Digital Mergers Motta Peitz IEP - v4
Digital Mergers Motta Peitz IEP - v4
Digital Mergers Motta Peitz IEP - v4
Abstract
Big tech mergers are frequently occurring events. What are the competitive effects
of these mergers? With the help of a simple model we identify the acquisition of
potential competitors as a pressing issue for merger control in digital industries. We
also sketch a few novel theories of harm of horizontal and conglomerate mergers
that are potentially relevant in digital industries. Finally, we draw some policy
recommendations on how to deal with mergers in such industries.
∗
We received helpful comments from Chiara Fumagalli. Massimo Motta gratefully acknowledges
financial aid from the Spanish Agencia Estatal de Investigación (AEI) and FEDER (project ECO2016-
76998-P). Martin Peitz gratefully acknowledges financial support from the Deutsche Forschungsgemein-
schaft through CRC TR 224 (project B05). Motta was Chief Competition Economist at the European
Commission between 2013 and 2016, and in that capacity he worked on several cases involving digital
platforms, including the Facebook/Whatsapp merger and the investigations on Google. Peitz was ad-
vising the Australian Competition and Consumer Commission (ACCC) in its Digital Platforms Inquiry.
†
ICREA-Universitat Pompeu Fabra and Barcelona Graduate School of Economics.
‡
University of Mannheim and MaCCI.
Big Tech Mergers 1
1 Introduction
Big tech mergers happen frequently. Taking a look at the acquisitions by the “Big Five”
over the last five years (Amazon, Apple, Facebook, Google, and Microsoft), Amazon is
reported to have made 42 acquisitions, Apple 33, Facebook 21, Google (Alphabet) 48 and
Microsoft 53.1 The vast majority of these and earlier mergers were not under the radar of
Antitrust Authorities (AAs)2 and the very few that have come under their scrutiny have
been approved, among them the prominent mergers of Google/Youtube, Google/Waze,
Google/Doubleclick, Facebook/Instagram, Facebook/Whatsapp, Microsoft/Linkedin.3
AAs and governments have become increasingly nervous at the perceived concentra-
tion in some digital markets and the persistent and increasing market power of some firms
operating in digital industries. This has led to a number of recent high-profile reports
on digital markets—e.g. ACCC (2019), Crémer et al. (2019), Furman et al. (2019), or
Scott Morton et al. (2019)—with a view to understand their functioning better and for-
mulate possible ways to promote competition.4 There is also concern that recent mergers
were investigated using an inadequate methodology possibly leading to wrong decisions;
for instance, in a report commissioned by the UK Competition and Markets Authority
(CMA), Argentesi et al. (2019) provide a critical ex-post evaluation of recent UK merger
cases in digital industries.
Policy proposals may include a mix of regulatory measures (e.g. obligations for certain
firms regarding data portability and interoperability, transparency, as well as not to
discriminate), stricter antitrust enforcement, and—relevant for the scope of the present
paper—possible changes in merger control. Indeed, some of the (many) mergers in digital
1
Sources: For Amazon: Wikipedia, https://en.wikipedia.org/wiki/List_of_mergers_and_
acquisitions_by_Amazon; for all other companies: www.mergr.com, last accessed December 24, 2019.
Other numbers regarding Google and Facebook are reported by Tim Wu and Stuart Thompson, The
Roots of Big Tech Run Disturbingly Deep, New York Times, June 7, 2019, https://www.nytimes.com/
interactive/2019/06/07/opinion/google-facebook-mergers-acquisitions-antitrust.html
2
In most jurisdictions, mergers should be notified and can be reviewed by antitrust authorities only if
certain turnover thresholds are met. Since in the digital industry firms often start to monetise only when
they have developed a very strong customer base, such thresholds may not be met even by companies
with very large market value.
3
As Furman et al. (2019, p. 11) put it, “[o]ver the last 10 years the 5 largest firms have made over
400 acquisitions globally. None has been blocked and very few have had conditions attached to approval,
in the UK or elsewhere, or even been scrutinised by competition authorities.” It is also noteworthy that
the large number of acquisitions is not matched by a large number of divestitures, we have seen very few
of the latter by big digital companies (one of the few instances is the spinoff of Paypal by Ebay).
4
We believe it is not controversial that the combination of economies of scale and scope, network
effects, switching costs and behavioural biases is contributing to the persistence of market power by big
tech, and make it more difficult for entrants to contest their market positions. For an analysis of barriers
to entry in the digital industry, see e.g. Franck and Peitz (2019) and the aforementioned reports.
Big Tech Mergers 2
industries may well have favoured the entrenchment of large firms’ market positions. This
applies not only to acquisition of horizontal nature, but also to acquisition of firms which
may appear as conglomerate or vertical. Indeed, challenges to an established incumbent
may also arise from often small, but quickly growing firms in adjacent markets. The
acquisition of potential competitors is therefore a pressing issue for merger control.
In Section 2, we develop a simple reduced-form framework to address the possible
anti- and pro-competitive effects of such acquisitions. In our setting, a start-up can
develop a project that succeeds with some probability. Whenever the start-up has the
ability to pursue its project, the merger will be anti-competitive. The acquisition then
becomes either a “killer acquisition” or an upgrade with suppressed competition. The
merger can only be pro-competitive if the start-up would not be able to pursue its project
absent the merger and if the incumbent will have an incentive to develop the project after
acquiring the start-up. We shall also see that the acquisition may also have beneficial ex
ante innovation effects: a merger may increase the expected benefit from innovation, and
hence stimulate effort to obtain it. In extensions we address conglomerate mergers, the
presence of outside investors, and exclusionary conduct by the incumbent.
In Section 3, we look at six novel theories of harm of big tech mergers which remove
actual competitors. These novel theories rely on some specific features of digital industries
(which include network effects; two-sidedness and free services to one side; the prominence
of big data). First, mergers may be consumer welfare decreasing even if consumers pay a
zero price. Second, when a merger enables the firm to combine the installed consumer base
through a merger, the privately most valuable merger may lead to a worse outcome from
a consumer welfare perspective than a privately less valuable merger. Third, a merger
of firms operating two-sided platforms can induce price effects that make it consumer
welfare decreasing. Fourth, in the context of big data, a conglomerate merger is anti-
competitive under some conditions. Fifth, a conglomerate merger that enables one-
stop shopping can be consumer welfare decreasing despite the efficiencies that arise from
one-stop shopping. Sixth, a profitable conglomerate merger that allows for bundling of
a “free” service may make a more-efficient competitor non-viable and, as a result, be
consumer welfare decreasing.
In Section 4, we conclude by making some policy recommendations based on our
analysis and some further considerations. This article draws on a number of recent
contributions and can be seen as a guide to these works. We give references to them
throughout the text.
Big Tech Mergers 3
so it would always have the ability to develop, but—as we shall see below—it may not
have the incentive to do so, if it bought E.
Payoffs If E remains independent, has enough resources, and successfully develops the
project, it will make duopoly profits πEd and the incumbent will make πId . If I takes over
E and successfully develops the project, it will be a monopolist with two products (or
services, or applications) and will earn πIM .7 If the project is shelved, or is not successfully
developed, the incumbent will remain a single-product monopolist and will earn πIm .
We assume that pπEd − K ≥ 0, or:
K
πEd ≥ , (A1)
p
which ensures that an entrant with sufficient resources will want to develop the project.
For the time being, we shall also assume that:
The LHS inequality of (A2) amounts to saying that industry profits are higher under
a monopoly than a duopoly, and the RHS inequality corresponds to the condition un-
der which the Arrow replacement effect holds, namely that—because of cannibalisation
effects—an entrant would have higher incentives to invest than a monopolist.8
• At time t = 2, the firm that owns the project decides whether to develop it or not.
• At time t = 3, active firms sell in the product market, and payoffs are realised.
We look for sub-game perfect Nash equilibria and, as usual, solve the game by back-
ward induction.
7
We do not impose a particular structure on the duopoly interaction. We implicitly assume that the
duopoly equilibrium exists and is unique. We also make assumptions on some properties of equilibrium
profits compared to monopoly outcomes below. An extension to an oligopoly with more than two firms
is straightforward.
8 d
The entrant’s incentive to invest is given by pπE ≥ K, the incumbent’s by pπIM +(1−p)πIm −K ≥ πIm ,
M m
which can be rewritten as p(πI − πI ) ≥ K. The former is larger than the latter when the RHS of (A2)
holds.
Big Tech Mergers 5
It is convenient to see first what happens if mergers were prohibited. This also coincides
with the continuation of the game which follows a rejection of the offer by the entrant.9
It is straightforward that:
• If RE < R, the entrant will not be able to develop the project. At equilibrium,
only the incumbent will be active: it will make a profit equal to πIm , and expected
consumer surplus will be CS m .
• If RE ≥ R, then the entrant will develop the project. Its expected profits and the
incumbent’s are, respectively, pπEd − K and pπId + (1 − p)πIm . Expected consumer
surplus is pCS d + (1 − p)CS m , which is larger than CS m .
We turn to the solution of the whole game. Let us move backwards, and look at the
sub-games starting from t = 2. First, suppose that E was acquired by I at t = 1. In this
case, I will own the project, and will develop it if and only if pπIM + (1 − p)πIm − K ≥ πIm ,
which can be rewritten as:
K
πIM − πIm ≥ . (1)
p
Note that πIM − πIm < πEd by A2. Therefore, if (1) holds, both the incumbent and a
sufficiently endowed entrant would develop the project, whereas if (1) is violated, only the
latter would develop it. This is the main force behind the so-called “killer acquisitions”
and it consists of the well-known Arrow replacement effect, whereby an incumbent has
less incentive to develop an innovation because it would cannibalise (part of) its current
profits.
Next, suppose that no merger took place at t = 1. In this case, the continuation
equilibrium is given by lemma 1.
Let us now move to the first stage of the game. If RE < R, then E’s expected profits
are VE = 0, which is also E’s reservation price for selling out. Whether it will develop
or not, the incumbent (which has the bargaining power in the takeover negotiation) will
9
As we shall show below, though, the takeover will always occur at equilibrium.
Big Tech Mergers 6
For the takeover to take place, it must be VI ≥ VE , or: p(πIM − πId ) − K ≥ pπEd − K,
which is always satisfied under assumption (A2).
K
(ii) If instead πIM − πIm < p
, the incumbent will not invest after the takeover, and its
willingness to bid for the takeover is given by:
For the takeover to take place, it must be VI0 ≥ VE , or: p(πIm − πId ) ≥ pπEd − K. Note
that under (ii) we can write πIm + K/p > πIM > πId + πEd , which is again satisfied: the
takeover will take place.
Note that the takeover will always occur at equilibrium because the incumbent has a
strong incentive to protect its monopoly profits (Gilbert and Newbery, 1982).
We can now build on the analysis carried out so far and look at what are the effects of
the acquisition relative to the (crude) benchmark where the incumbent cannot acquire
the potential competitor.
10
If it develops the project (i.e., if πIM −πIm ≥ K
p ), the incumbent’s gain from the acquisition is positive
M m m
(it will be given by pπI + (1 − p)πI − K − πI ). Otherwise it will be zero. If the acquisition entailed
some arbitrarily small but positive transaction cost, in the latter case the acquisition would not take
place, but for simplicity we abstract from such a cost.
Big Tech Mergers 7
Proposition 1. Relative to the benchmark where the takeover is prohibited, the acquisi-
tion of the potential competitor would:
K
• (i.a) (dead project) If πIM − πIm < p
and RE < R, the acquisition would be compet-
itively neutral: neither absent nor with the takeover, would the project be developed,
and in both cases consumer surplus would be CS m .
K
• (i.b) (killer acquisition) If πIM − πIm < p
and RE ≥ R, the acquisition would be
anticompetitive: absent the takeover, E would develop the project, and if success-
ful compete with the incumbent, whereas the takeover would lead to suppression of
competition and shelving of the project. The acquisition would result in consumer
surplus lowering from pCS d + (1 − p)CS m to CS m .
K
• (ii.a) (efficient upgrade) If πIM − πIm ≥ p
, and RE < R, the acquisition would be
efficient: absent the takeover, E would not invest in the project, whereas the takeover
will allow to get hold of a project which will be developed. Hence, the acquisition
would result in consumer surplus increasing from CS m to pCS M + (1 − p)CS m .
K
• (ii.b) (upgrade with suppressed competition) If πIM − πIm ≥ p
and RE ≥ R, the ac-
quisition would be anticompetitive: absent the takeover, E would develop the project,
and if successful compete with the incumbent, whereas under the takeover the in-
cumbent would invest in the project, but would suppress competition. The acquisi-
tion would hence result in consumer surplus lowering from pCS d + (1 − p)CS m to
pCS M + (1 − p)CS m .
RE < R RE ≥ R
(πIM − πIm < K/p) (i.a) dead project (i.b) killer acquisition
(πIM − πIm ≥ K/p) (ii.a) efficient upgrade (ii.b) upgrade, but suppressed competition
Big Tech Mergers 8
Within our simple model one can see that acquisitions of start-ups which may become po-
tential competitors a priori may be pro- or anti-competitive, so that simplistic policies—
say, continue with the laissez-faire approach that de facto antitrust agencies have adopted,
or at the opposite extreme the prohibition of all takeovers by big tech—would not be war-
ranted.
We will provide a more thorough policy discussion later on, but Proposition 1 already
gives us some useful hints. In particular, our analysis reveals that to understand the
likely effects of any given acquisition, it is crucial to identify the likely counterfactual to
that acquisition. Indeed, it is straightforward that whenever the start-up has the ability
to pursue its project, the merger will be anti-competitive. (In terms of Table 1, whenever
RE ≥ R, the acquisition will lower consumer surplus.) In other words, it will be inevitable
for AAs to ask what would happen if the merger was prohibited: Would the target firm
be able to invest and have the chances to develop the project? (We shall also come back
to this question in section 2.3 below, when considering an extension where the incumbent
may face outsiders bidding for the target.)
Unlike what has happened over the last decade, where AAs have developed simple
quantitative tools (such as GUPPI, IPR and the like), it looks like that a possibly more
complex qualitative analysis would be necessary, investigating — through the analysis
of internal documents, business plans, financial analyses — the likely scenarios for the
target should the merger not go ahead.
Another preliminary observation is that a merger may be pro-competitive (and hence
the AA and the Courts may make a mistake if prohibiting it) if and only if two conditions
are satisfied: first, the takeover target would not be able to pursue its project absent the
merger; second, the incumbent will have an incentive to pursue, rather than shelving, the
project. Like the first condition, also the second condition may not be easily checked, but
presumably the acquiring company may be able to convince the AAs that—should the
merger go ahead—it would really have an incentive to continue the project, for instance
because it would not lead to a close substitute but rather to a significant upgrade of its
core offering.
Before discussing policies more thoroughly, we extend the basic model in several
directions.
Big Tech Mergers 9
2.1.5 Extensions
In this subsection, we modify the base model with respect to profit ordering, the number
of insiders, the information about RE , and the availability of the start-up’s project.
Alternative profit ordering The assumption that duopoly profits are smaller than
monopoly profits (see LHS inequality of assumption A2) is innocuous in the sense that
should it not hold, quite simply the takeover would not take place.11
Less innocuous is instead the “Arrow Replacement effect” assumption (RHS inequality
of assumption A2): if it did not hold, then there may exist projects that an entrant
would not carry out, while an incumbent would—namely, the reverse of what we find
here. We are not aware of theoretical results which prove under what circumstances this
condition holds in general (or does not). In several standard oligopoly models, it holds
(for instance, one can show that it holds for general demand functions with homogenous
goods and cost-reducing innovation, and for simple differentiated goods models with linear
demand, e.g. Shubik-Levitan functions), but it is unclear to us whether it would hold
for general quality-enhancing innovations. As for the empirical point of view, the best
available evidence is provided by Cunningham et al. (2018), who study the development of
pharmaceutical projects12 and find that when incumbents acquire firms which have been
developing competing drugs, they are less likely to develop such drug projects (and that
the incentive to abandon such projects increases with the incumbent’s market power).
Such results may be interpreted as providing some support for the “Arrow replacement
effect” assumption.
empirical evidence.
The start-up’s ex ante incentives to innovate So far, the existence of the start-
up’s project was always exogenously given. However, it is conceivable that the project is
itself the outcome of a previous “ex ante” decision, which may well be influenced by ex-
13
They model the contracting game where the entrant resorts to the financial market, and hence they
endogenise the equivalent of our threshold R above.
14
If takeovers are allowed after the project is realised, the incumbent would want to take over a
successful start-up later on, just before it starts marketing its substitute product.
Big Tech Mergers 11
pectations about what would happen in the continuation game. Here, we endogenise the
probability that the start-up’s project comes into existence: the entrepreneur’s expected
profits is a positive function of its effort.
To account for the ex ante incentives to innovate, consider our base model and extend
it so as to have a pre-stage where the start-up’s owner decides the level of effort e she
makes to innovate. This effort will positively affect the probability the innovation will
be successful and will become the “project”. For simplicity, assume that the level of
effort is exactly equal to the probability of “discovery” of the project. The cost of effort
c(e) is increasing and convex in e, and is such that at equilibrium e ∈ [0, 1]. Expected
profits from effort will be Π(e) = eπe − c(e), where πe is the start-up’s expected payoff
in the continuation equilibrium of the game, and will consist of the takeover price it
obtains (we know at equilibrium it will decide to sell out).15 Equilibrium effort is then
given by e∗ = arg maxe {eπe − c(e)}, which is increasing in πe : on the basis of this simple
formalisation, whatever raises the expected continuation payoff of a project will raise the
probability that a project exists in the first place.
The introduction of this innovation effort stage of the game will allow us to deal with
the ex ante effects of mergers. In particular, many commentators argue that takeovers
by big tech are pro-competitive because they stimulate innovations, and we shall come
back to this issue at the end of section 2.3, as well as when drawing policy conclusions.
In the following sections, we extend the model in three further directions. In section
2.2 we show that our model can be reinterpreted to consider “conglomerate” mergers,
according to which the entrant is active in an independent market but that from there
it could invade the incumbent’s market with some positive probability. In section 2.3 we
consider the case where there is an outsider who also competes for the start-up. Finally,
in section 2.4 we endogenise the resources of the start-up.
an environment in which our incumbent firm, I, is active in market 1, and the potential
entrant, E, has a project in another market, 2. If successfully developed, the project
gives E a monopoly position in market 2, but might also give it, with some probability q,
the ability to launch a new product which would challenge the incumbent on the latter’s
turf, i.e. market 1. (Note that if I acquires E, this may be seen as a “conglomerate
merger”, since E’s is ostensibly operating in unrelated market 2.)
As before, the project development will cost K, and the probability of success is p,
and E needs to have at least resources R) to develop the project. The model will be
the same as in the previous section, after reinterpreting the payoffs as follows. If E
remains independent, has enough resources, and successfully develops the project, it will
make expected profits πEd = πE2 + qπE1d and the incumbent will make in expectations
πId = (1 − q)πI1m + qπI1d . If I takes over E and successfully develops the project, it will
be a monopolist with two products in market 1 and one in market 2, and its expected
profits will be πIM = πI1m + πE2 + q(πI1M − πI1m ). If the project is shelved, or is not
successfully developed, the incumbent will remain a single-product monopolist and will
earn πIm = πI1m .
Note that the assumption that an entrant with sufficient resources will want to develop
K
the project can be rewritten as πEd = πE2 + qπE1d ≥ p
.
Further, assumption (A2) becomes, after using the above definitions and simplifying:
meaning that the original inequalities in market 1 payoffs given by (A2) suffice for the
present extension.
Once re-interpreted this way, all the results obtained in the base model follow through.
First, it is immediate to check that if it acquires E, the incumbent will want to invest if
and only if:
p[πI1m + πE2 + q(πI1M − πI1m )] + (1 − p)πI1m − K ≥ πI1m , (5)
Remark 1. At equilibrium, the “conglomerate” takeover will always take place, and:
K
• If πE2 + q(πI1M − πI1m ) < p
, the incumbent will not develop the project.
K
• If πE2 + q(πI1M − πI1m ) ≥ p
, then the incumbent will develop the project after the
Big Tech Mergers 13
takeover.
Proof. E’s expected value at t = 1, absent the takeover, will be given by:
To find the incumbent’s valuation for the purchase of firm E, consider two cases.
First, suppose that (6) holds. In this case, I’s willingness to pay for E will be:
VI = πI1m +p[πE2 +q(πI1M −πI1m )]−K−[πI1m −pq(πI1m −πI1d )] = p[πE2 +q(πI1M −πI1d )]−K. (8)
The takeover will take place as long as VI0 ≥ VE , which after simplifying becomes:
which is always satisfied when condition (9). Indeed, recall that πI1M ≥ πI1d + πE1d ; hence,
it must be pπE2 − K + pq(πI1d + πE1d ) < pπE2 − K + pqπI1M < pqπI1m .
To summarise, this model shows an incumbent may want to acquire a start-up whose
complement product to the incumbent may become a substitute to it. It may therefore
provide a rationalisation of Crémer et al. (2019, p. 116), who argue that “[c]ompetitive
threats will typically come from the fringe ... Buying up promising start-ups that offer
fringe products or services may therefore result in early elimination of potential compet-
itive threats—which may be particularly problematic if done systematically.”
Several other commentators have identified in the risk of dis-intermediation a frequent
reason why incumbents may want to remove successful start-ups. (An incumbent typically
enjoys a bottleneck position which obliges users of complementary services to access the
platform under the control of the incumbent firm to be able to use them. The incumbent
Big Tech Mergers 14
firm may then fear dis-intermediation, which occurs whenever the other firm offering
complementary services manages to bypass the platform.)16
they might have (some would say they had already) developed into competitors of Face-
book in the social network apps market. And in the case of Waze, it was clear that it
was already active in turn-by-turn navigation apps, and as such it was already competing
with Google Maps (and perhaps it might have expanded to compete more generally with
Google Maps).
Based on our model, it is also worth noting that a laissez-faire approach with respect
to a takeover “war” between the insider and outsiders would be unlikely to deliver a good
outcome: the incumbent—pushed by the objective of protecting its incumbency profits—
20
would be more likely to win. An example of this may come from the Google/Waze
merger. Press articles reported that both Apple and Facebook had shown interest in
purchasing Waze. But, in our view unsurprisingly in the light of the discussion above,
Google was willing to pay more and eventually secured the purchase of Waze.
A few caveats on the policy of prohibiting incumbents from taking over firms whose
trajectory may lead them to enter the incumbent’s core business may be at issue, though.
(1) There may exist cases where no outsider is likely to have an interest in taking over E.
Or (2) such an outsider may exist but might not be endowed with sufficient resources to
guarantee that the start-up’s project will be developed. These are all possibilities that
an AA should assess on the facts of the case. Note, though, that unless the incumbent
would be able to show that it has the ability and above all the incentive to develop the
project in case of takeover, a priori there would be no reason to prefer the incumbent to
the outsiders. (We shall come back to the questions of burden and standard of proof in
merger control in the conclusion.) (3) As Fumagalli et al. (2020) note, a policy prohibiting
the incumbent from taking over a potential competitor may have some adverse ex ante
effects. Indeed, note that if the incumbent cannot bid for E, then the takeover price
d
received by E will fall from πO (the price paid by I when competing with O in the
takeover) to πEd or 0 (when O can takeover E by simply offering it the outside value).
This may have an ex ante effect, if—as posited in the previous section—equilibrium effort
e∗ is a nondecreasing function of the expected payoff of the innovator, πe .
endogenises them by making them depend on market outcomes which take place before
the acquisition offer is made. More concretely, we return to the base model and extend
the game studied there by adding a product market interaction at a time t = 0, and by
allowing the incumbent to take a strategic decision which may affect such an interaction.
In this way, we intend to capture, within a stylised model, the idea that an incumbent
may engage in exclusionary practices aimed at preventing an entrant from challenging
its core market. Such practices may be refusal to supply, degradation of interoperability,
tying or bundling, or imitation of the entrant’s products.21 They may be costly for the
incumbent in the short-run, but—to the extent that they will reduce the ability of the
entrant to contest its dominant position in the core market—they may turn out to be
profitable in the long-run. Our treatment below can be seen as the reduced-form of
more elaborated and better formalised models, such as Carlton and Waldman (2002) on
tying, Fumagalli and Motta (2019a, 2019b) respectively on tying and vertical foreclosure
(refusal to supply but also other exclusionary practices including exclusive dealing and
predation), and—the closest paper in spirit to our treatment here—Motta and Shelegia
(2020) on denial of interoperability and imitation. In all these papers, an incumbent has
a “dynamic” incentive to exclude, in the sense that it engages in a practice aimed at
excluding or marginalising a rival (by decreasing its scale, customer base, or profits) to
protect or enhance future profits, in settings where the market today is different from the
market tomorrow. The extension below shares the same feature.
The extended game Suppose that when the game starts, the entrant is active in
a certain market, which may be the same as the core market of the incumbent, or an
adjacent market, or an independent one. At t = 0(i), firm I decides whether it wants
to engage in some exclusionary practice (call it the “fighting” strategy) or “accommo-
date”. If the former, the incumbent reduces the visibility, sales and/or profits—that we
describe with the generic term “resources”—that the entrant obtains in the market at
time t = 0(ii); if the latter, the entrant will instead have higher “resources”. We indicate
L H
these realisations respectively by RE and RE , where L stands for “low”, H stands for
L H
“high”, and RE < R ≤ RE . However, fighting may be costly for the incumbent, so that
the incumbent’s profit realisation at t = 0(ii) will be lower when fighting than when ac-
f a
commodating: πI,0 = π I ≤ πI,0 = π I . The rest of the game is unchanged. In particular,
the resources obtained at t = 0(ii) and available to E at t = 1 will determine its ability
21
Note that while the first three practices may result in an abuse of dominance conduct and sanctioned
by competition law, for imitation to be unlawful it would be necessary to go beyond competition law,
and find an infringement of Intellectual Property Rights laws.
Big Tech Mergers 17
• At t = 2, the firm that owns the project decides whether to develop it or not.
• At t = 3, active firms sell in the product market, and payoffs are realised.
Remark 2. If the incumbent can take a strategic decision affecting E’s resources at t = 0,
but acquisitions are prohibited:
• If δ[p(πIm − πId )] ≥ π I − π I , the incumbent will “fight”, profits will be πIm and 0, and
no project will be developed.
• If δ[p(πIm − πId )] < π I − π I , the incumbent will “accommodate”, the entrant will
invest in its project and profits will be respectively pπId + (1 − p)pπIm and pπEd − K.
Proof. By accommodating, the total profits of the incumbent will be VIa = π I + δ[pπId +
(1 − p)pπIm ]. By fighting, they will be: VIf = π I + δπIm . It is immediate that VIf ≥ VIa iff
δ[p(πIm − πId )] ≥ π I − π I .
gains from exclusion, equal to δ[p(πIm − πId )]. In particular, the higher the weight of the
future market realisation, δ, the higher the chance the entrant could come up with a
substitute, p, and the greater the foregone profits from future competition, πIm − πId , the
stronger the incentive to engage in exclusionary practices.
Note also that not all the exclusionary practices may be equally costly. For instance,
Motta and Shelegia (2020) show that imitation may actually increase the short-run prof-
its of the incumbent: in their model, the entrant is initially active in a market which
is complementary to the primary product of the incumbent, and by copying E’s com-
plementary product, which reduces its price and expands its sales, the profits made by
the incumbent on its primary product would actually increase in the short-run (unless
the cost of imitation are too high), while the entrant’s sales and profits would fall. But
practices such as tying, predation, and vertical exclusion would typically harm both the
rival and the incumbent.
Solution, when acquisitions are possible Let us now solve the whole game, for
the case where acquisitions are possible. Given the incumbent’s strategies at t = 0, we
already know that at the continuation equilibria the takeover will always take place, and
that the payoffs will also depend on whether the incumbent has or not an incentive to
develop, as summarised by Table 2.
In particular, note that if the incumbent fights, E would not have enough resources to
develop, and as a result I will be able to acquire E at zero price; instead, if the incumbent
accommodates, absent the takeover E would be able to develop and hence the incumbent
will have to make a bid equal to pπEd − K.
Remark 3. If the incumbent can take a strategic decision affecting E’s resources at t = 0,
and acquiring it at t = 1:
Proof. Consider first the case where πIM − πIm ≥ K/p. By accommodating, the total
profits of the incumbent will be VIa = π I + δ[pπIM + (1 − p)πIm − pπEd ]. By fighting,
they will be: VIf = π I + δ[pπIM + (1 − p)πIm − K]. It is immediate that VIf ≥ VIa iff
δ[pπEd − K] ≥ π I − π I . Consider next the case where πIM − πIm < K/p.
By accommodating, the total profits of the incumbent will be VIa = π I + δ[πIm −
(pπEd − K)]. By fighting, they will be: VIf = π I + δ[πIm ]. It follows that VIf ≥ VIa iff
δ[pπEd − K] ≥ π I − π I .
Discussion and the “kill zone” argument This extension allows us to focus our
attention on exclusionary conduct as an alternative strategy for an incumbent to remove
a potential competitor.23
While there are reasons to think that a takeover may be a more convenient way to
exclude a potential rival, because it does not entail the dissipation of market profits,
there may be reasons for an incumbent to select an exclusionary conduct.24 In some
cases, the start-up might have very optimistic predictions about its future, or may want
to have a fair share of the profits that its elimination would save to the incumbent;25
or the exclusionary conduct may not be costly in the short-run, as it would be the case
when the incumbent can imitate the start-up’s features.
23
As discussed in section 2.2, the threat may come from a complement, and a platform may want to
use foreclosure to avoid any risk of dis-intermediation, as discussed inter alia by Bourreau and de Streel
(2019) and Scott Morton et al. (2019, pp. 72-73). The latter mentions the example of Facebook, which
excluded the video-capture-and-sharing app Vine when Vine attempted to link its users to their Facebook
friends. A discussion of abuse of dominance cases involving big tech firms is beyond the scope of this
paper. The European Commission has certainly been at the forefront of competition enforcement in this
area, as witnessed by its decisions against Google (Google Shopping, 27 June 2017; Google Android, 18
July 2018, Google AdSense, 20 March 2019), and its ongoing investigations on Amazon’s Marketplace
and on Apple’s App Store’s policies towards Spotify.
24
In some jurisdictions, including the US, antitrust enforcement against exclusionary conduct and
merger control have been notoriously lenient, so competition law alone would not explain preference for
either strategy.
25
In terms of our model, the incumbent may not necessarily have all the bargaining power.
Big Tech Mergers 20
But note that exclusionary conduct and acquisition may actually be complementary,
rather than substitute, strategies. As we have seen, an exclusionary strategy allows
the incumbent to reduce the acquisition price: as showed in Table 2, by “fighting” the
potential entrant, I reduces the price at which it will take it over. This is very similar
to Saloner (1987)’s model, which formalises the old argument that predation allows to
acquire a rival at a lower price.
This discussion also bears on the so-called “kill zone” argument, according to which
new firms would stay away from the core market of large digital platforms, not to be
killed by their aggressive behaviour or acquisition strategies.26 First of all, note that the
prospect of being acquired should actually push a start-up towards, rather than deter
from, the territory that an incumbent intends to protect. This way, the start-up would be
able to get at least a fraction of the monopolist profits that the incumbent saves through
the acquisition.27 The existence of a “kill zone”, if any, could therefore not be explained
by the possibility of being acquired, but rather by the threat that the incumbent may
engage in exclusionary practices to prevent competition from taking place. And if the
start-up ends up being acquired, it will be at a much lower price, which is driven down
by the fact that an exclusionary practice reduces the outside value of the start-up.
Motta and Shelegia (2020) formalise a model where a start-up operating in a com-
plementary market to an incumbent’s primary product may choose between a trajectory
leading it to a substitute or to a complement of the incumbent, and show that if the
start-up expects the incumbent to have an incentive to engage in an exclusionary strat-
egy (refusal of interoperability or imitation), then it will prefer to stay away from the
incumbent’s primary product and develop a complement instead.
mergers being scrutinised by AAs additional theories of harm are likely to be put forward
in the future. Insights into the effects of mergers developed in other contexts may well
apply to big tech mergers. Here, we point to a few specific market environments in which
big tech mergers can be investigated under the following novel theories of harm. First,
mergers might be detrimental to consumer welfare in markets in which consumers pay a
zero price. Second, when increasing the installed base through a merger the prohibition
of the privately most valuable merger can lead to a more efficient outcome. Third, a
merger of firms operating two-sided platforms can be anticompetitive because of induced
price effects. Fourth, a conglomerate merger that allows the combination of data may
be anti-competitive. Fifth, a conglomerate merger that enables one-stop shopping can
inflict consumer harm. Sixth, a conglomerate merger with bundling of a “free” service
may be profitable but make a more-efficient competitor non-viable.
We acknowledge that there are also various efficiency defences that do not apply to
standard industries. In particular, many big tech firms provide intermediation services
that are subject to positive network effects. With positive network effects a more con-
centrated market may bring benefits to consumers. Thus, mergers may be seen in a
more favourable light.29 We account for some such efficiencies in the specific theories we
consider below, but do not aim to provide a set of theories of efficiency defences that are
relevant in the context big tech mergers.
(e.g. because of additional network effects or because it leads to an improvement of the ecosystem), but
adverse long-run effects may arise from lower innovation or from making it more difficult to contest a
platform’s dominance.
29
For a formal analysis in this vein with two-sided platforms, see Correia-da-Silva et al. (2019).
30
See e.g. Motta and Tarantino (2017).
Big Tech Mergers 22
equal terms. Denoting duopoly profits conditional on acquired user base by π(ni , nj ) for
a firm with an installed base of ni and the competitor with a base nj after the merger,
the acquisition of firm 3 is worth π(3, 1) − π(2, 2) to firm 1 and π(2, 2) − π(1, 3) to
firm 2. The condition that firm 1 is willing to bid more than firm 2 can be rewritten
as π(3, 1) + π(1, 3) > 2π(2, 2) saying that duopoly industry profits are larger if assets
are distributed more asymmetrically. Thus, blocking the merger between firms 1 and
3, would lead to a symmetric duopoly outcome and, without network effects, consumer
welfare could be evaluated in a standard duopoly model in which the merger may affect
product differentiation or product quality. With network effects, quality depends on
network size and thus the comparison between the two mergers depends on how network
size affects network benefits.
Suppose first that network benefits are linear in installed base. If platform 1’s acquisi-
tion is approved, consumers on platform 1 enjoy large network benefits, while consumers
on platform 2 only small benefits. For given prices, consumers are better off if firm 1
acquires firm 3 than if firm 2 did, since network benefits are overall larger in the former
case. For numerical illustration, let network benefits be equal to the installed base and
used as our measure of consumer welfare (when taken as the sum over all consumers).
Then consumer welfare is 10 in the former (because three consumers enjoy a network
benefit of 3 and one consumer of 1) and 8 in the latter case (because each of the four
consumers enjoys a network benefit of 2).
Consider now the case that network effects are concave but not linear. Then a large
firm in terms of user base does not gain much by adding some additional users (through
an acquisition), but a smaller competitor would. To constuct a simple numerical example,
suppose that the benefit offered to consumers is 1 with network size 1 and 3 with network
size 2 or 3. Firm 1’s acquisition then generated consumer welfare of 10 while firm 2’s
acquisition would generate consumer welfare of 12. In this case, consumers are better
off, if the proposed acquisition by firm 1 is prohibited, as this opens the door to firm 2
acquiring firm 3.
This example may be criticised on the ground that we did not perform a proper
equilibrium analysis. In an equilibrium assessment we would need to take into account
how firms adjust their strategies. However, the following would do. Suppose that firms
offer their service for free and installed base maps into consumer benefits as outlined
above. Firms are also horizontally differentiated such that in duopoly with equal service
quality half of the consumers go to firm 1 and the other half to firm 2, while with service
qualities 3 and 1, respectively, three consumers go to the firm with higher quality. This
Big Tech Mergers 24
leads to consumer demand as calculated in the examples.32 But what about firm profits?
Firm monetise on the advertiser side. A larger installed base allows them to better target
advertising. In this case, we must have π(3, 1)/3 > π(2, 2)/2 > π(1, 3), which implies
that π(3, 1) + π(1, 3) > 2π(2, 2) must hold. Hence, in this case, firm 1 outbids firm 2 and
it depends on the way in which installed based maps into service quality whether the
merger is consumer welfare increasing or decreasing compared to the alternative merger
in which firm 2 acquires firm 3.
independent the former format applies. After a merger, the common owner can decide
which format to choose. Each platform carries only one ad (with two or more ad slots
advertisers would bid zero under either format). The monopoly profit with intermediated
trade is assumed to be 500 and the duopoly profit is assumed to be 200. Advertisers are
asymmetric in the sense that advertiser X does not need access to consumers through
a platform, e.g., because it is a well established brand that does not need to advertise
to consumers. Thus, if advertiser X does not advertise so that advertiser Y obtains
advertising slots on the two platforms advertiser X obtains duopoly profits.
With sequential auctions, we start with the second auction. If advertiser X won the
first auction, the analysis from above applies. Thus advertiser X wins the second auction
at price 200. If advertiser Y won the first auction, both advertisers will bid zero as it
does not matter who wins the auction. In the first auction, advertiser X obtains 300 if
she wins and 200 if she loses. Thus she is willing to bid 100. Advertiser Y obtains 200
if she wins the first auction and 0 if she loses. Thus she is willing to bid 200. Hence,
advertiser Y will win the first auction and pay 100; both advertisers then bid zero in the
second auction. If it is equally likely that firms 1 and 2 run the first auction the expected
profit of each independent intermediary is 50.
If the two platforms have one owner, the owner may decide to auction off the bundle
of one slot on each platform. Winning the auction gives advertiser X the profit 500
while losing gives 200. Thus advertiser X bids 300. Since advertiser Y is only willing
to bid 200, advertiser X wins the auction. It has to pay 200 and obtains a net profit of
300. This implies that the owner of the two platforms will make a profit of 200. With
sequential auctions the joint profit of the two platforms is 100 which is less than in the
case in which the bundle is auctioned off.
This finding can be interpreted as follows. A merger between platforms (which co-
ordinate their selling of attention) is profitable because it reduces product market com-
petition. The ensuing higher seller profits are partly extracted by the intermediary who
runs the platforms. The merger is consumer welfare and total welfare decreasing as it
preserves the monopoly position of one of the advertisers in the respective product cat-
egory. This issue arises if the attention market becomes rather concentrated after the
merger (as is clearly the case in our example).33
33
Looking beyond the two-platform example, another take-away for AAs is that it matters how many
consumers are using a certain set of platforms; that is, the actual and counterfactual homing decisions
of consumers are important to evaluate the effect of the merger.
Big Tech Mergers 26
aA (δA (1), 0, δC (λC (sB , sC ))) − K(sA ) + λC (sB , sC )aC (δA (1), 0, δC (λC (sB , sC )), 0) − K(sC ).
For given behaviour of firm B the merger is profitable for two reasons. First, in market
1 the firm now obtains higher ad revenues aA (δA (1), δB (λB ), 0). Second, the firm also
obtains higher profits in market 2. The firm has an incentive to increase quality sC .
By increasing its quality (given the competitor’s quality) it increases its market share.
This affects profits in two ways: First, it generates additional ad revenues in market
1; second, the per-consumer ad revenue is higher after the merger; this makes it more
attractive to gain market share. How does the competitor respond? The competitor is
now in a weaker position as it cannot compete with firm C on equal terms. If sB and
sC are strategic complements, after the merger both firms will provide higher qualities;
the proposed merger would then be consumer-welfare increasing, and advertisers would
not be affected.37 If they are strategic substitutes, the stand-alone firm provides lower
quality after the merger (see below).
Take a special case that results in strategic substitutes: Suppose that ad revenues aB
and aC do not depend on the volume of data in market 2, but only on the volume of
data in market 1. The setting then gives rise to profit functions πA = aA (δA (1), 0, 0) and
πC = λC (sB , sC )aC (0)−K(sC ) before the merger and π M = aA (δA (1), 0, δC (λC (sB , sC )))−
K(sA ) + λC (sB , sC )aB (δA (1)) − K(sC ) after the merger; firm B’s profit function in both
cases is πB = λB (sB , sC )aB (0) − K(sB ). Clearly, firm B’s profit maximising quality sB
is decreasing in sC . Hence, in this case, qualities are strategic substitutes. This implies
that due to the data advantage of the merged firm, firm B will provide less quality sB
to consumers after the merger, while firm A will provide higher quality sC . The profit
of the merged firm is higher, while the profit of firm B is lower after the merger. If
operating in the market is costly, firm B may be forced to exit the market. This is
37
More data may go hand-in-hand with better extracting advertiser surplus. If this were the case then
advertisers would suffer from the merger (while consumers continue to be better off).
Big Tech Mergers 28
a version of the efficiency offence argument: Due to the advantage in data the merger
makes the merged firm a stronger competitor; this is consumer-welfare increasing in the
short run (for a given number of firms), but may be harmful to consumers (and society)
in the long term, as firms may be induced to exit. Even with exit, consumers may well
be better off as they may enjoy a high service quality and the conglomerate merger is
then consumer-surplus increasing. However, this is not always the case. Suppose that
aA (0, δB (λB ), 0) = 0 for λB sufficiently small. Since an equilibrium with a merger has to
feature λB < 1/2, firm B may not be able make positive profit as the revenues on the
advertiser side in the equilibrium candidate do not cover its costs. In the long term, firm
B exits and the merged firm then chooses the unconstrained monopoly solution. This
solution may feature a lower service quality than the one under competition prior to the
merger.38
De Corniere and Taylor (2019) provide a richer environment in which data can be used
for various purposes. While we specified a setting in which data are pro-competitive in
the sense that si is increasing in δi , they also provide examples for the opposite situation,
namely that a larger volume of data decreases the utility offered to consumers and thus
data are anticompetitive. In such a situation the merger between firms A and C reduces
consumer welfare in market 2; consumers in market 1 are not affected.
Furthermore, within a pro-competitive environment, suppose that prior to the merger
firm A can share its data with firms B and C. De Corniere and Taylor (2019) focus on
the case in which firm A maximises its profit by sharing its data with only one of the two
firms prior to the merger and analyse the situation in which data collection is endogenous.
Additional data collection in market 1 here not only increases the profits of the firm with
data access in market 2, but also reduces the profit of the firm 2 without data access.
Data collection before the merger is then excessive from the point of view of maximising
firm A and C’s joint profit. After the merger of A and C data from market 1 can still
be given to firm B. However, the merged firm refuses to share these data with firm B
after the merger and reduces its efforts to collect data, as C’s profits are fully internalised
when choosing how much data to collect.39 As a result of less data collection, the merger
is consumer-surplus decreasing in the short term in markets 1 and 2.
Our conclusion is that it is possible to construct theories of consumer harm based on
the combination of data. However, such theories will only be applicable to very specific
cases.
38
We are not suggesting that the firm actually reduces its quality after exit, but in an evolving market
it may simply relax and stop to upgrade its service quality (or slow down this process).
39
Therefore, the merger is profitable.
Big Tech Mergers 29
Absent synergies or any strategic effects across markets, this merged firm would make
profit πAM = πA1 + πC2 . Clearly, firm A is willing to bid πC2 and the firm is indifferent as to
whether the merger goes through. Also, the merger does not change the allocation. We
will now look beyond the simple reduced-form setting and provide a leverage theory that
is of particular relevance in the context of big tech to show that a merger can be profitable
but total welfare reducing. The underlying economic mechanism has been elaborated by
Choi and Jeon (2020), which we embed into the context of a merger. We describe the
basic argument and develop a simple numerical example.
Suppose that firm A attracts consumer attention by offering a product in market
1 at the monopoly price. In the other market the two firms offer their product at a
non-negative price and on top monetise on the advertiser side (which consumers do not
mind). Firm B is more efficient in the sense that it offers higher quality at the same cost
(marginal cost equal to zero). Firms B and C are Bertrand competitors in case both
have developed their product. In this case, firm B makes the asymmetric Bertrand profit
and firm C obtains zero profit. Prior to the merger the two markets are independent.
If firm A merges with a firm with a project in market 2, it may sell a product bundle to
consumers. We restrict attention to firm A acquiring firm C. Then, in case of competition
in the market 2, it can offer a better deal to consumers than firm B by setting a low
price for the bundle. It also has an incentive to do so if the quality difference between B
and C is not too large. Since consumer would like to consume both types of content they
choose the bundle if the bundled price is not too high (here we suppose that a consumer
cannot buy product B if she buys the bundle). In return the merged firm offering the
bundle attracts all consumers and obtains the advertising revenues whenever C is active.
The point is that because of advertising opportunities there is a positive surplus on the
table in market 2. The novel insight arises because of a monetization opportunity on a
different market and the fact that the competitor to the merged firm does not have a
sufficiently flexible strategic instrument to attract consumers (it cannot set a negative
price to consumers).
The specific timing we have in mind is the following: first, firms B and C decide
whether to develop their projects and in case they do Nature decides which if any of
the projects by B and C are successful; second, firm A makes a take-it-or-leave-it offer
to firm C, which firm C accepts or rejects. If the merger goes through, A decides
whether to commit to bundling. And finally before advertisers decide where to advertise
and consumers decide which if any products to buy, firms simultaneously set prices to
consumers and advertisers.
For concreteness, we develop a numerical example. There are 100 identical consumers
Big Tech Mergers 31
who all attach the value of 2 monetary units to product A; 2 to the product offered by
firm B; and 1 to the product offered by C. The success probability of firms B and C is
1/2, which is independent across firms. In addition, each consumer generates advertising
revenue of 3 in market 2 leading to total advertising revenues of 300. All firms incur zero
cost.
Suppose that firms B and C developed their products. Absent merger, firm A sells its
product at price 2 and makes profit of 200; firm B sells at price 1 and makes profit of 400
(revenues on the consumer side of 100 plus revenues on the advertiser side of 300) in case
both firms B and C are successful, which occurs with probability 1/4. With probability
1/4, it is the only one successful and then sells the product at price 2 making overall
profit of 500 in this case. Hence, the expected profit of firm B is 225. With probability
1/4, firm C is the only one successful and then sells the product at price 1 making overall
profit of 400 in this case. Hence, the expected profit of firm C is 100.
We continue with the case that firms B and C developed their products and consider
now the possibility of a merger. With probability 1/4, firm C is the only one successful
and then sells the bundle at price of 3 making overall profit of 600 in this case (300 from
selling the bundle and 300 on the advertiser side). This corresponds to the combined
profits of A and C absent the merger and thus there are no strict incentives to engage in
the merger. When project C is not successful (which occurs with probability 1/2), firm
A clearly has no incentive to merge; in continues to sell in market 1 at the price of 2 and
makes profit 200. The interesting case is when projects B and C are successful. If firm
A acquires C and commits to sell a bundle, it has an incentive to sell the bundle at a
price slightly less than 1. When buying the bundle consumer obtain a net benefit of a
little bit more than 2. Alternatively, consumers could decide not to buy the bundle and
only buy B. The lowest price firm B can charge is zero. In this case, consumers would
get a net benefit of 2. Hence, consumers decide to buy the bundle and firm A’s profit is
400. Thus, with the merger firm A gains an additional 200. Absent the merger, firm C
would have earned zero. Hence, for any acquisition price between 0 and 200 the merger
is profitable for both parties and the merger will take place unless it is prohibited by the
AA.41 Advertiser surplus is unaffected (it is zero); consumer surplus increases from 100 to
41
We observe that also a merger between B and C would be profitable, as this would lead to the
monopoly outcome. We do not consider a merger between B and C on the ground that a competition
authority would reject a merger that removes the only competitor. If we also allowed for A acquiring
B the latter would agree to be acquired at any positive price if it anticipated that otherwise A were to
acquire C for sure. However, it it considers it sufficiently unlikely that A will be successful in acquiring
C it will reject an offer that is much below 400. For example, if it anticipates that the AA will prohibit
the acquisition of C by A with probability 1/2 (or that for other reasons the probability that the merger
between A and C will be consumed is 1/2) firm B will reject any offer by A that is in the latter’s interest,
Big Tech Mergers 32
200. Bundling allows firm A to leverage its monopoly position into a competitive market
and to drive a more-efficient competitor out of the market when this firm has developed
a higher-quality product. As a result, the merger is total welfare decreasing.
We now take a quick look at the ex ante incentives of firm B. Take the special case
that the acquisition price is 100.42 Absent the merger, firm B’s expected surplus is 225
whereas with the possibility of firm A acquiring C it is 100. Thus, if the development
cost is between 0 and 100 firm B will develop the project in any case; if the development
cost is between 100 and 225 it will only develop if a merger is not possible. Absent the
merger, firm B’s expected surplus is 100 whereas with the possibility of firm A acquiring
C it is 125 (400 with probability 1/4 and 100 with probability 1/4).
Hence, the above characterisation holds along the equilibrium path for the develop-
ment cost between 0 and 100. When the development cost passes the threshold of 100
then, with the possibility of a merger, firm B knows at the development stage that its
development costs exceed the expected profit. Then, firm B does not invest. As a result,
the merger is an off-equilibrium threat that chills investment. Firms A and C extract
monopoly rents from their respective markets: firm A obtains 200 and firm C expects
200 (400 with probability 1/2); this is the outcome for development costs between 100
and 200. Then, a merger that includes the acquired free offer into a bundle can possibly
leverage monopoly power from one market into another; this can be seen as creating a
“kill zone”, as more-efficient competitors would decide not to enter market 2.
4 Conclusion
We have argued in this paper that some big tech mergers may well have adverse compet-
itive effects, and hence require the scrutiny of Antitrust Authorities (AAs). Foremost,
the risk that the merger removes a potential competitors often deserves careful considera-
tion. We have developed a simple formal framework that allows us to address some basic
considerations as to when such mergers are likely to be anti-competitive. We have also
provided an informal account of novel theories of harm suggesting that certain mergers
in digital industries, including conglomerate mergers, raise anti-competitive concerns.
What are the implications for merger policy? For sure, the current situation where
acquisitions of big tech firms are systematically not investigated cannot go on. This
implies first that the relevant mergers should be notified, and second that AAs should
have the means to stop those mergers that are deemed anti-competitive. We discuss each
only when efficiency gains are strong enough, the burden of proof that their merger is pro-
competitive should be placed on the merging firms instead.46 This should hold for any
sector, and we submit that a fortiori merger policy would benefit from a reversal of the
burden of proof in case one of the merging parties has an entrenched dominant position,
as it is the case for some of the big tech firms. The merging parties would then need
to provide evidence that either the merger does not raise any significant competitive
issue (this will be the case for instance when the target is a small company which is
developing tools and/or may be acquired only for its human capital — and we would
expect such cases to be the majority) or that expected efficiency gains (which include
dynamic efficiency gains) are sufficiently strong to justify the acquisition.47
Note that the balancing of efficiency gains with competitive harms needs some re-
thinking too. It cannot be the case that an extremely small improvement in a platform’s
offering due to the merger may be sufficient to justify the acquisition of a potential
competitor just because the probability that the latter firm will become an effective com-
petitor in the future is considered small. Even more worrying, the standard of proof for
blocking a merger with a potential competitor appears to be that it is “more likely than
not” that the acquired firm would become an effective competitor. Further, Antitrust Au-
thorities typically need to substantiate such a finding with documentary evidence which
is unlikely to be easily obtained.48
If there is a concern that a potential competitor is acquired, and even if there is no
“smoking gun” proof that indeed at the counterfactual the acquired firm would have be-
come an effective competitor, the optimal merger policy should be to prohibit the merger
(or impose remedies) even if the probability of the entrant being a successful competitor
is considered to be not very large, provided that the expected gains in consumer welfare
from competition are large relative the gains that would come from the upgraded offer of
46
See Motta and Peitz (2019).
47
Scott Morton et al. (2019) suggest to create digital authority (DA), a sector regulator which would
also have additional power over merger review. “These specific merger regulations should require merging
firms to demonstrate that the combination will affirmatively promote competition. This shifting of the
burden of proof from the government (to prove harm) to the parties (to prove benefit) will assist the DA
by placing the job of demonstrating efficiencies on the parties, who have a greater ability to know what
they are.” (Scott Morton et al., 2019, p. 111) The ACCC (2019, p. 199) contemplates that “it may be
worthwhile to consider whether a rebuttable presumption should also apply, in some form, to merger
cases in Australia. ... [A]bsent clear and convincing evidence put by the merger parties, the starting
point for the court is that the acquisition will substantially lessen competition.”
48
Crémer et al. (2019, p. 119) write: “In the Facebook/WhatsApp merger, the Commission found
no documentary evidence that WhatsApp was planning to become a fully-fledged social network in the
future.[..] Such proof that the start-up is planning to enter the acquirer’s core market will generally also
be difficult to obtain in other cases. Clear plans for doing so will rarely exist when start-ups are being
bought up at an early point of their life.”
Big Tech Mergers 35
the merging firm (see Section 2). Such a balance-of-harm approach has been proposed
by Furman et al. (2019), and we fully agree with it.
As discussed in section 2.3, we are aware that the prospects of being acquired by a
large digital platform may increase the incentives of start-ups to innovate in the first
place. However, as we discussed there, prohibiting a dominant firm does not necessarily
mean that a start-up will be unable to sell out. In many instances, there will be other
firms interested in taking it over.49
Broader concerns Big tech may not only develop strong and entrenched positions in
several markets. Their eco-systems may play (and in some case are playing) an important
role in absorbing a large part of consumer attention and accumulating personal data from
many different activities. While this is convenient for many consumers—they can live
their digital live in one or very few eco-systems—they also become dependent on the
masters of these eco-systems.
This raises a number of issues for society, which go beyond standard market power
issues. In particular, some big tech firms can dominate their eco-systems becoming a
private regulator. This means that they also impose rules of conduct, i.e. which type of
behaviour on the platform is proper and which one is either not possible or punished. In
particular, they may exclude participants who allegedly misbehaved. While this is often
in the participants’ overall interest, there is room for negligence and abuse.
Clearly, merger control (and competition policy more broadly) is a highly imperfect
instrument to deal with these and other threats.50 Nevertheless, we would subscribe to
the view that vigilant merger control can sometimes help to avoid problems (or reduce
their severity) that are likely to arise from big tech firms expanding their sphere of
control. This does not make merger control a silver bullet, but an enlightened merger
control appears to us an important component of a policy on how society should deal
with big tech.
49
The fact that the incumbent may be willing to pay a higher acquisition price to defend its dominance
position should not be a reason to allow that acquisition if there are superior alternatives from the
competition point of view. Recall that a similar approach is followed by AAs in case of the so-called
“failing firm” mergers. If there exists another route for the failing firm’s assets to stay in the market,
the failing firm defence would not be accepted.
50
Regulatory instruments and interventions by the legislator may be the more appropriate instruments
to deal with such issues. In particular, regarding digital eco-systems public regulators may provide
oversight.
Big Tech Mergers 36
References
[1] ACCC (2019). Digital platforms inquiry. Final Report, June 2019.
[2] Argentesi, E., P. Buccirossi, E. Calvano, T. Duso, A. Marrazzo, and S. Nava (2019).
Ex-post assessment of merger control decisions in digital markets. Lear Report com-
missioned by the UK Competition and Markets Authority.
[4] Anderson, S. P. and M. Peitz (2019). Media see-saws: Winners and losers in platform
markets. Journal of Economic Theory, forthcoming.
[7] Carlton, Dennis and Robert Waldman (2002). The strategic use of tying to preserve
and create market power in evolving industries. Rand Journal of Economics 33,
194-220.
[8] Choi, J. P. and D.-S. Jeon (2020). A leverage theory of tying in two-sided markets
with non-negative price constraints. American Economic Journal: Microeconomics,
forthcoming.
[9] Correia-da-Silva, J., B. Jullien, Y. Lefouili and J. Pinho (2019). Horizontal merg-
ers between multi-sided platforms: Insights from Cournot competition. Journal of
Economics and Management Strategy 28, 109-124.
[10] Crémer, J., Y.-A. de Montjoye, and H. Schweitzer (2019). Competition policy for
the digital era. Final report presented to the European Commission.
[13] Deneckere, R. and C. Davidson (1985). Incentives to form coalitions with Bertrand
competition. Rand Journal of Economics 16, 473-486.
[14] Franck, J.-U. and M. Peitz (2019). Market definition and market power in the plat-
form economy. CERRE report. May 2019. Available at www.cerre.eu
Big Tech Mergers 37
[15] Fumagalli, C. and M. Motta (2019a). Tying in evolving industries, when future entry
cannot be deterred. International Journal of Industrial Organization, forthcoming.
[16] Fumagalli, C. and M. Motta (2019b). Dynamic vertical foreclosure. CEPR DP No.
12498.
[18] Furman, J., D. Coyle, A. Fletcher, D. McAuley and P. Marsden (2019). Unlocking
digital competition. Report of the Digital Competition Expert Panel, March 2019.
[19] Gilbert, R. J. and D. M. G. Newbery (1982). Preemptive patenting and the persis-
tence of monopoly. American Economic Review 72, 514-526.
[20] Motta, M. and M. Peitz (2019). Challenges for EU merger control. Concurrences No
2-2019, 44-49.
[21] Motta, M. and S. Shelegia (2020). “Kill zone”? Copying and refusal of interoper-
ability to exclude potential competitors. Unpublished manuscript.
[22] Motta, M. and E. Tarantino (2017). The effect of horizontal mergers when firms
compete in prices and investments. CEPR Discussion Paper No. 11550.
[23] Nörback P.J., and L. Persson (2009). The organization of the innovation industry:
Entrepreneurs, venture capitalists, and oligopolists. Journal of the European Eco-
nomic Association 7, 1261-90.
[25] Rhodes, A. and J. Zhou (2019). Consumer search and retail market structure. Man-
agement Science 65, 2607–2623.
[26] Saloner, G. (1987). Predation, mergers, and incomplete information. Rand Journal
of Economics 18, 156-186.