Competition authorities have started expressing concerns about horizontal mergers that give rise to a “significant impediment to effective innovation competition” (“SIEIC”) as a result of a reduction in post-merger R&D efforts (including lower R&D expenditure). At the heart of the development of SIEIC analysis lies a fundamental question of competition theory: under what conditions can variations of existing economic models be applied in merger cases?
This paper, available here, concludes that SIEIC constitutes a small but significant change in merger policy, which applies standard unilateral effects analysis but shifts the focus from price to innovation effects. This is not problematic: agencies should remain free to rely on new or adapted pre-existing economic models in merger control reviews. However, this conclusion is subject to the proviso that agencies must still discharge their “burden of persuasion”, and this may be hard to do based solely on theoretical models.
Section I describes and contextualises the role of innovation in European merger control.
It is useful to distinguish between different research stages when bringing a product to the market. The first stage focuses on pure research, the second concerns the “development stage” of a product, and the last stage covers R&D activities in relation to “early pipeline products”. Traditionally, merger control has focused on these latter stages, where product market overlaps, and actual or potential competition as regards identifiable products, can be observed. SIEIC, a theory developed by the European Commission in the Dow/DuPont merger, is based instead on the idea that firms compete in “innovation spaces” encompassing all stages of research, including the “discovery stage” where firms fund early lines of research.
SIEIC starts from the Commission’s guidelines acknowledgement that parameters other than price, and including innovation, can be relevant for merger control. However, SIEIC’s place in the EU’s merger policy is difficult to assess. On one hand, an original SIEIC “model” of innovation competition was formulated by Commission economists in parallel to the review of Dow/DuPont. On the other hand, Dow/DuPont is heavily evidence-based, and the main difference from previous cases seems to be that the Commission’s assessment was primarily based on the parties’ internal documents.
Section I also discusses the change brought about by theories of “significant impediment to effective innovation competition”.
Historically, innovation competition has been assessed by reference to R&D activities with specific ties to well-defined current or future product markets. Instead, the SIEIC theory of harm dispenses with the delineation of current or future product markets and examines competition in “innovation spaces”, a concept not dissimilar to an innovation market that “correspond[s] to the discovery targets” over which firms compete. In Dow/DuPont, innovation spaces were framed by reference to overlapping “lines of research” and “early pipeline products”, with the result of expanding the scope of merger review to early stage R&D efforts, where products are several years away from reaching the market. In effect, a merger may lead to a SIEIC even though one “may not be able to identify precisely which early pipeline products or lines of research the parties would likely discontinue.”
Compare this with the US approach: US antitrust agencies occasionally focus on mergers injurious to competition on separate upstream “innovation markets” that are not directly linked to a downstream current or future product market. The development of an “innovation markets” framework in US merger control enabled regulators to look at the effect of a merger on the “R&D process itself”, and on firms’ ability and incentives to commit resources to R&D activities. However, and in contrast to the Commission’s approach in Dow/DuPont, the delineation of separate upstream R&D markets where firms compete through investments for future technological applications is a key component of the “innovation markets” framework. While quite rare, reviews of innovation competition tend to focus on instances in which specific R&D activities that could have a potentially significant impact on specific downstream product markets could be identified.
In other words, the adoption of SIEIC marks a change in the Commission’s practice. Before Dow/DuPont, innovation competition was systematically assessed by reference to current or future downstream product markets as opposed to upstream innovation spaces/markets. On the other hand, SIEIC follows established merger control practice by adopting standard unilateral effects analysis, even if the shift from a focus on pricing incentives to innovation incentives represents a small but significant change in the analytical framework.
Section II discusses the economics of significant impediment to effective innovation competition” theories.
SIEIC raises the question of whether standard unilateral effects models can be applied seamlessly to innovation competition in the same way they have been applied to price competition. The risk of post-merger exit from R&D through the discontinuation, redirection or deferment of R&D projects is a function of the “innovation diversion ratio”. If the innovation diversion ratio is sufficiently high, the merger puts a “cannibalisation tax on the fruits of R&D investments by firm A” which may lower innovation. Conceptually, innovation diversion is similar to sales diversion as a result of postmerger price competition in markets for differentiated consumer goods.
But do the findings of a standard unilateral effects model involving price effects analysis hold true when non-pricing decisions, and in particular R&D decisions, are considered? There are three factors that call this hypothesis into question.
- First, SIEIC effectively assumes that pricing and R&D decisions are fungible variables. However, the validity of this assumption is questionable. Unlike prices and output which can – to some extent – be adjusted in the short term, firms are unable to shut down R&D programmes without friction, since R&D capital largely comprises inflexible assets. R&D capital must first be amortised, divested or (if labour) fired or reallocated, which is not seamless.
- Second, a SIEIC may also be predicated on an R&D exit scenario whereby the merged entity has an incentive to “redirect” R&D, thus eliminating rivalry between the merging parties and diminishing competition in the innovation space. Admittedly, if competing R&D capabilities are redirected towards distinct R&D targets, there is a lessening of competition. But is this decrease in competition detrimental to welfare in terms of overall innovation output? The answer to this question lies in whether or not the welfare costs of reduced competition within one innovation space are outweighed by the welfare benefits brought about by the increased ability of the merged entity to deploy its R&D resources across a higher number of innovation spaces. There is therefore a trade-off between competition in, and coordination across innovation spaces. This is an empirical question.
- Third, SIEIC may amount to the “deferment” of R&D by the merged firm. The problem with this is that standard unilateral effects analysis pays no heed to the organisation of R&D within the merged entity. As a result, SIEIC does not contemplate the possibility that there may be post-merger intrafirm R&D competition in the same innovation space on account of the merged firms’ organisational structure. In practice, an assessment of innovation competition should examine the merging parties’ post-closing organisation plans in a way that pricing analysis need not.
Taken together, these factors can plausibly decrease the opportunity cost of innovation, as well as outweigh the internalised cost of innovation diversion. Moreover, the SIEIC model fails to take merger-specific innovation efficiencies into account.
Section III puts SIEIC within the context of the economic methodology of merger control.
Merger control has become rather model-dependent in recent years. Economic models enable antitrust regulators to draw non-arbitrary inferences as to a merger’s effects on competition. At the same time, the ability of antitrust regulators to select and apply economic models should be subject to certain safeguards in the public interest. Models enable one to tell a ‘story’ about the case, but their applicability depends on certain conditions occurring; and even when those conditions occur, there can be disputes about the accurateness of the model – i.e. whether the model’s predictions are consistent witn reality.
Reflecting this, EU Courts have required the Commission to satisfy the “burden of persuasion” in relation to “complex economic appraisals.” In Tetra Laval, the court held that the Commission must be able to demonstrate that: ‘[t]he evidence relied on is factually accurate, reliable and consistent but also whether that evidence contains all the information which must be taken into account in order to assess a complex situation and whether it is capable of substantiating the conclusions drawn from it’. If these conditions – which have been elaborated in subsequent case law, and explicitly applied to economic models in RyanAir – are fulfilled, the economic evidence satisfies the burden of persuasion.
While the Tetra Laval only requires the Commission to tell a convincing story, granting it a certain amount of freedom to select economic models and other analytical tools to assess a merger, there are limits to this freedom. EU courts have previously rejected the application of controversial economic theories (e.g. GE/Honeywell). And the more speculative the theory of harm, the stricter the evidentiary requirements under the Tetra Laval test. When the anticompetitive nature of a merger is not readily apparent, “there is a need for more convincing evidence as compared to more easily expected anticompetitive effects.”
To date, the SIEIC framework has not been fully applied by the Commission in a merger control review – even in Dow/DuPont, where the Commission’s case primarily built on qualitative evidence (i.e. internal documents) regarding the closeness of innovation competition between the parties. Nonetheless, it must be assessed whether the SIEIC frameworks meets the Tetra Laval standards of accuracy, reliability and completeness. As regards accuracy, the SIEC model should only be applied if it fully takes into account the drivers of innovation in the industry under consideration. This may vary from industry to industry, but there is significant evidence that R&D is driven by factors other than firm rivalry. As regards reliability, SIEIC remains a conjecture – a theory that equates a post-merger loss in rivalry with lower innovation competition – which is not consensual under mainstream economic analysis. As regards consistency, the SIEIC seems to overlook the potential benefits of the merged entity’s ability and incentive to coordinate R&D programmes across innovation spaces, and seems to confuse a predicted reduction in R&D inputs with merger control’s necessary focus on R&D outputs. Finally, completeness means that the Commission’s assessment must take into account information that is both supportive and unsupportive of a theory of a harm, including information provided by the parties and other relevant information – which is to be assessed on a case-by-case basis.
This paper provides a comprehensive overview of the treatment of innovation in merger control practice. While focusing on the EU, it is also one of the few papers to look at the US. The lack of attention to the US is a common feature of the recent (European) debate on the topic – which is a bit shocking to me, since the concept of ‘innovation space’ comes from the US.
Ultimately, this is a long, somewhat sprawling paper, which as a result covers a lot of ground. One of the main virtues of the article is that correctly identifies the main step taken thus far under ‘new’ theories of harm based on innovation, which is to move away from identifiable products and markets and to focus on ‘innovation spaces’, identified as (early) lines of research not connected to any specific product or imminent market entry – and, hence, dissociated from ‘potential competition’ as commonlly understood. Furthermore, its analysis of the need of economic theories of harm to innovation needing to meet legal standards is rather good – and sorely absent from most of the literature in the topic, as we shall see in coming weeks.
At the same time, I think the paper has some issues. It is is structured around the Dow/DuPont merger, and hence it is focused on crop protection; or, at least, uses this market as an example throughout. In doing so, it approaches theories of harm based on innovation as if they were accurately reflected in this merger control decision – despite the Commission decision adopting an orthodox approach based on threats to pipeline products, even if simultaneously making noises in favour of SIEIC. The reason I mention this is because, as should be clear from the discussion of the economic literature over the coming weeks, theories of harm focusing on innovation can have a much wider scope than those discussed in this paper – and, as a result, so should SIEIC theories.
As such, I was not particularly convinced by section II, except in one respect that is often overlooked in economic discussions: the demonstration that a merger can lead to harmful effects on competition is a matter of evidence, and, as such, challenges to SIEIC on the grounds that standard unilateral effects’ for innovation are different from those for prices need to be addressed up to the requisite evidentiary standard. At the same time, the focus on Dow/DuPont also inquinates the section on legal standards, since this section reads like a criticism of Dow/DuPont rather than a discussion of how legal standards should apply to SIEIC – and I think that is a shame, since the latter is much more important but almost drowned by the discussion of an individual case.