ConcurrencesRevue des droits de la concurrence
MMaarrkkeett ddeeffiinniittiioonn iinn rreecceenntt EECC
mmeerrggeerr iinnvveessttiiggaattiioonnss::
TThhee rroollee ooff eemmppiirriiccaall aannaallyyssiiss
LLaaww && EEccoonnoommiiccss R Concurrences N° 3-2009
wwwwww..ccoonnccuurrrreenncceess..ccoomm
AAnnddrreeaa AAMMEELLIIOO
andrea.amelio@ec.europa.eu
l Chief Economist Team, DG COMP
DDaanniieell DDOONNAATTHH
daniel.donath@ec.europa.eu
l Chief Economist Team, DG COMP
11.. Over the past few years, the Commission has increasingly used sophisticated tools
for market definition purposes, and the use of empirical evidence is now common
practice for market definition. Quantitative evidence can often provide important
insights to the SSNIP test that examines whether a hypothetical monopolist would
profitably and permanently increase prices by 5-10% in a given candidate market. This
is because it can directly assess the reactions of customers and competing suppliers to
the price increase. In particular, it aims to evaluate whether customers would switch a
sufficient amount of their purchases from the hypothetical monopolist in a candidate
market to competing products of suppliers outside the candidate market and whether
these suppliers would be able to and have sufficient incentives to supply the customers
of the hypothetical monopolist.
22.. Consider a stylized example in which there are only two firms producing a particular
product, and these two firms merge. If this product forms a relevant product market,
then the merger would lead to a monopoly. If the market is wider (as for example
customers may substitute to other alternatives), there may be other suppliers that
sufficiently constrain the merging firms to alleviate competition concerns. Quantitative
analysis can be particularly useful in determining whether the potentially substitutable
products are “close enough” to the products of the merging parties so that they could
exert sufficient constraint on them post-merger. This is because although there may be
two products that prima facie may seem as “reasonable” substitutes for the products of
the merging parties, one of the products may be more substitutable for the products of
the merging parties than the other. Thus, even if one concludes that both of these
competing products are part of the same relevant product market along with the
products of the merging parties, the competitive constraint that these products exert on
the merging parties may be different. One should thus think of market definition as an
initial “discrete” exercise that identifies whether a particular product is in a relevant
product market or not, which however does not imply that all the products in the
market exert the same constraint on the products of the merging parties.1
33.. The aim of this article is to provide an overview of the type of empirical tests that the
Commission is routinely using for delineating markets. In particular, this article
discusses the advantages and disadvantages of the two basic types of empirical methods
that lend themselves particularly well to the assessment of the hypothetical monopolist
test and illustrates how they were used in a number of recent EC merger cases. The first
type of analysis, the critical loss analysis, directly assesses whether a price increase of 5-
10% is profitable by comparing by how much the hypothetical monopolist’s quantity
sales would have to decrease to make a price increase unprofitable (i.e. the critical loss)
with an actual loss that the hypothetical monopolist would incur in response to the same
price increase. The second type of analysis, the pricing analysis, uses the key intuition
that if two products are in the same relevant market, then competition between them
would be sufficiently strong to ensure that any “misalignment” between their prices
would only be temporary, as consumers would switch from the “high price” product to
the “low price” product. Thus, this analysis clearly only approximates the SSNIP test,
as it does not provide a direct answer for whether a price increase of 5-10% would be
profitable or not for the hypothetical monopolist.
MMaarrkkeett ddeeffiinniittiioonn iinn rreecceenntt
EECC mmeerrggeerr iinnvveessttiiggaattiioonnss::
TThhee rroollee ooff eemmppiirriiccaall aannaallyyssiiss
@@ LLaaww && EEccoonnoommiiccss
AAbbssttrraacctt
Over the past few years, the Commission has
increasingly used sophisticated tools for
market definition purposes, and the use of empirical
evidence is now common practice for market definition.
Market definition is based on the SSNIP test that examines
whether a hypothetical monopolist would profitably and
permanently increase prices by 5-10% in a given candidate
market. There are two types of empirical analyses (critical
loss analysis and pricing analysis) that have by now
become standard parts of the Commission’s “toolkit” for
defining markets. The aim of this article is to provide an
overview of these empirical tests and to illustrate how they
were used by the Commission in a number of recent EC
merger cases.
Depuis plusieurs années, la Commission européenne
utilise de plus en plus des outils sophistiqués pour
la définition du marché pertinent; l’utilisation des
techniques empiriques est maintenant une pratique
commune. La définition du marché pertinent est basée sur
le test SSNIP (augmentation faible mais significative et
durable du prix). Ce test examine si un “monopoleur
hypothétique”(hypotetical monopoly test) augmenterait les
prix de 5-10 % de façon rentable et durable dans un
marché candidat donné. Deux types d’analyses empiriques
font désormais partie de la boîte à outils de la Commission
européenne dans son appréciation du marché pertinent :
l’analyse des pertes critiques (critical loss analysis) et
l’analyse des prix (pricing analysis). Le but de cet article
est de donner un aperçu de ces techniques empiriques et
d’illustrer la façon dont elles ont été utilisées par la
Commission européenne dans des cas récents de fusion
soumis à son analyse.
AAnnddrreeaa AAMMEELLIIOO*
andrea.amelio@ec.europa.eu
Chief Economist Team, DG COMP
DDaanniieell DDOONNAATTHH*
daniel.donath@ec.europa.eu
Chief Economist Team, DG COMP
* The views expressed in this article are those of the two
authors and should not be considered as representative of
the official position of DG COMP or the European
Commission. The authors would also like to thank Svend
Albaek, Penelope Papandropoulos and, in particular,
Raphaël De Coninck for helpful comments and discussions.
Concurrences N° 3-2009 l Law & Economics l A. Amelio, D. Donath, Merger market definition... 1
1 See J. Farrell and C. Shapiro, “Antitrust Evaluation of Horizontal Mergers: An Economic Alternative to
Market Definition”, November 2008, that discuss the limitations of (and alternatives to) market definition
in differentiated products industries (http://www.econ.berkeley.edu/~farrell/ftp/Unilateral73.pdf).
Concurrences N° 3-2009 l Law & Economics l A. Amelio, D. Donath, Merger market definition... 2
II.. TThhee uunnddeerrlliinngg ddaattaa
44.. The quality of any type of empirical analysis is dependent
upon the quality of the underlying data. Hence, considerable
care should be taken in constructing and cleaning the data
prior to undertaking the market definition tests. Essentially, the
market definition tests reviewed in this paper can be performed
with data on product sales (both in terms of revenue and
quantities) along with at least some product characteristics.
Such data can be obtained from marketing agencies such as
AC Nielsen or Gfk as these companies collect scanner data in
supermarkets and report aggregate statistics for each country
such as the total quantities sold of a given product in a given
country and the average market price at which this product is
sold.2 Alternatively, the merging parties’ own accounting
systems store information for every sale that a firm makes (the
so-called transaction-level data), including the date of sale, the
type of the product sold, customer name and its location, the
quantity of the product sold, sales revenues and the production
costs. This data can also be used to create average product
prices by aggregating over all the sales that the merging parties
make.3
66.. The key difference between the data from the marketing
agencies and the data from the merging parties is that while the
data from the marketing agencies report market-level prices
and quantities that can usually be directly used for the SSNIP
test, the merging parties’ data only pertains to information on
the parties’ own prices and quantities, and thus such data does
not usually correspond to market-level statistics. This however
does not mean that such data would not be useful for market
definition, because the merging parties’ price and quantity
decisions provide useful evidence as to the competitive
constraints that they face in each market in which they
operate.4 Moreover, it is also noteworthy that the Commission
recently obtained the transaction-level data from all the main
competitors (in addition to the merging parties) in a number of
transactions and was thus able to reconstruct market-level
prices for the market definition tests.5 Thus, even if no
publically available data exists on market-level prices, this
information can still be collected during the merger
proceedings.
IIII.. CCrriittiiccaall lloossss aannaallyyssiiss
11.. OOvveerrvviieeww
77.. The concept of Critical Loss Analysis (CLA) is derived
directly from the definition of an antitrust market contained in
the Commission’s notice on market definition: a relevant
antitrust market is the smallest group of producers that, if they
behaved as a single hypothetical profit-maximizing firm,
would impose at least a small but significant and nontransitory
price increase of 5-10% (SSNIP).6,7 The logic of the
test is to identify a group of producers that would be able to
exercise market power if they could coordinate their pricing
and output behaviour. The standard hypothetical monopolist
test starts with the smallest possible candidate market (i.e.
products of the merging parties), asks whether a hypothetical
monopolist could profitably impose a SSNIP, and
progressively broadens the market by adding the nearest
substitute products up to the point when such a price increase
is profitable for the first time.
88.. The price increase contemplated by the SSNIP test has two
opposing effects on the hypothetical monopolist’s profits. On
one hand, it has a negative effect on profits because sales will
fall as some consumers buy less or substitute to rival firms’
products in response to the increase in price. On the other
hand, there is an offsetting positive effect on profits as the
hypothetical monopolist now earns higher margins on all of
the remaining sales. The purpose of CLA is to evaluate this
trade-off and thus to determine whether the price increase is
profitable or not by way of comparing the critical loss and the
actual loss due to the price increase. The critical loss is the
percentage reduction in quantity such that the two effects just
balance out, i.e. the gains from the price increase are exactly
offset by the losses. The actual loss is the actual percentage
reduction in quantity that the hypothetical monopolist would
realize in the candidate market in the event of such a price
increase. If the actual reduction in unit sales is greater than the
critical loss, then the price increase will be unprofitable (as
customers would switch to other competitors that would be
willing to supply enough sales to them) and the candidate
market has to be expanded (as there are other credible
competing products in the market). If the actual reduction in
unit sales is less then the critical loss, the price increase will be
profitable, and the candidate market is indeed the relevant
market (as customers are unable to switch to other competing
products).
2 Such data is usually constructed by collecting the required information in
supermarkets that account for a large percentage of total sales and is
extrapolated to construct the total amount of sales of the given product in a
given country.
3 While the marketing data has been usually cleaned by the agencies prior to
constructing the country-level statistics, the transaction-level data may
possibly be subject to a number of outliers. For example, customers may
often return the products they purchased, and such returns are often
recorded with negative revenues and negative quantities sold. If such a
return is large enough as a percentage of the total sales in any given month,
this may influence the weighted average price in that particular month and,
as a result, the pricing analysis. It is thus customary prior to constructing
the average prices to purge from the analysis all observations with nonsensical
values such as negative revenues and quantities, and observations
whose prices are either too high or too low compared to what we would
expect the customers to pay for the products.
4 For example, if a firm sets prices of two products such that their prices
move closely together over time, this would be consistent with the two
products being in the same market.
5 Examples include COMP/M. 4513 Arjowiggins/M-real Zanders Reflex,
COMP/M. 5153 Arsenal/DSP, COMP/M. 4980 ABF/GBI and COMP/M.
4989 Alo/MX.
6 See the Commission Notice on the definition of the relevant market for the
purposes of Community competition law (Official Journal C 372,
09.12.1997):
http://ec.europa.eu/competition/mergers/legislation/notices_on_substance.ht
ml#relevant_market.
7 It is important to note that in merger analysis the relevant counterfactual for
the market-level prices are the currently observed prices. This is different
from Article 81/Article 82 investigations, where it first must be estimated
how the prices would look like absent the infringement to correctly define
how wide the market is.
Concurrences N° 3-2009 l Law & Economics l A. Amelio, D. Donath, Merger market definition... 3
1100.. Undertaking CLA thus requires the computation of two
values (critical loss and actual loss). The critical loss
computation is rather straightforward, as it is based on a
simple formula and only requires information on contribution
margins.8 However, as the contribution margin depends on the
average product price and variable costs, correct information
on these variables must be collected. In particular, contribution
margin should identify the cost savings realized by a reduction
in output in the relevant time period that is considered to be
around 2 years for merger control purposes. It is thus
imperative that total costs are correctly divided between
variable and fixed costs, as a poor identification of variable
costs can lead to relevant markets which are broader than they
actually are (i.e. when margins are too high) or to markets that
are defined as too narrow (i.e. when margins are too low).
Additionally, the contribution margin must reflect the costs of
all firms in the provisional market, and thus it may be
necessary to obtain comparable price and cost information for
all market participants and not just the parties. It may also be
important in some specific instances to make adjustments to
the basic critical loss formula to account for special features of
a particular industry.9 Thus, all these elements must be taken
into account when calculating the critical loss.
1111.. To compute the actual loss, reaction of consumers (and
competitors) must be modelled to the 5-10% price increase.
The most direct way to model this reaction is by estimating the
elasticity of demand that captures how much demand for a
product changes when the price of the product changes.
Demand estimation however has very high data requirements
and may sometimes place quite restrictive assumptions on
consumer behaviour and thus may often result in rather nonrobust
estimates or no estimates at all.10 Although less precise
than demand estimation, demand elasticity can also be
approximated by taking advantage of exogenous price shocks
(if they exist) such as for example exchange rate shocks, as
one can compute by how much the demand changed in
response to this particular price change.11
1122.. There are also other ways how actual loss can be estimated.
For example,, it may also be estimated with a use of an accurate
customer survey that directly asks a large enough group of
customers whether they would switch to a different product in
response to a 5-10% price increase of the product of interest.12
By collating the number of customers that would switch as
opposed to the number of customers that would not switch in
response to the price increase, the actual percentage loss of
customers can be calculated. Qualitative evidence on how
customers responded to sudden shocks or on past switching
customer behaviour between different products can also be used
as a “rough guide” for the magnitude of actual loss, although the
evidentiary value of such analyses tends to be lower.
22.. AApppplliiccaattiioonnss iinn rreecceenntt EECC mmeerrggeerr
iinnvveessttiiggaattiioonnss:: CCOOMMPP//MM.. 44773344
IInneeooss//KKeerrlliinngg aanndd CCOOMMPP//MM.. 55333355
LLuufftthhaannssaa//SSNN AAiirrhhoollddiinngg
1133.. The issues raised above have been recently discussed in two
merger cases that are discussed below. In particular, COMP/M.
4734 Ineos/Kerling demonstrates the difficulties in the
estimation of the actual loss using (partial) demand analysis
and what type of evidence can be obtained to convincingly
approximate the actual loss even if one cannot rely on the
results of the demand analysis. On the other hand, while it was
relatively straightforward to estimate the actual loss with a
survey in a recent airline merger COMP/M. 5335 Lufthansa/SN
Airholding, the calculation of critical loss relied on some very
restrictive assumptions, and thus the findings were not
considered to be appropriate for market definition purposes.13
1144.. The main issue in the Ineos/Kerling transaction was the
assessment of whether the UK forms its own market for SPVC
(as the parties were the only two UK producers and
would thus be merging to monopoly), or whether competing
producers that export S-PVC to the UK are a sufficiently
strong competitive force, and thus the UK is part of a wider
market that encompasses other regions such as Western
Europe.14 Calculating the critical loss amounted to using the
standard formula based on the transaction-level data (such as
product prices and variable costs for each sale) that was
collected from the merging parties and third parties. However,
to compute the actual loss, it was necessary to undertake
residual demand analysis to estimate the relevant demand
elasticities to capture the relationship between the prices of the
merging parties and the quantity of imports. This estimation
however did not provide any robust estimates, as the behaviour
of both the merging parties and the potentially competing
importers were subject to very similar shocks (as S-PVC is an
oil derivate with world-wide prices), and thus the residual
demand analysis could not disentangle the relationship
between Ineos and Kerling from the relationship between the
merging parties and the importers from Western Europe.15
8 The basic formula for the critical loss ( Q/Q) is ( P/P) / ( P/P + CM)
where P/P is equal to the hypothesized price increase (e.g. 5%) and CM
is equal to the contribution margin of the producers in the group. The
contribution margin is defined as the percentage of the product price (prior
to the price increase) that the difference between the product price (prior to
the price increase) and the average variable costs accounts for.
9 For example, when a by-product is created by the production of the product
in the provisional market from which revenues are derived, the critical loss
must reflect the lost revenues associated with the lost sales of the byproducts
adjusted for any cost savings.
10 For an extensive overview of demand estimation and the problems
associated with this empirical method, see Ian Small, “Econometric
analyses of unilateral effects in merger cases”, Nº 2-2009, nº 25835,
www.concurrences.com.
11 These calculations are less precise, because they do not control for all the
other elements that may have an effect on changes in demand and also
pertain to only a given point in time.
12 It should be noted that customer surveys are often costly in terms of time
and resources.
13 COMP/M. 5141 KLM/Martinair is another airline merger, in which the use
of CLA was discussed.
14 S-PVC (“suspension polyvinyl chloride”) is for example used for the
production of plastic pipes and window frames.
15 To see this, note that the prices of the merging parties and the importers were
all moving in the same way as a result of the common oil shocks, and thus
for example the effect of Kerling’s own price changes on its sales were very
similar to those of Ineos and of the importers. As a result, the estimation
procedure could not properly identify the relevant elasticity parameters that
precisely capture the relationship between the quantity and prices.
Concurrences N° 3-2009 l Law & Economics l A. Amelio, D. Donath, Merger market definition... 4
1155.. The Commission thus had to rely on other pieces of
evidence that suggested that, when combined together, it is
likely that the actual loss would be higher than the critical loss.
In particular, this evidence suggested that (i) the competing
producers had uncommitted capacity that was comparable to
the amount of critical loss, (ii) switching analysis suggested
that the customers of the merging parties switched an
important amount of their purchases to competing producers in
the past, and (iii) when one of Ineos’ plants unexpectedly shut
down, third parties (in addition to Kerling) were able to
respond by increasing supplies to the UK with no evidence of
price increases or changes of the merging parties’ margins.
There was thus enough evidence to conclude that the market
for S-PVC is likely broader than the UK.
1166.. One of the key issues in the Lufthansa/SN Airholding
airline merger was the product market definition on the
Brussels-Frankfurt route, and in particular whether air travel
competes with train travel. While it was relatively
straightforward to estimate the actual loss with a survey that
was conducted at the airport gates in Brussels by asking
respondents whether they would have decided to travel by train
if airfare was 5-10% more expensive, the calculation of the
critical loss brought up a number of issues. First, while critical
loss analysis assumes that a single price is charged to all
customers, the airline industry is characterized by significant
price discrimination (as for example business passengers tend
to pay significantly higher prices than leisure passengers). It
thus does not seem to be reasonable to expect (as the CLA
assumes) a hypothetical monopolist to increase the average
price by 5-10%.16
1177.. Second, notwithstanding the fact that the gross margins
will differ significantly across the different customers
according to what they paid for their tickets (and thus the
average gross margin on which the critical loss is based is
likely not appropriate for the CLA calculations), the
application of CLA to an airline industry that is characterized
by high fixed costs also brings up an important question of
what the appropriate measures of variable costs and hence
gross margins are. In particular, while it is true that in the
short-run, variable costs per seat are likely best approximated
by catering and fuel costs along with departure taxes, the
5-10% price increase may actually (over the two-year time
horizon that is relevant for merger control) have an effect on
the number of frequencies/airplanes that the airline may use on
a given route. It is thus important to properly measure the
effect of the price increase on the behaviour of the airline,
otherwise the estimate of the variable costs may be too low
(and the margins too high), which may incorrectly imply that
the market is wide, and air travel competes with train travel (as
the critical loss may be lower than it would be in reality). For
example, sensitivity analysis can be used for that purpose to
test what effect different assumptions on costs have on the
magnitude of the critical loss.
IIIIII.. PPrriicciinngg aannaallyyssiiss
11.. OOvveerrvviieeww
1188.. As was seen in the preceding section, although critical loss
analysis directly evaluates the SSNIP test, it is fairly difficult
to implement. Pricing analysis, on one hand, does not provide
a direct answer to the SSNIP test, as for example a finding that
prices of two products move closely together does not provide
any insights as to the causality of the relationship between the
prices.17. On the other hand, pricing analysis uses
straightforward and easy-to-implement empirical techniques.
There is thus a trade-off between the evidentiary value of the
findings from the pricing analysis and the ease with which the
pricing analysis can be implemented. To alleviate such
concerns, it is thus always important to complement the
findings from the pricing analysis with some factual evidence
that explains how the competing producers constrain the
hypothetical monopolist.
1199.. Two techniques can be used to examine the extent to which
prices move together over time. The first technique, the
correlation analysis, measures the extent (summarized by the
correlation coefficient) to which the movements in the price of
one product are closely associated with the movements in the
price of another product. If the prices of the two products
move perfectly in line with each other, the correlation
coefficient is one. If there is no relationship between the
prices, the correlation coefficient is zero.
2200.. As the correlation coefficient can vary between zero and
one, to assess whether prices are sufficiently correlated to
consider two products or geographic areas to be in the same
market, it is typical to use as a benchmark the correlation
between two products or areas that can be safely considered to
be part of the same market. For example, if it is accepted that
the geographic market is EEA-wide, and the issue to examine
is whether the market is world-wide, this implies that the
correlations between the prices of the different Member states
may be used as useful benchmark against which the price
correlations of the different continents can be compared.18
2211.. Stationarity analysis, the second technique used for pricing
analysis, examines whether the price of one product relative to
a price of another product oscillates around a constant value
that is close to one over time. If it does (i.e. the relative price is
stationary), this implies that the two products are in the same
market, as the prices can only deviate from each other for a
16 Instead, it is likely that the hypothetical monopolist would increase prices to
(price-insensitive) business passengers by a larger amount than to leisure
(price-sensitive) passengers.
17 Consider for example a hypothetical monopolist in country A that produces
a particular product. Finding that the prices of this product move closely
with prices of products of competing producers in surrounding countries
does not provide any evidence for whether the competing producers provide
a constraint on the hypothetical monopolist, or whether the competing
producers simply follow the prices set by the hypothetical monopolist.
Obviously, for merger control purposes, the latter rather than the former is
required.
18 Of course, there may also be instances, in which no benchmark is readily
available. In such a case, a view must be taken on what level of correlation
is high enough to indicate that two products or geographic areas are in the
same market.
Concurrences N° 3-2009 l Law & Economics l A. Amelio, D. Donath, Merger market definition... 5
short period of time. Consider, for example, Figure 1 below
that displays a hypothetical price of a product in Member State
A relative to a hypothetical price of a product in Member State
B. As the relative price oscillates around 1, this implies that the
prices in the two Member States are on average the same,
which is consistent with the two Member States being in the
same market.
2222.. Both analyses should be viewed as complements rather
than substitutes, as each has its own advantages and
disadvantages. While the key advantage of correlation analysis
is that it is fairly easy to implement, it suffers from some
important shortcomings that need to be taken into account
when interpreting the results.19 Stationarity tests, on one hand,
avoid most of the issues that correlation analysis is suspect to
and also do not require any benchmarks. On the other hand,
they involve sophisticated econometric tests that are more
difficult to implement and can also result in misleading
findings due to for example a presence of a number of
structural breaks in the relative prices.
22.. AApppplliiccaattiioonnss iinn rreecceenntt EECC mmeerrggeerr
iinnvveessttiiggaattiioonnss:: CCOOMMPP//MM.. 44551133
AArrjjoowwiiggggiinnss//MM--rreeaall ZZaannddeerrss RReefflleexx
2233.. The issues that arise while defining markets were discussed
in great detail in the COMP/M. 4513 Arjowiggins/M-real
Zanders Reflex and COMP/M. 5153 Arsenal/DSP decisions.
These decisions both contain detailed annexes presenting the
Commission’s analyses and preferred empirical techniques but
also discussing the arguments and alternative empirical
analyses that were put forward by the notifying parties. In
addition, pricing analysis played an important role in the Phase
II investigations in COMP/M. 4989 Älo/MX and COMP/M.
4799 OMV/MOL, as well as in Phase I investigation of
COMP/M. 5190 Nordic Capital/Convatec. Given that the
Arjowiggins/Reflex transaction case nicely demonstrates the
use of all the different tools of pricing analysis, the remainder
of this section discusses this case in a greater detail to
demonstrate the technical framework of pricing analysis for
delineating markets.
2244.. The key market reviewed in this transaction was the
carbonless paper market. Carbonless paper comprises of three
layers of paper (CB top sheet, CFB middle sheet, and CF
bottom sheet) and is used to make duplicate copies without a
carbon layer. It is sold either in reels or sheets (sheets are
created by cutting reels into smaller pieces). In an allcarbonless
EEA-wide market, the combined entity would have
around half of the market (with the top five producers
accounting for about 90% of the market). The position of the
merged entity would however be significantly more
pronounced if national markets for reels and sheets were
considered separately, as the market share data showed
significant variations among the different national markets.20
Thus, the Commission undertook pricing analysis to examine
whether (i) the carbonless market should be split into separate
reels and sheets markets, and (ii) whether national boundaries
defined geographic markets, or whether there was a single
EEA-wide market.
2255.. To determine whether reels and sheets are in the same
product market, correlation coefficients were calculated for six
Member States to see how closely the market prices of reels
and sheets move.21 The resulting correlations ranged from
-0.32 to 0.50 and on their own suggested that prices of reels
and sheets do not move closely together. This conclusion was
further strengthened with the use of benchmarking, as it was
accepted that the three layers of carbonless paper (CB, CF and
CFB) are all part of the same market based on the qualitative
results of the market investigation and the finding that almost
all correlation coefficients between the three layers of
carbonless paper (CB, CF and CFB) ranged from 0.8 to 0.98.
It thus followed that reels are sheets were unlikely to be in the
same market, as the correlation coefficients between reels and
sheets were significantly lower than the correlations between
the three layers of carbonless paper.22
2266.. While there was a natural benchmark to use for the
delineation of the product market, there was no obvious
benchmark with respect to the geographic market is. The
market definition analysis thus primarily relied on stationarity
tests, particularly for the reels market, given that the evidence
from the correlation analysis was somewhat mixed. As Figure
2 suggests, the French, Polish and Italian prices appear to
move more closely together, which may be consistent with
these three Member States being in the same market, although
this finding can be also driven by movements in input costs
(such as for example paper pulp).23 Taking France as an
example, the stationarity analysis tested whether the reels
prices in the other five chosen Member States relative to the
French prices that are plotted in Figure 3 below are stationary,
which would be consistent with France being part of a wider
market. The downward trends in the relative reels prices that
are visible in Figure 3 are however not consistent with a
geographic market that would be wider than France (as the
prices of French reels are getting over time relatively more
18 There are, for example, instances, in which high correlations could be
entirely driven by changes in common elements (such as common cost
movements or currency movements), and thus two markets may be defined
as belonging to the same market, although the prices may in reality not be
directly related to each other (but rather driven by a common cost element).
Alternatively, prices of two markets may be related to each other but one of
the price series may be subject to significant random disturbances at some
point in time or may respond to changes in market conditions with a time
lag, which may result in very low correlation coefficients, as correlation
analysis examines contemporaneous movements over time.
20 For example, compared to about half of the market at the EEA-level, the
merged entity would have 70-80% of the German sheets market, 60-70% of
the French sheets market, and 70-80% of the Italian reels market.
21 Monthly volume and value sales of reels and sheets for each of the three
layers of carbonless paper in each Member State were submitted to the
Commission by the five largest producers of carbonless paper for a three
year period. Given that six out of the twenty-nine EEA countries accounted
for around three quarters of the total carbonless paper sales in the EEA over
the period, the analysis concentrated on those six countries (Germany,
France, Italy, Poland, Spain and the UK).
22 The stationarity analysis led to the same finding.
23 The price series could also be adjusted for paper pulp costs that were
however not available to the Commission, and such adjusted prices could
have been used for the correlation analysis.
Concurrences N° 3-2009 l Law & Economics l A. Amelio, D. Donath, Merger market definition... 6
expensive compared to the other countries), which formal
stationarity tests further confirmed. As similar results were
found for the other Member States, the stationarity analysis did
not provide any robust evidence that would be consistent with
the different Member States being in the same relevant
geographic market.
IIVV.. CCoonncclluussiioonnss
2277.. This article describes two important types of empirical
tools used for delineating markets: critical loss analysis and
pricing analysis. Each of these methods has its own advantages
and disadvantages and they are thus in principle
complementary. In particular, on one hand, critical loss
analysis provides a direct answer to the SSNIP test, while
pricing analysis only approximates it and thus must be
complemented with other qualitative findings to provide
convincing evidence as to how wide the market is. On the
other hand, critical loss analysis requires the estimation of the
actual loss of customers in response to the 5-10% price
increase, which is fairly complicated and can be extremely
costly in terms of time and data requirements, while pricing
analysis uses empirical techniques that are easy to implement.
2288.. Pricing analysis seems to be particularly well-suited for
investigations during Phase I proceedings due to the timing
constraints of Phase I and the inherent difficulties with the
estimation of the actual loss in the critical loss analysis.
Undertaking critical loss analysis is more suitable (subject to
the caveats regarding the complications that this method brings
about) for Phase II investigations, when more time is available
to perform sophisticated empirical tests. However, more
importantly, both of these complementary methods are by now
standard parts of the “toolkit” used by the Commission in
merger investigations to complement the market investigation
when defining markets. I
AAnnnneexx
FFiigguurree 1:: SSttaattiioonnaarryy rreellaattiivvee pprriiccee
Source: Sales data from the five largest producers of
carbonless paper
Source: Sales data from the five largest producers of
carbonless paper
FFiigguurree 2:: AAvveerraaggee nneett pprriicceess ooff rreeeellss
FFiigguurree 3:: AAvveerraaggee ccoouunnttrryy rreeeellss pprriicceess rreellaattiivvee ttoo nneett pprriicceess
ooff rreeeellss iinn FFrraannccee
Concurrences est une revue trimestrielle couvrant l’ensemble des questions de droits
communautaire et interne de la concurrence. Les analyses de fond sont effectuées sous forme
d’articles doctrinaux, de notes de synthèse ou de tableaux jurisprudentiels. L’actualité
jurisprudentielle et législative est couverte par neuf chroniques thématiques.Concurrences Editorial
Elie Cohen, Laurent Cohen-Tanugi,
Claus-Dieter Ehlermann, Ian Forrester,
Eleanor Fox, Laurence Idot, Frédéric Jenny,
Jean-Pierre Jouyet, Hubert Legal,
Claude Lucas de Leyssac, Mario Monti,
Louis Vogel, Denis Waelbroeck...
Interview
Sir Christopher Bellamy, Dr. Ulf Böge,
Nadia Calvino, Thierry Dahan, Frédéric Jenny,
William Kovacic, Neelie Kroes, Christine
Lagarde, Mario Monti, Viviane Reding,
Robert Saint-Esteben, Sheridan Scott...
Tendances
Marie-Laure Allain, Jacques Barrot, Jean-François
Bellis, Murielle Chagny, Claire Chambolle,
Luc Chatel, Dominique de Gramont,
Damien Géradin, Christophe Lemaire,
Pierre Moscovici, Jorge Padilla, Emil Paulis,
Joëlle Simon, Richard Whish...
Doctrines
Guy Canivet, Emmanuel Combe, Thierry Dahan,
Luc Gyselen, Daniel Fasquelle, Barry Hawk,
Laurence Idot, Frédéric Jenny, Bruno Lasserre,
Anne Perrot, Catherine Prieto, Patrick Rey,
Didier Theophile, Joseph Vogel...
Pratiques
Tableaux jurisprudentiels : Bilan de la pratique
des engagements, Droit pénal et concurrence,
Legal privilege, Cartel Profiles in the EU...
Horizons
Allemagne, Belgique, Canada, Chine, Hong-Kong,
India, Japon, Luxembourg, Suisse, Sweden, USA...
Droit et économie
Emmanuel COMBE, Philippe CHONÉ,
Laurent FLOCHEL, Penelope PAPANDROPOULOS,
Etienne PFISTER, Francisco ROSATI, David SPECTOR...
Chroniques
EEnntteenntteess
Michel DEBROUX
Laurence NICOLAS-VULLIERME
Cyril SARRAZIN
PPrraattiiqquueess uunniillaattéérraalleess
Catherine PRIETO
Anne-Lise SIBONY
Anne WACHSMANN
PPrraattiiqquueess rreessttrriiccttiivveess
eett ccoonnccuurrrreennccee ddééllooyyaallee
Mireille DANY
Daniel FASQUELLE
Marie-Claude MITCHELL
CCoonncceennttrraattiioonnss
Jean-Mathieu COT
Jérôme PHILIPPE
Stanislas MARTIN
AAiiddeess dd’’ÉÉttaatt
Jean-Yves CHÉROT
Jacques DERENNE
Christophe GIOLITO
PP rrooccéédduurreess
Pascal CARDONNEL
Christophe LEMAIRE
Agnès MAÎTREPIERRE
Chantal MOMÈGE
RRéégguullaattiioonnss
Denis LESCOP
Jean-Paul TRAN THIET
Thierry TUOT
SSeecctteeuurr ppuubblliicc
Bertrand du MARAIS
Stéphane RODRIGUES
Jean-Philippe KOVAR
PPoolliittiiqquuee iinntteerrnnaattiioonnaallee
Frédérique DAUDRET-JOHN
François SOUTY
Stéphanie YON
Revue des revues
Christelle ADJÉMIAN
Umberto BERKANI
Alain RONZANO
Bibliographie
Centre de Recherches sur l’Union Européenne
(Université Paris I – Panthéon-Sorbonne)
Tarifs2009J Abonnement annuel - 4 n° (version papier) 420 € 428,80 €
1 year subscription (4 issues) (print version)
J Abonnement annuel - 4 n° (version électronique + accès libre aux e-archives) 510 € 609,96 €
1 year subscription (4 issues) (electronic version + free access to e-archives)
J Abonnement annuel - 4 n° (versions papier & électronique accès libre aux e-archives) 620 € 741,52 €
1 year subscription (4 issues) (print & electronic versions + free access to e-archives)
J 1 numéro (version papier) 125 € 127,62 €
1 issue (print version)
BBuulllleettiinn éélleeccttrroonniiqquuee ee--CCoommppeettiittiioonnss l ee--bbuulllleettiinn ee--CCoommppeettiittiioonnss
J Abonnement annuel + accès libre aux e-archives 560 € 669,76 €
1 year subscription + free access to e-archives
RReevvuuee CCoonnccuurrrreenncceess ++ bbuulllleettiinn ee--CCoommppeettiittiioonnss l RReevviieeww CCoonnccuurrrreenncceess ++ ee--bbuulllleettiinn ee--CCoommppeettiittiioonnss
J Abonnement annuel revue (version électronique) + e-bulletin 730 € 873,08 €
1 year subscription to the review (online version) and to the e-bulletin
J Abonnement annuel revue (versions papier & électronique) + e-bulletin 820 € 980,72 €
1 year subscription to the review (print & electronic versions) + e-bulletin
RReennsseeiiggnneemmeennttss l SSuubbssccrriibbeerr ddeettaaiillss
Nom-Prénom l Name-First name : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . e-mail : . . . . . . . . . . . . . . . . . . . . . . . . . . .
Institution l Institution : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Rue l Street : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Ville l City : . . . . . . . . . . . . . . . . . . . . . . .
Code postal l Zip Code : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Pays l Country : . . . . . . . . . . . . . . . . . . .
N° TVA intracommunautaire/VAT number (EU) : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
PPaaiieemmeenntt l PPaayymmeenntt ddeettaaiillss
Vouspouvezpayerdirectementsurwww.concurrences.com(accèsimmédiatàvotrecommande)oubienutiliserceformulaire :
For instant access to your order, pay on-line on www.concurrences.com. Alternatively :
J Veuillez m’adresser une facture d’un montant de . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . €
Please bill me for the sum of . . . . . . . . . . . . . . . €
J Veuillez débiter ma carte MasterCard/Visa/American Express d’un montant de . . . . . . . . . . . €
Please debit the sum of . . . . . . . . . . . . . . . . . . € from my MasterCard/Visa/American Express
Numéro de carte/Card n° : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Date d’expiration/Expiry date : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Nom-Prénom/Name-First name : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
J J’ai transféré au compte bancaire dont références ci-dessous la somme de . . . . . . € à la date du . . . . . . . . . . .
I have transferred the sum of . . . . . . € to the bank account below on . . . . . . . . . . . . . . .(date)
IBAN (International Bank Account Number) BIC (Bank Identifier Code)
FFRR7766 33000066 66110088 55440000 00110033 55444400 110011 CCMMCCIIFFRRPPPP
CCoommppttee 3300006666 1100885544 0000001100335544440011 0011
BBaannkk :: CIC - Paris le Marais l 58 rue du Temple - 75 004 Paris l France
FFoorrmmuullaaiirree àà rreettoouunneerr àà l SSeenndd yyoouurr oorrddeerr ttoo
IInnssttiittuutt ddee ddrrooiitt ddee llaa ccoonnccuurrrreennccee
25 rue Balard - 75 015 Paris - France l contact: webmaster@concurrences.com
CCoonnddiittiioonnss ggéénnéérraalleess ((eexxttrraaiitt)) l SSuubbssccrriippttiioonn iinnffoorrmmaattiioonn
Les commandes sont fermes. L’envoi de la revue ou des articles de Concurrences et l’accès électronique aux bulletins ou articles de e-Competitions ont lieu dès réception
du paiement complet. Tarifs pour licences monopostes; nous consulter pour les tarifs multipostes. Consultez les conditions d’utilisation du site sur
www.concurrences.com (“Notice légale”).
Orders are firm and payments are not refundable. Reception of Concurrences and on-line access to e-Competitions and/or Concurrences require full prepayment.
Tarifs for 1 user only. Consult us for multi-users licence. For “Terms of use”, see www.concurrences.com.
FFrraaiiss dd''eexxppééddiittiioonn CCoonnccuurrrreenncceess hhoorrss FFrraannccee :: 3300 € l 30 € extra charge for sending hard copies outside France
Signature
RReevvuuee CCoonnccuurrrreenncceess l RReevviieeww CCoonnccuurrrreenncceess
HHTT TTTTCC
WWiitthhoouutt ttaaxx TTaaxx iinncclluuddeedd
((FFrraannccee oonnllyy))