Digital jigsaw pieces
VoxEU Column Competition Policy Industrial organisation

'Ecosystem’ theories of harm in digital mergers: New insights from network economics, part 1

The traditional antitrust analysis of acquisitions by large digital conglomerates relies on a handful of mechanisms to leverage market power from one narrow market into another. Yet large digital conglomerates often own fungible assets, and capabilities that can be deployed across markets. This first in a series of two columns argues that we need new approaches to articulate how an existing constellation of assets and capabilities may matter to the analysis of a deal.

Merger enforcement involving the acquisition by a large digital conglomerate of a complementary or unrelated asset, frequently a startup or a much smaller firm, has evolved into two competing views of the world.  On the one hand, the parties and their advisors (plus the ‘type 1 error’/’but dynamic competition!’ crowd) argue that these acquisitions are innocuous and in fact pro-competitive as they allow the target to benefit from the financial muscle and technical support of the acquiror, and/or ensure a product/service will be improved, offered at scale, or combined with other services, in ways that would not have been possible ‘standalone’.  On the other hand, multiple regulators are taking the stance that the permissive approach to merger control of the past few decades has been a major factor in the creation of sprawling digital conglomerates, which now straddle multiple markets and whose multiple advantages have erected impregnable fortresses (or secured ‘moats’) around them – so these deals should be challenged to pre-empt these companies from ‘gating’ the future. 

From an assessment perspective, the question enforcers are engaging with is whether the unique position of a handful of incumbents benefiting from network effects and economies of scale and scope, with an expanding constellation of activities across multiple markets, embarked on the acquisition of multiple targets, is adequately captured by the ‘traditional’ antitrust playbook.  Defining narrow markets and evaluating the prospect for harm in those markets in terms of ‘traditional’ categories and well-understood economic mechanisms is what has taken the agencies to approving multiple deals (from Google/DoubleClick to Facebook/Whatsapp and Facebook/Instagram, even to Google/Fitbit), and this is now recognised ex post as most often harmful. 1

Is the ‘bag of tools’ we currently use in antitrust – in particular, the recognised catalogue of ‘foreclosure/ exclusion’ mechanisms in the industrial organisation (IO) literature, which is the usual ‘go to’ playbook for non-horizontal deals – adequate, or too narrow? Multiproduct digital giants do not think ‘market by market’ but have broad vision and fungible assets and capabilities that can be deployed across markets: AI, machine-learning, cloud, data, content, software, user bases. And they often use terminology like ‘flywheel’ to suggest these assets create virtuous cycles – i.e. they accelerate adoption and growth.  This suggests market power may need to be understood not just in terms of a product position in a given market, but as a function of controlling multiple ‘levers’ (assets) at the same time. And then new acquisitions need to be assessed for how they fit in and contribute to this network of activities, assets, and capabilities.  Can those ‘amplify’ the effects of a deal in some way?  Are there workable theories of harm to be articulated with economic support?

In the US there is case law of the 1950s-60s prohibiting a few conglomerate deals on grounds the acquirer was already ‘big’ in multiple areas, 2 but these ‘conglomerate theories’ are now regarded as legally very challenging as they have not been pursued for so long (e.g. Kowarski and Fortes 2023). The UK Competition and Markets Authority (CMA) (see further below) has been recently experimenting with ‘ecosystem’ theories of harm to capture the idea that the collection of a conglomerate firm’s existing assets and capabilities matters when assessing the acquisition of an unrelated asset. How can existing disparate capabilities form part of the ‘theory of harm’ for a specific deal, when they are in place before the deal and not directly affected by it? The question is how does one articulate and formalise a possible theory in a way that is more than impressionistic, and what are the limiting principles?  What insights should we be drawing on?

Traditional IO is the theory of markets underpinning antitrust analysis, but it remains focused on a limited number of ‘leveraging’ mechanisms from a market to another (tying/bundling, raising rivals’ costs). The strategic management literature has discussed ‘ecosystems’ for a long time, but mainly through the lens of ‘value creation’ and ‘value capture’ to optimise a competitive position, and not so much with a view to detecting possible competitive harm.  We suggest in this first in a series of two columns that a possible contribution could come from network economics, which has started to look at the connection between firms’ assets and capabilities, and how their evolution through acquisitions can affect competition across markets. It has formal modelling tools that can be used to examine the effects of a combination of these networks. It makes the point that enforcers should not simply look at products and prices, as merger effects can arise from a combination of these ‘networks of capabilities’ rather than just ‘product relations’.  Implementation is hard, however, and we are very much at the starting blocks in terms of developing both theory and a practical methodology to evaluate and quantify capabilities and their combinations. As regulators want to move increasingly away from classic ‘foreclosure’ theories and pursue broader concerns around agglomerations of capabilities, first-mover advantages, and data-driven harms, this work needs to progress much further.

Beyond the ‘incentive/ability’ framework

The classic framing of ‘theories of harm’ in conglomerate deals today requires agencies to articulate an ‘incentive’ and ‘ability’ of the merged entity to foreclose competition. This in turn requires defining a specific market from which significant power can be leveraged into another (target) market, and articulating a clear ‘foreclosure/exclusion’ mechanism for this to happen. The mainstream IO literature has contributed multiple ‘mechanisms’ through which such leveraging can occur – for instance, by tying the product/service being acquired to the acquiror’s product, and making the sale of the former contingent on the latter, share can be shifted away from ‘single product’ rivals and power can be leveraged from one market into another. Bundling can have the same effect.  Or ‘raising rivals’ costs’: if the product being acquired is an input into a rivals’ service, does the deal create ‘incentives and ability’ to make the input available at worse conditions, making it harder for rivals to compete?

As IO has gained traction in merger analysis, the expectation today is that agencies need a reasonably tight story that articulates (and preferably models) how foreclosure would occur in a specific market based on a particular mechanism, and formulates this in terms of ‘incentive’ and ‘ability’. Analyses range from calibrated theory models at the more sophisticated end, to bargaining models, to ‘bread and butter’ static ‘vertical arithmetic’ exercises (i.e. ‘all else equal’ quantification of the costs and benefits of foreclosure). In dynamic environments, a simple cost/benefit quantification ‘all else equal’ does not work and the analysis becomes more hand-wavy. But again in the current practice, agencies are expected to put forward their case (and parties to defend against it) by articulating a specific exclusion ‘mechanism’ and showing there are ‘incentives’ to foreclose (meaning the strategy would be profitable) and an ‘ability’ to foreclose (meaning rivals would be actually harmed).  Sometimes this involves formal modelling, although in practice it is often difficult to overcome the powerful logic of the ‘one monopoly profit’ (i.e. the idea that foreclosure is not necessary for the owner of a key asset to extract all profit). 

The question now is whether this market-by-market, mechanism-focused approach should be revisited when considering acquisitions by multiproduct digital firms with presence in multiple spaces; and whether it is possible to articulate an economic approach which can make the analysis reasonably credible and tight. 

One important shift that needs to be made is to think about ‘assets and capabilities’ instead of just ‘products’. In antitrust, we traditionally focus on ‘products’ and proxy market power with some measure of market share. But market share in a specific product will not tell us anything about the position of a multiproduct conglomerate firm across multiple spaces. Is there not more to its competitive role (and market power) which arises from its collection of assets and capabilities (IP, R&D, engineers, software, data, etc). Are these diffuse, scarce, unique? Are they replicable? What advantage do they confer collectively? How can they be used to develop new products and enter new markets? Or to gain first-mover advantage? When acquiring a target, the acquiror is getting title to a set of assets and capabilities which will interact with its existing ones and can be combined and redeployed. It seems reductive to just focus of narrow ‘product markets’ and overlook the broader mesh of assets/capabilities and their role in expansion/diversification strategies. 

At the same time, agencies will face major pushback if they just formulate their theory of harm as ‘you have multiple assets and a strong position in multiple markets’. The challenge is to go further and establish a merger-specific effect, i.e. how controlling multiple relevant assets/capabilities, and augmenting that portfolio with an acquisition, will create additional market power by insulating the buyer from current and future competition – for example, by creating greater asymmetries, hoarding relevant assets, and increasing barriers to entry and expansion.  This requires a way to score and rank the relevant capabilities, understanding how they interact and where the target fits, and whether the combination of the target’s capabilities with the acquiror’s can facilitate entry into new markets that may be pre-emptive and competition-reducing in future (rather than innovative and market-expanding).  How could such concerns be made operational/actionable, and what are the limiting principles? We sketch below a few cases where this has been (or currently is) part of the live debate.

A few recent cases

The question of how to assess acquisitions by firms which are conglomerate constellations of assets and capabilities have come to the fore rapidly in the current climate of concern about failure to enforce effectively in the past.  Google/Fitbit in 2020 saw the first big public debate about this type of concern (Caffarra and Valletti 2020, Bria et al. 2020, Bourreau et atl. 2020, Caffarra et al. 2021). Except that at the time, the agencies were significantly behind in terms of posture and appetite to pursue an ‘unconventional’ case. The acquisition by Google of data-collecting wearable manufacturer Fitbit raised issue of standard exclusion/ foreclosure ( of other businesses from Fitbit’s data, of other wearable manufacturers from Google’s Bluetooth, etc.), but critically also around the very personal data being collected by Fitbit being combined with Google’s own data and capabilities and used to extract more rents from consumers in areas such as health insurance and employment.  In practice, the deal ended up being cleared in Europe with conventional access remedies to a traditional theory of harm (foreclosure), plus a ‘data siloing’ remedy ostensibly to ensure that Fitbit’s data would not be exploited by Google in digital advertising by mingling it with Google’s ‘data firehose’. This was a small concession to the outcry by civil society that the deal allowed Google to contribute Fitbit to its extractive advertising complex, but the European Commission argued it could not do more because it did not have the tools to pursue the case further (Regibeau 2021). Yet the concerns expressed in this case were an early manifestation of an ‘ecosystem theory of harm’.

The UK CMA blocked Meta/Giphy in 2022 on grounds that the inclusion of Giphy’s GIF business into Meta’s ‘ecosystem’ could have been used together with Meta’s other assets to disadvantage rivals (for example, by manipulating access to GIFs in all sorts of ways and reducing their monetisation opportunities while raising their costs). In practice, the formulation of the theory of harm was quite conventional (vertical foreclosure) likely in anticipation that a non-standard framing would have made a defence against an appeal harder (in the end, the case was remitted to the CMA by the CAT and the prohibition was confirmed.

An ‘ecosystem theory of harm’ has also come to the fore recently with Microsoft/Activision in the UK, 3 with the CMA experimenting with the idea that as a result of controlling a list of assets (cloud, Windows operating system, first and third-party games as well as ‘important’ content), Microsoft would be ‘uniquely advantaged’ at a critical time of pivoting towards new cloud gaming technologies. The final (prohibition) decision mentions an ‘ecosystem’ but then uses a conventional input foreclosure ‘incentive/ability’ framework to conclude that Microsoft would likely foreclose nascent game streaming competitors (by denying to them potentially access to an ‘important’ game). The case is under appeal.

Meta/Within 4  is another recent case challenged by the Federal Trade Commission (FTC) ultimately because of concerns around prospective early steps towards monopolisation of a new space (the metaverse), but where the complaint was formulated in terms of a narrower theory of ‘loss of potential competition’ in a narrow market for virtual reality (VR) fitness apps. 5 As Meta had been making multiple acquisitions of VR businesses to power its metaverse ambitions, Within as a dedicated VR fitness app was potentially an important user case that could help create an early advantage. One way of formulating the concern could have been that by acquiring Within, Meta (1) gained a ready-made capability in this space and could abandon its own stunted efforts at producing a VR fitness app of its own (a ‘reverse killer acquisition’, whereby the ecosystem owner buys an asset instead of using its resources to build it, thus abandoning its own innovation effort and reducing dynamic competition; Caffarra et al. 2020); and (2) set down a building block for its push into the metaverse, which it could further power up through its collection of other assets (especially the app store and the headset). Thus while the acquisition itself was small and the overlap with Meta’s own fitness app was weak (Meta’s Beat Sabre app was not directly competing with Within), Meta’s ‘ecosystem’ could power up Within into a meaningful starting point for a shift into the metaverse as a new form factor.  In practice the concern as formulated in the FTC’s complaint was ‘loss of potential competition in VR dedicated fitness apps’, to fit with a more traditional approach (Oldale et al. 2020), presumably for litigation expediency. The judge eventually rejected the complaint on the facts, while upholding ‘loss of potential competition’ as a viable theory of harm.

Will Amazon/iRobot (under review in multiple jurisdictions) be another testing ground? The same agencies appear concerned about Amazon’s acquisition, and it seems possible that they will use it to delve into Amazon’s smart home device strategy and ‘Sidewalk’ project (the investment in low bandwidth home/neighbourhood mesh networking and smart devices).

Overall, while in these cases (some) regulators have brought up the concept of an ‘ecosystem’, they have stopped short of articulating an economic ‘ecosystem theory’ and often used the concept just as ‘mood music’ before falling back to traditional analysis.  How could merger control be tooled up to explore how large digital firms with distinct assets and capabilities in multiple spaces can leverage them collectively when making a new acquisition, and create additional market power?  The issues that need to be addressed are:

  1. How to make this more concrete: how to assess, score and rank relevant capabilities?
  2. What is a viable theory of how one would assess their role in the context of an acquisition?
  3. How to compare the potential benefits from the creation of innovative products that the combination of assets/capabilities can generate, against the potential harm of creating first-mover advantage and bottlenecks? 

As we will discuss in the second column in this series, combining network economics with IO approaches can be a promising avenue.


Bria, F, C Caffarra, G Crawford, W Christl, T Duso, J Ryan and T Valletti (2020), “Europe must not rush Google-Fitbit deal”, Politico, 22 July.

Bourreau, M et al. (2020), “Google/Fitbit will monetize health data and harm consumers”,, 30 September 2020.

Caffarra, C and T Valletti (2020), “Google.Fitbit review: Privacy IS a competition issue”,, 4 May.

Caffarra, C, G Crawford and T Valletti (2020), “How tech rolls’: Potential competition and ‘reverse’ killer acquisitions”,, 11 May.

Caffarra, C, G Crawford and J Ryan (2021), “The antitrust orthodoxy is blind to real data harms”,, 22 April.

Kowarski, I and F Fortes (2023), “US FTC's Amgen-Horizon challenge could chill healthcare M&A, but has high hurdle with risky conglomerate theory”, MLex Comment, 26 May. 

Mekki, D (2023), “Principal Deputy Assistant Attorney General Doha Mekki of the Antitrust Division Delivers Remarks at Mercatus Center Second Annual Antitrust Forum: Policy in Transition”, Arlington, VA, 26 January.

Oldale, A, B Sayyed and A Sweeting (2020), “A review of cases involving the loss of potential and nascent competition at the FTC, with particular reference to vertigal mergers”, Competition Law and Policy Debate 6(2).

Regibeau, P (2021), “Why I agree with the Google-Fitbit decision”,, 13 March.


  1. See for instance the speech of US DOJ’s Principal Deputy AAG Doha Mekki at the Mercatus conference in January 2023, listing several deals that were incorrectly assessed by agencies or judges over the past few years (Mekki 2023).
  2. See for instance United States v. Grinnell Corp., 384 U.S. 563 (1966).
  3. Cristina Caffarra advised Microsoft on the Activision deal, though the discussion in this column is based only on publicly available information.
  4. Cristina Caffarra had a role advising the FTC in Meta/Within. Again, the discussion here only reflects information in the public domain.
  5. See

2,108 Reads