Data has a tendency to be misunderstood. It is often compared to a raw material like oil, waste, or capital. Elettra Bietti argues that misconceptions of data percolate to antitrust, producing impoverished regulatory approaches to digital markets.

In early February, I attended “The Internet’s Midlife Crisis” conference at UC Boulder. I was asked to speak about data in competition policy. I realized that I never formulated comprehensive views on data’s role in competition policy.

Here is a brief articulation of these views, organized in three parts.

Data Acquires Value in Particular Contexts

Although talk of data in competition and antitrust policy is ubiquitous, very few in competition and antitrust circles have asked the question: what actually is “data”? One might define data as an a-contextual set of 1s and 0s, a material “thing” or “object” that exists “out there” in a fixed material form. Yet in practice data acquires social, legal, and economic value in particular contexts. Data acquires the shape, functions and purposes that are projected onto it by those who collect, store, process, organize, and use it. Data is more often than not the by-product of an infrastructural process of production and accumulation designed with particular ends in mind. 

Taking data’s context-dependence seriously complicates the analysis of data’s value in digital markets and its role in competition policy. If data isn’t a mere “thing” that can be exchanged, shared, moved, ported, and anonymized, but is something whose value is context-dependent and partial, then how to develop a pro-competitive understanding of the phenomenon we call “data”? I haven’t seen this important question treated so far. Let’s leave it aside for a moment and turn to what antitrust agencies and regulators are currently saying and doing about data.

Antitrust Regulators’ Approaches to Data: Between Optimism and Misunderstanding

As I argue with my co-author Reuben Binns, for a long time antitrust authorities – in the US and Europe, at least – did not consider data to be relevant from a competition perspective. Information was thought to lead to efficiencies which did not warrant competition law scrutiny. The Federal Trade Commission’s decision not to investigate the Google/Doubleclick merger in 2007 is a salient example of antitrust agencies’ early reluctance to consider privacy and data a relevant competition law concern: “the sole purpose of federal antitrust review of mergers and acquisitions is to identify and remedy transactions that harm competition.” (emphasis added).

More recently, authorities have started to take interest in data as a relevant factor in antitrust law. In Europe, in particular, the advent of the GDPR has awakened competition law enforcers to the importance of data as an element of market power and a source of anticompetitive harm. In Apple/Shazam, the European Commission scrutinized Apple’s potential access to Shazam’s data from a competition perspective concluding that “access to Shazam’s data would not materially increase Apple’s ability to target music enthusiasts.” In Google/FitBit, the European Commission imposed a health data silo requirement on Google. In Germany, the Bundeskartellamt found that Facebook had abused its dominance by failing to give its users options to unbundle different types of data collection and use. Concretely, the agency decided that Facebook should have given users the option to disaggregate data related to their usage of Facebook from data related to their usage of Instagram or WhatsApp.

In the US, data has been less salient in competition cases. The recent lawsuits against Google and Facebook touch on data as a background concern. These cases’ primary focus, however, is on infrastructural power. For example, the case brought by Texas and others against Google in the advertising space points to Google’s efforts to leverage the discourse around data security and data privacy to enclose users and limit competition through an initiative called “Privacy Sandbox.”

These approaches illustrate two characterizations of data’s role in competition policy.

First, agencies are recognizing that the ability to control, collect and use data contributes to a firm’s market power. Data provides market actors with knowledge about their customers, users and competitors. This knowledge can lead to unfair and exclusionary market practices. For example, data enables self-preferencing behavior, as illustrated by allegations that Amazon engages in preferential treatment against its customers, favoring its own private label products or shipping methods over the products and shipping methods of its sellers and competitors. Data can also help firms gain asymmetric knowledge about an industry which can facilitate strategic decisions and acquisition activities, as illustrated for example by the UK prohibition of the Facebook/Gephy merger.

Second, agencies sometimes equate data to a non-price (quality) factor relevant to competition analysis. This possibility was mentioned in the early Google/DoubleClick case and led to a number of speculations about treating privacy as a quality dimension of digital products. European scholars have argued that excessive data collection should be treated as an “excessive pricing” offense under EU competition law. It might seem uncontroversial to say that surveillance erodes a digital service’s quality. Yet in practice surveillance and data cannot be reduced to price equivalents. Often, surveillance-based services are priced higher than equivalent non-digital products. A “smart” phone will tend to cost more than a “dumb” phone. A smart fitness tracker like Fitbit Sense 2 costs more than the less connected and surveillance dependent Fitbit Inspire 2. Privacy or good data practices are not something we can expect the market to produce, they also aren’t something that consumers can easily opt in or out of. Theorizing data’s competitive value in terms of a quality dimension equivalent to price is misleading. It misses the importance of structuring digital platform markets and competition in ways that facilitate, instead of obscuring, the contextual assessment of data and privacy concerns.

The focus of antitrust authorities in developing these theories has been to avoid two issues. One concern has been to disaggregate data rents derived from control over infrastructure from data acquired as a result of research, innovation, skill and labor. The second concern has been to avoid incumbents from expanding vertically into other market segments through exclusionary and anticompetitive data advantages.

The remedies imposed in these cases have included the unbundling of consent (e.g. the German Facebook case), more choices and opt-ins and opt-outs for consumers, data siloes and structural separation of databases (e.g. FitBit), data pools, data commons and data sharing/interoperability measures, as well as possibilities for competitors to access valuable data in an anonymized format.

Agencies and regulators have therefore started to weave data considerations into their analysis, but the normative rationales and effects of their interventions remain piecemeal and sometimes puzzling. Turning back to the question of what “data” represents, are these efforts by antitrust agencies and competition regulators sufficient? What approach ought competition policy aspire to with regard to data?

Clearing Confusions About Data in Antitrust

If data is a material set of 1s and 0s and acquires value as part of contextual interactions between persons and technological infrastructures, the definition of data-related harms and the articulation of remedies in this space requires revisiting.

The following are three suggestions for moving forward:

Data power is infrastructural power: data is an element of market power, yet market power is often best framed in terms of infrastructural capability. What matters is the ability to produce and capture data not how much data a firm controls at a given point in time. The ability to “datafy,” that is to transform human or company behavior into metrics and classifications capable of being analyzed and monetized, is a key dimension of market power, and such ability is primarily connected to control over infrastructure.

Clearing confusions about data’s value: data has been compared to many things: oil, dirt, capital, as asset. Some of these metaphors are plainly wrong. Some of them clash with each other. Understood as an asset, data increases a business’ value. But if seen through the lens of harmful surveillance, data decreases or erodes a service’s quality and consequently diminishes the overall business’ value. Any theory that takes data as an asset or a liability in the abstract is misleading. One cannot take the value of data as an objective measure that can be quantified and exists “out there.” Data’s value is relational and contextual. Prevalent objectivizing views produce bad solutions to competition policy questions.

Context-sensitive remedies: agencies tend to frame solutions to the problem of excessive or disproportionate control over data in two ways, and each should be revisited.

First, agencies and courts are inclined to push for solutions that give users and competitors more choices, the ability to opt-in or out of datasets, and that create an ocean of options for consumers to decide between. This individual-centric approach is unsatisfactory because individuals are not best placed to make decisions about data processes in an ecosystem largely opaque to them and controlled by platform gatekeepers. I would like to see a richer reflection around forms and levels of decision-making that do not assume the centrality of atomistic individual decision-making. Collective and agency-based governance can be combined with individual networked decision-making as part of a richer vision of data remedies in antitrust policy.

Second, agencies and courts think of data as an a-contextual asset that can be moved and shared between entities and across contexts. Taking data’s contextuality seriously may require inquiring about what happens when part of a database is separated from the other part. Does the data lose its economic value or privacy-salience? How does moving and sharing data transform its economic, social and legal value? The implementation and feasibility of data siloeing and data sharing remedies is likely to require a lot more sociological and evaluative scrutiny than antitrust experts have been willing to acknowledge.

Articles represent the opinions of their writers, not necessarily those of the University of Chicago, the Booth School of Business, or its faculty.