In this fourteenth article in our series on "Big Data & Issues & Opportunities" (see our previous article here), we will focus on the impact of big data on different aspects of EU competition law and seek to create more clarity on when and how the ownership or (mis)use of (big) data can give rise to competition law issues. Specific illustrations in the transport sector will be provided.
Big data has been a hot topic in competition law for several years now. It has been on the radar at national level for quite some time, and given significant attention by the European Commission more recently in the context of shaping competition policy in the era of digitisation. Also recently, the German Bundeskartellamt ("FCO") has ruled that Facebook was exploiting consumers by requiring them to agree to data collection and its use without fully informed consent (by requiring them to agree to data collection in order to have an account). This ground-breaking decision has triggered a wide ranging debate on the relationship between competition and data protection law. The FCO's theory is that Facebook's dominance is what allows it to impose on users' contractual terms that require them to allow Facebook to track them everywhere. Maurice Stucke, an antitrust expert and professor at the University of Tennessee, noted that the old myth that antitrust was concerned with price and if a product was free then consumers could not be harmed was being discredited. The FCO's decision straddles the line between competition policy and data protection (which lies outside of the FCO's competence) and is being appealed by Facebook.
As such, big data aggregation in the transport sector can give rise to a variety of competition law issues that suggest that certain aspects of competition law may not be fit for purpose. Abuse of dominance, merger control and anticompetitive behaviour have all seen challenges in the face of big data, AI and digitisation and will be addressed in this article as well as in the context of the transport sector.
Abuse of dominance
The challenge in the context of big data and abuse of dominance lies in measuring market power and subsequently the potential for abuse of dominance. The simple fact that a company has access to large amounts of data does not automatically provide it with a dominant market position. Important factors that need to be taken into account to determine the existence of dominance include:
- Do other competitors have access to the same data?
- Is there data which can substitute the data collected by the company?
- Does the company have the ability to analyse and monetise the collected data?
- Is the data held by the company raw data or fully analysed data?
The trend in current competition analyses seems to focus primarily on the amount of data, with limited attention being given to the aspects listed above. These aspects may lead to the conclusion that, in a given case, even access to a very large amount of data does not provide a company with market power.
The main criteria to determine whether access to certain data gives market power include:
- Quantity: Once a certain volume of data has been gathered, the collection of additional data will not necessarily lead to any significant additional findings or benefits for the collecting company (so-called diminishing returns theory). The level above which the returns decrease will obviously differ between companies and industry sectors;
- Quality: Not all collected data has the same value. Raw data which cannot be processed and thus cannot be immediately monetised has a lower value than data which is ready for use;
- Availability: As mentioned above certain data is readily available to multiple companies since consumers typically use their personal data in different manners for different purposes (multi-homing).
The joint study published by the French and German competition authorities suggests that future cases could be based on the logic that abuse of dominance can arise from a firm’s ability to derive market power from big data that a competitor is unable to match. Particularly, they propose two questions to be examined in such cases:
- Whether there is a scarcity of data and whether competitors are able to easily obtain/replicate this data.
- Whether the scope and scale of the relevant data matter for the assessment of market power.
The important, and constantly evolving role of big data in the digitised transport sector in general, and in transport companies in particular, also has an impact on the competition law issues faced by companies active in that sector. An example is companies selling their data to third parties who can then make use of it in an exploitative manner. For instance, a transport company which tracks, collects and aggregates users' location and specific routes, can sell this data to insurance companies which then justifiably raise their customers' car insurance premiums if they perceive them to regularly drive above the speed limit, take more dangerous routes or use their vehicle more frequently than the average user.
Transport companies that enjoy a dominant position on a specific market and who have in their possession large amounts of data on their customers, could very easily exploit such data with the view to cementing their dominant position in that market and to excluding rivals. However, to be classified as an abuse of dominance, there are legal tests to be met and account must always be taken of the specific legal and economic circumstances in each individual case and whether there is any objective justification for the particular conduct.
Mergers between a large undertaking and small emerging companies can have a huge effect on data-related markets resulting in an increase in concentration or differentiated access if the newcomer possesses data or large access to data in a different market. Here, again, there are suggestions that merger control rules fall short on the basis that they are often based on meeting certain financial thresholds and/or market share thresholds without which merger control rules may not be triggered, leaving the acquisition outside the scope of merger control altogether.
The essential question is whether the competition tools available to the Commission are sufficient to properly analyse the effects of a given merger or possession of data on future competition. In a merger control analysis, the Commission has to conduct a predictive assessment of the future market with and without the proposed merger. So, a merger may be cleared, only to prove anticompetitive later on down the line, based on the assessment under the current rules.
The first EU in-depth probe to consider the power of data came about with the European Commission's investigation of Apple's proposed acquisition of Shazam Entertainment (Apple/ Shazam case). Shazam is a popular app used to identify a song. The use of the app is often brief and many of its users are anonymous. The Commission was concerned that Apple, by combining its data with Shazam, might obtain an unassailable competitive advantage over rivals. It also had concerns that Apple could gain access to commercially sensitive data on the customers of rival streaming services. After a five-month probe, the Commission concluded that Shazam's app was not unique and that rival streaming services would still have the opportunity to access and use similar databases. The clear message from this case is what matters is the kind of data you are acquiring, how unique it is, whether it can be easily replicated and whether you can shut out rivals.
This area of competition law has seen particular challenges and forecasted issues in the transport sector and big data.
When it comes to big data and possible price fixing in an online environment, critical questions are now being asked as to whether price setting by algorithm amounts to an "agreement" or "concerted practice". If algorithms -which need big data- are purposefully programmed to exchange pricing information or other data between competitors or enforce collusion, this will clearly be seen as an agreement or concerted practice between human representatives of the colluding competitors. The more difficult question is to where to draw the line between actions that can be attributed to humans and those that may arise through machines using algorithms employing artificial intelligence technology such as deep learning.
As pricing algorithms become more widespread amongst firms across all industries, the question arises whether algorithms signal the end of cartels or whether they create new and more difficult-to-detect cartels. The main concern in this area is with cartels and price collusion between competitors which cannot be proven following the traditional definitions of collusion (despite the definition of ‘agreements’ itself being rather broadly construed under EU law).
The UK's Competition & Market Authority ("CMA") in a report to the OECD noted that alongside substantive legal challenges, certain features of algorithms may also make it more difficult as a practical matter to detect and investigate unlawful collusive, abusive or harmful conduct, or to distinguish such unlawful conduct from lawful independent commercial actions. These include the complexity of algorithms and the challenge of understanding their exact operation and effects can make it more difficult for consumers and enforcement agencies to detect algorithmic abuses and gather relevant evidence. In addition, such challenges of detection may be heightened by the ability of algorithms to rapidly evolve, whether through constant refinement by developers or because self-learning is built into them. Or indeed by the fact that – in a world where most businesses have instant access to pricing data and where market transparency is high – unlawful collusion and “mere” conduct parallelism may look very similar.
Illustration in the transport sector: Airlines have in their possession large amounts of data on their customers including whether or not a customer prefers to compare prices prior to booking a ticket, or whether he/she books their tickets through a travel agency or an app.
Upon this basis, it has been suggested that it is not inconceivable that airlines could take advantage of big data analytics and machine-learning mechanisms in order to engage in price setting through "parallel-pricing" or tacit collusion amongst them. Such behaviours could be found to be contrary to the competition rules as anti-competitive agreements or concerted practices between competitors.
Indeed, airlines are able to decide how to price their airfares upon the basis of different sets of data, such as the expected behaviour of the customer; the price competition; and objective operational factors (such as the aircraft capacity, remaining seats, etc.).
In light of the above, holding crucial information on customers' preferences can be key in setting the airfare price. The possibility of analysing and using this mass amount of information through computer algorithms and other machine-learning mechanisms could lead the airlines to de facto align on price (through the use of the algorithms, which would be in a position to automatically set the price at an optimal level for each type of customer), as the airlines would realise that they do not need to compete to attract customers who are already willing to pay the specific prices set by the algorithm, irrespective of the airline. Competition authorities could be faced with substantial evidentiary obstacles to prove a competition law infringement in the absence of neither human contacts nor human agreement between airlines but rather a tacit collusion between machines.
Assessing the market conduct of companies with access to large volumes of data raises complex issues under competition law. The difficulty of the exercise is compounded by the fact that the analysis also needs to take into account data privacy and consumer protection issues that are intimately linked to the questions under competition law.
Both the European Commission and various national competition authorities are continuing to invest significant time and effort into the competition law analysis of big data, and there is extensive and increasing legal literature on the topic. The recent public consultation on shaping competition policy in the age of digitisation has yielded some interesting insight on how to mould competition law to address these topical issues. However, many issues remain unexplored and new issues will arise as a result of on-going technological developments. An effective response to these developments will require close cooperation in particular between the European competition and data protection authorities and the use of thorough economic analysis to avoid an over-enforcement that could stifle innovation and the emergence of new services and business models.