Introduction

On 29 February 2024, I took part in the sixth episode of the “Let’s Talk Competition” webcast conversation series, organised by the European Commission. Alongside Guillaume Loriot and Vicky Robertson, we discussed the modifications and improvements brought about by the European Commission’s (the “Commission”) new Market Definition Notice (the “Notice”). The conversation focused on the Notice’s reflection of new market realities and the Commission’s recent case practice. This article is a recap of the points I made during the/our discussion.

Overview

For starters, I can only commend those involved in the drafting of the Notice. I have first-hand experience in working on Commission Guidelines and I know the challenges of committee writing. Overall, the Notice is much more comprehensive and conceptually grounded than its predecessor, representing a clear improvement.

The clarification that market definition is a means to an end, as opposed to an end in itself, is a key point in the Notice. The purpose of market definition is to produce initial “quick and dirty” metrics that will shed some light on the likelihood of a certain theory of harm (i.e. the ability and incentives a firm may have to engage in a specific anticompetitive practice). The selection of a relevant market involves defining a market that will have significance in conducting this initial evaluation.

The pertinence of this initial assessment is case-dependent. It can be useful when assessing competition between firms selling homogeneous products to homogeneous customers. In this case, computed market shares can be a useful proxy for the likelihood of a significant impediment to effective competition (“SIEC”). However, when considering firms that compete with differentiated products and/or serve customers with asymmetric preferences, it becomes evident that market shares are not necessarily indicative of SIEC. Furthermore, the time and resources (as well as the complex techniques) employed in this initial assessment would be better spent on the competitive assessment, which is effectively the end goal and may even require the same data and conceptual framework as the market definition stage.

For instance, consider a merger where the merging parties have market shares of 20% and 25% in a high-end segment and 10% and 15% in the wider market. In a case like this, the determination of the relevant market level will often spark fierce disagreement: is it the high-end segment or the wider market? To reach a conclusion, it may be necessary to conduct customer surveys and econometric analysis, among others, which may require a significant allocation of resources. It is evident that neither of the two metrics, neither 20+25 nor 10+15, fully represents the probable effect of the merger, which lies somewhere in between. The crucial point is to determine where This depends on the closeness of competition and is really part of the competitive assessment: an answer to this would require considering customer choices and diversion ratios, estimating demand, etc.

The recently issued 2023 US Merger Guidelines place greater emphasis on the competitive assessment than on market definition, which I find appropriate. Market definition should not be burdensome and sophisticated, but rather “quick and dirty”. Overall, practitioners should strive for a simple procedure and avoid importing aspects of the competitive assessment into this initial assessment, which is a mistake. Simplicity should be its main quality.

Practitioners may argue that market definition is important because of the reliance of EU law on market share thresholds. However, safe harbours are also meant to be “quick and dirty”: if the undertaking is not clearly and unequivocally below the safe harbour levels, then it is most likely not in the safe harbour. In such cases, it seems absurd to opt for a complex econometric exercise to prove compliance with the thresholds, when the same machinery would prove much more useful in the competitive assessment. Falling outside a general or vertical block exemption regulation is not necessarily problematic. The European Court of Justice’s Super Bock judgment clarified recently that even hardcore restrictions are not necessarily object restrictions.^[1]

Digitalisation and Ecosystems

There are important clarifications in the Notice concerning multi-sided platforms. It is worth noting, however, that this is not a new issue in competition policy.

At some point, there was a debate on whether the market for “free TV” should also include “paid TV”. The market in question is two-sided, with advertisers on one side and viewers/subscribers on the other. Each side may have different alternatives and preferences, and it may be the case that advertisers deem free TV and paid TV as closer substitutes than viewers do. In such a case, the discussion between “market for free TV” and “market for all TV” is incomplete and imprecise. Digital multi-sided markets are subject to the same unproductive debates. Moreover, the presence of indirect network effects ensures that changes in substitution patterns of viewers on one side will inevitably affect patterns of the advertisers on the other side, which is yet another element to be considered in the definition of the relevant market in the digital multi-sided markets (which the Notice indeed mentions in the corresponding section).

How to deal with (digital) conglomerates offering related complementary products is another hot topic. In the case of Article 102 cases, I share the idea that a company might have market power in several connected markets, and that therefore the assessment of dominance over all markets together is the right course of action. The main example I can think of is Microsoft’s position in business applications. However, this assessment of dominance is not a market definition exercise. In fact, to the extent that the dominant company twists the nature of competition in its favour by forcing suite (or “system”) competition to avoid competing directly with best-of-breed solutions, one should not define a market for suite products. This would be akin to the cellophane fallacy.

Now, it is clear that term “ecosystem markets” has also a merger (and not just abuse-of-dominance) connotation. We might assume that the common ownership of complementary products will allow a merged entity to either bundle, tie, fuel consumer inertia and myopia, increase switching costs and transaction costs for alternatives, or reinforce aftermarket dynamics, and ultimately add barriers to entry. In such a case, it is necessary to consider market definition metrics that provide information about the coverage of the merging parties’ and their competitors’ offerings.

The relevant question here is the link between the metrics and the likelihood of SIEC. Structural presumptions do not work with differentiated products, as I explained before. There is even less of a direct link between “shares” in ecosystems made of complementary products and the likelihood of SIEC. For this reason, a holistic competitive assessment should be conducted before making a decision on complex cases involving complementary products in digital markets. It is important that the competition authorities avoid making decisions based solely on market definition or on whether there is a 40% share with a 5% overlap.

Innovation

The aspects concerning innovation included in the Notice are also a celebrated addition.

The debate on innovation started as a debate on remedies. It began with product market overlaps in innovative markets. The divestment of certain products had to be future-proof as a remedy, which meant that there was an obligation to impose divestment in R&D as well. This was the case, for example, in the HDD ^[2] or GE/Alstom ^[3]

When two merging firms have two well-identified investment projects, the merger will (i) reduce competition in the corresponding future product market in case both projects succeed and might (ii) alter their incentives to invest in these two competing projects (emphasis on “might” instead of “will”, which will be explained later).

The Commission has been confronted with cases in which innovation was cross-product, so that focusing on particular product lines might underestimate overlaps; and where innovation was not product-specific, but involved a particular product space, so that it was necessary to look for an overlap of products that could be affected by this innovation process.

Up until this point, I stand behind the Commission’s concerns. If a merger is likely to change how firms invest structurally, then markets have to be defined accordingly and you need to use a proxy that allows you to focus on overlaps at the right level (i.e. the level of products affected by that innovation).

The first difference between the Commission’s view and mine is that I believe that innovation is a means to an end, not an end in itself. The main impact is on future product markets, where the deterioration of conditions, e.g. price and/or quantity, hampers consumer welfare.

Second, unlike unilateral effects for substitute products, the impact of mergers on innovation is ambiguous. I am not referring to efficiencies but rather to the first-order structural effect of the merger. Even when the impact on innovation is negative, the effect can be relatively small, which raises a debate over the S in “SIEC”.

For these reasons, although it is useful to calculate high-level metrics to check for overlaps, structural presumptions are not applicable in this subject. My disagreement is not with the market definition, but with the possible use of these metrics in the competitive assessment. I am reticent towards inferring the existence of a SIEC based on market definition or, for example, a 40% share with a 5% increment in an innovation space.

Geographic Markets

In terms of geographic market definition, I believe we need to separate two lines of debate: the macroeconomic aspect (“Are the Chinese coming?”) and the microeconomic aspect (“How far am I willing to travel to buy bread?”).

From a macroeconomic perspective, it is important to note that certain markets, particularly those dealing with commodities, view imports as a constraint. In some cases, markets are connected (e.g. pricing on regional or worldwide platforms or indexes). Otherwise, imports are an imperfect substitute for domestic production. Therefore, the first-order issue here is how close a substitute imports are.

Typically, the Commission’s approach is to consider that closer substitutability will translate into higher levels of imports. This is sound. Hence, considering imports as part of the market leads to an informative, quick-and-dirty metric. Naturally, imports could be more or less of a constraint to the parties than to their competitors. This will depend on where they are located and on product characteristics, which means that a competitive assessment is still necessary.

This discussion pertains to commodities that are relatively uniform, such as coal or stainless steel, which are currently traded on the international market. However, you cannot deal with the “threat from Chinese trains” following this approach given that the Chinese trains are not there yet. Hence, you need a prospective assessment.

There is no quick and dirty metric here. Therefore, it is not a market definition question, but a matter of competitive assessment based on internal documents, (e.g., board meetings, saying that the main driver of product development is not the threat of existing competitors but of Chinese ones).

The microeconomic aspect is related to local markets, which are present in many cases at national and European levels. The cases with local markets are a rare example in which the heterogeneity of consumer tastes can be directly assessed. In these cases, it is clear that the computed metrics are typically uninformative for a SIEC.

The Notice clarifies that where customers are price-discriminated by locations, markets should not be defined around the supplier’s location, but should look directly at these customers. However, this is also the case in classical retail where customers are not discriminated by locations. In this case, geographical zones are highly heterogeneous: customers’ location can be closer to outlets, customers may have different options or the merger might affect them differently. Even though prices are the same for all consumers, they bear different transportation costs, so the effective price is different. Here, diversion ratios are not proportional to “market shares”.

Moreover, in practice, the metrics used for initial screening in local markets are not true market shares. Presence-based indexes cannot possibly be market shares. Indexes based on overlapping zones are better in practice and conceptually. However, they require very strong assumptions to consider them as market shares (i.e. consumers must be uniformly distributed within the catchment areas). Even assuming the assumptions hold, shares computed on this basis are not useful due to the asymmetric conditions of competition in the area.

Hence, mirroring what I argued in relation to Innovation, these cases should not be decided on the basis of market definition. They should focus on developing a competitive assessment. The starting point should be customers, so that we can directly assess how the merger affects the alternatives available to them.

It is true that competition authorities often lack information on the location of customers . In these cases, I uphold the use of small homogeneous zones, defined by statistical institutes as representative consumers. As such, one can assess the impact on all possible customers directly. There may be a large number of these zones, which may cause the assessment to be burdensome. This could be pointed out as a flaw in this approach. On the other hand, there are often plenty of stores to consider when adopting a supplier-centric approach. Hence, this alternative method should not be much more costly.

Conclusion

I welcome the improvements brought about by the Notice. It recognises important aspects of market definition that have long been discussed. In doing so, the Commission is providing more legal certainty for companies and their advisors.

Admittedly, there are certain aspects in which my opinion is at odds with the Commission’s case practice and the Notice, but the main idea is shared: the definition of relevant antitrust markets is a means to an end, not an end in itself.

Overall, I believe that the Commission is taking steps in the right direction, but still lacks some pragmatism in its procedures, as well as flexibility. As such, I argue that the Commission should focus on adopting a more effects-based and less structural approach. The Commission would benefit from focusing on the specific issues of each case, rather than spending time and resources on self-imposed rules that may not be pertinent to the overall assessment and could lead to a narrow-minded approach.