What is a digital gatekeeper (Part 2)? The relevance of the single-homing v. multi-homing debate

As the Commission has announced that it will issue its proposal for ex ante regulation for so-called digital gatekeepers in the beginning of December, there are growing rumours about the form this regulation will take, the platforms that will be captured and the substantive rules that will be contained in this regulation.

As to the form, it is now suggested that the ex ante regulation and the New Competition Tool (“NCT”) will be combined in a single instrument that would be called the Digital Markets Act (“DMA”). Commissioner Vestager last week described the combined initiative as having “two complementary pillars, a combination of ex ante regulation and case-by-case enforcement.” If that is the plan, this means that the NCT would be limited to digital markets, however defined in the proposal.

As to the platforms covered by ex ante regulation, it has been suggested that the regulation would capture up to 20 companies, with a definition of digital gatekeepers combining quantitative and qualitative criteria. While it has been obvious from the start that companies, such as Google and other so-called GAFAs, would likely be covered by the regulation, the question now is which other companies or platforms would fall under its scope.

Finally, a list of practices that might form part of possible “black” or “grey” lists has been intentionally or unintentionally leaked by the Commission. While digital gatekeepers would apparently be banned from pursuing the practices described in the black list, the status of the grey list is unclear.

In this blog post, I would like to return to the threshold issue of how one should define digital gatekeepers for the purpose of an ex ante regulation. In my first post on the subject, I argued that the core element of the definition of a digital “gatekeeper” is that it “controls access” to critical online services which allow business users to reach a large category of users. I also explained that the other factors referred to by the Commission (e.g., in its Digital Services Act consultation), such as the presence of barriers to entry (e.g., large economies of scale and significant network effects) and market power, can be the cause or the consequence of the gatekeeping position enjoyed by some platforms, and therefore be somewhat redundant for the purposes of a definition.

I however explained that, from a policy standpoint, although they may be closely interlinked, it would not be a bad thing to include several factors in the definition of a digital “gatekeeper”, such as the facts that (i) the platform “controls access” to critical online services for business users to reach a large category of users, (ii) is protected by high barriers to entry and (iii) holds market power on the market for the critical services it delivers to business users. To avoid ending up with an overinclusive definition, I would make these factors cumulative, and possibly combine them with quantitative criteria (e.g., turnover, number of users, etc.).

Several readers of this blog reached out to me to point out that one important dimension was missing from my definition: the presence or absence of multi-homing. Users “single-home” when they use only one platform, while they “multi-home” when they use two or more platforms (or I should say “primarily” single-home or multi-home as users are not necessarily homogeneous). Whether there is multi-homing or not will often depend on how costly it is for users to do so. At a superficial level, the presence of multi-homing might seem good from a competition standpoint in that multi-homers do not seem “prisoners” to a single platform. Thus, one could be tempted to say that in the presence of multi-homing, a platform cannot be a gatekeeper, presumably because it will be constrained by alternative platforms. That approach would be too simplistic, however, because it ignores that digital platforms typically have two sides, and that a combination of multi-homing on one side with single-homing on the other side can be a source of competitive problems. Fortunately, there is a large economic literature on multi-homing, which offers some important clues as to the relevance of that concept with respect to the definition. More precisely, the teachings of this literature can at least help us to identify platforms that cannot be digital gatekeepers, hence creating some sort of “safe harbour”.

The starting point of the analysis is that in the case of two-sided platforms you can observe three different scenarios: (i) single-homing on both sides; (ii) single-homing on one side, with multi-homing on the other side; and (iii) multi-homing on both sides.

Most of the economic literature (including Armstrong (2006)) has focused on the second scenario, i.e. single-homing on one side, with multi-homing on the other side. The main takeaway of this literature is captured by the European Commission in the following fashion:

“Whether agents at both sides of a platform participate in multiple platforms or just one has important implications for market power. If one side of a market practices single-homing, then the only way for the other side to reach those agents is through their preferred platform. Thus, platforms have monopoly power over providing access to their single-homing customers for the multi-homing side. This monopoly power naturally leads to high prices being charged to the multi-homing side and typically there will be too few agents on this side being served from a social welfare point of view.”

Such a scenario is illustrated by the Apple App Store, where iPhone owners typically single home (there are indeed very few people who own both an iPhone and an Android phone), while app developers typically multi-home (apps such as Uber, Tinder, Twitter or Facebook need to be present on both the App Store and the Play Store to maximize their customer reach and strengthen the network effects on which they rely). As illustrated by the recent Commission investigation into the App Store, as well as related lawsuits in the United States, there is a significant danger that in this case the platform can take advantage of the multi-homing side (by, for instance charging a premium fee for granting access to the multi-homing side). Thus, platforms with single-homing on one side and multi-homing on the other side can raise a gatekeeping problem because they control access to their single-homing customers for the multi-homing side. There may, however, be circumstances where the market power of the platform can be constrained (e.g., when switching costs on the single-homing side are low), in which case the risk of harm is lowered.

In the case of single-homing on both sides of the platform, each side can only interact with the users on the other side of that platform. Assume for instance a ride-sharing or a food-delivery application, with most of the users single-homing. Does this create a gatekeeping problem? Well it depends on a variety of factors. For instance, is there significant competition for the users on each side of the market? It is indeed not because a user single-homes that there is no alternative to the service. It may simply mean that the user has chosen to use only one platform. Yet, the existence of alternatives somewhat constrains the market power of the platform if switching costs are low, as in that case users can always switch to another platform if they are unhappy with the service they receive. Other factors, such as asymmetry of information or collective action problem on one side of the platform can also have market power implications.

But the scenario on which I would like to particularly focus is when the platform is characterized by multi-homing on both sides. In that scenario, both sides can meet on alternative platforms, so any attempt of the platform to exploit one side would be likely counteracted by users switching to rival platforms. For instance, hotel reservation platforms have multi-homing on both sides in that most prospective travellers will consult more than one platform before booking a suitable room, while hotels will typically offer their rooms on multiple platforms. In that case, these platforms cannot be gatekeepers as both sets of users can reach out to each other through various platforms. For instance, a prospective traveller seeking to book a room at the Sofitel hotel in London can make that booking through various channels (including various OTAs, the website of the hotel itself, etc.), while the Sofitel in London can reach prospective travellers through a variety of channels. This does not mean that nothing can go wrong as these platforms – just like any business – can always engage in anticompetitive behaviours. For instance, the authors of the Competition policy for the digital era expert report note that practices of digital platforms that restrict multi-homing should be viewed with a sceptical eye. But it would be wrong to regulate these platforms as digital gatekeepers while they raise no gatekeeping issues. Thus, in the case of a platform with multi-homing on both sides, there should be a presumption that the platform does not hold a gatekeeper position.

While the above analysis would certainly require further elaboration (and I invite those who disagree to reach out), it could be used to bring clarity to the debate on which type of platforms should be defined as gatekeepers, falling under the scope of ex ante regulation. As the Commission is now considering a definition that would capture a wider group of platforms, the above analysis suggests that a “safe harbour” could be applied to platforms with multi-homing on both sides. To the extent that competition issues arise in that setting, they can be dealt with through competition law.

Economic theory suggests that platforms with single-homing on one side and multi-homing one the other side (competitive bottlenecks) are often – but not always – problematic as these platforms have monopoly power over providing access to their single-homing customers for the multi-homing side. Whether platforms with single homing on both sides are problematic depends on a variety of factors that need to be analysed. Does this just justify a rule whereby platforms in those settings should be presumptively considered as digital gatekeepers? Not necessarily, but the above safe harbour cannot apply to them.

In sum, while I still strongly believe that a definition based on cumulative criteria combined with quantitative criteria (to avoid capturing too many platforms) is the right way to go, considering the role of multi-homing brings a useful dimension to the debate. In my view, the presence of primarily single-homing users on at least one side is a necessary, but not sufficient condition for a platform to be considered as a digital gatekeeper. To put it differently, platforms with multi-homing on both sides should presumptively fall outside the definition of digital gatekeepers.

****