As the DOJ’s antitrust case against Google begins, all eyes are focused on whether Google violated antitrust law by, among other things, entering into exclusionary agreements with equipment makers like Apple and Samsung or web browsers like Mozilla. Per the District Court’s Memorandum Opinion, released August 4, “These agreements make Google the default search engine on a range of products in exchange for a share of the advertising revenue generated by searches run on Google.” The DOJ alleges that Google unlawfully monopolizes the search advertising market.
Aside from matters relating to antitrust liability, an equally important question is what remedy, if any, would work to restore competition in search advertising in particular and online advertising generally?
Developments in the UK might shed some light. The UK Treasury commissioned a report to make recommendations on changes to competition law and policy, which aimed to “help unlock the opportunities of the digital economy.” The report found that Big Tech’s monopolizing of data and control over open web interoperability could undermine innovation and economic growth. Big Tech platforms now have all the data in their hands, block interoperability with other sources, and will capture more of it, through their huge customer-facing machines, and so can be expected to dominate the data needed for the AI Period, enabling them to hold back competition and economic growth.
The dominant digital platforms currently provide services to billions of end users. Each of us has either an Apple or Android device in our pocket. These devices operate as part of integrated distribution platforms: anything anyone wants to obtain from the web goes through the device, its browser (often Google’s search engine), and the platform before accessing the Open Web, if not staying on an app on an apps store within the walls of the garden.
Every interaction with every platform product generates data, refreshed billions of times a day from multiple touch points providing insight into buying intent and able to predict people’s behavior and trends.
All this data is used to generate alphanumeric codes that match data contained in databases (aka “Match Keys”), which are used to help computers interoperate and serve relevant ads to match users’ interests. These were for many years used by all from the widely distributed Double Click ID. They were shared across the web and were used as the main source of data by competing publishers and advertisers. After Google bought Double Click and grew big enough to “tip” the market, however, Google withdrew access to its Match Keys for its own benefit.
The interoperability that is a feature of the underlying internet architecture has gradually been eroded. Facebook collected its own data from user’s “Likes” and community groups and also withdrew access for independent publishers to its Match Key data, and recently Apple has restricted access to Match Key data that is useful for ads for all publishers, except Google has a special deal on search and search data. As revealed in U.S. vs Google, Apple is paid over $10 billion a year by Google so that Google can provide its search product to Apple users and gather all their search history data that it can then use for advertising. The data generated by end user interactions with websites is now captured and kept within each Big Tech walled garden.
If the Match Keys were shared with rival publishers for use in their independent supply channel and used by them for their own ad-funded businesses, interoperability would be improved and effective competition could be generated with the tech platforms. Competition probably won’t exist otherwise.
Both Google and Apple currently impose restrictions on access to data and interoperability. Cookie files also contain Match Keys that help maintain computer sessions and “state” so that different computers can talk to each other and help remember previous visits to websites and enable e-commerce. Cookies do not themselves contain personal data and are much less valuable than the Match Keys that were developed by Double Click or ID for advertisers, but they do provide something of a substitute source of data about users’ intent to purchase for independent publishers.
Google and Apple are in the process of blocking access to Match Keys in all forms to prevent competitors from obtaining relevant data about users needs and wants. They also prevent the use of the Open Web and limit the inter-operation of their apps stores with Open Web products, such as progressive web apps.
The UK’s Treasury Report refers to interoperability 8 times and the need for open standards as a remedy 43 times; the Bill refers to interoperability and we are expecting further debate about the issue as the Bill passes through Parliament.
A Brief History of Computing and Communications
The solution to monopolization, or lack of competition, is the generation of competition and more open markets. For that to happen in digital worlds, access to data and interoperability is needed. Each previous period of monopolization involved intervention to open-up computer and communications interfaces via antitrust cases and policy that opened market and liberalized trade. We have learned that the authorities need to police standards for interoperability and open interfaces to ensure the playing field is level and innovation can take place unimpeded.
IBM’s activity involved bundling computers and peripherals and the case was eventually solved by unbundling and unblocking interfaces needed by competitors to interoperate with other systems. Microsoft did the same, blocking third parties from interoperating via blocking access to interfaces with its operating system. Again, it was resolved by opening-up interfaces to promote interoperability and competition between products that could then be available over platforms.
When Tim Berners Lee created the World Wide Web in the early 1990s, it took place nearly ten years after the U.S. courts imposed a break-up of AT&T and after the liberalization of telecommunications data transmission markets in the United States and the European Union. That liberalization was enabled by open interfaces and published standards. To ensure that new entrants could provide services to business customers, a type of data portability was mandated, enabling numbers held in incumbent telecoms’ databases to be transferred for use by new telecoms suppliers. The combination of interconnection and data portability neutralized the barrier to entry created by the network effect arising from the monopoly control over number data.
The opening of telecoms and data markets in the early 1990s ushered in an explosion of innovation. To this day, if computers operate to the Hyper Text Transfer Protocol then they can talk to other computers. In the early 1990s, a level playing field was created for decentralized competition among millions of businesses.
These major waves of digital innovation perhaps all have a common cause. Because computing and communications both have high fixed costs and low variable or incremental costs, and messaging and other systems benefit from network effects, markets may “tip” to a single provider. Competition in computing and communications then depends on interoperability remedies. Open, publicly available interfaces in published standards allow computers and communications systems to interoperate; and open decentralized market structures mean that data can’t easily be monopolized.
It’s All About the Match Keys
The dominant digital platforms currently capture data and prevent interoperability for commercial gain. The market is concentrated with each platform building their own walled gardens and restricting data sharing and communication across. Try cross-posting among different platforms as an example of a current interoperability restriction. Think about why messaging is restricted within each messaging app, rather than being possible across different systems as happens with email. Each platform restricts interoperability preventing third-party businesses from offering their products to users captured in their walled gardens.
For competition to operate in online advertising markets, a similar remedy to data portability in the telecom space is needed. Only, with respect to advertising, the data that needs to be accessed is Match Key data, not telephone numbers.
The history of anticompetitive abuse and remedies is a checkered one. Microsoft was prohibited from discriminating against rivals and had to put up a choice screen in the EU Microsoft case. It didn’t work out well. Google was similarly prohibited by the EU in Google search (Shopping) from (1) discriminating against rivals in its search engine results pages, (2) entering exclusive agreements with handset suppliers that discriminated against rivals, and (3) showing only Google products straight out of the box in the EU Android case. The remedies did not look at the monopolization of data and its use in advertising. Little has changed and competitors claim that the remedies are ineffective.
Many in the advertising publishing and ad tech markets recall that the market worked pretty well before Google acquired Double Click. Google uses multiple data sources as the basis for its Match Keys and an access and interoperability remedy might be more effective, proportionate and less disruptive.
Perhaps if the DOJ’s case examines why Google collects search data from its search engine, its use of search histories, browser histories and data from all interactions with all products for its Match Key for advertising, the court will better appreciate the importance of data for competitors and how to remedy that position for advertising-funded online publishing.
Following Europe’s Lead
The EU position is developing. Under the EU’s Digital Markets Act (DMA), which now supplements EU antitrust law as applied in the Google Search and Android Decisions, it is recognized that people want to be able to provide products and services across different platforms or cross-post or communicate with people connected to each social network or messaging app. In response, the EU has imposed obligations on Big Tech platforms in Articles 5(4) and 6(7) that provide for interoperability and require gatekeepers to allow open access to the web.
Similarly, Section 20.3 (e) of the UK’s Digital Markets, Competition and Consumers Bill (DMCC) refers to interoperability and may be the subject of forthcoming debate as the bill passes further through Parliament. Unlike U.S. jurisprudence with its recent fixation on consumer welfare, the objective of the Competition and Markets Authority is imposed by the law. The obligation to “promote competition for the benefit of consumers” is contained in EA 2013 s 25(3). This can be expressly related to intervention opening up access to the source of the current data monopolies: the Match Keys could be shared, meaning all publishers could get access to IDs for advertising (i.e., operating systems generated IDs such as Apple’s IDFA or Google’s Google ID or MAID).
In all jurisdictions it will be important for remedies to stimulate innovation, and to ensure that competition is promoted between all products that can be sold online, rather than between integrated distribution systems. Moreover, data portability needs to apply with reference to use of open and interoperable Match Keys that can be used for advertising, and that way address the data monopolization risk. As with the DMA, the DMCC should contain an obligation for gatekeepers to ensure fair reasonable and nondiscriminatory access, and treat advertisers in a similar way to that through which interoperability and data potability addressed monopoly benefits in previous computer, telecoms, and messaging cases.
Tim Cowen is the Chair of the Antitrust Practice at the London-based law firm of Preiskel & Co LLP.
If I were to draft new Merger Guidelines, I’d begin with two questions: (1) What have been the biggest failures of merger enforcement since the 1982 revision to the Merger Guidelines?; and (2) What can we do to prevent such failures going forward? The costs of under-enforcement have been large and well-documented, and include but are not limited to higher prices, less innovation, lower quality, greater inequality, and worker harms. It’s high time for a course correction. But do the new Merger Guidelines, promulgated by Biden’s Department of Justice (DOJ) and Federal Trade Commission (FTC), do the trick?
Two Recent Case Studies Reveal the Problem
Identifying specific errors in prior merger decisions can inform whether the new Guidelines will make a difference. Would the Guidelines have prevented such errors? I focus on two recent merger decisions, revealing three significant errors in each for a total of six errors.
The 2020 approval of the T-Mobile/Sprint merger—a four-to-three merger in a highly concentrated industry—was the nadir in the history of merger enforcement. Several competition economists, myself included, sensed something was broken. Observers who watched the proceedings and read the opinion could fairly ask: If this blatantly anticompetitive merger can’t be stopped under merger law and the existing Merger Guidelines, what kind of merger can be stopped? Only mergers to monopoly?
The district court hearing the States’ challenge to T-Mobile/Sprint committed at least three fundamental errors. (The States had to challenge the merger without Trump’s DOJ, which embraced the merger for dubious reasons beyond the scope of this essay.) First, the court gave undue weight to the self-serving testimony of John Legere, T-Mobile’s CEO, who claimed economies from combining spectrum with Sprint, and also claimed that it was not in T-Mobile’s nature to exploit newfound market power. For example, the opinion noted that “Legere testified that while T-Mobile will deploy 5G across its low-band spectrum, that could not compare to the ability to provide 5G service to more consumers nationwide at faster speeds across the mid-band spectrum as well.” (citing Transcript 930:23-931:14). The opinion also noted that:
T-Mobile has built its identity and business strategy on insulting, antagonizing, and otherwise challenging AT&T and Verizon to offer pro-consumer packages and lower pricing, and the Court finds it highly unlikely that New T-Mobile will simply rest satisfied with its increased market share after the intense regulatory and public scrutiny of this transaction. As Legere and other T-Mobile executives noted at trial, doing so would essentially repudiate T-Mobile’s entire public image. (emphasis added) (citing Transcript at 1019:18-1020:1)
In the court’s mind, the conflicting testimony of the opposing economists cancelled each other out—never mind such “cancelling” happens quite frequently—leaving only the CEO’s self-serving testimony as critical evidence regarding the likely price effects. (The States’ economic experts were the esteemed Carl Shapiro and Fiona Scott Morton.) It bears noting that CEOs and other corporate executives stand to benefit handsomely from the consummation of a merger. For example, Activision Blizzard Inc. CEO Bobby Kotick reportedly stands to reap more than $500 million after Microsoft completes its purchase of the video game publishing giant.
Second, although the primary theory of harm in T-Mobile/Sprint was that the merger would reduce competition for price-sensitive customers of prepaid service, most of whom live in urban areas, the court improperly credited speculative commitments to “provide 5G service to 85 percent of the United States rural population within three years.” Such purported benefits to a different set of customers cannot serve as an offset to the harms to urban consumers who benefited from competition between the only two facilities-based carriers that catered to prepaid customers.
Third, the court improperly embraced T-Mobile’s proposed remedy to lease access to Dish at fixed rates—a form of synthetic competition—to restore the loss in facilities-based competition. Within months of the consummated merger, the cellular CPI ticked upward for the first time in a decade (save a brief blip in 2016), and T-Mobile abandoned its commitments to Dish.
The combination of T-Mobile/Sprint represented the elimination of actual competition across two wireless providers. In contrast, Facebook’s acquisition of Within, maker of the most popular virtual reality (VR) fitness app on Facebook’s VR platform, represented the elimination of potential competition, to the extent that Facebook would have entered the VR fitness space (“de novo entry”) absent the acquisition. In disclosure, I was the FTC’s economic expert. (I commend everyone to read the critical review of the new Merger Guidelines by Dennis Carlton, Facebook’s expert, in ProMarket, as well as my thread in response.) The district court sided with the FTC on (1) the key legal question of whether potential competition was a dead letter (it is not), (2) market definition (VR fitness apps), and (3) market concentration (dominated by Within). Yet many observers strangely cite this case as an example of the FTC bringing the wrong cases.
Alas, the court did not side with the FTC on the key question of whether Facebook would have entered the market for VR fitness apps de novo absent the acquisition. To arrive at that decision, the court made three significant errors. First, as Professor Steve Salop has pointed out, the court applied the wrong evidentiary standard for assessing the probability of de novo entry, requiring the FTC to show a probability of de novo entry in excess of 50 percent. Per Salop, “This standard for potential entry substantially exceeds the usual Section 7 evidentiary burden for horizontal mergers, where ‘reasonable probability’ is normally treated as a probability lower than more-likely-than-not.” (emphasis in original)
Second, the court committed an error of statistical logic, by crediting the lack of internal deliberations in the two months leading up to Facebook’s acquisition announcement in June 2021 as evidence that Facebook was not serious about de novo entry. Three months before the announcement, however, Facebook was seriously considering a partnership with Peloton—the plan was approved at the highest ranks within the firm. Facebook believed VR fitness was the key to expanding its user base beyond young males, and Facebook had entered several app categories on its VR platform in the past with considerable success. Because de novo entry and acquisition are two mutually exclusive entry paths, it stands to reason that conditional on deciding to enter via acquisition, one would expect to see a cessation of internal deliberation on an alternative entry strategy. After all, an individual standing at a crossroads would consider alternative paths, but upon deciding which path to take and embarking upon it, the previous alternatives become irrelevant. Indeed, the opinion even quoted Rade Stojsavljevic, who manages Facebook’s in-house VR app developer studios, testifying that “his enthusiasm for the Beat Saber–Peloton proposal had “slowed down” before Meta’s decision to acquire Within,” indicating that the decision to pursue de novo entry was intertwined with the decision to entry via acquisition. In any event, the relevant probability for this potential competition case was the probability that Facebook would have entered de novo in the absence of the acquisition. And that relevant probability was extremely high.
Third, like the court in T-Mobile/Sprint, the district court again credited the self-serving testimony of Facebook’s CEO, Mark Zuckerberg, who claimed that he never intended to enter VR fitness apps de novo. For example, the court cited Mr. Zuckerberg’s testimony that “Meta’s background and emphasis has been on communication and social VR apps,” as opposed to VR fitness apps. (citing Hearing Transcript at 1273:15–1274:22). The opinion also credited the testimony of Mr. Stojsavljevic for the proposition that “Meta has acquired other VR developers where the experience requires content creation from the developer, such as VR video games, as opposed to an app that hosts content created by others.” (citing Hearing Transcript at 87:5–88:2). Because this error overlaps with one of the three errors identified in the T-Mobile/Spring merger, I have identified five distinct errors (six less one) needing correction by the new Merger Guidelines.
Although the court credited my opinion over Facebook’s experts on the question of market definition and market concentration, the opinion did not cite any economic testimony (mine or Facebook’s experts) on how to think about the probability of entry absent the acquisition.
The New Merger Guidelines
I raise these cases and their associated errors because I want to understand whether the new Merger Guidelines—thirteen guidelines to be precise—will offer the kind of guidance that would prevent a future court from repeating the same (or similar) errors. In particular, would either the T-Mobile/Sprint or Facebook/Within decision (or both) have been altered in any significant way? Let’s dig in!
The New Guidelines reestablish the importance of concentration in merger analysis. The 1982 Guidelines, by contrast, sought to shift the emphasis from concentration to price effects and other metrics of consumer welfare, reflecting the Chicago School’s assault on the structural presumption that undergirded antitrust law. For several decades prior to the 1980s, economists empirically studied the effect of concentration on prices. But as the consumer welfare standard became antitrust’s north star, such inquiries were suddenly considered off-limits, because concentration was deemed to be “endogenous” (or determined by the same factors that determine prices), and thus causal inferences of concentration’s effect on price were deemed impossible. This was all very convenient for merger parties.
Guideline One states that “Mergers Should Not Significantly Increase Concentration in Highly Concentrated Markets.” Guideline Four states that “Mergers Should Not Eliminate a Potential Entrant in a Concentrated Market,” and Guideline Eight states that “Mergers Should Not Further a Trend Toward Concentration.” By placing the word “concentration” in three of thirteen principles, the agencies make it clear that they are resuscitating the prior structural presumption. And that’s a good thing: It means that merger parties will have to overcome the presumption that a merger in a concentrated or concentrating industry is anticompetitive. Even Guideline Six, which concerns vertical mergers, implicates concentration, as “foreclosure shares,” which are bound from above by the merging firms’ market share, are deemed “a sufficient basis to conclude that the effect of the merger may be to substantially lessen competition, subject to any rebuttal evidence.” The new Guidelines restore the original threshold Herfindahl-Hirschman Index (HHI) of 1,800 and delta HHI of 100 to trigger the structural presumption; that threshold had been raised to an HHI of 2,500 and a change in HHI of 200 in the 2010 revision to the Guidelines.
This resuscitation of the structural presumption is certainly helpful, but it’s not clear how it would prevent courts from (1) crediting self-serving CEO testimony, (2) embracing bogus efficiency defenses, (3) condoning prophylactic remedies, (4) committing errors in statistical logic, or (5) applying the wrong evidentiary standard for potential competition cases.
Regarding the proper weighting of self-serving employee testimony, error (1), Appendix 1 of the New Guidelines, titled “Sources of Evidence,” offers the following guidance to courts:
Across all of these categories, evidence created in the normal course of business is more probative than evidence created after the company began anticipating a merger review. Similarly, the Agencies give less weight to predictions by the parties or their employees, whether in the ordinary course of business or in anticipation of litigation, offered to allay competition concerns. Where the testimony of outcome-interested merging party employees contradicts ordinary course business records, the Agencies typically give greater weight to the business records. (emphasis added)
If heeded by judges, this advice should limit the type of errors we observed in T-Mobile/Sprint and Facebook/Within, with courts crediting the self-serving testimony by CEOs and other high-ranking employees.
Regarding the embrace of out-of-market efficiencies, error (2), Part IV.3 of the New Guidelines, in a section titled “Procompetitive Efficiencies,” offers this guidance to courts:
Merging parties sometimes raise a rebuttal argument that, notwithstanding other evidence that competition may be lessened, evidence of procompetitive efficiencies shows that no substantial lessening of competition is in fact threatened by the merger. When assessing this argument, the Agencies will not credit vague or speculative claims, nor will they credit benefits outside the relevant market. (citing Miss. River Corp. v. FTC, 454 F.2d 1083, 1089 (8th Cir. 1972)) (emphasis added)
Had this advice been heeded, the court in T-Mobile/Sprint would have been foreclosed from crediting any purported merger-induced benefits to rural customers as an offset to the loss of competition in the sale of prepaid service to urban customers.
Regarding the proper treatment of prophylactic remedies offered by merger parties, error (3), footnote 21 of the New Guidelines state that:
These Guidelines pertain only to the consideration of whether a merger or acquisition is illegal. The consideration of remedies appropriate for otherwise illegal mergers and acquisitions is beyond its scope. The Agencies review proposals to revise a merger in order to alleviate competitive concerns consistent with applicable law regarding remedies. (emphasis added)
While this approach is very principled, the agencies cannot hope to cure a current defect by sitting on the sidelines. I would advise saying something explicit about remedies, including mentioning the history of their failures to restore competition, as Professor John Kwoka documented so ably in his book Mergers, Merger Control, and Remedies (MIT Press 2016).
Finally, regarding courts’ committing errors in statistical logic or applying the wrong evidentiary standard for potential competition cases, errors (4) and (5), the New Merger Guidelines devote an entire guideline (Guideline Four) to potential competition. Guideline Four states that “the Agencies examine (1) whether one or both of the merging firms had a reasonable probability of entering the relevant market other than through an anticompetitive merger.” Unfortunately, there is no mention that reasonable probability can be satisfied at less than 50 percent, per Salop, and the agencies would be wise to add such language in the Merger Guidelines. In defining “reasonable probability,” the Guidelines state that evidence that “the firm has successfully expanded into other markets in the past or already participates in adjacent or related markets” constitutes “relevant objective evidence” of a reasonable probably. In making its probability assessment, the court in Facebook/Within did not credit Facebook’s prior de novo entry in other app categories on Facebook’s VR platform. The Guidelines also state that “Subjective evidence that the company considered organic entry as an alternative to merging generally suggests that, absent the merger, entry would be reasonably probable.” Had it heeded this advice, the court would have ignored, when assessing the probability of de novo entry absent the merger, the fact that Facebook did not mention the Peloton partnership two months prior to the announcement of its acquisition of Within.
A Much Needed Improvement
In summary, I conclude that the new Merger Guidelines offer precisely the kind of guidance that would have prevented the courts in T-Mobile/Sprint and in Facebook/Within from committing significant errors. The additional language suggested here—taking a firm stance on remedies and defining reasonable probability—is really fine-tuning. While this review is admittedly limited to these two recent cases, the same analysis could be undertaken with respect to a broader array of anticompetitive mergers that have approved by courts since the structural presumption came under attack in 1982. The agencies should be commended for their good work to restore the enforcement of antitrust law.