The general anxiety which animates current efforts to regulate ‘Big Tech’ reflects a coalescence of concerns about tracking, ‘surveillance’, privacy violations, misinformation and fake news, addiction, hate speech, leveraging of market power, cascading monopolies, killer acquisitions and reverse variant, exploitation, foreclosure, undermining of innovation – amplified by giant size, extravagant profitability and phenomenal market capitalisations. And by concerns about de facto ‘unelected governments’ taking decisions of systemic importance while engaging in huge lobbying efforts. Antitrust enforcement (slow, backward looking, constrained by precedent and clueless on remedies) has failed to make a dent into the power of the main actors, and while there is something of an ‘antitrust awakening’ (especially in the US after 20 years of near-death) this failure has led to a strong pivot towards ex-ante regulation. We decided that “we need smart ex-ante regulation of digital, bright rules, because antitrust just cannot get us there”.   

However, the current regulation proposals will have limited upside for greater digital diversity and meaningful competition. First, the business models of the main digital actors are distinct and their incentives very different – from conduct towards complementary businesses and users, to approach to privacy and data protection. Yet the proposals winding their way through the European Parliament, meant to set out ‘self-executing’ obligations applying to all ‘gatekeepers’, are just a handful of rules extrapolated from a batch of current/recent antitrust cases. It is curious to be drawing up a shiny new regulatory machine essentially by repurposing as general presumptions theories of harm from a few ad hoc antitrust cases (some still not decided). Because they are general, and do not recognise the specifics of business models, the current formulations will not be ‘self-executing’ but will consume time in endless challenges. Second, rules derived from a narrow antitrust syllabus can help protect some competition ‘on the platform’ but will not empower entry and growth to challenge current incumbents, bringing about a more diverse digital world. Technological innovation will most likely remain tethered to the current incumbents' paths, relying on the same set of ingredients (users to be monetised, data to be gathered, subscriptions to be sold, advertisers and app developers to be attracted).Transformation of the digital landscape will require something else: effective merger moratoria, breakups of past anticompetitive mergers, wall-to-wall interoperability, data protection, and new thinking around shared digital infrastructures to create scale and favour new opportunities. 

1 A few home truths

There is established widespread concern around a handful of digital platforms (the ‘GAFAM’) arising from their size, profitability, pervasiveness, ‘systemic’ role, market power, and domination of technological change, which motivates current efforts at establishing a new regulatory regime. I do not try to document here the basis for that concern, which I will take as read, but discuss instead the shape of the regulatory effort underway and what it can and cannot hope to achieve.1 I argue that the proposed rules will not take us as far as many hope they will, and we need both more structure and more ambitious thinking.2  

First, some basic truths. 

Regulation plans for digital platforms in Europe/US are targeting ‘the GAFAM’ only. For all the ‘thresholds’ and ‘criteria’ laid out in pages of proposals, the bottom line is that regulation plans being laid out in Europe (and their US counterparts) are intended to catch only the ‘Big Five’ – and not because of anti-American sentiment (the US proposals would not make sense in that view) but because of their size, market power, profitability and ‘systemic’ importance.  No one else is (or will be) in scope.

Antitrust has (broadly) failed to curb the market power of the largest digital platforms. We are pivoting to regulation because we recognise that antitrust has broadly failed in this space so far. It did not need to be this way (the Microsoft case did teach an enduring lesson to Microsoft and changed it permanently) but ten years of engagement by the European Commission with Google have not moved the dial towards deterrence. Enforcement against Facebook has been too little, too late. Cases against Amazon and Apple are moving at glacial pace. There has been an upswell of cases over the last 18 months (Attorneys General and federal agencies in the US, national agencies in Europe as well as the Commission), but outcomes are years away. The reasons for the failure are multiple (bad case selection driven by complaints, dealing by definition with ‘yesterday’s problem’, straitjacket of precedent, useless remedies) but overall antitrust has not put any brake on the progress of players which now account for swathes of digital GDP. 

Merger control in particular has most spectacularly failed. Acquisitions have been the main tool for the accumulation of size and power by some of the current giants. What can put a stop to boundless growth can only be obsolescence, as the user base moves on to the next thing and new generations start off on something else. Sometimes the ‘next innovation’ is organic (Apple’s sequence of wondertoys), sometimes it is secured by acquisition (Facebook famously increased its longevity by buying Instagram and WhatsApp where it had failed to innovate itself).  Acquisitions that pre-empt challengers or bolt on new functionalities that can morph into future promising rivals have absolutely sealed the position of certain players. Merger control over the past two decades has been an absolute triumph for the defence bar over the agencies. 

We are pivoting to ex ante regulation to overcome the limitations of piecemeal, ad hoc antitrust cases, but we are doing so by putting in place rules that just veto selected conducts taken from the same ad hoc antitrust rules. The rationale for pushing ex ante regulation has been that “we need to complement piecemeal, ad hoc antitrust enforcement with rules capable of being self-executing”. However, it is curious to be doing so by taking ex post antitrust concerns in a handful of cases, and converting them into all-purpose ex ante rules common to all platforms. This generalisation does not work, the rules are not ‘self executing’, and enforcement of the regulation will involve individual litigated cases.

But nothing (at least in the European regulation) speaks to merger control. While some of the US proposals involved explicit designs to ban mergers, the European proposals do not include anything on merger control. We know why this is (technical point about the scope of the legal tool), but a gaping hole.

Very little also speaks about the role of data, the control of data, and the need for a coherent take that recognises abuses of privacy and data protection in preserving power. 

Yet this regulation train has left the station. The regulation juggernaut has got going, and proposals are set to be approved that are not very dissimilar from the original ones of last December (at least in Europe – the US may go through more permutations). The original formulation of the Digital Markets Act (DMA) has been somewhat clarified and amended by the European Parliament’s IMCO and ECON committees,3 but the structure and plan have remained much the same. The US has seen a broad-ranging set of bills sponsored by Congressmen Cicilline and Buck last June,4 and more recently the Open App Markets Act specifically on app stores.5 Commentators have set the DMA and the ‘Cicilline’ bills side by side and found them similar, praising the importance of ‘regulatory coherence’.6 All is well, then?

In this blog I argue that these rules cannot achieve much as formulated. For two reasons: first, because creating ex ante rules that are intended to be self executing by setting out obligations that paraphrase a few past (not very successful) and current (unfinished) antitrust cases is not going work, and there will be scope for endless challenges. Second, because given levels of incumbency and entrenchment of the main players the scope for actually creating new, serious competition with these rules is modest. They can somewhat constrain further expansion (good in itself) and redistribute some rents (though often just to other incumbents), yet to favour real new entry capable of creating diversity we need something else again. 

2 Digital platforms are highly heterogeneous in terms of their incentives, and the problems we are trying to solve differ across them

The concerns we associate with large digital platforms are not homogeneous. The unifying feature is their size and scope, which imply vast bargaining power and enables them to impose one-sided terms on third parties. The DMA’s reference to fairness as one of its animating principles is indeed about this: protecting businesses and users from unfair bargains arising from market power.  But beyond this, there are vast differences in the way that power is wielded. I have expressed this before in terms that correlate with monetisation, or business model:7 digital platforms that monetise in different ways will also tend to have different incentives driving what they do and how they engage with rivals and customers.    

Google is an advertising empire based around its Search function. It monetises on one side of its platform, by exploiting data from user searches to sell advertising – search ads on its Search page or display ads it places on its own properties (e.g. YouTube) and third parties’ estate (publishers). Its goal is protecting the search monopoly, ‘occupying’ new environments consumers may be performing searches from: mobile after desktop, and then wearables and IoT and cars. It also secured control of the vertical chain of intermediation services between online advertisers and publishers. Its incentives are on the user side, to extract data to enrich profiles and place targeted ads; to maximise revenues by directing traffic to its own properties rather than third parties’, especially where it faces a ‘zero price constraint’ and cannot monetise directly; to extract as much rent as possible from the vertical stack; and dynamically, to pre-empt challenges by ensuring no vertical could become a broader search rival, and no one could succeed in getting a chunk of search as it swings to a new environment.

Facebook is also an advertising empire based around its social network. Facebook also monetises on the advertisers’ side, by exploiting data from users’ private life, connections and app use, and placing ads on both its own properties and third-party websites. It operates advertising intermediation services allowing advertisers to buy ads on Facebook’s social platforms as well as on a network of third-party apps and websites. The main concerns around Facebook (see FTC Amended Complaint)8 have been documented acquisitions to pre-empt the growth of rival social networks which could challenge its position. There is also a concern around selectively degrading interconnection or shutting down access to core Facebook services (APIs) to nascent apps viewed as potential threats. One important dimension of these concerns is around extraction, wholesaling, and retailing of private personal life details and amplification of toxic content which increases engagement and creates additional monetization opportunities.

Amazon is a giant ecommerce platform that intermediates between consumers and third-party sellers, and also sells on its own behalf. Amazon is a transaction platform that monetises both through a commission on 3P sales and a margin on its own branded products (i.e. it makes money either way). This makes a big difference when thinking of incentives, relative to an ad-funded model subject to a zero-price constraint.  When it launches its own version of a product, is it because it has incentives to displace 3P sellers, giving itself priority position through its own selection algorithm (e.g. surfacing them as Buy-Box recommendations, which get most conversions), and crowding out/excluding 3P sellers in the process?  Or is it because it can offer a cheaper/better version?  Other questions include whether it has incentives to extract too high commission from 3P sellers to increase their price and marginalise them, or to force them to use its own logistics instead of their own fulfilment/logistic to harm logistic operators. Dynamically, there are questions about whether it is using ‘seller data’ to launch its own competing versions, and whether this undermines sellers’ innovation effort; and questions about the pricing strategy and the role of Prime.  

Apple is a device seller with a very successful complement, the App Store, monetising through sale of devices, services and commissions from developers.  The business model involves a walled garden for the OS and the App Store, with monetisation taking place through multiple levers (device sales, commission from developers on digital sales and subscriptions, and own services). As the app economy has grown hugely in recent years, commission revenues for Apple on the App Store side have soared, driven mostly by games. As Apple also launched a few apps of its own (e.g. music, TV, games), is there an inevitable incentive for it to favour its own apps and disadvantage third parties? Is charging a commission by an integrated operator a way of ‘raising rivals’ costs’ to foreclose rival 3P apps? What about services where it does not have its own app and the commission is the same? Then there are questions about whether the level of the commission is excessive and developers are foreclosed or their innovation undermined. Should regulation aim to shift rent from Apple to app developers? How much, what is ‘fair’? Do we want to eliminate commissions? How should the App Store be monetised? If Apple is to open up to third party app stores, who sets and guarantees acceptable quality levels?  

Microsoft is a software and increasingly cloud provider. It also operates a search engine and has a growing games business. It thus monetises variously through the sale of software, the sale of cloud services, as well as the sale of consoles, games and subscriptions, and through some advertising. 

There are vast differences across these business models that make it hard to discipline the exercise of market power and foster meaningful entry through a handful of general rules extrapolated from a handful of antitrust cases (some recent, some just underway). 

3 Yet the ‘rules’ for regulation are just synopses of recent antitrust cases, somewhat ‘anonymised’ to apply beyond their original setting 

Articles 5 and 6 of the DMA contain a total of about 15 rules (including the IMCO/ECON proposed amendments), each a mini-synopsis of a recent/current antitrust investigation, somewhat paraphrased and generalised and then listed in no particular order. While we can surmise where several of them come from (one is clearly Google Shopping, others are Google Android, Amazon Marketplace, Apple App Store Rules, Google Adtech etc.) there is no discernible organisation and limiting principles. We described it as a ‘game of charades’ in a previous article,9 and the basic structure has not much changed with recent amendments: the list still jumps around from mingling consumer data to MFNs to app store practices, to transparency of terms to publishers and advertisers, to self-preferencing, to interoperability, to data portability, to discrimination.10 The issues in each particular case have been ‘generalised’ to suggest general application and set presumptions.  

But a general rule against ‘conduct X’ is not going to be ‘self executing’ when incentives are very different. Take for example self-preferencing. The DMA text is quite loaded with the general presumption that all integrated platforms will always want to do this, and there is an obvious ‘conflict of interest’ whenever a platform owner is selling its own services too: “Gatekeepers are often vertically integrated and offer certain products or services to end users through their own core platform services. (They) can reserve a better position or differentiated treatment to their own offering, in terms of ranking, as opposed to the products of third parties also operating on that core platform service (…) the gatekeeper is in a dual-role position as intermediary for third party providers and as direct provider of products or services, leading to conflicts of interest” (Rec. 48). In the revised IMCO version, new proposed language in Rec. 49 goes further and adds: “to avoid any conflicts of interest, the gatekeeper should be required to treat its own product/service as a separate commercial entity that is commercially viable as a stand-alone service”. 

But we cannot presume a general incentive to steer consumers away from third party services whenever a platform is vertically integrated and offers a service of its own. 

The incentive for steering will be strong for an ad-funded model with a ‘zero-price constraint’ on one side of the platform, in which case shifting traffic to one’s own properties is the main way to monetize. This was indeed the core of the EC’s Google Shopping case,11 where the ‘zero-price constraint’ was binding as Google had committed to offer organic traffic for free, and a comparison shopping site could appear in organic search results at no cost. In addition, a search engine has strong incentives to divert traffic to build up content to prevent rivals from building their own and become a challenger. Furthermore, being able to track users in different settings that cannot all be monetized creates strong incentives to attract users into the ecosystem and keep them.

The position is different when monetisation occurs through a commission on each sale, for instance. Indeed, a growing economic literature – spurred by the prominence of the issue in the public discourse – has studied the incentives of hybrid ecommerce marketplaces like Amazon, monetising on 3P sales to consumers via commissions, as well as on a margin on its own branded sales.  This work says that (much as in other vertical settings) it is not obvious nor inevitable that a hybrid marketplace will have adverse incentives to displace 3P sellers, or to raise the commission to disadvantage them, or to promote in ranking the marketplaces’ product, or to use their data to enter with an own brand product and displace them, or expropriate their innovation.12 Because as in other vertical settings, the ‘conflict of interest’ is not inevitable when 3P businesses create value that is extracted anyway on each transaction through a commission.  Of course, it is legitimate for the regulator to demand assurances, for instance mandate that a hybrid e-commerce platform regularly shows that a sample of product recommendations is unbiased. But self-preferencing requiring vertical separation is not necessarily baked in for a large hybrid platform. 

Or take Apple, integrated into devices and own services on the App Store. Does it have inevitable incentives to direct consumers to its own apps, at the expense of others, and thus undermine 3P apps? Foreclose them? Again, recent research on this argued that in a business model where the economic value created by the App Store is extracted part through greater sales of devices and part through commission revenues, promoting own apps over 3P apps would only make sense if this could generate higher total surplus. If marginalising 3P apps reduces total surplus, then revenues on own services do not offset this and hampering 3P apps makes no sense.13 

‘Self-preferencing’ ultimately needs to mean a promotion of platform services over 3P party services that also harms consumers, including in the long run through entry effects. Not every form of vertical integration leads to foreclosure incentives, and the purpose of intervention cannot be just to redistribute rents to third party businesses.  Critically, the business model is not neutral about whether the platform engages in self-preferencing.  So how does a prohibition of self-preferencing need to be understood? What should be the presumptions? Why proceed on the basis of listing generic conducts and vetoing across the board, when the issue is what specific incentives for hampering contestability and fairness are at play in each case?  Why can’t we organise the rules in a more useful and effective way, taking into account differences? 

4 A more structured approach 

The DMA rules as currently laid out would benefit from some organisation that made them contingent on specific incentives associated with business models. What constitutes exploitation or extraction of excess rents from an ecosystem (fairness) will vary with business models. Likewise, since entry strategies also vary with business models (the way a new entrant could seek to compete against a social network is different to how one may challenge search, or an operating system, or an ecommerce business), regulations that aims to keep entry barriers low (contestability, essentially low entry barriers) will need to differ across models. Specialising the rules somewhat makes it more likely that one could tailor problematic conduct to a specific platform, and that the platforms could recognise what rules apply to them.  

Let’s take three main types of platforms – by no means a definitive categorisation nor one that excludes hybrids or an evolution of business models in different directions. The first is ad-funded content, which includes video content suppliers (Youtube), social networks (Facebook, Instagram), and search engines (Google search). The second is ecommerce and other marketplaces (Amazon, Uber). The third is operating systems and app stores (Apple, Android, Microsoft Windows), software platforms on which other software applications can be developed.  

In the case of ‘ad-funded’ services, the ‘flywheel’ (i.e. the virtuous cycle that builds user engagement) relies on building up a user base for an interesting service first that then attracts advertisers. The entry path for an entrant thus involves various ways of trying to scale up quickly on the user side, and losses before advertisers arrive. An entry strategy may involve for instance a distribution deal with someone who accounts for a large block of users, to become a default (mobile OSs have been an obvious example); or a deal with an intermediary that can shift customers from one platform to the other; or an acquisition of another player with some existing presence. 

How does a gatekeeper of this type defend its position from those potential entry efforts? By ‘throwing grit’ in the ability to build up user engagement, entrants can be prevented from gaining ‘escape velocity’ (i.e. building enough of a user base on one side of the platform to attract enough paying customers on the other side).  Anything that affects the ability of an entrant to gain some sort of scale, especially in the early stages of growth, can have big effects: growth is slow at first, before accelerating as more users means more sharing activity and hence more user growth, and then slowing again as the addressable market becomes saturated with users and there is less scope to add new users. Exclusionary conduct has the greatest prospect of success when applied early in the lifetime of a service as it can prevent the achievement of critical mass; it will also be more effective against services which are less ‘viral’ (while a service triggering very high levels of sharing activity might be able to still grow, one which is less ‘viral’ may ‘flat line’ and not be able to get off the ground at all). Consumer harm arises because a service which creates value for consumers (in the sense it would be widely used but for the conduct) may never be able to grow enough to harness the network effects necessary for its survival. 

If we want to do something about potential entry ('contestability' is an ambitious concept in these markets), we need to prevent conduct that can slow diffusion by denying scale to rivals or entrants. This includes: contracts to sell default position, because establishing defaults is a way to ensure persistence of users with the platforms; entering into exclusivity deals with distributors that are then unavailable to help potential challengers gain scale quickly;  making/buying a service and promoting it or advantaging it to take away customers from the competing vertical, by manipulating traffic to oneself or using integration and bundling/tying to force users to adopt it (i.e. 'self-preferencing' to steer demand away from the rival to the incumbent). 

In terms of fairness, since these business models are premised on data extraction, we need extensive rules on data collection and use – for instance, no mingling of data, no use of data without consent (to complement GDPR), no opaque rules of engagement.

The rules will be different in the case of operating systems or app stores. Realistically, new entry at the platform level is especially hard in these cases – aside from the large investment required, an entrant at the platform level would need to attract developers/apps and users, which is a hard ‘chicken and egg’ problem. It could perhaps start developing some apps in house to create a user base, and then try to incentivise users to switch from other platforms in various ways. The fact that platforms are sometimes bundled with hardware (Apple iOS and App Store) of course makes de novo entry more challenging.  That said, there could be perhaps the potential for competition from a platform built on top of an existing device (e.g. Samsung) or some other type of platform that can perhaps further extend as an app store (Amazon’s Fire OS). 

But while it is hard to think of real ‘system competition’, it is still important in these ‘operating systems or app store’ models to protect free and easy multihoming. Does a platform make it difficult for developers to distribute in a different and competing channel or make it difficult for users to port their content across platforms and to multihome?  Does a dominant platform tie up apps/developers with exclusives (so they are unavailable to others)? Does it place obstacles for developers (e.g. through the use of tools) to operate across platforms (restricting interoperability)? Does it make it difficult for developers to add consumers through other channels outside the platform? 

As to fairness, a legitimate question is whether the commission charged for distribution and intermediation is ‘excessive’. This requires however an understanding of the business model, how commission schedules are structured, and the interplay with other monetisation levers.  Another fairness question is whether the process of vetting and curating apps and 3P businesses before allowing them on the platform is exceedingly strict and cumbersome. This cannot be entirely left to unilateral disposition and there needs to be minimum standards and recourse. 

Different again is the case of marketplaces and transaction platforms. The business model in this case usually earns a fee or percentage from each sale. The platform therefore has strong incentives to make a match in order to get sales, and (as explained) not necessarily incentives to 'self-preference'.

Here again entrants require the ability to scale up rapidly to match enough demand and supply to provide adequate service quality (e.g. achieve enough density of drivers and riders/eaters in a ride-hailing/food delivery business). Multihoming is again key on either or both sides of the platform in these businesses. In many cases the addressable markets are large, and switching in some popular applications tends to be easier. Strategies to block entry or disadvantage rivals can include schemes to tie up one side into effective exclusivity, or to use incentive contracts in various ways to pre-empt multihoming; and MFN-type clauses that prohibit rivals from benefiting from better terms/prices and reduce the ability of sellers to give better deals to others. 

In terms of fairness, given its market power, does the platform adopt conduct that exploits some form of advantage over others that depend on it (for instance, privileged access to data in a manner which is not shared with businesses that contribute to generate that data)? That would be problematic, so it is right to rule it out and request that systems are put in place to avoid it. And while one cannot presume that there is an inevitable incentive to 'self-preference', we would still want to make sure that the incentives operate the right way in practice. It is perfectly fair to request evidence that a marketplace relying on an algorithm for surfacing its recommendation to consumers is designed in a way that does not favour its own operation over third party sellers that rely on it. 

Overall, the formless list that is the DMA needs to be parsed – at a minimum – into rules that are not just a rendition of current/recent antitrust cases (some unsuccessful some underway) but reflect where the bottlenecks to competition are in each case, and what specific intervention can benefit consumers and businesses (and not than just redistribute some rents).  And do so in ways that are specific enough and intelligible by each, or we will be stuck in endless challenges.  If the DMA cannot be purposed in the direction of the UK CMA’s bespoke codes of conduct, some specialisation is still needed.

5 What else do we need to think about for a chance at more diversity?

If clearly and thoughtfully organised, conduct regulation of the status quo players can help pre-empt too much more flexing of power, and preserve some competition ‘on the platform’ – at the margin.  Will it lead to actual new entry and new models, new technologies, new solutions capable of eventually overtaking today’s incumbents? Of course not. They are too large, too entrenched, too successful. We can protect and perhaps enhance competition ‘on the platform’ a bit, but if what we object to is size, persistence, power, technological change that is shaped by the GAFAM’s needs, a regulation regime that is basically repurposing the usual ex-post competition rules will not create serious challenges. Nor will corrupt class actions on the back of antitrust claims, and damages actions from those who believe they have been wronged. 

A set of rules like the DMA will be ‘navigated’, reduce and redistribute some rents, preserve some competition at the margin, but will not lead to anything approximating a landscape change, with new innovative businesses capable of creating more diversity and serious choice, and real innovation which is not dependent and subordinate to the incumbents: with every new digital institution born as a startup or an app, the same set of ingredients (users, advertisers, app developers) and the same set of economic imperatives: users need to be monetized; data needs to be gathered; subscriptions need to be sold.14 For this we will need something a lot more forward looking. 

At the most immediate level we need to think seriously about putting serious brakes on mergers by gatekeepers, and forcing targets to find a way to market instead of always selling to the monopolist. A regulation regime that is silent about merger control is pointless. The world is awash with cash, and the single most harmful development to competition in recent times has been how startups and new innovative businesses have come accustomed to think just about exit by buyout.15 By implication, we also need break ups which undo mergers we know ex post to have been anticompetitive. This is not for the faint hearted, although possible in Europe, but US regulators and legislators can certainly do this. 

Second, we need wall-to-wall interoperability obligations at each pinchpoint and bottleneck: only if new entrants can connect and leverage existing platforms and user bases can they possibly stand a chance to develop critical mass. Data portability alone is just a marginal solution (and no wonder it is the ‘go to’ remedy for GAFAM when they want to flag their good intentions). That interoperability needs to be the central plank of any regulation regime has been said multiple times now.16 We need the ‘privacy people’ involved, to stop saying there is a tension with privacy, and get it done.

Third, bring data protection and privacy rules much more to the fore, and work with these as complements to address directly the sources of power.17 There is a body of unenforced data protection rules in Europe (the GDPR as the key plank) which can complement regulation by getting directly to the source of exploitation – the ability to harvest and use data across services without restrictions. Think of purpose limitation rules in the GDPR for instance, and how they come close to ‘line of business restrictions’.18 Again it is time to end the narrative that there is a ‘tension’ between competition and privacy goals, and look for complementarities. 

Finally, and critically, we need new thinking and new initiatives to take us past the current paradigm of unregulated private digital infrastructures operated as fiefdoms. With the current distribution of digital assets and infrastructures it is hard to see how incumbency based on huge economies of scale and network effects cannot be preserved almost indefinitely, and no replacement candidate is likely to come to the fore that can diversify the landscape. There is developing thinking around ideas like new digital commons, and shared digital infrastructures – enabling new businesses to emerge by leveraging scale on standardised and interoperable open platforms for an open web, without toll bridges and bottlenecks. This requires multilateral cooperation between governments, open source software and civil society. But note that the idea of common-pool digital public goods is not that revolutionary: parts of the digital world already function as commons (Internet protocols, governed by international institutions and open standards, the open-source software that enables these protocols, and much of the information layer of the Internet). Some are thinking about these commons-based idea of technology as opposed to proprietary structures that protect the current incumbents. And perhaps ideas like data commons too.19 Backward-looking regulation which is no more than ersatz competition law, enforced by regulators with no idea about cutting-edge technology, is not going to move the dial significantly beyond GAFAM. 

With current proposals, the GAFAM will successfully manage change and DMA-like regulation over the next ten years.  Nothing much will happen. There will be challenges, compliance programmes, lengthy delays (some legitimate, as there are real ambiguities to be clarified), but nothing will fundamentally change in the direction of more diversity and choice without new opportunities for building new systems and challenges. And for this we need to think beyond rules that are only capable of catching some of the conducts but still require antitrust-like analysis and processes. 

Disclosure: I have assisted clients and agencies in matters adverse to Google over the past 8 years.  Over the past 3 years I have also advised Amazon, Apple, Microsoft, Newscorp, Uber and Netflix among others on antitrust and regulatory matters.  


1 There is a vast literature at this point on concerns around large digital platforms, which extends from foundational reports (StiglerFurman) to competition agencies digital reports (UK CMA, Australian ACCC, French Adlc etc.), to academic research on market concentration (Autor, D., D. Dorn, L. F. Katz, C. Patterson, and J. Van Reenen, (2020): “The fall of the labor share and the rise of superstar firms,” The Quarterly Journal of Economics, 135, 645–709, J Eeckout, The Profit Paradox, Princeton University Press 2021), to a huge body of papers, articles, blogs, and analyses of antitrust violations, distortions in ad tech markets, concerns around ecommerce, concerns about app stores, of privacy violations, all the way to  distortions of the public discourse.  For a series of articles on several dimensions of the problem see the work of Cory Doctorow at @doctorow. I will not attempt to summarise or reference this material further in this note.

2 Mark Lemley also very recently wrote on The Contradictions of Platform Regulation, and while I do not agree with all of his analysis, he also makes good point about contradictions in current designs.

3 See the Draft Report of the Committee on the Internal Market and Consumer Protection, and the Draft Opinion of the Committee on Economic and Monetary Affairs.

4 These include the American Choice and Innovation Online Act, (H.R. 3816), Augmenting Compatibility and Competition by Enabling Service Switching (ACCESS) Act (H.R. 3849), Ending Platform Monopolies Act (H.R. 3825),  see


6 See Yale Tobin Centre for Economic Policy, International Coherence in Digital Platform Regulation: an Economic Perspective in on the US and EU Proposals, 6 August 2021.

7 See Caffarra 2019, Caffarra, Etro, Latham and Scott-Morton 2020, Caffarra and Scott Morton 2021.

8 See

9 See Caffarra and Scott Morton 2021.

10 Post-amendments, as of end-August, Art. 5(a) starts with a prohibition of combining users’ data without GDPR-like consent; 5(b) goes on to prohibit a platform from restricting the ability of business users to offer their products under more favourable conditions on other channels (MFN-like clauses); 5(c) prohibits anti-steering provisions; 5 (e) prohibits restrictions to just use the payment service of the gatekeeper. 5(g) is an obligation to provide information to publishers and advertisers; 5(ga) prohibits the use “in competition with business users, (of) any data which is generated through activities by those business users”, 5(gb) is about uninstalling apps and software.  Art 6(c) creates the obligation to “allow and technically enable the installation and effective use of third-party software applications or app stores using, or interoperating with, operating systems of that gatekeeper”. 6(d) is about self preferencing by gatekeepers that are vertically integrated and offer (their) products/services to end users through their own core platform services; 6(e) is about “refraining from technically restricting the ability of end users to switch between and subscribe to different software applications and services to be accessed using the operating system”. 6(f) is about interoperability: 6(g and i) are about access to data and data portability. Finally 6(k) requires the gatekeeper “to apply access conditions that are FRAND or not less favourable than the conditions applied to its own services”..

11 See

12 Early work on this includes Jiang, B., Jerath, K. & Srinivasan, K. (2011). Firm strategies in the “mid tail” of platform-based retailing, Marketing Science, 30(5), 757-775, and Hagiu, A. & Wright, J. (2015). Marketplace or reseller?, Management Science, 61, 1, 184-203. Since then there have been Etro, F. (2021). Product selection in online marketplaces, Journal of Economics and Management Strategy, Vol. 30, 3 (Fall), pp. 1-25, and Hagiu, A., Teh, T.H. & Wright, J. (2020). Should platforms be allowed to sell on their own marketplaces?, SSRN 3606055. Work showing that the effect of a hybrid marketplace depends on the nature of the demand system, and the commission can be higher or lower are respectively Anderson, S. & O. Bedre-Defolie (2021). Hybrid platform model, CEPR DP 16243, and Etro, F. (2021), Platform competition with free entry of sellers, SSRN 3901080. For a discussion of entry and imitation incentives see Masden, E. & Vellodi, N. (2021). Insider imitation, mimeo, New York University.

13 See Etro, F. (2021). Device-funded vs ad-funded paltforms, International Journal of Industrial Organization, Vol. 75 (March), 102711 and (2021), Platform competition with free entry of sellers, SSRN 3901080. Also Jeon, Doh-Shin and Patrick Rey, 2021, Platform competition, ad valorem commissions and app development, mimeo, Toulouse School of Economics

14 See Daron Acemoglu, Antitrust Alone Won't Fix the Antitrust Problem, Project Syndicate October 2020; and Evgeny Morozov, Privacy activists are winning fights with tech giants. Why does victory feels hollow?, The Guardian 15 May 2021.

15 See for instance Valletti 2021 and Valletti and Kwoka

16 See Cory Doctorow, Adversarial Interoperability, EFF, October 2019, or Tech Monopolies and the Insufficient Necessity of Interoperability, Locusmag July 2021.  Also Crawford et al., Equitable Interoperability: the “Super Tool” of Digital Platform Governance, 14 July 2021, Yale Tobin Digital Regulation, also Michael Kades and Fiona Scott Morton, Interoperability as a Competition Remedy for Digital Networks, Equitable Growth September 2020.

17 See        

18 See, and

19 See and