Bargaining over Remedies in Merger Regulation

Inefficient Offers to an Agency Subject to
Judicial Review: an econometric test of
remedy agreement in EC merger regulation
Luke Garrod
Bruce Lyons
Andrei Medvedev
Bargaining theory suggests mutually
beneficial early agreement
If both parties are rational and have complete
information, mutually beneficial agreement will…
Definitely be reached and
Be reached immediately
Incomplete information can explain some delay…
Screening for other’s type
Signalling of own type
Evidence from…
Experiments
Labour bargaining
EC remedy agreements involve…
Bargaining strategy from firms
More passive competition agency
Merger remedies are agreements over the line
between good and harmful parts of a merger
ECMR allows merger unless it impedes competition
Firms self-select mergers to avoid proposals that would certainly
fail this test
Nearly all merger proposals either cannot be shown to impede
competition or can be modified to this effect
Agency (DG Competition) has power to either
Prohibit merger or
Accept an offer to remedy harmful parts of merger (e.g.
divestment)
Cannot make counter-offers
The institutions of EC remedy agreement provide a
natural experiment to test theory of strategic offers
Two parties
Competition agency and merging firms
Discrete ‘rounds’ (2-phase investigation)
More information gathered in second round
Allows early or late agreement (or no agreement)
Legally specified…
Time limits to each phase (i.e. limited evidence gathering)
Order of who can make offers and who can accept/reject
Administrative system with judicial review
Agency decision must be based on evidence
No signalling in offers – would not stand up to JR
Strategic behaviour only on one side
Isolates strategy of firms – agency more passive
29% of remedies accepted only after delay
7% qualifying mergers are remedied
As opposed to either no competition problem (89%) or
withdrawn during proceedings (3%) or complete prohibition
(0.6%)
Given remedies are agreed (in either Phase I or Phase II)…
…Probability of Phase I = 71% (1998-07)
1998 ECMR revision formalised Phase I remedies;
modest upward trend in early agreement
100%
90%
80%
70%
60%
50%
Ph.I/(Ph.I+Ph.II) remedies
40%
30%
20%
10%
2007
2006
2005
2004
2003
2002
2001
2000
1999
1998
1997
1996
1995
1994
1993
1992
1991
0%
The EC procedure lends itself to simple
modelling: assumptions
Timing of moves
Phase I remedy offer αO (firms)  accept or Phase II (agency) 
Phase II remedy offer αOO (firms)  accept or prohibit (agency)
Representation of remedies
 T  0, 1 where higher αT means more of merger is OK
Information
Common knowledge of variance of agency’s evidence:
uniform with range 2σ ; but only firms know αT
Agency
More evidence in Phase II: σ2 < σ1
T
T
Evidence supports agency estimates: 1     1 ,    1 
JR-robust decision rule: accept Phase I offer iff: αO < α1 (or αOO < α2)
Objective of merging firms


Max O , OO PrApp. in Ph. I O   1  PrApp. in Ph.I * PrApp. in Ph. II OO  K

Phase I approval probabilities and expected
decision errors are directly related
Probability of approval
 
Pr Approval in Phase I   Pr 
O
O
T





 1 2 1 


1


If offer accepted in Phase I
Type 1 error (too much remedy) if αO < αT
Type 2 error (too little remedy) if αO > αT
Optimal offers and consequent probability of
delayed agreement are determined by same factors
Offers depend on whether firms find it optimal to make
‘for sure’ acceptable offer or risk disagreement
3 ranges depending on whether play completely safe in both
phases, or only in Phase I, or to risk Phase II prohibition
Example of intermediate case:
Optimal offer and Type of error if agreement in Phase I

 *   T  12  1 1 

2 K 

 1 1 
Probability of failure to agree in Phase I
 
1  Pr  * 
1
 2
K 
3


4

  1 1 
 Same factors determine: (1) optimal offers; (2) type of error in
Ph.I approvals; and (3) probability of failure to agree in Ph.I
Empirical predictions from the model
Delay to Phase II more likely if
Complex or imprecise merger appraisal (high σ1)
High number of markets raising concern
Inexperienced agency (few previous cases)
Vertical issues, coordinated effects, entry barriers, …?
Delay is relatively less costly to the firms (K/π)
Large proportion of markets raising concern
Model does not predict any effect of
Obvious harm of the merger (αT)
Combined market shares; rival shares
Political impact
Merger size per se
Nationality of merging parties
Data
Sample
EU remedied mergers 1999-2006
N = 133
i.e. all remedied mergers except 27 due to lack of
reported data or predominantly vertical
Unit = merger
Aggregated from many markets per merger
Mean 13; max 142
Variables expressed as:
Market count (e.g. 13 markets under review)
Share of markets (e.g. 52% markets created ‘concern’)
Average across markets (e.g. mean combined market
share = 64%)
Treatment of potential reporting bias
Different style of reports in Phase I and II
Reporting of barriers to entry more likely in Phase II
Selection of markets for which market shares are reported seems
higher in Phase I
To derive a consistent figure for markets under review,
we applied a market share filter
Only count markets for which combined market share >25%
Consistent with EC ‘checklist’ filter
Applying filter drops more Ph.I than Ph.II markets
25% supported by sensitivity analysis
The filter removes more markets in Phase I, thus
supporting our worry of reporting bias
0.9
Proportion of Markets Removed
0.8
0.7
0.667
0.6
Phase I
0.5
Phase II
0.4
0.333
0.3
0.2
0.1
0
0
5
10
15
20
25
Merged Entity's Market Share Filter (%)
30
35
40
The market share filter mainly removes markets
without concern
900
Number of Markets Removed
800
700
Markets with concern
600
Markets without concern
500
400
300
200
100
0
0
5
10
15
20
25
30
Merged Entity's Market Share Filter (%)
35
40
As predicted, Phase II mergers are more complex
and have higher share of controversial markets
Variable
Phase I
Phase II
Markets
12.7
16.8
Mkts with concern
6.1 (53%)
10.5 (72%)
High barriers
2½ (25%)
8 (45%)
Vertical concern
12% mergers
28% mergers
Also as predicted, only negligible
differences relating to size and mkt share
Variable
Phase I
Phase II
Combined revenue
€40b
€38b
Combined mkt share
59%
63%
Share of leading rival &
concern
21%
16%
Share lead rival & no
concern
25%
27%
Probit regression results to look out for
Delay if…
Complexity of appraisal by agency
many markets to appraise
Lower opportunity cost of delay for forms
high proportion of merger under scrutiny
(1)
dependent variable: Phase
II = 1
constant
number of markets
(non-coordinated)
coefficient
(standard error)
(2)
marginal
effects
-0.8527
-1.0628
(0.5391)
(0.6900)
0.0148**
0.0046
(0.0070)
number of markets
(coordinated)
-0.0986
0.8801
-0.0307
-0.0002*
0.3166
1.4407***
-0.0001
0.0789
0.4484
-0.7222*
0.0251
-
-0.0003*
1.4180***
0.0533
-0.1828
-0.7001*
0.1636
-0.0001
0.4412
0.0168
-0.1783
(0.4025)
-
0.0039
(0.0080)
R-squared
0.3365
(0.4658)
(0.3999)
merged entity’s mean
market share
0.9331
(0.4354)
(0.4624)
EEA and US†
-0.0305
(0.0002)
(0.4333)
US only†
-0.0979
(0.6329)
(0.0001)
% markets with concern
0.0045
(0.0716)
(0.6208)
merger number
0.0145**
marginal
effects
(0.0069)
(0.0713)
coordinated effects†
coefficient
(standard error)
0.651
0.0012
Market Share
Threshold
Dependent variable:
phase II
constant
number of markets
(non-coordinated)
number of markets
(coordinated)
coordinated effects
merger number
% markets with concern
US only
EEA and US
merged entity’s mean
market share
0%
20%
(1)
(2)
(1)
(2)
(1)
(2)
-0.7230
-0.9704
0.8127
-1.0708
-0.7306
-1.2481*
(0.5013)
(0.6654)
(0.5226)
(0.6808)
(0.5763)
(0.7582)
0.0092*
0.0093*
0.0122*
0.0122*
0.0171**
0.0164**
(0.0054)
(0.0537)
(0.0063)
(0.0063)
(0.0079)
(0.0078)
-0.0891
-0.0878
-0.0860
-0.0852
-0.1154
-0.1186
(0.0590)
(0.0592)
(0.0656)
(0.0658)
(0.0897)
(0.0920)
0.9633
1.0249
0.7653
0.8331
0.8795
1.0061
(0.6537)
(0.6665)
(0.6091)
(0.6226)
(0.6013)
(0.6216)
-0.0003*
-0.0003*
-0.0003*
-0.0002*
-0.0002
-0.0002*
(0.0002)
(0.0002)
(0.0001)
(0.0001)
(0.0001)
(0.0001)
1.5903***
1.5681***
1.5057***
1.4837***
1.0795***
1.1167**
(0.4084)
(0.4102)
(0.4290)
(0.4307)
(0.4484)
(0.4518)
0.1106
0.0752
0.1037
0.0705
0.0756
0.0356
(0.4651)
(0.4699)
(0.4634)
(0.4674)
(0.4529)
(0.4575)
-0.6839*
-0.6627*
-0.7226*
-0.6985*
-0.7129*
-0.6657*
(0.3970)
(0.3991)
(0.3973)
(0.3997)
(0.3914)
(0.3950)
-
0.0045
-
0.0047
-
0.0083
(0.0080)
R-squared
30%
0.1853
0.1873
(0.0079)
0.1669
0.1691
(0.0079)
0.1327
0.1396
Conclusions
Delay in reaching agreement arises when competition
issues are complex and delay is costly to the firms
Firms act strategically
Not just greater potential harm merger (e.g. high shares)
Remedies agreed in Phase I are likely to be
Insufficient (Type 2 error) if competition issues are complex
and/or much for the firms to fight for
But too stringent (Type 1 errors) if competition issues are
relatively straightforward and/or delay is costly to firms
Issues on which we would particularly
welcome discussion
Any improvements on this paper!
Consistency of Ph.I and Ph.II reports as data sources
E.g. our market share filter, reporting of entry barriers
Appropriate econometric techniques for our dataset
Including when we move to direct use of market data
c150 mergers * ave.14 markets per merger = >2,000
obs.