A formal economic analysis of the potential anticompetitive effects of contractual restrictions between Google and manufacturers of mobile devices based on the Android operating system.
The model identifies a mechanism through which “tying” of Google’s app store with its search app can protect and increase Google’s profits from search advertising, and help it outbid other search engines.
A few points of principle that might be useful to develop remedies are discussed.
The Android case, still pending before the European Commission and other authorities (but concluded with an infringement decision by the Russian Federal Antimonopoly Service), has attracted considerable interest in antitrust circles, both from a legal and an economic perspective. In a recent paper, my co-author (Prof. Federico Etro) and I provide an economic rationalisation for concerns that restrictions in the contracts between Google and manufacturers of mobile devices based on the Android operating system have anti-competitive effects. We offer a new perspective based on an extension of recent important modelling work by Professors Choi and Jeon.
Our analysis considers certain features of the contractual agreements that Google enters into with original equipment manufacturers (OEMs) that produce mobile devices with an Android operating system (namely, the “Mobile Application Distribution Agreements” (MADAs), the Revenue Sharing Agreements (RSAs) and the Anti-Fragmentation Agreements (AFAs)). The concern is that taken together, these agreements de facto allow Google to tie its app store “Google Play” with its search engine Google Search, giving OEMs financial incentives to pre-install them on their devices, and this can foreclose rivals and protect Google’s dominance in search advertising. The model shows formally how the use of these agreements by Google can be motivated as a means to ensure exclusive pre-installation of its search engine application as the default search engine on Android devices, and foreclose installation of rival search engine providers, as well as the growth of original equipment manufacturers (OEMs) producing so-called “forked” Android devices (i.e. devices using the Android operating system without Google Play and other key apps by Google) – in a way that reduces both consumer surplus and welfare to its own advantage.
The article sets out a coherent economic mechanism for foreclosure inspired by the Choi and Jeon model, but seeking to reflect more closely the salient features of the case. We assume that both Google and a “more efficient” search engine can pay OEMs to pre-install their search apps, but Google cannot fully extract the entire surplus that consumers obtain from its application store and other apps as a standalone product – either because of the initial commitment to zero price charged to buyers, or because consumers have heterogeneous preferences and even optimal pricing cannot extract all consumer rents in full. In such cases, tying is shown in the model to be attractive to Google because it improves profits – either by outbidding rival search engines for pre-installation of the search app with full foreclosure; or (when preferences are heterogeneous) by attracting consumers with the highest valuation of its application store and marginalising the rival search engine on forked devices. We show that, under weak conditions, tying is profitable and reduces consumer surplus by inducing consumers to favour more expensive devices even with a “less efficient” search app.
We also review complementary motivations behind the conduct at issue: in particular, we discuss how RSAs with selected OEMs can generate a typical mechanism of “naked exclusion” which reinforces our main argument. In practice, selected OEMs in can be paid separately or sequentially to gain exclusivity, limiting at the same time the amount of payments through a “divide-and-conquer” strategy.
An entry-deterring strategy such as that articulated in the paper can generate at least three sources of consumer harm. First, entry deterrence eliminates gains from the provision of better or differentiated search engines. In spite of Google’s technological lead, different search engines could provide results that are differentiated for depth and range of outcomes, with different specificities, for different types of queries. Constraints to multi-homing thus create losses for consumers. Second, the conduct can make it difficult for rivals to build scale in search and compete effectively, eliminating pressure on Google to reduce its margins in online advertising. This has negative consequences for advertisers and, ultimately, for consumers. It can also translate into more intrusive advertising and larger collection of consumer data by Google without viable alternatives for consumers. And third, as well known and accepted from Microsoft days, entry deterrence can reduce a dominant firm’s incentive to invest in innovation, as well as the incentive for potential entrants.
What could a potential solution look like?
While remedies in these cases are often an opaque process which results from complex negotiations and trade-offs behind the scenes, one can think of a few points of principle.
A first measure to address the tying aspect could be to require the app store suite (GooglePlay together with its GooglePlay Services) to be offered on a standalone basis. This would extend to offering directly or indirectly a bundle that included other services at prices that are lower than the sum of the prices at which it offers the app store suite on a standalone basis (in order to prevent evasion of the unbundling remedy with a pricing that would make standalone products artificially non-attractive). Other commentators have noted that Google could then charge a positive fee for some of its products. Indeed our analysis suggests that Google could charge OEMs a positive price for the app store suite alone; however, the benefit for OEMs and consumers would be in the opportunity to have devices with the app store suite matched with alternative search engines as well as other applications, generating product differentiation in mobile devices and applications, and intensifying competition.
A second measure could focus on the requirement that OEMs are forbidden from commercialising forked Android devices if they also commercialise “Google compatible” Android devices. Undoing this could potentially enhance product differentiation and competition on the merit between OEMs without necessarily undermining the “Google compatible” ecosystem, and it would allow manufacturers and app developers to diversify their production through forked Android devices.
A third measure could be to require that Google does not offer payments to OEMs (and Mobile Network Operators) conditional on exclusivity of its search engine (or pre-installation of its app store or other Google’s product). This would allow rival search engines to pay OEMs for pre-installation and outbid Google for RSAs. Competition between search engines would be on the merit, with multi-homing and payments to OEMs shifted entirely or partially to consumers through lower prices for the Android mobiles. This would also enhance investments in innovation for the creation of new and better search engines and other software applications.
Of course, as the Microsoft Windows Media Player case has shown an unbundling remedy can be extremely ineffective when it is not complemented by further measures. A possible measure here could be to allow OEMs that wished to engage in exclusivity arrangements with competing search providers to do so; at the same time, OEMs that wished to do so could also offer consumers the opportunity to set GoogleSearch as the default search engine by presenting them with a “choice screen” where consumers should select one of multiple options. This would allow consumers to choose Google Search or any other alternative as a default – as part of a conscious choice. In effect a choice screen remedy would move the default decision from OEMs to consumers, and this would protect from concerns about possible retaliation against OEMs willing to opt for a different default search engine.