Google and the DMA: Balancing deserved advantages and market contestability
This image was my first disastrous attempt at image generation using ChatGPT. Ironically, Google Gemini was way worse at doing this, or so it seemed.
On April 16th, the European Commission issued a preliminary decision that specifies how Google will be expected to comply with Article 6(11) of the European Digital Markets Act (The DMA). In a nutshell: The Commission reaffirmed that Google is supposed to share its “Search Data” with any competitors that operate online search engines (including competitors who operate AI chatbots with search engine functionalities) in the EU and the European Economic Area.
Why is the decision such a big deal?
If it were to become final, the decision would leave Google with no choice but to share with competitors the search data that it has amassed in the course of some 20 years of dominance in the search engine space. Having access to detailed data on what users search for, which search results they click on and how they interact with search result pages will allow Google’s competitors (including OpenAI) to use the data for the purpose of refining their own ranking algorithms and their ability to monetize them by, for example, better targeted advertising. This ability to use Google’s historical data for their own products would in theory remove Google’s so-called “Data Moat”, which is a barrier for entering and becoming a serious competitor in the search engine market.
What is the motivation of the European Commission?
The underlying rationale is the DMA’s market contestability pillar, which is a primary objective of the regulation, enshrined in Art. 1(1):
“The purpose of this Regulation is to contribute to the proper functioning of the internal market by laying down harmonised rules ensuring for all businesses, contestable and fair markets in the digital sector across the Union where gatekeepers are present, to the benefit of business users and end users."
The notion of “market contestability”, according to Recital 32 of the DMA, “(…) should relate to the ability of undertakings to effectively overcome barriers to entry and expansion and challenge the gatekeeper on the merits of their products and services”. Therefore, the same Recital goes, the obligations in the regulation “should also address situations where the position of the gatekeeper may be entrenched to such an extent that inter-platform competition is not effective in the short term, meaning that intra-platform competition needs to be created or increased”.
In simple terms, the EC’s rationale is that Google’s historic search data constitutes such an advantage that it makes it impossible for competitors to actually challenge Google for the dominance of the search engine space. Teresa Ribera, Executive VP for a Clean, Just and Competitive Transition at the Commission put it this way in the press release:
“Data is a key input for online search and for developing new services, including AI. Access to this data should not be restricted in ways that could harm competition. In fast-moving markets, small changes can quickly have a big impact. We will not allow practices that risk closing markets or limiting choice.”
Is this the right decision?
At this point, it is critical to start by accepting that the DMA market contestability ship has sailed. Fostering market contestability is the law of the European Union. The more interesting question that this discussion poses is about the limits of that “contestability” mandate and how they interact with other important values at stake.
In fact, the Chicago School has submitted a very Chicagoan form of “success is not a crime” argument to the discussion. Scholars like Richard Epstein make a rather obvious point: Google didn’t build this advantage by a succession of evil anti-competitive practices but by building the objectively best search engine product. He makes this point very poignantly in a critical piece for the Civitas Institute at the University of Texas:
“(…) since Google had started from nothing, none of its initial moves could have been driven by clever monopolization tactics, but instead by its superior technology, which no one thinks is an antitrust offense. When, therefore, did its monopoly maintenance program begin? And how did it supplant or displace its earlier technical prowess?”
As much as I hate to agree with the Chicago School about anything, I think Epstein’s argument carries a lot of weight because it is not just built on the efficiency-is-the-only-value narrative that is so common and trite among law and economic scholars. In fact, viewed from the right angle, Epstein’s position sounds a lot like a Dworkinian argument about fairness. No matter how impenetrable Google’s “data moat” looks at the moment, it is critical to remember that online searches used to be dominated by the Yahoos of this world. It is not much of an exaggeration to state that before Google’s hegemony, search engines were mere directories of websites created by humans who visited websites and manually categorised them into thematic hierarchies.
Now, to be historically accurate, it is critical to note that Google was not the first mover in trying to solve the scalability problem posed by the attempt to manually index the internet. Altavista & Lycos, the Googles before Google, built the first generation of web crawlers able to use powerful hardware to crawl big portions of the internet and create an index out of it. Google’s big break consisted in adopting this more scalable approach and refining it: It became the one search engine that we loved to use because it built the best technology to show us not just “the whole internet” but the parts of the internet that were most relevant. They called it Pagerank, and it was a relatively simple idea that became a (if not -the-) precursor for the algorithmic world we now live in: While Altavista relied on matching keywords to determine what showed first in our search result pages, Google looked at the web of links that pointed to a certain website and assigned a ranking to it based on the number of sites pointing to it and the relevance of those sites themselves. This is in many ways how scholar success is measured nowadays: How many citations a work receives and by whom.
And so, it is not overly Chicagoan to say that Google’s advantage in today’s search engine and adjacent markets is based on its own merit. That it is essentially a “deserved” advantage. And yes, the more economic arguments carry a lot of weight too: Forcing Google to grant access to its search data removes the incentive for competitors to innovate that advantage away by their own means. Those two are in fact some of the biggest bones to pick with this version of market contestability proposed in the EC’s preliminary decision.
The Disincentive to Innovation Argument
This argument basically holds that if the market leader is forced to hand over its most valuable asset (Google’s Search Data) to its competitors, it removes the incentive to innovate in the first place. On the one hand, if a firm knows its most valuable trade secrets will be handed to rivals once it reaches a certain size, there is very little incentive to invest billions in research and development, which would stifle innovation. From the competitor’s perspective, the Chicago School similarly argues that decisions like the ones taken by the European Commission create a scenario akin to the classic “Free-rider problem”, where Google’s competitors will simply stop trying to build an alternative superior technology because the law allows them to rely on Google’s data assets. In a more moderate form of the argument, one can see competitors shifting away from investing in alternative technologies to collect and “understand” data about users’ intent more efficiently if they can simply train their models with somebody else’s hard-won data. This may lead to “homogenisation” as opposed to vibrant competition, argues Epstein, a scenario where every search product starts to look and behave like Google. In some of his most recent critiques, he speculates about a possible paradoxical reality in which another firm uses Google’s search data to become the new Default incumbent, benefiting from the same “data moat” to earn an advantage that they obtained by regulatory decree as opposed to technical superiority. I am not entirely sure this is not some kind of caricaturized depiction of, well, vibrant competition, but Epstein’s innovation arguments are very swaying overall.
The fairness argument (Google competes on its own merits)
I hate to make it sound like the always odious “success is not a crime” narrative that some on the right use to caricaturize certain views of the left, but I am afraid it is impossible to avoid the resemblance. The European Commission is probably right in pointing out that Google’s advantage and scale have reached a point where the firm can easily “kill” or “absorb” any competitor that it deems as a threat. It happens very often in Big-tech, like the time when Meta (formerly known as Facebook) acquired Instagram. However, there is a general principle that most can agree upon: a firm should not be punished for its oversized success in a market insofar as said success is the result of building great products and was obtained without resorting to anticompetitive or other pernicious practices. As a corollary to that principle, one might suggest that forcing a firm to sell the assets that it has acquired as the result of its own research and development should require a very compelling rationale. Well, the most vulgarly neoliberal may argue that there is no way the European Commission can set a fair price for sharing the data, so the measure is unavoidably expropriatory or at least confiscatory in nature.
Google’s legal team has skilfully advanced a version of this argument about merits by pointing out that EU law is informed by a “competition on the merits” ethos. They start by making the point that Article 102 of the Treaty on the Functioning of the European Union (TFEU) does not prohibit or censor dominant positions per se but actually the abuse of those positions. They point to ECJ cases such as Post Danmark I (C-209/10) and Intel (C-413/14P) that affirm the fact that EU competition law is not meant to protect “less efficient” competitors from the consequences of competition. Moreover, the Oscar Bronner case (C-7/97) set a very high bar for when a company must share its infrastructure with rivals by limiting those scenarios to the infrastructure that is so “indispensable” that the refusal to share would “eliminate all competition”.
I am slightly more interested in the question of what is the right thing to do in a Dworkinian sense, and I cannot help but find little fairness on a decision that punishes the advantage that Google has earned by building great products without a substantiated allegation of wrongdoing or abuse of market dominance. While market contestability is the law, I tend to think that any exception to the general principle that firms should not be punished for their meritorious success can only be justified by the balancing of another more weighty value in tension. In other words, I tend to think that a decision like the one the European Commission has preliminarily issued must be justified by robust evidence of the restriction in market contestability created by Google’s “data moat” and the harm it creates. Extreme exceptions to good principles should meet a very high bar.
All in all, this discussion is fascinating and very consequential. I am looking forward to see how it evolves in the course of the public consultation that started on 16th of April. Anyone who is investing significantly in AI applications in the EU should be chipping in.
Disclaimer: The views expressed in this post are solely my own. They do not represent the views of my employer or any other third party.
