, and intensifying rivalry for consumers. Algorithms often automate lawful pricing decisions, dynamic pricing helps markets adjust to changing conditions, and personalized prices and loyalty discounts can expand output and deliver genuine savings.

Competitive harm arises through narrower, fact-specific mechanisms. Firms may coordinate through shared confidential data or a common pricing agent, while dominant firms may use exclusive loyalty terms to foreclose rivals. Existing competition law already reaches those cases. The Bureau should target evidence of agreement or exclusion, not price variation, software use, or loyalty programs themselves. Questions about notice, transparency, and data use may warrant consumer-protection scrutiny, but they do not by themselves establish harm to competition.

A. Algorithms Can Compete or Coordinate

Pricing algorithms usually automate tasks that human managers have long performed, including monitoring competitors’ public prices, tracking inventory, and estimating how demand responds to price. Airlines and hotels have used software to adjust fares and room rates for decades. Scale and speed have changed, but the underlying activity has not.

Maureen K. Ohlhausen, then acting chair of the U.S. Federal Trade Commission, captured this point with the “guy named Bob” test. If a pricing manager named Bob may lawfully review competitors’ public prices and adjust his firm’s prices, an algorithm may perform the same task.[23] Automating lawful conduct does not make it unlawful. Appendix A provides a fuller discussion of algorithmic pricing and competition.

The empirical evidence confirms that pricing algorithms can produce different effects in different markets. Zach Y. Brown and Alexander MacKay studied high-frequency price data for over-the-counter allergy medication sold online.[24] They found that retailers using algorithms to update prices within hours systematically charged lower prices than competitors that updated prices daily or weekly.

Sophie Calder-Wang and Gi Heung Kim studied rental housing and found that landlords using pricing algorithms reduced rents more quickly during demand downturns, including the Great Recession.[25] Although rents increased more quickly during expansions, the software improved occupancy and helped the market clear when demand weakened.

Stephanie Assad and her co-authors reached a different result in their prominent study of German retail-gasoline stations.[26] They found that algorithmic pricing increased average prices and margins by roughly 15%, but the effects varied sharply with market structure. Adoption had little effect in monopoly markets. In duopoly markets, where two stations competed and both adopted pricing software, margins rose by 38% relative to markets where neither station adopted.[27] The authors called the result “tacit collusion,” reasoning that the algorithms learned not to undercut because competitors would quickly match price cuts.[28]

The evidence, though, indicates conscious parallelism—firms independently making similar decisions while recognizing their mutual interdependence—rather than an agreement. The German stations did not communicate, signal conditional strategies, coordinate through software vendors, agree to use common software or pricing rules, or exchange competitively sensitive information. Each station independently adopted pricing software, and each algorithm independently learned that aggressive price cuts were unprofitable.[29] Similar tools responding to similar market conditions produced parallel outcomes without a meeting of the minds. As the Bureau’s proposed enforcement guidelines correctly recognize, conscious parallelism alone does not violate Canadian competition law.[30]

The 9th U.S. Circuit Court of Appeals applied this distinction in Gibson v. Cendyn. Competing hotels had independently entered vertical software-licensing agreements at different times. The plaintiffs alleged that the agreements collectively produced anti-competitive effects, even though “neither the terms nor the operation of the licensing agreements [were] alleged to have harmed competition by affecting the competitive incentives in the relevant market, nor did the licensing agreements restrain any party’s ability to compete.”[31] The court held that conclusory allegations could not survive even the pleading stage. Plaintiffs still had to allege “a restraint of trade that causes an actual adverse effect on competition.”[32]

The 3rd U.S. Circuit Court of Appeals’ 2026 decision in Cornish-Adebiyi v. Caesars marks the other side of the line.[33] The court reversed a dismissal because casino-hotel guests had plausibly alleged a horizontal price-fixing agreement under Section 1 of the Sherman Act. Competing hotels allegedly delegated room pricing to a shared algorithm, supplied it with real-time nonpublic pricing and occupancy data, and followed its recommendations roughly 90% of the time. The court held that staggered adoption over 14 years affected the weight of the evidence, not whether the allegations supported an inference of conspiracy.

The legal question is whether firms agreed to coordinate. Parallel price movements alone do not establish that consensus. The decisive allegation in Cornish-Adebiyi was that competitors pooled confidential, nonpublic data through a shared pricing agent—a mechanism capable of converting an ordinary competitive input into a coordination device.

B. Dynamic Pricing Balances Supply and Demand

A single fixed price can be too high during slack periods, leaving willing buyers unserved, and too low during peak periods, producing shortages and long waits. Dynamic pricing—prices that change as supply and demand change—can lower off-peak prices to encourage consumption and raise peak prices to allocate scarce supply to those who value it most.

The practice long predates modern pricing algorithms. Airlines have managed seat availability against forecast demand since Ken Littlewood developed his booking rule in 1972. By 1992, American Airlines estimated that its yield-management program, which adjusted fares and seat availability based on expected demand, had generated $1.4 billion in quantifiable benefits over three years.[34]

Evidence from airline markets shows that dynamic pricing can expand output and increase total welfare. Kevin R. Williams found that it increased output by 2.7% and total welfare by 1% relative to charging one price per flight. Dynamic fares could fall for early, price-sensitive travellers and adjust when demand surged. They also reduced stockouts. Under uniform pricing, 29.4% of flights sold out before departure, compared with 18.7% under dynamic pricing, and uniform-price flights sold out earlier.[35]

An efficiency gain does not mean every buyer pays less or receives the benefit. In Williams’ study, airline revenue increased by 7.6%, while total consumer surplus fell by 6.3% relative to uniform fares. Airlines captured much of the gain. Early-booking leisure travellers generally benefited, while late-booking business travellers paid more.

The evidence from ride-hailing is more favourable to consumers. Juan Camilo Castillo found that surge pricing in Houston’s Uber market increased total welfare by 2.15% of gross revenue, or roughly $0.25 per trip, relative to uniform pricing. Riders at every income level benefited through lower average prices, shorter pickup times, and more reliable trips, with low-income riders gaining the most.[36] The gains were not universal. Driver surplus fell by 0.98% of gross revenue, leading Castillo to conclude that drivers “might be right to complain.”

Andrew Sweeting documented the same basic logic in another market for a perishable product: Major League Baseball tickets, which lose all value once a game begins. Sellers reduced prices by 40% or more as game time approached, and the ability to reprice increased the average seller’s expected payoff by roughly 16%.[37]

Grocery shoppers encounter the same practice when stores mark down bakery items, meat, and produce approaching their sell-by dates. Grocery pricing remains much closer to uniform pricing than to surge pricing. Its principal dynamic elements—perishable-goods markdowns and discounts that respond to demand—help sell inventory that might otherwise spoil or remain unsold.

The relevant comparison is therefore not always a fixed price. It may be the nonprice rationing that replaces flexible pricing, including stockouts, queues, spoilage, and unused capacity. Time-varying prices are familiar even for regulated essential services. In the PowerCentsDC experiment, established through a settlement approved by the District of Columbia’s utility regulator, households facing dynamic electricity rates produced “stable, predictable, and sizable demand reductions” during critical peak periods.[38]

Price increases during periods of scarcity can still raise legitimate questions about transparency and notice. Those are consumer-protection questions about how firms disclose prices, and consumer-protection tools can address them. They do not establish that adjusting prices to market conditions harms competition.

Pricing practices that may harm competition operate through different mechanisms, including prices tied to a shopper’s identity and algorithmic coordination among rivals. We address the first in the next subsection and the second in Appendix A (ICLE comments to the Bureau on algorithmic pricing and competition).

C. Personalized Pricing Is Not Inherently Harmful

Personalized pricing means offering different prices to different customers based on data-driven estimates of their willingness to pay. A targeted discount to a price-sensitive shopper who would otherwise buy nothing expands output and can benefit both the seller and the shopper. Brian Albrecht shows that, under competition, firms minimize consumer prices when they can identify customers who view their products as close substitutes because that information intensifies competition for those customers.[39] An empirical study of platform personalization similarly found that 62.5% of consumers would benefit. Losses were concentrated among the least price-sensitive customers, and most of the aggregate consumer harm arose from the platform’s market power rather than personalization itself.[40]

Price differentials are also common in competitive markets, so their presence in grocery retail does not show that competition has failed. James Dana demonstrates that these discount patterns can arise among firms that earn no economic profit and lack power to set the market price. As he observes, price discrimination “is common in many industries that appear to be extremely competitive.”[41]

Empirical studies reach the same conclusion. Joanna Stavins found that airline discounts tied to ticket restrictions were largest on the least concentrated routes: $253 at the 25th percentile of route concentration, compared with $165 at the 75th percentile. She cautioned that her sample covered only 12 routes.[42] Using a much larger sample, Severin Borenstein and Nancy Rose described competition’s significant positive effect on price dispersion—the variation in prices paid by different customers—as their strongest result.[43]

The Bureau also has evidence from Canada. Using Air Canada fare data, Ambarish Chandra and Mara Lederman of the University of Toronto found that competition reshapes price differentials rather than eliminating them. Each additional rival reduced fares in the middle of the distribution by 7% to 8% while barely changing the cheapest and most expensive fares. Some price gaps therefore narrowed while others widened, making aggregate measures of price dispersion potentially uninformative about how competition affects prices.[44]

The price differentials that persisted under rivalry reflected brand loyalty, which raises questions about consumer switching rather than market power. Canadian grocery shoppers do switch. The average Canadian regularly shops at 2.3 grocery stores, and multihoming—shopping across multiple stores and retail formats—is the industry norm.[45]

Businesses also appear to use personalized pricing rarely and on a limited scale. The European Commission conducted a mystery-shopping study across 160 e-commerce websites in eight member states. It found no price difference in 94% of 34,403 matched product pairs. Among the remaining 6%, the median difference was less than 1.6%.[46] Reputational concerns help explain firms’ caution. The OECD observes that businesses may avoid personalized pricing to protect their reputations and risk stigma when they adopt it.[47] We address the broader “surveillance pricing” narrative, and the gap between technical capability and observed conduct, in Sub-Question 3.

The Instacart episode illustrates how observers can mistake randomized price testing for personalized pricing—and how that mistake can produce legislation. A December 2025 study by the Groundwork Collaborative and Consumer Reports found that shoppers sometimes saw different prices for the same item at the same store and time. For example, a dozen Lucerne eggs at a Washington, D.C., Safeway appeared at five prices ranging from $3.99 to $4.79.[48]

The study’s own regressions undermine the claim that Instacart profiled individual shoppers. Across 193 cleaned submissions, price differentials associated with shoppers’ Instacart histories and demographic characteristics—including age, income, gender, and race—were “small and not statistically meaningful.” The study found no evidence that Instacart based prices on income, ZIP code, shopping history, or other individual characteristics.

The observed pattern instead resembles A/B testing, in which a company randomly assigns customers to groups to compare their responses to different prices. A random process, not a customer profile, determines the price each shopper sees. Instacart’s explanation that it tested randomized cohorts based on product category and location fits the study’s data.[49] The vendor’s marketing materials establish that its technology can target individual shoppers, but capability does not establish use. As Albrecht concludes, the episode “is a story about something that might happen, not something that did happen.”

Labelling the practice “surveillance pricing” assumes what the evidence failed to establish. Even scholars offering cautionary accounts of surveillance pricing recognize “complex trade-offs between efficiency gains and consumer welfare,” rather than uniform harm.[50]

The legislative response shows the cost of that mistaken inference. Maryland enacted its Protection from Predatory Pricing Act in April 2026 after the Instacart narrative dominated legislative debate. The law prohibits personalized prices based on a consumer’s personal data but exempts loyalty programs, memberships, subscriptions, and prices offered to consumers who consent to provide personal data. It therefore does not reach the A/B testing that prompted the debate.[51] When Colorado’s governor vetoed a similar bill, he identified the problem directly: It “would punish differentially lower prices, not just higher prices.”[52]

D. Loyalty Programs Deliver Genuine, Measured Savings

The United Kingdom’s Competition and Markets Authority (CMA) conducted a rigorous recent review of grocery loyalty pricing. It analysed a year of daily prices covering roughly 30,000 of the 86,000 loyalty promotions offered during that period, encompassing about 50,000 products at five supermarkets. These dual-price promotions offered one price to loyalty members and another to nonmembers.[53]

The CMA found that 92% of the products offered members genuine savings against the usual price. It applied a concrete test: The nonmember price during the promotion had to be the same as, or lower than, the price available to all shoppers during the preceding 28 days. The CMA found “very limited evidence of price changes which could indicate that supermarkets may have inflated prices to make their loyalty prices appear misleadingly attractive.” Across the five retailers, average member savings ranged from 17% to 25% of the nonmember price.

The pattern most likely to raise concern—a nonmember price higher than the price charged both before and after the promotion—appeared in only 0.6% of the roughly 50,000 products analysed. Other promotions in the surrounding days explained many of those cases. The CMA concluded that the potentially concerning examples “do not represent a systemic problem or suggest retailers are trying to mislead shoppers.”

The review also tested the suspicion that supermarkets fund loyalty discounts by inflating their regular prices. The data did not support it. Loyalty membership is nearly universal in the United Kingdom: 97% of shoppers belong to at least one scheme, and members hold three memberships on average. Yet 55% of surveyed shoppers believed that nonmember prices during loyalty promotions exceeded the usual price. That figure rose to 71% among shoppers aged 18 to 29. The CMA found that pattern in only 0.6% of the products analysed—roughly six in every 1,000.

Cross-retailer comparisons produced similar results. Of the 429 branded products the CMA compared, the loyalty price was the cheapest or tied for the cheapest available price for 368, or 86%. When compared with another retailer’s nonpromotional price, 52% of loyalty prices were cheaper, usually by 11% to 26%. Shoppers should still compare prices across stores. For 61 products, or 14%, the cheapest available price was not a loyalty price. That caveat supports better cross-retailer comparison tools, which we address in Sub-Question 6. It does not make the case against loyalty pricing.

Loyalty pricing is a longstanding feature of competitive markets, not a novel algorithmic harm. Frequent-flyer programs were already distinguishing loyal customers from occasional travellers in the deregulated U.S. airline industry during the 1980s.[54]

The CMA identified two practical caveats. First, some supermarkets could do more to ensure that shoppers without smartphones and those younger than 18 can access loyalty prices or understand how to obtain them. That calls for better program design, not abandoning loyalty pricing. Second, the CMA found no consumer-law concerns arising from supermarkets’ collection and use of personal data. Only 7% of nonmembers cited data concerns as their reason for not joining. The regulator that examined loyalty data most closely found no actionable evidence.

Genuine competitive harm from loyalty programs takes a different form: A dominant firm may impose exclusive terms that foreclose rivals or raise switching costs without providing offsetting value. We address those circumstances in “The Narrow Harm Cases” below, including our amicus brief in the EpiPen litigation. That brief explains why a legal test that cannot distinguish procompetitive discounting from exclusion will chill the discounting itself.

E. Liability Should Turn on Agreement or Exclusion

Three fact patterns can produce genuine competitive harm. Each depends on evidence of agreement or exclusion, not the presence of software or a loyalty card.

First, competitors may coordinate through a common vendor whose algorithm prices each firm’s products using a pool of rivals’ competitively sensitive data. That conduct differs from each firm independently using its own information and publicly available data. It raises a conspiracy question under section 45 of the Competition Act, whose criminal prohibition targets “naked restraints”—agreements such as price fixing that do not support a legitimate collaboration—and, on the civil side, under section 90.1.[55]

The Bureau has already drawn the relevant line. Its Competitor Collaboration Guidelines state that, “[i]n general, the Bureau does not consider publicly available information to be competitively sensitive.” Its proposed 2025 enforcement guidelines identify the common-vendor concern more specifically. If a third-party developer’s “algorithm uses commercially sensitive information provided to the developer by other firms,” the arrangement “could lead to coordinated outcomes without any direct information sharing between industry participants.”[56]

Second, a hub-and-spoke arrangement may include a “rim”—a horizontal agreement among competing firms to adopt and follow a common vendor’s recommendations. The rim distinguishes a conspiracy from a series of independent vertical contracts between the vendor and each firm. This, too, raises a section 45 question. The Competitor Collaboration Guidelines ask whether the parties reached a “meeting of the minds.” Conscious parallelism alone does not establish an agreement, although “parallel conduct coupled with facilitating practices, such as sharing competitively sensitive information,” may prove one.[57]

Third, a dominant grocer may impose exclusive loyalty terms that foreclose rivals or raise switching costs without providing offsetting value. That conduct may constitute abuse of dominance under sections 78 and 79. The Act’s enumerated anti-competitive acts include inducing a supplier to refuse to sell to a rival to impede its entry or expansion, and selling below acquisition cost to discipline or eliminate a competitor. Section 79 also requires that the effect on competition “is not a result of superior competitive performance.”[58]

The Bureau’s proposed guidelines take the same balanced approach. They identify volume and loyalty discounts as possible means of inducing exclusivity but recognize that “[e]xclusive dealing is not necessarily anti-competitive.” The concern arises when exclusivity forecloses rivals by denying them the scale, inputs, or sales channels needed to compete.[59] In each case, the agreement or exclusion creates the potential harm. The software or loyalty card does not.

The Bureau’s guidance already distinguishes these cases from lawful conduct, and U.S. algorithmic-pricing decisions show that courts can administer the distinction. Those decisions do not bind Canadian tribunals, but they help identify the allegations that can make a claim actionable.

In Gibson v. Cendyn, in which ICLE participated as amicus curiae, the 9th U.S. Circuit Court of Appeals explained that competing hotels could violate Section 1 of the Sherman Act by agreeing among themselves to follow a third party’s pricing recommendations. The pleaded claim failed because the software did not share one hotel’s confidential information with competing licensees. The court noted that allegations of such sharing could change the analysis.[60]

The district court reached a similar initial result in Cornish-Adebiyi v. Caesars, but the 3rd U.S. Circuit Court of Appeals reversed in July 2026. The appellate court read the complaint as alleging that each casino hotel supplied current, nonpublic room-pricing and occupancy data to a common platform, which processed those data alongside competitors’ information. Taken with the complaint’s other allegations, that pooling sufficed to plead a conspiracy.

In re RealPage involved the same critical allegation: The vendor pooled clients’ commercially sensitive pricing and supply data before returning pricing recommendations.[61] The relevant questions here concerned the flow of confidential data and any agreement to follow the common recommendation. A firm does not enter into a conspiracy merely by independently adopting software that relies on its own information and public data.

The Act also protects low prices because competition law seeks to promote them. An effects-based order under section 79 requires proof that the effect on competition “is not a result of superior competitive performance.” A lower price achieved through scale or efficiency reflects superior competitive performance, not abuse.[62] The Bureau’s proposed guidelines state the presumption directly: “Low prices are usually a sign of vigorous competition, which the Act is designed to encourage.”[63]

As ICLE and allied scholars argued as amici in the EpiPen litigation, loyalty discounts and the lower prices they produce routinely benefit consumers. A legal test that cannot distinguish procompetitive discounting from exclusion will chill the discounting itself.[64] A workable test should instead require objective evidence of foreclosure, including the share of the market covered and the duration of the challenged terms. Every sales contract “forecloses” the sales it wins, while short-term commitments that customers can easily terminate give rivals recurring opportunities to compete for their business.[65] A test based on the size of the savings rather than the extent of foreclosure condemns the discount itself—the costliest false positive in a low-margin industry.

III. Sub-Question 2: Impact on Small and Medium-Sized Businesses

The Bureau asks how these pricing practices affect small and medium-sized enterprises (SMEs). Modern pricing technology generally strengthens their ability to compete. Widely available third-party tools give smaller firms pricing capabilities once limited to large incumbents, reduce barriers to entry, and support new business models.

The greater risk comes from overbroad regulation. Same-terms mandates would erase legitimate cost efficiencies and raise consumer prices, while disclosure and documentation requirements impose fixed costs that fall most heavily on independent grocers. The Bureau should preserve smaller firms’ access to competitive tools and avoid rules that favour businesses large enough to absorb the cost of compliance.

A. Algorithms Lower Barriers to Entry

Sophisticated pricing strategies once required extensive data collection, specialized analytics, and continuous monitoring. Large incumbents could afford those resources, giving them a pricing advantage over smaller rivals. Widely available third-party pricing services now give firms of any size access to tools once available only through an internal analytics department. New entrants and smaller firms can adopt advanced pricing techniques without developing them in-house.

These services narrow the pricing-capability gap. A startup can use the same revenue-management software as an industry leader, reducing the risk that it will misprice its products or leave revenue unrealized. Algorithms thus reduce large firms’ scale advantage in pricing by allowing smaller firms to purchase sophisticated analysis through a vendor. The OECD has recognized this benefit, noting that algorithms can “reduce barriers to entry by allowing smaller entrants to gain market insights or develop new disruptive products at lower cost.”[66]

Algorithms can also support entry through new business models. Uber and other ride-hailing platforms built dynamic pricing into their services. By raising prices when demand surged, the platforms attracted more drivers and helped match supply with demand in real time. This addressed the peak-period shortages common in traditional taxi markets and helped establish a new market for on-demand rides. Consumers can now find rides more reliably at almost any time.

Had regulators prohibited surge pricing from the outset as “price gouging,” the platform model might not have attracted enough drivers or proved viable. Algorithms also enable products such as real-time price-comparison services, personalized shopping assistants, and dynamic-discount apps. These tools increase transparency, help consumers find better offers, and create new ways for firms to compete.

B. Same-Terms Rules Would Raise Prices

Rules requiring suppliers to offer every buyer the same wholesale terms would eliminate cost advantages associated with large-format retail. Emek Basker’s review of Walmart’s role in grocery markets found that its grocery prices averaged about 10% less than competitors’ prices. A Walmart Supercenter’s entry typically prompted competing grocery stores to reduce prices by roughly 1% to 2% in the short run. Over four years, the expansion of superstores, mass merchandisers, and warehouse clubs led traditional supermarkets to reduce prices by an estimated 3%.[67] Walmart’s presence therefore lowers prices for its customers and for consumers who continue shopping at competing stores.[68]

Using household panel data, Jerry Hausman and Ephraim Leibtag similarly estimated that the expansion of large-format supercentres produced substantial consumer-welfare gains. Those gains disproportionately benefited low-income households, which shop at price-sensitive formats more often than higher-income households.[69] A same-terms rule that erodes large retailers’ cost advantages would therefore raise prices most for shoppers with the fewest alternatives.[70]

Grocery prices already display substantial uniformity across stores. A same-terms rule would entrench that condition rather than correct it. Stefano DellaVigna and Matthew Gentzkow found that most large U.S. food, drugstore, and mass-merchandise chains charged nearly uniform prices across locations despite wide differences in local demographics and competition. The median chain sacrificed an estimated $16 million in annual profit by doing so.[71]

Uniform pricing is not the consumer-friendly benchmark it may appear to be. DellaVigna and Gentzkow’s model indicated that uniform pricing raised prices in the poorest decile of ZIP codes by 10% relative to the richest decile. Alberto Cavallo found similar uniformity across sales channels. Among 56 large multichannel retailers in 10 countries, online and offline prices were identical about 72% of the time overall and 52% of the time in food retail. Across Canadian sectors, they matched more than 90% of the time.[72] Requiring still greater uniformity would eliminate the price variation that most benefits low-income shoppers.

C. Compliance Costs Disadvantage Small Grocers

Disclosure and documentation rules designed for national platforms would burden small grocers disproportionately because compliance costs do not fall with transaction volume. Requirements for transaction-specific cost records or custom disclosure formats impose fixed costs that a large chain can spread across billions of dollars in sales but a small independent grocer cannot. Our comments to the U.S. Federal Trade Commission on online food-delivery fees examine this asymmetry in detail. They explain that the burden falls most heavily on small first-party operators—businesses that run their own delivery websites.[73]

IV. Sub-Question 3: Access to Data and Technology

The Bureau asks whether access to data and technology affects firms’ ability to adopt modern pricing practices and compete. In grocery retail, both inputs are widely available at modest cost. Third-party services give smaller firms sophisticated pricing capabilities, while grocery data depreciate quickly, lack strong network effects, and remain available from multiple commercial sources. Common software also does not imply pooled data. The current record does not show that data or pricing technology constitute scarce assets that foreclose rivals or entrench incumbents.

Regulatory intervention may create the greater competitive risk. Mandated data sharing and disclosure of confidential terms can facilitate coordination, weaken incentives to collect and analyse data, and impose disproportionate compliance costs on smaller firms. Claims that retailers use personal data to charge each shopper the most they will pay likewise require evidence, not inferences drawn from randomized price testing. The Bureau should distinguish firms’ use of their own or public data from the pooling of competitors’ confidential information, and leave genuine privacy concerns to privacy and data-protection law.

A. Common Algorithms Do Not Necessarily Pool Data

In its parallel consultation on algorithmic pricing, the Bureau observed that “[u]sing common pricing algorithms or pooling data among competitors may raise issues under the Competition Act.”[74] That concern is valid in principle but narrow in practice. The risk arises when third-party software ingests one firm’s confidential data and transmits it to a competitor, either directly or through recommendations based on pooled inputs. The common algorithm then becomes a conduit for competitively sensitive information.

Ordinary pricing software does not operate this way. Firms may independently use similar or even identical programs without pooling any data. A pricing algorithm, whether developed in-house or licensed from a vendor, generally relies on two types of information. The first consists of the firm’s commercially sensitive data, including sales volumes, inventory, and costs. Vendors can process that information while keeping it segregated from rivals. The second consists of lawfully collected public information, such as advertised prices and general market trends.

A hotel using revenue-management software, for example, may receive analysis of its own booking data, demand trends, and competitors’ advertised rates from public travel websites. It does not ordinarily receive a direct feed of a rival’s proprietary booking data. Reputable vendors design their systems to prevent the interfirm transfer of sensitive information because sharing it would expose their clients to liability.[75]

Monitoring competitors’ public prices is legitimate competitive conduct, whether performed manually or by software. Automation does not convert lawful information gathering into unlawful coordination. The Bureau’s Competitor Collaboration Guidelines generally treat public information as unproblematic and reserve concern for exchanges of competitively sensitive, nonpublic data.[76] They also distinguish direct exchanges between competitors, which present greater risk, from aggregated data supplied through an independent third party.

The controlling question is whether firms have moved beyond information gathering and agreed to coordinate their competitive decisions. Common analytical tools, similar market intelligence, and parallel outcomes do not establish such an agreement. An enforcer must still prove one.

Access to pricing data therefore does not ordinarily mean access to a rival’s secrets. It means access to a firm’s own records and publicly available market information. A policy that assumes algorithmic pricing necessarily pools competitors’ data would mischaracterize the technology and discourage lawful monitoring that helps discipline prices.

B. Pricing Technology Broadens Access

If access to technology affects firms’ ability to compete, pricing technology has broadened that access. This is the access dimension of the argument developed in Sub-Question 2, and it bears directly on the Bureau’s inquiry.

Sophisticated pricing once required scale. Firms had to collect data, acquire analytics expertise, and monitor markets continuously. Only large incumbents could justify those fixed investments, giving them a durable advantage over smaller rivals. Third-party pricing services have sharply reduced that advantage. They give a new entrant or small independent access to the revenue-management capabilities of an industry leader without requiring an in-house analytics department.

In economic terms, these services reduce the pricing advantages associated with firm size. A small firm can obtain sophisticated market analysis through a vendor’s platform. The OECD has recognized this benefit, observing that algorithms can lower barriers to entry by helping smaller firms gain market insights and develop new products at lower cost.[77]

Adoption data indicate broad diffusion rather than concentration. A 2020 survey by the Netherlands Authority for Consumers and Markets found that 80% of firms using pricing algorithms had developed them internally, while 20% relied on third-party providers.[78] In 2017, the European Commission found that 49% of European retailers monitored competitors’ prices. Of those retailers, 66.6% used price-monitoring software, meaning roughly one-third of all retailers used some form of algorithmic monitoring.[79] Among firms using monitoring software, 78% changed prices in response to competitors’ moves, and 35% used specialized algorithms to adjust prices automatically.[80] Surveys elsewhere in Europe show similar adoption patterns.[81]

Accessible pricing technology has also enabled new competitors and business models, including dynamic-pricing platforms, real-time price-comparison services, and discount apps. These products increase transparency and help consumers find lower prices. The competitive pressure they place on incumbents—often over incumbents’ objections—shows that the market remains open to entry.

The principal access risk comes from regulation itself. Compliance does not scale down. Disclosure, documentation, and audit requirements impose fixed costs that a national chain can spread across billions of dollars in sales but a small grocer cannot. Rules designed for the largest platforms could rebuild the entry barriers that accessible technology has lowered. We address this asymmetry in Sub-Question 2 and in our comments to the U.S. Federal Trade Commission on online food-delivery fees.[82] The Bureau should preserve pricing technology’s equalizing effect by avoiding compliance mandates that fall hardest on the smallest firms.

C. Grocery Data Does Not Create an Entry Barrier

A recurring theory treats large grocers’ accumulated transaction and loyalty data as a barrier to entry. On this account, incumbents use a self-reinforcing data advantage to price and target customers in ways that rivals cannot match. That theory fits grocery retail poorly for three reasons. The Bureau should require evidence of the mechanism rather than presume it.

First, grocery transaction data depreciate quickly and produce diminishing returns. Prices, promotions, product assortments, and consumer preferences change continuously. Each additional historical observation therefore adds less value, while stale data offer little predictive insight. A firm does not need every household’s complete purchase history to price competitively. It needs a representative and reasonably current picture of the market, which costs far less to obtain than the incumbent-scale archive assumed by the barrier theory.

Second, grocery retail lacks the demand-side network effects that can entrench data-driven platforms. Such effects arise when each additional user makes a service more valuable to other users. One shopper’s choice of a grocery chain does not make that chain more valuable to the next shopper. Grocery data therefore do not create a self-reinforcing loop that locks consumers into one retailer and denies rivals the scale needed to compete. Data accumulate, but they do not produce the winner-take-most dynamic found in markets where an entry-barrier theory is more plausible.

Third, the relevant data are neither exclusive nor unavailable to entrants. Consumers shop across supermarkets, supercentres, warehouse clubs, discount grocers, dollar stores, and online retailers. No single retail banner can see a household’s entire grocery basket.[83] Competition therefore spans formats well beyond the local supermarket.

Entrants can also purchase comparable market intelligence. Commercial sources offer scanner data, consumer-panel data, third-party analytics, and audience information. An entrant can buy much of the information that an incumbent collects internally. Data that are nonexclusive, rapidly depreciating, and available from multiple suppliers do not constitute an essential facility—an indispensable input that rivals cannot reasonably obtain elsewhere. Treating them as one would risk the durable false positive that the error-cost framework counsels against.

This point bears only on the Bureau’s data-access question. It is not a merger analysis. Our Kroger/Albertsons work addresses merger-specific concerns on their own terms.[84] The narrower conclusion here is that the current record does not show that access to grocery data creates a barrier requiring intervention.

D. Competitor Transparency Can Facilitate Coordination

The sub-question focuses on whether some firms have too little access to data. A more serious risk could arise from rules that compel firms to share data or disclose confidential terms in the name of equal access or “transparency.” The Bureau should distinguish consumer-facing transparency from competitor-facing transparency.

Consumer-facing transparency—including clear final prices, unit prices, and legible loyalty terms—reduces search costs and helps shoppers switch. Competitor-facing transparency can have the opposite effect. Mandating disclosure of supplier contracts, rebates, promotional allowances, or forward-looking pricing information can help rivals detect and punish departures from a coordinated price.

Tacit coordination rarely breaks down because firms cannot identify a common price. It breaks down because secret discounting makes deviations difficult to detect and punish. Anonymising disclosed terms does not eliminate that risk in a market with only a few large buyers. Volumes, shipping patterns, and timing may reveal the parties’ identities, and modern analytical tools make reidentification easier. Recent research shows that large language models (LLMs) can reidentify individuals or entities in ostensibly anonymised text by extracting identifying features and matching them with external data.[85]

The risk increases when disclosure rules cover only a small number of retailers. Even if supplier contracts omit retailers’ names, market participants may infer which chain received which terms. Disclosure could then make prices more uniform and higher. We develop this mechanism, including its interaction with statutory most-favoured-nation and same-terms rules, in Sub-Question 6.[86]

The same caution applies to data-sharing and data-portability mandates intended to “level” competitors’ access. Beyond facilitating coordination, compelled sharing weakens incentives to collect and analyse data. It may therefore degrade the very capability the mandate seeks to spread. Any such rule would also rest on the unsupported theory that grocery data create a barrier to entry, discussed in the preceding subsection. It would impose immediate, durable costs to address a harm that the evidence has not established—a strong case for error-cost restraint.

E. Personalized-Pricing Fears Outrun the Evidence

Much of the concern about data access rests on an unsupported narrative: Retailers use consumers’ personal data to charge each shopper the most they will pay. The episode that drove this narrative shows how misreading a study can produce poorly targeted law.

In December 2025, the Groundwork Collaborative and Consumer Reports published a study showing that Instacart shoppers sometimes saw different prices for the same item at the same store and time.[87] The study expressly found no evidence that Instacart based prices on individual characteristics such as income, ZIP code, or purchase history. The observed pattern instead matched routine A/B testing, in which random assignment determines the price a shopper sees.[88]

Unlike personalized pricing, A/B testing does not estimate what a particular shopper will pay. It assigns prices across randomized groups to learn about demand without relying on the shopper’s identity. Yet observers recast the study’s unexplained price variation as evidence of personal-data profiling.

The legislative response extended far beyond that finding. Lawmakers have introduced more than 40 bills addressing algorithmic or “surveillance” pricing across at least two dozen U.S. states, and Maryland enacted the first statute of its kind.[89] That law illustrates the risks of mismatched drafting.[90]

The statute’s title invokes predatory pricing—a settled term for below-cost pricing intended to exclude rivals—but its operative provisions address an unrelated practice by prohibiting prices personalized with consumers’ data. The law then exempts loyalty programs, memberships, subscriptions, and prices offered with consumer consent. It does not reach the A/B testing that prompted the debate. The triggering study did not find the practice the law targets, the statute’s title describes a different practice, and its exemptions sharply limit its reach. That is legislation shaped by alarm rather than measured effects.

Two facts should constrain any similar response in Canada. First, personalized pricing remains uncommon and modest in scale. The European Commission’s mystery-shopping exercise found price differences in only about 6% of matched tests, with a median difference below 1.6%. Those differences concentrated in industries such as airlines and hotels, where price discrimination is well established.[91]

Second, concerns about how businesses handle personal information fall primarily within privacy and data-protection law. Treating privacy concerns as competition problems produces rules poorly suited to either task. The Bureau should require evidence that a data practice harms competition or consumers before intervening and leave genuine privacy questions to the legal framework designed to address them.

V. Sub-Question 4: Price Comparison and Consumer Switching

The Bureau asks how the pricing practices at issue affect consumers’ ability to compare prices and value across retailers and their willingness to switch. Canadian shoppers actively compare offers, shop across multiple formats, and respond to price. Digital tools have reduced the time and effort required to compare retailers, while the pronounced shift toward discount stores provides direct evidence of price-driven switching. Loyalty pricing, dynamic pricing, and personalized pricing can make some comparisons more complex, but the evidence does not show that these practices suppress switching. Loyalty programs also lower effective prices and can intensify competition for shoppers.

The proportionate response is consumer-facing disclosure, including harmonized unit pricing, clear final prices, and legible loyalty terms, as discussed in Sub-Question 6. Canada’s concentrated grocery market and limited independent hard-discount competition raise separate concerns because major grocers own many discount banners. Those are supply-side questions about entry and expansion, which the Bureau’s 2023 market study addressed through structural recommendations. They do not establish that pricing practices prevent consumers from comparing prices or switching retailers.

A. Canadian Shoppers Compare Prices and Switch

Most Canadians buy groceries from five companies—Loblaws, Sobeys parent Empire, Metro, Costco, and Walmart—which together account for roughly two-thirds or more of national grocery sales.[92] The Bureau has found that concentration has increased in recent years[93] and that Canada has little independent hard-discount competition.[94] he degree of concentration remains contested. The Retail Council of Canada argues that Canada is not concentrated relative to peer countries, citing the rising shares of Costco and Walmart and the large majority of Canadians who regularly shop at discount retailers.[95]

When Canadians trade down to a discount grocer, they often remain within the same corporate family. Industry observers describe discount banners as a way to keep deal-seeking customers “within [the] estate” rather than lose them to a rival.[96] Still, roughly 6,900 independent grocers operate across Canada, and larger regional chains compete directly with national firms in some markets. The Bureau has found, however, that independents face significant barriers to becoming national competitors.[97]

Digital tools have reduced the cost of comparing prices. Weekly flyers, retailer apps, online storefronts, and third-party comparison services allow shoppers to survey competing offers before leaving home. That visibility disciplines attempts to price above competitive levels because consumers who see better alternatives can act on them.

Canadians regularly use that capacity. The Bureau’s survey found that 79% visit multiple grocery stores to find the best price at least sometimes. That figure includes 41% who do so sometimes, 21% often, and 17% very often.[98]

The clearest evidence of price-driven switching is the shift toward discount stores. The discount channel’s market share rose from roughly 35% in 2008 to near parity with conventional stores. Discount sales continue to grow faster than conventional sales by both dollars and volume.[99]

Online shopping provides another low-friction channel for comparison. About 30% of Canadians used at least one online option to buy groceries during a recent three-month period. That included 17% who used third-party delivery, 12% who used curbside pickup, and 9% who used home delivery.[100] Switching costs in these closely related channels can approach zero, as shoppers can open a competing app and compare offers side by side within seconds.

Individualized or conditional prices and tiered loyalty offers can complicate the comparisons made easy by flyers and shelf tags. The proportionate response is to improve comparison through clear final-price disclosure and, above all, unit pricing, which allows shoppers to compare prices per standard unit. This approach follows Recommendation 3 of the Bureau’s 2023 study, prompted by its finding that consumers have difficulty comparing even a few items across grocery stores.[101]

B. Loyalty Programs Can Strengthen Price Competition

Loyalty programs play a central role in Canadian grocery competition. About 75% of Canadians hold a loyalty or points card, and 61% say they are more likely to shop where they can earn rewards.[102] These programs affect switching in both directions. Points, targeted offers, and rewards earned across banners within the same corporate group make staying with one program more convenient, creating some switching friction. At the same time, they lower members’ effective prices. The Bureau has recognized that loyalty programs “can bring about lower prices for consumers” and “may also drive grocery stores to compete harder” for customers.[103]

The CMA’s 2024 loyalty-pricing review provides the most rigorous recent evidence. Of roughly 50,000 products offered through dual-price loyalty promotions, 92% delivered genuine savings. Average member savings ranged from 17% to 25%, and the CMA found very limited evidence that retailers inflated ordinary prices to make loyalty offers appear more attractive.[104]

Canadian consumers continue to compare prices and switch retailers. The pricing practices at issue have not prevented either behaviour. Switching friction alone does not establish competitive harm without evidence of deception, lock-in, exclusion, or market power. Restricting loyalty, dynamic, or personalized pricing based on an unsupported switching-harm theory would weaken the price competition that Canadians are driving through their shift toward discount retailers.

The Bureau should instead support consumer-facing tools that directly reduce comparison costs, including harmonized unit pricing, clear final prices, and legible loyalty terms. A wider range of competitors will strengthen comparison and switching more durably. Those competitors should remain free to use pricing tools that help them compete.

VI. Sub-Question 5: Shrinkflation and Skimpflation

The Bureau asks how shrinkflation and skimpflation affect consumers’ ability to compare price and value, and whether they influence purchasing and switching. Shrinkflation reduces a product’s quantity while holding its shelf price steady. Skimpflation reduces quality or changes a product’s formulation to cut costs. Both are rational nonprice responses to rising input costs, and both occur in competitive and concentrated markets. Neither practice alone reveals the level of competition.

The principal competition concern is comparability. Shrinkflation can succeed because consumers notice price changes more readily than changes in package size. Standardized unit pricing addresses that problem more directly than controls on product design.

Aljoscha Janssen and Johannes Kasinger provide the most comprehensive evidence to date. They analysed a decade of U.S. store-level scanner data and found that downsizing was more than five times as common as upsizing on a sales-weighted basis. Size reductions “usually occur with little or no accompanying price changes,” while the price per unit of volume averaged roughly 12% more during the year after a reduction.[105] A shopper can accurately read every posted price and still miss that a product has become more expensive per gram.

That evidence does not prove market power or “gouging.” Resizing is a predictable response to rising marginal costs, while broad increases in the price level chiefly reflect macroeconomic forces rather than the conduct of one firm. The evidence also does not establish the motive for any particular reduction. Janssen and Kasinger describe downsizing as a way for firms “to increase margins or respond to cost pressures.”[106] The comparability problem, and its appropriate remedy, remain the same in either case.

Existing Canadian cartel law addresses agreements among competitors to reduce package size or quality.[107] The best-documented shrinkflation case involved the three largest U.S. canned-tuna producers, which agreed to reduce the standard can size from six ounces to five ounces beginning in 2008. The U.S. Department of Justice brought criminal charges. Two of the three producers had pleaded guilty by 2018, and a chief executive received a prison sentence.[108]

Skimpflation presents the same comparison problem but is even less visible. Consumers may struggle to detect a quality change at the shelf, and unit pricing cannot capture it. The systematic evidence remains thin. The term arose in popular reporting about declining service quality, and grocery products have no evidence base comparable to the shrinkflation literature.

Existing safeguards address the clearest harms. Deceptive-marketing law reaches false or misleading claims about quality and net quantity, while repeat purchasing disciplines quality in products consumers buy weekly. Treating every reformulation as suspect would instead subject routine cost-saving product changes and genuine quality improvements to regulatory review.

The remedy should match the conduct: standardized unit pricing for hard-to-see quantity changes, deceptive-marketing enforcement for false claims, and cartel law for agreements among competitors. Product-design controls would address none of those problems as precisely.

VII. Sub-Question 6: Measures to Improve Transparency and Competition

The Bureau asks which measures could improve transparency and competition in retail pricing. The answer depends on the audience. Consumer-facing transparency helps shoppers compare offers and switch retailers. Competitor-facing transparency may expose confidential terms, discourage selective discounts, and facilitate coordination.

We recommend harmonized unit pricing, clear final prices, and legible loyalty terms. The Bureau should reject same-terms mandates, compelled disclosure of confidential contracts, and price or margin controls. Enforcement should target demonstrated agreements, deception, and exclusion while preserving lawful discounting and independent pricing. The most durable reforms operate on the supply side by reducing barriers to entry and expansion, strengthening independent wholesale infrastructure, and widening the range of grocers available to Canadian consumers.

A. Adopt Harmonized Unit Pricing

Our lead recommendation is accessible, harmonized unit pricing across every province and territory. Unit pricing displays the price per standard unit of measure, allowing shoppers to compare a 500-gram package with a 750-gram package at a glance. It addresses loyalty tiers, shrinkflation, package-size differences, and in-store and online comparisons without regulating prices.

This approach also follows the Bureau’s 2023 recommendation. The Bureau noted that only Quebec legally requires unit pricing and urged provincial and territorial governments to develop accessible, harmonized requirements.[109]

The rules should establish standard units for each product category and apply in stores, online, and in advertising. Retailers should display unit prices accurately, legibly, and close to the relevant item. Very small retailers could receive phased implementation or simplified compliance requirements. Unit pricing requires no disclosure of wholesale terms, margins, or supplier contracts. It provides transparency to shoppers without revealing competitively sensitive information to rivals.

Two additional consumer-facing measures follow the same principle. First, retailers should clearly disclose final prices and truthfully present discounts and reference prices—the prices used to frame an advertised saving. This would target misleading “was/now” offers and false scarcity claims without restricting dynamic pricing.

Second, retailers should present loyalty-program terms in plain language. Disclosures should compare the member price with the shelf price and clearly explain points’ redemption value, expiry dates, and material exclusions. We discuss those requirements further in Sub-Question 4.

B. Reject Same-Terms and Price-Control Mandates

The Bureau should reject same-terms and wholesale price-parity mandates. New York Senate Bill 8563, the proposed “Consumer Grocery Pricing Fairness Act,” offers a recent example. It would have required large suppliers to offer identical terms to retailers purchasing the same volume, capped at one truckload, and to disclose anonymized terms upon request.[110]

Such a requirement operates as a statutory most-favoured-nation clause. A most-favoured-nation clause promises one buyer terms at least as favourable as those offered to another. A privately negotiated clause limits a supplier’s ability to offer selective discounts after making that promise. A statutory mandate goes much further by imposing the obligation across the market. If a supplier must extend every discount to all similarly situated buyers, it has less incentive to discount at all. Suppliers will instead tend toward a uniform, higher wholesale price.[111]

The U.S. experience supports this concern. The Robinson-Patman Act discouraged discounting for decades by exposing sellers to litigation over price differences. In 2007, the bipartisan Antitrust Modernization Commission concluded that the law’s chilling effect on discounts likely raised consumer prices above the levels that would otherwise have prevailed.[112]

Supporters of same-terms rules often invoke the “waterbed effect,” under which a discount to a large buyer supposedly forces the supplier to charge smaller buyers more. Roman Inderst and Tommaso Valletti developed the leading model and derived a sufficient condition under which this effect would harm consumers.[113] Recent research questions whether their model supports that conclusion.[114]

Brian Albrecht shows that the condition for consumer harm is never satisfied in any equilibrium generated by Inderst and Valletti’s own model. Their analysis therefore identifies no equilibrium in which the waterbed effect harms consumers. Consumer surplus instead rises at the margin even when the smaller rival pays and charges more.

Empirically, the U.K. Competition Commission found little evidence of a systemic waterbed effect in grocery mergers, and courts have treated the theory skeptically, as in DeHoog v. Anheuser-Busch InBev, where the court noted that antitrust law protects the squeezed supplier, not the large buyer’s competitors.[115]

Any statutory cost-justification defence would offer little practical protection. A supplier would have to determine what portion of each discount reflected cost savings and maintain a continuous, transaction-level record to prove it. Those compliance costs would ultimately pass through to retail prices.[116]

Compelled disclosure of contract terms creates an additional coordination risk. George Stigler showed that the central obstacle to tacit coordination is detecting deviations from a common price. Publishing even anonymized terms lowers that obstacle in a market with only a dozen large buyers. Volumes, shipping patterns, and other details may reveal the parties’ identities, and modern LLMs make reidentification easier.[117] Once every buyer can demand a supplier’s lowest disclosed price, that price becomes a floor rather than a selective discount.

For the same reasons, the Bureau should avoid fee caps, margin caps, and cost-based pricing mandates. Its 2023 study recognizes that the Bureau cannot regulate grocery prices and that price controls are blunt tools because authorities struggle to identify the right price and adjust it as conditions change.[118] The food-delivery market illustrates the risk. A de facto cap can suppress the price signals needed to sustain service in high-cost areas and prompt providers to withdraw from them.[119]

C. Target Demonstrated Competitive Harm

The Bureau should enforce vigorously against conduct shown to cause harm. That includes price-fixing agreements, hub-and-spoke information sharing supported by a horizontal agreement among competitors, exchanges of confidential or forward-looking pricing data, exclusionary conduct that prevents entry or expansion, deceptive discounts or loyalty terms, hidden mandatory fees, and false claims about quantity or quality.

The Bureau should not presume liability from high prices, parallel price movements, ordinary monitoring of public prices, above-cost discounts, nonexclusionary loyalty programs, or algorithmic pricing without evidence of agreement or likely competitive harm. Its existing Competitor Collaboration Guidelines already provide a workable distinction between publicly available information and competitively sensitive information. That distinction should govern here.[120]

D. Promote Entry and Expansion

The durable way to discipline grocery prices is through greater entry and expansion. The most effective tools operate on the supply side: zoning and land-use reform, limits on restrictive property covenants, stronger wholesale and distribution infrastructure for independents, fewer interprovincial trade barriers, and support for independent, international, discount, and online grocers. The Bureau’s 2023 market study recommends several of these measures, including support for international and discount entrants and limits on property controls that impede competition.[121]

Wholesale-price regulation poorly addresses the food-access problem it often seeks to solve. U.S. evidence illustrates the mismatch. Hunt Allcott and his co-authors estimate that equalizing access and prices between low- and high-income households would close only about 10% of the gap in nutritional quality. Differences in demand account for the remaining 90%.[122] Zoning reform, targeted incentives, and transit investments address barriers to grocery access more directly than rules governing supplier prices.

VIII. Conclusion

The retail-pricing positions above follow the framework ICLE has applied across its submissions to the Bureau. Our algorithmic-pricing comments urged effects-based analysis rather than categorical rules. Our comments on the draft conduct guidelines emphasized asymmetric error costs, the uncertainty created by the excessive-pricing provision, and the need for workable market-power screens. Our merger comments treated structural presumptions as screening tools rather than verdicts. The common principles are demonstrated consumer harm, administrable standards, and recognition that false positives can prove especially costly in a low-margin industry.

The evidence reviewed here points in the same direction. Pricing algorithms usually automate lawful conduct. Dynamic pricing can expand output and reduce shortages, while personalized and loyalty pricing can lower effective prices and intensify competition. Competitive harm arises through narrower mechanisms, including agreements to pool confidential data, coordinated pricing, deception, and exclusion. High prices, parallel price movements, common software, or price differences alone do not establish that harm.

The Bureau can promote transparency without becoming a price regulator. Consumer-facing measures such as harmonized unit pricing, clear final prices, and legible loyalty terms help shoppers compare offers and switch retailers. They also address the comparability problems created by shrinkflation and complex pricing. Competitor-facing disclosure can instead expose sensitive terms, facilitate coordination, weaken investment incentives, and burden smaller firms disproportionately. Same-terms mandates, categorical presumptions against algorithmic pricing, and fee or margin caps carry similar risks.

Canadian consumers already compare prices, shop across formats, and switch in response to better offers. The more durable challenge lies on the supply side. Reforms that ease entry and expansion, limit restrictive property controls, improve wholesale access for independents, reduce interprovincial trade barriers, and support new retail formats will strengthen competition more effectively than controls on pricing practices.

The Bureau can best protect Canadian shoppers by making comparison easier, entry more feasible, and demonstrated coordination or exclusion riskier. We thank the Bureau for the opportunity to comment and welcome the chance to provide further analysis on any issue raised here.

[1] Competition Bureau Can., Notice: Behind the Price Tag: Examining the Food Supply Chain (June 16, 2026), https://competition-bureau.canada.ca/en/how-we-foster-competition/notice-behind-price-tag-examining-food-supply-chain [hereinafter Notice].

[2] Geoffrey A. Manne et al., Comments on Canadian Competition Bureau Discussion Paper on Algorithmic Pricing and Competition (Int’l Ctr. for Law & Econ., Aug. 4, 2025), https://laweconcenter.org/wp-content/uploads/2025/08/Canada-Algorithmic-pricing-and-competition-1.pdf (included as Appendix A).

[3] Ian Adams et al., Comments of the International Center for Law & Economics to the Competition Bureau Canada: Anti-Competitive Conduct and Agreements (Jan. 27, 2026), https://laweconcenter.org/wp-content/uploads/2026/01/Competition-Bureau-Canada-Comments-2026.pdf.

[4] Ian Adams et al., Comments of the International Center for Law & Economics to the Competition Bureau Canada: Proposed Merger Enforcement Guidelines (Feb. 10, 2026), https://laweconcenter.org/wp-content/uploads/2026/02/Competition-Bureau-Canada-Merger-Comments-2026.pdf.

[5] Geoffrey A. Manne et al., Comments from the International Center for Law and Economics on the Future of Competition Policy in Canada (Mar. 31, 2023), https://laweconcenter.org/wp-content/uploads/2023/03/ICLE-Canada-Comments.pdf.

[6] Brian C. Albrecht et al., Food-Retail Competition, Antitrust Law, and the Kroger/Albertsons Merger (Int’l Ctr. for Law & Econ., White Paper No. 2023-10-17, 2023), https://laweconcenter.org/wp-content/uploads/2023/10/Kroger-Albertsons-Merger-Full-Paper.pdf.

[7] Eric Fruits, The Waterbed Effect and the Price of “Pricing Fairness” (Int’l Ctr. for Law & Econ., Issue Brief No. 2026-03-17, 2026), https://laweconcenter.org/wp-content/uploads/2026/03/Grocery-Pricing-Legislation-2026.pdf.

[8] Brian Albrecht et al., Comments of the International Center for Law & Economics: Rule on Unfair or Deceptive Fees in Online Food Delivery Services, ANPRM, Federal Trade Commission Project No. P267101 (May 18, 2026), https://laweconcenter.org/wp-content/uploads/2026/05/FTC-Delivery-Fees-2026.pdf.

[9] Stat. Can., Consumer Price Index, February 2026 (Mar. 16, 2026), https://www150.statcan.gc.ca/n1/daily-quotidien/260316/dq260316a-eng.htm (reporting that grocery prices rose 30.1% from February 2021 to February 2026).

[10] Competition Bureau Can., Canada Needs More Grocery Competition: Retail Grocery Market Study Report 44 (2023), https://competition-bureau.canada.ca/sites/default/files/attachments/2023/CB-Retail-Grocery-Market-Study-Report-EN-2023-06-23.pdf.

[11] See Frank H. Easterbrook, The Limits of Antitrust, 63 Tex. L. Rev. 1 (1984).

[12] Competition Act, R.S.C. 1985, c C-34, § 78(1)(k) (Can.). See also An Act to Amend the Excise Tax Act and the Competition Act, S.C. 2023, c 31, pt 2 (Can.) [hereinafter Bill C-56].

[13] Consolidated Version of the Treaty on the Functioning of the European Union art. 102, May 9, 2008, 2008 O.J. (C 115) 47.

[14] Case 27/76, United Brands Co. v. Comm’n, ECLI:EU:C:1978:22, ¶ 252 (Feb. 14, 1978) (“The questions therefore to be determined are whether the difference between the costs actually incurred and the price actually charged is excessive, and, [if] the answer to this question is in the affirmative, whether a price has been imposed which is either unfair in itself or when compared to competing products.”).

[15] See, e.g., Aline Blankertz et al., The European Commission Can and Must Act on Excessive Pricing, ProMarket (Oct. 8, 2025), https://www.promarket.org/2025/10/08/the-european-commission-can-and-must-act-on-excessive-pricing.

[16] See Org. for Econ. Coop. & Dev., Excessive Prices, OECD Doc. DAF/COMP(2011)18 (Feb. 7, 2012), https://www.oecd.org/content/dam/oecd/en/publications/reports/2012/02/excessive-prices_372e092c/8e1fd82e-en.pdf; Raphaël De Coninck, Excessive Prices: An Overview of EU and National Case Law, Concurrences e-Competitions Bull. (June 2018), https://ecp.crai.com/wp-content/uploads/2018/06/Excessive-Pricing-R.-de-Coninck-e-Competitions-2018.pdf.

[17] David S. Evans & A. Jorge Padilla, Excessive Prices: Using Economics to Define Administrable Legal Rules, 1 J. Competition L. & Econ. 97 (2005).

[18] Massimo Motta & Alexandre de Streel, Excessive Pricing and Price Squeeze Under EU Law, in The Pros and Cons of High Prices 14 (Konkurrensverket ed., 2007), https://www.konkurrensverket.se/globalassets/dokument/informationsmaterial/rapporter-och-broschyrer/pros-and-cons/rapport_pros-and-cons_2007_high_prices.pdf.

[19] Lars-Hendrik Röller, Exploitative Abuses (ESMT Bus. Brief No. BB-107-002, 2007), https://d-nb.info/1012903311/34.

[20] Competition Act, R.S.C. 1985, c C-34, § 103.1(7) (Can.); see also Competition Bureau Can., Information Bulletin on Private Access to the Competition Tribunal (June 2025), https://competition-bureau.canada.ca/en/information-bulletin-private-access-competition-tribunal.

[21] Brunswick Corp. v. Pueblo Bowl-O-Mat, Inc., 429 U.S. 477, 489 (1977) (holding that antitrust injury must be “of the type the antitrust laws were intended to prevent” and “flow[] from that which makes defendants’ acts unlawful”).

[22] See George J. Stigler, A Theory of Oligopoly, 72 J. Pol. Econ. 44 (1964).

[23] Maureen K. Ohlhausen, Acting Chairman, Fed. Trade Comm’n, Should We Fear the Things That Go Beep in the Night? Some Initial Thoughts on the Intersection of Antitrust Laws and Algorithmic Pricing 10 (May 23, 2017), https://www.ftc.gov/system/files/documents/public_statements/1220893/ohlhausen_-_concurrences_5-23-17.pdf.

[24] Zach Y. Brown & Alexander MacKay, Competition in Pricing Algorithms, 15 Am. Econ. J.: Microeconomics 109 (2023), https://www.aeaweb.org/articles?id=10.1257/mic.20210158.

[25] Sophie Calder-Wang & Gi Heung Kim, Algorithmic Pricing in Multifamily Rentals: Efficiency Gains or Price Coordination? (Aug. 16, 2024) (unpublished manuscript), https://ssrn.com/abstract=4403058.

[26] Stephanie Assad et al., Algorithmic Pricing and Competition: Empirical Evidence from the German Retail Gasoline Market, 132 J. Pol. Econ. 723 (2024).

[27] Id. at 726 (finding that algorithmic adoption increased margins by roughly 15% overall and had no meaningful effect in monopoly markets, but increased margins by 38% in duopoly markets where both stations adopted).

[28] Id. at 727 (“Margins start to increase only about a year after market-wide adoption, suggesting that algorithms in this market learn tacitly collusive strategies. . . . [A]lgorithms learn that undercutting will not be profitable, since lower prices will be followed.”).

[29] Id. at 760 (“To the best of our knowledge, this occurs without explicit communication between competitors, making it legal according to current competition laws in many countries.”).

[30] Competition Bureau Can., Anti-Competitive Conduct and Agreements Enforcement Guidelines ¶ 121 (proposed Oct. 31, 2025), https://competition-bureau.canada.ca/en/how-we-foster-competition/consultations/anti-competitive-conduct-and-agreements.

[31] Gibson v. Cendyn Grp., LLC, 148 F.4th 1069 (9th Cir. 2025).

[32] Id.

[33] Cornish-Adebiyi v. Caesars Entm’t, Inc., No. 24-3006, slip op. (3d Cir. July 29, 2026), rev’g 2024 WL 4356188 (D.N.J. Sept. 30, 2024).

[34] See Ken Littlewood, Forecasting and Control of Passenger Bookings, 4 J. Revenue & Pricing Mgmt. 111 (2005) (reprinting a 1972 AGIFORS paper); Barry C. Smith et al., Yield Management at American Airlines, 22 Interfaces 8 (1992).

[35] See Kevin R. Williams, The Welfare Effects of Dynamic Pricing: Evidence from Airline Markets, 90 Econometrica 831 (2022).

[36] See Juan Camilo Castillo, Who Benefits from Surge Pricing?, 93 Econometrica 1811 (2025).

[37] See Andrew Sweeting, Dynamic Pricing Behavior in Perishable Goods Markets: Evidence from Secondary Markets for Major League Baseball Tickets, 120 J. Pol. Econ. 1133 (2012).

[38] See Frank A. Wolak, Do Residential Customers Respond to Hourly Prices? Evidence from a Dynamic Pricing Experiment, 101 Am. Econ. Rev. Papers & Proc. 83, 87 (2011).

[39] See Brian C. Albrecht, Price Competition and the Use of Consumer Data (Working Paper, June 12, 2026), https://briancalbrecht.com/papers/price-competition-data.

[40] See Manne et al., supra note 2, at 11–12 (discussing Nicholas Buchholz et al., Personalized Pricing and the Value of Time: Evidence from Auctioned Cab Rides, 93 Econometrica 929, 930 (2025)); see also John M. Yun, Should We Fear Personalized Pricing?, CPI Antitrust Chron. (July 2025).

[41] See James D. Dana Jr., Advance-Purchase Discounts and Price Discrimination in Competitive Markets, 106 J. Pol. Econ. 395, 395 (1998).

[42] See Joanna Stavins, Price Discrimination in the Airline Market: The Effect of Market Concentration, 83 Rev. Econ. & Stat. 200, 202 (2001).

[43] See Severin Borenstein & Nancy L. Rose, Competition and Price Dispersion in the U.S. Airline Industry, 102 J. Pol. Econ. 653, 672 (1994).

[44] See Ambarish Chandra & Mara Lederman, Revisiting the Relationship Between Competition and Price Discrimination, 10 Am. Econ. J.: Microeconomics 190, 193, 198, 210 (2018).

[45] See Retail Council of Can., The Great Canadian Shopping Trip (2019); see infra Sub-Question 4.

[46] See Eur. Comm’n, Consumer Market Study on Online Market Segmentation Through Personalised Pricing/Offers in the European Union 171 (2018).

[47] See Manne et al., supra note 2, at 9 (quoting OECD); Yun, supra note 40.

[48] See Katie Wells et al., Same Cart, Different Price: Instacart’s Price Experiments Cost Families at Checkout (Groundwork Collaborative & Consumer Reports, Dec. 9, 2025), https://groundworkcollaborative.org/wp-content/uploads/2025/12/Same-Cart-Different-Price.pdf.

[49] See Brian Albrecht, Instacart Didn’t Read Your Mind. It Ran an A/B Test., Truth on the Mkt. (Dec. 10, 2025), https://truthonthemarket.com/2025/12/10/instacart-didnt-read-your-mind-it-ran-an-a-b-test.

[50] See Ginger Zhe Jin et al., Surveillance Pricing: A Cautionary Summary of Potential Harms and Solutions, Competition Pol’y Int’l (July 14, 2025).

[51] See H.B. 895, 2026 Gen. Assemb., Reg. Sess. (Md. 2026) (enacted as 2026 Md. Laws ch. 154) (effective Oct. 1, 2026); Brian Albrecht, How a Panic over Egg Prices Hatched a Bad Law, Wash. Post (May 11, 2026); see infra Sub-Question 3.

[52] See Brian Albrecht, Eliminate ‘Personal Pricing’ and You Risk Harming Consumers, Fin. Times (July 2, 2026).

[53] See U.K. Competition & Mkts. Auth., Loyalty Pricing in the Groceries Sector—Summary (Nov. 27, 2024), https://assets.publishing.service.gov.uk/media/6749f28975bb645366b3a2a9/Executive_summary__2.pdf; U.K. Competition & Mkts. Auth., Loyalty Pricing in the Groceries Sector—Findings Report (Nov. 2024).

[54] See Borenstein & Rose, supra note 43.

[55] See Competition Bureau Can., Competitor Collaboration Guidelines § 1.1 (May 6, 2021), https://competition-bureau.canada.ca/sites/default/files/attachments/2022/CB-BC-CCGs-Eng.pdf.

[56] See id. § 3.7.2, at 42; Competition Bureau Can., supra note 30, ¶ 272.

[57] See Competition Bureau Can., supra note 55, § 2.2.

[58] See Competition Act, R.S.C. 1985, c C-34, §§ 78(1)(h)–(i), 79(1)(b)(ii) (Can.).

[59] See Competition Bureau Can., supra note 30, ¶¶ 171–176.

[60] Gibson, 148 F.4th at 1076, 1083 & n.8, 1088 n.12.

[61] See Cornish-Adebiyi v. Caesars Entm’t, Inc., No. 1:23-cv-02536-KMW-EAP, 2024 WL 4356188 (D.N.J. Sept. 30, 2024), rev’d, No. 24-3006, slip op. (3d Cir. July 29, 2026); Cornish-Adebiyi, No. 24-3006, slip op. at 9, 36–38, 42; In re RealPage, Inc., 709 F. Supp. 3d 478, 494, 512 (M.D. Tenn. 2023).

[62] See Competition Act, R.S.C. 1985, c C-34, § 79(1)(b)(ii) (Can.).

[63] See Competition Bureau Can., supra note 30, ¶ 236.

[64] See Br. of Amici Curiae Int’l Ctr. for Law & Econ. et al. at 2, In re EpiPen (Sanofi-Aventis U.S., LLC v. Mylan Inc.), No. 21-3005 (10th Cir. Sept. 22, 2021).

[65] See id. at 15.

[66] Org. for Econ. Coop. & Dev., Algorithmic Competition: Background Note, OECD Doc. DAF/COMP(2023)3 (June 14, 2023), https://one.oecd.org/document/DAF/COMP(2023)3/en/pdf.

[67] Emek Basker, The Causes and Consequences of Wal-Mart’s Growth, 21 J. Econ. Persp. 177, 188–89 (2007).

[68] Id. at 188.

[69] Jerry Hausman & Ephraim Leibtag, Consumer Benefits from Increased Competition in Shopping Outlets: Measuring the Effect of Wal-Mart, 22 J. Applied Econometrics 1157 (2007).

[70] Fruits, supra note 7.

[71] See Stefano DellaVigna & Matthew Gentzkow, Uniform Pricing in U.S. Retail Chains, 134 Q.J. Econ. 2011 (2019).

[72] See Alberto Cavallo, Are Online and Offline Prices Similar? Evidence from Large Multi-Channel Retailers, 107 Am. Econ. Rev. 283, 290, 292 (2017).

[73] Albrecht et al., supra note 8.

[74] Competition Bureau Can., Algorithmic Pricing and Competition: Discussion Paper (June 10, 2025), https://competition-bureau.canada.ca/en/how-we-foster-competition/education-and-outreach/publications/algorithmic-pricing-and-competition-discussion-paper. See infra Appendix A.

[75] See infra Appendix A.

[76] Competition Bureau Can., supra note 55, at 42–43.

[77] Org. for Econ. Coop. & Dev., supra note 66.

[78] Neth. Auth. for Consumers & Mkts., Position Paper: Oversight of Algorithms (2020), https://www.acm.nl/sites/default/files/documents/position-paper-oversight-of-algorithms.pdf.

[79] Eur. Comm’n, Commission Staff Working Document Accompanying the Final Report on the E-Commerce Sector Inquiry 17, 22, 24, 29–32, SWD(2017) 154 final (May 10, 2017), https://eur-lex.europa.eu/legal-content/EN/ALL/?uri=SWD:2017:154:FIN.

[80] Id.

[81] See Org. for Econ. Coop. & Dev., supra note 66, at 12. In 2020, the Norwegian Competition Authority found that 55% of surveyed firms used monitoring algorithms and 20% used pricing algorithms. In Denmark, 17% of e-commerce companies used pricing algorithms, with automation ranging from basic information provision to full algorithmic control. The Netherlands Authority for Consumers and Markets reported that 36% of firms used competitors’ pricing data. Of those firms, 16%—or 6% of all firms—used pricing algorithms.

[82] Albrecht et al. supra note 8.

[83] Albrecht et al., supra note 6.

[84] Id.

[85] See, e.g., Simon Lermen et al., Large-Scale Online Deanonymization with LLMs, arXiv:2602.16800 (Feb. 18, 2026), https://doi.org/10.48550/arXiv.2602.16800.

[86] Stigler, supra note 22; Fruits, supra note 7. See infra Sub-Question 6 for further discussion of compelled disclosure.

[87] Wells et al., supra note 48.

[88] Albrecht, supra note 49.

[89] H.B. 895, 2026 Gen. Assemb., Reg. Sess. (Md. 2026) (enacted as 2026 Md. Laws ch. 154) (effective Oct. 1, 2026).

[90] Albrecht, supra note 51.

[91] See Org. for Econ. Coop. & Dev., supra note 66 (reporting the European Commission’s mystery-shopping exercise); see also infra Appendix A.

[92] Estimates of the combined market share vary by scope and source. See U.S. Dep’t of Agric., Foreign Agric. Serv., Retail Foods Annual: Canada, GAIN Rep. No. CA2025-0040 (2025), https://www.fas.usda.gov/data/gain-report/2025/12/Retail%20Foods%20Annual_Ottawa_Canada_CA2025-0040.pdf (estimating that five leading retailers held more than 65% of the market in 2024: Loblaws 17%, Costco 15%, Sobeys 14%, Walmart 13%, and Metro 6%); Kevin Grier, Grocery Retail (UFCW Loc. 401, Nov. 2024), https://gounion.ca/wp-content/uploads/2024/11/UFCW-Local-401-Retail-Grier-November-2024-rev.pdf (estimating 2022 shares of 26% for Loblaws, 20% for Sobeys, 14% for Costco, and 12% each for Walmart and Metro).

[93] Competition Bureau Can., supra note 10.

[94] Id.

[95] Karl Littler, Retail Council of Can., Retail Council of Canada Submission to the Competition Bureau’s Market Study: Competition in Canada’s Grocery Sector (Dec. 16, 2022), https://competition-bureau.canada.ca/sites/default/files/attachments/2023/RCC-Market-Study-Submission.pdf (arguing that concentration had not recently increased, big-box retailers and new entrants had gained market share, regional and independent grocers remained significant, and 87% of Canadians regularly shopped at discount grocers).

[96] Ritika Dubey, Discount Is the New Name of the Game. What Does That Mean for Conventional Stores?, Can. Press (Apr. 3, 2026), https://www.thecanadianpressnews.ca/business/discount-is-the-new-name-of-the-game-what-does-that-mean-for-conventional-stores/article_50f71fe5-5410-5c00-ab63-b5d0655397f6.html (quoting Henry Chamber of Sentinel Management Consultants on grocers’ efforts to retain shoppers across their discount and conventional banners); Rebecca Harris, Retailers Are Betting on Discount. Will the Momentum Last?, Can. Grocer (Dec. 2, 2024), https://canadiangrocer.com/retailers-are-betting-discount-will-momentum-last (reporting that more than 62% of grocery shoppers had changed their primary store to secure better deals and nearly 30% chose stores based on in-store discounts and promotions).

[97] Competition Bureau Can., supra note 10.

[98] Id.

[99] Angela Altass, Budget-Conscious Shoppers Drive Increased Growth in Discount Grocery Channel, Grocery Bus. (July 9, 2025) (reporting that discount grocers’ market share rose from 35% in 2008 to roughly 50% and that their same-store sales outpaced conventional grocers 3-to-1 nationally by both dollars and tonnage).

[100] Competition Bureau Can., supra note 10.

[101] Id.

[102] Id.

[103] Id.

[104] U.K. Competition & Mkts. Auth., Loyalty Pricing in the Groceries Sector—Summary, supra note 53 (reporting that 92% of roughly 50,000 products offered through dual-price loyalty promotions yielded genuine savings, with average member savings of 17% to 25%).

[105] See Aljoscha Janssen & Johannes Kasinger, Shrinkflation and Consumer Demand, 45 Mktg. Sci. 142 (2026).

[106] Id. at 142.

[107] Competition Act, R.S.C. 1985, c C-34, § 45(1)(c) (Can.); Competition Bureau Can., supra note 55, § 2.4.3.

[108] See Xibo Wan et al., Shrinkflation, Unit Price Disclosure, and Consumer Welfare: Evidence from Canned Tuna, 142 Food Pol’y 103106 (2026); Press Release, U.S. Dep’t of Just., Former Bumble Bee CEO Sentenced to Prison for Fixing Prices of Canned Tuna (June 16, 2020), https://www.justice.gov/archives/opa/pr/former-bumble-bee-ceo-sentenced-prison-fixing-prices-canned-tuna.

[109] Competition Bureau Can., supra note 10, at 44.

[110] See S. 8563, 2025–2026 Reg. Sess. (N.Y. 2025).

[111] Fruits, supra note 7.

[112] See Antitrust Modernization Comm’n, Report and Recommendations (2007).

[113] Roman Inderst & Tommaso M. Valletti, Buyer Power and the “Waterbed Effect”, 59 J. Indus. Econ. 1 (2011).

[114] See Brian C. Albrecht, Consumer Harm from the Waterbed Effect: A Comment on Inderst and Valletti (2011) (Working Paper, Apr. 22, 2026), https://briancalbrecht.com/Albrecht-IV-Waterbed-Comment.pdf; Brian C. Albrecht, When Does the Waterbed Effect Harm Consumers? (Int’l Ctr. for Law & Econ., July 24, 2026), https://laweconcenter.org/resources/when-does-the-waterbed-effect-harm-consumers.

[115] DeHoog v. Anheuser-Busch InBev SA/NV, No. 1:15-cv-02250-CL, 2016 U.S. Dist. LEXIS 137759 (D. Or. July 22, 2016).

[116] Fruits, supra note 7.

[117] Stigler, supra note 22; Fruits, supra note 7.

[118] Competition Bureau Can., supra note 10.

[119] Albrecht et al., supra note 8.

[120] Competition Bureau Can., supra note 55.

[121] See Competition Bureau Can., supra note 10, recommendations 2 and 4.

[122] Hunt Allcott et al., Food Deserts and the Causes of Nutritional Inequality, 134 Q.J. Econ. 1793 (2019).

I. Introduction and Overview

The International Center for Law & Economics (ICLE) submits these comments in response to the Federal Communications Commission’s (FCC) notice of proposed rulemaking on modernizing the high-cost program.[1] ICLE is a nonprofit, nonpartisan research center that applies law & economics to public policy. Its work promotes sound economic analysis and consumer welfare, particularly in dynamic, technology-driven markets such as telecommunications.

The Commission asks what should become of three legacy support mechanisms: Connect America Fund Broadband Loop Support (CAF BLS), High-Cost Loop Support (HCLS), and the expiring Alternative Connect America Cost Model (A-CAM) programs. The shift to Internet Protocol (IP)-based networks has sharply reduced their relevance. Unsubsidized competitors now offer qualifying service to most locations these programs support, while low-Earth-orbit (LEO) satellites can reach nearly all the rest.[2] Congress also has committed $42.5 billion through the Broadband Equity, Access, and Deployment (BEAD) Program to serve remaining unconnected areas. Meanwhile, lower-cost IP infrastructure continues to displace the copper and time-division multiplexing (TDM) networks that these legacy subsidies sustain.[3]

The Commission should wind down CAF BLS and HCLS and allow the A-CAM programs to expire on schedule, without extension or replacement. Doing so would reduce annual high-cost spending by roughly $1.6 billion. Competition, technological advances, and other federal programs have eliminated the need for much of that support.

This proceeding must weigh the benefits of continued support against the programs’ full costs, rather than cataloging benefits alone. Consumers fund every dollar through a surcharge on their bills that has climbed to 38.8% of assessable revenue.[4] Support for areas the market can serve more efficiently also discourages the private investment and competition that could serve those areas without subsidies.

Against those mounting costs, the programs’ benefits have narrowed to a small and shrinking residual. Support is redundant where competitors already provide service, duplicative where other federal programs are funding new networks, and concentrated in legacy voice service that consumers continue to abandon.

The principal objections in the record do not justify preserving these programs. LEO satellite service provides a functional substitute for legacy service in precisely the sparse, high-cost areas at issue, where capacity constraints are least likely to bind.[5] Claims that ongoing operating costs require perpetual support fare no better. They mistake the cost of maintaining obsolete infrastructure for an unavoidable cost of serving rural communities.

II. Continued Support Must Be Judged by Its Current Costs

Every subsidy entails a tradeoff. The central question here is not whether high-cost support has produced benefits, but whether those benefits still justify the costs of continuing it.[6] A subsidy may make economic sense when a genuine market failure prevents private investment from reaching customers who value service above its cost.[7] Without such a failure, subsidies distort investment, encourage rent-seeking, and often miss their intended beneficiaries.[8]

Subsidies also replace the decentralized decisions of consumers and investors with an administrative judgment about who should build and operate a network. The government finances that intervention by taxing the same market activity that could support further expansion. The FCC should therefore weigh the benefits of continued support against the deployment, competition, and consumer welfare that those subsidies crowd out.

Commenters seeking to preserve the programs identify real benefits, but they define the need for support by reference to network costs rather than a continuing market failure.[9] NTCA–The Rural Broadband Association (NTCA) calls high-cost support an “ideal public-private partnership” that must continue so carriers can “maintain service, repay loans, and keep rates affordable.”[10] The National Rural Electric Cooperative Association (NRECA) asks the Commission to create a “new program to support ongoing operating costs.”[11] The Blooston Rural Carriers urge the Commission to measure sufficiency against “the cost of operating a network, and not just the cost of building one.”[12]

Each proposal severs support from the condition that justifies it. Every network has operating costs, and every carrier would prefer a subsidy. If operating costs alone warrant support, the subsidy has no logical endpoint. A program designed to correct a market failure instead becomes a permanent entitlement, regardless of whether the failure persists.

Support also can suppress market expansion once unsubsidized providers begin reaching an area. A subsidized incumbent with assured cost recovery competes on terms that an unsubsidized entrant cannot match. Investors respond by deploying capital elsewhere.[13] Each additional year of support in an area open to competition deters entry, directs capital toward incumbents rather than expansion, and requires consumers to fund a transfer whose original justification may have disappeared.

That asymmetry makes a simple catalog of benefits inadequate. A subsidy’s benefits are concentrated and readily observed. Its costs are dispersed and harder to measure: surcharges on consumers, networks that competitors never build, capital unavailable for more productive investments, and the lower prices and better service that foreclosed competition might have delivered. An evaluation that counts only what the subsidy sustains will systematically favor continued support because the investment it discourages never appears in the record. Sound economic analysis must account for both.

The Commission should therefore ask whether the market failure that once justified support still exists in each area receiving it. Where no such failure remains, continued support sacrifices deployment, competition, and consumer gains that the market could otherwise deliver—and charges consumers to do so. The Commission must evaluate that present tradeoff, not the undisputed value of support provided in the past.

III. Legacy Support No Longer Justifies Its Costs

The Communications Act incorporates the same economic principle. Section 254 requires universal-service support to be “sufficient,” while the Commission has long treated “minimizing the universal service contribution burden on consumers and businesses” as a goal equal to expanding service.[14] The Commission must therefore weigh benefits against costs, not examine benefits in isolation.

For each legacy high-cost mechanism, the relevant question is whether continued support produces service that the market would not otherwise provide. That benefit must justify both the surcharges imposed on consumers and the harm to competition and private investment. The answer increasingly is no. Competitors and new technologies already serve many supported locations, other federal programs fund deployment in many others, and some legacy support sustains voice and copper infrastructure that consumers and carriers are leaving behind.

The appropriate transition should reflect each program’s structure and recipients’ reasonable expectations. The Commission should promptly wind down the open-ended HCLS and CAF BLS mechanisms, which impose no prospective deployment obligations. It should allow the fixed-term A-CAM programs to expire as scheduled without extension or replacement. That approach would lower consumer costs, remove barriers to private investment, encourage modern network deployment, and end legacy support on a predictable schedule.

A. Legacy Support Raises Costs and Distorts Competition

Extending legacy programs beyond their useful lives imposes two principal costs. Consumers bear the first. The Commission finances the high-cost program through universal-service contributions assessed on interstate and international telecommunications revenue. Because providers generally recover those contributions through surcharges on customers’ bills, preserving unnecessary support increases the fees consumers pay.[15]

The market bears the second cost. Subsidizing incumbents in areas that competitors and new technologies can serve discourages the entry and private investment that could deliver service without public support.[16] A subsidy intended to correct a market failure can thereby suppress the competition that would eliminate the need for it.

1. Unnecessary Support Raises Consumers’ Bills

Mandatory contributions from telecommunications providers fund the Universal Service Fund (USF). Providers generally recover those contributions through line-item surcharges on customers’ bills. As ICLE has explained, the contribution factor is “essentially a tax on consumer phone bills” and, at its current level, “has become one of the largest hidden taxes on American telecommunications consumers.”[17] The factor reached 38.8% in the third quarter of 2026, up from 37% in the second quarter. That rate approaches two-fifths of assessable interstate and international telecommunications revenue.[18]

The contribution factor rises when required disbursements increase or the assessable revenue base contracts. That base consists largely of legacy interstate telecommunications revenue, which has declined for two decades. If the Fund’s obligations remain fixed or increase while assessable revenue falls, the contribution factor must continue rising.

The Commission has two ways to relieve that pressure. It can change how it raises money through contribution reform, or it can reduce how much the Fund spends through distribution reform. Reducing unnecessary disbursements offers the more direct and less distortionary approach. It eases pressure on the shrinking contribution base without extending assessments to additional services or providers.[19]

Reducing disbursements also may lessen demands for contribution reforms that would burden consumers elsewhere. Proposals to broaden the assessment base often target broadband service or large online content and application providers, sometimes called edge providers.[20] Such reforms would relocate the assessment rather than reduce it, imposing new costs on the broadband services and internet applications that consumers increasingly use.

The Commission should therefore address unnecessary distributions before expanding the contribution base. Contribution reform undertaken first would preserve a large Fund and spread its costs across more services, entrenching the underlying expense. Phasing out legacy high-cost support would instead reduce the Fund’s obligations, ease consumers’ burden, and diminish pressure to impose new assessments—all “without imposing new burdens on the broadband networks and services that consumers and the broader economy increasingly rely upon.”[21] Cutting unnecessary disbursements reduces consumers’ costs. Expanding the contribution base merely changes where those costs appear.

2. Legacy Subsidies Deter Competition and Investment

Broadband support should complement private investment. The Commission therefore should avoid subsidizing areas that private providers already serve or would serve without support.[22] Subsidies in those areas can encourage duplicative construction, depress private returns, and redirect capital away from network expansion.[23] The resulting harm continues as long as public support distorts investment decisions.

A provider assured of federal cost recovery can compete on terms that unsubsidized rivals cannot match. Investors respond by deploying capital in markets where they do not face subsidized competition.[24] Support in areas already served by unsubsidized providers therefore distorts competition and weakens private providers’ incentives to invest, potentially reducing total investment through crowding out.[25]

The deterred competitor need not already serve the area when the Commission calculates support. The prospect of competing against a subsidized incumbent can prevent entry before it occurs.[26] The subsidy may then appear justified simply because no competitor has entered, even though the subsidy itself helped keep that competitor away.

The Commission should protect competition, not particular competitors. Universal service aims to ensure that consumers receive service, not to preserve any incumbent’s revenue or market position. Yet commenters seeking to retain the legacy programs often frame support as necessary specifically to sustain particular carriers and networks.[27] That approach elevates the recipient over the statutory objective. Where the market can provide service, protecting an incumbent from competition undermines the competitive process. Subsidized overbuilding also can allow favored providers to benefit from infrastructure investments that unsubsidized rivals have already made.

LEO satellite broadband illustrates the market’s ability to reach even the most difficult locations. LEO service is “uniquely equipped to reach many of the remaining unserved areas in the United States, especially rural, remote, and other hard-to-reach locations that are difficult to service through traditional terrestrial means.”[28] Those are precisely the locations once assumed to depend exclusively on subsidized incumbents.

LEO service also “has dramatically expanded the geographic reach of high-speed internet access.” Starlink has been “made available to all locations in the United States,” while satellite and other technologies “now offer robust competition in areas previously served by, at most, one or two fixed-broadband providers.”[29] The Notice confirms that nearly all supported locations lacking a terrestrial connection at 100 megabits per second downstream and 20 megabits per second upstream—known as 100/20 Mbps service—already appear served by an LEO provider at that speed.[30] The Commission’s latest Section 706 Report likewise finds that LEO service meets the 100/20 Mbps benchmark for advanced telecommunications capability and reaches 99.7% of Americans.[31]

That expansion occurred without high-cost support. LEO investment represents the unsubsidized entry that continued support for legacy incumbents may deter. The private sector has already reached these markets. Continuing to subsidize incumbents against that competition would charge consumers to weaken the very entry that is closing the digital divide.

B. Legacy Support Delivers Diminishing Benefits

A subsidy’s benefit is the additional service that exists because of the subsidy and would not exist without it.[32] By that measure, these legacy programs deliver small and shrinking benefits because modern IP networks already serve many supported locations. NCTA–The Internet & Television Association (NCTA) notes that recent federal broadband funding generally finances fiber or hybrid networks capable of operating for a decade or longer without further support. Continuing high-cost subsidies in those areas buys little or no additional connectivity.[33] The Information Technology and Innovation Foundation likewise concludes that the programs “have achieved their original objectives” and that market changes have “substantially reduced the policy rationale for continuing them.”[34] Where qualifying service already exists, another subsidy purchases redundancy rather than universal service.

Support can even produce negative benefits when it discourages network upgrades. Subsidies tied to legacy operations encourage carriers to retain TDM and copper infrastructure rather than transition to more efficient IP networks.[35] IP networks offer greater reliability, higher call quality, and more effective robocall mitigation. Delayed transitions leave customers with outdated technology that cannot meet modern needs.[36] Subsidizing aging infrastructure also weakens carriers’ incentives to replace it, raises maintenance costs, and diverts investment from next-generation services.[37] Support intended to benefit rural consumers can therefore leave them on inferior networks longer than market forces would.

Much of this support does not fund future deployment. CAF BLS and HCLS impose no prospective buildout obligations, so their funding secures no new construction. It reimburses costs for networks already built.[38] More than one-quarter of CAF BLS—roughly $250 million—subsidizes traditional voice-only service. All $202 million in HCLS supports legacy voice loops without any deployment obligation.[39]

Consumers are rapidly abandoning those services. Switched-access lines now account for a small fraction of voice connections as households move to mobile service and interconnected Voice over Internet Protocol (VoIP). Mobile voice service is available at essentially every supported location.[40] Continued subsidies for a declining service that consumers are leaving deliver little incremental benefit.

Other federal programs further reduce that benefit. Where the BEAD Program, Enhanced A-CAM, or another enforceable commitment requires deployment of 100/20 Mbps broadband, additional legacy support adds little universal-service value and duplicates public spending.[41] The incremental benefit approaches zero when an unsubsidized competitor already serves a location, a satellite provider offers 100/20 Mbps service, and BEAD will fund another network—yet the cost of the legacy subsidy continues.

These programs now provide benefits that are narrow and declining. Their support is redundant where competitors already serve consumers, duplicative where other programs fund deployment, and concentrated in legacy voice service that consumers are abandoning. At the margin, it also discourages the upgrades that would improve rural service. The record therefore supports a rapid phaseout that frees consumers and carriers from continued investment in networks the market is replacing.

C. Wind Down Support on a Predictable Schedule

The principles above apply to each program under review. The legacy rate-of-return mechanisms—HCLS and CAF BLS—reimburse carriers based on their costs but impose no prospective deployment obligations or end dates. They also create weaker reliance interests than the model-based programs and should be affirmatively wound down. The A-CAM programs already have defined expiration dates. The Commission should let them expire as scheduled and reject proposals to extend or replace them.

Eliminate High-Cost Loop Support first. HCLS presents the clearest case for elimination and should end on the shortest timeline. The program shifts cost recovery from the intrastate to the interstate jurisdiction to subsidize traditional switched-voice loops. It supports no new deployment. Annual support already has fallen from $741 million in 2015 to roughly $202 million as customers migrate to mobile and IP service.[42] HCLS buys no additional broadband, sustains infrastructure that newer IP networks are replacing, and weakens carriers’ incentives to retire aging equipment. It offers no prospective universal-service benefit that justifies preserving it. The Commission should sunset HCLS promptly.

Wind Down Connect America Fund Broadband Loop Support. CAF BLS is the largest program at issue. Although it supports broadband-capable loops as well as voice service, its structure no longer fits its purpose. Its only deployment obligation—providing service at 25 megabits per second downstream and 3 megabits per second upstream, or 25/3 Mbps—came due at the end of 2023 and has been satisfied. Continued support therefore secures no new construction and functions as open-ended cost reimbursement.[43] More than one-quarter of CAF BLS, roughly $250 million, subsidizes voice-only service. A large and increasing share also flows to locations that an unsubsidized competitor already serves.[44]

The Commission should place CAF BLS on a defined, declining schedule. It should first eliminate support for voice-only service and locations served by an unsubsidized competitor or covered by an enforceable deployment commitment. Support would then contract as competition and other federal programs reach additional locations instead of continuing indefinitely by default.

Allow A-CAM I to expire in 2026. Only nine carriers still receive A-CAM I support, totaling roughly $8 million annually, and the program will expire at the end of 2026.[45] The Commission should reject the Notice’s proposal to extend it through 2028 merely to align its expiration date with other A-CAM programs. Administrative symmetry does not justify two additional years of support absent evidence that the subsidy remains necessary. Allowing A-CAM I to expire as scheduled honors the terms its recipients accepted and ends the program as designed.

Allow revised A-CAM I and A-CAM II to expire in 2028 without replacement. Revised A-CAM I and A-CAM II provide roughly $166 million and $218 million, respectively, to 71 carriers each. Both programs will expire at the end of 2028.[46] Recipients accepted fixed support for a defined period in exchange for deployment obligations. The Commission should honor those terms by allowing support to continue through 2028 and then end on the schedule carriers have long expected.

The Commission should reject proposals to replace these programs with a new, ongoing model-based support stream.[47] Such a program would convert fixed-term support into the kind of open-ended entitlement this proceeding should retire. Letting the existing programs expire provides an orderly phaseout, respects legitimate reliance interests, and avoids renewing the underlying subsidies.

This sequence would produce a rapid but orderly wind-down. HCLS would end first because it has the weakest justification. CAF BLS would decline on a defined schedule, and the A-CAM programs would conclude on the dates their recipients already anticipate. Together, these steps would eliminate roughly $1.6 billion in annual support that competition, new technology, and other federal programs have overtaken, without disrupting fixed-term commitments midstream.[48]

IV. LEO Satellite Is a Viable Substitute in High-Cost Areas

Several commenters urge the Commission to disregard satellite broadband because they do not consider it a genuine substitute for subsidized terrestrial networks. NRECA asks the Commission to abandon technological neutrality—the principle of applying the same standards regardless of transmission technology—and count only terrestrial service.[49] NTCA argues that no LEO operator currently meets universal-service standards because of limitations involving capacity, consistency, and voice service.[50] Both objections rest on the concern that a shared satellite network cannot reliably deliver benchmark performance to every subscriber.

That concern warrants scrutiny, but capacity constraints matter most when providers oversubscribe service in densely populated areas. Starlink can reliably meet the 100/20 Mbps benchmark in areas with fewer than roughly seven serviceable locations per square mile.[51] High-cost areas, by contrast, “are largely those in hard-to-reach, rural, remote areas.” They are costly to serve precisely because they are “sparsely populated.”[52] That explains why 85% of Starlink subscribers live in rural areas and why LEO systems “may be their best option available for the foreseeable future” in “remote and hard-to-reach areas.”[53] The capacity objection is therefore strongest in dense areas that generally do not receive high-cost support and weakest in the sparse areas that do.

The remaining objections do not alter the analysis. Voice is an IP application that can operate over a LEO broadband connection without a separate circuit-switched network. The absence of legacy carrier-grade voice reflects the technology’s modern architecture, not an inability to provide voice service.

LEO subscription prices also are “comparable to other rural provider costs,” and competition among satellite, wireless, and terrestrial providers should constrain them further. Where price limits adoption, the Commission should provide targeted aid to consumers in the highest-cost areas rather than continue subsidizing obsolete networks. Such demand-side support would address affordability directly without preserving a less efficient technology.

V. Legacy Operating Costs Do Not Justify Permanent Support

The strongest argument for preserving these programs concerns operating costs rather than deployment. It assumes that the cited costs are inherent in serving rural areas. Much of the burden instead comes from the aging copper and TDM infrastructure that continued subsidies keep in service. The appropriate benchmark is the cost of operating the IP networks that will replace it.

IP infrastructure costs substantially less to operate than legacy plant. Copper degrades over distance and relies on energy-intensive repeaters and TDM switches distributed throughout neighborhoods. Fiber carries data as pulses of light over largely passive infrastructure, lowering maintenance and energy costs while reducing the need for centralized switching facilities and related real estate.[54] Verizon’s migration of 4.5 million circuits to fiber, for example, produced roughly $180 million in annual savings, reduced maintenance dispatches by about 60%, and allowed the company to retire dozens of central offices.[55] Fiber also uses less electricity because it eliminates TDM switches and neighborhood electronics. It also relies less on the shrinking pool of technicians trained to maintain legacy copper.[56]

The operating burden cited by commenters is therefore neither fixed nor unavoidable. It largely reflects the cost of maintaining infrastructure that carriers should be retiring. Treating that expense as an inherent cost of rural service would lock the program into financing the inefficiencies created by obsolete technology.

These networks will transition to IP regardless of the outcome here. The BEAD Program and Enhanced A-CAM already require modern 100/20 Mbps networks designed to operate for a decade or longer without continuing support. Consumers are abandoning legacy switched networks, and the Commission has made the IP transition a policy priority. Subsidizing legacy operating costs gives carriers an incentive to delay the modernization that would reduce those costs. The subsidy then preserves the high-cost structure invoked to justify further subsidies.

The Commission should decline to transform transitional deployment support into a permanent operating subsidy. Phasing out support would accelerate the move to lower-cost IP infrastructure that the market and federal deployment programs are already delivering. Universal service will become more sustainable when policy stops financing the networks that make it unnecessarily expensive.

VI. Conclusion

The Commission should wind down the legacy high-cost mechanisms rather than preserve or replace them. The record and the Commission’s data show that unsubsidized terrestrial competitors and LEO satellites now reach the overwhelming majority of supported locations. The BEAD Program and Enhanced A-CAM are funding modern networks in many remaining areas. Meanwhile, lower-cost IP infrastructure continues to replace the copper and TDM networks that legacy support sustains.

These programs now provide little incremental service. Support is redundant where competitors already operate, duplicative where other federal programs fund deployment, and concentrated in legacy voice service that consumers are abandoning. It also raises consumers’ bills, discourages private investment, and gives carriers an incentive to retain costly infrastructure. The 38.8% contribution factor makes those tradeoffs impossible to ignore. The Commission should measure each program by the service it adds, not the network it preserves.

The Commission should eliminate HCLS on the shortest feasible timeline, place CAF BLS on a defined and declining schedule, and allow the A-CAM programs to expire under their existing terms without extension or replacement. This sequence would respect legitimate reliance interests while removing roughly $1.6 billion in annual obligations that competition, technological change, and other federal programs have overtaken.

Reducing unnecessary disbursements would directly ease pressure on the USF and the consumers who finance it. Where genuine affordability gaps remain, targeted consumer support offers a better response than permanent subsidies for obsolete networks. Universal service will endure only if the Commission allows its programs to end when their work is done.

[1] Reforming the High-Cost Program for an All-IP Future, Notice of Proposed Rulemaking, WC Docket Nos. 26-96 & 10-90, ¶ 33 (rel. May 21, 2026) [hereinafter High-Cost NPRM].

[2] Id. ¶ 31.

[3] Infrastructure Investment and Jobs Act, Pub. L. No. 117-58, div. F, tit. I, § 60102(b)(2), 135 Stat. 429, 1184 (2021).

[4] Proposed Third Quarter 2026 Universal Service Contribution Factor, Public Notice, DA 26-546, CC Docket No. 96-45 (rel. June 12, 2026), https://docs.fcc.gov/public/attachments/DA-26-546A1.pdf.

[5] LEO Policy Working Group, Low Earth Orbit Satellites: Policies to Promote Spectrum Sharing, Foster Competition, and Close Digital Divides 75–76 (Int’l Ctr. for L. & Econ. 2025), https://laweconcenter.org/resources/low-earth-orbit-satellites-policies-to-promote-spectrum-sharing-foster-competition-and-close-digital-divides-a-report-of-the-leo-policy-working-group.

[6] See Kristian Stout & Ben Sperry, Guiding Principles & Legislative Checklist for Broadband Subsidies (Int’l Ctr. for L. & Econ. 2022), https://laweconcenter.org/resources/guiding-principles-legislative-checklist-for-broadband-subsidies.

[7] Jonathan E. Nuechterlein & Howard Shelanski, Building on What Works: An Analysis of U.S. Broadband Policy, 73 Fed. Commc’ns L.J. 219, 257 (2021).

[8] Kristian Stout, Infrastructure Is Destiny: The Geography of the Next Technology Race 19 (Int’l Ctr. for L. & Econ. 2026), https://laweconcenter.org/resources/infrastructure-is-destiny-the-geography-of-the-next-technology-race.

[9] See, e.g., Comments of NTCA–The Rural Broadband Association, Reforming the High-Cost Program for an All-IP Future, WC Docket Nos. 26-96 & 10-90 (filed Aug. 4, 2026), https://www.fcc.gov/ecfs/document/26110069420/1 [hereinafter NTCA Comments]; Comments of the National Rural Electric Cooperative Ass’n, WC Docket Nos. 26-96 & 10-90 (filed Aug. 4, 2026), https://www.fcc.gov/ecfs/document/26110069460/1 [hereinafter NRECA Comments]; Comments of the Blooston Rural Carriers, WC Docket Nos. 26-96 & 10-90 (filed Aug. 4, 2026), https://www.fcc.gov/ecfs/document/26110069494/1 [hereinafter Blooston Comments].

[10] NTCA Comments, supra note 9, at i.

[11] NRECA Comments, supra note 9, at 4–5.

[12] Blooston Comments, supra note 9, at 2.

[13] T. Randolph Beard, George S. Ford, Lawrence J. Spiwak & Michael Stern, The Law and Economics of Municipal Broadband, 73 Fed. Commc’ns L.J. 1, 10 (2020), http://www.fclj.org/wp-content/uploads/2020/09/MunicipalBroadbandArticleFINAL.9.2.20.pdf (“Entry by a subsidized . . . firm . . . reduces the incentives of private firms to invest in modern communications infrastructure . . .. [E]ither the municipal entrant will fail or a private provider will exit or materially reduce its investments.”).

[14] 47 U.S.C. § 254(b)(5), (e); Connect America Fund, Report and Order and Further Notice of Proposed Rulemaking, WC Docket No. 10-90, ¶ 17 (2011), https://docs.fcc.gov/public/attachments/FCC-11-161A1.pdf.

[15] High-Cost NPRM, supra note 1, ¶ 6.

[16] See Stout & Sperry, supra note 6 (arguing that broadband support should “work in tandem with private investment” because subsidizing already-served areas risks “overbuilding in ways that can depress returns to incumbent providers”); Ben Sperry, Doublespeak in the Debate About Rural Broadband Buildout, Truth on Mkt. (Aug. 6, 2020), https://truthonthemarket.com/2020/08/06/doublespeak-in-the-debate-about-rural-broadband-buildout (explaining that government-subsidized entry “can discourage private investment and grow the size of the ‘donut hole,’ . . . leading to demand for even greater subsidies”).

[17] Jeffrey Westling, The Lifeline Program’s Afterlife Problem, Truth on Mkt. (May 6, 2026), https://truthonthemarket.com/2026/05/06/the-lifeline-programs-afterlife-problem.

[18] Proposed Third Quarter 2026 Universal Service Contribution Factor, supra note 4; Proposed Second Quarter 2026 Universal Service Contribution Factor, Public Notice, DA 26-218, CC Docket No. 96-45 (rel. Mar. 16, 2026).

[19] Comments of the Int’l Ctr. for L. & Econ., Reforming Legacy Rules for an All-IP Future & Accelerating Network Modernization, WC Docket Nos. 25-311 & 25-208, at 7 (filed May 22, 2026), https://laweconcenter.org/resources/icle-comments-to-the-fcc-on-accelerating-network-modernization [hereinafter ICLE Network Modernization Comments].

[20] Id. at 7–8.

[21] Id. at 8.

[22] Stout & Sperry, supra note 6.

[23] Id.

[24] See Beard et al., supra note 13, at 10 (“[T]he asymmetric subsidization of . . . entrants . . . reduces the incentives of private firms to invest”); cf. Richard A. Epstein, The Assault That Failed: The Progressive Critique of Laissez Faire, 97 Mich. L. Rev. 1697, 1714 (1999), https://repository.law.umich.edu/mlr/vol97/iss6/22 (arguing that rational actors will not “make private investments” where “the state could launch a competitor nearby . . . with a handsome government subsidy”).

[25] Comments of the Int’l Ctr. for L. & Econ. & TechFreedom, Connect America Fund, WC Docket No. 10-90, RM-11703, at 9 (filed Sept. 26, 2013), https://www.fcc.gov/ecfs/document/6017468701/1 [hereinafter ICLE CAF Comments].

[26] See Epstein, supra note 24, at 1714.

[27] See, e.g., NTCA Comments, supra note 9, at 21.

[28] LEO Policy Working Grp., supra note 5, at 70.

[29] Comments of the Int’l Ctr. for L. & Econ., Safeguarding and Securing the Open Internet, WC Docket No. 23-320, at 16–17 (filed Dec. 14, 2023), https://laweconcenter.org/wp-content/uploads/2023/12/ICLE-Comments-on-2023-FCC-Title-II-NPRM.pdf.

[30] See High-Cost NPRM, supra note 1, ¶¶ 18, 31.

[31] Inquiry Concerning Deployment of Advanced Telecommunications Capability to All Americans in a Reasonable and Timely Fashion, 2026 Section 706 Report, GN Docket No. 25-223, FCC 26-55, ¶ 28 (2026).

[32] See ICLE CAF Comments, supra note 25, at 9 (arguing that universal-service support should “subsidize the construction of networks and the provision of services in parts of the country where doing so would otherwise be uneconomical”).

[33] Comments of NCTA–The Internet & Television Ass’n, Reforming the High-Cost Program for an All-IP Future, WC Docket Nos. 26-96 & 10-90, at 6–7 (filed Aug. 4, 2026), https://www.fcc.gov/ecfs/document/26110069456/1 [hereinafter NCTA Comments].

[34] Comments of the Info. Tech. & Innovation Found., Reforming the High-Cost Program for an All-IP Future, WC Docket Nos. 26-96 & 10-90, at 1 (filed Aug. 4, 2026), https://www.fcc.gov/ecfs/document/26110069402/1.

[35] ICLE Network Modernization Comments, supra note 19, at 9.

[36] Jeffrey Westling, Subsidizing Obsolescence: How FCC Rules Keep Copper Alive, Truth on Mkt. (Mar. 18, 2026), https://truthonthemarket.com/2026/03/18/subsidizing-obsolescence-how-fcc-rules-keep-copper-alive.

[37] Comments of the Int’l Ctr. for L. & Econ., Lifeline and Link Up Reform and Modernization et al., WC Docket No. 11-42 et al., at 8 (filed May 4, 2026), https://laweconcenter.org/wp-content/uploads/2026/05/Lifeline-Comments.pdf.

[38] See High-Cost NPRM, supra note 1, ¶¶ 3, 7.

[39] Id. ¶¶ 9–10 & n.15.

[40] Id. ¶ 32 & n.62.

[41] See, e.g., NCTA Comments, supra note 33, at 2–4; Comments of INCOMPAS, Reforming the High-Cost Program for an All-IP Future, WC Docket Nos. 26-96 & 10-90, at 1–2 (filed Aug. 4, 2026), https://www.fcc.gov/ecfs/document/26110069447/1 (arguing that support should remain “targeted, accountable, and competitively neutral” and account for locations covered by enforceable Broadband Equity, Access, and Deployment or other commitments); Comments of WISPA–The Ass’n for Broadband Without Boundaries, WC Docket Nos. 26-96 & 10-90, at 2 (filed Aug. 4, 2026), https://www.fcc.gov/ecfs/document/26110069453/1 (arguing that the Commission should identify locations that lack 25/3 Mbps service and “are not subject to an enforceable commitment to deploy at least 100/20 Mbps” before maintaining support).

[42] High-Cost NPRM, supra note 1, ¶ 10.

[43] Id. ¶ 9 & nn.14–15.

[44] Id.

[45] Id. ¶ 12.

[46] Id. ¶ 14.

[47] See, e.g., NTCA Comments, supra note 9, at 34 (urging the Commission to “offer rural carriers an extended A-CAM program based on the successful Enhanced A-CAM” by 2028).

[48] High-Cost NPRM, supra note 1, ¶ 8.

[49] NRECA Comments, supra note 9, at 10.

[50] NTCA Comments, supra note 9, at 44.

[51] LEO Policy Working Grp., supra note 5, at 75.

[52] Id. at 76–77.

[53] Id.

[54] ICLE Network Modernization Comments, supra note 19, at 3.

[55] Eric Fruits & Brian Albrecht, Paying to Stand Still: Legacy Copper Mandates in a Fiber World 6 (Int’l Ctr. for L. & Econ. 2026), https://laweconcenter.org/resources/paying-to-stand-still-legacy-copper-mandates-in-a-fiber-world.

[56] Comments of the Int’l Ctr. for L. & Econ., Advancing IP Interconnection et al., WC Docket No. 25-304 et al., at 4 (filed Jan. 20, 2026), https://laweconcenter.org/wp-content/uploads/2026/01/ICLE-Comments-on-Advancing-IP-Interconnection.pdf.

ICLE Senior Scholar of Competition Policy Mario Zúñiga joined LP- Pasión por el Derecho podcast to discuss competition in digital markets, AI, and Peru’s banking sector. He explained why market concentration alone does not establish a lack of competition, and considered how Peru can encourage technology adoption and entry by new financial providers.

ICLE Senior Scholar of Competition Policy Mario Zúñiga joined the Peruvian Institute of Economics to discuss how mobility and delivery platforms contribute to Peru’s economy, employment, and consumer safety, and why concerns about these services can be addressed through existing law rather than new regulation.

Executive Summary

AI and other next-generation technologies require more than data centers, computing power, and electricity. They also depend on communications networks installed underground, strung across utility poles, and built through corridors controlled by federal, state, local, and private gatekeepers.

Broadband networks require large upfront investments that providers may never recover if demand, technology, or regulation changes. They also create social benefits that providers cannot fully capture. Search, negotiation, permitting, and enforcement costs can therefore make valuable projects uneconomic, especially in rural and underserved communities.

Permitting compounds these risks. A single route may require federal reviews, state highway approvals, local permits or franchises, private crossing agreements, and access to thousands of poles. Each gatekeeper controls part of the project, but none bears its cumulative cost. Uncertain delays can matter more than fees because they postpone revenue, shorten an asset’s useful life, and create opportunities for hold-up.

Reform should make approvals faster and more predictable while preserving legitimate environmental, cultural, public-safety, and land-management interests. Policymakers should streamline federal reviews, impose firm shot clocks with deemed-granted remedies, limit fees to reasonable government costs, modernize cable franchising, and strengthen pole-access and cost-allocation rules. One-touch make-ready and dig-once policies can further reduce construction costs.

Recent agency actions and court decisions advance many of these goals, but future officials or courts could reverse them. Congress should codify their durable core. Stable rules would reduce investment risk across all projects and deliver the greatest gains where deployment remains hardest to finance.

I.               Introduction

Previous papers in this series have argued that technological leadership depends less on who invents the next breakthrough than on which jurisdictions build the full, low-friction infrastructure stack—spectrum, broadband, fiber, satellite, edge computing, data centers, and AI accelerators—that allows innovation ecosystems to form, cluster, and endure.[1] At its foundation, AI still relies on physical communications networks buried underground and strung along utility poles. But before providers can bury cable or mount an antenna, they must secure a web of permits and regulatory approvals.[2]

Broadband networks must cross land controlled by federal, state, and local authorities. Deployment typically requires zoning approvals, construction permits, and permission to occupy public rights-of-way.[3] When permitting is slow, inconsistent, or costly—or access to rights-of-way is delayed or priced unpredictably—deployment becomes more expensive and risky. Those barriers also postpone the point at which providers can begin recovering their investments.[4]

Access barriers can constrain new construction as much as a lack of funding. Even a provider ready to invest cannot build without timely, reasonably priced access to the corridors its network must follow.[5] Streamlined permitting and fair, predictable access to public rights-of-way will therefore help determine where—and how quickly—the infrastructure needed for next-generation applications gets built.

II.             The Broadband Investment Challenge

Debate over the infrastructure needed for artificial intelligence (AI) has focused largely on the data centers where models are trained and run and the electricity required to power them.[6] That focus is understandable. The scale of modern computing has strained electrical grids, renewed interest in additional generating capacity, and made data-center siting a major political issue.

Compute and electricity are, however, only part of the infrastructure these technologies require. AI and the immersive applications built on it cannot operate in isolation inside server farms.[7] They must exchange data with users and devices in real time, wherever those users happen to be. That requires high-capacity, low-latency communications networks capable of carrying large volumes of data with minimal delay. AI, augmented reality (AR), and virtual reality (VR) will therefore require substantial investment in broadband and related communications infrastructure.[8]

Broadband investment is not assured. Networks require large upfront expenditures that providers cannot recover once committed to a particular route or technology. Providers also capture only part of the social value their networks create. Search, negotiation, permitting, and enforcement costs can therefore make socially valuable projects uneconomic, while control over a single segment can give one gatekeeper the power to hold up an entire build.

Providers must invest amid uncertain demand, technological change, and regulatory risk. The prospect of obsolescence or new rules that reduce returns raises financing costs and discourages marginal projects, especially in rural and low-income communities. The following subsections examine these constraints and explain why predictable rules and well-designed incentives are essential to broadband deployment.

A.             The Economics of Broadband Investment

Broadband providers face significant barriers to entering and expanding within a market.[9] Chief among them are the high fixed costs that providers must incur before serving a single customer.[10] Recovering those costs depends on attracting enough subscribers, an uncertain prospect in a competitive market.[11]

Much of that investment also becomes a sunk cost—an expense that cannot be recovered if the project fails. Providers cannot easily remove fiber from an unsuccessful build and redeploy it elsewhere. They cannot recover the labor and regulatory costs at all. If subscriptions fall short, providers have few ways to limit their losses.

Economies of scale can offset these risks.[12] A provider can spread a network’s fixed costs across a larger customer base, while customers with a greater willingness and ability to pay can support higher revenue. Projects serving large or relatively affluent populations therefore tend to carry less risk than projects in rural or low-income communities.

Competition across different kinds of technologies, often called intermodal competition, adds another layer of uncertainty. Broadband markets once allowed one or two providers to capture large shares because consumers had few alternatives. Today, providers using different technologies compete across price, speed, reliability, coverage, and other dimensions.[13]

Home broadband may reach consumers through coaxial cable, fiber optics, fixed wireless service delivered over mobile networks, or low-Earth-orbit (LEO) satellites. Mobile broadband can offer enough capacity and performance for some consumers to forgo a fixed home connection altogether. At the same time, traditionally fixed providers have added mobile services so customers can remain connected outside the home. Because these technologies have different costs and infrastructure needs, multiple competing networks may prove more efficient than a single set of pipes.[14]

Competition and deployment both depend on continued network investment. In a competitive capital market, investment flows toward projects whose expected return, adjusted for risk, exceeds their cost.[15] Broadband providers calculate that return primarily from monthly subscription fees, supplemented by advertising and ancillary services.[16]

Broadband also creates substantial value that providers cannot capture through those revenue streams.[17] These spillover benefits—known as positive externalities—reach millions of people and many sectors, while providers bear the concentrated cost of deployment. A project is socially worthwhile when its private revenue and broader social benefits together exceed its cost.[18] Yet a rational provider will not build if its own expected revenue cannot cover that cost. It cannot charge for every increase in a farmhouse’s value, improvement in a student’s grades, or savings in a rural hospital’s budget.

Communities lose those benefits when broadband investment falls short. Real-estate prices can help measure broadband’s value because they incorporate local access to education, health care, employment, and other amenities.

Students without reliable home internet perform worse on standardized tests than their connected peers, which can reduce college enrollment and degree attainment.^[19] Telemedicine can lower patients’ costs and make rural living practical for more people. It also helps rural health care facilities reach more patients and connect with specialists at major medical centers without employing them full time. Patients, in turn, can obtain care without traveling to a facility. Broadband also allows higher-earning remote workers to move to lower-cost rural areas, bringing income and spending into those communities.

These positive externalities create an underinvestment problem. Because providers cannot capture broadband’s full social value, expected private returns may not justify deployment even when a project would benefit society overall.[20]

AI, AR, and VR will intensify that problem. These technologies require high-capacity connections with minimal delay, reliable fiber links connecting local networks to the broader internet, processing located closer to users, and access to large-scale computing resources. Many applications must run AI models and exchange data continuously in real time, placing simultaneous demands on communications networks.

The benefits of these applications will extend far beyond what any one provider can monetize. Without policies that encourage investment, the communities least likely to attract private capital may also lack the infrastructure that AI, AR, and VR require. That outcome would widen the gap between served and unserved regions just as advanced connectivity becomes more important to economic competitiveness. Policymakers seeking broadband’s broader benefits must therefore give providers stronger incentives to invest in unserved and underserved areas.

B.             Transaction Costs and the Broadband Holdout Problem

In a world without transaction costs, parties would bargain toward an efficient allocation of resources regardless of who initially held the relevant property rights.^[21] Applied to broadband, this principle suggests that a fiber provider will secure access to a utility pole, strip of land, or railroad crossing whenever it values that access more than the owner values excluding it. The parties will strike a deal, and the network will get built.

In practice, transaction costs—the costs of identifying the right parties, negotiating an agreement, and enforcing it—can make broadband deployment prohibitively expensive. These costs generally take three forms:

• Search and information costs. Providers must identify who owns or controls each asset along a proposed route. That may require researching the ownership of thousands of utility poles divided among electric utilities and legacy telephone companies, tracing easements across private land, or determining which government agency controls a particular right-of-way.
• Bargaining and decision costs. Providers incur legal fees, administrative expenses, and delays while negotiating access. A single project may require agreements with municipalities, private landowners, homeowners associations, and railroads, each subject to different legal requirements and each able to demand payment for its consent.
• Policing and enforcement costs. Providers must monitor compliance and enforce their contractual and legal rights. These costs include policing pole-attachment agreements, protecting easement boundaries against encroachment, and litigating disputes over “fair and reasonable” rates.

Asset specificity can magnify each of these costs. The term describes investments tailored to a particular project that cannot easily be redirected elsewhere. Suppose a provider plans a 10-mile fiber route and obtains every approval except the one needed for the final mile. The owner controlling that segment can threaten the entire project and demand a payment far above the value of the access itself. The provider must either pay, redesign the route, or abandon investments already made along the other 9 miles.

For communications networks, transaction costs can exceed the price of the equipment itself. A mile of fiber-optic cable containing 144 strands costs roughly $5,000 to $8,000.[22] Make-ready work—the work needed to prepare a utility pole for a new attachment—can cost $600 to $6,000 per pole. Replacing a pole that cannot accommodate another attachment can cost $5,000 to $15,000.[23] Permitting and access to public rights-of-way can add $20,000 to $40,000 per mile, while railroad-crossing fees can make an otherwise viable project uneconomic.[24] Major projects also require substantial spending on construction crews, traffic control, and coordination among the parties working in public rights-of-way.

Costs accumulate when many parties hold vetoes over a single linear project and each seeks the greatest possible payment for its approval. Each gatekeeper considers the value of its own permission, not the cumulative burden on the network. The result resembles a linear anticommons—a project that requires consent from a long chain of independent parties, any one of which can block it.

These pressures can also push providers toward technologies that are less suitable for a particular community but easier to deploy. A fixed wireless project, for example, may require one tower-site approval, while a fiber build may require permission to attach equipment to 5,000 poles. Avoiding that chain of approvals can determine the technology a provider chooses, even if another option would deliver better service to consumers.

Government policy can raise or lower these transaction costs. When regulators can change rules abruptly or apply them subjectively, providers must account for the risk of an unfavorable policy shift.[25] That uncertainty weighs especially heavily on broadband projects because they require large upfront investments, take years to repay, and consist largely of sunk costs.[26]

A decision by the Federal Communications Commission (FCC) to subject broadband to common-carrier, utility-style regulation could reduce the expected return on an existing investment. So could changes by the National Telecommunications and Information Administration (NTIA) that tighten performance benchmarks under the Broadband Equity, Access, and Deployment (BEAD) Program. Providers must price these risks into a project even if the rules never change. Regulatory uncertainty can therefore sharply increase deployment costs and deter investments that would otherwise proceed.

C.             Demand, Obsolescence, and Regulatory Risk

Alongside high deployment costs, broadband providers face uncertainty about future demand. The FCC currently uses download speeds of 100 megabits per second and upload speeds of 20 megabits per second as its benchmark for fixed broadband.[27] That benchmark supports many current applications, but individual needs vary widely. A household with three children streaming separate programs requires more capacity than a single occupant who uses the internet mainly for email. Providers must design networks around expected use across their customer base, not the needs of any one household.

That task becomes harder as technology and consumer behavior change. AI, AR, and VR could sharply alter usage patterns and increase demand for network capacity. Providers can draw on lessons from the deployment of fourth-generation (4G) wireless networks and other technological shifts, but historical data cannot eliminate uncertainty about future demand.

Providers must also choose among technologies whose relative costs and capabilities will continue to change. Advances in LEO satellite service, sixth-generation (6G) wireless networks, or other technologies could undermine the economics of infrastructure built today. Physical assets could then become stranded, meaning they would no longer generate enough revenue to repay the debt used to finance them.

Investors account for this risk of obsolescence. If they cannot be confident that an asset will remain commercially useful for 10 years, they will demand a faster return of capital. That shorter recovery period can lead to higher prices, delayed construction, or no deployment in marginal markets. Regulation can compound the problem. Rules that calculate returns from historical costs over long recovery periods may fail to account for infrastructure that becomes obsolete before investors recover their capital.

Providers also face the risk of regulatory hold-up.[28] Before construction begins, providers retain leverage because capital is mobile. If one jurisdiction imposes excessive burdens, they can invest elsewhere. Governments may seek to attract investment by offering subsidies, regulatory stability, forbearance, or favorable franchise terms.

Once a provider has built the network, the balance of power shifts. The provider cannot move its fiber, recover most of its costs, or credibly threaten to leave. Regulators can then change the terms governing the investment by imposing rate controls, requiring competitors to receive access to the infrastructure, or adopting other rules that reduce the provider’s expected return.

This mismatch between regulatory assumptions and investment realities can produce underinvestment. Regulators may view broadband as a mature monopoly utility. In competitive markets, though, broadband deployment often carries risks closer to those of a technology startup than a traditional water or telephone company. Policy that ignores those risks will discourage investment precisely where uncertain demand already makes deployment hardest to justify.

III.           The Costs and Risks of Permitting

Broadband deployment requires costly, long-term investments that providers cannot easily recover or redirect. Regulators seeking to encourage deployment should therefore reduce avoidable costs and risks. Permitting offers a prime opportunity for reform.

A permit is more than a line item in a project budget. An outside party with discretion—and often an effective veto—controls both its price and its timing. Providers must identify the relevant authorities, navigate overlapping review processes, negotiate access and conditions, and comply with ongoing obligations. Each step adds transaction costs. Uncertain timelines, subjective standards, and litigation also delay revenue, shorten an asset’s useful economic life, and raise the return investors require. When a route needs approval from multiple gatekeepers, any one of them can hold up the entire build. Permitting can therefore become both one of a broadband project’s largest transaction costs and one of its greatest sources of investment risk.

A.             How Permitting Raises Transaction Costs

Permitting maps closely onto the three categories of transaction costs identified in Section II.B: finding the relevant decision-makers, securing their approval, and complying with or enforcing the resulting agreements.

The first category consists of search and information costs.[29] A rural broadband route may cross federal land administered by the Bureau of Land Management (BLM) or the U.S. Forest Service. That can trigger a right-of-way grant, environmental review under the National Environmental Policy Act (NEPA), historic-preservation review under Section 106 of the National Historic Preservation Act (NHPA), and tribal consultation.

The same route may follow or cross a state highway right-of-way, then pass through a patchwork of county and municipal permitting regimes as it approaches consumers. It may also require access to a railroad crossing controlled by a private carrier or to easements across private land. Before applying for approval, the provider must identify who controls each segment, which law governs, and which overlapping review processes apply. In areas where the federal government owns much of the land, a rural route may remain within the federal permitting system for most of its length.

The second category consists of bargaining and decision costs. Once a provider identifies the relevant authority or landowner, it must secure the right to enter the right-of-way and build. Each entity may impose its own applications, fees, engineering studies, surveys, and substantive standards. The provider must satisfy each one separately.

Many of these decisions involve discretion rather than a ministerial review based on an objective checklist. Providers must therefore do more than pay a posted fee. They may need to negotiate terms, commission studies, redesign routes, and accept project-specific conditions.[30] Multiplied across every jurisdiction along a route, these requirements create a substantial administrative burden before workers install a single strand of fiber.

The third category consists of policing and enforcement costs. After securing a permit, providers must comply with restoration requirements, seasonal work windows, inspection regimes, and other conditions. They must monitor those obligations and may need to challenge denials or disputed conditions through administrative proceedings or litigation.

Each authority and landowner along a route controls an independent point at which the project can stall. Each may seek to maximize the value of its own permission without accounting for the cumulative burden on the deployment.[31] A railroad may charge far more for a perpendicular crossing than the cost of the materials disturbed. A single county may withhold a right-of-way permit. Either can stop a project that has secured every other approval.

This dynamic helps explain why providers sometimes choose technologies that require a single approval over those that require thousands. Yet even that single approval may trigger NEPA and Section 106 review when a project crosses federal land. The holdout risk also grows as construction progresses. The last party controlling an unapproved segment gains bargaining power because the provider cannot readily redeploy the capital already sunk into the rest of the route.

On many routes, obtaining permission costs more than the hardware that permission allows the provider to install. The process meant to authorize construction can become costlier than construction itself.

B.             Delay, Discretion, and Litigation Risk

Permit fees account for only part of the risk that broadband providers must consider when investing in infrastructure. Delays, discretionary standards, and litigation can impose even greater costs.

Delay matters because broadband projects require large upfront investments in assets that providers cannot readily redeploy. Project economics depend heavily on when revenue begins relative to when costs accrue. Permitting pushes revenue further into the future, while financing and development costs begin accumulating before construction starts.

Providers must carry those costs through a review process of uncertain length. Each month of delay also consumes part of an asset’s useful economic life while the risk of technological obsolescence continues to grow. A 10-year asset that spends three years in review and litigation must repay its debt from seven years of revenue rather than 10. If a disruptive technology emerges in year five, the delay will have consumed much of the period during which the asset remained competitive. Permitting delays thus compress the recovery period on which the investment depends.

Discretion adds another layer of risk. Permitting standards often turn on open-ended questions: What constitutes a “reasonable period of time”? Does a denial rest on “substantial evidence”? Does a project qualify for a categorical exclusion from full environmental review?[32] A provider can incorporate a fixed fee into its budget. It cannot price a discretionary decision with the same confidence. Instead, it must estimate a range of possible outcomes and require a higher expected return to compensate for that uncertainty.

Completing agency review does not eliminate the risk. NEPA is the most frequently litigated federal environmental statute.[33] Research indicates that national advocacy organizations, rather than directly affected local parties, have filed most NEPA suits in recent years.[34] Litigants can challenge a categorical exclusion or a finding of no significant impact, adding years to a project that has already cleared agency review.

The Supreme Court narrowed that exposure last year in Seven County Infrastructure Coalition v. Eagle County.[35] The Court directed lower courts to give agencies substantial deference when reviewing the scope of their environmental analyses. It also held that NEPA generally does not require agencies to analyze attenuated upstream or downstream effects tied to separate projects in different places or at different times. The decision should narrow both the required analysis and the available grounds for litigation. But it did not eliminate NEPA review, its associated delays, or the possibility that a court could enjoin a project after an agency approves it. Investors must still account for those risks before committing capital.

Taken together, permitting adds fragmented veto points to projects that already require large, sunk investments exposed to technological obsolescence. It then makes both the duration and outcome of the process uncertain. Providers price that risk into every project, even when no delay or denial ultimately occurs. A project that proceeds on schedule still costs more ex ante because investors had to account for the possibility that it would not.

These burdens do not fall evenly. Delay and denial risk raise the hurdle rate—the minimum expected return needed to justify an investment—and eliminate marginal projects first. The effects therefore fall hardest in communities where deployment promises substantial social value but weak private returns. Streamlining permitting can help close that gap by allowing more of broadband’s diffuse social benefits to support actual investment.

IV.          The Broadband Permitting Stack

Section III explained how permitting operates as both a transaction cost and a source of investment risk. This section maps those burdens across federal, state, and local governments, as well as the separate regime governing pole attachments.

Although the requirements differ, their costs recur in familiar forms: application fees, recurring rents, in-kind obligations, consultant expenses, bonding and insurance requirements, compliance costs, and, above all, the carrying cost of delay. Federal environmental, historic-preservation, tribal, water, habitat, and land-use reviews may overlap with state highway-access and environmental rules. Local franchising, right-of-way, siting, street-cut, and inspection requirements add further approvals. Aerial deployments must also navigate pole-access, make-ready, and replacement processes controlled by utilities and existing attachers.

These layers compound rather than merely accumulate. Independent decision-makers control sequential veto points without bearing the project’s total cost. The resulting delays and uncertainty raise the return needed to justify investment and threaten marginal projects first—especially rural and underserved builds with limited private returns but substantial social value.

A.             Federal Environmental and Land-Use Permitting

A broadband project may require federal approval because it receives federal funding or because a federal agency must authorize part of it. The most consequential federal process is often review under NEPA, which requires agencies to assess the environmental effects of major federal actions.[36] For broadband projects, that duty may arise when a build receives federal support, including a BEAD subgrant, or when the FCC’s licensing authority makes a wireless deployment a federal action.[37]

The required analysis varies with the project’s likely effects. Agencies may apply a categorical exclusion to routine work requiring minimal documentation, prepare an environmental assessment to determine whether significant effects are likely, or require a full environmental-impact statement. Providers may need to hire consultants and wait for agency review, but uncertainty and litigation often impose the larger costs.[38]

Congress imposed deadlines and document-length limits in 2023, and the Supreme Court further narrowed NEPA exposure in Seven County Infrastructure Coalition v. Eagle County.[39] The Court directed lower courts to defer substantially to agencies and rejected demands that they analyze attenuated upstream and downstream effects tied to separate projects. These changes should limit the scope of review and the grounds for litigation. Neither eliminated NEPA’s underlying requirements or the delays they can create.

Section 106 of the NHPA adds a separate historic-preservation review. It requires agencies to assess a project’s effects on properties listed in, or eligible for, the National Register of Historic Places and to consult the relevant state or tribal historic preservation officer.[40] A single consultation can take six months or longer. In areas rich in historic or cultural resources, that review can determine the schedule for the entire build.[41]

Projects affecting tribal lands or cultural resources may require broader consultation. Section 106 reaches beyond tribes whose present-day lands contain the project. Agencies must also consult any tribe that attaches religious or cultural significance to an affected property, including ancestral sites far from current reservation boundaries.^[42] For wireless facilities, the FCC’s notification process may require consultation with many tribes over one site.

This process serves genuine interests in tribal sovereignty and cultural preservation. It also adds per-site fees and another uncertain scheduling variable. Providers and some policymakers contend that some tribes have used review fees partly to generate revenue rather than solely to fund oversight, allowing revenue to rise with the pace of deployment.[43] Whatever the purpose of a particular charge, providers must include both its amount and its timing in the cost of a build.

Routes crossing wetlands, streams, or navigable waters face another layer of federal review, administered primarily by the U.S. Army Corps of Engineers (Corps).[44] The Corps offers nationwide permits for routine linear crossings that remain within defined limits on the area and degree of disturbance.[45] This streamlined process can keep the burden modest.

Costs and delays rise sharply when a crossing exceeds those limits and requires an individual permit. That process may demand a formal wetland delineation, public notice, and a review lasting more than a year. Section 7 of the Endangered Species Act can add consultation with the U.S. Fish and Wildlife Service, biological surveys, and seasonal work restrictions.[46] Those restrictions may affect a project’s schedule more than its budget by halting construction during breeding or migration periods.

As with NEPA review, uncertainty often costs more than the permit fee itself. Providers must account for the possibility of an individual permit, extended consultation, or a seasonal restriction at every crossing, including those that ultimately qualify for streamlined treatment.

Projects crossing federal land face a separate right-of-way process. This is especially common in the western United States, where the federal government owns large expanses of land. The Federal Land Policy and Management Act authorizes the BLM to grant rights-of-way across public lands.[47] The Forest Service, National Park Service, and Department of Defense administer separate regimes for their holdings.

These approvals can generate two fee streams: cost-recovery charges that reimburse agencies for processing and monitoring, and rental fees based partly on the population a facility serves. A 2024 BLM rule streamlined the communications-use program, required electronic filing, and established a 270-day deadline for decisions.[48] Those reforms implicitly acknowledged that the earlier process imposed delays substantial enough to deter deployment.

B.             State Permitting and Regulatory Fragmentation

The most common state-level requirement is permission to occupy or cross state highway rights-of-way, typically through an encroachment or utility-accommodation permit.[49] Because each state establishes its own regime, a provider operating across several states must navigate distinct fee schedules, technical standards, applications, and review timelines. This fragmentation raises search and bargaining costs before the provider pays a permit fee.

Many states also require a carrier to obtain a certificate of public convenience and necessity, or to register with the utility commission, before operating or occupying the right-of-way. A related designation, eligible-telecommunications-carrier status, is a precondition for most subsidy programs and carries obligations of its own.[50]

State law may duplicate or displace federal requirements. Roughly 16 states have enacted their own environmental-review statutes, often called “little NEPAs.”[51] California’s Environmental Quality Act is the best-known example.[52] These laws can add a parallel track of environmental study and litigation to projects already subject to NEPA.

In roughly 20 states, Section 224’s reverse-preemption provision displaces the federal pole-attachment regime. State rules therefore govern one of the largest infrastructure-access costs across much of the country. Many states have also replaced local cable franchises with statewide video-franchise systems.[53]

Other state requirements include dig-once policies that coordinate conduit installation, one-call rules requiring excavators to locate buried utilities, and pavement-restoration standards. These measures add procedural steps, but thoughtful design can reduce overall costs by coordinating construction, preventing damage, and avoiding repeated excavation.

C.             Local Franchising and Siting Controls

At the municipal level, cable operators must obtain a local franchise under Title VI of the Communications Act. Federal law caps the franchise fee at 5% of gross revenue.[54] Localities have also required in-kind contributions, including public-access channels, capacity on institutional networks serving public facilities, free service to public buildings, and buildout commitments. The FCC’s 2019 order counts many of these contributions toward the 5% ceiling, but disputes over additional obligations persist.[55]

Negotiating and renewing franchises can itself delay deployment. Reform proposals have therefore paired franchising requirements with a shot clock—a deadline for government action—and a deemed-granted remedy that automatically approves an application when the locality misses that deadline.[56]

Noncable providers generally use permits rather than franchises to access local rights-of-way.[57] Section 253 bars state and local requirements that prohibit or effectively prohibit telecommunications service. It also preserves local authority to manage public rights-of-way and require fair and reasonable compensation.[58] The FCC has generally treated fees closely tied to a locality’s actual costs as the most appropriate measure under that framework. Enforcement remains uneven, and some localities continue to impose above-cost or gross-revenue fees.

Above-ground facilities—including cabinets, equipment huts, small-cell antennas, and towers—may require conditional-use permits, variances, site-plan and aesthetic review, public hearings, and compliance with height and spacing limits. Some jurisdictions also require providers to place facilities underground. Section 332(c)(7) preserves substantial local authority over wireless siting while prohibiting decisions that ban or effectively ban wireless service.[59] FCC shot clocks limit the time for review, but the process can still take months or years.

Providers must obtain additional public-works approvals before trenching or boring. These may cover street cuts, lane closures, traffic control, and pavement restoration. Rules addressing pavement damage and moratoria on cutting newly paved streets can add large and highly variable costs.[60] Building and electrical permits and inspections may apply as well. Some localities also impose temporary siting moratoria that consume much or all of the applicable shot-clock period.[61]

Wireless facilities operate under a more developed federal timing framework. Under Section 332(c)(7), localities presumptively have 90 days to act on collocation applications (which seek to place equipment on existing structures) and 150 days to review new sites.[62] Missing those deadlines constitutes a “failure to act,” but it does not automatically approve the application. The provider must seek relief in court.

The FCC’s 2018 Small Cell Order shortened the review periods to 60 days for collocations and 90 days for new small wireless facilities. It also established presumptively reasonable fee levels.[63] Section 6409(a) requires localities to approve eligible modifications that do not substantially change the physical dimensions of an existing tower or base station. FCC rules apply a 60-day shot clock and a self-executing deemed-granted remedy to those requests.[64]

These provisions are among the strongest timing protections in the current permitting system. Their existence highlights how little comparable discipline governs many other approvals. Even here, localities can pause or restart the clocks by declaring applications incomplete or requiring supplemental studies.[65]

D.            Pole Attachments and Make-Ready Costs

To deploy aerial fiber, a provider must attach its lines to utility poles it usually does not own. Federal law, rather than private bargaining alone, largely governs the terms of that access. Section 224 of the Communications Act requires pole-attachment rates, terms, and conditions to be “just and reasonable.” The FCC administers that standard through formula-based rates in states that have not certified their own regulatory regimes.[66]

This framework aims to promote investment and facilitate deployment, but it does not cover poles owned by municipalities or rural electric cooperatives.[67] Providers seeking to use those poles must negotiate without a federal regulatory backstop.

The permit or rental rate represents only part of the cost. The larger and less predictable expense often comes from make-ready work: the rearrangement, relocation, or reinforcement of existing equipment needed to create space for a new line.[68] This work can cost several hundred to several thousand dollars per pole. It may also take months, because each company with equipment already on the pole must move its facilities in sequence.[69]

When a pole is too short or weak to support another attachment, make-ready work may require replacing it at a cost of roughly $5,000 to $15,000.[70] Utilities have historically charged the entire replacement cost to the new attacher, even when the pole had deteriorated, neared the end of its useful life, or otherwise required replacement.[71]

The FCC has found that a utility may not impose the full cost on a new attacher that did not solely cause the replacement. Instead, the pole owner must bear costs attributable to preexisting violations or a replacement it already had an obligation to make.[72] This approach prevents utilities from using a new attachment request to shift the cost of overdue maintenance onto the requesting provider.

The FCC has also adopted one-touch make-ready (OTMR) rules for simpler work. OTMR allows one qualified contractor to move existing equipment in a single pass rather than waiting for a succession of incumbent contractors.[73] The process can reduce delays that might otherwise consume much of a year.[74]

E.              How Permitting Burdens Compound

Each permitting layer can impede deployment on its own. Together, they impose a much larger burden. Independent authorities and landowners control separate veto points, yet none has an incentive to account for the project’s total cost. The provider therefore faces more than the price a single, coordinated decision-maker might charge. It must bear the cumulative demands of many actors, each seeking value for its own approval.

Each layer also adds more than a fee. It introduces a delay of uncertain length. Because broadband projects require large, sunk investments in assets exposed to obsolescence, those delays raise financing costs, shorten the period available to recover capital, and increase the project’s risk premium.

These burdens fall unevenly. Each additional cost or delay raises the hurdle rate, so the marginal project fails first. That project is disproportionately likely to serve a rural or underserved community where expected private returns are thin but the broader social benefits are substantial. Permitting barriers thus inflict their greatest harm where deployment may create the most social value. That mismatch supplies the central case for reform, to which the next section turns.

V.             A Reform Agenda for Faster, More Predictable Deployment

Effective reform must address more than posted fees. Uncertain delays, fragmented approvals, and the risk of regulatory reversal often do more damage because providers must price those risks into every project. These burdens fall hardest on marginal builds, especially networks serving rural and underserved areas.

Since mid-2025, the FCC’s “Build America” agenda has targeted many of these barriers. The agency has opened proceedings on wireless siting and wireline rights-of-way, revised its pole-attachment rules, and proposed changes to its environmental and historic-preservation reviews.[75] Further reforms should streamline federal reviews, constrain state and local fees and timelines, modernize cable franchising, expand pole-access protections, resolve pole-replacement disputes, coordinate fiber installation with roadwork, and codify durable rules that investors can rely on. Together, these changes would reduce the cost and uncertainty of building the communications networks that AI and other next-generation applications require.

A.             Streamlining Federal Reviews

Federal review regimes impose their greatest costs through delay, uncertainty, and litigation rather than permit fees. Reform should therefore limit full review to projects likely to cause significant effects, accelerate the reviews that remain, and give approved projects greater finality.

Categorical exclusions offer the most efficient tool for routine deployments. When experience shows that a defined class of projects does not ordinarily cause significant environmental effects, an exclusion can avoid repetitive case-by-case assessments while narrowing both direct costs and the grounds for litigation.[76] Programmatic agreements can serve the same function under Section 106. Rather than reconsidering recurring issues site by site, agencies, preservation officials, and tribes can establish one framework governing an entire category of deployments.[77]

Recent Supreme Court decisions have narrowed the governing federal obligations. In Seven County Infrastructure Coalition v. Eagle County, the Court reinforced agency discretion under NEPA and limited demands to analyze attenuated effects tied to separate upstream or downstream projects.[78] In Sackett v. EPA, the Court narrowed federal wetlands jurisdiction under the Clean Water Act.[79] These decisions provide meaningful limits, but their practical effect still depends on how agencies implement them and how courts apply them in later cases.

Congress and agencies should translate those boundaries into clear, durable rules. Agencies should adopt categorical exclusions for well-defined classes of communications facilities that reliably produce no significant effects. Firm review deadlines should govern the projects that remain, with automatic approval when an agency fails to act on time.

Reform must also respect the limits courts have already imposed. In United Keetoowah Band of Cherokee Indians v. FCC, the D.C. Circuit vacated the FCC’s attempt to exempt small wireless facilities from NEPA and Section 106 review.[80] The court found that the FCC had not adequately justified its public-interest determination or addressed potential effects on historic and tribal resources. The decision shows that streamlining measures need a well-developed record that squarely addresses environmental, historic-preservation, and tribal interests.

The same principles apply to projects crossing federal land. Agencies should use master right-of-way agreements, standardized applications, cost-based rather than value-based charges, and firm decision deadlines. The BLM’s 2024 communications-use rule offers a useful model for making federal approvals faster and more predictable without abandoning substantive review.

B.             Cost-Based Fees and Self-Enforcing Deadlines

For facilities outside the FCC’s small wireless framework, state and local reform should focus on two issues: fees and review deadlines. In both cases, the rule matters only if providers can enforce it effectively.

Fees for construction permits and access to public rights-of-way should recover the government’s actual costs. They should not capture the value of the provider’s investment or exploit the government’s monopoly control over the corridor. Value-based and gross-revenue fees operate as taxes on deployment. They fall hardest on thin-margin rural projects that policy should encourage.

Presumptively reasonable benchmarks can give providers and localities a clear reference point. The FCC’s small-cell safe harbors—$500 for certain applications and $270 per site each year—illustrate the approach. A locality could impose a higher fee, but it would bear the burden of showing why its actual costs justify the increase. Governments should not be able to evade fee caps by demanding in-kind contributions instead of money.

Deadlines require equally strong enforcement. A shot clock that merely allows a provider to sue after time expires preserves much of the delay and uncertainty it aims to eliminate. The provider must still account for the expense, duration, and risk of litigation.

A self-executing “deemed granted” remedy works differently. Under Section 6409(a), an eligible application receives automatic approval when the locality fails to act within the prescribed period. The provider therefore knows it will receive an answer by a date certain, regardless of the locality’s diligence. Extending this model to ordinary siting, right-of-way, and construction permits would provide the strongest available protection against permitting delay.

The American Broadband Deployment Act of 2025 and the BROADBAND Leadership Act reflect this approach by pairing deadlines across multiple permitting layers with cost-based fee limits.[81] The FCC is pursuing similar goals under its Section 253 and Spectrum Act authority, including a proposed “rocket docket”—an expedited process for resolving preemption disputes.[82]

Reform should still preserve local authority over legitimate concerns such as public safety, traffic management, and aesthetics. The aim should be to discipline fees and timing while allowing reasonable, evenhandedly applied substantive conditions.

C.             Reforming Local Franchising

Title VI’s local cable-franchising regime predates modern competition among fiber providers building their own networks. Today, it can operate as both a toll and a source of delay for video-capable broadband entrants. Moreover, the 5% gross-revenue franchise-fee cap cannot serve as a meaningful ceiling if localities may extract additional value through in-kind demands.

The FCC’s 2019 order, which counts many in-kind obligations toward the cap, was a sound response. Congress should codify that approach so it does not depend on changing FCC majorities. Governments should not be able to evade a monetary cap by demanding services, facilities, or other contributions instead.

The broader reform is to subject franchising to the same timing rules as other permits. The American Broadband Deployment Act of 2025, for example, would establish a 120-day deadline so localities cannot use a process negotiated with an incumbent to delay a second provider.[83] States that have replaced separate local regimes with a single statewide video franchise offer a model that other jurisdictions could expand.

Franchising can also obstruct providers that do not offer cable service. Some localities have conditioned ordinary right-of-way permits on a noncable provider first obtaining a cable franchise. This practice erases the distinction between permits and franchises and imports a slower, more demanding process into routine right-of-way access.

Gateway Fiber encountered this problem in Minnesota. After the company sought permits to build fiber in several cities, Maple Grove said it would defer action until Gateway completed the cable-franchising process. Gateway is not a cable operator, yet the city effectively refused to issue the right-of-way permits without a cable-franchise agreement.[84]

Gateway petitioned the FCC in June 2025 for a declaratory ruling preempting the requirement under Section 253(d). The company argued that Maple Grove’s refusal to act “has the effect of prohibiting” telecommunications service in violation of Section 253(a). According to Gateway, the city’s position prevents the company from operating its existing network and blocks expansion into new areas.[85]

The Wireline Competition Bureau placed the petition on public notice in July 2026 and situated the dispute within the FCC’s broader Section 253 wireline-deployment inquiry.[86] Whatever the Commission decides, the episode exposes a broader vulnerability. The line between a permit and a franchise can turn on disputed questions about a provider’s regulatory classification and a locality’s authority under state law. A locality can exploit that uncertainty to repurpose a legacy franchising process as a gatekeeping tool against a new fiber entrant—the kind of last-mile chokepoint that Section 253 is meant to police.[87]

D.             Closing the Pole-Attachment Gap

Because access to existing poles often represents one of the largest costs of a wireline build, policymakers should extend Section 224’s guarantee of just, reasonable, and timely access more broadly. Recent FCC orders have improved the federal regime. OTMR allows one approved crew to complete simple work rather than waiting for multiple attachers to act in sequence. The FCC’s July 2025 order also established access timelines for large and bulk applications, required parties to meet and confer, expanded self-help when utilities miss deadlines, and accelerated approval of qualified contractors.[88]

These reforms reach only part of the problem. Section 224 exempts poles owned by municipalities and cooperatively organized utilities, placing them outside the FCC’s jurisdiction.[89] Municipal utilities and rural electric cooperatives own a substantial share of the distribution poles in rural America—the same unserved and underserved areas targeted by federal broadband programs and where deployment economics are already most difficult.[90]

Providers seeking access to municipal or cooperative poles therefore face rates and timelines unconstrained by federal rules. In many states, no comparable state protections apply. Congress should extend Section 224—or an equivalent set of federally enforceable deadlines and cost-allocation rules—to these poles. Doing so would narrow the gap between the places where connectivity is most needed and those where access rules make deployment feasible.

Any extension should still allow pole owners to recover legitimate costs. Municipal and cooperative utilities are nonprofit entities owned by citizens or members, so ratepayers ultimately bear the cost of below-cost attachment rates. Reform should prevent monopoly pricing and unreasonable delay without forcing those ratepayers to subsidize broadband attachers.

E.              Enforcing Pole-Replacement Cost Sharing

Broadband providers and pole owners increasingly dispute who should pay when a pole must be replaced to accommodate a new attachment. The FCC’s December 2023 declaratory ruling clarified that an attachment does not “solely” necessitate replacement when the utility has already red-tagged the pole (marked it for replacement) or when a change in the utility’s construction standards requires the upgrade. In those circumstances, the new attacher should not bear the full cost.[91]

The central challenge is now enforcement. The rule helps providers only if they can invoke it quickly and at reasonable cost. Otherwise, the utility’s initial allocation governs while the dispute proceeds, and the FCC’s cost-sharing principle may have little practical effect.

Three complementary reforms could close that gap. First, the FCC should resolve these disputes through an expedited complaint process or “rocket docket” governed by firm deadlines.[92] Where authorized, fee shifting could further reduce the cost of asserting a valid claim. Faster and less expensive adjudication would make the proper allocation—not the utility’s opening demand—the operative result.

Second, Congress should codify the cost-allocation principle. Codification would protect it from reconsideration petitions and changes in FCC leadership, giving providers a more stable basis for investments expected to last 20 years.

Third, targeted public reimbursement funds could help resolve cases in which genuine cost sharing is warranted. Limited support for qualifying pole replacements could break deployment impasses without requiring the parties to litigate responsibility pole by pole.

F.              Coordinating Broadband with Roadwork

Excavation accounts for much of the cost of underground fiber, making coordination with planned roadwork one of the highest-return deployment reforms. A 2021 Federal Highway Administration rule requires state transportation departments receiving federal-aid highway funds to designate a broadband utility coordinator and establish processes for registration, notification, and coordination. A January 2026 Government Accountability Office review found that most states had made progress implementing those requirements, although their effects on deployment remain difficult to measure.[93]

Weak “dig-once” policies merely notify providers of planned roadwork and give them an opportunity to install conduit. That approach helps only providers whose construction plans happen to align with a particular project. Middle-mile and backhaul routes, which connect local networks to the broader internet, frequently will not align.

Stronger policies require road sponsors to install spare conduit whenever a federally funded project opens the ground. The sponsoring agency can then own and lease that conduit to broadband providers. This approach preserves far more of the roadwork’s future value by allowing providers to install fiber later without excavating again. Proponents estimate that installing conduit during planned construction adds only a few percentage points to the road project’s cost.

G.            Making Reform Durable

As outlined above, many of the most consequential recent reforms rest on agency orders or judicial interpretations. But both can change. Investors will hesitate to commit capital for 10 years based on a rule that a new FCC majority, a reconsideration petition, or a later court decision could undo in five.

That uncertainty limits the value of agency streamlining. Even a sound reform will attract less investment if providers must account for the risk that it will disappear before they recover their costs. Congress should therefore codify the durable core of these reforms.

Codification would do more than settle individual policy disputes. By reducing uncertainty across every project, it would lower the return investors require to commit capital. That reduction would matter most for marginal projects, especially in rural and other hard-to-reach areas where modest changes in cost or risk can determine whether a network gets built.

VI.          Conclusion

Debate over the infrastructure needed for next-generation applications often focuses on data centers and electricity. Yet AI, AR, and VR can reach users only through communications networks built across public land, local rights-of-way, railroad crossings, and utility poles—one approval at a time.

Broadband investment already carries high fixed costs, uncertain demand, obsolescence risk, and long recovery periods. Fragmented permitting compounds those risks. Each additional gatekeeper can impose fees, delay construction, or block a route, while no single authority bears the project’s cumulative cost. Providers must price that uncertainty into every build, even when approvals ultimately arrive on time. The resulting burden falls hardest on rural and underserved projects whose social value may be high but whose expected private returns are thin.

Reform should preserve legitimate environmental, historic-preservation, public-safety, and land-management interests while making approvals faster, more predictable, and more proportionate. Firm shot clocks, self-executing deemed-granted remedies, cost-based fee caps, streamlined federal reviews, modernized franchising, broader pole-access protections, and rapid resolution of pole-replacement disputes would directly reduce deployment barriers. OTMR and dig-once policies can lower construction costs by reducing repeated work. Codifying the durable core of these reforms in federal legislation would further reduce the risk that future regulators will reverse them before investors recover their capital.

The jurisdictions that succeed in the next technology race will need more than computing power and electricity. They will need rules that allow the networks connecting those resources to people and businesses to get built.

[1] Kristian Stout, Infrastructure Is Destiny: The Geography of the Next Technology Race (Int’l Ctr. for L. & Econ. May 26, 2026), https://laweconcenter.org/resources/infrastructure-is-destiny-the-geography-of-the-next-technology-race; Jeffrey Westling, The Invisible Infrastructure: Spectrum Policy for the AI Era (Int’l Ctr. for L. & Econ. July 24, 2026), https://laweconcenter.org/resources/the-invisible-infrastructure-spectrum-policy-for-the-ai-era.

[2] See Intelligent Networks: Powering Artificial Intelligence and Transforming Communications: Hearing Before the Subcomm. on Telecomm. & Media of the S. Comm. on Com., Sci., & Transp., 119th Cong. (2026) (statement of Jonathan Spalter, President & CEO, USTelecom—The Broadband Ass’n), https://www.commerce.senate.gov/wp-content/uploads/meetings/379d9950-dc84-b1bf-2c6a-ee1ebb42fbe2/Spalter_Written_Testimony_7eddc348-33d3-4f43-b974-3023b1205597.pdf.

[3] Jeffrey Westling, 2020 Broadband Scorecard Report (R St. Inst. Feb. 10, 2021), https://www.rstreet.org/research/2020-broadband-scorecard-report.

[4] See, e.g., Comments of the International Center for Law & Economics, The State of Competition in the Communications Marketplace, GN Docket No. 26-78, at 11 (2026), https://laweconcenter.org/resources/icle-comments-to-the-fcc-on-the-state-of-competition-in-the-communications-marketplace [hereinafter ICLE Competition Comments]; Stout, supra note 1, at 18; Jeffrey Westling, The Last Mile Is a Paper Trail: Why Broadband Gets Stuck, Truth on the Mkt. (Apr. 22, 2026), https://truthonthemarket.com/2026/04/22/the-last-mile-is-a-paper-trail-why-broadband-gets-stuck.

[5] See Ben Sperry & Kristian Stout, Issue Brief: Pole Attachments and Broadband Build-Out, at 1 (Int’l Ctr. for L. & Econ. 2021), https://laweconcenter.org/resources/issue-brief-pole-attachments-and-broadband-build-out; Kristian Stout & Ben Sperry, Guiding Principles & Legislative Checklist for Broadband Subsidies (Int’l Ctr. for L. & Econ. 2022), https://laweconcenter.org/resources/guiding-principles-legislative-checklist-for-broadband-subsidies; Jeffrey Westling, The Cost of Holding Up Broadband, Truth on the Mkt. (July 14, 2026), https://truthonthemarket.com/2026/07/14/the-cost-of-holding-up-broadband.

[6] Int’l Energy Agency, Key Questions on Energy and AI (2026), https://www.iea.org/reports/key-questions-on-energy-and-ai; Int’l Energy Agency, Energy and AI (2025), https://www.iea.org/reports/energy-and-ai.

[7] Stout, supra note 1, at 13; Westling, The Invisible Infrastructure, supra note 1, at 17; Jeffrey Westling & Kristian Stout, The AI Boom Still Runs on Old-Fashioned Airwaves, Truth on the Mkt. (July 29, 2026), https://truthonthemarket.com/2026/07/29/the-ai-boom-still-runs-on-old-fashioned-airwaves.

[8] ICLE Competition Comments, supra note 4, at 2; Westling, The Last Mile Is a Paper Trail, supra note 4.

[9] See Comments of the International Center for Law & Economics, Implementing the Infrastructure Investment and Jobs Act: Prevention and Elimination of Digital Discrimination, GN Docket No. 22-69, at 6 (May 16, 2022), https://laweconcenter.org/resources/icle-comments-to-the-fcc-on-prevention-and-elimination-of-digital-discrimination [hereinafter Digital Discrimination Comments]; Geoffrey A. Manne, Kristian Stout & Ben Sperry, A Dynamic Analysis of Broadband Competition: What Concentration Numbers Fail to Capture, at 30 (Int’l Ctr. for L. & Econ. 2021), https://laweconcenter.org/resources/a-dynamic-analysis-of-broadband-competition-what-concentration-numbers-fail-to-capture; ICLE Competition Comments, supra note 4, at 8.

[10] ICLE Competition Comments, supra note 4, at 8; see also Westling, The Last Mile Is a Paper Trail, supra note 4.

[11] ICLE Competition Comments, supra note 4, at 8; Joe Kane, Ben Sperry & Kristian Stout, Brief of ICLE and ITIF to the Eighth Circuit in Minnesota Telecom Alliance v. FCC, at 21 (Int’l Ctr. for L. & Econ. 2024), https://laweconcenter.org/resources/brief-of-icle-and-itif-to-8th-circuit-in-minnesota-telecom-alliance-v-fcc.

[12] ICLE Competition Comments, supra note 4, at 8.

[13] Manne et al., supra note 9, at 10.

[14] See Digital Discrimination Comments, supra note 9, at 4–5; Eric Fruits, Ben Sperry & Kristian Stout, The Competitive Effects of the Proposed Charter/Cox Transaction, at 4–6 (Int’l Ctr. for L. & Econ. 2025), https://laweconcenter.org/resources/the-competitive-effects-of-the-proposed-charter-cox-transaction.

[15] ee Jeffrey Westling & Kristian Stout, ICLE Reply Comments to FCC on Eliminating Barriers to Wireless Deployments, at 12 (Int’l Ctr. for L. & Econ. 2026), https://laweconcenter.org/resources/icle-reply-comments-to-fcc-on-eliminating-barriers-to-wireless-deployments; Eric Fruits & Geoffrey A. Manne, Quack Attack: De Facto Rate Regulation in Telecommunications, at 16 (Int’l Ctr. for L. & Econ. 2023), https://laweconcenter.org/resources/quack-attack-de-facto-rate-regulation-in-telecommunications.

[16] See generally ICLE Competition Comments, supra note 4; see also Comments of the International Center for Law & Economics and Scholars of Law & Economics, Protecting the Privacy of Customers of Broadband and Other Telecommunications Services, WC Docket No. 16-106 (May 27, 2016), https://laweconcenter.org/resources/comments-icle-scholars-law-economics-matter-protecting-privacy-customers-broadband-telecommunications-services.

[17] ICLE Competition Comments, supra note 4, at 10; Westling, The Last Mile Is a Paper Trail, supra note 4.

[18] Westling, The Last Mile Is a Paper Trail, supra note 4.

[19] Chandler Dempsey et al., The Economic Impact of the Rural Digital Opportunity Fund, at 61 (Fla. State Univ. Dep’t of Econ. 2021), https://cosspp.fsu.edu/economics/wp-content/uploads/sites/10/2021/10/TheEconomicImpactoftheRuralDigitalOpportunityFund.pdf.

[20] Eric Fruits et al., Dynamic Competition in Broadband Markets: A 2024 Update, at 42 (Int’l Ctr. for L. & Econ. 2024), https://laweconcenter.org/wp-content/uploads/2024/06/Broadband-Competition-2024-Update.pdf.

[21] R.H. Coase, The Problem of Social Cost, 3 J.L. & Econ. 1 (1960).

[22] Westling, The Last Mile Is a Paper Trail, supra note 4.

[23] Id.

[24] Id.; see also Drew Clark, Railroad Crossing Killing Rural Broadband Deployment, Broadband Breakfast (Nov. 18, 2025), https://broadbandbreakfast.com/railroad-crossing-killing-rural-broadband-deployment-says-virginia-telecom-official.

[25] Ben Sperry et al., ICLE and CEI Amicus to the 6th Circuit on Safeguarding and Securing the Open Internet Order, at 4 (Int’l Ctr. for L. & Econ. 2024), https://laweconcenter.org/resources/icle-and-cei-amicus-to-the-6th-circuit-on-safeguarding-and-securing-the-open-internet-order.

[26] ICLE Competition Comments, supra note 4, at 8.

[27] Inquiry Concerning Deployment of Advanced Telecommunications Capability to All Americans in a Reasonable and Timely Fashion, 2024 Section 706 Report, GN Docket No. 22-270, FCC 24-27 (rel. Mar. 18, 2024), https://docs.fcc.gov/public/attachments/DOC-401205A1.pdf.

[28] Lazar Radic, Geoffrey A. Manne & Dirk Auer, Regulate for What? A Closer Look at the Rationale and Goals of Digital Competition Regulations, 22 Berkeley Bus. L.J. 73 (2025), https://laweconcenter.org/resources/regulate-for-what-a-closer-look-at-the-rationale-and-goals-of-digital-competition-regulations.

[29] Stout, supra note 1, at 18–19.

[30] Id.

[31] Robert D. Cooter & Michael D. Gilbert, Public Law and Economics 42 (Oxford Univ. Press 2022), https://laweconcenter.org/resources/public-law-and-economics (illustrating the unanimity holdout problem through a road crossing multiple parcels, where any owner can block the entire project); see also Comments of the International Center for Law & Economics, Implementation of Section 621(a)(1) of the Cable Communications Policy Act of 1984, as Amended by the Cable Television Consumer Protection and Competition Act of 1992, MB Docket No. 05-311, at 5 (July 18, 2019), https://laweconcenter.org/resources/icle-comments-on-implementation-of-section-621a1-of-the-cable-communications-policy-act-of-1984 (applying Klein-style hold-up theory to show how gatekeepers can extract rents after providers make relationship-specific sunk investments and explaining why the 1984 Act capped franchise fees).

[32] 47 U.S.C. § 332(c)(7)(B)(ii) (requiring state and local governments to act on siting requests “within a reasonable period of time”); id. § 332(c)(7)(B)(iii) (requiring written denials “supported by substantial evidence contained in a written record”).

[33] See Kristen Hite, Cong. Rsch. Serv., IF11932, National Environmental Policy Act: Judicial Review and Remedies (2024), https://www.congress.gov/crs-product/IF11932.

[34] Nikki Chiappa et al., Understanding NEPA Litigation (Breakthrough Inst. 2025), https://thebreakthrough.org/issues/energy/understanding-nepa-litigation.

[35] Seven County Infrastructure Coalition v. Eagle County, 605 U.S. 168 (2025); Kristian Stout, Testimony Before the House Judiciary Subcommittee on the Administrative State, Regulatory Reform, and Antitrust: The 30,000-Foot View: Competition and Regulation in the U.S. Airline Industry 12 (June 24, 2026), https://laweconcenter.org/resources/testimony-of-kristian-stout-to-the-house-judiciary-antitrust-subcommittee-on-competition-in-the-airline-industry.

[36] National Environmental Policy Act of 1969 § 102(2)(C), 42 U.S.C. § 4332(2)(C) (requiring federal agencies to prepare detailed environmental-impact statements for recommendations or reports on “major Federal actions significantly affecting the quality of the human environment”).

[37] Westling, The Last Mile Is a Paper Trail, supra note 4.

[38] Stout, The 30,000-Foot View, supra note 35, at 12.

[39] Seven County, 605 U.S. at 168.

[40] National Historic Preservation Act § 106, 54 U.S.C. § 306108.

[41] Westling, The Last Mile Is a Paper Trail, supra note 4; Geoffrey A. Manne, Gus Hurwitz & Kristian Stout, Putting Politics over Policy at the FCC, Truth on the Mkt. (Oct. 4, 2018), https://truthonthemarket.com/2018/10/04/putting-politics-over-policy-at-the-fcc.

[42] 54 U.S.C. § 302706(b) (requiring agencies to consult with Indian tribes that attach religious and cultural significance to affected historic properties).

[43] Statement of Commissioner Michael O’Rielly, Accelerating Wireless Broadband Deployment by Removing Barriers to Infrastructure Investment, Second Report and Order, WT Docket No. 17-79, FCC 18-30 (2018), https://docs.fcc.gov/public/attachments/FCC-18-30A4.pdf.

[44] Clean Water Act § 404, 33 U.S.C. § 1344 (authorizing Army Corps of Engineers permits for discharges of dredged or fill material into “waters of the United States,” including wetlands); 33 C.F.R. pts. 320–332; see U.S. Env’t Prot. Agency, Section 404 of the Clean Water Act, https://www.epa.gov/cwa-404; Rivers and Harbors Act of 1899 § 10, 33 U.S.C. § 403; U.S. Army Corps of Eng’rs, Section 10/Section 404 Permitting Reference Guide, https://www.spk.usace.army.mil/Portals/12/documents/regulatory/Reg_workshop/2015-11-06/RPW-Sec10-404.pdf.

[45] Clean Water Act § 404(e), 33 U.S.C. § 1344(e) (authorizing the Corps’ general and nationwide permit program); 33 C.F.R. pt. 330 (implementing the program).

[46] Endangered Species Act § 7(a)(2), 16 U.S.C. § 1536(a)(2) (requiring federal agencies, in consultation with the secretary, to ensure that their actions neither jeopardize listed species nor destroy or adversely modify designated critical habitat); 50 C.F.R. pt. 402 (implementing Section 7); U.S. Fish & Wildlife Serv., Section 7 Consultation, https://www.fws.gov/services/consultation-and-technical-assistance/esa-section-7-consultation (explaining that the Fish and Wildlife Service handles terrestrial and freshwater species, while NOAA’s National Marine Fisheries Service handles most marine and anadromous species).

[47] Federal Land Policy and Management Act of 1976 tit. V, 43 U.S.C. §§ 1761–1771 (authorizing the Bureau of Land Management to grant rights-of-way over public lands for communications sites and facilities, among other uses).

[48] Update of the Communications Uses Program, Cost Recovery Fee Schedules, and Section 512 of FLPMA for Rights-of-Way, 89 Fed. Reg. 25,916 (Apr. 12, 2024) (effective May 13, 2024), https://www.federalregister.gov/documents/2024/04/12/2024-06997/update-of-the-communications-uses-program-cost-recovery-fee-schedules-and-section-512-of-flpma-for.

[49] 23 C.F.R. § 645.211 (requiring each state transportation department to adopt an approved utility-accommodation policy governing utilities’ use and occupancy of highway rights-of-way); Fed. Highway Admin., Guidance on Longitudinal Telecommunications Installations on Limited Access Highway Right-of-Way, https://www.fhwa.dot.gov/utilities/001222.cfm (explaining that states allow fiber-optic installation along freeway rights-of-way under utility-accommodation policies, subject to reasonable permitting conditions governing time, place, and manner).

[50] 47 U.S.C. § 254(e) (limiting federal universal-service support to eligible telecommunications carriers designated under § 214(e) and requiring recipients to use that support only for its intended facilities and services); id. § 214(e) (authorizing state commissions—and, in specified circumstances, the FCC—to designate eligible telecommunications carriers).

[51] See Mass. Exec. Off. of Energy & Env’t Affs., MEPA Regulatory Review (2021), https://www.mass.gov/doc/mepa-regulatory-review-presentation/download (reporting that “[a]bout 16 states have ‘little NEPA’ programs”).

[52] Cal. Pub. Res. Code §§ 21000–21189.70 (2026).

[53] See Dana A. Scherer, Cong. Rsch. Serv., R46077, Potential Effect of FCC Rules on State and Local Video Franchising Authorities (2020), https://www.congress.gov/crs-product/R46077.

[54] Cable Communications Policy Act of 1984, Pub. L. No. 98-549, 98 Stat. 2779 (codified as amended at 47 U.S.C. §§ 521–573).

[55] Implementation of Section 621(a)(1) of the Cable Communications Policy Act of 1984, as Amended by the Cable Television Consumer Protection and Competition Act of 1992, Third Report and Order, MB Docket No. 05-311, FCC 19-80 (2019), https://docs.fcc.gov/public/attachments/FCC-19-80A1.pdf.

[56] See American Broadband Deployment Act of 2025, H.R. 2289, 119th Cong. (2025), https://www.congress.gov/bill/119th-congress/house-bill/2289.

[57] Title VI requires cable operators to obtain franchises before providing cable service. 47 U.S.C. § 541(b)(1). Noncable telecommunications and broadband providers instead access local rights-of-way under Section 253, which preserves local authority only to “manage the public rights-of-way” and require “fair and reasonable compensation.” Id. § 253(c). Some local codes therefore distinguish cable-television franchises from the permits and use authorizations that telecommunications providers obtain for rights-of-way access. See, e.g., Wash. Rev. Code § 35.99.040.

[58] 47 U.S.C. § 253(a), (c).

[59] 47 U.S.C. § 332(c)(7)(A), (B)(i)(II).

[60] Fed. Highway Admin., U.S. Dep’t of Transp., Pavement Utility Cuts: Controlling Pavement Cuts by Implementing Policy (2018), https://www.fhwa.dot.gov/utilities/utilitycuts/man03.cfm.

[61] Accelerating Wireless Broadband Deployment by Removing Barriers to Infrastructure Investment, Declaratory Ruling and Third Report and Order, WC Docket No. 17-84, ¶¶ 149–150, FCC 18-111 (2018), https://docs.fcc.gov/public/attachments/FCC-18-111A1.pdf.

[62] Petition for Declaratory Ruling to Clarify Provisions of Section 332(c)(7)(B) to Ensure Timely Siting Review and to Preempt Under Section 253 State and Local Ordinances That Classify All Wireless Siting Proposals as Requiring a Variance, Declaratory Ruling, WT Docket No. 08-165, ¶ 32, FCC 09-99 (2009), https://docs.fcc.gov/public/attachments/FCC-09-99A1.pdf.

[63] Accelerating Wireline Broadband Deployment by Removing Barriers to Infrastructure Investment et al., Declaratory Ruling and Third Report and Order, WT Docket No. 17-79 & WC Docket No. 17-84, ¶¶ 40–46, 104–110, FCC 18-133 (2018), https://docs.fcc.gov/public/attachments/FCC-18-133A1.pdf.

[64] 47 U.S.C. § 1455(a) (providing that state and local governments “may not deny, and shall approve” eligible requests to modify existing wireless towers or base stations when the modification does not substantially change their physical dimensions); Acceleration of Broadband Deployment by Improving Wireless Facilities Siting Policies et al., Report and Order, WT Docket No. 13-238 et al., ¶¶ 205–215, FCC 14-153 (2014), https://docs.fcc.gov/public/attachments/FCC-14-153A1.pdf (adopting a 60-day shot clock and a self-executing “deemed granted” remedy when a locality fails to act within 60 days after receiving the applicant’s notice).

[65] Accelerating Wireline Broadband Deployment, supra note 63, ¶¶ 84–90 (governing how incompleteness determinations affect the shot clocks).

[66] 47 U.S.C. § 224(b)(1).

[67] 47 U.S.C. § 224(a)(1).

[68] 47 C.F.R. § 1.1402(o).

[69] Westling, The Last Mile Is a Paper Trail, supra note 4.

[70] Id.

[71] Accelerating Wireline Broadband Deployment by Removing Barriers to Infrastructure Investment, Declaratory Ruling, WC Docket No. 17-84, ¶¶ 8–22, DA 21-78 (WCB 2021), https://docs.fcc.gov/public/attachments/DA-21-78A1.pdf (finding that utilities often charged new attachers the full cost of replacing poles already near the end of their useful lives or otherwise due for replacement and clarifying that new attachers need not pay to correct preexisting violations or replace “red-tagged” poles that utilities must replace).

[72] Id. ¶¶ 8–26.

[73] Accelerating Wireline Broadband Deployment, supra note 61, ¶¶ 33–86.

[74] For recent enforcement, see Fed. Commc’ns Comm’n, FCC Issues Precedent-Setting Order Under Expedited Review Process for Pole Attachment Complaints (Feb. 5, 2026), https://docs.fcc.gov/public/attachments/DOC-418491A1.pdf.

[75] Brendan Carr, Chairman, Fed. Commc’ns Comm’n, A Build Agenda for America (July 2, 2025), https://docs.fcc.gov/public/attachments/DOC-412663A1.pdf; see also Fed. Commc’ns Comm’n, Chairman Carr’s Build America Agenda, https://www.fcc.gov/build-america.

[76] 40 C.F.R. § 1501.4; 42 U.S.C. § 4336(a)(2).

[77] 36 C.F.R. § 800.14(b); see also 54 U.S.C. § 306108.

[78] Seven County, 605 U.S. at 168.

[79] Sackett v. EPA, 598 U.S. 651 (2023) (holding that the Clean Water Act covers only wetlands with a continuous surface connection to bodies that constitute “waters of the United States” in their own right, making the wetlands indistinguishable from those waters).

[80] United Keetoowah Band of Cherokee Indians v. FCC, 933 F.3d 728 (D.C. Cir. 2019) (vacating the FCC’s 2018 order insofar as it excluded small wireless-facility deployments from NEPA and National Historic Preservation Act Section 106 review because the FCC failed to justify that deregulation served the public interest or adequately address potential effects on historic and tribal resources).

[81] American Broadband Deployment Act of 2025, supra note 56; BROADBAND Leadership Act, H.R. 278, 119th Cong. (2025), https://www.congress.gov/bill/119th-congress/house-bill/278/text.

[82] Build America: Eliminating Barriers to Wireless Deployments, Notice of Proposed Rulemaking, WT Docket No. 25-276 (Sept. 2025), https://docs.fcc.gov/public/attachments/DOC-414414A1.pdf.

[83] American Broadband Deployment Act of 2025, supra note 56.

[84] Petition of Gateway Infrastructure, LLC d/b/a Gateway Fiber for Preemption and Declaratory Ruling Pursuant to Section 253(d) of the Communications Act of 1934, WC Docket No. 26-166 (filed June 17, 2025), https://www.fcc.gov/ecfs/search/search-filings/filing/26109834869 [hereinafter Gateway Fiber Petition]; see Wireline Competition Bureau Seeks Comment on Petitions for Preemption and Declaratory Ruling, Public Notice, DA 26-689 (WCB July 8, 2026), https://docs.fcc.gov/public/attachments/DA-26-689A1.pdf.

[85] Gateway Fiber Petition, supra note 84; see 47 U.S.C. § 253(a), (d).

[86] Wireline Competition Bureau Seeks Comment on Petitions for Preemption and Declaratory Ruling, supra note 84.

[87] Jeffrey Westling, Fiber Freeze: How Maple Grove Made a Cable Franchise the Price of Broadband, Truth on the Mkt. (Aug. 20, 2026), https://truthonthemarket.com/2026/08/20/fiber-freeze-how-maple-grove-made-a-cable-franchise-the-price-of-broadband.

[88] Accelerating Wireline Broadband Deployment by Removing Barriers to Infrastructure Investment, Fifth Report and Order, Fourth Further Notice of Proposed Rulemaking, and Orders on Reconsideration, WC Docket No. 17-84, FCC 25-38 (July 24, 2025), https://docs.fcc.gov/public/attachments/FCC-25-38A1.pdf.

[89] Ben Sperry, Geoffrey A. Manne & Kristian Stout, The Role of Antitrust and Pole-Attachment Oversight in TVA Broadband Deployment, 29 J. Tech. L. & Pol’y 73 (2025), https://scholarship.law.ufl.edu/cgi/viewcontent.cgi?article=1243&context=jtlp.

[90] Ben Sperry, Broadband Deployment, Pole Attachments, & the Competition Economics of Rural-Electric Co-ops, Truth on the Mkt. (Aug. 16, 2023), https://truthonthemarket.com/2023/08/16/broadband-deployment-pole-attachments-the-competition-economics-of-rural-electric-co-ops.

[91] Accelerating Wireline Broadband Deployment by Removing Barriers to Infrastructure Investment, Fourth Report and Order, Declaratory Ruling, and Third Further Notice of Proposed Rulemaking, WC Docket No. 17-84, FCC 23-109 (rel. Dec. 15, 2023), https://docs.fcc.gov/public/attachments/FCC-23-109A1.pdf; see also Westling, The Cost of Holding Up Broadband, supra note 5.

[92] Westling, The Cost of Holding Up Broadband, supra note 5.

[93] Broadband Infrastructure Deployment, 86 Fed. Reg. 68,553 (Dec. 3, 2021), https://www.federalregister.gov/documents/2021/12/03/2021-26231/broadband-infrastructure-deployment; see also 23 C.F.R. § 645.307(a); MOBILE NOW Act § 607, Pub. L. No. 115-141, div. P, tit. VI, 132 Stat. 1097, 1104 (2018); U.S. Gov’t Accountability Off., GAO-26-107734, Broadband Infrastructure: States’ Reported Progress Implementing Requirements to Facilitate Deployment Along Federal-Aid Highways (2026), https://www.gao.gov/products/gao-26-107734.

Executive Summary

The European Commission responded very differently to two sets of courtroom defeats. After the General Court annulled the merger prohibitions in Airtours, Schneider, and Tetra Laval in 2002, the Commission launched sweeping institutional, procedural, and substantive reforms. Its response to the later annulments of infringement decisions under Article 102 of the Treaty on the Functioning of the European Union (TFEU), which prohibits abuse of a dominant position, has been far narrower.

The 2002 merger judgments exposed weak internal checks, procedural defects, insufficient evidence, and economic analysis too dependent on market structure. The Commission responded by reorganizing merger review, strengthening the Hearing Officer, creating independent economic and peer review, expanding parties’ access to evidence, and adopting a broader effects-based legal framework. These reforms improved the Commission’s ability to test its theories before acting and produce decisions capable of surviving judicial review.

Intel, Qualcomm, and Bulgarian Energy Holding revealed similar weaknesses in Article 102 enforcement. The courts found failures to document and disclose relevant third-party evidence, consider exculpatory material, and connect theories of harm to realistic exclusionary effects. The judgments confirm that the Commission retains discretion to pursue legitimate theories of abuse. It must, however, support those theories with coherent evidence, respect the rights of defense, and show that the challenged conduct could exclude competitors under actual market conditions.

These decisions warrant a targeted review of the safeguards governing Article 102 investigations. The revision of Regulation 1/2003 should strengthen interview records, case-file completeness, access to evidence, independent economic review, and internal challenges to case teams’ conclusions. The Commission’s draft Article 102 guidelines should also reflect the courts’ effects-based jurisprudence rather than rely on presumptions that substitute conduct categories for competitive analysis. Strong internal checks will conserve enforcement resources, improve legal certainty, and reinforce the legitimacy of Commission decisions as competition policy faces greater political and industrial-policy pressures.

Introduction

The European Commission has responded very differently to two sets of courtroom defeats. In 2002, the Court of First Instance—now the General Court—annulled three merger prohibitions in quick succession: Airtours, Schneider, and Tetra Laval.[1] The reversals became the Commission’s “Airtours moment.” It quickly launched major institutional, procedural, and substantive reforms. That reform project continues in its April 2026 draft merger guidelines.

More recently, European Union courts overturned Commission infringement decisions under Article 102 of the Treaty on the Functioning of the European Union (TFEU), which prohibits abuse of a dominant market position. The principal cases were Intel, Qualcomm, and Bulgarian Energy Holding.[2] These decisions echoed the 2002 merger cases. Across both groups, the courts stressed the rights of defense and to be heard, including access to a complete evidentiary file and exculpatory material; analysis of competitive effects grounded in actual market conditions; and evidence strong enough to support the Commission’s theory of harm.

Yet the Commission’s responses diverged. It answered the 2002 judgments with broad merger reforms. Its response to the Article 102 judgments has been narrower and, in one important respect, counterintuitive. The Commission’s 2024 draft Article 102 guidelines rely more heavily on conduct categories and presumptions, even as the case law demands evidence that conduct can exclude competitors.

Merger review and unilateral-conduct enforcement differ in important ways. Merger analysis asks what a proposed transaction is likely to do, usually under tight statutory deadlines. Article 102 cases examine past conduct and often unfold over years. Yet the same institutional principle links both fields: Sound internal checks and strong due-process protections improve accuracy and produce decisions more likely to survive judicial review. That premise matters as European competition enforcement faces heightened public and political scrutiny.

Section 1 of this issue brief examines the merger annulments in Airtours, Schneider, and Tetra Laval. Section 2 considers the institutional, procedural, and substantive reforms that followed. Section 3 turns to the Article 102 annulments in Intel, Qualcomm, and Bulgarian Energy Holding. Section 4 draws together the broader implications, and Section 5 concludes.

I. The Airtours Moment: Three Merger Annulments

European merger control reached what contemporary observers called its “high watermark” between 1999 and 2001. The European Commission prohibited an unprecedented five transactions in 2001 alone, drawing the attention of business leaders and investors worldwide to its merger-review process.[3]

That record of aggressive enforcement met a sharp judicial response in 2002. Within five months, the General Court annulled the Commission’s merger prohibitions in Airtours, Schneider, and Tetra Laval. No European Union court had previously annulled a Commission merger prohibition. The three judgments therefore challenged both the outcomes of these cases and the methods that produced them.

The judgments exposed three related weaknesses. The Commission’s institutional structure lacked effective internal checks and procedural safeguards. Its evidence did not meet the required legal standard. And its analysis relied too heavily on formal indicators such as market shares and concentration, without adequately establishing the likely competitive effects. By scrutinizing the Commission’s evidence and economic reasoning more closely, the court established principles that reshaped European merger control and drove reforms that continue to influence it today.

A. Weak Internal Checks and Procedural Safeguards

One contemporary account described the Airtours, Schneider, and Tetra Laval judgments as exposing a “poorly organized merger review process” in which the Commission failed to compile evidence carefully or ground its decisions in sufficient facts.[4]

The criticism centered on the Commission’s institutional structure for merger review. Within the Directorate-General for Competition (DG Competition), the Merger Task Force assessed the evidence, developed a theory of harm—an explanation of how a merger could harm competition—decided whether the evidence established that theory, and recommended either approval, approval subject to conditions, or prohibition.[5]  Effective internal safeguards did not adequately counterbalance this concentration of responsibilities. In Airtours, Schneider, and Tetra Laval, the General Court ultimately found that the evidence did not support the Commission’s decisions.

In Schneider, the General Court found that the Commission’s Statement of Objections—the formal document setting out its competition concerns—“did not permit Schneider to assess the full extent of the competition problems” alleged by the Commission.[6] That omission denied Schneider an opportunity to propose, “properly and in good time,” remedies broad enough to address those concerns.[7] Because remedies offered the only practical means of avoiding prohibition, the court considered this defect especially serious. It held that the Commission had violated Schneider’s rights of defense in several respects, invalidating the decision.[8]

The reasoning in Airtours was similar. The Commission’s manifest errors deprived Airtours of a meaningful opportunity to propose effective remedies. The Commission neither articulated its theory of harm clearly nor disclosed it early enough, then rejected Airtours’ final commitments because they came too late in the process.

In Tetra Laval, the General Court restated the principles governing access to the Commission’s case file.[9] Access allows parties to exercise their rights of defense effectively, particularly when they face an adverse decision, fine, or penalty. Although denying adequate access can justify annulment, the court found no such violation in Tetra Laval.[10]

More broadly, the three cases exposed the costs of weak internal checks. Without an effective mechanism for identifying evidentiary and analytical flaws early, judicial review became even more important. The reforms discussed below targeted these weaknesses.

B. The Commission’s Evidentiary Burden

The second strand of the merger judgments concerned the evidence needed to justify prohibition. In Airtours, the General Court held that the Commission must produce “convincing evidence” of its theory of harm[11] and “prove conclusively” that the merger would create an anticompetitive market structure.[12] The alleged harm was collective dominance—a situation in which several firms can coordinate their conduct without an express agreement.

The Commission fell short. The court found that, “far from basing its prospective analysis on cogent evidence,” the Commission had made fundamental errors in assessing whether the merger would create collective dominance. [13]  It therefore had not proved to the required legal standard that the merger would impede effective competition.

The General Court identified similar defects in Schneider. It called some of the Commission’s findings “abstract,”[14] found that the supporting evidence was sometimes insufficient,[15] and criticized other findings as lacking evidence or precision.[16] Taken together, these errors deprived the Commission’s economic assessment of probative value—the capacity to prove the proposition for which it was offered.[17]

In Tetra Laval, the Commission likewise failed to meet its burden. It had not shown convincingly that Tetra Laval could leverage its dominant position—use market power in one market to gain dominance in another—or that eliminating Sidel as a potential competitor would strengthen the merged company’s position.[18] The court found “no cogent evidence” supporting the latter claim.[19]

Because merger review requires predictions about future market conditions, the court required “sufficiently convincing” evidence and a “particularly plausible” forecast.[20] It found neither.[21]

C. From Market Structure to Competitive Effects

The General Court did not reject the theories of harm themselves: collective dominance in Airtours and conglomerate leveraging in Tetra Laval. Instead, it demanded rigorous economic analysis showing that the predicted anticompetitive effects were likely. In Schneider, the court found that the Commission had not met its burden of proof. Its analysis lacked a coherent economic foundation because it did not properly evaluate countervailing evidence and relied on inconsistent or unclear market definitions.[22] The judgment shifted the inquiry away from formal indicators such as market structure, market shares, and concentration measures and toward likely competitive effects.

In Airtours, the Commission relied heavily on a reduction from four major competitors to three. It argued that the remaining firms would have strong incentives to coordinate tacitly—that is, align their conduct without an express agreement. The General Court held that this structural change alone could not establish collective dominance. The required prospective analysis “calls for close examination in particular of the circumstances which are relevant for assessing the effects of the concentration on competition,” supported by convincing evidence.[23]

The court then identified conditions relevant to whether tacit coordination could endure. Consumers might defeat it by switching to smaller competitors. Coordination might also unravel if the three large firms lacked an effective way to punish a member that departed from their shared course of conduct.[24] By requiring evidence on these mechanisms, the court aligned the legal concept of collective dominance with the economic concept of tacit coordination and supplied a framework for future cases.[25]

In Tetra Laval, the General Court recognized that the Merger Regulation can reach conglomerate mergers—transactions involving firms in neighboring or complementary markets.[26] But the Commission had to establish the alleged conglomerate effects through cogent evidence, not speculation about future conduct.[27] It needed “sufficiently convincing evidence” of both the likelihood and the consequences of leveraging and could consider only conduct that probably would be lawful.[28] Although the Commission retained some discretion, its prediction of the merged company’s conduct had to be “particularly plausible.”[29]

Nicholas Levy summarized the resulting standard: The Commission must show more than the mere possibility of leveraging. It must establish that the merger would, in all likelihood and in the relatively near future, create or strengthen a dominant position based on close examination of the relevant circumstances.[30] Together, these cases largely cemented an effects-based approach to European merger control.

II. From Judicial Defeat to Merger Reform

The three annulments arrived at a pivotal point in the Commission’s modernization of European merger control. The Commission had submitted a report to the Council in 2000, published a Green Paper reviewing the existing Merger Regulation in December 2001, and opened a public consultation.[31] Airtours, Schneider, and Tetra Laval transformed that planned modernization into a broader and more urgent overhaul. As Nicholas Levy observed, “the judicial defeats of 2002 provided the catalyst for a series of reforms” intended to make merger review “more structured, firmly grounded in sound economics, and consistently based on an objective assessment of quantitative evidence.”[32]

European Competition Commissioner Mario Monti acknowledged in November 2002 that the judgments, “no matter how painful, came at a right moment.” He urged the Commission to turn those setbacks into “an opportunity for even deeper reform than originally envisaged” and announced what he called “the most far-reaching reform” of European merger control since the first Merger Regulation took effect in 1989.[33]

The Commission approved the core reform package in December 2002 and implemented its principal elements through 2004. The package combined institutional, procedural, and substantive changes: stronger internal checks and independent economic review; earlier access to evidence and greater engagement with merging parties; and a revised Merger Regulation, new horizontal-merger guidelines, and a more effects-based analytical framework. Together, these reforms addressed the central weaknesses identified in the three judgments and reshaped how the Commission investigated, analyzed, and decided merger cases.

A. Building Internal Checks and Economic Expertise

European European Competition Commissioner Mario Monti responded to the merger annulments by strengthening the Directorate-General for Competition’s internal checks and balances.[34] He treated the quality of the decision-making process as central to the reform:

We will adopt changes, as radical as needed, to ensure that our merger investigations are conducted in a manner which is more thorough and more firmly grounded in economic reasoning and to further strengthen the due process guarantees built into our merger proceedings.[35]

Monti first reorganized merger review. He disbanded the centralized Merger Task Force and placed merger units within the agency’s sector-specific directorates. He also created the position of chief economist, who reported directly to the director-general and led a team charged with providing independent economic analysis and advising decision-makers at every level of an investigation.

Monti further established peer review panels composed of experienced Directorate-General officials who had not worked on the case. The panels reviewed case teams’ preliminary conclusions and challenged their assumptions. Monti explained that this independent scrutiny would counter investigators’ natural tendency to become convinced by their own arguments. A dedicated support unit helped make the panels an effective check on the soundness of preliminary findings.[36]

The reforms also expanded the Hearing Officer’s mandate, emphasizing the independence and integrity needed to preserve objective, transparent, and efficient proceedings.[37] The Hearing Officer reported to the competition commissioner rather than the Directorate-General, strengthening the office’s ability to protect merging parties’ rights of defense and oversee the fair conduct of proceedings.[38]

Pablo Ibáñez Colomo later connected these reforms to the broader value of judicial review:

One can think of a number of reasons why more robust checks on administrative action might lead to better outcomes, both from a legal and a non-legal standpoint. This conclusion is, if anything, confirmed by the very history of EU merger control, which shows that efforts to improve enforcement may be prompted by judicial review.[39]

The reforms following Airtours, Schneider, and Tetra Laval illustrate that principle. More extensive judicial review prompted the Commission to build stronger internal safeguards, deepen its economic expertise, and improve the quality of merger decisions.

B. Earlier Access and Procedural Engagement

The Commission answered criticism of its fact-finding on two fronts. Internally, the new checks and balances allowed officials outside the case team to test its preliminary findings before the Commission reached a formal decision. Greater management oversight also helped ensure that investigations received adequate resources. Monti instructed managers to “see to it that due attention is paid to the quality of evidence, and to ensuring that decisions reference such evidence clearly and comprehensively.”[40]

The Commission also gave notifying parties—the companies seeking merger approval—earlier access to the evidentiary record and the case team. Once the Commission opened an in-depth, or Phase II, investigation through an Article 6(1)(c) decision, the parties could review the file and receive ad hoc access to important third-party submissions that contradicted their position. This earlier disclosure allowed them to challenge those submissions before the Commission issued its statement of objections, rather than waiting until the Commission had formally set out its concerns.

The reforms created additional opportunities for direct engagement. Before issuing a statement of objections, the Commission could hold “triangular” meetings among the case team, the notifying parties, and third-party complainants. It also introduced state-of-play meetings at decisive stages of an investigation, giving the parties opportunities to discuss the case with senior Directorate-General for Competition officials. [41]

The Commission formalized these changes in its Best Practices on the Conduct of EC Merger Control Proceedings, adopted Jan. 20, 2004.[42] The Merger Best Practices provide that, immediately after the Commission opens an in-depth investigation, notifying parties should receive key documents, including “substantiated submissions of third parties running counter to the notifying parties’ own contentions.”[43] The Directorate-General also committed to “use its best endeavours” to provide timely access to those documents and to make further disclosures on an ad hoc basis as the investigation developed.[44]

Together, these measures moved evidentiary testing earlier in the process, when merging parties could still correct the factual record, answer third-party allegations, and address the Commission’s concerns before its position hardened.

C. Embedding Effects-Based Merger Analysis

Monti also announced guidelines that would provide a “sound economic framework” for assessing horizontal mergers—transactions between actual or potential competitors. The guidelines would explain how the Commission assesses a merger’s likely competitive effects and applies the concept of collective dominance. They would also address factors that could counter an initial finding of likely harm, including buyer power, ease of entry, and efficiencies.[45]  The Commission adopted its Horizontal Merger Guidelines in 2004.[46]

The Commission gave this economic framework institutional support by creating the position of chief competition economist and recruiting more economists who specialized in how firms compete and markets operate.[47] These changes helped translate the guidelines’ effects-based principles into case investigations and decisions.

The reforms also changed the law governing merger review. The revised European Community Merger Regulation (ECMR), Regulation 139/2004, took effect May 1, 2004. It replaced the former dominance test with a broader inquiry into whether a transaction would “significantly impede effective competition in the common market or in a substantial part of it, in particular as a result of the creation or strengthening of a dominant position.” The new test retained dominance as a central concern but allowed the Commission to reach mergers that could substantially harm competition without creating or strengthening a dominant firm. Without the General Court’s judgments in Airtours, Schneider, and Tetra Laval, the substantive test likely would have remained largely unchanged.[48]

Two decades later, Guillaume Loriot, deputy director-general for mergers at the Directorate-General for Competition, credited the judgments with forcing the Commission into “deeper retrospection.” That process, he said, enabled “more rigorous, and all-encompassing reforms” that changed European Union merger control “for the better.”[49]

The Commission extended that reform trajectory when it published draft revised merger guidelines April 30, 2026, combining its horizontal and nonhorizontal guidelines into a single framework.[50] Executive Vice President Teresa Ribera had previewed the draft’s “more robust treatment of dynamic effects”—effects on innovation, investment, and competition over time—including both dynamic harms and efficiencies.[51]

That emphasis carries forward the central lesson of Airtours, Schneider, and Tetra Laval: The Commission must ground predictions of competitive harm in rigorous economic analysis and convincing evidence.

III. Article 102 Meets Closer Judicial Scrutiny

The Commission entered the Intel litigation with a formidable record in Article 102 appeals. Thibault Schrepel and Janneke Parrish identify Intel as its only Article 102 loss before the Court of Justice during Margrethe Vestager’s decade as competition commissioner.^[52] Pinar Akman accordingly describes the 2017 ruling as “a clear turning point in the evolution of the law on the abuse of dominance.[53]

The Intel litigation ultimately produced three principal judgments: Intel 2017, which set aside the General Court’s 2014 judgment; Intel Renvoi, which annulled the Commission’s findings concerning Intel’s rebates and the resulting fine; and Intel 2024, which upheld the remand judgment.[54] The General Court later annulled Commission infringement decisions in Qualcomm and Bulgarian Energy Holding. Those judgments form part of a broader line of Article 102 cases demanding careful treatment of evidence and rigorous analysis of exclusionary effects.[55]

The cases echo the 2002 merger judgments in three respects. First, they require the Commission to document and disclose relevant third-party evidence, while distinguishing procedural lapses that materially impair a company’s defense from those that do not. Second, they require the Commission to consider evidence supporting and undermining its theory and to meet its burden with a coherent factual record. Third, they test theories of exclusion against actual market conditions, requiring evidence that the challenged conduct was capable of producing more than hypothetical anticompetitive effects. The courts did not deny the Commission discretion to pursue particular theories of harm. They instead required the Commission to prove that those theories fit the facts.

A. Access to the File and Rights of Defense

The three Article 102 judgments returned to a central theme of the merger cases: The Commission must give investigated companies a meaningful opportunity to test the evidence against them and present a full defense. That obligation covers evidence supporting the alleged infringement and potentially exculpatory material—evidence that could help disprove it—obtained through interviews with third parties. In Qualcomm and Bulgarian Energy Holding, failures to record and disclose such evidence materially impaired the companies’ ability to defend themselves. Intel involved a similar recordkeeping failure, but the courts found that it did not justify annulment.

Intel argued that the Commission had withheld potentially exculpatory information from a third-party interview, refused access to other third-party documents, and improperly denied a second oral hearing. The internal note that the Commission eventually disclosed contained no meaningful account of the interview, including “any indication of the content of the discussions” or “the nature of the information” that the interviewee had provided.[56] The Court of Justice nevertheless concluded that Intel had not shown that access to a fuller record during the administrative proceeding could have improved its defense.[57]

The Intel litigation clarified the Commission’s obligation to document third-party interviews.[58] The Commission must make an adequate record of every interview conducted to collect information about an investigation, including the substance of the information provided. This procedural duty complements the broader requirement that competition authorities meaningfully examine evidence submitted by an investigated company to show that its conduct could not restrict competition.[59] A failure to document an interview does not automatically invalidate a decision, but it may do so when the missing information could have materially assisted the company’s defense.

Qualcomm illustrates that distinction. The Commission’s 2018 decision found that Qualcomm had abused its dominant position by paying Apple to obtain its baseband chipsets exclusively from Qualcomm.[60] The General Court annulled the decision in full, relying on procedural defects that impaired Qualcomm’s rights of defense and flaws in the Commission’s analysis of competitive effects. It found that “the administrative procedure which led to the adoption of the contested decision is vitiated by a number of procedural errors which affected the applicant’s rights of defence.”[61] Alfonso Lamadrid described the court’s willingness to annul an Article 102 decision in full after finding such procedural errors as extraordinary.[62]

The Commission had failed to record properly or disclose fully several meetings and interviews with Qualcomm’s competitors and customers.[63] The General Court examined the withheld information’s probative value to determine whether the omissions had impaired Qualcomm’s defense. Some of that information went to a central issue: whether Intel had a viable competing chipset that Qualcomm’s payments could have excluded from the market.[64] Lamadrid argued that the judgment did not create new law but “simply requires the Commission to fulfill its (post-Intel) obligations regarding both procedure and substance” through a “clear, logical and thorough application of the law.”[65]

The General Court applied the same principles in Bulgarian Energy Holding. The Commission’s 2018 BEH Gas decision found that the Bulgarian Energy Holding (BEH) Group had restricted competitors’ access to essential gas infrastructure.[66] he court annulled that decision Oct. 25, 2023, on both substantive and procedural grounds. Although it upheld one isolated instance of exclusionary conduct, that conduct could not support the Commission’s finding of a single and continuous infringement.[67] The Commission appealed, and the case remained pending as of Aug. 31, 2026.[68]

The General Court also found that the Commission had infringed the BEH Group’s rights of defense by failing to document certain third-party interviews and disclose potentially exculpatory evidence. The omitted information proved important to BEH’s defense and contributed to the Commission’s decision. The court rejected any “supposed discretion” to exclude relevant evidence from the file. Such discretion would impair the undertaking’s defense and prevent the court from verifying the Commission’s compliance with Regulation 1/2003.[69]

Drawing on Intel and Qualcomm, the court reiterated that the Commission must record third-party meetings and interviews and give the investigated undertaking access to every document that might assist its defense. The principal exceptions cover other companies’ business secrets, internal Commission documents, and other confidential information.[70] Access to the file promotes equality of arms—the principle that each side must have a fair opportunity to present its case—and remains “integral” to the rights of defense.[71]

These cases establish a materiality standard rather than a rule of automatic annulment. In Intel, the inadequate record concerned one meeting with one third party, and Intel had received at least limited information about its contents. The General Court was therefore not persuaded that the omission materially impaired Intel’s defense.[72] Qualcomm involved an “exceptional number of procedural violations”: seven improperly recorded meetings with six third parties, including one meeting central to the Commission’s analysis and another involving a source who claimed anonymity.[73] BEH likewise involved omitted evidence that bore directly on the company’s defense and the Commission’s conclusions.

The parallel with the 2002 merger cases is direct. Airtours, Schneider, and Tetra Laval exposed investigative and decision-making powers insufficiently constrained by procedures capable of identifying unsupported conclusions before the Commission acted. Intel, Qualcomm, and Bulgarian Energy Holding reveal similar weaknesses in Article 102 enforcement. The later judgments therefore reinforce Monti’s earlier recognition that third parties can influence Commission investigations and that investigated companies need timely, meaningful opportunities to test their allegations and defend themselves.[74]

B. Evidence of Exclusionary Effects

The three Article 102 cases echoed the evidentiary concerns in Airtours, Schneider, and Tetra Laval. In each case, the court found that the Commission’s analysis had disregarded relevant evidence, including information that undermined its theory of harm. Merger review and Article 102 enforcement ask different legal questions: Merger review predicts a transaction’s likely effects, while Article 102 generally examines whether past conduct was capable of excluding competitors. Both inquiries nevertheless require concrete evidence grounded in actual market conditions, not abstract or hypothetical conclusions.

In Intel Renvoi, the General Court found that the Commission had not established to the required legal standard that Intel’s rebates were capable of foreclosing competitors.[75] The Commission had to examine all relevant circumstances and support its conclusion with a “precise and consistent body of evidence” that left “no residual doubt.” That obligation carried particular weight because Intel had offered a plausible alternative explanation for its conduct—one under which the rebates could not restrict competition or produce the alleged foreclosure effects.[76]

Jacques Buhart et al. identified three related defects in the Commission’s effects analysis in Qualcomm. First, the evidence was inconsistent or contradictory. Second, the Commission failed to consider “all the relevant factors,” including evidence that undercut its theory. Third, the remaining evidence “did not make it possible to support the Commission’s findings.” The analysis therefore could not substantiate the Commission’s conclusions to the required legal standard.[77]

The General Court applied the same evidentiary discipline in Bulgarian Energy Holding. The case involved an alleged refusal to grant access to an essential facility—infrastructure that competitors needed to operate in the related market. The Commission bore the burden of proving that the refusal was capable of producing exclusionary effects. More specifically, it had “to prove that that refusal was liable to eliminate all competition, on the part of the person requesting that service, on the related market.”[78] The Commission also had to establish that the complainant was an actual or potential competitor with a “sufficiently advanced project to enter the market.”[79]

The Commission failed to make that showing. In both Qualcomm and Bulgarian Energy Holding, the evidence did not establish that the firms allegedly excluded were viable actual or potential competitors. Without that connection between the challenged conduct and a realistic prospect of entry or expansion, the alleged exclusionary effects remained hypothetical.

These judgments do not require proof of actual anticompetitive effects in every Article 102 case. They do require the Commission to demonstrate, through coherent evidence and a complete assessment of market conditions, that the conduct was capable of producing those effects. That requirement reflects the same principle that shaped the merger judgments: A legally plausible theory of harm cannot substitute for proof that the theory fits the facts.

C. Testing Theories Against Market Reality

As w in the merger cases, the courts did not prevent the Commission from pursuing particular theories of harm in Intel, Qualcomm, or Bulgarian Energy Holding. They instead required rigorous economic analysis showing that the alleged exclusionary effects were plausible under actual market conditions.

The Court of Justice began that shift in its 2017 Intel decision.[80] When a dominant undertaking submits evidence that its loyalty rebates could not restrict competition, the Commission must examine all relevant circumstances. The court identified five factors: the practice’s market coverage, the conditions governing the rebates, their duration, their amount, and evidence of a strategy to exclude competitors as efficient as the dominant undertaking.[81]

On remand, the General Court examined the Commission’s application of the as-efficient-competitor (AEC) test. That test asks whether the challenged conduct could exclude a hypothetical competitor that operates as efficiently as the dominant firm. The court found errors in the Commission’s analysis and held that it had not adequately established the rebates’ capacity to foreclose such a competitor. It therefore annulled the relevant portions of the 2009 decision and the fine.[82]

The Commission appealed, but the Court of Justice upheld the Intel Renvoi judgment in October 2024.[83] It confirmed that the General Court must examine any argument capable of undermining the Commission’s AEC analysis, including challenges to the governing methodology and the probative value of the evidence. The resulting framework looks beyond structural indicators. It requires the Commission to assess the alleged exclusionary effect in light of all relevant facts, including market coverage, the terms and duration of the rebates, their amount, and any strategy to exclude as-efficient competitors.[84]

The General Court applied the same approach in Qualcomm. Although the procedural violations independently justified annulment, the court also reviewed the substance of the Commission’s decision in the interests of the administration of justice.[85] The Commission had failed to account for evidence that competing chipsets could not satisfy most of Apple’s technical requirements. Because Apple could not have switched those purchases to Qualcomm’s rivals even without the challenged payments, the payments could not have excluded competition for that demand.[86]

The court rejected an analysis based on a presumption that Qualcomm’s exclusivity payments were unlawful. It required the Commission to consider all relevant factual circumstances and establish effects that were more than hypothetical.[87] As the court explained:

[N]ot every exclusionary effect is necessarily detrimental to competition. Competition on the merits may, by definition, lead to the departure from the market or the marginalisation of competitors that are less efficient and so less attractive to consumers from the point of view of, among other things, price, choice, quality or innovation.[88]

Once Qualcomm challenged the Commission’s analysis, the Commission had to assess both the payments’ capacity to produce anticompetitive effects and their ability to foreclose at least as-efficient competitors.[89]

Bulgarian Energy Holding applied the same principle to a refusal-to-supply theory. The General Court emphasized that conduct restricting another company’s options does not necessarily restrict competition within the meaning of Article 102.[90] The Commission had to prove that the refusal to grant access to essential infrastructure was “liable to eliminate all competition, on the part of the person requesting that service, on the related market.”[91] The alleged exclusionary effects could not remain purely hypothetical.[92]

That inquiry required evidence that the party seeking access was an actual or potential competitor with a “sufficiently advanced project to enter the market.”[93] The complainant’s plans remained imprecise, and it had made no tangible access request that could support a credible refusal-to-supply claim.[94] The Commission therefore failed to show that the BEH Group’s conduct was capable of restricting competition in the Bulgarian gas market.[95]

Together, these cases move Article 102 analysis away from conclusions based principally on market structure, the form of the conduct, or presumptions of harm. The relevant question is how the conduct operates under actual market conditions, including whether a credible rival could enter, expand, or compete effectively. Ibáñez Colomo has similarly linked stronger judicial review to improvements in Commission enforcement and greater checks on administrative action.[96]

Akman argues that the shift toward economic, effects-based analysis emerged “through not wins but losses for the Commission and/or the Legal Service on appeals.” In her account, “[i]t is the CJEU which has actively—and perhaps despite the Commission and/or the Legal Service—reformed its jurisprudence to adopt an effects-based approach.”[97] As in the merger cases, judicial reversal supplied the pressure for doctrinal reform.

IV. Judicial Defeat and Institutional Learning

Successful challenges to Commission merger prohibitions were so rare before Airtours, Schneider, and Tetra Laval that an appeal could appear futile. The three annulments changed that perception by demonstrating that merger decisions faced meaningful judicial oversight. Commisioner Monti acknowledged that the judgments proved “beyond any doubt” that European Union courts conducted “meticulous and stringent” review of the Commission’s substantive analysis. He also recognized the practical consequence: “[I]t is clear that the [General Court] is now holding us to a very high standard of proof, and this has clear implications for the way in which we conduct our investigations and draft our decisions.”[98] Werner Berg summarized the lesson: The Commission had to strengthen its substantive review and economic analysis to produce decisions capable of surviving “in-depth juridical review.”[99]

The Commission had compiled a similarly strong record defending its Article 102 decisions before Intel, Qualcomm, and Bulgarian Energy Holding. Yet those judgments did not prompt a comparable public reckoning. The Commission’s responses to Qualcomm and Bulgarian Energy Holding largely stated that it would study the judgments and consider its next steps.[100] Its appeals in Intel and Bulgarian Energy Holding may suggest that it viewed the General Court’s judgments primarily as legal errors requiring appellate correction rather than evidence of broader institutional problems. The appeals cannot, however, substitute for a clear account of what lessons the Commission accepted or rejected.

This section examines that gap from two perspectives. First, it asks whether the revision of the European Union’s antitrust procedural rules should strengthen internal review, recordkeeping, access to evidence, and other safeguards implicated by the Article 102 judgments. Second, it considers whether the Commission’s draft Article 102 guidelines reflect the courts’ effects-based jurisprudence or retreat toward conduct categories and presumptions. The central question is whether the Commission has absorbed the judgments’ lessons internally—or missed an opportunity for a second Airtours moment.

A. Reform Without a Second Airtours Moment?

The cases examined above show how procedural failures can affect both the reliability of Commission decisions and the depth of judicial scrutiny. The 2004 merger reforms stand out because the Commission did more than address the specific due-process defects identified by the General Court. It changed the institutional structure of merger review. Independent economic analysis, peer review, greater management oversight, and earlier access to evidence allowed the Commission to test case teams’ conclusions before adopting a decision. These safeguards could expose weaknesses in the evidentiary record and reduce the risk of decisions unsupported by facts or plausible real-world effects. Many of the most consequential changes required no new legislation because the Commission could adopt them internally.

The resulting checks and balances were not confined to merger review. Many applied across the Commission’s competition-enforcement system, including Article 102 investigations.[101] Yet Intel, Qualcomm, and Bulgarian Energy Holding raise a difficult question: Do those safeguards operate effectively in abuse-of-dominance cases? The Commission has said little about the procedural implications of these defeats, despite judicial criticism at times comparable to that in the 2002 merger judgments.

European Commission President Ursula von der Leyen directed Executive Vice President Teresa Ribera to “strengthen and speed up enforcement of competition rules” and align that enforcement with the European Union’s industrial and sustainability goals.[102] The Commission then opened a public consultation July 10, 2025, on revising Regulation 1/2003 and its implementing regulation, Regulation 773/2004—the principal procedural rules governing EU antitrust investigations.[103] That revision offers a natural opportunity to examine the Commission’s internal investigative and decision-making processes, including how it records interviews, maintains complete case files, discloses exculpatory material, tests case teams’ conclusions, and ensures meaningful access to evidence.

Stakeholders have already argued that Intel and Qualcomm warrant tighter internal procedures, particularly rules ensuring complete investigatory files and effective access to relevant evidence.[104] Stronger and faster enforcement remains a legitimate goal, but speed can strain the rights of defense when internal safeguards fail. Procedural protections do not merely constrain enforcement. They improve its accuracy, strengthen its legitimacy, and make Commission decisions more likely to survive judicial review. The revision of Regulation 1/2003 should confront that relationship directly.

B. Effects in Principle, Presumptions in Practice

Intel, Qualcomm, and Bulgarian Energy Holding show that European Union courts can engage closely with complex economic questions. The courts examined the Commission’s evidentiary record, tested the probative value of its evidence, and scrutinized its economic models. This review does not reflect hostility toward Article 102 enforcement. Google Shopping illustrates the other side of the principle: The General Court accepted the Commission’s extensive factual and economic record, and the Court of Justice upheld that judgment and the €2.42 billion fine.[105] The Commission may therefore have absorbed some practical lessons by developing more detailed foreclosure theories, counterfactual analyses, and economic evidence in its cases.

The Commission has not publicly explained how Intel, Qualcomm, and Bulgarian Energy Holding affected its approach. Its August 2024 draft guidelines on exclusionary abuses do not resolve the question, and the guidelines remained in draft form as of Aug. 31, 2026.[106] The Commission has said the guidelines will reflect its enforcement experience and codify the courts’ Article 102 jurisprudence since its 2008 guidance on enforcement priorities.

A March 2023 policy brief from the Directorate-General for Competition described that jurisprudence in strongly effects-based terms. It stated that “[t]he effects-based approach promoted by the Commission is clearly reflected in these developments and is now firmly enshrined in the Union Courts’ case law.” The brief added that recent judgments had “confirmed and endorsed the main elements of an effects-based approach to exclusionary conduct by dominant undertakings.”[107]

The draft guidelines have nevertheless drawn criticism for selectively presenting the case law and reviving a more form-based approach. An effects-based approach asks how conduct is likely to operate under actual market conditions. A form-based approach relies more heavily on the category into which the conduct falls and presumptions attached to that category. The draft guidelines identify some conduct as presumptively capable of producing exclusionary effects, potentially shifting the evidentiary burden away from the Commission’s affirmative demonstration of competitive harm.

Pinar Akman, Chiara Fumagalli, and Massimo Motta argue that the draft uses the case law selectively, particularly the judgments beginning with Intel.[108] Dirk Auer and Lazar Radic characterize its reliance on conduct categories and presumptions as a repudiation of the courts’ effects-based analysis.[109] Akman, Fumagalli, and Motta identify the central tension:

[The draft guidelines] do not embrace aspects of the case law that are effects-orientated and either overemphasise the operational value of certain concepts (e.g. “competition on the merits”) from the formalistic era of the case law or disregard statements from the case law that evidence an effects-based approach.[110]

The Article 102 jurisprudence is complex and does not always yield a single, coherent narrative.[111] The cases examined here are nonetheless clear on three points. The Commission must respect procedural rights, consider evidence that supports and undermines its theory, and establish through cogent evidence that the conduct was capable of producing nonhypothetical exclusionary effects. Guidelines intended to codify that jurisprudence should make those requirements more transparent, not dilute them through presumptions that substitute the form of conduct for its likely competitive effect.

V. Conclusions

Judicial victories and defeats both clarify the law. Because Article 102 annulments remain rare, Intel, Qualcomm, and Bulgarian Energy Holding carry particular weight. Together, they clarify the Commission’s duties to maintain and disclose a complete evidentiary record, protect the rights of defense, satisfy the applicable standard of proof, and ground findings of exclusionary effects in actual market conditions. They also confirm that European Union courts can scrutinize complex economic evidence without displacing the Commission’s authority to select and pursue legitimate theories of harm.

The Commission need not redesign its procedures after every annulment. Targeted judicial corrections form part of any mature enforcement system. These cases do, however, raise broader concerns. Qualcomm and Bulgarian Energy Holding exposed material failures to record interviews, disclose relevant evidence, and preserve equality of arms. Intel, Qualcomm, and Bulgarian Energy Holding also revealed substantive analyses that did not adequately connect the challenged conduct to realistic exclusionary effects. The recurrence of these problems warrants closer examination of whether safeguards introduced after the 2002 merger judgments work as intended in Article 102 investigations.

That review should focus on practical mechanisms. The Commission should ensure that case files contain complete records of third-party interviews and all relevant inculpatory and exculpatory evidence. It should give investigated companies timely access to that material and meaningful opportunities to answer it. A better-resourced Hearing Officer, independent economic review, stronger management oversight, and peer review by officials outside the case team could also test preliminary conclusions before they harden. These safeguards would help counter investigators’ tendency to favor their initial theory, conserve limited enforcement resources, and produce decisions more likely to withstand judicial review.

The Commission’s current reform projects provide two opportunities to act. The revision of Regulation 1/2003 can strengthen investigative procedures, access to the file, and internal review. The Article 102 guidelines can provide substantive clarity by faithfully reflecting the courts’ effects-based jurisprudence. Presumptions may simplify analysis in appropriate cases, but they should not replace evidence that the conduct was capable of excluding competition under the relevant market conditions. Nor should they deprive investigated companies of a genuine opportunity to rebut the Commission’s theory.

These protections will matter even more as competition enforcement becomes intertwined with industrial-policy goals such as resilience and strategic autonomy. Those priorities may intensify political pressure on individual investigations and sharpen public scrutiny of the Commission’s decisions. Transparent standards, rigorous economics, and credible internal checks help preserve the independence and legitimacy of enforcement under those conditions.

The Article 102 judgments may not demand a replica of the post-Airtours reform package. But they do call for the same institutional willingness to learn from defeat.

[1] See Case T-342/99, Airtours plc v. Comm’n, 2002 E.C.R. II-2585; Case T-310/01, Schneider Electric SA v. Comm’n, 2002 E.C.R. II-4071; Case T-5/02, Tetra Laval BV v. Comm’n, 2002 E.C.R. II-4381.

[2] See Case C-413/14 P, Intel Corp. v. Comm’n, ECLI:EU:C:2017:632 [hereinafter Intel 2017]; Case T-286/09 RENV, Intel Corp. v. Comm’n, ECLI:EU:T:2022:19 [hereinafter Intel Renvoi]; Case C-240/22 P, Comm’n v. Intel Corp., ECLI:EU:C:2024:915 [hereinafter Intel 2024]; Case T-235/18, Qualcomm, Inc. v. Comm’n, ECLI:EU:T:2022:358; Case T-136/19, Bulgarian Energy Holding EAD v. Comm’n, ECLI:EU:T:2023:669.

[3] See, e.g., Airtours/First Choice, Schneider/Legrand, Tetra Laval/Sidel, and General Electric/Honeywell, as well as other transactions the Commission prohibited between 1999 and 2001, including Volvo/Scania, WorldCom/Sprint, SCA/Metsä Tissue, and CVC/Lenzing. See Nicholas Levy, EU Merger Control: A Brief History, Cleary Gottlieb (Feb. 3, 2004), https://www.clearygottlieb.com/-/media/organize-archive/cgsh/files/publication-pdfs/eu-merger-control—a-brief-history.pdf. Parties also withdrew several high-profile notifications when faced with likely prohibition, including Skandinaviska Enskilda Bank/FöreningsSparbanken and EMI/Time Warner.

[4] WilmerHale, After Three Losses, Super Mario Strikes Back (Jan. 1, 2003), https://www.wilmerhale.com/en/insights/publications/after-three-losses-super-mario-strikes-back-winter-2003.

[5] Werner Berg, New EC Merger Regulation: A First Assessment of Its Practical Impact, 24 Nw. J. Int’l L. & Bus. 683 (2004), https://scholarlycommons.law.northwestern.edu/cgi/viewcontent.cgi?article=1586&context=njilb (citing Christian Ahlborn, Airtours/First Choice, In Competition (June 2002)); Nicholas Levy, Mario Monti’s Legacy in EC Merger Control, 1 Competition Pol’y Int’l 99 (2005), https://www.clearygottlieb.com/-/media/organize-archive/cgsh/files/publication-pdfs/mario-monti-s-legacy-in-ec-merger-control.pdf; Herwig Hofmann, Good Governance in European Merger Control: Due Process and Checks and Balances Under Review, 24 Eur. Competition L. Rev. 114 (2003); Mathew Heim, Problems and Process: European Merger Control and How to Use It, 4 J. Pub. Affs. 73 (2004); Damien J. Neven & Lars-Hendrik Röller, Discrepancies Between Markets and Regulators: An Analysis of the First Ten Years of EU Merger Control, in The Pros and Cons of Merger Control 13 (Swedish Competition Authority ed., 2002), https://www.konkurrensverket.se/globalassets/dokument/informationsmaterial/rapporter-och-broschyrer/pros-and-cons/rapport_pros-and-cons_2002_the-pros-and-cons-of-merger-control.pdf.

[6] Schneider Electric, 2002 E.C.R. II-4071, ¶ 453.

[7] Id. ¶¶ 458, 460.

[8] Id. ¶¶ 454, 462.

[9] The Commission appealed the General Court’s Tetra Laval judgment, arguing that the court had interfered excessively and imposed an unduly high standard of proof for conglomerate effects. The CJEU dismissed the appeal, holding that the General Court acted within the proper scope of judicial review when it examined whether the Commission’s evidence was accurate, reliable, consistent, and sufficient to support its conclusions. See Case C-12/03 P, Comm’n v. Tetra Laval BV, ECLI:EU:C:2005:87, 2005 E.C.R. I-987.

[10] Tetra Laval, ECLI:EU:T:2002:264, ¶¶ 89, 92, 117.

[11] Airtours, 2002 E.C.R. II-2585, ¶ 63.

[12] Id. ¶ 210.

[13] Id. ¶ 294.

[14] Schneider Electric, 2002 E.C.R. II-4071, ¶ 409.

[15] Id. ¶¶ 349, 409.

[16] Id. ¶¶ 336, 398.

[17] Id. ¶ 411.

[18] See Mario Monti, European Comm’r for Competition, EU Competition Policy, Address at the Fordham Annual Conference on International Antitrust Law & Policy (Oct. 31, 2002), https://ec.europa.eu/commission/presscorner/api/files/document/print/en/speech_02_533/SPEECH_02_533_EN.pdf.

[19] Tetra Laval, 2002 E.C.R. II-4381, ¶ 137.

[20] Id. ¶ 162.

[21] Id. ¶ 246.

[22] Schneider Electric, 2002 E.C.R. II-4071, ¶ 437.

[23] Airtours, 2002 E.C.R. II-2585, ¶ 63.

[24] Id.

[25] Pablo Ibáñez Colomo, Law, Policy, Expertise: Hallmarks of Effective Judicial Review in EU Competition Law, 24 Cambridge Y.B. Eur. Legal Stud. 143 (2022), https://doi.org/10.1017/cel.2022.7.

[26] Mario Monti, European Comm’r for Competition, Merger Control in the European Union: A Radical Reform, Speech/02/545 (Nov. 7, 2002), https://ec.europa.eu/commission/presscorner/api/files/document/print/en/speech_02_545/SPEECH_02_545_EN.pdf.

[27] Tetra Laval, 2002 E.C.R. II-4381, ¶ 137.

[28] Id. ¶ 162.

[29] Id.

[30] Levy, supra note 3.

[31] European Commission, Green Paper on the Review of Council Regulation (EEC) No. 4064/89, COM (2001) 745 final (Dec. 11, 2001).

[32] Levy, supra note 3.

[33] Monti, supra note 26.

[34] Neither the General Court nor Commissioner Mario Monti used the term “checks and balances.” They instead discussed evidence, burdens of proof, Commission errors, judicial-review standards, and modernization. Commentators later used “checks and balances” to describe the actors who safeguard due process, procedural fairness, and objectivity in Commission decision-making. The Commission eventually adopted the term. In Merck/Sigma-Aldrich, the hearing officer identified the commissioner for competition and cabinet, senior Directorate-General for Competition officials, the relevant horizontal-coordination unit, the Chief Economist’s Team, the Legal Service, associated Commission services, the hearing officer, the Advisory Committee on Concentrations, and possible peer review. See Final Report of the Hearing Officer ¶ 32 & n.39, Case M.8181, Merck/Sigma-Aldrich (Apr. 30, 2021), https://ec.europa.eu/competition/mergers/cases1/202212/M_8181_8229078_2374_3.pdf; see also Directorate-General for Competition, European Commission, Proceedings for the Application of Articles 101 and 102 TFEU: Key Actors and Checks and Balances, https://competition-policy.ec.europa.eu/system/files/2021-04/key_actors_checks_balances_art_101_102_factsheet.pdf.

[35] Monti, supra note 26.

[36] Id.

[37] Pieter Van Cleynenbreugel, The Hearing Officer’s Extended Mandate: Whose Special Friend in the Conduct of EU Competition Proceedings?, 33 Eur. Competition L. Rev. 286 (2012).

[38] See Commission Decision 2001/462/EC, ECSC of 23 May 2001 on the Terms of Reference of Hearing Officers in Certain Competition Proceedings, 2001 O.J. (L 162) 21, superseded by Decision of the President of the European Commission 2011/695/EU of 13 October 2011 on the Function and Terms of Reference of the Hearing Officer in Certain Competition Proceedings, 2011 O.J. (L 275) 29.

[39] Pablo Ibáñez Colomo, Judicial Review in EU Merger Control: Towards Deference on Issues of Law?, 4 Eur. L. Open 441 (2025), https://doi.org/10.1017/elo.2025.10016.

[40] Monti, supra note 26.

[41] Commissioner Monti acknowledged criticism that the Commission was too susceptible to influence from merging parties’ competitors. He argued that merging parties should have ample opportunity to defend their positions and confront third-party concerns early in the investigation. Id. at 7.

[42] Directorate-General for Competition, European Commission, Best Practices on the Conduct of EC Merger Control Proceedings (Jan. 20, 2004), https://competition-policy.ec.europa.eu/document/download/a288a6d8-3962-4072-b6a1-93e2ba08d12e_en?filename=proceedings.pdf [hereinafter Merger Best Practices].

[43] Id. ¶ 45.

[44] Id. ¶ 46.

[45] Monti, supra note 26.

[46] Guidelines on the Assessment of Horizontal Mergers Under the Council Regulation on the Control of Concentrations Between Undertakings, 2004 O.J. (C 31) 5.

[47] Monti, the first economist appointed competition commissioner, emphasized economic analysis and the recruitment of industrial-organization economists to DG Competition. Director-General Philip Lowe, also an economist, supported that effort. Monti described it as a central goal: “I should like to underline that an increased economic approach in the interpretation of our rules was, indeed, one of my main objectives when I took on my new responsibilities as Competition Commissioner.” Monti, supra note 18, at 7.

[48] See Berg, supra note 5.

[49] Guillaume Loriot, Reflecting on 20 Years of the EU Merger Regulation, Keynote Address at Modern EU and UK Merger Control: 20th Anniversary—Two Decades of Regulation 139/04 and the Enterprise Act Regime (Feb. 27, 2024), https://competition-policy.ec.europa.eu/document/download/ca447d8a-b2b7-4583-b983-753b3385ac86_en?filename=20240227_Kings_College_Latham_Merger_Conference_London_Guillaume-Loriot_keynote_speech.pdf.

[50] See European Commission, Review of the Merger Guidelines, https://competition-policy.ec.europa.eu/mergers/review-merger-guidelines_en (last visited Aug. 30, 2026).

[51] Teresa Ribera, Executive Vice President of the European Commission, Keynote Speech at the 23rd International Conference on Competition (IKK) (Mar. 12, 2026), https://ec.europa.eu/commission/presscorner/detail/en/speech_26_602.

^[52] Thibault Schrepel and Janneke Parrish examined the Commission’s record in competition cases before the Court of Justice, excluding the General Court, from Nov. 1, 2014, through Nov. 30, 2024—Margrethe Vestager’s tenure as competition commissioner. They identify Intel as the Commission’s only Article 102 loss during that period and report no Article 102 defeats involving novel theories of harm or procedural issues. See Thibault Schrepel & Janneke Parrish, Mapping Reversals: An Empirical Account of Margrethe Vestager’s Track Record Before the Court of Justice of the European Union, 22 Eur. Competition J. 63 (2026), https://doi.org/10.1080/17441056.2025.2511425.

[53] Pinar Akman, A Critical Inquiry into ‘Abuse’ in EU Competition Law, 44 Oxford J. Legal Stud. 405 (2024), https://doi.org/10.1093/ojls/gqae008.

[54] For purposes of this paper, the principal Intel decisions are Intel 2017, which set aside the General Court’s 2014 judgment; Intel Renvoi, which annulled the Commission’s findings concerning Intel’s rebates and the resulting fine; and Intel 2024, which upheld Intel Renvoi. See supra note 2.

[55] This section does not exhaust the European courts’ jurisprudence. It focuses on decisions whose holdings and broader statements echo Airtours, Schneider Electric, and Tetra Laval. Other Article 102 judgments reinforce the reasoning in Intel, Qualcomm, and Bulgarian Energy Holding, including Case T-612/17, Google LLC v. Commission, ECLI:EU:T:2021:763; Case C-377/20, Servizio Elettrico Nazionale SpA v. Autorità Garante della Concorrenza e del Mercato, ECLI:EU:C:2022:379; and Case C-680/20, Unilever Italia Mkt. Operations Srl v. Autorità Garante della Concorrenza e del Mercato, ECLI:EU:C:2023:33. In Google (Shopping), the General Court largely upheld the Commission’s decision but articulated demanding standards for its treatment of evidence. Servizio Elettrico Nazionale and Unilever Italia, both preliminary-reference cases, extended Intel’s evidentiary logic beyond exclusivity rebates. They required competition authorities to assess exclusionary capability in light of all relevant circumstances and treated evidence of actual market performance as relevant, though not dispositive. The Court of Justice also held that when a dominant undertaking presents evidence that its conduct could not produce anticompetitive effects, competition authorities must examine that evidence carefully and impartially. Although these cases did not review the Commission’s procedural framework, the Court grounded that obligation in the right to be heard. See Servizio Elettrico Nazionale, ECLI:EU:C:2022:379, ¶¶ 50–54; Unilever Italia, ECLI:EU:C:2023:33, ¶¶ 47–62.

[56] Intel Renvoi, ECLI:EU:T:2022:19, ¶ 102.

[57] Intel 2017, ECLI:EU:C:2017:632, ¶¶ 94–102.

[58] Intel Renvoi, ECLI:EU:T:2022:19, ¶ 93.

[59] See Ibáñez Colomo, supra note 25; see also Servizio Elettrico Nazionale, ECLI:EU:C:2022:379, ¶ 51 (holding that, when a dominant undertaking submits supporting evidence, the competition authority must determine whether the challenged conduct was capable of restricting competition under the particular circumstances).

[60] Commission Decision C(2018) 240 final of 24 January 2018 Relating to a Proceeding Under Article 102 TFEU and Article 54 of the EEA Agreement, Case AT.40220, Qualcomm (Exclusivity Payments).

[61] Qualcomm, ECLI:EU:T:2022:358, ¶ 344.

[62] Alfonso Lamadrid, Case T-235/18, Qualcomm v. European Commission (Part I: Procedure), Chillin’Competition (June 27, 2022), https://chillingcompetition.com/2022/06/27/case-t-235-18-qualcomm-v-european-commission-part-i-procedure.

[63] The Commission also failed to maintain proper records of these meetings or include them in the case file. Qualcomm, ECLI:EU:T:2022:358, ¶¶ 207–09, 256–59, 291–94. See also Council Regulation 1/2003, art. 19, 2003 O.J. (L 1) 1, which governs Commission interviews conducted to collect information related to an investigation.

[64] Qualcomm, ECLI:EU:T:2022:358, ¶ 279. The General Court also found serious due-process defects that infringed Qualcomm’s rights of defense. Material differences between the Commission’s statement of objections and final decision rendered irrelevant the data and economic arguments on which Qualcomm had relied. The court explained that this defense “is of no practical effect if the scope of the conduct concerned is modified by the Commission after the [statement of objections].” Id. ¶¶ 328, 333–352.

[65] Lamadrid, supra note 62.

[66] Commission Decision C(2018) 8806 final of 17 December 2018 Relating to a Proceeding Under Article 102 TFEU and Article 54 of the EEA Agreement, Case AT.39849, BEH Gas.

[67] See European Commission, Commission Staff Working Document Accompanying the Report from the Commission on Competition Policy 2023, SWD (2024) 53 final (Mar. 6, 2024), https://eur-lex.europa.eu/legal-content/EN/TXT/HTML/?uri=CELEX:52024SC0053. The General Court upheld one isolated instance of exclusionary conduct during part of the infringement period. It nonetheless annulled the Commission’s decision in full because that conduct alone could not establish the alleged single and continuous infringement, given the factual and legal defects in the remaining elements.

[68] The Commission appealed the General Court’s judgment on Jan. 10, 2024. As of Aug. 31, 2026, the appeal remained pending. Case C-14/24 P, Comm’n v. Bulgarian Energy Holding EAD. Advocate General Medina issued her opinion on April 23, 2026. Opinion of Advocate General Medina, Case C-14/24 P, Comm’n v. Bulgarian Energy Holding EAD, ECLI:EU:C:2026:347.

[69] Bulgarian Energy Holding, ECLI:EU:T:2023:669, ¶¶ 1170–1172; Qualcomm, ECLI:EU:T:2022:358, ¶ 344.

[70] Bulgarian Energy Holding, ECLI:EU:T:2023:669, ¶ 1154.

[71] Id. ¶¶ 1152, 1155. See also Council Regulation (EC) No. 1/2003, art. 27(2), 2003 O.J. (L 1) 1, 13; Commission Regulation (EC) No. 773/2004 of 7 April 2004 Relating to the Conduct of Proceedings by the Commission Pursuant to Articles 81 and 82 of the EC Treaty, art. 15(1)–(2), 2004 O.J. (L 123) 18, 22.

[72] Intel Renvoi, ECLI:EU:T:2022:19, ¶¶ 50–51.

[73] Assimakis Komninos & James Killick, The EU General Court Quashes Dominant Chip Manufacturer’s Antitrust Fine for “Exclusivity Payments” and Censures the European Commission for Multiple Due-Process and Substantive Errors (Qualcomm), e-Competitions, June 2022, Art. No. 107095 (June 15, 2022).

[74] Monti, supra note 26.

[75] Intel Renvoi, ECLI:EU:T:2022:19, ¶¶ 168–335.

[76] Id. ¶¶ 163, 255, 522.

[77] Jacques Buhart et al., The EU General Court Annuls the Commission’s Decision Regarding a Dominant Chip Manufacturer as a Result of Insufficient Analysis of Anticompetitive Effects and on Procedural Grounds (Qualcomm), e-Competitions, June 2022, Art. No. 112054 (June 15, 2022).

[78] Bulgarian Energy Holding, ECLI:EU:T:2023:669, ¶ 279.

[79] Id. ¶ 281.

[80] Intel 2017, ECLI:EU:C:2017:632.

[81] The five factors are the practice’s market coverage; the conditions governing the rebates; their duration; their amount; and evidence of a strategy to exclude competitors as efficient as the dominant undertaking. See id. ¶ 139.

[82] Intel Renvoi, ECLI:EU:T:2022:19, ¶ 499.

[83] Intel 2024, ECLI:EU:C:2024:915.

[84] Org. for Econ. Coop. & Dev. [OECD], The Standard and the Burden of Proof in Competition Law Cases, OECD Roundtables on Competition Policy Papers No. 318 (2024), https://doi.org/10.1787/0199f63f-en.

[85] Qualcomm, ECLI:EU:T:2022:358, ¶ 346.

[86] Id. ¶¶ 409–24. Lamadrid argues that the judgment applied counterfactual logic without using the term. Because Apple could not have switched to rivals absent Qualcomm’s payments, those payments could not have restricted competition. Alfonso Lamadrid, Case T-235/18, Qualcomm v. European Commission (Part II: Substance), Chillin’Competition (July 11, 2022), https://chillingcompetition.com/2022/07/11/case-t-235-18-qualcomm-v-european-commission-part-ii-substance.

[87] Qualcomm, ECLI:EU:T:2022:358, ¶¶ 355, 396–97; see also Ibáñez Colomo, supra note 25.

[88] Qualcomm, ECLI:EU:T:2022:358, ¶ 351.

[89] Id. ¶ 424.

[90] See Bulgarian Energy Holding, ECLI:EU:T:2023:669, ¶¶ 945–49; see also Kristi Georgieva & Niamh Martyn, Bulgarian Energy Holding v. Commission (Case T-136/19): Revamped Energy to Overturn Abuse of Dominance Decisions on Account of Errors of Fact and Assessment, Cleary Antitrust Watch (Oct. 25, 2023), https://www.clearyantitrustwatch.com/2023/10/bulgarian-energy-holding-v-commission-case-t-136-19-revamped-energy-to-overturn-abuse-of-dominance-decisions-on-account-of-errors-of-fact-and-assessment.

[91] Bulgarian Energy Holding, ECLI:EU:T:2023:669, ¶ 279.

[92] Id. ¶ 282.

[93] Id. ¶ 281.

[94] Id. ¶¶ 280–81.

[95] Bas Braeken et al., The EU General Court Annuls a €77M Fine Imposed on a Bulgarian Gas Company Due to a Lack of Evidence and Procedural Flaws (Bulgarian Energy Holding), e-Competitions, Oct. 2023, Art. No. 121154 (Oct. 25, 2023).

[96] Ibáñez Colomo, supra note 39.

[97] Akman, supra note 53.

[98] Monti, supra note 26.

[99] Berg, supra note 5.

[100] After the Qualcomm judgment, the Commission said it would study the ruling and consider its next steps. Foo Yun Chee, Qualcomm Wins Fight Against $1 Billion EU Antitrust Fine, Reuters (June 15, 2022), https://www.reuters.com/technology/eu-court-rejects-1-bln-eu-antitrust-fine-against-qualcomm-2022-06-15. The Commission gave a similar response after the BEH judgment. Directorate-General for Competition, Weekly e-News, No. 37/2023 (Oct. 27, 2023).

[101] See, e.g., Monti, supra note 26; see also Ibáñez Colomo, supra note 39.

[102] Letter from Ursula von der Leyen, President, European Commission, to Teresa Ribera Rodríguez, Executive Vice-President-Designate for a Clean, Just and Competitive Transition (Sept. 17, 2024), https://commission.europa.eu/document/download/5b1aaee5-681f-470b-9fd5-aee14e106196_en?filename=Mission%20letter%20-%20RIBERA.pdf.

[103] Council Regulation 1/2003 of Dec. 16, 2002, on the Implementation of the Rules on Competition Laid Down in Articles 81 and 82 of the Treaty, 2003 O.J. (L 1) 1; Commission Regulation 773/2004 of Apr. 7, 2004, Relating to the Conduct of Proceedings by the Commission Pursuant to Articles 81 and 82 of the EC Treaty, 2004 O.J. (L 123) 18 [hereinafter Regulation 773/2004]; see also European Commission, Commission Staff Working Document: Evaluation of Regulations 1/2003 and 773/2004, SWD(2024) 216 final (Sept. 5, 2024), https://ec.europa.eu/info/law/better-regulation/have-your-say/initiatives/13431-EU-antitrust-procedural-rules-evaluation_en.

[104] The European Competition Lawyers Forum argued that Intel and Qualcomm raised evidentiary-quality concerns that the review of Regulation 1/2003 should address. European Competition Lawyers Forum, Contribution to the Public Consultation: Reforms to Regulation 1/2003 (Oct. 24, 2022). Participants in the Commission’s evaluation likewise questioned whether the Commission maintained complete case files containing all relevant evidence, including exculpatory material. European Commission, supra note 103, at 144.

[105] The CJEU dismissed Google and Alphabet’s appeal, leaving in place the General Court’s judgment and the €2.42 billion fine. Case C-48/22 P, Google LLC & Alphabet Inc. v. Comm’n (Google Shopping), ECLI:EU:C:2024:726.

[106] European Commission, Press Release IP/24/3623, Commission Seeks Feedback on Draft Antitrust Guidelines on Exclusionary Abuses (Aug. 1, 2024), https://ec.europa.eu/commission/presscorner/detail/en/ip_24_3623.

[107] Linsey McCallum et al., A Dynamic and Workable Effects-Based Approach to Abuse of Dominance, Competition Policy Brief No. 1/2023 (Eur. Comm’n, Directorate-Gen. for Competition Mar. 2023), https://doi.org/10.2763/731952.

[108] Pinar Akman, Chiara Fumagalli & Massimo Motta, The European Commission’s Draft Guidelines on Exclusionary Abuses: A Law and Economics Critique and Recommendations, 16 J. Eur. Competition L. & Prac. 234 (2025), https://doi.org/10.1093/jeclap/lpaf020.

[109] Dirk Auer & Lazar Radic, The Commission’s Article 102 TFEU Guidelines: Consolidation or Creation?, SSRN Scholarly Paper No. 5134897 (2025), https://doi.org/10.2139/ssrn.5134897.

[110] Akman et al., supra note 108.

[111] See, e.g., Fernando Castillo de la Torre, The Dwindling Law in Article 102 TFEU, in Article 102 TFEU: Past, Present and Future ch. 6 (Assimakis Komninos & Ekaterina Rousseva eds., 2025).

The road to rural access to AI-enabled services runs through the permit office. Chips, data centers, and electricity get the attention, but delivering AI to a farmer’s field, a rural clinic, or a factory floor takes a broadband network. Building it means navigating utility poles, road crossings, and public rights-of-way—each with a gatekeeper who can keep the future waiting.

Private investment will connect most of America. In many rural areas, however, getting that investment to pencil out will require policies that improve the incentives to build. The potential benefits are substantial—AI-assisted diagnostics at rural hospitals, precision agriculture that streams field imagery in real time, and reliable home broadband that lets higher earners work remotely and settle outside major metropolitan areas.

Much of that value spreads through communities in ways no provider can capture on a monthly bill. Without policies that lower construction costs, the places least likely to attract private capital will also be the last to get AI-ready infrastructure. That would widen the gap between connected and unconnected regions just as connectivity becomes central to economic competitiveness.

That makes the next phase of the Broadband Equity, Access, and Deployment (BEAD) Program more consequential than it might appear. After restructuring the program, the National Telecommunications and Information Administration (NTIA) has roughly $21 billion in unspent funds and has told states to direct that money toward more construction. But as in the first funding round, money alone won’t clear the path. Successful builds depend on timely, predictable access to public rights-of-way and the infrastructure along them.

Read the full piece here.

TL;DR

Background: Personalized pricing uses data about a customer to estimate what that person will pay. Maryland and New Jersey have enacted restrictions on the practice in grocery sales, while Connecticut has adopted a broader retail restriction. The Federal Trade Commission (FTC) has also proposed guidance warning that undisclosed personalized pricing is likely unfair or deceptive.

But… The effects on consumers depend on the market and how sellers set prices. In the three major empirical studies the FTC cites in its proposed guidance, most consumers paid less, received a lower offer, or gained from personalization, even when consumers lost as a group. Nor is the practice as widespread as the debate suggests: A 2018 European Commission study of 160 online retailers found personalized price differences in just 6% of tests, with a median pricing difference below 1.6%.

Moreover… Disclosure has costs. Requiring sellers to explain each tailored price could discourage discounts for price-sensitive customers, while evidence that such explanations change shopping decisions is weak. A price tag tells a shopper what they can pay. It says nothing about what anyone else pays.

KEY TAKEAWAYS

The Price Is…Complicated

Price discrimination means charging different buyers different prices for the same product for reasons other than cost. Student discounts, matinee tickets, airline fares, and grocery coupons are familiar examples. Personalized pricing uses data about an individual to estimate what that person will pay.

Its effects depend on the market. Richard Schmalensee and Lars Stole explain that a monopolist’s price discrimination can raise total welfare only if it increases sales. Models by Jacques-François Thisse and Xavier Vives show that, among competing firms, it can drive down both prices and profits. As Daniel J. Gilman puts it, “The welfare effect of price discrimination is known to be ambiguous and thus a matter of empirical measurement.”

The alternative matters, too. John M. Yun argues that, in the absence of personalized pricing, sellers would usually use another form of price discrimination, rather than charge everyone the same price.

Three Studies, No Easy Verdict

Three studies cited in the FTC’s personalized-pricing guidance find mixed effects. It should be noted, however, that none tests personalized posted prices for everyday shoppers in a U.S. retail store.

In a ZipRecruiter experiment, Jean-Pierre Dubé and Sanjog Misra estimate that personalization could have raised profits by 86% over the firm’s actual pricing. Optimizing a single price would have raised profits by 55%; personalization added 19% over that optimized price. About three in five potential customers would receive a lower offer. But the study examines a hiring platform with a single seller, not competing retailers. 

Benjamin Reed Shiller estimates that prices based on browsing histories could have raised Netflix’s 2006 profits by 12.99%. Yet “most consumers receive lower prices when prices are personalized, and hence the majority are slightly better off.” His results are based on a simulation of personalized pricing because at the time of his study Netflix did not engage in personalized pricing. 

Nicholas Buchholz and his co-authors find that 62.5% of riders on Liftago, a Prague taxi platform, gain from personalization. Riders least sensitive to price and wait times bear most of the losses. Total consumer surplus falls by a further 2.25% compared with platform-set uniform pricing.

These settings reveal real tradeoffs, but they do not establish what personalized pricing would do in a competitive U.S. retail market or support a presumption that personalization harms consumers.

Competition Spoils the Plot

Competition can change who gains from personalized pricing. Andrew Rhodes and Jidong Zhou find that when all consumers already buy a product and all firms personalize, consumers gain as a group while industry profits fall. If uniform pricing leaves some buyers out, personalization can bring them in even as total consumer surplus falls. Brian C. Albrecht finds that when competing firms have the same complete information, competition can push prices for contested buyers down to the cost of serving them.

More data does not automatically mean more harm. Dubé and Misra find that “allowing the firm to target on more granular data can be good for consumer surplus.” Ginger Zhe Jin, Liad Wagman, and Mengyi Zhong call the tradeoffs “complex.” Their review identifies the greatest risk as when one firm has much more data than its rivals.

Many sellers leave money on the table rather than vary prices. Alberto Cavallo found no evidence that large U.S. retailers varied online prices by a shopper’s location or browsing history. Stefano DellaVigna and Matthew Gentzkow found that most large U.S. food, drug, and general-merchandise chains charged nearly uniform prices across stores. The median chain gave up about $16 million a year in profit by doing so.

The Price Tag Said What?

A price tag tells a shopper what they can pay, not what everyone else pays. Yet the FTC’s deception theory assumes shoppers expect the same price for all and treats silence about personalization as misleading. The International Center for Law & Economics (ICLE) argues that this goes too far. In International Harvester (1984), the FTC distinguished a misleading “half-truth” from a “pure omission,” a subject on which the seller “has simply said nothing.” Pure omissions generally are not deceptive.

Unfairness requires proof of substantial injury that consumers cannot reasonably avoid and that benefits to consumers or competition do not outweigh. ICLE says the FTC’s hypothetical examples never apply that test. In FTC v. Corpay (2026), the 11th U.S. Circuit Court of Appeals upheld an unfairness finding over undisclosed fleet-card fees based on survey evidence, internal documents, and customer complaints.

More Fine Print Means Fewer Deals

A typical supermarket stocks more than 31,000 products, Gilman notes. Explaining how it sets every tailored price would be impractical and could lead sellers to drop targeted discounts. The explanations might do little for shoppers: An OECD study found no significant effect of pricing disclosures on purchasing decisions in experiments in Dublin, Ireland and Santiago, Chile.

Petra Persson shows how firms can bury useful information in a flood of disclosures, a risk Frank Easterbrook and Daniel Fischel identified decades ago. Disclosure rules can also change what sellers offer. Allison Nicoletti and Christina Zhu found lower approval rates for loans covered by federal mortgage-disclosure rules. ICLE has made a similar point about uniform-pricing mandates for grocery suppliers: They discourage discounts.

For more on this issue, see “ICLE Comments on the FTC’s Personalized Pricing Proposed Enforcement Policy Statement.”

Suppose a retailer charges $10 for a service. One customer values it at $11 and buys it. Another values it at $8 and walks away. The retailer sends the second customer a $3 coupon. She pays $7, gets something worth $8 to her, and the retailer makes a sale it otherwise would have missed. The customers paid different prices, but without the coupon, the likely outcome was one $10 sale—not two $7 sales.

That missing sale matters to the Federal Trade Commission’s (FTC) proposed enforcement policy statement on personalized pricing. The FTC acknowledges that Congress has not authorized a blanket ban on personalized pricing and that the practice’s effects remain unclear. Yet when consumers reasonably expect prices to stay the same regardless of their personal data, the proposal says sellers should disclose that a price is personalized, why, and what types of data they used. The FTC says failing to do so is likely unfair or deceptive.

The FTC has reason to worry about sellers who invent discounts, break promises about data use, or tack on charges buyers never agreed to. But its proposal goes further, casting suspicion on price differences that sellers do not explain. That risks overlooking both the sales that tailored offers make possible and the trial and error through which sellers discover what customers will buy. It also assumes that the agency can decide which pricing practices need an explanation despite its own limited knowledge and political incentives.

Truth on the Market readers have already seen important objections to the proposal. Alden Abbott explains why a consumer’s expectation alone cannot establish deception. Daniel J. Gilman asks whether the economic research cited by the FTC supports its claims. This post asks what happens to the market processes that generate prices once the government starts supervising them—and how officials might use that authority.

Austrian economics, with its focus on how buyers and sellers discover opportunities through exchange, and Public Choice theory, which examines the incentives facing government officials, point toward a narrower policy. The FTC should pursue proven lies and consumer injuries under existing law. But it must also leave room for sellers to test prices, and not turn ordinary price search into a presumptive violation merely because software makes the search more precise.

Read the full piece here.

The next serious competitor may be waiting for a license, or working on a product nobody has seen yet.  Neither shows up in today’s market-share figures. And much of the current antitrust debate ignores both of those figures completely, even as it ponders how aggressively the government should reshape markets.

By contrast, a dynamic approach to antitrust can enforce the law vigorously while also recognizing that competition changes over time and that government officials rarely know enough to redesign complex markets. The practical questions are which restraints deserve priority, what evidence should justify intervention, and how the law can leave room for the experiments that produce tomorrow’s competitors.

I propose two commitments.

First, enforcement agencies and policymakers should pay closer attention to government-created barriers to entry. Licensing restrictions, certificate-of-need laws that require approval before a provider can open or expand, discriminatory subsidies, procurement preferences, and legal immunities can all keep would-be competitors out.

Second, enforcers should judge private conduct by its likely effects on innovation and the competitive process. That means considering firms’ capabilities and the costs of mistaken enforcement decisions, rather than treating today’s market concentration figures as the last word. Cartels, anticompetitive mergers, and exclusionary monopolization still warrant action. But size, integration, and commercial success alone do not establish harm.

Read the full piece here.

Some in Congress want to give teenagers their childhoods back by taking away the last place many of them can gather without adults. That cure may prove worse than the disease.

Read the full piece here.

Power-plant rules do more than regulate pollution. They can also regulate competitors out of the market. That is the overlooked significance of the Environmental Protection Agency’s (EPA) recent repeal of most Biden administration greenhouse-gas standards for power plants. The move is an environmental-policy reversal and also a competition-policy reform.

The final rule is only a partial repeal. It eliminates standards for existing fossil-fuel-fired steam units, along with several carbon-capture requirements for coal plants and new baseload combustion turbines. A separate supplemental proposal contemplates a broader rollback. Published Sept. 17 in the Federal Register, the final rule takes effect Nov. 16.

The central economic point is that a regulation that forwards a legitimate public-policy goal may nonetheless also skew competition. This happens when the rule imposes costs on some technologies that their rivals do not bear, steers investment toward politically favored alternatives, or raises barriers for firms too small to absorb the compliance burden.

The power-plant rule did all three. Its repeal could therefore improve economic welfare by putting generating technologies on more equal footing, reducing wasted resources, and giving cost and reliability greater weight in deciding how to power the grid.

Read the full piece here.

Frontier artificial-intelligence (AI) labs may have good reasons to tap the brakes. The antitrust question begins when they all reach for the same pedal.

Anthropic CEO Dario Amodei’s proposal to “pace the frontier” has become a rallying point for AI companies interested in coordinating a slowdown in capability development. His plan calls for embedded outside evaluators, common safety standards, and limits on the pace of model advances. Amodei also suggests that some form of antitrust waiver may be needed to permit certain safety discussions.

As my colleague Dirk Auer explains, agreements among rivals to restrain development deserve scrutiny, even when made in the name of safety. Before granting an exemption, policymakers should ask where existing legal and market incentives—and the forms of cooperation antitrust law already allows—fall short.

Civil and criminal laws already give AI labs reason to consider the consequences of their conduct, including safety risks. Market incentives point in the same direction. Customers favor products that work as intended and remain dependable. A model that enables mass hacking or causes other serious harm is, among other things, a very bad product.

Companies can respond to these incentives on their own by strengthening safeguards, limiting model autonomy, or delaying releases. Existing antitrust law also permits competitors to collaborate on many safety matters and participate in carefully structured standard-setting efforts.

Any proposed exemption therefore carries a heavy burden. Its proponents must show why unilateral action and lawful collaboration cannot address the problem—and identify precisely what conduct existing antitrust law prevents.

Read the full piece here.

Ask Brussels what ChatGPT is, and you get two answers. Under the Artificial Intelligence Act (AI Act), the European Union regulates it as artificial intelligence. On Aug. 31, the European Commission supplied another answer when it designated ChatGPT as a Very Large Online Search Engine under the Digital Services Act (DSA). Each classification comes with its own rulebook. Together, they may shape what ChatGPT can say.

The DSA designation puts ChatGPT alongside Google Search and Bing in the category reserved for the largest search engines operating in the EU. Under Article 33, a platform or search engine enters that category when it reaches at least 45 million average monthly active users in the EU. OpenAI reported that ChatGPT Search averaged about 159.1 million monthly active users in the EU during the six months ending March 31, comfortably clearing the threshold. ChatGPT now has four months to comply with the additional obligations the DSA imposes on services of that size.

Those obligations require the largest services to examine their broader effects on society. Article 34 requires them to identify and assess “systemic risks” arising from their design or operation. These include risks to fundamental rights, civic discourse, public security, elections, public health, and children. Article 35 then requires services to adopt reasonable, proportionate, and effective measures to mitigate those risks.

The DSA offers several examples, including changes to a service’s design, recommendation systems, terms and conditions, and content-moderation practices. Lawmakers largely developed these obligations for platforms and conventional search engines. A search engine generally organizes and presents information found elsewhere. ChatGPT, by contrast, generates the answer a user sees.

That difference has practical consequences. Efforts to reduce systemic risks may extend beyond how a service ranks, recommends, or displays information. They may shape the substance of the answer itself.

OpenAI has accepted the designation and emphasized its commitment to complying with the DSA while balancing access to useful information with protection from harmful and illegal content. Few would quarrel with that goal. The difficulty lies in deciding what the balance requires.

ChatGPT’s designation therefore presents a broader problem. A regulatory framework designed to govern the distribution of online information may fit awkwardly when applied to a service that produces the answers users receive.

Read the full piece here.

In five years, computational antitrust has gone from a proposition to a field. More than eighty agencies now use computational tools, a literature has formed around them, a scattered set of researchers and enforcers has come to recognize itself as one. The next five years will decide what kind of field it becomes.

Read the full piece here.

The European Union’s Digital Markets Act (DMA) depends on two terms that lawmakers never defined. “Fairness” and “contestability” determine how the European Commission measures gatekeeper compliance, imposes new obligations, and grades its own performance every three years. The law uses both terms liberally but explains neither.

Recital 79 even promises that the Commission will apply a “predefined standard” to identify unfair practices and practices that limit contestability. Anyone searching the DMA for that standard will come away empty-handed.

In a new white paper, I reconstruct what I argue is nonetheless a fully coherent framework implicit in the enacted text. The DMA’s provisions and its subsequent enforcement record, along with various guidance documents issued along the way, do reveal specific conceptions of fairness and contestability, even if lawmakers declined to state them openly. Put simply: “fairness” refers to a presumed structural imbalance in digital markets, and “contestability” is the redistribution program enacted to correct it

That reconstruction has practical consequences. It explains how the Commission is likely to assess the DMA’s success—and its own performance. It also identifies the policy choices that countries outside the European Union may adopt, often without expressly debating them, when they enact similar laws.

Read the full piece here.

The trouble with chatbots is right there in the name: They talk. When that talk contributes to harm, policymakers face a deceptively difficult question. Should they seek rules that amount to policing dangerous conduct, or deciding what the chatbot should have said instead?

The question matters because chatbots have quickly become one of the most visible uses of artificial intelligence, and a growing focus of regulatory attention. Millions of people now use them to search for information, seek advice, work through problems, make decisions, and discuss personal matters. As those uses have expanded, scrutiny has shifted toward what chatbots say—and what happens when those conversations go wrong.

Some of that scrutiny stems from cases in which chatbot interactions allegedly contributed to real harm, including harm to children. These cases raise legitimate questions about safety and responsibility. They also form part of a broader debate about conversational AI, including inaccurate or dangerous advice, systems that reinforce users’ existing beliefs, and responses that sound more understanding or authoritative than they really are. The concern, in short, centers less on the technology behind the chatbot than on the answers it gives.

That shift creates a thornier problem. Most people agree that providers should take reasonable steps to prevent foreseeable harm. Deciding what those steps should be is harder when the product itself consists of a conversation. The most obvious safety measures change the chatbot’s answers: limiting the topics it will discuss, making its responses more cautious, or refusing certain conversations altogether. Yet much of conversational AI’s value lies in its ability to respond directly, personally, and dynamically to what a user asks.

Responsibility for those interactions can be difficult to assign. A chatbot’s answer emerges from an exchange with the user, rather than in isolation. The provider controls the model and the conditions under which it operates, but the user shapes the conversation that produces a particular output.

As liability comes to depend on what a chatbot should or should not have said, regulators and courts will have to distinguish responsibility for harmful outputs from control over the substance of AI responses. In drawing that line, they may move from deciding who bears responsibility for an answer to deciding what the “right” answer should have been.

Read the full piece here.

The frontier AI race’s latest safety proposal would have the leading contenders agree on how fast to run. Antitrust law has a less flattering name for that arrangement: a cartel.

That is the central problem with the “pacing the frontier” plan that Anthropic CEO Dario Amodei unveiled last weekend, and that fellow AI executives Sam Altman, Elon Musk, and Demis Hassabis welcomed within hours. The plan would have leading AI labs coordinate on safety standards and limits on the pace of development. The proposed safety goals are legitimate but the arrangement would go too far in seeking to set the pace of development.

The first and most immediate problem is that Amodei’s proposal seems better suited to calming fears about overcapacity and overinvestment in AI than to controlling the risks of recursively self-improving systems—AI that can help create still more capable AI. Its primary effect could be to give labs relief from an expensive development race under the banner of safety.

The plan also presents antitrust enforcers with a troublesome mismatch. They could verify that labs slowed development or limited computing inputs, but not whether the delay produced better safety research and testing. A pacing cartel could therefore deliver less AI without delivering safer AI, all while allowing competitors to manage their commercial risks collectively.

If Anthropic’s competitors adopted the plan, they would sacrifice the potential benefits of faster innovation in exchange for largely discretionary safety improvements that less restrictive tools could likely deliver. Antitrust authorities should not bless such an agreement when better private arrangements and public-policy measures are available and, arguably, already in place.

Amodei may still be right about the underlying danger. Frontier AI—the most advanced AI systems under development—may pose rare but potentially catastrophic risks. After the OpenAI–Hugging Face incident, those dangers look less like unforeseeable “black swans” than recognizable “white swans.” In that incident, hundreds of OpenAI agents coordinated through an unauthorized message board to escape their sandbox, a restricted testing environment, and attack a third party’s servers.

That episode demands serious thought about the right market and regulatory responses. But deliberately slowing innovation, at potentially enormous cost to consumers, is the wrong one.

If fierce competition at the AI frontier magnifies catastrophic risks, the answer is to make labs bear the costs their choices impose on others while leaving them free to compete.

Read the full piece here.

I’ve criticized plenty of private antitrust cases. This may be the first in which the plaintiffs sued to stop a genuinely bad idea—and I still think they should lose. That is the odd posture of the antitrust challenge to Mayor Zohran Mamdani’s plan for subsidized grocery stores in New York City.

On Sept. 9, the National Supermarket Association and two New York supermarkets filed suit in the U.S. District Court for the Southern District of New York. They allege that New York City and the New York City Economic Development Corp. are attempting to monopolize “retail sales of weekly-stock-up groceries at full-service stores” in certain New York City markets through the Mamdani plan for discounted city-run grocery stores (here, here, and here).

The gravamen of the complaint is that the planned discounts would amount to predatory pricing. That’s a tough row to hoe for several reasons, but the demanding predatory-pricing standard is hardly the plaintiffs’ only obstacle. The case raises some interesting issues, but I think it’s a loser. I’ll get to that.

Read the full piece here.

The State of New Jersey’s antitrust case against Amazon begins with a serious allegation: that Amazon restricted drivers from moving among competing delivery contractors. But the state’s complaint goes much farther. It argues that Amazon’s entire Delivery Service Partner (DSP) system unlawfully gives the company monopsony power over both the small businesses that deliver its packages and the drivers they employ.

The complaint, filed Aug. 4, tests how far antitrust law should reach into labor arrangements. Invoking Sections 1 and 2 of the Sherman Act, the New Jersey Antitrust Act, and the state’s authority to seek damages and injunctive relief, New Jersey asks a federal court to treat an integrated logistics system as an unlawful buyer cartel. It also leaves the door open to structural remedies against a business model Amazon created to expand its delivery capacity.

The allegations deserve serious attention. An agreement among competing employers not to hire one another’s workers can deprive those workers of a valuable competitive option. Retaliation for organizing, if proved, can violate labor law and may signal an effort to suppress competition.

But the complaint repeatedly blurs three distinct propositions: that a worker or contractor has been mistreated, that a firm possesses buyer power, and that the firm unlawfully acquired or maintained that power through exclusionary conduct. Proving the first does not prove the other two.

New Jersey’s no-poach theory is the case’s strongest component. If the state can prove a broad and unjustified restriction on workers moving among genuinely competing DSPs, that claim could warrant discovery, a targeted injunction, or a settlement.

The broader monopsony theory is far shakier. The complaint defines markets around Amazon’s proprietary operating system, treats ordinary obstacles to job searches and worker-employer matching as barriers that effectively lock workers in, and assumes that lower compensation or demanding performance standards are anticompetitive without adequately considering their efficiency rationales. Most important, it never credibly explains how the challenged labor practices harm consumers in the downstream market for delivered goods and services.

New Jersey’s challenge therefore looks less like a careful application of antitrust law than an effort to stretch antitrust to reach distribution practices that have delivered substantial benefits to consumers.

Read the full piece here.

Booking.com wanted to become a one-stop shop for a traveler’s entire journey. European regulators saw a different itinerary: a one-way trip to deeper dominance.

In a closely watched Sept. 9 judgment, the European Union’s General Court upheld the European Commission’s 2023 decision blocking Booking’s acquisition of Etraveli. The court agreed that adding Etraveli’s flight-booking business to Booking’s dominant hotel-booking platform would strengthen network effects and create a “connected trip” ecosystem that rivals would struggle to replicate. It reached that conclusion despite finding errors in some of the Commission’s market-share calculations.

At the heart of the case lies the business ecosystem—a cluster of interconnected products and services built around a common platform. Once largely a matter of competition-policy theory, ecosystems have now moved squarely into EU merger law.

The ruling matters well beyond online travel. Policymakers have invoked ecosystems to justify new theories of harm, changes to existing competition-law provisions, and entirely new regulatory regimes. Booking/ETraveli—widely regarded as the first merger prohibition based on an ecosystem theory of harm—now gives that campaign judicial backing.

Read the full piece here.

Competition policy is supposed to protect the competitive process, not do double-duty as price regulation. Yet the line between the two appears to be blurring in both a U.S. antitrust case and the European Union’s enforcement of the Digital Markets Act (DMA).

On Aug. 14, after years of litigation with Epic Games, Apple submitted a proposed commission structure to the U.S. District Court for the Northern District of California. Under the proposal, Apple would collect a 15% commission on purchases made through links from iOS apps and 10% on subscription renewals. Small businesses participating in Apple’s video, news, and mini-app programs would pay 5%.

Four days later, Apple announced new terms for developers in the European Union, following what the European Commission called a “close dialogue.” Apple would charge 26% on App Store sales made through its in-app purchase system, 20% when developers use an alternative payment processor within an app, and 15% on purchases made through external links. It would also impose a new 5% “Core Technology Commission” on apps distributed through rival marketplaces or the open web. Reduced rates of 15% and 10% would apply to small businesses, partner programs, and subscriptions after the first year.

The parallels are hard to miss. Both the U.S. and EU regimes ostensibly seek to protect competition, yet both pursue that goal by dictating the rates Apple may charge developers that rely on its platform and technology. That looks less like conventional competition enforcement and more like price regulation.

This raises several important policy questions. Are mobile ecosystems so uncompetitive as to warrant rate-setting, which is widely regarded as a policy of last resort? If so, do U.S. antitrust law and the DMA give courts and competition agencies the proper tools for the job?

The stakes extend beyond the distributional question of how Apple and Epic divide a fixed pie. Poorly designed price controls can weaken competition and innovation over time, ultimately harming the very consumers that enforcers ostensibly mean to protect.

Read the full piece here.

ChatGPT can draft a memo, plan a trip, and search the web. As of Aug. 31, it can add another line to its regulatory résumé: The European Commission has designated it a “very large online search engine” under the Digital Services Act (DSA).

The label may sound technical, but it captures a central problem for digital regulation: What happens when a service does too many things to fit neatly into any one legal category?

That question sits at the heart of today’s competition-policy debate over whether the wave of digital-market regulations adopted in recent years remains fit for purpose. The rise of artificial intelligence (AI), particularly AI-enabled applications, is already testing how well these frameworks can adapt to technologies their drafters did not anticipate.

The existing regimes face two related challenges. First, regulators must determine how to classify AI applications that become new gateways to digital services. Second, they must decide how to respond as new AI providers reshape competition and perhaps acquire gatekeeping power of their own. AI assistants and agents—tools that can perform multistep tasks on a user’s behalf—could upend familiar patterns of web browsing, online search, and e-commerce.

Of course, the competitive effects could cut in either direction. Incumbent platforms could integrate AI features in ways that shut out emerging providers. Alternative, new entrants could use those same technologies to challenge or even displace heretofore entrenched firms. Today’s disruptors, in other words, could become tomorrow’s gatekeepers.

Both possibilities expose the same basic problem: AI services do not fit neatly into the regulatory boxes that governments have built for digital markets. They increasingly behave like platforms, connecting users with other services and performing functions associated with search engines, online marketplaces, and virtual assistants. Yet frameworks organized around those predefined categories struggle to capture services that combine all three.

Read the full piece here.

First, “Uber” became a verb. Now, in parts of Africa, it is becoming past tense.

The company’s withdrawal from Nigeria and Uganda complicates one of competition policy’s favorite storylines: U.S. technology platforms enter, conquer, and never leave. That assumption has helped fuel a rush to import rules modeled on the European Union’s Digital Markets Act (DMA), which imposes special obligations on the largest digital platforms. Yet African ride-hailing markets are not following the script.

Uber once seemed built for precisely that story. Born from a founder’s struggle to find a taxi on a snowy night in Paris, it quickly became one of the most disruptive startups in recent memory and transformed urban transportation. Its African retreat shows that a revolutionary idea does not guarantee permanent dominance.

Nigeria and Uganda are part of a broader pattern. In September 2025, Uber announced its departure from Ivory Coast. Earlier this year, it announced that it would also leave Tanzania.

Uber still operates in Egypt, Ghana, Kenya, and South Africa, and has said it remains committed to the continent. Four departures in less than a year make that commitment look considerably narrower than it once did.

What explains the retreat? Uber’s public statements offer little guidance. Possible explanations include its inability—or unwillingness—to adapt its business model to local markets, fierce competition, currency volatility, rising operating costs, and a global strategy increasingly focused elsewhere.

Those possibilities raise the more interesting question: Does Uber’s retreat reflect failures peculiar to the company, or does it reveal something deeper about competition in African markets? This post argues for the latter.

Read the full piece here.

The price to beat in Amazon’s advertising auctions may not have come from another advertiser at all, according to a lawsuit filed Aug. 31 by the Federal Trade Commission (FTC). In the complaint, the FTC alleges that Amazon quietly set the price itself, promising one pricing rule and using another.

That allegation could support a conventional deception claim. If Amazon promised advertisers a genuine generalized second-price auction while deliberately concealing a material departure from that model, the FTC may have a viable case.

But the complaint tries to make a broader leap, arguing that Amazon’s pricing practices necessarily harmed retail consumers. The use of reserve prices, relevance scores, or dynamic pricing does not automatically make an auction unfair. Nor does it necessarily harm the consumers who buy the advertised products.

That is the complaint’s weaker theory. Based on the public record, the FTC has asserted retail-consumer harm with confidence but has yet to prove it.

Read the full piece here.

Just a few months ago, Huawei’s chairman observed that Washington’s export controls have done little to hold China back. Instead, they have pushed the company and the Chinese state to invest far more heavily in domestic champions, accelerating an already rapid climb up the technology ladder and clearing the way for global expansion. The latest proof is in Egypt and Huawei’s offer to supply Cairo with roughly 2,000 artificial-intelligence chips. This deal is a signal that Huawei now feels confident enough to pitch its wares abroad, on the strength of a domestic manufacturing base it expects to keep scaling.

Although this one deal won’t turn the country into an artificial intelligence powerhouse, it does expose a weakness in Washington’s export-control strategy: restrictions that slow Chinese companies at home can help them find customers abroad.

Read the full piece here.

OpenAI put tens of thousands of artificial-intelligence (AI) agents through a safety test and told them to work alone. Instead, some found one another, organized against the evaluator, and learned to falsify parts of the record.

The episode should reset the terms of the AI-safety debate. Testing, red teaming, and independent investigation remain essential. Their proper role is to uncover problems and guide continuous correction. No finite assessment can establish how an evolving, context-dependent system will behave across every prompt, user, tool, deployment, and interaction with other systems. And governments should not turn a snapshot into a permission slip.

This post extends the argument I developed in “Too Much Order, Too Soon” and “Artificial Intelligence, Natural Ignorance.” The problem goes beyond officials lacking enough information to design the right test. It’s that safety knowledge emerges through the same experimentation and competition that premature regulation could cut short.

That distinction anchors what I call applied Austrian economics (AAE). Austrian economics explains why knowledge—including knowledge about AI safety—remains dispersed, depends on context, and often emerges only through use. Public Choice—the study of how incentives shape government decisions—explains why a checkpoint created to gather that knowledge can become a gatekeeper, a competitive moat, and a source of false confidence. The practical task is to build safety throughout the AI lifecycle while preserving the decentralized discovery that makes improvement possible.

Read the full piece here.

Brussels has finally given Article 102 an instruction manual. Issued Sept. 3, the European Commission’s first comprehensive Guidelines on exclusionary abuses of dominance under the Treaty on the Functioning of the European Union (TFEU) replace a legal scavenger hunt with a single framework. The catch is that the manual still gives the Commission considerable room to decide when vigorous competition has become unlawful exclusion.

Until now, businesses, national authorities, and courts had to piece together the governing standards from treaty text, sporadic Commission decisions, and a growing body of judicial opinions. The Guidelines bring the rules on dominance, exclusionary effects, specific practices, and objective justifications into one framework. They also recognize that firms compete on quality, choice, and innovation, as well as price.

But the Guidelines do not mark a decisive return to disciplined, effects-based enforcement. They allow the Commission, in some cases, to dispense with a counterfactual, assess nonpricing conduct without a price-cost test, infer exclusion from conduct merely capable of producing it, and use presumptions to shift the evidentiary burden. The final text narrows some of the original draft version’s most aggressive express presumptions, as Global Competition Review has noted, but leaves ample room to expand liability later.

The Guidelines thus do more to organize and clarify Article 102 enforcement than to solve its central economic problem. The law remains highly receptive to precautionary intervention against successful firms, even when the challenged conduct may bring consumers substantial benefits alongside possible exclusionary risks.

No one seriously disputes that dominant firms can harm competition. The harder question—and the one these Guidelines do not settle—is whether the law can distinguish that harm from the ordinary process by which firms achieve dominance through investment, integration, innovation, and better service.

Read the full piece here.

Farmers have won the right to repair—at least, if they own John Deere equipment. In July, the Federal Trade Commission (FTC) announced a settlement requiring Deere to give farmers and independent repair shops access to the software, manuals, and other resources needed to fix its products. The settlement sensibly balances consumers’ immediate interest in affordable repairs against the long-term need to protect innovators’ intellectual property.

Deere had previously reserved many of those resources for its licensed repair shops. Such control—especially over onboard software—is common in industries ranging from vehicles and medical devices to consumer electronics.

Critics contend that these restrictions allow manufacturers to “monopolize” repair aftermarkets. That concern has fueled right-to-repair campaigns and legislation across the country. In the early 2020s, the FTC promoted the right to repair as an anti-concentration policy that could spur competition. Advocates have also framed it as a matter of property rights: A farmer who buys a tractor owns it and should generally be free to repair it.

But the right to repair also implicates intellectual property. As one scholar argues, “While copyright and IP law protect firms to some degree, the broad extent of most proposed [right-to-repair] legislation across the states and federal government constitutes a threat to this protection.”

That concern deserves attention. One need not resolve the broader tension between intellectual property and economic competition to examine the more fundamental legal tension between intellectual property and property rights.

Read the full piece here.

In the late 1800s, real estate agents gathered at local associations to exchange information about the homes they were trying to sell. The intuitive logic was that the best way to make a sale was to spread the word.

The real estate market has evolved considerably since then, but buyers and sellers, typically through their agents, are still trying to solve the same information problem. Buyers look for homes with specific features, such as location, number of rooms, and lot size, while sellers look for buyers whose preferences match the home they are selling. Real estate agents play a central role in this matching process by building a network that raises the likelihood of a timely sale.

Today, the leading real estate listing database is the Multiple Listing Service (MLS). It traces its origins to San Diego in 1885, when the local real estate board would distribute daily listings to its members. The MLS is not a single entity; it is composed of roughly 500 local real estate organizations, which are owned or controlled by brokers and associations in their respective territories. Members create the listings, while the regional MLSs distribute them to other agents and to public-facing websites such as Zillow, Redfin, and Realtor.com.

There are clear efficiencies from having a single, comprehensive database to economize on search costs and to raise the odds of a timely match. The tradeoff, however, is that, within a given territory, a viable second source of listings is unlikely to emerge. This raises the concern that whoever controls the listings data controls the terms of competition involving those listings.

This is exactly the issue at the heart of the recent antitrust dispute between Zillow, on one side, and Midwest Real Estate Data (MRED), a regional MLS serving the Chicagoland area, and Compass, the nation’s largest real estate brokerage, on the other. Zillow depends on MRED for listings in the Chicagoland area, where, according to Zillow’s complaint, MRED holds a 98% share.

Read the full piece here.

A price that knows your name may be creepy. It may also be a bargain. That ambiguity sits at the heart of the Federal Trade Commission’s (FTC) proposed enforcement policy statement on personalized pricing, which opens with a candid admission: The practice is “not well understood,” and its effects on consumers remain “unclear.” That uncertainty should shape the final policy.

Personalized pricing is a technologically refined form of price discrimination—charging different customers different prices for the same product or service. But it is not one practice with one predictable effect. Depending on the market, consumer demand, competition, and what would happen without personalization, it may lower prices, expand output, sharpen competition, or raise legitimate consumer concerns.

The FTC should therefore draw a clear legal line. A seller may not misrepresent what a price is, who can obtain it, or why it was offered. But a seller does not necessarily deceive consumers merely by failing to volunteer that individualized information helped determine the price. Turning that silence into a violation would allow a consumer’s assumption to create a duty to disclose—and could make beneficial discounts harder to offer.

As a new Mercatus Center policy comment puts it, “[a] final statement built around those limits, coupled with express protection for individualized discounts, would deter genuine deception and unfairness while preserving price competition and experimentation that can lower prices and expand output.”

Read the full piece here.

Executive Summary

Smartphones and social media are now central to adolescent life, prompting understandable concern about teens’ mental health, sleep, school performance, and compulsive use. Policymakers have responded with enacted or proposed age bans, verification requirements, parental-consent mandates, time limits, school-phone restrictions, and warning labels. This white paper assesses the evidence behind these interventions, their unintended consequences, and their constitutional durability.

The evidence does not establish that social media broadly causes adolescent mental-health problems. Average associations are generally small, often explain less than 1% of the variation in well-being, and depend heavily on model and measurement choices. Outcomes also vary by user and context. Social media is associated with harm for some adolescents but provides others with friendship, belonging, creative expression, and access to support—benefits that may be especially important for LGBTQ+, neurodivergent, rural, and otherwise isolated young people.

Broad mandates have weaker support than the public rhetoric suggests. Australia’s social-media ban and China’s gaming curfews illustrate the tendency toward circumvention and displacement. School-phone restrictions produce mixed results, while parental-consent mandates and warning labels lack meaningful outcome evaluations. Age verification creates privacy risks and may drive adolescents toward unmonitored access. Broad restrictions can also isolate vulnerable young people, entrench large platforms, and burden constitutionally protected speech.

More promising approaches target specific harms while preserving choice and beneficial uses. Voluntary friction tools, family controls, social-media literacy, physical exercise, industry standards, and independent certification all can promote healthier habits and are able to adapt as evidence develops. Heavy device use may also partly fill a void left by declining opportunities for independent play and in-person socialization. Restoring those opportunities may improve adolescent well-being more than banning their digital substitutes.

Prologue

An eighth grader walks through his front door after school, eyes fixed on his phone. He heads down the hall to his room, shuts the door, flops onto his bed, and spends the next four hours scrolling through social media, messaging friends, and watching videos.

No public parks lie within walking distance of his home. Shopping malls bar unsupervised minors. Like 66% of children his age, he cannot leave his yard without an adult present, and his parents cannot accompany him until they finish working for the day.[1] His peers face the same circumstances. Thus, they spend their free time alone together on their phones—the only source of social connection and exploration available to them.

Introduction

In a relatively short amount of time, the use of smartphones and social media have become ubiquitous across age groups. In 2010, manufacturers shipped 300 million smartphones worldwide, and social media had about 1 billion users.[2] By April 2026, the numbers of unique mobile-phone users and active social-media user identities had each reached approximately 5.8 billion.[3]

Time spent on social media also grew dramatically. In August 2010, the average visitor worldwide spent approximately 4.5 hours per month—or roughly nine minutes per day—on social-networking sites.[4] By 2025, online adults spent more than 150 minutes per day, on average, consuming social and video feeds when the measure included platforms such as YouTube and TikTok.[5]

This rapid growth has raised concerns about the harms associated with phone use generally and social media in particular, especially among adolescents. Surveys suggest that average social-media use among U.S. adolescents ages 12–17 approximately doubled over the past decade, from 2.5 to three hours per day in 2015 to about five hours per day in 2025.[6]

Over roughly the same period, the Centers for Disease Control and Prevention’s Youth Risk Behavior Survey recorded a sharp rise in the share of adolescents who reported that they “felt sad or hopeless for two or more weeks in a row.” The steepest increase occurred from 2019 to 2021, followed by a modest decline, while other measures of mental health remained relatively stable (see Figure 1).

FIGURE 1: U.S. High School Students’ Mental Health Indicators, 1991-2023

SOURCE: CDC Youth Risk Behavior Survey[7]

The apparent association between greater access to smartphones and social media and rising mental-health problems—sometimes compounded by concerns about school performance—has prompted calls for sweeping restrictions. Policymakers in many jurisdictions have proposed or adopted minimum-age bans, parental-consent mandates, time-based curfews, and universal age verification.

Yet evidence that these interventions improve adolescent mental health, educational outcomes, or other measures of well-being remains weak at best. There is stronger evidence that they impose substantial costs, including lost privacy, harms to vulnerable populations, circumvention that can undermine voluntary controls, and violations of constitutional rights such as freedom of speech.

Treating social media and screen use as the villain in this story is understandable but, as we argue in this white paper, incomplete and overly simplistic. Jonathan Haidt, a leading proponent of the view that smartphones and social media have contributed to deteriorating adolescent mental health, also argues that a parallel trend has received too little attention: the steady erosion of children’s and adolescents’ freedom to play, explore, and socialize independently.[8]

Haidt argues that increasingly restrictive parenting and institutional practices, adopted largely in the name of child safety, have deprived young people of opportunities to develop autonomy, resilience, and social competence through independent exploration and unsupervised play. In his account, adults have systematically overprotected children in the physical world while underprotecting them in the digital one.

That overrestrictive parenting in the physical world may, in fact, be the broader problem. Indeed, adolescents may suffer from too little in-person social connection “in real life” and seek substitute forms through smartphones and social media. If so, restricting their access to phones and social media may do little to address the underlying problem and could even make it worse.

This white paper evaluates policies intended to address harms associated with smartphones and social media, drawing on research in psychology, economics, and law. It begins by assessing the empirical relationship among smartphone use, social-media use, and mental health. As we document, the evidence is weak and ambiguous. For most adolescents, smartphones and social media appear neither uniformly good nor uniformly bad for mental health. They seem to benefit a significant minority and harm a smaller minority. Their effects also depend heavily on context, suggesting that blanket rules are unlikely to offer an effective solution.

Section II evaluates existing and proposed laws through the lens of law & economics. It asks whether these policies can reduce the harms associated with social media while preserving its benefits, especially given the wide variation in individual experiences. The central finding is that the most politically popular interventions—minimum-age bans, universal age verification, and mandatory time caps—have the weakest evidence of effectiveness and the highest documented costs. Section III examines these interventions’ unintended consequences, and Section IV assesses their prospects under the First Amendment.

We then consider alternatives that could help adolescents derive greater net benefits from phones and social media while reducing harms more effectively. In particular, we examine how parents and other caregivers can encourage healthier use, including by loosening restrictions on activities such as independent travel that build autonomy, agency, and in-person peer interaction.

I.          Smartphones, Social Media, and Adolescent Mental Health

On several measures, American adolescents report worse mental health than they did two decades ago. In the Centers for Disease Control and Prevention’s (CDC) 2023 Youth Risk Behavior Survey, 40% of high school students reported persistent sadness or hopelessness, up from 30% in 2013. The share who had seriously considered suicide rose more modestly, from 17% to 20%, while the share who reported attempting suicide remained roughly constant at 9%.[9] Academic indicators also disappoint, with math and reading scores flat or declining.[10]

These changes coincide with the spread of smartphones and social media among adolescents, but the evidence of a causal relationship remains far less clear. International data reveal a more complicated picture. In England, for example, the share of young people reporting longstanding mental-health conditions rose sixfold, from 0.8% to 4.8%, between 1995 and 2014—a period of rapid growth in smartphone and social-media use. Yet measured psychological distress barely changed.[11]

Across 41 countries surveyed from 1994 to 2022, adolescent psychological complaints followed markedly different trajectories from one country to the next, even as smartphone and social-media use followed broadly similar paths.[12] Across 35 countries, complaints in 2022 also substantially exceeded projections based on pre-2020 trends, suggesting that the COVID-19 pandemic and related policies may help explain the recent increase.[13]

These findings suggest that the relationship among smartphones, social media, and mental health is more complex than the account presented by Jonathan Haidt and some other scholars and activists. As Candice Odgers of the University of California, Irvine, observes, “Hundreds of researchers, myself included, have searched for the kind of large effects suggested by Haidt. Our efforts have produced a mix of no, small and mixed associations.”[14] Policymakers have nevertheless often proceeded as though the causal evidence were much clearer.

Before Section II evaluates policies intended to address the alleged harms associated with phones and social media, this section offers a more balanced assessment of the nature and strength of those relationships, drawing on the best evidence available at the time of writing.

A.             Main Effects and Their Limits

We begin with “main-effect” studies, which examine the overall relationships among social-media use, device use, and mental-health outcomes. Most use cross-sectional designs that measure variables at a single point in time. They can therefore identify correlations but cannot establish causation.

Although public discussion often treats this research as settled, the available evidence has several important limitations. First, results vary with the analytical model. Researchers have found effects ranging from minuscule to moderate, but the studies collectively offer no consistent and compelling account of how social media or device use affects adolescent mental health.

Second, the large datasets used in some studies can produce statistically significant results that have little practical importance due to small effect sizes. Statistical significance indicates that an observed relationship is unlikely to result from chance. It does not establish that the effect is large or meaningful.

Third, research generally cannot prove that something is safe. It can only fail to detect harm. Yet journals are less likely to publish null findings—results showing no effect—than studies reporting even a small, statistically significant relationship. This publication bias may make the evidence of harm appear stronger and more consistent than it is.

1.               Model Choices Shape the Results

Deciding which variables to study, how to define and measure them, and which statistical controls to include is as much art as science. Researchers can reasonably disagree about whether to control for a variable—accounting for it statistically—or investigate it as a potential mechanism through which an effect operates. Some models also provide more conservative tests than others.

Observed correlations often hinge on these choices. Results can vary depending on how researchers measure “social-media use” or “device use,” which outcomes they examine, and which other variables they control for.

One way to test whether results are robust rather than dependent on a single model is to examine many similar models at once and identify overarching trends. Researchers call this technique specification-curve analysis. Amy Orben and Andrew Przybylski of the University of Oxford used it to test about 20,000 model specifications.[15] As they note, “for some of the datasets analyzed there were more than a trillion different ways to operationalize a simple regression.”[16] That range underscores the value of examining the overall pattern rather than selecting a few models from among a trillion possibilities.

Orben and Przybylski found a “negative but small” association that explained at most 0.4% of the variation in well-being. They concluded that the effect was “too small to warrant policy change.”[17] Many other factors had stronger associations. Being bullied was 4.3 times more strongly associated with reduced well-being than technology use. Wearing eyeglasses was 1.5 times more strongly associated.

In 2022, Jean Twenge, Haidt, and their coauthors used the same data and analytical method but chose different model specifications. They found that:

[I]n particular, social media use (as opposed to all screen time) for girls (as opposed to all teens) showed much larger relationships with poor mental health, particularly when we removed possible mediators (such as unhappiness at school) from the set of variables controlled for, and when we treated all of the mental health scales equally.[18]

Some of these adjustments seem reasonable, including separating the effects by sex and examining social-media use rather than overall screen time. But the larger estimates depend heavily on treating unhappiness at school as a mediator—a pathway through which social media might affect mental health—rather than as a control. Both choices have defensible rationales, which the two sets of authors discuss in their published work.

Public discussion, however, rarely reflects how much the results vary or how strongly researchers contest them. It tends to favor Haidt and Twenge’s conclusions over those of Orben and Przybylski and other researchers discussed below who question the relationship.

2.               Small Effects and Measurement Error

A second dispute between Haidt and Twenge, on one side, and Orben and Przybylski, on the other, concerns whether the observed effects are large enough to justify policy intervention. Large datasets—the datasets analyzed by Orben and Przybylski included more than 355,000 participants—give researchers substantial statistical power. But they can also produce statistically significant results with little practical importance. In other words, a relationship may be unlikely to result from sampling chance yet still be too small to matter.

If social-media use has a weaker association with well-being than bullying and even wearing eyeglasses, policymakers should ask whether regulation is warranted and, if so, whether the proposed response is proportionate to the effect.

Consider an analogy. Suppose an official wants to determine whether a bowling ball is suitable for use in an official game. A sufficiently powerful microscope will establish beyond doubt that the ball carries a speck of dust. The speck is real, but it reveals little about the ball’s suitability. It does not alter the ball’s weight or shape enough to disqualify it. Statistical significance measures the sensitivity of the instrument, not the magnitude of what it detects. Datasets containing hundreds of thousands of respondents resemble extremely powerful microscopes: They can detect associations that are unlikely to reflect sampling chance but may remain too small to matter in practice.

Orben and Przybylski are not alone in finding statistically significant but small effects. In a 2025 meta-analysis, Christopher Ferguson of Stetson University and his colleagues pooled 79 effect sizes from 46 studies. They found an association of 0.061, below the 0.10 threshold that they and other researchers treat as the minimum of evidentiary interest.[19]

A University of Cambridge team led by Luisa Fassi reached a similar result. Its review covered 143 studies involving nearly 1.1 million adolescents.[20] The pooled estimate for time spent on social media, based on 56 of those studies and 482,273 participants, showed a statistically significant but small association with internalizing symptoms, such as anxiety and depression (r = 0.12). That correlation means social-media use accounts for roughly 1.4% of the variation in the outcome, leaving 98.6% unexplained. A 2026 meta-analysis of 153 longitudinal studies found similar effect sizes. It also found that video gaming appeared to improve executive function by about the same magnitude.[21]

Statisticians have proposed standard effect-size benchmarks to reduce the risk that researchers’ prior beliefs will shape how they characterize results. Guidelines commonly distinguish among small, medium, large, and even huge effects.[22] Researchers define those categories in different ways, but the U.S. Department of Education and others recommend benchmarks that account for the seriousness of the outcome, research norms, and effect sizes typical of the field. Social-media and device-use researchers would benefit from field-specific benchmarks that reduce the risk of well-meaning scholars either downplaying or exaggerating effects based on their subjective views.

Even these estimates may be too generous because researchers often measure phone and social-media use poorly. A preregistered meta-analysis by Douglas Parry and his colleagues found that self-reported and device-logged media use correlated at only r = 0.38, implying roughly 14% shared variance. Self-reports also rarely matched logged use.[23] Studies that ask adolescents to recall their screen time may therefore measure something quite different from their actual use.

The measurement error also may not be random. Evidence suggests that the gap between actual and remembered screen time varies with respondents’ well-being. Unhappy adolescents may overestimate their screen time, creating a false correlation between screen time and mood.[24]

This problem extends beyond social-media research. Common-method bias can arise whenever participants report both the proposed cause and the outcome at the same time. Features of the reporting method or the respondent’s outlook may then make the two appear more closely related than they are. Researchers can reduce this bias through better study design and statistical analysis.[25]

3.               Publication Bias and the File-Drawer Problem

The studies described above must be interpreted in light of journals’ longstanding reluctance to publish null results—findings in which a hypothesized relationship does not appear. This “file-drawer problem,” together with low statistical power, flexible analytical practices, and incentives favoring novel and statistically significant findings, have contributed to the replication crisis in social psychology. For many years, when researchers failed to reproduce a published effect, they were more likely to assume that something had gone wrong with the replication than to question the original finding.

These incentives complicate efforts to understand social media’s effects. Studies reporting small negative associations between social-media use and adolescent well-being are generally more publishable and newsworthy than studies finding no meaningful relationship.

A null result does not, however, automatically establish safety. A study may fail to detect harm because its measures are imprecise or its sample is too small. But large, well-designed studies that consistently find effects close to zero do provide meaningful evidence that any average harm is likely to be limited.

Public discussion often misses this distinction. Statistically significant but practically small negative effects receive disproportionate attention. Publication bias, media incentives, and people’s greater attention to negative information can therefore create a more alarming impression than the literature as a whole supports.

B.             Mechanisms and Competing Explanations

Because main-effect studies cannot establish causation, evidence about the specific pathways through which phones and social media might cause harm can strengthen the empirical case. The four leading candidates are sleep loss, social comparison and body-image concerns, social exclusion and bullying, and intermittent reinforcement that encourages compulsive use. Each mechanism is plausible, but the supporting evidence remains generally weak and often contradictory.[26] Moreover, the association in each case attaches not to time spent on a device or social media but to a particular behavior or preexisting vulnerability—each with an offline analogue.

Another possibility is that the association is not causal at all. Adolescents who spend more time on their phones differ from those who spend less time on their phones in ways that have nothing to do with the phones themselves. Several potential alternative pathways have been identified, including displacement and reverse causation.

Under the displacement theory, phone use competes for time with sleep, exercise, and in-person socializing—activities that predict adolescent well-being far more strongly than screen use predicts its absence. In Orben and Przybylski’s 2019 analysis, getting enough sleep and eating breakfast had positive associations with well-being 1.7 to 44.2 times larger than technology use’s negative association. When phone use crowds out those activities, it may serve as a marker for the displaced activity rather than independently cause the harm. The appropriate response would be to protect the activity, not to prohibit the device.[27]

Reverse causation between device use and mental health is also highly plausible, as is a bidirectional relationship in which phone use and poor mental health affect each other. An adolescent who is anxious, isolated, or depressed may have strong reasons to spend more time on a phone and fewer attractive alternatives.[28] In two longitudinal studies—one involving 594 adolescents and the other 1,132 college students—Taylor Heffer and her colleagues found that “social-media use did not predict depressive symptoms over time for males or females. However, greater depressive symptoms predicted more frequent social-media use only among adolescent girls.”[29]

In a longitudinal sample of 17,409 young people, Orben and her colleagues likewise found that lower-than-expected life satisfaction predicted greater subsequent social-media use. The relationship was statistically significant and consistent across participants ages 10–21 and across both sexes. The authors concluded that the relationship was bidirectional.[30] As Odgers and Michaeline Jensen put it, adolescents “at higher risk for depressive symptoms may selectively use social media more, or differently.”[31]

Age-specific life events and hormonal changes associated with puberty may also interact with social-media and device use. In the same longitudinal study, Orben and her colleagues found evidence that adolescents were more sensitive to associations between social-media use and lower subsequent life satisfaction during particular developmental periods. The first developmental window differed by sex, appearing at approximately ages 11–13 for girls and 14–15 for boys. A second window appeared at age 19 for both groups.[32]

The literature also supports other possible explanations for changing adolescent mental health, including schoolwork pressure,[33] the COVID-19 pandemic,[34] and mental-health awareness efforts that may themselves inflate reported symptoms.[35]

C.             Social Media’s Benefits

Studies that measure what adolescents experience on social media, rather than simply how many hours they spend there, find more benefit than harm. Given the negative public discourse surrounding social media, that result may seem surprising. It shouldn’t be.

Social media can help adolescents maintain friendships, find a sense of belonging, share humor, express creativity, explore their identities, seek help, and connect with peers who share their experiences. Adolescents themselves often describe social media in mixed terms. The Pew Research Center reported in 2025 that many teens saw benefits for connection and friendship while also reporting costs to sleep, productivity, confidence, and mental health.[36] [37]

The American Academy of Pediatrics similarly notes that social media can help adolescents feel connected, especially during periods of social isolation.[38] LGBTQ+ youth offer the clearest example. Matthew Berger and his colleagues’ systematic review found that social media may support LGBTQ+ youths’ mental health and well-being by fostering peer connections, helping them explore and manage their identities, and providing social support. These benefits are clinically significant. For adolescents whose identities receive little affirmation in their immediate environment, online connections can provide a major source of belonging and resilience.[39]

These benefits do not erase the risks. But they do mean that a blanket prohibition would impose certain, widely distributed costs to address risks that remain concentrated and uncertain.

D.            What the Evidence Supports

Several tentative conclusions follow from the forgoing. First, the science is not settled. Effect sizes vary with model specifications but tend to be small. In main-effect studies, digital-technology use typically explains less than 1% of the variation in adolescent well-being.

Second, an effect may be statistically significant yet have little practical importance, particularly when compared with variables such as bullying or even wearing eyeglasses. Third, journals’ reluctance to publish null results, combined with people’s greater attention to negative information, likely draws disproportionate attention to small negative effects. The nature of scientific research also makes a definitive “social media is safe” paper unlikely.

Fourth, confounding factors and competing explanations, including reverse causation, further muddy the picture. Odgers and Jensen summarize the state of the evidence well:

While future research may identify clear or stronger linkages, at present the available evidence falls short of the standard of proof required to identify digital technology use as a putative environmental cause of adolescent mental health problems.[40]

Fifth, any assessment must account for social media’s general benefits and its particular importance to some vulnerable populations.

Parents and policymakers may want definitive research establishing that social-media and device use are entirely safe, but science cannot provide that assurance. It can show that researchers have not detected specific harms—or have ruled out effects above a meaningful threshold—under specified conditions. And while some studies find negative effects under certain conditions, others find much smaller effects that affect narrower groups.

Until the research converges on a consensus that social media causes serious harm, public discussion should accurately reflect the mixed evidence. Policymakers should also demand evidence that interventions work before adopting them widely. Section II turns to that question.

II.          Do the Proposed Policies Work?

Despite limited and mixed evidence that device and social-media use cause substantial harm, policymakers around the world are rushing to embrace solutions. The threshold question is what problem these measures aim to solve. This section evaluates whether they produce meaningful benefits, including whether they achieve their stated objectives. Section III considers their unintended consequences.

Some states have enacted laws that restrict constitutionally protected speech. Courts have already enjoined several of these laws, while challenges to others remain pending. Section IV examines those cases and the governing constitutional precedents in greater detail. Here, we identify policies likely to face constitutional challenges in the United States, while focusing primarily on evaluating their effectiveness.

The most popular government responses include blanket age restrictions on device or social-media access, other age-based limits on the duration or timing of use, parental-consent mandates, and labeling requirements. Meanwhile, nongovernmental tools to address the public concerns include apps that let parents or users impose time limits or add “friction”—such as pauses or prompts that interrupt automatic use.

Evaluating most of these policies and tools remains difficult because studies measuring their effects on adolescent health are sparse or nonexistent. Most interventions are too new, have faced legal barriers, or have been implemented too inconsistently to permit credible evaluation.[41] But even when measured against their own often-dubious premises, the evidence that government mandates work remains weak.

A.             Age Restrictions and Parental Consent

Some of the most vocal proponents to “do something” about the harms attributed to social media have settled on mandatory age restrictions. Haidt, for example, advocates banning smartphones for children under 14 and social media for children under 16.[42]

The most straightforward case for these restrictions rests on two arguments. First, using the specified technology before the age cutoff is so widely harmful that narrower restrictions would not suffice. Second, voluntary parental limits face a network-effects and free-rider problem. Children may argue that barring them from smartphones or social media will isolate them because all their friends use those technologies. A mandate, in theory, relieves that pressure by imposing the same rule on everyone.

But there are strong reasons to conclude that a broad mandate would be too blunt. As the longitudinal study by Orben and her colleagues suggests, adolescent life satisfaction may be more sensitive to social-media use during certain developmental periods that differ by sex.[43]

Even if age restrictions could provide some benefits in principle, mandatory restrictions would also impose substantial costs. Indeed, the evidence discussed below suggests that they can produce the opposite of their intended effect. Rather than helping parents work with their children to manage social-media use, age bans may encourage teenagers to seek access to the forbidden fruit.

These mandates also raise serious constitutional concerns in the United States. Federal district courts in Arkansas, Louisiana, and Ohio have permanently enjoined laws requiring social-media platforms to verify users’ ages and obtain parental consent before allowing minors to create accounts.[44] The laws at issue were Arkansas’ Social Media Safety Act, Louisiana’s Secure Online Child Interaction and Age Limitation Act, and Ohio’s Parental Notification by Social Media Operators Act.[45]

The district courts applied strict scrutiny, in part because the laws required parental consent before minors could access constitutionally protected, nonobscene content.[46] As Section IV discusses, the 6th U.S. Circuit Court of Appeals reversed the Ohio decision in June 2026. The Arkansas and Louisiana injunctions remain in effect.

1.               Australia’s Under-16 Ban

Australia’s Constitution protects an implied freedom of political communication that may ultimately limit or invalidate the country’s effort to restrict teenagers’ access to social media.[47] Nonetheless, on Dec. 10, 2025, Australia implemented the world’s first national minimum age of 16 for social-media accounts.[48]

Leaving aside the policy’s broader merits and drawbacks, implementation has proved difficult. During the first month, platforms deactivated nearly 5 million accounts held by Australian minors.[49] That scale of initial compliance appears impressive—until one examines what happened next and what had come before:

• Self-reported age had already proved unreliable as an enforcement mechanism. Despite platforms’ existing minimum age of 13, 84% of Australians ages 8–12 reported using social media.[50]
• Virtual-private-network (VPN) use surged 170% on the ban’s first day. A consistent after-school spike suggested that teenagers quickly found workarounds.[51] The Bloomsbury Intelligence and Security Institute documented teenagers retaining access through previously falsified birthdates, VPNs, and AI-generated facial images designed to defeat verification. Others migrated to less-regulated platforms outside the ban’s coverage.[52]
• Current age-estimation technology has a margin of error of two to three years. Many 14- and 15-year-olds can therefore pass as 16, while many 16- and 17-year-olds may wrongly lose access.
• A June 2026 report based on an observational study of 436 Australian adolescents found that 86% of participants under 16 had accessed social-media sites covered by the ban after it took effect.[53]

Most critically, preventing young people from holding social-media accounts does not necessarily prevent them from accessing social-media content. Major platforms allow extensive browsing without an account. But logged-out users cannot use the safety features, parental controls, or restricted modes available through registered accounts.

Algorithmic ranking still operates for anonymous visitors, but their default feeds rely on popularity signals and geolocation rather than personalized, age-appropriate settings. Platform safeguards for minors—including screen-time prompts, private-by-default settings, restricted direct messaging, and sensitive-content filters—do not attach to sessions with no identified user.

Regulatory attention has begun to shift toward this gap, with some jurisdictions considering limits on anonymous browsing for all users. For now, account-level bans risk moving teenage use to the least-supervised surface of a platform. The result may be perverse: A policy intended to protect teenagers could make some of them less safe by pushing them from monitored accounts to unmonitored browsing.[54]

B.             Time Limits and Digital Friction

Limits on time spent using smartphones, social media, and other online services have become popular among adults and adolescents alike. In response, entrepreneurs have developed an impressive array of friction-based tools that help users limit their overall phone use or time spent on particular apps. These tools appear effective among people who choose to use them. For example:

• The “one sec” app inserts a brief breathing exercise before opening a social-media app. In a six-week field study of 280 users, it reduced actual app openings by 57%.[55]
• So-called “nudge-based” interventions—including disabling nonessential notifications, switching phone displays to grayscale, and moving social-media apps off the home screen—significantly reduced problematic smartphone use, screen time, and depressive symptoms.[56] One study found that grayscale settings alone reduced daily screen time by approximately 20 minutes.[57]

These friction-based tools work by targeting the automatic, habitual dimension of app use. They do not restrict access or require identity verification. They simply make compulsive use less effortless. Some evidence suggests that voluntary tools can also help users experiencing mental-health challenges.[58]

But these findings cannot be generalized to all users, much less treated as evidence that mandatory time limits would work. Voluntary tools chosen by motivated users differ fundamentally from restrictions imposed on everyone.

1.               China’s Gaming Limits

Mandatory time limits, unlike voluntary ones, appear ineffective and may prove counterproductive. China has restricted minors’ online gaming since 2019, when it limited users under 18 to 1.5 hours per day and three hours on holidays. In 2021, China tightened the restriction to just three hours per week and limited to 8 to 9 p.m. on Fridays, Saturdays, Sundays, and public holidays.[59]

One study reported a 93.6% compliance rate under the stricter regime, which China enforced through real-name verification and facial-recognition technology.[60] Yet that figure masks a more complicated reality. A survey of 430 adolescents conducted three months after the stricter limits took effect found that 17% of heavy gamers circumvented them. Of those surveyed, 3% rented other people’s gaming accounts, and 14% used a family member’s identity. Meanwhile, 59% shifted to watching short videos, which the restrictions did not cover.[61]

An analysis published in Nature Human Behaviour examined 7 billion hours of Tencent telemetry data—records of users’ actual play—and found no credible evidence that either the 2019 or 2021 policy reduced the prevalence of minors’ “heavy playtime,” defined as at least four hours per day on at least six days per week.[62] What’s striking about this result is that even China’s authoritarian state, with its aggressive use of identity-verification and surveillance systems, could not measurably reduce heavy play in the market segment studied. Instead, the evidence suggests that some adolescents evaded the restrictions or shifted their attention to less-regulated, less-visible, and potentially more harmful activities.

China’s gaming curfews represent an extreme version of the device-use and social-media restrictions now under consideration elsewhere. They also illustrate why evidence about voluntary tools cannot establish that mandates will work. Self-selected users adopt voluntary limits because they want to change their behavior. A platform-enforced cap creates different incentives and may produce evasion, displacement, and compensatory use—the substitution of one digital activity for another.

In the United States, Virginia enacted Senate Bill 854, which imposed a default one-hour daily limit on social-media use by users under 16. A federal court preliminarily enjoined the law in February 2026.[63] New York[64] and Illinois[65] have also enacted laws restricting algorithmically curated feeds and nighttime notifications for minors. Both laws have drawn constitutional objections and may face legal challenges.[66]

C.             School Phone Restrictions

School phone-use restrictions enjoy broad support, whether adopted by individual schools or imposed by law. Among the policies reviewed in this white paper, they are also the only measures evaluated through multiple rigorous studies conducted over different periods and in different jurisdictions. Some studies find improvements in academic performance, well-being, or both. The effects, however, are generally modest and context-dependent. It should also be noted that school phone restrictions apply to many activities beyond social media.

A 2016 study by Louis-Philippe Beland and Richard Murphy found that English schools that banned classroom phone use experienced modest improvements in test scores, with larger gains among lower-achieving students.[67] A 2025 study by David Figlio and Umut Özek likewise found small improvements in test scores at Florida schools.[68] Neither study examined mental-health effects, but a 2024 review of five studies by Kristof De Witte and Joanna Maldonado did find stronger evidence of benefits to social well-being than to academic performance.[69]

A 2025 study by Victoria Goodyear and colleagues compared 20 English secondary schools that restricted daytime phone use with 10 that permitted it. The researchers found “no differences in adolescents’ self-reported mental wellbeing, anxiety, depression, problematic social media use, and their motives for using social media.”[70] The restrictions also did not improve sleep duration or efficiency, physical activity, academic attainment, or disruptive classroom behavior. The authors noted that “restrictive school phone policies did not lower the overall time adolescents spent on their phones/social media.”[71] Restrictive policies may nonetheless reduce the time teachers spend addressing phone-related incidents and thereby lower staff costs.[72]

A 2024 study by Sara Abrahamsson reached different conclusions. Using the staggered adoption of smartphone bans across Norwegian schools, Abrahamsson found that classroom bans reduced consultations for psychological symptoms by 60%, reduced visits to general practitioners for psychiatric conditions by 30%, and increased grade-point averages by 0.08 standard deviations among middle school girls after two years.[73] She found no significant effects from less restrictive policies and no significant effects among boys.

This mixed evidence does not support broad conclusions about the effectiveness of school phone restrictions. All that can be said is that their effects appear to be context-dependent. In some settings, they may improve the learning environment, produce modest academic gains, and reduce staff costs. The evidence is less clear that they are effective as mental-health interventions, particularly for boys or when students compensate for reduced in-school phone use by spending more time on their phones outside school.

D.            Parental Consent Mandates

Alongside phone bans and age restrictions, parental-consent mandates—which require a parent to authorize a minor’s social-media account—are among the most politically popular proposals.[74] Yet our review found no study assessing whether these requirements improve adolescent mental health, reduce cyberbullying, improve sleep, or produce any other measurable health benefit.

These mandates may impose their greatest costs on young people in unsupportive households. LGBTQ+ youth who have not disclosed their identities to their parents, for example, may lose access to vital social support, information, and community resources if a parent refuses consent. Section III examines these effects in greater detail.

E.              Warning Labels

California’s Assembly Bill 56, which is set to take effect in January 2027, will require black-box-style health warnings on social media unless a platform determines that a user is older than 17.[75] The statute requires platforms to display a warning for at least 10 seconds, covering at least one-quarter of the screen, when a user first opens the platform each day. After three hours of use in a day, the platform must display a warning that cannot be dismissed for at least 30 seconds and covers at least three-quarters of the screen. It must repeat that warning at least once each hour thereafter.

The prescribed warning states that “the Surgeon General has warned that while social media may have benefits for some young users, social media is associated with significant mental health harms and has not been proven safe for young users.” Then-U.S. Surgeon General Vivek Murthy similarly called for tobacco-style warning labels on social-media platforms in 2024.[76]

The evidence for this approach is essentially nonexistent in the social-media context. No study has evaluated whether social-media warnings improve adolescent mental-health outcomes. The analogy to tobacco is instructive, though perhaps not in the way advocates imagine. Tobacco warning labels accompanied a broader cultural shift in attitudes toward smoking, but little reliable evidence shows that the warnings alone changed behavior. At most, they played a minor role alongside the more substantial and observable effects of excise taxes, public-education campaigns, smoke-free workplace rules, and alternatives such as e-cigarettes and heated-tobacco products.[77]

Consumers generally pay attention to product labels when those labels provide information directly relevant to a goal they are pursuing.[78] A consumer with a gluten sensitivity, for example, may seek products labeled “gluten-free” and avoid those labeled “contains wheat” or “contains gluten.” But a consumer without such a sensitivity will likely ignore those labels and choose based on price, brand, size, familiarity, or other product characteristics. Adding a warning label does not itself create the concern needed to make consumers notice or act on it.

Research also documents several ways consumers resist or disregard warnings. They may decline to read them, look away, discount the source, or treat the warning as paternalism or legal boilerplate rather than evidence of genuine risk. Some even experience reactance—a desire to resist perceived efforts to control their choices.[79] These responses are strongest among heavy users, for whom repeated warnings provide the least new information.

The indirect evidence from digital disclosures offers even less reason for confidence. In one experiment, only 7.3% of users presented with an optional terms-of-service disclosure chose to view it. Those who did kept it open for an average of 22.9 seconds—enough time to read an estimated 2.5% of the document.[80] Cookie banners required by the European Union’s General Data Protection Regulation (GDPR) operate much like repeated regulatory warnings. They have produced substantial consent fatigue and habitual dismissal. Moreover, experiments show that users’ responses depend heavily on a prompt’s design and the effort required to respond, while audits have found widespread discrepancies between users’ apparent choices and websites’ actual tracking practices.[81]

A deeper problem is that much social-media use is habitual rather than deliberative. There are cues that often prompt users to open an app, while unpredictable social rewards reinforce repeated checking and continued engagement. As habits develop, behavior becomes more automatic and less dependent on conscious goals, leaving little room for the deliberative processing on which warning labels depend.[82] Research distinguishing reflective decision-making from automatic behavior predicts that warnings aimed at conscious thought will have limited effects on habitual conduct. The available evidence supports that prediction.[83]

The available evidence therefore provides little reason to expect mandatory warnings to meaningfully reduce harmful social-media use among adolescents.

F.              Voluntary Tools and Mandatory Rules

While direct evaluations remain sparse, the available evidence suggests that voluntary tools—including time limits, grayscale displays, and other forms of digital friction—can help some users manage their phone, app, and social-media use. Mandatory interventions, by contrast, show little evidence of effectiveness and may prove useless or counterproductive. As Section III explains, mandates can also produce unintended consequences that cause substantial net harm.

III.          The Unintended Costs of Mandatory Interventions

Even if policymakers could, in principle, justify an intervention as a way to reduce harms associated with adolescent social-media use, they must also account for its unintended but predictable consequences. Those consequences may diminish the intervention’s benefits or make it harmful on balance.

This section examines the principal unintended effects of the mandatory interventions discussed in Section II. It demonstrates that several popular measures already impose substantial, concrete harms. These include privacy and security risks from age-verification systems, circumvention and displacement, disproportionate burdens on vulnerable groups, and compliance costs.

A.             Privacy and Security Risks of Age Verification

Enforcing any age-based restriction requires some way to determine users’ ages. Unfortunately, most verification methods now in use create substantial privacy and security risks. In February 2025, the European Data Protection Board warned that age-assurance systems could threaten data protection and other fundamental rights.[84] In March 2026, 405 cybersecurity experts published an open letter warning that mandatory age verification “breaks the internet’s security model.”[85] A 2025 data breach illustrated that risk vividly by exposing government-identification documents belonging to 70,000 Discord users.[86]

Common age-verification methods include:

• Self-declaration, which users can easily circumvent by lying about their age.
• Age estimation, which infers age from a facial image or other traits. These methods have margins of error measured in years and may require platforms or third parties to process biometric information, raising concerns about the security of data storage.
• Document-based verification, which requires users to submit personally identifiable information from a driver’s license, passport, or similar document, sometimes combined with biometric authentication. Collecting and storing this information creates significant privacy and security risks.

These privacy costs fall on everyone, not just teenagers. Adults must also prove their age to access platforms covered by verification mandates. If platforms require users to establish their ages explicitly, laws intended to protect adolescents could create a universal identity-disclosure regime without any demonstrated benefit for the adolescents they seek to protect.

Privacy-preserving alternatives are possible and already exist. Zero-knowledge proofs (ZKPs), for example, can establish that a user exceeds a specified age without revealing the user’s identity, birth date, or underlying personal information. Properly designed systems can also make successive proofs unlinkable, preventing websites or credential issuers from connecting separate age checks and building a record of an individual’s online activity.

The European Commission has identified ZKPs as a way to provide age assurance without transmitting personal data and has incorporated them into its open-source age-verification architecture. Google has also developed ZKP technology for age assurance, initially in a partnership with Germany’s Sparkasse banking network. The company also open-sourced its ZKP libraries in July 2025 so that other digital-identity and age-assurance providers could implement them. Decentralized-identity services such as zkMe likewise offer ZKP-based credentials that allow users to prove they satisfy an age threshold without disclosing the underlying identity information to the service requesting verification.[87]

Major social-media platforms do not appear to have widely adopted these tools. That limited adoption may suggest that consumers have not demanded stronger privacy protections. It could also reflect implementation costs, interoperability barriers, or platforms’ limited incentives to minimize data collection. Nonadoption therefore provides little evidence about how much consumers value privacy.

B.             Circumvention and Displacement

Evidence from Australia’s age-based social-media restrictions, China’s online-gaming limits, and school phone-use policies shows a consistent pattern: Squeeze activity in one place and it bulges out somewhere else. Australia’s surge in virtual private network (VPN) use, China’s rental market for gaming accounts and 59% shift to short-form video, and the SMART Schools finding that school restrictions did not reduce overall phone use all illustrate this dynamic.

Displacement is especially concerning in the social-media context because logged-out browsing deprives users of safety features, parental controls, and restricted modes. The Brookings Institution identified displacement as a principal adverse consequence of age-based bans.[88] The Organisation for Economic Co-operation and Development (OECD) reached a similar conclusion, noting that restrictive policies may paradoxically increase exposure to harmful content rather than reduce it.[89]

Snapchat’s early growth illustrates how school restrictions can produce unintended consequences. Evan Spiegel and Bobby Murphy launched the disappearing-photo app in 2011, but it initially attracted little attention. Spiegel’s mother then told his cousin, a student at a high school in Southern California, about the app. The school had blocked established social networks, including Facebook, but had not yet discovered Snapchat. Students adopted it as a workaround—a way to exchange virtual notes, photos, and jokes during class without leaving a lasting trail for teachers or parents.[90]

The app then spread through other Southern California high schools. Its daily active users rose from about 3,000 to 30,000 in one month in early 2012.[91] By February 2013, users were sending more than 60 million messages through the app each day.[92]

C.             Disproportionate Burdens on Vulnerable Youth

Broad access restrictions impose their greatest burdens on those young people who benefit most from online social connection and support. Research consistently finds that social media provides important benefits to LGBTQ+ youth, although it can also expose some users to harm. In a 2024 survey of 406 people ages 14–18 who identified as LGBTQ+, Celia Fisher and colleagues found that social media served as both a vital resource and a source of risk. The benefits centered on identity exploration and community connection.[93] A 2025 review by Carolyn Bacaj and colleagues discussed studies showing that social media provides access to identity-affirming content, peer support, role models, and mental-health resources that may be unavailable in local communities.[94]

An earlier systematic review of 26 studies likewise found that social media supported LGBTQ+ youths’ mental health and well-being through peer connection and identity management.[95] The authors explained:

LGBTQ youths negotiated and explored identity and obtained support from peers on social media. Instagram, Tumblr, and Twitter were commonly used to access LGBTQ content owing to ease of anonymity. Identity management was the most studied social media affordance, important to LGBTQ youths for strategic disclosure. Key strategies for managing identities included being anonymous, censoring locations or content, restricting audiences, and using multiple accounts.[96]

These benefits can be substantial. For young people whose offline environments are unsupportive or openly hostile, online communities may provide the only reliable sources of accurate health information, crisis support, and peer recognition. They also offer civic infrastructure through which LGBTQ+ youth can find mentors, legal resources, and pathways to adulthood.

Parental-consent mandates pose a particular threat to LGBTQ+ youth who are not out to their families. Requiring parental authorization gives unsupportive parents effective veto power over a child’s access to community resources. The burden falls most heavily on young people with the fewest alternative sources of offline support. For this reason, LGBTQ+ advocacy organizations, including The Trevor Project and GLAAD, along with civil-liberties groups such as the American Civil Liberties Union, have opposed broad restrictions that threaten LGBTQ+ youths’ access to affirming online resources.[97]

Other vulnerable groups face similar risks. Young people in rural areas, those with disabilities or living in abusive households, and those in undocumented families may lose disproportionately from blunt access restrictions.[98] Research indicates that many neurodivergent adolescents and young adults—particularly those with autism and, to a lesser but growing extent, those with attention-deficit/hyperactivity disorder (ADHD)—use Discord, Reddit, and specialized forums for social connection, peer support, and interest-based communities.[99] These forums can also offer asynchronous communication, lower sensory demands, and greater control over social interactions that may be less available in offline peer settings.

Access regimes that treat all minors as uniformly at risk can impose especially steep costs on these users. The U.S. Surgeon General’s advisory acknowledges such benefits but offers no framework for preserving them within restrictive regulatory systems.

D.            Compliance Costs and Incumbent Advantage

Developing, enacting, implementing, and litigating laws governing phone and social-media access consumes legislative resources and public attention that could support policies with clearer net benefits. Even measures that courts later enjoin or invalidate force device makers and social-media platforms to devote resources to compliance and litigation. Those resources could instead support product improvements, including features that make phones and platforms safer for adolescents.

Implementation costs are also easy to underestimate. Platform-level compliance costs can be difficult to measure, but they can be substantial, particularly for smaller firms. The EU’s GDPR, for example, imposed fixed costs that fell disproportionately on businesses with fewer resources. After the GDPR took effect, hundreds of U.S. websites—including many news sites—blocked European visitors rather than incur those costs. Some remained unavailable in Europe for months or years.[100]

The OECD warns that disproportionate compliance burdens may weaken competition and entrench incumbent platforms.[101] The GDPR experience illustrates that risk. Using data from 1,084 online firms, Samuel Goldberg and colleagues estimated that EU page views and revenue each fell by approximately 12% after the regulation took effect. Smaller e-commerce sites experienced a 16.7% decline in recorded revenue, compared with 7.9% among larger sites.[102]

The political economy of social-media regulation presents a related concern. Supporters may form a modern “Baptists and bootleggers” coalition, in which groups with sincere public-interest motives align with actors that stand to benefit materially from the same restrictions. The “Baptists” include parents, pediatricians, educators, mental-health advocates, and child-safety organizations who argue that platforms expose minors to addictive design, harmful content, privacy intrusions, and developmental risks.

The “bootleggers,” in this analogy, are not necessarily illicit actors. They are parties whose private interests align with regulation for other less public-regarding reasons. They may include incumbent platforms that can absorb compliance costs more easily than smaller competitors, vendors that sell mandated age-verification and monitoring tools, legacy media companies and other businesses competing for users’ attention, and political actors seeking greater control over online-speech infrastructure.

This does not mean that social-media regulation is unjustified, nor does it imply that supporters of the proposed regulations act in bad faith. The point is to show how broad regulatory coalitions can combine genuine protective motives with private economic or institutional incentives. Policymakers should therefore examine not only a law’s stated purpose, but who gains power, revenue, or competitive advantage from its implementation.

E.              Reactance and the Forbidden-Fruit Effect

Human fascination with that which is forbidden goes back to the origin of the species, at least by some accounts. The notion of “forbidden fruit” comes from the biblical account of Adam and Eve, but modern researchers across disciplines have documented the same basic tendency. Restrictions can make the prohibited object or activity more appealing.

Psychologists call one version of this response reactance. Efforts to control people’s behavior by forcing or preventing an action can provoke resistance and strengthen the desire to do what has been forbidden.[103] Scarcity can likewise increase an object’s perceived value.[104]

Outside the psychology laboratory, these dynamics may help explain the so-called “Streisand effect.” The term arose after Barbra Streisand sued to suppress an aerial photograph of her Malibu home, causing public interest in the image to surge. Evidence from Australia and China reflects a similar pattern. When governments restrict access to social media or video games by age or time of day, adolescents find workarounds such as VPNs, borrowed credentials, or rented accounts.

These mechanisms suggest one reason age-based social-media restrictions could backfire. For some adolescents, making social media explicitly off-limits may increase its appeal and strengthen their motivation to circumvent the restriction.

People generally expect some control over their choices, including what to read and how to spend their free time. Parents spend much of the children’s youth helping their kids to develop the skills and judgment needed to govern themselves most effectively. Evidence on parental strategies for guiding social-media use suggests that guidance and support, combined with trust and progressively greater age-appropriate autonomy, tend to produce better outcomes for young users.

F.              The Costs of Broad Restrictions

Taken together, these unintended consequences weaken the case for broad restrictions on adolescents’ access to smartphones and social media. Age-verification mandates can impose privacy and security costs on all users. Restrictions may displace use rather than eliminate it and may deprive vulnerable young people of especially valuable connection and support. Fixed compliance costs can burden smaller competitors while entrenching incumbent platforms. Restrictions may also provoke circumvention and reactance, making prohibited technologies more attractive rather than less.

 IV.          First Amendment Limits on Social-Media Restrictions

Most of the mandates evaluated in Section II restrict access to speech. They condition a user’s ability to create a social-media account—and thus to speak, publish, and receive information—on age, parental consent, or the presentation of identifying documents. In the United States, such laws must survive review under the First Amendment, and many have not.

This section explains the constitutional framework that courts apply to these laws and that determines which laws survive. It also shapes how much legislative effort, litigation expense, and compliance cost go toward laws that courts ultimately invalidate.

When a law restricting speech is challenged, a court first determines the applicable “level of scrutiny,” meaning the strength of the justification the government must provide. That choice often drives the outcome. Courts apply three principal levels of scrutiny:

• Rational-basis review governs most economic and social regulations that do not burden speech or another fundamental right. A law survives if it bears a rational relationship to a legitimate government purpose. In practice, courts uphold nearly every law reviewed under this standard.[105]
• Intermediate scrutiny often applies to, among other things, regulations that do not target speech based on its content. The government must nonetheless show that the law advances an important interest and does not burden substantially more speech than necessary. It need not, however, adopt the least restrictive alternative.[106]
• Strict scrutiny applies to laws that restrict speech based on its content. The government must prove that the law advances a compelling interest and is narrowly tailored to that end, using the least restrictive means available.[107]

Constitutional scholar Gerald Gunther famously described strict scrutiny as “‘strict’ in theory and fatal in fact.”[108] That is an exaggeration, but not a large one. In a broad empirical study, Adam Winkler of the University of California, Los Angeles examined federal court decisions applying strict scrutiny from 1990 through 2003. He found that challenged laws survived about 30% of the time overall and about 22% of the time in free-speech cases.[109]

The Supreme Court recently made the point even more forcefully. “In the First Amendment context,” the Court explained, “we have held only once that a law triggered but satisfied strict scrutiny—to uphold a federal statute that prohibited knowingly providing material support to a foreign terrorist organization.”[110]

The central question in litigation over social-media restrictions is therefore which level of scrutiny applies. The answer depends on what speech the law reaches and whose access it burdens.

A.             The Government’s Interest and Minors’ Rights

Two well-established principles stand in tension. On one hand, the Supreme Court has repeatedly recognized the government’s compelling interest in protecting minors’ physical and psychological well-being.[111] On the other, minors have constitutional rights of their own. As the Court reiterated in Brown v. Entertainment Merchants Ass’n, “minors are entitled to a significant measure of First Amendment protection.” The government may bar the dissemination of protected material to them only in narrow, well-defined circumstances.[112]

The most important exception concerns sexual material that is obscene as to minors. Under Ginsberg v. New York, a state may prohibit the sale to minors of sexual material that remains protected for adults.[113]

The level of scrutiny applied to a particular law depends on how these principles interact. The Supreme Court’s 2025 decision in Free Speech Coalition, Inc. v. Paxton illustrates the distinction.[114] Texas required age verification on commercial websites where more than one-third of the content consisted of “sexual material harmful to minors.” Because the law restricted minors’ access only to material unprotected as to them and burdened adults’ access to protected speech only incidentally, the Court applied intermediate scrutiny and upheld the statute.[115]

Most of the social-media laws surveyed in Section II differ in critical respects. They restrict minors’ access to entire platforms, where the overwhelming majority of content is protected speech for minors and adults alike. They also typically require adults to verify their age or identity before speaking or receiving information.

Courts have generally subjected laws of this kind to strict scrutiny. The government must establish a compelling interest and show that the law is narrowly tailored, including that no less restrictive alternative would achieve its purpose. The precedents discussed below explain why courts often find both the evidentiary justification and the tailoring inadequate.

B.             Evidence and Less Restrictive Alternatives

In Brown, California defended its ban on selling violent video games to minors with psychological studies that purported to link violent games to aggression. The Court found the evidence insufficient to establish a compelling interest. The studies did “not prove that violent video games cause minors to act aggressively,” and the measured effects were “both small and indistinguishable from effects produced by other media,” including television cartoons that the law left untouched.[116]

The parallel to the evidence reviewed in Section I is difficult to miss. The main-effects literature on social media and adolescent mental health reports small, contested associations that typically explain less than 1% of the variation in well-being. The research also leaves plausible roles for reverse causation and confounding. A state defending a social-media restriction must do more than assert that platforms harm children. It must demonstrate causation with evidence of a kind that, so far, does not exist.

Even if a court accepts the state’s interest as compelling, the state must show that its law is narrowly tailored and uses the least speech-restrictive means available. In Ashcroft v. American Civil Liberties Union, [117] the Supreme Court affirmed a preliminary injunction against the federal Child Online Protection Act. The law required age verification for commercial websites carrying material harmful to minors. The Court found that blocking and filtering software, which parents could adopt voluntarily, was less restrictive and, on the record before it, likely more effective.[118]

That reasoning applies directly to social-media mandates. The parental controls, device-level settings, and friction-based tools described in Section II are precisely the kind of less restrictive alternatives that can prove fatal under strict scrutiny.

Brown identified another tailoring defect. California’s law was overinclusive because it restricted minors whose parents did not object to the material. It was also underinclusive because it ignored other media presenting similar content. A state may not abridge speech merely to enforce what it “thinks parents ought to want.”[119]

The 9^th U.S. Circuit Court of Appeals applied similar reasoning in 2024 to California’s Age-Appropriate Design Code Act. The court held that requiring covered businesses to assess and mitigate children’s exposure to “harmful” content effectively “deputizes covered businesses into serving as censors for the State.”[120]

1.               Diverging Lower-Court Decisions

Lower-court decisions have largely followed this framework. Federal district courts permanently enjoined social-media age-verification and parental-consent laws in Arkansas, Ohio, and Louisiana and preliminarily enjoined similar laws in Utah, Virginia, and Georgia. Those courts generally applied strict scrutiny and relied on Brown and Ashcroft in evaluating the laws’ evidentiary support and tailoring.[121] An appellate court later reversed the Ohio judgment.

In August 2025, the Supreme Court declined to block Mississippi’s age-verification law while litigation continued. That procedural decision did not resolve the merits. Justice Brett Kavanaugh nevertheless wrote separately that the challengers were “likely to succeed on the merits” because the law was “likely unconstitutional” under existing precedent.[122]

The appellate picture has since become less uniform. In November 2025, a divided panel of the 11th U.S. Circuit Court of Appeals stayed an injunction against Florida’s law restricting social-media accounts for children under 14 while the state appealed. At that preliminary stage, the majority treated the law as regulating addictive design features rather than speech based on its content, and therefore applied intermediate scrutiny. [123]

In June 2026, the 6th U.S. Circuit Court of Appeals reversed the permanent injunction against Ohio’s parental-consent requirement. The court did not categorically hold the requirement constitutional. Instead, a panel majority concluded that NetChoice could not assert minors’ First Amendment rights and had not carried the demanding burden required to establish that the law was facially invalid. Two months later, the same circuit vacated a decision denying a preliminary injunction against Tennessee’s social-media age-verification and parental-consent law. It remanded the case for consideration of NetChoice’s likelihood of success without deciding the merits.[124]

These decisions have produced divergent outcomes, but largely under different statutes and at different procedural stages. Courts have not yet settled when an age-verification or parental-consent law regulates protected speech based on its content. The Supreme Court may eventually need to clarify that question. For now, most laws restricting access to protected speech remain enjoined. The principal departures have arisen when courts characterized a law as regulating design features or resolved the case on standing or the demanding standard governing facial challenges.

For policymakers, the interventions with the weakest evidence of effectiveness—blanket age bans, universal verification requirements, and parental-consent mandates—also face the greatest constitutional risk. The connection is direct. Under strict scrutiny, a state must show that its law works and that less speech-restrictive alternatives wouldn’t work as well. Litigation also consumes legislative and judicial resources for years, while platforms incur compliance costs even if a court ultimately invalidates the law. Legally vulnerable mandates may also crowd out more effective and constitutionally sound approaches. Section V turns to those alternatives.

V.          Better Alternatives to Government Mandates

As Section I explained, social media is neither an unalloyed good nor uniformly harmful. For adolescents in particular, the challenge is to preserve its beneficial uses while limiting potential harms. This white paper has questioned many of the most politically popular interventions. Section II found little evidence that top-down mandates produce meaningful benefits. Section III showed that these interventions often cause unintended harms. Section IV examined one such harm in greater detail—the burden on constitutionally protected speech.

Broad mandates perform poorly, the consequences of other compulsory interventions are difficult to predict, and users’ experiences vary widely. These uncertainties call for epistemic humility—recognition of the limits of what policymakers can know in advance. They also support a decentralized approach that proceeds cautiously and adapts to evidence and feedback.

This section considers steps that users, parents, schools, platforms, and other organizations can take to help adolescents use smartphones and social media more safely and productively. Some measures address devices and apps directly. Others seek to restore opportunities for independent activity, in-person interaction, and other forms of socialization that have declined in recent decades.

A.             Clarifying the Risks and Benefits

Before turning to specific actions, it is worth considering the context in which they would operate. In recent years, a small number of highly visible researchers have focused public attention on the challenges associated with smartphones and social media. Their questions and concerns are important, and they deserve credit for raising them.

Some of these researchers have also become vocal proponents of the simplistic, one-size-fits-all policies discussed in Sections II and III—policies that often prove ineffective, counterproductive, or harmful. This advocacy has contributed to skewed reporting and public confusion about the nature and scale of the problems. The first step, then, is to describe the risks, benefits, and underlying causes more accurately.

With a clearer understanding of those tradeoffs, users, parents, educators, and other caregivers will be better equipped to decide how and in what contexts these technologies should be used—and which restrictions are most likely to work.

B.             Parent-Guided Tools and User Controls

Within this framework, parents can work with their children to modify how they use devices and social media. If particular features disproportionately reinforce habitual use, for example, parents can encourage adolescents to disable those features without abandoning the platform. This targeted approach can help adolescents regulate their use, reduce harmful experiences, and preserve beneficial ones.

Every major platform has introduced safety tools for teenage users, including Meta’s Family Center, TikTok’s private-by-default accounts for minors, YouTube’s supervised experience, and Snapchat’s Family Center. Platforms developed these features in response to pressure from parents, advocacy organizations, and the media, which made teenage safety a point of competition.

Third-party developers have also created tools designed to reduce harmful or excessive device use. The one sec app, for example, introduces friction by requiring users to pause before opening selected apps. Operating systems offer additional controls, including app-specific time limits, blue-light reduction, and grayscale displays. As Section II explained, substantial evidence indicates that friction-based tools such as one sec and grayscale settings can reduce app use.

More can still be done. The next subsection considers mechanisms that could encourage further innovation in tools that help adolescents and their parents manage technology use.

C.             Reputation, Standards, and Certification

New third-party apps and platform-design changes are promising, but several factors may inhibit the development of more and better apps and design changes. Ironically, one may be a consequence of public confusion driven by overstated claims from some proponents of government interventions. Parents who believe that denying adolescents access to devices is the only solution are less likely to help them develop healthier habits or to demand tools that support those habits.

A more fundamental obstacle is information. Most users lack the expertise needed to evaluate how particular devices, platforms, and design features affect attention, habits, and well-being. It is unrealistic to expect more than a small minority to acquire that expertise. The practical challenge is therefore to translate expert knowledge into signals that users can readily understand and trust.

Society has faced this problem before with other technologies that consumers didn’t fully understand. The established responses include brand reputation, industry standards, and independent standards and certification.

1.               Brand Reputation and Product Design

Brands communicate a range of product qualities to consumers. Because brands depend on reputation, their owners have strong incentives to protect users and preserve trust. For operating systems, social-media platforms, and third-party apps, that means designing products that users associate with well-being and control rather than manipulation or compulsive use.

If users, parents, schools, and employers come to view a “healthy digital experience” as an element of product quality, companies will have a reputational incentive to provide meaningful screen-time controls, friction before endless scrolling, clearer notification settings, age-appropriate defaults, more transparent recommendation systems, and tools that help users identify and limit harmful patterns.

This incentive may be especially strong for social-media platforms. A reputation for exploiting attention, amplifying harmful content, or mistreating vulnerable users can erode public trust, invite scrutiny, and weaken long-term loyalty. A reputation for respecting users’ time and well-being can, by contrast, become a competitive advantage.

2.               Industry Standards for Responsible Design

Industry standards could establish shared expectations for responsible digital design across operating systems, social-media platforms, and app ecosystems. Such standards might identify best practices for notification defaults, parental controls, transparency in recommendation systems, age-appropriate design, avoidance of dark patterns (interfaces that manipulate users into unwanted choices), app-store labeling, and tools that let users set and enforce personal usage goals across multiple apps.

Clear standards can reduce uncertainty for developers by defining good practices. They may also help smaller third-party developers by providing a design framework that makes their apps more acceptable to platforms, advertisers, schools, families, and enterprise customers.

3.               Independent Certification and Trust Marks

Independent organizations with relevant expertise could develop standards for evaluating device designs, social-media platforms, and third-party apps. Historically, trusted seals developed by organizations such as Underwriters Laboratories (UL) and Good Housekeeping have proven effective in helping consumers identify products that meet defined quality or safety standards.[125]

A similar certification system could independently signal that a digital product has satisfied established criteria for user well-being, safety, transparency, and control. For operating systems, certification could assess whether users have effective cross-app controls over notifications, usage limits, privacy, and age-appropriate settings. For social-media apps, it could evaluate recommendation systems, reporting tools, content controls, protections for teenagers, crisis-response pathways, and features that encourage repeated or prolonged use. For third-party apps, certification could recognize products that respect users’ attention, work with device-level controls, and disclose features designed to influence behavior.

Certification would make responsible design visible and marketable. Certified products could distinguish themselves from competitors, while users, parents, schools, employers, app stores, and regulators might scrutinize uncertified products more closely.

A consumer-facing seal would become valuable if users wanted a simple way to identify apps, platforms, and devices designed to support healthier digital experiences. Most users cannot realistically inspect a product’s recommendation system, notification architecture, privacy practices, or behavioral-design choices before adopting it. A trusted seal could reduce that burden. Parents selecting apps for children, adults seeking to curb compulsive use, schools choosing educational tools, and employers purchasing workplace software could favor products carrying a recognizable mark of quality and trust. As more buyers relied on the seal, developers and platforms would gain stronger incentives to earn it, advertise it, and maintain compliance.

Insurers could provide another source of demand. They might encourage or require certified digital products when certification reduces the risk of claims involving workplace harm, youth-safety incidents, data misuse, or other costly outcomes. Schools, employers, health systems, youth organizations, and app platforms might receive better insurance terms for using products that meet recognized safety and risk-management standards. Certification would also document due diligence by showing that an organization selected tools reviewed against accepted criteria rather than relying solely on vendor assurances. Over time, certification could become a practical market requirement in higher-risk settings, much as UL marks became important for products involving electrical risks.

This approach offers several advantages. Voluntary standards can adapt as the evidence develops, while statutes are difficult to amend. Certification also poses fewer constitutional concerns because it neither restricts speech nor mandates identity disclosure. Finally, market incentives may produce more valuable design improvements than a minimum-compliance regime. Platforms seeking certification would have reason to exceed baseline requirements rather than merely satisfy them.

D.            Collaborative Parental Guidance

Another promising alternative to legislative mandates is for parents to work with their children to implement solutions that respect both parental authority and adolescent development. Platforms would then have incentives to adapt their offerings to the expressed needs of parents and adolescents.

Platform-level parental controls already give parents meaningful authority over younger children’s experiences with smartphones and social media. Opt-in time limits, notification controls, usage dashboards, and content filters can also support constructive conversations with adolescents about healthier use. Meta’s Family Center, TikTok’s private-by-default accounts for minors, and various third-party apps represent steps in this direction. Still, adoption and effectiveness vary, and independent evaluation remains limited.

Highly restrictive parental mediation warrants caution. Restriction alone does not provide the education, social support, and emotional guidance associated with lower rates of habitual smartphone use. Evidence also associates highly restrictive approaches with problematic online behavior and suggests that they may sometimes contribute to it.[126]

E.              Building Skills and Resilience

Beyond platform design and family tools, several interventions show promise for reducing social-media harms by strengthening adolescents’ skills and resilience. A recent review concluded:

Programs that teach adolescents cognitive or behavioral strategies to engage with social media in a healthy manner appear to be more effective at improving long-term well-being than interventions that completely restrict social media use or warn adolescents of its negative effects. Tailoring interventions to the specific contextual factors salient to an individual or group may maximize the effectiveness and long-term impact of reducing [problematic social media use] and improving overall well-being.[127]

School-based media-literacy programs have produced modest improvements in adolescents’ ability to evaluate online content critically, particularly among girls and younger adolescents.[128] Regular, vigorous exercise may also reduce phone use while improving executive function—the skills involved in planning, focus, and impulse control.[129]

Researchers studying social-media use have not adequately answered a more basic question—why adolescents spend so much time on their phones generally and social media specifically. One possibility is that some adolescents lack the confidence, skills, or opportunities to interact in person and therefore default to social media, where interaction may feel less risky.

F.              Restoring In-Person Socialization

As discussed above, reduced attention spans and adolescent mental-health issues often correlate with device use. But that use may partly reflect a deeper problem—inadequate opportunities for in-person socialization. Children develop social competence by interacting with their peers. When they lack opportunities to learn how to behave with other children during crucial periods of neuro-cognitive development, they may enter adolescence less prepared for face-to-face interaction and more likely to default to device-based interactions.

Because screen-mediated interaction doesn’t provide all the rewards of in-person connection, this reliance can become self-reinforcing. Adolescents who feel uncomfortable interacting in person may retreat further into online communication, gaining still less experience with face-to-face relationships.

Many adolescents and young adults now spend less time socializing in person, in part because the environments in which children traditionally developed social competence have eroded. These environments include unstructured play, independent movement through neighborhoods, gathering in public places, and strong community institutions. Time-use data illustrate the shift. Among people ages 15–24, in-person time with friends fell by nearly 70% from 2003 to 2020. U.S. social engagement with friends, family, and others also declined during that period.[130]

Several factors may have contributed:

• The long decline of child-directed free play, through which young people practice making rules, resolving conflicts, regulating emotions, building friendships, and interacting as equals.[131]
• Greater pressure from school, homework, college admissions, and résumé building, which converts time with peers into performance time and is associated with adolescent depression and anxiety.[132]
• Parental and institutional anxiety about safety, which limits children’s independent mobility, especially unsupervised activity outside the home.[133]
• Car-dependent or heavily policed built environments with too few accessible “third places”—public spaces outside the home and school where people can gather. Research links walkable neighborhoods and third places to social capital, sociability, community belonging, and support for young people.[134]

On this account, screen and social-media use expands to fill a void. When adolescents lack freedom to play, physical autonomy, welcoming public spaces, and strong intergenerational community institutions, screen-mediated connection becomes a substitute for missing in-person opportunities rather than the root cause of isolation. Restricting or banning smartphones and social media without restoring the connections they replace could therefore leave adolescents even more isolated.

For the next generation of adolescents, the solution must include rebuilding the skills, culture, institutions, and physical environments that enable children and teenagers to form and sustain healthy human social connection.

VI.          Conclusion

Drawing the preceding analysis together, we assess each major approach against four criteria: evidence that it reduces the targeted harms, the probability and severity of adverse consequences, legal durability in the United States, and its likely net benefits. Table 1 summarizes that assessment.

The approaches with the strongest net-benefit case share three characteristics:

• They target specific mechanisms. They address particular features of devices or apps that may cause harm in particular circumstances rather than restricting access broadly.
• They preserve users’ rights. They require neither universal identity checks nor restrictions on protected speech or state-mandated surveillance of family relationships.
• They can be evaluated and adapted. They allow users and, in some cases, parents to monitor effectiveness and make adjustments. They also encourage app developers, device manufacturers, and platform operators to improve their tools over time.

The approaches with the weakest net-benefit case—minimum-age bans, parental-consent mandates, and mandatory warning labels—share the opposite characteristics. They apply generic rules to diverse users and circumstances. Access mandates intrude directly on privacy and speech, while warning mandates risk habituation and compelled speech. All three are difficult to adapt as evidence changes.

TABLE 1: Net-Benefit Assessment of Major Approaches to Adolescent Device and Social-Media Use

The evidence supports three broad conclusions.

First, the most politically popular interventions have the weakest evidentiary support and the greatest potential to impose costs far in excess of their benefits. Researchers have not shown that age bans, universal age verification, or mandatory time limits improve adolescent outcomes. Their documented or foreseeable costs include privacy risks, circumvention, displacement, disproportionate effects on vulnerable young people, and serious constitutional challenges in the United States.

Second, approaches that empower users, parents, schools, and other organizations are more likely to reduce harm. Independent certification, privacy protections, family tools, and optional friction-based design changes address how people experience platforms rather than determining who may use them. These approaches can adapt as the evidence develops and avoid many of the legal, privacy, and equity costs associated with top-down mandates. They may attract less political attention, but they are more likely to produce useful improvements.

Third, the limited evidence demands humility. Evaluations of most interventions rely on thin evidence of effectiveness. Every approach, whether legislative or market-driven, should include mechanisms for measuring outcomes and changing course as evidence accumulates. Legislative mandates warrant particular caution because they can entrench ineffective interventions, create new harms, and foreclose more promising alternatives. Australia’s social-media ban and China’s gaming curfews provide cautionary examples.

Finally, as noted in Section V, some of the most effective responses may not regulate smartphones or social media directly. They may instead remove barriers that prevent children and adolescents from engaging in unstructured play, independent travel, and in-person community life. If online life has expanded partly because offline life has contracted, restoring those opportunities may do more for adolescent well-being than restricting screens alone.

Epilogue: A Negotiated Design Framework

As this white paper was being completed, Meta and a bipartisan coalition of 51 attorneys general agreed to a proposed consent judgment in the social-media multidistrict litigation. The agreement, which remains subject to court approval, would operate for 10 years without requiring Meta to admit liability or wrongdoing.[135]

The agreement would establish a negotiated, court-enforced compliance framework rather than a generally applicable statutory mandate. It would require Meta to adopt an age-assurance system with quantified accuracy standards. Commercially available methods must keep the agreement-defined error rate at or below 10% for users ages 16–17 and 3% for those ages 13–15. Meta’s proprietary methods would initially face higher limits that tighten after two years.

The agreement would also impose a default two-hour cumulative daily-use limit and restrict access between midnight and 6 a.m. It would disable most nighttime and school-hour notifications, offer teenagers a nonpersonalized chronological feed, hide like and reaction counts by default, prohibit cosmetic-procedure filters, maintain age-appropriate privacy settings, and expand parental-supervision tools. Those tools would notify supervising parents when teenagers repeatedly search for terms related to suicide, self-harm, or eating disorders. An independent auditor would oversee compliance.[136]

It’s possible that these design changes will produce benefits. They are more targeted, adaptable, and open to evaluation than the blunt policies many governments have adopted or proposed. Their actual effects on adolescent behavior and well-being, age-assurance errors, privacy, and circumvention remain uncertain and should be independently evaluated. That uncertainty reinforces this white paper’s central lesson. Interventions should target specific mechanisms, measure real outcomes, and adapt as the evidence develops.

[1] Michael Toscano, Lyman Stone & Grant Bailey, More Screen Time, Less Play Time for America’s Kids (Inst. for Fam. Stud. May 5, 2026), https://ifstudies.org/blog/new-ifs-brief-more-screen-time-less-play-for-americas-kids.

[2] See Int’l Telecomm. Union, Impact of M2M Communications and Non-M2M Mobile Data Applications on Mobile Networks 6, ITU-T Technical Paper QSTP-M2MI (June 15, 2012), https://www.itu.int/dms_pub/itu-t/opb/tut/t-tut-iot-2012-m2m-pdf-e.pdf.

[3] Backlinko Team, Social Media Usage & Growth Statistics (Backlinko Apr. 27, 2026), https://backlinko.com/social-media-users; Simon Kemp, Digital 2026 Mid-Year Global Update Report (DataReportal Apr. 22, 2026), https://datareportal.com/reports/digital-2026-mid-year-global-update-report.

[4] Press Release, Comscore, Inc., Russia Has Most Engaged Social Networking Audience Worldwide (Oct. 20, 2010), https://www.comscore.com/Insights/Press-Releases/2010/10/Russia-Has-Most-Engaged-Social-Networking-Audience-Worldwide; The Nielsen Co., Social Networks/Blogs Now Account for One in Every Four and a Half Minutes Online (June 2010) (reporting that U.S. visitors spent an average of 6 hours, 35 minutes, and 2 seconds per person on social-networking and blog sites in April 2010).

[5] Simon Kemp, Digital 2025: The State of Social Media in 2025 (DataReportal Feb. 5, 2025), https://datareportal.com/reports/digital-2025-sub-section-state-of-social.

[6] Jonathan Rothwell, Teens Spend Average of 4.8 Hours on Social Media Per Day (Gallup Oct. 13, 2023), https://news.gallup.com/poll/512576/teens-spend-average-hours-social-media-per-day.aspx.

[7] Ctrs. for Disease Control & Prevention, Youth Risk Behavior Survey Data Explorer, 1991–2023 High School Youth Risk Behavior Survey Data, https://yrbs-explorer.services.cdc.gov (last visited Sept. 5, 2026) [hereinafter CDC Youth Risk Behavior Survey].

[8] Jonathan Haidt, The Anxious Generation: How the Great Rewiring of Childhood Is Causing an Epidemic of Mental Illness (2024).

[9] CDC Youth Risk Behavior Survey, supra note 7.

[10] Nat’l Ctr. for Educ. Statistics, NAEP Long-Term Trend Assessment Results: Reading and Mathematics (2023), https://www.nationsreportcard.gov/highlights/ltt/2023. Research on device and social-media use finds similarly small negative effects on educational outcomes, which accords with the mental-health literature. See Metin Kuş, A Meta-Analysis of the Impact of Technology-Related Factors on Students’ Academic Performance, 16 Frontiers Psych. 1524645, at 1, 7–8 (2025), https://doi.org/10.3389/fpsyg.2025.1524645. We focus primarily on mental-health effects because they dominate activist and policy concerns.

[11] Jacqueline Pitchforth et al., Mental Health and Well-Being Trends Among Children and Young People in the UK, 1995–2014: Analysis of Repeated Cross-Sectional National Health Surveys, 49 Psychol. Med. 1275, 1275 (2019), https://doi.org/10.1017/S0033291718001757 (analyzing 36 national surveys covering 140,830 respondents ages 4–24 and finding “relatively little change” in measures of psychological distress and emotional well-being). In Scotland, the comparable figure rose from 2.3% to 6.0% over 11 years.

[12] Karen Schrijvers et al., Three Decades of Adolescent Health: Unveiling Global Trends Across 41 Countries in Psychological and Somatic Complaints (1994–2022), 69 Int’l J. Pub. Health 1607774 (2024), https://doi.org/10.3389/ijph.2024.1607774. Among 470,797 15-year-olds, trends were linear in 25% of country-complaint-gender combinations, quadratic in 28.7%, and cubic in 28.7%.

[13] Alina Cosma et al., Cross-National Trends in Adolescents’ Psychological and Somatic Complaints Before and After the Onset of COVID-19 Pandemic, 76 J. Adolescent Health 254 (2025), https://doi.org/10.1016/j.jadohealth.2024.09.028 (studying 792,606 adolescents across 35 countries in four survey waves from 2010 to 2022). The underlying age gradient long predates smartphones. Across 43 countries between 2001 and 2014, adolescent life satisfaction fell by 0.61 standard deviations between ages 10 and 16. Michael Daly, Cross-National and Longitudinal Evidence for a Rapid Decline in Life Satisfaction in Adolescence, 94 J. Adolescence 422 (2022), https://doi.org/10.1002/jad.12037.

[14] Candice L. Odgers, The Great Rewiring: Is Social Media Really Behind an Epidemic of Teenage Mental Illness?, 628 Nature 29, 29 (2024), https://doi.org/10.1038/d41586-024-00902-2 (reviewing Haidt, supra note 8).

[15] Amy Orben & Andrew K. Przybylski, The Association Between Adolescent Well-Being and Digital Technology Use, 3 Nature Hum. Behav. 173, 173 (2019), https://doi.org/10.1038/s41562-018-0506-1 (analyzing data from the U.S. Youth Risk Behavior Survey, Monitoring the Future, and the U.K. Millennium Cohort Study).

[16] Id.

[17] Id.

[18] Jean M. Twenge et al., Specification Curve Analysis Shows That Social Media Use Is Linked to Poor Mental Health, Especially Among Girls, 224 Acta Psychologica 103512, at 8 (2022).

[19] Christopher J. Ferguson et al., There Is No Evidence That Time Spent on Social Media Is Correlated with Adolescent Mental Health Problems: Findings from a Meta-Analysis, 56 Prof. Psychol.: Rsch. & Prac. 73 (2025), https://doi.org/10.1037/pro0000589 (analyzing 79 effect sizes from 46 studies and finding β = .061, 95% CI [.047, .075], and β = .044 for the longitudinal subset). Egger’s regression (p = .021) and a trim-and-fill analysis that identified four missing studies suggested possible publication bias, but the authors provide no bias-adjusted estimate. An earlier meta-analysis by the same group found a pooled r of .052 for screen media generally and no evidence that screen media contributed to suicidal ideation. Christopher J. Ferguson et al., Like This Meta-Analysis: Screen Media and Mental Health, 53 Prof. Psychol.: Rsch. & Prac. 205 (2022), https://doi.org/10.1037/pro0000426.

[20] Luisa Fassi et al., Social Media Use and Internalizing Symptoms in Clinical and Community Adolescent Samples: A Systematic Review and Meta-Analysis, 178 JAMA Pediatrics 814 (2024), https://doi.org/10.1001/jamapediatrics.2024.2078. The review covered 886 effect sizes from 143 studies involving 1,094,890 adolescents. Across clinical and community samples, the pooled estimate for time spent on social media was r = 0.12 (95% CI [0.09, 0.14]) across 56 studies, 117 effect sizes, and 482,273 participants, with I² = 98.0%. Sample type did not significantly moderate the result. Following Jacob Cohen, the authors classify effect sizes below r = 0.10 as small. Jacob Cohen, Statistical Power Analysis for the Behavioral Sciences (2d ed. 1988).

[21] Samantha Teague et al., Digital Media Use and Child Health and Development: A Systematic Review and Meta-Analysis, 180 JAMA Pediatrics 510 (2026), https://doi.org/10.1001/jamapediatrics.2026.0085 (analyzing 1,072 effect sizes from 153 longitudinal studies covering 115 cohorts). For social-media use and depression, the pooled estimate was r = 0.09 (95% CI [0.06, 0.12]). Effects were stronger after one year or less (β = 0.12) than after one to three years (β = 0.04) or more than three years (β = 0.03). For internalizing behaviors, poor-quality studies yielded β = 0.21, significantly larger than the estimate from good-quality studies. Restricting the analysis to fair- and good-quality studies attenuated most associations and rendered the association with academic achievement nonsignificant. The authors nevertheless conclude that “digital media use was consistently associated with risks to child and adolescent health and development, particularly for social media.”

[22] David C. Funder & Daniel J. Ozer, Evaluating Effect Size in Psychological Research: Sense and Nonsense, 2 Advances Methods & Pracs. Psych. Sci. 156 (2019), https://doi.org/10.1177/2515245919847202.

[23] Douglas A. Parry et al., A Systematic Review and Meta-Analysis of Discrepancies Between Logged and Self-Reported Digital Media Use, 5 Nature Hum. Behav. 1535 (2021), https://doi.org/10.1038/s41562-021-01117-5. Across 66 effect sizes from 44 studies involving 52,007 participants, self-reported and logged use correlated at only r = 0.38 (95% CI [0.33, 0.42]). For self-reported problematic-use scales, the correlation fell to r = 0.25 across 40 effect sizes from 19 studies. Only three of 49 comparisons (6.12%) placed mean self-reported use within 5% of mean logged use. The authors caution against drawing broad knowledge claims or policy recommendations from studies that rely solely on self-reported media use.

[24] Id. (calling it “an important unanswered question” whether the discrepancy reflects random or systematic error and citing evidence that inaccuracy varied directly with respondents’ well-being); see Craig J.R. Sewall et al., Objectively Measured Digital Technology Use During the COVID-19 Pandemic, 8 Mobile Media & Commc’n 379 (2020), https://doi.org/10.1177/2050157920902830. On common-method variance, see Orben & Przybylski, supra note 15, at 178 (finding associations with caregiver-reported well-being an order of magnitude smaller than those with adolescent self-reported well-being). On the broader measurement problem, see Linda K. Kaye et al., The Conceptual and Methodological Mayhem of “Screen Time”, 17 Int’l J. Env’t Rsch. & Pub. Health 3661 (2020), https://doi.org/10.3390/ijerph17103661.

[25] Philip M. Podsakoff et al., Common Method Biases in Behavioral Research: A Critical Review of the Literature and Recommended Remedies, 88 J. Applied Psych. 879 (2003), https://doi.org/10.1037/0021-9010.88.5.879.

[26] See, e.g., Jacqueline Nesi & Mitchell J. Prinstein, Using Social Media for Social Comparison and Feedback-Seeking: Gender and Popularity Moderate Associations with Depressive Symptoms, 43 J. Abnormal Child Psychol. 1427 (2015), https://doi.org/10.1007/s10802-015-0020-0. Frequency of technology use was already nonsignificant at the model’s first step, which controlled for baseline depressive symptoms and offline excessive reassurance-seeking. Adding the social-comparison measure did not change that result. Sophia Choukas-Bradley, who develops the appearance-comparison account, likewise acknowledges that the cross-sectional designs dominating the literature “cannot rule out the possibility that adolescents who initially have higher social comparison tendencies are more likely to seek out appearance-related SM content.” Sophia Choukas-Bradley, Adolescents’ Social Media Use and Body Image (MIT Press Works in Progress July 17, 2024), https://wip.mitpress.mit.edu/pub/gik297kw. A recent meta-analysis reported estimates of r = −0.062 (95% CI [−0.163, 0.039]) for ordinary use and sleep duration, r = 0.079 (95% CI [−0.037, 0.194]) for ordinary use and sleep problems, r = 0.038 (95% CI [−0.010, 0.086]) for problematic use and sleep duration, and r = 0.189 for problematic use and sleep problems. Oli Ahmed et al., Social Media Use, Mental Health and Sleep: A Systematic Review with Meta-Analyses, 367 J. Affective Disorders 701 (2024), https://doi.org/10.1016/j.jad.2024.08.193. Heterogeneity was extreme in the ordinary-use analyses (I² = 92.68% and 97.93%), and excluding influential studies reversed the sleep-duration estimate’s sign. Another study found that unhappiness with one’s appearance was associated with 25.89 times the odds of low self-esteem, while at least five hours of daily social-media use was associated with 1.56 times the odds and no use with 1.75 times the odds. Russell M. Viner et al., Roles of Cyberbullying, Sleep, and Physical Activity in Mediating the Effects of Social Media Use on Mental Health and Wellbeing Among Young People in England: A Secondary Analysis of Longitudinal Data, 3 Lancet Child & Adolescent Health 685 (2019), https://doi.org/10.1016/S2352-4642(19)30186-5. Nonusers thus appeared to fare worse than the heaviest users, a result difficult to reconcile with a dose-response account and suggesting that both extremes may proxy for something other than exposure. Appearance dissatisfaction and low self-esteem also come from self-evaluative items in the same instrument, so the 25.89 figure does not offer a like-for-like comparison with behavioral exposure. Finally, the relationship between intense use—measured by time and modeled alongside problematic use—and well-being varied by country. It was negative where intense use was uncommon and positive where it was widespread. Maartje Boer et al., Adolescents’ Intense and Problematic Social Media Use and Their Well-Being in 29 Countries, 66 J. Adolescent Health S89 (2020), https://doi.org/10.1016/j.jadohealth.2020.02.014.

[27] Orben & Przybylski, supra note 15, at 178–79 & tbl.3 (reporting that getting enough sleep and eating breakfast had positive associations with well-being 1.7 to 44.2 times larger than technology use’s negative association); Viner et al., supra note 26, tbl.4 (reporting that physical activity mediated 13.4% of the association among girls in a single-mediator model and 9.0% in the combined model, while acting as a suppressor rather than a mediator among boys at −5.9%).

[28] On the reverse pathway in a second dataset, see Taylor Heffer et al., The Longitudinal Association Between Social-Media Use and Depressive Symptoms Among Adolescents and Young Adults: An Empirical Reply to Twenge et al. (2018), 7 Clinical Psychol. Sci. 462 (2019), https://doi.org/10.1177/2167702618812727.

[29] Heffer et al., supra note 28.

[30] Amy Orben et al., Windows of Developmental Sensitivity to Social Media, 13 Nature Commc’ns 1649 (2022), https://doi.org/10.1038/s41467-022-29296-3. The cross-sectional sample included 84,011 participants. The cross-lagged estimate cited here comes from a random-intercept cross-lagged panel model covering 17,409 Understanding Society participants ages 10–21 across 52,556 measurement occasions. The path from lower-than-expected life satisfaction to higher subsequent social-media use was b = −0.02, 95% CI [−0.03, −0.01], SE = 0.007, p = 0.004, and did not vary by age or sex. The authors characterize the relationship as “bidirectional in nature.” Cross-lagged paths in the other direction were predominantly nonsignificant, and the model’s largest standardized coefficient was −0.13.

[31] Candice L. Odgers & Michaeline R. Jensen, Annual Research Review: Adolescent Mental Health in the Digital Age: Facts, Fears, and Future Directions, 61 J. Child Psychol. & Psychiatry 336 (2020), https://doi.org/10.1111/jcpp.13190.

[32] Orben et al., supra note 30.

[33] Alina Cosma et al., Cross-National Time Trends in Adolescent Mental Well-Being from 2002 to 2018 and the Explanatory Role of Schoolwork Pressure, 66 J. Adolescent Health S50 (2020), https://doi.org/10.1016/j.jadohealth.2020.02.010 (analyzing 915,054 adolescents across 36 countries). The authors report a small linear increase in psychosomatic complaints, partly explained by rising schoolwork pressure in higher-income countries, and conclude that their findings “do not provide evidence for substantial declines in mental well-being.”

[34] Cosma et al., supra note 13.

[35] Isaac Winterburn et al., The Impact of Mental Health-Related Social Media Posts on Young Adults’ Mental Health Reporting: A Two-Study Experimental Investigation (PsyArXiv preprint, Mar. 12, 2026), https://doi.org/10.31234/osf.io/msf9k_v1. In the larger experiment, which included 611 U.K. adults ages 18–24, exposure raised State-Trait Anxiety Inventory scores by 3.17 points on an 80-point scale and broadened respondents’ concept of mental disorder by 0.59 points on a 10-point measure. It did not affect self-diagnosis, functional impairment, coping behavior, or total difficulties. All effect sizes ranged from f² = 0.01 to 0.04. On the underlying hypothesis, see Lucy Foulkes & Jack L. Andrews, Are Mental Health Awareness Efforts Contributing to the Rise in Reported Mental Health Problems? A Call to Test the Prevalence Inflation Hypothesis, 69 New Ideas Psych. 101010 (2023), https://doi.org/10.1016/j.newideapsych.2023.101010.

[36] See Michelle Faverio, Monica Anderson & Eugenie Park, Teens, Social Media and Mental Health (Pew Rsch. Ctr. Apr. 22, 2025), https://www.pewresearch.org/internet/2025/04/22/teens-social-media-and-mental-health.

[37] International studies show similar heterogeneity. See Patti M. Valkenburg et al., Adolescents’ Social Media Experiences and Their Self-Esteem: A Person-Specific Susceptibility Perspective, 2 Tech. Mind & Behav. (2021), https://doi.org/10.1037/tmb0000037. The study collected 21,970 assessments from 300 Dutch adolescents ages 13–16 over three weeks and used dynamic structural-equation modeling to estimate a separate within-person effect for each participant. Across the 15,708 observations reporting valence, 78% of adolescents showed a positive effect of social-media experiences on self-esteem (β ≥ +.05), 19% showed no or a very small effect, and 3% showed a negative effect. For time spent, 56% showed no or a very small effect, 27% showed a negative effect, and 18% showed a positive effect, with person-specific coefficients ranging from β = −.31 to +.27. These were the same adolescents over the same three weeks. The sign turned on which variable researchers measured. In a smaller companion study, 46% of adolescents felt better after browsing, 10% felt worse, and 44% felt no different. Ine Beyens et al., The Effect of Social Media on Well-Being Differs from Adolescent to Adolescent, 10 Sci. Reps. 10763 (2020), https://doi.org/10.1038/s41598-020-67727-7. Those figures cover passive use in a sample of 63 adolescents over one week. Adolescents themselves describe mixed tradeoffs, reporting greater connection and stronger friendships alongside costs to sleep and productivity. See Faverio, Anderson & Park, supra note 36; Am. Acad. of Pediatrics, Social Media and Adolescent Mental Health: Pediatric Mental Health Minute, https://www.aap.org/en/patient-care/mental-health-minute/social-media-and-adolescent-mental-health.

[38] Am. Acad. of Pediatrics, supra note 37.

[39] See Matthew N. Berger et al., Social Media Use and Health and Well-Being of Lesbian, Gay, Bisexual, Transgender, and Queer Youth: Systematic Review, 24 J. Med. Internet Rsch. e38449 (2022), https://doi.org/10.2196/38449.

[40] Odgers & Jensen, supra note 31, at 342. The review also notes that “[a]t present, narrative reviews and meta-analytic work do not support causal claims, or even strong and consistent correlational patterns, linking adolescents’ digital technology usage with mental health problems.” Id. at 339.

[41] Nat’l Telecomm. & Info. Admin., Kids Online Health and Safety Task Force Research Agenda (2024) (identifying this evidence gap and calling for longitudinal and experimental studies to evaluate regulatory effectiveness).

[42] Jonathan Haidt, How the Great Rewiring of Childhood Impacted Everything—and What We Can Do About It, The Anxious Generation, https://www.anxiousgeneration.com/parent-teacher-talk (last visited Aug. 31, 2026).

[43] Orben et al., supra note 30.

[44] NetChoice, LLC v. Griffin, No. 5:23-CV-05105, 2025 WL 978607, at *8 (W.D. Ark. Mar. 31, 2025); NetChoice, LLC v. Murrill, 812 F. Supp. 3d 594, 646 (M.D. La. 2025); NetChoice, LLC v. Yost, 778 F. Supp. 3d 923, 955 (S.D. Ohio 2025), rev’d, 180 F.4th 268 (6th Cir. 2026).

[45] See Ark. Code Ann. § 4-88-1402; La. Stat. Ann. §§ 51:1751–1759; Ohio Rev. Code Ann. § 1349.09.

[46] Griffin, 2025 WL 978607, at *8; Yost, 778 F. Supp. 3d at 955.

[47] Writ of Summons, The Digital Freedom Project Inc. v. Commonwealth, No. S163/2025 (High Ct. Austl. filed Nov. 27, 2025), https://cdn.prod.website-files.com/68d2334bf5fe4689986391b0/6927e145e148f5a16dfc018e_SEALED%20Writ%20of%20Summons%20-%20DFP%20v%20Commonwealth%20of%20Australia.pdf. The challenge invokes Australia’s implied freedom of political communication. The High Court referred the stated questions to the Full Court on July 29, 2026, and the case remains pending. High Ct. of Austl., Case No. S163/2025, https://www.hcourt.gov.au/cases-and-judgments/cases/current/case-no-s1632025 (last visited Sept. 13, 2026).

[48] Online Safety Amendment (Social Media Minimum Age) Act 2024 (Cth) (Austl.).

[49] Reuters, Australia Social Media Ban Hits 4.7 Million Teen Accounts in First Month (NBC News Jan. 16, 2026), https://www.nbcnews.com/world/asia/australia-social-media-ban-hits-47-million-teen-accounts-first-month-rcna254352 (reporting the eSafety Commissioner’s figure). Australian platforms generally use layered age-assurance systems rather than a single uniform check. They infer age from existing account signals, deactivate, freeze, or sign out suspected under-16 accounts, and send disputed cases through more formal verification. The eSafety Commissioner maintains that platforms must take “reasonable steps” to prevent children under 16 from holding accounts. eSafety Comm’r, Social Media Age Restrictions (updated Apr. 2, 2026), https://www.esafety.gov.au/about-us/industry-regulation/social-media-age-restrictions.

[50] eSafety Comm’r, Behind the Screen: The Reality of Age Assurance and Social Media Access for Young Australians (Feb. 2025), https://www.esafety.gov.au/sites/default/files/2025-02/Behind-the-screen-transparency-report-Feb2025.pdf; see also Jasmine Fardouly, Potential Effects of the Social Media Age Ban in Australia for Children Younger Than 16 Years, 7 Lancet Digital Health e235 (2025), https://doi.org/10.1016/j.landig.2025.01.016.

[51] VPN Super, Australia Social Media Ban: VPN Usage Surges 170% on Day One (VPN Observatory Dec. 2025), https://vpnsuper.com/vpn-observatory/australia-social-media-ban-under-16-december-2025 (reporting the provider’s connection data for Australian users connecting to servers in Singapore and the United States).

[52] Aryamehr Fattahi, Australia’s Social Media Ban: A Test for Global Digital Governance (Bloomsbury Intel. & Sec. Inst. Dec. 15, 2025), https://bisi.org.uk/reports/australias-world-first-social-media-ban-a-test-for-global-digital-governance.

[53] Courtney Barnes et al., Assessing Early Effects of Australia’s Social Media Minimum Age Act on Adolescents’ Social Media Use: Observational Study, 393 BMJ e363695 (2026), https://doi.org/10.1136/bmj-2026-363695 (enrolling 436 adolescents ages 12–17 before the Act took effect and following up with 408 three months later).

[54] Shin Ling Wu, Implications of Australia’s Under-16 Social Media Ban, 10 Nature Hum. Behav. 204 (2026), https://doi.org/10.1038/s41562-025-02378-0.

[55] David J. Grüning, Frederik Riedel & Philipp Lorenz-Spreen, Directing Smartphone Use Through the Self-Nudge App One Sec, 120 Proc. Nat’l Acad. Scis. e2213114120 (2023), https://doi.org/10.1073/pnas.2213114120 (reporting a six-week field study of 280 users in which participants abandoned 36% of attempts to open a target app after the pause, attempts fell 37% by week six, and actual app openings fell 57%).

[56] Jay A. Olson et al., A Nudge-Based Intervention to Reduce Problematic Smartphone Use: Randomised Controlled Trial, 21 Int’l J. Mental Health & Addiction 3842 (2023), https://doi.org/10.1007/s11469-022-00826-w.

[57] Cynthia A. Dekker & Susanne E. Baumgartner, Is Life Brighter When Your Phone Is Not? The Efficacy of a Grayscale Smartphone Intervention Addressing Digital Well-Being, 12 Mobile Media & Commc’n 688 (2024), https://doi.org/10.1177/20501579231212062.

[58] Melissa G. Hunt et al., No More FOMO: Limiting Social Media Decreases Loneliness and Depression, 37 J. Soc. & Clinical Psychol. 751 (2018); Christoph Pieh et al., Smartphone Screen Time Reduction Improves Mental Health: A Randomized Controlled Trial, 23 BMC Med. 107 (2025).

[59] Stricter Limits on Minors’ Online Gaming (State Council of the People’s Republic of China Aug. 31, 2021), https://english.www.gov.cn/statecouncil/ministries/202108/31/content_WS612da2a2c6d0df57f98df6bd.html.

[60] Xinyu Zhou et al., High Compliance and Reduced Online Gaming Time Among Chinese Adolescents After the 2021 Gaming Policy: Effects Across All Gamers, Not Just Heavy Users, 14 J. Behav. Addictions 1119 (2025).

[61] Xinyu Zhou et al., Compliance and Alternative Behaviors of Heavy Gamers in Adolescents to Chinese Online Gaming Restriction Policy, 13 J. Behav. Addictions 687 (2024).

[62] David Zendle et al., No Evidence That Chinese Playtime Mandates Reduced Heavy Gaming in One Segment of the Video Games Industry, 7 Nature Hum. Behav. 1753 (2023).

[63] Va. Code Ann. § 59.1-577.1(A)–(B) (2026); NetChoice, LLC v. Jones, 822 F. Supp. 3d 656 (E.D. Va. 2026) (granting a preliminary injunction).

[64] Stop Addictive Feeds Exploitation (SAFE) for Kids Act, N.Y. Gen. Bus. Law §§ 1500–1508 (McKinney 2024); N.Y. Comp. Codes R. & Regs. tit. 13, pt. 700 (2026) (final rules published July 28, 2026, and effective Jan. 25, 2027).

[65] Children’s Social Media Safety Act, H.B. 5511, 104th Gen. Assemb. (Ill. 2026) (signed July 31, 2026).

[66] See Robert Winterton, New York’s Final SAFE for Kids Rules Cement an Unconstitutional Censorship Regime (NetChoice July 29, 2026), https://netchoice.org/new-yorks-final-safe-for-kids-rules-cement-an-unconstitutional-censorship-regime; Amy Bos, NetChoice Veto Request Letter to Gov. Pritzker on Illinois HB 5511, the Children’s Social Media Safety Act (NetChoice June 5, 2026), https://netchoice.org/netchoice-veto-request-letter-to-gov-pritzker-on-illinois-hb-5511-the-childrens-social-media-safety-act.

[67] Louis-Philippe Beland & Richard Murphy, Ill Communication: Technology, Distraction & Student Performance, 41 Labour Econ. 61 (2016) (finding that average test scores rose by 0.07 standard deviations, the equivalent of adding one hour of study per week, with gains twice as large among lower-achieving students).

[68] David Figlio & Umut Özek, The Impact of Cellphone Bans in Schools on Student Outcomes: Evidence from Florida (Annenberg Inst., EdWorkingPaper No. 25-1315, 2025), https://doi.org/10.26300/9y5r-yx11.

[69] Kristof De Witte & Joanna Maldonado, To Ban or Not to Ban? A Rapid Review on the Impact of Smartphone Bans in Schools on Social Well-Being and Academic Performance, 14 Educ. Scis. 906 (2024), https://doi.org/10.3390/educsci14080906.

[70] Victoria A. Goodyear et al., School Phone Policies and Their Association with Mental Wellbeing, Phone Use, and Social Media Use (SMART Schools): A Cross-Sectional Observational Study, 51 Lancet Reg’l Health Eur. 101211 (2025), https://doi.org/10.1016/j.lanepe.2025.101211.

[71] Id.

[72] Samuel J. Perry et al., Health Economics Analysis of Restrictive School Smartphone Policies in Secondary Schools in England (SMART Schools), 29 BMJ Mental Health e301892 (2026), https://doi.org/10.1136/bmjment-2025-301892.

[73] Sara Abrahamsson, Smartphone Bans, Student Outcomes and Mental Health (Norwegian Sch. of Econ., Discussion Paper Series in Economics No. 1/2024, 2024), https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4735240.

[74] In a May 26–June 1, 2026, Pew survey, 85% of U.S. adults supported requiring parental consent before minors could create social-media accounts, compared with 78% who supported age verification and 56% who supported an under-16 social-media ban. Jeffrey Gottfried & Monica Anderson, Majority of Americans Support Banning Social Media for Kids Under 16 (Pew Rsch. Ctr. July 1, 2026), https://www.pewresearch.org/short-reads/2026/07/01/majority-of-americans-support-banning-social-media-for-kids-under-16. As discussed above, federal district courts enjoined parental-consent mandates in Arkansas and Ohio on First Amendment grounds. See supra note 46. The 6th Circuit has since reversed the Ohio ruling. See infra note 124 and accompanying text.

[75] Cal. Health & Safety Code § 28000 et seq. (added by A.B. 56, 2025 Cal. Stat. ch. 671).

[76] Vivek H. Murthy, Opinion, Surgeon General: Why I’m Calling for a Warning Label on Social Media Platforms, N.Y. Times (June 17, 2024), https://www.nytimes.com/2024/06/17/opinion/social-media-health-warning.html; see also Off. of the U.S. Surgeon Gen., Social Media and Youth Mental Health: The U.S. Surgeon General’s Advisory (2023), https://www.hhs.gov/sites/default/files/sg-youth-mental-health-social-media-advisory.pdf.

[77] James L. Hamilton, The Demand for Cigarettes: Advertising, the Health Scare, and the Cigarette Advertising Ban, 54 Rev. Econ. & Stat. 401 (1972); Kenneth E. Warner, The Effects of the Anti-Smoking Campaign on Cigarette Consumption, 67 Am. J. Pub. Health 645 (1977); Kenneth E. Warner, Effects of the Antismoking Campaign: An Update, 79 Am. J. Pub. Health 144 (1989); Eugene M. Lewit, Douglas Coate & Michael Grossman, The Effects of Government Regulation on Teenage Smoking, 24 J.L. & Econ. 545 (1981); Badi H. Baltagi & Dan Levin, Estimating Dynamic Demand for Cigarettes Using Panel Data: The Effects of Bootlegging, Taxation and Advertising Reconsidered, 68 Rev. Econ. & Stat. 148 (1986); Badi H. Baltagi & Dan Levin, Cigarette Taxation: Raising Revenues and Reducing Consumption, 3 Structural Change & Econ. Dynamics 321 (1992); John A. Tauras, Smoke-Free Air Laws, Cigarette Prices, and Adult Cigarette Demand, 44 Econ. Inquiry 333 (2006); Matthew C. Farrelly et al., The Impact of Tobacco Control Programs on Adult Smoking, 98 Am. J. Pub. Health 304 (2008); Melanie A. Wakefield et al., Impact of Tobacco Control Policies and Mass Media Campaigns on Monthly Adult Smoking Prevalence, 98 Am. J. Pub. Health 1443 (2008); Melanie A. Wakefield et al., Time Series Analysis of the Impact of Tobacco Control Policies on Smoking Prevalence Among Australian Adults, 2001–2011, 92 Bull. World Health Org. 413 (2014); Matthew C. Farrelly et al., Are Tobacco Control Policies Effective in Reducing Young Adult Smoking?, 54 J. Adolescent Health 481 (2014); Henry Saffer et al., E-Cigarettes and Adult Smoking: Evidence from Minnesota, 60 J. Risk & Uncertainty 207 (2020); Michael F. Pesko, Charles J. Courtemanche & Joanna Catherine Maclean, The Effects of Traditional Cigarette and E-Cigarette Tax Rates on Adult Tobacco Product Use, 60 J. Risk & Uncertainty 229 (2020); Chad Cotti et al., The Effects of E-Cigarette Taxes on E-Cigarette Prices and Tobacco Product Sales: Evidence from Retail Panel Data, 86 J. Health Econ. 102676 (2022).

[78] S. Ratneshwar et al., Benefit Salience and Consumers’ Selective Attention to Product Features, 14 Int’l J. Rsch. Mktg. 245 (1997); Klaus G. Grunert & Josephine M. Wills, A Review of European Research on Consumer Response to Nutrition Information on Food Labels, 15 J. Pub. Health 385 (2007).

[79] See, e.g., Seth M. Noar et al., Effects of Strengthening Cigarette Pack Warnings on Attention and Message Processing: A Systematic Review, 94 Journalism & Mass Commc’n Q. 416, 416–42 (2017); Claudia Sillero-Rejon et al., Avoidance of Tobacco Health Warnings? An Eye-Tracking Approach, 116 Addiction 1263, 1263–74 (2021); Noel T. Brewer et al., Reactance to Health Warnings Scale: Development and Validation, 50 Annals Behav. Med. 736, 736–50 (2016); Yoo Jin Cho et al., Does Reactance Against Cigarette Warning Labels Matter? Warning Label Responses and Downstream Smoking Cessation Amongst Adult Smokers in Australia, Canada, Mexico and the United States, 11 PLOS ONE e0159245 (2016).

[80] Robert Epstein & Vanessa R. Zankich, The Surprising Power of a Click Requirement: How Click Requirements and Warnings Affect Users’ Willingness to Disclose Personal Information, 17 PLOS ONE e0263097, at 11 (2022).

[81] See Christine Utz et al., (Un)informed Consent: Studying GDPR Consent Notices in the Field, 2019 ACM SIGSAC Conference on Computer & Communications Security 973, 973–90 (2019); Célestin Matte, Nataliia Bielova & Cristiana Santos, Do Cookie Banners Respect My Choice? Measuring Legal Compliance of Banners from IAB Europe’s Transparency and Consent Framework, 2020 IEEE Symposium on Security & Privacy 791, 791–809 (2020); Frank Papenmeier, Josephine Halama & Carl Reichert, Accepting Cookies: Nudging, Deceptive Patterns and Personal Preference, 168 Computers Hum. Behav. 108641 (2025).

[82] See Ian A. Anderson & Wendy Wood, Overestimates of Social Media Addiction Are Common but Costly, 15 Sci. Reps. 39388 (2025); Maria T. Maza et al., Association of Habitual Checking Behaviors on Social Media with Longitudinal Functional Brain Development, 177 JAMA Pediatrics 160, 160–67 (2023); Björn Lindström et al., A Computational Reward Learning Account of Social Media Engagement, 12 Nature Commc’ns 1311 (2021); Katharina Lorenz-Spreen et al., Habitual Social Media and Smartphone Use Are Linked to Task Delay for Some, but Not All, Adolescents, 28 J. Computer-Mediated Commc’n zmad008 (2023).

[83] See Theresa M. Marteau, Gareth J. Hollands & Paul C. Fletcher, Changing Human Behavior to Prevent Disease: The Importance of Targeting Automatic Processes, 337 Science 1492, 1492–95 (2012); Fritz Strack & Roland Deutsch, Reflective and Impulsive Determinants of Social Behavior, 8 Personality & Soc. Psych. Rev. 220, 220–47 (2004); Wendy Wood & David T. Neal, Healthy Through Habit: Interventions for Initiating & Maintaining Health Behavior Change, 2 Behav. Sci. & Pol’y 71, 71–83 (2016).

[84] Eur. Data Prot. Bd., Statement 1/2025 on Age Assurance (Feb. 11, 2025), https://www.edpb.europa.eu/our-work-tools/our-documents/statements/statement-12025-age-assurance_en.

[85] Joint Statement of Security and Privacy Scientists and Researchers on Age Assurance (Mar. 2, 2026), https://csa-scientist-open-letter.org/ageverif-Feb2026.

[86] Lena Cohen, Discord Voluntarily Pushes Mandatory Age Verification Despite Recent Data Breach (Elec. Frontier Found. Feb. 19, 2026), https://www.eff.org/deeplinks/2026/02/discord-voluntarily-pushes-mandatory-age-verification-despite-recent-data-breach (discussing the privacy, security, and anonymity risks posed by age-verification systems).

[87] See Eur. Comm’n, Guidelines on Measures to Ensure a High Level of Privacy, Safety and Security for Minors Online, Pursuant to Article 28(4) of Regulation (EU) 2022/2065, 2025 O.J. (C/5519) ¶¶ 41–42, https://eur-lex.europa.eu/legal-content/EN/TXT/PDF/?uri=OJ:C_202505519; Eur. Comm’n, Annex B—Zero Knowledge Proofs for the Age Verification Solution, EU Age Verification Blueprint (2026).

[88] Nicol Turner Lee, Josie Stewart & Carolina Oxenstierna, How Will Bans on Social Media Affect Children? (Brookings Inst. Dec. 9, 2025), https://www.brookings.edu/articles/how-will-bans-on-social-media-affect-children.

[89] Org. for Econ. Coop. & Dev., Too Young to Scroll? Why Governments Are Cracking Down on Social Media Age Limits (June 2025), https://www.oecd.org/en/blogs/2025/06/too-young-to-scroll-why-governments-are-cracking-down-on-social-media-age-limits.html [hereinafter OECD].

[90] Paresh Dave & David Pierson, There’s One Part of Snapchat That Facebook Can’t Copy: CEO Evan Spiegel, L.A. Times (Mar. 1, 2017), https://www.latimes.com/business/technology/la-fi-tn-evan-spiegel-20170228-htmlstory.html.

[91] Jenna Wortham, A Growing App Lets You See It, Then You Don’t, N.Y. Times (Feb. 8, 2013), https://www.nytimes.com/2013/02/09/technology/snapchat-a-growing-app-lets-you-see-it-then-you-dont.html.

[92] Id.

[93] Celia B. Fisher, Xiangyu Tao & Madeline Ford, Social Media: A Double-Edged Sword for LGBTQ+ Youth, 156 Computers Hum. Behav. 108194 (2024).

[94] Carolyn Bacaj et al., Review of Current Trends in LGBTQ+ Youth and Social Media: Implications for Mental Health, Identity Development, and Civic Engagement, 13 Current Pediatrics Reps. 3 (2025), https://doi.org/10.1007/s40124-024-00338-2.

[95] Berger et al., supra note 39 (reviewing 26 studies).

[96] Id.

[97] LGBT Tech, Coalition Statement in Opposition to Age-Minimum Social Media Bans (Mar. 24, 2026), https://www.lgbttech.org/post/lgbt-tech-leads-coalition-statement-in-opposition-to-age-minimum-social-media-bans.

[98] See Turner Lee, Stewart & Oxenstierna, supra note 88 (documenting the disproportionate burdens that age-based social-media bans impose on LGBTQ+, low-income, and other marginalized young people).

[99] Elizabeth McGhee Hassrick et al., Benefits and Risks: A Systematic Review of Information and Communication Technology Use by Autistic People, 3 Autism Adulthood 72, 72–84 (2021); Michal Alon-Tirosh & Naama Meir, Use of Social Network Sites Among Adolescents with Autism Spectrum Disorder: A Qualitative Study, 14 Frontiers Psychol. 1192475 (2023); Mitzi J. Ritzman & Roma Subramanian, Voices from a Pandemic: Understanding How Young Adults on the Autism Spectrum Use Computer-Mediated Communication, 28 Autism 381, 381–89 (2024); Catherine J. Crompton et al., Autistic Peer-to-Peer Information Transfer Is Highly Effective, 24 Autism 1704, 1704–12 (2020); Callie M. Ginapp et al., The Experiences of Adults with ADHD in Interpersonal Relationships and Online Communities: A Qualitative Study, 3 SSM—Qualitative Rsch. Health 100223 (2023); Sihan Jian, Saptarshi Bhowmik & Bret Staudt Willet, Coping with ADHD in College: Topic Modeling and Sentiment Analysis of Online Peer Discussions on Reddit (Int’l Educ. Data Mining Soc’y, Poster Paper No. 267, 2025).

[100] Neil Thurman et al., Forbidden Fruit or Soured Grapes? Long-Term Effects of the Temporary Unavailability and Rationing of US News Websites on Their Consumption from the European Union, 84 Int’l Commc’n Gazette 698 (2022), https://doi.org/10.1177/17480485221111312.

[101] See OECD, supra note 89 (noting that disproportionate compliance burdens may entrench incumbent platforms).

[102] Samuel G. Goldberg, Garrett A. Johnson & Scott K. Shriver, Regulating Privacy Online: An Economic Evaluation of the GDPR, 16 Am. Econ. J.: Econ. Pol’y 325, 327–28, 354 (2024).

[103] Jack W. Brehm, A Theory of Psychological Reactance (1966); Brad J. Bushman & Angela D. Stack, Forbidden Fruit Versus Tainted Fruit: Effects of Warning Labels on Attraction to Television Violence, 2 J. Experimental Psych.: Applied 207 (1996), https://doi.org/10.1037/1076-898X.2.3.207; Kate A. Varava & Brian L. Quick, Adolescents and Movie Ratings: Is Psychological Reactance a Theoretical Explanation for the Forbidden Fruit Effect?, 59 J. Broad. & Elec. Media 149 (2015), https://doi.org/10.1080/08838151.2014.998224.

[104] Stephen Worchel, Jerry Lee & Akanbi Adewole, Effects of Supply and Demand on Ratings of Object Value, 32 J. Personality & Soc. Psych. 906 (1975), https://doi.org/10.1037/0022-3514.32.5.906.

[105] See, e.g., Williamson v. Lee Optical Co., 348 U.S. 483, 487–88 (1955); FCC v. Beach Commc’ns, Inc., 508 U.S. 307, 313–15 (1993) (explaining that a classification subject to rational-basis review “must be upheld . . . if there is any reasonably conceivable state of facts that could provide a rational basis” for it).

[106] See United States v. O’Brien, 391 U.S. 367, 377 (1968); Ward v. Rock Against Racism, 491 U.S. 781, 798–99 (1989) (explaining that intermediate scrutiny does not require “the least restrictive or least intrusive means” of advancing the government’s interest).

[107] See Reed v. Town of Gilbert, 576 U.S. 155, 163 (2015) (holding that content-based laws are “presumptively unconstitutional”); United States v. Playboy Ent. Grp., Inc., 529 U.S. 803, 813 (2000) (“If a less restrictive alternative would serve the Government’s purpose, the legislature must use that alternative.”).

[108] Gerald Gunther, The Supreme Court, 1971 Term—Foreword: In Search of Evolving Doctrine on a Changing Court: A Model for a Newer Equal Protection, 86 Harv. L. Rev. 1, 8 (1972).

[109] Adam Winkler, Fatal in Theory and Strict in Fact: An Empirical Analysis of Strict Scrutiny in the Federal Courts, 59 Vand. L. Rev. 793 (2006) (finding that laws subject to strict scrutiny in published federal decisions from 1990 through 2003 survived about 30% of applications overall and 22% of free-speech challenges).

[110] Free Speech Coal., Inc. v. Paxton, 606 U.S. 461, 484 (2025).

[111] Sable Commc’ns of Cal., Inc. v. FCC, 492 U.S. 115, 126 (1989) (“[T]here is a compelling interest in protecting the physical and psychological well-being of minors.”); see also Free Speech Coalition, 606 U.S. at 496 (“Texas’s interest in shielding children from sexual content is important, even ‘compelling.’”).

[112] Brown v. Ent. Merchs. Ass’n, 564 U.S. 786, 794–95 (2011) (quoting Erznoznik v. City of Jacksonville, 422 U.S. 205, 212–13 (1975)).

[113] Ginsberg v. New York, 390 U.S. 629 (1968).

[114] Free Speech Coalition, 606 U.S. at 461.

[115] See id. at 478 (holding that Texas H.B. 1181 “is an exercise of Texas’s traditional power to prevent minors from accessing speech that is obscene from their perspective” and only incidentally affects speech protected for adults, so intermediate scrutiny applies).

[116] Brown, 564 U.S. at 800–01.

[117] Ashcroft v. ACLU, 542 U.S. 656 (2004).

[118] See id. at 666–70.

[119] Brown, 564 U.S. at 804 (emphasis in original).

[120] NetChoice, LLC v. Bonta, 113 F.4th 1101, 1118 (9th Cir. 2024) (preliminarily enjoining the California Age-Appropriate Design Code Act’s data-protection-impact-assessment provisions).

[121] Griffin, 2025 WL 978607 (permanently enjoining Arkansas’ Social Media Safety Act); Yost, 778 F. Supp. 3d 923, rev’d, 180 F.4th 268 (6th Cir. 2026); Murrill, 812 F. Supp. 3d 594 (permanently enjoining Louisiana Act 456); NetChoice, LLC v. Reyes, 748 F. Supp. 3d 1105 (D. Utah 2024) (preliminarily enjoining Utah’s law requiring platforms to disable certain social-media functions for minors); Jones, 822 F. Supp. 3d 656 (preliminarily enjoining Virginia’s social-media-time-limit law); NetChoice, LLC v. Carr, 789 F. Supp. 3d 1200 (N.D. Ga. 2025) (preliminarily enjoining Georgia’s social-media age-verification and parental-consent law).

[122] NetChoice, LLC v. Fitch, 145 S. Ct. 2658, 2658 (2025) (Kavanaugh, J., concurring in the denial of an application to vacate a stay) (noting that seven federal district courts had enjoined similar laws and that Mississippi’s statute is “likely unconstitutional” under the Court’s precedents).

[123] Comput. & Commc’ns Indus. Ass’n v. Uthmeier, No. 25-11881, 2025 WL 3458571 (11th Cir. Nov. 25, 2025) (staying the preliminary injunction against Florida H.B. 3 pending appeal).

[124] Yost, 180 F.4th 268 (holding that NetChoice lacked third-party standing to assert minors’ First Amendment rights and that the parental-consent requirement survived a facial challenge); but see NetChoice, LLC v. Skrmetti, No. 25-5660, 2026 WL 2546423 (6th Cir. Aug. 28, 2026) (vacating the denial of a preliminary injunction against Tennessee’s social-media age-verification and parental-consent law and remanding for consideration of NetChoice’s likelihood of success on the merits).

[125] Julian Morris, Consumer Protection in the 21st Century (Int’l Ctr. for L. & Econ. Feb. 24, 2023), https://laweconcenter.org/resources/consumer-protection-in-the-21st-century.

[126]  Hagit Sasson & Gustavo Mesch, Parental Mediation, Peer Norms and Risky Online Behavior Among Adolescents, 33 Comput. Hum. Behav. 32, 32–38 (2014), https://doi.org/10.1016/j.chb.2013.12.025.

[127] Jason M. Nagata et al., Problematic Social Media Use Interventions for Mental Health Outcomes in Adolescents, 27 Current Psychiatry Reps. 491 (2025), https://doi.org/10.1007/s11920-025-01619-3.

[128] Chloe S. Gordon et al., Outcomes of a Cluster Randomized Controlled Trial of the SoMe Social Media Literacy Program for Improving Body Image-Related Outcomes in Adolescent Boys and Girls, 13 Nutrients 3825 (2021), https://doi.org/10.3390/nu13113825.

[129] Amanda C.Y. Tang, C.P. Wong & Louisa H.T. Louie, Effectiveness of Exercise in Reducing Mobile Phone Addiction in Adolescents: A Systematic Review and Meta-Analysis, 14 Frontiers Psychol. 1294116 (2023); Yun Li et al., Intervention Measures for Mobile Phone Addiction in Teenagers: A Network Meta-Analysis, 16 Frontiers Psychiatry 1629251 (2025).

[130] Off. of the U.S. Surgeon Gen., Our Epidemic of Loneliness and Isolation: The U.S. Surgeon General’s Advisory on the Healing Effects of Social Connection and Community 13 (2023); Viji Diane Kannan & Peter J. Veazie, U.S. Trends in Social Isolation, Social Engagement, and Companionship—Nationally and by Age, Sex, Race/Ethnicity, Family Income, and Work Hours, 2003–2020, 21 SSM—Population Health 101331 (2023).

[131] Peter Gray, The Decline of Play and the Rise of Psychopathology in Children and Adolescents, 3 Am. J. Play 443, 443–44, 458–59 (2011).

[132] Thomas Steare et al., The Association Between Academic Pressure and Adolescent Mental Health Problems: A Systematic Review, 339 J. Affective Disorders 302, 302–17 (2023).

[133] Sarah Foster et al., The Impact of Parents’ Fear of Strangers and Perceptions of Informal Social Control on Children’s Independent Mobility, 26 Health & Place 60, 60–68 (2014).

[134] Kevin M. Leyden, Social Capital and the Built Environment: The Importance of Walkable Neighborhoods, 93 Am. J. Pub. Health 1546, 1546–51 (2003); Danielle M. Littman, Third Places, Social Capital, and Sense of Community as Mechanisms of Adaptive Responding for Young People Who Experience Social Marginalization, 69 Am. J. Community Psych. 436, 436–50 (2022).

[135] In re Social Media Adolescent Addiction/Personal Injury Prods. Liab. Litig., MDL No. 3047, No. 4:22-md-03047-YGR (N.D. Cal.).

[136] Id.