Shaping competition policy in the era of digitisation

Executive Summary

This briefing note summarises some of the key points discussed at the European Commission's conference on "Shaping competition policy in the era of digitisation" held on 17 January 2019. The topics of the panel discussions were:

• Competition, Data and Artificial Intelligence.
• Digital Platforms' Market Power.
• Preserving Digital Innovation.

The conference spanned a wide range of topics around the theme of digitisation and competition law. This note touches on some of the main competition concerns emerging from the discussions, namely:

• big data and network effects as drivers of market power;
• the risk of consolidation and foreclosure through abuse and buy-outs;
• the connection between innovation and public welfare; and

• the measures proposed to "re-tool" competition policy and enforcement to address these issues, including: treating data as an "essential facility", shifting the burden of proof in big tech acquisitions, collective bargaining by businesses on dominant platforms, and a greater emphasis on consumer behaviour in online environments.

The panellists and speakers comprised policy makers, academics, CEOs and journalists, but did not extend to representatives from the technology sector. This was reflected in the approach taken to some of the issues discussed. A list of speakers are listed here.

This note seeks to summarise the views of the panellist at the conference and does not represent Ashurst's views on any of the issued discussed. A full video of the conference is available here.

Welcome Speech

Commissioner Vestager (EU Competition Commissioner).

Commissioner Vestager's welcome speech focused on the question of how competition policy should respond to digitisation. She suggested that it is important to make these decisions now, given the impact digitisation is likely to have on everyone's future. She argued that, collectively, all stakeholders had a responsibility toward future generations to get this right.

Panel I:- Competition, Data, Privacy, and Artificial Intelligence

Panel chair: Yves-Alexandre de Montjoye (Assistant Professor of Data Science, Imperial College, and Special Adviser to Commissioner Vestager). Panel moderator: Hal Hodson (The Economist). Panelists: Ariel Ezrachi (Slaughter and May Professor of Competition Law, University of Oxford), Alessandro Acquisti (Professor of Information Technology and Public Policy, and PwC, William W). Cooper (Professor of Risk and Regulatory Innovation, Carnegie Mellon University), Karen Yeung (Interdisciplinary Professorial Fellow in Law, Ethics and Informatics, Birmingham Law School).

Data as a unique commodity: 

• Data is difficult to define: Data is an essential input in an increasingly digitalised market which evades conventional economic analysis. Primarily,, data presents different contours and qualities which makes it difficult to precisely define. 

• Data is difficult to value: There is also a general lack of empirical data on the value of data. Consumers are generally unaware of the value of the data they share and there is also a considerable lack of transparency on the part of the companies which harvest and analyse it. However, the value of data displays a unique link with privacy as most data is personal. When connected with other data its value in terms of predictive ability increases as do the privacy risks associated with its use. By consequence measures taken to improve privacy tend to degrade the quality of the input.
• Data ownership is difficult to establish: Furthermore, data is an input not entirely owned by the holder, which is subject to the providers consent. Nevertheless, data is understood to have important re-allocative effects. creating winners and losers.

Data as an essential facility:

• Access to data as a competitive advantage: In an era of digitisation access to large data grants provides a major competitive advantage. Data provides key access to consumers, notably through targeted advertising, and can be used to exert considerable influence on decision making. Machine learning has also become an important component in innovative products and services. Whilst hardware and software are required to implement these techniques, access to large data sets is by far the most important ingredient. This tilts the competitive landscape in favour of incumbents who have been able to shape consumer ecosystems in ways that allow them to harvest a larger share of that data, creating bottlenecks.
• Granting access to date: However, as discussed above, data is not a typical input. EU rules on data privacy (as set out in the GDPR) set important limits on wider access to this data. Responses must also balance the incentive to invest in innovation with the need for greater access. While mandatory data sharing remedies might have some merit, any such approach should be assessed on a case-by-case basis and take into account: (a) whether the data is collected on the company's primary or secondary market; (b) whether the data was provided willingly or the result of a "take it or leave it" ecosystem; and (c) the impact of these measures on innovation as well the kind of innovation pursued.

Digital environments provide their users with the illusion of freedom and control:

• Consumer control: Platforms use consumer data to engineer "triggers" and "friction" to guide consumer choice and lock the consumer in to that particular platform. Some forms of innovation, whilst convenient, may further limit consumer control. Voice recognition applications are one example of this. These technologies give even greater control to the platform provider in guiding consumer choice.
• "Free" services fuel network effects: It was discussed that "free" services and network effects tend to lock consumers into using a particular platform, which facilitates a cycle of greater data harvesting improved predictive ability and, thereby greater network effects. Insofar as large scale data harvesting is at the heart of the digital economy, closer competition law scrutiny of consumer lock-in mechanisms may be required.

• Assumptions about consumer benefit from "free" services lack an empirical basis: The correlation between higher levels of data collection and more free services does not entail causality. There is a lack of data on whether increments of data collection are in fact necessary for these free services to exist, and further whether these increments always lead to improved services or just greater profits.

Innovation's dark side:

• Public vs Private: Innovation can lead to public and/or private good. The latter may in some circumstances be detrimental to the first (by making it easier to manipulate consumers into giving away more data for instance). Innovation can have a dark side, impinging on privacy, welfare, human dignity, the rule of law, democracy and respect for human rights.
• Regulation does not necessarily stifle innovation: If designed well, regulation can nurture socially beneficial forms of innovation. Well-designed regulation has knock-on benefits for industry, for instance in greater consumer confidence in innovative products (the failure of Google Glass is an illustration of this point).

• Avoiding a race to the bottom: Innovation should be viewed not as a mechanical process, but as a human and social process which requires difficult choices which are value driven. Whilst all states may not share a common vision for socially beneficial innovation, an international regulatory race to the bottom does not further the common public good.

Panel I Keynote Speech:- Competition Policy in the Digital Age

Jean Tirol, Chairman of the Toulouse School of Economics, laureate of the 2014 Sveriges Riksbank prize in economic sciences in memory of Alfred Nobel

Public intervention in the digitalised economy is unavoidable:

• A combination of competition enforcement, regulation, industrial policy and innovation policy are likely.

• More regulation is required as opposed to self-regulation, which is often self-serving. Solutions should err on the side of competition policy, however the competition toolbox will need to be updated.
• Competition policy tends to be slow, arrives too late and does not have obvious remedies at its disposal. An adapted competition toolbox will require more participative antitrust regulation, characterised by greater industry input, more regular guidance and more coordinated action between agencies.
• A better understanding of new business models is required (e.g. with respect to multi-sided markets).

• More drastic solutions, such as breaking up monopolistic providers of digital infrastructure may be necessary, although there are no well formulated proposals to do so at this stage.

Contestable monopolies are not necessarily bad for competition:

• … so long as efficient rivals are "able to enter" and "enter when able". Contestability is the key to competition in digital markets, characterised by high levels of concentration.
• Monopolies often fail to innovate: there is little incentive to compete with one's own products and services, and little competitive pressure on management.

Efficient rivals must be able to enter the market:

• Digital platforms are characterised by very strong network effects which lead to a winner-takes-all phenomenon and market power caused by direct and indirect. Dominant firms are able to create and shore-up market power through preference for their own services, predation and deterrence of multi-homing.
• Consumers are not generally inclined to switch away entirely from an incumbent platform, which underlines the importance of market overlaps for potential competition. Nevertheless, whilst it is cheap to multi-home, consumers often do not do this, a fact which is partly explained by inertia on the part of consumers.

Contestability also requires that firms enter when able:

• "Entry for buyout" is liable to suppress price competition but also innovation (though "killer acquisitions"). Given the pace of innovation, the common lack of data of the effects of tech acquisitions and the clear incentive for incumbents to buy-out potential competitors at an early stage (before that data can be generated), it may be necessary to shift the burden of proving the pro-competitive effects of acquisitions on the dominant undertaking. Speakers in later panels commented on how the "entry for buyout" trend skews potential rival's incentives toward "rent-seeking" innovation – innovation to attract incumbent buy-outs – which does not contribute to the public good.
• Facebook/WhatsApp and Facebook/Instagram: The Nobel laureate cited the Facebook/WhatsApp and Facebook/Instagram acquisitions as examples of acquisitions which his "gut feeling" tells him are likely to be anti-competitive but which regulators lack sufficient data to assess.

Industrial policy is part of the solution to the EU's poor performance in global rankings of top global tech firms:

• The top 20 positions are split evenly between the US and China. Targeted industrial policy has been successful, taking DARPA as an example, but must be designed properly.

• Eight key recommendations to policy makers:
  — identify the market failure so as to design the proper policy
  — use independent high-level experts to select projects and recipients of public funds
  — pay attention to the supply side (talents, infrastructure) and not only the demand side
  — adopt a competitively neutral policy
  — do not prejudge the solution
  — evaluate ex post and disseminate the results, include a "sunset clause" in the program, forcing its closure in the event of a negative assessment
  — involve the private sector in risk taking, so as to avoid white elephants
  — strengthen universities

GDPR is "well meaning" but may be counterproductive:

• GDPR runs risk of further enshrining dominance: The merit of the regulation is making the resale of data harder and giving consumers greater control over their data. However, as currently formulated, it further concentrates data holdings in the hands of the incumbents.
• Further coordination is required between the different agencies to make a more effective policy landscape. In particular, data protection rules do not currently account for limited cognition. More must be done to reduce information and transaction costs for consumers, for instance through standard data protection templates.

Panel II:- Platforms and Market Power, Bottleneck Power, Network Effects and the Rick of Lock in

Panel chair: Jacques Crémer (researcher and former Professor of Economics at the Toulouse School of Economics, and Special Adviser to Commissioner Vestager). Panel moderator: Lewis Crofts (MLex). Panelists: Fiona Scott Morton (Theodore Nierenberg Professor of Economics, Yale University School of Management), Sandeep Vaheesan (Legal Director, Open Markets Institute), Monique Goyen. 

Downstream cooperation to counteract platform power may be necessary:

• Collective bargaining solution: Platforms share certain traits with governments, and abuse of market power could be dealt with through collective bargaining. From the platform participant's perspective, platforms such as Amazon resemble a form of government. Participants rely on its infrastructure, pay taxes in the form of fees and are required to abide by the platform's rules. Platforms also determine the user's access to markets and are able to make or break businesses. Platforms have, and often exercise, the power to discriminate (based on vertical integration or contract), to charge monopolistic fees and closely monitor platform users. One solution to tame and contain market power is to allow retailers to develop collective bargaining power: coordination between sellers "banding together" to obtain concessions.
• Antitrust exemptions required: Whilst this collective bargaining solution may weaken competition between platform users, antitrust exemptions should be considered for smaller companies so long as these exemptions are accompanied by greater oversight.

Competition policy must pay greater attention to consumer behaviour:

• Consumers are locked in: Competition policy which considers consumers as economic units will fail to curb market power in digitised industries. Competition policy must be further integrated with consumer protection and data protection rules. Since data is a driving force behind market power in the digital market, competition policy needs a clear understanding of the behaviour which leads to the accumulation of data. Often consumers are presented with a "take it or leave it" choice when using digital platforms. The personal data acquired through this conduct compounds market dominance. The direct and indirect network effects these platforms benefit from have the effect of locking consumers in to the services they provide.
• Consumers are not rational: The current solutions to this problem do not sufficiently take into account the "limited rationality" of consumers, who do not have the time to make meaningful choices about sharing their data based on complex terms and conditions. In the digital market, dominance can be fuelled by convenience. This underlines the link between competition policy and data protection; combating market power entails building privacy by default into digital environments and enhancing consumer ability to make meaningful choices on the use of their data, given the time and cognitive space available to them in these environments.

• Competition policy needs to move away from the online/offline dichotomy: Ever increasing spill-over effects into offline markets make the distinction less useful. Competition authorities must take a more holistic approach, which may include wider market definitions, in recognition of this dynamic.

Panel III:- Competing with Data - A business Prespective

Moderator: Jennifer Baker. Panelists: Oliver Bäte (CEO of Allianz SE), Vittorio Colao (former CEO of Vodafone Group plc).

Data disrupts conventional market analysis:

• Large data-sets = market power: Large data-sets allow undertakings to exert considerable influence on consumers, both within the undertaking's primary market and without. This ability to influence consumer choice translates into market power. Competition law analysis of dominance should not be limited to well defined markets, and should be more focussed on consumer behaviour and access to consumers. Undertakings with a low market share in a particular market may in fact be able to influence consumer decisions in a number of markets.
• Competition policy must adjust: Forward looking competition policy and enforcement must be able to factor in this new form of market power. This has important implications for the way competition policy deals with the "gatekeepers" of the digital ecosystems, who, with developments such as the Internet Of Things, are increasing their capacity to influence consumers.

Data intermediaries are required to monetise consumer data:

• Fair price for data: Consumers are not aware of the value of the data they share, but access to data is a key component for competition. In order to create a true market place, intermediaries should play a role in ensuring consumers obtain a fair price for their data. Intermediaries, auctioning data on behalf of a large number of consumers, would also ensure a more competitive allocation of the rents flowing from this data and ensure wider access to the data itself.

• Data protection is key: Data protection must be at the centre of this market driven solution, as consumers must trust these intermediaries for the model to work. Indeed, greater privacy could be considered a competitive advantage in this model. The GDPR provides the foundation for the monetisation of data (through data ownership) but we should also be realistic about the way consumers behave (without realising the true value of their data). Monetisation is key to ensuring effective competition between platform providers; only then will consumers be able to assess the cost of the "free" services platforms provide.

Panel IV:- Preserving Digital Innovation Through Competition Policy 

(Panel chair: Heike Schweitzer. Panel moderator: Natalia Drozdiak. Panelists: Professor John Van Reenen, Professor Rupprecht Podszun, Professor Mariana Mazzucato).

Innovation in the digital economy differs from more traditional markets:

• This requires new theories of innovation-related harm: Innovation in this field is different from innovation in pharmaceutical or agrochemical industries. It is typically more short term, less asset driven, less based on intellectual property rights and more ad hoc.
• Updated tool-kit: In light of these differences, an updated competition tool-kit is required to preserve innovation in digitised markets. This may include greater emphasis on data pools, data traders and data sharing agreements, given the importance of access to larger data sets to innovation in this sector.

Digitised markets require novel theories of innovation-related harm:

• Novel theories of innovation-related harm are required to accommodate for the way innovation tends to occur in these markets. In terms of merger control, a broader analysis of potential competition is necessary to account for the fact that market boundaries are constantly evolving. Concerns over "killer acquisitions" may also require enhanced scrutiny of strategies to buy-up potential competitors.
• Burden of proof on dominant undertakings: Furthermore, in digitised markets characterised by uncertainty, we may need to move away from an assessment of whether a merger is more likely than not to lead to a significant impediment of effective competition. One solution, also put forward by Jean Tirol, is to place the burden of proof on dominant undertakings. In this regard, where there are strong and entrenched positions, as with digital platforms, competition policy should be less concerned with false positives (i.e. an over-protective stance to mergers) than false negatives (i.e. letting harmful mergers through).

In the face of uncertainty, value judgements used to shape competition policy should be guided by key presumptions:

• The key presumptions:
  — that product innovation should prevail over platform innovation (as the former is often more beneficial to the consumer);
  — that competition policy should strive to preserve diversity (given that digital markets naturally tend toward concentration); and
  — that intervention should ensure that consumers are able to choose which innovations succeed and which fail, not the gatekeepers
• New approach needed: The current approach to enforcement under Articles 101 and 102 TFEU displays a price competition bias which makes it ill-suited for digitised economies. The remedy to this is more innovative and proactive enforcement, pushing pioneering theories of harm to the fore. Competition enforcement should embrace its role as the primary fora for addressing novel and complex economic problems.
• Legal uncertainty: However, the legal uncertainty generated by this bold evolutionary approach should be balanced with lower sanctions.

Digitisation as a driver for market power:

• "Superstar" firms: Digitation has led to increased concentration and increase of price/cost margins, shifting market shares towards "superstar" firms in all industries. The causes are twofold:
  — First, platform competition and network effects create winner-takes-all effects.
  — Second, a fall in price and improvements in the quality of software and ICT enables large firms to make greater investments in data collection and analysis, which in turn leads to higher revenues, compounding the "superstar" competitive advantage (the "Walmart effect").

• Innovation may consolidate market power: Even if market power is gained through competition on the merits, there is a real risk that innovation will not lead to consumer benefits, but rather serve to consolidate these positions of power.

Policy makers should think more about "shaping":

• Competition enforcement is reactive by nature: The EU must consider how it can accommodate enhanced state intervention which does not only focus on fixing market failures but sets out a market shaping agenda. Innovation is often a product of mission-oriented state policies: the state can and should play an important role in crowding-in the private sector to solve big problems (such as climate change). This may require changes to current rules on state aid.
• Role of the state in innovations: Conversely, it is important to recognise the role of the state in important innovations such as GPS and the Internet, particularly when discussing the fairness of taxation on tech giants such as Google or Uber.

Closing Remarks

Johannes Laitenberger (Director-General for DG Competition).

EU competition law is not unfit for purpose. But regulation cannot continue on a "business as usual" basis. The TFEU rules provide a broad base for policy and enforcement, which are able to accommodate changes and new phenomena. But it is clear that tests, theories and procedures need to be updated to adapt to the digitised economy.

The Commission must adopt a "falsificationist" approach to these issues, constantly testing prevailing theories for error. The Commission must be prepared to revisit assumptions and theories in the light of new facts. The complexity and novelty means that the Commission must place even more emphasis on understanding new markets.

Next steps: The Commission's report on shaping competition policy in the era of digitisation, drawing on the discussions held in the framework of this conference and the written stakeholder contributions is expected this spring.

With thanks to Zac Davies of Ashurst for his contribution.