This paper – which you can find here – asks whether big data should be treated as an essential facility. Without getting into matters relating to accuracy, freedom of choice, pluralism, or even whether regulation is adequate, this paper simply explores whether and how antitrust law could impose a duty on dominant firms to share their big data.
The paper is structured as follows:
- Section 2 seeks to debunk the current (antitrust) narrative about big data. To summarise, the argument is that while data is an important input, it is not that different from other inputs. Unlike what certain authors seem to hold: (i) the economic utility of big data does not depend on the data as such, but on the material and intellectual resources that a firm invests in developing the analytics necessary to draw reliable and grounded inferences out of that data; (ii) data is not a single input like oil, but is instead a concept that contains multiple distinct inputs that may be specific to each firm; and (iii) when looking at the industrial process that leads from big data to new products and services, antitrust scholars should distinguish not only the many markets for digital data to which firms have access, but also the different markets for the many kinds of information that can be extracted from those data.
- Section 3 attempts to clarify the meaning of “exclusivity” and “barriers to entry” in order to determine how they apply to big data. In short, the authors argue that there is no reason to apply them any differently to big data than how they are applied in other industries.
- Regarding exclusivity, while digital data is ubiquitous, there are only two scenarios where only one firm has access to some digital data: (i) when there are legal restrictions regarding access to data; (ii) factually, when only one company has access to the specific data. While this data will then belong exclusively to some company, that does not mean that the data is essential (this would require a downstream market for which the data is essential), nor that the firm has monopoly power in the generation and collection of that data (that would require a market to be identified in respect of this data).
- Regarding barriers to entry, there are two ways data can be collected: (i) by acquisition in data marketplaces; (ii) by entering into vertically the upstream markets for the generation and collection of the data that the companies desire. Evidently, acquiring data implies incurring costs. Other costs include those related to the acquisition of the infrastructure and related managerial and engineering skills.
With big data access, antitrust scholars have stressed that firms may work to prevent their rivals from generating and collecting some digital data by concluding exclusive deal agreements, by merging to create a data bottleneck, or – when dominant – by adopting foreclosing strategies. However, this breakdown of barriers to entry is not new or restricted to digital markets – it applies to any industry. The same goes for the observation that firms operating with big data often have high upfront costs with zero-marginal costs, which can result in economies of scale and scope for those established firms.
- Some scholars argue that big data gives rise to a new set of concerns regarding barriers to entry, however. This argument takes two forms: (i) some argue that big data consolidates the indirect network effect that already characterizes two-sided business models, thereby reinforcing the market power of digital platforms who can anticipate market demand and improve their products; (ii) a slightly different argument is that big data puts digital platforms in the position of benefiting from a strong information asymmetry that prevents customers from determining whether platforms’ goods are as good as they could be and whether the platforms are somehow manipulating the information that they provide. However, the authors argue that the first form of the argument is not supported by empirical evidence, and that the second form is not concerned with matters relevant to competition law.
- Section 4 then reviews the application of the essential facilities doctrine to data. This section begins by reviewing the essential facilities doctrine in the US and in Europe. Its assessment, in short, is that: “the EU and U.S. approaches diverge with respect to antitrust liability in cases of refusal to deal. In the U.S. the sole exception to the principle that firms have no duty to aid competitors by granting them access to their resources applies when a monopolist seeks to achieve an anticompetitive end by terminating a voluntary course of dealing. In contrast, the EU case law has defined a framework of exceptional circumstances according to which a refusal to deal may involve abusive conduct.” Despite these differences, the authors conclude that under either variant of this doctrine data cannot be considered an essential facility.
Even if an infringement were found, the authors point out that additional issues would arise regarding the appropriate remedy. One would assume that these remedies would take the form of compulsory licenses. However, it is unclear how one could apply such a remedy to big data because: (i) one would have to identify which data is subject to the duty to share, but the requesting firm cannot know ex ante the relevant dataset needed within the entire raw database available to the incumbent; (ii) issues regarding the updating of data would arise; (iii) major concerns would arise regarding the terms and conditions for the licences.
- Section 5 argues that big data is the wrong target of competition: instead, the focus should be, if on anything, on the information extracted from big data. It is only information (and the analytical tools used to extract it) that should be considered an essential facility. “Hence, the notion of big data is misleading twice over. Firstly, it suggests that big data are a homogeneous resource whose features are given and whose applications are known in advance, when they are nothing of the kind. Secondly, it draws attention towards a step of the data value chain (data collection) that is less important than others (data processing and information production). Likewise, the attention given to the definition of data markets is malfunctioning, especially when it is not accompanied by an equal focus on the definition of the intermediate markets for the pieces information extracted from the big data.”
This paper provides a fairly comprehensive review of the literature on antitrust and big data. It is rigorous, detailed and very much worth reading for anyone interested in big data.
While I really enjoyed the paper, I do not think it is without issues. I’m not sure I agree with the authors’ view that the essential facilities doctrine is “the major point of the EU/U.S. transatlantic divergence” or that “it represents the main antitrust tool in the EU for overseeing intellectual property” (for that matter, I’m not sure I agree with the authors’ reading of the essential facilities doctrine in the US following Trinko either). I also don’t see how difficulties about the application of compulsory licenses are relevant to the issue of whether big data can amount to an essential facility.
Perhaps more importantly, the paper does not really address whether big data poses problems for market structure (nor does it seek to do so). But, arguably, the problem of big data from a competition perspective is that it allows for the entrenchment of corporations with market power. I am not familiar with any previous case or article that argues that a refusal to supply data is a breach of competition law (which is where the essential facilities doctrine would kick in). I am very familiar – as no doubt is everyone reading this – with literature and authors expressing concerns about the potential impact of big data on the level of competition in the market.
Thus, the very rigour of the paper ensures that it does not look at one of (if not the) most important questions regarding big data and its use, i.e. whether regulatory responses may be appropriate. What the paper makes clear, however, are the limitations of competition law in dealing with such a complex issue.