Spotlight

Introduction

Judge Amit Mehta’s decision in the Google Search case[1] is commendable in many respects. He seems to strive to credit counterarguments wherever doing so is sensible, rather than trying to “bullet-proof” his opinion (as other “Big Tech”-related decisions often do) by discounting every argument put forward by Google. He is not gratuitously dismissive of Google’s experts. And, as he did in dismissing the entirety of the states’ “self-preferencing” claims at summary judgment,[2] he is willing to reject various of the plaintiffs’ arguments here—finding, for example, that some of their proposed relevant markets were not relevant markets and/or that Google was not a monopolist in them. The decision is also very clearly written and, at 275 pages, admirably thorough.

That’s the good. Unfortunately, where it counts the most, Judge Mehta’s decision is seriously lacking, to the point that his primary legal conclusion—that Google’s default search distribution deals were anticompetitive—is untenable. In this paper, I explain why.

I. The Core Legal Defect: Misapplication of the Causation Standard for Exclusionary Conduct

A. Misreading Microsoft

I start with a quote from the decision that brings us directly to the heart of the matter. Quoting the D.C. Circuit Court of Appeals’ decision in Microsoft,[3] Judge Mehta holds that, as a matter of law:

[C]ausation does not require but-for proof. The plaintiff is not required to show that but for the defendant’s exclusionary conduct the anticompetitive effects would not have followed. Such a standard would create substantial proof problems, as “neither plaintiffs nor the court can confidently reconstruct… a world absent the defendant’s exclusionary conduct.” “To some degree, ‘the defendant is made to suffer the uncertain consequences of its own undesirable conduct.’”[4]

But, as Judge Ginsburg—widely regarded to be the primary author of the per curiam Microsoft opinion—has written, that is a misreading of Microsoft.[5] In fact, this decision—this case—is the paradigmatic example of why that is a misreading of Microsoft.

As a general matter, the plaintiff must establish that the defendant’s conduct has the “requisite anticompetitive effect”[6]—that is, that it caused the alleged competitive harm.[7] The “reasonably capable of” standard has “limited applicability”[8] and permits an inference of causation only in special circumstances: 1) when the competitive threat allegedly affected by the defendant’s conduct is nascent; 2) when that conduct was already proven to have anticompetitive effect; and 3) in a government (as opposed to private) enforcement action.[9] “Only when these conditions are met may the government avoid having to show that the threat would have become a real competitor but for the alleged exclusionary conduct.”[10]

The intuition behind this is as follows: There might be multiple reasons—including ones not involving the defendant’s exclusionary conduct—that prevent a competitive threat from materializing. Normally, a plaintiff has to show that the defendant’s conduct—and not one or more of these other factors (like, e.g., that the rival was of such low quality that it couldn’t realistically have mounted a real challenge)—was the but-for cause of the challenger’s failure. But where that threat is inchoate, proving the hypothetical course of future competition in the market is effectively impossible. Thus, in circumstances where we’re confident that lessening the plaintiff’s burden won’t systematically lead to erroneous outcomes, we allow it to make out its prima facie case without demonstrating but-for causation:

The court pointed out that “neither plaintiffs nor the court can confidently reconstruct a product’s hypothetical technological development in a world absent the defendant’s exclusionary conduct.” Given this “underlying proof problem,” the Court may infer causation.[11]

According to Judge Ginsburg’s clarification of what the Microsoft court held, the limited circumstances that satisfy this standard are: 1) as noted, the threat is of the sort that can’t actually be proven—that is, the allegedly thwarted competition must arise from a speculative, but realistic, process that can’t be falsified; 2) the defendant’s conduct must be proven to be, in fact, sufficient to impede the materialization of such a threat (“[o]f critical importance is that the court’s causation standard was conditioned on its having found anticompetitive effects”[12]); and 3) the plaintiff must be a government enforcer.[13]

Obviously, the Google Search case involves a government enforcement action. But for Microsoft’s lighter causation standard to apply, it must also involve a nascent threat and conduct proven sufficient to prevent rivals from achieving minimum efficient scale. Arguably, neither is true in Google Search.

The holding in Microsoft that causation need not be perfectly proven was a function of those specialized facts. It was not, contra Judge Mehta’s approach to this case (and that of many others before him), “a matter of general tendency.”[14]

1. Causation may be inferred only when the competitive process is speculative

The key question in Microsoft was the foreclosure of competition by a nascent competitor—where the entire theory of the case was built on a set of suppositions about the progress of technology, speculation about the unpredictable role the nascent competitor could play in disrupting competition in an established market, and uncertainty about whether Microsoft viewed it (Netscape Navigator) as a competitive threat.

That is not really the case in Google Search. Bing, obviously, and Yahoo! and others before and after it are/were not nascent competitors. Nor is Bing bringing an innovative disruption to the market that will follow an unknown course. Rather, it is a direct, close substitute for Google Search. And, of course, this is obvious to Google. It is distributed the same way; it is used the same way; it is not unknown or uncertain in its competitive relationship with Google Search. None of which means it was actually a competitive threat to Google (more on this later). But it does mean that the competitive process is well understood.

Why does this matter? Because it means that much less speculation is required about whether and how Bing could act as a competitive constraint on Google. We do have actual competition and consumer behavior to assess in understanding the extent of its competitive threat and whether Google’s challenged conduct impaired it. And we do not have to create an entirely hypothetical world in which establishing causation would be impossible.

None of that was true in Microsoft.

We know this, in part, because of another case the same court decided a few years after Microsoft: Rambus v. FTC.[15] With respect to the speculative extent of the but-for world and the viability of demonstrating causation, the facts in Rambus are more similar to the facts here than are the facts in Microsoft. In Rambus (which involved the selection by a standard-setting organization (SSO) of technology to be included in an industry standard) competition was direct and well-understood. The sole question was whether Rambus’s conduct (deception over its patent holdings, which led to the inclusion of its technology in the industry standard) enabled its technologies to monopolize the relevant markets to the exclusion of its rivals, or whether, had it disclosed and the SSO obtained assurances it would license its technology on RAND terms, it would still have obtained its dominant market position.

Of course, in Rambus—as everywhere—it was impossible to truly know the but-for world (i.e., what Rambus’s market position and that of its competitors would have looked like under different licensing terms). But the competitive process by which such an outcome could arise was straightforward, and its competitive alternatives were known. The same could not really be said of Microsoft.

B. Misreading Rambus

In Rambus, the question was whether Rambus’s behavior led to the exclusion of rivals, or whether, even absent its behavior, rivals would have been excluded. That is the identical question that should have been asked here: Was Bing’s failure to gain more market share proved to be a function of Google’s distribution deals, or was it a function of consumer and distributor preferences for Google over its rivals?

As the court in Rambus found, the FTC’s reasoning (from which the court heard the appeal) was logically flawed, in exactly the same way the court’s reasoning is flawed in this case. It is therefore odd that Judge Mehta distinguishes Rambus and rejects the applicability of its legal standard on grounds that are superficial and unrelated to the relevant question of when a given legal standard should apply.

The consequence of this decision for the Google Search case was enormous. By relieving the plaintiffs of having to show but-for causation, Judge Mehta relieved them of the burden of proving their case. This is exactly why Rambus is so important.

1. What Rambus Says

So, here’s what the court said in Rambus. First, the court laid out the standard:

The critical question is whether Rambus engaged in exclusionary conduct, and thereby acquired its monopoly power in the relevant markets unlawfully.

To answer that question, we adhere to two antitrust principles that guided us in Microsoft. First, “to be condemned as exclusionary, a monopolist’s act must have ‘anticompetitive effect.’ That is, it must harm the competitive process and thereby harm consumers. In contrast, harm to one or more competitors will not suffice.” Microsoft, 253 F.3d at 58…. Second, it is the antitrust plaintiff—including the Government as plaintiff—that bears the burden of proving the anticompetitive effect of the monopolist’s conduct. Microsoft, 253 F.3d at 58-59.[16]

Applying these principles, however, the implication of the FTC’s argument was ambiguous:

The Commission’s conclusion that Rambus’s conduct was exclusionary depends, therefore, on a syllogism: Rambus avoided one of two outcomes by not disclosing its patent interests; the avoidance of either of those outcomes was anticompetitive; therefore Rambus’s non-disclosure was anticompetitive.[17]

The Rambus court acknowledged that the first of these possible outcomes would be anticompetitive,[18] but on the second it found the causal link unclear.[19] And because this alternative was not inherently anticompetitive, the Rambus court rejected the FTC’s argument based on its failure to prove that Rambus’s conduct, and not simply its inclusion in the standard (even on less-favorable terms), led to its market position:

Here, the Commission expressly left open the likelihood that [the SSO] would have standardized Rambus’s technologies even if Rambus had disclosed its intellectual property. Under this hypothesis, [the SSO] lost only an opportunity to secure a RAND commitment from Rambus. But loss of such a commitment is not a harm to competition from alternative technologies in the relevant markets….

…Thus, if [the SSO], in the world that would have existed but for Rambus’s deception, would have standardized the very same technologies, Rambus’s alleged deception cannot be said to have had an effect on competition in violation of the antitrust laws.[20]

If the court was correct that only one of the two outcomes was anticompetitive, then its conclusion was inescapable. As Josh Wright has long (since 2009) maintained, “the D.C. Circuit’s causation standard [in Rambus] should not be controversial and appears eminently reasonable.”[21]

Both the Commission and the D.C. Circuit accept that there must be a causal showing that deception significantly contributes to some anticompetitive effect. The disagreement is over whether both possible paths actually involve anticompetitive effects. If one agrees with the Commission that both causal paths violate Section 2, a requirement that a plaintiff specify precisely which path resulted in an anticompetitive effect is unnecessary and likely unwise. However, if one believes that only one causal path constitutes a violation of Section 2, such a requirement is necessary….[22]

Importantly, the Rambus court held this despite recognizing that Rambus’s deception made the inclusion of its technology in the standard “somewhat more likely.”[23] That wasn’t enough, because the FTC failed to show that that wouldn’t have happened even without—that is, but for—Rambus’s deception. “The critical point is that the Commission bore the burden of demonstrating that Rambus’s deception caused the unlawful acquisition of monopoly power.”[24]

C. Misconstruing the Legal Standard Under Microsoft and Rambus

Here, we can rewrite Rambus’s conclusion using the facts of the Google Search case and readily see its applicability:

Thus, if [distributors like Apple and Mozilla], in the world that would have existed but for [Google’s default distribution deals], would have [chosen] the very same [default search provider], [Google’s conduct] cannot be said to have had an effect on competition in violation of the antitrust laws.[25]

This is not how Judge Mehta assesses the Rambus decision, however. Instead, he distinguishes it on the tenuous grounds that it involved a different type of exclusionary conduct, in a different factual setting, and that, as a result, it is inconsistent with his reading of Microsoft:

Rambus does not establish a categorical rule that the anticompetitive effects of an exclusive agreement must be measured against a but-for world. That case involved deception to a standards-setting organization, a form of exclusionary conduct particularly susceptible to a finding of materiality…. In such circumstances, the D.C. Circuit deemed it appropriate to demand proof that Rambus’s deception in fact resulted in competitive harm. Nowhere, however, did the court suggest that such a strict standard of proof was required to demonstrate anticompetitive effects for other forms of exclusionary conduct, particularly exclusive dealing arrangements. Such a holding would be contrary to Microsoft, and the court in Rambus nowhere questioned that precedent. Rambus therefore does not require Plaintiffs to prove substantial foreclosure against a but-for world.[26]

But “deception to a standards-setting organization” is not a circumstance “particularly susceptible” to a but-for analysis compared to that of default search distribution deals. Indeed, it has more in common with the circumstances here than Microsoft does. That’s because, as noted, the but-for world in that case is easy to understand and analyze (even if, as in all cases, the but-for world is never a simple or certain calculation). It is in the Google Search case, as well. The but-for world in this case, as in Rambus, involves an essentially binary choice among known alternatives with a direct line between the relevant conduct and that choice. In Rambus, it was a choice by an SSO between two licensing regimes (one less restrictive, and one more restrictive) for Rambus’s patents in its standards, and Rambus’s deception could clearly affect that choice. Here, it is a choice by users between (essentially) two search engines, and the choice of default search provider can clearly affect that choice (because the cost of using the non-default is inherently higher).

Moreover (and as noted above), according to Judge Ginsburg, Rambus is not a special exception to Microsoft’s general rule; Microsoft is a special exception to Rambus. It is the nascency of the threat in Microsoft (which didn’t exist in Rambus) that leads to its uniquely truncated analysis and not the specific context of Rambus that somehow cabins its applicability:

Reading Microsoft and Rambus together, the key takeaway is that only when anticompetitive effects are shown (as required by Microsoft and Rambus) does the “reasonably capable of” causation standard apply to allegations that exclusionary conduct killed a nascent threat. Only when these conditions are met may the government avoid having to show that the threat would have become a real competitor but for the alleged exclusionary conduct.[27]

Nevertheless, misconstruing the legal standard under Microsoft and Rambus, Judge Mehta holds that Microsoft’s lighter, “reasonably capable of” standard applies:

The key question then is this: Do Google’s exclusive distribution contracts reasonably appear capable of significantly contributing to maintaining Google’s monopoly power in the general search services market? The answer is “yes.” Google’s distribution agreements are exclusionary contracts that violate Section 2 because they ensure that half of all GSE [general search engine] users in the United States will receive Google as the preloaded default on all Apple and Android devices, as well as cause additional anticompetitive harm.[28]

D. How We Know It Is Wrong

It is immediately obvious why, even if it weren’t a misreading of the case law, this cannot possibly be the correct standard—and why it makes no sense to suggest that a less strict standard of proof is particularly appropriate for exclusive-dealing arrangements.

We don’t need to ask if the agreements “reasonably appear capable of significantly contributing to maintaining Google’s monopoly power” because of course they are “reasonably… capable” of contributing to Google’s monopoly power. That’s why we have scores of antitrust cases looking at the effects of distribution agreements. If all that were required to win such a case were the reasonable capability of an agreement to contribute to a dominant firm’s competitive position, then no exclusive or quasi-exclusive agreement would ever be legal.[29]

We ask whether exclusive agreements “reasonably appear capable” of maintaining monopoly, instead of asking whether they actually maintained monopoly, only when the connection between what is being excluded and monopoly maintenance is unclear—as in Microsoft, where it was unclear if Netscape Navigator could actually constitute a competitive threat to Microsoft’s operating-system dominance. But here that is not a question.[30] Where the rival is a direct competitor with a close substitute product, an exclusive deal by a dominant incumbent is always capable of foreclosing the rival. In such circumstances it is simply not consistent with the plaintiff’s burden of proof to allow them to show only that the challenged conduct is the sort that could maintain monopoly, rather than that the defendant’s conduct in fact caused anticompetitive harm.

Difficult as it may be, demonstrating this in an actual competition case like Google Search is not impossible. Indeed, as I discuss below, there is copious evidence in the case that the cause of Bing’s limited market share was not Google’s default distribution deals, but Bing’s lack of quality. Any comparable evidence in Microsoft would have been wholly speculative—but not here.

As Judge Ginsburg notes (again challenging the general applicability of Microsoft’s truncated legal standard):

As in Microsoft, the “but-for” world in Rambus was highly uncertain.[[31]] In both cases, one could reasonably find the defendant’s conduct may have caused the defendant to acquire or maintain its monopoly power. At the same time, it was also possible that the defendants in those cases would have acquired or maintained their monopoly power even absent their anticompetitive behavior. The court in Rambus held the government must bear the burden of that uncertainty. This burden applies in all Section 2 cases….[32]

But it is also arguably the case that, properly construed, Google’s default distribution deals were not capable of excluding Bing from the market. We turn to this next.

II. The Failure To Prove That Defaults Are Exclusive

Judge Mehta holds that Google’s distribution deals had anticompetitive effects “because they ensure that half of all [general search engine] users in the United States will receive Google as the preloaded default on all Apple and Android devices.”[33] He derives this conclusion (which he repeats several times) from the testimony of one of the plaintiffs’ economic experts, Michael Whinston, who finds that “50% of all queries in the United States are run through the default search access points covered by the challenged distribution agreements.”[34]

Judge Mehta’s claim is that, because users don’t switch away from defaults very often, and because the “market realities” of the search market—given Google’s default distribution deals—are that half of all relevant searches occur through access points covered by those deals, we can conclude that those deals foreclose competitors from 50% of the general search market (a big enough amount to constitute anticompetitive foreclosure).

But this is not how you measure foreclosure, and the assertion that that number shows that Google’s default deals—and not something else—“significantly contribut[e] to maintaining Google’s monopoly power”[35] is fallacious.

Just because the government shows there is “significant” usage of Google’s default services ex post—meaning, given Google’s default deals, and after consumers have chosen which search engine to use—does not mean it has proven that 50% of the market was foreclosed from access by competitors. Nor does it mean that the government has met its burden of proving that this was caused by the agreements. Perhaps all of those consumers are inframarginal consumers who would have chosen Google Search anyway, even if it weren’t the default. Perhaps all of them were perfectly capable of accessing Bing, but simply chose not to. In that case, the ex-post usage data would tell us nothing about the extent of foreclosure.

Demonstrating foreclosure requires comparison to the but-for world; it requires showing that, absent Google’s deals, Bing would have had access to and been used by substantially more marginal consumers (those who view Bing and Google as effective substitutes and wouldn’t expend extra cost to use one search engine or the other).[36] This is so for three reasons.

First, these agreements are not, in fact, “exclusive.” That matters, because it is much harder to infer that it was the agreements, and not consumer preferences for a particular product, that caused Google’s dominance when consumers have ample opportunity to exercise their preferences to not use Google Search.

Second, and relatedly, looking at the share of searches ex post that go through these defaults is less telling, and the number can’t simply be accepted at face value, when searching via the default service is not the only option. As in Rambus, Google’s maintenance of a large share of these searches is just as consistent with a non-problematic set of facts (consumers simply prefer Google Search and, knowing that, distributors offer Google as the default) as with a problematic one (consumers use Google Search only because it is the default service, and distributors offer Google Search as the default only because Google pays them to do so).

Finally, failing to demand that the plaintiffs demonstrate that Google’s default distribution deals actually foreclosed competitors, and allowing them effectively to prove their case by showing only that the deals made Google’s large market share “somewhat more likely,”[37] erases the requirement that plaintiffs can win only if they prove that challenged conduct causes anticompetitive harm. This is an invitation for dramatically erroneous decisions.[38] As Judge Ginsburg and Koren Wong-Ervin write:

Without requiring proof of but-for causation, there is great risk of erroneously condemning [conduct] that may be procompetitive. Consider, for example, Herbert Hovenkamp’s proposal presumptively to condemn acquisitions by a monopolist of “any firm that has the economic capabilities for entry and is a more-than-fanciful possible entrant, unless the acquired firm is no different from many other firms in these respects.” “More-than-fanciful” is an invitation to speculate, not a standard of proof.[39]

Let’s examine these problems with Judge Mehta’s legal standard based on Michael Whinston’s market-share numbers a little more thoroughly.

A. Establishing the Amount of Competition: Minimum Efficient Scale

So much of Judge Mehta’s conclusion that Google’s default deals were exclusive (despite not actually being “exclusive”) turns on his contention that defaults are the best way for search engines to distribute themselves,[40] and, therefore, that Google tying up default access was sufficient to establish the requisite exclusivity for an exclusive-dealing claim.

But as the Rambus court points out, “conduct [does] not violate antitrust laws where absent that conduct consumers would still receive the same product and the same amount of competition.”[41] This is a statement that the but-for world matters, and the relevant question is the relative “amount of competition.”

So, how do we measure the “amount of competition”? Well, we can’t measure it as Judge Mehta and Prof. Whinston do, by looking at what people choose ex post.[42] This is a fairly useless statistic. It says next to nothing about what share of marginal searchers use Google because it is the default, and what share use Google because they prefer it. (And, of course, it says nothing about what share are inframarginal consumers who would use Google regardless of the cost of accessing it). That, in turn, tells you nothing about the amount of competition that existed before they made their choices.[43]

Nor can we simply look at the scope of the default distribution agreements. If people can still easily choose other search providers despite the default deals, those deals cannot be said to foreclose competition—at the very least, not in proportion to the share of the market covered by those deals.

We can, however, examine whether rival search providers were able to achieve minimum efficient scale in such an environment—that is, whether the conduct at issue was capable of precluding otherwise viable competitors from gaining enough customers to sustain themselves as a competitive alternative. Indeed, this is exactly what Google argued was required: “[Google] contends that Plaintiffs have failed to establish a link between the agreements, the denial of sufficient scale to rivals, and anticompetitive effects….”[44] It is also what Judge Ginsburg says is required: “The court [in Microsoft] inferred harm to the competitive process from these findings, in essence recognizing that minimum-efficient scale is the mechanism by which exclusionary conduct harms competition.”[45]

Yet nowhere does Judge Mehta effectively grapple with the minimum-efficient-scale question. He does discuss the importance of scale in search, and he holds that Google is of higher quality than Bing and other competitors, in part, because of its scale. But he never really asks or answers the questions 1) whether this difference is uniquely attributable to Google’s default distribution deals, and 2) whether those deals preclude rivals from effectively competing, or simply make it harder for them to compete because they raise the cost of achieving comparable quality.

Now, the evidence here, as far as we can assess it from the decision, is not entirely clear-cut. But the answer isn’t really the issue. The real issue is that this is an essential question, on which the government bears the burden of proof, and it was simply missing from the opinion. In other words, this means the holding in the government’s favor is unsupported as a matter of law.

B. De Facto Exclusivity

Even so, let’s assess how well the evidence supports the conclusion that the default deals were really “de facto” exclusive, and that they prevented rivals from achieving the minimum efficient scale.

We have no idea if the default deals had anything to do with it, but we do know that Neeva, a once-promising general search engine, apparently had a hard time competing for users and went out of business after about four years.[46] On the other hand, Bing, Yahoo!, DuckDuckGo, Ecosia, and Brave all exist and continue to compete in this environment. Yes, they have relatively small market shares, but apparently they have enough scale for viability.

Similarly, on the one hand, there has been clear “competition for the market” between Bing and Google with respect to the default access points on Apple devices and in Mozilla’s Firefox browser. On the other hand, there is less clear competition with respect to Android OEMs (in Google’s favor)—but there is also less clear competition (actually, there is none) between them for default placement in Edge (in Bing’s favor).

1. Misunderstanding the ‘power of defaults’

With respect to the conclusion that the cost to users of choosing the non-default option is higher, that is inherently true, of course. But it is arguably trivially so. Judge Mehta spends a fair amount of time on this question (although not in the proper context of this but-for assessment) before arriving at his conclusion that being a non-default is tantamount to being excluded. His analysis, however, is unconvincing.

First, the analysis is heavily influenced by the assertions of the government’s behavioral expert, Antonio Rangel.[47] As I will discuss below, some empirical data specific to the context at hand is used to bolster the more general behavioral claims of the government’s expert (which I believe cuts in many respects against Judge Mehta’s conclusions). But it is clear that any ambiguity was resolved by Judge Mehta in favor of asserted general behavioral patterns:

That users overwhelmingly use Google through preloaded search access points is explained in part by default bias or the “power of defaults.” The field of behavioral economics teaches that a consumer’s choice can be heavily influenced by how it is presented. The consensus in the field is that “defaults have a powerful impact on consumer decisions.”[48]

There are access points other than the default that can be used to distribute a GSE, but those channels are far less effective at reaching users. That is due in part to users’ lack of awareness of these options and the “choice friction” required to reach these alternatives.[49]

[A]s Dr. Rangel convincingly explained, the combination of user habit, Google’s brand, and choice friction creates a powerful default effect that drives most consumers to use the default search access points occupied by Google.[50]

The main problem with this is not so much that behavioral science is wrong (surely, it is correct that the more friction there is to switch from a default, the less likely someone will switch), but that it is not dispositive. This makes it a weak basis for meeting the plaintiffs’ burden. It is also not clear that general behavioral theories have the same traction in the specific environment at issue. As my colleague Dan Gilman has discussed, the learnings of behavioral science were established in settings quite different than search engine defaults.[51] “Generalizing findings from, e.g., cereal-box placement to the durability of search engine defaults seems a stretch (or entirely speculative).”[52]

To be sure, Rangel’s testimony did purport to apply those learnings in context.[53] But what really matters in this case is not the direction of the behavioral assumptions, but the magnitude. (Again, no one disputes that defaults grant some benefit, nor that promoting one’s products—i.e., marketing—can influence consumer choice). The claim here is that the availability of switching does not sufficiently negate the effective exclusivity of defaults to permit rivals to compete. That claim depends on the extent to which users tend not to switch away from defaults, not just the fact that they sometimes don’t.

Among other things (more of which are discussed below), it must be noted that, even when users are presented with a neutral option (e.g., a “choice screen”), they appear to make essentially the same choices as when presented with a default. In Europe, where Google has since 2020 implemented a search engine choice screen on Android following the EU’s 2018 antitrust decision against it,[54] Google’s share of the search engine market has barely budged.[55]

By the same token (at least when Google is the non-default) users are apparently quick to switch from a less-preferred default in order to get access to Google Search:

In a 2016 experiment, Mozilla switched the default GSE on both new and existing users from Google to Bing. By the twelfth day, Bing had kept only 42% of the search volume. After some additional time, those numbers dropped to 20–35%….[56]

It is exceedingly difficult to square these facts with the court’s conclusions on the functional irrelevance of non-default options.

C. Overlooking Evidence of User Switching Behavior and Impressionistic, Not Quantitative, Conclusions

The majority of the decision’s discussion of consumers’ behavior around defaults is largely impressionistic, not quantitative. For example, Judge Mehta notes that:

Another non-default search access point is the bookmarks page on a browser. The Safari “Favorites” page, for instance, contains preloaded icons to access Google, Bing, and Yahoo. A user also can add a new search engine on that page. But few consumers use this channel, as it first requires finding the Favorites page in a new Safari tab, which requires an “extra click.” Google itself receives only 10% of its searches on Safari through the bookmark.[57]

Strictly speaking, 10% is “quantitative,” but the decision’s conclusions based on this data are decidedly impressionistic. Judge Mehta asserts that Google receiving “only” 10% of searches on Safari through the bookmark is an insignificant volume. But in that setting—where Google Search is already the default on Safari and can be accessed simply by typing in Safari’s URL bar, and in which it is alleged that virtually no one ever uses anything other than default search on Safari—why would there be any searches on Google via the bookmark? If that number of searches is at all different from zero, it would appear to demonstrate that it is indeed a relevant channel by which consumers can find search engines, including non-default ones. In other words, 10% may be “insignificant” as a share of Google searches, but it is quite significant with respect to the relevant legal standard.

Indeed, “nearly 40% of queries on Apple’s mobile devices flow through non-default search access points, such as default bookmarks or organic search.”[58] Judge Mehta dismisses this by arguing that “the fact that some consumers access search on non-default access points is not dispositive on exclusivity.”[59] “Not dispositive” is not quantitative. While the statement is true, the burden of proof is on the plaintiffs, and this not being dispositive cuts against them, not against Google.

Elsewhere, Judge Mehta also rejects the actual evidence of people switching to the non-default on PC desktops. It turns out that a lot of Windows desktop users download Chrome and use Google Search there, rather than relying on Bing, which is the default search engine in the Edge browser:

To be sure, downloads of an alternative browser occur with greater frequency on Windows desktop computers. On such devices, Edge is the default browser and Bing is the default search engine. Yet, Google’s search share on Windows devices is 80%, with most of the queries flowing through the Chrome default, which means Chrome was downloaded onto the device.[60]

Despite this, Judge Mehta is quick to note that, for those users who still use Edge, Bing is the most used search engine:

The power of defaults is evident, however, from the share of Bing users on Edge. Bing’s search share on Edge is approximately 80%; Google’s share is only 20%. Even if one assumes that some portion of those Bing searches are performed by Microsoft-brand loyalists, Bing’s uniquely high search share on Edge cannot be explained by that alone. The default on Edge drives queries to Bing.[61]

One might suggest that all this shows is that people really prefer Chrome to Edge, not that they prefer Google Search to Bing enough to switch away from the default (on either browser). Except that, as the opinion points out, “Google’s dominance on Windows cannot, however, be attributed simply to the popularity of Chrome. Google had an 80% search share on Windows when Chrome first launched, and that share has remained steady ever since.”[62] If that’s the case, it can mean only that the default search on Windows desktops isn’t very sticky—and it isn’t just because users prefer Chrome to Edge; apparently it’s because they prefer Google Search to Bing.

So how does the court conclude that it supports the “power of defaults” that, of those users who don’t switch to Chrome on Windows desktops, approximately 80% use Edge’s default? If most Windows users who prefer Google Search to Bing switch from the default by downloading Chrome instead of by switching the default in Edge, then of course most of those who remain on Edge will use Bing. If they preferred Google to Bing, they would have switched to Chrome.

As Judge Mehta notes elsewhere, “[m]any users do not know that there is a default search engine, what it is, or that it can be changed.”[63] Perhaps. But then again, apparently, many users do know that Chrome gets them access to their preferred search engine. Whatever “choice friction” impedes the movement away from the default search engine on Windows desktops, it is not strong enough to prevent people from maneuvering around it in spades—they just don’t often do so directly by switching the default search engine in Edge.[64]

The opinion also brushes off these examples of default switching by asserting that they merely “confirm that the default effect is weaker when the alternative is a dominant firm with high brand recognition backed by a quality product.”[65] First, this is pretty hand-wavy and impressionistic. Maybe it’s true; in fact, I’m sure it’s true to some extent. But for an opinion that otherwise regularly says we have to look at “market realities,” not the world as it might be, this is a weak basis to conclude that evidence of people switching away from defaults doesn’t really show that people switch away from defaults.

Regardless—isn’t the ability to attract users because you are widely used, have good brand recognition, and have a demonstrably high-quality product pretty much the definition of competition on the merits? Indeed, one could recast Judge Mehta’s statement as precisely the opposite of the decision’s holding: Google’s default agreements can’t be deemed to have caused anticompetitive harm because defaults are readily overcome by high-quality, reputable alternatives.

It should also be noted (but, unfortunately, Judge Mehta doesn’t say it) that Microsoft is also a “dominant firm with high brand recognition backed by a quality product.”[66] What’s good for the goose is good for the gander: Microsoft has plenty of market heft to ensure that its products don’t languish in obscurity in the face of consumer inertia.[67]

III. The Scale and Quality Argument: A Double-Edged Sword

All of which raises the question: Is Bing a comparably high-quality product or not? Determining that seems like a pretty important prerequisite to determining whether its small market share is a function of anticompetitive exclusion or a failure to compete on the merits. Yet Judge Mehta is, at best, equivocal on this. First, he notes that:

Everyone agrees that Google’s distribution agreements did not cause Microsoft’s past underinvestment in search. Microsoft “missed” the mobile revolution and was unable to improve its browser, Internet Explorer, until it used Google’s rendering engine, Chromium. Some of Microsoft’s quality issues also were attributable to its poor index.[68]

Yet, “[u]ltimately, Microsoft committed significant capital to search.”[69] And “[t]hat investment (combined with secured distribution on Windows devices) has allowed Bing to achieve quality parity with Google on Windows desktop devices.”[70]

Elsewhere, however, Judge Mehta concludes that “Google’s exclusive agreements… deny rivals access to user queries, or scale, needed to effectively compete,”[71] and that “[t]his perpetual scale and quality deficit means that Microsoft has no genuine hope of displacing Google as the default GSE on Safari. As Apple’s Eddy Cue testified, there was ‘no price that Microsoft could ever offer [Apple]’ to prompt a switch to Bing, because it lacks Google’s quality.”[72]

I admit that it’s unclear to me why Bing’s apparent quality parity in desktop search doesn’t redound to its benefit in mobile search. Indeed, it has to be noted that the court did not identify separate relevant markets for mobile and desktop search; it identified a single “general search services” market.[73] So, it’s a little unclear, but it seems that, according to the court, ultimately Bing simply isn’t up to Google Search’s quality standard.

A. The Failure to Distinguish Between Exclusion and Low Quality: A Catastrophic Legal Blunder

The problem is that it is precisely past decisions and their alleged influence on current outcomes that the court uses to establish the proposition that Google’s default deals are anticompetitive. As I keep pointing out, however—and as Judge Mehta appears here to recognize—plaintiffs cannot meet their burden of proof that Google’s deals were exclusionary by pointing to Bing’s limited success, if the court agrees that Bing’s low quality could also have caused it. And here, Judge Mehta also concedes that those deals didn’t cause Bing’s lower quality, either (“Google’s distribution agreements did not cause Microsoft’s past underinvestment in search.”).[74]

For the court to sustain its claim that Microsoft is the appropriate guiding precedent (and thus, that the government has made its case under Microsoft’s “edentulous” legal standard), it has to be the case that Bing could have outcompeted Google on quality if not for the agreements—that is, that it “reasonably constituted [a] threat.”[75]

By conceding that Bing was unable to secure distribution deals comparable to Google’s because of its low quality, however, the opinion (and the government, of course) fails to do this. As such, they fall right into the trap explained by Greg Werden:

But if operating at a much smaller scale than Google makes rival search engines uncompetitive, their fate was sealed when Google achieved a dominant share. The government posits no scenario in which any rival search engine could have substantially closed the scale gap…. If the government’s scale contentions are fully credited, the conduct that is at the heart of the case did not maintain Google’s dominant share. And any conduct that could not have maintained dominance most likely served a legitimate purpose. One way or another, the elements of the monopolization offense cannot be established under the government’s view of the facts…. But the government does not contend that rival search engines ever posed a real threat to Google’s monopoly. Indeed, it claims to have proved just the opposite.[76]

That’s a catastrophic problem for the opinion’s holding. Nevertheless, Judge Mehta does find that Bing is not a viable competitor on mobile. Yet he refutes Google’s claim that this is because of Microsoft’s own business failures, rather than its inability to gain scale:

Google also maintains that the quantity of user data is less important than how it is used, and if its rivals had Google’s business foresight and drive to innovate, they too could win default distribution. But that position blinks reality. Apple’s flirtation with Microsoft best illustrates this point. Microsoft has invested $100 billion in search in the last two decades and its quality now matches Google’s on desktop search. Yet, Microsoft’s failure to anticipate the emergence of mobile search caused it to fall behind, and with Google guaranteed default placement on all mobile devices, Microsoft has never achieved the mobile distribution that it needs to improve on that platform.[77]

Isn’t Microsoft’s “failure to anticipate the emergence of mobile search” precisely the sort of competitive failure that Google is talking about? How is Microsoft’s diminished scale attributable to Google’s conduct if it was Microsoft’s independent business decisions that denied it the ability to compete effectively?

This is exactly why a plaintiff must prove that the defendant’s conduct, and not an excluded rival’s own mistakes, were the cause of the rival’s inability to compete. Otherwise, the law would be enlisted to rectify competitors’ poor business decisions, rather than to protect the competitive process.

1. Even Judge Mehta knows ‘reasonably appears capable of’ is the wrong standard

It also bears noting that Judge Mehta already—and properly—threw out exactly this sort of claim on summary judgment when he dismissed the plaintiff states’ claims that “Google’s targeting of SVPs [specialized vertical providers] caused anticompetitive effects in the proposed markets.”[78] But the basis on which he did so is shockingly at odds with the basis for his decision in the government’s favor in this case. Citing Microsoft, in fact, Judge Mehta held in his summary judgment opinion that:

Speculation that Google’s conduct “can reasonably be expected,” “might,” or “could potentially” degrade SVPs and make them less attractive partners to Google’s rivals is not evidence of anticompetitive effects in the relevant markets. Plaintiffs are required to show with proof “that the monopolist’s conduct indeed has the requisite anticompetitive effect,” and they have fallen well short.[79]

And, as he notes elsewhere in his summary judgment opinion, also citing Microsoft, “[t]he sole issue for the court to resolve is whether Google has maintained monopoly power in the relevant markets through ‘exclusionary conduct’ as opposed to procompetitive means.”[80]

The words “can reasonably be expected”—rejected by Judge Mehta in his summary judgment decision—might ring a bell, as they are awfully close to the “reasonably appear capable of” standard adopted by the court in this decision.

B. Less-Efficient Channels of Distribution: Misapplying Microsoft Again

Finally, there is another problem with the legal sufficiency of the exclusivity claims, and it stems, once again, from a misapplication of Microsoft.

Judge Mehta claims that “mere user access to these less efficient channels of distribution does not render the browser agreements non-exclusive.”[81] A significant part of the defense of this position turns on an analogy to Microsoft and the argument there that it was sufficient that Microsoft foreclosed access to the best method of distribution. Indeed, the next sentence after the quote above is, “Microsoft again illustrates the point.”[82] But does it?

Judge Mehta says this case is the same as Microsoft where the court “reject[ed] the argument that Microsoft’s licensing agreements with OEMs were not exclusive ‘because Netscape is not completely blocked from distributing its product,’ as ‘although Microsoft did not bar its rivals from all means of distribution, it did bar them from the cost-efficient ones.’”[83] He then asserts that “[t]he record here resembles that in Microsoft. Users are free to navigate to Google’s rivals through non-default search access points, but they rarely do.”[84]

But this elides a couple of key things.

First, the Microsoft court didn’t look ex post at what consumers did (which, as I tire of pointing out, could be attributable to either anticompetitive conduct or consumer preferences); it looked at which channels of distribution were available and if they were viable substitutes, regardless of whether they were actually used or not.

The analogy to Microsoft fails most obviously on the point that the “market realities” have changed a lot since the late 1990s. Downloading Netscape from the internet was wholly unfamiliar, exceedingly complex, and truly difficult for PC users back then—a real “choice friction” and thus not really a viable alternative. But downloading a competing search engine or browser today is trivially easy, and users do it all the time (to the tune of 12.6 billion app downloads in the United States in 2023 alone).[85] In this environment, the fact that users don’t download or use competing general search engines sufficiently to displace Google Search despite the ease of doing so suggests that it is consumer preference for Google Search, not the relative inefficiency of the channel of distribution, that causes this result.

Instead, Judge Mehta concludes that, while “a user can download Chrome, Edge, or [DuckDuckGo] onto an Apple device,” “[t]his, too, is not an easily accessible search point, as it involves similar choice friction as acquiring a search application. Google receives only 7.6% of all queries on Apple devices through user-downloaded Chrome.”[86]

Not only is downloading an application trivially easy, but the fact that Google receives only 7.6% of search queries on Apple devices through Chrome, but “most”[87] of its queries on Windows desktops through user-downloaded Chrome is decidedly ambiguous. Maybe that shows that people download Chrome on Windows not to get easy access to Google Search but because the Chrome browser is superior to the Edge browser, while it is not any better than Safari. But it is also consistent with the conclusion that people aren’t prevented from accessing their preferred search provider (Google Search)—they just don’t need to download Chrome on Apple devices to get easy access to it, while they do need to do so on Windows devices.

Second, the opinion says that “[t]he court in Microsoft did not say that these contracts caused zero market foreclosure merely because Internet Explorer had other, less-efficient means of reaching users.”[88] True. But the court in Microsoft also didn’t say that any amount of difference in distribution efficiency was sufficient to maintain that a non-exclusive agreement was effectively exclusive. As noted, it is now trivially easy to switch search providers on virtually every platform and at multiple decision points on each. Defaults don’t prevent that, and prioritized placement (from, e.g., a spot on the Android home screen) doesn’t even crowd out alternatives once they are downloaded (which can then be similarly accessed from priority positions on the home screen). “Very slightly less efficient” could still be “efficient.” The fact that the difference between the foreclosed and available channels of distribution in Microsoft was large enough to matter does not mean that the difference between them in Google Search is big enough to matter.

1. So, Dentsply is good law, but Rambus isn’t?

In response, Judge Mehta goes back to ex post user conduct to hold that the fact that users don’t often use these alternatives shows that the difference does matter here, and that Google’s default distribution deals are effectively exclusive and lead to foreclosure:

Sure, users can access Google’s rivals by switching the default search access point or by downloading a rival search app or browser. But the market reality is that users rarely do so. The fact that exclusive agreements allow users to reach rivals through other means does not make the foreclosure number zero.[89]

But it cannot be a sufficient argument that “the market reality is that users rarely do so.” That market reality is exactly what is at issue in the case. Using the lack of user uptake from trivially easy alternative distribution channels as evidence that those alternative distribution channels aren’t relevant assumes the conclusion. It’s poor legal reasoning.

Judge Mehta is correct, however, that “‘[t]he mere existence of other avenues of distribution is insufficient without an assessment of their overall significance to the market.’”[90] If only he had demanded such an assessment.

The Dentsply case that Judge Mehta cites for this proposition was (in my opinion) wrongly decided. It shouldn’t be held up as the standard of analysis and, in any case, it was in the 3^rd U.S. Circuit Court of Appeals and not binding precedent. But even so, Dentsply dealt with exclusive agreements that included a term explicitly prohibiting authorized distributors from selling rivals’ products, thus arguably making it extremely difficult for those products to be accessed by the ultimate consumer. This case is different. None of Google’s agreements include terms prohibiting its counterparties from dealing with anyone else. And here, competing products are available to the ultimate consumer, and they show up on users’ devices in locations virtually identical to Google’s.

In any case, the Dentsply court does not rely on ex-post uptake to support its claim that alternative distribution channels are insignificant (although it does look at that statistic on occasion). Instead, it describes in detail the qualitative differences between the channel of distribution foreclosed by Dentsply and the alternatives, finding that the alternatives are decidedly less attractive. Here, by contrast, the only thing that distinguishes default placement from the other channels of distribution is alleged “choice friction.” Otherwise, they are, quite literally, identical (or, as in the difference between, say, search bar integration and a home-page bookmark, trivially different). That makes assessing their “overall significance to the market” dependent on what is being distributed, and not solely the channel of distribution itself.

2. In fact, we know from other parts of the decision that ‘less-efficient’ alternatives can’t be dismissed

Later, also quoting Dentsply, Judge Mehta asserts that:

In the end, Google’s dismissal of the importance of scale is inconsistent with market realities. Google often warns that competition is “only a click away.” However, “[t]he paltry penetration in the market by competitors over the years has been a refutation of [that] theory by tangible and measurable results in the real world.”[91]

This misses the mark for the same reason. There is plenty of evidence to demonstrate that competition is just a click away. In fact, some of it was evidence the court used to exclude specialized vertical search providers (e.g., Amazon and TripAdvisor) from the relevant market. Without challenging that conclusion here (although I do think it has problems), it appears eminently “tangible and measurable” to the question of whether users will switch to alternative search engines that, when the alternative is demonstrably superior for the query at issue, they do so in droves:

Google views competition from SVPs as “intense for commercial clicks.” A 2020 Bank of America study reported that 58% of users search Amazon first when they seek to make an online purchase, as opposed to only 25% who go first to Google, demonstrating Google’s secondary status as a starting point for users with high commercial intent….

…Microsoft recognizes that “if Bing or Google were not doing vertical searches well, or at least not having organic results that people could click to get to vertical search engines,” users might bypass GSEs and instead search directly on Amazon from the outset….

…[A]nalysis show[s] that a query sample of Google’s top 25 non-navigational shopping queries attracts more queries weekly on Amazon (3.7 million) than Bing (0.4 million)…, [and] that Yelp’s local query volume is higher than Google’s and much higher than Bing’s.[92]

None of these alternative vertical search engines is installed as the default. And yet, when consumer preference is strong enough—when they produce better results—consumers have no trouble using them. Whether or not this is sufficient to affect the court’s relevant market or market-share analysis, it is surely enough to demonstrate that users are not locked into defaults when the “choice friction” required to switch from them is small relative to the benefit. That, in turn, is a function of the quality of the search provider, not the method by which it is accessed.

IV. Getting It Wrong on the Substantiality of Foreclosure, Too

The “substantiality” of foreclosure must also be briefly addressed, for similar reasons. While the opinion downplays its significance as a search engine distribution channel, Windows desktops constitute a substantial share of the distribution market. Windows accounts for 64% of desktop operating systems and almost 30% of all operating systems across all platforms in the United States.[93] On these devices, Bing is the default search engine. So, right out of the box, the share of the market that Google could even possibly foreclose is reduced by Windows’ 30% market share.

Of the remaining 70%, we know that small but non-trivial portions are not actually foreclosed to competitors. We know this from the ex-post data showing, for example, that “5.1% of all searches on iPhones are conducted on a GSE other than Google [where it is the default].”[94] We also know that, on Android, “[a]lthough OEMs must preload the Google Search Widget, users can delete it. As of 2016, there were about 200,000 logged widget deletions daily.”[95] Also, “Samsung already preloads a second browser—its proprietary S browser—on all Samsung devices.”[96]

We also know that “nearly 40% of queries on Apple’s mobile devices flow through non-default search access points, such as default bookmarks or organic search.”[97] Of course, a great number of these searches are performed on Google Search anyway.[98] But these are searches performed by users who demonstrably navigate around the default. By definition, they are not foreclosed to Google’s competitors.

Indeed, simultaneous with the Google default deals, Bing is, in fact, distributed by these counterparties to Google’s deals. Thus, as the result of an agreement with Microsoft, Bing shows up as an option on Safari’s homepage and on the Safari “Favorites” page, “which contains preloaded icons to access Google, Bing, and Yahoo.”[99] Mozilla has a “this time, search with” feature on Firefox “which allows users to select a different search product from its ‘Awesome Bar’ for a given query.”[100]

Again, Android is actually a somewhat unique case, and there the absence of true foreclosure is almost entirely dependent on the availability of end-consumer choice (which, again, is far from irrelevant). But even if we assume zero distribution of Bing on Android devices,[101] it still has unfettered access to distribution on Windows devices and is still distributed alongside Google on Apple devices and in Firefox.

The bottom line is that, even measured by ex-post consumer behavior, rivals are not foreclosed from access to consumers. And measured by the availability of access to rival search engines (whether consumers choose to use them or not), competitors are not actually foreclosed from distribution on any devices or in any browsers at all. To be sure, the remaining effective foreclosure level could be “substantial.” But nowhere does the court’s opinion demonstrate this. As plaintiffs have the burden of proof, the existence of meaningful consumer usage and availability, despite purported exclusive agreements, should have been deemed by the court to undermine the government’s case, not support it.

V. The Fateful Conclusion that Bing Isn’t a Real Competitor and the Problem of Remedy

Finally, I have to say a word on remedy here, although I do so for now only insofar as it bears on what I have been arguing; there are many other arguments about remedy that make this holding problematic. But here is one, and it hearkens back to Greg Werden’s Catch-22.[102]

The jig was up for the plaintiffs in this case once they argued that Bing was not a viable competitor to Google Search. In the world of that “market reality,” no reasonable remedy would do any good to rectify the allegedly anticompetitive circumstances. And the court accepted the government’s quality arguments pretty much wholesale:

The market reality is that Google is the only real choice as the default GSE. Apple’s Senior Vice President of Services, Eddy Cue, put it succinctly when, in a moment of (perhaps inadvertent) candor, he said: “[T]here’s no price that Microsoft could ever offer [Apple] to” preload Bing. “No price.” Mozilla stated something similar in a letter to the Department of Justice prior to the filing of this lawsuit. It wrote that switching the Firefox default to a rival search engine “would be a losing proposition” because no competitor could monetize search as effectively as Google. A “losing proposition.” If “no price” could entice a partner to switch, or if doing so is viewed as a “losing proposition,” Google does not face true market competition in search.[103]

But if “no price” could entice a partner to switch to Bing, and Bing is not truly a competitor to Google in search, then, as Greg Werden says, “the conduct that is at the heart of the case did not maintain Google’s dominant share.”[104]

The Microsoft decision relies on the contention that, although unproven, Netscape Navigator was a viable competitive threat to Windows. Thus, the government had to prove in that case that the threats to Microsoft’s operating-system monopoly were real, even if it didn’t have to prove the threats would have succeeded but for Microsoft’s conduct. The government’s burden is at least as high here.

And yet, in the quote above, Judge Mehta essentially finds that the government didn’t meet even this burden. He finds, in effect, that it wasn’t the nature of Google’s agreements that contributed to Google’s continued monopoly power; it was the fact that no distributor would ever choose Bing as the default—at any price. That conclusion means that it was Bing’s low quality that excluded it from default distribution and the reason “Google does not face true market competition in search”[105] is a function of quality, not Google’s deals.

That’s already fatal to the case. The remedy point is this: That same market reality means that no remedy prohibiting Google from entering into such agreements will rectify the situation. It means that Apple, Mozilla, Samsung, et al. will still choose Google as the default, even if Google is forbidden from paying them a revenue share (or even a set price) to do so—they will just forego the revenue from doing so, and Google will get a windfall.[106]

Yet it is hard to conceive of any other remedy that follows from Judge Mehta’s analysis in this decision. The decision is laser-focused on the determination that Google’s default distribution deals were (effectively) exclusive and thus foreclosed a substantial share of the market and deprived rivals of scale. Everything in the decision comes down to the default nature of the deals. It stands to reason that any remedy would be limited to removing the one-deal characteristic that, according to the court, makes the agreements anticompetitive.

Cutting against this somewhat is Judge Mehta’s conclusion that, having been deprived of scale by Google’s distribution deals, no rival is in a position to secure a default deal of its own.[107] But it is by no means clear from the decision that, in the absence of default deals with Google, rivals would be unable to compete effectively through other channels of distribution or compete for such deals in the future. The problem is that, because the court has no ability to prohibit Apple and Mozilla from offering default search engines without a Google deal, even prohibiting Google from entering into those deals doesn’t mean it won’t still be offered as the default, and this may not change the competitive landscape enough to enable Bing and other rivals to compete effectively.

Perhaps one might think that Google should just be compelled to share its data (and/or other “secret sauce”) with rivals so they have the quality necessary to actually win default placement deals. But that doesn’t work either. In the first place, this would be a clear acknowledgment that it is quality, not default distribution deals, that impedes rivals’ commercial success. If that is the only effective remedy, then the necessary legal basis of the holding is undermined.

Secondly, implementing that remedy would entail mandating that Google enter into deals with competitors to help them compete. This is anathema to U.S. antitrust law.[108] So much so that Judge Mehta himself threw out one of the plaintiff states’ claims in this case on exactly that basis:

Plaintiff States seek to bypass the “no duty to deal” doctrine entirely….

…The concerns that animate the no-duty-to-deal principle are equally applicable here. Primarily, adjudicating Plaintiff States’ claim would require the court to act as a “central planner” that endeavors to identify the proper “terms of dealing.” Their claim requires grappling with a host of questions that the court is ill-equipped to handle…. And those thorny questions foreshadow the challenges the court would face in administering a remedy…. A favorable outcome for Plaintiff States thus would mire the court in Google’s day-to-day operations…. The court has learned a lot about Google, but it is “ill suited” for that role.[109]

It is extremely difficult to see how Judge Mehta would countenance a forced-sharing arrangement for Google’s data as a remedy for the remaining claims in this case, given his unequivocal dismissal of other claims on precisely that basis.

As I noted, there is more to say about potential remedies in this case.[110] But for now, the most important thing is that the absence of viable remedies strongly supports the arguments I have presented here that the court’s liability finding was improper.

 

 

[1] United States, et al. v. Google LLC, Memorandum Opinion, Case No. 20-cv-3010 (APM) (D.D.C., Aug. 5, 2024), https://www.tn.gov/content/dam/tn/attorneygeneral/documents/pr/2024/pr24-59-Google.pdf ( “Google Search decision”).

[2] United States, et al. v. Google LLC, Memorandum Opinion, Case No. 20-cv-3010 (APM) (D.D.C., Aug. 4, 2023) (“Google Search summary judgment”), https://storage.courtlistener.com/recap/gov.uscourts.dcd.223205/gov.uscourts.dcd.223205.626.0_3.pdf.

[3] United States v. Microsoft, 253 F.3d 34 (D.C. Cir. 2001) (“Microsoft”).

[4] Google Search decision at 216 (quoting Microsoft, 253 F.3d at 79 and id. (quoting Areeda)).

[5] See Douglas H. Ginsburg & Koren Wong-Ervin, Challenging Consummated Mergers Under Section 2, CPI North America Column 2 (May 25, 2020), https://www.pymnts.com/cpi-posts/challenging-consummated-mergers-under-section-2-2 (“[T]he assertion that Section 2 does not require proof of anticompetitive effects is based upon a misreading of the D.C. Circuit’s decision in United States v. Microsoft. The assertion conflates the Microsoft court’s standard for proving competitive effects with its standard for establishing causation.”).

[6] Microsoft, 253 F.3d at 58 (“[T]he plaintiff, on whom the burden of proof of course rests… must demonstrate that the monopolist’s conduct indeed has the requisite anticompetitive effect.”).

[7] See Ginsburg and Wong-Ervin, supra note 5, at 2 (“The court went on to devote fully 20 pages to a careful analysis of the actual effects of each type of Microsoft’s allegedly anticompetitive conduct. Only after finding that each type of conduct indeed had an anticompetitive effect did the court turn to the separate and distinct question of causation.”).

[8] Id. at 3 (“[I]t is important to understand the limited applicability of the “reasonably capable of” standard.”)

[9] See id. at 2 (“It was in addressing this question that the court said it was appropriate, in a government enforcement action, to ‘infer causation when exclusionary conduct is aimed at producers of nascent competitive technologies as well as when it is aimed at producers of established substitutes.’”) (quoting Microsoft, 253 F.3d at 79).

[10] Id. at 4.

[11] Id. at 3 (quoting Microsoft, 253 F.3d at 79).

[12] Id. at 3.

[13] Presumably to maintain the requirement of antitrust injury for private plaintiffs, which helps ensure they are not abusing the courts just to harm a competitive rival.

[14] D. Bruce Hoffman, Antitrust in the Digital Economy: A Snapshot of Federal Trade Commission Issues, Remarks at GCR Live Antitrust in the Digital Economy 10 (May 2019), https://www.ftc.gov/system/files/documents/public_statements/1522327/hoffman__gcr_live_san_francisco_2019_speech_5-22-19.pdf.

[15] Rambus Inc. v. FTC, 522 F.3d 456 (D.C. Cir. 2008) (“Rambus”).

[16] Rambus, 522 F.3d at 463 (emphasis added).

[17] Id.

[18] Id. at 463 (“[I]f Rambus’s more complete disclosure would have caused [the SSO] to adopt a different (open, non-proprietary) standard, then its failure to disclose harmed competition and would support a monopolization claim.”).

[19] Id. at 463-64 (“But while we can assume that Rambus’s non-disclosure made the adoption of its technologies somewhat more likely than broad disclosure would have, the Commission made clear… that there was insufficient evidence that [the SSO] would have standardized other technologies had it known the full scope of Rambus’s intellectual property. Therefore, for the Commission’s syllogism to survive—and for the Commission to have carried its burden of proving that Rambus’s conduct had an anticompetitive effect—we must also be convinced that if Rambus’s conduct merely enabled it to avoid… RAND licensing terms, such conduct, alone, could be said to harm competition.”).

[20] Id. at 466-67 (emphasis added).

[21] Joshua D. Wright, Why the Supreme Court Was Correct to Deny Certiorari in FTC v. Rambus, Mar-09(2) Global Competition Policy 1, 6 (2009), https://www.competitionpolicyinternational.com/file/view/5880.

[22] Id.

[23] Rambus, 522 F.3d at 463.

[24] Wright, supra note 21, at 6 (emphasis added).

[25] Rambus, 522 F.3d at 467.

[26] Google Search decision at 219 (emphasis added).

[27] Ginsburg & Wong-Ervin, supra note 5, at 4.

[28] Google Search decision at 216.

[29] A quick aside: There are those who actually hold this view. They are radical outliers in the antitrust community and their position is unsupported by the sensible reading of any case law or economic analysis. See, e.g., Petition for Rulemaking to Prohibit Exclusionary Contracts by Open Markets Institute, et al., Fed. Trade Comm’n Doc. ID FTC-2021-0036-0002 (Jul. 20, 2021), https://downloads.regulations.gov/FTC-2021-0036-0002/content.pdf. They include stalwart neo-Brandeisian ideologues like the Open Markets Institute, the American Economic Liberties Project, and Marshall Steinbaum. They also include entities decidedly foreign (as far as I know) to the antitrust community like Friends of Family Farmers, Friends of the Earth, and the People’s Parity Project. Their position doesn’t represent what the law or the economics actually says.

[30] At least, it is not a question in the abstract given the obvious and direct competition from alternative general search engines. As I discuss below, however, the court makes this a question by challenging the relative quality of rival search engines. In this case, that actually constitutes another, separate basis for reversing the decision. See infra Section III.A.

[31] NB: “highly uncertain” in that Rambus’s market position could have been the result of either of two causes; not “highly uncertain” in the Microsoft sense that the competitive process was speculative.

[32] Ginsburg & Wong-Ervin, supra note 5, at 4 (emphasis added).

[33] Google Search decision at 216.

[34] Id. at 217.

[35] Id. at 216.

[36] See generally Joshua D. Wright, Moving Beyond Naïve Foreclosure Analysis, 19 Geo. Mason L. Rev. 1163 (2012); id. at 1181-82 (“The primary thrust of this Article is that accurately measuring the foreclosure produced by any allegedly exclusionary agreement requires foreclosure to be measured relative to what would be obtained but for that agreement.”); Thomas G. Krattenmaker and Steven C. Salop, Anticompetitive Exclusion: Raising Rivals’ Costs to Achieve Power over Price, 96 Yale L.J. 209, 259 (1986) (defining a “net foreclosure rate” as “the percentage of the suppliers’ capacity that was available to rivals before the exclusionary rights agreement was adopted but that is no longer available as a result of the agreement”).

[37] Rambus, 522 F.3d at 464.

[38] See generally Geoffrey A. Manne, Error Costs in Digital Markets, in Global Antitrust Institute Report on the Digital Economy (Joshua D. Wright and Douglas H. Ginsburg eds. 2020), https://gaidigitalreport.com/wp-content/uploads/2020/11/Manne-Error-Costs-in-Digital-Markets.pdf.

[39] Ginsburg & Wong-Ervin, supra note 5, at 4.

[40] See, e.g., Google Search decision at 24 (“The most efficient channel of GSE distribution is, by far, placement as the preloaded, out-of-the-box default GSE.”).

[41] Rambus, 522 F.3d at 466 (citing Schuylkill Energy Res., Inc. v. Penn. Power Light Co., 113 F.3d 405, 414 (3d Cir. 1997)).

[42] Google Search decision at 228 (“That means only 30% of all GSE queries in the United States come through a search access point that is not preloaded with Google.”).

[43] See also infra note 101, noting that it might be relevant to assess whether the size of Google’s payments was sufficient to induce distributors to promote it as the default to an extent out of proportion to its relative quality advantage (and thus, presumably, consumer preferences). But the court did not undertake this analysis.

[44] Id. at 227.

[45] Ginsburg & Wong-Ervin, supra note 5, at 3 (emphasis added). See also Microsoft, 253 F.3d at 71 (“[Microsoft’s conduct] help[ed] keep usage of Navigator below the critical level necessary for Navigator or any other rival to pose a real threat to Microsoft’s monopoly.’”)

[46] See Google Search decision at 163 (“As for Neeva, it entered and exited within four years.”).

[47] Who is also, perhaps tellingly, the Bing Professor of Neuroscience, Behavioral Biology & Economics at Caltech. See Antonio Rangel, Caltech (last accessed Aug. 13, 2024), https://www.hss.caltech.edu/people/antonio-rangel.

[48] Id. at 26.

[49] Id. at 31.

[50]  Id. at 229.

[51] See Daniel J. Gilman, Google, Amazon, Switching Costs, and Red Herrings, Truth on the Market (Nov. 22, 2023), https://truthonthemarket.com/2023/11/22/google-amazon-switching-costs-and-red-herrings.

[52] Id.

[53] See Prof. Antonio Rangel Presentation (Sept. 12-13, 2023), U.S. v. Google LLC Trial Exhibit No. UPXD104, available at https://www.justice.gov/d9/2023-09/416682.pdf.

[54] See Google Android (Case COMP/AT.40099) Commission Decision of 18 July 2018, https://competition-cases.ec.europa.eu/cases/AT.40099.

[55] See Mobile Search Engine Market Share in Europe — July 2024, StatCounter (last visited Aug. 12, 2024), https://gs.statcounter.com/search-engine-market-share/mobile/europe/#monthly-202001-202408 (showing that Google’s EU mobile search engine market share was 97.32% in January 2020 (just before the choice screen was implemented) and 95.96% in July 2024—a change of 1.37 percentage points).

[56] Google Search decision at 117.

[57] Id. at 32.

[58] Id. at 207.

[59] Id. at 208.

[60] Id. at 33.

[61] Id.

[62] Google Search decision at 32.

[63] Id. at 27.

[64] Or, at least, they don’t anymore. But 80% of searches on Windows desktops ran through Google Search even before Chrome existed. Id. at 32. That means that, apparently, a substantial majority of users were perfectly willing and able to overcome the “choice friction” and switch to Google Search even within Windows’ default browser.

[65] Id. at 229.

[66] By which I mean Office and Windows, not necessarily Bing. But the point holds: Microsoft is certainly all these things.

[67] It should also be noted that Microsoft did not just passively rely on its name recognition to overcome apparent consumer disinterest; rather, it actively tried to induce users to switch by offering “Bing Rewards” (now “Microsoft Rewards”) for using Bing, redeemable for gift cards, Microsoft services, and the like. See Microsoft Rewards, Microsoft (last visited Aug. 12, 2024), https://www.microsoft.com/en-us/rewards/about.

[68] Google Search decision at 238 (emphasis added).

[69] Id.

[70] Id. (emphasis added).

[71] Id. at 226.

[72] Id. at 232 (emphasis added).

[73] Id. at 152.

[74] Google Search decision at 238.

[75] Microsoft, 253 F.3d at 79 (“Given this rather edentulous test for causation, the question in this case is not whether Java or Navigator would actually have developed into viable platform substitutes, but (1) whether as a general matter the exclusion of nascent threats is the type of conduct that is reasonably capable of contributing significantly to a defendant’s continued monopoly power and (2) whether Java and Navigator reasonably constituted nascent threats at the time Microsoft engaged in the anticompetitive conduct at issue.”) (emphasis added).

[76] Gregory J. Werden, The Missing Element in the Google Case, Truth on the Market (Apr. 15, 2024), https://truthonthemarket.com/2024/04/15/the-missing-element-in-the-google-case (emphasis added).

[77] Id. at 232 (emphasis added)].

[78] Google Search decision at 6. See generally Google Search summary judgment.

[79] Google Search summary judgment at 48-49 (quoting Microsoft, 253 F.3d at 58-59) (emphasis in original).

[80] Id. at 21 (quoting Microsoft, 253 F.3d at 58).

[81] Google Search decision at 209.

[82] Id.

[83] Google Search decision at 209 (citing Microsoft, 253 F.3d at 64) (emphasis added).

[84] Id.

[85] See, e.g., Laura Ceci, Number of Mobile App Downloads Worldwide from 2021 to 2023, by Country (in Billions), Statista (Jun. 24, 2024), https://www.statista.com/statistics/1287159/app-downloads-by-country/#:~:text=In%202023%2C%20mobile%20apps%20in,generated%20approximately%2012.6%20billion%20downloads (“In 2023…, users in the United States generated approximately 12.6 billion downloads.”).

[86] Google Search decision at 32.

[87] See id. at 33 (“Google’s search share on Windows devices is 80%, with most of the queries flowing through the Chrome default.”).

[88] Id. at 221 (emphasis added).

[89] Id. (emphasis added).

[90] Id. at 209 (quoting United States v. Dentsply, 399 F.3d 181, 196 (3d Cir. 2005) (“Dentsply”)).

[91] Id. at 236 (quoting Dentsply, 399 F.3d at 194).

[92] Id. at 53-54 (emphasis added).

[93] See Desktop Operating System Market Share in United States Of America — July 2024, StatCounter (last accessed Aug. 12, 2024), https://gs.statcounter.com/os-market-share/desktop/united-states-of-america/#monthly-202001-202408 and Operating System Market Share in United States Of America — July 2024, StatCounter (last accessed Aug. 12, 2024), https://gs.statcounter.com/os-market-share/all/united-states-of-america/#monthly-202001-202408.

[94] Google Search decision at 103.

[95] Id. at 121.

[96] Id. at 122. Although it currently offers Google Search as the default, it need not, and this is subject to a different distribution deal than the Android one.

[97] Id. at 207.

[98] See id. at 209 (“In 2020 only 5.1% of all search queries on iOS devices went to a rival GSE through a non-default access point…. Most non-default queries still go through Google.”).

[99] Id. at 32.

[100] Google Search decision at 205.

[101] Notably, there is not, in fact, zero distribution of Bing on Android devices. Edge is currently the eighth-most downloaded free communications app on the Google Play store and the 189th most downloaded free app overall. See Microsoft Edge Daily Category Rankings: Google Play — Jul 12, 2024 – Aug 10, 2024 — All Categories — US — All Types, Sensor Tower (last accessed Aug. 13, 2024), https://app.sensortower.com/category-rankings?os=android&app_id=com.microsoft.emmx&start_date=2024-07-12&end_date=2024-08-10&countries=US&category=communication&category=application&category=all&chart_type=free&chart_type=grossing&device=android&hourly=false&selected_tab=charts&date=2024-08-10&summary_chart_type=topselling_free.

[102] See Werden, supra note 71.

[103] Google Search decision at 201 (emphasis added, except the term “no price” in the quote from Eddy Cue).

[104] Werden, supra note 71.

[105] Google Search decision at 201.

[106] It should be noted that “no price” perhaps doesn’t really mean “no price”; perhaps it means “no price relative to Google’s payment.” Maybe if Google is prohibited from paying anything, there would be some price that Microsoft could pay relative to Google’s legally mandated $0 that Apple would accept. Maybe. That’s not actually what this says, but I suppose it is possible. Indeed, this is actually what the decision should have assessed—whether the size of Google’s payments, not the fact of its defaults, was sufficient to effectively foreclose access by Microsoft for a product that was of demonstrably lower quality, but maybe not that much lower quality. To make this stark, if Google paid distributors $1 per year instead of billions, would we accept the argument that the default nature of those deals foreclosed competition? Despite Eddy Cue’s and the court’s hyperbole, I think it extremely unlikely we would accept such a claim at face value.

[107] Although it must be noted that, at least on mobile, the court concluded that “Google’s distribution agreements did not cause Microsoft’s past underinvestment in search.” Google Search decision at 238. So, it’s not clear that Google’s deals really did deprive Microsoft of scale.

[108] See Verizon Commc’ns Inc. v. Law Off. of Curtis V. Trinko LLP, 540 U.S. 398, 407-08 (2004) (“Firms may acquire monopoly power by establishing an infrastructure that renders them uniquely suited to serve their customers. Compelling such firms to share the source of their advantage is in some tension with the underlying purpose of antitrust law, since it may lessen the incentive for the monopolist, the rival, or both to invest in those economically beneficial facilities.”).

[109] Google Search decision at 268 (quoting Trinko, 540 U.S. at 408) (emphasis added).

[110] I will just add that arguments by competitors (naturally) and neo-Brandeisians (same) for breaking up Google by forcing it to divest Chrome or Android run squarely into the holding in Microsoft. Whether or not it was appropriate for the court to base its holding on Microsoft’s “edentulous” causation standard, having done so, that standard does not support a divestiture remedy: “In devising an appropriate remedy, the District Court also should consider whether plaintiffs have established a sufficient causal connection between Microsoft’s anti-competitive conduct and its dominant position in the OS market…. [S]tructural relief, which is ‘designed to eliminate the monopoly altogether . . . require[s] a clearer indication of a significant causal connection between the conduct and creation or maintenance of the market power.’ Absent such causation, the antitrust defendant’s unlawful behavior should be remedied by ‘an injunction against continuation of that conduct.’… [W]e have found a causal connection between Microsoft’s exclusionary conduct and its continuing position in the operating systems market only through inference.” Microsoft 253 F.3d at 106-07 (quoting Areeda & Hovenkamp) (emphasis in original).

I. Introduction

Atlanta’s Mercedes-Benz Stadium in 2018 became the first major sports venue in the United States to switch to a fully cashless payment system. At the end of the new payment model’s first year of operations, the stadium reported that wait times had fallen by 20 to 30 seconds and per-capita food and beverage sales had risen by 16%, while saving more than $350,000 in operating expenses.[1]

Many other stadiums have since followed the Mercedes-Benz example,[2] and a growing number of restaurants and retail outlets are likewise going cashless. While some of these decisions were precipitated by COVID-19, the trend predated the pandemic and has continued in its wake, driven by a desire to reduce wait times and other costs associated with cash, such as counting and depositing it at the bank and mitigating the risk of robbery.[3]

Other merchants are keeping cash payments, for now, but many are no longer accepting personal checks. Target announced in early July that it would cease accepting checks July 15.[4] Several others—ranging from Whole Foods to Old Navy—recently have announced similar policies. The reasons for dropping checks are similar to those for cash: the cost of acceptance outweighs the benefits. Check transactions take much longer than cash, card, or mobile payments, and they come with a significant risk that the check will “bounce.” For Target, however, the final nail in the coffin was the “extremely low volumes” of check writing, which meant it no longer made sense to maintain check-acceptance facilities, which come with fixed costs.

Merchants are meeting their customers where they find them. In a 2022 Pew poll, 41% of Americans said they didn’t use cash in a typical week, while only 14% said they used cash for most or all their purchases—down from 24% in 2015.[5] These numbers are consistent with a Gallup poll the same year that found only 13% Americans stated that they used cash for most purchases.[6] It’s possible that the decline in cash use was exaggerated by the COVID-19 pandemic; a more recent Forbes poll found that 22% said they use cash most often for making purchases.[7] Nonetheless, the same Forbes poll found that 70% of Americans use cards most often, while 7% used digital wallets and 1% said they most often use buy-now-pay-later schemes.[8]

The Federal Reserve Board’s Diary of Consumer Payment Choice has also found a consistent decline in the proportion of payments made using cash. In volume terms, cash has largely been replaced by credit and debit cards (Figure 1).[9]

FIGURE 1: Share of Payment-Instrument Use for All US Payments (2016-2022)

SOURCE: Federal Reserve Board Diary of Consumer Payment Choice

Moreover, according to surveys by Pew (Figure 2), Americans at all income levels have reduced their use of cash and increased their use of cashless payments over the past decade.[10] In 2015, those with an annual income of $30,000 or less made substantially all their purchases using cash, while only 15% of that group made no purchases using cash. By 2022, the proportion only using cash had fallen to 30%, while the proportion not using cash at all had risen to 24%. In all other income groups, the proportion not using cash is now higher than the proportion only using cash.

FIGURE 2: Proportion of US Adults Who Use Cash for Their Purchases

SOURCE: Pew Research Center

Clearly, consumers have a strong and growing preference for electronic payments in general and cards in particular. But stories of the death of cash are premature. A recent YouGov poll found that, when asked to list all payment methods used in the past 30 days, cash was the one used by the largest proportion of respondents (67%).[11] This result does not necessarily contradict the other surveys: it is likely that most people who use cash do so only intermittently and for smaller purchases.

Despite this strong and growing preference for electronic payments by both merchants and consumers, there remains some confusion about the costs and benefits of different payment methods. The purpose of this white paper is to summarize the existing literature on the relative costs of cash, other paper-based payments (primarily checks), and electronic payments.[12]

In short, the evidence shows that, when all costs and all parties to a transaction are considered, electronic payments (debit cards, credit cards, and mobile payments) are more cost-effective than cash for most transactions. The main reason for this is that electronic payments enable consumers to spend more than they have in their wallet, which results in “ticket lift” for merchants. Card rewards, including cashback and merchant-specific loyalty programs, further increase this ticket lift.[13] In addition, “tap-and-pay” contactless payments can reduce the time it takes to tender payment relative to cash, especially when cash payments are eliminated altogether. This increases throughput, improving the customer experience and reducing labor costs. Finally, electronic payments enable merchants to sell online, including for in-store pickup.

It should be noted that this is not an argument for eliminating cash, which is likely to continue to play an important (if smaller) role in payments for many years to come. Rather, it is intended to offer a more balanced perspective on the role of cash and electronic payments in retail and other merchant settings—and to consider the implications for payments regulation.

A. Organization of the Review

Payments typically involve at least three parties: a seller, a buyer, and a bank. Where the buyer and seller use different banks, there will be at least four parties (unless one of those parties has chosen the self-custody option—also known as “cash under the mattress”). Depending on the payment method used, there may be other parties involved; for example, card payments may involve other processors, while cash payments may involve security companies moving physical cash to and from the merchant’s bank.

Each payment also typically involves a series of actions. For example, when making a purchase at a store, after items have been recorded at the register, the cashier tells the customer the total that is owed; the customer then proffers a means of payment (typically card, cash, or mobile); and the cashier processes the payment. In the case of cash, this will likely include calculating and returning change; for a chip-based card, it may involve dipping and either entering a PIN or rendering customer signature; or, if it is a contactless payment (card or phone) and the amount is below the floor limit, the customer may simply tap and go.

Most studies of the cost of payments use, in part at least, a version of “activity-based costing” (“ABC”) that seeks to account for all the costs associated with a particular payment type by assessing all of the associated activities.[14] There are, however, significant differences among the studies, both in what types of costs are included and how those costs are allocated. Broadly speaking, studies can be divided into those that focus exclusively on one party in the payment system (merchant, bank, or consumer) and those that seek to account for the costs to all (or, at least, most) parties—and hence to society. Reflecting these differences, the paper is organized as follows:

Section II reviews partial-cost studies, including those that focus exclusively on costs to merchants and costs to consumers.

Section III reviews social-cost studies that seek to evaluate the costs and benefits of different payment systems more broadly.

Section IV offers some conclusions.

II. Partial-Cost Studies

Many studies of the cost of payments focus primarily, if not exclusively, on the costs incurred by one part of the payment system—usually merchants. This section considers those studies, looking first at merchants, then consumers, then banks.

A. Merchant-Cost Studies

While merchant-cost studies are inherently narrow in scope and should not, by themselves, form the basis of public policy, they can offer valuable insights. For example, a 1983 study by the Federal Reserve noted that:

Many retailers tend to view the costs of handling cash transactions as equivalent to the cost of doing business—a sales clerk, for instance, must be on hand to conduct transactions of whatever type. Thus there is a tendency to regard the marginal cost of selling for cash as zero, but this view should not be adopted without critical examination.

There are many elements of cost associated with the handling of a sales transaction. Some costs may be higher for check or credit card transactions, but others may be higher for cash.[15]

The report went on to list many of the costs that should be considered, including the time to conclude the transaction, security costs, and counterparty risk.[16] Nonetheless, the study concluded that cash was generally less costly for merchants than other forms of payment, including cards, in part because it assumed that the use of payment cards does not result in a net increase in sales.

1. Robert M. Grant’s pioneering study

Another study published in 1983 (but undertaken prior to the Federal Reserve study) considered the costs of several different retail-payment methods in the United Kingdom: “Cash; Cheques; Bank credit cards (Access and Visa); Travel and entertainment (T&E) credit cards (American Express and Diners Club); and In-house credit accounts.”[17] The line items considered in this study can be seen in Table 1. Of note, in contrast to the Federal Reserve study, author Robert M. Grant concluded that, while the direct cost of credit cards and in-house credit were higher than the direct costs of cash and checks, these costs were more than offset by increases in sales, which the author accounted for as a reduction in the unit cost of overhead.[18] One reason Grant found such significant increases in sales associated with payment cards and store credit was that such payments represented 11% of sales in stores that took credit, compared with about 6% for all stores. Grant therefore reasonably assumed that between 20% and 30% of sales made using credit constituted “additional” sales.[19]

TABLE 1: Average Cost of Payment Methods as % of Retailer’s Revenue (1981)

SOURCE: Robert M. Grant

2. Accounting for ‘ticket lift’ and increased throughput: Layne Farrar (2011)

Many subsequent studies have sought to assess the relative costs to merchants of accepting different forms of payment. Unlike Grant’s original study, however, few of have attempted to account for the effect the method of payment might have on demand. One that did take such an approach was a 2011 study by Anne Layne-Farrar, who found that the use of payment cards were associated with higher per-transaction (“ticket”) amounts; this is known as “ticket lift.”[20]

Layne-Farrar looked at the costs and benefits of different payment methods at a range of different retail outlets: quick-serve restaurants (QSR), supermarkets, discount retailers, retail-gasoline outlets, and travel retailers (stores at train stations and airports). She begins her account of the QSR analysis with the following observation:

In 1998, Sonic Inc., an Oklahoma City based drive-in chain, became one of the first QSRs to accept cards at its 2,200 restaurant locations. According to an article published three years later, in 2001, the increasing relative costs of handling cash as compared to card payments was the primary motivation for Sonic. Technological advances over time have lowered the network and equipment costs of processing card transactions while the costs of handling cash appear to have remained flat. Sonic found that customer orders (tickets) paid by card were 80 percent higher than cash tickets. In other words, although Sonic decided to accept cards in order to lower its cash handling costs, it found direct benefits from card acceptance in the form of dramatically higher sales.

KFC began accepting cards in 2001, three years after Sonic. In contrast to Sonic, as its motivation KFC cited specific benefits expected from cards, rather than solely the savings derived from reduced cash handling. Specifically, KFC began accepting payment cards as a way to sell its higher priced group meals, such as large buckets of chicken with side dish containers and packages of biscuits.[21]

Following this prelude, Layne-Farrar provides a detailed analysis of the various costs associated with accepting different payment methods, the average sizes of transaction made with those methods, and the benefits of increased revenue resulting from the use of payment cards relative to cash.

Layne-Farrar found that, in addition to ticket lift, the use of payment cards reduced average transaction time, resulting in increased throughput of customers, which resulted in a further increase in revenue. Table 2 provides a summary of Layne-Farrar’s estimates of the costs and benefits from the use of different payment methods for an average transaction assuming ticket lift of 10%, which Layne-Farrar saw as a low amount (she also calculated the effects of 20% ticket lift).[22] While cash transactions were less costly to process at the average ticket size, when taking into account ticket lift and throughput improvements, the net benefits per-transaction were significantly higher for debit cards.

TABLE 2: Per-Transaction Costs and Benefits for QSRs (2011)

SOURCE: Layne-Farrar

Tables 3-7 are Layne-Farrar’s estimates of the net benefits of using different transaction methods at big-box discount retailers, supermarkets, gas-station retail, non-fuel convenience stores, and travel retail.[23] In each case, only the low-ticket-lift option is given (the net benefits of the high-ticket-lift option are always greater); nonetheless, the net benefits of debit cards exceed those of cash and, where checks are evaluated, those as well. The clear conclusion of this work is that, by the time Layne-Farrar undertook her analysis in 2011, acceptance of debit cards generated per-transaction net benefits for merchants that exceeded those of cash.

TABLE 3: Per-Transaction Costs and Benefits for Big-Box Discount Stores

SOURCE: Layne-Farrar (2011)

TABLE 4: Per-Transaction Costs and Benefits for Supermarkets

SOURCE: Layne-Farrar (2011)

TABLE 5: Net Benefit Per Fuel Transaction ($) for Gas-Station Retail

SOURCE: Layne-Farrar (2011)

TABLE 6: Net Benefit Per Non-Fuel Transaction ($) for Convenience Stores

SOURCE: Layne-Farrar (2011)

TABLE 7: Net Benefit Per Fuel Transaction ($) for Travel Retail

SOURCE: Layne-Farrar (2011)

3. Economists Inc.

In 2014, Economists Inc. carried out a study similar to Layne-Farrar’s, but went into even more granular detail regarding the merchants’ processes for accepting payments, focusing on five merchants: a fast-food restaurant, a full-serve restaurant, a gas station, a convenience store, and a small independent grocery store.[24] They also included both credit and debit cards in their analysis.

Table 8 shows tender times for cash and credit (including debit run as “credit”) at the five merchants.[25] It is noteworthy that both the mean and median tender times at the fast-food restaurant and grocery store were slightly lower than for cash, whereas the in-person tender times at other merchants were slightly higher for credit.

These differences may reflect different rates of throughput and associated investment in technology and training. At higher-throughput merchants, such as grocery stores and fast-food restaurants, there are likely greater returns on investments in technology that integrates checkouts with point-of-service (POS) machines, for example, such that the checkout operator does not need to re-input information to the POS machine.[26]

The tender time for self-service gas pumps is, perhaps unsurprisingly, considerably shorter than the time when paying in-store, as this avoids the shoe-leather time involved with walking to the store to pay and back.

TABLE 8: Tender Times at the Merchants Studied by Economists Inc

SOURCE: Economists Inc. (2014)

Like Layne-Farrar, Economists Inc. found that the use of payment cards resulted in significantly higher purchase amounts relative to cash. This ticket lift can be seen in Table 9, which shows that, for every establishment, the minimum, maximum, mean, and median payments are nearly all higher for payments made using credit cards (or debit run as “credit”) than for cash.[27] (The one exception is the minimum for in-store gas and joint sales at the gas station.)

Economists Inc. also reported that merchants themselves had indicated that, when they began accepting card payments, they noticed a significant increase in sales.[28]

TABLE 9: Amounts Spent When Using Cash or Credit

SOURCE: Economists Inc. (2014)

4.  Ticket lift and increased throughput from contactless

Contactless payments use RFID to transmit tokenized payment information from a card or from a smartphone. While the first contactless payment cards were introduced in the mid-1990s, and an EMV contactless standard was first developed in 1996, their uptake by both card issuers and merchants was initially low, especially in the United States.[29] However, the vast majority of U.S. debit and credit cards now support contactless payments, more than 150 million Americans have used contactless payment apps on their smartphones and, as of 2020, 58% of U.S. merchants accepted them.[30] Evidence from other jurisdictions suggest this switch to contactless is likely to result in both ticket lift and increased throughput.

A 2020 study by David Bounie and Youssouf Camara looked at the effects on 2018 sales of the shift to contactless payments at 275,580 merchants in France.[31] The researchers found that merchants who accepted contactless payments had card sales that were 15.3% higher, on average, than merchants who did not accept contactless payments.[32]

A 2022 study by Sumit Agarwal, Wenlan Qian, Yuan Ren, Hsin-Tien Tsai, and Bernard Yeung looked at the effect of the introduction of “quick-response” (QR) code mobile payments in Singapore in 2017.[33]  The researchers found that “monthly business creation among business-to-consumer industries increased by 8.9% more than among business-to-business industries.”[34] Meanwhile, use of mobile payments tripled, while use of automated-teller machines (ATMs) fell dramatically. Of particular note, consumers increased their credit-card spending by 3.3%, driving most of the increase in consumer spending that led to the increase in B2C business formation.[35]

5. Fumiko Hayashi’s comparison of debit and cash

In a 2021 study, Federal Reserve Bank of Kansas economist Fumiko Hayashi compared the cost of merchant acceptance of cash and debit cards in different countries. Table 10 reproduces the summary table in her study.[36]

TABLE 10: Merchant Acceptance Costs in Selected Countries, Various Years

SOURCE: Hayashi

Hayashi’s data for the Unted States are based on two studies: the first is the Food Marketing Institute (FMI) study from 2000 that formed the basis of the GHL analysis discussed at-length in Section III of this white paper, and the second is a Bank of Canada study that Hayashi coauthored with Marie-Hélène Felt, Joanna Stavins, and Angelika Welte[37].

There are several problems with both of these studies: first, as GHL point out, the FMI study does not consider costs associated with either theft or counterfeit when calculating the costs of cash.[38] Second, the “2018” data from the United States is based on 2015 survey data on the cost of acceptance in Canada that has been adjusted by using certain U.S.-specific data.[39] This was done despite Hayashi herself arguing in a previous paper coauthored with William Keeton that such practices are inappropriate, especially where the United States is concerned, noting that:

The danger of relying on other countries’ cost studies is particularly apparent for the United States, where checks and credit cards are used on a larger scale and more parties are involved in the payments process.[40]

Among other things, these differences in the mix of payment type will have an effect on the optimal fees charged by parties to the system, due to the two-sided nature of payments.[41] For example, the “interchange” fees retained by issuing banks might be higher in the United States due to issuers seeking to increase their share of the payments market by offering incentives for consumers to switch from checks to cards (for example, in the form of cashback or other rewards).[42] (These fees and incentive payments are essentially transfers within the system, not a net cost. Indeed, to the extent that they result in a shift to more socially efficient modes of payment, they are, on net, beneficial.)

At a more general level, Hayashi’s survey suffers from inappropriate aggregation. That is to say, whereas it is likely true that debit-acceptance costs exceed those of cash for some proportion of payments, it is unlikely to be true for all merchants and all amounts. Indeed, the study of the Netherlands that Hayashi used notes that, although the average cost of acceptance for debit was notionally higher than for cash in 2002, the “breakeven point” (when taking into account costs for both retailers and banks) was €11.63, which was below the average ticket size for debit (€44).[43]  Meanwhile, in 2009, the merchant-acceptance cost for debit was on average lower than for cash; the breakeven point had, however, fallen to €3.06, not zero.[44]

Finally, and related to the problem of inappropriate aggregation, Hayashi fails to consider the ticket-lift effect identified by Grant and confirmed by Layne-Farrar, Economists Inc, Bounie & Camara, and others.

6. Problems with generalizing across jurisdictions

There is broad agreement that the cost of using different payment types for a transaction of a given amount varies significantly across jurisdictions and sectors. For example, in their 2003 study of the Norwegian payment system, using data from a 2001 survey, Olaf Gresvik and Grete Øwre found that Norwegian banks operated payments at a loss (although technological improvements had decreased the cost of the payment system over time).[45] By contrast, at least until 2010, debit payments in the United States were used to cross-subsidize checking accounts.[46]

Within the United States, costs can vary considerably from state to state. For example, a cashier in San Jose, California, might earn $20/hour, while a cashier in Louisville, Kentucky, might earn $10/hour. If both cashiers take the same amount of time to tender payments, the tender costs in San Jose could be twice what they are in Louisville. Even within a state, salaries can vary somewhat, with cashiers in Smyrna, Tennessee earning an average of $14.36, according to Indeed.com, while those in Jackson, Tennessee earn only $11.45.[47]

But that is far from the whole story. While salaries might be higher in, e.g., California, the average ticket is also likely to be higher. Thus, the number of sales required per-dollar of income will be lower. This means that the average time to tender any specific dollar amount will be lower in California. How this affects the net costs to tender per-dollar depends on the ratio of spending to salaries, which is not only location-specific, but also merchant-specific.

7. Tender-time studies

A subcategory of merchant-cost studies focuses more narrowly on the time taken to tender a payment. These are essentially “time and motion studies” of the parts of the retail-checkout process that involves payment. Table 11 provides some examples of such studies. (These studies also have implications for consumers for whom the time taken to process payments and, relatedly, the time spent queuing has an opportunity cost, as discussed in Section III.)

TABLE 11: Tender Time Studies (Time to Complete Tender in Seconds)

[48][49][50][51][52] SOURCE: Polasik et al., adapted by author

8. Studies of the cost of accepting cash

A final subcategory of merchant cost-of-acceptance studies looks in greater detail exclusively at the cost of accepting cash. A relatively recent example in the U.S. context was produced by IHL Group, which undertook a very detailed activity-based costing.[53] Specifically, IHL identified nine activities associated with the cost of cash:

1. Start/Rebuild Drawer-Functions related to opening drawer from initial deposit to rebuild of for next cashier.

2. Closing Drawer-This includes functions related to closing out a drawer. Time for cashiers, managers, or cash office personnel to count and reconcile the drawer with POS or cash register totals.

3. Pickups-This is inclusive of pickups during a shift for too much cash in the drawer or bills that are large denominations.

4. Change Orders-Cost associated with a cashier requiring change throughout the shift

5. Audits/Discrepancies– These are costs of redoing counts, auditing tills, and time associated with recounts for any discrepancies.

6. Prepare/Coordinate Deposits-Costs associated with preparing or coordinating deposits.

7. CIT/Deposit Costs-These are costs of Cash In Transit companies (armored trucks) or cost for managers or other employees to go to bank to make the deposits.

8. Bank Charges-These are charges surrounding bank fees. This includes statement fees, reconciliation, cash value fees and change orders among other things.

9. Cash Shrink-This is the cost of theft, fraud or other cash loss activities. This is cash that just disappears in the process. For this study we used previous data from other studies as a value.[54]

IHL then undertook several hundred interviews and numerous modeling exercises to establish the amount of time associated with actions 1-6 and, using wage estimates for each relevant job in each relevant location, they calculated the total cost for each such activity. They then added costs 7-9 to arrive at a total for each business.

IHL found that the average cost of cash across all segments was 9.1% of the revenue of the businesses studied.[55] Of this, the largest component was “close drawer,” accounting for about 40% of the total, whereas bank charges were only 4.3%.[56]

Table 12 shows the range of costs incurred by different businesses for accepting cash.[57] Even the lowest of these (food/grocery) has a cost of cash (4.7%) that is higher than the highest merchant-discount rates charged by acquiring banks.

TABLE 12: Cost of Cash by Segment

SOURCE: IHL

B. Consumer-Cost Studies

While most partial-cost studies focus on merchants, some look at consumers. Several studies have shown that transaction value is a key determinant of the method of payment. One explanation is that consumers want to avoid receiving a significant amount of change in the form of coins. Thus, Heng Chen, Kim P. Huynh, and Oz Shy of the Bank of Canada found that “a significant number of cash users … switch to paying with debit or credit cards at transaction values marginally above $5 and $10.” [58] They attribute this to “the burden of receiving coins as change associated with the currency denomination structure.”[59] (The Bank of Canada withdrew the $1 note in 1989 and the $2 note in 1996, leaving $5 as the smallest denomination bill.[60])

C. Bank-Costs Studies

A third category of partial-cost studies considers the costs to banks. The most notable example of this is the study undertaken by Olaf Gresvik and Grete Øwre mentioned above, which formally introduced the concept of activity-based costing (ABC) to payments and applied it to the processing of payments by Norwegian banks, based on a 2001 survey.[61]

III. Social-Cost Studies

The second category of payment-cost studies seeks to evaluate not only the costs to merchants, but the costs to society as a whole. An early focus of such studies was checks, which at the time were the dominant form of noncash payment. A 1990 paper by David Humphrey and Allen Berger considered the divergence between the private and social costs of payments in the United States, with a particular focus on checks, which the authors argued were overused because payor businesses (in particular) benefitted from the float[62] associated with checks that had not yet cleared.[63]

A 1996 paper by Kirstin Wells, using data from 1993, compared the social cost of checks with that of automated-clearinghouse (ACH) payments and concluded that, in contrast to the 1987 data used by Humphrey and Berger, there was not a significant difference in float cost between using checks and ACH.[64] Indeed, Wells estimated that the value of float fell 91.3% from an average of $1.04 to $0.09, mainly due to increases in the efficiency of check processing.[65]

Starting in the early 2000s, the focus of social-cost-of-payments studies shifted to retail payments and, specifically, to the relative cost of cash and payment cards (although checks were also evaluated in some studies in the 2000s, as they were then still quite widely used). This section focuses on such studies.

A. Garcia-Swartz, Hahn, and Layne-Farrar

In a study originally published by the AEI Brookings Joint Center for Regulatory Studies in 2004, and subsequently updated and published in the Review of Network Economics in 2006, Daniel Garcia-Swartz, Robert Hahn, and Anne Layne-Farrar (“GHL”) undertook arguably the first and still one of the most comprehensive social-cost (or benefit-cost) assessments of retail payments.[66]

1.  Merchant costs

The starting point for GHL was a study undertaken by the Food Marketing Institute (FMI) in 1998 that sought to calculate the direct costs of accepting various payment types, namely: cash, checks (verified and nonverified), credit cards, and debit cards (signature and PIN). These direct costs (as categorized by the FMI) were:

1. “Tender-time”: this is the cost of the time spent by cashiers processing a transaction after ringing up all the items (this was based on another FMI study, from 2000, at which time cash remained quicker than card, and wages from the 2002 Bureau of Labor Statistics survey);

2. “Deposit preparation”: this is the cost of the time taken to prepare a cash deposit (e.g. counting cash, reconciling the register drawer, preparing a deposit slip, etc.);

3. “Bank charges”: these are the explicit fees charged by banks, such as a deposit fee for cash and checks, or the merchant discount rate for cards;

4. “Other direct costs”: these include costs such as using armored cars to transport cash, collection costs and losses on “bounced” checks, and credit card chargebacks.[67]

Table 13 reproduces GHL’s summary table showing these initial calculations.[68] As can be seen, based only on these costs, the per-transaction cost of cash is lower than that of other payment types. The amount tendered in the average cash transaction is, however, much lower than the amount tendered in other transaction types. When scaled to $100 of sales, cash is the second-most costly for the merchant, after credit cards, while verified checks are the least costly.

GHL then note that the FMI analysis omits two potentially significant costs: (1) theft and counterfeit losses for cash and (2) float loss for all payment types.

TABLE 13: Grocery Stores’ Per-Transaction Processing Costs for Various Payment Instruments, Modified ($), (2003)

SOURCE: Garcia-Swartz, et al.

2. Consumer costs

GHL identify the following payment-related consumer costs:

1. Processing time: this is the opportunity cost of the consumer’s time while waiting for the transaction to be processed.

2. Queue time: this is assumed to be equal to processing time

3. Explicit price: this is the explicit bank service charge associated with withdrawing cash and processing cheques and debit transactions.

4. Implicit price: this is the “shoe-leather” costs of obtaining cash (i.e. the opportunity cost of the time taken to travel to/from an ATM and withdraw cash).

5. Seigniorage: the profit made by the central bank from printing money (basically the difference between the face value of the currency and the costs of production)

3. Bank costs

In addition to merchants and consumers, banks also incur costs in the form of ATM maintenance (applies to cash); production of cards (applies only to cards); transaction processing (applies to all payment methods); and card rewards (applies mainly to credit cards, but also to a lesser extent to debit cards).

4. Central bank costs

A fourth set of costs arise from the involvement of the central bank in producing and processing banknotes and coins, as well as in processing checks (approximately $0.0015 and $0.03, respectively, for a transaction of $11.52). The cost of check processing, however, is recovered from banks, is therefore included in banks’ processing costs.

TABLE 14: Grocery Stores’ Per-Transaction Processing Costs for Various Payment Instruments, Cash Transaction of $11.52 ($), (2003)

SOURCE: Garcia-Swartz, et al. (2006)

Putting these all together, GHL calculate the total marginal cost for transactions of $11.52 (the average size of a cash transaction) and $52.24 (the average size of a check transaction).[69] Table 14 replicates the figures for the $11.52 transaction.[70] Once double-counting is eliminated, the social cost of paying with cash and card are roughly the same.

Meanwhile, for grocery-store transactions of $54.24, debit cards have the lowest social cost ($0.94 for PIN-authorized transactions and $1.00 for signature-authorized transactions), followed by verified check ($1.08), credit card ($1.32), nonverified check ($1.40) and, finally, cash ($1.98).[71]

5. Accounting for (social) benefits

But the story does not end there. GHL note that there is a range of benefits arising from the use of certain payment methods that, at least partially, offsets these costs. For consumers these include:

1. Float: while checks, credit and charge cards impose float costs on merchants, they provide consumers with some float (in the case of credit cards used purely transactionally this may be quite large).

2. Credit option: the option to use the credit function of credit cards

3. Record keeping: electronic transactions and even cancelled checks provide a record that is valuable to many consumers.

4. Cashback at POS saves a trip to the ATM (but the amount is limited)

5. Rewards cards (mainly credit) provide marginal benefits about twice their cost

6. Discover cards provide marginal benefits equal to their marginal cost

7. Privacy, the exclusive domain of cash, offers users significant benefits [albeit at a cost in terms of dramatically increased costs of recourse] [72]

For banks, the benefits include a small amount of float (this is basically the counterpart of the float costs incurred by consumers who use cards) and processing revenue, which is part of the amount earned by banks for processing transactions. Meanwhile, central banks earn seigniorage and a small amount for processing transactions (which, as noted earlier, is netted out).

The marginal benefits associated with a typical check-size transaction (for 2003, when such transactions were more common) of $52.24 are shown in Table 15. The payment methods providing the greatest marginal benefit are credit ($1.61 per transaction) and cash ($0.92 per transaction).

By adding the marginal benefits to the marginal costs, GHL are then able to calculate the net social marginal cost associated with the different payment types. For a transaction of this size, in 2003, the authors estimate that credit cards would have the lowest net social cost, followed by PIN debit, signature debit, verified check, cash, and nonverified check.

TABLE 15: Adding Benefits to Grocery-Store Transactions of $52.24

SOURCE: Garcia-Swartz, et al.

GHL use the same methodology to calculate the net social marginal cost for transactions at two other types of merchant: discount stores and specialty electronics stores.

TABLE 16: Net Social Marginal Cost for Transactions at Discount and Electronics Stores

SOURCE: Garcia-Swartz, et al.

In the case of discount stores, for purchase amounts of $15.49, the social cost of cash, bank credit, American Express, and debit are about equal. Meanwhile, for larger amounts, cash is more costly than all other payment methods except check.

6. Sensitivity analysis

GHL then undertake a “sensitivity analysis,” in which they adjust some of the parameters of their estimates. For example, increasing the number of people queueing for checkout increases the net social cost of checks relative to cash quite significantly and increases the net social cost of cards slightly. The result is that, for example, if there are three people queuing at GHL’s average grocery store spending the average cash amount at such a store ($11.52), cash becomes marginally socially beneficial.[73] Payment cards, however, remain superior at the amounts typical for those purposes ($33 for signature debit, $41.05 for PIN debit, and $44.59 for credit). In other words, the “break even” point for those payment types shifts to the right.

7. GHL’s conclusions

It is worth repeating the conclusions GHL drew from their research (in a separate study published alongside the detailed analysis described above):

First, transaction size assumptions are critical in analyzing payment-processing costs. At smaller transaction sizes, the net social marginal cost of all payment instruments – paper and electronic alike – are remarkably similar. No one instrument stands out as more socially efficient. At larger transaction sizes, however, significant differences emerge. For grocery store transactions, electronic payments are considerably less costly on net for society than paper methods. Yet another pattern emerges for the larger transactions conducted at electronics stores. Here credit cards with a large proportion of reward cardholders have the lowest net social marginal cost. This pattern is consistent with observed behavior: namely that cash use dominates smaller transaction sizes but drops precipitously as transaction size increases.

Second, retailer type influences the individual cost elements and thus affects private cost calculations. Since the distribution of transaction sizes differs across venues, this result follows naturally from our first finding.  Added to the transaction size effect are apparent differences in merchant costs, such as point of sale time and back-office processing costs.

Finally, and most importantly, the relative merits of different payment methods change significantly when all parties are counted and benefits are included. Merchant studies have found that paper methods are the cheapest for merchants. This is confirmed in our study of the distribution of private costs and benefits. But what is cheap for merchants is relatively expensive for other parties to a transaction. Certain parties, especially consumers, receive considerable benefits from payment cards, which tip their net private costs in favor of that method of payment.[74]

B. Shampine Critique of GHL and GHL’s Response

Allan Shampine undertook a critical review of GHL, questioning their assumptions regarding the amount of time taken to obtain cash from ATMs, the value of card rewards, the range of nonpecuniary benefits that consumers derive from different payment methods, and the appropriateness of some other cost categories, such as seigniorage.[75] He then applied his own sensitivity analysis—which, by making very different assumptions, found that, for certain payment sizes that GHL had identified as lower cost for cards, cash may be lower cost.

GHL responded by noting that, of course, if one makes different assumptions, it is possible to achieve different outcomes.[76] But where the assumptions that GHL made were at least supported as far as possible by empirical evidence, most of the adjustments Shampine made had no empirical basis and should therefore not be treated as reliable. Moreover, as they note, there are very significant individual differences among consumers, merchants, and banks regarding the costs and benefits of any particular payment type.

Ironically, one of the GHL assumptions that Shampine criticizes as insufficiently generous is the privacy benefits of cash. While it is no doubt true that some consumers benefit from the anonymity of cash payments, for most consumers, that is not the main priority. Moreover, there is a social cost to privacy when consumers use cash to engage in illegal activity. Indeed, security is typically more important, and since cards are far more secure than cash, it is possible that the relative benefits are tipped even more toward cards for most consumers.

C. Other Social-Cost-of-Payments Studies

Numerous researchers have undertaken studies to estimate the social cost of payments in other jurisdictions. While several of these studies seek to account for costs borne by consumers, none of those we identified has been as comprehensive or detailed as GHL.[77] Specifically, none adequately account for the social benefits of different payment modes. Also, to our knowledge, none of them focus on the United States. Nonetheless, the studies have introduced some valuable insights. Perhaps most notable is the importance of differentiating fixed and variable costs (although, as discussed below, these costs change over time).

1. Fixed v variable costs, and the ‘breakeven’ point

In their analysis of the Dutch payments system, Hans Brits and Carlo Winder show that payment cards have relatively high fixed costs and much lower variable costs. As a result, in 2002, the “breakeven” point for debit transactions occurred at €11.36.[78] Above that amount, it is more socially cost efficient to pay with debit than with cash. This can be seen in Figure 3, which shows the relative cost of making a payment with cash, debit card, or “e-purse” (this last is a rechargeable smart card that can be used to pay for a range of goods and services in the Netherlands).

FIGURE 3: Breakeven Points for Different Payment Types, Netherlands (2002)

SOURCE: Brits & Winder (2005)

In a study of the private and social costs of payments in Sweden in 2002, Mats Bergman, Gabriella Guibourg, and Björn Segendorf of Sweden’s central bank found that the unit transaction costs of cash (4.6 SEK) was higher than for either credit cards (4.4 SEK) or debit cards (3.1 SEK for PIN, 3.2 SEK for signature).[79] They then sought to identify the breakeven point for each payment method, and found that debit cards became more cost-effective than cash at about 72 SEK (U.S. $7), while credit cards became more cost-effective than cash at about 160 SEK (U.S. $16).[80]

Technological improvements have reduced both the fixed and variable costs associated with card-based payments. For example, the fixed cost associated with the time taken to process a card-based transaction has generally fallen.[81] Meanwhile, both the fixed and variable costs associated with card-based fraud has fallen by more than 75% as a result of the introduction of the Europay, Mastercard and Visa (EMV) Chip.[82]

FIGURE 4: Breakeven Points for Different Payment Types, Sweden (2002)

SOURCE: Bergman, Guibourg, & Segendorf (2007)

2. Technological change lowers the breakeven point

While technological improvements have also increased the efficiency of processing cash, there are some human aspects to cash processing that are almost impossible to eliminate, and that are inherently proportional to the transaction amount.

Increased use of a payment method can create a virtuous circle for that method, while having the opposite effect for other methods. So, for example, the shift from cash to debit in the Netherlands resulted in lower average debit-acceptance costs, but increased the average acceptance costs of cash. Thus, Nicole Jonker found the breakeven point for merchant acceptance of debit in the Netherlands had fallen from the €11.36 found by Brits & Winder in 2002 to €3.06 in 2009.[83]

A similar phenomenon appears to be happening with both cash and checks in the United States, as demonstrated by the examples given in the introduction. It is reasonable to ask why this shift seems to have happened later in the United States than in Europe. Aside from “culture,” one argument is that U.S. banks implemented more-efficient check processing in the early 1990s, thereby reducing the incentives for merchants to switch. Another is that the Durbin amendment in 2010 reduced consumers’ incentives to pay with debit, and resulted in their switching from debit to credit for lower-value payments, which increased the relative cost of acceptance for merchants.

D. Partial Social-Cost-of-Cash Studies

Many other studies advertised as assessing the social costs of payments have considered the effects on a somewhat narrower range of participants. For example, a study published by the European Central Bank titled “The Social and Private Costs of Retail Payment Instruments: A European Perspective” notes:

Due to the considerable effort necessary to collect viable data on the costs incurred by all of the parties in the payment chain, the analysis focuses on the most important parties issuing authorities, i.e.:

• central banks and governments;

• banks and interbank infrastructure providers (automated clearing houses, ATM networks, etc.);

• retailers and companies; and

• cash-in-transit companies.[84]

There would appear to be a rather significant omission from this study: consumers (indeed, the ECB acknowledges this deficit). The same omission is present in many other similar studies. This is both rather odd and rather troublesome, since consumers clearly are important participants in the payment system—indeed, without them, the system would be pointless. Moreover, if GHL are correct, consumers tend to benefit more from electronic payments than from cash—indeed, cash is typically a net cost for consumers—so omitting them from the analysis seems more than a mere oversight.

IV. Conclusion

This review shows that both the partial and social costs of different modes of payments vary considerably by location, type of merchant, and over time. Nonetheless, several broad conclusions emerge:

First, retailers that accept card payments tend to experience ticket lift; many also benefit from increased throughput. As a result, retailers such as quick-serve restaurants that sell low-ticket items and might be below the “breakeven” point for cards relative to cash (if considering only the direct transaction costs) but benefit on net from adding cards because of the significant ticket lift and increase in throughput.

Second, over time, there has generally been a reduction in the “breakeven” point for electronic payments. This has likely been driven by such innovations as the EMV Chip and contactless payments, which have reduced fraud and tender-time costs, and increased benefits to all parties.

Third, while innovations in cash management have also reduced the cost of accepting cash in general, the cost of multimodal payment acceptance means that the relative cost of continuing to accept cash has increased, especially in locations where throughput is of the essence, such as ballparks and quick-serve restaurants. This has led some such merchants to drop cash acceptance.

[1] Mercedes-Benz Stadium Achieves Success in First Year of Stadium-Wide Cashless Initiative, Mercedes-Benz Stadium (Mar. 9, 2020), https://www.mercedesbenzstadium.com/news/mercedes-benz-stadium-achieves-success-in-first-year-of-stadium-wide-cashless-initaitive (“MBS’s expected first-year results have been realized, once again locking in the No. 1 spot for food and beverage including speed of service across all NFL venues for the third consecutive year.  Due to the new cashless model, roughly 95 percent of fans noticed the same or an increase in speed at concession lines and at peak times a 20-30 second reduction in wait times. Results also include an increase in food and beverage per capita numbers for close to 50 events at MBS through 2019 including a combined 16 percent increase for Atlanta Falcons and Atlanta United all while saving more than $350,000 in operational expenses… Since going cashless, more than 2.5 million guests have attended events at MBS. Of those, only 1.2 percent have used the cash-to-cards kiosks, showing that fans are bringing their own credit cards or using mobile payment options.”).

[2] See Ben Gran, More Stadiums Are Going Cashless. What Does This Mean for Your Personal Finances?, the ascent (Feb. 13, 2024), https://www.fool.com/the-ascent/personal-finance/articles/more-stadiums-are-going-cashless-what-does-this-mean-for-your-personal-finances (“America is quickly becoming a more cashless society, and sports venues are leading the charge. As of 2022, nearly all Major League Baseball ballparks had gone cashless, along with most NFL stadiums and NBA/NHL arenas like the United Center in Chicago.”).

[3] See, e.g., Benjamin Gottlieb, Why Some Restaurants in LA Are Going Cash-Free, Marketplace (Aug. 19, 2019), https://www.marketplace.org/2019/08/19/why-some-restaurants-in-la-are-going-cash-free.

[4] Nate Delesline III, Target to Stop Accepting Personal Checks, Retail Dive (Jul. 9, 2024), https://www.retaildive.com/news/target-to-stop-accepting-personal-checks/720792.

[5] Michelle Faverio, More Americans Are Joining the ‘Cashless Economy’, Pew Research Center (Oct. 5, 2022), https://www.pewresearch.org/short-reads/2022/10/05/more-americans-are-joining-the-cashless-economy.

[6] Jeffrey M. Jones, Americans Using Cash Less Often; Foresee Cashless Society, Gallup (Aug. 25, 2022), https://news.gallup.com/poll/397718/americans-using-cash-less-often-foresee-cashless-society.aspx.

[7] Ketherine Haan, People Are Twice as Likely to Spend More Money When Using Card than Cash in 2024, Forbes Advisor (May 16, 2024), https://www.forbes.com/advisor/business/software/people-twice-likely-spend-using-card-than-cash.

[8] Id.

[9] See Emily Cubides & Shaun O’Brien, 2023 Findings from the Diary of Consumer Payment Choice, The Fed. Res. Fin. Serv. (Jul. 2023), at 6, available at https://www.frbsf.org/cash/wp-content/uploads/sites/7/2023-Findings-from-the-Diary-of-Consumer-Payment-Choice.pdf (which notes that “The category ‘other’ includes payments made with pre-paid [debit], checks, mobile payment apps, money orders.”).

[10] Faverio, supra note 5.

[11] Kineree Shah, Cash Remains King – 67% of Americans Still Use Traditional In-Store Payment, YouGov (Feb. 12, 2024), https://business.yougov.com/content/48650-cash-remains-king-67-of-americans-still-prefer-traditional-in-store-payment.

[12] A survey of the available literature was conducted using EconLit, the Social Science Research Network (www.ssrn.com), the IDEAS database (ideas.repec.org), and Google Scholar. This enabled us to identify the primary methodologies used to evaluate the value, costs, and benefits of different payment technologies. Initially, we used search terms such as “cost of cash” and “cost of payments” (with and without the restrictive use of speech marks). As the research progressed, we expanded this to a range of other terms, including “speed of payments,” so as to capture a wider range of studies addressing the issues under consideration.

[13] Credit cards take this one step further, enabling consumers to spread their spending out, reducing temporary liquidity constraints without the need to arrange an overdraft or other loan, thereby further increasing ticket lift—especially for higher-value items.

[14] In a traditional activity-based costing study, the aim is to account for the costs of each activity undertaken by a business and thereby enable management to make better decisions on resource allocation, investments in innovation, and so on.

[15] Board of Governors of the Federal Reserve System, Credit Cards in the U.S. Economy: Their Impact on Costs, Prices, and Retail Sales 36 (Jul. 27, 1983), available at https://fraser.stlouisfed.org/title/credit-cards-us-economy-5331.

[16] The study notes some of the costs: “Included among the relevant cost concepts, for example, would be the time required to complete a trans­ action, which may in turn influence the number of check-out stations and sales clerks that a store needs. Credit card transactions absorb time because credit slips must be written and frequently some sort of authorization procedure undertaken. Personal checks usually trigger certain time-consuming precautionary steps, such as inspecting and copying down identification data or summoning a manager from elsewhere in the store to approve acceptance of the check. Cash transactions most likely consume less time than check or credit card transactions, but the counting of cash received, the making of change, and the stocking and replenishment of cash registers with currency and coin are cash-related activities that occupy an employee’s time. Time consumed in reconciling sales records with cash, checks, and credit slips on hand may vary with the proportion of sales transacted by each means, and from one business to another. Security-related expenses comprise a large family of costs in which further variation may be found among the different means of payment. Included in such a concept would be both direct expenses of security precautions plus an allowance for any uncovered risk associated with each transaction medium. An obvious risk, for example, is the possibility of theft. This particular risk is likely to be more pronounced for cash because the full negotiability of cash makes it an attractive target. Acceptance of personal checks entails the risk that the check may be uncollectable, because the writer may not have sufficient funds on deposit or for some other reason.  Security risks borne by operators of in-house credit card plans include the costs associated with delinquent and uncollectable accounts.” Id. at 36-37.

[17] See Robert M. Grant, Transaction Costs to Retailers of Different Methods of Payment: Results of a Pilot Study, 4(2) Managerial & Decision Econ. 89-96 (Jun. 1983).

[18] Id. at 91. Grant is one of the most-cited social scientists, with nearly 100,000 citations and an H index of 58. See Google Scholar Profile of Robert M. Grant, Google Scholar (last accessed Aug. 22, 2024), https://scholar.google.com/citations?user=CQ8P0PcAAAAJ&hl=en.

[19] Grant, supra note 17, at 93, 95-96.

[20] Anne Layne-Farrar, Are Debit Cards Really More Costly for Merchants? Assessing Retailers’ Costs and Benefits of Payment Instrument Acceptance (SSRN Working Paper Sep. 9, 2011), available at https://ssrn.com/abstract=1924925.

[21] Id. at 6.

[22] Id. at 16-17.

[23] See id. at 23, 27, 34, 35, 39.

[24] Retailer Payment Systems: Relative Merits of Cash and Payment Cards, Economists Inc. (Nov. 19, 2014), available at https://ei.com/wp-content/uploads/2015/01/Cost_of_Cash_Study.pdf.

[25] Id. at 52.

[26] Id. at 21.

[27] Id. at 5.

[28] Id. at 58-60.

[29] See Tom Akana & Wei Ke, Contactless Payment Cards: Trends and Barriers to Consumer Adoption in the U.S. (Discussion Paper 20-03, May 2020), available at https://www.philadelphiafed.org/-/media/frbp/assets/consumer-finance/discussion-papers/dp20-03.pdf.

[30] See US Contactless Payment Statistics, Finical Holdings LLC (last accessed Aug. 22. 2024),  https://finicalholdings.com/us-contactless-payment-statistics.

[31] David Bounie & Youssouf Camara, Card-Sales Response to Merchant Contactless Payment Acceptance, 119 J. Banking & Fin. 105938 (Oct. 2020).

[32] Id.

[33] Sumit Agarwal, Wenlan Qian, Yuan Ren, Hsin-Tien Tsai, & Bernard Yeung, The Real Impact of FinTech: Evidence from Mobile Payment Technology (Working Paper, NUS Business School, National University of Singapore, September 2022), available at https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3556340.

[34] Id. at 4.

[35] Id. at 5.

[36] Fumiko Hayashi, Cash or Debit Cards? Payment Acceptance Costs for Merchants, 106(3) Econ. Rev. Fed. Res. Bank of Kansas City 53 (Aug. 2021), available at https://www.kansascityfed.org/Economic%20Review/documents/8213/EconomicReviewV106N3Hayashi.pdf.

[37] See Fumiko Hayashi, Marie-Hélène Felt, Joanna Stavins, & Angelika Welte, Distributional Effects of Payment Card Pricing and Merchant Cost Passthrough in the United States and Canada (Fed. Res. Bank of Kansas City, Research Working Paper no. 20-18, Dec. 2020), available at https://www.kansascityfed.org/Research%20Working%20Papers/documents/7595/rwp20-18.pdf.

[38] See Daniel Garcia-Swartz, Robert Hahn, and Anne Layne-Farrar, The Move Toward a Cashless Society: Calculating the Costs and Benefits, 5(2) Rev. of Network Econ. 202 (2006) [hereinafter “GHL”] (“The FMI study omits… theft and counterfeit loss for cash…”).

[39] Specifically: “The total cost of accepting cash, debit card, and credit card payments is based on the Bank of Canada Retailer Survey, but by using United States–specific information as follows: merchant service charge, the average wage for cashiers and back-office workers (from the U.S. Bureau of Labor Statistics Retail Trade Earnings and Hours), cash theft and fraud as a percentage of cash sales (from the National Retail Security Survey), the fraud and chargeback rate for cards (derived from Hayashi et al. 2018 and FRPS), and the average terminal rental cost (using the low end of $30 to $100 per month listed on merchant acquirers’ websites). We then calculate the average fixed cost and proportional cost per transaction for each of the three payment methods.” Hayashi, et al., Distributional Effects, supra note 37, at 59.

[40] Fumiko Hayashi & William R. Keeton, Measuring the Costs of Retail Payment Methods, 97(2) Econ. Rev. Fed. Res. Bank of Kansas City 38 (2012).

[41] See Todd J. Zywicki, The Economics of Payment Card Interchange Fees and the Limits of Regulation, (ICLE Financial Regulatory Program White Paper Series, Jun. 2, 2010), at 36-38, available at https://laweconcenter.org/images/articles/zywicki_interchange.pdf.

[42] Id. at 16-18.

[43] See Hans Brits & Carlo Winder, Payments Are No Free Lunch, 3(2) DNB Occasional Studies 27 (2005).

[44] See Nicole Jonker, Social Costs of POS Payments in the Netherlands 2002-2012: Efficiency Gains from Increased Debit Card Usage, 11(2) DNB Occasional Studies 32 (2013).

[45] Olaf Gresvik & Grete Øwre, Costs and Income in the Norwegian Payment System 2001. An Application of the Activity Based Costing Framework, (Working Paper No. 2003/8, Norges Bank, Sep. 17, 2003), at 1, available at https://hdl.handle.net/11250/2498619.

[46] Todd J. Zywicki, Geoffrey A. Manne, & Julian Morris, Price Controls on Payment Card Interchange Fees: The U.S. Experience, (ICLE Financial Regulatory Research Program White Paper 2014-2), at 5-8, available at https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2446080 (noting the loss of free banking accounts, higher monthly maintenance fees, and average minimum holdings required to avoid fees post-Durbin amendment).

[47] See, Cashier Salary in Tennessee, Indeed (last accessed Jul. 24, 2024), https://www.indeed.com/career/cashier/salaries/TN?from=top_sb (salaries based on averages posted on the website on Jul. 24, 2024).

[48] Contactless Payments: Delivering Merchants and Customer Benefits, A Smart Card Alliance Report (2004).

[49] Elizabeth Klee, Paper or Plastic? The Effect of Time on the use of Checks and Debit Cards at Grocery Stores (Finance and Economics Discussion Series, No. 2006-02, Washington Board of Governors of the Federal Reserve System).

[50] Guy Quaden, Costs, Advantages And Disadvantages of Different Payment Methods, Report, Bank of Belgium (2005).

[51] Brits & Winder, supra note 43.

[52] See Michal Polasik et al., Time Efficiency of Point-of-Sale Payment Methods: Empirical Results for Cash, Cards and Mobile Payments, 141 Lecture Notes in Business Information Processing 312 (2013), (Figures given are the mean times for the merchant).

[53] See Greg Buzek, Cash Multipliers: How Reducing the Costs of Cash Handling Can Enable Retail Sales and Profit Growth, IHL Group (2018).

[54] Id. at 7.

[55] Id. at 10.

[56] Id. at 11.

[57] Id. at 10.

[58] Heng Chen, Kim P. Huynh, & Oz Shy, Cash Versus Card: Payment Discontinuities and the Burden of Holding Coins (Bank of Canada Staff Working Paper 2017-47), at ii, available at https://www.econstor.eu/bitstream/10419/197853/1/1011056011.pdf.

[59] Id.

[60] About Legal Tender, Bank of Canada (last accessed Aug. 22, 2024), https://www.bankofcanada.ca/banknotes/about-legal-tender.

[61] Gresvik & Øwre, supra note 45.

[62] Float is working capital; float loss is the cost associated with the time taken for transactions to clear and settle, which means that additional working capital is required to cover outgoings.

[63] David B. Humphrey & Allen N. Berger, Market Failure and Resource Use: Economic Incentives to Use Different Payment Instruments, in The US. Payment system: Efficiency, risk and the role of the Federal Reserve: Proceedings of a symposium on the U.S. payment system sponsored by the Federal Reserve Bank of Richmond (1990), at 45-86.

[64] Kirstin E. Wells, Are Checks Overused?, 20(4) Quarterly Rev. Fed. Res. Bank of Minneapolis 2-12 (1996).

[65] See id. at 4.

[66] GHL, supra note 38.

[67] See id. at 200.

[68] See id. at 201. The authors make certain modifications to the data from FMI, including updating the processing time for transactions and the cost of armored cars.

[69] Id. at 202.

[70] Id. at 204.

[71] Id. at 208.

[72] See id. at 209-12.

[73] See id. at 225.

[74] Garcia-Swartz, et al., supra note 38 at 196.

[75] See Alan Shampine, Another Look at Payment Instrument Economics, Rev. of Network Econ., vol. 6(4), at 495-508 (2007).

[76] Daniel Garcia-Swartz, Robert Hahn, & Anne Layne-Farrar, Further Thoughts on the Cashless Society: A Reply to Dr. Shampine, 6(4) Rev. of Network Econ. 509-524 (2007).

[77] See, e.g., Kerstin Junius, et al., Costs of Retail Payments – An Overview of Recent National Studies in Europe (ECB Occasional Paper No. 294, May 2022), available at https://www.ecb.europa.eu/pub/pdf/scpops/ecb.op294~8ac480631a.en.pdf.

[78] See Brits & Winder, supra note 43, at 27.

[79] Mats Bergman, Gabriela Guibourg, & Björn Segendorf, The Costs of Paying – Private and Social Costs of Cash and Card Payments, at 15 (Sverges Riksbank Working Paper No. 212, Sep. 2007),  available at https://archive.riksbank.se/Upload/Dokument_riksbank/Kat_publicerat/WorkingPapers/WP212.pdf.

[80] Id. at 2.

[81] Brits & Winder, supra note 43, at 27; Layne-Farrar, supra note 20, at 7; Smart Card Alliance, supra note 48; Polasik et al., supra note 52.

[82]  Visa EMV Chip Cards Help Reduce Counterfeit Fraud by 87 Percent, Visa (Sep. 3, 2019), https://usa.visa.com/visa-everywhere/blog/bdp/2019/09/03/visa-emv-chip-1567530138363.html.

[83] Jonker, supra note 44, at 32.

[84] Heiko Schmiedel, Gergana Kostova, & Wiebe Ruttenberg, The Social and Private Costs of Retail Payment Instruments: A European Perspective (ECB Occasional Paper Series No. 137, Sept. 2012), at 12, available at https://www.ecb.europa.eu/pub/pdf/scpops/ecbocp137.pdf.

ICLE Senior Scholar Ben Sperry took part in a recent Washington Legal Foundation-TechFreedom webinar on the legal limits to the principle of common carriage, from railroads to the internet. Video of the full event is embedded below.

Introduction

We thank the Australian Competition & Consumer Commission (ACCC) for the invitation to comment on its July 25 issues paper on recent developments and emerging issues in digital-platform markets, in anticipation of the 10th and final report of its Digital Platform Services Inquiry.[1] The International Center for Law & Economics (ICLE) is a nonprofit, nonpartisan global research and policy center founded with the goal of building the intellectual foundations for sensible, economically grounded policy. ICLE promotes the use of law & economics methodologies to inform public-policy debates and has longstanding expertise in the evaluation of competition law and policy. ICLE’s interest is to ensure that competition law remains grounded in clear rules, established precedent, a record of evidence, and sound economic analysis.

We commend the ACCC’s cautious approach to intervening in digital markets by conducting a comprehensive inquiry before suggesting digital market regulations or reforms to competition law to address alleged competition problems in those markets. As highlighted by the ACCC chair in the press release announcing this paper, “in many cases international legislation is only recently enacted so the full impact may not be clear.”[2]

In these comments, we explore, from a competition and regulation perspective, some of the issues addressed in the issues paper. We respectfully suggest careful consideration before approving any sectoral regulation of digital markets in Australia. Digital markets are generally dynamic, competitive, and beneficial to consumers. Those benefits derive mainly from increased productivity and relatively cheap access to information. While there are always possible competition issues and anticompetitive behavior, these are neither pervasive nor sufficiently unique to justify strict, sui generis preemptive rules. We acknowledge that the ACCC:

[A]grees with the growing international consensus that digital platforms require specific and tailored regulation. While various jurisdictions are taking different approaches to implementing such measures, it is clear to the ACCC that enforcing existing competition and consumer laws ‘ex post’ (i.e. after conduct has occurred) cannot by itself address the systemic and significant problems arising in markets for digital platform services.[3]

We respectfully disagree and posit, instead, that existing antitrust laws are sufficient to address potential anticompetitive practices in digital markets, as we explain below. Furthermore, we think that the ACCC already has the necessary tools and expertise to handle these cases.

Of course, applying antitrust laws to digital markets can be challenging. For example, it can be difficult to define relevant markets and dominant positions on multisided platforms and in the fast-changing digital landscape. The identity of relevant competitors and competing products and services is not always clear, and the boundaries between the digital and non-digital worlds are sometimes overstated. Those challenges, however, can be properly addressed through the existing legal framework. Moreover, many limitations on the effectiveness of ex-post enforcement can be overcome with institutional measures, such as equipping the ACCC with more resources to incorporate advanced, state-of-the-art technical expertise.

Ex-ante regulations like the European Union’s Digital Markets Act (DMA) can have serious unintended consequences, such as stifling innovation, reducing consumer welfare, and increasing compliance costs. They can also lead to increased risks of regulatory capture and rent-seeking, as the verdict on whether a gatekeeper has complied with the law often comes down to the degree to which rivals are satisfied. Of course, rivals have a clear personal stake in never being satisfied. By tethering intervention to a comparatively clear public-benefit standard—consumer welfare—competition laws minimize the potential for error costs and decrease the chances that the law will be coopted for private gain.

Our comments also express the view that policymakers’ present concerns about competition in “AI markets” may be unwarranted. This is particularly true of the notions that data-driven network effects shield incumbents in AI markets from competition; that Web 2.0’s most successful platforms leverage their competitive positions to dominate generative-AI markets; that these same platforms use strategic partnerships with AI firms to insulate themselves from competition; and that generative-AI services occupy narrow markets that leave firms with significant market power.

In fact, we are still far from understanding the boundaries of antitrust-relevant markets in AI. Three primary notions should be at the forefront of competition authorities’ minds when they think about market definitions surrounding AI products and services. First, the “AI market” is not unitary, but instead comprises many distinct goods and services. Second, and relatedly, despite AI marketing hype, this extremely heterogeneous product landscape intersects with equally variegated consumer demand.

In other words: AI products and services may, in many instances, be substitutable for non-AI products, which would mean that, for the purposes of antitrust law, AI and non-AI products contend in the same relevant markets. Getting this relevant product-market definition right is important in antitrust, because incorrect market definitions could lead to wrong inferences about market power. While either an overly broad or overly narrow market definition could lead to erroneous enforcement, we believe the former currently represents the bigger threat.

Third, overenforcement in the field of generative AI could paradoxically engender the very harms that policymakers are seeking to avert. As we explain in greater detail below, preventing so-called “Big Tech” firms from competing in AI markets (for example, by threatening competition intervention whenever they forge strategic relationships with AI startups, launch their own generative-AI services, or embed such services in their existing platforms) may thwart an important source of competition and continued innovation. Competition in AI markets is important,^[4] but trying naïvely to hold incumbent tech firms back out of misguided fears they will come to dominate the AI space is likely to do more harm than good. It is essential to acknowledge how little we know about these nascent markets and that the most important priority at the moment is to ask the right questions to ensure sound competition policy.

The remainder of these comments proceeds as follows: Section II considers international regulatory developments and major market developments discussed in the issues paper’s Topics 1 and 2. Section III considers the emerging issue of competition and AI, aiming to provide a balanced view in a discussion in which proponents of “preemptive enforcement” may be amplifying potential competitive threats that are currently speculative.

II. International Regulatory Developments and Major Developments in Digital Platforms: Rationales for Digital Competition Regulation Are Uncertain

The issues paper presents a detailed description of recent regulatory developments (including both the enactment of new regulations and the discussion of bills and expert reports) in the European Union, Germany, India, Japan, South Korea, and the United Kingdom. As it notes, some key themes have emerged, including “addressing anti-competitive self-preferencing, tying and bundling practices, addressing barriers to switching, and promoting access to third-party applications on a platform or service”.[5]

Whether these are explicit regulations (such as the Digital Markets Act (DMA) in the European Union) or reforms to competition law (as in Germany), all of these reforms seek to ban some or all of the aforementioned conduct or business models due to their allegedly negative effects on competition.

In general terms, “digital markets” or digital platforms are not good candidates for ex-ante regulation, given that the markets for such services are reasonably competitive. As Herbert Hovenkamp has explained:

With Big Tech, we’re looking at probably the most productive part of the economy. The rate of innovation is high. They spend a lot of money on R&D. They are among the largest patent holders. There’s very little evidence of collusion. They seem to be competing with each other quite strongly. They pay their workers relatively well and have fairly educated workforces. None of this is a sign that these are industries we should be pursuing. That doesn’t mean they don’t do some anti-competitive things. But the whole idea that we should be targeting Big Tech strikes me as fundamentally wrong-headed.[6]

According to economic theory and long-tested economic principles, ex-ante regulation^[7] is justified only in the presence of clear market failures^[8]. Digital markets, however, do not present the kind of market failures that warrant ex-ante regulation. For example, digital markets do not generally foment natural monopolies, significant externalities, public goods, or informational asymmetries.

To be sure, one can find some level of informational asymmetries or externalities in some digital markets, but not of such magnitude that they couldn’t be addressed through market competition (actual or potential) or through general rules, such as data-protection or consumer-protection laws. A more plausible argument can be made regarding the presence of “network effects” in online platforms. If a firm moves fast and is the first to attract customers, that customer base will, in turn, attract more customers and sellers. This network growth could, so the story goes, result in a single firm monopolizing the market. However, as Evans and Schmalensee have pointed out, that result is far from inevitable:

Systematic research on online platforms by several authors, including one of us, shows considerable churn in leadership for online platforms over periods shorter than a decade. Then there is the collection of dead or withered platforms that dot this sector, including Blackberry and Windows in smartphone operating systems, AOL in messaging, Orkut in social networking, and Yahoo in mass online media.^[9]

Some regulations and proposals such as the DMA or the proposed American Innovation and Choice Online Act (AICOA) in the United States mention alleged failures of antitrust law (i.e., “too slow” and “too hard for plaintiffs”) as the primary rationale to regulate digital markets. As Giuseppe Colangelo has explained:

Against this background, the regulatory approaches recently advanced do not seem to reflect the distinctive features of digital markets, but rather the need to design enforcement short-cuts to cope with growing concerns that antitrust law is unable to address potential anticompetitive practices by large online platforms. Hence, in most of the mentioned reports, the revival of regulation seems supported more by an alleged antitrust enforcement failure rather than true a market failure. The goal is indeed to fill alleged enforcement gaps in the current antitrust rules by introducing tools aimed at lowering legal standards and evidentiary burdens in order to address anti-competitive practices that standard antitrust analysis would struggle to tackle.^[10]

The ACCC’s fifth interim report does not rely on the presence of market failures to support the need for new competition measures and/or regulations. Rather, resembling the discussion about the above-mentioned regulations and proposals in the European Union and the United States, the ACCC considers that “the enforcement of traditional competition laws against digital platforms in Australia would likewise generally be a slow process. Further, the immense scale and financial resources of large digital platforms may impede traditional enforcement through the courts, resulting in protracted litigated outcomes.”[11]

Sluggish procedure could be a plausible justification for piecemeal reform. The ACCC’s fifth interim report acknowledges this, giving several examples of recent antitrust cases in both Australia and Europe.[12] Antitrust cases could possibly be better expedited. Competition agencies and courts should generally have more resources and faster procedures to adjudicate cases before market structures or markets in general change, rendering any potential intervention useless. This can, however, be resolved via institutional and legal reforms within existing competition law.

The fact that cases are deemed “hard to win” is not a valid justification to jettison the law. Indeed, this might be a strength, not a shortcoming, of antitrust law, particularly in the context of “abuse of dominance” or monopolization cases.^[13] Antitrust cases start from the premise that it is not always immediately obvious whether certain conduct is pro- or anticompetitive.[14] What follows is a structured inquiry operating within a burden-shifting framework that aims to elucidate this question. Regulations like the DMA jettison this cautious approach by condemning broad swathes of specific conduct with strict ex-ante rules that cover many different companies, in different sectors, and with a range of varied products and services. Competitive conditions vary widely, as do the competitive consequences of the business models employed by these firms. DMA-style regulations substitute theoretically robust concepts like “relevant markets” and “market power” or “dominant position” for cruder ones like “core platforms services” or “gatekeeper,” with the express intent of providing shortcuts to condemn business models and practices. But these “shortcuts” have a cost: they can easily lead to condemnation of business models and practices that provide benefits for consumers, such as lower prices and a safer user experience, among others.

Even those open to considering digital-markets regulation acknowledge that there are considerable challenges, especially if the intent is to regulate digital platforms in a manner similar to “essential facilities”:

In the tech industry, the first challenge is to identify a stable essential facility. It must be stable because divestitures take a while to perform, and the cost of implementing them would not be worth its while if the location of the essential facility kept migrating. This condition may not be met, though. While the technology and market segments of electricity, railroads, and (up to the 1980s) telecoms had not changed much since the early twentieth century, digital markets are fastmoving. Rapidly morphing technologies and demands make it difficult for regulators to identify, collect data on, and regulate essential facilities.^[15]

Moreover, even if warranted, regulations create barriers to entry and regulatory risks, and they restrict the monetization of business assets. They also tend to make markets less attractive and could deter potential competitors from entering them. It is possible that the DMA is already producing such consequences. As Alba Ribera has explained:

One of the greatest examples of the dichotomy that arises between the different types of consequences that can be generated by the regulatory capture of digital ecosystems can be found in Meta’s recent decision not to launch its new service Threads in the European Economic Space. To the extent that its service could be interpreted as falling within the definition of a “core platform service” belonging to the category of “online social networks” (listed by the DMA), Meta decided to refrain from entering the European market, due to the disproportionate burden that the demanding obligations imposed by the DMA would entail. It should be noted that Threads is still an entrant service in the online social networking market, in contrast to the predominant position occupied by X (previously known as Twitter). In this way, we observe that the categorization as a core platform service unifies and eliminates all the nuances that free competition entails with respect to incoming services in the markets.^[16]

Some of these unintended consequences were observable in the EU even before the DMA fully entered into force. From users’ perspective, regulation can serve to make services and products more expensive. Facebook is already experimenting with a new business model in the EU in which the consumer would see no ads (thus, there would be no data collection, or less collection of data for marketing purposes, at any rate), but would have to pay for subscriptions.[17] If this business model would generalize, some privacy-minded users may prefer it and probably would be able to afford it. But other consumers that may prefer and have benefitted the most from digital platforms with zero price or otherwise affordable products, such as WhatsApp and Facebook, would be worse off.

From the perspective of the companies that own and operate digital platforms and services, if regulations like the DMA make their platforms less profitable, some could choose not to enter or, indeed, to leave such markets. As Geoffrey Manne and Dirk Auer have explained, “to regulate competition, you first need to attract competition.”^[18]

While these considerations are especially pertinent in the context of developing countries, which rely heavily on attracting foreign direct investment, they could also affect the competitiveness of countries like Australia.

The DMA entered into effect in full force in March 2024. While it may be too early to reach definitive conclusions about its effects, consumers are already reporting that they have experienced a degraded user experience. For example, the French newspaper Liberation has detailed how Google Maps’ map results are not showing directly in search-results pages in the same ways they once did (See Figures 1 and 2).

FIGURE 1: US Search Results for ‘Crepes in Paris’

SOURCE: Chamber of Progress ^[19]

FIGURE 2: French VPN Search Results for ‘Crepes in Paris’

SOURCE: Chamber of Progress ^[20]

Presumably, this is happening because a direct link to Google Maps would constitute “self-preferencing” wherein the search engine, Google, would be “unfairly” directing traffic to its own digital-navigation service. Such conduct is prohibited by Art. 6(5) of the DMA. But this kind of integration is very convenient for consumers, who can search for a restaurant and then quickly find the directions to walk or commute to it (and sometimes even book a table).

While removing some features, Google is also adding more results to its results pages, because it assumes that it is required under the DMA to provide “fair” links to competing sites like Yelp and TripAdvisor.^[21] In theory, the consequence of such requirements is “more options” for consumers. In practice, what consumers have is less relevant results, a more cluttered results page, and thus, a downgraded user experience.

Apple highlights another quality-degrading consequence of the DMA: the obligation imposed on platforms like iOS to allow competing app stores and to allow apps to be downloaded directly from their websites (commonly known as “sideloading”).^[22] This “openness,” however, would allow third-party applications to bypass controls and protections implemented to safeguard users’ security and privacy.^[23] This is already happening in Europe, where Apple has been forced to allow Epic Games to launch an alternative app store on its iOS operating system.[24] While this may seem a positive development for (some) developers and consumers, it could also harm user trust in the platform and thus decrease the total number of transactions, to the detriment of all parties involved (business users, consumers, and the owner of the platform). Indeed, “[p]hishers are using a novel technique to trick iOS and Android users into installing malicious apps that bypass safety guardrails built by both Apple and Google to prevent unauthorized apps.”[25] This sort of attack will be more effective in the absence of the protections in Apple’s App Store.[26]

The recent Microsoft/CrowdStrike outage—which affected many services around the world, including airlines and public services—is a good example of the tradeoffs between “openness” and the security and reliability of digital platforms. At least in part, the outage is explained by the fact that Microsoft must give access to its operating system to CrowdStrike and other developers. As explained in a Financial Times note:

Giving software companies that kind of access to an operating system is dangerous — it means you can quickly lose control of your computer if any of the software providers you rely on makes a mistake or is compromised. That is why Apple began informing third-party developers in 2020 that it would no longer grant them kernel-level access to the MacOS operating system (and also quite possibly why the CrowdStrike problem didn’t affect Apple devices).

But not all the fault lies with Microsoft. A 2009 agreement between the company and the European Commission requires it to grant outside developers the same access to Windows that its own security software has. The idea was to make it possible for other software companies to compete with Microsoft by ensuring many of its products and services are interoperable with outside software and tools. That’s a worthy goal, and many provisions in the agreement are entirely reasonable, such as requiring that Outlook support common calendar event and scheduling formats.

But the 2009 agreement is profoundly flawed in requiring Microsoft to make all of the APIs, or programming functions, that its own security software products use available to manufacturers of third-party security software products. This is the provision that requires Microsoft to give kernel-level access to companies such as CrowdStrike. Until it is changed, it’s not clear that Microsoft can implement the chief lesson of this debacle and start phasing out access, as Apple did four years ago.[27]

The problem with ex-ante regulations like the DMA, which prohibit some types of conduct absolutely, is that they don’t capture the complexities and tradeoffs generally present in market dynamics and, of course, specifically in digital markets. As Dirk Auer has pointed out, the alleged benefits of the DMA are looking more and more like broken promises:

When it was passed, European policymakers like Margrethe Vestager and Thierry Breton assured the public that the far-reaching regulation would not compromise security, lead to costlier services, or otherwise degrade users’ online experience. They also argued that it would be fast and easy to apply, thus avoiding the lengthy litigation that has come to be associated with competition enforcement.

As the effects of the DMA start to play out, however, these promises appear increasingly fanciful.

The biggest concern is that Europeans’ online safety is being compromised. Apple has warned that it will not be able to guarantee the safety of rival app stores and payment systems that can now access its ecosystem. If this sounds abstract, it is worth noting that these sorts of security flaws facilitated the Oct. 7 attacks carried out by Hamas. They also increase more mundane risks of identity theft and fraud.

Similarly, Amazon will struggle to exclude nefarious goods, sellers, and shippers from its online marketplace. Commenting on similar issues in the United States, the company surmised that it risked losing “customer trust by advertising something that is not a good deal for them.” This loss of consumer trust would, in turn, harm the bottom lines of the roughly two million businesses that rely on the platform.[28]

As indicated above, the DMA’s unintended consequences affect not only consumers, but also business users. Since Google began to implement the DMA in January 2024, early estimates suggest that clicks from Google ads to hotel websites fell by 17.6%.^[29] Presumably, this ought to be considered a failure, even by the DMA’s own (uncertain) standards.

Australia itself has experience with another regulation intended to address a “significant bargaining power imbalance” between digital platforms and businesses interacting with them. The News Media and Digital Platforms Mandatory Bargaining Code was enacted in 2021 to allow news organizations to collectively bargain with platforms like Google and Facebook for remuneration for news content featured on those platforms without breaching Australian competition laws.[30] It has, however, apparently resulted in less traffic for smaller media outlets and independent publishers. As reported by The Guardian: “Smaller publishers are already feeling the effects of a potential ban on news on Facebook, a parliamentary committee has heard, as news outlets small and large make the case for social media companies to be compelled to pay for news.”[31]

This is why, even when their relative lenity is considered, antitrust laws are more flexible and more likely to be appropriate for various markets and business models.^[32] As Posner illustrates, while rules are generally simpler and cheaper to enforce, they are often underinclusive or overinclusive. When facts like the mere size or number of users—which are disconnected from competitive goals—are determinative of their application, they are “especially apt to fail.”^[33]

All in all, as the ACCC acknowledges, most digital competition regulation “has only recently been enacted, and the full impact of recent international regulatory developments cannot yet be observed.”[34] In that vein, it would be wise to study markets, perform proper regulatory-impact analysis, and keep learning from the experience of other regulators and markets. It may be advisable, for instance, to follow the example of South Korea, which has hit the pause button on its proposal to regulate digital markets.[35]

Regarding major developments in digital-platform markets, the issues paper provides updated information about two markets: online private messaging and app marketplaces.

With respect to the online private-messaging market, the issues paper reminds that:

The ACCC’s September 2020 report found that both Facebook’s and Apple’s services benefited from identity-based network effects, providing them with significant competitive advantages over smaller suppliers of standalone services in Australia. However, because the use of Apple’s online private messaging services was limited to users of Apple devices, the ACCC found that iMessage was likely to impose weaker competitive constraints on Facebook Messenger and WhatsApp than those services imposed on iMessage.[36]

The paper also highlights that “Australians’ usage of these services is increasing. Based on recent ACMA survey data, in the 6 months prior to June 2023, 72% of Australian adults used Facebook Messenger (up from 66% in 2020), 51% used WhatsApp (up from 39% in 2020) and 35% used FaceTime (up from 33% in 2020).”[37]

It is important, at this point, to consider that, while the aforementioned messaging platforms (Facebook’s Messenger and Apple’s iMessage) have a large number of users and, therefore, a relevant market share and the advantage of network effects, this does not mean that the messaging-apps market is not competitive. The issues paper itself recognizes that:

Online private messaging services encompass a range of services, including text, audio and video messaging services, and are offered by a wide variety of platforms. The ACCC observed a wide range of online private messaging services available to Australian users, which were often highly differentiated, offering different features and functionalities, and used by consumers for a number of different purposes.[38]

Besides Facebook’s Messenger and Apple’s iMessage, several actors with smaller but not irrelevant market shares create competitive pressure on both.[39] Moreover, this is a market where consumers can use more than one application simultaneously (i.e., there is “multi-homing”). This allows consumers to try different services at little cost, which means that both current competitors and new entrants can challenge the market leaders. The fact that the use of these services is increasing in Australia provides an opportunity for these challengers.

Furthermore, although network effects offer an important advantage, they do not guarantee a comfortable monopolist position.[40] Catherine Tucker explains how network effects can be a double-edged sword, as they can also lead to a sudden decline of users to a marginally superior competitor.[41] Actual economic studies of data-network effects have been few and far between, with scant empirical evidence to support the theory that the control of personal data creates an insurmountable barrier to entry.^[42]

Regarding app marketplaces, the issues paper finds that:

In its March 2021 report, the ACCC found that Apple and Google were the predominant mobile app marketplace operators in Australia, offering the App Store and Play Store respectively.

The report referenced data from Statcounter, estimating that in December 2020, Apple iOS held 54% of the market share of mobile operating systems (OS) in Australia, while Android held 46%. Based on updated data from Statcounter, these shares appear to have remained stable – as of June 2024, Apple iOS again held 54% market share, while Android held 45%. The next closest competitor, Samsung, held less than 1% (0.86%).

In the March 2021 report, the ACCC considered it likely that Apple and Google held significant market power in mobile app distribution in Australia, due to their control of the iOS and Android mobile OS. This meant the App Store and the Play Store were ‘must haves’ for the majority of mobile app developers in Australia.[43]

The fact that both Apple and Google (Android) both have relatively high (and similar) market shares does not necessarily mean that the market is not competitive. Two powerful actors can discipline each other’s efforts to exercise market power.[44] Our point here is not that there cannot be any competition problem or anticompetitive conduct in this market, of course, but rather that more evidence than merely concentration or market shares is required in order to determine that there are even possible competition issues in this market. The rivalry between Apple and Android in the market has largely benefited consumers with innovation and increased access to a large array of features.[45]

III. Potential Emerging Issues: Competition and AI

Although the issues paper also identifies emerging issues in the online-gaming and cloud-computing markets, it identifies very specific (if somewhat theoretical) risks in the generative-AI market that echo the concerns raised by several competition agencies around the world. Drawing on a discussion paper by the Digital Platforms Regulators Forum, the issues paper notes that:

• Developing and operating large language models (LLMs) requires a large upfront financial investment, access to vast datasets, long development lead times, access to sophisticated AI systems and talent, and substantial ongoing computing costs and access to computing resources (such as cloud storage), which together create high barriers to entry.
• LLMs are likely to have features common to digital-platform services that make such markets tend toward concentration, including a positive feedback loop involving the collection and use of user data, economies of scale, and access to large volumes of high-quality user data. Because of these characteristics, new entrants could find it difficult to compete with digital-platform services that use LLMs as part of new and existing services.
• Generative AI could lead users to interact more with particular digital platforms. Over time, this may make it more difficult for users to leave these platforms. LLMs could allow big digital platforms to strengthen and expand their market power by continuing to engage in the allegedly anti-competitive practices the ACCC has previously observed. These include anti-competitive self-preferencing, tying, and restrictions on data access.[46]

We address these concerns in the following subsections. Subsection A summarizes recent calls for competition intervention in generative-AI markets. Subsection BB argues that many of these calls are underpinned by fears of data-related incumbency advantages (often referred to as “data-network effects”), including in the context of mergers. Subsection CC explains why these effects are unlikely to play a meaningful role in generative AI markets. Finally, subsection D offers five key takeaways to help policymakers better weigh the tradeoffs inherent in competition enforcement interventions in generative AI markets.

A. Calls for Intervention in AI Markets Should Acknowledge Actual Market Developments

It was once (and frequently) said that Google’s “data monopoly” was unassailable: “If ‘big data’ is the oil of the information economy, Google has Standard Oil-like monopoly dominance—and uses that control to maintain its dominant position.”[47] Similar claims of data dominance have been attached to nearly all large online platforms, including Facebook (Meta), Amazon, and Uber.[48]

While some of these claims continue even today (for example, “big data” is a key component of the U.S. Justice Department’s (DOJ) Google Search and adtech antitrust suits),[49] a shiny new data target has emerged in the form of generative AI. The launch of ChatGPT in November 2022, as well as the advent of AI image-generation services like Midjourney and Dall-E, have dramatically expanded the public’s conception of what is—and what might be—possible to achieve with generative-AI technologies built on massive datasets.

While these services remain both in the early stages of mainstream adoption and in the throes of rapid, unpredictable technological evolution, they are nevertheless already on the radar of competition policymakers around the world. Several antitrust enforcers appear to believe that, by acting now, they can avoid the “mistakes” that purportedly were made during the formative years of Web 2.0.[50] These mistakes, critics assert, include failing to appreciate the centrality of data in online markets, as well as letting mergers go unchecked and allowing early movers to entrench their market positions.^[51] As Federal Trade Commission (FTC) Chair Lina Khan has put it: “we are still reeling from the concentration that resulted from Web 2.0, and we don’t want to repeat the missteps of the past with AI.”^[52]

This response from the competition-policy world is deeply troubling. Rather than engage in critical self-assessment and adopt an appropriately restrained stance, the enforcement community appears to be champing at the bit. Rather than assessing their prior assumptions based on the current technological moment, enforcers’ top priority appears to be figuring out how to rapidly and almost reflexively deploy existing competition tools to address the presumed competitive failures presented by generative AI.^[53]

It has for some time been increasingly common for competition enforcers to argue that “data-network effects” serve not only to entrench incumbents in those markets where the data is collected, but also to confer similar, self-reinforcing benefits in adjacent markets. Several enforcers have, for example, prevented large online platforms from acquiring smaller firms in adjacent markets, citing the risk that they could use their vast access to data to extend their dominance into these new markets.^[54]

They have concomitantly launched consultations to ascertain the role that data plays in AI competition. For instance, in a recent consultation, the European Commission asked: “What is the role of data and what are its relevant characteristics for the provision of generative AI systems and/or components, including AI models?”^[55] The FTC has likewise been hypervigilant about the risks ostensibly posed by incumbents’ access to data. In comments submitted to the U.S. Copyright Office, for example, the FTC argued that:

The rapid development and deployment of AI also poses potential risks to competition. The rising importance of AI to the economy may further lock in the market dominance of large incumbent technology firms. These powerful, vertically integrated incumbents control many of the inputs necessary for the effective development and deployment of AI tools, including cloud-based or local computing power and access to large stores of training data. These dominant technology companies may have the incentive to use their control over these inputs to unlawfully entrench their market positions in AI and related markets, including digital content markets.[56]

Likewise, Jonathan Kanter, assistant U.S. attorney general for antitrust, recently claimed that:

We also see structures and trends in AI that should give us pause AI relies on massive amounts of data and computing power, which can give already dominant firms a substantial advantage. Powerful networks and feedback effects may enable dominant firms to control these new markets, and existing power in the digital economy may create a powerful incentive to control emerging innovations that will not only impact our economy, but the health and well-being of our society and free expression itself.[57]

On an even more hyperbolic note, Andreas Mundt, the head of Germany’s Federal Cartel Office, called AI a “first-class fire accelerator” for anticompetitive behavior and argued it “will make all the problems only worse.”[58] He further argued that “there’s a great danger that we’ll get an even deeper concentration of digital markets and power increase at various levels, from chips to the front end.”[59] In short, Mundt is one of many policymakers who believe that AI markets will enable incumbent tech firms to further entrench their market positions.

Certainly, it makes sense that the largest online platforms—including Alphabet, Meta, Apple, and Amazon—should have a meaningful advantage in the burgeoning markets for generative AI services. After all, it is widely recognized that data is an essential input for generative AI.^[60] This competitive advantage should be all the more significant, given that these firms have been at the forefront of AI technology for more than a decade. Over this period, Google’s DeepMind and AlphaGo and Meta’s NLLB-200 have routinely made headlines.[61] Apple and Amazon also have vast experience with AI assistants, and all of these firms deploy AI technologies throughout their platforms.[62]

Contrary to what one might expect, however, the tech giants have, to date, been largely unable to leverage their vast troves of data to outcompete startups like OpenAI and Midjourney. At the time of writing, for instance, OpenAI’s ChatGPT appears to be, by far, the most successful chatbot,[63] despite the large tech platforms’ apparent access to far more (and more up-to-date) data.

There are important lessons to glean from these developments, if only enforcers would stop to reflect. The meteoric rise of consumer-facing AI services should offer competition enforcers and policymakers an opportunity for introspection. As we explain, the rapid emergence of generative AI may undercut many core assumptions of today’s competition-policy debates, which have focused largely on the rueful after-effects of the purported failure of 20^th-century antitrust to address the alleged harms of 21^st-century technology. These include the notions that data advantages constitute barriers to entry and can be leveraged to project dominance into adjacent markets; that scale itself is a market failure to be addressed by enforcers; and that the use of consumer data is inherently harmful to those consumers.

B. Data-Network Effects Theory and Enforcement

Proponents of more extensive intervention by competition enforcers into digital markets often cite data-network effects as a source of competitive advantage and a barrier to entry (although terms like “economies of scale and scope” may offer more precision).^[64] The crux of the argument is that “the collection and use of data creates a feedback loop of more data, which ultimately insulates incumbent platforms from entrants who, but for their data disadvantage, might offer a better product.”^[65] This self-reinforcing cycle purportedly leads to market domination by a single firm. Thus, it is argued, e.g., that Google’s “ever-expanding control of user personal data, and that data’s critical value to online advertisers, creates an insurmountable barrier to entry for new competition.”^[66]

But it is important to note the conceptual problems these claims face. Because data can be used to improve products’ quality and/or to subsidize their use, if possessing data constitutes an entry barrier, then any product improvement or price reduction made by an incumbent could be problematic. This is tantamount to an argument that competition itself is a cognizable barrier to entry. Of course, it would be a curious approach to antitrust if competition were treated as a problem, as it would imply that firms should under-compete—i.e., should forego consumer-welfare enhancements—in order to inculcate a greater number of firms in a given market, simply for its own sake.[67]

Meanwhile, actual economic studies of data-network effects have been few and far between, with scant empirical evidence to support the theory.^[68] Andrei Hagiu and Julian Wright’s theoretical paper offers perhaps the most comprehensive treatment of the topic to date.[69] The authors ultimately conclude that data-network effects can be of differing magnitudes and have varying effects on firms’ incumbency advantage.^[70] They cite Grammarly (an AI writing-assistance tool) as a potential example: “As users make corrections to the suggestions offered by Grammarly, its language experts and artificial intelligence can use this feedback to continue to improve its future recommendations for all users.”^[71]

This is echoed by economists who contend that “[t]he algorithmic analysis of user data and information might increase incumbency advantages, creating lock-in effects among users and making them more reluctant to join an entrant platform.”^[72] Crucially, some scholars take this logic a step further, arguing that platforms may use data from their “origin markets” in order to enter and dominate adjacent ones:

First, as we already mentioned, data collected in the origin market can be used, once the enveloper has entered the target market, to provide products more efficiently in the target market. Second, data collected in the origin market can be used to reduce the asymmetric information to which an entrant is typically subject when deciding to invest (for example, in R&D) to enter a new market. For instance, a search engine could be able to predict new trends from consumer searches and therefore face less uncertainty in product design.^[73]

This possibility is also implicit in Hagiu and Wright’s paper.[74] Indeed, the authors’ theoretical model rests on an important distinction between “within-user” data advantages (that is, having access to more data about a given user) and “across-user” data advantages (information gleaned from having access to a wider user base). In both cases, there is an implicit assumption that platforms may use data from one service to gain an advantage in another market (because what matters is information about aggregate or individual user preferences, regardless of its origin).

Our review of the economic evidence suggests that several scholars have, with varying degrees of certainty, raised the possibility that incumbents may leverage data advantages to stifle competitors in their primary market or in adjacent ones (be it via merger or organic growth). As we explain below, however, there is ultimately little evidence to support such claims. Policymakers have nonetheless been keenly receptive to these limited theoretical findings, basing multiple decisions on these theories, often with little consideration given to the caveats that accompany them.^[75]

Indeed, it is remarkable that, in its section on “[t]he data advantage for incumbents,” the “Furman Report” created for the UK government cited only two empirical economic studies, and they offer directly contradictory conclusions with respect to the strength of data advantages.[76] The report nevertheless concluded that data “may confer a form of unmatchable advantage on the incumbent business, making successful rivalry less likely,”[77] and it adopted without reservation what it deemed “convincing” evidence from non-economists that have no apparent empirical basis.[78]

In the Google/Fitbit merger proceedings, the European Commission found that the combination of data from Google services with that of Fitbit devices would reduce competition in advertising markets:

Giving [sic] the large amount of data already used for advertising purposes that Google holds, the increase in Google’s data collection capabilities, which goes beyond the mere number of active users for which Fitbit has been collecting data so far, the Transaction is likely to have a negative impact on the development of an unfettered competition in the markets for online advertising.^[79]

As a result, the Commission cleared the merger only on the condition that Google refrain from using data from Fitbit devices for its advertising platform.^[80] The Commission also appears likely to focus on similar issues in its ongoing investigation of Microsoft’s investment in OpenAI.[81]

Along similar lines, in its complaint to enjoin Meta’s purchase of Within Unlimited—makers of the virtual-reality (VR) fitness app Supernatural—the FTC relied on, among other things, the fact that Meta could leverage its data about VR-user behavior to inform its decisions and potentially outcompete rival VR-fitness apps: “Meta’s control over the Quest platform also gives it unique access to VR user data, which it uses to inform strategic decisions.”^[82]

The DOJ’s twin cases against Google also implicate data leveraging and data barriers to entry. The agency’s adtech complaint charges that “Google intentionally exploited its massive trove of user data to further entrench its monopoly across the digital advertising industry.”[83] Similarly, in its Google Search complaint, the agency argued that:

Google’s anticompetitive practices are especially pernicious because they deny rivals scale to compete effectively. General search services, search advertising, and general search text advertising require complex algorithms that are constantly learning which organic results and ads best respond to user queries; the volume, variety, and velocity of data accelerates the automated learning of search and search advertising algorithms.[84]

Finally, updated merger guidelines published in recent years by several competition enforcers cite the acquisition of data as a potential source of competition concerns. For instance, the FTC and DOJ’s 2023 guidelines state that “acquiring data that helps facilitate matching, sorting, or prediction services may enable the platform to weaken rival platforms by denying them that data.”^[85] Likewise, the UK Competition and Markets Authority warned against incumbents acquiring firms in order to obtain their data and foreclose other rivals:

Incentive to foreclose rivals…

7.19(e) Particularly in complex and dynamic markets, firms may not focus on short term margins but may pursue other objectives to maximise their long-run profitability, which the CMA may consider. This may include… obtaining access to customer data….^[86]

In short, competition authorities around the globe have taken an increasingly aggressive stance on data-network effects. Among the ways this has manifested is in enforcement decisions based on fears that data a platform collects in one market might confer decisive competitive advantages in adjacent markets. Unfortunately, these concerns rest on little to no empirical evidence, either in the economic literature or the underlying case records.

C. Data-Incumbency Advantages in Generative AI

Given the assertions detailed in the previous section, it would be reasonable to assume that firms such as Google, Meta, and Amazon should be in pole position to meet the burgeoning demand for generative AI. After all, these firms have not only been at the forefront of the field for the better part of a decade, but they also have access to vast troves of data, the likes of which their rivals could only dream when they launched their own services. Thus, the authors of the Furman Report caution that “to the degree that the next technological revolution centres around artificial intelligence and machine learning, then the companies most able to take advantage of it may well be the existing large companies because of the importance of data for the successful use of these tools.”[87]

To date, however, this is not how things have unfolded (although it bears noting that these technologies remain in flux and the competitive landscape is susceptible to change). The first significantly successful generative AI service was arguably not from either Meta—which had been working on chatbots for years and had access to, arguably, the world’s largest database of actual chats—or Google. Instead, the breakthrough came from a previously unknown firm called OpenAI.

OpenAI’s ChatGPT service currently accounts for an estimated 60% of visits to online AI tools (although reliable numbers are somewhat elusive).[88] It broke the record for the fastest online service to reach 100 million users (in only a couple of months), more than four times faster than TikTok, the previous record holder.^[89] Based on Google Trends data, ChatGPT is nine times more popular worldwide than Google’s own Bard service, and 14 times more popular in the United States.^[90] In April 2023, ChatGPT reportedly registered 206.7 million unique visitors, compared to 19.5 million for Google’s Bard.^[91] In short, at the time we are writing, ChatGPT appears to be the most popular chatbot. The entry of large players such as Google Bard or Meta AI appear to have had little effect thus far on its leading position.^[92]

The picture is similar in the field of AI image generation. As of August 2023, Midjourney, Dall-E, and Stable Diffusion appear to be the three market leaders in terms of user visits.^[93] This is despite competition from the likes of Google and Meta, who arguably have access to unparalleled image and video databases by virtue of their primary platform activities.^[94]

This raises several crucial questions: how have these AI upstarts managed to be so successful, and is their success just a flash in the pan before Web 2.0 giants catch up and overthrow them? While we cannot answer either of these questions dispositively, we offer what we believe to be some relevant observations concerning the role and value of data in digital markets.

A first important observation is that empirical studies suggest that data exhibits diminishing marginal returns. In other words, past a certain point, acquiring more data does not confer a meaningful edge to the acquiring firm. As Catherine Tucker put it, following a review of the literature: “Empirically there is little evidence of economies of scale and scope in digital data in the instances where one would expect to find them.”^[95]

Likewise, following a survey of the empirical literature on this topic, Geoffrey Manne and Dirk Auer conclude that:

Available evidence suggests that claims of “extreme” returns to scale in the tech sector are greatly overblown. Not only are the largest expenditures of digital platforms unlikely to become proportionally less important as output increases, but empirical research strongly suggests that even data does not give rise to increasing returns to scale, despite routinely being cited as the source of this effect.^[96]

In other words, being the firm with the most data appears to be far less important than having enough data. Moreover, this lower bar may be accessible to far more firms than one might initially think possible. Furthermore, obtaining sufficient data could become easier still—that is, the volume of required data could become even smaller—with technological progress. For instance, synthetic data may provide an adequate substitute to real-world data,[97] or may even outperform real-world data.[98] As Thibault Schrepel and Alex Pentland surmise:

[A]dvances in computer science and analytics are making the amount of data less relevant every day. In recent months, important technological advances have allowed companies with small data sets to compete with larger ones.^[99]

Indeed, past a certain threshold, acquiring more data might not meaningfully improve a service, where other improvements (such as better training methods or data curation) could have a large impact. In fact, there is some evidence that excessive data impedes a service’s ability to generate results appropriate for a given query: “[S]uperior model performance can often be achieved with smaller, high-quality datasets than massive, uncurated ones. Data curation ensures that training datasets are devoid of noise, irrelevant instances, and duplications, thus maximizing the efficiency of every training iteration.”[100]

Consider, for instance, a user who wants to generate an image of a basketball. Using a model trained on an indiscriminate range and number of public photos in which a basketball appears surrounded by copious other image data, the user may end up with an inordinately noisy result. By contrast, a model trained with a better method on fewer, more carefully selected images could readily yield far superior results.[101] In one important example:

The model’s performance is particularly remarkable, given its small size. “This is not a large language model trained on the whole Internet; this is a relatively small transformer trained for these tasks,” says Armando Solar-Lezama, a computer scientist at the Massachusetts Institute of Technology, who was not involved in the new study…. The finding implies that instead of just shoving ever more training data into machine-learning models, a complementary strategy might be to offer AI algorithms the equivalent of a focused linguistics or algebra class.[102]

Platforms’ current efforts are thus focused on improving LLMs’ mathematical and logical reasoning, rather than maximizing the size of training datasets.^[103] Two points stand out. The first is that firms like OpenAI rely largely on publicly available datasets—such as GSM8K—to train their LLMs.^[104] Second, the real challenge to creating innovative AI lies not so much in collecting data, but in creating innovative AI-training processes and architectures:

[B]uilding a truly general reasoning engine will require a more fundamental architectural innovation. What’s needed is a way for language models to learn new abstractions that go beyond their training data and have these evolving abstractions influence the model’s choices as it explores the space of possible solutions.

We know this is possible because the human brain does it. But it might be a while before OpenAI, DeepMind, or anyone else figures out how to do it in silicon.^[105]

Furthermore, it is worth noting that the data most relevant to startups in a given market may not be those held by large incumbent platforms in other markets. They might instead be data specific to the market in which the startup is active or, even better, to the given problem it is attempting to solve:

As Andres Lerner has argued, if you wanted to start a travel business, the data from Kayak or Priceline would be far more relevant. Or if you wanted to start a ride-sharing business, data from cab companies would be more useful than the broad, market-cross-cutting profiles Google and Facebook have. Consider companies like Uber, Lyft and Sidecar that had no customer data when they began to challenge established cab companies that did possess such data. If data were really so significant, they could never have competed successfully. But Uber, Lyft and Sidecar have been able to effectively compete because they built products that users wanted to use—they came up with an idea for a better mousetrap. The data they have accrued came after they innovated, entered the market and mounted their successful challenges—not before.[106]

The bottom line is that data is not the be-all and end-all that many in competition circles make it out to be. While data may often confer marginal benefits, there is little evidence that these benefits are ultimately decisive.[107] As a result, incumbent platforms’ access to vast numbers of users and troves of data in their primary markets might only marginally affect their competitiveness in AI markets.

A related observation is that firms’ capabilities and other features of their products arguably play a more important role than the data they own.[108] Examples of this abound in digital markets. Google overthrew Yahoo in search, despite initially having access to far fewer users and far less data. Google and Apple overcame Microsoft in the smartphone operating-system market, despite having comparatively tiny ecosystems (at the time) to leverage. TikTok rose to prominence despite intense competition from incumbents like Instagram, which had much larger userbases. In each of these cases, important product-design decisions (such as the PageRank algorithm, recognizing the specific needs of mobile users,^[109] and TikTok’s clever algorithm) appear to have played far more significant roles than the firms’ initial user and data endowments (or lack thereof).

All of this suggests that the early success of OpenAI likely has more to do with its engineering decisions than with what data it did or did not possess. Going forward, OpenAI and its rivals’ relative abilities to offer and monetize compelling use cases by offering custom versions of their generative AI technologies will arguably play a much larger role than (and contribute to) their ownership of data.[110] In other words, the ultimate challenge is arguably to create a valuable platform, of which data ownership is a consequence, not a cause.

It is also important to note that, in those instances where it is valuable, data does not just fall from the sky. Instead, it is through smart business and engineering decisions that firms can generate valuable information (which does not necessarily correlate with owning more data). For instance, OpenAI’s success with ChatGPT is often attributed to its more efficient algorithms and training models, which arguably have enabled the service to improve more rapidly than its rivals.^[111] Likewise, the ability of firms like Meta and Google to generate valuable data for advertising arguably depends more on design decisions that elicit the right data from users, rather than the raw number of users in their networks.

Put differently, setting up a business so as to gather and organize the right information is more important than simply owning vast troves of data.[112] Even in those instances where high-quality data is an essential parameter of competition, it does not follow that having vaster databases or more users on a platform necessarily leads to better information for the platform. Indeed, if data ownership consistently conferred a significant competitive advantage, these new AI firms would not be where they are today.

Moreover, it is important not to neglect the role that open-source models currently play in fostering innovation and competition. As former DOJ Chief Antitrust Economist Susan Athey pointed out in a recent interview, “[the AI industry] may be very concentrated, but if you have two or three high quality—and we have to find out what that means, but high enough quality—open models, then that could be enough to constrain the for-profit LLMs.”[113] Open-source models are important because they allow innovative startups to build upon models already trained on large datasets—therefore entering the market without incurring that initial cost. Apparently, there is no lack of open-source models, since companies like xAI, Meta, and Google offer their AI models for free.[114]

This does not, of course, mean that data is worthless. Rather, it means that competition authorities should not assume that the mere possession of data is a dispositive competitive advantage, absent compelling empirical evidence to support such a finding. In this light, the current wave of decisions and competition-policy pronouncements that rely on data-related theories of harm are premature.

D. Five Key Takeaways: Reconceptualizing the Role of Data in Generative-AI Competition

As we explain above, data-network effects are not the source of barriers to entry that they are sometimes made out to be. The picture is far more nuanced. Indeed, as economist Andres Lerner demonstrated almost a decade ago (and the assessment is only truer today):

Although the collection of user data is generally valuable for online providers, the conclusion that such benefits of user data lead to significant returns to scale and to the entrenchment of dominant online platforms is based on unsupported assumptions. Although, in theory, control of an “essential” input can lead to the exclusion of rivals, a careful analysis of real-world evidence indicates that such concerns are unwarranted for many online businesses that have been the focus of the “big data” debate.[115]

While data can be an important part of the competitive landscape, incumbents’ data advantages are far less pronounced than today’s policymakers commonly assume. In that respect, five primary lessons emerge:

1. Data can be (very) valuable, but beyond a certain threshold, those benefits tend to diminish. In other words, having the most data is less important than having enough;
2. The ability to generate valuable information does not depend on the number of users or the amount of data a platform has previously acquired;
3. The most important datasets are not always proprietary;
4. Technological advances and platforms’ engineering decisions affect their ability to generate valuable information, and this effect swamps those that stem from the amount of data they own; and
5. How platforms use data is arguably more important than what data or how much data they own.

These lessons have important ramifications for policy debates over the competitive implications of data in technologically evolving areas.

First, it is not surprising that startups, rather than incumbents, have taken an early lead in generative AI (and in Web 2.0 before it). After all, if data-incumbency advantages are small or even nonexistent, then smaller and more nimble players may have an edge over established tech platforms. This is all the more likely given that, despite significant efforts, the biggest tech platforms were unable to offer compelling generative-AI chatbots and image-generation services before the emergence of ChatGPT, Dall-E, Midjourney, etc.

This suggests that, in a process akin to Clayton Christensen’s “innovator’s dilemma,”[116] something about the incumbent platforms’ existing services and capabilities might have been holding them back in this emerging industry. Of course, this does not necessarily mean that those same services or capabilities could not become an advantage when the generative-AI industry starts addressing issues of monetization and scale.[117] But it does mean that assumptions about a firm’s market power based primarily on its possession of data are likely to be off the mark.

Another important implication is that, paradoxically, policymakers’ efforts to prevent Web 2.0 platforms from competing freely in generative-AI markets may ultimately backfire and lead to less, not more, competition. Indeed, OpenAI is currently acquiring a sizeable lead in generative AI. While competition authorities might like to think that other startups will emerge and thrive in this space, it is important not to confuse those desires with reality. While there currently exists a vibrant AI-startup ecosystem, there is at least a case to be made that significant competition for today’s AI leaders will come from incumbent Web 2.0 platforms—although nothing is certain at this stage.

Policymakers should beware not to stifle that competition on the misguided assumption that competitive pressure from large incumbents is somehow less valuable to consumers than those that originate from smaller firms. This is particularly relevant in the context of merger control. An acquisition (or an “acqui-hire”) by a “Big Tech” company does not only, in principle, entail a minor risk to harm competition (it is not a horizontal merger),[118] but could create a stronger competitor to the current market leaders.

Finally, even if there were a competition-related market failure to be addressed in the field of generative AI (which is anything but clear), the remedies under contemplation may do more harm than good. Some of the solutions that have been put forward have highly ambiguous effects on consumer welfare. Scholars have shown that, e.g., mandated data sharing—a solution championed by EU policymakers, among others—may sometimes dampen competition in generative AI.[119] This is also true of legislation like the General Data Protection Regulation (GDPR), which makes it harder for firms to acquire more data about consumers—assuming such data is, indeed, useful to generative AI services.[120]

In sum, it is a flawed understanding of the economics and practical consequences of large agglomerations of data that has led competition authorities to believe data-incumbency advantages are likely to harm competition in generative AI—or even in the data-intensive Web 2.0 markets that preceded it. Indeed, competition or regulatory intervention to “correct” data barriers and data network and scale effects is liable to do more harm than good.

[1] Australian Competition & Consumer Commission, Digital Platform Services Inquiry —March 2025 — Final Report Issues Paper (25 Jul. 25 2024), available at https://www.accc.gov.au/system/files/dpsi-10-final-report-issues-paper.pdf?ref=0&download=y (hereinafter “issues paper” or “final report issues paper”). See also Press Release, Final Digital Platforms Report to Focus on Global Developments and Emerging Competition and Consumer Issues, Australian Competition & Consumer Commission (25 Jul. 2024), https://www.accc.gov.au/media-release/final-digital-platforms-report-to-focus-on-global-developments-and-emerging-competition-and-consumer-issues.

[2] Press release, id.

[3] Digital Platform Services Inquiry: Interim Report No. 5 — Regulatory Reform, Australian Competition & Consumer Commission (Sep. 2022), at 47, available at https://www.accc.gov.au/system/files/Digital%20platform%20services%20inquiry%20-%20September%202022%20interim%20report.pdf.

[4] Artificial intelligence is, of course, not a market (at least, not an antitrust-relevant market). Within the realm of what is called “AI,” companies offer myriad products and services, and specific relevant markets would need to be defined before assessing harm to competition in specific cases.

[5] Issues paper, supra note 1, at 4.

[6] Robert Armstrong & Ethan Wu, What Big Tech Antitrust Gets Wrong. An Interview with Herbert Hovenkamp, Financial Times (19 Jan. 2024), https://www.ft.com/content/4eec8bc3-c892-4704-ae66-a4432c6d4fd7.

[7] By ex-ante regulation, we mean specific rules and duties that are sector-specific (“digital markets”), whose application would not be based on the effects of the conduct regulated, and where fines would apply in case of non-compliance. See Bruce H. Kobayashi & Joshua D. Wright, Antitrust and Ex-Ante Sector Regulation, The GAI Report on the Digital Economy 25 (2020); see also id. at Table 1 at 869.

[8] See Robert Cooter & Tomas Ulen, Law and Economics 40-43 (2000); W. Kip Viscusi, Joseph E. Harrington, Jr. and John M. Vernon, Economics of Regulation and Antitrust 376-79 (2005).

[9] David S. Evans & Richard Schmalensee, Debunking The “Network Effects” Bogeyman, Regulation 36, 39 (Winter 2017-2018) available at https://www.cato.org/sites/cato.org/files/serials/files/regulation/2017/12/regulation-v40n4-1.pdf.

[10] Giuseppe Colangelo, Evaluating the Case for Regulation of Digital Platforms, The GAI Report on the Digital Economy 905, 930 (2020) https://gaidigitalreport.com/2020/10/04/evaluating-the-case-for-ex-ante-regulation-of-digital-platforms.

[11] Interim Report No 5, supra note 3, at 48.

[12] Australian Competition & Consumer Commission, supra note 3, at 48-49.

[13] We often run the risk of condemning business practices and models we don’t fully understand. Sometimes, even the businesses that implement them don’t fully know or understand the impact of such practices. See Frank H. Easterbrook, The Limits of Antitrust, 63 Tex. L. Rev. 1 (1984).

[14] Dirk Auer & Lazar Radic, The Growing Legacy of Intel, 14 J. Eur. Comp. L. & Pract. 15 (2023), (“Competition cases routinely hinge on the fundamental distinction between conduct that anti-competitively serves to exclude competitors, on the one hand, and competition on the merits that may lead firms to exit the market, on the other. (…) anticompetitive foreclosure and competition on the merits both ultimately result in the same observable outcome: namely, that rivals exit the market. In order to draw the line, policymakers must infer both the root causes and the effects of firms’ market exit”.)

[15] Jean Tirole, Competition and the Industrial Challenge for the Digital Age, 15 Annual Rev. of Econ. 573, 581 (2023), available at https://www.annualreviews.org/content/journals/10.1146/annurev-economics-090622-024222.

[16] Alba Ribera, La Regulación de los Ecosistemas Digitales Frente a las Relaciones Complejas se los Operadores Económicos, Centro Competencia (18 Oct. 2023), https://centrocompetencia.com/regulacion-ecosistemas-digitales-relaciones-complejas-operadores-economicos (Free translation of the following text in Spanish: “Uno de los mayores ejemplos de la dicotomía que se erige entre los distintos tipos de consecuencias que se pueden generar por la captura regulatoria de los ecosistemas digitales lo podemos encontrar en la reciente decisión de Meta, de no lanzar su nuevo servicio Threads en el Espacio Económico Europeo. En la medida en que su servicio podría interpretarse de forma que cayera dentro de la definición de un “servicio básico de plataforma” perteneciente a la categoría de redes sociales en línea” (listada por la LMD), Meta decidió abstenerse de entrar en el mercado europeo, por la carga desproporcionada que le supondría las exigentes obligaciones impuestas por la LMD. Cabe notar que Threads es aún un servicio entrante en el mercado de redes sociales en línea, en contraste con la posición predominante ocupada por la actual X (anteriormente conocida como Twitter). De esta forma, observamos que la categorización como servicio básico de plataforma unifica y elimina todos los matices que el propio juego de la libre competencia opera respecto de servicios entrantes en los mercados.”).

[17] Press Release, Facebook and Instagram to Offer Subscription for No Ads in Europe, Meta (30 Oct. 2023), https://about.fb.com/news/2023/10/facebook-and-instagram-to-offer-subscription-for-no-ads-in-europe.

[18] Geoffrey Manne & Dirk Auer, Brussels Effect or Brussels Defect: Digital Regulation in Emerging Markets, Truth on the Market (20 Dec. 2022), https://truthonthemarket.com/2022/12/20/brussels-effect-or-brussels-defect-digital-regulation-in-emerging-markets.

[19] Adam Kovacevich, Europe’s Digital Market Act Fails Consumers, Chamber of Progress (4 Mar. 2024), https://medium.com/chamber-of-progress/europes-digital-market-act-fails-consumers-dcaf70cc548c.

[20] Id.

[21] Id.

[22] See Jon Porter & David Pierce, Apple Is Bringing Sideloading and Alternate App Stores to the iPhone, The Verge (25 Jan. 2024), https://www.theverge.com/2024/1/25/24050200/apple-third-party-app-stores-allowed-iphone-ios-europe-digital-markets-act.

[23] See Apple, Complying with the Digital Markets Act (2024), available at https://developer.apple.com/security/complying-with-the-dma.pdf.

[24] Kim Mackrael, Apple’s Hold on the App Store Is Loosening, at Least in Europe, Wall St. J. (16 Aug. 2024), https://www.wsj.com/tech/epic-games-apple-app-store-europe-44ceda50.

[25] Dan Goodin, Novel Technique Allows Malicious Apps to Escape iOS and Android Guardrails, ArsTechnica (21 Aug. 2024), https://arstechnica.com/security/2024/08/novel-technique-allows-malicious-apps-to-escape-ios-and-android-guardrails.

[26] See id. (“Both mobile operating systems employ mechanisms designed to help users steer clear of apps that steal their personal information, passwords, or other sensitive data. iOS bars the installation of all apps other than those available in its App Store, an approach widely known as the Walled Garden.”).

[27] Josephine Wolff, Software crash Exposes Tensions Between Security and Competition, Financial Times (28 Jul. 2024), https://www.ft.com/content/60dde560-194a-40d1-8c98-1d96d6d019a0.

[28] Dirk Auer, The Broken Promises of Europe’s Digital Regulation, Truth on the Market (12 Mar. 2024), https://truthonthemarket.com/2024/03/12/the-broken-promises-of-europes-digital-regulation.

[29] Mirai, DMA’s Negative Impact (Feb. 2024), available at LinkedIn, https://www.linkedin.com/feed/update/urn:li:activity:7161330551709138945 (“Since the implementation of the DMA on the 19th of January, the number of clicks from Google Hotel Ads to hotel websites has decreased by 17.6% in EU countries compared to the rest of the world.”).

[30] See Press Release, Parliament Passes News Media and Digital Platforms Mandatory Bargaining Code, Treasury Portfolio Ministers and Minister for Communications, Urban Infrastructure, Cities and the Arts, (25 Feb. 2021), https://ministers.treasury.gov.au/ministers/josh-frydenberg-2018/media-releases/parliament-passes-news-media-and-digital-platforms.

[31] Josh Taylor, Facebook’s Potential News Ban Already Affecting Smaller Australian Media Outlets, Inquiry Told, The Guardian (21 Jun. 2024), https://www.theguardian.com/media/article/2024/jun/21/facebooks-potential-news-ban-already-affecting-smaller-australian-media-outlets-inquiry-told.

[32] See, e.g., Thomas Lambert, Tech Platforms and Market Power: What’s the Optimal Policy Response?, Mercatus Working Paper (Nov. 2021), at 14, available at https://www.mercatus.org/research/working-papers/tech-platforms-and-market-power-whats-optimal-policy-response (“Because they are more rigid and prescriptive than antitrust’s flexible standards, and thus less likely to be appropriate for a broad range of diverse firms, ex ante rules addressing market power concerns tend to be limited in scope. They are usually tailored for a particular industry or group of firms. Antitrust’s standards are focused on ends rather than specific means, and are therefore less likely to ‘misfire’ when applied broadly.”).

[33] See Richard Posner, Antitrust Law (2nd. ed. 2001), at 39 (“Rules are generally simpler and cheaper to enforce than standards and provide clearer guidance both to the people subject to them and to the courts that administer them. But they are often either underinclusive or overinclusive, and sometimes they are both at the same time. They are especially apt to fail as a sensible method of lawmaking when the relation of the rule’s purpose to the fact of facts that it makes determinative or legality is unclear. In such cases, the decision whether to characterize the case as falling within the domain of the rule may depend on the same factors that would determine legality under a standard.”)

[34] Issues paper, supra note 1, at 8.

[35] Kwon Soon-wan &Yeom Hyun-a, South Korea Hits Pause on Anti-Monopoly Platform Act Targeting Google, Apple, The Chosun Daily (8 Feb. 2024), https://www.chosun.com/english/national-en/2024/02/08/A4U4X6TWEFFOXF7ITCS5K6SZN4.

[36] Issues paper, supra note 1, at 10.

[37] Id. at 9-10.

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

[39] See, Most Used Messenger by Brand in Australia as of June 2024, Statista (31 Jul. 2024), https://www.statista.com/forecasts/1187986/most-used-messenger-by-brand-in-australia.

[40] See supra note 9 and accompanying text.

[41] Catherine Tucker, Network Effects and Market Power: What Have We Learned in the Last Decade?, Antitrust 72, 75-76 (2018) (“the rise and fall of such platforms may also be more dramatic and renders such platforms far more vulnerable to a marginally superior competitor. The sudden decline of MySpace can be explained by the idea that given limited ability to spend time on two similar social networks, users switched far more quickly to a competitor.”).

^[42] For a review of the literature on increasing returns to scale in data (a topic somewhat broader than data-network effects), see Geoffrey Manne & Dirk Auer, Antitrust Dystopia and Antitrust Nostalgia: Alarmist Theories of Harm in Digital Markets and Their Origins, 28 Geo. Mason L. Rev. 1281, 1344 (2021).

[43] Issues paper, supra note 1, at 10.

[44] See generally William J. Baumol, Contestable Markets: An Uprising in the Theory of Industry Structure, 72 Am. Econ. Rev. 1 (1982); William J. Baumol, John Panzar & Robert D. Willig, Contestable Markets and the Theory of Industry Structure (revised ed. 1988).

[45] See, e.g., Andrew Lanxon, Android vs. iPhone: 15 Years of Innovation Through Rivalry, CNET (24 Apr. 2024), https://www.cnet.com/tech/mobile/smartphone-showdown-15-years-of-android-vs-iphone.

[46] Issues paper, supra note 1, at 13-14.

[47] Nathan Newman, Taking on Google’s Monopoly Means Regulating Its Control of User Data, Huffington Post (24 Sep. 2013), http://www.huffingtonpost.com/nathan-newman/taking-on-googlesmonopol_b_3980799.html.

[48] See, e.g., Lina Khan & K. Sabeel Rahman, Restoring Competition in the U.S. Economy, in Untamed: How to Check Corporate, Financial, and Monopoly Power (Nell Abernathy, Mike Konczal, & Kathryn Milani, eds., 2016), at 23. (“From Amazon to Google to Uber, there is a new form of economic power on display, distinct from conventional monopolies and oligopolies…, leverag[ing] data, algorithms, and internet-based technologies… in ways that could operate invisibly and anticompetitively.”); Mark Weinstein, I Changed My Mind—Facebook Is a Monopoly, Wall St. J. (1 Oct. 2021), https://www.wsj.com/articles/facebook-is-monopoly-metaverse-users-advertising-platforms-competition-mewe-big-tech-11633104247 (“[T]he glue that holds it all together is Facebook’s monopoly over data…. Facebook’s data troves give it unrivaled knowledge about people, governments—and its competitors.”).

[49] See, generally, Abigail Slater, Why “Big Data” Is a Big Deal, The Reg. Rev. (6 Nov. 2023), https://www.theregreview.org/2023/11/06/slater-why-big-data-is-a-big-deal; Amended Complaint at ¶36, United States v. Google, 1:20-cv-03010- (D.D.C. 2020); Complaint at ¶37, United States v. Google, 1:23-cv-00108 (E.D. Va. 2023), https://www.justice.gov/opa/pr/justice-department-sues-google-monopolizing-digital-advertising-technologies (“Google intentionally exploited its massive trove of user data to further entrench its monopoly across the digital advertising industry.”).

[50] See, e.g., Press Release, Commission Launches Calls for Contributions on Competition in Virtual Worlds and Generative AI, European Commission (9 Jan. 2024), https://ec.europa.eu/commission/presscorner/detail/en/IP_24_85; Krysten Crawford, FTC’s Lina Khan Warns Big Tech over AI, SIEPR (3 Nov. 2020), https://siepr.stanford.edu/news/ftcs-lina-khan-warns-big-tech-over-ai (“Federal Trade Commission Chair Lina Khan delivered a sharp warning to the technology industry in a speech at Stanford on Thursday: Antitrust enforcers are watching what you do in the race to profit from artificial intelligence.”) (emphasis added).

[51] See, e.g., John M. Newman, Antitrust in Digital Markets, 72 Vand. L. Rev. 1497, 1501 (2019) (“[T]he status quo has frequently failed in this vital area, and it continues to do so with alarming regularity. The laissez-faire approach advocated for by scholars and adopted by courts and enforcers has allowed potentially massive harms to go unchecked.”); Bertin Martins, Are New EU Data Market Regulations Coherent and Efficient?, Bruegel Working Paper 21/23 (2023), https://www.bruegel.org/working-paper/are-new-eu-data-market-regulations-coherent-and-efficient (“Technical restrictions on access to and re-use of data may result in failures in data markets and data-driven services markets.”); Valéria Faure-Muntian, Competitive Dysfunction: Why Competition Law Is Failing in a Digital World, The Forum Network (24 Feb. 2021), https://www.oecd-forum.org/posts/competitive-dysfunction-why-competition-law-is-failing-in-a-digital-world.

[52] See Rana Foroohar, The Great US-Europe Antitrust Divide, Financial Times (5 Feb. 2024), https://www.ft.com/content/065a2f93-dc1e-410c-ba9d-73c930cedc14.

[53] See, e.g., Press Release, supra note 50.

[54] See infra, Section III.B. Commentators have also made similar claims; see, e.g., Ganesh Sitaram & Tejas N. Narechania, It’s Time for the Government to Regulate AI. Here’s How, Politico (15 Jan. 2024), https://www.politico.com/news/magazine/2024/01/15/sitaraman-artificial-intelligence-regulation-00134873 (“All that cloud computing power is used to train foundation models by having them “learn” from incomprehensibly huge quantities of data. Unsurprisingly, the entities that own these massive computing resources are also the companies that dominate model development. Google has Bard, Meta has LLaMa. Amazon recently invested $4 billion into one of OpenAI’s leading competitors, Anthropic. And Microsoft has a 49 percent ownership stake in OpenAI—giving it extraordinary influence, as the recent board struggles over Sam Altman’s role as CEO showed.”).

[55] Press Release, supra note 50.

[56] Comment of U.S. Federal Trade Commission to the U.S. Copyright Office, Artificial Intelligence and Copyright, Docket No. 2023-6 (30 Oct. 2023), at 4, https://www.ftc.gov/legal-library/browse/advocacy-filings/comment-federal-trade-commission-artificial-intelligence-copyright (emphasis added).

[57] Jonathan Kanter, Remarks at the Promoting Competition in AI Conference (30 May 2024), https://youtu.be/yh–1AGf3aU?t=424.

[58] Karin Matussek, AI Will Fuel Antitrust Fires, Big Tech’s German Nemesis Warns, Bloomberg (26 Jun. 2024), https://www.bloomberg.com/news/articles/2024-06-26/ai-will-fuel-antitrust-fires-big-tech-s-german-nemesis-warns?srnd=technology-vp.

[59] Id.

[60] See, e.g., Joe Caserta, Holger Harreis, Kayvaun Rowshankish, Nikhil Srinidhi, & Asin Tavakoli, The Data Dividend: Fueling Generative AI, McKinsey Digital (15 Sep. 2023), https://www.mckinsey.com/capabilities/mckinsey-digital/our-insights/the-data-dividend-fueling-generative-ai (“Your data and its underlying foundations are the determining factors to what’s possible with generative AI.”).

[61] See, e.g., Tim Keary, Google DeepMind’s Achievements and Breakthroughs in AI Research, Techopedia (11 Aug. 2023), https://www.techopedia.com/google-deepminds-achievements-and-breakthroughs-in-ai-research; see, e.g., Will Douglas Heaven, Google DeepMind Used a Large Language Model to Solve an Unsolved Math Problem, MIT Technology Review (14 Dec. 2023), https://www.technologyreview.com/2023/12/14/1085318/google-deepmind-large-language-model-solve-unsolvable-math-problem-cap-set; see also, A Decade of Advancing the State-of-the-Art in AI Through Open Research, Meta (30 Nov. 2023), https://about.fb.com/news/2023/11/decade-of-advancing-ai-through-open-research; see also, 200 Languages Within a Single AI Model: A Breakthrough in High-Quality Machine Translation, Meta, https://ai.meta.com/blog/nllb-200-high-quality-machine-translation (last visited 18 Jan. 2023).

[62] See, e.g., Jennifer Allen, 10 Years of Siri: The History of Apple’s Voice Assistant, Tech Radar (4 Oct. 2021), https://www.techradar.com/news/siri-10-year-anniversary; see also Evan Selleck, How Apple Is Already Using Machine Learning and AI in iOS, Apple Insider (20 Nov. 2023), https://appleinsider.com/articles/23/09/02/how-apple-is-already-using-machine-learning-and-ai-in-ios; see also, Kathleen Walch, The Twenty Year History Of AI At Amazon, Forbes (19 Jul. 2019), https://www.forbes.com/sites/cognitiveworld/2019/07/19/the-twenty-year-history-of-ai-at-amazon.

[63] See infra Section III.C.

[64] See, e.g., Cédric Argenton & Jens Prüfer, Search Engine Competition with Network Externalities, 8 J. Comp. L. & Econ. 73, 74 (2012).

[65] John M. Yun, The Role of Big Data in Antitrust, The GAI Report on the Digital Economy 220, 233 (2020), https://gaidigitalreport.com/2020/08/25/big-data-and-barriers-to-entry/#_ftnref50; see also, e.g., Robert Wayne Gregory, Ola Henfridsson, Evgeny Kaganer, & Harris Kyriakou, The Role of Artificial Intelligence and Data Network Effects for Creating User Value, 46 Acad. of Mgmt. Rev. 534 (2020) (final pre-print version at 4), http://wrap.warwick.ac.uk/134220 (“A platform exhibits data network effects if, the more that the platform learns from the data it collects on users, the more valuable the platform becomes to each user.”); Karl Schmedders, José Parra-Moyano, & Michael Wade, Why Data Aggregation Laws Could be the Answer to Big Tech Dominance, Silicon Republic (6 Feb. 2024), https://www.siliconrepublic.com/enterprise/data-ai-aggregation-laws-regulation-big-tech-dominance-competition-antitrust-imd.

[66] Nathan Newman, Search, Antitrust, and the Economics of the Control of User Data, 31 Yale J. Reg. 401, 409 (2014) (emphasis added); see also id. at 420 & 423 (“While there are a number of network effects that come into play with Google, [“its intimate knowledge of its users contained in its vast databases of user personal data”] is likely the most important one in terms of entrenching the company’s monopoly in search advertising…. Google’s overwhelming control of user data… might make its dominance nearly unchallengeable.”).

[67] See also Yun, supra note 65 at 229 (“[I]nvestments in big data can create competitive distance between a firm and its rivals, including potential entrants, but this distance is the result of a competitive desire to improve one’s product.”).

^[68] See Manne & Auer, Antitrust Dystopia and Antitrust Nostalgia, supra note 42.

[69] Andrei Hagiu & Julian Wright, Data-Enabled Learning, Network Effects, and Competitive Advantage, 54 RAND J. Econ. 638 (2023).

^[70] Id. at 639. The authors conclude that “Data-enabled learning would seem to give incumbent firms a competitive advantage. But how strong is this advantage and how does it differ from that obtained from more traditional mechanisms… .”

^[71] Id.

^[72] Bruno Jullien & Wilfried Sand-Zantman, The Economics of Platforms: A Theory Guide for Competition Policy, 54 Info. Econ. & Pol’y 10080, 101031 (2021).

^[73] Daniele Condorelli & Jorge Padilla, Harnessing Platform Envelopment in the Digital World, 16 J. Comp. L. & Pol’y 143, 167 (2020).

[74] See Hagiu & Wright, supra note 6969.

^[75] For a summary of these limitations, see generally Catherine Tucker, Network Effects and Market Power, supra note 41; see also Manne & Auer, supra note 42, at 1330.

[76] See Jason Furman, Diane Coyle, Amelia Fletcher, Derek McAuley, & Philip Marsden (Dig. Competition Expert Panel), Unlocking Digital Competition (2019) at 32-35 (“Furman Report”), available at https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/785547/unlocking_digital_competition_furman_review_web.pdf.

[77] Id. at 34.

[78] Id. at 35. To its credit, it should be noted, the Furman Report does counsel caution before mandating access to data as a remedy to promote competition. See id. at 75. That said, the Furman Report maintains that such a remedy should remain on the table because “the evidence suggests that large data holdings are at the heart of the potential for some platform markets to be dominated by single players and for that dominance to be entrenched in a way that lessens the potential for competition for the market.” Id. The evidence, however, does not show this.

^[79] Case COMP/M.9660 — Google/Fitbit, Commission Decision (17 Dec. 2020) (Summary at O.J. (C 194) 7), available at https://ec.europa.eu/competition/mergers/cases1/202120/m9660_3314_3.pdf, at 455.

^[80] Id. at 896.

[81] See Natasha Lomas, EU Checking if Microsoft’s OpenAI Investment Falls Under Merger Rules, TechCrunch (9 Jan. 2024), https://techcrunch.com/2024/01/09/openai-microsoft-eu-merger-rules.

^[82] Amended Complaint at 11, Meta/Zuckerberg/Within, Fed. Trade Comm’n. (2022) (No. 605837), available at https://www.ftc.gov/system/files/ftc_gov/pdf/D09411%20-%20AMENDED%20COMPLAINT%20FILED%20BY%20COUNSEL%20SUPPORTING%20THE%20COMPLAINT%20-%20PUBLIC%20%281%29_0.pdf.

[83] Amended Complaint (D.D.C), supra note 49, at ¶37.

[84] Amended Complaint (E.D. Va), id., at ¶8.

^[85] Merger Guidelines, US Dep’t of Justice & Fed. Trade Comm’n (2023) at 25, available at https://www.ftc.gov/system/files/ftc_gov/pdf/2023_merger_guidelines_final_12.18.2023.pdf.

[86] Merger Assessment Guidelines, Competition and Mkts. Auth (2021) at ¶7.19(e), available at https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/1051823/MAGs_for_publication_2021_–_.pdf.

[87] Furman Report, supra note 76, at ¶4.

[88] See, e.g., Chris Westfall, New Research Shows ChatGPT Reigns Supreme in AI Tool Sector, Forbes (16 Nov. 2023), https://www.forbes.com/sites/chriswestfall/2023/11/16/new-research-shows-chatgpt-reigns-supreme-in-ai-tool-sector/?sh=7de5de250e9c; Sujan Sarkar, AI Industry Analysis: 50 Most Visited AI Tools and Their 24B+ Traffic Behavior, Writerbuddy (last visited 15 Jul. 2024), https://writerbuddy.ai/blog/ai-industry-analysis.

^[89] See Krystal Hu, ChatGPT Sets Record for Fastest-Growing User Base, Reuters (2 Feb. 2023), https://www.reuters.com/technology/chatgpt-sets-record-fastest-growing-user-base-analyst-note-2023-02-01; Google: The AI Race Is On, App Economy Insights (7 Feb. 2023), https://www.appeconomyinsights.com/p/google-the-ai-race-is-on.

^[90] See Google Trends, https://trends.google.com/trends/explore?date=today%205-y&q=%2Fg%2F11khcfz0y2,%2Fg%2F11ts49p01g&hl=en (last visited 12 Jan. 2024) and https://trends.google.com/trends/explore?date=today%205-y&geo=US&q=%2Fg%2F11khcfz0y2,%2Fg%2F11ts49p01g&hl=en (last visited 12 Jan. 2024).

^[91] See David F. Carr, As ChatGPT Growth Flattened in May, Google Bard Rose 187%, Similarweb Blog (5 Jun. 2023), https://www.similarweb.com/blog/insights/ai-news/chatgpt-bard.

^[92] See Press Release, Introducing New AI Experiences Across Our Family of Apps and Devices, Meta (27 Sep. 2023), https://about.fb.com/news/2023/09/introducing-ai-powered-assistants-characters-and-creative-tools; Sundar Pichai, An Important Next Step on Our AI Journey, Google Keyword Blog (6 Feb. 2023), https://blog.google/technology/ai/bard-google-ai-search-updates.

^[93] See Ion Prodan, 14 Million Users: Midjourney’s Statistical Success, Yon (19 Aug. 2023), https://yon.fun/midjourney-statistics; see also Andrew Wilson, Midjourney Statistics: Users, Polls, & Growth [Oct 2023], ApproachableAI (Oct. 13, 2023), https://approachableai.com/midjourney-statistics.

^[94] See Hema Budaraju, New Ways to Get Inspired with Generative AI in Search, Google Keyword Blog (12 Oct. 2023), https://blog.google/products/search/google-search-generative-ai-october-update; Imagine with Meta AI, Meta (last visited 12 Jan. 2024), https://imagine.meta.com.

^[95] Catherine Tucker, Digital Data, Platforms and the Usual [Antitrust] Suspects: Network Effects, Switching Costs, Essential Facility, 54 Rev. Indus. Org. 683, 686 (2019).

^[96] Manne & Auer, supra note 42, at 1345.

[97] See, e.g., Stefanie Koperniak, Artificial Data Give the Same Results as Real Data—Without Compromising Privacy, MIT News (3 Mar. 2017), https://news.mit.edu/2017/artificial-data-give-same-results-as-real-data-0303 (“[Authors] describe a machine learning system that automatically creates synthetic data—with the goal of enabling data science efforts that, due to a lack of access to real data, may have otherwise not left the ground. While the use of authentic data can cause significant privacy concerns, this synthetic data is completely different from that produced by real users—but can still be used to develop and test data science algorithms and models.”).

[98] See, e.g., Rachel Gordon, Synthetic Imagery Sets New Bar in AI Training Efficiency, MIT News (20 Nov. 2023), https://news.mit.edu/2023/synthetic-imagery-sets-new-bar-ai-training-efficiency-1120 (“By using synthetic images to train machine learning models, a team of scientists recently surpassed results obtained from traditional ‘real-image’ training methods.).

[99] Thibault Schrepel & Alex ‘Sandy’ Pentland, Competition Between AI Foundation Models: Dynamics and Policy Recommendations, MIT Connection Science Working Paper (Jun. 2023), at 8.

[100] Igor Susmelj, Optimizing Generative AI: The Role of Data Curation, Lightly (last visited 15 Jan. 2024), https://www.lightly.ai/post/optimizing-generative-ai-the-role-of-data-curation.

[101] See, e.g., Xiaoliang Dai, et al., Emu: Enhancing Image Generation Models Using Photogenic Needles in a Haystack, ArXiv (27 Sep. 2023) at 1, https://ar5iv.labs.arxiv.org/html/2309.15807 (“[S]upervised fine-tuning with a set of surprisingly small but extremely visually appealing images can significantly improve the generation quality.”); see also, Hu Xu, et al., Demystifying CLIP Data, ArXiv (28 Sep. 2023), https://arxiv.org/abs/2309.16671.

[102] Lauren Leffer, New Training Method Helps AI Generalize Like People Do, Sci. Am. (26 Oct. 2023), https://www.scientificamerican.com/article/new-training-method-helps-ai-generalize-like-people-do (discussing Brendan M. Lake & Marco Baroni, Human-Like Systematic Generalization Through a Meta-Learning Neural Network, 623 Nature 115 (2023)).

^[103] Timothy B. Lee, The Real Research Behind the Wild Rumors about OpenAI’s Q* Project, Ars Technica (8 Dec. 2023), https://arstechnica.com/ai/2023/12/the-real-research-behind-the-wild-rumors-about-openais-q-project.

^[104] Id.; see also GSM8K, Papers with Code (last visited 18 Jan. 2023), https://paperswithcode.com/dataset/gsm8k; MATH Dataset, GitHub (last visited 18 Jan. 2024), https://github.com/hendrycks/math.

^[105] See Lee, supra note 103.

[106] Geoffrey Manne & Ben Sperry, Debunking the Myth of a Data Barrier to Entry for Online Services, Truth on the Market (26 Mar. 2015), https://truthonthemarket.com/2015/03/26/debunking-the-myth-of-a-data-barrier-to-entry-for-online-services (citing Andres V. Lerner, The Role of ‘Big Data’ in Online Platform Competition (26 Aug. 2014), https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2482780.).

[107] See Catherine Tucker, Digital Data as an Essential Facility: Control, CPI Antitrust Chron. (Feb. 2020), at 11 (“[U]ltimately the value of data is not the raw manifestation of the data itself, but the ability of a firm to use this data as an input to insight.”).

[108] Or, as John Yun put it, data is only a small component of digital firms’ production function. See Yun, supra note 65, at 235 (“Second, while no one would seriously dispute that having more data is better than having less, the idea of a data-driven network effect is focused too narrowly on a single factor improving quality. As mentioned in supra Section I.A, there are a variety of factors that enter a firm’s production function to improve quality.”).

^[109] Luxia Le, The Real Reason Windows Phone Failed Spectacularly, History–Computer (8 Aug. 2023), https://history-computer.com/the-real-reason-windows-phone-failed-spectacularly.

[110] Introducing the GPT Store, Open AI (10 Jan. 2024), https://openai.com/blog/introducing-the-gpt-store.

^[111] See Michael Schade, How ChatGPT and Our Language Models Are Developed, OpenAI, https://help.openai.com/en/articles/7842364-how-chatgpt-and-our-language-models-are-developed; Sreejani Bhattacharyya, Interesting Innovations from OpenAI in 2021, AIM (1 Jan. 2022), https://analyticsindiamag.com/interesting-innovations-from-openai-in-2021; Danny Hernadez & Tom B. Brown, Measuring the Algorithmic Efficiency of Neural Networks, ArXiv (8 May 2020), https://arxiv.org/abs/2005.04305.

[112] See Yun, supra note 65 at 235 (“Even if data is primarily responsible for a platform’s quality improvements, these improvements do not simply materialize with the presence of more data—which differentiates the idea of data-driven network effects from direct network effects. A firm needs to intentionally transform raw, collected data into something that provides analytical insights. This transformation involves costs including those associated with data storage, organization, and analytics, which moves the idea of collecting more data away from a strict network effect to more of a ‘data opportunity.’”).

[113] Josh Sisco, POLITICO PRO Q&A: Exit Interview with DOJ Chief Antitrust Economist Susan Athey, Politico Pro (2 Jul. 2024), https://subscriber.politicopro.com/article/2024/07/politico-pro-q-a-exit-interview-with-doj-chief-antitrust-economist-susan-athey-00166281.

[114] Belle Lin, Open-Source Companies Are Sharing Their AI Free. Can They Crack OpenAI’s Dominance?, Wall St. J. (21 Mar. 2024), https://www.wsj.com/articles/open-source-companies-are-sharing-their-ai-free-can-they-crack-openais-dominance-26149e9c.

[115] Lerner, supra note 106, at 4-5 (emphasis added).

[116] See Clayton M. Christensen, The Innovator’s Dilemma: When New Technologies Cause Great Firms to Fail (2013).

[117] See David J. Teece, Dynamic Capabilities and Strategic Management: Organizing for Innovation and Growth (2009).

[118] Antitrust merger enforcement has long assumed that horizontal mergers are more likely to cause problems for consumers than vertical mergers. See Geoffrey A. Manne, Dirk Auer, Brian Albrecht, Eric Fruits, Daniel J. Gilman, & Lazar Radic, Comments of the International Center for Law and Economics on the FTC & DOJ Draft Merger Guidelines, International Center for Law & Economics (18 Sep. 2023), https://laweconcenter.org/resources/comments-of-the-international-center-for-law-and-economics-on-the-ftc-doj-draft-merger-guidelines.

[119] See Hagiu & Wright, supra note 69, at 69 (“We use our dynamic framework to explore how data sharing works: we find that it in-creases consumer surplus when one firm is sufficiently far ahead of the other by making the laggard more competitive, but it decreases consumer surplus when the firms are sufficiently evenly matched by making firms compete less aggressively, which in our model means subsidizing consumers less.”); see also Lerner, supra note 106.

[120] See, e.g., Hagiu & Wright, id. (“We also use our model to highlight an unintended consequence of privacy policies. If such policies reduce the rate at which firms can extract useful data from consumers, they will tend to increase the incumbent’s competitive advantage, reflecting that the entrant has more scope for new learning and so is affected more by such a policy.”); Jian Jia, Ginger Zhe Jin, & Liad Wagman, The Short-Run Effects of the General Data Protection Regulation on Technology Venture Investment, 40 Marketing Sci. 593 (2021) (finding GDPR reduced investment in new and emerging technology firms, particularly in data-related ventures); James Campbell, Avi Goldfarb, & Catherine Tucker, Privacy Regulation and Market Structure, 24 J. Econ. & Mgmt. Strat. 47 (2015) (“Consequently, rather than increasing competition, the nature of transaction costs implied by privacy regulation suggests that privacy regulation may be anti-competitive.”).

Judge Ada Brown of the U.S. District Court for the Northern District of Texas issued an Aug. 20 order effectively striking down the Federal Trade Commission’s (FTC) April 2024 rule barring noncompete clauses (“noncompetes”) in employment contracts. Ryan LLC, a global tax-services and software provider, had challenged the rule, which had been scheduled to take effect in September.

Read the full piece here.

After its antitrust victory against Google this month, the U.S. Justice Department became the dog that finally caught the car. After years of barking about Big Tech’s market power, enforcers have secured a ruling against Google for how it distributes its search engine through contracts with smartphone manufacturers like Apple and browsers like Mozilla. But the path forward is murkier than headlines suggest.

Read the full piece here.

To butcher a Winston Churchill quote, it’s not yet clear if this is the beginning of the end, or just the end of the beginning, for children’s online-safety bills.

Such legislation has been all the rage in recent years, earning bipartisan support at both the federal and state level. A version of the Kids Online Safety Act (KOSA)—with legislative text that was merged with the somewhat-related Children and Teen’s Online Privacy Protection Act (COPPA 2.0)—passed the U.S. Senate earlier this summer, although it now appears that it will face legislative roadblocks in the U.S. House.

Read the full piece here.

In an Aug. 5 order, the U.S. District Court for the District of Columbia held that Google engaged in illegal monopolization of internet “general search services” and “general text search advertising.” This decision, dubbed “an historic win for the American people” by U.S. Attorney General Merrick Garland, may face tough sledding on appeal.

Read the full piece here.

Wall Street gets worked up over each jobs report or move by the Federal Reserve, but Main Street’s prosperity hinges on a less flashy metric: productivity growth. For nearly two decades, America’s economic engine has been sputtering, with labor-productivity growth crawling at just 1% annually, down from the more than 2% growth in the late 1990s and early 2000s. This decline has cost the average American household around $10,000 per year in income.

Read the full piece here.

Yep, more about noncompetes. I’ve been at this a bit. I’m aware. Just last week, and then again here, here, here, here, and here at Truth on the Market; here in a more formal journal article; and here with my International Center for Law & Economics (ICLE) colleagues and scholars of law and economics.

Maybe “a bit” doesn’t quite cut it. But I do have a few excuses…

Read the full piece here.

If you’ve been keeping up with the machinations of the Federal Communications Commission’s (FCC) latest attempt to regulate broadband under Title II of the Communications Act, you know that providers are challenging the rules in the 6th U.S. Circuit Court of Appeals. The court has put that case on the fast track.

Earlier this week, the International Center for Law & Economics (ICLE) and the Competitive Enterprise Institute (CEI) submitted an amicus brief urging the court to vacate the FCC’s order.

Read the full piece here.

Readers of Truth on the Market are no doubt aware of Judge Amit Mehta’s Aug. 5 decision in the Google search antitrust case—that is, his 286-page memorandum and order finding Google liable for violating Section 2 of the Sherman Act (specifically, illegal monopoly maintenance in two markets: general search services and general text advertising).

Read the full piece here.

Noncompetes have been a subject of much interest here at Truth on the Market (see Alden Abbott, Brian Albrecht, Corbin Barthold, Gus Hurwitz, Richard Pierce Jr., and your humble-if-obsessive scribe here, here, here, here, and here). They’re also something I’ve studied independently—first, at the Federal Trade Commission (FTC) and then, for example, here.

Read the full piece here.

The past four years have been heady times for those pursuing a more aggressive antitrust-enforcement regime. Critics of the Biden administration’s competition policy have charged that Federal Trade Commission Chair Lina Khan and Assistant U.S. Attorney General Jonathan Kanter have pursued a reflexive “big is bad” policy that will hurt consumers and the American economy’s dynamism and global competitiveness.

Anewly released draft of the European Union’s proposed monopolization guidelines suggest they could pose a new threat to innovative business practices that promote high-tech economic growth. The EU should scrap the draft and U.S. antitrust enforcers should likewise reject its approach.

Read the full piece here.

As in the Public Enemy song that gives this post its title, the hype about alleged competition risks in the artificial intelligence (AI) “market” is a sequel—and not a good one—to the hyperbolic and dystopian view that has informed several recent antitrust-policy proposals and demands for tougher enforcement of competition laws, particularly in digital markets. As we will explain, the evidence tells a different story, and there are plenty of reasons to be rather cautious before taking any action in the AI sector.

Read the full piece here.

Google recently announced that it has changed its plans to phase out third-party cookies in the Chrome web browser. The company had previously planned to disable third-party cookies in Chrome, a change supported by many in the privacy-stakeholder community, but which was met with criticism from the adtech industry and competition lawyers. Google’s new plans face similar criticism, while raising additional concerns from privacy advocates.

In short, Google is caught between EU privacy laws like the General Data Protection Regulation (GDPR) and competition laws like the Digital Markets Act (DMA). And despite claims that those laws are not in conflict, this episode very clearly shows how different interpretations are clearly in tension.

Read the full piece here.

Over the past decade, the intersection of privacy and antitrust has become a significant focus in the literature on data’s role in digital markets. In a landscape where platforms acquire data to strategically offer sellers preferential access to consumers’ attention, personal data is indeed the most valuable asset in a platform’s information arsenal. As a result, privacy has moved to the forefront, with policy makers keen to evaluate whether data accumulation strategies could compromise individuals’ privacy and reinforce platforms’ market power. Consequently, there is a growing argument that the unique characteristics of digital markets and the potential uses of data in the digital economy require an approach that integrates privacy considerations into antitrust enforcement, fostering close collaboration between antitrust authorities and data protection regulators.

Read the full piece here.

TL;DR

Background: Capital One Financial Corp.’s proposed acquisition of Discover Financial Services could transform competition in the payments space and confer significant benefits to consumers, especially those underserved by traditional banks.

While the merged entity would be the sixth-largest U.S. bank by assets, those would account for just 3% of total domestic bank assets. More significantly, the merged company would be the third-largest credit-card issuer by purchase volume.

But… The deal requires approval from the Federal Reserve, the U.S. Office of the Comptroller of the Currency, and the Federal Deposit Insurance Corp. In addition, the U.S. Justice Department (DOJ) has authority to block the merger under antitrust law.

However… The merger will likely increase competition, innovation, and financial inclusion. For example, it would likely facilitate expanded free checking for lower-income consumers and foster continued innovation in serving sub- and near-prime borrowers. 

The merger will spark competition in payment networks by expanding Discover’s reach in both debit and credit cards to better compete with Visa, Mastercard, and American Express. This aligns with the frequently cited policymaker goal of increased competition in this space.

KEY TAKEAWAYS

GREATER COMPETITION IN PAYMENT NETWORKS

Discover’s credit-card network is the fourth-largest in the United States, accounting for only about 4% of payment volume. Discover has languished at that figure for roughly two decades, trailing far behind Visa, Mastercard, and American Express. 

For years, consumer advocates and policymakers have expressed concern about a perceived lack of competition in the credit-card network market, referring to Visa and Mastercard as a “duopoly” and calling for legislation to expand competition. 

Capital One may be able to use its innovative culture and marketing savvy to leverage Discover’s card network and allow it to compete more successfully.

A POSSIBLE RETURN TO DEBIT REWARDS CARDS

By switching its debit cards to Discover’s payment networks, Capital One might offer more attractive products to depositors. In particular, it could expand access to free checking accounts with no minimum balance requirements to a wider range of low-income consumers.  It also could offer debit cards with cashback to lower-income consumers who would not qualify for credit cards. The benefits for this important underserved community could be tremendous.

A NOT-SO-MEGA MERGER

Critics of the merger note a combined Capital One-Discover would be the sixth-largest bank by assets and the third-largest credit-card issuer by purchaser volume. But the merged company would hold only 3% of all domestic bank assets, a trivial amount compared to such industry behemoths as JPMorgan Chase, Citibank, and Bank of America.

While the merged firm would be the largest holder of credit-card debt, accounting for nearly 22% of outstanding credit-card loans by dollar amount, that amount remains below the merger threshold of all conventional antitrust analysis.

IMPROVED DATA SECURITY AND FRAUD PROTECTION

The merger has the potential to improve fraud detection in several ways. When Capital One’s debit cards are moved to Discover’s networks, they will no longer be subject to routing requirements under the so-called “Durbin amendment.” As a result, all transactions on those cards will be monitored directly by Capital One’s systems. In addition, Capital One will be able to implement its highly innovative fraud-detection and prevention systems across all Discover networks.

WHAT ABOUT SUBPRIME BORROWERS?

Critics of the merger focus much of their concern on potential harms to “near-prime” or “subprime” segments of borrowers. 

There is, however, little evidence that “near-prime” or “subprime” borrowers constitute a distinct market for antitrust purposes. In particular, a consumer’s credit status is rarely static over time. Due to changes in income and other circumstances, a subprime borrower today may be a prime borrower next year, and vice versa. 

Moreover, Capital One gradually gained its market share in the “subprime” space through its data-driven strategy. This has enabled the company to identify lower-risk individuals in (otherwise) higher-risk groups, thereby serving otherwise underserved consumers, while limiting default risk. It also provides opportunities for these consumers to migrate toward a lower-risk category by gradually increasing the size of their credit lines as they demonstrate creditworthiness.

For more on this issue, see The Capital One-Discover Merger: A Law and Economics Analysis.

U.S. District Court Judge Amit Mehta ruled in an Aug. 5 order that Google violated antitrust law by improperly maintaining a monopoly. The case focused on “general search engines” (GSEs) used for internet search, and the impropriety was the manner through which Google secured distribution in partnering with internet-browser developers, mobile-device manufacturers, and wireless carriers.

Read the full piece here.

Even in the era of a Federal Trade Commission (FTC) led by Lina Khan, antitrust law can be a pretty hum-drum affair, consumed with minutiae about relevant markets, market shares, the Herfindahl-Hirschman Index (HHI), unilateral effects, coordinated effects, and possible efficiencies. Things get just a bit more interesting when there are allegations of a conspiracy to restrict output and raise prices.

Then, there are the once-in-a-lifetime cases that turn knobs all the way up to 11. And that’s the recent National Football League (NFL) “Sunday Ticket” litigation. Plaintiffs were a class of DirecTV “NFL Sunday Ticket” subscribers who claimed NFL teams colluded with broadcasters to suppress the availability of out-of-market NFL game telecasts, driving up the price paid for Sunday Ticket. After a three-week trial, the jury came back with an eye-popping $4.7 billion award against the NFL that could have been trebled to more than $14 billion.

Read the full piece here.

About a month ago, the Wall Street Journal reported that the Federal Trade Commission (FTC) was preparing an antitrust suit against the nation’s three largest pharmaceutical benefit managers (PBMs), the intermediaries who negotiate drug prices on behalf of insurers and who manage benefits for nearly nine in 10 insured Americans. 

This development followed a two-year investigation of the top six largest American PBMs, and a recent FTC interim report that suggested PBMs engage in anticompetitive practices that hike prices for patients. Those were alleged to include negotiating and accepting rebates from pharmaceutical companies to favorably place their on-brand drugs on insurer formularies—the lists of drugs that an insurer will cover—while imposing onerous conditions for the coverage of competing substitute drugs.

Read the full piece here.

Ferris Bueller famously said: “Life moves pretty fast. If you don’t stop and look around once in a while, you could miss it.”

The same could be said for the Federal Communications Commission’s (FCC) latest effort to regulate broadband providers under Title II of the Communications Act, under what is commonly referred to as “net neutrality.”

Read the full piece here.

Among the high-profile First Amendment cases heard by the Supreme Court this past term were NRA v. Vullo, which yielded a 9-0 opinion against jawboning—or the “use of official speech to inappropriately compel private action”—and the NetChoice cases, Moody v. NetChoice and NetChoice v. Paxton—which made clear that social media sites are First Amendment speakers. Nestled between these was Murthy v. Missouri, in which a six-justice majority found that the plaintiffs lacked standing to challenge alleged government efforts to pressure social media platforms to suppress their speech.

Read the full piece here.

The U.S. Justice Department (DOJ) and Federal Trade Commission (FTC), in tandem with their fellow competition-law enforcers from Europe (the European Commission) and the United Kingdom (the Competition and Markets Authority, or CMA), issued a joint statement July 23 titled “Joint Statement on Competition in Generative AI Foundation Models and AI Products.”

Read the full piece here.

The European Commission appears resolved to tear down Apple’s “walled garden.” Following a complaint filed by Spotify, the Commission has already fined Apple an exorbitant €1.8 billion for allegedly abusing its dominant position in the market for distributing streaming-music apps to iPhone and iPad users (a case where Apple was found to be dominant in a narrowly defined product market encompassing only its own products).

Read the full piece here.

The next president, whether a Democrat or a Republican, should place a high priority on federal regulatory reform, in order to promote good jobs, vibrant firms, and a stronger American economy.

Regulation is, of course, appropriate to address such problems as public health and safety and environmental pollution. But a growing body of scholarship finds that overregulation has slowed economic growth and reduced American prosperity. Fortunately, revamped federal regulatory review, centered in the White House, can play a key role in helping tame the regulatory behemoth.

Read the full piece here.

INTEREST OF AMICI CURIAE*

The International Center for Law & Economics (“ICLE”) is a nonprofit, non-partisan global research and policy center that builds intellectual foundations for sensible, economically grounded policy. ICLE promotes the use of law and economics methodologies and economic learning to inform policy debates and has longstanding expertise in evaluating law and policy.

ICLE scholars have written extensively in the areas of telecommunications and broadband policy. This includes white papers, law journal articles, and amicus briefs touching on issues related to the provision and regulation of broadband Internet service.

The Competitive Enterprise Institute (“CEI”) is a nonprofit organization headquartered in Washington, D.C., dedicated to promoting the principles of free markets and limited government. Since 1984, CEI has carried out its mission through policy analysis, commentary, and litigation. This case interests CEI due to the massive negative effects of arbitrary government enforcement over the internet.

SUMMARY OF THE ARGUMENT

Since the rise of the commercial internet in the nineties, the regulatory approach to broadband regulation in the United States has been relatively “hands-off.” This was an intentional choice by Congress and regulators. After the Telecommunications Act of 1996, the Stevens Report concluded that “Internet access services are appropriately classed as information, rather than telecommunications, services.” Federal-State Joint Board on Universal Service, 13 FCC Rcd. 11501, 11536 (1998). This approach was maintained by the FCC for all but 3 of the last 28 years, when, between 2015 and 2018, the Commission decided to reclassify broadband access as a telecommunications service. The FCC again returned to classifying broadband access as a Title I information service in 2018. See Restoring Internet Freedom, 33 FCC Rcd. 311 (2018).

Under the current “light-touch” Title I approach to broadband regulation, the marketplace for broadband in the U.S. is competitive, dynamic, and generally serves consumers well. Since 2018, more households are connected to the internet; broadband speeds have increased, while prices have fallen; more households are served by multiple providers; and new technologies like satellite and 5G have expanded internet access and intermodal competition among providers. See Eric Fruits, Geoffrey A. Manne, Ben Sperry, & Kristian Stout, Dynamic Competition in Broadband Markets: A 2024 Update (ICLE White Paper, Jun. 2024), https://laweconcenter.org/wp-content/uploads/2024/06/Broadband-Competition-2024-Update.pdf. The FCC’s own data suggests that 91% of Americans have access to high-speed broadband under its new and faster definition adopted earlier this year. See 2024 706 Report, FCC 24-27, GN Docket No. 22-270, ¶¶ 20, 22 (Mar. 18, 2024). The evidence would suggest that continuing the current regulatory approach is the best path forward.

Instead, the FCC has decided again to reclassify broadband as a Title II telecommunications service. See Safeguarding and Securing the Open Internet; Restoring Internet Freedom, 89 Fed. Reg. 45404 (May 22, 2024) [hereinafter “Order”]. While the FCC does use its forbearance authority to “tailor” its rules to broadband, the FCC retains significant regulatory authority that will dampen investment incentives and harm consumers. The rules themselves, particularly the general conduct rule and Section 214 authority over licenses, leave the FCC with nearly unbounded authority over commercial business decisions by broadband providers. As a result, broadband providers are highly likely to avoid possible litigation by getting their business plans pre-approved by the FCC. This sets up the FCC to be a de facto central planner over broadband.

The Order fails in two important respects.

First, the Order violates the Major Questions Doctrine (“MQD”) by claiming expansive authority to reclassify broadband as a Title II service while rewriting the statute to fit under its forbearance authority. The MQD is violated here because, even after the Supreme Court’s decision in Loper Bright Enterprises v. Raimondo, 144 S. Ct. 2244 (2024), that doctrine’s requirement of “clear congressional authorization” for major policy decisions remains in force. West Virginia v. EPA, 597 U.S. 697, 723 (2022). The question of broadband classification under the Communications Act is, almost undeniably, one of vast economic and political significance. The Communications Act does not clearly classify broadband as a Title II telecommunications service, as many courts have now ruled. In fact, the “tailoring” of the statute through a broad application of forbearance authority to many of its main provisions shows that this is the wrong interpretative turn.

Second, the Order’s vague rules would give the FCC nearly unlimited authority over the business decisions of broadband providers—authority that would allow the FCC to engage in arbitrary and discriminatory enforcement. Despite forbearance, the FCC retains the ability to regulate broadband providers as they see fit under Title II. This is particularly the case under the general conduct rule as well as under the Section 214 authority over licenses for entry along with the associated advisory opinion process. The Order is arbitrary and capricious because it fails to justify its vague rules: those rules will result in regulatory uncertainty that will dampen investment incentives for broadband providers to extend, maintain, and upgrade their networks. Ultimately, that harms consumers.

The FCC’s solution in search of a problem should be rejected. Title II reclassification is bad policy, but even worse law.

ARGUMENT

I. The Order Violates the Major Questions Doctrine.

A. The Major Questions Doctrine Still Applies Post-Loper Bright to Ambiguous Grants of Authority Over Questions of Vast Economic and Political Significance.

After the Supreme Court officially overruled Chevron, some argued that this ruling raised questions about the status of the MQD. Indeed, the FCC argued in its briefs that the MQD can’t be read to “supplant the best-reading analysis required by Loper Bright.” See Respondents Supp. Br. Regarding Loper Bright, at 8. However, this court’s decision to apply the MQD and grant a stay in this case is entirely consistent with Loper Bright. Courts must presume that “Congress intends to make major policy decisions itself, not leave those decisions to agencies.” West Virginia v. EPA, 597 U.S. 697, 723 (2022) (quoting United States Telecom Ass’n. v. FCC, 855 F.3d 381, 419 (DC Cir. 2017) (Kavanaugh, J., dissenting from denial of rehearing en banc)). When an agency’s advocates argue that the agency has immense powers that are unspoken in statute, “both separation of powers principles and a practical understanding of legislative intent” should make judges “ ‘reluctant to read into ambiguous statutory text’ the delegation claimed to be lurking there.” Id. (quoting Util. Air Regul. Grp. v. EPA, 573 U.S. 302, 324 (2014) (“UARG”)). In such cases, “something more than a merely plausible textual basis for the agency action is necessary”; instead, the agency “must point to ‘clear congressional authorization’ for the power it claims.” Id. In short, the MQD requires agencies to receive clear expressions of authority in order to regulate in areas of vast economic and political significance.

The application of the MQD is not contingent on the existence of Chevron. Chevron required deference to agency action when the statute was ambiguous. Loper Bright, 144 S. Ct. at 2254 (“Since our decision in Chevron U.S.A. v. Natural Resources Defense Council, Inc… we have sometimes required courts to defer to ‘permissible’ agency interpretations of statutes who agencies administer—even when a reviewing court reads the statute differently.”). The MQD, on the other hand, requires clear and express delegation to agencies when they assert authority over areas of vast economic and political significance. Indeed, in the pre-Loper Bright world, the MQD had the effect of displacing Chevron deference in some circumstances, but that was not its primary doctrinal function. Loper Bright only overruled Chevron and its progeny that gave agencies deference under ambiguous statutes. By and large, Loper Bright left the MQD alone.

Justice Gorsuch’s concurrence in West Virginia v. EPA illuminates certain aspects of the MQD, and this court should give that concurrence appropriate weight. As Justice Gorsuch explained, the major questions doctrine is appropriately understood as a “clear statement rule”; he compared it to other canons of statutory construction that help avoid constitutional problems, such as canons guarding against retrospective application of laws and protecting sovereign immunity. See West Virginia v. EPA, 597 U.S. at 736-37 (Gorsuch, J., concurring). “The major questions doctrine works in much the same way to protect the Constitution’s separation of powers.” Id. at 737. As such, the MQD can be understood as a rule of statutory construction that applies to agency assertions of authority, not simply an exception to Chevron deference.

Nothing in Loper Bright changes this analysis. In Loper Bright, the Court overruled Chevron, finding that courts must determine what the law is. Loper Bright, 144 S. Ct at 2273. Animating the Court’s concerns were separation of powers issues: “Courts must exercise their independent judgment in deciding whether an agency has acted within its statutory authority…” Id. The separation-of-powers rationale operates identically for both the MQD and for the rejection of Chevron: Congress must do its job and clearly delegate authority to the executive branch. The MQD remains foundational for courts when dealing with agency rules that implicate questions of vast economic or political significance.

B. Broadband Classification Under the Communications Act Is a Question of Vast Economic and Political Significance.

How do we identify what is, or what isn’t, a major question? In his concurrence in West Virginia v. EPA, Justice Gorsuch suggested a set of “triggers,” two of which are relevant here. West Virginia v. EPA, 597 at 744 (Gorsuch, J., concurring). The doctrine is triggered when an agency seeks to regulate “a significant portion of the American economy.” Id. (quoting UARG, 573 U.S. at 324). The doctrine is also triggered when an “agency claims the power to resolve a matter of great ‘political significance.’” Id. at 743 (quoting NFIB v. OSHA, 595 U.S. 109, 117 (2022)).[1]

The FCC purports to offer several justifications for reclassifying broadband, but they all reduce to the same fundamental premise: The FCC (now) thinks broadband is an essential service and should be regulated as such. Order, ¶ 2 (broadband internet connections “are absolutely essential to modern day life” and “[i]t has therefore never been more important that the Commission have both the necessary authority to oversee this essential service”). In other words, the essential nature of broadband access points to both its economic and political significance.

Of course, many essentials of modern-day life—shelter, food, clothing—are provided by numerous suppliers in competitive markets. So is broadband internet. The present marketplace for broadband in the U.S. is competitive, dynamic, and generally serves consumers well. Since 2018, more households are connected to the internet; broadband speeds have increased, while prices have fallen; more households are served by multiple providers; and new technologies like satellite and 5G have expanded internet access and intermodal competition among providers. See Fruits, Manne, Sperry, & Stout, supra. The FCC’s own data suggests that 91% of Americans have access to high-speed broadband under its new and faster definition adopted earlier this year. See 2024 706 Report, FCC 24-27, GN Docket No. 22-270, ¶¶ 20, 22 (Mar. 18, 2024). The light-touch approach of Title I has served the United States well.

What is relevant here is that broadband access represents a substantial share of the U.S. economy:

• In 2023, broadband providers generated $140 billion in revenue. See Internet Service Providers in the US – Market Size (2005–2030), IBISWorld (June 9, 2024), https://www.ibisworld.com/industry-statistics/market-size/internet-service-providers-united-states/.
• Broadband providers have invested $2.1 trillion in building, maintaining, and improving their networks since 1996, including $102.4 billion in 2022 alone. See USTelecom, 2022 Broadband Capex Report (Sept. 8, 2023), https://www.ustelecom.org/research/2022-broadband-capex/.
• In 2023, the U.S. had 131 million households subscribing to fixed broadband. See International Telecommunication Union, United States Fixed Broadband Subscriptions (2023), https://datahub.itu.int/data/?e=USA&c=701&i=19303&u=count.

Regulating broadband access is indisputably a “significant portion of the American economy” akin to “regulating tobacco products, eliminating rate regulation in the telecommunications industry, subjecting private homes to Clean Air Act restrictions, and suspending local housing laws and regulations.” See West Virginia v. EPA, 597 U.S. at 744 (Gorsuch, J., concurring).

There is extensive—perhaps extraordinary—evidence of the political significance of classifying broadband as a Title II telecommunications service. Congress and state legislatures have debated for years about whether and how to regulate broadband access, including net neutrality. Cf. NFIB v. OSHA, 595 U.S. at 121-122 (Gorsuch, J., concurring) (noting the power over public policy, such as public health, largely resides with the states and localities who have employed “a variety of measures in response to the current pandemic”). In fact, politicians on both sides of the aisle have continued to debate whether it was appropriate to classify broadband as a Title II or Title I service. See Dissenting Statement of Brendan Carr 1-3. Members of Congress introduced a Congressional Review Act resolution to stop the Restoring Internet Freedom Order. See Markey Net Neutrality Resolution Reaches 40-Vote Milestone in the Senate (Jan. 9, 2018), https://www.markey.senate.gov/news/press-releases/markey-net-neutrality-resolution-reaches-40-vote-milestone-in-the-senate. This all points not only to the major political ramifications of classification, but to the fact that even Congress has not yet spoken clearly on the classification question of broadband providers under the Communications Act.

C. The Communications Act Does Not Give Clear Authority to the FCC to Classify Broadband as a Title II Telecommunications Service.

The question then becomes whether the Communications Act gives the FCC express or clear authority to classify broadband as a Title II service. The Court has repeatedly noted that Congress must “speak clearly when authorizing an agency to exercise powers of vast economic and political significance.” NFIB v. OSHA, 595 U.S. at 117 (quoting Ala. Ass’n of Realtors v. Dep’t of Health & Human Servs., 594 U.S. 758, 764 (2021)). The authorization must be plain or clear so agencies can’t “exploit some gap, ambiguity, or doubtful expression in Congress’s statutes to assume responsibilities far beyond its initial assignment.” Id. at 669.

For instance, in Alabama Association of Realtors v. Department of Health & Human Services, the Court rejected the Centers for Disease Control and Prevention’s (CDC) attempt to impose a moratorium upon residential evictions due to COVID-19. The Court emphasized that “[e]ven if the text were ambiguous, the sheer scope of the CDC’s claimed authority… would counsel against the Government’s interpretation.” 594 U.S. at 764. The Court was concerned that the government’s reading of the statute would give them “a breathtaking amount of authority” with virtually “no limit… beyond the requirement that CDC deem a measure ‘necessary.’” Id. at 764-65.

Much like the agency actions in NFIB and Realtors, the scope of authority claimed by the FCC through reclassification is staggering, allowing the Commission to regulate nearly the entire internet infrastructure through Title II’s expansive regulatory provisions.

The FCC points to National Cable & Telecommunications Association v. Brand X Internet Services, 545 U.S. 967 (2005) to argue that it has already been decided that the FCC has the authority to classify broadband as a Title II service. But this is a misunderstanding of what that case represents.

At most, in his Brand X dissent, Justice Scalia believed that it would be appropriate to apply a telecommunications classification to the access/delivery component of broadband internet service. Brand X, 545 U.S. at 1005-14 (Scalia, J., dissenting). If this was the Court’s opinion, then there would be a strong argument that it is settled law that Congress spoke clearly to the issue. But it wasn’t. The majority rejected Justice Scalia’s arguments and found the statute ambiguous as to classifying cable-modem service. In other words, Brand X did not foreclose a challenge under the MQD.

On the contrary, Brand X and the D.C. Circuit’s decisions upholding the 2015 and 2018 Orders stand for the proposition that the classification of broadband service under the Communications Act is ambiguous. See US Telecom Ass’n v. FCC, 825 F.3d 674 (D.C. Cir. 2016) (2015 Order); Mozilla Corp v. FCC, 940 F.3d 1 (D.C. Cir. 2019) (2018 Order). That means the answer to the second part of the MQD inquiry, whether Congress clearly spoke to the issue, must be “no.”

D. The Use of Forbearance to Rewrite the Communications Act’s Provisions Shows the FCC Took a Wrong Interpretive Turn.

The Commission is obligated to refrain from enforcing Title II provisions when it concludes that such enforcement is not necessary and that forbearance would serve the public interest. 47 U.S.C. § 160(a). However, in effect, the FCC has essentially reserved for itself the power to decide which provisions of Title II will apply—and to whom. This reservation of authority cannot be anything except arbitrary.

Further, the Order’s attempt to nominally minimize the reach of its claimed authority under Title II through forbearance, much like the 2015 Order in which the Commission noted that it was “tailor[ing]” Title II “for the 21st Century,” 2015 Order ¶ 5,  amounts to rewriting the act to make it more palatable, including by forbearing from rate regulation, network-unbundling requirements, and Section 214 exit certification requirements. See Order ¶¶ 308, 413. See also FCC Fact Sheet, Safeguarding and Securing the Open Internet (Sept. 28, 2023) (“Propose to forbear from 26 Title II provisions, and clarify that the Commission will not regulate rates or require network unbundling.”). The FCC’s “tailored forbearance” is an attempt to rewrite the statute to make it work. See Order ¶¶ 157, 265, 296, 297, 314, 332, 358, 370, 371, 372, 381, 382, 414, 418, 419, 430, 632.

This is very similar to the attempted “tailoring” by the EPA that the Court rejected in UARG, 573 U.S. at 328 (“We affirm the core administrative-law principle that an agency may not rewrite clear statutory terms to suit its own sense of how the statute should operate.”). There, the Tailoring Rule was an attempt to make it such that small entities with the potential to emit greenhouse gasses would not be subject to lawsuits that the Clean Air Act would otherwise allow. See id. at 326 (“The Tailoring Rule is not just an announcement of the EPA’s refusal to enforce the statutory permitting requirements; it purports to alter those requirements and to establish with the force of law that otherwise-prohibited conduct will not violate the Act. This alteration of the statutory requirements was crucial to the EPA’s ‘tailoring’ efforts. Without it, small entities with the potential to emit greenhouse gases in amounts exceeding the statutory thresholds would have remained subject to citizen suits—authorized by the Act.”). The Court rejected this attempt to rewrite the statute, concluding that an agency “has no power to ‘tailor’ legislation to bureaucratic policy goals by rewriting unambiguous statutory terms.” Id. at 325. Much like the EPA in UARG, the FCC’s “need to rewrite clear provisions” of the Communications Act “should have alerted” them “that it had taken a wrong interpretive turn.” Id. at 328.

Moreover, the ability to forbear under Title II also gives the FCC the ability to stop forbearing once Title II reclassification is made. Thus, the decision to reclassify will have huge economic and political implications, as the public and those regulated will have to pay special attention to the forbearance and possible un-forbearance of the FCC’s decisions going forward. Further, the fact of a “tailored” reclassification under Title II would always remain as a sword of Damocles hanging over providers’ heads. Even forborne rules can be put into force, meaning that providers will always have to act in the shadow of this regulatory authority, and will, whether explicitly or implicitly, be guided by that knowledge.

II. The Order’s Vagueness Creates Major Negative Effects.

A. The Order’s Vagueness Imposes Economic Uncertainty on a Large Portion of the U.S. Economy.

The Commission claims that its Order is “the best mix of bright-line rules and case-by-case review.” Order ¶ 390. The bright-line rules prohibit blocking, throttling, and paid or affiliated prioritization, while case-by-case review will be applied under the Order’s general conduct rule as well as provisions of Section 214. But, even the bright-line rules are not so bright, as the bans on blocking and throttling are subject to “reasonable network management” exceptions and the ban on paid prioritization may be waived if a provider can demonstrate the practice “provide[s] some significant public interest benefit and would not harm the open nature of the internet.” 47 CFR §§ 8.3(a), (b),  (c)(2). Under this framework, broadband providers can’t be sure the FCC will accept that they are engaging in reasonable network management.

The uncertainty and costs of complying with the Order’s general conduct rule are even greater than these “bright-line” rules. The Order describes the general conduct rule as “a backstop mechanism to respond to attempts by BIAS providers to wield their gatekeeper power in ways that do not violate the bright-line rules, but nevertheless may compromise the open internet.” Order ¶ 636. Rather than a “backstop,” the general conduct rule is a “catch-all”, see Order ¶ 502, that would allow the Commission to intervene when it finds that a provider’s conduct generally threatened end users or content providers under some principle of net neutrality. See 47 CFR §§ 8.3(d). As “guidance,” the Commission proposes a non-exhaustive list of factors that could possibly (but not necessarily) be used to prove a violation. Order ¶ 507. The factors comprise an uncertain mashup of competition law, consumer-protection law, and First Amendment law and include: effects on end-user control; competitive effects; effects on innovation, investment, or broadband deployment; effects on free expression; whether the conduct is application-agnostic; and whether the conduct conforms to standard industry practices. Id. That is to say, essentially any action implicating anyone or any firm interacting with any part of the internet.

When the FCC relies on a vast expansion of discretionary power based on a list of non-exhaustive factors, this creates enormous uncertainty  for firms that must invest billions of dollars in infrastructure over the course of decades. Even on the relatively shorter timescale required to offer innovative new service packages to consumers, a tremendous volume of negotiations are required among the broadband networks, rights holders, and any other third parties. The only practical way to comply with the general conduct rule would be to involve the FCC in business decisions at every level. Indeed, the Commission anticipates such involvement with its promised “creation of an advisory opinion process.” Order ¶ 636. Both the general conduct rule and a time-consuming and costly advisory opinion process cannot help but chill innovation and ultimately harm consumers through higher prices, reduced quality, and limited choice.

As it did in the 2015 Open Internet Order, in the latest Order the Commission uses its authority under Section 10 of the Communications Act to “forbear” from applying a wide range of Title II regulations. In other words, the FCC selectively chose which common carrier-style requirements to apply, exempting broadband internet providers from many of the more burdensome regulations that traditional telecommunications companies must follow. See Order ¶ 265 (describing the rule as “carefully tailored to avoid the potential issues that commenters claim are problematic”).

The ability to forbear under Title II, however, also gives the Commission the ability to stop forbearing once Title II reclassification is made. Order ¶ 372 (“Although we adopt firm forbearance from all direct rate regulation, with respect to other provisions from which we forbear here, we note that it also is within the Commission’s discretion to proceed incrementally, and we find that adopting an incremental approach here—by virtue of the forbearance granted here—guards against any unanticipated and undesired detrimental effects on broadband deployment that could arise.”) Thus, the decision to reclassify will have huge economic and political implications, as the public and those regulated will have to pay special attention to the forbearance and possible un-forbearance of the FCC’s decisions going forward.

Section 214 may be the biggest source of costly uncertainty. Section 214(a) of the Act prohibits any carrier from constructing, acquiring, or operating any line, and from engaging in transmission through any such line, without first obtaining a certificate from the Commission ‘‘that the present or future public convenience and necessity require or will require the construction, or operation, or construction and operation, of such … line ….’’ 47 U.S.C. § 214(a). Section 214(a) also prohibits any carrier from discontinuing, reducing, or impairing service to a community without first obtaining a certificate from the Commission “that neither the present nor future public convenience and necessity will be adversely affected ….” Id. The Order describes these latter requirements as “exit certification requirements.”

The Order forbears from Section 214 exit certification requirements regarding the discontinuance, reduction, or impairment of broadband internet access services as well as the Commission’s implementing section 214(a)–(d) rules. Order ¶ 308. The Order, however, does not forbear from certification requirements regarding authority for entry, acquisitions (including transfers of control and assignments), and temporary or emergency service and related requirements, instead granting blanket authority to all current and future broadband providers (with the exception of five Chinese companies), “subject to the Commission’s reserved power to revoke such authority.” Order ¶¶ 308, 777. The Commission argues that granting blanket authority is superior to forbearance, in part, because it preserves the Commission’s “ability to protect consumers and the public interest by withdrawing such grants on an individual basis.” Order ¶ 10. Despite this blanket authority, the Commission reserves the right to “withdraw[] such grants on an individual basis.” Order ¶ 10. Commentors made clear that this would lead to considerable uncertainty and longer wait times for broadband providers seeking to maintain, upgrade, or expand their networks.[2]

Blanket authority will create more problems due to its ambiguous nature than either full regulation or forbearance: regrettably, blanket authority appears to be a compromise that failed. Blanket authority creates a gray area where broadband providers are neither fully regulated nor fully exempt. This can lead to confusion about which aspects of Section 214 still apply and to what extent, especially with the Commission’s ability to “revoke” or “withdraw” such authority “on an individual basis.” Thus, despite the Commission’s assurances that blanket authority “removes barriers to entry,” the specter of case-by-case revocation of such authority will lurk behind every investment or acquisition subject to Section 214.

As such, providers will be unsure about the limits of their authority, leading to hesitation in making investment, merger, or acquisition decisions. Moreover, blanket authority may be applied differently to providers or providers might interpret the blanket authority differently, leading to uneven practices across the industry. In particular, the Order anticipates future rulemaking in which Section 214 may be applied differently to small providers. Order ¶ 329 (“The Commission expects to release a further notice of proposed rulemaking (FNPRM) at a future time to examine whether any section 214 rules specifically tailored to BIAS, including for small providers, are warranted.”)

In short, the Commission cannot assert that its forbearance or granting of “blanket authority” under Section 214 allays vagueness concerns, because of the uncertainty that has been introduced  at every step in the process..

B. The Order’s Vagueness Gives the FCC Unbounded Power Over Broadband.

The Order is vague because it does not have “sufficient definitiveness that ordinary people can understand what conduct is prohibited.” Kolender v. Lawson, 461 U.S. 352, 357 (1983). As a result, the FCC has claimed unbounded power to engage in “arbitrary and discriminatory enforcement.” Id.

While a broadband provider could argue that they are engaging in reasonable network management, the case-by-case nature of enforcement outlined in the Order means that no one can be sure they are on the right side of the law. See Order ¶ 506 (“Consistent with our proposal, we adopt a case-by-case approach that will consider the totality of the circumstances when analyzing whether conduct satisfies the general conduct standard to protect the open internet.”).

This problem is exacerbated by the slight guidance that is offered in the Order. For example, while the Order does not ban zero rating, data caps, and usage-based billing, the Order is clear that these practices will be scrutinized and evaluated on a case-by-case basis under the general conduct rule. See Order ¶¶ 521-30. In sum, the FCC retains nearly unlimited ability to punish a business for a practice it deems “unreasonable interference” or a “disadvantage” to end users or edge providers, while providers must operate in a legal gray area where some practices are neither banned nor clearly permitted until they receive review by the FCC.

This vagueness is not cured by the presence of the Order’s advisory opinion process because, even after issuing an opinion, the FCC retains the right to bring a subsequent enforcement action after reconsidering, rescinding, or revoking it. See 47 CFR § 8.6(b)(3) (“An advisory opinion states only the enforcement intention of the Enforcement Bureau as of the date of the opinion, and it is not binding on any party. Advisory opinions will be issued without prejudice to the Enforcement Bureau or the Commission to reconsider the questions involved, or to rescind or revoke the opinion”). In other words, there is no basis for concluding a covered entity has “the ability to clarify the meaning of the regulation by its own inquiry, or by resort to an administrative process.” Cf. Village of Hoffman Estates v. Flipside, Hoffman Estates, Inc., 455 U.S. 489, 498 (1982). The FCC may engage in utterly arbitrary and discriminatory enforcement under the Order.

The FCC’s proposal is similar to the EPA’s rule in Sackett v. EPA, 598 U.S. 651, 669 (2023). There, the EPA rule “assesses the aggregate effect of that group based on a variety of open-ended factors.” Id. at 681. The regulated community is thus “left ‘to feel their way on a case-by-case basis.’” Id. When asked how the regulated community was to know their own obligations, the “EPA recommends asking the Corps” for a “written decision,” quite reminiscent of the FCC’s proposed advisory opinions. Id at 670. The Court found that the “EPA’s interpretation gives rise to serious vagueness concerns.” Id. at 680.

C. The Order Is Arbitrary and Capricious.

The Order’s vagueness creates regulatory uncertainty that will dampen investment incentives for broadband providers, harming consumers in the process. This was not accounted for in the cost-benefit analysis, thus making the Order arbitrary and capricious. See 5 U.S.C. § 706(2)(A); Nat’l Ass’n of Home Builders v. EPA, 682 F.3d 1032, 1040 (D.C. Cir. 2012) (“[When an] agency decides to rely on a cost-benefit analysis as part of its rulemaking a serious flaw undermining that analysis can render the rule unreasonable”). Moreover, in making this decision, the FCC failed to make a reasoned response to the evidence presented in the record. This is a further reason the Order is arbitrary and capricious. As the Supreme Court recently summarized the law in Ohio v. EPA, 144 S. Ct. 2040, 2053 (2024):

An agency action qualifies as “arbitrary” or “capricious” if it is not “reasonable and reasonably explained.” FCC v. Prometheus Radio Project, 592 U. S. 414, 423 (2021)… the agency must offer “a satisfactory explanation for its action[,] including a rational connection between the facts found and the choice made…” Motor Vehicle Mfrs. Assn. of United States, Inc. v. State Farm Mut. Automobile Ins. Co., 463 U. S. 29, 43 (1983)… [and] cannot simply ignore “an important aspect of the problem.” Ibid.

There is a growing body of evidence that Title II reclassification will hinder broadband investment due to regulatory uncertainty. The see-sawing between Title I and Title II regulation over the years has already injected regulatory uncertainty into the broadband market. But reimposing Title II regulations—particularly with the vast discretion in the new rules as noted above—will inject additional uncertainty as successive Commissions change objectives or identify new objectives under Title II.

In comments to the Commission, ICLE notes that firms’ investment decisions can be thought of as an assembly line, where investment opportunities are investigated and evaluated. Opportunities with negative returns on investment are rejected and those with positive returns are further evaluated and ranked. Comments of International Center for Law & Economics, WC Docket No. 23-320, at 25-26 (Dec. 14, 2023) (“ICLE Comments”). Because firms have limited resources, some of the investments with positive returns are rejected. Once a firm decides to pursue an investment opportunity, the project is further evaluated throughout the deployment timeframe. Just as a product can be pulled from the assembly line for defects, investments can be pulled for economic or technical defects. Generally speaking, the further down the assembly line the project goes, the less likely it is to be pulled. Thus, an interruption at the end of the assembly line is likely to be less disruptive than an interruption at the beginning. Similarly, a shift in the regulatory regime would be expected have little impact in the short-run for projects already underway, but have a larger impact in the long-run as firms plan out new investments.

Title II classification can turn projects with positive expected returns into projects with negative expected returns. In addition, the regulatory uncertainty that is endemic to Title II regulation reduces firms’ confidence in the reliability of their return-on-investment projections. Because of the well-known and widely accepted risk-return tradeoff, firms facing increased uncertainty in investment returns will demand higher expected returns from the investments they pursue. See Edwin J. Elton & Martin J. Gruber, Modern Portfolio Theory and Investment Analysis (4th ed., 1991). In this way, even if the return on a potential investment is unchanged, the increase in the uncertainty of those returns would discourage investment.

This is not mere theory. ICLE’s comments to the Commission cite empirical research finding that net neutrality regulations have a significant negative impact on fiber-optic network investment by internet service providers. ICLE Comments, at 26-27 (citing Wolfgang Briglauer, Carlo Cambini, Klaus Gugler, & Volker Stocker, Net Neutrality and High-Speed Broadband Networks: Evidence from OECD Countries, 55 Eur. J. Law Econ. 533–571 (2023)). The Commission notes, “the underlying data for this study were not available to us in our analysis,” yet claims to make “corrections” to the paper to conclude “there is no empirical evidence in the record that Title II reclassification would have any significant negative impact on broadband investment.” Order ¶ 628. The Commission provides no explanation of what “corrections” were necessary, how the corrections were made, or how those corrections could have been made without access to the underlying data. The Commission’s unexplained rejection of published peer-reviewed academic research has the hallmarks of “a serious flaw undermining that analysis” that “render[s] the rule unreasonable.” See Nat’l Ass’n of Home Builders v. EPA, 682 F.3d at 1040.

It is axiomatic that any increase in regulation must be associated with a cost of complying with the increased regulation. Although the Order acknowledges comments claiming the reclassification of broadband internet under Title II will result in higher regulatory compliance costs, it dismisses these concerns, noting “no commenter provided quantitative estimates of the magnitude of these potential compliance costs” and concluding “[i]n our predictive judgment, and based on qualitative analysis, however, we believe that these compliance costs are likely to be small and are outweighed by the benefits of reclassification.” Order ¶ 629. The Order focuses solely on the “direct increase in compliance costs.” Order ¶ 630. Even so, the Commission relies on “qualitative analysis,” rather than the “quantitative estimates” it demands of its critics.

In practice, the direct costs of complying with the Order will be orders of magnitude smaller than the indirect costs borne by providers and consumers. These indirect costs are unmeasurable at this time because the Order has provided only vague guidance on how the regulation will be applied and enforced, or how providers can comply. For example, research suggests that both ISPs and consumers benefit from paid prioritization “under severe congestion and high-value content.” Axel Gautier & Robert Somogyi, Prioritization vs Zero-Rating: Discrimination on the Internet, 73 Int’l. J Ind. Org. 102662 (Dec. 2020). While the Order specifies a “bright-line” ban on paid prioritization, it also provides the Commission an option to waive the ban, if there is a “significant public interest benefit.” The Order does not indicate, and ISPs can’t know, however, whether the Commission will consider delivering any particular high-value content under severe congestion will be such a public interest benefit.

These are substantial costs. They need to be evaluated by the FCC and incorporated into the cost-benefit analysis. The failure to do so is arbitrary and capricious, because there was neither a substantive response to comments in this area nor any consideration of this important aspect of the problem.

CONCLUSION

For the foregoing reasons, the Court should set aside the Order as unlawful.

[1] The MQD may also “apply when an agency seeks to ‘intrude into an area that is the particular domain of state law,’” id. at 744, but that is not relevant here.

[2] See, e.g., Comments of WISPA – Broadband Without Boundaries, In the Matter of Safeguarding and Securing the Open Internet, WC Docket No. 23-320, at  (Dec. 14, 2023), available at https://www.fcc.gov/ecfs/document/121440645853/1 (“Broadband providers across the country who never needed to obtain prior consent when adding a General Partner, refinancing, or engaging in numerous other kinds of purely domestic transactions would now need to file for and obtain prior Commission consent to an assignment or transfer of control. This alone will result in many more applications filed under Section 214 each year than have ever been filed before. . . . Even assuming the majority of those applications are subject to streamlined treatment, this would constitute a huge administrative burden on Commission resources, likely leading to far longer processing times than are already faced by applicants.”).

* No party’s counsel authored any part of this brief. No one, apart from amici and their counsel, contributed money intended to fund the brief’s preparation or submission. All parties have consented the brief’s filing.

Introduction

We thank the Australian Competition & Consumer Commission (ACCC) for the invitation to comment on its July 25 issues paper on recent developments and emerging issues in digital-platform markets, in anticipation of the 10th and final report of its Digital Platform Services Inquiry.[1] The International Center for Law & Economics (ICLE) is a nonprofit, nonpartisan global research and policy center founded with the goal of building the intellectual foundations for sensible, economically grounded policy. ICLE promotes the use of law & economics methodologies to inform public-policy debates and has longstanding expertise in the evaluation of competition law and policy. ICLE’s interest is to ensure that competition law remains grounded in clear rules, established precedent, a record of evidence, and sound economic analysis.

We commend the ACCC’s cautious approach to intervening in digital markets by conducting a comprehensive inquiry before suggesting digital market regulations or reforms to competition law to address alleged competition problems in those markets. As highlighted by the ACCC chair in the press release announcing this paper, “in many cases international legislation is only recently enacted so the full impact may not be clear.”[2]

In these comments, we explore, from a competition and regulation perspective, some of the issues addressed in the issues paper. We respectfully suggest careful consideration before approving any sectoral regulation of digital markets in Australia. Digital markets are generally dynamic, competitive, and beneficial to consumers. Those benefits derive mainly from increased productivity and relatively cheap access to information. While there are always possible competition issues and anticompetitive behavior, these are neither pervasive nor sufficiently unique to justify strict, sui generis preemptive rules. We acknowledge that the ACCC:

[A]grees with the growing international consensus that digital platforms require specific and tailored regulation. While various jurisdictions are taking different approaches to implementing such measures, it is clear to the ACCC that enforcing existing competition and consumer laws ‘ex post’ (i.e. after conduct has occurred) cannot by itself address the systemic and significant problems arising in markets for digital platform services.[3]

We respectfully disagree and posit, instead, that existing antitrust laws are sufficient to address potential anticompetitive practices in digital markets, as we explain below. Furthermore, we think that the ACCC already has the necessary tools and expertise to handle these cases.

Of course, applying antitrust laws to digital markets can be challenging. For example, it can be difficult to define relevant markets and dominant positions on multisided platforms and in the fast-changing digital landscape. The identity of relevant competitors and competing products and services is not always clear, and the boundaries between the digital and non-digital worlds are sometimes overstated. Those challenges, however, can be properly addressed through the existing legal framework. Moreover, many limitations on the effectiveness of ex-post enforcement can be overcome with institutional measures, such as equipping the ACCC with more resources to incorporate advanced, state-of-the-art technical expertise.

Ex-ante regulations like the European Union’s Digital Markets Act (DMA) can have serious unintended consequences, such as stifling innovation, reducing consumer welfare, and increasing compliance costs. They can also lead to increased risks of regulatory capture and rent-seeking, as the verdict on whether a gatekeeper has complied with the law often comes down to the degree to which rivals are satisfied. Of course, rivals have a clear personal stake in never being satisfied. By tethering intervention to a comparatively clear public-benefit standard—consumer welfare—competition laws minimize the potential for error costs and decrease the chances that the law will be coopted for private gain.

Our comments also express the view that policymakers’ present concerns about competition in “AI markets” may be unwarranted. This is particularly true of the notions that data-driven network effects shield incumbents in AI markets from competition; that Web 2.0’s most successful platforms leverage their competitive positions to dominate generative-AI markets; that these same platforms use strategic partnerships with AI firms to insulate themselves from competition; and that generative-AI services occupy narrow markets that leave firms with significant market power.

In fact, we are still far from understanding the boundaries of antitrust-relevant markets in AI. Three primary notions should be at the forefront of competition authorities’ minds when they think about market definitions surrounding AI products and services. First, the “AI market” is not unitary, but instead comprises many distinct goods and services. Second, and relatedly, despite AI marketing hype, this extremely heterogeneous product landscape intersects with equally variegated consumer demand.

In other words: AI products and services may, in many instances, be substitutable for non-AI products, which would mean that, for the purposes of antitrust law, AI and non-AI products contend in the same relevant markets. Getting this relevant product-market definition right is important in antitrust, because incorrect market definitions could lead to wrong inferences about market power. While either an overly broad or overly narrow market definition could lead to erroneous enforcement, we believe the former currently represents the bigger threat.

Third, overenforcement in the field of generative AI could paradoxically engender the very harms that policymakers are seeking to avert. As we explain in greater detail below, preventing so-called “Big Tech” firms from competing in AI markets (for example, by threatening competition intervention whenever they forge strategic relationships with AI startups, launch their own generative-AI services, or embed such services in their existing platforms) may thwart an important source of competition and continued innovation. Competition in AI markets is important,^[4] but trying naïvely to hold incumbent tech firms back out of misguided fears they will come to dominate the AI space is likely to do more harm than good. It is essential to acknowledge how little we know about these nascent markets and that the most important priority at the moment is to ask the right questions to ensure sound competition policy.

The remainder of these comments proceeds as follows: Section II considers international regulatory developments and major market developments discussed in the issues paper’s Topics 1 and 2. Section III considers the emerging issue of competition and AI, aiming to provide a balanced view in a discussion in which proponents of “preemptive enforcement” may be amplifying potential competitive threats that are currently speculative.

II. International Regulatory Developments and Major Developments in Digital Platforms: Rationales for Digital Competition Regulation Are Uncertain

The issues paper presents a detailed description of recent regulatory developments (including both the enactment of new regulations and the discussion of bills and expert reports) in the European Union, Germany, India, Japan, South Korea, and the United Kingdom. As it notes, some key themes have emerged, including “addressing anti-competitive self-preferencing, tying and bundling practices, addressing barriers to switching, and promoting access to third-party applications on a platform or service”.[5]

Whether these are explicit regulations (such as the Digital Markets Act (DMA) in the European Union) or reforms to competition law (as in Germany), all of these reforms seek to ban some or all of the aforementioned conduct or business models due to their allegedly negative effects on competition.

In general terms, “digital markets” or digital platforms are not good candidates for ex-ante regulation, given that the markets for such services are reasonably competitive. As Herbert Hovenkamp has explained:

With Big Tech, we’re looking at probably the most productive part of the economy. The rate of innovation is high. They spend a lot of money on R&D. They are among the largest patent holders. There’s very little evidence of collusion. They seem to be competing with each other quite strongly. They pay their workers relatively well and have fairly educated workforces. None of this is a sign that these are industries we should be pursuing. That doesn’t mean they don’t do some anti-competitive things. But the whole idea that we should be targeting Big Tech strikes me as fundamentally wrong-headed.[6]

According to economic theory and long-tested economic principles, ex-ante regulation^[7] is justified only in the presence of clear market failures^[8]. Digital markets, however, do not present the kind of market failures that warrant ex-ante regulation. For example, digital markets do not generally foment natural monopolies, significant externalities, public goods, or informational asymmetries.

To be sure, one can find some level of informational asymmetries or externalities in some digital markets, but not of such magnitude that they couldn’t be addressed through market competition (actual or potential) or through general rules, such as data-protection or consumer-protection laws. A more plausible argument can be made regarding the presence of “network effects” in online platforms. If a firm moves fast and is the first to attract customers, that customer base will, in turn, attract more customers and sellers. This network growth could, so the story goes, result in a single firm monopolizing the market. However, as Evans and Schmalensee have pointed out, that result is far from inevitable:

Systematic research on online platforms by several authors, including one of us, shows considerable churn in leadership for online platforms over periods shorter than a decade. Then there is the collection of dead or withered platforms that dot this sector, including Blackberry and Windows in smartphone operating systems, AOL in messaging, Orkut in social networking, and Yahoo in mass online media.^[9]

Some regulations and proposals such as the DMA or the proposed American Innovation and Choice Online Act (AICOA) in the United States mention alleged failures of antitrust law (i.e., “too slow” and “too hard for plaintiffs”) as the primary rationale to regulate digital markets. As Giuseppe Colangelo has explained:

Against this background, the regulatory approaches recently advanced do not seem to reflect the distinctive features of digital markets, but rather the need to design enforcement short-cuts to cope with growing concerns that antitrust law is unable to address potential anticompetitive practices by large online platforms. Hence, in most of the mentioned reports, the revival of regulation seems supported more by an alleged antitrust enforcement failure rather than true a market failure. The goal is indeed to fill alleged enforcement gaps in the current antitrust rules by introducing tools aimed at lowering legal standards and evidentiary burdens in order to address anti-competitive practices that standard antitrust analysis would struggle to tackle.^[10]

The ACCC’s fifth interim report does not rely on the presence of market failures to support the need for new competition measures and/or regulations. Rather, resembling the discussion about the above-mentioned regulations and proposals in the European Union and the United States, the ACCC considers that “the enforcement of traditional competition laws against digital platforms in Australia would likewise generally be a slow process. Further, the immense scale and financial resources of large digital platforms may impede traditional enforcement through the courts, resulting in protracted litigated outcomes.”[11]

Sluggish procedure could be a plausible justification for piecemeal reform. The ACCC’s fifth interim report acknowledges this, giving several examples of recent antitrust cases in both Australia and Europe.[12] Antitrust cases could possibly be better expedited. Competition agencies and courts should generally have more resources and faster procedures to adjudicate cases before market structures or markets in general change, rendering any potential intervention useless. This can, however, be resolved via institutional and legal reforms within existing competition law.

The fact that cases are deemed “hard to win” is not a valid justification to jettison the law. Indeed, this might be a strength, not a shortcoming, of antitrust law, particularly in the context of “abuse of dominance” or monopolization cases.^[13] Antitrust cases start from the premise that it is not always immediately obvious whether certain conduct is pro- or anticompetitive.[14] What follows is a structured inquiry operating within a burden-shifting framework that aims to elucidate this question. Regulations like the DMA jettison this cautious approach by condemning broad swathes of specific conduct with strict ex-ante rules that cover many different companies, in different sectors, and with a range of varied products and services. Competitive conditions vary widely, as do the competitive consequences of the business models employed by these firms. DMA-style regulations substitute theoretically robust concepts like “relevant markets” and “market power” or “dominant position” for cruder ones like “core platforms services” or “gatekeeper,” with the express intent of providing shortcuts to condemn business models and practices. But these “shortcuts” have a cost: they can easily lead to condemnation of business models and practices that provide benefits for consumers, such as lower prices and a safer user experience, among others.

Even those open to considering digital-markets regulation acknowledge that there are considerable challenges, especially if the intent is to regulate digital platforms in a manner similar to “essential facilities”:

In the tech industry, the first challenge is to identify a stable essential facility. It must be stable because divestitures take a while to perform, and the cost of implementing them would not be worth its while if the location of the essential facility kept migrating. This condition may not be met, though. While the technology and market segments of electricity, railroads, and (up to the 1980s) telecoms had not changed much since the early twentieth century, digital markets are fastmoving. Rapidly morphing technologies and demands make it difficult for regulators to identify, collect data on, and regulate essential facilities.^[15]

Moreover, even if warranted, regulations create barriers to entry and regulatory risks, and they restrict the monetization of business assets. They also tend to make markets less attractive and could deter potential competitors from entering them. It is possible that the DMA is already producing such consequences. As Alba Ribera has explained:

One of the greatest examples of the dichotomy that arises between the different types of consequences that can be generated by the regulatory capture of digital ecosystems can be found in Meta’s recent decision not to launch its new service Threads in the European Economic Space. To the extent that its service could be interpreted as falling within the definition of a “core platform service” belonging to the category of “online social networks” (listed by the DMA), Meta decided to refrain from entering the European market, due to the disproportionate burden that the demanding obligations imposed by the DMA would entail. It should be noted that Threads is still an entrant service in the online social networking market, in contrast to the predominant position occupied by X (previously known as Twitter). In this way, we observe that the categorization as a core platform service unifies and eliminates all the nuances that free competition entails with respect to incoming services in the markets.^[16]

Some of these unintended consequences were observable in the EU even before the DMA fully entered into force. From users’ perspective, regulation can serve to make services and products more expensive. Facebook is already experimenting with a new business model in the EU in which the consumer would see no ads (thus, there would be no data collection, or less collection of data for marketing purposes, at any rate), but would have to pay for subscriptions.[17] If this business model would generalize, some privacy-minded users may prefer it and probably would be able to afford it. But other consumers that may prefer and have benefitted the most from digital platforms with zero price or otherwise affordable products, such as WhatsApp and Facebook, would be worse off.

From the perspective of the companies that own and operate digital platforms and services, if regulations like the DMA make their platforms less profitable, some could choose not to enter or, indeed, to leave such markets. As Geoffrey Manne and Dirk Auer have explained, “to regulate competition, you first need to attract competition.”^[18]

While these considerations are especially pertinent in the context of developing countries, which rely heavily on attracting foreign direct investment, they could also affect the competitiveness of countries like Australia.

The DMA entered into effect in full force in March 2024. While it may be too early to reach definitive conclusions about its effects, consumers are already reporting that they have experienced a degraded user experience. For example, the French newspaper Liberation has detailed how Google Maps’ map results are not showing directly in search-results pages in the same ways they once did (See Figures 1 and 2).

FIGURE 1: US Search Results for ‘Crepes in Paris’

SOURCE: Chamber of Progress ^[19]

FIGURE 2: French VPN Search Results for ‘Crepes in Paris’

SOURCE: Chamber of Progress ^[20]

Presumably, this is happening because a direct link to Google Maps would constitute “self-preferencing” wherein the search engine, Google, would be “unfairly” directing traffic to its own digital-navigation service. Such conduct is prohibited by Art. 6(5) of the DMA. But this kind of integration is very convenient for consumers, who can search for a restaurant and then quickly find the directions to walk or commute to it (and sometimes even book a table).

While removing some features, Google is also adding more results to its results pages, because it assumes that it is required under the DMA to provide “fair” links to competing sites like Yelp and TripAdvisor.^[21] In theory, the consequence of such requirements is “more options” for consumers. In practice, what consumers have is less relevant results, a more cluttered results page, and thus, a downgraded user experience.

Apple highlights another quality-degrading consequence of the DMA: the obligation imposed on platforms like iOS to allow competing app stores and to allow apps to be downloaded directly from their websites (commonly known as “sideloading”).^[22] This “openness,” however, would allow third-party applications to bypass controls and protections implemented to safeguard users’ security and privacy.^[23] This is already happening in Europe, where Apple has been forced to allow Epic Games to launch an alternative app store on its iOS operating system.[24] While this may seem a positive development for (some) developers and consumers, it could also harm user trust in the platform and thus decrease the total number of transactions, to the detriment of all parties involved (business users, consumers, and the owner of the platform). Indeed, “[p]hishers are using a novel technique to trick iOS and Android users into installing malicious apps that bypass safety guardrails built by both Apple and Google to prevent unauthorized apps.”[25] This sort of attack will be more effective in the absence of the protections in Apple’s App Store.[26]

The recent Microsoft/CrowdStrike outage—which affected many services around the world, including airlines and public services—is a good example of the tradeoffs between “openness” and the security and reliability of digital platforms. At least in part, the outage is explained by the fact that Microsoft must give access to its operating system to CrowdStrike and other developers. As explained in a Financial Times note:

Giving software companies that kind of access to an operating system is dangerous — it means you can quickly lose control of your computer if any of the software providers you rely on makes a mistake or is compromised. That is why Apple began informing third-party developers in 2020 that it would no longer grant them kernel-level access to the MacOS operating system (and also quite possibly why the CrowdStrike problem didn’t affect Apple devices).

But not all the fault lies with Microsoft. A 2009 agreement between the company and the European Commission requires it to grant outside developers the same access to Windows that its own security software has. The idea was to make it possible for other software companies to compete with Microsoft by ensuring many of its products and services are interoperable with outside software and tools. That’s a worthy goal, and many provisions in the agreement are entirely reasonable, such as requiring that Outlook support common calendar event and scheduling formats.

But the 2009 agreement is profoundly flawed in requiring Microsoft to make all of the APIs, or programming functions, that its own security software products use available to manufacturers of third-party security software products. This is the provision that requires Microsoft to give kernel-level access to companies such as CrowdStrike. Until it is changed, it’s not clear that Microsoft can implement the chief lesson of this debacle and start phasing out access, as Apple did four years ago.[27]

The problem with ex-ante regulations like the DMA, which prohibit some types of conduct absolutely, is that they don’t capture the complexities and tradeoffs generally present in market dynamics and, of course, specifically in digital markets. As Dirk Auer has pointed out, the alleged benefits of the DMA are looking more and more like broken promises:

When it was passed, European policymakers like Margrethe Vestager and Thierry Breton assured the public that the far-reaching regulation would not compromise security, lead to costlier services, or otherwise degrade users’ online experience. They also argued that it would be fast and easy to apply, thus avoiding the lengthy litigation that has come to be associated with competition enforcement.

As the effects of the DMA start to play out, however, these promises appear increasingly fanciful.

The biggest concern is that Europeans’ online safety is being compromised. Apple has warned that it will not be able to guarantee the safety of rival app stores and payment systems that can now access its ecosystem. If this sounds abstract, it is worth noting that these sorts of security flaws facilitated the Oct. 7 attacks carried out by Hamas. They also increase more mundane risks of identity theft and fraud.

Similarly, Amazon will struggle to exclude nefarious goods, sellers, and shippers from its online marketplace. Commenting on similar issues in the United States, the company surmised that it risked losing “customer trust by advertising something that is not a good deal for them.” This loss of consumer trust would, in turn, harm the bottom lines of the roughly two million businesses that rely on the platform.[28]

As indicated above, the DMA’s unintended consequences affect not only consumers, but also business users. Since Google began to implement the DMA in January 2024, early estimates suggest that clicks from Google ads to hotel websites fell by 17.6%.^[29] Presumably, this ought to be considered a failure, even by the DMA’s own (uncertain) standards.

Australia itself has experience with another regulation intended to address a “significant bargaining power imbalance” between digital platforms and businesses interacting with them. The News Media and Digital Platforms Mandatory Bargaining Code was enacted in 2021 to allow news organizations to collectively bargain with platforms like Google and Facebook for remuneration for news content featured on those platforms without breaching Australian competition laws.[30] It has, however, apparently resulted in less traffic for smaller media outlets and independent publishers. As reported by The Guardian: “Smaller publishers are already feeling the effects of a potential ban on news on Facebook, a parliamentary committee has heard, as news outlets small and large make the case for social media companies to be compelled to pay for news.”[31]

This is why, even when their relative lenity is considered, antitrust laws are more flexible and more likely to be appropriate for various markets and business models.^[32] As Posner illustrates, while rules are generally simpler and cheaper to enforce, they are often underinclusive or overinclusive. When facts like the mere size or number of users—which are disconnected from competitive goals—are determinative of their application, they are “especially apt to fail.”^[33]

All in all, as the ACCC acknowledges, most digital competition regulation “has only recently been enacted, and the full impact of recent international regulatory developments cannot yet be observed.”[34] In that vein, it would be wise to study markets, perform proper regulatory-impact analysis, and keep learning from the experience of other regulators and markets. It may be advisable, for instance, to follow the example of South Korea, which has hit the pause button on its proposal to regulate digital markets.[35]

Regarding major developments in digital-platform markets, the issues paper provides updated information about two markets: online private messaging and app marketplaces.

With respect to the online private-messaging market, the issues paper reminds that:

The ACCC’s September 2020 report found that both Facebook’s and Apple’s services benefited from identity-based network effects, providing them with significant competitive advantages over smaller suppliers of standalone services in Australia. However, because the use of Apple’s online private messaging services was limited to users of Apple devices, the ACCC found that iMessage was likely to impose weaker competitive constraints on Facebook Messenger and WhatsApp than those services imposed on iMessage.[36]

The paper also highlights that “Australians’ usage of these services is increasing. Based on recent ACMA survey data, in the 6 months prior to June 2023, 72% of Australian adults used Facebook Messenger (up from 66% in 2020), 51% used WhatsApp (up from 39% in 2020) and 35% used FaceTime (up from 33% in 2020).”[37]

It is important, at this point, to consider that, while the aforementioned messaging platforms (Facebook’s Messenger and Apple’s iMessage) have a large number of users and, therefore, a relevant market share and the advantage of network effects, this does not mean that the messaging-apps market is not competitive. The issues paper itself recognizes that:

Online private messaging services encompass a range of services, including text, audio and video messaging services, and are offered by a wide variety of platforms. The ACCC observed a wide range of online private messaging services available to Australian users, which were often highly differentiated, offering different features and functionalities, and used by consumers for a number of different purposes.[38]

Besides Facebook’s Messenger and Apple’s iMessage, several actors with smaller but not irrelevant market shares create competitive pressure on both.[39] Moreover, this is a market where consumers can use more than one application simultaneously (i.e., there is “multi-homing”). This allows consumers to try different services at little cost, which means that both current competitors and new entrants can challenge the market leaders. The fact that the use of these services is increasing in Australia provides an opportunity for these challengers.

Furthermore, although network effects offer an important advantage, they do not guarantee a comfortable monopolist position.[40] Catherine Tucker explains how network effects can be a double-edged sword, as they can also lead to a sudden decline of users to a marginally superior competitor.[41] Actual economic studies of data-network effects have been few and far between, with scant empirical evidence to support the theory that the control of personal data creates an insurmountable barrier to entry.^[42]

Regarding app marketplaces, the issues paper finds that:

In its March 2021 report, the ACCC found that Apple and Google were the predominant mobile app marketplace operators in Australia, offering the App Store and Play Store respectively.

The report referenced data from Statcounter, estimating that in December 2020, Apple iOS held 54% of the market share of mobile operating systems (OS) in Australia, while Android held 46%. Based on updated data from Statcounter, these shares appear to have remained stable – as of June 2024, Apple iOS again held 54% market share, while Android held 45%. The next closest competitor, Samsung, held less than 1% (0.86%).

In the March 2021 report, the ACCC considered it likely that Apple and Google held significant market power in mobile app distribution in Australia, due to their control of the iOS and Android mobile OS. This meant the App Store and the Play Store were ‘must haves’ for the majority of mobile app developers in Australia.[43]

The fact that both Apple and Google (Android) both have relatively high (and similar) market shares does not necessarily mean that the market is not competitive. Two powerful actors can discipline each other’s efforts to exercise market power.[44] Our point here is not that there cannot be any competition problem or anticompetitive conduct in this market, of course, but rather that more evidence than merely concentration or market shares is required in order to determine that there are even possible competition issues in this market. The rivalry between Apple and Android in the market has largely benefited consumers with innovation and increased access to a large array of features.[45]

III. Potential Emerging Issues: Competition and AI

Although the issues paper also identifies emerging issues in the online-gaming and cloud-computing markets, it identifies very specific (if somewhat theoretical) risks in the generative-AI market that echo the concerns raised by several competition agencies around the world. Drawing on a discussion paper by the Digital Platforms Regulators Forum, the issues paper notes that:

• Developing and operating large language models (LLMs) requires a large upfront financial investment, access to vast datasets, long development lead times, access to sophisticated AI systems and talent, and substantial ongoing computing costs and access to computing resources (such as cloud storage), which together create high barriers to entry.
• LLMs are likely to have features common to digital-platform services that make such markets tend toward concentration, including a positive feedback loop involving the collection and use of user data, economies of scale, and access to large volumes of high-quality user data. Because of these characteristics, new entrants could find it difficult to compete with digital-platform services that use LLMs as part of new and existing services.
• Generative AI could lead users to interact more with particular digital platforms. Over time, this may make it more difficult for users to leave these platforms. LLMs could allow big digital platforms to strengthen and expand their market power by continuing to engage in the allegedly anti-competitive practices the ACCC has previously observed. These include anti-competitive self-preferencing, tying, and restrictions on data access.[46]

We address these concerns in the following subsections. Subsection A summarizes recent calls for competition intervention in generative-AI markets. Subsection BB argues that many of these calls are underpinned by fears of data-related incumbency advantages (often referred to as “data-network effects”), including in the context of mergers. Subsection CC explains why these effects are unlikely to play a meaningful role in generative AI markets. Finally, subsection D offers five key takeaways to help policymakers better weigh the tradeoffs inherent in competition enforcement interventions in generative AI markets.

A. Calls for Intervention in AI Markets Should Acknowledge Actual Market Developments

It was once (and frequently) said that Google’s “data monopoly” was unassailable: “If ‘big data’ is the oil of the information economy, Google has Standard Oil-like monopoly dominance—and uses that control to maintain its dominant position.”[47] Similar claims of data dominance have been attached to nearly all large online platforms, including Facebook (Meta), Amazon, and Uber.[48]

While some of these claims continue even today (for example, “big data” is a key component of the U.S. Justice Department’s (DOJ) Google Search and adtech antitrust suits),[49] a shiny new data target has emerged in the form of generative AI. The launch of ChatGPT in November 2022, as well as the advent of AI image-generation services like Midjourney and Dall-E, have dramatically expanded the public’s conception of what is—and what might be—possible to achieve with generative-AI technologies built on massive datasets.

While these services remain both in the early stages of mainstream adoption and in the throes of rapid, unpredictable technological evolution, they are nevertheless already on the radar of competition policymakers around the world. Several antitrust enforcers appear to believe that, by acting now, they can avoid the “mistakes” that purportedly were made during the formative years of Web 2.0.[50] These mistakes, critics assert, include failing to appreciate the centrality of data in online markets, as well as letting mergers go unchecked and allowing early movers to entrench their market positions.^[51] As Federal Trade Commission (FTC) Chair Lina Khan has put it: “we are still reeling from the concentration that resulted from Web 2.0, and we don’t want to repeat the missteps of the past with AI.”^[52]

This response from the competition-policy world is deeply troubling. Rather than engage in critical self-assessment and adopt an appropriately restrained stance, the enforcement community appears to be champing at the bit. Rather than assessing their prior assumptions based on the current technological moment, enforcers’ top priority appears to be figuring out how to rapidly and almost reflexively deploy existing competition tools to address the presumed competitive failures presented by generative AI.^[53]

It has for some time been increasingly common for competition enforcers to argue that “data-network effects” serve not only to entrench incumbents in those markets where the data is collected, but also to confer similar, self-reinforcing benefits in adjacent markets. Several enforcers have, for example, prevented large online platforms from acquiring smaller firms in adjacent markets, citing the risk that they could use their vast access to data to extend their dominance into these new markets.^[54]

They have concomitantly launched consultations to ascertain the role that data plays in AI competition. For instance, in a recent consultation, the European Commission asked: “What is the role of data and what are its relevant characteristics for the provision of generative AI systems and/or components, including AI models?”^[55] The FTC has likewise been hypervigilant about the risks ostensibly posed by incumbents’ access to data. In comments submitted to the U.S. Copyright Office, for example, the FTC argued that:

The rapid development and deployment of AI also poses potential risks to competition. The rising importance of AI to the economy may further lock in the market dominance of large incumbent technology firms. These powerful, vertically integrated incumbents control many of the inputs necessary for the effective development and deployment of AI tools, including cloud-based or local computing power and access to large stores of training data. These dominant technology companies may have the incentive to use their control over these inputs to unlawfully entrench their market positions in AI and related markets, including digital content markets.[56]

Likewise, Jonathan Kanter, assistant U.S. attorney general for antitrust, recently claimed that:

We also see structures and trends in AI that should give us pause AI relies on massive amounts of data and computing power, which can give already dominant firms a substantial advantage. Powerful networks and feedback effects may enable dominant firms to control these new markets, and existing power in the digital economy may create a powerful incentive to control emerging innovations that will not only impact our economy, but the health and well-being of our society and free expression itself.[57]

On an even more hyperbolic note, Andreas Mundt, the head of Germany’s Federal Cartel Office, called AI a “first-class fire accelerator” for anticompetitive behavior and argued it “will make all the problems only worse.”[58] He further argued that “there’s a great danger that we’ll get an even deeper concentration of digital markets and power increase at various levels, from chips to the front end.”[59] In short, Mundt is one of many policymakers who believe that AI markets will enable incumbent tech firms to further entrench their market positions.

Certainly, it makes sense that the largest online platforms—including Alphabet, Meta, Apple, and Amazon—should have a meaningful advantage in the burgeoning markets for generative AI services. After all, it is widely recognized that data is an essential input for generative AI.^[60] This competitive advantage should be all the more significant, given that these firms have been at the forefront of AI technology for more than a decade. Over this period, Google’s DeepMind and AlphaGo and Meta’s NLLB-200 have routinely made headlines.[61] Apple and Amazon also have vast experience with AI assistants, and all of these firms deploy AI technologies throughout their platforms.[62]

Contrary to what one might expect, however, the tech giants have, to date, been largely unable to leverage their vast troves of data to outcompete startups like OpenAI and Midjourney. At the time of writing, for instance, OpenAI’s ChatGPT appears to be, by far, the most successful chatbot,[63] despite the large tech platforms’ apparent access to far more (and more up-to-date) data.

There are important lessons to glean from these developments, if only enforcers would stop to reflect. The meteoric rise of consumer-facing AI services should offer competition enforcers and policymakers an opportunity for introspection. As we explain, the rapid emergence of generative AI may undercut many core assumptions of today’s competition-policy debates, which have focused largely on the rueful after-effects of the purported failure of 20^th-century antitrust to address the alleged harms of 21^st-century technology. These include the notions that data advantages constitute barriers to entry and can be leveraged to project dominance into adjacent markets; that scale itself is a market failure to be addressed by enforcers; and that the use of consumer data is inherently harmful to those consumers.

B. Data-Network Effects Theory and Enforcement

Proponents of more extensive intervention by competition enforcers into digital markets often cite data-network effects as a source of competitive advantage and a barrier to entry (although terms like “economies of scale and scope” may offer more precision).^[64] The crux of the argument is that “the collection and use of data creates a feedback loop of more data, which ultimately insulates incumbent platforms from entrants who, but for their data disadvantage, might offer a better product.”^[65] This self-reinforcing cycle purportedly leads to market domination by a single firm. Thus, it is argued, e.g., that Google’s “ever-expanding control of user personal data, and that data’s critical value to online advertisers, creates an insurmountable barrier to entry for new competition.”^[66]

But it is important to note the conceptual problems these claims face. Because data can be used to improve products’ quality and/or to subsidize their use, if possessing data constitutes an entry barrier, then any product improvement or price reduction made by an incumbent could be problematic. This is tantamount to an argument that competition itself is a cognizable barrier to entry. Of course, it would be a curious approach to antitrust if competition were treated as a problem, as it would imply that firms should under-compete—i.e., should forego consumer-welfare enhancements—in order to inculcate a greater number of firms in a given market, simply for its own sake.[67]

Meanwhile, actual economic studies of data-network effects have been few and far between, with scant empirical evidence to support the theory.^[68] Andrei Hagiu and Julian Wright’s theoretical paper offers perhaps the most comprehensive treatment of the topic to date.[69] The authors ultimately conclude that data-network effects can be of differing magnitudes and have varying effects on firms’ incumbency advantage.^[70] They cite Grammarly (an AI writing-assistance tool) as a potential example: “As users make corrections to the suggestions offered by Grammarly, its language experts and artificial intelligence can use this feedback to continue to improve its future recommendations for all users.”^[71]

This is echoed by economists who contend that “[t]he algorithmic analysis of user data and information might increase incumbency advantages, creating lock-in effects among users and making them more reluctant to join an entrant platform.”^[72] Crucially, some scholars take this logic a step further, arguing that platforms may use data from their “origin markets” in order to enter and dominate adjacent ones:

First, as we already mentioned, data collected in the origin market can be used, once the enveloper has entered the target market, to provide products more efficiently in the target market. Second, data collected in the origin market can be used to reduce the asymmetric information to which an entrant is typically subject when deciding to invest (for example, in R&D) to enter a new market. For instance, a search engine could be able to predict new trends from consumer searches and therefore face less uncertainty in product design.^[73]

This possibility is also implicit in Hagiu and Wright’s paper.[74] Indeed, the authors’ theoretical model rests on an important distinction between “within-user” data advantages (that is, having access to more data about a given user) and “across-user” data advantages (information gleaned from having access to a wider user base). In both cases, there is an implicit assumption that platforms may use data from one service to gain an advantage in another market (because what matters is information about aggregate or individual user preferences, regardless of its origin).

Our review of the economic evidence suggests that several scholars have, with varying degrees of certainty, raised the possibility that incumbents may leverage data advantages to stifle competitors in their primary market or in adjacent ones (be it via merger or organic growth). As we explain below, however, there is ultimately little evidence to support such claims. Policymakers have nonetheless been keenly receptive to these limited theoretical findings, basing multiple decisions on these theories, often with little consideration given to the caveats that accompany them.^[75]

Indeed, it is remarkable that, in its section on “[t]he data advantage for incumbents,” the “Furman Report” created for the UK government cited only two empirical economic studies, and they offer directly contradictory conclusions with respect to the strength of data advantages.[76] The report nevertheless concluded that data “may confer a form of unmatchable advantage on the incumbent business, making successful rivalry less likely,”[77] and it adopted without reservation what it deemed “convincing” evidence from non-economists that have no apparent empirical basis.[78]

In the Google/Fitbit merger proceedings, the European Commission found that the combination of data from Google services with that of Fitbit devices would reduce competition in advertising markets:

Giving [sic] the large amount of data already used for advertising purposes that Google holds, the increase in Google’s data collection capabilities, which goes beyond the mere number of active users for which Fitbit has been collecting data so far, the Transaction is likely to have a negative impact on the development of an unfettered competition in the markets for online advertising.^[79]

As a result, the Commission cleared the merger only on the condition that Google refrain from using data from Fitbit devices for its advertising platform.^[80] The Commission also appears likely to focus on similar issues in its ongoing investigation of Microsoft’s investment in OpenAI.[81]

Along similar lines, in its complaint to enjoin Meta’s purchase of Within Unlimited—makers of the virtual-reality (VR) fitness app Supernatural—the FTC relied on, among other things, the fact that Meta could leverage its data about VR-user behavior to inform its decisions and potentially outcompete rival VR-fitness apps: “Meta’s control over the Quest platform also gives it unique access to VR user data, which it uses to inform strategic decisions.”^[82]

The DOJ’s twin cases against Google also implicate data leveraging and data barriers to entry. The agency’s adtech complaint charges that “Google intentionally exploited its massive trove of user data to further entrench its monopoly across the digital advertising industry.”[83] Similarly, in its Google Search complaint, the agency argued that:

Google’s anticompetitive practices are especially pernicious because they deny rivals scale to compete effectively. General search services, search advertising, and general search text advertising require complex algorithms that are constantly learning which organic results and ads best respond to user queries; the volume, variety, and velocity of data accelerates the automated learning of search and search advertising algorithms.[84]

Finally, updated merger guidelines published in recent years by several competition enforcers cite the acquisition of data as a potential source of competition concerns. For instance, the FTC and DOJ’s 2023 guidelines state that “acquiring data that helps facilitate matching, sorting, or prediction services may enable the platform to weaken rival platforms by denying them that data.”^[85] Likewise, the UK Competition and Markets Authority warned against incumbents acquiring firms in order to obtain their data and foreclose other rivals:

Incentive to foreclose rivals…

7.19(e) Particularly in complex and dynamic markets, firms may not focus on short term margins but may pursue other objectives to maximise their long-run profitability, which the CMA may consider. This may include… obtaining access to customer data….^[86]

In short, competition authorities around the globe have taken an increasingly aggressive stance on data-network effects. Among the ways this has manifested is in enforcement decisions based on fears that data a platform collects in one market might confer decisive competitive advantages in adjacent markets. Unfortunately, these concerns rest on little to no empirical evidence, either in the economic literature or the underlying case records.

C. Data-Incumbency Advantages in Generative AI

Given the assertions detailed in the previous section, it would be reasonable to assume that firms such as Google, Meta, and Amazon should be in pole position to meet the burgeoning demand for generative AI. After all, these firms have not only been at the forefront of the field for the better part of a decade, but they also have access to vast troves of data, the likes of which their rivals could only dream when they launched their own services. Thus, the authors of the Furman Report caution that “to the degree that the next technological revolution centres around artificial intelligence and machine learning, then the companies most able to take advantage of it may well be the existing large companies because of the importance of data for the successful use of these tools.”[87]

To date, however, this is not how things have unfolded (although it bears noting that these technologies remain in flux and the competitive landscape is susceptible to change). The first significantly successful generative AI service was arguably not from either Meta—which had been working on chatbots for years and had access to, arguably, the world’s largest database of actual chats—or Google. Instead, the breakthrough came from a previously unknown firm called OpenAI.

OpenAI’s ChatGPT service currently accounts for an estimated 60% of visits to online AI tools (although reliable numbers are somewhat elusive).[88] It broke the record for the fastest online service to reach 100 million users (in only a couple of months), more than four times faster than TikTok, the previous record holder.^[89] Based on Google Trends data, ChatGPT is nine times more popular worldwide than Google’s own Bard service, and 14 times more popular in the United States.^[90] In April 2023, ChatGPT reportedly registered 206.7 million unique visitors, compared to 19.5 million for Google’s Bard.^[91] In short, at the time we are writing, ChatGPT appears to be the most popular chatbot. The entry of large players such as Google Bard or Meta AI appear to have had little effect thus far on its leading position.^[92]

The picture is similar in the field of AI image generation. As of August 2023, Midjourney, Dall-E, and Stable Diffusion appear to be the three market leaders in terms of user visits.^[93] This is despite competition from the likes of Google and Meta, who arguably have access to unparalleled image and video databases by virtue of their primary platform activities.^[94]

This raises several crucial questions: how have these AI upstarts managed to be so successful, and is their success just a flash in the pan before Web 2.0 giants catch up and overthrow them? While we cannot answer either of these questions dispositively, we offer what we believe to be some relevant observations concerning the role and value of data in digital markets.

A first important observation is that empirical studies suggest that data exhibits diminishing marginal returns. In other words, past a certain point, acquiring more data does not confer a meaningful edge to the acquiring firm. As Catherine Tucker put it, following a review of the literature: “Empirically there is little evidence of economies of scale and scope in digital data in the instances where one would expect to find them.”^[95]

Likewise, following a survey of the empirical literature on this topic, Geoffrey Manne and Dirk Auer conclude that:

Available evidence suggests that claims of “extreme” returns to scale in the tech sector are greatly overblown. Not only are the largest expenditures of digital platforms unlikely to become proportionally less important as output increases, but empirical research strongly suggests that even data does not give rise to increasing returns to scale, despite routinely being cited as the source of this effect.^[96]

In other words, being the firm with the most data appears to be far less important than having enough data. Moreover, this lower bar may be accessible to far more firms than one might initially think possible. Furthermore, obtaining sufficient data could become easier still—that is, the volume of required data could become even smaller—with technological progress. For instance, synthetic data may provide an adequate substitute to real-world data,[97] or may even outperform real-world data.[98] As Thibault Schrepel and Alex Pentland surmise:

[A]dvances in computer science and analytics are making the amount of data less relevant every day. In recent months, important technological advances have allowed companies with small data sets to compete with larger ones.^[99]

Indeed, past a certain threshold, acquiring more data might not meaningfully improve a service, where other improvements (such as better training methods or data curation) could have a large impact. In fact, there is some evidence that excessive data impedes a service’s ability to generate results appropriate for a given query: “[S]uperior model performance can often be achieved with smaller, high-quality datasets than massive, uncurated ones. Data curation ensures that training datasets are devoid of noise, irrelevant instances, and duplications, thus maximizing the efficiency of every training iteration.”[100]

Consider, for instance, a user who wants to generate an image of a basketball. Using a model trained on an indiscriminate range and number of public photos in which a basketball appears surrounded by copious other image data, the user may end up with an inordinately noisy result. By contrast, a model trained with a better method on fewer, more carefully selected images could readily yield far superior results.[101] In one important example:

The model’s performance is particularly remarkable, given its small size. “This is not a large language model trained on the whole Internet; this is a relatively small transformer trained for these tasks,” says Armando Solar-Lezama, a computer scientist at the Massachusetts Institute of Technology, who was not involved in the new study…. The finding implies that instead of just shoving ever more training data into machine-learning models, a complementary strategy might be to offer AI algorithms the equivalent of a focused linguistics or algebra class.[102]

Platforms’ current efforts are thus focused on improving LLMs’ mathematical and logical reasoning, rather than maximizing the size of training datasets.^[103] Two points stand out. The first is that firms like OpenAI rely largely on publicly available datasets—such as GSM8K—to train their LLMs.^[104] Second, the real challenge to creating innovative AI lies not so much in collecting data, but in creating innovative AI-training processes and architectures:

[B]uilding a truly general reasoning engine will require a more fundamental architectural innovation. What’s needed is a way for language models to learn new abstractions that go beyond their training data and have these evolving abstractions influence the model’s choices as it explores the space of possible solutions.

We know this is possible because the human brain does it. But it might be a while before OpenAI, DeepMind, or anyone else figures out how to do it in silicon.^[105]

Furthermore, it is worth noting that the data most relevant to startups in a given market may not be those held by large incumbent platforms in other markets. They might instead be data specific to the market in which the startup is active or, even better, to the given problem it is attempting to solve:

As Andres Lerner has argued, if you wanted to start a travel business, the data from Kayak or Priceline would be far more relevant. Or if you wanted to start a ride-sharing business, data from cab companies would be more useful than the broad, market-cross-cutting profiles Google and Facebook have. Consider companies like Uber, Lyft and Sidecar that had no customer data when they began to challenge established cab companies that did possess such data. If data were really so significant, they could never have competed successfully. But Uber, Lyft and Sidecar have been able to effectively compete because they built products that users wanted to use—they came up with an idea for a better mousetrap. The data they have accrued came after they innovated, entered the market and mounted their successful challenges—not before.[106]

The bottom line is that data is not the be-all and end-all that many in competition circles make it out to be. While data may often confer marginal benefits, there is little evidence that these benefits are ultimately decisive.[107] As a result, incumbent platforms’ access to vast numbers of users and troves of data in their primary markets might only marginally affect their competitiveness in AI markets.

A related observation is that firms’ capabilities and other features of their products arguably play a more important role than the data they own.[108] Examples of this abound in digital markets. Google overthrew Yahoo in search, despite initially having access to far fewer users and far less data. Google and Apple overcame Microsoft in the smartphone operating-system market, despite having comparatively tiny ecosystems (at the time) to leverage. TikTok rose to prominence despite intense competition from incumbents like Instagram, which had much larger userbases. In each of these cases, important product-design decisions (such as the PageRank algorithm, recognizing the specific needs of mobile users,^[109] and TikTok’s clever algorithm) appear to have played far more significant roles than the firms’ initial user and data endowments (or lack thereof).

All of this suggests that the early success of OpenAI likely has more to do with its engineering decisions than with what data it did or did not possess. Going forward, OpenAI and its rivals’ relative abilities to offer and monetize compelling use cases by offering custom versions of their generative AI technologies will arguably play a much larger role than (and contribute to) their ownership of data.[110] In other words, the ultimate challenge is arguably to create a valuable platform, of which data ownership is a consequence, not a cause.

It is also important to note that, in those instances where it is valuable, data does not just fall from the sky. Instead, it is through smart business and engineering decisions that firms can generate valuable information (which does not necessarily correlate with owning more data). For instance, OpenAI’s success with ChatGPT is often attributed to its more efficient algorithms and training models, which arguably have enabled the service to improve more rapidly than its rivals.^[111] Likewise, the ability of firms like Meta and Google to generate valuable data for advertising arguably depends more on design decisions that elicit the right data from users, rather than the raw number of users in their networks.

Put differently, setting up a business so as to gather and organize the right information is more important than simply owning vast troves of data.[112] Even in those instances where high-quality data is an essential parameter of competition, it does not follow that having vaster databases or more users on a platform necessarily leads to better information for the platform. Indeed, if data ownership consistently conferred a significant competitive advantage, these new AI firms would not be where they are today.

Moreover, it is important not to neglect the role that open-source models currently play in fostering innovation and competition. As former DOJ Chief Antitrust Economist Susan Athey pointed out in a recent interview, “[the AI industry] may be very concentrated, but if you have two or three high quality—and we have to find out what that means, but high enough quality—open models, then that could be enough to constrain the for-profit LLMs.”[113] Open-source models are important because they allow innovative startups to build upon models already trained on large datasets—therefore entering the market without incurring that initial cost. Apparently, there is no lack of open-source models, since companies like xAI, Meta, and Google offer their AI models for free.[114]

This does not, of course, mean that data is worthless. Rather, it means that competition authorities should not assume that the mere possession of data is a dispositive competitive advantage, absent compelling empirical evidence to support such a finding. In this light, the current wave of decisions and competition-policy pronouncements that rely on data-related theories of harm are premature.

D. Five Key Takeaways: Reconceptualizing the Role of Data in Generative-AI Competition

As we explain above, data-network effects are not the source of barriers to entry that they are sometimes made out to be. The picture is far more nuanced. Indeed, as economist Andres Lerner demonstrated almost a decade ago (and the assessment is only truer today):

Although the collection of user data is generally valuable for online providers, the conclusion that such benefits of user data lead to significant returns to scale and to the entrenchment of dominant online platforms is based on unsupported assumptions. Although, in theory, control of an “essential” input can lead to the exclusion of rivals, a careful analysis of real-world evidence indicates that such concerns are unwarranted for many online businesses that have been the focus of the “big data” debate.[115]

While data can be an important part of the competitive landscape, incumbents’ data advantages are far less pronounced than today’s policymakers commonly assume. In that respect, five primary lessons emerge:

1. Data can be (very) valuable, but beyond a certain threshold, those benefits tend to diminish. In other words, having the most data is less important than having enough;
2. The ability to generate valuable information does not depend on the number of users or the amount of data a platform has previously acquired;
3. The most important datasets are not always proprietary;
4. Technological advances and platforms’ engineering decisions affect their ability to generate valuable information, and this effect swamps those that stem from the amount of data they own; and
5. How platforms use data is arguably more important than what data or how much data they own.

These lessons have important ramifications for policy debates over the competitive implications of data in technologically evolving areas.

First, it is not surprising that startups, rather than incumbents, have taken an early lead in generative AI (and in Web 2.0 before it). After all, if data-incumbency advantages are small or even nonexistent, then smaller and more nimble players may have an edge over established tech platforms. This is all the more likely given that, despite significant efforts, the biggest tech platforms were unable to offer compelling generative-AI chatbots and image-generation services before the emergence of ChatGPT, Dall-E, Midjourney, etc.

This suggests that, in a process akin to Clayton Christensen’s “innovator’s dilemma,”[116] something about the incumbent platforms’ existing services and capabilities might have been holding them back in this emerging industry. Of course, this does not necessarily mean that those same services or capabilities could not become an advantage when the generative-AI industry starts addressing issues of monetization and scale.[117] But it does mean that assumptions about a firm’s market power based primarily on its possession of data are likely to be off the mark.

Another important implication is that, paradoxically, policymakers’ efforts to prevent Web 2.0 platforms from competing freely in generative-AI markets may ultimately backfire and lead to less, not more, competition. Indeed, OpenAI is currently acquiring a sizeable lead in generative AI. While competition authorities might like to think that other startups will emerge and thrive in this space, it is important not to confuse those desires with reality. While there currently exists a vibrant AI-startup ecosystem, there is at least a case to be made that significant competition for today’s AI leaders will come from incumbent Web 2.0 platforms—although nothing is certain at this stage.

Policymakers should beware not to stifle that competition on the misguided assumption that competitive pressure from large incumbents is somehow less valuable to consumers than those that originate from smaller firms. This is particularly relevant in the context of merger control. An acquisition (or an “acqui-hire”) by a “Big Tech” company does not only, in principle, entail a minor risk to harm competition (it is not a horizontal merger),[118] but could create a stronger competitor to the current market leaders.

Finally, even if there were a competition-related market failure to be addressed in the field of generative AI (which is anything but clear), the remedies under contemplation may do more harm than good. Some of the solutions that have been put forward have highly ambiguous effects on consumer welfare. Scholars have shown that, e.g., mandated data sharing—a solution championed by EU policymakers, among others—may sometimes dampen competition in generative AI.[119] This is also true of legislation like the General Data Protection Regulation (GDPR), which makes it harder for firms to acquire more data about consumers—assuming such data is, indeed, useful to generative AI services.[120]

In sum, it is a flawed understanding of the economics and practical consequences of large agglomerations of data that has led competition authorities to believe data-incumbency advantages are likely to harm competition in generative AI—or even in the data-intensive Web 2.0 markets that preceded it. Indeed, competition or regulatory intervention to “correct” data barriers and data network and scale effects is liable to do more harm than good.

[1] Australian Competition & Consumer Commission, Digital Platform Services Inquiry —March 2025 — Final Report Issues Paper (25 Jul. 25 2024), available at https://www.accc.gov.au/system/files/dpsi-10-final-report-issues-paper.pdf?ref=0&download=y (hereinafter “issues paper” or “final report issues paper”). See also Press Release, Final Digital Platforms Report to Focus on Global Developments and Emerging Competition and Consumer Issues, Australian Competition & Consumer Commission (25 Jul. 2024), https://www.accc.gov.au/media-release/final-digital-platforms-report-to-focus-on-global-developments-and-emerging-competition-and-consumer-issues.

[2] Press release, id.

[3] Digital Platform Services Inquiry: Interim Report No. 5 — Regulatory Reform, Australian Competition & Consumer Commission (Sep. 2022), at 47, available at https://www.accc.gov.au/system/files/Digital%20platform%20services%20inquiry%20-%20September%202022%20interim%20report.pdf.

[4] Artificial intelligence is, of course, not a market (at least, not an antitrust-relevant market). Within the realm of what is called “AI,” companies offer myriad products and services, and specific relevant markets would need to be defined before assessing harm to competition in specific cases.

[5] Issues paper, supra note 1, at 4.

[6] Robert Armstrong & Ethan Wu, What Big Tech Antitrust Gets Wrong. An Interview with Herbert Hovenkamp, Financial Times (19 Jan. 2024), https://www.ft.com/content/4eec8bc3-c892-4704-ae66-a4432c6d4fd7.

[7] By ex-ante regulation, we mean specific rules and duties that are sector-specific (“digital markets”), whose application would not be based on the effects of the conduct regulated, and where fines would apply in case of non-compliance. See Bruce H. Kobayashi & Joshua D. Wright, Antitrust and Ex-Ante Sector Regulation, The GAI Report on the Digital Economy 25 (2020); see also id. at Table 1 at 869.

[8] See Robert Cooter & Tomas Ulen, Law and Economics 40-43 (2000); W. Kip Viscusi, Joseph E. Harrington, Jr. and John M. Vernon, Economics of Regulation and Antitrust 376-79 (2005).

[9] David S. Evans & Richard Schmalensee, Debunking The “Network Effects” Bogeyman, Regulation 36, 39 (Winter 2017-2018) available at https://www.cato.org/sites/cato.org/files/serials/files/regulation/2017/12/regulation-v40n4-1.pdf.

[10] Giuseppe Colangelo, Evaluating the Case for Regulation of Digital Platforms, The GAI Report on the Digital Economy 905, 930 (2020) https://gaidigitalreport.com/2020/10/04/evaluating-the-case-for-ex-ante-regulation-of-digital-platforms.

[11] Interim Report No 5, supra note 3, at 48.

[12] Australian Competition & Consumer Commission, supra note 3, at 48-49.

[13] We often run the risk of condemning business practices and models we don’t fully understand. Sometimes, even the businesses that implement them don’t fully know or understand the impact of such practices. See Frank H. Easterbrook, The Limits of Antitrust, 63 Tex. L. Rev. 1 (1984).

[14] Dirk Auer & Lazar Radic, The Growing Legacy of Intel, 14 J. Eur. Comp. L. & Pract. 15 (2023), (“Competition cases routinely hinge on the fundamental distinction between conduct that anti-competitively serves to exclude competitors, on the one hand, and competition on the merits that may lead firms to exit the market, on the other. (…) anticompetitive foreclosure and competition on the merits both ultimately result in the same observable outcome: namely, that rivals exit the market. In order to draw the line, policymakers must infer both the root causes and the effects of firms’ market exit”.)

[15] Jean Tirole, Competition and the Industrial Challenge for the Digital Age, 15 Annual Rev. of Econ. 573, 581 (2023), available at https://www.annualreviews.org/content/journals/10.1146/annurev-economics-090622-024222.

[16] Alba Ribera, La Regulación de los Ecosistemas Digitales Frente a las Relaciones Complejas se los Operadores Económicos, Centro Competencia (18 Oct. 2023), https://centrocompetencia.com/regulacion-ecosistemas-digitales-relaciones-complejas-operadores-economicos (Free translation of the following text in Spanish: “Uno de los mayores ejemplos de la dicotomía que se erige entre los distintos tipos de consecuencias que se pueden generar por la captura regulatoria de los ecosistemas digitales lo podemos encontrar en la reciente decisión de Meta, de no lanzar su nuevo servicio Threads en el Espacio Económico Europeo. En la medida en que su servicio podría interpretarse de forma que cayera dentro de la definición de un “servicio básico de plataforma” perteneciente a la categoría de redes sociales en línea” (listada por la LMD), Meta decidió abstenerse de entrar en el mercado europeo, por la carga desproporcionada que le supondría las exigentes obligaciones impuestas por la LMD. Cabe notar que Threads es aún un servicio entrante en el mercado de redes sociales en línea, en contraste con la posición predominante ocupada por la actual X (anteriormente conocida como Twitter). De esta forma, observamos que la categorización como servicio básico de plataforma unifica y elimina todos los matices que el propio juego de la libre competencia opera respecto de servicios entrantes en los mercados.”).

[17] Press Release, Facebook and Instagram to Offer Subscription for No Ads in Europe, Meta (30 Oct. 2023), https://about.fb.com/news/2023/10/facebook-and-instagram-to-offer-subscription-for-no-ads-in-europe.

[18] Geoffrey Manne & Dirk Auer, Brussels Effect or Brussels Defect: Digital Regulation in Emerging Markets, Truth on the Market (20 Dec. 2022), https://truthonthemarket.com/2022/12/20/brussels-effect-or-brussels-defect-digital-regulation-in-emerging-markets.

[19] Adam Kovacevich, Europe’s Digital Market Act Fails Consumers, Chamber of Progress (4 Mar. 2024), https://medium.com/chamber-of-progress/europes-digital-market-act-fails-consumers-dcaf70cc548c.

[20] Id.

[21] Id.

[22] See Jon Porter & David Pierce, Apple Is Bringing Sideloading and Alternate App Stores to the iPhone, The Verge (25 Jan. 2024), https://www.theverge.com/2024/1/25/24050200/apple-third-party-app-stores-allowed-iphone-ios-europe-digital-markets-act.

[23] See Apple, Complying with the Digital Markets Act (2024), available at https://developer.apple.com/security/complying-with-the-dma.pdf.

[24] Kim Mackrael, Apple’s Hold on the App Store Is Loosening, at Least in Europe, Wall St. J. (16 Aug. 2024), https://www.wsj.com/tech/epic-games-apple-app-store-europe-44ceda50.

[25] Dan Goodin, Novel Technique Allows Malicious Apps to Escape iOS and Android Guardrails, ArsTechnica (21 Aug. 2024), https://arstechnica.com/security/2024/08/novel-technique-allows-malicious-apps-to-escape-ios-and-android-guardrails.

[26] See id. (“Both mobile operating systems employ mechanisms designed to help users steer clear of apps that steal their personal information, passwords, or other sensitive data. iOS bars the installation of all apps other than those available in its App Store, an approach widely known as the Walled Garden.”).

[27] Josephine Wolff, Software crash Exposes Tensions Between Security and Competition, Financial Times (28 Jul. 2024), https://www.ft.com/content/60dde560-194a-40d1-8c98-1d96d6d019a0.

[28] Dirk Auer, The Broken Promises of Europe’s Digital Regulation, Truth on the Market (12 Mar. 2024), https://truthonthemarket.com/2024/03/12/the-broken-promises-of-europes-digital-regulation.

[29] Mirai, DMA’s Negative Impact (Feb. 2024), available at LinkedIn, https://www.linkedin.com/feed/update/urn:li:activity:7161330551709138945 (“Since the implementation of the DMA on the 19th of January, the number of clicks from Google Hotel Ads to hotel websites has decreased by 17.6% in EU countries compared to the rest of the world.”).

[30] See Press Release, Parliament Passes News Media and Digital Platforms Mandatory Bargaining Code, Treasury Portfolio Ministers and Minister for Communications, Urban Infrastructure, Cities and the Arts, (25 Feb. 2021), https://ministers.treasury.gov.au/ministers/josh-frydenberg-2018/media-releases/parliament-passes-news-media-and-digital-platforms.

[31] Josh Taylor, Facebook’s Potential News Ban Already Affecting Smaller Australian Media Outlets, Inquiry Told, The Guardian (21 Jun. 2024), https://www.theguardian.com/media/article/2024/jun/21/facebooks-potential-news-ban-already-affecting-smaller-australian-media-outlets-inquiry-told.

[32] See, e.g., Thomas Lambert, Tech Platforms and Market Power: What’s the Optimal Policy Response?, Mercatus Working Paper (Nov. 2021), at 14, available at https://www.mercatus.org/research/working-papers/tech-platforms-and-market-power-whats-optimal-policy-response (“Because they are more rigid and prescriptive than antitrust’s flexible standards, and thus less likely to be appropriate for a broad range of diverse firms, ex ante rules addressing market power concerns tend to be limited in scope. They are usually tailored for a particular industry or group of firms. Antitrust’s standards are focused on ends rather than specific means, and are therefore less likely to ‘misfire’ when applied broadly.”).

[33] See Richard Posner, Antitrust Law (2nd. ed. 2001), at 39 (“Rules are generally simpler and cheaper to enforce than standards and provide clearer guidance both to the people subject to them and to the courts that administer them. But they are often either underinclusive or overinclusive, and sometimes they are both at the same time. They are especially apt to fail as a sensible method of lawmaking when the relation of the rule’s purpose to the fact of facts that it makes determinative or legality is unclear. In such cases, the decision whether to characterize the case as falling within the domain of the rule may depend on the same factors that would determine legality under a standard.”)

[34] Issues paper, supra note 1, at 8.

[35] Kwon Soon-wan &Yeom Hyun-a, South Korea Hits Pause on Anti-Monopoly Platform Act Targeting Google, Apple, The Chosun Daily (8 Feb. 2024), https://www.chosun.com/english/national-en/2024/02/08/A4U4X6TWEFFOXF7ITCS5K6SZN4.

[36] Issues paper, supra note 1, at 10.

[37] Id. at 9-10.

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

[39] See, Most Used Messenger by Brand in Australia as of June 2024, Statista (31 Jul. 2024), https://www.statista.com/forecasts/1187986/most-used-messenger-by-brand-in-australia.

[40] See supra note 9 and accompanying text.

[41] Catherine Tucker, Network Effects and Market Power: What Have We Learned in the Last Decade?, Antitrust 72, 75-76 (2018) (“the rise and fall of such platforms may also be more dramatic and renders such platforms far more vulnerable to a marginally superior competitor. The sudden decline of MySpace can be explained by the idea that given limited ability to spend time on two similar social networks, users switched far more quickly to a competitor.”).

^[42] For a review of the literature on increasing returns to scale in data (a topic somewhat broader than data-network effects), see Geoffrey Manne & Dirk Auer, Antitrust Dystopia and Antitrust Nostalgia: Alarmist Theories of Harm in Digital Markets and Their Origins, 28 Geo. Mason L. Rev. 1281, 1344 (2021).

[43] Issues paper, supra note 1, at 10.

[44] See generally William J. Baumol, Contestable Markets: An Uprising in the Theory of Industry Structure, 72 Am. Econ. Rev. 1 (1982); William J. Baumol, John Panzar & Robert D. Willig, Contestable Markets and the Theory of Industry Structure (revised ed. 1988).

[45] See, e.g., Andrew Lanxon, Android vs. iPhone: 15 Years of Innovation Through Rivalry, CNET (24 Apr. 2024), https://www.cnet.com/tech/mobile/smartphone-showdown-15-years-of-android-vs-iphone.

[46] Issues paper, supra note 1, at 13-14.

[47] Nathan Newman, Taking on Google’s Monopoly Means Regulating Its Control of User Data, Huffington Post (24 Sep. 2013), http://www.huffingtonpost.com/nathan-newman/taking-on-googlesmonopol_b_3980799.html.

[48] See, e.g., Lina Khan & K. Sabeel Rahman, Restoring Competition in the U.S. Economy, in Untamed: How to Check Corporate, Financial, and Monopoly Power (Nell Abernathy, Mike Konczal, & Kathryn Milani, eds., 2016), at 23. (“From Amazon to Google to Uber, there is a new form of economic power on display, distinct from conventional monopolies and oligopolies…, leverag[ing] data, algorithms, and internet-based technologies… in ways that could operate invisibly and anticompetitively.”); Mark Weinstein, I Changed My Mind—Facebook Is a Monopoly, Wall St. J. (1 Oct. 2021), https://www.wsj.com/articles/facebook-is-monopoly-metaverse-users-advertising-platforms-competition-mewe-big-tech-11633104247 (“[T]he glue that holds it all together is Facebook’s monopoly over data…. Facebook’s data troves give it unrivaled knowledge about people, governments—and its competitors.”).

[49] See, generally, Abigail Slater, Why “Big Data” Is a Big Deal, The Reg. Rev. (6 Nov. 2023), https://www.theregreview.org/2023/11/06/slater-why-big-data-is-a-big-deal; Amended Complaint at ¶36, United States v. Google, 1:20-cv-03010- (D.D.C. 2020); Complaint at ¶37, United States v. Google, 1:23-cv-00108 (E.D. Va. 2023), https://www.justice.gov/opa/pr/justice-department-sues-google-monopolizing-digital-advertising-technologies (“Google intentionally exploited its massive trove of user data to further entrench its monopoly across the digital advertising industry.”).

[50] See, e.g., Press Release, Commission Launches Calls for Contributions on Competition in Virtual Worlds and Generative AI, European Commission (9 Jan. 2024), https://ec.europa.eu/commission/presscorner/detail/en/IP_24_85; Krysten Crawford, FTC’s Lina Khan Warns Big Tech over AI, SIEPR (3 Nov. 2020), https://siepr.stanford.edu/news/ftcs-lina-khan-warns-big-tech-over-ai (“Federal Trade Commission Chair Lina Khan delivered a sharp warning to the technology industry in a speech at Stanford on Thursday: Antitrust enforcers are watching what you do in the race to profit from artificial intelligence.”) (emphasis added).

[51] See, e.g., John M. Newman, Antitrust in Digital Markets, 72 Vand. L. Rev. 1497, 1501 (2019) (“[T]he status quo has frequently failed in this vital area, and it continues to do so with alarming regularity. The laissez-faire approach advocated for by scholars and adopted by courts and enforcers has allowed potentially massive harms to go unchecked.”); Bertin Martins, Are New EU Data Market Regulations Coherent and Efficient?, Bruegel Working Paper 21/23 (2023), https://www.bruegel.org/working-paper/are-new-eu-data-market-regulations-coherent-and-efficient (“Technical restrictions on access to and re-use of data may result in failures in data markets and data-driven services markets.”); Valéria Faure-Muntian, Competitive Dysfunction: Why Competition Law Is Failing in a Digital World, The Forum Network (24 Feb. 2021), https://www.oecd-forum.org/posts/competitive-dysfunction-why-competition-law-is-failing-in-a-digital-world.

[52] See Rana Foroohar, The Great US-Europe Antitrust Divide, Financial Times (5 Feb. 2024), https://www.ft.com/content/065a2f93-dc1e-410c-ba9d-73c930cedc14.

[53] See, e.g., Press Release, supra note 50.

[54] See infra, Section III.B. Commentators have also made similar claims; see, e.g., Ganesh Sitaram & Tejas N. Narechania, It’s Time for the Government to Regulate AI. Here’s How, Politico (15 Jan. 2024), https://www.politico.com/news/magazine/2024/01/15/sitaraman-artificial-intelligence-regulation-00134873 (“All that cloud computing power is used to train foundation models by having them “learn” from incomprehensibly huge quantities of data. Unsurprisingly, the entities that own these massive computing resources are also the companies that dominate model development. Google has Bard, Meta has LLaMa. Amazon recently invested $4 billion into one of OpenAI’s leading competitors, Anthropic. And Microsoft has a 49 percent ownership stake in OpenAI—giving it extraordinary influence, as the recent board struggles over Sam Altman’s role as CEO showed.”).

[55] Press Release, supra note 50.

[56] Comment of U.S. Federal Trade Commission to the U.S. Copyright Office, Artificial Intelligence and Copyright, Docket No. 2023-6 (30 Oct. 2023), at 4, https://www.ftc.gov/legal-library/browse/advocacy-filings/comment-federal-trade-commission-artificial-intelligence-copyright (emphasis added).

[57] Jonathan Kanter, Remarks at the Promoting Competition in AI Conference (30 May 2024), https://youtu.be/yh–1AGf3aU?t=424.

[58] Karin Matussek, AI Will Fuel Antitrust Fires, Big Tech’s German Nemesis Warns, Bloomberg (26 Jun. 2024), https://www.bloomberg.com/news/articles/2024-06-26/ai-will-fuel-antitrust-fires-big-tech-s-german-nemesis-warns?srnd=technology-vp.

[59] Id.

[60] See, e.g., Joe Caserta, Holger Harreis, Kayvaun Rowshankish, Nikhil Srinidhi, & Asin Tavakoli, The Data Dividend: Fueling Generative AI, McKinsey Digital (15 Sep. 2023), https://www.mckinsey.com/capabilities/mckinsey-digital/our-insights/the-data-dividend-fueling-generative-ai (“Your data and its underlying foundations are the determining factors to what’s possible with generative AI.”).

[61] See, e.g., Tim Keary, Google DeepMind’s Achievements and Breakthroughs in AI Research, Techopedia (11 Aug. 2023), https://www.techopedia.com/google-deepminds-achievements-and-breakthroughs-in-ai-research; see, e.g., Will Douglas Heaven, Google DeepMind Used a Large Language Model to Solve an Unsolved Math Problem, MIT Technology Review (14 Dec. 2023), https://www.technologyreview.com/2023/12/14/1085318/google-deepmind-large-language-model-solve-unsolvable-math-problem-cap-set; see also, A Decade of Advancing the State-of-the-Art in AI Through Open Research, Meta (30 Nov. 2023), https://about.fb.com/news/2023/11/decade-of-advancing-ai-through-open-research; see also, 200 Languages Within a Single AI Model: A Breakthrough in High-Quality Machine Translation, Meta, https://ai.meta.com/blog/nllb-200-high-quality-machine-translation (last visited 18 Jan. 2023).

[62] See, e.g., Jennifer Allen, 10 Years of Siri: The History of Apple’s Voice Assistant, Tech Radar (4 Oct. 2021), https://www.techradar.com/news/siri-10-year-anniversary; see also Evan Selleck, How Apple Is Already Using Machine Learning and AI in iOS, Apple Insider (20 Nov. 2023), https://appleinsider.com/articles/23/09/02/how-apple-is-already-using-machine-learning-and-ai-in-ios; see also, Kathleen Walch, The Twenty Year History Of AI At Amazon, Forbes (19 Jul. 2019), https://www.forbes.com/sites/cognitiveworld/2019/07/19/the-twenty-year-history-of-ai-at-amazon.

[63] See infra Section III.C.

[64] See, e.g., Cédric Argenton & Jens Prüfer, Search Engine Competition with Network Externalities, 8 J. Comp. L. & Econ. 73, 74 (2012).

[65] John M. Yun, The Role of Big Data in Antitrust, The GAI Report on the Digital Economy 220, 233 (2020), https://gaidigitalreport.com/2020/08/25/big-data-and-barriers-to-entry/#_ftnref50; see also, e.g., Robert Wayne Gregory, Ola Henfridsson, Evgeny Kaganer, & Harris Kyriakou, The Role of Artificial Intelligence and Data Network Effects for Creating User Value, 46 Acad. of Mgmt. Rev. 534 (2020) (final pre-print version at 4), http://wrap.warwick.ac.uk/134220 (“A platform exhibits data network effects if, the more that the platform learns from the data it collects on users, the more valuable the platform becomes to each user.”); Karl Schmedders, José Parra-Moyano, & Michael Wade, Why Data Aggregation Laws Could be the Answer to Big Tech Dominance, Silicon Republic (6 Feb. 2024), https://www.siliconrepublic.com/enterprise/data-ai-aggregation-laws-regulation-big-tech-dominance-competition-antitrust-imd.

[66] Nathan Newman, Search, Antitrust, and the Economics of the Control of User Data, 31 Yale J. Reg. 401, 409 (2014) (emphasis added); see also id. at 420 & 423 (“While there are a number of network effects that come into play with Google, [“its intimate knowledge of its users contained in its vast databases of user personal data”] is likely the most important one in terms of entrenching the company’s monopoly in search advertising…. Google’s overwhelming control of user data… might make its dominance nearly unchallengeable.”).

[67] See also Yun, supra note 65 at 229 (“[I]nvestments in big data can create competitive distance between a firm and its rivals, including potential entrants, but this distance is the result of a competitive desire to improve one’s product.”).

^[68] See Manne & Auer, Antitrust Dystopia and Antitrust Nostalgia, supra note 42.

[69] Andrei Hagiu & Julian Wright, Data-Enabled Learning, Network Effects, and Competitive Advantage, 54 RAND J. Econ. 638 (2023).

^[70] Id. at 639. The authors conclude that “Data-enabled learning would seem to give incumbent firms a competitive advantage. But how strong is this advantage and how does it differ from that obtained from more traditional mechanisms… .”

^[71] Id.

^[72] Bruno Jullien & Wilfried Sand-Zantman, The Economics of Platforms: A Theory Guide for Competition Policy, 54 Info. Econ. & Pol’y 10080, 101031 (2021).

^[73] Daniele Condorelli & Jorge Padilla, Harnessing Platform Envelopment in the Digital World, 16 J. Comp. L. & Pol’y 143, 167 (2020).

[74] See Hagiu & Wright, supra note 6969.

^[75] For a summary of these limitations, see generally Catherine Tucker, Network Effects and Market Power, supra note 41; see also Manne & Auer, supra note 42, at 1330.

[76] See Jason Furman, Diane Coyle, Amelia Fletcher, Derek McAuley, & Philip Marsden (Dig. Competition Expert Panel), Unlocking Digital Competition (2019) at 32-35 (“Furman Report”), available at https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/785547/unlocking_digital_competition_furman_review_web.pdf.

[77] Id. at 34.

[78] Id. at 35. To its credit, it should be noted, the Furman Report does counsel caution before mandating access to data as a remedy to promote competition. See id. at 75. That said, the Furman Report maintains that such a remedy should remain on the table because “the evidence suggests that large data holdings are at the heart of the potential for some platform markets to be dominated by single players and for that dominance to be entrenched in a way that lessens the potential for competition for the market.” Id. The evidence, however, does not show this.

^[79] Case COMP/M.9660 — Google/Fitbit, Commission Decision (17 Dec. 2020) (Summary at O.J. (C 194) 7), available at https://ec.europa.eu/competition/mergers/cases1/202120/m9660_3314_3.pdf, at 455.

^[80] Id. at 896.

[81] See Natasha Lomas, EU Checking if Microsoft’s OpenAI Investment Falls Under Merger Rules, TechCrunch (9 Jan. 2024), https://techcrunch.com/2024/01/09/openai-microsoft-eu-merger-rules.

^[82] Amended Complaint at 11, Meta/Zuckerberg/Within, Fed. Trade Comm’n. (2022) (No. 605837), available at https://www.ftc.gov/system/files/ftc_gov/pdf/D09411%20-%20AMENDED%20COMPLAINT%20FILED%20BY%20COUNSEL%20SUPPORTING%20THE%20COMPLAINT%20-%20PUBLIC%20%281%29_0.pdf.

[83] Amended Complaint (D.D.C), supra note 49, at ¶37.

[84] Amended Complaint (E.D. Va), id., at ¶8.

^[85] Merger Guidelines, US Dep’t of Justice & Fed. Trade Comm’n (2023) at 25, available at https://www.ftc.gov/system/files/ftc_gov/pdf/2023_merger_guidelines_final_12.18.2023.pdf.

[86] Merger Assessment Guidelines, Competition and Mkts. Auth (2021) at ¶7.19(e), available at https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/1051823/MAGs_for_publication_2021_–_.pdf.

[87] Furman Report, supra note 76, at ¶4.

[88] See, e.g., Chris Westfall, New Research Shows ChatGPT Reigns Supreme in AI Tool Sector, Forbes (16 Nov. 2023), https://www.forbes.com/sites/chriswestfall/2023/11/16/new-research-shows-chatgpt-reigns-supreme-in-ai-tool-sector/?sh=7de5de250e9c; Sujan Sarkar, AI Industry Analysis: 50 Most Visited AI Tools and Their 24B+ Traffic Behavior, Writerbuddy (last visited 15 Jul. 2024), https://writerbuddy.ai/blog/ai-industry-analysis.

^[89] See Krystal Hu, ChatGPT Sets Record for Fastest-Growing User Base, Reuters (2 Feb. 2023), https://www.reuters.com/technology/chatgpt-sets-record-fastest-growing-user-base-analyst-note-2023-02-01; Google: The AI Race Is On, App Economy Insights (7 Feb. 2023), https://www.appeconomyinsights.com/p/google-the-ai-race-is-on.

^[90] See Google Trends, https://trends.google.com/trends/explore?date=today%205-y&q=%2Fg%2F11khcfz0y2,%2Fg%2F11ts49p01g&hl=en (last visited 12 Jan. 2024) and https://trends.google.com/trends/explore?date=today%205-y&geo=US&q=%2Fg%2F11khcfz0y2,%2Fg%2F11ts49p01g&hl=en (last visited 12 Jan. 2024).

^[91] See David F. Carr, As ChatGPT Growth Flattened in May, Google Bard Rose 187%, Similarweb Blog (5 Jun. 2023), https://www.similarweb.com/blog/insights/ai-news/chatgpt-bard.

^[92] See Press Release, Introducing New AI Experiences Across Our Family of Apps and Devices, Meta (27 Sep. 2023), https://about.fb.com/news/2023/09/introducing-ai-powered-assistants-characters-and-creative-tools; Sundar Pichai, An Important Next Step on Our AI Journey, Google Keyword Blog (6 Feb. 2023), https://blog.google/technology/ai/bard-google-ai-search-updates.

^[93] See Ion Prodan, 14 Million Users: Midjourney’s Statistical Success, Yon (19 Aug. 2023), https://yon.fun/midjourney-statistics; see also Andrew Wilson, Midjourney Statistics: Users, Polls, & Growth [Oct 2023], ApproachableAI (Oct. 13, 2023), https://approachableai.com/midjourney-statistics.

^[94] See Hema Budaraju, New Ways to Get Inspired with Generative AI in Search, Google Keyword Blog (12 Oct. 2023), https://blog.google/products/search/google-search-generative-ai-october-update; Imagine with Meta AI, Meta (last visited 12 Jan. 2024), https://imagine.meta.com.

^[95] Catherine Tucker, Digital Data, Platforms and the Usual [Antitrust] Suspects: Network Effects, Switching Costs, Essential Facility, 54 Rev. Indus. Org. 683, 686 (2019).

^[96] Manne & Auer, supra note 42, at 1345.

[97] See, e.g., Stefanie Koperniak, Artificial Data Give the Same Results as Real Data—Without Compromising Privacy, MIT News (3 Mar. 2017), https://news.mit.edu/2017/artificial-data-give-same-results-as-real-data-0303 (“[Authors] describe a machine learning system that automatically creates synthetic data—with the goal of enabling data science efforts that, due to a lack of access to real data, may have otherwise not left the ground. While the use of authentic data can cause significant privacy concerns, this synthetic data is completely different from that produced by real users—but can still be used to develop and test data science algorithms and models.”).

[98] See, e.g., Rachel Gordon, Synthetic Imagery Sets New Bar in AI Training Efficiency, MIT News (20 Nov. 2023), https://news.mit.edu/2023/synthetic-imagery-sets-new-bar-ai-training-efficiency-1120 (“By using synthetic images to train machine learning models, a team of scientists recently surpassed results obtained from traditional ‘real-image’ training methods.).

[99] Thibault Schrepel & Alex ‘Sandy’ Pentland, Competition Between AI Foundation Models: Dynamics and Policy Recommendations, MIT Connection Science Working Paper (Jun. 2023), at 8.

[100] Igor Susmelj, Optimizing Generative AI: The Role of Data Curation, Lightly (last visited 15 Jan. 2024), https://www.lightly.ai/post/optimizing-generative-ai-the-role-of-data-curation.

[101] See, e.g., Xiaoliang Dai, et al., Emu: Enhancing Image Generation Models Using Photogenic Needles in a Haystack, ArXiv (27 Sep. 2023) at 1, https://ar5iv.labs.arxiv.org/html/2309.15807 (“[S]upervised fine-tuning with a set of surprisingly small but extremely visually appealing images can significantly improve the generation quality.”); see also, Hu Xu, et al., Demystifying CLIP Data, ArXiv (28 Sep. 2023), https://arxiv.org/abs/2309.16671.

[102] Lauren Leffer, New Training Method Helps AI Generalize Like People Do, Sci. Am. (26 Oct. 2023), https://www.scientificamerican.com/article/new-training-method-helps-ai-generalize-like-people-do (discussing Brendan M. Lake & Marco Baroni, Human-Like Systematic Generalization Through a Meta-Learning Neural Network, 623 Nature 115 (2023)).

^[103] Timothy B. Lee, The Real Research Behind the Wild Rumors about OpenAI’s Q* Project, Ars Technica (8 Dec. 2023), https://arstechnica.com/ai/2023/12/the-real-research-behind-the-wild-rumors-about-openais-q-project.

^[104] Id.; see also GSM8K, Papers with Code (last visited 18 Jan. 2023), https://paperswithcode.com/dataset/gsm8k; MATH Dataset, GitHub (last visited 18 Jan. 2024), https://github.com/hendrycks/math.

^[105] See Lee, supra note 103.

[106] Geoffrey Manne & Ben Sperry, Debunking the Myth of a Data Barrier to Entry for Online Services, Truth on the Market (26 Mar. 2015), https://truthonthemarket.com/2015/03/26/debunking-the-myth-of-a-data-barrier-to-entry-for-online-services (citing Andres V. Lerner, The Role of ‘Big Data’ in Online Platform Competition (26 Aug. 2014), https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2482780.).

[107] See Catherine Tucker, Digital Data as an Essential Facility: Control, CPI Antitrust Chron. (Feb. 2020), at 11 (“[U]ltimately the value of data is not the raw manifestation of the data itself, but the ability of a firm to use this data as an input to insight.”).

[108] Or, as John Yun put it, data is only a small component of digital firms’ production function. See Yun, supra note 65, at 235 (“Second, while no one would seriously dispute that having more data is better than having less, the idea of a data-driven network effect is focused too narrowly on a single factor improving quality. As mentioned in supra Section I.A, there are a variety of factors that enter a firm’s production function to improve quality.”).

^[109] Luxia Le, The Real Reason Windows Phone Failed Spectacularly, History–Computer (8 Aug. 2023), https://history-computer.com/the-real-reason-windows-phone-failed-spectacularly.

[110] Introducing the GPT Store, Open AI (10 Jan. 2024), https://openai.com/blog/introducing-the-gpt-store.

^[111] See Michael Schade, How ChatGPT and Our Language Models Are Developed, OpenAI, https://help.openai.com/en/articles/7842364-how-chatgpt-and-our-language-models-are-developed; Sreejani Bhattacharyya, Interesting Innovations from OpenAI in 2021, AIM (1 Jan. 2022), https://analyticsindiamag.com/interesting-innovations-from-openai-in-2021; Danny Hernadez & Tom B. Brown, Measuring the Algorithmic Efficiency of Neural Networks, ArXiv (8 May 2020), https://arxiv.org/abs/2005.04305.

[112] See Yun, supra note 65 at 235 (“Even if data is primarily responsible for a platform’s quality improvements, these improvements do not simply materialize with the presence of more data—which differentiates the idea of data-driven network effects from direct network effects. A firm needs to intentionally transform raw, collected data into something that provides analytical insights. This transformation involves costs including those associated with data storage, organization, and analytics, which moves the idea of collecting more data away from a strict network effect to more of a ‘data opportunity.’”).

[113] Josh Sisco, POLITICO PRO Q&A: Exit Interview with DOJ Chief Antitrust Economist Susan Athey, Politico Pro (2 Jul. 2024), https://subscriber.politicopro.com/article/2024/07/politico-pro-q-a-exit-interview-with-doj-chief-antitrust-economist-susan-athey-00166281.

[114] Belle Lin, Open-Source Companies Are Sharing Their AI Free. Can They Crack OpenAI’s Dominance?, Wall St. J. (21 Mar. 2024), https://www.wsj.com/articles/open-source-companies-are-sharing-their-ai-free-can-they-crack-openais-dominance-26149e9c.

[115] Lerner, supra note 106, at 4-5 (emphasis added).

[116] See Clayton M. Christensen, The Innovator’s Dilemma: When New Technologies Cause Great Firms to Fail (2013).

[117] See David J. Teece, Dynamic Capabilities and Strategic Management: Organizing for Innovation and Growth (2009).

[118] Antitrust merger enforcement has long assumed that horizontal mergers are more likely to cause problems for consumers than vertical mergers. See Geoffrey A. Manne, Dirk Auer, Brian Albrecht, Eric Fruits, Daniel J. Gilman, & Lazar Radic, Comments of the International Center for Law and Economics on the FTC & DOJ Draft Merger Guidelines, International Center for Law & Economics (18 Sep. 2023), https://laweconcenter.org/resources/comments-of-the-international-center-for-law-and-economics-on-the-ftc-doj-draft-merger-guidelines.

[119] See Hagiu & Wright, supra note 69, at 69 (“We use our dynamic framework to explore how data sharing works: we find that it in-creases consumer surplus when one firm is sufficiently far ahead of the other by making the laggard more competitive, but it decreases consumer surplus when the firms are sufficiently evenly matched by making firms compete less aggressively, which in our model means subsidizing consumers less.”); see also Lerner, supra note 106.

[120] See, e.g., Hagiu & Wright, id. (“We also use our model to highlight an unintended consequence of privacy policies. If such policies reduce the rate at which firms can extract useful data from consumers, they will tend to increase the incumbent’s competitive advantage, reflecting that the entrant has more scope for new learning and so is affected more by such a policy.”); Jian Jia, Ginger Zhe Jin, & Liad Wagman, The Short-Run Effects of the General Data Protection Regulation on Technology Venture Investment, 40 Marketing Sci. 593 (2021) (finding GDPR reduced investment in new and emerging technology firms, particularly in data-related ventures); James Campbell, Avi Goldfarb, & Catherine Tucker, Privacy Regulation and Market Structure, 24 J. Econ. & Mgmt. Strat. 47 (2015) (“Consequently, rather than increasing competition, the nature of transaction costs implied by privacy regulation suggests that privacy regulation may be anti-competitive.”).

I. Introduction

In its request for testimony related to the Sept. 16 and 17, 2024 hearings, the Surface Transportation Board (STB) has asked for commentary on “how the industry has grown and intends to grow in the future.”^[1] While this line of inquiry reflects the STB’s interest in understanding industry dynamics, it raises questions about the most effective approach to fulfilling the STB’s regulatory role under the Staggers Rail Act of 1980 (“the Act”).

The STB’s focus on the rail industry’s growth plans merits careful consideration. It is important to reflect on whether scrutinizing or directing industry growth falls within the purview of a regulatory agency. The STB’s primary role, as defined by its organic statute, is to ensure that firms behave within appropriate constraints, rather than to guide commercial strategies. Additionally, the Act’s Rail Transportation Policy section and operative language do not explicitly mention industry growth as a goal. Instead, the Act emphasizes that the STB should make an “adequate and continuing effort” to provide rail carriers with adequate revenue levels to earn “a reasonable and economic profit or return (or both) on capital employed in the business.”^[2] This suggests that the industry should be guided primarily by market forces, rather than regulatory direction.

By concentrating on growth plans, the STB may inadvertently be missing opportunities to foster technological and business-model innovation crucial to achieve the Act’s policy goals of creating a sound, safe, and efficient rail-transportation system. Such innovation and experimentation are essential to realize these objectives. In examining these issues, we aim to offer constructive feedback that may help the STB refine its regulatory approach to align more closely with the Act’s intended purposes.

The following sections will elaborate on each of these points, providing evidence and arguments based on economic principles, industry examples, and the Act’s statutory language. Our goal is to contribute to a productive dialogue on how best to achieve the STB’s important regulatory objectives.

II. STB Focus Should Center on Where Regulation Impedes Profitability

The STB’s concern with the decline in carload volumes over the past decade, while generally understandable, is misguided in at least two respects. First, the STB’s remit is not to judge private firms’ commercial strategies, but to determine the proper, minimal regulatory environment necessary, pursuant to the Act. Second, “growth” of rail carriage, however judged, is not the same thing as profitability, the actual metric by which firms should be judged when determining their long-term viability.

Following these two points, the correct question for the STB to ask is the extent to which existing regulations prevent rail carriers from experimenting with business models and technologies that could help them to achieve long-term profitability. It is inappropriate and outside the proper scope of a regulatory agency like the STB to question firms about their plans for “growth.”

A. The Staggers Rail Act Calls for Stanching Overregulation

The Staggers Rail Act was designed to allow market forces to guide the rail industry’s development, with minimal regulatory intervention. A primary purpose of the Act is “[t]o allow, to the maximum extent possible, competition and the demand for services to establish reasonable rates for transportation by rail.”[3] Thus, market forces, and not regulatory fiat, are and should remain the primary drivers of rail carriers’ commercial decisions. Indeed, as Bernard Sharfman elaborates, a proper interpretation of the Act must consider the overall deregulatory thrust of its provisions:

Based on an analysis of the fifteen policy statements found in the Rail Transportation Policy section of the Act, it is clear that minimizing regulation in the freight rail industry is the primary objective of the Act. This is supported by viewing the Act in its historical context and, most importantly, the Act’s operative language.[4]

Other provisions of the Act that empower the STB to promulgate rules or otherwise conduct business derive from this primary objective.[5] That is to say, the STB’s ability to act, whether through informal hearings or formal rulemaking, must always be constrained by the goal of minimizing onerous regulations on rail carriers.

Given this background, the STB’s current hearings are curious. Inquiring about how private firms plan to grow suggests an interest in intervening in the firms’ core conduct—their selection of business strategies, technologies, etc.—that go beyond the limited remit of the Act to, e.g., ensure reasonable compliance with common-carriage requirements, permitted rate regulation, and so forth. So long as a firm’s output complies with the STB’s properly crafted regulations, it is inappropriate to inquire into “growth” strategies.

Beyond the fact that it exceeds the proper scope of STB authority, questioning firms about their growth plans may lead to unintended consequences. If viewed as a form of jawboning from their regulator, it may lead companies to focus on short-term “growth” metrics—such as adding unprofitable service or hiring more personnel than necessary—rather than on long-term profitability.

B. Profitability Is the Proper Metric to Evaluate Firms

Even on its own terms, the STB’s focus on “growth” is misplaced. It’s crucial to recognize that raw volume growth is not always the most relevant metric to assess industry health or economic efficiency. In many cases, profitability and returns on invested capital (ROIC) are more important indicators of either a firm’s or an industry’s performance, as well as how they allocate resources. In other words, profit-maximizing firms are interested in increasing sales volumes only inasmuch as the increased volumes are associated with an increase in ROIC or profits.

In industries with high fixed costs and capital intensity, such as freight rail, firms often focus on maximizing profitability, rather than pure volume growth. This approach aligns with the economic concept of allocative efficiency, where resources are directed to their most valuable uses.[6] Like all firms, railroads have limited resources with which to make their investments. While profitability is a necessary precondition for investment, not all profitable investments can be undertaken.

Among the universe of potentially profitable projects, firms are likely to give priority to those that promise greater returns on investment relative to those with lower ROIC. Thus, any evaluation of railroad investments must examine not only whether a given investment is likely to be profitable, but also how its expected returns compare to other investment opportunities. For railroads, this may mean focusing on high-margin traffic, optimizing network efficiency, or investing in technology to reduce operating costs. Such strategies can lead to improved financial performance and long-term sustainability, even if overall carload volumes stall or decrease.

Recent empirical research by Susanna Mansikkamäki on the relationship between firm growth and profitability finds: “A carefully planned growth strategy that avoids non-profitable growth becomes increasingly important with age and size.”[7] Mansikkamäki concludes: “[L]arger firms may have a size-based competitive advantage but only if they stay profitable. When profitability drops, the advantage can become a disadvantage.”[8]

In the context of the rail industry, profitability is particularly important given the capital-intensive nature of the business. Railroads require significant ongoing investments in infrastructure maintenance and upgrades. Without sustainable profitability, these investments become difficult or impossible to make, potentially leading to deterioration in service quality and safety. Indeed, even the Act itself contemplates regulated carriers’ profitability as a concern, stating that one its goals is to “promote a safe and efficient rail transportation system by allowing rail carriers to earn adequate revenues.”[9]

By contrast, policies aimed at promoting volume growth might have unintended consequences. For example, encouraging unprofitable traffic could lead to underinvestment in infrastructure and reduced service quality for all customers. Moreover, a single-minded focus on volume growth might divert attention and investments from innovations that could improve efficiency, reduce operating costs, and place downward pressure on freight rates.

C. STB Should Investigate Regulations that Interfere with Profitability

The Staggers Rail Act was enacted in 1980 in response to a ruinous state of affairs in the rail industry, largely due to years of overregulation.[10] As President Jimmy Carter said when signing the Act into law: “At the heart of this legislation is freeing the railroad industry and its customers from… excessive [regulatory] control.”[11] The Act therefore explicitly focuses on both deregulation[12] and ensuring that carriers receive adequate revenue.[13]

In a recent working paper, Bernard Sharfman recommends an approach to statutory interpretation that can better align the STB’s activity with the Act’s overall aims. He argues it should be viewed as an optimization problem, where regulation is to be minimized subject to the constraints found in the Rail Transportation Policy section.[14] Sharfman notes:

If the fifteen listed policy statements were all weighted equally, then it would be problematic in how and when to apply them. Conflicts would constantly arise. However, reading the Rail Transportation Policy section in its historical context makes clear there is one and only one primary objective— minimizing regulation in the freight rail industry.[15]

Indeed, as Sharfman notes, courts have approached the Act in a very similar manner. In CSX Transp. v. U.S., the D.C. Circuit held that:

Congress made its [deregulatory] intent absolutely clear by adopting a statement of Rail Transportation Policy section, that began as follows:

(1) to allow, to the maximum extent possible, competition and the demand for services to establish reasonable rates for transportation by rail; (2) to minimize the need for Federal regulatory control over the rail transportation system and to require fair and expeditious regulatory decisions when regulation is required.[16]

Implicit in allowing maximal competition is creating an environment in which profit and loss signals are undistorted by regulation, and profit opportunities draw new competitive entries into the market. Further, the Act itself explicitly forbids the STB from interfering in private contracting and rate setting between carriers and shippers (outside of situations where “market dominance” exists).[17]

Thus, instead of scrutinizing “growth” plans, the STB should focus on ensuring that its regulatory actions do not unnecessarily interfere with the industry’s ability to operate efficiently and profitably. This is not to say that the STB has no role in helping firms to “grow” or otherwise be profitable. The deregulatory thrust of the Act, and the provisions noted above, suggest that the STB has an enduring obligation to determine when regulatory activity might interfere with new business models or technological innovations that could enable regulated entities to grow profitably.

For instance, the STB should evaluate how its regulations impact rail carriers’ ability to price their services competitively. The Act states that rail carriers “may establish any rate for transportation or other service provided by the rail carrier.”[18] Regulatory actions that hinder this freedom can directly affect profitability.

The STB could also evaluate how regulations might constrain carriers’ ability to adapt their operations to changing market conditions,[19] or how regulatory uncertainty or overly burdensome rules could discourage capital investments needed for long-term viability.[20] And, of course, technological innovation—as in every sector of the economy—will remain crucial to ensure the future profitability of the freight-rail industry.[21] The STB should be doing everything within its remit to ensure that rail operators are empowered to experiment with new technologies that could increase consumer welfare dramatically.

For example, one study suggests that if crew sizes were to be reduced from two to one, carriers could save $2.5 billion annually.[22] Finding ways to promote automated trains could help make this a reality, and the STB may be in a position to determine where existing regulations make this kind of automation difficult or impracticable. By contrast, the recent reciprocal-switching rule[23] could have a negative effect on a carrier’s profitability and capability to invest in long-range capital investment. The STB should reconsider this rule in light of this negative effect on long-term incentives.

Looking forward, intermodal competition is only expected to increase, particularly as technology improves in trucking (including the introduction of automated trucking).[24] This will continue to put pressure on rail freight as a shipping option, and put even more importance on the deregulatory aims of the Act.

One possible approach for the STB is to enact a “dynamic regulatory” regime, as outlined by Geoffrey A. Manne and Gus Hurwitz.[25] In essence, they advocate for regulators to construct information-feedback mechanisms between a regulator and a regulated industry.[26] Using these feedback mechanisms, the regulator serves as an important information gatherer who can then use the data on the performance of regulated entities to incrementally modify (improve) regulations.[27] This will necessarily require a flexible view of regulation; as information is collected, some rules will need to be attenuated, while in other cases, new extensions may be required. But ultimately, the goal is for the regulator to view itself as a dynamic entity that bases its behavior on empirical market realities.

Relevant to the spirit of this proceeding, STB could be well-positioned to, among other things, collect information on the profit challenges that carriers face. The STB could develop and utilize performance indicators that reflect the industry’s long-term financial sustainability. These metrics should consider such factors as return on invested capital, operating-ratio improvements, and long-term investment capacity, which can then feed back into STB’s review of its regulations and how they intersect with these factors.

Along these lines, another recommendation to improve the STB’s regulatory decision-making would be to adopt formal requirements to conduct cost-benefit analysis in its rulemakings, similar to those used by executive agencies. Specifically, the STB could amend its procedures to: (1) require the inclusion of a cost-benefit analysis for proposed and final rules, including consideration of reasonable alternatives; (2) explicitly consider the cumulative impact of new rules in light of existing regulatory burdens; and (3) ensure the use of reliable, up-to-date data that reflects current market realities.

Such reforms would help ensure that board members have the information needed to make well-informed decisions, increase transparency, and reduce the likelihood of adopting rules that impose costs exceeding their benefits. By institutionalizing rigorous economic analysis, the STB can better fulfill its statutory obligations to minimize unnecessary regulation, while promoting a healthy rail industry.

III. Conclusion

The STB’s recent focus on industry growth plans represents a concerning departure from its mandated role under the Act. This approach not only misinterprets the Act’s primary objective of minimizing regulation, but also risks interfering with the market forces that should guide the rail industry’s development.

We urge the STB to reconsider its regulatory philosophy and realign its actions with the Act’s intended purposes. The STB should refrain from scrutinizing or attempting to direct private firms’ growth strategies, as such inquiries fall outside the board’s regulatory purview and could lead to unintended consequences, potentially encouraging short-term thinking at the expense of long-term profitability.

Instead of focusing on raw volume growth, the STB should recognize that profitability and returns on invested capital are more meaningful indicators of industry health and economic efficiency. The board should also consider how its own activities might hinder the industry’s ability to achieve profitable growth. Toward this end, the STB should prioritize investigating how existing regulations might interfere with rail carriers’ ability to innovate, adapt to changing market conditions, and maintain long-term profitability. This includes evaluating the impact of regulations on pricing flexibility, operational efficiency, and technological advancement.

We recommend that the STB adopt a more dynamic regulatory approach by establishing feedback mechanisms to gather empirical data on industry performance, as well as using this information to refine its regulatory framework continuously. By shifting its focus from growth plans to regulatory impediments, the STB can better fulfill its role in fostering a competitive, efficient, and financially sustainable rail industry. This approach would not only align more closely with the Staggers Act’s intent, but also better serve the long-term interests of rail carriers, shippers, and the broader economy.

We respectfully submit that the STB’s energies would be better directed towards creating a regulatory environment that allows rail carriers to innovate, compete effectively, and achieve sustainable profitability. This, rather than a focus on predetermined growth metrics, is the surest path to a thriving and efficient rail transportation system in the United States. By embracing this approach, the STB can ensure that it remains true to its statutory mandate, while fostering an environment conducive to the long-term health and success of the rail industry.

 

 

[1] Surface Transportation Board, Notice of Public Hearing, Docket No. EP 775, 89 Fed. Reg. 135 (Jul. 15, 2024).

[2] 49 U.S.C. § 10704(a)(2).

[3] 49 U.S.C. § 10101(1)

[4] Bernard S. Sharfman, Using ‘Enacted Purposes’ to Interpret a Regulatory Statute, SSRN (Jul. 29, 2024), at 4, https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4905077.

[5] Id. at 3.

[6] See, William J. Baumol & Alan S. Blinder, Microeconomics: Principles and Policy (6^th ed., 1994), 245-246.

[7] Susanna Mansikkamäki, Firm Growth and Profitability: The Role of Age and Size in Shifts Between Growth–Profitability Configurations, 19 J. Bus. Venturing Insights e00372 (Jun. 2023).

[8] Id.

[9]  49 U.S.C. § 10101(3)

[10] Sharfman, supra, note 4 at 4.

[11] Jimmy Carter, Staggers Rail Act of 1980 Statement on Signing S. 1946 Into Law, (Oct. 14, 1980), Online by Gerhard Peters and John T. Woolley, THE AMERICAN PRESIDENCY PROJECT, https://www.presidency.ucsb.edu/documents/staggers-rail-act-1980-statement-signing-s-1946-into-law.

[12] 49 U.S.C.  § 10101(2)

[13] Id.

[14]  Sharfman, supra, note 4 at 2-4.

[15] Id. at 7.

[16] CSX Transp. v. U.S., 867 F. 2d 1439 (D.C. Cir. 1988) (emphasis added); see also Village of Ridgefield Park v. NY, Susqu. & Western Rail. Corp., 750 A. 2d 57 (N.J. 2000) (noting that a major purpose of the act is minimize federal regulatory control over rail service).

[17] Sharfman, supra, note 4 at 9.

[18] 49 U.S.C. § 10701(c).

[19] This is not an uncommon concern. For example, in the EU, the General Data Protection Regulation (GDPR) has been fairly controversial, with many observers noting that costs have risen for many firms without significant offsetting benefits for consumers. See, e.g., Andreas Streim & Isabelle Stroot, After 5 Years: GDPR Only Receives the Grade “Sufficient”, Bitkom (Oct. 5, 2023), https://www.bitkom.org/EN/List-and-detailpages/Press/5-years-GDPR-receives-grade-sufficient.

[20] See, e.g., Wolfgang Briglauer, Carlo Cambini, Klaus Gugler, & Volker Stocker, Net Neutrality and High-Speed Broadband Networks: Evidence from OECD Countries, 55 Eur. J. L. & Econ. 533 (2022) (Statistical analysis indicating that net-neutrality regulations slow broadband investment, as measured by the number of fiber connections deployed).

[21] See, e.g., Christopher Mims, How to Move More Goods Through America’s Clogged Infrastructure? Robot Trains, Wall St. J. (Oct. 9, 2021), https://www.wsj.com/articles/how-to-move-more-goods-through-americas-clogged-infrastructure-robot-trains-11633812291.

[22] Analysis of North American Freight Rail Single-Person Crews: Safety and Economics, Oliver Wyman (2015) at 47.

[23] See, STB Adopts Final Rule For Reciprocal Switching, Surface Transportation Board (Apr. 20, 2024), https://www.stb.gov/news-communications/latest-news/pr-24-20.

[24] See, e.g., McCall Macomber, Truck Platooning Moving Freight into the Future, Illinois Center for Transportation (Jul. 28, 2021), https://ict.illinois.edu/news/newsletters/more-newsletters/august-2021/truck-platooning-moving-freight-into-the-future.

[25] Justin (Gus) Hurwitz & Geoffrey A. Manne, Pigou’s Plumber (or Regulation as a Discovery Process), SSRN (Feb. 8, 2024), https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4721112.

[26] Id. at 36-40.

[27] Id.

Introduction

We thank the Treasury of the Australian Government for the opportunity to comment on the exposure draft of the Treasury Laws Amendment Bill 2024: Acquisitions.[1] The International Center for Law & Economics (ICLE) is a nonprofit, nonpartisan global research and policy center founded with the goal of building the intellectual foundations for sensible, economically grounded policy. ICLE promotes the use of law & economics methodologies to inform public-policy debates and has longstanding expertise in the evaluation of competition law and policy. ICLE’s interest is to ensure that competition law remains grounded in clear rules, established precedent, a record of evidence, and sound economic analysis.

While some of the reforms included in the exposure draft are positive (clearer review timelines may bring more legal certainty to merging parties), others threaten to erode the current merger-review regime by unduly focusing on market structure rather than competition; significantly decreasing legal certainty; increasing the number of unnecessary merger notifications; incorporating untested concepts that fall outside the scope of traditional competition; subverting time-honed theories of harm; punishing benign and even procompetitive transactions; and, ultimately, harming consumers. Our overriding concern is that intellectually coherent antitrust policy must focus on safeguarding competition and the interests of consumers, rather than competitors or a predetermined market structure.

Our comments focus primarily on five modifications to the Australian Competition and Consumer Act of 2010[2] (CCA):

1. Setting notification thresholds based on market-concentration metrics;
2. Including a presumption that the acquisition of 20% or more voting power concedes control of the body corporate and is therefore subject to the mandatory notification regime;
3. Modifying the “substantial lessening of competition” (“SLC”) test to include operations that merely create or strengthen market power;
4. Changing the relevant factors to be considered under the SLC test; and
5. Reviewing so-called “serial acquisitions” by considering the combined effect of all acquisitions by the merging parties within a three-year period.

In its ongoing efforts to ensure that antitrust law in general, and merger control in particular, remain tethered to sound principles of economics, law, and due process, ICLE has submitted responses to consultations and published papers, articles, and reports in a number of jurisdictions, including the European Union, the United States, Brazil, the Republic of Korea, the United Kingdom, and India. These and other publications are available on ICLE’s website.[3]  In January 2024, ICLE submitted comments[4] to the Competition Taskforce’s Reform Consultation on Merger Reform (“Consultation”).[5] Our comments on the exposure draft to a significant extent build on the arguments raised therein.

I. Notification Thresholds Based on Market Concentration

A notification threshold based on market-concentration metrics would run counter to one of the stated objectives of the merger reform process, which is to achieve “a merger control system that is faster, stronger, and simpler.”[6]

According to the International Competition Network’s (ICN) Recommended Practices for Merger Notification and Review Procedures, notification thresholds should be “clear and understandable”[7] and “should be based on information that is readily accessible to the parties to the proposed transaction.”[8] A notification threshold based on market concentration does not meet these requirements. Most companies do not readily have information about market shares and concentration levels. Even where they do, such information might be measured in terms of narrowly defined markets that do not readily translate into the antitrust context. Such notification criteria could significantly increase parties’ compliance costs.

The problem is compounded by the discretion afforded in setting notification thresholds. According to the exposure draft, notification thresholds are set either by regulation or the minister (s.51ABG). The latter’s power, however, is entirely discretionary and is checked only (albeit very vaguely) by the minister’s ability to consult the ACCC. The ensuing lack of legal certainty as to why, when, and how the minister might change the merger-notification thresholds is likely to increase the cost of doing business in Australia, as well as inject political influence in what has hitherto remained a largely predictable and technical exercise.

A second problem with concentration-based notification thresholds is that they unduly emphasize market structure. Our concern is that, by instituting market concentration as a notification criterion, the ACCC’s subsequent merger-review process will remain committed to the analysis of market structure as the prime indicator of whether a merger should be allowed (this conclusion also follows from Question 9 of the Consultation, which asks whether Australia’s merger regime should focus more on the overall structure of the market).[9] This would be a mistake. Market structure is, at best, an imperfect proxy for competitive effects and, at worst, a misleading one.

To start, the assumption that “too much” concentration is harmful presumes both that the structure of a market is what determines economic outcomes, and that anyone knows what the “right” amount of concentration is.[10] But as economists have understood since at least the 1970s (despite an extremely vigorous, but ultimately futile, effort to show otherwise), market structure does not determine competition outcomes.[11] This view is well-supported, and held by scholars across the political spectrum.[12]

The absence of correlation between increased concentration and both anticompetitive causes and deleterious economic effects is also demonstrated by a recent, influential empirical paper by Shanat Ganapati. Ganapati finds that the increase in industry concentration in U.S. non-manufacturing sectors between 1972 and 2012 was “related to an offsetting and positive force—these oligopolies are likely due to technical innovation or scale economies. [The] data suggests that national oligopolies are strongly correlated with innovations in productivity.”[13] In the end, Ganapati found, increased concentration resulted from beneficial growth in firm size in productive industries that “expand[s] real output and hold[s] down prices, raising consumer welfare, while maintaining or reducing [these firms’] workforces.”[14] Sam Peltzman’s research on increasing concentration in manufacturing finds that it has, on average, been associated with both increased productivity growth and widening margins of price over input costs. These two effects offset each other, leading to “trivial” net price effects.[15]

Further, the presence of harmful effects in industries with increased concentration cannot be readily extrapolated to other industries. Thus, while some studies have plausibly shown that an increase in concentration in a particular case led to higher prices (which has been found true in only a minority of the relevant literature), assuming the same result from an increase in concentration in other industries or other contexts is simply not justified:

The most plausible competitive or efficiency theory of any particular industry’s structure and business practices is as likely to be idiosyncratic to that industry as the most plausible strategic theory with market power.[16]

As Chad Syverson recently summarized:

Perhaps the deepest conceptual problem with concentration as a measure of market power is that it is an outcome, not an immutable core determinant of how competitive an industry or market is… As a result, concentration is worse than just a noisy barometer of market power. Instead, we cannot even generally know which way the barometer is oriented.[17]

In other words, depending on the nature and dynamics of the market in question, competition may well be protected under conditions that preserve a certain number of competitors in the relevant market. But competition may also be protected under conditions in which a single winner takes all on the merits of their business.[18] It is reductive, and bad policy, to presume that a certain number of competitors is always and everywhere conducive to better economic outcomes, or indicative of anticompetitive harm.

This does not mean that concentration measures have no use in merger enforcement. Instead, it demonstrates that market concentration is often unrelated to antitrust enforcement, because it is driven by factors endogenous to each industry. In revamping its merger-control rules, Australia should be careful not to rely too heavily on structural presumptions based on concentration measures, as these may be poor indicators of those cases where antitrust enforcement would be most beneficial to consumers.

In sum, market structure should remain only a proxy for determining whether a transaction significantly lessens competition. It should not be at the forefront of merger review. And it should certainly not be the determining factor in deciding whether to block a merger. Similarly, it is not an appropriate notification threshold in merger control.

Our view is that there is no need to reinvent the wheel. Turnover has typically been used as a proxy for a merger’s competitive impact because it offers a first indicator of the parties’ relative position on the market. Despite the Consultation’s claim that “mergers of all sizes are potentially capable of raising competition concerns,”[19] where the parties (and especially the target company) have either no or only negligible turnover in Australia, it is highly unlikely that the merger will significantly lessen competition. Again, as recommended by the ICN:

Examples of objectively quantifiable criteria are assets and sales (or turnover). Examples of criteria that are not objectively quantifiable are market share and potential transaction-related effects. Market share-based tests and other criteria that are inherently subjective and fact-intensive may be appropriate for later stages of the merger control process (e.g., determining the scope of information requests or the ultimate legality of the transaction), but such tests are not appropriate for use in making the initial determination as to whether a transaction requires notification.

II. Presumption of ‘Control’ for the Acquisition of 20% or More Voting Power

The exposure draft introduces a rebuttable presumption stating that “if the person’s voting power in the target is 20% or more at a particular time, the person is taken to control the target at that time.” While it is understood that, under certain conditions, minority shareholdings may have anticompetitive effects, this presumption may unjustifiably increase the number of merger notifications. The concept of “control” in competition law, and particularly in merger control, normally involves the ability to exercise decisive influence on an undertaking,[20] which normally should be linked to the power to take decisions unilaterally (more than 50% of voting rights, or right to appoint a majority of the Board of Directors, for instance).

There are, admittedly, jurisdictions that set lower thresholds for triggering the duty to notify a transaction, but this is generally reserved for specific instances where other, special conditions are met that increase the likelihood the acquirer will influence the target’s competitive strategy.  As noted by the OECD:

…when percentage thresholds are established at the lower end of the scale, a combination with additional criteria that indicate a closer relationship between the two parties involved in the transaction might be preferable. The Japanese merger review regime illustrates this point well. In addition to a 50% interest threshold for share acquisitions it has two lower thresholds of a 10% or 20% interest in the target. But in each case a review will be triggered only if additional indicators suggest some influence over the target: in the case of a 20% interest, the holder must be the largest shareholder; in the case of a 10% interest, the holder must be among the three largest holders of voting rights and a number of other criteria must be taken into account that suggest some ability to influence the target (emphasis added).[21]

In most cases, however, a shareholder’s acquisition of 20% of voting rights for a given target should not give rise to competition concerns. This specific change, therefore, would increase the administrative costs of the Australian merger-control regime without adding any significant benefits relevant to protecting competition and, ultimately, consumers.

The presumption that the acquisition of 20% of the voting rights in a company constitutes “control,” however, is arguably a symptom of a broader issue with the exposure draft. Namely, the exposure draft stretches the definition of “acquisition” beyond its natural limits. According to the exposure draft, the mandatory notification and suspension of a transaction is triggered by an acquisition of shares or assets, including property; legal or equitable rights that are not property; goodwill, or an interest in goodwill; as well as interests acquired through partnerships. This list is too broad and could lead to unintended consequences, such as capturing employee contracts. Whether this is the intent of the exposure draft is unclear, but it would be unfortunate if a reform aimed at achieving “a merger control system that is faster, stronger, and simpler” inadvertently stifled an employee’s ability to leave a struggling and failing business.

III. The Modification of the SLC Test

The proposed draft amends Section 4G of the CCA to amplify the meaning of “substantially lessening competition” to include the creation, strengthening, or entrenching of market power. According to the original consultation: “(u)nder the current substantial lessening of competition test, it may be difficult to stop acquisitions that lead to a dominant firm extending their market power into related or adjacent markets.”[22]

The Consultation assumed this to be a problem, particularly in digital markets. Preventing dominant firms from leveraging their market power in one market to restrict competition in an adjacent one is a legitimate concern. We should, however, be clear about what is meant by “materially increase or materially extend a position of substantial market power.”

Merger control should not, as a matter of principle, seek to prevent incumbents from entering adjacent markets. Large firms moving into the core business of competitors from adjacent markets often represents the biggest source of competition for incumbents, as it is often precisely these firms who have the capacity to contest competitors’ dominance in their core businesses effectively. This scenario is prevalent in digital markets, where incumbents must enter multiple adjacent markets, most often by supplying highly differentiated products, complements, or “new combinations” of existing offerings.[23]

Moreover, it is unclear why the SLC test in its current state is insufficient to curb the misuse of market power. The SLC test is a standard used by regulatory authorities to assess the legality of proposed mergers and acquisitions. Simply put, it examines whether a prospective merger is likely to substantially lessen competition in a given market, with the purpose of preventing mergers that increase prices, reduce output, limit consumer choice, or stifle innovation as a result of a decrease in competition.

The SLC test is one of the two major tests deployed by competition authorities to determine whether a merger is anticompetitive, the other being the dominance test. Most merger-control regimes today use the SLC test, and for two good reasons. The first is that, under the dominance test, it is difficult to assess coordinated effects and non-horizontal mergers.[24] The other, mentioned in the Consultation, is that the SLC test allows for more robust effects-based economic analysis.[25]

The SLC test examines likely coordinated and non-coordinated effects in all three types of mergers: horizontal, vertical, and conglomerate. Horizontal mergers may substantially lessen competition by eliminating a significant competitive constraint on one or more firms, or by changing the nature of competition such that firms that had not previously coordinated their behavior will be more likely to do so. Vertical and conglomerate mergers tend to pose less of a risk to competition.[26] Still, there are facts and circumstances under which they can substantially lessen competition by, for example, foreclosing rivals from necessary inputs, supplies, or markets. These outcomes will often be associated with an increase in market power. As the OECD has written:

The focus of the SLC test lies predominantly on the impact of the merger on existing competitive constraints and on measuring market power post-merger.[27]

In other words, the SLC test already accounts for increases in market power that are capable and likely of harming competition. As to whether the “entrenchment” of market power—in line with the 2022 amendments to Canadian competition law—should be added to the SLC test, there is no reason to believe that this is either necessary or appropriate in the Australian context. The 2022 amendments to Canadian competition law mentioned in the Consultation[28] largely align Canada’s merger-control regime with its abuse-of-dominance provision. That provision prohibits anticompetitive activities that damage or eliminate competitors and that “preserve, entrench or enhance their market power.”[29] But in Australia, Section 46 (the equivalent of the Canadian abuse-of-dominance provision) prohibits conduct “that has the purpose, or has or is likely to have the effect, of substantially lessening competition.” The proposed amendment would thus create a discrepancy between merger control and Section 46, where the latter would remain tethered to an SLC test, and the former would shift to a new standard. Additionally, since it remains unclear what the results of Canada’s 2022 merger-control amendments have been or will be, it would be wiser for Australia to adopt a “wait and see” approach before rushing to replicate them.

Lastly, there is the question of defining “materiality” in the context of an increase or entrenchment of market power. Currently, Section 50 of the CCA prohibits mergers that “substantially lessen competition,” with no mention of materiality.[30] The Merger Guidelines do, however, state that:

The term “substantial” has been variously interpreted as meaning real or of substance, not merely discernible but material in a relative sense and meaningful.[31] (emphasis added)

The proposed amendment follows suit, referring to the concepts of “material increase” and “material extension” of market power. What does this mean? How does a “material increase” in market power differ from a non-material one? In its comments on the proposed American Innovation and Choice Online Act (“AICOA”), the American Bar Association’s Antitrust Law Section criticized the bill for using amorphous terms such as “fairness,” “preferencing,” and “materiality,” or the “intrinsic” value of a product. Because these concepts were not defined either in the legislation or in existing case law, the ABA argued that they injected variability and indeterminacy into how the legislation would be administered.[32] The same argument applies here.

As drafted, the new SLC test could be interpreted so broadly that any incremental increase in the market share of a company that already holds some degree of market power would “substantially lessen competition.” This is misguided, and could capture swathes of procompetitive conduct. Indeed, there are many mergers that would—if permitted—benefit consumers, either immediately or in the longer term, but that may have some effect on enhancing market share or market power. Indeed, improving a firm’s products and thereby increasing its sales will often lead to increased market share and market power. This is not a competition problem per se; the problem, rather, is when market power is misused, or is likely to be misused. Whether or not this is effectively the case is what competition authorities strive to ascertain. As drafted, the modified SLC test could substitute that judicious approach for a blunt, de facto prohibition of mergers and acquisitions by firms with market power.

IV. The Relevant Factors Under the SLC Test

According to the exposure draft, the ACCC must consider not just the object of the CCA when reviewing mergers, but also “all relevant matters, including the interests of consumers” (S. 51ABX(2)-(4), emphasis added). While the direct reference to consumers is warranted, as consumer welfare has traditionally been at the forefront of Australian competition law and policy, the exposure draft clarifies that these additional factors (or relevant matters) would include concepts that traditionally have not fallen within the remit of antitrust law, and for which no cogent antitrust theories of harm currently exist, such as “financial and economic power.” Financial and economic power deviate in subtle but important ways from the time-honed concept of “market power,” which assesses a company’s ability to profitably raise prices, reduce output, or deprecate the quality of a product (including through decreased innovation).

Market power dovetails well with the consumer-welfare standard, which forms the basis of the SLC (and thus the cornerstone of any antitrust theory of harm or efficiency defense), and which is properly understood as:

Offer[ing] a tractable test that is broad enough to contemplate a variety of evidence related to consumer welfare but also sufficiently objective and clear to cabin discretion and honor the principle of the rule of law. Perhaps most significantly, it is inherently an economic approach to antitrust that benefits from new economic learning and is capable of evaluating an evolving set of commercial practices and business models.[33]

By contrast, “financial and economic power” could mean nearly anything, and could penalize companies merely for being profitable or having high turnovers—i.e., for competing successfully on the market. As such, they are poor indicators of competitive harm and could, in fact, stifle the very conduct that Australian competition law aims to promote.

V. Serial Acquisitions

The merger-reform paper states that multiple or serial acquisitions will be treated as follows:

…to respond to concerns regarding serial or creeping acquisitions and roll up strategies, all mergers within the previous three years by the acquirer or the target will be aggregated for the purposes of assessing whether a merger meets the notification thresholds, irrespective of whether those mergers were themselves individually notifiable.^[34]

We understand that multiple small acquisitions can, under some circumstances, create a cumulative risk to competition, especially in highly concentrated markets. There remains the question of when this is likely to occur. While serial acquisitions and roll-up strategies merit further study, there is no apparent basis, in either the economic literature or enforcement experience, for any general changes to the procedures or substantive standards by which serial acquisitions are scrutinized. We urge the Treasury to consider whether the proposed change to the notification thresholds is well-tailored to identifying so-called “midnight mergers” or efforts to evade “merger control obligations by transaction structuring, for example, dividing or staggering the merger into several smaller transactions.”^[35]

As the Treasury is well-aware, any substantial lowering of notification thresholds will impose costs on both merging firms and the enforcers called on to scrutinize noticed acquisitions.^[36] Moreover, bundling all mergers “by the acquirer or the target” across any moving three-year window will, in effect, greatly lower the threshold for those firms engaged in multiple acquisitions over time. We also question whether Treasury is aware of any theoretical or empirical basis for stipulating a three-year window. While any single three-year period may be clear enough, a moving window may create unnecessary uncertainty for consummated transactions well after operations or assets have been knit together, such that there is no efficient way to “unscramble the eggs.”

Note, first, that serial acquisitions may range across product, service, or geographic markets. While there are circumstances under which vertical or conglomerate acquisitions may prove anticompetitive, such transactions typically prove procompetitive or benign.^[37] Cumulative anticompetitive effects across geographic markets may be rarer still.

More broadly, many of the activities described as “serial acquisitions” are indistinguishable from normal patterns of business growth and consolidation that occur in maturing industries. As a general matter, it is not clear why a company growing through multiple small acquisitions should be viewed differently than one growing “organically” or through fewer, larger acquisitions. This raises important questions about the underlying theory of harm. If the concern is market concentration, this can occur through various means, not just serial acquisitions. If the concern is about the specific process of multiple small acquisitions, it is unclear why this would be inherently more problematic than other forms of growth.

Recent research by Cohn, Hotchkiss, and Towery sheds light on the motivations behind roll-up strategies in private-equity buyouts of private firms.[38] Their study suggests that these strategies are often driven by two primary motives: unlocking growth potential in capital-constrained firms and improving operational performance in underperforming firms. They find that acquired firms often experience significant increases in sales growth and moderate improvements in profitability post-acquisition. Such findings support the view that these strategies can create value through both growth and operational improvements. They also suggest that properly executed roll-up strategies can serve legitimate business purposes beyond mere market consolidation.

Given the legitimate business reasons for acquisitions (serial or not), we are aware of no theoretical or empirical grounds on which to suppose that multiple acquisitions are typically anticompetitive. At the same time, there is no reason to suppose that the organic growth of a firm precludes anticompetitive conduct. The competitive effects of growth—whether through acquisition or internal expansion—depend on various factors, including market structure, barriers to entry, and the specific capabilities and assets being acquired or developed. For example, in some cases, serial acquisitions might allow a firm to quickly assemble complementary assets and capabilities, leading to increased innovation and more robust competition. In other instances, organic growth might allow a firm to build market power in ways that are difficult for competitors to challenge.

To be clear, we do not suggest that there are no circumstances under which serial acquisitions raise competitive concerns. Rather, we believe that considerable work remains to be done if competition enforcers seek to tailor notice requirements in a manner that is efficient for both commercial development and enforcement alike. As described in the Merger Reform Paper, the serial-notice requirement appears to us conspicuously overbroad, and decidedly at odds with Treasury’s stated goal of a simpler, clearer, and more “targeted” merger-review process.

Finally, while enforcers would do well to consider all relevant market factors—and likely pro- and anticompetitive effects—in scrutinizing those mergers that raise serious competitive concerns, we wonder whether the addition of the supplementary principles “replacing the ‘merger factors’ currently in section 50(3) of the CCA”^[39] will further compound the problems of scope raised above. That is, will the open-ended list of potential considerations, and their extension to potential competition, together with the bundling of multiple acquisitions across markets, further undermine the goals of simpler, clearer, and more “targeted” and predictable merger review? One thing is clear: by implying that no sale is ever final, the new rules will create additional (and in our view, unnecessary) uncertainty for businesses, shareholders, and employees.

[1] Reforming Mergers and Acquisitions – Exposure Draft,  Australian Government, The Treasury (24 Jul. 2024),  https://treasury.gov.au/consultation/c2024-554547.

[2] Competition and Consumer Act 2010, No. 51, 1974 (Aus.).

[3] International Center for Law & Economics, https://laweconcenter.org.

[4] Dirk Auer, Geoffrey A. Manne, & Lazar Radic, ICLE Response to the Australian Competition Taskforce’s Merger Reform Consultation, Int’l. Ctr. Law Econ. (19 Jan. 2024), https://laweconcenter.org/resources/icle-response-to-the-australian-competition-taskforces-merger-reform-consultation.

[5] Merger Reform Consultation Paper, Australian Government, The Treasury (20 Nov. 2023), https://treasury.gov.au/consultation/c2023-463361.

[6] Exposure Draft, supra note 1, at 1.

[7] Recommended Practices for Merger Notification & Review Procedures, International Competition Network 5 (2002-2018), https://www.internationalcompetitionnetwork.org/portfolio/merger-np-recommended-practices.

[8] Id., at 7.

[9] Consultation, supra note 4.

[10] The following section is adopted from Geoffrey A. Manne, et al., Comments of the International Center for Law & Economics on the FTC & DOJ Draft Merger Guidelines, Int’l. Ctr. Law Econ. 38 (18 Sep. 2023) https://laweconcenter.org/wp-content/uploads/2023/09/ICLE-Draft-Merger-Guidelines-Comments-1.pdf.

[11] See, e.g., Harold Demsetz, Industry Structure, Market Rivalry, and Public Policy, 16(1) J. Law Econ. 1-9 (Apr. 1973).

[12] Nathan Miller, et al., On the Misuse of Regressions of Price on the HHI in Merger Review, 10(2) J. Antitrust Enforc. 248-259 (28 May 2022); See, e.g., Steven Berry, Martin Gaynor, & Fiona Scott Morton, Do Increasing Markups Matter? Lessons from Empirical Industrial Organization, 33(3) J. Econ. Perspect. 44-68, 48 (2019).

[13] Shanat Ganapati, Growing Oligopolies, Prices, Output, and Productivity, 13(3) Am. Econ. J. Microecon. 309-327, 324 (Aug. 2021).

[14] Id., at 309.

[15] Sam Peltzman, Productivity, Prices and Productivity in Manufacturing: a Demsetzian Perspective, Coase-Sandor Working Paper Series in Law and Economics 917, (19 Jul. 2021).

[16] Timothy F. Bresnahan, Empirical Studies of Industries with Market Power, in Handbook of Industrial Organization, Richard Schmalensee & Robert Willig (eds.), 1011, 1053-54 (1989).

[17] Chad Syverson, Macroeconomics and Market Power: Context, Implications, and Open Questions, 33(3) J. Econ. Perspect. 23-43, 26 (2019).

[18] Nicolas Petit & Lazar Radic, The Necessity of the Consumer Welfare Standard in Antitrust Analysis, ProMarket (18 Dec. 2023), https://www.promarket.org/2023/12/18/the-necessity-of-a-consumer-welfare-standard-in-antitrust-analysis.

[19] Consultation, supra note 4,  at 24.

[20] See, e.g., Council Regulation (EC) No 139/2004 On the Control of Concentrations Between Undertakings (2004) Official Journal L24 1-22, Article 3 (2) (EU).

[21] Definition of Transaction for the Purpose of Merger Control Review, OECD 15 (24 Jan. 2014), https://one.oecd.org/document/DAF/COMP(2013)25/en/pdf.

[22] Consultation, supra note 4, at 19.

[23] Nicolas Petit, Big Tech and the Digital Economy: The Moligopoly Scenario (2020); see also Walid Chaiehoudj, On “Big Tech and the Digital Economy”: Interview with Professor Nicolas Petit, Competition Forum (11 Jan. 2021), https://competition-forum.com/on-big-tech-and-the-digital-economy-interview-with-professor-nicolas-petit.

[24] Standard for Merger Review, OECD 6 (11 May 2010), https://www.oecd.org/daf/competition/45247537.pdf.

[25] Id.; see also Consultation, supra note 4, at 31 (indicating that “[SLC test] would enable mergers to be assessed on competition criteria but not prescriptively identify which competition criteria should be taken into account. It may permit more flexible application of the law and a greater degree of economic analysis in merger decision-making” (emphasis added).)

[26] See, e.g., Guidelines on the Assessment of Non-Horizontal Mergers Under the Council Regulation on the Control of Concentrations Between Undertakings,  (2008/C 265/07), paras 11-13 (EU).

[27] OECD, supra note 24, at 16; see also Guidelines on the Assessment of Horizontal Mergers Under the Council Regulation on the Control of Concentrations between Undertakings (2004/C 31/03)(2004) Official Journal C 265, 6-25 (EU).

[28] Consultation, supra note 4, at 30-31.

[29] Competition Act, R.S.C., 1985, c. C-34, at ss. 78 and 79 (Can.).

[30] Section 44G CCA, however, does mention a “material increase in competition.” (emphasis added).

[31] ACCC, Merger Guidelines (2008) (Aus.); see also Australia, Senate 1992, Debates, vol. S157, p. 4776, as cited in the Merger Guidelines (2008).

[32] Geoffrey A. Manne & Lazar Radic, The ABA’s Antitrust Law Section Sounds the Alarm on Klobuchar-Grassley, Truth on the Market (12 May 2022), https://truthonthemarket.com/2022/05/12/the-abas-antitrust-law-section-sounds-the-alarm-on-klobuchar-grassley.

[33] Elyse Dorsey, et al., Consumer Welfare & The Rule of Law: The Case Against the New Populist Antitrust Movement, 47 Pepperdine Law Rev. 861 (1 Jun. 2020).

^[34] Merger Reform: A Faster, Stronger, and Simpler System for a More Competitive Economy, Australian Government, The Treasury 5 (10 Apr. 2024), https://treasury.gov.au/sites/default/files/2024-05/p2024-518262-merger-reforms-paper.pdf (“Merger Reform Paper”).

^[35] Id.

^[36] See, generally, Brian Albrecht, Dirk Auer, Daniel J. Gilman, Gus Hurwitz, & Geoffrey A. Manne, Comments of the International Center for Law & Economics on Proposed Changes to the Premerger Notification Rules, Int’l Ctr Law Econ. (27 Sept. 2023), https://laweconcenter.org/resources/comments-of-the-international-center-for-law-economics-on-proposed-changes-to-the-premerger-notification-rules.

^[37] See, e.g., Francine Lafontaine & Margaret Slade, Exclusive Contracts and Vertical Restraints: Empirical Evidence and Public Policy, 10 Handbook of Antitrust Economics 391, 408-09 (2008); see also Francine Lafontaine & Margaret E. Slade, Transaction Cost Economics and Vertical Market Restrictions—Evidence, 55(3) Antitrust Bull. 587 (2010); James Cooper, Luke Froeb, Daniel O’Brien, & Michael Vita, Vertical Antitrust Policy as a Problem of Inference, 23 Int’l. J. Ind. Organ. 639-664 (2005); David Reiffen & Michael Vita, Comment: Is There New Thinking on Vertical Mergers?, 63 Antitrust L.J. 917, 920 (1995); Henry Ogden Armour & David Teece, Vertical Integration and Technological Innovation, 62 Rev. Econ. & Stat. 470, 470 (1980); Dennis W. Carlton, Transaction Costs and Competition Policy, 73 Int’l J. Indus. Org. 1, 7 (2019); regarding conglomerate mergers, see, e.g., Conglomerate Effects of Mergers – Note by the United States, OECD (10 Jun. 2020), at 2.

[38] See Jonathan B. Cohn, Edith Hotchkiss, & Erin Towery, Sources of Value Creation in Private Equity Buyouts of Private Firms, 26 Rev. of Fin. 257 (2022).

^[39] Merger Reform Paper, supra note 34, at 9-10.

PORTLAND, Ore. (Aug. 5, 2024) – The International Center for Law & Economics (ICLE) offers the following statement on today’s order from the U.S. District Court for the District of Columbia in United States v. Google and State of Colorado et al. v. Google finding that Google’s distribution agreements with browser developers and Android device manufacturers and carriers are anticompetitive in the market for general search text advertising and violate Section 2 of the Sherman Act.

The following quote can be attributed to ICLE President Geoffrey A. Manne:

The court’s order, which relies heavily on contested theories from the field of behavioral economics about the supposed power of defaults, fails to demonstrate how the contractual agreements at-issue harm consumers or competition. Moreover, the court overlooks the broader competitive landscape in search and the vigorous competition in which Google has been engaged to become the default search engine.

The fact that Google search has an 80% market share even on Windows devices, where Edge is the default browser and Bing is the default search engine, demonstrates that consumers go out of their way to use Google because they believe it is the best option. A default placement is worth very little if your product isn’t any good. By the same token, Google hasn’t been ousted as the default anywhere, because it has a superior product. The opinion offers no evidence to suggest that Bing would have become a viable competitor under any other set of facts. And that is fatal to the claims in this case, for which the plaintiffs, not Google, bear the burden of proof.

The court’s decision turns on a finding that Google won its place in the market through anticompetitive exclusive agreements and not “competition on the merits.” But the court struggles to maintain this conclusion, and ultimately its decision falters on that basis. Did Google win its place as a default search provider for Apple, for example, because it is the best search engine or by paying Apple for unwarranted exclusivity? The court’s conclusion that it was the latter is difficult to sustain—as even the court itself recognizes.

For more on the topic, see ICLE’s 2020 explainer; “The FTC Did Not ‘Fumble the Future’ in Its Google Search Investigation” by Geoff Manne and Dirk Auer; and “The Missing Element in the Google Case” by Greg Werden.

To schedule an interview with Geoff about the topic, contact ICLE Media and Communications Manager Elizabeth Lincicome at 919-744-8087 or [email protected].

Executive Summary

Smoking kills hundreds of Americans every day. Transitioning to vaping is one of best methods to stop smoking. While the risk from vaping is not zero, it is negligible for those who use well-made products. The U.K. government sends vaping kits to smokers to encourage them to quit, and permits the sale of vapes from vending machines. By contrast, U.S. public-health policy toward vaping is equivocal, and bogged down by concerns that children might start the habit. U.S. regulators have only approved a few vaping products, and they largely are not the ones that consumers most want. Nonetheless, the market does supply the products that adult consumers demand, regardless of their legal status. This paper describes a survey of a small subset of the legal market: 14 gas stations in the Philadelphia suburbs that sell large quantities of unapproved vaping products every day. It then reviews recent court decisions regarding the U.S. Food and Drug Administration’s (FDA) actions relevant to the sector, and suggests that U.S. policymakers should replace the FDA’s current arbitrary and capricious approach to vaping regulation with objective and reasonable quality standards.

I. Background

Burning tobacco releases hundreds of substances that contribute to more than 480,000 premature deaths annually in the United States from a range of causes, including cancer and cardiovascular disease.[1] While nicotine is the primary substance in cigarettes that creates physiological dependence, it is not dangerous in normal doses.[2] Nicotine patches and gum were developed to provide smokers with nicotine without the dangers of tobacco combustion. But while helping some smokers to quit, patches fail to meet the ritual, social habit, and other needs of most smokers.

Independent researchers and entrepreneurs understood this and developed alternatives to combustible products that better met smokers’ needs in the form of nicotine vaping products. Independent companies like NJoy and Pax Labs were among those that drove the early U.S. vaping market, with products such as the latter’s Juul pod-based e-cigarette, which used patented nicotine salts. Some users have preferred traditional tobacco flavors, but most report they enjoy alternate flavors, such vanilla, blueberry, and mango. Some vapes even contain no nicotine at all.[3]

Dating back to the 1980s, the major tobacco companies brought noncombustible “heat-not-burn” tobacco products to the market in markets like the U.K. and Japan.[4] Some also invested their capital to buy the more successful vaping-product companies, including Altria’s notable $13 billion partial acquisition of Juul in 2018, after it was spun off from Pax Labs.[5]

Vaping and heat-not-burn tobacco products have been shown to satisfy some of the ritual, sensory, pharmacokinetic, and social aspects that many smokers enjoy, while lowering the risks for the user. Vaping products are not riskless,[6] but they are substantially safer than smoking; the U.K. government suggests they are at least 95% safer, based on the available evidence.[7]

U.K. health authorities are sufficiently convinced of the relative safety of vaping that they encourage it as an alternative to patches and gum where those have not proved effective. Hospitals hand out free vaping starter kits to recruit smokers into a cessation program.[8] And in April 2023, the U.K. government decided to send vaping kits to roughly 1 million people—20% of the nation’s identified smokers.[9] U.K. health officials’ aim is for the U.K. to be a smoke-free society (that is, one with less than 5% regular smokers) by 2030. The U.K.’s support of vaping as an alternative to smoking is such that it even allows the products to be sold through vending machines.

A recent study by the MUSC Hollings Cancer Center found that smokers given vaping tools will reduce the amount they smoke, even if they are neither encouraged to use the tools nor even shown how to use them.[10] With 638 smoker participants, this was the largest study of its type concluded in the United States. Of the 638 smokers, 427 actively participated (211 in the control group did not receive vaping products). While some participants stated that they wanted to quit smoking and others stated that they did not, the reduction and even cessation in smoking was universal, including among those who had no intention of quitting.

This study adds to the weight of evidence that vaping options can reduce smoking.[11] It also answers a concern raised by vaping critics that vaping only works effectively in smoking cessation in an artificial study setting with smokers who say they want to quit.

Yet U.S. government policy remains equivocal about the relative benefits of vaping. The U.S. Food and Drug Administration (FDA) has approved a few vaping products,[12] but FDA communications (along with communications from other relevant federal agencies) focus more on the potential risks of vaping, especially to youth, rather than its role in smoking cessation in adults.[13]

As leading tobacco-regulation analyst Clive Bates has put it, the FDA has created “almost insurmountable regulatory hurdles and barriers to entry for vapes.”[14] The strict legal sanctions the FDA levied in 2019 against U.S. vape pioneer and previous market leader Juul,[15] preventing it from selling flavored products, have not been followed up with further significant action against manufacturers or retailers selling similar products from companies based outside the United States. The FDA has instead effectively left a legal void by denying approval to thousands of products, but failing to prevent those and similar products from reaching the market.[16]

There was also a period during which products manufactured using synthetic nicotine seemed to be able to circumvent the FDA’s rules. This ended in April 2022, when Congress passed legislation effectively deeming synthetic nicotine to be a “tobacco product” and thereby authorizing the FDA to regulate vapes that contained it.[17] Most producers, retailers, and consumers are, unsurprisingly, unsure of what is legal.

In a recent analysis of discarded vaping packs conducted across several states by the market-research group WSPM,[18] an astonishing more than 97% were not legal in the United States. This suggests that vast numbers of imported vaping products are brought into the country illegally. But are these products available only on the black market, or are they available in major retail outlets?

This issue brief aims to look closely at one subset of retailers: gas stations in several suburbs of Philadelphia. The gas stations are legal businesses that sell fuel and other quick-serve products, including drinks, food, cigarettes, and vaping products. Additionally, a more anecdotal analysis of three illicit markets, which primarily sell illegal drugs, is undertaken to observe what vaping products are available.

II. The Market for Vaping Products

When the field research for this project was undertaken, from December 2023 through June 2024, only 23 vaping products were approved for sale by the FDA. All of the approved products were tobacco-flavored.[19] By contrast, in the WSPM discarded-pack analysis, more than 98% of the recovered products were non-tobacco flavors. In other words, consumer demand appears to strongly favor non-tobacco-flavored vaping products, even though the FDA has not approved any such products for sale in the United States.

With such obvious demand for non-approved products, one would obviously expect a black market to emerge. Many food stores, gas stations, tobacco outlets, and other “convenience” locations, which have sold cigarettes for decades, now sell vaping products. While they do sell the FDA-approved products, I decided to investigate whether, and how many non-approved products are sold through these legal outlets.

Through personal contacts in the suburbs of Northeast Philadelphia, I examined the vaping products and sales over the previous year at 14 gas stations, some within city limits and some outside. Two of these gas stations, owned by the father of a friend, provided me with more than a decade of sales information, dating back to before these locations sold vaping products at all. This sample of convenience was not randomly created and is not intended to be representative of either this region of the country, nor even of Southeast Pennsylvania, specifically. It may contain unknown biases and may not reflect other locations within this geographic zone.

Over the six-month period that the survey was conducted, a total of 459 brand-name products were found across the 14 gas stations, with one station selling 169 individual brands. Two brands (Elf Bar and Lost Mary) dominated half the market, each with dozens of branded products. More than 30 different Elf Bar and Lost Mary flavors were identified, including myriad fruit flavors (blueberry, kiwi, watermelon, cherry, etc.), as well as flavors like bubble gum.

Figure 1 demonstrates the most popular brand groups. Note that Vuse, at 2% of the market, is the only brand with products (tobacco- or menthol-flavored) approved by FDA. Juul, also at 2%, has not been approved by the FDA, but its status remains in a kind of legal limbo.[20] Of the 459 identified branded products, only three were approved by the FDA. Therefore, more than 99% of the market is either illegal, or exists in a legal grey area.

FIGURE 1: Market Share of Most Popular Vapor Brands (%)

A. Are These Findings Unusual?

Over the past two years, on various visits to Washington, D.C., and New York City, I have visited gas stations up and down the I-95 corridor. These locations appear to sell similar products to those in the Philadelphia suburbs, often with different brands from within the same brand group. The most dominant product was Geek Bar in Maryland and D.C., while it was Elf Bar in Pennsylvania and Lost Mary in the New York metro area. These brands were also identified in the discarded-pack analysis, which noted that Geek Bar was the most discarded product group in Washington, D.C. All of these products are made by the same manufacturer: the China-based Shenzhen iMiracle, whose products are not approved for sale in the United States.

B. Market Value

Figure 2 demonstrates the growth in average monthly per-store revenues from vapor products. (For 2023, this is across all 14 gas stations; for 2016-2022, revenue was averaged across just the two stations for which data was available.)

The vast majority of this revenue comes from products that are not approved by the FDA. The much smaller revenue streams from approved vapes did also grow over time, but not at the same rates as for the non-approved products.

As demonstrated in Figure 2, average monthly sales have grown from a few hundred dollars to more than $10,000. The number of products sold has also grown from less than a dozen to hundreds. Every outlet in the survey sold more than 100 vape products. Non-approved products represented more than 99% of sales volume and 98% of revenue.

FIGURE 2: Average Monthly Vaping Revenue ($)

The two stations within Philadelphia city limits had slightly lower sales, presumably because their tax rates are slightly higher, although sample sizes were not sufficiently large to rule out other causes (e.g., lower average income in those locations).

III. Why Legal Businesses Sell Non-Approved Products

Five store managers/owners agreed to be interviewed on condition of anonymity. Of the five, none were entirely sure which, if any, products they sold were illegal. They reported that they assumed their suppliers operated legally and that they sell a wide variety of products. In short, they follow what customers buy, and rely on their suppliers to undertake legal due diligence.

One store owner commented that they might advertise vaping products more by displaying them more prominently in future, but were aware of the “legal limbo” of some products and that they “still made far more money from cigarette sales.” Cigarette (and cigar) sales were, on average, about 14 times larger than vaping-product sales in 2023 and 2024. One proprietor noted, however, that cigarette sales had been roughly 20 times more than vape sales only a few years ago, suggesting that vape sales are growing faster than cigarette sales.

Another owner reported that, while he obviously did not want to “break the law,” he has found that demand for vaping products is significant; that there is no “strong guidance on illegality”; and that there is “absolutely no enforcement.” Therefore, he said: “I’ll keep selling them until I’m told not to and that is enforced.”

A. How the Vaping Market Got to This Point

While distributors and retailers both are bound by the law, a significant amount of the blame for so many non-approved products being sold in the putatively legal market can be attributed to U.S. health policy and product regulation. While the FDA has denied application after application for useful vaping products, it takes no action against the sale of non-approved products.

One obvious concern is that the non-approved products currently available on the market are not required to meet any quality or safety standards. In denying approval to the vast majority of vapor products, the FDA has not only limited consumer choice, but it has induced many of those consumers to purchase products that may be of lower quality.

Vape manufacturers and importers want repeat business. They therefore have incentive not to harm users. But the FDA’s nominal claim of regulatory oversight for vaping products may have the effect of crowding out other means by which consumers might obtain information about the quality of vaping products. These could include, for example, industry standards or online vendors applying their own quality controls—both of which existed prior to the FDA deeming vapes to be tobacco products.[21]

Meanwhile, in evading its responsibility as an enforcer, the FDA has effectively misled consumers. Of course, the agency’s record in ensuring quality manufacturing processes contains other blemishes, including in its oversight of generic pharmaceuticals.[22] The potential for unnecessary harms from unauthorized vaping products (e.g., due to adulteration) is not negligible. While it is possible that these problems are magnified by the fact that most vapes are made in China, most authorized vapes are also manufactured in China. The more fundamental problem is that unauthorized vapes, by their nature, evade any regulatory approval. After all, domestically manufactured illegal cannabis vapes adulterated with e-acetate caused the death of 68 Americans in 2019 and 2020.[23]

On a more positive note, many of the larger manufacturers, including Shenzhen iMiracle, produce versions of their products that have been approved for sale by the U.K. government and therefore have cleared U.K. quality rules.[24] Indeed, Lost Mary appears to be the most ubiquitous brand in London, as it is on much of the Eastern seaboard of the United States. But the versions of these products marketed in the United States have higher levels of nicotine and may have been produced using different standards. Moreover, many other products sold in United States do not have equivalents that are currently approved in the U.K., or possibly anywhere.[25]

B. Attempts to Comply with FDA Guidance

While manufacturers have tried to comply with FDA rules, in the absence of clear guidance and objective standards, they have been faced with a nearly impossible task. For its part, the FDA has shown little interest in approving its backlog of product applications, leaving manufacturers to resort to litigation. In a recent decision, the 5^th U.S. Circuit Court of Appeals found that the FDA’s decisionmaking processes for approving vaping products were “arbitrary and capricious.”[26] The agency has also repeatedly changed its mind regarding what it would require from manufacturers. As the court found:

FDA refused to review petitioners’ marketing restrictions, which it had repeatedly stated were key to discouraging youthful use of the products and were thus critical components of [applications]. FDA repeatedly counselled applicants that long term studies were likely unnecessary and it said nothing about comparative efficacy studies—until the PMTA deadline was long gone; and then it refused petitioners the opportunity to conduct such studies. Finally, FDA’s defense against petitioners on the merits of their applications is loaded with post hoc rationalizations. Any of these errors is a “fatal flaw.” Taken together, they are mortal wounds.[27]

The FDA has appealed the 5^th Circuit’s ruling to the U.S. Supreme Court, which granted certiorari in July 2024.[28] This has left the legal limbo in place. David Spross of the National Association of Tobacco Outlets testified at a June 2024 hearing of the U.S. Senate Judiciary Committee about his members’ desire to sell legal vape products, and encouraged the FDA to authorize more, but most other witnesses and committee members were much more focused on the potential risks of vaping to youth.[29]

As Clive Bates has explained:

The technical, scientific, creative, and legal resources of American vape businesses are mired in endless, exhausting efforts to prove the acceptability of products that are already 8-10 years old to the satisfaction of the unyielding, never-satisfied FDA. In contrast, the Shenzhen innovation system is tirelessly focused on the consumer – what the consumer wants today and what they need to get around pointless regulation….FDA is nurturing demand for illicitly supplied Chinese products.[30]

C. Illicit Sellers of Illicit Products

Through previous work on opioids,[31] I have come to know several drug dealers whose primary market is selling products such as oxycodone, fentanyl and, occasionally, heroin. None sell vaping products to any scale, primarily because they are so easy to buy through ordinary retail outlets. When these drug dealers have sold vapes, it has only been flavored disposable vapes—products similar, if not identical, to those sold in convenience stores. The consensus they reported to me is that if there were a police crackdown on legal sales, they might sell more illicit vapes, but for now it remains a tiny market for them.

This is quite a telling finding. In a market for illegal products, illicit sellers with active buyers barely engage in the market because legal sellers dominate sales of non-approved products.

IV. Conclusion

It is hard to think of a public-health policy more poorly managed in the United States than vaping. Given that the alternative to vaping is often a far deadlier smoking habit, public policy should encourage vaping products manufactured to a high standard, and marketed and sold exclusively to adults. That means establishing and enforcing clear, objective, and reasonable quality standards that apply to all vapes on the market, as the U.K. has done,[32] not imposing arbitrary and capricious requirements on vape manufacturers and importers, as the FDA has done.

Thousands of U.S. outlets currently sell vaping products of unknown quality. While the FDA has threatened action against retailers with warning letters,[33] there has been little follow-through on enforcement. This failure undermines not only the rule of law, but also public health.

[1] See, Adult Data on Smoking, Ctr. for Disease Control and Prevention, https://www.cdc.gov/tobacco/data_statistics/fact_sheets/adult_data/cig_smoking/index.htm#references (last visited May 5, 2023)

[2] See Nicotine Dosage, Drugs.com, https://www.drugs.com/dosage/nicotine.html (last visited Aug. 16, 2023).

[3] Average Revenue Per Unit (ARPU) in the E-cigarettes Segment of the Tobacco Products Market Worldwide from 2019 to 2029, statista, https://www.statista.com/forecasts/1438523/average-revenue-per-unit-arpu-e-cigarettes-tobacco-products-market-worldwide (last visited Aug. 30, 2024).

[4] See Tax Treatment of Heated Tobacco Products, HM Treas., https://www.gov.uk/guidance/e-cigarettes-regulations-for-consumer-products (last updated Mar. 13, 2018).

[5] David Goldman, The Biggest American Cigarette Company Buys a $13 billion Stake in the Biggest E-cigarette Startup, CNN Bus. (Dec. 20, 2018), https://www.cnn.com/2018/12/19/business/altria-juul/index.html.

[6] Akihiro Kishimoto et al., Forecasting Vaping Health Risks Through Neural Network Model Prediction of Flavour Pyrolysis Reactions, 14 Sci. Rep. 9591, 10 (2024), https://www.nature.com/articles/s41598-024-59619-x.

[7] See Ann McNeill et al., Vaping in England: Evid. Update Feb. 2021, Public Health England (2021), https://www.gov.uk/government/publications/vaping-in-england-evidence-update-february-2021.

[8] See, Starter Vape Packs to be Handed Out in Hospitals, U.E. Anglia (Apr. 29, 2021), https://www.uea.ac.uk/about/news/article/starter-vape-packs-to-be-handed-out-in-hospitals.

[9] Neil O’Brien, Smokers Urged to Swap Cigarettes for Vapes in World First Scheme, Dep’t Health & Soc. Care (Apr. 11, 2023), https://www.gov.uk/government/news/smokers-urged-to-swap-cigarettes-for-vapes-in-world-first-scheme.

[10] See Leslie Cantu, Largest U.S. Study of E-cigarettes Shows Their Value as Smoking Cessation Aid, M.U.S.C. Health, https://muschealth.org/health-professionals/progressnotes/2023/fall/e-cigarettes-study (last visited Aug. 22, 2024).

[11] See Jeremy Y. Levett et al., Efficacy and Safety of E-cigarette Use for Smoking Cessation: A Systematic Review and Meta-analysis of Randomized Controlled Trails, 136 Am. J. Med. 802, 802-805 (2024), https://www.amjmed.com/article/S0002-9343%2823%2900295-4/fulltext#:~:text=io%2F26fkq).-,Results,CI%2C%201.29%2D2.44.

[12] See, FDA Permits Marketing of E-Cigarette Products, Marking the First Authorization of Its Kind by the Agency, U.S. Food & Drug Admin. (Oct. 12, 2021), https://www.fda.gov/news-events/press-announcements/fda-permits-marketing-e-cigarette-products-marking-first-authorization-its-kind-agency.

[13] Clive Bates, Fixing U.S. Vape Regulation- Twelve Proposals, The Counterfactual (Jun. 12, 2024), https://clivebates.com/fixing-u-s-vape-regulation-twelve-proposals.

[14] Id. at 11.

[15]  James Ducharme, How Juul Got Vaporized, Time (May 17, 2021), https://time.com/6048234/juul-downfall.

[16] Matthew Perrone, Thousands of Unauthorized Vapes Are Pouring Into the US Despite the FDA Crackdown On Fruity Flavors, Associated Press (Jun. 26, 2023), https://apnews.com/article/fda-vapes-vaping-elf-bar-juul-80b2680a874d89b8d651c5e909e39e8f.

[17]  See Requirements for Products Made With Non-tobacco Nicotine Take Effect Apr. 4, U.S. Food & Drug Admin. (Apr. 13, 2022), https://www.fda.gov/tobacco-products/ctp-newsroom/requirements-products-made-non-tobacco-nicotine-take-effect-april-14.

[18] See Cami Mondeaux, Majority of Disposed Vapes in D.C. Exported from China Despite Bad: Study, Wash. Exam’r (May 29, 2024), https://www.washingtonexaminer.com/policy/healthcare/3020885/majority-disposed-vapes-dc-exported-china-study.

[19] E-Cigarettes, Vapes, and other Electronic Nicotine Delivery Systems (ENDS), U.S. Food & Drug Admin., https://www.fda.gov/tobacco-products/products-ingredients-components/e-cigarettes-vapes-and-other-electronic-nicotine-delivery-systems-ends (last visited Aug. 30, 2024).

[20] Update on FDA’s Scientific Review of Juul Product Applications, U.S. Food & Drug Admin. (Jun. 6, 2024), https://www.fda.gov/tobacco-products/ctp-newsroom/update-fdas-scientific-review-juul-product-applications.

 

[21] Guy Bentley & Julian Morris, The FDA Has Decimated the E-cigarette Market, Reason Foundation (Sep. 22, 2021), https://reason.org/commentary/the-fda-has-decimated-the-e-cigarette-market.

[22] David Dobbs, A New Book Argues That Generic Drugs Are Poisoning Us, N.Y. Times (May 13, 2019), https://www.nytimes.com/2019/05/13/books/review/bottle-of-lies-katherine-eban.html.

[23] See, Severe Lung Disease Associated With Using E-cigarette Products, Ctr. for Disease Control & Prevention, https://archive.cdc.gov/www_cdc_gov/tobacco/basic_information/e-cigarettes/severe-lung-disease.html (last updated Feb. 18, 2020).

[24] See E-cigarettes: Regulations for Consumer Products, Med. & Healthcare Products Regul. Agency (Feb. 29, 2016), https://www.gov.uk/guidance/e-cigarettes-regulations-for-consumer-products (last updated Aug. 16, 2024).

[25] Russ Ware, Vaping and the FDA: An Updated Timeline, Versed Vaper (Jul. 26, 2024), https://versedvaper.com/vaping-and-the-fda-an-updated-timeline.

[26] See Jonathan H. Adler, En Banc Fifth Circuit Rejects FDA’s Vaping Regulation “Surprise Switcheroo”, Reason (Mar. 1, 2024), https://reason.com/volokh/2024/01/03/en-banc-fifth-circuit-rejects-fdas-vaping-regulation-surprise-switcheroo.

[27] Wages & White Lion Invs. v. Food & Drug Admin., 90 F.4th 357, 362 (5th Cir. 2024).

[28] Jessie Hellmann, Fight Over Flavored Vapes Lands at Supreme Court, Roll Call (Jul. 2, 2024), https://rollcall.com/2024/07/02/fight-over-flavored-vapes-lands-at-supreme-court.

[29] See, Combatting the Youth Vaping Epidemic by Enhancing Enforcement Against Illegal E-cigarettes, U.S. S. Comm on the Judiciary, 118^th Cong. (2024), https://www.judiciary.senate.gov/committee-activity/hearings/combatting-the-youth-vaping-epidemic-by-enhancing-enforcement-against-illegal-e-cigarettes.

[30] Id. at 11.

[31]  See Roger Bate, A Field Study of the Opioid Market: Authenticity and Price from Pharmacy to Street, AEI, available at https://www.aei.org/wp-content/uploads/2018/10/Bate-final-wp.pdf (last visited Aug. 23, 2024).

[32] Id. at 19.

[33] David Spross, FDA Steps Up Enforcement Actions Against Tobacco Manufacturers, Wholesalers and Retailers, CSP (Feb. 7, 2024), https://www.cspdailynews.com/tobacco/fda-steps-enforcement-actions-against-tobacco-manufacturers-wholesalers-retailers.

I. Introduction

Atlanta’s Mercedes-Benz Stadium in 2018 became the first major sports venue in the United States to switch to a fully cashless payment system. At the end of the new payment model’s first year of operations, the stadium reported that wait times had fallen by 20 to 30 seconds and per-capita food and beverage sales had risen by 16%, while saving more than $350,000 in operating expenses.[1]

Many other stadiums have since followed the Mercedes-Benz example,[2] and a growing number of restaurants and retail outlets are likewise going cashless. While some of these decisions were precipitated by COVID-19, the trend predated the pandemic and has continued in its wake, driven by a desire to reduce wait times and other costs associated with cash, such as counting and depositing it at the bank and mitigating the risk of robbery.[3]

Other merchants are keeping cash payments, for now, but many are no longer accepting personal checks. Target announced in early July that it would cease accepting checks July 15.[4] Several others—ranging from Whole Foods to Old Navy—recently have announced similar policies. The reasons for dropping checks are similar to those for cash: the cost of acceptance outweighs the benefits. Check transactions take much longer than cash, card, or mobile payments, and they come with a significant risk that the check will “bounce.” For Target, however, the final nail in the coffin was the “extremely low volumes” of check writing, which meant it no longer made sense to maintain check-acceptance facilities, which come with fixed costs.

Merchants are meeting their customers where they find them. In a 2022 Pew poll, 41% of Americans said they didn’t use cash in a typical week, while only 14% said they used cash for most or all their purchases—down from 24% in 2015.[5] These numbers are consistent with a Gallup poll the same year that found only 13% Americans stated that they used cash for most purchases.[6] It’s possible that the decline in cash use was exaggerated by the COVID-19 pandemic; a more recent Forbes poll found that 22% said they use cash most often for making purchases.[7] Nonetheless, the same Forbes poll found that 70% of Americans use cards most often, while 7% used digital wallets and 1% said they most often use buy-now-pay-later schemes.[8]

The Federal Reserve Board’s Diary of Consumer Payment Choice has also found a consistent decline in the proportion of payments made using cash. In volume terms, cash has largely been replaced by credit and debit cards (Figure 1).[9]

FIGURE 1: Share of Payment-Instrument Use for All US Payments (2016-2022)

SOURCE: Federal Reserve Board Diary of Consumer Payment Choice

Moreover, according to surveys by Pew (Figure 2), Americans at all income levels have reduced their use of cash and increased their use of cashless payments over the past decade.[10] In 2015, those with an annual income of $30,000 or less made substantially all their purchases using cash, while only 15% of that group made no purchases using cash. By 2022, the proportion only using cash had fallen to 30%, while the proportion not using cash at all had risen to 24%. In all other income groups, the proportion not using cash is now higher than the proportion only using cash.

FIGURE 2: Proportion of US Adults Who Use Cash for Their Purchases

SOURCE: Pew Research Center

Clearly, consumers have a strong and growing preference for electronic payments in general and cards in particular. But stories of the death of cash are premature. A recent YouGov poll found that, when asked to list all payment methods used in the past 30 days, cash was the one used by the largest proportion of respondents (67%).[11] This result does not necessarily contradict the other surveys: it is likely that most people who use cash do so only intermittently and for smaller purchases.

Despite this strong and growing preference for electronic payments by both merchants and consumers, there remains some confusion about the costs and benefits of different payment methods. The purpose of this white paper is to summarize the existing literature on the relative costs of cash, other paper-based payments (primarily checks), and electronic payments.[12]

In short, the evidence shows that, when all costs and all parties to a transaction are considered, electronic payments (debit cards, credit cards, and mobile payments) are more cost-effective than cash for most transactions. The main reason for this is that electronic payments enable consumers to spend more than they have in their wallet, which results in “ticket lift” for merchants. Card rewards, including cashback and merchant-specific loyalty programs, further increase this ticket lift.[13] In addition, “tap-and-pay” contactless payments can reduce the time it takes to tender payment relative to cash, especially when cash payments are eliminated altogether. This increases throughput, improving the customer experience and reducing labor costs. Finally, electronic payments enable merchants to sell online, including for in-store pickup.

It should be noted that this is not an argument for eliminating cash, which is likely to continue to play an important (if smaller) role in payments for many years to come. Rather, it is intended to offer a more balanced perspective on the role of cash and electronic payments in retail and other merchant settings—and to consider the implications for payments regulation.

A. Organization of the Review

Payments typically involve at least three parties: a seller, a buyer, and a bank. Where the buyer and seller use different banks, there will be at least four parties (unless one of those parties has chosen the self-custody option—also known as “cash under the mattress”). Depending on the payment method used, there may be other parties involved; for example, card payments may involve other processors, while cash payments may involve security companies moving physical cash to and from the merchant’s bank.

Each payment also typically involves a series of actions. For example, when making a purchase at a store, after items have been recorded at the register, the cashier tells the customer the total that is owed; the customer then proffers a means of payment (typically card, cash, or mobile); and the cashier processes the payment. In the case of cash, this will likely include calculating and returning change; for a chip-based card, it may involve dipping and either entering a PIN or rendering customer signature; or, if it is a contactless payment (card or phone) and the amount is below the floor limit, the customer may simply tap and go.

Most studies of the cost of payments use, in part at least, a version of “activity-based costing” (“ABC”) that seeks to account for all the costs associated with a particular payment type by assessing all of the associated activities.[14] There are, however, significant differences among the studies, both in what types of costs are included and how those costs are allocated. Broadly speaking, studies can be divided into those that focus exclusively on one party in the payment system (merchant, bank, or consumer) and those that seek to account for the costs to all (or, at least, most) parties—and hence to society. Reflecting these differences, the paper is organized as follows:

Section II reviews partial-cost studies, including those that focus exclusively on costs to merchants and costs to consumers.

Section III reviews social-cost studies that seek to evaluate the costs and benefits of different payment systems more broadly.

Section IV offers some conclusions.

II. Partial-Cost Studies

Many studies of the cost of payments focus primarily, if not exclusively, on the costs incurred by one part of the payment system—usually merchants. This section considers those studies, looking first at merchants, then consumers, then banks.

A. Merchant-Cost Studies

While merchant-cost studies are inherently narrow in scope and should not, by themselves, form the basis of public policy, they can offer valuable insights. For example, a 1983 study by the Federal Reserve noted that:

Many retailers tend to view the costs of handling cash transactions as equivalent to the cost of doing business—a sales clerk, for instance, must be on hand to conduct transactions of whatever type. Thus there is a tendency to regard the marginal cost of selling for cash as zero, but this view should not be adopted without critical examination.

There are many elements of cost associated with the handling of a sales transaction. Some costs may be higher for check or credit card transactions, but others may be higher for cash.[15]

The report went on to list many of the costs that should be considered, including the time to conclude the transaction, security costs, and counterparty risk.[16] Nonetheless, the study concluded that cash was generally less costly for merchants than other forms of payment, including cards, in part because it assumed that the use of payment cards does not result in a net increase in sales.

1. Robert M. Grant’s pioneering study

Another study published in 1983 (but undertaken prior to the Federal Reserve study) considered the costs of several different retail-payment methods in the United Kingdom: “Cash; Cheques; Bank credit cards (Access and Visa); Travel and entertainment (T&E) credit cards (American Express and Diners Club); and In-house credit accounts.”[17] The line items considered in this study can be seen in Table 1. Of note, in contrast to the Federal Reserve study, author Robert M. Grant concluded that, while the direct cost of credit cards and in-house credit were higher than the direct costs of cash and checks, these costs were more than offset by increases in sales, which the author accounted for as a reduction in the unit cost of overhead.[18] One reason Grant found such significant increases in sales associated with payment cards and store credit was that such payments represented 11% of sales in stores that took credit, compared with about 6% for all stores. Grant therefore reasonably assumed that between 20% and 30% of sales made using credit constituted “additional” sales.[19]

TABLE 1: Average Cost of Payment Methods as % of Retailer’s Revenue (1981)

SOURCE: Robert M. Grant

2. Accounting for ‘ticket lift’ and increased throughput: Layne Farrar (2011)

Many subsequent studies have sought to assess the relative costs to merchants of accepting different forms of payment. Unlike Grant’s original study, however, few of have attempted to account for the effect the method of payment might have on demand. One that did take such an approach was a 2011 study by Anne Layne-Farrar, who found that the use of payment cards were associated with higher per-transaction (“ticket”) amounts; this is known as “ticket lift.”[20]

Layne-Farrar looked at the costs and benefits of different payment methods at a range of different retail outlets: quick-serve restaurants (QSR), supermarkets, discount retailers, retail-gasoline outlets, and travel retailers (stores at train stations and airports). She begins her account of the QSR analysis with the following observation:

In 1998, Sonic Inc., an Oklahoma City based drive-in chain, became one of the first QSRs to accept cards at its 2,200 restaurant locations. According to an article published three years later, in 2001, the increasing relative costs of handling cash as compared to card payments was the primary motivation for Sonic. Technological advances over time have lowered the network and equipment costs of processing card transactions while the costs of handling cash appear to have remained flat. Sonic found that customer orders (tickets) paid by card were 80 percent higher than cash tickets. In other words, although Sonic decided to accept cards in order to lower its cash handling costs, it found direct benefits from card acceptance in the form of dramatically higher sales.

KFC began accepting cards in 2001, three years after Sonic. In contrast to Sonic, as its motivation KFC cited specific benefits expected from cards, rather than solely the savings derived from reduced cash handling. Specifically, KFC began accepting payment cards as a way to sell its higher priced group meals, such as large buckets of chicken with side dish containers and packages of biscuits.[21]

Following this prelude, Layne-Farrar provides a detailed analysis of the various costs associated with accepting different payment methods, the average sizes of transaction made with those methods, and the benefits of increased revenue resulting from the use of payment cards relative to cash.

Layne-Farrar found that, in addition to ticket lift, the use of payment cards reduced average transaction time, resulting in increased throughput of customers, which resulted in a further increase in revenue. Table 2 provides a summary of Layne-Farrar’s estimates of the costs and benefits from the use of different payment methods for an average transaction assuming ticket lift of 10%, which Layne-Farrar saw as a low amount (she also calculated the effects of 20% ticket lift).[22] While cash transactions were less costly to process at the average ticket size, when taking into account ticket lift and throughput improvements, the net benefits per-transaction were significantly higher for debit cards.

TABLE 2: Per-Transaction Costs and Benefits for QSRs (2011)

SOURCE: Layne-Farrar

Tables 3-7 are Layne-Farrar’s estimates of the net benefits of using different transaction methods at big-box discount retailers, supermarkets, gas-station retail, non-fuel convenience stores, and travel retail.[23] In each case, only the low-ticket-lift option is given (the net benefits of the high-ticket-lift option are always greater); nonetheless, the net benefits of debit cards exceed those of cash and, where checks are evaluated, those as well. The clear conclusion of this work is that, by the time Layne-Farrar undertook her analysis in 2011, acceptance of debit cards generated per-transaction net benefits for merchants that exceeded those of cash.

TABLE 3: Per-Transaction Costs and Benefits for Big-Box Discount Stores

SOURCE: Layne-Farrar (2011)

TABLE 4: Per-Transaction Costs and Benefits for Supermarkets

SOURCE: Layne-Farrar (2011)

TABLE 5: Net Benefit Per Fuel Transaction ($) for Gas-Station Retail

SOURCE: Layne-Farrar (2011)

TABLE 6: Net Benefit Per Non-Fuel Transaction ($) for Convenience Stores

SOURCE: Layne-Farrar (2011)

TABLE 7: Net Benefit Per Fuel Transaction ($) for Travel Retail

SOURCE: Layne-Farrar (2011)

3. Economists Inc.

In 2014, Economists Inc. carried out a study similar to Layne-Farrar’s, but went into even more granular detail regarding the merchants’ processes for accepting payments, focusing on five merchants: a fast-food restaurant, a full-serve restaurant, a gas station, a convenience store, and a small independent grocery store.[24] They also included both credit and debit cards in their analysis.

Table 8 shows tender times for cash and credit (including debit run as “credit”) at the five merchants.[25] It is noteworthy that both the mean and median tender times at the fast-food restaurant and grocery store were slightly lower than for cash, whereas the in-person tender times at other merchants were slightly higher for credit.

These differences may reflect different rates of throughput and associated investment in technology and training. At higher-throughput merchants, such as grocery stores and fast-food restaurants, there are likely greater returns on investments in technology that integrates checkouts with point-of-service (POS) machines, for example, such that the checkout operator does not need to re-input information to the POS machine.[26]

The tender time for self-service gas pumps is, perhaps unsurprisingly, considerably shorter than the time when paying in-store, as this avoids the shoe-leather time involved with walking to the store to pay and back.

TABLE 8: Tender Times at the Merchants Studied by Economists Inc

SOURCE: Economists Inc. (2014)

Like Layne-Farrar, Economists Inc. found that the use of payment cards resulted in significantly higher purchase amounts relative to cash. This ticket lift can be seen in Table 9, which shows that, for every establishment, the minimum, maximum, mean, and median payments are nearly all higher for payments made using credit cards (or debit run as “credit”) than for cash.[27] (The one exception is the minimum for in-store gas and joint sales at the gas station.)

Economists Inc. also reported that merchants themselves had indicated that, when they began accepting card payments, they noticed a significant increase in sales.[28]

TABLE 9: Amounts Spent When Using Cash or Credit

SOURCE: Economists Inc. (2014)

4.  Ticket lift and increased throughput from contactless

Contactless payments use RFID to transmit tokenized payment information from a card or from a smartphone. While the first contactless payment cards were introduced in the mid-1990s, and an EMV contactless standard was first developed in 1996, their uptake by both card issuers and merchants was initially low, especially in the United States.[29] However, the vast majority of U.S. debit and credit cards now support contactless payments, more than 150 million Americans have used contactless payment apps on their smartphones and, as of 2020, 58% of U.S. merchants accepted them.[30] Evidence from other jurisdictions suggest this switch to contactless is likely to result in both ticket lift and increased throughput.

A 2020 study by David Bounie and Youssouf Camara looked at the effects on 2018 sales of the shift to contactless payments at 275,580 merchants in France.[31] The researchers found that merchants who accepted contactless payments had card sales that were 15.3% higher, on average, than merchants who did not accept contactless payments.[32]

A 2022 study by Sumit Agarwal, Wenlan Qian, Yuan Ren, Hsin-Tien Tsai, and Bernard Yeung looked at the effect of the introduction of “quick-response” (QR) code mobile payments in Singapore in 2017.[33]  The researchers found that “monthly business creation among business-to-consumer industries increased by 8.9% more than among business-to-business industries.”[34] Meanwhile, use of mobile payments tripled, while use of automated-teller machines (ATMs) fell dramatically. Of particular note, consumers increased their credit-card spending by 3.3%, driving most of the increase in consumer spending that led to the increase in B2C business formation.[35]

5. Fumiko Hayashi’s comparison of debit and cash

In a 2021 study, Federal Reserve Bank of Kansas economist Fumiko Hayashi compared the cost of merchant acceptance of cash and debit cards in different countries. Table 10 reproduces the summary table in her study.[36]

TABLE 10: Merchant Acceptance Costs in Selected Countries, Various Years

SOURCE: Hayashi

Hayashi’s data for the Unted States are based on two studies: the first is the Food Marketing Institute (FMI) study from 2000 that formed the basis of the GHL analysis discussed at-length in Section III of this white paper, and the second is a Bank of Canada study that Hayashi coauthored with Marie-Hélène Felt, Joanna Stavins, and Angelika Welte[37].

There are several problems with both of these studies: first, as GHL point out, the FMI study does not consider costs associated with either theft or counterfeit when calculating the costs of cash.[38] Second, the “2018” data from the United States is based on 2015 survey data on the cost of acceptance in Canada that has been adjusted by using certain U.S.-specific data.[39] This was done despite Hayashi herself arguing in a previous paper coauthored with William Keeton that such practices are inappropriate, especially where the United States is concerned, noting that:

The danger of relying on other countries’ cost studies is particularly apparent for the United States, where checks and credit cards are used on a larger scale and more parties are involved in the payments process.[40]

Among other things, these differences in the mix of payment type will have an effect on the optimal fees charged by parties to the system, due to the two-sided nature of payments.[41] For example, the “interchange” fees retained by issuing banks might be higher in the United States due to issuers seeking to increase their share of the payments market by offering incentives for consumers to switch from checks to cards (for example, in the form of cashback or other rewards).[42] (These fees and incentive payments are essentially transfers within the system, not a net cost. Indeed, to the extent that they result in a shift to more socially efficient modes of payment, they are, on net, beneficial.)

At a more general level, Hayashi’s survey suffers from inappropriate aggregation. That is to say, whereas it is likely true that debit-acceptance costs exceed those of cash for some proportion of payments, it is unlikely to be true for all merchants and all amounts. Indeed, the study of the Netherlands that Hayashi used notes that, although the average cost of acceptance for debit was notionally higher than for cash in 2002, the “breakeven point” (when taking into account costs for both retailers and banks) was €11.63, which was below the average ticket size for debit (€44).[43]  Meanwhile, in 2009, the merchant-acceptance cost for debit was on average lower than for cash; the breakeven point had, however, fallen to €3.06, not zero.[44]

Finally, and related to the problem of inappropriate aggregation, Hayashi fails to consider the ticket-lift effect identified by Grant and confirmed by Layne-Farrar, Economists Inc, Bounie & Camara, and others.

6. Problems with generalizing across jurisdictions

There is broad agreement that the cost of using different payment types for a transaction of a given amount varies significantly across jurisdictions and sectors. For example, in their 2003 study of the Norwegian payment system, using data from a 2001 survey, Olaf Gresvik and Grete Øwre found that Norwegian banks operated payments at a loss (although technological improvements had decreased the cost of the payment system over time).[45] By contrast, at least until 2010, debit payments in the United States were used to cross-subsidize checking accounts.[46]

Within the United States, costs can vary considerably from state to state. For example, a cashier in San Jose, California, might earn $20/hour, while a cashier in Louisville, Kentucky, might earn $10/hour. If both cashiers take the same amount of time to tender payments, the tender costs in San Jose could be twice what they are in Louisville. Even within a state, salaries can vary somewhat, with cashiers in Smyrna, Tennessee earning an average of $14.36, according to Indeed.com, while those in Jackson, Tennessee earn only $11.45.[47]

But that is far from the whole story. While salaries might be higher in, e.g., California, the average ticket is also likely to be higher. Thus, the number of sales required per-dollar of income will be lower. This means that the average time to tender any specific dollar amount will be lower in California. How this affects the net costs to tender per-dollar depends on the ratio of spending to salaries, which is not only location-specific, but also merchant-specific.

7. Tender-time studies

A subcategory of merchant-cost studies focuses more narrowly on the time taken to tender a payment. These are essentially “time and motion studies” of the parts of the retail-checkout process that involves payment. Table 11 provides some examples of such studies. (These studies also have implications for consumers for whom the time taken to process payments and, relatedly, the time spent queuing has an opportunity cost, as discussed in Section III.)

TABLE 11: Tender Time Studies (Time to Complete Tender in Seconds)

[48][49][50][51][52] SOURCE: Polasik et al., adapted by author

8. Studies of the cost of accepting cash

A final subcategory of merchant cost-of-acceptance studies looks in greater detail exclusively at the cost of accepting cash. A relatively recent example in the U.S. context was produced by IHL Group, which undertook a very detailed activity-based costing.[53] Specifically, IHL identified nine activities associated with the cost of cash:

1. Start/Rebuild Drawer-Functions related to opening drawer from initial deposit to rebuild of for next cashier.

2. Closing Drawer-This includes functions related to closing out a drawer. Time for cashiers, managers, or cash office personnel to count and reconcile the drawer with POS or cash register totals.

3. Pickups-This is inclusive of pickups during a shift for too much cash in the drawer or bills that are large denominations.

4. Change Orders-Cost associated with a cashier requiring change throughout the shift

5. Audits/Discrepancies– These are costs of redoing counts, auditing tills, and time associated with recounts for any discrepancies.

6. Prepare/Coordinate Deposits-Costs associated with preparing or coordinating deposits.

7. CIT/Deposit Costs-These are costs of Cash In Transit companies (armored trucks) or cost for managers or other employees to go to bank to make the deposits.

8. Bank Charges-These are charges surrounding bank fees. This includes statement fees, reconciliation, cash value fees and change orders among other things.

9. Cash Shrink-This is the cost of theft, fraud or other cash loss activities. This is cash that just disappears in the process. For this study we used previous data from other studies as a value.[54]

IHL then undertook several hundred interviews and numerous modeling exercises to establish the amount of time associated with actions 1-6 and, using wage estimates for each relevant job in each relevant location, they calculated the total cost for each such activity. They then added costs 7-9 to arrive at a total for each business.

IHL found that the average cost of cash across all segments was 9.1% of the revenue of the businesses studied.[55] Of this, the largest component was “close drawer,” accounting for about 40% of the total, whereas bank charges were only 4.3%.[56]

Table 12 shows the range of costs incurred by different businesses for accepting cash.[57] Even the lowest of these (food/grocery) has a cost of cash (4.7%) that is higher than the highest merchant-discount rates charged by acquiring banks.

TABLE 12: Cost of Cash by Segment

SOURCE: IHL

B. Consumer-Cost Studies

While most partial-cost studies focus on merchants, some look at consumers. Several studies have shown that transaction value is a key determinant of the method of payment. One explanation is that consumers want to avoid receiving a significant amount of change in the form of coins. Thus, Heng Chen, Kim P. Huynh, and Oz Shy of the Bank of Canada found that “a significant number of cash users … switch to paying with debit or credit cards at transaction values marginally above $5 and $10.” [58] They attribute this to “the burden of receiving coins as change associated with the currency denomination structure.”[59] (The Bank of Canada withdrew the $1 note in 1989 and the $2 note in 1996, leaving $5 as the smallest denomination bill.[60])

C. Bank-Costs Studies

A third category of partial-cost studies considers the costs to banks. The most notable example of this is the study undertaken by Olaf Gresvik and Grete Øwre mentioned above, which formally introduced the concept of activity-based costing (ABC) to payments and applied it to the processing of payments by Norwegian banks, based on a 2001 survey.[61]

III. Social-Cost Studies

The second category of payment-cost studies seeks to evaluate not only the costs to merchants, but the costs to society as a whole. An early focus of such studies was checks, which at the time were the dominant form of noncash payment. A 1990 paper by David Humphrey and Allen Berger considered the divergence between the private and social costs of payments in the United States, with a particular focus on checks, which the authors argued were overused because payor businesses (in particular) benefitted from the float[62] associated with checks that had not yet cleared.[63]

A 1996 paper by Kirstin Wells, using data from 1993, compared the social cost of checks with that of automated-clearinghouse (ACH) payments and concluded that, in contrast to the 1987 data used by Humphrey and Berger, there was not a significant difference in float cost between using checks and ACH.[64] Indeed, Wells estimated that the value of float fell 91.3% from an average of $1.04 to $0.09, mainly due to increases in the efficiency of check processing.[65]

Starting in the early 2000s, the focus of social-cost-of-payments studies shifted to retail payments and, specifically, to the relative cost of cash and payment cards (although checks were also evaluated in some studies in the 2000s, as they were then still quite widely used). This section focuses on such studies.

A. Garcia-Swartz, Hahn, and Layne-Farrar

In a study originally published by the AEI Brookings Joint Center for Regulatory Studies in 2004, and subsequently updated and published in the Review of Network Economics in 2006, Daniel Garcia-Swartz, Robert Hahn, and Anne Layne-Farrar (“GHL”) undertook arguably the first and still one of the most comprehensive social-cost (or benefit-cost) assessments of retail payments.[66]

1.  Merchant costs

The starting point for GHL was a study undertaken by the Food Marketing Institute (FMI) in 1998 that sought to calculate the direct costs of accepting various payment types, namely: cash, checks (verified and nonverified), credit cards, and debit cards (signature and PIN). These direct costs (as categorized by the FMI) were:

1. “Tender-time”: this is the cost of the time spent by cashiers processing a transaction after ringing up all the items (this was based on another FMI study, from 2000, at which time cash remained quicker than card, and wages from the 2002 Bureau of Labor Statistics survey);

2. “Deposit preparation”: this is the cost of the time taken to prepare a cash deposit (e.g. counting cash, reconciling the register drawer, preparing a deposit slip, etc.);

3. “Bank charges”: these are the explicit fees charged by banks, such as a deposit fee for cash and checks, or the merchant discount rate for cards;

4. “Other direct costs”: these include costs such as using armored cars to transport cash, collection costs and losses on “bounced” checks, and credit card chargebacks.[67]

Table 13 reproduces GHL’s summary table showing these initial calculations.[68] As can be seen, based only on these costs, the per-transaction cost of cash is lower than that of other payment types. The amount tendered in the average cash transaction is, however, much lower than the amount tendered in other transaction types. When scaled to $100 of sales, cash is the second-most costly for the merchant, after credit cards, while verified checks are the least costly.

GHL then note that the FMI analysis omits two potentially significant costs: (1) theft and counterfeit losses for cash and (2) float loss for all payment types.

TABLE 13: Grocery Stores’ Per-Transaction Processing Costs for Various Payment Instruments, Modified ($), (2003)

SOURCE: Garcia-Swartz, et al.

2. Consumer costs

GHL identify the following payment-related consumer costs:

1. Processing time: this is the opportunity cost of the consumer’s time while waiting for the transaction to be processed.

2. Queue time: this is assumed to be equal to processing time

3. Explicit price: this is the explicit bank service charge associated with withdrawing cash and processing cheques and debit transactions.

4. Implicit price: this is the “shoe-leather” costs of obtaining cash (i.e. the opportunity cost of the time taken to travel to/from an ATM and withdraw cash).

5. Seigniorage: the profit made by the central bank from printing money (basically the difference between the face value of the currency and the costs of production)

3. Bank costs

In addition to merchants and consumers, banks also incur costs in the form of ATM maintenance (applies to cash); production of cards (applies only to cards); transaction processing (applies to all payment methods); and card rewards (applies mainly to credit cards, but also to a lesser extent to debit cards).

4. Central bank costs

A fourth set of costs arise from the involvement of the central bank in producing and processing banknotes and coins, as well as in processing checks (approximately $0.0015 and $0.03, respectively, for a transaction of $11.52). The cost of check processing, however, is recovered from banks, is therefore included in banks’ processing costs.

TABLE 14: Grocery Stores’ Per-Transaction Processing Costs for Various Payment Instruments, Cash Transaction of $11.52 ($), (2003)

SOURCE: Garcia-Swartz, et al. (2006)

Putting these all together, GHL calculate the total marginal cost for transactions of $11.52 (the average size of a cash transaction) and $52.24 (the average size of a check transaction).[69] Table 14 replicates the figures for the $11.52 transaction.[70] Once double-counting is eliminated, the social cost of paying with cash and card are roughly the same.

Meanwhile, for grocery-store transactions of $54.24, debit cards have the lowest social cost ($0.94 for PIN-authorized transactions and $1.00 for signature-authorized transactions), followed by verified check ($1.08), credit card ($1.32), nonverified check ($1.40) and, finally, cash ($1.98).[71]

5. Accounting for (social) benefits

But the story does not end there. GHL note that there is a range of benefits arising from the use of certain payment methods that, at least partially, offsets these costs. For consumers these include:

1. Float: while checks, credit and charge cards impose float costs on merchants, they provide consumers with some float (in the case of credit cards used purely transactionally this may be quite large).

2. Credit option: the option to use the credit function of credit cards

3. Record keeping: electronic transactions and even cancelled checks provide a record that is valuable to many consumers.

4. Cashback at POS saves a trip to the ATM (but the amount is limited)

5. Rewards cards (mainly credit) provide marginal benefits about twice their cost

6. Discover cards provide marginal benefits equal to their marginal cost

7. Privacy, the exclusive domain of cash, offers users significant benefits [albeit at a cost in terms of dramatically increased costs of recourse] [72]

For banks, the benefits include a small amount of float (this is basically the counterpart of the float costs incurred by consumers who use cards) and processing revenue, which is part of the amount earned by banks for processing transactions. Meanwhile, central banks earn seigniorage and a small amount for processing transactions (which, as noted earlier, is netted out).

The marginal benefits associated with a typical check-size transaction (for 2003, when such transactions were more common) of $52.24 are shown in Table 15. The payment methods providing the greatest marginal benefit are credit ($1.61 per transaction) and cash ($0.92 per transaction).

By adding the marginal benefits to the marginal costs, GHL are then able to calculate the net social marginal cost associated with the different payment types. For a transaction of this size, in 2003, the authors estimate that credit cards would have the lowest net social cost, followed by PIN debit, signature debit, verified check, cash, and nonverified check.

TABLE 15: Adding Benefits to Grocery-Store Transactions of $52.24

SOURCE: Garcia-Swartz, et al.

GHL use the same methodology to calculate the net social marginal cost for transactions at two other types of merchant: discount stores and specialty electronics stores.

TABLE 16: Net Social Marginal Cost for Transactions at Discount and Electronics Stores

SOURCE: Garcia-Swartz, et al.

In the case of discount stores, for purchase amounts of $15.49, the social cost of cash, bank credit, American Express, and debit are about equal. Meanwhile, for larger amounts, cash is more costly than all other payment methods except check.

6. Sensitivity analysis

GHL then undertake a “sensitivity analysis,” in which they adjust some of the parameters of their estimates. For example, increasing the number of people queueing for checkout increases the net social cost of checks relative to cash quite significantly and increases the net social cost of cards slightly. The result is that, for example, if there are three people queuing at GHL’s average grocery store spending the average cash amount at such a store ($11.52), cash becomes marginally socially beneficial.[73] Payment cards, however, remain superior at the amounts typical for those purposes ($33 for signature debit, $41.05 for PIN debit, and $44.59 for credit). In other words, the “break even” point for those payment types shifts to the right.

7. GHL’s conclusions

It is worth repeating the conclusions GHL drew from their research (in a separate study published alongside the detailed analysis described above):

First, transaction size assumptions are critical in analyzing payment-processing costs. At smaller transaction sizes, the net social marginal cost of all payment instruments – paper and electronic alike – are remarkably similar. No one instrument stands out as more socially efficient. At larger transaction sizes, however, significant differences emerge. For grocery store transactions, electronic payments are considerably less costly on net for society than paper methods. Yet another pattern emerges for the larger transactions conducted at electronics stores. Here credit cards with a large proportion of reward cardholders have the lowest net social marginal cost. This pattern is consistent with observed behavior: namely that cash use dominates smaller transaction sizes but drops precipitously as transaction size increases.

Second, retailer type influences the individual cost elements and thus affects private cost calculations. Since the distribution of transaction sizes differs across venues, this result follows naturally from our first finding.  Added to the transaction size effect are apparent differences in merchant costs, such as point of sale time and back-office processing costs.

Finally, and most importantly, the relative merits of different payment methods change significantly when all parties are counted and benefits are included. Merchant studies have found that paper methods are the cheapest for merchants. This is confirmed in our study of the distribution of private costs and benefits. But what is cheap for merchants is relatively expensive for other parties to a transaction. Certain parties, especially consumers, receive considerable benefits from payment cards, which tip their net private costs in favor of that method of payment.[74]

B. Shampine Critique of GHL and GHL’s Response

Allan Shampine undertook a critical review of GHL, questioning their assumptions regarding the amount of time taken to obtain cash from ATMs, the value of card rewards, the range of nonpecuniary benefits that consumers derive from different payment methods, and the appropriateness of some other cost categories, such as seigniorage.[75] He then applied his own sensitivity analysis—which, by making very different assumptions, found that, for certain payment sizes that GHL had identified as lower cost for cards, cash may be lower cost.

GHL responded by noting that, of course, if one makes different assumptions, it is possible to achieve different outcomes.[76] But where the assumptions that GHL made were at least supported as far as possible by empirical evidence, most of the adjustments Shampine made had no empirical basis and should therefore not be treated as reliable. Moreover, as they note, there are very significant individual differences among consumers, merchants, and banks regarding the costs and benefits of any particular payment type.

Ironically, one of the GHL assumptions that Shampine criticizes as insufficiently generous is the privacy benefits of cash. While it is no doubt true that some consumers benefit from the anonymity of cash payments, for most consumers, that is not the main priority. Moreover, there is a social cost to privacy when consumers use cash to engage in illegal activity. Indeed, security is typically more important, and since cards are far more secure than cash, it is possible that the relative benefits are tipped even more toward cards for most consumers.

C. Other Social-Cost-of-Payments Studies

Numerous researchers have undertaken studies to estimate the social cost of payments in other jurisdictions. While several of these studies seek to account for costs borne by consumers, none of those we identified has been as comprehensive or detailed as GHL.[77] Specifically, none adequately account for the social benefits of different payment modes. Also, to our knowledge, none of them focus on the United States. Nonetheless, the studies have introduced some valuable insights. Perhaps most notable is the importance of differentiating fixed and variable costs (although, as discussed below, these costs change over time).

1. Fixed v variable costs, and the ‘breakeven’ point

In their analysis of the Dutch payments system, Hans Brits and Carlo Winder show that payment cards have relatively high fixed costs and much lower variable costs. As a result, in 2002, the “breakeven” point for debit transactions occurred at €11.36.[78] Above that amount, it is more socially cost efficient to pay with debit than with cash. This can be seen in Figure 3, which shows the relative cost of making a payment with cash, debit card, or “e-purse” (this last is a rechargeable smart card that can be used to pay for a range of goods and services in the Netherlands).

FIGURE 3: Breakeven Points for Different Payment Types, Netherlands (2002)

SOURCE: Brits & Winder (2005)

In a study of the private and social costs of payments in Sweden in 2002, Mats Bergman, Gabriella Guibourg, and Björn Segendorf of Sweden’s central bank found that the unit transaction costs of cash (4.6 SEK) was higher than for either credit cards (4.4 SEK) or debit cards (3.1 SEK for PIN, 3.2 SEK for signature).[79] They then sought to identify the breakeven point for each payment method, and found that debit cards became more cost-effective than cash at about 72 SEK (U.S. $7), while credit cards became more cost-effective than cash at about 160 SEK (U.S. $16).[80]

Technological improvements have reduced both the fixed and variable costs associated with card-based payments. For example, the fixed cost associated with the time taken to process a card-based transaction has generally fallen.[81] Meanwhile, both the fixed and variable costs associated with card-based fraud has fallen by more than 75% as a result of the introduction of the Europay, Mastercard and Visa (EMV) Chip.[82]

FIGURE 4: Breakeven Points for Different Payment Types, Sweden (2002)

SOURCE: Bergman, Guibourg, & Segendorf (2007)

2. Technological change lowers the breakeven point

While technological improvements have also increased the efficiency of processing cash, there are some human aspects to cash processing that are almost impossible to eliminate, and that are inherently proportional to the transaction amount.

Increased use of a payment method can create a virtuous circle for that method, while having the opposite effect for other methods. So, for example, the shift from cash to debit in the Netherlands resulted in lower average debit-acceptance costs, but increased the average acceptance costs of cash. Thus, Nicole Jonker found the breakeven point for merchant acceptance of debit in the Netherlands had fallen from the €11.36 found by Brits & Winder in 2002 to €3.06 in 2009.[83]

A similar phenomenon appears to be happening with both cash and checks in the United States, as demonstrated by the examples given in the introduction. It is reasonable to ask why this shift seems to have happened later in the United States than in Europe. Aside from “culture,” one argument is that U.S. banks implemented more-efficient check processing in the early 1990s, thereby reducing the incentives for merchants to switch. Another is that the Durbin amendment in 2010 reduced consumers’ incentives to pay with debit, and resulted in their switching from debit to credit for lower-value payments, which increased the relative cost of acceptance for merchants.

D. Partial Social-Cost-of-Cash Studies

Many other studies advertised as assessing the social costs of payments have considered the effects on a somewhat narrower range of participants. For example, a study published by the European Central Bank titled “The Social and Private Costs of Retail Payment Instruments: A European Perspective” notes:

Due to the considerable effort necessary to collect viable data on the costs incurred by all of the parties in the payment chain, the analysis focuses on the most important parties issuing authorities, i.e.:

• central banks and governments;

• banks and interbank infrastructure providers (automated clearing houses, ATM networks, etc.);

• retailers and companies; and

• cash-in-transit companies.[84]

There would appear to be a rather significant omission from this study: consumers (indeed, the ECB acknowledges this deficit). The same omission is present in many other similar studies. This is both rather odd and rather troublesome, since consumers clearly are important participants in the payment system—indeed, without them, the system would be pointless. Moreover, if GHL are correct, consumers tend to benefit more from electronic payments than from cash—indeed, cash is typically a net cost for consumers—so omitting them from the analysis seems more than a mere oversight.

IV. Conclusion

This review shows that both the partial and social costs of different modes of payments vary considerably by location, type of merchant, and over time. Nonetheless, several broad conclusions emerge:

First, retailers that accept card payments tend to experience ticket lift; many also benefit from increased throughput. As a result, retailers such as quick-serve restaurants that sell low-ticket items and might be below the “breakeven” point for cards relative to cash (if considering only the direct transaction costs) but benefit on net from adding cards because of the significant ticket lift and increase in throughput.

Second, over time, there has generally been a reduction in the “breakeven” point for electronic payments. This has likely been driven by such innovations as the EMV Chip and contactless payments, which have reduced fraud and tender-time costs, and increased benefits to all parties.

Third, while innovations in cash management have also reduced the cost of accepting cash in general, the cost of multimodal payment acceptance means that the relative cost of continuing to accept cash has increased, especially in locations where throughput is of the essence, such as ballparks and quick-serve restaurants. This has led some such merchants to drop cash acceptance.

[1] Mercedes-Benz Stadium Achieves Success in First Year of Stadium-Wide Cashless Initiative, Mercedes-Benz Stadium (Mar. 9, 2020), https://www.mercedesbenzstadium.com/news/mercedes-benz-stadium-achieves-success-in-first-year-of-stadium-wide-cashless-initaitive (“MBS’s expected first-year results have been realized, once again locking in the No. 1 spot for food and beverage including speed of service across all NFL venues for the third consecutive year.  Due to the new cashless model, roughly 95 percent of fans noticed the same or an increase in speed at concession lines and at peak times a 20-30 second reduction in wait times. Results also include an increase in food and beverage per capita numbers for close to 50 events at MBS through 2019 including a combined 16 percent increase for Atlanta Falcons and Atlanta United all while saving more than $350,000 in operational expenses… Since going cashless, more than 2.5 million guests have attended events at MBS. Of those, only 1.2 percent have used the cash-to-cards kiosks, showing that fans are bringing their own credit cards or using mobile payment options.”).

[2] See Ben Gran, More Stadiums Are Going Cashless. What Does This Mean for Your Personal Finances?, the ascent (Feb. 13, 2024), https://www.fool.com/the-ascent/personal-finance/articles/more-stadiums-are-going-cashless-what-does-this-mean-for-your-personal-finances (“America is quickly becoming a more cashless society, and sports venues are leading the charge. As of 2022, nearly all Major League Baseball ballparks had gone cashless, along with most NFL stadiums and NBA/NHL arenas like the United Center in Chicago.”).

[3] See, e.g., Benjamin Gottlieb, Why Some Restaurants in LA Are Going Cash-Free, Marketplace (Aug. 19, 2019), https://www.marketplace.org/2019/08/19/why-some-restaurants-in-la-are-going-cash-free.

[4] Nate Delesline III, Target to Stop Accepting Personal Checks, Retail Dive (Jul. 9, 2024), https://www.retaildive.com/news/target-to-stop-accepting-personal-checks/720792.

[5] Michelle Faverio, More Americans Are Joining the ‘Cashless Economy’, Pew Research Center (Oct. 5, 2022), https://www.pewresearch.org/short-reads/2022/10/05/more-americans-are-joining-the-cashless-economy.

[6] Jeffrey M. Jones, Americans Using Cash Less Often; Foresee Cashless Society, Gallup (Aug. 25, 2022), https://news.gallup.com/poll/397718/americans-using-cash-less-often-foresee-cashless-society.aspx.

[7] Ketherine Haan, People Are Twice as Likely to Spend More Money When Using Card than Cash in 2024, Forbes Advisor (May 16, 2024), https://www.forbes.com/advisor/business/software/people-twice-likely-spend-using-card-than-cash.

[8] Id.

[9] See Emily Cubides & Shaun O’Brien, 2023 Findings from the Diary of Consumer Payment Choice, The Fed. Res. Fin. Serv. (Jul. 2023), at 6, available at https://www.frbsf.org/cash/wp-content/uploads/sites/7/2023-Findings-from-the-Diary-of-Consumer-Payment-Choice.pdf (which notes that “The category ‘other’ includes payments made with pre-paid [debit], checks, mobile payment apps, money orders.”).

[10] Faverio, supra note 5.

[11] Kineree Shah, Cash Remains King – 67% of Americans Still Use Traditional In-Store Payment, YouGov (Feb. 12, 2024), https://business.yougov.com/content/48650-cash-remains-king-67-of-americans-still-prefer-traditional-in-store-payment.

[12] A survey of the available literature was conducted using EconLit, the Social Science Research Network (www.ssrn.com), the IDEAS database (ideas.repec.org), and Google Scholar. This enabled us to identify the primary methodologies used to evaluate the value, costs, and benefits of different payment technologies. Initially, we used search terms such as “cost of cash” and “cost of payments” (with and without the restrictive use of speech marks). As the research progressed, we expanded this to a range of other terms, including “speed of payments,” so as to capture a wider range of studies addressing the issues under consideration.

[13] Credit cards take this one step further, enabling consumers to spread their spending out, reducing temporary liquidity constraints without the need to arrange an overdraft or other loan, thereby further increasing ticket lift—especially for higher-value items.

[14] In a traditional activity-based costing study, the aim is to account for the costs of each activity undertaken by a business and thereby enable management to make better decisions on resource allocation, investments in innovation, and so on.

[15] Board of Governors of the Federal Reserve System, Credit Cards in the U.S. Economy: Their Impact on Costs, Prices, and Retail Sales 36 (Jul. 27, 1983), available at https://fraser.stlouisfed.org/title/credit-cards-us-economy-5331.

[16] The study notes some of the costs: “Included among the relevant cost concepts, for example, would be the time required to complete a trans­ action, which may in turn influence the number of check-out stations and sales clerks that a store needs. Credit card transactions absorb time because credit slips must be written and frequently some sort of authorization procedure undertaken. Personal checks usually trigger certain time-consuming precautionary steps, such as inspecting and copying down identification data or summoning a manager from elsewhere in the store to approve acceptance of the check. Cash transactions most likely consume less time than check or credit card transactions, but the counting of cash received, the making of change, and the stocking and replenishment of cash registers with currency and coin are cash-related activities that occupy an employee’s time. Time consumed in reconciling sales records with cash, checks, and credit slips on hand may vary with the proportion of sales transacted by each means, and from one business to another. Security-related expenses comprise a large family of costs in which further variation may be found among the different means of payment. Included in such a concept would be both direct expenses of security precautions plus an allowance for any uncovered risk associated with each transaction medium. An obvious risk, for example, is the possibility of theft. This particular risk is likely to be more pronounced for cash because the full negotiability of cash makes it an attractive target. Acceptance of personal checks entails the risk that the check may be uncollectable, because the writer may not have sufficient funds on deposit or for some other reason.  Security risks borne by operators of in-house credit card plans include the costs associated with delinquent and uncollectable accounts.” Id. at 36-37.

[17] See Robert M. Grant, Transaction Costs to Retailers of Different Methods of Payment: Results of a Pilot Study, 4(2) Managerial & Decision Econ. 89-96 (Jun. 1983).

[18] Id. at 91. Grant is one of the most-cited social scientists, with nearly 100,000 citations and an H index of 58. See Google Scholar Profile of Robert M. Grant, Google Scholar (last accessed Aug. 22, 2024), https://scholar.google.com/citations?user=CQ8P0PcAAAAJ&hl=en.

[19] Grant, supra note 17, at 93, 95-96.

[20] Anne Layne-Farrar, Are Debit Cards Really More Costly for Merchants? Assessing Retailers’ Costs and Benefits of Payment Instrument Acceptance (SSRN Working Paper Sep. 9, 2011), available at https://ssrn.com/abstract=1924925.

[21] Id. at 6.

[22] Id. at 16-17.

[23] See id. at 23, 27, 34, 35, 39.

[24] Retailer Payment Systems: Relative Merits of Cash and Payment Cards, Economists Inc. (Nov. 19, 2014), available at https://ei.com/wp-content/uploads/2015/01/Cost_of_Cash_Study.pdf.

[25] Id. at 52.

[26] Id. at 21.

[27] Id. at 5.

[28] Id. at 58-60.

[29] See Tom Akana & Wei Ke, Contactless Payment Cards: Trends and Barriers to Consumer Adoption in the U.S. (Discussion Paper 20-03, May 2020), available at https://www.philadelphiafed.org/-/media/frbp/assets/consumer-finance/discussion-papers/dp20-03.pdf.

[30] See US Contactless Payment Statistics, Finical Holdings LLC (last accessed Aug. 22. 2024),  https://finicalholdings.com/us-contactless-payment-statistics.

[31] David Bounie & Youssouf Camara, Card-Sales Response to Merchant Contactless Payment Acceptance, 119 J. Banking & Fin. 105938 (Oct. 2020).

[32] Id.

[33] Sumit Agarwal, Wenlan Qian, Yuan Ren, Hsin-Tien Tsai, & Bernard Yeung, The Real Impact of FinTech: Evidence from Mobile Payment Technology (Working Paper, NUS Business School, National University of Singapore, September 2022), available at https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3556340.

[34] Id. at 4.

[35] Id. at 5.

[36] Fumiko Hayashi, Cash or Debit Cards? Payment Acceptance Costs for Merchants, 106(3) Econ. Rev. Fed. Res. Bank of Kansas City 53 (Aug. 2021), available at https://www.kansascityfed.org/Economic%20Review/documents/8213/EconomicReviewV106N3Hayashi.pdf.

[37] See Fumiko Hayashi, Marie-Hélène Felt, Joanna Stavins, & Angelika Welte, Distributional Effects of Payment Card Pricing and Merchant Cost Passthrough in the United States and Canada (Fed. Res. Bank of Kansas City, Research Working Paper no. 20-18, Dec. 2020), available at https://www.kansascityfed.org/Research%20Working%20Papers/documents/7595/rwp20-18.pdf.

[38] See Daniel Garcia-Swartz, Robert Hahn, and Anne Layne-Farrar, The Move Toward a Cashless Society: Calculating the Costs and Benefits, 5(2) Rev. of Network Econ. 202 (2006) [hereinafter “GHL”] (“The FMI study omits… theft and counterfeit loss for cash…”).

[39] Specifically: “The total cost of accepting cash, debit card, and credit card payments is based on the Bank of Canada Retailer Survey, but by using United States–specific information as follows: merchant service charge, the average wage for cashiers and back-office workers (from the U.S. Bureau of Labor Statistics Retail Trade Earnings and Hours), cash theft and fraud as a percentage of cash sales (from the National Retail Security Survey), the fraud and chargeback rate for cards (derived from Hayashi et al. 2018 and FRPS), and the average terminal rental cost (using the low end of $30 to $100 per month listed on merchant acquirers’ websites). We then calculate the average fixed cost and proportional cost per transaction for each of the three payment methods.” Hayashi, et al., Distributional Effects, supra note 37, at 59.

[40] Fumiko Hayashi & William R. Keeton, Measuring the Costs of Retail Payment Methods, 97(2) Econ. Rev. Fed. Res. Bank of Kansas City 38 (2012).

[41] See Todd J. Zywicki, The Economics of Payment Card Interchange Fees and the Limits of Regulation, (ICLE Financial Regulatory Program White Paper Series, Jun. 2, 2010), at 36-38, available at https://laweconcenter.org/images/articles/zywicki_interchange.pdf.

[42] Id. at 16-18.

[43] See Hans Brits & Carlo Winder, Payments Are No Free Lunch, 3(2) DNB Occasional Studies 27 (2005).

[44] See Nicole Jonker, Social Costs of POS Payments in the Netherlands 2002-2012: Efficiency Gains from Increased Debit Card Usage, 11(2) DNB Occasional Studies 32 (2013).

[45] Olaf Gresvik & Grete Øwre, Costs and Income in the Norwegian Payment System 2001. An Application of the Activity Based Costing Framework, (Working Paper No. 2003/8, Norges Bank, Sep. 17, 2003), at 1, available at https://hdl.handle.net/11250/2498619.

[46] Todd J. Zywicki, Geoffrey A. Manne, & Julian Morris, Price Controls on Payment Card Interchange Fees: The U.S. Experience, (ICLE Financial Regulatory Research Program White Paper 2014-2), at 5-8, available at https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2446080 (noting the loss of free banking accounts, higher monthly maintenance fees, and average minimum holdings required to avoid fees post-Durbin amendment).

[47] See, Cashier Salary in Tennessee, Indeed (last accessed Jul. 24, 2024), https://www.indeed.com/career/cashier/salaries/TN?from=top_sb (salaries based on averages posted on the website on Jul. 24, 2024).

[48] Contactless Payments: Delivering Merchants and Customer Benefits, A Smart Card Alliance Report (2004).

[49] Elizabeth Klee, Paper or Plastic? The Effect of Time on the use of Checks and Debit Cards at Grocery Stores (Finance and Economics Discussion Series, No. 2006-02, Washington Board of Governors of the Federal Reserve System).

[50] Guy Quaden, Costs, Advantages And Disadvantages of Different Payment Methods, Report, Bank of Belgium (2005).

[51] Brits & Winder, supra note 43.

[52] See Michal Polasik et al., Time Efficiency of Point-of-Sale Payment Methods: Empirical Results for Cash, Cards and Mobile Payments, 141 Lecture Notes in Business Information Processing 312 (2013), (Figures given are the mean times for the merchant).

[53] See Greg Buzek, Cash Multipliers: How Reducing the Costs of Cash Handling Can Enable Retail Sales and Profit Growth, IHL Group (2018).

[54] Id. at 7.

[55] Id. at 10.

[56] Id. at 11.

[57] Id. at 10.

[58] Heng Chen, Kim P. Huynh, & Oz Shy, Cash Versus Card: Payment Discontinuities and the Burden of Holding Coins (Bank of Canada Staff Working Paper 2017-47), at ii, available at https://www.econstor.eu/bitstream/10419/197853/1/1011056011.pdf.

[59] Id.

[60] About Legal Tender, Bank of Canada (last accessed Aug. 22, 2024), https://www.bankofcanada.ca/banknotes/about-legal-tender.

[61] Gresvik & Øwre, supra note 45.

[62] Float is working capital; float loss is the cost associated with the time taken for transactions to clear and settle, which means that additional working capital is required to cover outgoings.

[63] David B. Humphrey & Allen N. Berger, Market Failure and Resource Use: Economic Incentives to Use Different Payment Instruments, in The US. Payment system: Efficiency, risk and the role of the Federal Reserve: Proceedings of a symposium on the U.S. payment system sponsored by the Federal Reserve Bank of Richmond (1990), at 45-86.

[64] Kirstin E. Wells, Are Checks Overused?, 20(4) Quarterly Rev. Fed. Res. Bank of Minneapolis 2-12 (1996).

[65] See id. at 4.

[66] GHL, supra note 38.

[67] See id. at 200.

[68] See id. at 201. The authors make certain modifications to the data from FMI, including updating the processing time for transactions and the cost of armored cars.

[69] Id. at 202.

[70] Id. at 204.

[71] Id. at 208.

[72] See id. at 209-12.

[73] See id. at 225.

[74] Garcia-Swartz, et al., supra note 38 at 196.

[75] See Alan Shampine, Another Look at Payment Instrument Economics, Rev. of Network Econ., vol. 6(4), at 495-508 (2007).

[76] Daniel Garcia-Swartz, Robert Hahn, & Anne Layne-Farrar, Further Thoughts on the Cashless Society: A Reply to Dr. Shampine, 6(4) Rev. of Network Econ. 509-524 (2007).

[77] See, e.g., Kerstin Junius, et al., Costs of Retail Payments – An Overview of Recent National Studies in Europe (ECB Occasional Paper No. 294, May 2022), available at https://www.ecb.europa.eu/pub/pdf/scpops/ecb.op294~8ac480631a.en.pdf.

[78] See Brits & Winder, supra note 43, at 27.

[79] Mats Bergman, Gabriela Guibourg, & Björn Segendorf, The Costs of Paying – Private and Social Costs of Cash and Card Payments, at 15 (Sverges Riksbank Working Paper No. 212, Sep. 2007),  available at https://archive.riksbank.se/Upload/Dokument_riksbank/Kat_publicerat/WorkingPapers/WP212.pdf.

[80] Id. at 2.

[81] Brits & Winder, supra note 43, at 27; Layne-Farrar, supra note 20, at 7; Smart Card Alliance, supra note 48; Polasik et al., supra note 52.

[82]  Visa EMV Chip Cards Help Reduce Counterfeit Fraud by 87 Percent, Visa (Sep. 3, 2019), https://usa.visa.com/visa-everywhere/blog/bdp/2019/09/03/visa-emv-chip-1567530138363.html.

[83] Jonker, supra note 44, at 32.

[84] Heiko Schmiedel, Gergana Kostova, & Wiebe Ruttenberg, The Social and Private Costs of Retail Payment Instruments: A European Perspective (ECB Occasional Paper Series No. 137, Sept. 2012), at 12, available at https://www.ecb.europa.eu/pub/pdf/scpops/ecbocp137.pdf.

Introduction

Judge Amit Mehta’s decision in the Google Search case[1] is commendable in many respects. He seems to strive to credit counterarguments wherever doing so is sensible, rather than trying to “bullet-proof” his opinion (as other “Big Tech”-related decisions often do) by discounting every argument put forward by Google. He is not gratuitously dismissive of Google’s experts. And, as he did in dismissing the entirety of the states’ “self-preferencing” claims at summary judgment,[2] he is willing to reject various of the plaintiffs’ arguments here—finding, for example, that some of their proposed relevant markets were not relevant markets and/or that Google was not a monopolist in them. The decision is also very clearly written and, at 275 pages, admirably thorough.

That’s the good. Unfortunately, where it counts the most, Judge Mehta’s decision is seriously lacking, to the point that his primary legal conclusion—that Google’s default search distribution deals were anticompetitive—is untenable. In this paper, I explain why.

I. The Core Legal Defect: Misapplication of the Causation Standard for Exclusionary Conduct

A. Misreading Microsoft

I start with a quote from the decision that brings us directly to the heart of the matter. Quoting the D.C. Circuit Court of Appeals’ decision in Microsoft,[3] Judge Mehta holds that, as a matter of law:

[C]ausation does not require but-for proof. The plaintiff is not required to show that but for the defendant’s exclusionary conduct the anticompetitive effects would not have followed. Such a standard would create substantial proof problems, as “neither plaintiffs nor the court can confidently reconstruct… a world absent the defendant’s exclusionary conduct.” “To some degree, ‘the defendant is made to suffer the uncertain consequences of its own undesirable conduct.’”[4]

But, as Judge Ginsburg—widely regarded to be the primary author of the per curiam Microsoft opinion—has written, that is a misreading of Microsoft.[5] In fact, this decision—this case—is the paradigmatic example of why that is a misreading of Microsoft.

As a general matter, the plaintiff must establish that the defendant’s conduct has the “requisite anticompetitive effect”[6]—that is, that it caused the alleged competitive harm.[7] The “reasonably capable of” standard has “limited applicability”[8] and permits an inference of causation only in special circumstances: 1) when the competitive threat allegedly affected by the defendant’s conduct is nascent; 2) when that conduct was already proven to have anticompetitive effect; and 3) in a government (as opposed to private) enforcement action.[9] “Only when these conditions are met may the government avoid having to show that the threat would have become a real competitor but for the alleged exclusionary conduct.”[10]

The intuition behind this is as follows: There might be multiple reasons—including ones not involving the defendant’s exclusionary conduct—that prevent a competitive threat from materializing. Normally, a plaintiff has to show that the defendant’s conduct—and not one or more of these other factors (like, e.g., that the rival was of such low quality that it couldn’t realistically have mounted a real challenge)—was the but-for cause of the challenger’s failure. But where that threat is inchoate, proving the hypothetical course of future competition in the market is effectively impossible. Thus, in circumstances where we’re confident that lessening the plaintiff’s burden won’t systematically lead to erroneous outcomes, we allow it to make out its prima facie case without demonstrating but-for causation:

The court pointed out that “neither plaintiffs nor the court can confidently reconstruct a product’s hypothetical technological development in a world absent the defendant’s exclusionary conduct.” Given this “underlying proof problem,” the Court may infer causation.[11]

According to Judge Ginsburg’s clarification of what the Microsoft court held, the limited circumstances that satisfy this standard are: 1) as noted, the threat is of the sort that can’t actually be proven—that is, the allegedly thwarted competition must arise from a speculative, but realistic, process that can’t be falsified; 2) the defendant’s conduct must be proven to be, in fact, sufficient to impede the materialization of such a threat (“[o]f critical importance is that the court’s causation standard was conditioned on its having found anticompetitive effects”[12]); and 3) the plaintiff must be a government enforcer.[13]

Obviously, the Google Search case involves a government enforcement action. But for Microsoft’s lighter causation standard to apply, it must also involve a nascent threat and conduct proven sufficient to prevent rivals from achieving minimum efficient scale. Arguably, neither is true in Google Search.

The holding in Microsoft that causation need not be perfectly proven was a function of those specialized facts. It was not, contra Judge Mehta’s approach to this case (and that of many others before him), “a matter of general tendency.”[14]

1. Causation may be inferred only when the competitive process is speculative

The key question in Microsoft was the foreclosure of competition by a nascent competitor—where the entire theory of the case was built on a set of suppositions about the progress of technology, speculation about the unpredictable role the nascent competitor could play in disrupting competition in an established market, and uncertainty about whether Microsoft viewed it (Netscape Navigator) as a competitive threat.

That is not really the case in Google Search. Bing, obviously, and Yahoo! and others before and after it are/were not nascent competitors. Nor is Bing bringing an innovative disruption to the market that will follow an unknown course. Rather, it is a direct, close substitute for Google Search. And, of course, this is obvious to Google. It is distributed the same way; it is used the same way; it is not unknown or uncertain in its competitive relationship with Google Search. None of which means it was actually a competitive threat to Google (more on this later). But it does mean that the competitive process is well understood.

Why does this matter? Because it means that much less speculation is required about whether and how Bing could act as a competitive constraint on Google. We do have actual competition and consumer behavior to assess in understanding the extent of its competitive threat and whether Google’s challenged conduct impaired it. And we do not have to create an entirely hypothetical world in which establishing causation would be impossible.

None of that was true in Microsoft.

We know this, in part, because of another case the same court decided a few years after Microsoft: Rambus v. FTC.[15] With respect to the speculative extent of the but-for world and the viability of demonstrating causation, the facts in Rambus are more similar to the facts here than are the facts in Microsoft. In Rambus (which involved the selection by a standard-setting organization (SSO) of technology to be included in an industry standard) competition was direct and well-understood. The sole question was whether Rambus’s conduct (deception over its patent holdings, which led to the inclusion of its technology in the industry standard) enabled its technologies to monopolize the relevant markets to the exclusion of its rivals, or whether, had it disclosed and the SSO obtained assurances it would license its technology on RAND terms, it would still have obtained its dominant market position.

Of course, in Rambus—as everywhere—it was impossible to truly know the but-for world (i.e., what Rambus’s market position and that of its competitors would have looked like under different licensing terms). But the competitive process by which such an outcome could arise was straightforward, and its competitive alternatives were known. The same could not really be said of Microsoft.

B. Misreading Rambus

In Rambus, the question was whether Rambus’s behavior led to the exclusion of rivals, or whether, even absent its behavior, rivals would have been excluded. That is the identical question that should have been asked here: Was Bing’s failure to gain more market share proved to be a function of Google’s distribution deals, or was it a function of consumer and distributor preferences for Google over its rivals?

As the court in Rambus found, the FTC’s reasoning (from which the court heard the appeal) was logically flawed, in exactly the same way the court’s reasoning is flawed in this case. It is therefore odd that Judge Mehta distinguishes Rambus and rejects the applicability of its legal standard on grounds that are superficial and unrelated to the relevant question of when a given legal standard should apply.

The consequence of this decision for the Google Search case was enormous. By relieving the plaintiffs of having to show but-for causation, Judge Mehta relieved them of the burden of proving their case. This is exactly why Rambus is so important.

1. What Rambus Says

So, here’s what the court said in Rambus. First, the court laid out the standard:

The critical question is whether Rambus engaged in exclusionary conduct, and thereby acquired its monopoly power in the relevant markets unlawfully.

To answer that question, we adhere to two antitrust principles that guided us in Microsoft. First, “to be condemned as exclusionary, a monopolist’s act must have ‘anticompetitive effect.’ That is, it must harm the competitive process and thereby harm consumers. In contrast, harm to one or more competitors will not suffice.” Microsoft, 253 F.3d at 58…. Second, it is the antitrust plaintiff—including the Government as plaintiff—that bears the burden of proving the anticompetitive effect of the monopolist’s conduct. Microsoft, 253 F.3d at 58-59.[16]

Applying these principles, however, the implication of the FTC’s argument was ambiguous:

The Commission’s conclusion that Rambus’s conduct was exclusionary depends, therefore, on a syllogism: Rambus avoided one of two outcomes by not disclosing its patent interests; the avoidance of either of those outcomes was anticompetitive; therefore Rambus’s non-disclosure was anticompetitive.[17]

The Rambus court acknowledged that the first of these possible outcomes would be anticompetitive,[18] but on the second it found the causal link unclear.[19] And because this alternative was not inherently anticompetitive, the Rambus court rejected the FTC’s argument based on its failure to prove that Rambus’s conduct, and not simply its inclusion in the standard (even on less-favorable terms), led to its market position:

Here, the Commission expressly left open the likelihood that [the SSO] would have standardized Rambus’s technologies even if Rambus had disclosed its intellectual property. Under this hypothesis, [the SSO] lost only an opportunity to secure a RAND commitment from Rambus. But loss of such a commitment is not a harm to competition from alternative technologies in the relevant markets….

…Thus, if [the SSO], in the world that would have existed but for Rambus’s deception, would have standardized the very same technologies, Rambus’s alleged deception cannot be said to have had an effect on competition in violation of the antitrust laws.[20]

If the court was correct that only one of the two outcomes was anticompetitive, then its conclusion was inescapable. As Josh Wright has long (since 2009) maintained, “the D.C. Circuit’s causation standard [in Rambus] should not be controversial and appears eminently reasonable.”[21]

Both the Commission and the D.C. Circuit accept that there must be a causal showing that deception significantly contributes to some anticompetitive effect. The disagreement is over whether both possible paths actually involve anticompetitive effects. If one agrees with the Commission that both causal paths violate Section 2, a requirement that a plaintiff specify precisely which path resulted in an anticompetitive effect is unnecessary and likely unwise. However, if one believes that only one causal path constitutes a violation of Section 2, such a requirement is necessary….[22]

Importantly, the Rambus court held this despite recognizing that Rambus’s deception made the inclusion of its technology in the standard “somewhat more likely.”[23] That wasn’t enough, because the FTC failed to show that that wouldn’t have happened even without—that is, but for—Rambus’s deception. “The critical point is that the Commission bore the burden of demonstrating that Rambus’s deception caused the unlawful acquisition of monopoly power.”[24]

C. Misconstruing the Legal Standard Under Microsoft and Rambus

Here, we can rewrite Rambus’s conclusion using the facts of the Google Search case and readily see its applicability:

Thus, if [distributors like Apple and Mozilla], in the world that would have existed but for [Google’s default distribution deals], would have [chosen] the very same [default search provider], [Google’s conduct] cannot be said to have had an effect on competition in violation of the antitrust laws.[25]

This is not how Judge Mehta assesses the Rambus decision, however. Instead, he distinguishes it on the tenuous grounds that it involved a different type of exclusionary conduct, in a different factual setting, and that, as a result, it is inconsistent with his reading of Microsoft:

Rambus does not establish a categorical rule that the anticompetitive effects of an exclusive agreement must be measured against a but-for world. That case involved deception to a standards-setting organization, a form of exclusionary conduct particularly susceptible to a finding of materiality…. In such circumstances, the D.C. Circuit deemed it appropriate to demand proof that Rambus’s deception in fact resulted in competitive harm. Nowhere, however, did the court suggest that such a strict standard of proof was required to demonstrate anticompetitive effects for other forms of exclusionary conduct, particularly exclusive dealing arrangements. Such a holding would be contrary to Microsoft, and the court in Rambus nowhere questioned that precedent. Rambus therefore does not require Plaintiffs to prove substantial foreclosure against a but-for world.[26]

But “deception to a standards-setting organization” is not a circumstance “particularly susceptible” to a but-for analysis compared to that of default search distribution deals. Indeed, it has more in common with the circumstances here than Microsoft does. That’s because, as noted, the but-for world in that case is easy to understand and analyze (even if, as in all cases, the but-for world is never a simple or certain calculation). It is in the Google Search case, as well. The but-for world in this case, as in Rambus, involves an essentially binary choice among known alternatives with a direct line between the relevant conduct and that choice. In Rambus, it was a choice by an SSO between two licensing regimes (one less restrictive, and one more restrictive) for Rambus’s patents in its standards, and Rambus’s deception could clearly affect that choice. Here, it is a choice by users between (essentially) two search engines, and the choice of default search provider can clearly affect that choice (because the cost of using the non-default is inherently higher).

Moreover (and as noted above), according to Judge Ginsburg, Rambus is not a special exception to Microsoft’s general rule; Microsoft is a special exception to Rambus. It is the nascency of the threat in Microsoft (which didn’t exist in Rambus) that leads to its uniquely truncated analysis and not the specific context of Rambus that somehow cabins its applicability:

Reading Microsoft and Rambus together, the key takeaway is that only when anticompetitive effects are shown (as required by Microsoft and Rambus) does the “reasonably capable of” causation standard apply to allegations that exclusionary conduct killed a nascent threat. Only when these conditions are met may the government avoid having to show that the threat would have become a real competitor but for the alleged exclusionary conduct.[27]

Nevertheless, misconstruing the legal standard under Microsoft and Rambus, Judge Mehta holds that Microsoft’s lighter, “reasonably capable of” standard applies:

The key question then is this: Do Google’s exclusive distribution contracts reasonably appear capable of significantly contributing to maintaining Google’s monopoly power in the general search services market? The answer is “yes.” Google’s distribution agreements are exclusionary contracts that violate Section 2 because they ensure that half of all GSE [general search engine] users in the United States will receive Google as the preloaded default on all Apple and Android devices, as well as cause additional anticompetitive harm.[28]

D. How We Know It Is Wrong

It is immediately obvious why, even if it weren’t a misreading of the case law, this cannot possibly be the correct standard—and why it makes no sense to suggest that a less strict standard of proof is particularly appropriate for exclusive-dealing arrangements.

We don’t need to ask if the agreements “reasonably appear capable of significantly contributing to maintaining Google’s monopoly power” because of course they are “reasonably… capable” of contributing to Google’s monopoly power. That’s why we have scores of antitrust cases looking at the effects of distribution agreements. If all that were required to win such a case were the reasonable capability of an agreement to contribute to a dominant firm’s competitive position, then no exclusive or quasi-exclusive agreement would ever be legal.[29]

We ask whether exclusive agreements “reasonably appear capable” of maintaining monopoly, instead of asking whether they actually maintained monopoly, only when the connection between what is being excluded and monopoly maintenance is unclear—as in Microsoft, where it was unclear if Netscape Navigator could actually constitute a competitive threat to Microsoft’s operating-system dominance. But here that is not a question.[30] Where the rival is a direct competitor with a close substitute product, an exclusive deal by a dominant incumbent is always capable of foreclosing the rival. In such circumstances it is simply not consistent with the plaintiff’s burden of proof to allow them to show only that the challenged conduct is the sort that could maintain monopoly, rather than that the defendant’s conduct in fact caused anticompetitive harm.

Difficult as it may be, demonstrating this in an actual competition case like Google Search is not impossible. Indeed, as I discuss below, there is copious evidence in the case that the cause of Bing’s limited market share was not Google’s default distribution deals, but Bing’s lack of quality. Any comparable evidence in Microsoft would have been wholly speculative—but not here.

As Judge Ginsburg notes (again challenging the general applicability of Microsoft’s truncated legal standard):

As in Microsoft, the “but-for” world in Rambus was highly uncertain.[[31]] In both cases, one could reasonably find the defendant’s conduct may have caused the defendant to acquire or maintain its monopoly power. At the same time, it was also possible that the defendants in those cases would have acquired or maintained their monopoly power even absent their anticompetitive behavior. The court in Rambus held the government must bear the burden of that uncertainty. This burden applies in all Section 2 cases….[32]

But it is also arguably the case that, properly construed, Google’s default distribution deals were not capable of excluding Bing from the market. We turn to this next.

II. The Failure To Prove That Defaults Are Exclusive

Judge Mehta holds that Google’s distribution deals had anticompetitive effects “because they ensure that half of all [general search engine] users in the United States will receive Google as the preloaded default on all Apple and Android devices.”[33] He derives this conclusion (which he repeats several times) from the testimony of one of the plaintiffs’ economic experts, Michael Whinston, who finds that “50% of all queries in the United States are run through the default search access points covered by the challenged distribution agreements.”[34]

Judge Mehta’s claim is that, because users don’t switch away from defaults very often, and because the “market realities” of the search market—given Google’s default distribution deals—are that half of all relevant searches occur through access points covered by those deals, we can conclude that those deals foreclose competitors from 50% of the general search market (a big enough amount to constitute anticompetitive foreclosure).

But this is not how you measure foreclosure, and the assertion that that number shows that Google’s default deals—and not something else—“significantly contribut[e] to maintaining Google’s monopoly power”[35] is fallacious.

Just because the government shows there is “significant” usage of Google’s default services ex post—meaning, given Google’s default deals, and after consumers have chosen which search engine to use—does not mean it has proven that 50% of the market was foreclosed from access by competitors. Nor does it mean that the government has met its burden of proving that this was caused by the agreements. Perhaps all of those consumers are inframarginal consumers who would have chosen Google Search anyway, even if it weren’t the default. Perhaps all of them were perfectly capable of accessing Bing, but simply chose not to. In that case, the ex-post usage data would tell us nothing about the extent of foreclosure.

Demonstrating foreclosure requires comparison to the but-for world; it requires showing that, absent Google’s deals, Bing would have had access to and been used by substantially more marginal consumers (those who view Bing and Google as effective substitutes and wouldn’t expend extra cost to use one search engine or the other).[36] This is so for three reasons.

First, these agreements are not, in fact, “exclusive.” That matters, because it is much harder to infer that it was the agreements, and not consumer preferences for a particular product, that caused Google’s dominance when consumers have ample opportunity to exercise their preferences to not use Google Search.

Second, and relatedly, looking at the share of searches ex post that go through these defaults is less telling, and the number can’t simply be accepted at face value, when searching via the default service is not the only option. As in Rambus, Google’s maintenance of a large share of these searches is just as consistent with a non-problematic set of facts (consumers simply prefer Google Search and, knowing that, distributors offer Google as the default) as with a problematic one (consumers use Google Search only because it is the default service, and distributors offer Google Search as the default only because Google pays them to do so).

Finally, failing to demand that the plaintiffs demonstrate that Google’s default distribution deals actually foreclosed competitors, and allowing them effectively to prove their case by showing only that the deals made Google’s large market share “somewhat more likely,”[37] erases the requirement that plaintiffs can win only if they prove that challenged conduct causes anticompetitive harm. This is an invitation for dramatically erroneous decisions.[38] As Judge Ginsburg and Koren Wong-Ervin write:

Without requiring proof of but-for causation, there is great risk of erroneously condemning [conduct] that may be procompetitive. Consider, for example, Herbert Hovenkamp’s proposal presumptively to condemn acquisitions by a monopolist of “any firm that has the economic capabilities for entry and is a more-than-fanciful possible entrant, unless the acquired firm is no different from many other firms in these respects.” “More-than-fanciful” is an invitation to speculate, not a standard of proof.[39]

Let’s examine these problems with Judge Mehta’s legal standard based on Michael Whinston’s market-share numbers a little more thoroughly.

A. Establishing the Amount of Competition: Minimum Efficient Scale

So much of Judge Mehta’s conclusion that Google’s default deals were exclusive (despite not actually being “exclusive”) turns on his contention that defaults are the best way for search engines to distribute themselves,[40] and, therefore, that Google tying up default access was sufficient to establish the requisite exclusivity for an exclusive-dealing claim.

But as the Rambus court points out, “conduct [does] not violate antitrust laws where absent that conduct consumers would still receive the same product and the same amount of competition.”[41] This is a statement that the but-for world matters, and the relevant question is the relative “amount of competition.”

So, how do we measure the “amount of competition”? Well, we can’t measure it as Judge Mehta and Prof. Whinston do, by looking at what people choose ex post.[42] This is a fairly useless statistic. It says next to nothing about what share of marginal searchers use Google because it is the default, and what share use Google because they prefer it. (And, of course, it says nothing about what share are inframarginal consumers who would use Google regardless of the cost of accessing it). That, in turn, tells you nothing about the amount of competition that existed before they made their choices.[43]

Nor can we simply look at the scope of the default distribution agreements. If people can still easily choose other search providers despite the default deals, those deals cannot be said to foreclose competition—at the very least, not in proportion to the share of the market covered by those deals.

We can, however, examine whether rival search providers were able to achieve minimum efficient scale in such an environment—that is, whether the conduct at issue was capable of precluding otherwise viable competitors from gaining enough customers to sustain themselves as a competitive alternative. Indeed, this is exactly what Google argued was required: “[Google] contends that Plaintiffs have failed to establish a link between the agreements, the denial of sufficient scale to rivals, and anticompetitive effects….”[44] It is also what Judge Ginsburg says is required: “The court [in Microsoft] inferred harm to the competitive process from these findings, in essence recognizing that minimum-efficient scale is the mechanism by which exclusionary conduct harms competition.”[45]

Yet nowhere does Judge Mehta effectively grapple with the minimum-efficient-scale question. He does discuss the importance of scale in search, and he holds that Google is of higher quality than Bing and other competitors, in part, because of its scale. But he never really asks or answers the questions 1) whether this difference is uniquely attributable to Google’s default distribution deals, and 2) whether those deals preclude rivals from effectively competing, or simply make it harder for them to compete because they raise the cost of achieving comparable quality.

Now, the evidence here, as far as we can assess it from the decision, is not entirely clear-cut. But the answer isn’t really the issue. The real issue is that this is an essential question, on which the government bears the burden of proof, and it was simply missing from the opinion. In other words, this means the holding in the government’s favor is unsupported as a matter of law.

B. De Facto Exclusivity

Even so, let’s assess how well the evidence supports the conclusion that the default deals were really “de facto” exclusive, and that they prevented rivals from achieving the minimum efficient scale.

We have no idea if the default deals had anything to do with it, but we do know that Neeva, a once-promising general search engine, apparently had a hard time competing for users and went out of business after about four years.[46] On the other hand, Bing, Yahoo!, DuckDuckGo, Ecosia, and Brave all exist and continue to compete in this environment. Yes, they have relatively small market shares, but apparently they have enough scale for viability.

Similarly, on the one hand, there has been clear “competition for the market” between Bing and Google with respect to the default access points on Apple devices and in Mozilla’s Firefox browser. On the other hand, there is less clear competition with respect to Android OEMs (in Google’s favor)—but there is also less clear competition (actually, there is none) between them for default placement in Edge (in Bing’s favor).

1. Misunderstanding the ‘power of defaults’

With respect to the conclusion that the cost to users of choosing the non-default option is higher, that is inherently true, of course. But it is arguably trivially so. Judge Mehta spends a fair amount of time on this question (although not in the proper context of this but-for assessment) before arriving at his conclusion that being a non-default is tantamount to being excluded. His analysis, however, is unconvincing.

First, the analysis is heavily influenced by the assertions of the government’s behavioral expert, Antonio Rangel.[47] As I will discuss below, some empirical data specific to the context at hand is used to bolster the more general behavioral claims of the government’s expert (which I believe cuts in many respects against Judge Mehta’s conclusions). But it is clear that any ambiguity was resolved by Judge Mehta in favor of asserted general behavioral patterns:

That users overwhelmingly use Google through preloaded search access points is explained in part by default bias or the “power of defaults.” The field of behavioral economics teaches that a consumer’s choice can be heavily influenced by how it is presented. The consensus in the field is that “defaults have a powerful impact on consumer decisions.”[48]

There are access points other than the default that can be used to distribute a GSE, but those channels are far less effective at reaching users. That is due in part to users’ lack of awareness of these options and the “choice friction” required to reach these alternatives.[49]

[A]s Dr. Rangel convincingly explained, the combination of user habit, Google’s brand, and choice friction creates a powerful default effect that drives most consumers to use the default search access points occupied by Google.[50]

The main problem with this is not so much that behavioral science is wrong (surely, it is correct that the more friction there is to switch from a default, the less likely someone will switch), but that it is not dispositive. This makes it a weak basis for meeting the plaintiffs’ burden. It is also not clear that general behavioral theories have the same traction in the specific environment at issue. As my colleague Dan Gilman has discussed, the learnings of behavioral science were established in settings quite different than search engine defaults.[51] “Generalizing findings from, e.g., cereal-box placement to the durability of search engine defaults seems a stretch (or entirely speculative).”[52]

To be sure, Rangel’s testimony did purport to apply those learnings in context.[53] But what really matters in this case is not the direction of the behavioral assumptions, but the magnitude. (Again, no one disputes that defaults grant some benefit, nor that promoting one’s products—i.e., marketing—can influence consumer choice). The claim here is that the availability of switching does not sufficiently negate the effective exclusivity of defaults to permit rivals to compete. That claim depends on the extent to which users tend not to switch away from defaults, not just the fact that they sometimes don’t.

Among other things (more of which are discussed below), it must be noted that, even when users are presented with a neutral option (e.g., a “choice screen”), they appear to make essentially the same choices as when presented with a default. In Europe, where Google has since 2020 implemented a search engine choice screen on Android following the EU’s 2018 antitrust decision against it,[54] Google’s share of the search engine market has barely budged.[55]

By the same token (at least when Google is the non-default) users are apparently quick to switch from a less-preferred default in order to get access to Google Search:

In a 2016 experiment, Mozilla switched the default GSE on both new and existing users from Google to Bing. By the twelfth day, Bing had kept only 42% of the search volume. After some additional time, those numbers dropped to 20–35%….[56]

It is exceedingly difficult to square these facts with the court’s conclusions on the functional irrelevance of non-default options.

C. Overlooking Evidence of User Switching Behavior and Impressionistic, Not Quantitative, Conclusions

The majority of the decision’s discussion of consumers’ behavior around defaults is largely impressionistic, not quantitative. For example, Judge Mehta notes that:

Another non-default search access point is the bookmarks page on a browser. The Safari “Favorites” page, for instance, contains preloaded icons to access Google, Bing, and Yahoo. A user also can add a new search engine on that page. But few consumers use this channel, as it first requires finding the Favorites page in a new Safari tab, which requires an “extra click.” Google itself receives only 10% of its searches on Safari through the bookmark.[57]

Strictly speaking, 10% is “quantitative,” but the decision’s conclusions based on this data are decidedly impressionistic. Judge Mehta asserts that Google receiving “only” 10% of searches on Safari through the bookmark is an insignificant volume. But in that setting—where Google Search is already the default on Safari and can be accessed simply by typing in Safari’s URL bar, and in which it is alleged that virtually no one ever uses anything other than default search on Safari—why would there be any searches on Google via the bookmark? If that number of searches is at all different from zero, it would appear to demonstrate that it is indeed a relevant channel by which consumers can find search engines, including non-default ones. In other words, 10% may be “insignificant” as a share of Google searches, but it is quite significant with respect to the relevant legal standard.

Indeed, “nearly 40% of queries on Apple’s mobile devices flow through non-default search access points, such as default bookmarks or organic search.”[58] Judge Mehta dismisses this by arguing that “the fact that some consumers access search on non-default access points is not dispositive on exclusivity.”[59] “Not dispositive” is not quantitative. While the statement is true, the burden of proof is on the plaintiffs, and this not being dispositive cuts against them, not against Google.

Elsewhere, Judge Mehta also rejects the actual evidence of people switching to the non-default on PC desktops. It turns out that a lot of Windows desktop users download Chrome and use Google Search there, rather than relying on Bing, which is the default search engine in the Edge browser:

To be sure, downloads of an alternative browser occur with greater frequency on Windows desktop computers. On such devices, Edge is the default browser and Bing is the default search engine. Yet, Google’s search share on Windows devices is 80%, with most of the queries flowing through the Chrome default, which means Chrome was downloaded onto the device.[60]

Despite this, Judge Mehta is quick to note that, for those users who still use Edge, Bing is the most used search engine:

The power of defaults is evident, however, from the share of Bing users on Edge. Bing’s search share on Edge is approximately 80%; Google’s share is only 20%. Even if one assumes that some portion of those Bing searches are performed by Microsoft-brand loyalists, Bing’s uniquely high search share on Edge cannot be explained by that alone. The default on Edge drives queries to Bing.[61]

One might suggest that all this shows is that people really prefer Chrome to Edge, not that they prefer Google Search to Bing enough to switch away from the default (on either browser). Except that, as the opinion points out, “Google’s dominance on Windows cannot, however, be attributed simply to the popularity of Chrome. Google had an 80% search share on Windows when Chrome first launched, and that share has remained steady ever since.”[62] If that’s the case, it can mean only that the default search on Windows desktops isn’t very sticky—and it isn’t just because users prefer Chrome to Edge; apparently it’s because they prefer Google Search to Bing.

So how does the court conclude that it supports the “power of defaults” that, of those users who don’t switch to Chrome on Windows desktops, approximately 80% use Edge’s default? If most Windows users who prefer Google Search to Bing switch from the default by downloading Chrome instead of by switching the default in Edge, then of course most of those who remain on Edge will use Bing. If they preferred Google to Bing, they would have switched to Chrome.

As Judge Mehta notes elsewhere, “[m]any users do not know that there is a default search engine, what it is, or that it can be changed.”[63] Perhaps. But then again, apparently, many users do know that Chrome gets them access to their preferred search engine. Whatever “choice friction” impedes the movement away from the default search engine on Windows desktops, it is not strong enough to prevent people from maneuvering around it in spades—they just don’t often do so directly by switching the default search engine in Edge.[64]

The opinion also brushes off these examples of default switching by asserting that they merely “confirm that the default effect is weaker when the alternative is a dominant firm with high brand recognition backed by a quality product.”[65] First, this is pretty hand-wavy and impressionistic. Maybe it’s true; in fact, I’m sure it’s true to some extent. But for an opinion that otherwise regularly says we have to look at “market realities,” not the world as it might be, this is a weak basis to conclude that evidence of people switching away from defaults doesn’t really show that people switch away from defaults.

Regardless—isn’t the ability to attract users because you are widely used, have good brand recognition, and have a demonstrably high-quality product pretty much the definition of competition on the merits? Indeed, one could recast Judge Mehta’s statement as precisely the opposite of the decision’s holding: Google’s default agreements can’t be deemed to have caused anticompetitive harm because defaults are readily overcome by high-quality, reputable alternatives.

It should also be noted (but, unfortunately, Judge Mehta doesn’t say it) that Microsoft is also a “dominant firm with high brand recognition backed by a quality product.”[66] What’s good for the goose is good for the gander: Microsoft has plenty of market heft to ensure that its products don’t languish in obscurity in the face of consumer inertia.[67]

III. The Scale and Quality Argument: A Double-Edged Sword

All of which raises the question: Is Bing a comparably high-quality product or not? Determining that seems like a pretty important prerequisite to determining whether its small market share is a function of anticompetitive exclusion or a failure to compete on the merits. Yet Judge Mehta is, at best, equivocal on this. First, he notes that:

Everyone agrees that Google’s distribution agreements did not cause Microsoft’s past underinvestment in search. Microsoft “missed” the mobile revolution and was unable to improve its browser, Internet Explorer, until it used Google’s rendering engine, Chromium. Some of Microsoft’s quality issues also were attributable to its poor index.[68]

Yet, “[u]ltimately, Microsoft committed significant capital to search.”[69] And “[t]hat investment (combined with secured distribution on Windows devices) has allowed Bing to achieve quality parity with Google on Windows desktop devices.”[70]

Elsewhere, however, Judge Mehta concludes that “Google’s exclusive agreements… deny rivals access to user queries, or scale, needed to effectively compete,”[71] and that “[t]his perpetual scale and quality deficit means that Microsoft has no genuine hope of displacing Google as the default GSE on Safari. As Apple’s Eddy Cue testified, there was ‘no price that Microsoft could ever offer [Apple]’ to prompt a switch to Bing, because it lacks Google’s quality.”[72]

I admit that it’s unclear to me why Bing’s apparent quality parity in desktop search doesn’t redound to its benefit in mobile search. Indeed, it has to be noted that the court did not identify separate relevant markets for mobile and desktop search; it identified a single “general search services” market.[73] So, it’s a little unclear, but it seems that, according to the court, ultimately Bing simply isn’t up to Google Search’s quality standard.

A. The Failure to Distinguish Between Exclusion and Low Quality: A Catastrophic Legal Blunder

The problem is that it is precisely past decisions and their alleged influence on current outcomes that the court uses to establish the proposition that Google’s default deals are anticompetitive. As I keep pointing out, however—and as Judge Mehta appears here to recognize—plaintiffs cannot meet their burden of proof that Google’s deals were exclusionary by pointing to Bing’s limited success, if the court agrees that Bing’s low quality could also have caused it. And here, Judge Mehta also concedes that those deals didn’t cause Bing’s lower quality, either (“Google’s distribution agreements did not cause Microsoft’s past underinvestment in search.”).[74]

For the court to sustain its claim that Microsoft is the appropriate guiding precedent (and thus, that the government has made its case under Microsoft’s “edentulous” legal standard), it has to be the case that Bing could have outcompeted Google on quality if not for the agreements—that is, that it “reasonably constituted [a] threat.”[75]

By conceding that Bing was unable to secure distribution deals comparable to Google’s because of its low quality, however, the opinion (and the government, of course) fails to do this. As such, they fall right into the trap explained by Greg Werden:

But if operating at a much smaller scale than Google makes rival search engines uncompetitive, their fate was sealed when Google achieved a dominant share. The government posits no scenario in which any rival search engine could have substantially closed the scale gap…. If the government’s scale contentions are fully credited, the conduct that is at the heart of the case did not maintain Google’s dominant share. And any conduct that could not have maintained dominance most likely served a legitimate purpose. One way or another, the elements of the monopolization offense cannot be established under the government’s view of the facts…. But the government does not contend that rival search engines ever posed a real threat to Google’s monopoly. Indeed, it claims to have proved just the opposite.[76]

That’s a catastrophic problem for the opinion’s holding. Nevertheless, Judge Mehta does find that Bing is not a viable competitor on mobile. Yet he refutes Google’s claim that this is because of Microsoft’s own business failures, rather than its inability to gain scale:

Google also maintains that the quantity of user data is less important than how it is used, and if its rivals had Google’s business foresight and drive to innovate, they too could win default distribution. But that position blinks reality. Apple’s flirtation with Microsoft best illustrates this point. Microsoft has invested $100 billion in search in the last two decades and its quality now matches Google’s on desktop search. Yet, Microsoft’s failure to anticipate the emergence of mobile search caused it to fall behind, and with Google guaranteed default placement on all mobile devices, Microsoft has never achieved the mobile distribution that it needs to improve on that platform.[77]

Isn’t Microsoft’s “failure to anticipate the emergence of mobile search” precisely the sort of competitive failure that Google is talking about? How is Microsoft’s diminished scale attributable to Google’s conduct if it was Microsoft’s independent business decisions that denied it the ability to compete effectively?

This is exactly why a plaintiff must prove that the defendant’s conduct, and not an excluded rival’s own mistakes, were the cause of the rival’s inability to compete. Otherwise, the law would be enlisted to rectify competitors’ poor business decisions, rather than to protect the competitive process.

1. Even Judge Mehta knows ‘reasonably appears capable of’ is the wrong standard

It also bears noting that Judge Mehta already—and properly—threw out exactly this sort of claim on summary judgment when he dismissed the plaintiff states’ claims that “Google’s targeting of SVPs [specialized vertical providers] caused anticompetitive effects in the proposed markets.”[78] But the basis on which he did so is shockingly at odds with the basis for his decision in the government’s favor in this case. Citing Microsoft, in fact, Judge Mehta held in his summary judgment opinion that:

Speculation that Google’s conduct “can reasonably be expected,” “might,” or “could potentially” degrade SVPs and make them less attractive partners to Google’s rivals is not evidence of anticompetitive effects in the relevant markets. Plaintiffs are required to show with proof “that the monopolist’s conduct indeed has the requisite anticompetitive effect,” and they have fallen well short.[79]

And, as he notes elsewhere in his summary judgment opinion, also citing Microsoft, “[t]he sole issue for the court to resolve is whether Google has maintained monopoly power in the relevant markets through ‘exclusionary conduct’ as opposed to procompetitive means.”[80]

The words “can reasonably be expected”—rejected by Judge Mehta in his summary judgment decision—might ring a bell, as they are awfully close to the “reasonably appear capable of” standard adopted by the court in this decision.

B. Less-Efficient Channels of Distribution: Misapplying Microsoft Again

Finally, there is another problem with the legal sufficiency of the exclusivity claims, and it stems, once again, from a misapplication of Microsoft.

Judge Mehta claims that “mere user access to these less efficient channels of distribution does not render the browser agreements non-exclusive.”[81] A significant part of the defense of this position turns on an analogy to Microsoft and the argument there that it was sufficient that Microsoft foreclosed access to the best method of distribution. Indeed, the next sentence after the quote above is, “Microsoft again illustrates the point.”[82] But does it?

Judge Mehta says this case is the same as Microsoft where the court “reject[ed] the argument that Microsoft’s licensing agreements with OEMs were not exclusive ‘because Netscape is not completely blocked from distributing its product,’ as ‘although Microsoft did not bar its rivals from all means of distribution, it did bar them from the cost-efficient ones.’”[83] He then asserts that “[t]he record here resembles that in Microsoft. Users are free to navigate to Google’s rivals through non-default search access points, but they rarely do.”[84]

But this elides a couple of key things.

First, the Microsoft court didn’t look ex post at what consumers did (which, as I tire of pointing out, could be attributable to either anticompetitive conduct or consumer preferences); it looked at which channels of distribution were available and if they were viable substitutes, regardless of whether they were actually used or not.

The analogy to Microsoft fails most obviously on the point that the “market realities” have changed a lot since the late 1990s. Downloading Netscape from the internet was wholly unfamiliar, exceedingly complex, and truly difficult for PC users back then—a real “choice friction” and thus not really a viable alternative. But downloading a competing search engine or browser today is trivially easy, and users do it all the time (to the tune of 12.6 billion app downloads in the United States in 2023 alone).[85] In this environment, the fact that users don’t download or use competing general search engines sufficiently to displace Google Search despite the ease of doing so suggests that it is consumer preference for Google Search, not the relative inefficiency of the channel of distribution, that causes this result.

Instead, Judge Mehta concludes that, while “a user can download Chrome, Edge, or [DuckDuckGo] onto an Apple device,” “[t]his, too, is not an easily accessible search point, as it involves similar choice friction as acquiring a search application. Google receives only 7.6% of all queries on Apple devices through user-downloaded Chrome.”[86]

Not only is downloading an application trivially easy, but the fact that Google receives only 7.6% of search queries on Apple devices through Chrome, but “most”[87] of its queries on Windows desktops through user-downloaded Chrome is decidedly ambiguous. Maybe that shows that people download Chrome on Windows not to get easy access to Google Search but because the Chrome browser is superior to the Edge browser, while it is not any better than Safari. But it is also consistent with the conclusion that people aren’t prevented from accessing their preferred search provider (Google Search)—they just don’t need to download Chrome on Apple devices to get easy access to it, while they do need to do so on Windows devices.

Second, the opinion says that “[t]he court in Microsoft did not say that these contracts caused zero market foreclosure merely because Internet Explorer had other, less-efficient means of reaching users.”[88] True. But the court in Microsoft also didn’t say that any amount of difference in distribution efficiency was sufficient to maintain that a non-exclusive agreement was effectively exclusive. As noted, it is now trivially easy to switch search providers on virtually every platform and at multiple decision points on each. Defaults don’t prevent that, and prioritized placement (from, e.g., a spot on the Android home screen) doesn’t even crowd out alternatives once they are downloaded (which can then be similarly accessed from priority positions on the home screen). “Very slightly less efficient” could still be “efficient.” The fact that the difference between the foreclosed and available channels of distribution in Microsoft was large enough to matter does not mean that the difference between them in Google Search is big enough to matter.

1. So, Dentsply is good law, but Rambus isn’t?

In response, Judge Mehta goes back to ex post user conduct to hold that the fact that users don’t often use these alternatives shows that the difference does matter here, and that Google’s default distribution deals are effectively exclusive and lead to foreclosure:

Sure, users can access Google’s rivals by switching the default search access point or by downloading a rival search app or browser. But the market reality is that users rarely do so. The fact that exclusive agreements allow users to reach rivals through other means does not make the foreclosure number zero.[89]

But it cannot be a sufficient argument that “the market reality is that users rarely do so.” That market reality is exactly what is at issue in the case. Using the lack of user uptake from trivially easy alternative distribution channels as evidence that those alternative distribution channels aren’t relevant assumes the conclusion. It’s poor legal reasoning.

Judge Mehta is correct, however, that “‘[t]he mere existence of other avenues of distribution is insufficient without an assessment of their overall significance to the market.’”[90] If only he had demanded such an assessment.

The Dentsply case that Judge Mehta cites for this proposition was (in my opinion) wrongly decided. It shouldn’t be held up as the standard of analysis and, in any case, it was in the 3^rd U.S. Circuit Court of Appeals and not binding precedent. But even so, Dentsply dealt with exclusive agreements that included a term explicitly prohibiting authorized distributors from selling rivals’ products, thus arguably making it extremely difficult for those products to be accessed by the ultimate consumer. This case is different. None of Google’s agreements include terms prohibiting its counterparties from dealing with anyone else. And here, competing products are available to the ultimate consumer, and they show up on users’ devices in locations virtually identical to Google’s.

In any case, the Dentsply court does not rely on ex-post uptake to support its claim that alternative distribution channels are insignificant (although it does look at that statistic on occasion). Instead, it describes in detail the qualitative differences between the channel of distribution foreclosed by Dentsply and the alternatives, finding that the alternatives are decidedly less attractive. Here, by contrast, the only thing that distinguishes default placement from the other channels of distribution is alleged “choice friction.” Otherwise, they are, quite literally, identical (or, as in the difference between, say, search bar integration and a home-page bookmark, trivially different). That makes assessing their “overall significance to the market” dependent on what is being distributed, and not solely the channel of distribution itself.

2. In fact, we know from other parts of the decision that ‘less-efficient’ alternatives can’t be dismissed

Later, also quoting Dentsply, Judge Mehta asserts that:

In the end, Google’s dismissal of the importance of scale is inconsistent with market realities. Google often warns that competition is “only a click away.” However, “[t]he paltry penetration in the market by competitors over the years has been a refutation of [that] theory by tangible and measurable results in the real world.”[91]

This misses the mark for the same reason. There is plenty of evidence to demonstrate that competition is just a click away. In fact, some of it was evidence the court used to exclude specialized vertical search providers (e.g., Amazon and TripAdvisor) from the relevant market. Without challenging that conclusion here (although I do think it has problems), it appears eminently “tangible and measurable” to the question of whether users will switch to alternative search engines that, when the alternative is demonstrably superior for the query at issue, they do so in droves:

Google views competition from SVPs as “intense for commercial clicks.” A 2020 Bank of America study reported that 58% of users search Amazon first when they seek to make an online purchase, as opposed to only 25% who go first to Google, demonstrating Google’s secondary status as a starting point for users with high commercial intent….

…Microsoft recognizes that “if Bing or Google were not doing vertical searches well, or at least not having organic results that people could click to get to vertical search engines,” users might bypass GSEs and instead search directly on Amazon from the outset….

…[A]nalysis show[s] that a query sample of Google’s top 25 non-navigational shopping queries attracts more queries weekly on Amazon (3.7 million) than Bing (0.4 million)…, [and] that Yelp’s local query volume is higher than Google’s and much higher than Bing’s.[92]

None of these alternative vertical search engines is installed as the default. And yet, when consumer preference is strong enough—when they produce better results—consumers have no trouble using them. Whether or not this is sufficient to affect the court’s relevant market or market-share analysis, it is surely enough to demonstrate that users are not locked into defaults when the “choice friction” required to switch from them is small relative to the benefit. That, in turn, is a function of the quality of the search provider, not the method by which it is accessed.

IV. Getting It Wrong on the Substantiality of Foreclosure, Too

The “substantiality” of foreclosure must also be briefly addressed, for similar reasons. While the opinion downplays its significance as a search engine distribution channel, Windows desktops constitute a substantial share of the distribution market. Windows accounts for 64% of desktop operating systems and almost 30% of all operating systems across all platforms in the United States.[93] On these devices, Bing is the default search engine. So, right out of the box, the share of the market that Google could even possibly foreclose is reduced by Windows’ 30% market share.

Of the remaining 70%, we know that small but non-trivial portions are not actually foreclosed to competitors. We know this from the ex-post data showing, for example, that “5.1% of all searches on iPhones are conducted on a GSE other than Google [where it is the default].”[94] We also know that, on Android, “[a]lthough OEMs must preload the Google Search Widget, users can delete it. As of 2016, there were about 200,000 logged widget deletions daily.”[95] Also, “Samsung already preloads a second browser—its proprietary S browser—on all Samsung devices.”[96]

We also know that “nearly 40% of queries on Apple’s mobile devices flow through non-default search access points, such as default bookmarks or organic search.”[97] Of course, a great number of these searches are performed on Google Search anyway.[98] But these are searches performed by users who demonstrably navigate around the default. By definition, they are not foreclosed to Google’s competitors.

Indeed, simultaneous with the Google default deals, Bing is, in fact, distributed by these counterparties to Google’s deals. Thus, as the result of an agreement with Microsoft, Bing shows up as an option on Safari’s homepage and on the Safari “Favorites” page, “which contains preloaded icons to access Google, Bing, and Yahoo.”[99] Mozilla has a “this time, search with” feature on Firefox “which allows users to select a different search product from its ‘Awesome Bar’ for a given query.”[100]

Again, Android is actually a somewhat unique case, and there the absence of true foreclosure is almost entirely dependent on the availability of end-consumer choice (which, again, is far from irrelevant). But even if we assume zero distribution of Bing on Android devices,[101] it still has unfettered access to distribution on Windows devices and is still distributed alongside Google on Apple devices and in Firefox.

The bottom line is that, even measured by ex-post consumer behavior, rivals are not foreclosed from access to consumers. And measured by the availability of access to rival search engines (whether consumers choose to use them or not), competitors are not actually foreclosed from distribution on any devices or in any browsers at all. To be sure, the remaining effective foreclosure level could be “substantial.” But nowhere does the court’s opinion demonstrate this. As plaintiffs have the burden of proof, the existence of meaningful consumer usage and availability, despite purported exclusive agreements, should have been deemed by the court to undermine the government’s case, not support it.

V. The Fateful Conclusion that Bing Isn’t a Real Competitor and the Problem of Remedy

Finally, I have to say a word on remedy here, although I do so for now only insofar as it bears on what I have been arguing; there are many other arguments about remedy that make this holding problematic. But here is one, and it hearkens back to Greg Werden’s Catch-22.[102]

The jig was up for the plaintiffs in this case once they argued that Bing was not a viable competitor to Google Search. In the world of that “market reality,” no reasonable remedy would do any good to rectify the allegedly anticompetitive circumstances. And the court accepted the government’s quality arguments pretty much wholesale:

The market reality is that Google is the only real choice as the default GSE. Apple’s Senior Vice President of Services, Eddy Cue, put it succinctly when, in a moment of (perhaps inadvertent) candor, he said: “[T]here’s no price that Microsoft could ever offer [Apple] to” preload Bing. “No price.” Mozilla stated something similar in a letter to the Department of Justice prior to the filing of this lawsuit. It wrote that switching the Firefox default to a rival search engine “would be a losing proposition” because no competitor could monetize search as effectively as Google. A “losing proposition.” If “no price” could entice a partner to switch, or if doing so is viewed as a “losing proposition,” Google does not face true market competition in search.[103]

But if “no price” could entice a partner to switch to Bing, and Bing is not truly a competitor to Google in search, then, as Greg Werden says, “the conduct that is at the heart of the case did not maintain Google’s dominant share.”[104]

The Microsoft decision relies on the contention that, although unproven, Netscape Navigator was a viable competitive threat to Windows. Thus, the government had to prove in that case that the threats to Microsoft’s operating-system monopoly were real, even if it didn’t have to prove the threats would have succeeded but for Microsoft’s conduct. The government’s burden is at least as high here.

And yet, in the quote above, Judge Mehta essentially finds that the government didn’t meet even this burden. He finds, in effect, that it wasn’t the nature of Google’s agreements that contributed to Google’s continued monopoly power; it was the fact that no distributor would ever choose Bing as the default—at any price. That conclusion means that it was Bing’s low quality that excluded it from default distribution and the reason “Google does not face true market competition in search”[105] is a function of quality, not Google’s deals.

That’s already fatal to the case. The remedy point is this: That same market reality means that no remedy prohibiting Google from entering into such agreements will rectify the situation. It means that Apple, Mozilla, Samsung, et al. will still choose Google as the default, even if Google is forbidden from paying them a revenue share (or even a set price) to do so—they will just forego the revenue from doing so, and Google will get a windfall.[106]

Yet it is hard to conceive of any other remedy that follows from Judge Mehta’s analysis in this decision. The decision is laser-focused on the determination that Google’s default distribution deals were (effectively) exclusive and thus foreclosed a substantial share of the market and deprived rivals of scale. Everything in the decision comes down to the default nature of the deals. It stands to reason that any remedy would be limited to removing the one-deal characteristic that, according to the court, makes the agreements anticompetitive.

Cutting against this somewhat is Judge Mehta’s conclusion that, having been deprived of scale by Google’s distribution deals, no rival is in a position to secure a default deal of its own.[107] But it is by no means clear from the decision that, in the absence of default deals with Google, rivals would be unable to compete effectively through other channels of distribution or compete for such deals in the future. The problem is that, because the court has no ability to prohibit Apple and Mozilla from offering default search engines without a Google deal, even prohibiting Google from entering into those deals doesn’t mean it won’t still be offered as the default, and this may not change the competitive landscape enough to enable Bing and other rivals to compete effectively.

Perhaps one might think that Google should just be compelled to share its data (and/or other “secret sauce”) with rivals so they have the quality necessary to actually win default placement deals. But that doesn’t work either. In the first place, this would be a clear acknowledgment that it is quality, not default distribution deals, that impedes rivals’ commercial success. If that is the only effective remedy, then the necessary legal basis of the holding is undermined.

Secondly, implementing that remedy would entail mandating that Google enter into deals with competitors to help them compete. This is anathema to U.S. antitrust law.[108] So much so that Judge Mehta himself threw out one of the plaintiff states’ claims in this case on exactly that basis:

Plaintiff States seek to bypass the “no duty to deal” doctrine entirely….

…The concerns that animate the no-duty-to-deal principle are equally applicable here. Primarily, adjudicating Plaintiff States’ claim would require the court to act as a “central planner” that endeavors to identify the proper “terms of dealing.” Their claim requires grappling with a host of questions that the court is ill-equipped to handle…. And those thorny questions foreshadow the challenges the court would face in administering a remedy…. A favorable outcome for Plaintiff States thus would mire the court in Google’s day-to-day operations…. The court has learned a lot about Google, but it is “ill suited” for that role.[109]

It is extremely difficult to see how Judge Mehta would countenance a forced-sharing arrangement for Google’s data as a remedy for the remaining claims in this case, given his unequivocal dismissal of other claims on precisely that basis.

As I noted, there is more to say about potential remedies in this case.[110] But for now, the most important thing is that the absence of viable remedies strongly supports the arguments I have presented here that the court’s liability finding was improper.

 

 

[1] United States, et al. v. Google LLC, Memorandum Opinion, Case No. 20-cv-3010 (APM) (D.D.C., Aug. 5, 2024), https://www.tn.gov/content/dam/tn/attorneygeneral/documents/pr/2024/pr24-59-Google.pdf ( “Google Search decision”).

[2] United States, et al. v. Google LLC, Memorandum Opinion, Case No. 20-cv-3010 (APM) (D.D.C., Aug. 4, 2023) (“Google Search summary judgment”), https://storage.courtlistener.com/recap/gov.uscourts.dcd.223205/gov.uscourts.dcd.223205.626.0_3.pdf.

[3] United States v. Microsoft, 253 F.3d 34 (D.C. Cir. 2001) (“Microsoft”).

[4] Google Search decision at 216 (quoting Microsoft, 253 F.3d at 79 and id. (quoting Areeda)).

[5] See Douglas H. Ginsburg & Koren Wong-Ervin, Challenging Consummated Mergers Under Section 2, CPI North America Column 2 (May 25, 2020), https://www.pymnts.com/cpi-posts/challenging-consummated-mergers-under-section-2-2 (“[T]he assertion that Section 2 does not require proof of anticompetitive effects is based upon a misreading of the D.C. Circuit’s decision in United States v. Microsoft. The assertion conflates the Microsoft court’s standard for proving competitive effects with its standard for establishing causation.”).

[6] Microsoft, 253 F.3d at 58 (“[T]he plaintiff, on whom the burden of proof of course rests… must demonstrate that the monopolist’s conduct indeed has the requisite anticompetitive effect.”).

[7] See Ginsburg and Wong-Ervin, supra note 5, at 2 (“The court went on to devote fully 20 pages to a careful analysis of the actual effects of each type of Microsoft’s allegedly anticompetitive conduct. Only after finding that each type of conduct indeed had an anticompetitive effect did the court turn to the separate and distinct question of causation.”).

[8] Id. at 3 (“[I]t is important to understand the limited applicability of the “reasonably capable of” standard.”)

[9] See id. at 2 (“It was in addressing this question that the court said it was appropriate, in a government enforcement action, to ‘infer causation when exclusionary conduct is aimed at producers of nascent competitive technologies as well as when it is aimed at producers of established substitutes.’”) (quoting Microsoft, 253 F.3d at 79).

[10] Id. at 4.

[11] Id. at 3 (quoting Microsoft, 253 F.3d at 79).

[12] Id. at 3.

[13] Presumably to maintain the requirement of antitrust injury for private plaintiffs, which helps ensure they are not abusing the courts just to harm a competitive rival.

[14] D. Bruce Hoffman, Antitrust in the Digital Economy: A Snapshot of Federal Trade Commission Issues, Remarks at GCR Live Antitrust in the Digital Economy 10 (May 2019), https://www.ftc.gov/system/files/documents/public_statements/1522327/hoffman__gcr_live_san_francisco_2019_speech_5-22-19.pdf.

[15] Rambus Inc. v. FTC, 522 F.3d 456 (D.C. Cir. 2008) (“Rambus”).

[16] Rambus, 522 F.3d at 463 (emphasis added).

[17] Id.

[18] Id. at 463 (“[I]f Rambus’s more complete disclosure would have caused [the SSO] to adopt a different (open, non-proprietary) standard, then its failure to disclose harmed competition and would support a monopolization claim.”).

[19] Id. at 463-64 (“But while we can assume that Rambus’s non-disclosure made the adoption of its technologies somewhat more likely than broad disclosure would have, the Commission made clear… that there was insufficient evidence that [the SSO] would have standardized other technologies had it known the full scope of Rambus’s intellectual property. Therefore, for the Commission’s syllogism to survive—and for the Commission to have carried its burden of proving that Rambus’s conduct had an anticompetitive effect—we must also be convinced that if Rambus’s conduct merely enabled it to avoid… RAND licensing terms, such conduct, alone, could be said to harm competition.”).

[20] Id. at 466-67 (emphasis added).

[21] Joshua D. Wright, Why the Supreme Court Was Correct to Deny Certiorari in FTC v. Rambus, Mar-09(2) Global Competition Policy 1, 6 (2009), https://www.competitionpolicyinternational.com/file/view/5880.

[22] Id.

[23] Rambus, 522 F.3d at 463.

[24] Wright, supra note 21, at 6 (emphasis added).

[25] Rambus, 522 F.3d at 467.

[26] Google Search decision at 219 (emphasis added).

[27] Ginsburg & Wong-Ervin, supra note 5, at 4.

[28] Google Search decision at 216.

[29] A quick aside: There are those who actually hold this view. They are radical outliers in the antitrust community and their position is unsupported by the sensible reading of any case law or economic analysis. See, e.g., Petition for Rulemaking to Prohibit Exclusionary Contracts by Open Markets Institute, et al., Fed. Trade Comm’n Doc. ID FTC-2021-0036-0002 (Jul. 20, 2021), https://downloads.regulations.gov/FTC-2021-0036-0002/content.pdf. They include stalwart neo-Brandeisian ideologues like the Open Markets Institute, the American Economic Liberties Project, and Marshall Steinbaum. They also include entities decidedly foreign (as far as I know) to the antitrust community like Friends of Family Farmers, Friends of the Earth, and the People’s Parity Project. Their position doesn’t represent what the law or the economics actually says.

[30] At least, it is not a question in the abstract given the obvious and direct competition from alternative general search engines. As I discuss below, however, the court makes this a question by challenging the relative quality of rival search engines. In this case, that actually constitutes another, separate basis for reversing the decision. See infra Section III.A.

[31] NB: “highly uncertain” in that Rambus’s market position could have been the result of either of two causes; not “highly uncertain” in the Microsoft sense that the competitive process was speculative.

[32] Ginsburg & Wong-Ervin, supra note 5, at 4 (emphasis added).

[33] Google Search decision at 216.

[34] Id. at 217.

[35] Id. at 216.

[36] See generally Joshua D. Wright, Moving Beyond Naïve Foreclosure Analysis, 19 Geo. Mason L. Rev. 1163 (2012); id. at 1181-82 (“The primary thrust of this Article is that accurately measuring the foreclosure produced by any allegedly exclusionary agreement requires foreclosure to be measured relative to what would be obtained but for that agreement.”); Thomas G. Krattenmaker and Steven C. Salop, Anticompetitive Exclusion: Raising Rivals’ Costs to Achieve Power over Price, 96 Yale L.J. 209, 259 (1986) (defining a “net foreclosure rate” as “the percentage of the suppliers’ capacity that was available to rivals before the exclusionary rights agreement was adopted but that is no longer available as a result of the agreement”).

[37] Rambus, 522 F.3d at 464.

[38] See generally Geoffrey A. Manne, Error Costs in Digital Markets, in Global Antitrust Institute Report on the Digital Economy (Joshua D. Wright and Douglas H. Ginsburg eds. 2020), https://gaidigitalreport.com/wp-content/uploads/2020/11/Manne-Error-Costs-in-Digital-Markets.pdf.

[39] Ginsburg & Wong-Ervin, supra note 5, at 4.

[40] See, e.g., Google Search decision at 24 (“The most efficient channel of GSE distribution is, by far, placement as the preloaded, out-of-the-box default GSE.”).

[41] Rambus, 522 F.3d at 466 (citing Schuylkill Energy Res., Inc. v. Penn. Power Light Co., 113 F.3d 405, 414 (3d Cir. 1997)).

[42] Google Search decision at 228 (“That means only 30% of all GSE queries in the United States come through a search access point that is not preloaded with Google.”).

[43] See also infra note 101, noting that it might be relevant to assess whether the size of Google’s payments was sufficient to induce distributors to promote it as the default to an extent out of proportion to its relative quality advantage (and thus, presumably, consumer preferences). But the court did not undertake this analysis.

[44] Id. at 227.

[45] Ginsburg & Wong-Ervin, supra note 5, at 3 (emphasis added). See also Microsoft, 253 F.3d at 71 (“[Microsoft’s conduct] help[ed] keep usage of Navigator below the critical level necessary for Navigator or any other rival to pose a real threat to Microsoft’s monopoly.’”)

[46] See Google Search decision at 163 (“As for Neeva, it entered and exited within four years.”).

[47] Who is also, perhaps tellingly, the Bing Professor of Neuroscience, Behavioral Biology & Economics at Caltech. See Antonio Rangel, Caltech (last accessed Aug. 13, 2024), https://www.hss.caltech.edu/people/antonio-rangel.

[48] Id. at 26.

[49] Id. at 31.

[50]  Id. at 229.

[51] See Daniel J. Gilman, Google, Amazon, Switching Costs, and Red Herrings, Truth on the Market (Nov. 22, 2023), https://truthonthemarket.com/2023/11/22/google-amazon-switching-costs-and-red-herrings.

[52] Id.

[53] See Prof. Antonio Rangel Presentation (Sept. 12-13, 2023), U.S. v. Google LLC Trial Exhibit No. UPXD104, available at https://www.justice.gov/d9/2023-09/416682.pdf.

[54] See Google Android (Case COMP/AT.40099) Commission Decision of 18 July 2018, https://competition-cases.ec.europa.eu/cases/AT.40099.

[55] See Mobile Search Engine Market Share in Europe — July 2024, StatCounter (last visited Aug. 12, 2024), https://gs.statcounter.com/search-engine-market-share/mobile/europe/#monthly-202001-202408 (showing that Google’s EU mobile search engine market share was 97.32% in January 2020 (just before the choice screen was implemented) and 95.96% in July 2024—a change of 1.37 percentage points).

[56] Google Search decision at 117.

[57] Id. at 32.

[58] Id. at 207.

[59] Id. at 208.

[60] Id. at 33.

[61] Id.

[62] Google Search decision at 32.

[63] Id. at 27.

[64] Or, at least, they don’t anymore. But 80% of searches on Windows desktops ran through Google Search even before Chrome existed. Id. at 32. That means that, apparently, a substantial majority of users were perfectly willing and able to overcome the “choice friction” and switch to Google Search even within Windows’ default browser.

[65] Id. at 229.

[66] By which I mean Office and Windows, not necessarily Bing. But the point holds: Microsoft is certainly all these things.

[67] It should also be noted that Microsoft did not just passively rely on its name recognition to overcome apparent consumer disinterest; rather, it actively tried to induce users to switch by offering “Bing Rewards” (now “Microsoft Rewards”) for using Bing, redeemable for gift cards, Microsoft services, and the like. See Microsoft Rewards, Microsoft (last visited Aug. 12, 2024), https://www.microsoft.com/en-us/rewards/about.

[68] Google Search decision at 238 (emphasis added).

[69] Id.

[70] Id. (emphasis added).

[71] Id. at 226.

[72] Id. at 232 (emphasis added).

[73] Id. at 152.

[74] Google Search decision at 238.

[75] Microsoft, 253 F.3d at 79 (“Given this rather edentulous test for causation, the question in this case is not whether Java or Navigator would actually have developed into viable platform substitutes, but (1) whether as a general matter the exclusion of nascent threats is the type of conduct that is reasonably capable of contributing significantly to a defendant’s continued monopoly power and (2) whether Java and Navigator reasonably constituted nascent threats at the time Microsoft engaged in the anticompetitive conduct at issue.”) (emphasis added).

[76] Gregory J. Werden, The Missing Element in the Google Case, Truth on the Market (Apr. 15, 2024), https://truthonthemarket.com/2024/04/15/the-missing-element-in-the-google-case (emphasis added).

[77] Id. at 232 (emphasis added)].

[78] Google Search decision at 6. See generally Google Search summary judgment.

[79] Google Search summary judgment at 48-49 (quoting Microsoft, 253 F.3d at 58-59) (emphasis in original).

[80] Id. at 21 (quoting Microsoft, 253 F.3d at 58).

[81] Google Search decision at 209.

[82] Id.

[83] Google Search decision at 209 (citing Microsoft, 253 F.3d at 64) (emphasis added).

[84] Id.

[85] See, e.g., Laura Ceci, Number of Mobile App Downloads Worldwide from 2021 to 2023, by Country (in Billions), Statista (Jun. 24, 2024), https://www.statista.com/statistics/1287159/app-downloads-by-country/#:~:text=In%202023%2C%20mobile%20apps%20in,generated%20approximately%2012.6%20billion%20downloads (“In 2023…, users in the United States generated approximately 12.6 billion downloads.”).

[86] Google Search decision at 32.

[87] See id. at 33 (“Google’s search share on Windows devices is 80%, with most of the queries flowing through the Chrome default.”).

[88] Id. at 221 (emphasis added).

[89] Id. (emphasis added).

[90] Id. at 209 (quoting United States v. Dentsply, 399 F.3d 181, 196 (3d Cir. 2005) (“Dentsply”)).

[91] Id. at 236 (quoting Dentsply, 399 F.3d at 194).

[92] Id. at 53-54 (emphasis added).

[93] See Desktop Operating System Market Share in United States Of America — July 2024, StatCounter (last accessed Aug. 12, 2024), https://gs.statcounter.com/os-market-share/desktop/united-states-of-america/#monthly-202001-202408 and Operating System Market Share in United States Of America — July 2024, StatCounter (last accessed Aug. 12, 2024), https://gs.statcounter.com/os-market-share/all/united-states-of-america/#monthly-202001-202408.

[94] Google Search decision at 103.

[95] Id. at 121.

[96] Id. at 122. Although it currently offers Google Search as the default, it need not, and this is subject to a different distribution deal than the Android one.

[97] Id. at 207.

[98] See id. at 209 (“In 2020 only 5.1% of all search queries on iOS devices went to a rival GSE through a non-default access point…. Most non-default queries still go through Google.”).

[99] Id. at 32.

[100] Google Search decision at 205.

[101] Notably, there is not, in fact, zero distribution of Bing on Android devices. Edge is currently the eighth-most downloaded free communications app on the Google Play store and the 189th most downloaded free app overall. See Microsoft Edge Daily Category Rankings: Google Play — Jul 12, 2024 – Aug 10, 2024 — All Categories — US — All Types, Sensor Tower (last accessed Aug. 13, 2024), https://app.sensortower.com/category-rankings?os=android&app_id=com.microsoft.emmx&start_date=2024-07-12&end_date=2024-08-10&countries=US&category=communication&category=application&category=all&chart_type=free&chart_type=grossing&device=android&hourly=false&selected_tab=charts&date=2024-08-10&summary_chart_type=topselling_free.

[102] See Werden, supra note 71.

[103] Google Search decision at 201 (emphasis added, except the term “no price” in the quote from Eddy Cue).

[104] Werden, supra note 71.

[105] Google Search decision at 201.

[106] It should be noted that “no price” perhaps doesn’t really mean “no price”; perhaps it means “no price relative to Google’s payment.” Maybe if Google is prohibited from paying anything, there would be some price that Microsoft could pay relative to Google’s legally mandated $0 that Apple would accept. Maybe. That’s not actually what this says, but I suppose it is possible. Indeed, this is actually what the decision should have assessed—whether the size of Google’s payments, not the fact of its defaults, was sufficient to effectively foreclose access by Microsoft for a product that was of demonstrably lower quality, but maybe not that much lower quality. To make this stark, if Google paid distributors $1 per year instead of billions, would we accept the argument that the default nature of those deals foreclosed competition? Despite Eddy Cue’s and the court’s hyperbole, I think it extremely unlikely we would accept such a claim at face value.

[107] Although it must be noted that, at least on mobile, the court concluded that “Google’s distribution agreements did not cause Microsoft’s past underinvestment in search.” Google Search decision at 238. So, it’s not clear that Google’s deals really did deprive Microsoft of scale.

[108] See Verizon Commc’ns Inc. v. Law Off. of Curtis V. Trinko LLP, 540 U.S. 398, 407-08 (2004) (“Firms may acquire monopoly power by establishing an infrastructure that renders them uniquely suited to serve their customers. Compelling such firms to share the source of their advantage is in some tension with the underlying purpose of antitrust law, since it may lessen the incentive for the monopolist, the rival, or both to invest in those economically beneficial facilities.”).

[109] Google Search decision at 268 (quoting Trinko, 540 U.S. at 408) (emphasis added).

[110] I will just add that arguments by competitors (naturally) and neo-Brandeisians (same) for breaking up Google by forcing it to divest Chrome or Android run squarely into the holding in Microsoft. Whether or not it was appropriate for the court to base its holding on Microsoft’s “edentulous” causation standard, having done so, that standard does not support a divestiture remedy: “In devising an appropriate remedy, the District Court also should consider whether plaintiffs have established a sufficient causal connection between Microsoft’s anti-competitive conduct and its dominant position in the OS market…. [S]tructural relief, which is ‘designed to eliminate the monopoly altogether . . . require[s] a clearer indication of a significant causal connection between the conduct and creation or maintenance of the market power.’ Absent such causation, the antitrust defendant’s unlawful behavior should be remedied by ‘an injunction against continuation of that conduct.’… [W]e have found a causal connection between Microsoft’s exclusionary conduct and its continuing position in the operating systems market only through inference.” Microsoft 253 F.3d at 106-07 (quoting Areeda & Hovenkamp) (emphasis in original).

ICLE Senior Scholar Ben Sperry took part in a recent Washington Legal Foundation-TechFreedom webinar on the legal limits to the principle of common carriage, from railroads to the internet. Video of the full event is embedded below.

ICLE Senior Scholar Eric Fruits was a guest on NW Fresh to discuss the state of Portland, Oregon’s economy. Video of the full segment is embedded below.

ICLE research was cited in a story in Discourse magazine about the value of data in artificial intelligence. You can read the full piece here.

As summed up by the International Center for Law and Economics, “data may often confer marginal benefits [but] there is little evidence that these benefits are ultimately decisive.”

ICLE Senior Scholar Ben Sperry was quoted by Law360 about the 9th U.S. Circuit Court of Appeals’ Bonta decision and what it could mean for children’s online safety legislation. You can read the full piece here.

Ben Sperry, a senior scholar of innovation policy at the International Center for Law & Economics, wrote in a post on the scholarly blog Truth on the Market agreed that the Ninth Circuit ruling was likely to ripple beyond California, noting that it wouldn’t be surprising if the duty-of-care provisions in the Senate-passed version of KOSA were similarly deemed unconstitutional.

“To butcher a Winston Churchill quote: It’s not yet clear if this is the beginning of the end, or just the end of the beginning, for children’s online-safety bills,” he said.

ICLE Senior Scholar Ben Sperry was cited by Communications Daily about the 9th U.S. Circuit Court of Appeals’ Bonta decision and what it could mean for children’s online safety legislation. You can read the full piece here.

The 9th U.S. Circuit Court of Appeals’ decision to partially uphold an injunction against a California age-appropriate social media design law (see 2408160015) means similar legislation at the federal level is likely unconstitutional, a policy expert at the International Center for Law & Economics said Monday. Innovation policy scholar Ben Sperry argued that duty of care provisions in the Kids Online Safety Act, which the Senate passed last month 91-3 (see 2407300042), likely violate the First Amendment. The 9th Circuit found the Age-Appropriate Design Code Act’s (AB-2273) impact assessment requirement is violative because it requires that platforms make judgments about what online content could harm children. Sperry argued that under KOSA, platforms would be incentivized to censor all but “the most benign speech” to avoid triggering children’s anxiety or to avoid “bullying” claims.

 

ICLE Editor-in-Chief R.J. Lehmann was quoted by the Washington Examiner in a story on the role that home and auto rate increases could play in the 2024 election. You can read the full piece here.

“California’s Prop 103 regulatory system makes it extraordinarily difficult for insurers to achieve rate increases, and after enormous wildfire losses, a number of the largest carriers have either threatened to exit the market or have actually done so,” Ray Lehmann, editor-in-chief of the International Center for Law and Economics, told the Washington Examiner in an email.

He added that after “decades of state officials sticking their head in the sand about the nature of the problem, the governor and insurance commissioner have in the past year attempted to enact a series of reforms that would liberalize the market somewhat and keep the remaining companies from heading for the exits.”

…“According to the U.S. Bureau of Labor Statistics’ Producer Price Index (PPI), the average monthly cost of private passenger auto insurance has risen 18.8% from $164.50 (or $2,290 annually) in July 2020 to $190.80 (or $2,720 annually) in July 2024,” Lehmann said. “By contrast, the increase from July 2016 to July 2020 was about 16% and from July 2012 to July 2016 was 13.1%. So, auto has been on a path of steeper increases over time.”

According to the same metrics, homeowners insurance has climbed 14.3% “from $213.40 ($2,561 annually) in July 2020 to $243.96 ($2,927 annually) in July 2024.” That came on the heels of a very modest increase of 4.3% for the previous four years.

…One unanticipated advantage of Republicans picking Trump over Gov. Ron DeSantis (R-FL) in this election cycle is that the Sunshine State has created a serious hazard with its own property insurance market. “Florida and California have been the notable basket cases,” Lehmann said.

After storms devastated the state in 2004 and 2005, Florida adopted the strategy of loading a “huge amount of risk onto the state-backed insurer, Florida Citizens, and the state-backed reinsurer, the Florida Cat Fund.”

Consequently, the state market has “been dominated in the years since by dozens of tiny, thinly capitalized, domestic insurance companies that only wrote homeowners and only wrote it in Florida,” he said.

The state lucked out with a hurricane-light decade between 2006 and 2016 but has been hit hard in recent years. Homeowners insurance litigation plus those hurricanes have meant that “more than a dozen domestic insurers have already been declared insolvent and dozens more are on the brink.”

…“DeSantis and the legislature have passed reform legislation intended to address the runaway litigation issues, but it remains unclear whether that will be sufficient to stabilize the market” or if a few more bad storms ruin it for all 23 million and counting Floridians, Lehmann said.

ICLE Chief Economist Brian Albrecht was cited in an op-ed in Idaho State Journal on Vice President Kamala Harris’ price controls proposal. You can read the full piece here.

ICLE Chief Economist Brian Albrecht was quoted by The Pinnacle Gazette about Vice President Kamala Harris’ proposal to impose price controls. You can read the full piece here.

According to Brian Albrecht, another economist, allowing the government to define “fair” prices can lead to substantial market control, dangerously edging toward price control. Historical instances, like during World War II and the 1970s energy crisis, show price controls often lead to shortages and market disruption.

ICLE President Geoffrey A. Manne was quoted by Eurasia Review in an op-ed about Judge Amit Mehta’s decision in the Google search antitrust case. You can read the full piece here.

Google has announced that it will appeal the decision, and the commentary about how it could do so is already mushrooming.  Geoffrey A. Manne, president of the International Center for Law and Economics, is one, offering a detailed overview about where Judge Mehta is said to have misread or misunderstood such concepts as proof of anticompetitive conduct.

ICLE Chief Economist Brian Albrecht was quoted by Sinclair Media’s The National Desk in a story about Vice President Kamala Harris’ proposal to fight “unfair” pricing. You can read the full piece here.

Economists said that price controls are nothing new. According to Brian Albrecht, an economist, any policy that grants the government the power to determine what constitutes a “fair” or “unfair” price is effectively price control, regardless of how it is labeled. “When bureaucrats, not markets, determine acceptable prices, we’re dealing with price controls,” he said.

ICLE’s amicus brief in the challenge to the Federal Communications Commission’s rulemaking to reimpose Title II regulation on broadband providers was cited in a Communications Daily story about the case. You can read the full piece here.

The International Center for Law and Economics, Phoenix Center, Tech Freedom (see 2408140039) and U.S. Chamber of Commerce are among other groups that have filed amicus briefs opposing the FCC.

ICLE Chief Economist Brian Albrecht was quoted by The Dispatch in a story about Kamala Harris’ anti-price gouging proposals. You can read the full piece here.

After the speech, Harris tweeted something similar, and Sen. Bob Casey then bragged that Harris had just “endorsed” his and Elizabeth Warren’s broad “Price Gouging Prevention Act.” As economist Brian Albrecht explains in an excellent primer, these kinds of proposals differ from state anti-gouging laws (which narrowly apply only during emergencies) and are simply national price controls by another name…

ICLE President Geoffrey A. Manne was cited by the American Institute for Economic Research in a blog post about the U.S. District Court ruling in the recent Google search antitrust case. You can read the full piece here.

As Geoffrey Manne, President and Founder of the International Center for Law and Economics, notes in his review of the case, “Google’s default agreements can’t be deemed to have caused anticompetitive harm because defaults are readily overcome by high-quality, reputable alternatives.” The ruling itself even acknowledges that there was no “price that Microsoft could ever offer Apple to make the switch because of Bing’s inferior quality.” In other words, the price of the contract and Google’s market share are irrelevant because there is no other product that matches the quality of its search. Consumers clearly have options when it comes to search, but because they consistently choose one product over the other does not mean that the exclusive contracts indicate anticompetitive behavior. Companies invest in marketing, partnerships, and product placements to increase their visibility and, ultimately, their revenue. This is not anti-competitive; it is the essence of competition.

ICLE Chief Economist Brian Albrecht was quoted by Fortune in a story about recent revisions to the U.S. Bureau of Labor Statistics’ jobs report. You can read the full piece here.

Brian Albrecht, chief economist at the International Center for Law and Economics, said he’s not surprised by the lack of market movement. “It’s a big revision, but we expected a big revision,” he told Fortune. “Private forecasters were putting out numbers—anything from 350,000 up to a million—and this is on the upper side of it, but we expected it to be on the high side, which is why we had a bunch of coverage on it before it even came out.”

To Albrecht’s point, investors widely anticipated a large negative revision to payroll growth on Wednesday, with some fearing an even more dramatic 1 million reduction in jobs between March 2023 and March 2024, as Fortune previously reported.

Eric Wallerstein, chief markets strategist at Yardeni Research, echoed Albrecht’s somewhat sanguine view as well. “The market’s lack of reaction is pretty telling,” he said. “I think the revisions, you can say, they were priced in. We were kind of expecting some sort of big revision … It’s kind of a nothing burger.”

…Albrecht backed up that view. “The [Fed] governors know this stuff, and the voting members know this stuff,” he said. “So it shouldn’t change too much.”

…Albrecht explained that there are always common errors that can occur when the Bureau of Labor Statistics (BLS) reaches out to businesses for employment data. “They don’t get every establishment, and so they get some sampling error,” he said, noting that “sometimes they get HR folks who do a really good job; sometimes they don’t. Sometimes they send [the payroll survey] back; sometimes they don’t.”

..,“The model overestimated based on a big jump in new business formation from the year before,” Albrecht explained. “It turns out that that was overly optimistic. Business formation still is up, but it skyrocketed and then flattened out.”

The good news, according to Albrecht, is that this sampling issue may soon be a thing of the past, because we have settled into a more “steady state” of business births and deaths.

“This revision was huge, yes, but we shouldn’t expect some big change next year in the revision, because the big reason for the revision—the kind of mess-up of the birth-death model—shouldn’t be there anymore,” he said.

ICLE President Geoffrey A. Manne was cited by Dissident Voice in a blog post about the U.S. District Court’s decision in the Google search antitrust case. You can read the full piece here.

Google has announced that it will appeal the decision, and the commentary about how it could do so is already mushrooming.  Geoffrey A. Manne, president of the International Center for Law and Economics, is one, offering a detailed overview about where Judge Mehta is said to have misread or misunderstood such concepts as proof of anticompetitive conduct.