This article, available here, argues that a complete analysis of potential efficiencies from mergers should not only analyse how the merged entity’s prices, quantities and innovation incentives change (i.e., the direct effects of a merger), but also how these change for rival firms (indirect effects). While competition authorities sometimes analyse how mergers directly affect the merged firm’s innovation incentives, especially in high-tech industries, impacts on rivals’ innovation incentives have been rarely mentioned in merger guidelines or competition cases. This is unfortunate, since the effects of mergers on innovation in the relevant market depend on the reactions of non-merging competitors. While there is a growing literature on the effects of mergers on the innovation of the merging firms, evidence on the effects of mergers on outsiders’ innovation incentives is scarce.
Thus, this paper studies how horizontal mergers affect the innovation efforts of both the merged entity and its non-merging competitors. Using data on horizontal mergers among pharmaceutical firms in Europe, it finds that average patenting and R&D of merged entities and their rivals decline substantially in post-merger periods. This result is in line with the predictions of a model developed by the authors. Consistent with this model, negative effects of mergers on innovation are concentrated in markets with high R&D intensity, and where there are overlaps in the pre-merger innovation activities of merging and rival firms
Section 2 provides an overview of the literature.
While there is extensive literature on the relationship between competition and innovation, this literature does not specifically address how mergers change innovation incentives. Most research on this latter topic is much more recent, and has given rise to ambiguous results. The results of existing empirical studies on mergers and innovation in the merged entity have also been mixed, and seem to depend on product and technology market characteristics. Further, evidence on the effects of mergers on innovation activities of rivals is scarce.
Section 3 presents the authors’ theoretical model.
The model analyses a three-player oligopoly market in which firms compete on innovation and price. Two of these firms are more efficient, i.e. they are better at innovating, while a third firm faces higher innovation costs. It is assumed that, in the event of a merger, one of the efficient firms merges with the less efficient firm. Moreover, it is assumed that a merger leads to a (costless) technology transfer, i.e., the less efficient firm can use the efficient technology in case it is maintained as part of the merged entity. Finally, demand is assumed to be linear, i.e. it is assumed that firms’ innovations increase customers’ willingness to pay for their products.
The model compares profits and innovation levels (i) for the pre-merger oligopoly and (ii) for the post-merger market structure in which one of the firms with higher innovation levels has purchased the less innovative rival firm. The key results from the model are that a merger has (a) a negative effect on the merged entity’s innovation efforts in an industry with high research intensity and (b) a negative effect on non-merging competitors in an industry with high research intensity, provided the target firm conducts relatively little innovation compared to other firms before the merger. The model also predicts that negative effects of mergers on innovation are more likely to occur when pre-merger competition is intense, and less likely when an industry’s R&D intensity is low.
Sections 4 and 5 describe the data and empirical strategy, respectively.
The authors combined several data sources for their empirical analysis. First, they collected data on mergers between 1991 and 2007 from the website of the European Commission, with a focus on the pharmaceutical industry (as defined by NACE code). This data include market definitions by officials of the European Commission and the names of all competitors active in the relevant product markets. The authors also collected the names of all acquirers, targets and competitors from the reports, while deleting a few firms that mainly operate in other sectors like financial companies, hospitals and non-profit organizations. The resulting treatment group comprises 65 merger cases, which affected 381 firms.
Second, the authors matched firms from this sample with several other data sources, such as accounting data from the R&D scoreboard and the Amadeus database, and transaction activity from Bureau van Dijk’s Zephyr database. For a subsample of firms, the Entrepreneurial Studies Source, which extracts data on the firms’ main competitors from company accounts and industry reports, was used. Data on patent applications for the years 1978–2015 for all companies in the sample came from the PATSTAT database, developed by the European Patent Office (EPO) and the OECD. This is important, because the main innovation indicator used by the authors was the number of patents per year.
The assumptions made by the authors in their model – ruling out mergers between dominant players and concentrating on the effects of mergers between large and small firms – resemble the merger pattern observed in the empirical data. Acquirers are characterised by the highest innovation levels— as reflected in the number of patent applications, as well as by their cumulative patent stock. The same is true if we look at citation-weighted patents or R&D expenditures. Acquirers are, however, only slightly more innovative than their non-merging competitors, and these differences are not statistically significant. All innovation indicators suggest that target firms are less innovative than their acquirers and rivals.
The empirical strategy aims to identify the causal effect of mergers on merged entities and their rivals. For this purpose, the authors employ propensity score matching (to construct the counterfactual), and combine it with a DiD estimator in order to evaluate the impact of mergers. In particular, the authors are interested in a comparison between the actual post-merger outcome of firms in the relevant market and the situation had no merger taken place. Instead of comparing different groups, the authors focus on within-firm changes. This allows them to control for time-invariant unobservables through a DiD estimator, while time-constant and time-varying observables are controlled through the propensity score. To estimate the propensity scores, the authors use values of patenting lagged 1–3 years relative to merger events, as well as the pre-sample (1978–1989) average of the number of patents and a dummy indicating non-zero innovative activity during that time period. The authors also control for the number of citations per patent, the value of sales, the profit-to-sales ratio and the year of the merger. Propensity scores are used to construct control groups of firms with similar characteristics, including pre-merger levels and trends in patenting, through nearest neighbour matching.
Section 6 presents the results of the empirical analysis.
Estimates for patent counts confirm that within-firm (and within-market) variation in M&A activity is associated with a considerable decline in innovation. Innovation output by the merged entity and its competitors decreases on average by more than 30% and 7% compared to other firms, respectively. Profitability increases in post-merger periods for both acquirers and competitors (possibly due to a reduction in R&D spending and other investments). The correlation between M&As and sales is in line with the theoretical model: non-merging rivals increase their sales after a merger, while the merged entity decreases its operations compared to the combined pre-merger activities of acquirer and target. It takes some time until adjustments to firms’ research programs are fully realised. The merged entity’s patent output declines by almost 35% in year t + 3 and about 44% in year t + 4 . For merging firms’ rivals, there is almost no adjustment in innovation activity for the first two post-merger periods, but a decline in patenting of more than 15% and 25% in the 3rd and 4th post-merger year, respectively, can be observed.
When taking control groups into account, merger entities display very similar behaviour compared to control firms before the merger. After mergers, however, there is a drop in the merged entities’ average innovation outputs, while competitors of merging firms experience a small increase in patenting in the first two post-merger periods but a substantial decline in later years. This is in line with the results of the model, since pharmaceutical markets are arguably characterised by high research levels. Using controls where research intensity is low, the effects are less negative for the merged entity and even become positive for non-merging rivals. These results are in line with the hypotheses of positive effects for non-merging rivals of mergers on innovation in markets with low R&D intensity. Consistent with the interpretation that mergers reduce innovation due to a reduction in competition, there is little evidence ogf significant effects on non-merging competitors in technology fields without overlap of acquirer and target prior to the merger. This result indicates that the reduction in post-merger patenting is likely to be due to a reduction of competition in these fields. For the merged entity, there are significantly negative, albeit smaller, effects even in fields without overlap, possibly due to general downsizing or a reallocation of economic activities.
All in all, the empirical analysis indicates that mergers reduce innovation of both merging firms and non-merging competitors due to a reduction in competition in affected technology fields. However, the effects seem to be quite heterogeneous, as there is evidence for less negative—and in some cases even positive—effects in markets with low innovation intensity. These empirical results are consistent with several predictions of the theoretical model.
Section 7 concludes.
While merger policy in both the EU and the US increasingly discusses the effects of mergers on innovation, the focus has, until now, almost exclusively been on effects on merging parties’ innovation activities. This paper indicates that mergers can indeed not only have a negative impact on the merged firms’ innovation activities, but can also negatively affect rivals’ R&D incentives and, thereby, further reduce an industry’s innovativeness. This finding is especially relevant for research-intensive industries such as pharmaceutical markets. In contrast, in markets with low levels of pre-merger innovation activity, a merger can have positive effects on innovation outputs by the merged entity and its non-merging competitors.
As a consequence, the paper suggests that merger policy should pay closer attention to the effects that mergers can have on innovation incentives, not only of the merged entity, but also on rivals, particularly in high-tech industries. Focusing only on the merged entity’s innovation activities may well underestimate the negative effects that mergers can have on innovation.
To my (economically) untrained eye, this is an ambitious, rigorous and comprehensive paper that addresses a thorny topic with great care (one third of the paper are annexes, mainly on controls and robustness checks). I am, of course, unable to accept or challenge the results, but they seem plausible and well supported empirically. In any event, I think this is a good representative of a second wave of pro-enforcement papers. First, it takes on board the criticisms, made in papers which I will sample next week, to the first generation of pro-enforcement arguments. Second, it builds on the original arguments in favour of enforcement and develops new ones by zooming in on problematic sectors and looking for empirical evidence to support its case.
At the same time, the paper gives rise to a number of important questions from a policy standpoint – in particular, regarding its call for closer scrutiny of mergers in high-tech intensive industries. First, how are we to distinguish them from low intensity mergers? Further, is the authors’ proposal valid as a generic policy proposalv even for high-tech industries, given the model assumption of 3-to-2 mergers where an inneficient target is acquired – is this common across high-tech sectors? I also have questions about how the authors measure low intensity industries when their empirical work focuses on high innovation sectors, but I may well have missed something on this when reading the paper.
Second, and more fundamentally, I had some doubts about whether this paper adds as much to the merger/innovation debate as the authors seem to imply. On its face this might sound a startling question – it is clear that this is an important contribution to the academic debate. From a policy standpoint, however, there are good reasons to doubt that the pharma industry is able to provide a blueprint for other (technologically sophisticated) industries. Innovation in phama builds on scientific developments and is very well structured – to the point where pharma mergers have long been targetted on the basis of orthodox theories of harm concerning pipeline products. I think my doubt is particularly pertinent because the authors argue that the effects they identify only arise for horizontal mergers – which seems to require well delineated product markets, i.e. pipeline products. In other words, the paper seems to work well as an explanation of concerns with innovation – and merger control practice – in the pharmaceutical sector. How much of its findings extend to other sectors of the economy remains an open question.
A related question is whether patent counts provide a useful metric outside of pharma. Let’s consider big tech, which is the (sometimes unspoken) addressee of these debates, even though not the target of merger control intervention on this basis (yet). Many of digital developments may be protected by IP, but this does not often take the form of patents – instead, we are usually concerned about design choices, copyright and trade secrets. Would we be able to apply the authors model to such an environment?
In other words, this paper struck me as a valuable and insightful contribution to the academic literature, and to understanding the impact of horizontal mergers in the pharmaceutical sector – but I left it with the impression that more work is needed before its insights can be used to guide how innovation is taken into account in merger control more generally.