If EU policymakers are truly concerned about restoring competitiveness to digital markets, they need to adjust their expectations when it comes to data access as a silver bullet to promote competition.
At a roundtable in Brussels in May, NYU law Professor Harry First described the European Union’s proposed Digital Markets Act (DMA)—an ambitious regulatory proposal aimed at large platforms identified as “gatekeepers”—as “81 pages of dense, hard to understand language.” He half-jokingly added, “it is written in English … I know it’s written in English … but it takes some work to go through.” Professor First’s difficulties with the DMA text raise a simple question: beyond its stated ambition to improve “contestability” and “fairness,” what is the EU concretely trying to achieve in digital markets?
The answer lies in the fine print. According to the DMA, access to data is key to business success in the digital economy. The DMA’s prescriptions invite the idea that imposing data access, interoperability, and mobility obligations on “gatekeepers” can improve rivalry, promote entry, and ultimately innovation. And requiring gatekeepers to silo their data can restore a level playing field with existing and new firms.
So far, so good. There is, indeed, some truth to the idea that “data is the new oil.” But the directions taken in the DMA to implement this idea are based on implicit assumptions about the relationship between data and competition. Possible flaws in these assumptions could compromise the success of the DMA in delivering more “contestability” and “fairness.” They thus deserve to be made explicit and carefully reviewed.
The DMA’s Specific Rivalry Policy
The DMA embodies a specific approach to creating business rivalry through sectoral regulation. First, it wants to improve the competitive potential of “business users.” Perhaps, the DMA believes that data sharing obligations can promote competition between gatekeepers and their business. Business users include sellers on Amazon, advertisers on Google and Facebook, and influencers (or their agencies) in the case of “online social networking services” or “video sharing platforms”. Second, the weight of the DMA’s provisions leans towards a preference for disruption through substitutes. The DMA defines gatekeepers’ users in “core platform services” (CPS) as the bottlenecks that rivalrous efforts should contest.
The rivalry policy selected by the DMA begets the following question: Can business users who benefit from data sharing obligations under the DMA compete with the gatekeepers for a share of their CPS? Clearly, the DMA data obligations might lead to quality improvements in business users’ product offerings. But this is irrelevant from a “contestability” perspective. Instead, the relevant economic question is whether the DMA makes it relatively easy, and not too costly, for business users to reposition or extend their product lines in order to compete with the gatekeepers in their CPS.
Through these lenses, the DMA’s specific rivalry policy appears to entail risks and large sunk costs. Take, for example, online marketplaces: turning a merchant into the next Amazon would require an implausible, uneconomical, and inefficient duplication of fixed costs (e.g., in warehousing, transport, and server capabilities). The same is true of online advertisement. The DMA’s data transparency obligations improve advertisers’ information about the services they buy from Google and Facebook. But it’s hard to see how they can ever allow advertisers to put a dent in gatekeepers’ dominance in online advertisement services.
A more effective rivalry policy of disruption through substitutes would consist of promoting competition from other tech giants, not business users. In February, The Economist described “the growing overlap between America’s five tech behemoths” as the “the most salient feature of the new grammar of competition.” Established gatekeepers in other areas, and more generally other tech firms beyond the eye-catching FAANG (Facebook, Amazon, Apple, Netflix, and Google) have usually incurred large sunk costs and benefit from substantial economies of scope that reduce business risk. Unfortunately, the DMA works backwards on this front, imposing lines of business restrictions around gatekeepers by disincentivizing entry and/or expansion into other CPS. One of its key provisions, for example, forbids gatekeepers from combining data sources. Precaution should be taken to avoid that firms like IBM, SAP, Salesforce, Stripe, or Oracle be burdened with gatekeeper status, and in turn trapped behind soft lines of business restrictions. Most of them are active players in cloud computing, another CPS.
There is another problem with the DMA’s rivalry policy: Pressure from complements, not substitutes, is a substantial source of disruption for established firms. Combined with the firm-level disadvantages mentioned above, business users face strong prospects of value creation and competitive disruption if they follow indirect entry paths in new products and services. The DMA occasionally acknowledges this by talking of competition among gatekeepers’ “ancillary markets,” but most of the provisions in the DMA invite business users to contest established gatekeepers in CPS. There is probably more technological opportunity from leveraging data at the “edge” of networks than at the center of gatekeepers’ cloud platforms.
To put things clearly, we can expect higher contestability potential from business users building the next generation CPS that complement existing ones like voice assistants, virtual reality, self-driving cars, etc. This could mean, for example, that rather than competing with Google in online search engines, an entrant could disrupt online search by developing an entirely new way of indexing or finding information online. The DMA should embrace a more neutral approach toward entry paths and embody a clear and equal commitment to a rivalry-through-complements policy.
The Focus on Data Infrastructure, Not Human Capital
The emphasis of the DMA on data underestimates the critical role of non-technological inputs as the source of contestability in the digital economy. It is worth being very explicit about the nature of the problem. Most indicators show a long-term decline in the price of data infrastructure. In particular, the costs of cloud computing, for example for data storage or to train algorithms, have fallen. In a recent episode, Andreessen Horowitz venture capitalists and Dell Technologies founder Michael Dell concurred that they did not foresee further anticompetitive oligopoly consolidation in IT infrastructure, with substantial entry and competitive opportunities in distributed cloud, on-premise-computing, and edge intelligence.
By contrast, human capital is not cheap. The skills required to work with data are costly due to “a global shortage of engineers with the basic skills required to add IT‑ and software‑ driven functionality to new products and services.” Not every firm can afford science, technology, engineering, and mathematics (STEM) workers with relevant IT and software skills. Large US tech firms whose large R&D budgets are labor intensive tend to hire the lion’s share of data scientists, software engineers, and other experts. This is a problem because even in sectors “with a high degree of technological opportunity, recent advances in innovation have required a significant increase in human inputs.” A better distribution of “technical managers” (as Nobel Prize laureate Kenneth Arrow called them) across geographies and industry sectors might improve innovation, and the emergence of competitive firms outside of the US.
Why not, then, refocus the policy discussion on gatekeepers’ “acqui-hire” M&A transactions or non-compete clauses in employment contracts? Similarly, EU policymakers might further promote the immigration of high skilled workers and education in STEM fields into the EU if they want European companies to compete with the gatekeepers. Studies on innovation ecosystems suggest that geography and labor market mobility can boost innovation. In the digital economy, there is reason to believe that this property holds true, given (i) the substantial role of workers, managers, and technical personnel as repositories of tacit knowledge, and (ii) the highly combinatorial nature of innovation. Policymakers need to adjust their expectations from regulation such as the DMA to this reality: Strengthening the supply of human capital would be a welcome complementary policy to the DMA. Such policies could lift real constraints on the effectiveness of the DMA’s data obligations.
The Assumption of a Tradeoffs-Free Digital Economy
In any policy development exercise, tradeoffs must be considered. Yet, the DMA mostly adopts a tradeoff-free perspective. The idea underpinning the current approach can be summarized as follows: Improvements to competition deserve to be brought, and the market—seconded by other regulatory instruments—will supply all manners of good things after this. This approach can be seen in several key provisions of the DMA.
Consider first the tradeoff between contestability and privacy. The DMA prohibits gatekeepers from combining personal data sources from CPS with other personal data. Here, the provision does not appear based on privacy concerns—if this was the case, then non-gatekeeping firms would not be left free to combine personal data as they see fit either. Rather, the DMA’s ambition is to prevent anticompetitive leveraging by gatekeepers. More problematic, the various DMA’s data sharing requirements pay little heed to the privacy implications of increasing flows of personal information between gatekeepers and business users. This is a step backwards from the EU’s strong commitment to effective privacy protection in the digital economy, while economists are just grasping the full extent to which privacy considerations are important.
With this in mind, consider the specific case of online advertising. Here, the tradeoffs are even more complex. With more contestability in online advertising, prices for online ads should decrease, and advertisers’ demand should increase. This means more ads, not less. And this raises a hard policy question: are more ads a consumer welfare improvement? We believe that this is not obvious, yet competition authorities act like this is the case. For example, the UK Competition and Markets Authority’s (CMA) ongoing intervention against Facebook’s acquisition of Giphy is based on the preservation of competition to the primary benefit of advertisers, but it is less clear if and how UK consumers stand to benefit from this.
To be charitable to the DMA (and to the CMA), let us assume that contestability in online advertising leads to an equilibrium of platform services competing on quality to reduce ad targeting. Again, is this a consumer welfare improvement? Users’ preferences towards targeted ads are heterogeneous. Besides, if the DMA’s contestability scenario is that new, non-ad-based business models would emerge to compete with the gatekeepers, this implies that rivals rely on alternative business models, asking for example for a subscription or taking a cut of every transaction. Again, the jury is still out: is imposing on users a requirement to pay for a service a consumer welfare improvement?
Overall, it is very hard to see on the basis of which economic, social, or even ethical grounds the DMA would attempt to resolve the abovementioned tradeoffs, beyond a dogmatic belief in the merits of competition.
The DMA has not been adopted yet, so there is still time to improve it. With this piece, we have tried to show where the DMA might not address the true sources of market power in the digital economy effectively, and which targets policymakers should look at if they are truly concerned about restoring competitiveness. One takeaway is to lower expectations in relation to data access as a silver bullet to promote competition.
True, in trying to respond to Professor First’s puzzle, we might have given our own subjective reading of what the DMA tries to achieve. Alternative readings of the DMA exist. As we said elsewhere, the DMA shares common blood with the US consent decree with AT&T of 1956, which forced Bell Labs to openly license 8,600 patents and get out of all business unrelated to communications. The DMA’s spirit plausibly pursues a similar goal: make gatekeepers’ data, APIs, and other key assets widely available and trust that innovation will happen. Alternatively, the DMA might be rationalized as an attempt to allow alternative players to address unserved segments of market demand (e.g., assist search engine Ecosia in its attempt to serve search users interested in sustainability).
Whichever reading is right, the fact remains that much about the digital economy is still unknown. Statements of consensus in the research community are premature given the limited time economists have had for empirical observation. The claim that sharing gatekeepers’ proprietary assets is essential in order to promote contestability is contradicted on a daily basis by basic industry facts. The IT and software industry shows high numbers of firm entry and job creation. And while antitrust advocates and practitioners talk about gatekeeping positions, IT infrastructure experts and software engineers discuss Big Tech’s services in terms of “public clouds” and “edge computing.”
We close with a note of caution. A lesson from experience is to avoid picking winners and saving losers, a favorite pastime of governments enamored with industrial policy. How can we trust that political policymakers, who are knowingly bad at picking winners when they allocate subsidies, will do better when it comes to targeting digital gatekeepers with liability and remedies?
Learn more about our disclosure policy here.