Ex ante or ex post?
The role of information for the timing of merger
control
Andreea Cosnita-Langlais∗and Jean-Philippe Tropeano†
October 6, 2016‡
Abstract
We compare the pre- and post-closing merger control to conclude on the optimal timing of merger assessment. Controlling mergers involves costly evidence provision by both
firms and the antitrust agency. We show that the ex post enforcement has a strong impact both on the population of mergers and on the information provided. The ex post
assessment is optimal in case of highly polarized merger type distribution, i.e. many very
pro- and very anticompetitive mergers.
Keywords: merger control, information, evidence provision
JEL classification: L41, K21, D82
∗
EconomiX-CNRS and University Paris Ouest Nanterre La Défense, 200 Av. de la République, 92001
Nanterre cedex; +33 1 40 97 77 86, [email protected]
†
Paris School of Economics and University of Paris 1; [email protected]
‡
We are indebted to Daniel Rubinfeld, Eirik K. Kristiansen and Francis Bloch for their helpful comments
and suggestions. We also thank various participants attending the EARIE 2012 conference in Rome, the
BECCLE 2012 conference in Bergen, the MaCCI-ZEW 2013 conference in Mannheim, and the seminars at
the universities of Paris 1 and Liege. Jean-Philippe Topeano acknowledges financial support from the French
National Research Agency, grant ANR-POLICRE (2013-2016). All remaining errors are ours.
1
1
Introduction
Merger control is a pillar of competition policy throughout the world. However, the timing of
merger control is not the same across the various jurisdictions: in the EU merger projects are
solely examined ex ante, before consummation, whereas in the US they may be challenged
afterwards. Consummated merger challenges in the US even account for about one-fifth of
total merger challenges since 2001 (Rosch, 2012). This procedural choice of ex post versus ex
ante merger policy enforcement 1 is likely to substantially modify the decision of the antitrust
authority decision. The following examples illustrate this.
In January 2004, the Federal Trade Commission closed its investigation of the consummated merger between Genzyme and Novazyme2 . The FTC’s decision not to challenge this
merger to monopoly was based on evidence of both a lack of anticompetitive effects and synergies made possible by the merger during the two years that followed the merger3 . Had the
merger been examined before consummation, the mere fact that it led to monopoly would
have recommended its prohibition4 . In contrast, the European Commission prohibited in
2012 the Deutsche Börse/NYSE-Euronext merger to near monopoly on the European financial derivatives market5 , partly because most of the future efficiency gains argued by the
parties were not considered as verifiable at the time of the notification6 .
But challenging a consummated merger also lays a heavy burden on the merging firms.
Typically, unwinding a consummated merger found to be unlawful is deemed to involve
substantial costs for the merging parties, in particular when "unscrambling the eggs" is very
1
See the OECD (2014) for the whole range of possible combinations notification-assessment in the different
countries.
2
See details at http://www.ftc.gov/opa/2004/01/genzyme.htm.
3
See Chairman’s Muris statement, available at
http://www.ftc.gov/os/2004/01/murisgenzymestmt.pdf.
4
See Commissioner’s Thompson dissenting statement, available at
http://www.ftc.gov/os/2004/01/thompsongenzymestmt.pdf.
5
Case M.6166.
6
See in particular paragraphs 1187, 1203, 1298, 1324, 1335 and 1337 of the decision.
2
difficult due to the strong integration reached by the insiders7 . Following decisions such as
Chicago Bridge (2005)8 , various commentators have thus remarked that post-close challenges
have the potential to create uncertainty among companies, thereby possibly chilling business
activity9 .
The purpose of this paper is to examine the role of timing of merger control for the
information revealed during enforcement, so as to conclude which timing of merger control
performs better in terms of decision errors and eventually welfare, the pre- or the post-closing
enforcement. For this we use a stylized model accounting for both the evidence provision
by the parties and the risky prospect of a costlier post-consummation ban. Contrary to
conventional wisdom, according to which there might be a trade-off between the informational
advantage for the agency and the costly risk for firms associated with ex post assessment, we
find them to be complementary.
We base our analysis on a model where the merger has both pro- and anticompetitive
effects, but this is the firms’ private information. The key assumption is that the enforcement
relies on the confrontation of evidence, and this confrontation takes place either before (ex
ante) or after the closing (ex post). In order for the merger to be cleared, the merging firms
must provide verifiable costly evidence on pro-competitive arguments in favor of their merger,
such as efficiency gains for instance, enough to counterbalance the evidence provided by the
competition agency on the anticompetitive effects it has identified during the assessment.10
7
8
Again, see OECD (2014).
Chicago Bridge & Iron Co. v. FTC, 515 F.3d 447 (5th Cir. 2008), also at
http://www.ftc.gov/os/adjpro/d9300/index.htm.
9
See for instance http://www.deallawyers.com/blog/2005/01/undoing-done-deals-the-chicago-bridgedecision.html: "Post-close challenges paralyze markets because they threaten to strip companies of acquired
assets and of years of independent product development that they would have engaged in but for the futile
attempt to acquire a competitor. [...] Further, even if divestiture does return a company to the position
it was in before the merger, the company would be at a disadvantage compared to competitors who have
moved on and continued to develop next-generation products during that time. Lost opportunities could be
considerable.”
10
For instance, the European Merger Regulation makes clear that an otherwise anticompetitive merger can
be cleared provided that the benefits from the merger outweigh the negative effects, but the burden of proof
3
The merging firms’ evidence provision involves a direct cost but also a substantial opportunity
cost, due to the delay that the firms agree to incur when they decide to submit to the lengthy
assessment by the CA, and which will prevent them from turning to alternative projects for
instance (Ormosi 2012, p. 580).
We show it may be optimal for the competition agency (CA henceforth) to commit to
controlling mergers ex post despite the extra cost imposed to firms. This result is driven
by the twofold impact of the ex post control. First, the extra cost may help the CA to
induce self selection by leading certain merging firms to refrain from merging. This is the
selection effect of the ex post control, and it is always welfare-improving because the deterred
mergers are always anticompetitive. But the extra cost may also lead firms to invest more
in evidence provision ex post - this will be termed the informational effect of the ex post
control. The welfare impact of the informational effect is typically ambiguous, since both
pro- and anticompetitive merging firms may be induced to provide more evidence in favor
of their merger. Our analysis identifies the cases where the ex post enforcement is optimal,
thanks to a positive informational effect, and also the sufficient conditions ensuring that a
negative informational effect is lower than the positive selection effect, again making the ex
post control preferable.
In terms of policy implications, we obtain that the shape of the distribution of merger
types is critical. To put it short, the ex post control is preferable whenever the CA is very
likely to face very anticompetitive mergers and very efficient mergers. Instead, whenever
there is a low competitive risk associated with the merger, i.e. the merger type distribution
is concentrated around the medium range, where both competitive effects of mergers, positive
and negative, are relatively close, then the ex ante control of mergers is more efficient. Let
us briefly sketch the intuition. First, the ex post enforcement is more efficient for very
anticompetitive and very pro-competitive mergers, since the very anticompetitive mergers
will be deterred, while the very pro-competitive mergers will have incentives to invest in
evidence provision, thereby reducing the risk of type I errors in case of control by the CA.
is on the merging parties - see Paragraph 87 of the Horizontal Merger Guidelines.
4
Both effects are welfare-improving. Instead, mergers with intermediate anticompetitive effects
are less efficiently controlled ex post. These mergers are not deterred because they remain
profitable enough to incur the extra cost in case of ex post prohibition, and they even have
higher incentives to invest in evidence provision to reduce the risk of a ban. This increased
level of evidence in their favor makes the prohibition of these merger types more difficult ex
post than ex ante.
As such, this paper contributes to a recent literature dealing with the optimal timing of
competition policy. Barros (2003) and Berges et al. (2008) study the opportunity of mandatory notifications for the agreement exemptions under Art.101 TFEU. Choe and Shekhar
(2010) consider the same question, of compulsory ex ante notifications, but in the case of
mergers. Our analysis departs from these articles in two main aspects: we focus on the role
of evidence provision, both before and after consummation, while the existing papers ignore
this point, and moreover, we endogenize the evidence available. We also complement in a
dynamic framework Lagerlöf and Heidhues (2005), who discussed the costly evidence production, but in order to establish the desirability of an efficiency defense in merger control.
More recently, Dertwinkel-Kalt and Wey (2016) investigate the role of remedies for information acquisition in merger control under different institutional settings, i.e. adversarial
vs inquisitorial systems. To our knowledge, there is only one paper, Ottaviani and Wickelgren (2011), dealing with the impact of the information available on the timing of merger
control. Their paper shows that the ex post enforcement always performs better as long as
firms’ profits are not too risky. We offer the opposite policy recommendation, by considering alternative assumptions: asymmetric information between the agency and the merging
firms both with ex ante and ex post enforcement, and fully endogenizing the information
available ex post. Finally, our paper is also related to the literature studying the changes in
the population of mergers due to a change in enforcement (see Nocke and Whinston (2010),
Armstrong and Vickers (2010) or Burguet and Caminal (2015)), by considering the outcome
of a stricter, more severe merger control. However, and in contrast to these articles, we allow
for asymmetric information whenever the competition agency examines the merger.
5
The paper proceeds as follows: the model is presented next, and the main results derived
in Section 3. We briefly discuss possible policy implications before concluding.
2
The model
We consider a reduced-form merger control game between two risk-neutral agents: the merging firms and the competition authority. All projects in the population of potential mergers
enhance market power, and thereby give rise to an anticompetitive effect - typically, a price
increase. However, there may also be a pro-competitive effect from mergers, such as cost
savings or a quality increase due to synergies. We consider the population of mergers to be
heterogenous w.r.t. the size of this pro-competitive effect. For this we consider a continuoustype framework where e denotes the size of the pro-competitive effects. Let e be distributed
according to cdf F (x) on the interval [e, e] , with
e
e ef(e)de
the expected average level of
efficiency gains and f(e) the density function.
The payoffs
The profit gain from merger is denoted π(e) for type e. We assume that the profit gain
increases with e and that π(e) ≥ 0. Let W (e) stand for the change in consumer surplus due
to a merger of type e. W (e) is increasing with e and to avoid trivial cases we assume that
W (e) < 0 < W (e). We denote by e the welfare-neutral type, i.e. such that W (e) = 0.
Then we have e > e. These assumptions on the monotonicity of profit and consumer surplus
functions are compatible with Cournot competition with homogenous goods11 , as well as with
price competition for some specifications of product differentiation12 .
The CA is assumed to apply a consumer surplus welfare standard, which is in line with
the current practice in most antitrust jurisdictions.13
The information available and the production of evidence
11
See Neven and Röller (2005), p.833-834.
See Neven (2001), p.432-433.
13
In practice, mergers get cleared or banned depending on the expected competitive impact, which is
12
typically assessed in terms of expected post-merger price variation.
6
The type of a given merger is the insiders’ private information.
We model the merger assessment by the CA as a confrontation of evidence between
the CA and the insiders, that can take place either before ("ex ante") or after the merger
consummation ("ex post"). This confrontation may lead to errors because parties find and
provide more or less convincing pieces of evidence.
First, to challenge the merger, the CA must provide evidence on the anticompetitive
effects of the merger. To fix ideas, one can think about this as a simplified version of the
European setting, where the EC needs to argue convincingly likely anticompetitive effects in
order for a merger project to be subjected to a Phase II investigation. We assume that if
it invests an amount x to obtain evidence, the CA actually obtains it, and thereby actively
challenges the merger. This cost of evidence provision for the agency is distributed on the
interval [0, +∞[ according to the cdf G(x).
Faced with the merger challenge from the agency, and in order to secure the merger
approval, the merging firms need to provide in response enough evidence on the merger’s
pro-competitive effect to counterbalance the agency’s own evidence on the merger’s anticompetitive impact. Such convincing arguments in favor of their merger may be the merger
efficiency gains, an easy market access, a fast pace of innovation, or incentives to increase
or maintain product quality. We assume that if the insiders invest in information, incurring
the cost c, this will increase the probability to provide convincing sufficient evidence from 0
to h > 0. For ease of exposition, we do not allow the probability h to depend on the type
e. But doing so would not change qualitatively our results. We interpret c as encompassing
both the direct cost of evidence as well as the opportunity cost engaging in evidence provision instead of for instance delaying alternative profitable projects due to the lengthy merger
control process14 .
Finally, and so as to write off any other source of exogenous bias in the ex post vs. ex
ante comparison, we voluntarily assume the investment cost in evidence provision to be the
14
We should here that the ex post opportunity cost may be lower since the merger is ongoing. We keep it
unchanged for simplicity. Our results are reinforced in case of a lower opportunity cost.
7
same for the firms as well as for the agency, both ex ante and ex post. Importantly, if the
merger is banned after consummation, and thus the insiders are constrained to undo their
merger ex post, they incur an additional fixed cost equal to k > 0.
The merger control game
The timing of the game will be the following:
Stage 1: The CA decides whether to control mergers ex ante. This decision is observed
by the firms.
Stage 2: Merging firms observe their type e and decide whether to engage in merger or
not.
Stage 3: The CA observes its cost of control x and decides whether to challenge or not
the submitted merger.
Stage 4: If the merger is challenged, firms decide to provide evidence counterbalancing
the agency’s at cost c .
A merger project will be accepted if the CA’s evidence on the anticompetitive effect is
met by overwhelming proofs in favor of the pro-competitive effect brought by the insiders,
or if the CA does not challenge the merger because of lack of evidence. Otherwise, the CA
blocks the merger.
We determine next the Perfect Bayesian Equilibria of this game.
3
The optimal timing of merger control
In what follows we compare the outcome of the merger control depending on whether it takes
place before or after the merger consummation.
In case of ex ante merger control, given the lack of extra cost from an ex post unscrambling
of the merger, it is straightforward to see that it is optimal for the insiders, whatever their
type, to engage in merger at stage 2. The only decision left to discuss is that of evidence
provision at stage 4 in order to secure merger approval in case of a challenge by the agency.
With ex ante assessment, type e insiders will invest in evidence provision incurring a cost
8
c iff π(e) > hc . This leads to an efficiency gains threshold eI above which the merging firms
provide convincing information with probability h. Given that the merger profit is increasing
in efficiency gains, one has that the probability to provide convincing evidence increases with
e as well.
As far as the agency is concerned, at the previous stage (stage 3) it decides to challenge
the merger iff the expected benefit from controlling the merger is higher than the corresponding cost. This is the case whenever the welfare gain from merger control, equal to
e
Max(eI ,e) hW (e)f (e)de
−
e
e W (e)f(e)de,
exceeds its cost, equal to x. We denote by xea the
resulting threshold of merger control cost below which the agency invests in evidence provision
to challenge a merger.
The following proposition summarizes the Perfect Bayesian Equilibrium (PBE) and the
outcome of the ex ante control:
Proposition 1 In case of ex ante assessment, there is a unique Perfect Bayesian Equilibrium
where the CA challenges the merger iff x ≤ xea and where the firms always engage in merger.
At the equilibrium firms with e > Max(eI , e) invest in evidence provision in case of control.
Proof. Firms incur no cost to engage in mergers. Thus all firms engage in mergers.
A firm invests in evidence provision iff e > eI . The CA challenges the merger iff the exe
e W (e)f(e)de
pected welfare without control
e
Max(eI ,e) hW (e)f (e)de − x.
This leads to the threshold xea . We deduce the expected welfare
at the equilibrium: (1 − G(xea ))
xea ), where E(x/x < xea ) =
is lower than the expected welfare with control
xea
x
e
e W (e)f(e)de
g(u)
G(xea ) du. The
+ G(xea )
e
Max(eI ,e) hW (e)f(e)de
− E(x/x <
first term stands for the welfare outcome of
uncontrolled merger projects, the second one to the welfare outcome of controlled and cleared
merger projects, whereas the third is the cost of control or evidence production for the CA.
The intuition is direct: ex ante, because firms have always incentives to engage in merger,
the decision to challenge the merger is the result of a trade-off between the risk of prohibiting
pro-competitive merger types (e > e) and clearing anticompetitive projects (e < e). The
9
question is to what extent the ex post control is likely to relax this trade-off, and thereby
improve the expected welfare.
Let us now turn to the ex post control. The only exogenous difference between the ex
ante and the ex post assessment is the extra cost incurred by firms in case the CA blocks
the merger and they have to undo it after consummation. The merging firms may decide not
to merge to avoid that extra cost. Therefore we first discuss below the insiders’ decision to
engage in merger under the ex post assessment.
After merging, if the merger is challenged, the merger of type e invests in evidence provision iff π(e) >
c−k
h .
This yields a threshold eI (k) above which the merging firms provide
convincing evidence with probability h. Note that the extra cost incurred ex post increases
the incentive to invest in information: eI (k) decreases with k.
At the previous stage, anticipating this investment in evidence provision, type e insiders
choose whether to engage in merger or not. Clearly, whenever type e firms expect the agency
to challenge the merger with probability G(x), their merger is profitable iff G(x)Max(hπ(e)−
(1−h)k −c, −k)+(1−G(x))π(e) ≥ 0. This condition implicitly determines the cut-off merger
type, indifferent between merging and not merging, denoted e(x, k). Insiders decide to engage
in merger for enough efficiency gains, i.e. for e ≥ e(x, k). It is straightforward to see that the
higher the probability of control G(x), the higher the minimum level of efficiency gains that
insiders need in order to merge - thus, e(x, k) is increasing in x. It is worth stressing that the
CA induces an ex ante selection of mergers that will eventually be controlled ex post, and
this will prove crucial for the outcome of the ex post merger assessment.
Turning now to the CA’s behavior in case of ex post merger enforcement, the agency
will control mergers only if it is worth doing so, given the cost of evidence provision and the
expected population of mergers actually submitted. In other words, the intensity of the ex
post control chosen by the CA will depend on the expected cut-off merger type, e. Explicitly,
the CA exerts the ex post merger control iff
e
e
Max(e,eI (k)) hW (e)f(e)de − e W (e)f(e)de
≥ x.
Denote x(e) the optimal level of ex post enforcement chosen by the CA. Note that the higher
10
the expected marginal merger-type e, i.e. the more pro-competitive the submitted mergers,
the lower will be the optimal intensity of control x(e). The PBE of the ex post enforcement
game obtains at the intersection of best reply functions x(e) and e(k, x). Below we define
and characterize this PBE:
Proposition 2 a) In case of ex post assessment, there exists a unique Perfect Bayesian
Equilibrium where the CA challenges the merger iff x ≤ xep ≡ x(eep ) and the merging firms
decide to engage in merger iff e ≥ eep ≡ e(xep , k). At the equilibrium, the firms with e >
Max(eI (k), eep ) invest in evidence provision in case of control.
b) At the equilibrium: (i) eI (k) < eI , (ii) eep < e, and (iii) eep > e if the cost k of undoing
the merger is high enough.
Proof. a) The Best Reply function e(k, x) increases with x, with e(k, 0) = e and e(k, x) >
e if x → ∞. The Best Reply function xep (e) decreases with e, with xep (e) = 0 > e for e > e
and xep (e) > 0. Then, it straightforward to see that there is a unique intersect between both
functions denoted by xep and eep .
b) (i) Given the higher cost from ex post enforcement for the firms, i.e. k > 0, one has
that eI (k) < eI .
(ii) eep < e is implied by xep (e) = 0 for e > e.
(iii) Two cases may occur: either eep = e or eep > e. This depends in particular on the
relative position of e(x, k). An increase in k increases e(xep , k). Then for k high enough we
have e(x = ε, k) > e with ε close to 0. This ensures that eep > e. In addition, for k very high,
e(x, k) is very steep for x close to zero, so that eep becomes close to e.
Proposition 2 characterizes the outcome of the ex post merger enforcement.
First, the equilibrium is unique, defined by the couple formed by the optimal cost threshold
below which the CA controls mergers and the efficiency level above which the firms decide to
provide evidence. Not surprisingly, the higher cost in case of post-consummation ban induces
the merging firms to provide evidence for a lower level of efficiency gains than with ex ante
11
enforcement (eI (k) < eI ). In other words, provided that the firms engage in merger, they
will provide more information with ex post enforcement.
Importantly, and in contrast with the ex ante enforcement, there also exists an efficiency
threshold above which the merging firms decide to engage in merger, due to the higher cost
associated with a possible ex post ban. This extra cost k reduces the expected merger profitability, and thereby the incentive to merge. If k is high enough, it induces a selection
of types beforehand and only the insiders with sufficiency efficient gains will submit their
merger. However, the corresponding cut-off efficiency level will clearly be below e, meaning
that anticompetitive mergers will be submitted with ex post enforcement, or, equivalently,
the ex post enforcement necessarily involves underdeterrence. This is due to the fact that the
induced selection is not perfect, given the asymmetric information context - if it were, then,
by controlling mergers the CA would trigger only pro-competitive mergers being submitted,
but then it would no longer need to control mergers, i.e. that would not be an equilibrium.
Next we turn to the ultimate objective of our study, which is to compare the outcomes
of ex ante and ex post merger enforcement, so as to conclude on which one is optimal and
under which circumstances.
We examine below the welfare impact of the ex post control, for which we can distinguish
between two different effects.
The first one is a "selection" effect: when faced with the ex post enforcement, some
anticompetitive mergers are no longer submitted, because of the higher expected cost of a
post-consummation ban. Recall that eep < e, meaning that the marginal merger (i.e. the last
one to take place) is anticompetitive. In other words, the selection effect is always welfare
improving because all deterred mergers are anticompetitive.
The second effect is "informational": with ex post enforcement, the firms invest more in
evidence provision so as to avoid the extra cost of undoing the merger after consummation in
case of control and prohibition. The welfare outcome of this informational effect is however
ambiguous: on the one hand fewer pro-competitive mergers are blocked since they provide
12
more often information, but on the other hand, and due to the same increase in evidence
provision by the insiders, there may be also more anticompetitive mergers that get cleared.
Clearly, the welfare impact of the informational effect is closely related to the cost of evidence
provision. More precisely, three cases are possible.
If the cost of information is high enough, only pro-competitive mergers invest in evidence
provision in case of ex post enforcement, because they are the only ones for whom it is
profitable to do so. Then, the CA will make fewer type I errors (wrongful bans), and the
informational effect is positive, welfare-improving.
If, on the contrary, the cost of information is very low, then regardless of the type of
enforcement, ex ante or ex post, all merger types provide evidence. In other words, the ex post
control does not trigger supplementary evidence provision, and therefore its informational
effect is nil.
Finally, if the cost of information takes intermediate levels such that ex post, some anticompetitive merger types start investing in evidence provision while they did not ex ante, then
the informational effect is ambiguous for welfare, because in that case more anticompetitive
mergers will get cleared.
We can now compare the welfare outcomes of ex ante vs ex post control, and conclude
when the latter is optimal:
Proposition 3 The ex post control is always optimal if c is either high enough or low enough.
For intermediate levels of c, the ex post control is optimal if the distribution of efficiencies is
sufficiently polarized around e and e.
Proof. See the Appendix.
We show here that the optimality of the ex post enforcement relies on the size of and
interplay between its two welfare effects, the selection effect and the informational effect. Let
us focus first on cases where the ex post control leads to the higher expected welfare and
then discuss the opposite cases.
If the cost of evidence provision for firms is very high, both effects are positive, whereas
13
if this cost is very low, there is no informational effect. In both instances, the ex post control
dominates the ex ante control. For intermediate levels of the firms’ information cost, the CA
may fail to control mergers more efficiently ex post. Indeed, the CA will face more insiders
providing evidence in favor of their merger, but this evidence provision comes from both
pro- and anticompetitive mergers. In this case the informational effect of ex post control is
ambiguous and potentially even negative. However, the negative informational welfare effect
will be low if the share of anticompetitive mergers providing evidence is also low. Moreover,
recall that the ex post enforcement always yields a positive selection effect, since the expected
merger profitability is lower due to the extra cost k. The worst merger types are deterred
from merging under ex post assessment and the higher the share of such worst mergers
being deterred, the higher the positive selection effect. To sum up, the more concentrated
the distribution of merger types around its extreme values, the higher the positive selection
effect and the lower the negative informational effect, i.e. the more likely that the ex post
enforcement be optimal.
4
Concluding remarks and policy implications
This paper studied the optimal timing of merger control by comparing the pre- and postclosing enforcement. Mergers have both pro- and anticompetitive effects, and in order to avoid
the costly ban, the insiders need to provide enough evidence on the merger’s pro-competitive
effect to counterbalance the agency’s own evidence on the merger’s anticompetitive impact.
The relative efficiency of the ex post assessment relies on the size of and interplay between
its two welfare effects, the selection effect and the informational effect. The former is always
welfare-improving, since it results from the deterrence of the most anticompetitive mergers.
The latter however can potentially lower welfare, whenever anticompetitive mergers have
higher incentives to provide evidence in favor of their merger. Therefore, if the share of
such types is high, the expected benefit from the ex post control is lower. We thus conclude
that the welfare comparison between ex ante and ex post control depends on the shape of
14
the distribution of merger types. More precisely, we obtain that the ex post assessment is
optimal when there are a lot of very bad and very good mergers that may take place.
In terms of policy implications, we basically offer the opposite recommendation to that
of Ottaviani and Wickelgren (2011): the ex post control is preferable in case of a risky
distribution of merger types, highly polarized around its extremes values. In this case, the ex
post enforcement yields overwhelming benefits: the deterrence of the most anticompetitive
merger types, and incentives for the very pro-competitive merger types to invest more in
evidence provision. Instead, for a low-risk merger-type distribution, displaying the bulk of
types around median values, i.e. not very pro- nor very anticompetitive mergers, the ex
ante control dominates because the ex post control gives more incentives to anticompetitive
merger types to invest in evidence provision, thus making merger control more difficult ex
post.
To better grasp the diverging finding as compared with Ottaviani and Wickelgren (2011),
note that the uncertainty is ex ante symmetric in their paper, while we focus on uncertainty
on the CA’s side. More precisely, if there is an important uncertainty on the merger’s welfare
effect, but the firms are ex ante better informed than the CA, then we claim that the ex post
control is preferable. Instead, if the firms as well have poor information on the true welfare
impact of the merger, then Ottaviani and Wickelgren (2011) claim that the ex ante control
is more attractive. Consequently, both papers actually yield compatible predictions, to the
extent that they address different cases - see the Summarizing Table below:
Optimal regime
Symmetric uncertainty
Risky distribution
Low-risk
of types
distribution
(highly polarized)
of types
ex ante
ex post
ex post
ex ante
(Ottaviani and Wickelgren 2011)
Asymmetric uncertainty
(this paper)
Figure 1: Summarizing Table
15
References
[1] Armstrong, M. and J. Vickers, 2010, "A Model of Delegated Product Choice", Econometrica 78(1), 213-44.
[2] Barros, P., 2003, "Looking behind the curtain - effects of the modernization of the
European competition policy", European Economic Review 47(4), 613-624.
[3] Berges, F., F. Loss, E. Malavolti and T. Vergé, 2008, "European Competition Policy
Modernization: From Notification to Legal Exception", European Economic Review
52(1), 77-98.
[4] Burguet, R. and R. Caminal, 2015, "Bargaining Failures and Merger Policy", International Economic Review 56(3), 1019-41.
[5] Choe, C. and C. Shekhar, 2010, "Compulsory or voluntary pre-merger notification?
Theory and some evidence", International Journal of Industrial Organization 28(1), 1020.
[6] Dertwinkel-Kalt, M. and C. Wey, 2016, "Evidence Production in Merger Control: The
Role of Remedies", DICE Discussion Paper No. 217.
[7] Lagerlöf, J. and P. Heidhues, 2005, "On the desirability of an efficiency defense in merger
control", International Journal of Industrial Organization 23(9-10), 803-27.
[8] Neven, D., 2001, "How should protection be evaluated in Article III GATT", European
Journal of Political Economy, 17(2), 421-444.
[9] Neven, D. and L.H. Röller, 2005, "Consumer surplus vs. welfare standard in a political
economy model of merger control", International Journal of Industrial Organization 23
(9—10), 829—848.
[10] Nocke, V. and M. Whinston, 2010, " Dynamic Merger Review", Journal of Political
Economy 118(6), 1201-51.
16
[11] OECD,
able
2014,
mergers",
"Investigations
14
of
consummated
DAF/COMP/WP
and
3/WD(2014)23,
non
notifi-
available
at
http://www.oecd.org/officialdocuments/publicdisplaydocumentpdf/?cote
=DAF/COMP/WP3/WD%282014%2923&docLanguage=En
[12] Ormosi, P.L., 2012, "Claim efficiencies or offer remedies? An analysis of litigation strategies in EC mergers", International Journal of Industrial Organization 30, 578-592.
[13] Ottaviani, M. and A. Wickelgren, 2011, "Ex ante or ex post competition policy, A
progress report", International Journal of Industrial Organization 29(3), 356-35.
[14] Rosch, J.T., 2012, Remarks at the ABA Section of Antitrust Law Spring Meeting: Consummated Merger Challenges (Comm’r, F.T.C.,) - The Past is Never Dead (Mar. 29,
2012), available at http://www.ftc.gov/speeches/rosch/120329springmeetingspeech.pdf.
17
5
Appendix
Proof of Proposition 3. The ex post expected welfare writes:
G(xep )
e
e
ep
Max(eep ,eI (k)) hW (e)f(e)de + (1 − G(x )) eep
W (e)f(e)de − E(x/x < xep ).
The ex ante expected welfare writes:
G(xea )
e
e
ep
Max(e,eI ) hW (e)f(e)de + (1 − G(x )) e W (e)f (e)de
− E(x/x < xea ).
(i) When c is low and h high, all types invest in evidence provision ex post as well as ex
ante, i.e. eI = e.
For any x, the ex post expected welfare is higher than the ex ante expected welfare:
if the CA does not control:
while if the CA controls:
e
eep
e
eep
W (e)f(e)de >
hW (e)f(e)de >
e
e W (e)f(e)de,
e
e hW (e)f (e)de.
(ii) When c is high and h low so that only pro-competitive insiders provide evidence ex
post, i.e. eI (k) > e, then for any x, the ex post expected welfare is higher than the ex ante
expected welfare:
if the CA does not control:
whereas if the CA controls:
e
e
eep W (e)f(e)de > e W (e)f(e)de,
e
e
eI (k) hW (e)f(e)de > eI hW (e)f(e)de.
(iii) For the intermediate range of c and h, we show there always exists a distribution
of types e such that the ex post control yields a higher expected welfare than the ex ante
control.
For every k > 0, one has that eI (k) < eI and for high enough k one has that eep > e.
We also consider here the case where eI > e and eI (k) < e. The other cases were examined
before.
Note first that if eep > eI , which is the case for k high, the ex post expected welfare is
higher than the ex ante expected welfare.
We thus focus now on the case where eep < eI .
For a given x, the expected welfare is equal to:
e
e
Max(eep ,eI (k)) hW (e)f(e)de − x, eep W (e)f (e)de),
e
e
Max( eI hW (e)f(e)de − x, e W (e)f(e)de).
ex post: Max(
and ex ante:
18
Four different configurations emerge depending on the level of x :
(a) For x such that there is no control ex post nor ex ante (i.e. the CA does not challenge the merger, neither ex post nor ex ante), the ex post expected welfare is higher. The
difference between the ex post expected welfare and the ex ante expected welfare is given by:
eep
e
W (e)f(e)de. The higher (F (eep ) − F (e)) , the higher this difference.
(b) For x such that there is control ex post and ex ante (i.e. the CA challenges the merger
ex post and ex ante), the difference between the ex post expected welfare and the ex ante
expected welfare is given by:
eI
Max(eep ,eI (k)) hW (e)f(e)de.
If (F (e) − F (eep )) is low enough,
the difference is positive.
(c) For x such that there is no control ex post (i.e. the CA does not challenge the merger)
but there is one ex ante (i.e. the CA challenges the merger), the difference between the ex
post expected welfare and the ex ante expected welfare is equal to:
e
eep
W (e)f(e)de −
e
eI
hW (e)f (e)de + x
The lower (F (e) − F (eep )) and the higher (F (e) − F (e)), the higher that difference that
may be positive.
(d) For x such that there is control ex post (i.e. the CA challenges the merger ex
post) but no control ex ante, then
e
e W (e)f(e)de
>
e
eI
e
Max(eep ,eI (k)) hW (e)f(e)de
−x >
e
eep
W (e)f(e)de >
hW (e)f (e)de − x.
Therefore, the expected welfare is higher ex post.
Thus, if we consider distributions with large enough weight on the intervals [e, eep ] and
[e, e] and small enough weight on eI (k), e , then the ex post welfare is higher.
19