Economic Analysis and Competition Policy Research

Home   •   About   •   Analytics   •   Videos

Professor Tim Wu, former White House advisor on antitrust, offered remedies following Judge Mehta’s decision in the U.S. Google Search case. He identified both Google’s revenue-sharing agreements that exclude competitors and its access to certain “choke points” as a basis for remedies. A divestment order of Chrome and the Android operating system was proposed, as well as an access remedy to Google’s browser, data and A.I. technologies.

It is hard to see how the transfer of a browser monopoly into others’ hands, however, would facilitate access and use of it. That could repeat the mistake of the AT&T 1984 divestiture order that transferred local access monopolies into separate ownership without creating any competitive constraint or pressure on those local business to innovate and compete. In a follow up article, Julia Angwin pointed out the fundamental problem being the Google search results pages, facing no competitive pressure, are now “a pulsing ad cluttered endless scroll,” which masks relevant results. Google’s ad-fuelled profit maximisation leads it to promote that which is remunerative over that which is more relevant.   

Also, there remains a major issue with any access order: Will it be able to withstand future technology changes used by Google to circumvent their aims? A crucial issue in writing an order to a monopoly tech company to provide access to XYZ or supply XYZ interface, and the day after the order being written a technical change (or simply version control) making the order technically outdated and pointless.

Any remedy first needs to stop the infringement, prevent its reoccurrence and restore competition. So, the core problem now is to restore competition to the Google search monopoly. This means finding a competing consumer-facing search product that is ad-funded so that “free at the point of use” search can provide competitive pressure on Google’s own free at the point of use product. A possible optionis canvassed below.

The two-sided nature of the search market means any effective solution needs to create consumer-facing competition with Google Search pages and business-facing competition for Google’s Search Text Advertising offering. A starting point for remedies is then prohibiting the mechanisms used by Google that restricted competition from rivals. This means prohibiting the revenue-sharing and default-setting deals with Apple and other technology and telecoms companies that have acted as a moat to protect Google’s Search “castle.” However, restoring effective competition going forward also means enabling the use of data inputs and alternative access points (such as the browser) so competing search ads face competitive price pressure.

The proposal below is inspired by the BT Openreach settlement (and prior BT Consent decree). BT proposed an access remedy, which applied to the local loop. Non-discriminatory access to BT’s local loop (Openreach) business was supplied to third parties on the same terms as it was supplied to downstream parts of BT. The obligation applied to the BT Group of companies and its internal divisions, and corporate structure was subject to non-discrimination both on supply and use. This improved upon the AT&T divestiture remedy, which was in operation in the United States at the time. Avoiding the risk of technology change also means taking account of an often-overlooked Consent Decree, which was agreed among BT/MCI/Concert and the DOJ in 1994. That decree broke new ground as it imposed a non-discriminatory “use” obligation on the recipient of services supplied by the monopoly supplier. A similar obligation on non-discriminatory use of inputs could apply to the use by Google of inputs and would apply overtime irrespective of the technical means of supply. 

Scale of Google’s data inputs and sunk investments costs

Google now has unrivalled scale in data acquired from billions of users millions of times per day when they interact with Google’s many products. That data is obtained from its ownership of Chrome, the dominant web browser, providing Google with unrivalled browser history data. It also uses other interoperable code (such as that stored in cookies) to check which websites browsers have visited and has an unparalleled understanding of consumers interests and purchasing behavior. Per Judge Mehta’s Memorandum Opinion in USA v Google (Search), data from billions of search histories provides it with “uniquely strong signals” of intent to purchase data that is combined with all data from all other interactions with all of Google’s many products (see trial exhibit of Google presentation: Google is magical). Its knowledge from all data inputs is combined to provide it with high quality information for advertising. The Memorandum Opinion recognizes that “more users mean more advertisers and more advertisers mean more revenues,” and “more users on a GSE means more queries, which in turn means more ad auctions and more ad revenue.” These positive feedback loops suggest increasing returns to scale and returns to the scope of a range of products offered over the same platform using artificial intelligence as part of its systems. It has built one of the most recognized and valuable brands in the world.

The costs facing any competitor seeking to make an entirely new search engine from scratch are now enormous. This is referenced in evidence as the “Herculean problem.” Reference is made to the many billions of dollars that would be needed by Apple to build a new search engine of its own.

Any restoration of competition will now have to overcome these very considerable advantages and sunk costs, while at the same time competing with Google as the established, and well-knownsupplier of the best search engine in the world. That point about the costs being “sunk” for Google but not new entrants will be returned to below.  

The Memorandum Opinion refers to the uniqueness of Google’s Search and the search access choke points many times. Access to these unique facilities must now be on the cards as a remedy.

Third party access to data inputs, match keys and access points to support effective competition in “free at the point of use search results businesses”

Google uses data inputs to identify the user’s “purchasing intent” that inform its ads machine. Data inputs are combined from multiple consumer interactions with others digital content and has enabled Google to charge high prices for its search text ads. Google’s Search engine consists of at least three key components: (1) an index of media owner content cataloged by a web crawler, (2) a “relevance engine” to match consumer input to this catalog, (3) ranking and monetization of the search engine results. At a technical level, the online display advertising system relies on match keys that enable the matching of demand for ad inventory to match a supplier of ad inventory. Third parties need access to these data inputs currently uniquely available to Google, to derive user’s purchasing intent. Competing rivals could then employ the input data and match keys to match inventory supporting display advertising and competing search page results businesses using Google’s relevance engine.Use of such inputs would help drive down prices for ads in competing search businesses. 

Access points for search businesses include the Chrome browser. Here, the idea that the browser could be quarantined, as suggested by Professor Wu, could be picked up. The browser would also need to be monitored so that it provides a neutral gateway to the web. It is a choke point that can be enhanced with additional functionality – a wallet in the browser substitutes for decentralized wallets that could otherwise be deployed by competing websites. As was the case with the 1956 AT&T consent decree, AT&T was prevented from competing in areas that were open to competition – so too Google could be prevented from adding functionality to its gateway that could be provided by others elsewhere on the web. The browser then loses its position as gateway controller and becomes a neutral window on the web. 

Google is owned by Alphabet so there is an opportunity to apply an obligation to Alphabet not to discriminate in the supply of its relevance engine as between Google and third parties rather than its Search system as a whole. That would enable competition between pages and page presentations offered by different businesses. It would overcome the enormous costs and “Herculean” task of creating a new search engine from scratch. New players might then be tempted to enter that business and resell Alphabet’s relevance engine results combined with its own ads or ads from third parties, which would increase price pressure on search ads.

Currently, 80 percent of the SERP is composed of advertising of one sort or another. Enabling competition in the provision of search results could avoid the morass of current search pages and encourage both quality and price competition. This could benefit both consumers and advertisers.Alternative search businesses could be expected to innovate in the way that they provide and present ads; higher proportions of the results pages could be composed of relevant results and fewer ads.  If competing businesses had access to Alphabet’s relevance engine and data inputs they could use them for their own advertising, introducing price pressure on Google’s search text ads. New entrants could be expected to finance their businesses quickly given that they would be reselling a proven search product.

Availability of distribution deals with Google’s revenue sharing partners

The current agreements with Apple, OEMs and telecoms carriers operate as exclusive agreements. They contain contractual restrictions in the form of default settings and revenue sharing payments,which incentivize the parties to promote Google Search Ads. The scale of the payments operates as a disincentive and prevents the parties from offering products competing with Google in search. 

Removal of only the contractual default setting is likely to be insufficient to end the anticompetitive effect of the agreements and would go no way to restoring competition. The sharing of revenue from Google’s Search advertising must end if competition between new search advertising players is to be established.

Ending the current distribution deals on a revenue-sharing basis creates a problem of what is an acceptable replacement deal. If Google products are to compete on their merits no restriction at all should be imposed on distributors from providing competing alternatives. However, Google’s distributors will still need to be paid for distribution and the volumes and scale of payments is so large that current recipients are still likely to only sell Google products, even if the restrictive provisions are removed. They are unlikely to take the risk of backing a competitor search product if some form of competition in search is not restored. If access to Alphabet’s relevance engine is mandated as described above that would also help to restore competition at the distributor level. 

A proposed access remedy needs to underpin the restoration of competition 

A remedy order applicable to Alphabet could provide access to an independent and quarantined browser (access point) and search relevance engine. That would not restore competition alone. Overcoming the considerable barrier to entry of a new entrant seeking to build its own relevance engine and attracting new users while competing with Google is very hard. It is currently prohibitive,even for Apple.

When considering the issue further, it is important to appreciate that: 

• The relevance engine and index are currently both organizationally and technically separate from the ads and ad network organization. 

• Search is currently optimized by people working in a search business. There are separate groups of people that work on products and separate organizations for advertising. 

• Alphabet’s products (news, maps, images, shopping, etc.) are interweaved between relevant search results when the page is presented to end users. An effective remedy could build on these existing organizational and business boundaries. 

If third party competitors could access the relevance engine and its index on non-discriminatory terms, they may be able to create effective competition between new “Search Engine Results” businesses. Those businesses would access the substantial sunk investments already made in optimizing search relevant to users’ needs and overcome the substantial “Google” brand value. As noted above, that investment is sunk for Google but represents a considerable barrier to entry for others. Since much of that value has been obtained illegally, there would be a case for stripping Alphabet of that value. Perhaps a better solution here would be to enable the use of the brand to support entry. New competitors would be known to be using Google’s world-renowned relevance engine. The established reputation for quality would help entry. As this is central to restoring competition compensation for use is then a non-issue.

Moreover, Google currently offers access to its relevance engine to companies (like Duck Duck Go) that would resell them, so cannot easily suggest that the above proposal is unworkable.

How the proposal addresses technology changes over time

The law has been broken through the denial of access to data inputs and choke points, and thusdeprived rivals of scale. No other players have sufficient scale to replicate Google’s position. Access to the same data that is used in Google Ads would be a starting point for competitors to create competing search ads from. The solution is access to the IDs and the data inputs that Alphabet uses to fund its search business. Obligations can be crafted to access data feeds for non-discriminatory use of whatever Google uses.

To be clear, there are two critical data feeds that will be needed for competitors to function: (1) Access to the Google relevance engine. This would enable competitors to offer a highly relevant search product. Results would be from a proven and established, world renowned and high-quality source; and (2) Access to the data inputs and advertising IDs and match key data, which are used in Google search ads to identify purchasing intent that can be matched with available advertising inventory.

As a matter of U.S. law and practicality, a non-discrimination obligation on usage can be contained in an order addressed to Google as the user of a search engine or data source owned by Alphabet. As a usage–based non-discrimination obligation applicable to the user of assets owned by the head company, Alphabet, it is materially different from a requirement to supply. There is less of a risk of it offending the case law that defers to businesses deciding whether and with whom they contract – it is instead a requirement not to discriminate between what is received by Google’s search business and what is received by third parties’ search businesses. If Google’s monetization of search results uses no inputs from its relevance engine or data hoard, then it would have no obligation to supply.Conveniently the Alphabet holding company could also be the addressee of the obligation, as was the case with BT Group and its operating corporate entities such as Openreach.

Note that this approach also better addresses the issue of technology change over time. The more usual divestiture order and access obligation suffers from technology being defined at the time the order is written. Since it must be written as a remedy to a defined problem and so if the harm was bundling of interoperability or lack of access to XYZ APIs, then the order mandates unbundling and a requirement to supply XYZ APIs. If a new API is invented that achieves a same end by different means, or a new technology is introduced, there will not have been any case against the defendant for abuse with relation to that new API or technology and no order can easily force the supply of the new API. By contrast, where the addressee of the order is in the same group as the supplier an order can be crafted in terms of non-discrimination in the use of the monopoly asset owned by the group head company and used by a functionally separate business. 

Conclusions

This essay addresses the core problem for effective remedies identified in USA v Google (Search). Any remedy needs to address the scale of Google’s data inputs and sunk investments. This is remedied by providing third party access to data inputs and access points to support effective competition in “free at the point of use search results” but would also create competing ads businesses with pressure on ads prices. The current distribution and revenue-sharing partners need to be prohibited. The proposed access remedy enables the creation of competition between rival search engine results businesses, imposing market discipline on the promotion and presentation of search results. The proposal addresses technology changes over time by drawing on lessons learned from divestiture in telecommunication and from ensuing that non-discrimination in usage of key inputs is the focus of the remedy. 

Additionally, allowing Alphabet to continue to own its browser (even if quarantined) and provide access to search access points means that capital funding will continue to be in the interests of the Alphabet group. Divestiture would otherwise place monopoly assets in others’ hands with incentives to raise price and degrade quality for all those seeking to use them. Funding of divested assets that are currently cost centres in a vertically integrated business would otherwise also be a major issue to overcome. Here, the proposed non-discrimination remedy bites in a different way – so that technology change is not a problem with this type of remedy. 

The approach described here would need to be coupled with transparency obligations such that third parties have visibility of what data inputs the Google Search Co receives so that they can make comparisons. Agreements between different divisions of Alphabet – whether partially in separate ownership or otherwise – can be entered into between different corporate entities within Alphabet to more effectively enable oversight across both a corporate and technical boundary. If done carefully, addressing technology, financial and commercial terms, and the scope for technology change circumventing the remedy can be managed. In effect, it would aim to make the remedy future proof.

The Justice Department’s pending antitrust case against Google, in which the search giant is accused of illegally monopolizing the market for online search and related advertising, revealed the nature and extent of a revenue sharing agreement (“RSA”) between Google and Apple. Pursuant to the RSA, Apple gets 36 percent of advertising revenue from Google searches by Apple users—a figure that reached $20 billion in 2022. The RSA has not been investigated in the EU. This essay briefly recaps the EU law on remedies and explains why choice screens, the EU’s preferred approach, are the wrong remedy focused on the wrong problem. Restoring effective competition in search and related advertising requires (1) the dissolution of the RSA, (2) the fostering of suppressed publishers and independent advertisers, and (3) the use of an access remedy for competing search-engine-results providers.

EU Law on Remedies

EU law requires remedies to “bring infringements and their effects to an end.” In Commercial Solvents, the Commission power was held to “include an order to do certain acts or provide certain advantages which have been wrongfully withheld.”

The Commission team that dealt with the Microsoft case noted that a risk with righting a prohibition of the infringement was that “[i]n many cases, especially in network industries, the infringer could continue to reap the benefits of a past violation to the detriment of consumers. This is what remedies are intended to avoid.” An effective remedy puts the competitive position back as it was before the harm occurred, which requires three elements. First, the abusive conduct must be prohibited. Second, the harmful consequences must be eliminated. For example, in Lithuanian Railways, the railway tracks that had been taken away were required to be restored, restoring the pre-conduct competitive position. Third, the remedy must prevent repetition of the same conduct or conduct having an “equivalent effect.” The two main remedies are divestiture and prohibition orders.

The RSA Is Both a Horizontal and a Vertical Arrangement

In the 2017 Google Search (Shopping) case, Google was found to have abused its dominant position in search. In the DOJ’s pending search case, Google is also accused of monopolizing the market for search. In addition to revealing the contours of the RSA, the case revealed a broader coordination between Google and Apple. For example, discovery revealed there are monthly CEO-to-CEO meetings where the “vision is that we work as if we are one company.” Thus, the RSA serves as much more than a “default” setting—it is effectively an agreement not to compete.

Under the RSA, Apple gets a substantial cut of the revenue from searches by Apple users. Apple is paid to promote Google Search, with the payment funded by income generated from the sale of ads to Apple’s wealthy user base. That user base has higher disposable income than Android users, which makes it highly attractive to those advertising and selling products. Ads to Apple users are thought to generate 50 percent of ad spend but account for only 20 percent of all mobile users.

Compared to Apple’s other revenue sources, the scale of the payments made to Apple under the RSA is significant. It generates $20 billion in almost pure profit for Apple, which accounts for 15 to 20 percent of Apple’s net income. A payment this large and under this circumstance creates several incentives for Apple to cement Google’s dominance in search:

The RSA also gives Google an incentive to support Apple’s dominance in top end or “performance smartphones,” and to limit Android smartphone features, functions and prices in competition with Apple. In its Android Decision, the EU Commission found significant price differences between Google Android and iOS devices, while Google Search is the single largest source of traffic from iPhone users for over a decade.

Indeed, the Department of Justice pleadings in USA v. Apple show how Apple has sought to monopolize the market for performance smartphones via legal restrictions on app stores and by limiting technical interoperability between Apple’s system and others. The complaint lists Apple’s restrictions on messaging apps, smartwatches, and payments systems. However, it overlooks the restrictions on app stores from using Apple users’ data and how it sets the baseline for interoperating with the Open Web. 

It is often thought that Apple is a devices business. On the contrary, the size of its RSA with Google means Apple’s business, in part, depends on income from advertising by Google using Apple’s user data. In reality, Apple is a data-harvesting business, and it has delegated the execution to Google’s ads system. Meanwhile, its own ads business is projected to rise to $13.7 billion by 2027. As such, the RSA deserves very close scrutiny in USA v. Apple, as it is an agreement between two companies operating in the same industry.

The Failures of Choice Screens

The EU Google (Search) abuse consisted in Google’s “positioning and display” of its own products over those of rivals on the results pages. Google’s underlying system is one that is optimized for promoting results by relevance to user query using a system based on Page Rank. It follows that promoting owned products over more relevant rivals requires work and effort. The Google Search Decision describes this abuse as being carried out by applying a relevance algorithm to determine ranking on the search engine results pages (“SERPs”). However, the algorithm did not apply to Google’s own products. As the figure below shows, Google’s SERP has over time filled up with own products and ads.

To remedy the abuse, the Decision compelled Google to adopt a “Choice Screen.” Yet this isn’t an obvious remedy to the impact on competitors that have been suppressed, out of sight and mind, for many years. The choice screen has a history in EU Commission decisions.

In 2009, the EU Commission identified the abuse Microsoft’s tying of its web browser to its Windows software. Other browsers were not shown to end users as alternatives. The basic lack of visibility of alternatives was the problem facing the end user and a choice screen was superficially attractive as a remedy, but it was not tested for efficacy. As Megan Grey observed in Tech Policy Press, “First, the Microsoft choice screen probably was irrelevant, given that no one noticed it was defunct for 14 months due to a software bug (Feb. 2011 through July 2012).” The Microsoft case is thus a very questionable precedent.  

In its Google Android case, the European Commission found Google acted anticompetitively by tying Google Search and Google Chrome to other services and devices and required a choice screen presenting different options for browsers. It too has been shown to be ineffective. A CMA Report (2020) also identified failures in design choices and recognized that display and brand recognition are key factors to test for choice screen effectiveness.

Giving consumers a choice ought to be one of the most effective ways to remedy a reduction of choice. But a choice screen doesn’t provide choice of presentation and display of products in SERPs.  Presentations are dependent on user interactions with pages. And Google’s knowledge of your search history, as well as your interactions with its products and pages, means it presents its pages in an attractive format. Google eventually changed the Choice Screen to reflect users top five choices by Member State. However, none of these factors related to the suppression of brands or competition, nor did it rectify the presentation and display’s effects on loss of variety and diversity in supply. Meanwhile, Google’s brand was enhanced from billions of user’s interactions with its products.

Moreover, choice screens have not prevented rival publishers, providers and content creators from being excluded from users’ view by a combination of Apple’s and Google’s actions. This has gone on for decades. Alternative channels for advertising by rival publishers are being squeezed out.

A Better Way Forward

As explained above, Apple helps Google target Apple users with ads and products in return for 36 percent of the ad revenue generated. Prohibiting that RSA would remove the parties’ incentives to reinforce each other’s market positions. Absent its share of Google search ads revenue, Apple may find reasons to build its own search engine or enhance its browser by investing in it in a way that would enable people to shop using the Open Web’s ad funded rivals. Apple may even advertise in competition with Google.  

Next, courts should impose (and monitor) a mandatory access regime. Applied here, Google could be required to operate within its monopoly lane and run its relevance engine under public interest duties in “quarantine” on non-discriminatory terms. This proposal has been advanced by former White House advisor Tim Wu:

I guess the phrase I might use is quarantine, is you want to quarantine businesses, I guess, from others. And it’s less of a traditional antitrust kind of remedy, although it, obviously, in the ‘56 consent decree, which was out of an antitrust suit against AT&T, it can be a remedy. And the basic idea of it is, it’s explicitly distributional in its ideas. It wants more players in the ecosystem, in the economy. It’s almost like an ecosystem promoting a device, which is you say, okay, you know, you are the unquestioned master of this particular area of commerce. Maybe we’re talking about Amazon and it’s online shopping and other forms of e-commerce, or Google and search.

If the remedy to search abuse were to provide access to the underlying relevance engine, rivals could present and display products in any order they liked. New SERP businesses could then show relevant results at the top of pages and help consumers find useful information.

Businesses, such as Apple, could get access to Google’s relevance engine and simply provide the most relevant results, unpolluted by Google products. They could alternatively promote their own products and advertise other people’s products differently. End-users would be able to make informed choices based on different SERPs.

In many cases, the restoration of competition in advertising requires increased familiarity with the suppressed brand. Where competing publishers’ brands have been excluded, they must be promoted. Their lack of visibility can be rectified by boosting those harmed into rankings for equivalent periods of time to the duration of their suppression. This is like the remedies used for other forms of publication tort. In successful defamation claims, the offending publisher must publish the full judgment with the same presentation as the offending article and displayed as prominently as the offending article. But the harm here is not to individuals; instead, the harm redounds to alternative publishers and online advertising systems carrying competing ads. 

In sum, the proper remedy is one that rectifies the brand damage from suppression and lack of visibility. Remedies need to address this issue and enable publishers to compete with Google as advertising outlets. Identifying a remedy that rectifies the suppression of relevance leads to the conclusion that competition between search-results-page businesses is needed. Competition can only be remedied if access is provided to the Google relevance engine. This is the only way to allow sufficient competitive pressure to reduce ad prices and provide consumer benefits going forward.

The authors are Chair Antitrust practice, Associate, and Paralegal, respectively, of Preiskel & Co LLP. They represent the Movement for an Open Web versus Google and Apple in EU/US and UK cases currently being brought by their respective authorities. They also represent Connexity in its claim against Google for damages and abuse of dominance in Search (Shopping).

© 2024 The Sling