The Economics of Merger Control

The Economics of Merger Control

European Merger Policy: Theory vs. Practice

University of Groningen, The Netherlands

Final Draft Master’s Thesis Economics

Geert Sprik 1323474

Contact:

g.h.sprik@student.rug.nl

Abstract

For competition authorities, it is very hard to make a sound economic analysis in a merger investigation. Therefore, this thesis tries to analyze whether the European Commission has used economic theory correctly in determining the anticompetitive effects of a given merger. This problem is approached by discussing the standard oligopoly models, by analyzing the actual merger policy of the EC and by discussing the merger cases

Oracle/PeopleSoft and Gencor/Lonrho. It is evident that the traditional

dominance test was not in accordance with the economic concept unilateral effects. Efficiencies were also ignored in the old regime. After the 2004 reform, the use of economics was improved. Dominance was put aside to a large extent, cost savings were taken more seriously and empirical methods were used more frequently.

JEL codes: C 15, K21, L13, L49

1. Introduction

A merger takes place when two or more firms decide to join to become one entity.1 Companies merge because they anticipate that more profits can be made when the assets of the firms are combined. However, the economic literature has shown that profits of the merged firms after the merger may be lower than the combined profits of the two merging firms prior to the merger. This result is at odds with intuition; why is it then that we observe so many mergers in practice? This is known as the so-called “merger paradox”, first pointed out by Salant et al. (1983).

Many models and results in the merger theory literature have some merit for merger policy. Competition agencies (CA’s) implicitly and explicitly use economic theory in assessing a merger proposal. Furthermore, to investigate a specific merger case, a CA has to collect case-specific information. Market shares, marginal costs, and demand elasticities etc., are variables that play a fundamental role in a merger case. It is crucial that CA’s apply the appropriate theory and analysis, so that possible anticompetitive and welfare effects of a merger are correctly determined. Merger control is probably the most controversial part of antitrust policy. Besides the tendency of governments and competition agencies to create “national champions”, a decision in which a merger is prohibited is often heavily debated. Merger policy should be designed in such a way that mergers that are not detrimental to welfare will be approved and mergers that hurt welfare will be prohibited. Economists have acknowledged the fact that mergers might have various welfare implications. Not only can a merger result in increased market power of the merging firms, also pro-collusive effects must be taken into account. Furthermore, mergers will sometimes give rise to welfare-improving efficiency gains.

Nowadays, competition policy is much more sophisticated than thirty years ago. Laws and regulations precisely set out how competition authorities should assess merger proposals. For example, since 1989 the European Union (EU) has a Merger Regulation2_{, which was seriously}

reformed in 2004, and since 1984, the United States (US) has its Merger Guidelines.3 Many authors in the economic literature have pointed at misuse of economic theory with respect to specific merger cases or merger control in general.4 Especially the merger policy in the European Union was heavily criticized. This thesis is focused on the merger policy in Europe, using US merger policy as a benchmark.

The main economic issues in merger policy are market definition, unilateral effects analysis, pro-collusive effects analysis, efficiency analysis, and the use of econometric tools. The main question of this thesis is whether the European Commission (EC) uses economic theory in such a way that these components of merger policy are correctly implemented. This question is addressed, first, by identifying leaks in EC merger policy, thereby considering the existing literature on merger policy and, second, by studying some real-life merger cases. These cases are Oracle/PeopleSoft and

Gencor/Lonrho. I will study these cases by describing the procedure

employed by the EC, and, consequently, I will give my own view on the application of economic theory in those cases. Furthermore, since the reform of EC merger policy in 2004 was a clear turning point, I will explain whether and how this reform has changed the EC’s approach to the implementation of economics in merger policy.

Merger control has evolved rapidly the last ten years. In the past, merger analysis has mainly relied on a “documents approach”. Evidence in a merger case mainly consisted of statements from the parties involved in the merger,

2 _{Council Regulation (EC) No 139/2004 of 20 January 2004 on the control of} concentrations between undertakings (the EC Merger Regulation).

Available at:

http://europa.eu.int/eur-lex/pri/en/oj/dat/2004/l_024/l_02420040129en00010022.pdf. 3 _{US Horizontal Merger Guidelines of April 8 1997. Available at:} http://www.ftc.gov/bc/docs/horizmer.htm.

customer inquiries, questionnaires, etc. Nowadays, there is a tendency to use “a more economic approach”.5 _{The use of econometric tools has become}

fairly common. This indicates that the debate on the theoretical and scientific basis of merger policy will continue for years to come.

This thesis will start with an overview of the leading economic theories with respect to mergers; section 2 is devoted to this issue. Section 3 will give an overview of the general procedure in merger control. Section 4 discusses econometric tools that can be used in merger cases. Section 5 describes the merger regimes of the European Union and the US in detail. Section 6 will discuss the use of economics and econometrics in the merger policy of the EU. Finally, I conclude in section 7.

2. Merger Theory

From an economic point of view, a merger has to be beneficial for the merging firms in terms of (discounted) profits. Otherwise, it would not make sense to merge. This implies that CA’s can expect that mergers will result in higher profits for the merging firms, the so-called “insiders”. However, is this confirmed by economic theory? What happens to the profits of the firms in the industry that do not participate in a given merger (the “outsider” firms)? Moreover, how are consumers affected by the merger, i.e. what is the effect on consumer surplus? Mergers and their effects are most relevant in oligopoly models, because in those markets there is some degree of competition, but not enough to conclude that a merger has no effects at all.6

This section is devoted to the theoretical implications of a horizontal merger. A horizontal merger is a merger between firms with similar product lines. A merger can also be vertical, i.e. a merger between firms that are in the same industry but at another stage of the production cycle. Furthermore,

5 _{Christiansen (2006).}

a merger can be conglomerate, i.e. a merger between firms that have unrelated business activities.7 _{However, I will restrict the attention in this}

thesis to horizontal mergers.

Two main issues should be considered when discussing the effects of mergers. First, mergers may have unilateral effects. Firms may find it profitable to increase price, or: behave anticompetitive, in an independent way. Second, a merger might favour collusion in the industry. The merger may have pro-collusive effects.

Various models of oligopoly are considered in the next sections. Subsequently, miscellaneous theoretical aspects of mergers are discussed. 2.1 Unilateral Effects and Oligopolistic Markets: Cournot, Bertrand and Stackelberg

Unilateral effects are defined as the effects of a merger on the ability of the merged parties to behave unilaterally in a more anticompetitive way. This will mostly be reflected in a higher market price. In general, it is the case that a merger between two large firms will increase the market power of the merging firms, subsequently leading to higher prices and lower welfare, absent any efficiency gains. Market power is defined here as the ability of a firm to raise its price above some competitive level for a considerable period of time.

2.1.1 Cournot Competition

First, I consider the standard Cournot model. There are no capacity constraints, n identical firms, and a demand function of the form:

Q

p=α − , (2.1) where p is industry price,

α

is a demand parameter and Q is total industry demand. Firms have linear costs, leading to constant marginal costs, c. First, I consider the situation before the merger. After taking the usual first derivatives, using symmetry and rearranging, gives:

n c = q_b + − 1 ; n c c p_b + − + = 1 α _; 2 1+ − = n c b

α

π

, (2.2)

where the subscript “b” indicates that these expressions are equilibrium values before the merger. These are the “normal” results of the Cournot model. The expressions in (2.2) show, respectively, the output per firm, price in the industry and profit per firm before the merger.

Now, consider a merger of m + 1 firms. This means that there just will be

m firms less in the industry. The total, post-merger number of firms will be n-m. We can now recalculate the Cournot equilibrium by plugging in the

new number of firms, n-m, in the functions of (2.2):

m n c = q_a − + − 1 ; n m c c p_a − + − + = 1 ; 2 1+ − − = m n c a , (2.3)

where the subscript “a” indicates that these expressions are values after the merger. Note that the Cournot model implies that a merger will just lead to one or more firms disappearing from the market. This is a direct result of the assumptions of symmetric firms and constant marginal costs.

Studying the profit functions of the individual firms before and after the merger immediately makes clear that the outsiders to the merger always gain, since

π

a >

π

b, for all values of m. Turning to the insiders, we see that

there is only one firm earning profit

π

a, while before the merger there were

m + 1 firms earning profit

π

_b. Therefore, we have to calculate whether the post-merger profit of one insider firm is larger than m + 1 times the pre-merger profit of a firm. Hence, we have to determine whether:

However, firms that do not participate in the merger will earn a higher profit (compare the third expression in (2.2) and (2.3)).

Because of higher prices, it is immediately clear that consumer surplus is always lower after a merger (compare p and b p ).a 8 Roughly speaking,

given lower profits of the merging firms and lower consumer surplus, CA’s should be suspicious about a merger, whatever definition of (social) welfare standard is used,9 if it is believed that a Cournot model represents the structure of the market in question correctly.

An intuitively appealing way to show how a merger affects profits and consumer surplus is to show some figures of demand curves and reaction functions of firms. For the Cournot model, the reaction functions are downward sloping, since the decisions (quantities) are strategic substitutes. This means that if firm 1 will increase its quantity, firm 2 will decrease its quantity (ceteris paribus). After a merger, the merging firms will decrease their quantity unilaterally and outsiders will react, by increasing their quantity. Consider figure 1. R₀(Q) shows the reaction function of the outsiders to a quantity of q charged by the insiders. R1NC(q) shows the reaction curve of the insiders before they merge. This curve shifts downward to R₁c(q) _{after the merger. The curve of the outsiders does not} shift, because their response does not depend on the way the other firms, i.e. the merging firms, determine their quantity. The pre-merger Nash-equilibrium is at A and the post-merger Nash-equilibrium is at B. The insiders will now internalize the competitive losses they inflicted on each other before the merger and reduce output as a result. From figure 1, it becomes clear that the (joint) output of the insiders will decrease and the output of the outsiders will increase. Price will be higher.

8 _{Consumer surplus (CS) is 12}

(

_α−p

)

_{q. Since q is negatively related to p, CS is always} lower when price increases.

Figure 1: Cournot Reaction Functions

2.1.2 Bertrand Competition with Differentiated Products

It does not make sense to study the unilateral effects of a merger in the homogenous Bertrand model, because a merger will not change profits or consumer surplus, unless the merger is to monopoly.10 Therefore, I consider a model of differentiated products with price competition.11 _{Suppose that}

marginal costs are constant and equal to zero and that demand is symmetric. In particular, the demand function for firm i is given by:

(

)

= − − − = N j j i i n i i p p A p p _N p q 1 1 ,...,

γ

, i=1,...,N, (2.5)

where pi is the price and qi is the quantity demanded of firm i’s product. N is the number of products in the industry and γ is a parameter indicating the degree of substitution between the products. Doing the usual computations, where the profit function of a single firm before the merger is given by the

10 _{Recall that in the Bertrand model profits are zero.}

price-cost margin multiplied by the demand function (2.5), and noting that decisions are on price, gives the equilibrium pricep : b

₋ + = N N A pb ₁ 2

γ

(2.6) Now, let the number of firms that merge be denoted by M < N. The reaction function of the firms that merge has changed now (remember that in a price game, reaction functions are upward sloping). In fact, it has become flatter. It turns out that the new equilibrium price of the insiders,p , _i

and the new price of the outsiders, p , are given by: o

(

)

(

)

(

N M

)

A N M N M N N N N pi 2 2 1 3 2 4 1 2 2 2 − _{+ −} + − − + − + =

γ

γ

γ

_(2.7)

(

)

(

N M

)

A N M N M N N M N N po 2 2 ) 1 3 ( 2 4 2 2 2 − _{+ −} + − − + − + =

γ

γ

γ

_(2.8)

From equations (2.7) and (2.8) it becomes immediately clear that p > i p o

(since M > 1), and consumer surplus is lower. Furthermore, Motta (2004) shows that a merger in the differentiated Bertrand model will reduce the sum of industry profits and consumer surplus.

2.1.3 Stackelberg

A market characterized by Stackelberg competition is possibly very interesting for merger policy. This is because of lot of markets have one or two big firms, the “leaders”, that control the market by first setting their quantity and the “followers” react on that quantity by setting their own (less profitable) quantity.

Huck et al. (2001) show that a merger between a leader and a follower is always profitable in a Stackelberg market with quantity competition. Furthermore, they conclude that output and total welfare is always reduced for all types of mergers, i.e. mergers between exclusively followers, exclusively leaders, and mergers between leaders and followers.

However, there is a situation in which a merger that leads to Stackelberg leadership will enhance welfare. Suppose an industry with more than three firms, which behave, pre-merger, as in the Cournot model. Now suppose that two firms merge and that this new firm will become a Stackelberg leader. It can be shown that such a merger will result in higher profit for the merger firm, but price will be reduced. Consumers will benefit from such a merger.

2.1.4 Conclusions

The most striking conclusion from the models discussed above is that, in general, a merger will have a detrimental effect on total welfare through the increase of prices. Theory thus indicates that competition authorities should be concerned about the detrimental effects on welfare of mergers in oligopolistic markets.

To summarize, a merger between two firms within the Cournot, Bertrand and Stackelberg model: (1) increases prices, (2) reduces total output and, (3) decreases welfare.

2.2 Pro-Collusive Effects

there is no hard evidence that firms really made arrangements on price or output. For merger policy, this is not relevant. There is no hard evidence of collusive arrangements, because the merger is not operative yet. CA’s should compare the pre-merger and the post-merger situation and ask the question whether collusion is (more) likely post-merger. For a collusive agreement to be feasible, it has to be easy and timely to detect a deviation from the agreement. Punishment, by setting lower prices or higher quantities, must also be credible, because otherwise, firms would never have an incentive to punish and collusion would not be sustainable at all. Collusion is easier to sustain when short-term gains from a deviation are small compared to the cost of future retaliation.

In the next subsection, I will discuss some industry characteristics that are relevant to determine whether collusion is (more) likely after a merger. 2.2.1 Relevant Factors for Collusion12

Number of Firms (Concentration) Firms will find it easier to collude when there are only a few firms in the market. Intuitively this can be explained as follows. Suppose that there are one hundred firms in the market and they all set some collusive price. If one firm deviates from this agreement by setting a lower price, this will give this firm the whole market. This means that this firm gets all profits. The gains from deviating are thus extraordinarily large and a punishment might not be really effective and credible. Therefore, in markets with many firms it is hard to set up a collusive agreement. The opposite is true for a market with only a few firms. A merger automatically leads to a decline of the number of firms in an industry, and therefore, the number of firms in an industry is very relevant in determining the pro-collusive effects of a merger.

Symmetry of firms Firms that are highly asymmetric in terms of market share or capacity, may find it hard to collude. Consider a collusive agreement between a small and a large firm. The small firm in the agreement has very much to gain from deviating but does not lose much

when punishment is initiated. This makes the agreement unstable. A merger may lead to more symmetry and this issue is therefore very relevant for merger policy.

Homogeneity of products Prices can be compared more easily in a homogeneous market. It is easy to set a common price and a deviation can therefore be detected early. Mergers between firms with homogeneous products would be more likely to result in anticompetitive pro-collusive effects.

Entry When entry is very easy, new firms can enter and charge a lower price, causing the colluding incumbent firms to lose demand. Also the fact that entering firms will lead to more firms does hinder collusion. If entry is very easy in an industry, mergers will not likely provide collusive incentives.

Regularity and Frequency of Demand If demand is very irregular or orders are infrequent, collusion is hard to sustain. This is because in high demand periods, firms have a greater incentive to deviate since they can earn high profits. In a market with highly irregular demand, a merger will most likely not cause any collusion concerns.

Market Transparency A very important aspect in the sustainability of collusion is the degree of (price) transparency in a market. When transparency is high, firms communicate intensively with each other, deviation can be detected early and collusion can be sustained easily, ceteris paribus. However, when firms also communicate their prices to consumers, these consumers can easily choose the firm with the lowest prices. Consequently, a deviation from collusive behaviour might not be profitable. When reviewing a merger case, a CA has to determine whether important information is only known by firms or also available to consumers.

Ownership Links If a firm owns a part of another firm, and vice versa, the likelihood of collusion is increased. The incentive to set a collusive price is increased, because the firm will receive a share of the rivals’ profits.13

2.3 Efficiencies and Welfare Standards

A merger does not have to hurt welfare. This is because cost savings can offset price increases that are usually a consequence of mergers. To determine whether cost savings match up against post-merger market power increases, some kind of trade-off has to be considered.

A very basic, partial equilibrium model, originating from Williamson (1968), can be used to explain how the trade-off between price rises and efficiencies works. 14 Consider figure 2 below. There are two firms in the market. There is a demand curve, D, pre-merger marginal cost is mc1, and price is p1.

Figure 2: Trade-off between Efficiencies and Market Power

Price is at marginal cost pre-merger and there are no profits, assuming no fixed costs. Now, consider a merger between the two firms, resulting in a

monopoly and consequently a higher price p2. However, the merger also leads to cost savings, whether they are a result of synergies, scale economies or whatever. Marginal costs are reduced and are now mc2. Now a comparison between pre- and post-merger welfare can be made. If a total welfare standard is used, (see below), the loss in consumer surplus, given by the dotted area plus the familiar dead-weight loss, has to be compared to the change in producer surplus, i.e. profits. Since profits are initially zero, producer surplus goes up by the dotted area and the area denoted by “efficiencies”. Therefore, in conclusion, total welfare declines by the dead-weight loss, but increases by the area denoted by “efficiencies”. The net effect is ambiguous.

Cost savings can also be studied in the textbook models of Cournot and Bertrand. For example, Motta (2004) uses a differentiated Bertrand model to show that a merger will increase total welfare, i.e. profits plus consumer surplus, if efficiencies are large enough.15

In a merger case, CA’s require possible efficiencies to be merger-specific. When merging firms claim these kinds of efficiencies, we speak of an “efficiency defence”. On the other side, cost savings might give the merging firm an enormous competitive advantage. The merging firm might set such a low price as a result of which competitors will not be able to operate profitably anymore, and eventually would have to exit the industry. This might be a reason for a CA to prohibit a certain merger. This situation is known as an “efficiency offence”. However, this “offence” is only limitedly valid because efficiencies will often result in higher profits of the merging firm and higher consumer surplus, because of the lower prices the firm might charge post-merger.16 Welfare might thus increase as a result of the merger! Furthermore, an efficiency offence argument is not focused on the effects on competition, but on the effect on competitors. This is fundamentally wrong, because competition policy should maximize and

protect welfare and not (small) firms. These arguments explain why efficiency offences should be handled very carefully in merger control.

It is very hard for antitrust authorities to estimate dead-weight loss or to determine to what extent costs will be lower. However, the analysis above suggests a simple rule: when efficiencies exist which are higher than the potential anticompetitive effects from increased market power, CA’s should allow such a merger. However, in our simple model, consumer surplus strictly decreases. Consumers are worse off from the merger, and it seems strange that such a merger may go ahead. A very important aspect is brought up here. Should economic welfare be used, defined as consumer surplus plus producer surplus, or should CA’s take consumer surplus as a standard for evaluating the anticompetitive effects for a merger? There is no single good answer to this question, since there are many valid arguments to use either one of them.

In countries like Canada and Australia the authorities tend more towards a Total Welfare Standard (TWS), i.e. consumer surplus plus profits. But in many jurisdictions, like the US and the EU, the Consumer Surplus Standard (CSS) is used. What are the arguments in favour of using a TWS instead of a CSS? First, Firms are owned by people. Those people are at the same time customers. Ultimately, profits of a firm will flow back to these consumers. Second, theoretically, there would be no incentive for firms to produce any product at all if competition policy would be aimed at strictly maximizing consumer surplus, since CA´s could just oreder firms to sell at marginal costs. Remember that consumer surplus is maximized when prices are at marginal costs (assuming firms will never sell their products below marginal cost).

towards customers, i.e. a CA might want to use competition policy to redistribute income.

Motta (2004) states that theoretical economists generally prefer total welfare. Furthermore, he argues that: “it would not be wise for competition

authorities to adopt a consumer welfare objective, for several reasons.”

Most of these reasons are stated above. Motta adds the argument that firms might not have the incentive to invest, innovate, and introduce new products when CA’s use a CSS.

To summarize, the choice between CSS and TWS is not a trivial choice between right or wrong, but rather a choice between the welfare of individual consumers and total economic wellbeing.

2.4 Miscellaneous Issues and Advanced Literature

Especially the model of Cournot with respect to unilateral effects is not very well suited to describe real-life mergers. First, the model does not recognize that a merger will lead to a firm that is approximately as large as the merging firms combined before the merger, in terms of assets and capacity. A model that includes some measurement for assets or capacity would be more realistic and more relevant for merger policy. Second, the Salant et al. (1983) model does not give the result that a merger increases the profits of the merging firm. This is problematical, since firms would not have an incentive to merge if profits do not increase. Finally, the relevance for antitrust policies is limited because products are usually differentiated and there is no single price in an industry. Firms are never the same, neither in size nor in their products.

2.4.1 Convex Costs, Assets and Capacity

A model that recognizes the above problems is Perry and Porter (1985). In this model, firms act the same as in the Cournot model of Salant et al. (1983), but costs are now convex, i.e. costs are not linear in output. Demand in this model is linear and the pre-merger cost function is given by_{C = sg+dx+}

(

_{e 2 xs}

)

2_{, where s is the fraction of total capital that is} owned by a single firm, g is industry fixed costs, d is a parameter, x is output produced by a single firm and d and e are parameters.17 Marginal costs will thus increase with production, MC =d +

( )

e s x. It turns out that there exist values for s and e such that any merger is profitable for the merging firms. This simple adjustment to the Cournot model eliminates the merger paradox.

2.4.2 Asymmetric Firms and Mergers

Farrell and Shapiro (1990) study the welfare effects of mergers. They consider a Cournot model with homogeneous goods. An important assumption in their model is that firms are permitted to differ in costs. They find the familiar result that mergers that do not create efficiencies will raise price, but they also conclude that a merger will raise total welfare if the merger is profitable and if the joint pre-merger market share of the merging firms is small enough.

Summarizing, the Farrell-Shapiro criterion says that mergers are assumed to be profitable for themerging parties, thus a positive effect on the profits of the non-merging firm plus consumer surplus, Farrell and Shapiro refer to this as the external effect of the merger, is a sufficient condition for supporting a merger.

2.4.3 Foreclosure, Countervailing Buyer Power and Remedies

It is possible that a merger will give the merging firm a lot of leverage to foreclose competitors in the future. If two large firms join forces and have many financial resources at their disposal, they might charge low prices after the merger to force competitors out and, after the rivals have left the

market, charge higher prices afterwards. It is also possible that a merger will give the merged firm exclusive access to essential production facilities. Consequently, possible entrants will never be able to enter the market.

When a market has only a few buyers, these buyers can exercise market power over suppliers. Consider the case where a merger is potentially anticompetitive, but there are buyers that can exert a lot of influence on the merging firm. In such a case, the merger may not be anticompetitive.

Anticompetitive mergers do not always have to be prohibited. It is also possible to impose conditions on the merging firm that prevent the anticompetitive effects of the merger. These measures are called “remedies” and they are quite common. Remedies fall into two categories: structural and behavioural remedies. Examples of structural remedies are the divestiture of a daughter company and the divestitures of some assets. Behavioural remedies are usually engagements by the merged firm not to abuse their market power on a certain area of business, for example the obligation to license its products to competitors. Behavioural remedies require extensive monitoring by antitrust authorities and are, by definition, not structural. This is why in general, structural remedies are preferred above behavioural remedies.

3. Merger Policy

How do competition authorities analyze a specific merger case? We will see below what the merger control procedures in the EU and the US are. This section is purely positive in nature, i.e. only the merger control procedure and analysis will be described. No tests against economic theory, opinions and/or expert views are presented here; this is left for section 6.

case analysis is the investigation of possible anticompetitive effects resulting from a merger.

3.1 Defining the Relevant Market

A market is defined as the set of products and geographical areas to which the products of the merging firms belong. It is necessary to determine in which market a given merger will have an effect. The relevant market serves therefore as some kind of boundary. However, it is not immediately clear what “the relevant market” is. For instance, two large orange cultivators in Africa may decide to merge. They both produce premium oranges. Obviously, sellers of premium oranges in Africa are part of the relevant market. However, are all other orange producers also part of the relevant market? Moreover, the producers of bananas and mangos might also be part of the relevant market. Furthermore, exotic fruit producers in

Asia might also be part of the relevant market.

In the hypothetical example above, there is concern about the relevant

product market and the relevant geographic market. In practice, this is often

the case. What is the approach to determine the relevant market in these cases? At first glance, it seems quite easy to determine product markets. For the above example, it seems that the relevant market is the market for oranges in Africa, because oranges are easily distinguishable from other fruits and Africa is far away from other continents. But is this the case? Certainly the answer here is “no”. Oranges come in many types and, on the other extreme, may be good substitutes for other exotic fruits. Furthermore, oranges produced in Africa are sold all across the world, so it is quite unrealistic to expect that the geographic market is restricted to Africa.

grower, a monopolist. Would it be profitable for this orange grower to increase prices in a non-transitory way by, say, 5%? If this is indeed profitable, this will mean that premium oranges do not face very high competitive constraints. The monopolistic orange grower does not lose that much demand to a substitute product, say, mango growers. The SSNIP test has now found the relevant market, i.e. African premium oranges. If the SSNIP test reveals that a price increase is unprofitable, this implies that the relevant market is not given by premium oranges only. The test should then consider a wider market, for instance all oranges. If a price increase is still unprofitable, the market has to be widened further, including for instance all exotic fruits. This process continues until a price increase is profitable and a separate market has been found.

The SSNIP test relies on the concept of product substitutability. Consumers may find oranges and bananas very different. This is demand substitutability. However, also suppliers may substitute in production. For example, a cultivator that produces all sorts of exotic fruits may decide to switch to produce more premium oranges, by planting more orange trees. In this example, it may be very costly for such a grower to switch to oranges. For the SSNIP test, this means that the market is defined quite narrow, because these suppliers will not find it easy to react to a price rise of premium oranges (by producing more premium oranges). Another constraint on the profitability of a price rise is possible entry in the premium orange market. If it would be very easy for newcomers to set up an orange plantation, a price rise of the incumbent firms will not be profitable.

3.1.1 Product Market

To define a product market the (own)-price elasticity of a product is a useful piece of information. The price elasticity is defined as the percentage change in the quantity demanded resulting from a one percent increase in the price of that product. If the demand for oranges decreases by 10% because of a price rise of 5%, it is likely that this price rise is unprofitable. Furthermore, in the light of the SSNIP test, it is unlikely that only oranges are part of the relevant product market, since the price might cause customers to choose other products.

When there are other products that might exercise competitive constraints on the demand for oranges, it is useful to estimate the cross-price elasticity with respect to these substitute products. The cross-price elasticity between for example oranges and mangos is defined as the percentage change in the demand for mangos when the price of oranges rises by 1%. If this elasticity is high, mangos are a good substitute for oranges, and might be part of the relevant product market.

Analyzing the correlation between prices of products is also a helpful tool. When prices of oranges and mangos move in the same direction over some period of time this can be an indication of high substitutability on the demand side. However, correlation in itself is not evidence of a high degree of substitutability. For example, a year in which the weather is very bad and the harvest fails, it is likely that all exotic fruits will be more expensive. 3.1.2 Geographical Market

Elzinga and Hogarty (1973) have developed a test (EH test) that can identify a geographic market. The first step is to determine whether imports are small with respect to total domestic consumption in a given region. The next step is to investigate whether exports are a small part of local production. If the answer to both questions is affirmative, the relevant geographic is known.

The transportation costs compared to the price of a product give another piece of information on the scope of the geographic market. If transportation costs between two regions for some product market are high, it is not likely that both regions exercise a competitive constraint on each other. However, when transportation costs are not that large compared to the value of the product the geographic market may be very large. For instance, in the example above, the oranges are shipped to Europe and the US in large amounts at the same time, so high transaction costs do not hinder competition between continents.

3.2 Market Power, Market Share, HHI: Structural Indicators

When the relevant product and geographic market are determined and the firms in the market are identified, it is time to focus on the concept of market power. We defined market power as the ability of firms to keep prices above some competitive level for a considerable period of time. Market power is closely related to the analysis of the so-called structural

indicators. These indicators are very important for any antitrust organization

because they can exclude possible anticompetitive effects from the merger in an early stage of the merger investigation. For example, when the market share of two merging is not more than 10% in total, it is very unlikely that the merger will raise competition concerns. Why is this the case? Mergers that concern small firms will just not have a large enough impact on welfare. It is not worthwhile to prohibit a merger that has a marginal effect on welfare.

determining whether a merger is potentially anticompetitive.18 _The

characters HHI stand for Herfindahl-Hirschmann Index. The HHI of a market is calculated by summing the squares of the market shares held by the respective firms in a given industry;

= = N i i S HHI 1 2_{, where S} i is the market share of firm i. For example, an industry consisting of three firms with market shares 20%, 30% and 50% has an HHI of (20%)2 + (30%)2 + (50%)2 = 3800.

3.3 Competitive Effects Analysis and the Final Decision

The next step in the merger review process is to analyze whether unilateral effects and pro-collusive effects will likely be a problem. How do antitrust authorities engage in such an investigation? In economic terms, unilateral effects can be analyzed by looking at the ability of the merging firms to impose higher prices after the merger. In essence, this analysis is an empirical question, requiring market data and empirical techniques. However, often there is little data available and there is no possibility to collect data in time. Therefore, a unilateral or pro-collusive effects analysis is often a qualitative analysis, taking into account the factors that influence the market power of a firm. When it has been established that it is likely that will have high market power, possible issues like efficiency gains, failing firms and/or remedies are taken into account. When this is completed, a decision is taken whether or not to prohibit the merger. Ultimately, firms usually have the possibility to go to court when they disagree with the decision of the CA.

4. Econometric Tools

There is a tendency in antitrust policy, and in merger control in particular, to use more and more advanced empirical methods to asses the

competitive implications of a given case.19 _{This is due to the availability of}

more sophisticated techniques, greater availability of data and an increased interest in these methods. Econometric tools may include the estimation of demand elasticities, simulation and price analysis. These techniques are used either to identify the market in a merger case or to assess the anticompetitive effects of a merger. Epstein and Rubinfeld (1999), for example, discuss the practical applications of these tools. In sections 4.1 to 4.6 below, the most relevant tools are discussed in some detail.

4.1 Data and Econometric Research in Merger Cases

Four decades ago, empirical research was simply impossible due to lack of (sufficient) data. Nowadays, data is still scarce. However, there are large marketing firms that provide market data, like AC Nielsen and IRI.20 Furthermore, industry organizations and universities might have information about specific industries or consumer preferences. Furthermore, firms, like CRA International and Lexecon are specialized in the economics of antitrust cases. These firms are often consulted by CA’s to give an (econometric) analysis in a specific merger case.

4.2 Demand Elasticity Estimation

The most important variable in merger analysis is the (own-) price elasticity of demand. The effects on consumer surplus and profits all depend crucially on this variable. The own-price elasticity of demand is defined as the percentage change in the quantity demanded of product X as a response to the percentage change in the price of product X. If one wants to estimate demand elasticities quantitatively it is necessary to use some kind of regression model. The following log-linear demand system is often used:

y p p q_i ln _i ln _i ln ln =α +β +χ ₋ +δ , (4.1)

where q is the quantity produced by firm I, i p is the price of firm i, i p−i is

the vector of all prices in the industry except p , y is the vector of all other _i

demand shifting variables, and

α

, β, χ, and δ are parameters. From a demand equation like (4.1), elasticities can be derived immediately, because of the log-linear form.

A problem in estimating demand equations is the endogeneity of the variables. Suppose that a demand equation like (4.1) is estimated, where price is assumed exogenous. However, when quantity rises, price is usually affected as well. In fact, we need a demand system, where price is a function of demand and vice versa. This means that both variables are endogenous. Since pi is on the right-hand side of (4.1) and it is assumed in the standard

ordinary least squares (OLS) model that all RHS variables are exogenous, the estimator forβ will be inconsistent. A way to solve the endogeneity problem is to use so-called “instrumental variables” regression. An instrumental variable is some variable that is correlated with the explanatory variable of interest, but uncorrelated with the endogenous variable (i.e. uncorrelated with the error term). Two Stage Least Squares (TSLS) is often used as a way to estimate the “correct” coefficients. Consider now the demand equation (4.1). With TSLS, we first regress p_i on an instrumental variable. In a demand equation, marginal cost, c_i, is often used as a instrumental variable, because marginal costs are usually correlated with price, but not with quantity. This regression produces predicted values forp_i, call those valuesp . The second step is to regress _i q on_i p , where the _i

coefficient on p is the TSLS estimator. This new coefficient_i βi is a

consistent estimator ofβ _.

4.3 Measuring Market Power: Residual Demand

power estimation. They use the concept of residual demand facing a single firm. This residual demand just gives the relationship between price and quantity of a firm, taking into account the response of all other firms in the industry. For example, the question of whether a merger creates unilateral market power can be answered by calculating the merging firms’ residual elasticities of demand and the extent of pre-merger substitution between the merging firms’ products.

Suppose, inverse demand faced by firm i =1,….,n is:

(

_i, _i,ρ

)

i

i p q Q

p = ₋ , (4.2)

where p is the price of firm i’s product, _i q is the quantity, _i Q_−i is the quantity for all other firms’ products and ρ is a vector of exogenous variables (of size S) that influences the demand function of firm i. The first task is to find the maximum of the profit function of firm i, which will give the best-reply function. The optimal quantity produced by firm i, q , is a _i

function of the vector of all other quantities, Q−i , the cost of firm i, c , the i

vector of industry-specific cost variables (of size L), w, and the vector of exogenous variables that affect demand, ρ:

(

i i

)

i

i R Q w c

q = −,ρ, , (4.3) For all other firms, the same reaction functions can be obtained:

(

i i

)

i

i R q w c

Q₋ = ₋ ,ρ, , ₋ , (4.4) where c−i is the vector of all the firm-specific cost variables apart from

those specific to firm i. Residual demand can now be calculated, where the quantity of firm i is only a function of the price of firm i, by substituting (4.3) and (4.4) back into the direct demand function (4.2):

(

i i

)

i

i RD q w c

= = + + + + = S S L L il l i i s is i i i r i q w c p 1 1 ln ln α β δ ρ χ γ (4.6)

This is a regression in logarithms, which is convenient because the coefficientβi directly gives the residual demand elasticity. However, (4.6)

would not give a consistent estimator forβi, because pi and qi are

determined simultaneously in a demand system like this. The endogenity problem kicks in again, and it is appropriate to use the IV-method.

4.4 Merger Simulation

The technique that is known as merger simulation is a very promising tool to estimate the anticompetitive effects from a merger. Merger simulation is a set of quantitative techniques that uses the standard oligopoly models, like Cournot and Bertrand, to predict the (unilateral) effects on prices of a merger. The goal of a merger simulation is to predict the prices and quantities that will prevail post-merger, using only information about the pre-merger market situation. Merger simulation is a very direct application of economic theory, which is something we as economists should like.

First, the analyst has to make informed assumptions about the nature of the industry in question, i.e. whether competition is in quantity, prices etc. Mostly, the differentiated Bertrand model is used.

The next step is to estimate a specific demand system. A demand model has to satisfy some standard conditions. For example, the own-price elasticity has to be negative. For each demand model there is a trade-off between simplicity (manageability) and accuracy. In practice, the following alternative demand models are often used:

1. Log-linear demand (Baker and Bresnahan (1985)). Now demand is given by an equation like (4.1), where the demand elasticity is constant.

ij

U associated with product j: Uij =αj−βpj+εij. Utility is a function of

pricepj, with parameters αj andβ. εij is an error term. β is the price

coefficient and is assumed the same across all consumers and products. αj

represents the perceived quality of product j. Assuming that the error terms are identically and independently distributed, the consumers’ choice probabilities are distributed according to the extreme value distribution (i.e. the logistic distribution), where the probability of a consumer choosing product j is:

(

)

(

_k p_k

)

k n p j j , 1,..., exp ₌ + + = β α β α π . (4.7)

The next step is to differentiate between “inside” goods, i.e. the products of the merging firms, and “outside” goods, i.e. the aggregation of all other products. Let product n be the outside good, with a price, pn normalizedto zero. Next, “shares” for the inside goods are defined. Each share sj equals the choice probability, that issj =πj

(

1−πn

)

. Define εj as the own-price

elasticity of good j, εjk as the cross-price elasticity between goods j and k,

and

ε

as the market elasticity. Finally, the pre-merger share-weighted average price of the inside goods isp. Werden and Froeb (1994) show that the own-price and cross-price elasticities of the system are given by respectively:

( )

{

p_j s_j s_j

}

p_j p j

β

ε

ε

= 1− + (4.8)

(

p

)

p p s_{k k} jk β ε ε = − (4.9)

3. Proportionally Calibrated Almost Ideal Demand System (PCAIDS). A new demand model that is quite often used for merger simulation is the PCAIDS model, developed by Epstein and Rubinfeld (2002).21 _{PCAIDS is}

an approximation of AIDS, but requires much less data. In fact, the only data needed are market shares and two price elasticities: an elasticity for a specific brand and an industry price elasticity. AIDS explains the market share of a product/brand as a linear function of the prices of all goods in the market: = = n j ij j i p s 1 ln

β

, (4.10) where s_i is the market share of product i, α_i is a constant, p_j is the price of product j, Y is the vector of all other demand-shifting variables and

δ

and

ij

β are parameters. Each “own” coefficient βii gives the effect of a

product’s price on its market share and the “cross” coefficients βij give the

effects of a product j’s price on the market share of product i. We wish to predict the change in the share of a good as a result of the merger. Therefore, equation (4.10) can be differentiated, giving a linear equation relating the change in each brand’s market share to the change in the prices. However, estimation of the differentiated form of (4.10) would require the estimation of n2 _{elasticities. In a system with 100 brands, this would}

amount to the estimation of 10.000 elasticities. Therefore, Epstein and Rubinfeld (2002) invented PCAIDS. They summarize the huge advantage of this model by stating that: “… The share lost as a result of a price increase

is allocated to the other firms in the relevant market in proportion to their respective shares”.22 PCAIDS requires data only on market shares, the industry price elasticity and the price elasticity of one brand. Epstein and Rubinfeld (2002) show that the own-price elasticity of brand i and the cross-price elasticity of brand i with respect to the cross-price of brand j are equal to:

(

1

)

1+ + + − =

ε

ε

i i ii ii s s b (4.11)

(

+1

)

+ = ε ε j i ij ij _s s b (4.12) After a specific demand system is chosen, the cost structure of the market has to be determined. Usually, marginal costs are assumed to be constant (with respect to output). When costs and demand characteristics are known, pre-merger prices and quantities can be “simulated” by using the first-order condition (FOC) for profit maximization.

Realizing that the merged firms will likely set higher prices post-merger, elasticities, market shares and profit margins will also be different after the merger. Merger-specific efficiencies may yield cost savings, which also have to be incorporated in the analysis. At the end, the welfare effects of the merger can be determined.

With full knowledge of the demand system (own- and cross-price elasticities), the kind of oligopoly behaviour, and costs for all firms in an industry, merger simulation would become, theoretically, the only tool needed for antitrust policy. The change in welfare could be calculated with great accuracy. In practice, there are a couple of factors that hinder the use of merger simulation. First, the characteristics of a market never follow the textbook cases of linear demand, Bertrand competition and constant marginal costs. Demand is never exactly linear, competition is never purely Bertrand and costs are seldom constant. Second, a merger simulation has huge data requirements. The information demands are often too great. Ideally, it would be best to test all available demand models in a merger simulation analysis, because the sensitivity of the results with respect to the alternative demand systems may be large. If all demand systems give a similar result, merger simulation is likely to have an influence on the decision by the antitrust agency. When different demand systems give complete different outcomes, merger simulation is not very helpful.

nice.23 _{It allows anybody to fill in numbers and the program will compare}

the pre- and post-merger situation. 4.5 Critical Loss Analysis

Critical Loss Analysis (CLA) is an often-used econometrical tool in merger analysis, and especially in delineating markets. It was introduced by Harris and Simons (1989). The question posed by CLA is this: “What is the

smallest loss of sales that would make a given price increase unprofitable for a hypothetical monopolist”? This is the “critical loss” (CL). The next step for an analyst is to determine the actual loss of sales corresponding to the price increase. If this actual loss is larger than the CL, then the price increase is unprofitable. In essence, CLA is a way to make the SSNIP test operational.

CLA is quite easy to describe mathematically, see O’Brien and Wickelgren (2003). Consider a symmetric industry with several firms selling differentiated products. Marginal costs are constant, denoted by c, and demand is linear. The pre-merger price of firms A and B is p and total quantity of firms A and B is q. Suppose now that firm A and B merge. It is likely that such a merger will lead to a price increase and a quantity reduction. The critical gains and losses for a hypothetical monopolist in this market can be compared, and are equal to:

(

Gain

)

=∆p

(

q+∆q

) (

=− p−c

)

∆q =

(

Loss

)

(4.13) Gains are thus equal to the price increase multiplied by the post-merger quantity. The losses are equal to the pre-merger profit margin multiplied by the reduction in quantity. Dividing both sides of (4.13) by pq gives:

q q p c p q q p p ₊∆ ₌₋ − ∆ ∆ 1 (4.14)

Now we can solve for the critical loss, which is given by the percentage change in quantity, ∆q q, that satisfies (4.14):

(

p c

)

p p p p p q q CL − + ∆ ∆ = ∆ − = (4.15)

Realizing that ∆p pis just the percentage price increase, the critical loss is given by: m P P CL + = , (4.16) where P is the (given) percentage price increase of the hypothetical monopolist, ∆p p, and m is the pre-merger profit margin as a percentage of pre-merger price,

(

p−c

)

p.

Plugging in some numbers, say 5% for the price increase and a 30% margin, gives: 143 . 0 3 . 0 05 . 0 05 . 0 _≅ + = CL (4.17)

This means that the quantity has to fall with more than 14,3 percent to make the 5 percent price rise unprofitable. If the actual loss is smaller than this critical loss, the price increase is profitable. If a SSNIP is indeed profitable for a given market in a merger case, this implies that the current (narrow) market should be considered as a separate market.

What about the measurement of actual loss? This is an empirical question and it requires calculation of demand elasticities. It is helpful to think about breaking the price increase of the merging firms into two parts: first, A raises its price and consequently firm B raises its price equivalently. O’Brien and Wickelgren (2003) show that the actual loss is equal to:

(

−

)

= −

= own cross cross

m P P

AL ε ε 1 ε , (4.18)

where AL stands for Actual Loss, _εown_{is the own-price elasticity of product}

A, _εcross _{is the cross-price elasticity of demand for product B with respect to}

CL m P P m P AL cross ₌ + > − = 1 ε (4.19)

After some rearranging, this becomes:

(

)

m cross CL m P m P _{= >ε} + (4.20)

From equation (4.20) it becomes clear that a higher profit margin, m, makes a given price increase more profitable. Reconsidering our previous example reveals that the price increase of 5 percent will be unprofitable if the cross-price elasticity is smaller than 0,48. This is intuitive, since it is not likely that a merger between firms, which sell products with a low degree of substitutability, will be very profitable, since the products were not facing that much competition from each other before the merger. On the other side, a merger that involves firms with similar products, i.e. products with a high degree of substitutability, might be very profitable, since the merger will take away the competitive pressure of the substitute product.

4.6 Price Correlation Analysis

In its basic form, price correlation analysis involves the calculation of a correlation between two price series of two products. High correlation would indicate that the two products are close substitutes and should therefore exert quite some competitive pressure on each other and belong to the same market.

5. Merger Control in Europe and the US

5.1 EC Merger Policy

The 1957 the Treaty of Rome introduced a central authority, the European Commission, to regulate the European market. In 1962 article 81 (anticompetitive agreements) and article 82 (abuse of dominance) of the Treaty came into force. This was the first step in the EU to identify the implications of anticompetitive practices in an economy. However, this early text was very unclear with respect to their application to mergers. It was not until 1990 that a Merger Regulation came into force. This Regulation sets out some thresholds for which a concentration, i.e. a merger, has a so-called Community dimension. If the aggregate turnover of the merging firms is above 5 billion Euros, and if some other thresholds are met, the European Commission has the jurisdiction to review the merger. When a merger concerns small firms, firms with turnover below 5 billion Euros, the Member States of the EU are responsible for Merger Control. Mergers are only reviewed once within the EU. In this way, mergers can be assessed in a single procedure, and do not have to go through a number of different procedures in individual EU countries. This is known as the "one-stop shop" principle and it is quite successful. Furthermore, firms that are settled outside Europe, but which sell into Europe also fall under the scope of the ECHMR (European Commission Horizontal Merger Regulation).

merger will “raise serious doubts as to its compatibility with the common market” or not. This is the so-called Phase I procedure. When there are indeed “serious doubts”, the so-called Phase 2 procedure starts. When the Phase I-decision is taken and there are “serious doubts”, the Commission has four months to investigate the merger proposal thoroughly in Phase II. The Phase II decision will lead to a decision whether the concentration is compatible with the common market or not. When parties to a merger do not agree with the decision of the EC, it is possible to appeal to the Court of First Instance (CFI). A further appeal to the European Court of Justice (ECJ) is also possible. These courts do not engage in fact-finding, only procedural issues, ECJ, and new views on the (old) information, CFI, are considered.

Article 2.3 of the 1990 Regulation says that: “A concentration which

creates or strengthens a dominant position as a result of which effective competition would be significantly impeded in the common market or in a substantial part of it shall be declared incompatible with the common market”. This means that a merger is prohibited when a dominant position was created or strengthened. The so-called “dominance test” was born.

In 2004, a new Regulation came into force. Article 2.3 of the 2004 regulation says that: “A concentration which would significantly impede

effective competition, in the common market or in a substantial part of it, in particular as a result of a dominant position, shall be declared incompatible with the common market”. This is the new Significant Impediment to Effective Competition test (SIEC test) and it is the most striking change, from an economic point of view, of the Merger Regulation. The reforms that lead to the new Merger Regulation of 2004 and the new Horizontal Merger Guidelines (ECHMG)24 were initiated by the European Commission at the end of 2001 by the presentation of the so-called “Green Paper”.25 The reform was launched under the motto of a “more economic approach”,

24 _{Guidelines on the assessment of horizontal mergers under the Council Regulation on the}

control of concentrations between undertakings. (2004).

appropriately.26 _{In section 6, I will discuss whether this reform had a}

significant effect on the use of economics in EC merger policy. 5.1.1 Objective of EC Merger Policy

The EC Horizontal Merger Guidelines state that: “effective competition

brings benefits to consumers, such as low prices, high quality products a wide selection of goods and services, and innovation. Through its control of

mergers, the Commission prevents mergers that would be likely to deprive customers of these benefits by significantly increasing the market power of firms”.27 The EC thus clearly employs a consumer surplus standard.

5.1.2 Market Definition

The EC is focused on two sources of competitive constraints: demand substitutability and supply substitutability. To determine which products are viewed as substitutes by customers, the SSNIP test is sometimes used in the new regime. In the old regime, the SSNIP test was never used. In the investigation of supply substitution, a more qualitative analysis is used. Is it easy and cheap for producers to switch from one product to another?

5.1.3 Structural Indicators

In the new regime, market share data and HHI levels are used as indicators for market power. Mergers that result in firms that will have over 50% market share “may in themselves be evidence of a dominant market

position.”28 _{Furthermore, mergers that do not lead to a firm with more than}

25% of the market may be presumed to have no anticompetitive effects. Table 1 below gives the HHI-levels that are used in EC merger policy. Mergers with a concentration below 1000 HHI normally do not raise competitive problems. For large values of the HHI-index (>2000) and a high change in the HHI (> 150), adverse competition effects are likely. When the

“delta” (change in the HHI) is over 250, and post-merger HHI is above 1000, the Commission is also concerned about anticompetitive effects.

Table 1 EC Horizontal Merger Guidelines HHI Standards for Concentration

HHI Post-Merger Change in HHI Adverse Competitive Effects? < 1000 Irrelevant Unlikely < 250 Unlikely* 1000-2000 > 250 Likely < 150 Unlikely* > 2000 > 150 Likely

Source: EC Horizontal Merger Guidelines. * A merger is unlikely to cause competition concerns, except for some “special circumstances” (see text).

Table 2 below gives an overview of the most important statistics of merger cases treated by the EC form 2000 to 2006. The last row shows that only in a small percentage of all cases the EC decided to prohibit a merger. In the whole history of EC merger control, i.e. since 1990, only 19 mergers were really prohibited.29

Table 2 EC Merger Control Statistics 1998-2005

Year 2000 2001 2002 2003 2004 2005 2006 Notified Cases 330 335 277 211 247 313 356 Phase II Proceedings 18 21 7 9 8 10 13 Remedies 38 20 15 17 16 18 19 Prohibitions 2 5 0 0 1 0 0 Source: http://ec.europa.eu/comm/competition/mergers/statistics.pdf. 5.1.4 Competitive Analysis

The ECHMG recognizes that a merger might cause unilateral effects and pro-collusive effects, in EC terminology non-coordinated effects and

29 _See