Competition agencies are pivoting towards “loss of potential competition” theories of harm, especially in acquisitions by highly dominant companies of smaller/nascent players offering a product or functionality in related spaces. The concern is broader than implied by the recent debate on “killer acquisitions.” Many say “What’s not to like?” We argue that it’s time to be more cautious. Large platforms have exceptional abilities to enter new spaces but also to “roll up” (willing) startups, buying instead of putting their own effort into rival innovation. Foregoing such effort is rarely good for society. We call these “buy vs. build” cases a form of “reverse” killer acquisitions that allow the incumbent to do away with its own innovation effort. Can we trade off the benefits of integration against this potential loss of organic expansion or should we adopt simplified policy rules to minimize the type of error we think matters most, requiring economists to pursue productive analyses of integration efficiencies, rather than the normal advocacy?

By Cristina Caffarra, Gregory S. Crawford & Tommaso Valletti1



Competition agencies – the UK CMA, but also the EC, the U.S. agencies, and more – are pivoting heavily towards “loss of potential competition” as a theory of harm, particularly in acquisitions by highly dominant companies of smaller/nascent players offering a product or functionality in a related but not immediately overlapping space. The concern is actually broader than implied by the recent debate on “killer acquisitions” (the suggestion that the incumbent may be buying a startup to pre-empt the threat this may pose if it evolves into a future replacement for its core business). We do care about these possibilities, even if infrequent, because the “killing” of a potential replacement can lead to large welfare losses. Yet there are more common variants that drive concerns about “loss of potential competition.” Established incumbents in spaces like tech, digital payments, internet, pharma and more have embarked in bids to acquire features, businesses, and functionalities to shortcut the time and effort they would otherwise require for organic expansion. We have traditionally looked at these cases benignly, but it is now right to be much more cautious.

Large digital platforms, in particular, have exceptional abilities to pursue organic expansion but also opportunities to “roll up” (willing) startups to “get there faster,” “buying” instead of expending effort in rival innovation. Foregoing such effort is never good for consumers and society as a whole: while innovative effort is costly, it will often yield multiple providers and differentiated services, with socially desirable properties. The pivot towards “potential competition theories of harm” is significantly about these “buy vs. build” cases and their potential to represent “reverse” killer acquisitions that allow the incumbent to do away with its own innovation effort, and reduce innovation overall relative to a “no deal” scenario. The benefits of integration need to be set against this potential loss of organic expansion, the associated innovation effort, and a world with multiple providers of differentiated services. Can we do this trade off on a case-by-case basis? Or do we need to set down simplified policy rules to minimize the type of error we think matters most? Is it not time to correct the huge lobbying effort directed at government departments around the agencies, and set economists off to do something more interesting and creative around truly exploring efficiency arguments (instead of pseudo-legal advocacy)?



Agencies are paying more attention than ever before to mergers where there is no loss of immediate competition between the parties, but the deal may eliminate the prospect of future competition. “Potential competition” theories are increasingly focused on the notion that, “but for the transaction,” parties currently engaged in complementary activities would have eventually come to expand and compete in each other’s domain; and the deal entails the loss of that future competition and the dampening of overall innovation relative to an independent counterfactual. This is coming up particularly in deals involving tech platforms buying up (or building stakes in) younger/innovative specialists in areas they are not currently in, but where they want to expand (in different ways, high-profile UK CMA Phase 2 cases like PayPal/iZettle, Sabre/Farelogix, and Amazon/Deliveroo all had some of this flavor – with the CMA being sufficiently concerned that the acquirer/investor was planning or had potential to develop a similar service to the target’s, that the deal would have snuffed).2

Much greater attention to potential competition is indeed overdue, and is especially apt in deals involving large tech platforms with enormous capabilities to expand their reach into multiple adjacent markets through the “roll up” of smaller/nascent firms. Indeed, as recently shown in Argentesi et. al (2019), this is “the way tech rolls”: frequent relatively small acquisitions that provide complementary functionalities or services to be integrated into the platform – sometimes cannibalized, sometimes bolted on, but essentially no longer separate efforts.3 The long-held posture in traditional antitrust (assuming we could look at these cases, most of which have flown under the radar) has tended to be “Mergers of complements? No issue. In fact, great! Integration is efficient. Potential entry is too speculative to worry about.” And indeed integration into a platform that can provide execution and funding to a nascent firm could, in principle, be a good thing (we are all familiar with the notion that startups are very willing sellers, and indeed the prospect of a future sale is often the driver for their innovation effort in the first place). But this is only one side of the story. Suppose the target indeed offered a platform a “way to market” in a service or functionality where it is lacking, but suppose there was also evidence that the platform would have otherwise strived to build its own offering in that space organically. A deal then eliminates a future competitor in that functionality – the acquirer itself.

Note that – critically – the perspective here is broader than the recent debate on whether there are “killer acquisitions” in tech that we failed to analyze and catch, i.e. acquisitions for the purpose of killing or taming a potential future threat to the acquirer’s core business. This phenomenon no doubt exists and can have large social costs (Facebook/WhatsApp, Facebook/Instagram, and Google/DoubleClick being oft-quoted “poster children” for it), but it’s only one permutation. There is a much more common possibility – “reverse” killer acquisitions? – where one is asking what innovation by the buyer is being foregone as a result of it buying a business it could have built organically instead. Looking just for the possibly elusive “future replacement” to a core business misses out on multiple cases where the buyer discontinues or foregoes its own effort because it has appropriated the “next best thing.”



What we should care about is the overall intensity of innovation effort in the economy and the impact this ultimately has on consumers. And by “innovation” we don’t literally just mean scientists in white coats. What we mean is a new process, a new and better version of a product, or a new way of making or delivering or designing a service. An alternative payment functionality. An alternative delivery format. Perhaps even an alternative ecosystem in the long run.

The prospect of reaping future rewards by acquiring a strong position in a new market is always the driver for innovation – as much for existing platforms seeking to grow into new and attractive spaces as for newcomers with a new idea. Tech platforms have extraordinary means to engage in productive organic innovation (and no doubt already do a significant amount of it). Yet their documented modus operandi is also to identify and buy up businesses at an incredible rate, (ostensibly) to fuel even quicker growth and expansion.

Superficially, what’s not to like? There is no immediate loss of competition, and a platform can provide execution capabilities, scaling opportunities and ideas and means for expansion in ways the target could perhaps not have dreamt of. That’s good. But the point is that economic research has shown that we want as much rival innovation effort as possible. A buyout of a promising nascent/small innovator deprives the world of that innovator’s contribution in an alternative scenario – an IPO, a sale to another buyer, or some other version of the future – in which it would have competed with an innovation developed and implemented by the buyer. This is what we are missing with “reverse” killer acquisitions: social welfare is enhanced when the would-be buyer also develops a service or a product, with head-to-head competition in the final product market.4 So it is right to look at the incentives of both target and acquirer in the counterfactual.

Now the standard counterargument: how do we know they would have both ended up producing some viable version of competing products? Is it not better to allow for the acquisition to take place, even if it just improves the opportunity for one of the two to come forth? Wouldn’t this avoid duplicative innovation?

The answer is no, and for at least three reasons. First, consumers care about levels of innovation effort per se. Because it is only if there is the prospect of being chased, caught, and overtaken that there is a strong incentive to come up with better solutions. Arrow first provided this insight almost sixty years ago,5 and it is a robust conclusion of the contemporary academic literature (notwithstanding recent misleading efforts by merger advisors to dilute this message). Among others, Kokkoris & Valletti (2020, Section III) review this literature and show that while it is possible to find special cases where a merger enhances innovation incentives (e.g. when there is a strong demand-expanding effect of an innovation that cannot be internalized, or when there are merger-specific efficiency gains),6 the general conclusion of the academic literature is that consumers and society are better off when innovative firms are not permitted to merge.7

Second, the size and overwhelming dominance of some tech platforms is already thought to have dampening effects on the “invest for buyout” incentive that can provide one pro-innovation justification for acquisitions, at least around the “core businesses” of these platforms There has been an open discussion for some time of “kill zones,” the reduced willingness of venture capitalists to provide funding for startups that replicate the main functionalities and/or could be direct replacements to those offered by dominant tech platform.8 This suggests that, where markets have tipped, innovation efforts by potential challengers is weakened and we may already be foregoing competition that is “not even born” by challengers we will never know. Merger policy thus needs to lean towards preserving more, rather than less innovation effort.

Third, and relatedly, the welfare effect of foregoing one of two innovation efforts may be sizeable if the two were to turn into real competitors. The “prize” is larger (and potentially huge) in cases where the target would have been a real substitute, allowing us to get away from monopoly/super dominance in the “primary” market. But even short of that more extreme version, the welfare benefits of competition are likely to be sizeable.

None of this is wildly unusual, by the way. These are the very principles the U.S. agencies are already meant to follow when evaluating horizontal mergers that may lessen innovation. Already the 2010 U.S. Horizontal Merger Guidelines established that “(t)he Agencies may consider whether a merger is likely to diminish innovation competition by encouraging the merged firm to curtail its innovative efforts below the level that would prevail in the absence of the merger. That curtailment of innovation could take the form of reduced incentive to continue with an existing product-development effort or reduced incentive to initiate development of new products.” And it is also consistent with recent EC practice, for instance in the agrochemical sector.9



A major strand of the debate around the expanding power of tech platforms has been the failure of merger policy to examine at all hundreds of consummated deals that went below the radar (of the total of nearly 500 for “GAFAM” deal alone since 2010, only a very small number has been actually reviewed – and this does not include stake-building in companies, nor does it include acquisitions by other many large platforms). As we do know ex post that there have been a few spectacular misses, the notion that we have not vetted hundreds of deals has driven a diffuse concern that we have missed cases where the deal “killed” the “next big thing,” i.e. a serious challenger that could have potentially emerged out of one or more of these targets.

This version of the debate has sought to transpose the logic and insights of the seminal article of the same name by Cunningham, Ederer, & Ma, which used careful data analysis in pharma to establish that a conservative estimate of the number of deals involving the discontinuation (“killing”) of competing innovation projects by an incumbent is approximately 6 percent.10 Now 6 percent may not seem like a big number, but it is when one considers the potential benefit this could have translated into as a result of competition with a (potentially monopolist) incumbent. The transposition to tech has been motivated by the idea that – while there is no similarly direct empirical evidence (antitrust markets are not so easily defined as in pharma, nor are outcomes of historical innovation projects) – the concern that acquisitions by an incumbent can be motivated by a design to kill a potential future competitor is there.

Of course, testing for “killer acquisitions,” even ex post, remains difficult. Even careful ex post studies such as the one done by Lear for the CMA do not provide a clear roadmap.11 There are some filters one can perhaps apply to spot ex ante deals where the target has the potential for being a replacement to the acquirer in the future or for depriving a rival of an opportunity to partner with a third party that could turn them into proper challengers. For example, in her keynote to the FTC Hearings on Competition and Consumer Protection,12 Susan Athey argued we have to look at whether the target has the potential to provide faster scale and growth opportunities to others, such that an incumbent would want to block it from being acquired. Examples of such settings include when the target is a specialized vertical that can offer a rival an entry path, an intermediary that can offer a rival a block of users and accelerate growth, or a software tool which can be integrated in to help a rival scale quickly and make it a more effective competitor. The message is to think about how the target could help a rival be a stronger player and then consider whether its integration or acquisition by a dominant incumbent would stop this threatening path.

Even so, ex ante and true “killer acquisitions” are unlikely to account for most of the deals that have been consummated in the past decade. Agencies will (and should) remain on the lookout for these, and a number of recommendations to agencies are to be found in the Background Note on Killer Acquisitions just issued by the OECD Secretariat for its imminent Meeting of the Competition Committee.13 But there’s more to the story.



What seems empirically more prevalent are cases where we are not necessarily worried about the buyer acquiring and then neutering a future threat (either directly, or by making them unavailable to others). What seems to be more frequent are cases where the acquisition may effectively extinguish the standalone effort of the buyer to expand in a particular space because the target immediately provides it with those capabilities. This covers a broader set of possibilities as platforms continue to expand into adjacent fields by buying functionalities, capabilities, even whole businesses (see the recent example of Google/Fitbit).14 To date, this has been regarded as only benign: no overlap, no obvious foreclosure of existing competitors, quick time to market, and bonanza for the target.

But wait. What is in fact often apparent (particularly when one looks at internal documents) is that these acquisitions are often evaluated internally in terms of “buy vs build.” Which is to say that there is often an alternative path to expanding into a particular space through the acquisition: with sprawling capabilities, competences, and limitless internal funding, buyers are often already on the way to building a functionality themselves. Internal documents often show the incumbent making (or thinking about making) an organic foray into this new market. The opportunity to buy instead then comes along. Once bought, the target may be cannibalized for certain assets to power the incumbent’s own effort. Or the incumbent’s own project quietly may be shelved. Either way, the buyer’s innovative effort in the target’s market has been extinguished. It’s the killing of one of the two efforts, but not the target’s – the buyer’s. Hence a “reverse” killer acquisition. There was some of this flavor indeed in the CMA’s concerns in PayPal/iZettle, Sabre/Farelogix and Amazon/Deliveroo – where concerns included whether the acquisition allowed the buyer to forego its own efforts in the target’s area, eliminating the prospect of future competition.



The insight that we want to preserve innovation efforts of both current and future competitors, even if duplicative, is a robust conclusion of the academic literature. Consumers always like more competition in product and service markets, including through entry by large platforms – think of how the power of media giants like Disney, Comcast, and AT&T has been contained by genuine competition in video streaming, with Netflix the early innovator but also home-grown efforts from Amazon and Apple.

The first question is then what would an agency need to show in order to conclude that the acquisition of a nascent/small player creates concerns along the lines discussed in this paper. We want to add two final reflections – one about the economics and one about resources – to the thoughtful points made in their Background Note by the OECD Secretariat.

On the economics of running a case, there is of course uncertainty here, as agencies will not be able to know with a high degree of likelihood whether both innovation efforts (the buyer’s and the target’s) will lead to competing products or services in the counterfactual. There is judgment involved, as buyers will always insist they were never ever going to make it in that space on a standalone basis, nor invest in that innovation effort themselves. It is not clear what one could test for with data either: internal documents will tend to be more informative (at least until they are lawyered to death), and possibly of use can be financial analyses of valuations and “build vs. buy” scenarios. Remedies are also challenging here: for instance, one might think of the parties being asked to “clone” the target’s technology and let a third party receive a white label version and develop it. (Of course, the merging parties would hate it – but there is ample precedent in pharma, and again not an extravagant solution). But in principle this is doable: every merger assessment involves constructing and analyzing future counterfactuals, and our point is that those counterfactuals should be rich and imaginative, accounting both for killer and reverse killer acquisitions.

Are the agencies going to go down that route? Yes, in theory, if they had limitless resources and were willing to embrace “expected consumer surplus” as part of their toolkit. Then, there is reality. Resources are scarce, and agencies should not be tying them up in lengthy legal battles around these issues. As the OECD Background Note states, “(…) perhaps the most important proposal that has emerged from the debate over the acquisition of nascent firms has been to reverse the burden of proof and create a rebuttable presumption” (para 143). This has been advocated inter alia by Valletti in his 2018 Keynote at the CRA December Conference; and by Peitz & Motta (2020),15 as well as multiple expert reports in the past year or so. Twelve leading U.S. academics and practitioners have also very recently advocated for a change in the U.S. law in this direction in their submission to Congress.16

Ultimately, the kind of policy we want needs to reflect our views – as a society – on what type of enforcement errors we think matter most (not the views of the parties and the howls of anguish of their advisors). We have proceeded for years on grounds that “Type 1 errors” (the risk of overenforcement) are the most pernicious as they would “chill innovation” stone dead, while “Type 2 errors” (the risk of underenforcement) will quickly be corrected by the growth of rivals or new entrants. But the recent track record in tech put that argument convincingly in the ground: hundreds of acquisitions not investigated, failures to diagnose potential “killer” acquisitions, multiple “reverse” cases where the buyer turns off its own effort, and fewer incentives to invest in challengers “under the shadow” of giants.

There is therefore a strong case for agencies to “lean in” and aggressively protect innovation in a huge and growing sector of the economy. Yes, enforcers cannot be all-knowing, especially given their limited resources and the huge asymmetry of information. This would militate in favor of super-dominant firms being required to proactively show why they are pursuing the deal, and how consumers would benefit. The reality is also that the lobbying playing field is far from level: lots of well-paid advocates for merging companies; few poorly-paid advocates for consumers. Perhaps therefore the system should build in structural hurdles that big money would have to get over. It would be great to see a world where the focus was instead shifted into making positive cases for efficiencies. Think of how quick firms would be to make their data available (not only to consultants, but also maybe to academics) if it were the only way they could get a pet merger through? Even if there is no change in rules, more aggressive enforcement is called for: it’s time to accept we can live with a few false positives after twenty years of false negatives, and to see more innovation competition instead of “rollups” and shopping sprees.

1 Caffarra: Charles River Associates (CRA), London, email:; Crawford: Dept. of Economics, University of Zurich, email:; Valletti: Tommaso Valletti, Dept. of Economics & Public Policy, Imperial College London, email: A version of this paper appeared on May 11, 2020 on Vox Blog, CEPR’s Portal. See

2 iZettle was a growing provider of in-store payment service solutions to small merchants, a capability Paypal was intent on developing; Farelogix is a supplier of technology solutions for airlines, including distribution and merchandising, which Sabre intended to integrate to improve its own offering Deliveroo is a restaurant food and convenience grocery delivery service in which Amazon has taken a minority share, and a question the CMA considered was whether Amazon’s own future (re)entry into food delivery in the UK would have been deterred by this stake. Paypal/iZettle and Amazon/Deliveroo were unconditionally approved after a Phase 2 investigation, while Sabre/Farelogix was blocked by the CMA and challenged by the DOJ, with a Delaware District Court rejecting the suit. For disclosure, CRA advised on these cases, but the views expressed in this note are the authors’ only. Crawford and Valletti were not involved in these cases. Outside of tech there are also high-profile recent and equally relevant examples in pharma, e.g. the Roche/Spark deal approved both by the FTC and the UK CMA, the Illumina/PacBio deal, undone after opposition from the FTC and the CMA, and the J&J/Takeda deal where concerns from the EU led to the deal being withdrawn from notification and abandoned. In all of these cases the issue was whether the buyer was engaged in competing research efforts and the acquisition would prevent one or the other innovation efforts from going forward.

3 Argentesi, E., P. Buccirossi, E. Calvano, T. Duso, A. Marazo, & S. Nava (2019), “Merger policy in digital markets: An ex-post assessment,” CEPR Discussion Paper 14166 and its summary on VoxEU at Broadening the scope slightly, further show hundreds of acquisitions by “GAFAM” (Google, Amazon, Facebook, Apple and Microsoft) in the areas of mobile, search, advertising, media, e-commerce, social, hardware and software. Just for the period after 2010, Wikipedia reports in the region of 180 deals in the public domain for Google/Alphabet, 75 for Facebook, over 60 for Amazon, 84 for Apple, 94 for Microsoft, nearly 500 in total – see, and equivalent entries for Facebook, Amazon, Apple, and Microsoft.

4 There is of course the separate and important issue that authorities would also have to ensure that a dominant platform does not disadvantage the rival in favour of its own homegrown product, but that is a topic separate from those considered here.

5 Arrow, K (1962), “Economic Welfare and the Allocation of Resources for Invention,” in The Rate and Direction of Inventive Activity, (R.R. Nelson, ed., Princeton University Press).

6 One example of a merger-specific efficiency would be the ability of the merged firm to organize R&D more efficiently internally (as long as this could not be achieved, say, by licensing). It is, however, something that deeply depends on inside industry knowledge that only the merging parties can have and should therefore be able to demonstrate in front of an agency in an efficiency defence.

7 Kokkoris, I. & T. Valletti (2020), “Innovation Considerations in Horizontal Merger Control,” Journal of Competition Law and Economics. This conclusion is further supported by a complementary literature using dynamic methods, including Gowrisankaran, G. (1999), “A Dynamic Model of Endogenous Horizontal Mergers,” RAND Journal of Economics, and Mermelstein, B., V. Nocke, M. Satterthwaite, & M. Whinston (2020), “Internal versus External Growth in Industries with Scale Economies: A Computational Model of Optimal Merger Policy,” Journal of Political Economy. See also Federico, G., Scott Morton, F. and Shapiro, C. (2019), “Antitrust and Innovation: Welcoming and Protecting Disruption,” at

8 See among others Kamepalli, S., R. Rajan, & L. Zingales (2019), “Kill Zone,” Working Paper, University of Chicago. On top of an empirical analysis documenting the “kill zone” phenomenon among VCs, the paper develops a theoretical model where the prospect of a merger with a dominant platform reduces the price start-ups can hope to extract in a deal, and thus diminishes their incentive to invest ex ante.

9 See Case COMP/M.7932 Dow/Dupont (Commission decision of March 27, 2017), Case COMP/M.7962 ChemChina/Syngenta, (Commission decision of April 5, 2017), and Case COMP/M.8084 Bayer/Monsanto (Commission decision of March 21, 2018).

10 Cunningham, C., Ederer, F., & S. Ma (2020), “Killer Acquisitions,” Working Paper LBS & Yale

11 See

12 Hearings on Competition and Consumer Protection in the 21st Century, FTC, October 15, 2018.

13 See OECD (2020), Start-ups, Killer Acquisitions and Merger Control – Background Note, DAF/COMP(2020)5, focusing on acquisitions for purposes of killing or controlling a nascent potential rival,

14 For a discussion of Google/Fitbit, see Google/Fitbit review: Privacy IS a competition issue, Cristina Caffarra & Tommaso Valletti Vox EU, March 4, 2020, at

15 Martin Peitz & Massimo Motta (2020), “Big Tech Mergers,” CEPR Discussion Paper 14353, summarized at

16 Baker, Farrell, Gavil, Gaynor, Kades, Katz, Kimmelman, Melamed, Rose, Salop, Scott Morton & Shapiro, 2020, Joint Response to the House Judiciary Committee on the State of Antitrust Law and Implications for Protecting Competition in Digital Markets