The Digital Markets Act (“DMA’) will impose far-reaching obligations on some large technology firms operating in the European Union. The DMA’s definitions and obligations are designed to bite without effects-based analysis. Thus, the hope is to bypass the length and risk of traditional antitrust proceedings. We argue that this procedural goal will likely be missed. By being intentionally vague the DMA will invite challenges, both on gatekeeper designation and obligations, that can only be resolved through economic analysis. We argue that enforcers need to use soft law and guidelines effectively to achieve the goal of minimizing procedural delay due to “battles” between opposing economic experts.
By Philip Hanspach & Magdalena Viktoria Kuyterink[1]
The European Union’s Digital Markets Act (“DMA”) is a big policy bet against the tech giants.[2] Its critics point at arbitrary thresholds to identify gatekeepers, vague and contradictory obligations, and the lack of an efficiency defense for regulated parties. The proponents of the DMA accept potential inefficiencies with a shrug and reply: “Do you want economists to drag you through endless econometrics and modeling exercises?” Avoiding costs in terms of time, money, and agency resources associated with economic analysis seems to be an important reason why the DMA was designed the way it was.
Enforcers in the current antitrust system are sometimes exhausted with the perseverance with which defendants, for example in Article-102-cases, sow doubt onto theories of harm, or conjure up efficiencies that justify practices under scrutiny.[3] However, the DMA will likely fail to cut out lengthy economic analyses. In this piece, we argue that the DMA’s obligations will inevitably revert to effects-based analysis. This does not have to be a problem. If early enforcement practice and guidelines show the way, it might be possible to get the enforcement right without opening the door to endless back-and-forth between opposing economists.
First, what is the DMA? It is an ambitious legislative proposal that defines certain “core platform services” (“CPS”) and quantitative thresholds for large platforms that are designated as “gatekeepers” for these services. These gatekeepers have to follow a list of obligations, for example not to use their business customers’ data to compete against those same business customers. The DMA explicitly provides for the possibility to add new obligations in response to new developments.[4]
Within the two overarching substantial goals of “fairness” and “contestability,” there is a procedural goal of simplifying and speeding up the process of limiting big tech conduct. While economists usually much prefer to discuss substance rather than procedure,[5] we want to focus on the latter in this article. It is not hard to see where the desire for simple rules comes from. Antitrust odysseys such as Google Shopping[6] or Google Android[7] bind the resources of enforcers for years, create legal uncertainty, and are of no help for competitors who might have been harmed by a conduct ten years ago but who have since disappeared from the market. The DMA tries to be a tool apart from antitrust by adopting some ideas and language from the field of regulation.[8]
At first, the approach seems straightforward. It centers around prohibitions of conduct, without opportunity for gatekeepers to argue that their behavior is efficient (in the sense of benefiting final consumers). In principle, this should leave no room for meddling economists and their expert reports. Surprisingly, there seems to be no big outcry from private sector economists and lawyers about missing out on revenue.
Instead, less self-interested commentators have called for a greater consideration of efficiencies. For example, the German Monopolies Commission, an independent expert committee, recommended exceptions for conducts that benefit consumers.[9] A more cautious case-by-case approach was also favored during the consultation phase, for example by the Netherlands.[10]
Finally, the Joint Research Center of the European Commission published a report[11] authored by several economists (including the European Commission’s former Chief Competition Economist Tommaso Valletti), in which the authors warn of “an unfavorable trade-off between speed and quality of judgement” and chilling effects on innovation due to legal uncertainty. They suggest that effects-based analysis should play a role in the form of a grey list of conducts. Gatekeepers would then be allowed to demonstrate the efficiency of conducts on this grey list.
That does not create sleepless nights for our hypothetical DMA defender: “Perhaps we accidentally ban some efficient conduct but you can’t make an omelet without breaking a few eggs.” Ironically, while there is little room for a gatekeeper to defend conducts on the grounds that they create efficiencies for consumers, DMA article 9 does allow for exceptions in the case of obligations that turn out to be too disruptive to gatekeepers themselves (or their business customers). This provision is supposed to be used only under very rare circumstances, but it raises the question why demonstrably efficient conducts could not be treated similarly.
There are additional reasons why a dismissal of effects-based analysis in the context of the DMA would be premature. Upon closer inspection, the DMA has several provisions that will ultimately hinge on economic analysis. When the DMA will inevitably be litigated, its wording will require a judge to rule on issues that are fundamentally effects-based, no matter how many times we hear that the DMA does not care about effects. There are (at least) two main problems: first, the shifting nature of the activities described as CPS will inevitably invite fights over market definition. Second, some obligations are premised on the suspected effects as we show further below.
First, the DMA is not clear on delineating the limits of the CPS it seeks to regulate, so quantitative market definition will be unavoidable. Take the example of “online search engines”, one such CPS. Where does Google’s search business end and where does the rest of its business start? Google Search is embedded in websites and search functionalities are included in ancillary services such as Google Maps. Meanwhile, big tech firms are disrupting traditional online search, for example via voice-based search on smart speakers. It is not hard to imagine disagreement between a gatekeeper and the enforcer over what business lines and technologies to include in this CPS.
Due to this ambiguity, a gatekeeper that faces what it perceives as “regulatory overreach” might find it attractive to battle out in court which parts of its business are regulated by the DMA. The overarching problem in regulation with regards to innovative sectors is that business models and technologies change all the time. In the future, what will still be considered “online search”? Judges will end up making judgement calls on what is included in the wording of each CPS. We know from competition law that an analysis-free search for the spirit of the law and the meaning of words can result in absurd or at least arbitrary definitions. Think of early enforcement of the Sherman Act in the U.S., where it was not even clear if potential antitrust violations in manufacturing could be prosecuted on the grounds that the Act spoke of “commerce.”[12]
The well-known alternative from the antitrust toolkit is traditional market definition. Of course, this is precisely what the DMA is not interested in, as market definition requires data, painstaking analysis, and time. So, a DMA enforcer might confidently reply: “I’ll recognize an online search engine when I see one.” In the end, though, it is not clear that the shortcut offered by brief, verbal descriptions of CPS will speed things up. The descriptions are too vague not to be contested and gatekeepers have an obvious interest in minimizing the extent of their business that is affected by regulation.
Admittedly, market definition on platform markets is somewhat uncharted terrain. It is understandable that the DMA’s creators do not want to hinge their enforcement on new market definition tools for these markets, such as proposed modifications of SSNIP tests that account for two-sided markets. As the likely targeted gatekeepers, such as Google or Meta, are singular in their scope and without obvious comparison in their markets, we also have to accept that the correct data for market definition might not be readily available.
A possible solution: early enforcement or soft law, such as enforcement guidelines, should acknowledge the principles surrounding market definition. The main principle is that the relevant market should include products that provide a competitive constraint on the candidate market.[13] This principle, courtesy of competition economists, is flexible enough to account for future technological developments.
If the enforcer does not want to resort to quantitative analysis, be it due to a lack of resources or satisfactory data, at least a qualitative argument can be brought forward based on this logical and accepted principle. By introducing this principle early in the DMA’s lifespan, the enforcer can set predictable rules before technology changes too much. Agreeing on a theoretical principle might not sound like a lot. However, we believe it is superior to not having a systematic way of keeping the DMA’s CPS aligned with reality.
Second, some obligations simply make no sense without an effects-based interpretation. Interoperability comes to mind: After all, we care less about whether an API gives us the impression that service A is interoperable with service B than about whether it has the intended effect of turning single-homing customers into multi-homing customers. (Whether this makes sense economically is a different question altogether. In many standard models of two-sided markets, single-homing customers are valuable and courted, while multi-homing customers tend to have a hard time).[14]
Therefore, any litigation against a decision based on these obligations will inevitably require an effects-based analysis. Disagreements over other obligations may hinge on effects as well. There is also the suspension of obligations that should take into account effects (article 9.3). Finally, how else, if not by quantitative analysis, will the necessary evaluation of the DMA’s success a few years down the line assess “the chilling effects unfair conduct has on sales”?[15]
Consider a simplified and hypothetical example of self-preferencing in online advertisement: a business customer bidding for advertisement space accuses an online advertisement gatekeeper of self-preferencing. The claimant shows that it lost its bid against the gatekeeper’s ad aggregator 5 percent of the time when it made the higher bid. How would you start to investigate self-preferencing without quantitative analysis? You might ask the gatekeeper to explain its algorithm for determining the ranking of bids for ad aggregators. To conduct an independent analysis, you need to analyze outcomes, however. More than one variable might determine which bidder wins the ad slot: there might be good commercial reasons not to always let the highest bidder win but also to let the type of content and bidder play a role.
To be sure, what constitutes self-preferencing in this case would come down to case-specific details within an investigation. It makes sense not to discuss this at the level of the DMA text. And just because it is not written down in the DMA does not mean that we would not get an appropriate, effects-based analysis in this hypothetical self-preferencing case. Our point is a reminder that effects-based analysis will take place, so how can the DMA be enforced without the delays and costs of an antitrust investigation? We make three suggestions:
First, we suggest drawing a clear link between the obligation (no self-preferencing) and a verifiable violation of this obligation (for example, an unexplained difference in bidding outcomes). Depending on the obligation, this could be a general principle of outcomes that violate the obligation or a non-exhaustive list of examples. The latter approach could suffice if a general principle is too difficult to define and we are concerned about specific conducts. The appropriate place for this might be in guidelines on DMA enforcement.
Second, rather than ignoring the issue of economic evidence, enforcers would be well advised to anticipate it and to set standards of proof in their favor: make it easy for business customers to bring evidence of a violation of obligations and make gatekeepers explain in detail why their conducts are compliant. This avoids the conundrum of antitrust where the defendants too often get away with just sowing doubt onto a test or claim (as with the as-efficient competitor test in the Intel judgement), even if the substantial argument should still carry the day.
Third, the European Commission should use its discretion to engage in regulatory dialogue to specify gatekeeper-specific obligations. This regulatory dialogue can result in public guidelines on the principles that inform gatekeeper-specific obligations. Then, observers can generalize these principles to other and new gatekeepers in similar lines of business. Some critics have noted that the DMA obligations remain too close to cases that we have seen in the past.[16]
This ties into the concerns about the cost and uncertainty introduced by economic analysis. By being public and transparent about its principles, the Commission will guide discussions and frame disagreements, including the economics of DMA enforcement. This transparency will simplify discussions by making it clear which questions need to be answered to analyze conduct. In pursuing clearer and faster economic analyses, transparency will achieve more than secrecy or ambiguity which open the door to gatekeepers’ economists to come up with far-fetched arguments that might confuse judges.
For all of these reasons, we are skeptical whether the DMA can really speed up enforcement in the digital era. How can we avoid the dragged-out courtroom sagas of past antitrust cases without throwing notions of efficiency overboard?
Define the principles that will guide initial and future gatekeeper designations. They can be vague enough to allow for future technological development as long as they allow the European Commission to make a principled and defensible argument why some new and innovative services do or do not fall under the DMA. The German Federal Cartel Office’s application of article 19(a) of the German Competition Act, dealing with abusive conduct of undertakings of paramount significance for competition across markets, might be an example of how this can work in practice. Our examples illustrate that by demonstrating more rigor at the outset, the enforcer can reduce the risk of a quantitative stalemate later on. Either early enforcement or guidelines should make it clear what will happen during enforcement.
If the problems that we list above are not addressed, we believe that the first few years of the DMA will be painful and unsatisfactory for all sides involved, regulators and gatekeepers, lawyers, and economists alike. We might then ask; how did we end up here? We believe that lack of communication between lawyers and economists plays a role in this. Economists have sometimes done a bad job winning the trust of lawyers and helping the readers of their analyses to tell good from bad. This has sparked the desire among some people to sideline economic analysis altogether. Yet, as the DMA shows, even if you try, you can’t get around analyzing effects if you want to pursue specific economic goals.
[1] Both European University Institute, Florence, Italy. We thank Natalia Moreno Belloso, Linus Hoffmann, and Lola Montero Santos for helpful comments and suggestions. All errors remain our own.
[2] https://www.promarket.org/2021/11/02/the-european-unions-big-policy-bet-against-the-tech-giants/.
[3] https://www.promarket.org/2020/09/28/difference-between-research-academic-lobbying-hidden-funding/.
[4] In contrast, UK adopts an approach of “enforceable codes of conduct.” See Cusumano, Michael A., Annabelle Gawer & David B. Yoffie, “Can self-regulation save digital platforms?” Industrial and Corporate Change 30.5 (2021): 1259-1285. Some commentators see the UK approach as more tailored: https://voxeu.org/article/european-commission-digital-markets-act-translation.
[5] https://voxeu.org/article/digital-markets-act-economic-perspective-final-negotiations.
[6] https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3965639.
[7] https://www.politico.eu/article/google-eu-fine-court-hearing-android-antitrust-margrethe-vestager-european-commission/.
[8] https://www.gleisslutz.com/en/Digital_Markets_Act_und_Digital_Services_Act.html.
[9] https://www.monopolkommission.de/en/reports/special-reports/special-reporton-own-initiative/372-sr-82-dma.html.
[10] https://technologyquotient.freshfields.com/post/102gm5k/a-christmas-present-from-brussels-the-digital-markets-act.
[11] https://publications.jrc.ec.europa.eu/repository/handle/JRC122910.
[12] https://www.shrm.org/resourcesandtools/legal-and-compliance/employment-law/pages/sherman-anti-trust-act.aspx.
[13] Motta, Massimo. “Competition policy: theory and practice” Cambridge University Press (2004): 102.
[14] See, for example, Rochet, Jean-Charles & Jean Tirole. “Two-sided markets: a progress report.” The RAND Journal of Economics 37.3 (2006): 645-667. “[T]he more general insight [is] that the single-homing side receives a large share of the joint surplus, while the multi-homing one receives a small share.”
[15] This formulation was used to describe the “expected result(s) and impact” of the DMA in the original proposal of the law.
[16] https://cepr.org/voxeu/columns/european-commission-digital-markets-act-translation.
