Taking notice and looking ahead: a new draft market definition notice for the EU
On 22 November 2022, the European Commission launched a public consultation on its draft revised market definition notice. Oxera’s response to the consultation praised the Commission for the improvements on the current market definition notice, which dates from 1997. However, based on our direct experience with the economics behind market definition, we see several areas for further clarification and improvement.
Much has changed within competition policy practice since the existing European Commission market definition notice was drawn up.1 For example, digital markets have become more prominent—in 1997, Google, Facebook and Twitter did not even exist. Our understanding of multi-sided markets has also developed, more attention is now given to innovation and other non-price elements of competition, and practitioners have had more experience with quantitative techniques and other sources of evidence for defining markets.
To ensure that its market definition notice is aligned with current knowledge, best practice, and case law, the Commission is currently revising its 1997 notice.2 In April 2020, it launched an evaluation of its current notice, with the results set out in a staff working document in July 2021. Following a call for evidence in January 2022, it published its draft revised market definition notice on 22 November 2022. It also launched a public consultation on this draft that ran until 13 January 2023.
Building on our direct experience with the economics behind market definition, Oxera has responded to the consultation . This article provides an update on the main changes proposed in the draft revised market definition notice, the areas where the changes should be welcomed, and our recommendations to the Commission on areas where the draft could be further improved.
What has changed?
The current draft revision includes many important changes and additional guidance on various key market definition issues, including:
- a clear explanation of the principles that underly market definition, including the focus on product and geographic markets, the fact that market definition does not prejudge the outcome of a competitive assessment, and the distinction between competitive constraints from within and from outside a relevant market;
- greater emphasis on non-price elements of competition, including quality and innovation;
- introducing the SSNIP3 test as the ‘theoretical criterion’4 to determine whether a hypothetical monopolist in a candidate market could exercise market power. The SSNIP test—also called the ‘hypothetical monopolist test’—is always the relevant conceptual framework when thinking about demand-side substitution;
- clarifications on quantitative techniques for implementing the SSNIP test, such as critical loss analysis5 and expanded guidance on possible sources of evidence;
- separate sections containing guidance on specific issues in market definition, including the presence of significant differentiation between the products or services in question (where market definition becomes less informative), the impact of price discrimination (which can lead to separate relevant markets for different customer groups, even if the product and geography are the same), significant investments in R&D,6 and market definition in the presence of multi-sided platforms, after-markets, bundles or digital ecosystems.
In praise of the new notice
In our response to the public consultation, we have praised the Commission for the current draft and the work undertaken. We consider the current draft to be a welcome improvement on the existing guidance. It captures current best practice, and is comprehensive yet concise and clear. It encompasses the significant developments of the past 20 years in EU competition law, and reflects the enhanced role of economic analysis.
We note that the many references to relevant case law are helpful. Importantly, in the notice the SSNIP test remains as the key thought framework in defining relevant markets. Reflecting the growth in online markets where products are provided for free to one set of users, the guidance extends the idea of the SSNIP test to non-price parameters and introduces the idea of a small but significant non-transitory decrease in quality (SSNDQ) test as the non-price equivalent of the SSNIP test.
Points for further clarification or improvement
Points of praise notwithstanding, we see several areas for further clarification or improvement in the revised notice. Based on our practical experience with the economics behind market definition in Europe and beyond, we provided the Commission with a number of recommendations.
Below, we elaborate on a selection of the most important points from Oxera’s consultation response.
The revised notice should emphasise that the SSNIP test is, in the first instance, a thought framework.
The SSNIP test is referred to in the draft notice as a ‘theoretical criterion’—i.e. it is not limited to a quantitative test. However, we suggest emphasising how the SSNIP (and SSNDQ) test is, in the first instance, a thought framework, or principle, that precedes potential quantification. Alternatively, or additionally, it may help to refer to the thought framework as the ‘hypothetical monopolist test’, to help distinguish it from critical loss analysis (CLA).
The revised notice should explain how CLA (as a tool for the implementation of the SSNIP test) is conducted.
In the draft notice, CLA is given as an example of quantitative implementation of the SSNIP test, but is explained only in general terms. We suggest adding clarification that the critical loss is calculated as SSNIP / (SSNIP + m), where ‘m’ is the gross margin (or contribution margin). This simple formula is omitted from the draft notice, perhaps unnecessarily.
Based on our experience in practice, we also consider that it would be valuable to provide at least some definition of gross margin (as the only other input parameter of CLA that is not yet defined in the notice) and some guidance on how a reliable gross margin may be derived or inferred from financial data. Alternatively, it might be useful to note that gross margins may, under profit maximisation and absent capacity constraints, be inferred theoretically from observed elasticities (through the Lerner index).7 However, these elasticities would then also have implications for, and should be consistent with, the analysis of actual loss. This caveat also applies when, conversely, actual loss and gross margins are used to infer elasticity. Finally, we suggest that CLA could be identified as the primary and accepted mathematical implementation of the SSNIP test, rather than just ‘an example’ of it.
The revised notice should clarify that CLA considers a SSNIP in the original candidate market only.
If the SSNIP test is not passed on the first iteration, it may make a difference for subsequent iterations whether the SSNIP test is applied only to the original candidate (or ‘focal’) market, or to a potentially enlarged candidate market. We suggest clarifying that, for each iteration, the SSNIP should, in principle, apply only to the original focal market. Otherwise, the concern is that this would result in a wider market definition based on a chain-of-substitution logic, even though the chain is not sufficiently strong. For example, if A is constrained only by B, and B is constrained by C, a SSNIP on an enlarged market of A+B would lead to the inclusion of C. However, this would then occur despite the absence of constraint from C on A.
The revised notice should recognise that the absence of actual substitution does not necessarily imply separate markets—the threat of substitution is what really matters.
The notice refers to evidence of past substitution as potential key evidence for market definition. However, it is well recognised in economics that the threat of diversion to a close second-best option may still impose a strong competitive constraint on the best option, even if actual substitution does not take place. This is shown in, for example, standard spatial models of oligopoly in industrial organisation (e.g. Hotelling models).8 We suggest that there could be acknowledgement of the point that competition can occur even without actual substitution. Ultimately, what matters is not actual substitution, but substitutability or the threat of potential substitution (of which actual substitution can be indicative).
The notice should clarify the merits of having access to evidence on hypothetical substitution, and refer to conjoint surveys as a potentially helpful tool to increase reliability.
The draft notice rightly points out that evidence on hypothetical substitution may constitute the only available direct evidence on substitutability or potential substitution, ‘for instance in cases calling for a forward-looking assessment’.9
Based on our experience, the simple absence of reliable real-world data on substitution or demand is an equally relevant instance where evidence on hypothetical substitution has strong merit. In fact, as noted above, the variable of interest is not actual substitution but potential substitution (or substitutability), of which actual substitution can be indicative. As such, evidence on hypothetical substitution is of general relevance. We suggest that the revised notice should clarify this point. We also suggest noting that evidence on hypothetical substitution is generally complementary to data on actual substitution: both aim to answer the same question, but from different methodological approaches, and each with its own relative merits. In fact, data on actual substitution is a form of evidence on hypothetical substitution.
Finally, we suggest that the revised notice should make reference to conjoint surveys as a potentially helpful tool to increase reliability over surveys that simply ask respondents for their direct valuation.
The revised notice should clarify that market definition is typically a less useful tool in the presence of significant product differentiation.
The draft notice explains that, when products are differentiated, market shares may be a less reliable indicator of market power. This is correct, but does not tell the whole story. The notice also explains that, in the presence of product differentiation, the Commission can identify separate relevant markets within a continuum of differentiated products, or it can define a single broad market and then take into account differences between market segments within that broad market. This flexibility in the approach serves to highlight that, in situations where products are significantly differentiated, it is more difficult to come to a clear-cut and objective conclusion on how markets should be delineated.
Looking ahead
The revised market definition notice has re-affirmed the important role that market definition plays in competition law. It provides an updated, in-depth overview of the concepts and principles of market definition, and allows policymakers to be more adept at dealing with the new challenges presented by, for example, multisided platforms in the digital space with zero monetary prices.
Once finalised in the second half of 2023, the revised market definition notice will provide greater transparency to external stakeholders. It will mean that the Commission’s implementation of EU competition law is more predictable, and thereby bring greater legal certainty for firms and their advisers across the EU and beyond.
1 European Commission (1997), ‘Commission Notice on the definition of relevant market for the purposes of Community competition law’, Official Journal of the European Communities.
2 European Commission (2022), ‘Competition: Commission seeks feedback on draft revised Market Definition Notice’, 8 November.
3 Small but significant non-transitory increase in price.
4 European Commission (2022), op. cit., para. 31.
5 Ibid., para. 59.
6 On this issue, the draft notice takes stock of recent high-profile merger cases such as Dow / DuPont and Bayer / Monsanto.
7 The Lerner index captures the link between the demand conditions facing a company and the extent to which it prices above marginal costs. The condition states that, at the profit-maximising price, the price–cost margin (Lerner index) is equal to -1 divided by the own price elasticity of demand faced by the firm. So, if the elasticity is -2 (meaning that if the firm raises its price by 10%, it will lose 20% of its demand), we would expect to observe margins of 50% (-1 divided by -2). Hence the less elastic the demand a company faces, the higher it will price above its marginal costs.
8 Hotelling, H. (1929), ‘Stability in Competition’, The Economic Journal, 39:153.
9 European Commission (2022), op. cit., para. 54.
